5.2 Emerging Science: Genetics, Chronobiology, Imaging, Biomarkers (2023–2026)

Standing caveat that applies to this entire section: almost the whole modern CH evidence base rests on small-to-moderate samples (tens to a few hundred patients). Effect sizes are modest, replication across ancestries is patchy, and several headline findings below directly contradict one another. This is a map of active argument, not settled fact.

5.2.1 Genetics — the field’s most solid ground, with corrections

peer-reviewed The anchor finding is the 2023 International Consortium for Cluster Headache Genetics meta-analysis: 16 research groups across 13 countries (Norway, Netherlands, Sweden, Denmark, Germany, Greece, Spain, Italy, Taiwan, UK), 4,043 European cases and 21,729 controls (Ann Neurol 2023).

Correction: an earlier pass of this chapter listed the loci as MERTK, FHL5, SATB2-region, DUSP10, CAPN2, ADCYAP1R1, and MME. This is wrong. ADCYAP1R1 and MME come from an unrelated, much smaller 2016 Italian study and are not loci in the 2023 GWAS at all. The correct list, with three genuinely novel loci that the earlier pass omitted entirely:

Locus (gene)	Lead SNP	Odds ratio	p-value	Status
DUSP10	rs17011182	1.38	7.8×10⁻²¹	known
MERTK	rs13399108	1.41	1.7×10⁻³⁰	known
FTCDNL1 (SATB2 region)	rs6714578	1.53	2.8×10⁻³⁷	known
FHL5	rs9486725	1.29	2.5×10⁻¹⁷	known
WNT2	rs2402176	1.20	2.6×10⁻⁸	novel
PLCE1	rs57866767	1.18	4.5×10⁻⁹	novel
LRP1	rs11172113	1.18	5.2×10⁻⁹	novel
CAPN2	rs10916600	—	1.3×10⁻¹³	trans-ancestry only (driven by the Taiwanese cohort)

SNP heritability is 14.5% (SE 1.74%). Mendelian randomisation implicates smoking as causal for CH (IVW β=1.11, p=6.3×10⁻⁶), though with significant heterogeneity (Cochran’s Q p=0.03) that should temper confidence. Three loci — FHL5, PLCE1, LRP1 — are shared with migraine but have a larger effect in CH; migraine shares CH’s genetic correlations with pain and ADHD/depression but notably not with smoking or risk-taking.

peer-reviewed A parallel Taiwanese GWAS (734 patients) found the single largest locus effect reported in any ancestry — CAPN2, OR 1.59 — alongside MERTK, genuinely shared with the European cohorts (J Headache Pain 2022). Chinese registry data (816 patients) found a family-history rate of only 6.99%, markedly lower than European series — worth bearing in mind when interpreting heritability figures built almost entirely on European cohorts (Cephalalgia 2024).

peer-reviewed A 2026 Swedish functional-validation study proposed that CH risk genes converge on NLRP3-inflammasome regulation — a possible link between genetics and inflammation (§5.2.4) (J Headache Pain 2026). Correction: the earlier pass reported the protective MERTK odds ratio as 0.69; the correct figure is 0.67 (95% CI 0.57–0.80), and the effect showed substantial heterogeneity (I²=65.7%) — in the replication cohort alone, MERTK was not significant. This inflammasome argument is built on gene-expression annotation, not on measured inflammasome activity in patients, and should be read as a hypothesis, not a finding.

5.2.2 Chronobiology — real phenotype, unsupported genotype

peer-reviewed Correction: the earlier pass attributed the current chronobiology synthesis to “Brandt & Fronczek.” The correct authors are Ran, Spulber and Belin (Karolinska Institutet, Sweden) — “Chronobiology and cluster headache: insights into a hypothalamic disorder,” Curr Opin Neurol, February 2026 (PMID 41709685). Brandt and Fronczek (Netherlands) wrote a separate, unrelated editorial the same season.

The most important sentence in the whole chronobiology literature comes from a critical 2025 review with authors from the US, Korea, and Sweden: “Multiple small genetic studies have shown core circadian gene variants to be cluster headache susceptibility genes, whereas larger genetic studies have not shown core circadian gene variants” (Burish et al., Cephalalgia 2025 peer-reviewed). None of the eight loci in the 2023 international GWAS is a core clock gene. The clock hypothesis currently survives on phenotype and physiology, not on genetics. The same review’s pooled figures: circadian rhythmicity is reported in ~70% of patients (16 studies, 4,953 individuals), the peak attack window is 2–3am, and autumn is the most common bout-onset season (31% of 3,709 patients).

peer-reviewed Countering that partly, a April 2026 study using 707 CH cases and 682 controls genotyped clock genes not covered by standard GWAS panels (BMAL1, NPAS2, CLOCK, CRY1-2, PER1-3) and reported 258 of 897 marker combinations significantly associated with CH risk, concluding “molecular clock dysfunction [plays] a central role in the manifestation of cluster headache” (Cephalalgia 2026). This directly complicates the Burish et al. “large studies find nothing” conclusion — it depends heavily on which genes a study actually looks at. No therapeutic intervention has been tested on the strength of either finding.

peer-reviewed A Portuguese/UK team found that CH patients’ CLOCK gene expression fluctuates significantly less across the seasons than in healthy controls — a blunted circannual amplitude, unrelated to whether the patient was currently in or out of bout (Cephalalgia 2024). A Swedish/US rat study mapped melatonin receptors (MT1/MT2) directly onto the trigeminal and sphenopalatine ganglia, providing a peripheral anatomical rationale for melatonin in CH rather than the usual central/pineal one — animal data only, but mechanistically interesting (J Headache Pain 2025).

5.2.3 Imaging — an open, unresolved contradiction

This is a case where sources genuinely conflict and the conflict is presented explicitly rather than smoothed over, per this chapter’s standards.

peer-reviewed Positive finding (China): the largest CH imaging cohort ever assembled — 69 episodic CH patients vs 63 controls at 7T — found increased hypothalamic volume on the headache side (right anterior-inferior hypothalamus, p=0.019; right posterior hypothalamus, p=0.017) plus widespread functional changes across hippocampal and amygdalar subregions (J Headache Pain 2025). A caveat worth noting: patient-group standard deviations on the volume measures were 4–5x those of controls, suggesting a handful of extreme values may be driving the result, and the patient and control groups were badly sex-mismatched.

peer-reviewed Negative finding (Italy) — a direct contradiction, same question: a cleaner, tighter volumetric study — 26 in-bout episodic CH vs 20 matched controls, five hypothalamic subunits plus total volume, adjusted for age, sex and intracranial volume — found no significant difference on any measure and no correlation with clinical features, concluding CH is not a macrostructural hypothalamic disease (Radiol Med 2025). A second Italian multimodal study similarly found no subcortical volume differences at all, with the actual differentiating features being cortical (frontal cortical thinning) (J Headache Pain 2026). A larger Chinese 7T study directly comparing CH to migraine also found the headline differentiating features were cortical, cerebellar and brainstem — not hypothalamic (J Headache Pain 2026).

The honest reading: hypothalamic macrostructure in CH is unproven and contested across three independent groups; the more consistent signals across studies are functional and microstructural rather than volumetric — reduced hypothalamic fractional anisotropy, increased mean diffusivity, altered network connectivity. But even here, a 2025 Italian study explicitly found no correlation between hypothalamic microstructural change and cortical-network connectivity change in the same patients — undercutting a simple “the hypothalamus drives the cortex” story (J Headache Pain 2025).

5.2.4 Biomarkers and inflammation — a corrected and genuinely contested picture

peer-reviewed Correction: the earlier pass claimed IL-1β was elevated specifically in chronic CH. This is wrong, and the direction is actually inverted in the primary source. The largest CH cytokine study to date — the Danish Headache Center’s plasma biobank, 412 participants across chronic CH, episodic-in-bout, episodic-remission and controls, 45 cytokines measured (Ann Neurol 2025) — found IL-1β was decreased in episodic-CH-in-bout plasma, not elevated in chronic CH. The genuinely correct and directly relevant finding for a chronic-CH patient is the study’s broader pattern: episodic CH in bout looks broadly anti-inflammatory (suppressed cytokines); chronic CH looks broadly pro-inflammatory (elevated IL-6, CCL7, CXCL9, HGF, MMP12, TGFα). Oncostatin M is elevated across all three disease states.

peer-reviewed A separate Swedish study measuring both CSF and serum in the same patients found CSF cytokines elevated in CH both in bout and in remission (arguing for a persistent, trait-like neuroinflammatory state rather than a bout-limited one) — but serum cytokines from the same patients moved in the opposite direction, decreasing during attacks (J Headache Pain 2024). This means peripheral blood is not a reliable proxy for what is happening in the central nervous system in CH, and any inflammation claim based on blood alone should be read cautiously.

peer-reviewed As noted in §5.1.4, PACAP-38 is elevated across all CH states and most strongly in chronic CH (+49.8%) — currently the single most consistent circulating biomarker finding that specifically distinguishes chronic from episodic disease.

A blunt critical voice worth including directly, per this chapter’s mandate to surface dissent: a Chinese (Chengdu, traditional-medicine-affiliated) review states the CH-inflammation relationship “remains an inference based on changes in inflammatory mediators and neuropeptides in clinical studies; it is not established by direct causal evidence” (Front Neurol 2025 peer-reviewed).

5.2.5 Competing hypotheses — the field’s live arguments

• Against a single hypothalamic generator: Coppola, Abagnale, Sebastianelli and Goadsby (Italy/UK) argue for a distributed network model in which the posterior hypothalamus is better understood as a “crossroads” than a generator, and propose instead an inherited misalignment between the suprachiasmatic nucleus and peripheral clocks that lowers the threshold for a bout (Cephalalgia 2024 peer-reviewed).
• The “epiphenomenon” question, posed explicitly over a decade ago, remains unresolved: is the hypothalamus causal, or just along for the ride (PMID 21864072 peer-reviewed)?
• The Japanese Headache Society’s consensus position (published in Japanese) presents four parallel mechanisms rather than a single one: hypothalamic generator, neuropeptide/NO signalling, a peripheral origin near the internal carotid/cavernous sinus, and trigeminal hyperexcitability driving the sphenopalatine ganglion (Japanese Journal of Headache 2026 peer-reviewed, non-Anglophone source).

Diagram: competing pathophysiology models