This is the first study to demonstrate that microglial activation can be detected in patients with migraines, irrespective of their headache status. CX3CL1 in CSF thus provides a novel biomarker to study migraine pathophysiology in humans. We were furthermore able to clarify the state-dependent effects of serum t-Tau in the migraine cycle and found that its concentrations were elevated in the headache phase but not interictally, rendering it a potential biomarker of migraine attacks. T-tau elevations were not accompanied by an increase in other biomarkers for neurodegeneration, which argues against significant neuroaxonal deterioration in migraine patients. In line, the absence of an increase in GFAP in migraine patients, as well as the lack of an increase in NFL speaks against astrocytic or neuroaxonal damage in the CNS in migraine patients.

In contrast, CX3CL1 and t-Tau pose potential fluid biomarkers to study the pathophysiology of migraine and possibly inform clinical management.

Tau protein in peripheral blood samples has been tested as a marker of neuroaxonal damage in a variety of diseases [17]. Only one study investigated the significance of t-Tau in the serum of migraine patients in the interictal phase of the disease. Overeem et al. found increased concentrations of t-Tau in migraine patients compared with healthy controls [7]. Based on our data, we demonstrated that this difference is driven primarily by elevated concentrations in the headache phase of the disease. However, according to the paper by Overeem et al., there is no dependence of t-Tau release on the presence of aura symptoms [7]. The authors conclude that it is not cortical activity but neuronal pain-associated activity, e.g. in the trigeminal ganglion, that could be responsible for t-Tau release. Our finding of serum t-Tau elevation in the ictal phase of the disease supports this hypothesis. In conclusion, serum t-Tau could be a potential biomarker for disease activity, as it is already used in peripheral neuroinflammatory disorders such as Guillain-Barré syndrome [18]. Further studies are needed to evaluate the relationship between attack frequency, distance to migraine attacks, and dependence on specific migraine therapeutics and concentration of t-Tau protein in serum.

We demonstrated that t-Tau concentration is detectable only in serum and not in CSF. This offers room for pathophysiologically relevant speculation in the context of migraine attacks. It was speculated that t-Tau elevation in migraine patients represents functional changes in the trigeminal nervous system [7], but without further elucidation of a specific mechanism. Regardless of why t-Tau protein elevation occurs in migraine patients, the lack of difference in concentrations of other biomarkers of neuroaxonal degeneration such as NfL, GFAP and UCHL1 needs to be emphasized and refutes major neurodegenerative processes in migraine.

CX3CL1 in CSF but not in serum was significantly higher in migraine patients regardless of sample collection in the headache phase or interictally versus healthy controls. Even though there are some investigations in mouse models suggesting microglial activation in migraine, evidence in human subjects is lacking [19]. Most murine studies investigated purinergic receptors (e.g., P2X/P2Y), which is plausible given the metabolic aspects of migraine pathophysiology, including neuronal energy deficiency [20]. The involvement of the fraktalkine pathway was only recently suggested in another mouse model of migraine and was also associated with impaired neuronal signalling in thalamo-cortical networks [21]. We found that CX3CL1, the ligand to the fractalkine receptor, is increased in humans irrespective of the headache status. Based on the current evidence, it can only be speculated that increased microglial activity could represent a pro-neuroinflammatory trait, with implications for neuronal transmission in regions with dense microglial concentration such as the thalamus [21]. Additionally, microglia are thought to play an important role in the modulation of migraine, and this process may be responsible for the progression toward chronification [19]. Also, microglia activation has been associated with cortical spreading depolarization [22], a potential cause of microgliosis in migraine patients. Supporting this notion, patients undergoing CSF examination in the interictal phase of migraine in this study mostly presented with transient neurological symptoms and white matter abnormalities on MRI, which were eventually attributed to migraine with aura. It remains to be clarified whether CX3CL1 concentrations differ between patients with migraines with and without aura. Finally, neuron-microglia interactions play a role in pain generation associated with migraine [23, 24]. In our study, we detected an increase in CX3CL1 in comparison with healthy controls solely in the CSF, which is indicative of a central origin of CX3CL1.

Interestingly, we could not detect a difference in the levels of another microglial marker, sTREM2, between the groups studied. To date, no studies exist on the significance of sTREM2 in the context of migraine. Given that there are no indications of neurodegeneration in migraine, neither in neuroimaging nor in biomarker studies, a lack of elevations of sTREM2, which is associated with microglial activation in Alzheimer’s and Parkinson’s disease, is plausible [25]. It must be noted that micoglial activation is indeed not a simple binary state but rather a fine-tuned homeostatic process that may differ between the site of micoglial activation and its triggering receptor [26].

A main limitation of this study is the retrospective design, so that no more detailed information on further migraine characteristics could be collected, such as duration of the disease, or acute or preventive medication. On the other hand, CSF samples were specifically acquired in order to rule out other neurological disorders. Performing a lumbar puncture is nowadays rather unusual in migraine patients making it almost impossible to perform prospective CSF studies in this population. For the same reason, a higher number of cases is very difficult to achieve, even if this would certainly be desirable for a biomarker analysis, particularly with regard to the statistical significance between the individual groups. Body mass index (BMI) may be a covariate for at least NFL and GFAP; no values were available for this because of the retrospective design. For other covariates, such as age and gender, we controlled by choosing an age- and gender-matched control group.

Another limiting factor is the sometimes very long storage time of the samples before the above-mentioned biomarkers have been measured. On the other hand, it has been described that the measurement of t-Tau is reliable even after years of storage [27].

To assess the levels of the above biomarkers as diagnostic, other control groups would need to be studied. Since this study was exploratory in nature, larger studies with other patient cohorts would be desirable, specifically for t-Tau protein in serum.

To assess disease activity by a biomarker, longitudinal sample collections from the same patients would be desirable. This is also a goal of future studies.