Background
Migraine headache is a disabling neurological disorder that carries with it a significant burden which crosses over multiple domains on both personal and societal levels, including the negative impact on work productivity, family life, mental health, leisure activities, and the economy [
1‐
3]. The yearly prevalence of migraine is close to 15% in the general public. A systematic analysis of the 2016 global burden disease (GBD) study estimated that just above one billion people in the world had migraine in 2016 [
4]. Based on the latest GBD study in 2019, migraine headache is the second most frequent cause of global disability in all age groups combined (regardless of sex), and first most frequent cause of global disability in 15-49 year-old women, specifically [
5]. Persons with migraine disease suffer from either episodic migraine (EM), defined as recurrent headache disorder manifesting in at least five attacks lasting 4–72 h with typical characteristics, or chronic migraine (CM), diagnosed when the individual has headache occurring on 15 or more days/ month for more than three months, which, on at least eight days/month, has the features of migraine headache [
6‐
8].
Both environmental and genetic factors have been implicated in the development of migraine [
6,
9]. The pathophysiology of migraine is not fully understood, and multiple theories have been propounded to explain this pathogenesis including trigeminovascular pathway activation, vascular dysfunction, cortical spreading depression (CSD), and neuroinflammation [
6,
10,
11]. Migraine attacks have been suggested to have either a peripheral origin, in which first order trigeminovascular neurons are activated, or a central origin, where CSD and hypothalamic/brainstem activation play a major role [
9]. Pre-clinical data have suggested that during a migraine attack, neuropeptides such as calcitonin gene-related peptide (CGRP) and pituitary adenylate cyclase-activating peptide (PACAP) are increasingly released from afferent nociceptive fibers and act on vascular smooth muscle cells of meningeal arteries to trigger a cascade of intracellular signaling events mediated by cyclic adenosine monophosphate and cyclic guanosine monophosphate that ultimately lead to opening of potassium channels and vasodilation [
7,
9]. In addition, the release of CGRP and PACAP can lead to degranulation of meningeal mast cells, which in turn can increase levels of proinflammatory mediators such as prostaglandin E2 (PGE2) [
12,
13]. It has also been proposed that CSD can trigger mast cells to excrete such proinflammatory molecules [
14]. PGE2 enhances inflammatory pain and possibly nociceptor sensitization, and can itself increase the release of CGRP in preclinical models [
13,
15]. Moreover, the endothelial cells of the meningeal arteries can also release vascular endothelial growth factor (VEGF) that increases vascular leakage, nitric oxide (NO) synthesis, and mobilizes more macrophages and neutrophils to adjacent tissues [
16,
17]. These immune cells excrete cytokines that promote neuronal sensitization [
16]. In clinical models of migraine, infusion of CGRP, PACAP, PGE2, or glyceryl trinitrate appeared to provoke headache attacks [
7,
9,
13]. Also, studies have indicated that neurotrophins such as nerve growth factor (NGF) and brain-derived neurotrophic factor (BDNF) may have a role in modulating nociceptive pathways [
18]. To be specific, inflammation could give rise to increased production of NGF which can in turn increase the expression of pain receptors such as transient receptor potential vanilloid receptor 1 (TRPV1) on peripheral nociceptive fibers and the excretion of CGRP and BDNF [
18]. Other mechanisms that have been implicated in migraine are increased sensitivity to NO, serotonin secretion from platelets, increased levels of homocysteine, decreased vitamin D levels, and oxidative stress [
6,
11,
19‐
23]. Although the effectiveness of various prescribed medications in dampening migraine symptoms is generally acceptable, adverse effects such as gaining or losing weight, drop in blood pressure, decreased awareness, and sleepiness/lethargy restrict the use of such treatments. Moreover, suboptimal treatment of EM in persons with migraine disease can facilitate the transformation of EM to CM, which is a more burdensome state [
6]. Therefore, identifying more tolerable and efficacious options is indispensable [
24‐
26]. On the other hand, probing for biomarkers such as neurotransmitters, receptors, and inflammatory factors can potentially reveal mechanisms involved in development of migraine, provoking the attacks, and evolution of CM, and promote the development of novel and more effective anti-migraine medications. Taking this into account, we aimed to investigate potential biomarkers associated with migraine in EM and CM patients with a special focus on PGE2, VEGF, NGF and BDNF.
Discussion
With the aim to improve our understanding of the underlying mechanisms of migraine as a highly prevalent condition around the world, we sought to identify potential biomarkers involved in the pathogenesis of migraine. In line with this aim, we focused on the serum levels of PGE2, VEGF, NGF and BDNF in patients suffering from CM or EM. In addition to the detected lower levels of NGF, BDNF and PGE2 among EM patients compared to controls, NGF was the only biomarker in our study that showed significantly different (i.e., lower) serum levels in EM patients than that of CM patients, while none of the remaining three biomarkers investigated in our case control study were able to discern between CM and EM (although difference in PGE2 levels between EM and CM patients trended towards significance). Moreover, our results indicated that higher NGF and PGE2 serum levels have a moderate and weak positive correlation with headache frequency in migraine patients, respectively, while no significant correlation was replicated for BDNF and VEGF.
NGF is a neurotrophin mainly found in the limbic system [
31]. It is recognized to be involved in cognition, mood, protection of neurons, neuroplasticity and response to stress mechanisms [
32,
33]. Peripheral expression of NGF has also been associated with nociceptive sensitization; pain is conveyed from trigeminal ganglion to trigeminal nucleus caudalis through TRPV1 [
34,
35]. Our study demonstrated that peripheral blood NGF levels are significantly lower in EM patients compared to healthy controls and CM subjects. A study by Blandini et al. evaluating migraine patients, cluster headache patients, and controls corroborated our findings, as they showed a reduction in plasma and platelet levels of NGF in migraine patients in comparison to controls [
36]. In contrast, a study by Jang et al. on patients with CM and controls revealed that levels of NGF and other neuropeptides such as CGRP and substance P showed elevations in persons with migraine disease, a positive correlation was noted between NGF and these other neuropeptides, and levels of NGF and the aforementioned neuropeptides significantly correlated with intensity of pain [
37]. In a study conducted by Sarchielli et al. on patients with chronic daily headache with history of migraine and controls, elevated cerebrospinal fluid (CSF) levels of NGF were observed in patients with migraine [
38]. The same research group also demonstrated that CSF levels of NGF positively correlate with headache frequency, in line with our current study findings, but do not correlate with VAS scores representative of headache severity [
38,
39]. Another study by Sarchielli et al. also showed that CSF levels of NGF and glutamate are significantly higher in migraine patients compared to controls [
35]. Of note, a study by Martins et al. also showed no difference between migraine patients and controls in terms of plasma levels of NGF [
18]. Nevertheless, the inconsistencies in the literature regarding the levels of NGF in patients with migraine warrants the need for further investigations to reach a more definite conclusion.
BDNF is the most abundant neurotrophin found in different compartments of the nervous system [
40]. BDNF is involved in pain signaling and regulation, in addition to the roles in neuronal development and differentiation [
36,
41,
42]. With regards to pain, BDNF exerts a paradoxical, dose-dependent effect where low doses lead to hyperalgesia, while high doses result in analgesia [
43]. Its expression is associated with CGRP in trigeminal ganglion neurons. Researchers have suggested that these associations of BDNF may play a significant role in modulating susceptibility of patients to migraine [
44]. Our study showed that EM and CM patients have significantly lower levels of BDNF compared to healthy controls. We postulate that this finding might be explained by the hyperalgesic effect of BDNF in lower doses. Martins et al. and Blandini et al. have confirmed our BDNF findings in CM patients (compared to controls) [
18,
36]. However, most other studies in the literature have indicated the opposite. For example, Sarchielli et al. have indicated that CSF levels of BDNF are significantly higher in persons with migraine disease, CSF levels of NGF and BDNF positively correlate with each other, and BDNF levels in the CSF correlate with number of headache days in each month (although correlation was weaker compared to NGF and headache frequency) but not with headache severity [
35,
39]. Regarding their latter finding, our current study did not find a significant correlation between headache frequency and serum BDNF levels. In 2010, Tanure and colleagues demonstrated that BDNF serum levels are significantly elevated during migraine attacks [
42]. Another study also revealed higher levels of BDNF in migraine attacks compared to headache-free period and tension-type headaches [
45]. Regarding our finding of similar BDNF levels between CM and EM patients, we were not able to identify studies that have reached a similar conclusion as no study to our knowledge has reported on findings regarding direct comparison of BDNF levels between EM and CM patients. Regardless of the inverse or direct relationship between BDNF levels and propensity for migraine, these various levels of evidence emphasize the role of BDNF in nociceptive pathways. One explanation for the reduced levels of both neurotrophins (NGF and BDNF) in our study could be that a subgroup of the EM and CM patients had concomitant undiagnosed depression and therefore were not receiving antidepressants. It has been previously propounded that patients suffering from depression may have decreased levels of NGF and BDNF, and fluoxetine might increase those levels [
46].
In pathophysiology of migraine, release of vasoactive substances by meningeal and brain mast cells may possibly play a pivotal role because this can trigger trigeminovascular mechanisms leading to development of pain [
47]. VEGF is one of these substances which comprises an array of glycoproteins that contribute significantly towards cellular protection and angiogenesis. Additionally, these glycoproteins are regarded as potential proinflammatory cytokines [
48,
49]. Our study revealed higher peripheral levels of VEGF in both EM and CM patients compared to healthy controls. Rodriguez-Osorio X et al. [
50] have also reported that VEGF levels are significantly higher in patients with episodic migraines compared to controls. One proposed rationale for the increased VEGF levels in persons with migraine disease could be that this response is of a compensatory manner.[
50] Moreover, it could herald the onset of chronic endothelial dysfunction in apparently healthy individuals. In addition, SSRIs have been suggested to increase levels of VEGF both centrally (i.e., in the hippocampus and dentate gyrus) and peripherally in the blood, [
51,
52] and this could have been the case for our participants who were taking SSRIs or even SNRIs and TCAs. On the other hand, significant reductions in VEGF levels during the interictal period have been reported by Michalak et al. [
17] It is important to further investigate the association between VEGF and migraine in future studies. The importance of this association could be due to higher rates of cardiovascular accidents in patients suffering from migraine; increased risk of ischemic stroke in women and increased risk of myocardial infarction in men with migraine have been reported in the literature [
53,
54]. Some authors have speculated that this relationship is due to endothelial changes reflected in alterations of factors such as VEGF.
PGE2, as a member of the prostaglandin family, has been reported to be involved in underlying mechanisms of pain in migraine [
13]. The pain induced by PGE2 might possibly stem from the activation of TRPV1 receptors [
55]. Some studies have shown that infusion of PGE2 leads to headache in 83-100% of patients while this rate after prostaglandin F2-alpha infusion has been only 17% [
56‐
58]. However, our study showed that serum PGE2 levels are significantly lower in EM patients compared to controls, and while they are also lower in CM patients in contrast to controls, the difference is not statistically significant. Sarchielli et al. showed that internal jugular venous blood levels of PGE2 maximize within two hours of headache onset, plateau until hours 4-6 post-headache onset, and then decline [
59]. Mohammadian and colleagues showed that PGE2 levels in saliva and nasal lavage samples of migraine patients during the inter-ictal phase are comparable to that of controls [
60] Tuca et al. showed when saliva is collected during a headache attack from migraine patients, PGE2 levels are significantly higher than that of samples corresponding to the period between attacks [
61]. They also showed that PGE2 saliva levels declined in persons with migraine disease who consumed calcium channel blockers (CCBs) for two months, as opposed to those migraine patients who were given placebo, and this could explain the lower levels of PGE2 in our migraine study subjects who were taking CCBs at baseline. Similarly, Li et al. demonstrated that in migraine patients, cyclooxygenase-2 (COX-2; enzyme that gives rise to PGE2) levels are significantly higher in the attack period as opposed to the headache-free period, and comparable between controls and migraine patients during the inter-ictal phase [
62]. As close to 40% and 70% of our study participants in the CM and EM groups were taking NSAIDS, this could have affected PGE2 levels, especially if they consumed a selective COX-2 inhibitor in close proximity to the blood draw.
In our study, NGF and PGE2 serum levels positively correlated with headache frequency in migraine patients, while serum levels for BDNF and VEGF did not show such significant correlations. We were not able to find any study in the literature assessing the correlation between headache frequency and levels of
in vivo PGE2 and VEGF, and we found only two studies that evaluated the correlation between CSF levels (not serum) of NGF and BDNF and number of headache days [
38,
39]. Therefore, it is difficult to draw any meaningful conclusion. Nevertheless, based on our current findings and Sarchielli et al.’s studies, [
38,
39] there might be a possibility that higher levels of NGF and PGE2 could herald the “transformation” of EM to CM. If this is true, one could possibly consider a more prominent role of inflammation in migraine evolution/”chronification”. In 2018, we showed that serum levels of CRP and tumor necrosis factor alpha (TNF-α) did not correlate with frequency of migraine attacks [
63]. Most recently, however, we showed that serum levels of proinflammatory biomarkers including CRP, TNF-α, interleukin-6, NO, and malondialdehyde positively correlated with number of migraine headache days per month [
11,
12]. These recent findings of our research group and the postulated role for inflammation in CM support the role of inflammation in migraine evolution [
64].
For biomarkers in our study with significantly lower levels in EM and/or CM patients compared to controls (NGF, BDNF, and PGE2), one common possible explanation, at least for EM patients perhaps, could be that the minimum 72-hour lag between headache onset and blood collection may have been too long, and resulted in sizeable drops from peak levels during attacks. This may have also explained the comparability of serum levels of BDNF, VEGF, and PGE2 between EM and CM patients. One other explanation could be that although migraine patients and controls were age and sex-matched, they may not have been matched for silent proinflammatory states (for example, conditions with heightened levels of C-reactive protein (CRP) and other proinflammatory cytokines) that might influence NGF, BDNF, VEGF, and PGE2 levels. A combination of timing of blood draw in reference to most recent headache attack and confounding non-diagnosed proinflammatory comorbidities might possibly explain the significant difference of serum levels of NGF between EM and CM patients in our study. Finally, difference in study populations and laboratory assays used for measuring the biomarkers could have biased our results towards the opposite direction of some of the previously mentioned findings in the literature.
We believe that our study has several strengths; first, classification errors for EM and CM were at the minimum as diagnosis of eligible participants was confirmed by a neurologist specialized in headaches. Second, patients previously diagnosed with MOH were not included, hence decreasing confounding bias of medication effects on biomarker levels. Third, the one-month length of follow up for documenting headache characteristics of migraine patients was relatively robust. We also acknowledge that our study faced several limitations; the case-control nature of the study prohibited us from drawing conclusions about causality and directionality between migraine and NGF, BDNF, VEGF, and PGE2 peripheral blood levels. Our study was not powered to detect pre-defined differences between levels of biomarkers in EM and CM patients, and healthy controls as no formal sample size calculation was conducted. As mentioned above, we were not able to control for multiple confounding factors that might have influenced levels of biomarkers. It was not feasible for us to measure the concurrent CSF levels of our investigated biomarkers, so it should be clarified that peripheral levels of biomarkers in our study participants might not reflect changes in their central nervous systems. Distribution of females was different between EM and CM patients, although the difference was not statistically significant, and this may have masked any possible gender effect modification on serum biomarker levels.