Mitochondrial function and oxidative stress markers in higher-frequency episodic migraine

Increasing evidence points towards the role of mitochondrial functioning, energy metabolism, and oxidative stress in migraine. However not all previous research has been conclusive and some mitochondrial function/oxidative stress markers have not yet been examined. To this end, alpha-lipoic acid (ALA), total thiols, total plasma antioxidant capacity (TAC), lipid peroxide (PerOx), oxidised LDL (oxLDL), HbA1c and lactate were determined in the serum of 32 higher frequency episodic migraineurs (5–14 migraine days/ months, 19 with aura, 28 females) in this cross-sectional study. The majority of patients had abnormally low ALA and lactate levels (87.5% and 78.1%, respectively). 46.9% of the patients had abnormally high PerOx values, while for thiols and TAC over one third of patients had abnormally low values (31.2% and 37.5%, respectively). 21.9% of patients had abnormally low HbA1c and none had an HbA1c level above 5.6%. oxLDL was normal in all but one patient. This study provides further evidence for a role of oxidative stress and altered metabolism in migraine pathophysiology, which might represent a suitable therapeutic target. ALA, being too low in almost 90% of patients, might represent a potential biomarker for migraine. Further research is needed to replicate these results, in particular a comparison with a control group. This study is part of the trial registration: ClinicalTrials.gov: NCT03132233, registered on 27.04.2017, https://clinicaltrials.gov/ct2/show/NCT03132233.

www.nature.com/scientificreports/ Inclusion criteria. Patients were included, if they were previously diagnosed with migraine (with or without aura) in accordance with the ICHD-3 (International Classification of Headache Disorders version 3 Beta) Classification criteria 71 , were between the ages of 18 and 65 years, experienced between 5 and 14 migraine days per month (over the last 4 months), had an age of onset of migraine less than 50 years old and had not changed the type, dosage or frequency of any prophylactic medication (exclusive of medications taken for acute relief of migraine symptoms) for at least 3 months prior to study onset.
Exclusion criteria. Patients were excluded, if they had a history of any significant neurological, psychiatric or other medical condition or a known history of suspected secondary headache, if they were taking simple analgesics or non-steroidal anti-inflammatory drugs (NSAIDs) more than 14 days per 4 weeks or triptans on more than 10 days per 4 weeks for headaches or other body pain or any prescription opioids, if they had a previous diagnosis of medication overuse headache, which has reverted to episodic migraine within the last 6 months or met ICHD-3 Beta Classification criteria 71 for chronic migraine (> 15 headache days per month), if they had had a surgery for migraine prevention, if they had received botulinum toxin injections within the last 6 months or if they were pregnant.
Laboratory procedures. The  Venous blood samples were drawn from an antebrachial vein following overnight fasting. After 30-60 min at room temperature the serum was separated from the rest of the blood by centrifugation at 1300G for 10 min. Aliquots of serum were stored at − 80 °C. One aliquot contained 0.3 ml serum. Three aliquots per patient were sent for analysis. Blood samples for HbA1C and lactate were not stored, but immediately sent at room temperature to the inhouse laboratory for immediate analysis.
Total antioxidant capacity. TAC was measured using an ImAnOx-assay (Ganzimmun Diagnostic AG, Mainz, Germany) (inter-assay variation: 2.43%; intra-assay variation: 2.33%). This photometric test reflects the sum of all antioxidant components by measuring hydrogen peroxide (H 2 O 2 ) degradation by the serum antioxidants. Please refer to 73 for further details.
Oxidised LDL. The measurement of serum oxidised LDL was performed using a sandwich ELISA method (ox-LDL ELISA kit, Ganzimmun Diagnostic AG, Mainz, Germany) (inter-assay variation: 9-11%; intra-assay variation: 3.9-5.7%). No antioxidants were added to the plasma samples before collection. Please refer to 75 for further details.
Alpha-lipoic acid. ALA was determined in serum using the HPLC method (Ganzimmun Diagnostic AG, Mainz, Germany). The standards and the solutions were sourced from Merck KGaA. In brief, 100 µl of the serum sample was diluted in 1.9 ml acetone. The sample was mixed thoroughly for 5 s. After that, the sample was centrifuged for 10 min at 3500 U/min. 800 µl of the supernatant was evaporated under air at 45 °C for 10 min. The dry residue was dissolved in 400 µl 30/70 0.1% acetic acid/ acetone mix. The sample was mixed thoroughly for 5 s. 300 µl of the solution was transferred in a HPLC vial. A calibration curve in empty serum was prepared with five different standard concentrations. The highest standard was 200 µg/l and the lowest 12.5 µg/l. The preparation of the standard was equal to the sample preparation. The concentration of ALA was determent by LC-MS/MS with a Varian 320 in negative mode. For the HPLC method an Atlantis T3, 3 µm, 150 × 2.1 mm column from Waters GmbH was used. The isocratic gradient was 30% 0.1% acetic acid and 70% acetone with a flowrate of 0.3 ml/min and an injection volume of 40 µl. The runtime was 4 min with a retention time of ALA at 2.1 min. Several post-hoc analyses were performed. Correlations of biomarkers and migraine intensity were examined visually and Spearman's rank correlation coefficient was calculated. As measure of migraine intensity, the number of migraine days and the MIDAS score at baseline were considered. Subgroup comparisons were performed between patients with and without migraine prophylaxis, between patients with and without acute migraine attack at baseline ± 2 days and between MA and MO. Subgroups were tested for a difference using Wilcoxon's rank sum test (continuous outcomes) and Fisher's exact test (frequencies).
In accordance with the exploratory nature of the analyses, p values should not be interpreted as confirmative, but can be useful in identifying hypotheses worth of further investigation. In accordance with recent statistical guidelines, the term 'statistically significant' is not used (following the strong suggestions made in the ASA Editorial on 'Moving to a world beyond p < 0.05' 77 ).All analyses were conducted using the statistical software package R 78 .

Results
Study population. Thirty-two patients were included in the study 4,27 . The mean age was 34 ± 10.8 years.
Twelve patients had migraine without aura (MO) and 20 migraine with aura (MA). Patient characteristics and demographics' information are shown in Table 1. Eleven patients were using at least one stable migraine prophylaxis (no changes within at least 3 months prior to study onset) (see Table 2 for the migraine preventatives used).

Summary statistics.
Summary statistics of all endpoints are given in Table 3 For ALA and lactate, the majority of patients had abnormally low values (28/32 (88%) and 23/32 (72%) respectively). Only two patients' lactate levels were too high. For one patient, an extremely high level of ALA (13.25) was measured. For PerOx half of the patients (46.9%) had abnormally high values. For thiols and TAC about one third of patients had abnormally low values (31.2% and 37.5%, respectively). For HbA1c about 20% of patients (21.9%) had abnormally low values and no one had an HbA1c above 5.6%. For oxidated LDL, a very high, abnormal level was measured in one patient, while for all other patients, the levels were in the normal range.

Correlations of mitochondrial function biomarkers and migraine severity.
We found no indication for a correlation of the 7 mitochondrial function biomarkers with MIDAS score or number of migraine Table 1. Summary statistics of patient characteristics. Categorical variables are summarized as frequencies and percentages (%), numerical variables are summarized by mean and one standard deviation (sd). SD = standard deviation, m = meter, kg = kilogram.

Variables All patients
Age, mean (SD) 34 (10.8) Female, N (%) 28 (87.5) Male Comparison between patients studied during or outside of an attack. Summary statistics of absolute levels of the mitochondrial function biomarkers and the frequencies of patients with abnormal values according to acute migraine attack at baseline (baseline visit ± 2 days) are presented in supplementary information section 1.3. Most patients presented with acute migraine at baseline; for one patient this information is missing. Our data provide no evidence for any difference between these two groups.
Comparison between patients with and without aura. Summary statistics of absolute levels of the mitochondrial function biomarkers and the frequencies of patients with abnormal values according to aura are presented in supplementary information section 1.4. We found no evidence for differences between patients with or without aura, neither in the absolute values of the biomarkers nor in the proportions of patients with abnormal values.

Discussion
We have shown that apart from oxLDL and HbA1c most other markers of mitochondrial functioning showed abnormalities in a significant proportion (> 30%) of the patients examined.

ALA.
To the best of our knowledge ALA levels have not previously been determined in migraine. Almost 90% of patients in this sample had abnormally low values of ALA. ALA, also known as thioctic acid, is an eightcarbon, sulfur-containing compound that functions as a water-and fat-soluble antioxidant 79,80 . It can directly (by removing reactive species) and indirectly (by chelating transition metal ions) reduce oxidative stress 79,80 . The human body can synthesize small amounts of ALA 79 . ALA also plays an important role as co-enzyme in energy metabolism [79][80][81] . Furthermore, it is able to regenerate other antioxidants, such as vitamin C and E, CoQ10, it increases intracellular glutathione and activates endogenous antioxidant systems [82][83][84] . Apart from its anti-oxidant action, ALA seems to assist weight loss 85 , increase insulin sensitivity and decrease blood lipids 86 . All of these mechanisms are probably migraine relevant. Interestingly, ALA supplementation (300-600 mg) per day has been shown to significantly reduce migraine attack frequency, severity and duration [36][37][38] , which seems to align with our findings. Further research is needed to see, whether this finding is specific to our medium-high Table 2. Types and frequencies of migraine prophylactic treatments. Note that more than one type of prophylactic drug could be used.  www.nature.com/scientificreports/ frequency episodic migraine population or a general characteristic of migraine or even a general characteristic of other (neurological) diseases with a mitochondrial/oxidative stress component. Additionally, as ALA measurements are less established and standardised than other markers, these results should be replicated in a larger cohort, with a different laboratory and using a control group. Should this finding be replicated and migraine specific, ALA might represent a potential biomarker.
TAC . Serum (or plasma) concentrations of different antioxidants can be measured separately, but since the measurement of different antioxidant molecules individually is impractical and costly and their antioxidant effects are additive, the total antioxidant capacity of a sample is typically measured, and this is typically referred to as total antioxidant capacity (TAC), total antioxidant status (TAS) or other synonyms, which will be used interchangeably. Almost 40% of our patients had abnormally low TAC being in line with results of previous research. A study on 75 MO patients demonstrated that the levels of total antioxidants were decreased and the levels of total oxidants and the oxidative stress index were increased 55 . Another study found TAC to be significantly reduced in migraineurs compared to controls 66 . TAC levels increased after successful prophylactic treatment compared to the baseline, irrespective of treatment modality (rTMS versus amitriptyline) and the increase correlated with  www.nature.com/scientificreports/ treatment success 66 .We assume higher TAC with lower migraine severity, less recent oxidative stress exposure, and increased distance to previous and future migraine attack. These assumptions would have to be validated in future research.
PerOx. Lipid peroxidation is the oxidative degradation of lipids via free radical damage of the lipids in cell membranes, polyunsaturated fatty acids in particular. The end products of lipid peroxidation are reactive aldehydes, such as 4-hydroxynonenal (HNE) and malondialdehyde (MDA). Free radicals cause increased accumulation of these lipid peroxidation by-products in the blood. About half of the patients had abnormally high total PerOx levels, being in line with previous research. Several studies have found serum levels of MDAs to be significantly elevated in migraine patients 56,67 , even in the interictal phase 87 . oxLDL. Oxidized low-density lipoprotein (LDL) is a harmful type of cholesterol that is produced when normal LDL cholesterol is damaged by chemical interactions with ROS. All but one patient had normal levels for oxLDL, which is in contrast to the study of Bernecker et al. that found highly significantly elevated levels oxLDL in female migraineurs 57 . This result could be due to differences in study population, as the migraineurs of the Berecker et al. study tended to have metabolic syndrome and had generally higher BMIs as our migraine patient population.
Thiols. The term "thiol" refers to organic compounds containing sulfur (in form of the functional group -SH, the thiol group). Thiol groups are able to destroy ROS and other free radicals by enzymatic as well as nonenzymatic mechanisms 88 . Total thiol levels have previously been used to evaluate excess free radical generation, both in physiological and pathological conditions 89 . Protein thiol levels in serum have been shown to be a direct measure of the in vivo reduction/oxidation (redox) status in humans, because thiols react readily with ROS to form disulfides 76 . Thiol redox homeostasis plays an important role in neurogenerative diseases 90 and in nine other categories of human disorders serum protein thiols have been found to be significantly reduced compared to healthy controls 76 . About one third of patients had abnormally low serum thiol levels, but this seems to be in line with previous research. A larger study found significantly reduced thiol levels in 151 migraine patients (74 MO, 77 MA) compared to 70 healthy controls and there was a negative correlation with migraine disability 61 . A negative correlation between the levels of total thiols and the duration of the headaches has also been demonstrated 55 . However, others studies found no significant difference in thiol groups between patients and controls, even during attacks 68 and one study even found higher total (-SH + -S -S-) & native thiol (-SH) levels in serum of migraineurs, but this did not correlated with disease severity or migraine type 63 . Recent exposure to oxidative stress, migraine severity, time in the migraine cycle and similar aspects could explain the different results.
HbA1c. HbA1c (glycated hemoglobin) is an indication of the average blood glucose levels over the last two to three months. Just over 20% of patients had abnormally low HbA1c levels and none of them had HbA1c levels that were above 5.6%. To the best of our knowledge HbA1c has rarely been looked at in migraine. One study found no significant difference in HbA1c levels between CM, EM and healthy controls 19 .
However, magnetic resonance spectroscopy (MRS) studies in migraine have consistently shown abnormalities of mitochondrial oxidative phosphorylation (OXPHOS), such as hypometabolism between 3-9 and during migraine attacks 10 , in the resting brain and in the muscle following exercise 3,11,12 . A 16% decrease of absolute ATP levels in migraine without aura patients was also demonstrated interictally using 31P-MRS 13 . These findings are supported by early studies showing that metabolic changes induced by fasting, glucose or insulin administration can trigger migraine attacks; e.g. a 50 g glucose tolerance test (GTT) after a 10-h fast triggered a migraine in 6 out of 10 migraine patients reporting attacks associated with fasting 14 . Abnormal metabolic responses were also reported in GTT studies 14,15 and interictal impaired glucose tolerance and insulin resistance has been reported in various other studies [16][17][18][19][20] . While only 20% of our migraineurs had abnormally low HbA1c levels, all levels tended to be on the lower side, despite reported higher carbohydrate diets. As HbA1c levels correspond to an average blood sugar measurement, low average values despite probable highs after carbohydrate rich meals could be an indication that there might be lows as well. This would be in line with previous neuroimaging and GTT research results, but it is speculation only and these assumption need to be confirmed by future research.
Lactate. Lactate is typically measured to assess tissue oxygenation, arising from either decreased oxygen delivery or a disorder in oxygen use, both of which lead to increased anaerobic metabolism and increases in lactate levels. In certain types of migraine, especially migrainous stroke, elevated serum lactate and pyruvate levels have previously been reported 46,47 . In contrast to this, only 2 patients had abnormally high serum lactate levels in our cohort and over 70% of patients serum lactate levels were abnormally low.
While there is little data on serum lactate levels in migraine, data on brain lactate analysed with 1H-MRS have also been shown to vary due to patient selection (see review by Reyngoudt et al. (2012) for details 8 ). Elevated brain lactate levels were found in some studies of MA 91,92 , but not in MO [93][94][95][96] . Occipital baseline lactate levels were increased in patients with visual auras, but not in those having complex neurological auras. By contrast, during photic stimulation lactate increased significantly in the latter, but not in the former 91 . Stimulus-induced lactate increases are physiological 97 and can be explained by the neuron-astrocyte lactate shuttle 98 . Hence, their absence in migraine patients, whose neuronal activation is energetically more demanding 99 , could be considered pathological and might be contributing to an energetic crisis.
To the best of our knowledge, no recent studies have looked at baseline serum lactate levels in episodic migraine patients or subgroups thereof. More research is needed to replicate this finding; in particular a study www.nature.com/scientificreports/ combining lactate level quantification in the cortex with that of the periphery and with brain energetics seems warranted. We can only speculate as to why lactate levels were predominantly low in the majority of our patients. They all came rested, but fasted overnight to the trial site. Decreased baseline lactate levels might be a sign of increased cerebral lactate consumption and an indicator of an increased cerebral energy demand of the migraine brain, as in addition to ketone bodies, lactate constitutes the only other major alternative brain energy substrate from glucose and is used especially during times of high metabolic demands or hypoglycemia 100 . A study using 13C-L-lactate and magnetic resonance spectroscopy suggested that the contribution of plasma lactate to brain metabolism can be up to 60% 101 , which is very similar to ketone bodies. It could also be a sign of decreased lactate synthesis as demonstrated with 1H-MRS 91 . In summary, we have shown that apart from oxLDL and HbA1c most other markers of mitochondrial functioning are abnormal in at least > 30% of the patients examined. As oxidative stress is a complex mechanism including different sources of ROS and various pathways, differing results in previous research may at least be partially caused by different oxidative stress parameters examined, e.g. MDA versus HNE, as well as by different study groups investigated, e.g. adults versus children, MA versus MO, females versus males, and differences in migraine severity, recent oxidative stress exposure and the time within the individual migraine cycle, where measurements were taken. Genetic research examining oxidative stress related genes in larger homogenous migraine cohorts could be interesting future research that would hardly be influenced by these factors.
Our data provide no evidence for correlations between any of the seven mitochondrial function / oxidative stress markers and migraine severity. This could be due to our sample population being fairly homogenous or the sample size being too small. In addition, we found no evidence for an effect of migraine prophylaxis. This is not surprising, since patients were still suffering from a substantial number of migraine days/months despite the prophylactic treatment (5-14 days/months), suggesting that the critical migraine pathophysiological mechanisms remained active. Furthermore, no evidence for an effect of a preceding or subsequent migraine attack has been found. This might be due to only 5 patients being migraine attack free within 2 days before and after the venous puncture, making an analysis of the potential impact of an attack difficult. We also found no evidence for a difference between MA and MO patients. For a randomly selected migraine cohort mainly recruited via public advertisements, the number of MA patients was unusually high (62.5%) in our study population. We can only speculate as to why this might be the case. Since participants were part of the 9 months MigraKet intervention trial 70 , it seems plausible that MA patients might have been more motivated to take place in such a lengthy trial and this led to the observed over-representation.
While we found no correlation between these mitochondrial function/oxidative stress markers and disease severity, differences in methodologies used and patient characteristics, recent oxidative stress exposure and also time in the respective migraine cycle is likely to play a role. Future research examining these markers at different time points during the migraine cycle and in different migraine types would be interesting.
The most important limitation of this study is the absence of a matched control group. While abnormally low levels in 90% of patients in the case of ALA are likely to be of importance, we cannot be sure that PerOx, TAC and thiol level findings would have been significantly different from controls. Future research is needed to replicate these findings in the presence of a control group. Secondly, the sample size was fairly small, in particular with regards to the correlation analyses. In addition, one third of patients was using a migraine prophylaxis. While our data provide no evidence for an effect of migraine prophylaxis, the inclusion of patients who are using a prophylaxis is not ideal.

Conclusion
In conclusion, this study provides further support for metabolic abnormalities in migraine, in particular the role of increase oxidative stress and decreased antioxidant capacity respectively in migraine pathophysiology. The peripheral markers assessed here could easily be examined in most doctor's offices and might assist personalised migraine treatment that targets oxidative stress and mitochondrial functioning; however, further research is needed to replicate these findings, ideally in the presence of a control group.

Data availability
For inspection purposes, insight to the original data will be permitted to the members of the appropriate authorities and also for the members of the local ethics committee, EKNZ.