Our discussion is intended to examine the overlapping and complementary preclinical and clinical evidence to support human imaging findings in patients with MO and patients with MA. The strength and weakness of the data will be considered. When appropriate, recommendations will be made for how to proceed to advance the translational evidence from animal to human.
Migraine without aura
Evidence of macrophage involvement
Preclinical findings suggest that macrophages participate in a delayed inflammatory response that may take place in the meninges in migraine [7, 37]. This was investigated using the nitric oxide-donor glyceryl trinitrate (GTN), which induced migraine attacks in 70-80% of patients with migraine when given intravenously . When given to rodents, GTN activated dural macrophages along the middle meningeal artery to express proinflammatory and nociceptive inducible nitric oxide synthase (iNOS), interleukin-1β (IL-1β), and interleukin-6 (IL-6) . The macrophages activated with a delay of 2 hours, comparable to the timing of migraine headache induced after GTN infusion [7, 39]. As histological examination is not feasible in humans, clinical studies examined monocytes in blood taken from the jugular vein during spontaneous migraine attacks instead, finding increased iNOS expression and higher levels of IL-1β and IL-6 [40, 41].
To localize a site for this monocyte/macrophage activation in patients with MO, an imaging study subsequently used the MRI contrast agent USPIO, which is engulfed by activated macrophages . The study looked for lateralized differences in patients who experienced unilateral migraine headache after cilostazol ingestion. However, the study found no differences in side-to-side signal intensity in the middle cerebral artery (MCA) or cavernous segment of the internal carotid artery (ICAcavernous), for vascular territories supplied by the anterior cerebral artery (ACA), MCA, or posterior cerebral artery (PCA), for the pons or thalamus, or upon visual inspection of the brain parenchyma or dura mater. While patients who did not receive sumatriptan treatment for the attack had a higher uptake than patients who received treatment, this was investigated as a post-hoc analysis. It is possible that sumatriptan treatment could have attenuated macrophage activity in the primary analysis.
Of note, since the study did not compare the uptake of USPIO between patients with MO and HC, nor between during and outside an attack, no inferences can be made about changes in macrophage activity during attacks. To analyze attack-specific macrophage activity, uptake could be compared between patients who did and did not experience a migraine attack after receiving a migraine provoking substance, or between spontaneous migraine attacks and the interictal state. Furthermore, the timing of the study with scans 27 hours after transfer infusion, may not have permitted detection of transient changes in the beginning of the attack. Finally, statistical testing was not possible for the dura mater or the middle meningeal artery, where macrophage activation was initially implicated in preclinical studies.
Evidence of vascular permeability
Most imaging studies in MO have examined vascular permeability that could be the consequence of inflammation. In rats, antidromic trigeminal activation leads to release from C and Aδ fibers of inflammatory neuropeptides which induces dural plasma protein extravasation . Neuropeptides such as substance P directly induces extravasation [42, 43], while others, e.g. CGRP, do so indirectly [44,45,46] by stimulating mast cell degranulation and histamine release [47, 48]. GTN infusion in rats caused a similar, but delayed plasma protein extravasation, suggesting an identical response could occur in patients with migraine . Vascular permeability and BBB function could also deteriorate due to the activity of matrix metalloproteinases (MMPs), extracellular proteolytic enzymes that regulate inflammation and disrupt the BBB [49, 50]. While early findings suggested MMPs to be dysregulated in patients with MO [51,52,53], findings of increased MMP-9 levels during MO attacks [53, 54] could not be replicated in a study measuring in blood from the external jugular vein .
Neuroimaging offers the opportunity to visualize this extravasation in patients with MO. Two initial clinical cases reported an increased uptake in meningeal 99Tc-HSA ipsilateral to the patient’s migraine headache. The uptake was increased three hours after tracer injection, and not on early acquisition [12, 27], which suggests that the uptake corresponded to extravasation of the tracer and not only hyperemia . However, how vascular changes evolve over time during migraine has not specifically been evaluated. Another study measured high parasellar uptake of gallium , whose uptake is enhanced by increased blood flow and vascular membrane permeability . However, the uptake was also high in cluster headache, and the region is not relevant for migraine pathophysiology. To verify whether plasma protein extravasated during migraine attacks, structured human imaging studies subsequently used the transfer rate of gadolinium or hydrophilic molecules of a similar size to investigate and quantify BBB permeability [11, 16, 28]. However, scans of 25 MO patients during attacks (19 spontaneous, 6 GTN-provoked) found no difference in BBB permeability for gadolinium or the migraine treatment dihydroergotamine compared to the interictal state [11, 28]. Another study found no difference between patients with MO interictally and healthy controls, but substantial variation in the gadolinium transfer rate limited interpretation of this finding .
Importantly, the structured studies lacked the statistical power to detect minor differences in blood-to-brain leakage of gadolinium [11, 16], which may not be sufficiently sensitive to detect minor changes in vascular permeability. Furthermore, the structured studies could not directly examine the meninges, since the resolution of standard MRI is too low to explore the meninges specifically. However, a recent case series examined intradural vascular gadolinium uptake using vessel wall MRI during and outside attacks of migraine (mixed MO and MA). The time from headache onset until scan was not reported but was <24 hours for all patients . The study did not find vessel wall enhancement in intradural intracranial vessels, but all participants had consumed anti-inflammatory analgesics or triptans before their scan, and therefore one cannot exclude that they may have quenched an inflammatory signal. Finally, no statistical comparison was performed, and the case series did not include healthy controls.
Migraine with aura
Animal models of migraine with aura suggest a proinflammatory role of cortical spreading depression (CSD), the neural correlate of aura symptoms. In neurons, CSD activates a distinct inflammatory pathway by increasing Pannexin 1 megachannel opening and caspase-1 activation, subsequently resulting in neural high mobility group protein (HMGB1) release. This stimulates nuclear factor-κβ (NF-κβ) activation in astrocytes, inducing transcription of cytokines and proinflammatory enzymes and headache behavior in animal experiments .
Evidence of macrophage involvement
In rodents, manipulating the cortex with electrodes triggered CSD, which activated meningeal macrophages [58, 59]. In the dura, macrophage activation was delayed by 20 minutes, a time frame similar to the delay from aura onset until headache onset in the majority of patients with MA . Macrophages may release cytokines such as IL-1β and tumor necrotizing factor-α (TNF-α) that are reported as elevated after CSD  and in patients with MA . These could sensitize meningeal afferents directly (e.g. IL-1  and TNF-α ) or indirectly by increasing CGRP release .
Though no imaging studies directly examined macrophage activation in patients with MA, one study reported increased uptake of the tracer 11C-PBR28 in patients with multiple attacks of migraine with visual aura within the parameningeal tissues (Table 3). TSPO is the peripheral benzodiazepine receptor and upregulates during inflammation in several cell types, including macrophages. Enhanced uptake was associated with the total number of visual auras in the preceding 4 weeks . In these patients, averaging 8 attacks over the prior 30 days, enhanced tracer uptake persisted for at least 18 days after the last attack. It is unknown as yet whether this uptake occurs in migraine without aura .
Complementary to the above, a recent study using micro-computed tomography (μCT) demonstrated the existence of microvascular channels in human skull. These channels allow leucocytes derived from skull bone marrow to migrate towards the brain to reach the meninges , as demonstrated in chemical meningitis and stroke. To explain the enhanced 11C-PBR28 uptake in the bone marrow, it has been posited that inflammatory signal(s) generated in cortex following CSDs (e.g., cytokines and/or HMGB-1) are released and taken up by microvascular channels to reach the bone marrow. Within the bone marrow, these signals provoke the migration of myeloid cells towards the meninges overlying the occipital cortex, the source of CSD in visual auras. Indeed, the demonstration of enhanced tracer uptake in visual cortex overlying meninges, and adjacent bone marrow supports the above formulation after multiple migraine with aura attacks . Whether the findings and formulation relate to the pathogenesis of recurrent and frequent migraine with aura attacks remain to be determined.
Evidence of microglia involvement
Animal models of migraine with aura suggest neuroinflammatory activation of microglia. As delineated above, cortical spreading depression initiates a complex cascade involving pannexin 1 channel opening with subsequent microglial activation. Like macrophages, activated microglia may release proinflammatory cytokines [67, 68] which can sensitize perivascular nociceptive afferents. In addition, activated microglia express TSPO [6, 69], which exhibited sustained upregulation in animals after CSD, when examined with a PET tracer .
Similar pro-inflammatory upregulation was replicated in patients with MA interictally with the TSPO tracer 11C-PBR28 , suggesting pro-inflammatory microglial activation or recruitment . TSPO was upregulated in several regions previously implicated in MA, including occipital striate and extrastriate visual cortex, somatosensory cortex, insula, thalamus, and in the spinal trigeminal nucleus . There was an association between frequent migraine attacks and uptake of the tracer in the human neuroimaging study, which may reflect a cumulative impact of multiple CSDs on microglial activation or number, sufficient to detect clinically. Similar cumulative effects of multiple CSDs have been observed preclinically .
Of note, 11C-PBR28 binds non-specifically to activated macrophages, monocytes, neutrophils, dendritic cells, mast cells, and adipocytes besides microglia . Although this shortcoming limits the ability to dissect the contribution of individual cell types in the inflammatory process, it does have the advantage of detecting a composite of the inflammatory process in multiple cell types, thereby enhancing the overall signal intensity. The development of more cell type-specific ligand will add another dimension to these new and important findings.
Evidence of vascular permeability
In animal models of migraine, CSD induces extravasation of plasma proteins from dura mater vasculature, by activating trigeminal afferents [3, 57, 72]. In rats, trigeminal activation after CSD may activate and upregulate the BBB degrading enzyme MMP-9, which results in plasma protein leakage . Though attacks in patients with MA have been associated with increased MMP-9 concentrations , it is uncertain whether MMP-9 levels differ between patients with MA interictally and healthy controls [51, 54].
In MA, a BBB breakdown has been visualized in a few clinical cases with contrast extravasation on MRI [17,18,19,20,21]. However, these were prolonged auras [17, 21] or atypical cases in which differential diagnoses such as hemiplegic migraine, cerebral amyloid angiopathy or transient headache and neurological deficits with cerebrospinal fluid lymphocytosis (HANDL) syndrome cannot be completely excluded [18,19,20]. In a structured study of typical cases of MA in 19 patients, Hougaard et al. found no increase in BBB permeability for gadolinium during the migraine headache . However, comparisons to healthy controls were not made, and there was a lack of statistical power to detect minor permeability changes. In this study, patients were studied during hypoperfusion or hyperperfusion phases. Interestingly, in the case published by Rostein et al., the observed increase in BBB permeability observed was contemporaneous with ipsilateral hypoperfusion, which would suggest a BBB disruption unrelated to hyperemia . To our knowledge, this is a unique observation.
Hemiplegic migraines (HM) could be sporadic (SHM) or familial (FHM). Familial HM can be due to several genetic mutations, which all increase the susceptibility to CSD. In patients with HM, CSD could induce more pronounced extravasation of plasma proteins than in classical aura, since preclinical studies suggest multiple CSDs amplify plasma protein extravasation . This corresponds to the severe neurological paresis characterizing the disorder.
One structured imaging study (N = 2)  and 4 case reports examined changes in vascular permeability in FHM, while one case report examined changes in vascular permeability in SHM (Table 2). Contrast agent extravasated in 5 out of 13 hemiplegic migraine attacks. These were always prolonged auras, and the extravasation of contrast agent was extensive, sometimes associated with cortical oedema, and was seen in a phase of hyperemia [22,23,24,25,26].
The increased vascular permeability reported in human neuroimaging studies of FHM [22,23,24] (Table 2) supports preclinical findings of leaky vessels after CSD . However, there is a need for structured studies using sensitive methods to detect extravasation, to confirm these findings in a broader population of patients with FHM. Studying specific components of CNS inflammation, such as macrophage or microglial activation, as has been done in patients with MO and MA, would also be essential to determine if specific inflammatory pathways are involved in FHM.
Most human imaging studies have focused on changes in vascular permeability to detect a key feature of inflammation in brain and surrounding tissues. In MA, neuroimaging studies have found neuroinflammatory activity in the cortex (possibly microglial in origin) and parameningeal tissues (possibly monocytic in nature). Imaging signs of neuroinflammation in MO have been less convincing. Major disruption of the BBB has not been observed consistently, although subtle or transient changes cannot be ruled out. The presence of a BBB breakdown is mainly seen in cases of atypical and prolonged auras as in hemiplegic migraine. In the classic forms of aura this observation is exceptional, and in MO it is uncertain. Furthermore, it is not clear whether the observed barrier breaks are due to hyperemia or inflammation or both. Only one observation showing extravasation at a hypoperfusion phase supports the second hypothesis . It is possible that the common methods used lack the sensitivity to detect subtle disruptions of the BBB . Future studies should apply more sensitive methods to detect extravasation, e.g. dynamic contrast-enhanced MRI (DCE-MRI) with longer acquisition times, or methods other than those based on gadolinium tracer, e.g. detecting molecular diffusion of intra- and extra-cellular water with T1 and T2 mapping.
Novel tracers targeting molecules upregulated during specific inflammatory processes, such as the leukocyte trafficking molecule urokinase plasminogen activated receptor, could then define components of an inflammatory response further. However, there are still many relevant targets in migraine for whom tracers have not been developed yet.
Other methodological considerations apply to the study of neuroinflammation in migraine in general. Timing of scans with regard to migraine phase may be essential to detect transient changes; serial scans during the migraine prodrome, headache, and resolution should be conducted to determine the time window of such transient changes. Importantly, temporal considerations may depend on the individual inflammatory event. For example, preclinical and clinical studies suggest extravasation is most likely to occur a few hours into migraine attacks, whereas microglial activation may not begin until a few days after CSD . Furthermore, studies should be sufficiently powered to detect discrete inflammatory changes in macrophage activity or BBB permeability.
Meningeal and vascular inflammation is relatively unexamined. Preclinical studies suggest this location to be a prime candidate for inflammation in MO . The dura lacks a BBB and has a composition and resident inflammatory cells distinct from those in the brain. Imaging data show that the middle meningeal artery (MMA) dilates specifically on the pain side in cilostazol-induced attacks . However, several studies with MRI contrast agents did not report whether there was an overt dural enhancement or not, but this would probably have been mentioned if present. Future studies should explicitly report whether dural enhancement was observed or not. With future high-resolution imaging techniques, it may be possible to analyze the meninges separately. Neuroimaging studies should also continue to explore inflammatory changes in the cortex of patients with MA, particularly in relation to CSD.
Establishing the role of inflammatory pathways in migraine pathophysiology could help identify locations and targets for specific anti-migraine treatments and contrast the different migraine types. For example, the extent to which the BBB is altered or not in migraine is essential to determine the site of action of current and future migraine treatments. Neuroinflammatory imaging signals may also become possible future biomarkers for migraine. Therefore, it will be critical to determine whether microglial activation occurs in MO (especially whether the tracer 11C-PBR28 shows similarities in MO and MA), if macrophage activation occurs in MA, and if either occurs in FHM.
In summary, brain imaging circa 2022 provides an essential tool to understand the natural history of migraine and is a viable way to reconcile emerging preclinical and clinical data. It also holds great promise for discovering and interrogating key biological processes in human brain underlying this enigmatic neurovascular disorder. At this point, more studies are needed along with more specific and selective markers of cells and tissues as well as efforts to harmonize the protocols and newly acquired data sets between laboratories.