Cerebral Iron Concentration in Episodic Migraineurs

Objective: To investigate cerebral iron concentrations in patients with episodic migraine and investigate correlations with clinical parameters, such as monthly migraine days or disease duration. Methods: We included episodic migraineurs and healthy controls from 18 to 80 years; headache diaries were kept during a four-week baseline period. All participants underwent MRI scans, including a multi-echo 3D gradient recalled echo sequence that allowed calculating quantitative susceptibility maps. We performed whole-brain analyses comparing the iron level of healthy controls and migraineurs and searched for regions in which migraineurs’ iron concentrations correlated with their migraine frequency or disease duration. The signicance level was set at 0.001 (uncorrected), the extent threshold at ten voxels. Results: We included 15 patients and 18 controls. There were several brain regions such as the anterior cingulate cortex and the middle frontal gyrus, in which migraineurs stored more iron, but none in which controls had higher iron levels. Iron correlated positively with migraine frequency or disease duration in multiple brain regions. There was one region in which iron load correlated negatively with disease duration. Conclusions: Migraine predisposes to increased iron levels. Not every brain area with an altered iron concentration is active during migraine attacks, so perhaps the increased iron might not solely be due to migraine but to a common cause, such as a metabolic or information processing disorder.


Introduction
Iron plays an essential role in electron transport, glucose metabolism, and synthesis of neurotransmitters and is indispensable for the brain's normal functioning. (Piñero and Connor, 2016) Its highest concentrations accumulate in the red nucleus, the basal ganglia, and the dentate nucleus of healthy adults. (Piñero and Connor, 2016). Iron in ux and e ux are tightly regulated because free iron ions lead to oxidative stress implicated in the pathogenesis of neurodegenerative disorders. (Piñero and Connor, 2016) However, abnormal iron metabolism does not seem to be implicated only in neurodegenerative disorders.
Some studies investigated cerebral iron concentration in patients suffering from migraine. They found an increased load in the periaqueductal grey (PAG) and deep brain nuclei. ( The reason for and consequences of migraineurs' increased iron load is unknown. On the one hand, the increased metabolic activity of speci c brain areas during or outside migraine attacks could increase the concentration of free radicals, which then might lead to tissue damage and iron deposition. On the other hand, increased metabolism might increase the need for iron without causing permanent deposition. This study analyses the cerebral iron concentration of patients suffering from episodic migraine compared to healthy controls (HC). In addition, we search for brain regions in which the iron concentration correlates with monthly migraine days (MMD) or disease duration.

Patients
We included patients between 18 and 80 years with the diagnosis of episodic migraine according to criteria published in the third edition of the International Classi cation of Headache Disorders. (Olesen, 2018) We excluded patients in case of pregnancy, a neurodegenerative disorder, or contraindications against MRI examinations. Furthermore, we also excluded patients who had less than two migraine days and patients who had ≥ 15 headache days and ≥ eight migraine days during the baseline period.
As controls, we included healthy volunteers between 18 and 80 years, who did not suffer from any headache disorder.
We excluded patients and controls if the registration of the multi-echo 3D gradient recalled echo (GRE) sequence was unsuccessful. The available data determined the sample size.

Study Design
During a four-week baseline period, migraineurs kept a headache diary that allowed the identi cation of migraine days according to published criteria. (Tassorelli et al., 2018) Patients treated acute headache attacks as needed but did not change their prophylactic treatment during that period. The MRI examination de ned the end of the baseline period.

Statistical analysis
This study aimed primarily to investigate whether the migraine diagnosis affects iron load in the brain. To that end, we compared migraineurs' brains with healthy controls. The secondary aim of this study was to investigate whether MMD correlate with the cerebral iron load. To that end, we searched for signi cant correlations between the MMD during the baseline period and iron concentrations. Finally, we also searched for signi cant correlations between disease duration and the cerebral iron load.
We performed whole-brain analyses with the SPM12, and set the extent threshold at ten voxels and the signi cance level at a voxel-threshold of P<0.001 (uncorrected). We computed one-sided unpaired t-tests (e.g. controls > patients). All analyses were corrected for age and sex (whenever the sample comprised more than one sex). In addition, the analysis for correlations between the MMD and iron levels was corrected for disease duration, and the analysis for correlations between the disease duration and iron levels was corrected for MMD. We did not include headache days as independent variable because headache days and migraine days correlated highly (see below) and consequently cannot be regarded as independent. Inclusion of both variables into the model would potentially have resulted in an underestimation of the effect of migraine days.
The automated anatomical labelling atlas (AAL-3) allowed identifying brain regions. (Rolls et al., 2020) Continuous variables are reported as means and standard deviations (SD) and categorical variables as frequencies. We estimated correlations between two variables calculating Spearman's rho. Except for the whole-brain analyses for which we used the SPM software, we conducted the statistical analysis in SPSS version 25.

Data Availability
The datasets presented in this article are not readily available because due to Swiss law, the researchers must assess whether the use of the data and coded datasets are within the primary scope of the informed consent. Data is only available upon request, after the researchers have reviewed the purpose of the inquiry. Requests to access the datasets should be directed to Dr. Lars Michels, lars.michels@usz.ch

Results
Clinical data. Of 23 migraineurs who met the inclusion criteria, we had to exclude eight as no multi-echo 3D GRE sequence had been acquired. Thus, the data of 15 patients and 18 controls entered the analysis. The average age of migraineurs and HC was 35 ± 13 and 34 ± 18 years, respectively. All migraineurs (15/15, 100%) and 13 HC (13/18, 72.2%) were female. Most migraineurs reported suffering from migraine with and without aura (15/18, 80%); only three had never experienced an aura. The average disease duration was 21 ± 14 years. Average migraine, headache, and medication days were 6 ± 3, 8 ± 3, and 5 ± 3 days during the baseline period. Migraine days and headache days correlated highly (r = 0.562). Migraineurs had a mean total MIDAS score of 28 ± 23 points.
MRI data. First, we investigated differences in the iron concentrations of migraineurs and healthy controls (HC). There was no brain region in which HC accumulated more iron than migraine patients did. Conversely, we found several areas that contained signi cantly more iron in migraineurs than HC (see  Table 1).
Second, we investigated an association between iron content and MMD. We identi ed fourteen brain regions in which iron content correlated statistically signi cantly and positively with the number of MMD during the baseline period but no area with a negative correlation (see Table 1 and Fig. 2).
Finally, we searched for an association between iron content and disease duration. There was one brain region with a statistically signi cant negative correlation, and there were several regions with a positive correlation. All results are listed in Table 1 and are illustrated in Fig. 3. Table 1 Results of the analyses of the iron levels; x, y, and z are the MNI coordinates of the maximal iron accumulation; clusters comprising multiple regions are listed one below the other and marked with an asterisk. We set the extent threshold at ten voxels.
Peak Tscore Cluster size (voxel) x y z Hemisphere

Brain region
Areas with a signi cantly lower iron concentration in migraineurs than in HC after correction for age and sex Areas with a signi cantly higher iron concentration in migraineurs than in HC after correction for age and sex

Discussion
Our analysis identi ed several brain regions in which episodic migraineurs' iron load exceeds that of HC or correlates positively with MMD or disease duration. Conversely, we found no brain region in which migraine patients had a lower iron concentration than HC or iron load correlated negatively with MMD and only one region in which iron levels correlated negatively with disease duration. Thus, suffering from migraine predisposes to an increased cerebral iron load.
In many brain areas in which iron levels correlated with migraine frequency or disease duration, HC did not have less iron than did migraine patients (see Table 1). This nding suggests that these areas do not generally store more iron in migraineurs than in HC. Consequently, circumstances other than episodic migraine might lead to an increased iron load there, too.
Given that migraine frequency is not stable throughout life,(Straube and Andreou, 2019) our nding of iron correlating with MMD in some areas suggests that iron accumulations might not be stable either and change with the migraine frequency. Hence, in these areas, iron deposition might not be irreversible but accessible to transcellular transport.
Some of the areas in which HC stored less iron (see Table 1 Table 2 for further details.
Of the areas in which iron content correlated with MMD (see Table 1 Of the areas in which iron levels correlated with disease duration (see Table 1 Table 2 lists all identi ed brain areas. Table 2 Overview of brain areas with increased iron levels and their implication in migraine pathophysiology; MMD -monthly migraine days; IL -iron levels; DD -disease duration Interestingly, we did not nd altered iron levels in the insula, which holds a central place in pain processing, (Segerdahl et al., 2015) suggesting that it might not be pain per se that alters iron levels. Moreover, about the absence of correlation between iron load and MMD or disease duration in several areas (see Table 1), one may speculate that migraine and increased iron concentrations share a common cause instead of causing each other.
There was a signi cantly higher iron load in migraineurs' anterior and posterior cingulate cortex as well as precuneus (see Table 1). These are part of the default mode network (Buckner et al., 2008) that likely consumes the highest amount of energy of all large-scale networks in the brain. (Raichle et al., 2001) Consequently, the increased iron levels identi ed in this sample might be due to an underlying metabolic disorder. There is some evidence that mitochondrial dysfunction may occur in migraineurs (Gross et al., 2019) and that mitochondrial dysfunction is associated with iron accumulation. (Urrutia et al., 2014) Furthermore, a previous study suggested that the processing of sensory information might require more energy in migraineurs than in HC. (Gantenbein et al., 2013) So, in patients without mitochondrial pathology, a constantly increased energy demand might exceed mitochondria's maximum energy output.
We included many migraineurs with aura in this study (see above). Thus, a different sample might not have suggested the hypothesis of iron being linked to an underlying metabolic disorder, because previous studies indicated that -unlike in patients without aura -cerebral lactate concentrations are higher in Interestingly, not only does processing of sensory information require more energy in migraineurs (see above), (Gantenbein et al., 2013) previous studies also found atypical processing of sensory information. (Schwedt et al., 2015) Studies investigating brain activity changes in response to olfactory and visual stimuli, and heat pain found an increase in several areas with altered iron levels (see Table 3). Given that we found a positive correlation between iron concentration and MMD in some regions reacting abnormally to sensory stimulation (see Table 3), one might wonder whether these abnormalities increase with MMD. However, given the absence of a correlation between the extent of interictal phono-or photophobia and MMD, (Ashkenazi et al., 2009;Cucchiara et al., 2015) it seems less likely that an increasing iron load is associated with a growing impairment of normal functioning.
In this regard, the increased iron load in the temporal pole (see Table 1) is an intriguing nding. This area is implicated in processing sensory stimuli (see Table 3), (Demarquay et al., 2008;Moulton et al., 2011) indicating that iron deposition in that region might also be due to a disorder of information processing. Moreover, a case series suggests that the anterior temporal lobe might induce migraneous headaches, (Yankovsky et al., 2005) and one study documented increased perfusion of this area during migraine attacks. (Afridi et al., 2005) So, the positive correlation between iron and migraine frequency (see Table 1) might re ect the temporal pole's proneness for inducing attacks. However, the available knowledge is insu cient to verify this speculative hypothesis. We consider it unlikely that altered iron levels in our cohort are the result of a particular diet, as none of the patients reported being on a special diet during the study period. In addition, Domínguez and colleagues (2019) demonstrated that patients with chronic migraine have larger iron deposits in the PAG that were related to levels of biomarkers of in ammation (higher plasma levels of soluble tumour necrosis factor) and blood-brain barrier disruption (higher levels of cellular bronectin) but not diet.
Parallel mechanisms have been proposed to explain iron deposition in other neuroin ammatory conditions, such as multiple sclerosis, independent of dietary iron intake.

Strengths and Limitations
The strength of this study is that it is the rst to perform a whole-brain analysis and to use QSM sequences to analyse the iron concentrations in migraineurs' brains. Consequently, these analyses provide novel insights.
Some limitations need to be addressed. First, we only included episodic migraine patients with at least two monthly migraine days. If we had included participants with higher and lower migraine frequencies, we might have discovered additional brain areas in which iron levels correlate with migraine frequencies.
Given this limitation, we might only identify brain regions as correlating with MMD if iron concentration rises very steeply with an increasing number of monthly migraine days.
Second, our sample includes only female migraineurs. Thus, we do not know whether the results can be generalised to male patients.
Finally, even though the ideal sample size for studies analysing cerebral iron content is unknown, a larger sample might have yielded different or more reliable results. Hence, we encourage further research on that topic.

Conclusion
In this study, we showed that migraine attacks predispose to an increased cerebral iron load. In some brain areas, iron levels correlated with MMD but were comparable to HC, implying that circumstances other than episodic migraine might lead to an increased iron load there, too. This nding suggests that, in these regions, iron is not deposited irreversibly but accessible to transcellular transport in these areas.
Not every brain area with an altered iron concentration is active during migraine attacks suggesting that the consequences of the headache disorder reverberate widely -even in areas not directly implicated in the attacks.
Concerning the absence of any correlation between iron load and MMD in many areas, one may speculate whether the increased iron identi ed in this study might not solely be due to migraine but to a common cause such as an underlying metabolic disorder or a disorder of information processing.  Figure 1 Illustration of iron concentrations differences between migraineurs and healthy HC. For example, patients showed higher iron concentrations at prefrontal regions but also in the visual cortex. The full list of group differences is listed in Table 1.

Figure 2
Illustration of signi cant (p < 0.001 uncorrected, k ≥ 10 voxels) positive correlations between monthly migraine days and iron load in patients with episodic migraine. The full list of group differences is listed in Table 1.

Figure 3
Illustration of signi cant (p < 0.001 uncorrected, k ≥ 10 voxels) positive correlations between disease duration and iron load in patients with episodic migraine. The full list of group differences is listed in Table 1.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Appendix12.docx