Subsegmentation of the hippocampus in subgroups of migraine with aura patients: advanced structural neuroimaging study

Abstract

Background

This study investigated for a possible contributing role of hippocampus in the different clinical phenotypic manifestations of migraine aura.

Methods

Herein, patients were categorized as those with pure visual aura (MwAv), those who reported additional somatosensory and dysphasic symptoms (MwAvsd), and healthy controls (HCs). Neuroimaging data obtained using FreeSurfer-based segmentation of hippocampal subfields were compared between HCs and patients with migraine with aura, as well as between HCs and those with MwAv and MwAvsd. The average migraine aura complexity score (MACS) was calculated for each patient to investigate the correlation between hippocampal subfield volume and migraine aura complexity.

Results

Herein, 46 patients with migraine with aura (28 MwAvsd and 18 MwAv) and 31 HCs were included. There were no significant differences in the hippocampal subfields between HCs and patients with migraine with aura. The average MACS negatively correlated with the volumes of the left and right hippocampi, Cornu Ammonis (CA) 1, CA3, CA4, molecular layer, left granule cell layer of the dentate gyrus, hippocampal fissure, and hippocampus-amygdala transition area. The MwAvsd subgroup had significantly smaller whole hippocampal volumes in both hemispheres, as well as in both subicula, compared with the MwAv subgroup and HCs. In addition, the left molecular layer, right CA1, and hippocampal fissures were significantly smaller in the MwAvsd group than in the MwAv subgroup and HCs.

Conclusions

Smaller left and right hippocampal volumes, particularly of the subiculum/CA1 area, may play an important role in the pathophysiology of somatosensory and dysphasic symptoms in migraine with aura.

Introduction

Approximately 30% of patients with migraine experience migraine with aura (MwA), which exhibits diverse presentations such as complicated visual and somatosensory symptoms, dysphasia, and motor weakness [1]. Additionally, MwA patients with dysphasic aura symptoms are more likely to develop interictal impairment in cognitive processing [2]. Unfortunately, the reason some people experience pure visual aura (MwAv) during MwA attacks while others suffer abundant non-visual symptoms (MwAvsd) is unknown. MwA prevention and treatment will improve with a better knowledge of its intricacy [3].

Neuroimaging and electrophysiological studies have shown notable differences between MwAvsd and MwAv, suggesting that those MwA subgroups should be studied separately [2,3,4,5,6,7,8]. Compared to the MwAv group, the MwAvsd subgroup has different structural cortical features and functional connectivity in visual, somatosensory [3, 6, 8], and language processing regions [7, 8]. MwAvsd individuals may also have higher cortical excitability and cognitive impairment [2, 4]. The Migraine Aura Complexity Score (MACS) was introduced to assess the presence and quality of visual, somatosensory, dysphasic, and other higher cortical symptoms during MwA attacks [9] to stratify patients into more homogenous groups and investigate the connection between MwA complexity and brain changes. A higher MACS score indicates more aura symptoms beyond visual ones, that is a more complex aura. Thus, the MACS may aid neuroimaging studies of MwAvsd’s complex pathophysiology [10].

Interestingly, the hippocampus, which plays a key role in episodic memory and other cognitive functions [11], has only recently been investigated in migraine patients, and there are associations between hippocampus volume, migraine attack frequency, white matter lesions, and cognitive decline [12,13,14,15]. However, to our knowledge, the volume of the hippocampus in MwA patients has never been measured. The hippocampus may be involved in MwAvsd due to recent findings on its relationship with language cortical areas responsible for online language processing and production [16,17,18], and its role in functional integration of the visual and somatosensory cerebral cortex [19]. New hippocampal subfield segmentation methods may support this hypothesis [20].

Using structural neuroimaging, this study compared the volumes of hippocampal structures between healthy controls (HCs) and MwA patients, as well as between MwAv and MwAvsd subgroups and HCs. MACS and volumetric measurements of the hippocampus and its substructures were used to examine disease features, particularly aura symptoms. We hypothesized that the hippocampus region’s multimodal integrative role would contribute to migraine aura complexity based on previous findings. Consequently, MwAvsd patients may have more marked structural hippocampal abnormalities than HCs and MwAv patients.

Materials and methods

Participants

This was a comparative study of neuroimaging data obtained using FreeSurfer-based segmentation of hippocampal subfields from HCs and patients with episodic migraine with typical aura (ICHD-3, code 1.2.1). Most patients with MwA included in the study were from a previous migraine neuroimaging study cohort [7, 21]. However, this is our first attempt to analyze the volume of hippocampus and its subfields in this group of patients. Patients with MwA were recruited from among those who consecutively visited the Headache Center at the University Clinical Centre of Serbia. The sample size was based on the available data and previous literature [7, 21]. Additionally, because the focus of this investigation was on the subset of patients who experienced MwAvsd, we performed a sample size calculation (N = z² × (1 – p)/d²) based on a confidence level of 95% (z = 1.96), a margin of error (d) of 5%, and an estimated MwAvsd population proportion (p) of 1.5%, which showed that 23 or more participants were required for a confidence level of 95% for the measured values in the MwAvsd population. The MwAvsd population proportion was calculated according to the worldwide prevalence of migraine (14.7%) [22], the proportion of patients with migraine exhibiting aura (30%) and the proportion of patients with MwA exhibiting MwAvsd symptoms (visual plus somatosensory symptoms 22% + visual plus somatosensory and dysphasic symptoms 8.8% + visual plus dysphasic symptoms 3.1% = 33.9%) [23]. The inclusion criteria were as follows: patients (1) aged 18–55 years, (2) suffering from episodic MwA for more than 2 years before the enrolment in the study, (3) with at least 2 MwA attacks in the year before the enrolment in the study, 3) without a history of migraine preventive therapy for at least 1 year before the enrolment in the study, and (4) who provided consent for participation in the study. A trained headache specialist examined and interviewed all patients with MwA who agreed to participate in the study. Diagnosis of episodic migraine with typical aura was confirmed according to the 3rd International Classification of Headache Disorders criteria [1]. The exclusion criteria were as follows: (1) the presence of other types of migraine with aura (such as migraine with brainstem aura, hemiplegic, or retinal migraine), (2) the presence of any other neurological, cardiovascular, or endocrine diseases, (3) claustrophobia or inability to undergo magnetic resonance imaging (MRI), and (4) structural abnormalities on MRI. Additionally, during the interview, patients’ demographic characteristics and disease symptoms (the frequency of MwA per year, headache intensity [pain ranging from 1 to 10 on a numerical rating scale], presence of migraine without aura (MwoA) attacks, presence of somatosensory and dysphasic auras, the presence of photophobia, phonophobia, and nausea during MwA attacks, and disease duration in years) were noted. Based on these data, the patients were subdivided into MwAvsd (visual aura with somatosensory and/or dysphasia symptoms) and MwAv (only visual aura) subgroups for additional comprehensive analyses. In addition, from the electronic records (structured electronic questionnaires that patients regularly filled out after every MwA attack), the last six consecutive MwA attacks before MRI were used to calculate the average MACS for each patient. MACS electronic questionnaire collected data about the quality of visual and somatosensory symptoms, as well as symptoms related to higher cortical dysfunctions during the aura, such as micropsia or macropsia, prosopagnosia, dyspraxia, dysphasic and memory disturbances symptoms (Supplementary Table 1). Patients who had experienced visual aura also reported the level of involvement of the visual field, while patients who had experienced somatosensory symptoms also reported the number of body regions that were involved [3, 9]. These data were then utilized to calculate the MACS. HCs were voluntarily recruited among the hospital and university staff, as well as their friends and relatives, and their sex and age were balanced with those of patients with MwA. All HCs underwent interviews and physical examinations to check their health status and identify any exclusion criteria. Therefore, inclusion criteria for HCs were an absence of any systematic disease, including neurological disorders and any kind of headache (with exception of possible infrequent tension-type headache during their lifetime), as well as acceptance to participate in the study. Exclusion criteria were findings of structural brain abnormalities on MRI and inability to undergo MRI scan. Furthermore, HCs had no family members with migraine.

The Scientific Ethics Committee of the Clinical Center and Neurology Clinic approved this study (reference number: 23–690). Informed consent was obtained from eligible patients for participation in the study.

MRI data acquisition and post-processing

MRI was performed using a 3 T Scanner (MAGNETOM Skyra, Siemens, Erlangen, Germany). The protocol for MRI examinations was as follows: (1) 3D T1 (repetition time [TR] = 2300 ms, echo time [TE] = 2.98 ms, flip angle = 9°, 130 slices with a voxel size of 1 × 1 × 1 mm³, an acquisition matrix of 256 × 256, and field of view (FOV) = 256 × 256 mm²), (2) 3D FLAIR (TR = 5000 ms, TE = 398 ms, TI = 1800 ms, flip angle = 120^o, an acquisition matrix of 256 × 256, and FOV = 256 × 256 mm²), and (3) T2 weighted spin echo (T2W) in an axial plane (TR = 4800 ms, TE = 92 ms, flip angle = 90^o, an acquisition matrix of 384 × 265, a FOV = 186 × 229 mm², and slice thickness = 5 mm). T2W images were used to exclude the presence of brain lesions or structural abnormalities. Patients with MwA did not experience migraine 72 h before and after the MRI scan.

Analysis was performed using FreeSurfer (v6.0) on an HP DL850 server (Intel Xeon 3.2 MHz, eight cores, 16 GB RAM) using a recon-all script, combining 3D T1 and FLAIR images for automatic cortical reconstruction and segmentation of brain structures. Details of FreeSurfer and its routines have been previously reported [24, 25]. After the finalization of the script, the segmentation of the brain structures was manually inspected by an experienced neuroradiologist, blinded to the subject status, to ensure that the grey and white matter boundaries were correctly delineated. Furthermore, FreeSurfer-based segmentation of hippocampal subfields was performed [20]. The outputs of the hippocampal segmentation were left and right hemisphere images with label assignments for voxels in the hippocampal area to one of the following subregions: hippocampal tail, subiculum, presubiculum, parasubiculum, Cornu Ammonis (CA) subregions, such as CA1, CA3, and CA4, hippocampal fissure, molecular layer, granule cell layer of the dentate gyrus (GC-DG), hippocampus-amygdala transition area (HATA), and fimbria (Fig. 1). The volume of each extracted structure was measured in millimeters. The outputs of hippocampal segmentation for each subject were carefully reviewed by a radiologist (I.P.) and a trained analyst (M.D.). Moreover, the total intracranial volume was computed from each subject’s high-resolution T1W images using the fully automated Structural Image Evaluation with the Normalization of Atrophy tool in the FSL package.

Fig. 1

An example of hippocampal subregion segmentation. CA – Cornu Ammonis, GC-DG – granule cell layer of the dentate gyrus, HATA – hippocampus-amygdala transition area

Statistical analysis

Demographic data were compared between patients with MwA and HCs using the independent-sample t-test for continuous parametric data, Mann-Whitney U for continuous non-parametric data and chi-square test for categorical data, as appropriate. A general linear model (GLM) analysis was used to investigate differences in the volumes of the hippocampal subfields and the whole volume of the hippocampus between HCs and patients with MwA, while controlling for the effects of age, sex, and total intracranial volume to avoid spurious results. Additionally, a GLM analysis was used to investigate the differences between the MwAvsd and MwAv subgroups and HCs, controlling for the effects of age, sex, and the total intracranial volume. A partial correlation test, while controlling for the effects of age, sex, and the total intracranial volume, was used to assess the correlations between the volumes of the hippocampal subfields and the following characteristics of patients with MwA: frequency of MwA attacks, disease duration, headache intensity, presence of MwoA attacks and MACS. The GLM and partial correlation results were corrected for multiple comparisons using the false discovery rate with the Bonferroni–Holm correction (p < 0.05).

Results

Herein, 46 patients with MwA and 31 HCs were included. The groups were balanced regarding age (36.2 ± 9.1 vs. 36.0 ± 8.6 years, p = 0.916) and sex (69.6% vs. 71.0% of women, p = 1.000). The main clinical characteristics of the patients are shown in Table 1.

Table 1 Clinical characteristics of a total group of migraine with aura (MwA) patients

Multivariate GLM with the Bonferroni–Holm correction did not reveal any significant differences in hippocampal subfields between patients with MwA, without distinction of aura subtypes, and HCs (Table 2). The average MACS was negatively correlated with the volumes of the left and right hippocampi (r = -0.454, p = 0.002; r = -0.389, p = 0.010, respectively), as well as with the volumes of most hippocampal subregions (Table 3). There were no significant correlations between the number of MwA attacks per year, disease duration, headache intensity, presence of MwoA attacks and hippocampal or hippocampal subfield volumes (Supplementary Table 2).

Table 2 Comparison of the volumes of hippocampal subfields between MwA patients and HCs

Table 3 Total group of migraine with aura patients. Correlation between hippocampal volume and its subfields and migraine aura complexity score (MACS)

Subsequently, the hippocampal volume and the volume of the hippocampal subfields in the MwAvsd and MwAv subgroups were compared with each other and with HCs. The demographic and clinical characteristics of the MwA subgroups are shown in Table 4. Compared with the MwAv subgroup and HCs, the MwAvsd subgroup had significantly smaller whole hippocampal volumes in both hemispheres, as well as in both subicula (Table 5). In addition, the left molecular layer, right CA1, and hippocampal fissures were significantly smaller in the MwAvsd group than in the MwAv subgroup and the HCs. Additionally, GC-DG, CA3, and CA4, as well as the left CA1 and right molecular layers, were significantly smaller in the MwAvsd group than in the MwAv subgroup. In both subgroups (MwAv and MwAvsd), there was no significant correlation between the hippocampal subfield volumes and clinical features, including the MACS index (Supplementary Table 3).

Table 4 Characteristics of MwAvsd and MwAv subgroups of patients

Table 5 Comparison of the volumes of hippocampal subfields between MwAvsd, MwAv and HCs

Discussion

The present study utilized FreeSurfer-based segmentation of hippocampal subfields using 3T structural MRI data to investigate potential differences in hippocampal volume between HCs and patients with MwA, as well as between MwA subgroups and HCs. We observed significantly smaller volumes of both hippocampi, including the subicula, left molecular layer, right CA1, and hippocampal fissure, in the MwAvsd subgroup than in the MwAvsd subgroup and HCs. Moreover, smaller volumes of both the hippocampi, CA1, CA3, CA4, and the molecular layer, as well as the left hippocampal fissure, GC-DG, and HATA, indicated a higher MACS. However, there were no differences between HCs and patients with MwA, which may be a consequence of the different trends in the MwA subgroups relative to HCs. These findings, corroborating those of previous studies, suggest that MwAvsd should be studied separately from MwAv. Herein, we discuss the hippocampal subfield volume differences in MwAvsd relative to MwAv and HCs as well as their potential relevance to the multifaceted pathophysiology of MwA.

Hippocampal formation is critical for spatial navigation and episodic memory and has traditionally not been associated with language functions [18]. However, accumulating evidence from studies on patients with epilepsy [16, 17, 26] suggests that the hippocampus may also be important for language processing. In particular, the hippocampus is an active component of the naming network, providing associative links between different lexical-semantic representations in cortical areas such as the left lateral temporal lobe, and its dynamics are closely associated with efficient word production [16, 27]. In a study by Hamamé et al. [16], inefficient hippocampal activation was associated with “tip-of-the-tongue” states, which is similar to symptoms of dysnomia during MwA attacks, the most common symptom of language disturbances in the ictal stage of MwA [9]. Altogether, it can be considered that smaller hippocampal volumes, in particular smaller volumes of both subicula, the left molecular layer, right CA1, and hippocampal fissures, may predispose patients with MwA to more complex auras. Furthermore, given that there is no correlation between the duration of disease and hippocampal volume, our findings let us argue that a smaller hippocampal volume in MwAvsd is not likely a mere consequence of recurrent migraine attacks. This is in contrast to two cross-sectional investigations conducted on patients with migraine without aura [12, 28]. They observed that low-frequency migraine patients had a significantly higher bilateral hippocampus volume than the high-frequency and HCs groups, suggesting an initial adaptive plasticity that may become dysfunctional with increased frequency. After adjustment for headache frequency, right hippocampal volume was favourably related with excellent migraine outcomes at a 2-year follow-up [12, 28]. Moreover, memory problems during MwA attacks are among the rarest symptoms [9], which raises an additional argument against the hypothesis that a smaller volume of the hippocampus is due to degenerative processes caused by MwA attacks. Therefore, our findings suggest that a smaller hippocampal volume may be a specific feature of MwAvsd. However, this statement needs to be validated in future studies.

A widely accepted hypothesis regarding the pathophysiology of the migraine aura is centered toward a primary brain phenomenon called cortical spreading depolarization/depression (CSD), which predominately originates in the occipital cortex and propagates toward parietal and temporal cortices [29, 30]. Knowing that in some patients, somatosensory and/or dysphasic aura onset occurs at the same time or before the visual aura, it may be hypothesized that CSD can initiate simultaneously at distant cortical regions independently from the visual cortex, which is in agreement with Leão’s observation in animals [31]. This hypothesis adds to the interpretation of our findings, pointing to the possibility that a hippocampus with a smaller volume, like in our MwAvsd patients, may become one of the CSD originators. Reduced hippocampus volume may predispose individuals to abnormalities in long-term synaptic transmission that can initiate and propagate CSD. Specific regions within the hippocampus are highly connected with multiple cortical areas in the medial parietal, temporal, and occipital lobes [11], suggesting that the hippocampus could have a modulatory effect and, in some cases, maladaptive responses during MwA attacks. Nevertheless, both hypotheses must be investigated using functional and multimodal neuroimaging before further speculation. Our results indicate that the CA1, molecular layer, and subiculum are the most important hippocampal subfields that contribute to the smaller size of the hippocampi in patients with MwAvsd. It is important to note that volumetric analyses of structural neuroimaging have consistently confirmed age-related volume reductions in the subiculum and CA1 [32,33,34]. However, herein, age-related volume reductions in these regions were not different between MwAvsd, MwAv, and HCs, identifying the left and right subiculum, right CA1, and left molecular layer as specific features and new targets for the investigation of pathophysiological mechanisms in MwAvsd. Furthermore, knowing that the CA1/subiculum transition area appears to be the main area of anatomical connectivity in multiple cortical areas [11], including the hypothalamus, one of the so-called migraine generators, and that CSD could originate in the hippocampal cortex [35], it may be hypothesized that these regions play a significant role in MwAvsd and attack initiation. Additionally, in animal models, hippocampal spreading depression was associated with a linked and temporary alteration in spontaneous activity and blood flow in the ipsilateral neocortex, without the propagation of spreading depression to that region [35]. Moreover, given that the pyramidal cells from CA1 project to the temporal and parietal association areas, visual, auditory, olfactory, somatosensory, and prefrontal-orbital-agranular insular regions [36], and considering our finding of a negative correlation between CA1 volume and MACS, future functional neuroimaging studies on patients with MwAvsd should investigate functional connectivity between CA1 and the aforementioned regions to explain the nature of the findings in this study. Additionally, it should be taken into consideration that these regions are important in the pathophysiology of temporal lobe epilepsy where CA1 pyramidal axons sprout into the subiculum, as well as into the molecular layer. This sprouting could lead to the amplification and synchronization of epileptic discharges as they emerge from the hippocampus [37]. Some of these pathophysiological mechanisms in the subiculum/CA1 regions may also interplay with the multifaceted pathophysiology of MwAvsd, where bursting and/or specific gamma-aminobutyric acid-mediated functions could influence the complexity of MwA attacks.

This study represents an attempt to study the volume of the hippocampus and its subfields in patients with MwA and their subgroups; however, it has some limitations. First, the sample size was relatively small, which may have limited its statistical power. A larger dataset of participants is necessary to obtain more robust results. In addition, we did not investigate the relationship between the possible influence of frequency of MwoA attacks in MwA patients and volumes of hippocampi and its subfields which should be further explored in future studies. However, our cohort had active MwA attacks during the investigated period, did not take additional painkillers for other reasons, or was on any kind of preventive treatment, which makes selected patients suitable for testing our hypothesis. Furthermore, a previous study concluded that the best test-retest reproducibility was achieved for hippocampal subfields > 300 mm³_, suggesting that the measurement of smaller hippocampal subfields should be interpreted with caution [38]. Nevertheless, the strong statistical significance suggests that future studies are warranted. Furthermore, future studies using higher resolution and multimodal neuroimaging data may allow for more targeted investigations to explain the findings of our study. Finally, we recognize that the inclusion in the group of HCs of subjects with infrequent tension-type headaches may have biased the results. However, none of the participants from the HCs group reported any kind of headache 1-year before the investigation and none of them had any diagnosis regarding the primary headache types.

Herein, we report the first findings that suggest that differences in the volumes of the CA1, molecular layer, and subiculum may be related to the complexity of the migraine aura. Validation of these findings in future studies may eventually lead to a migraine-specific target for non-invasive neuromodulatory treatments [39, 40]. Studies on larger patient cohorts and with longitudinal, prospective follow-up, are needed to reveal any potential environmental and genetic relationship between the risk of being affected by complex forms of migraine with aura and the structure of the hippocampus [41].