Bout-associated intrinsic functional network changes in cluster headache: A longitudinal resting-state functional MRI study

Abstract

Background

Previous imaging studies on the pathogenesis of cluster headache (CH) have implicated the hypothalamus and multiple brain networks. However, very little is known regarding dynamic bout-associated, large-scale resting state functional network changes related to CH.

Methods

Resting-state functional magnetic resonance imaging data were obtained from CH patients and matched controls. Data were analyzed using independent component analysis for exploratory assessment of the changes in intrinsic brain networks and their relationship between in-bout and out-of-bout periods, as well as correlations with clinical observations.

Results

Compared to healthy controls, CH patients had functional connectivity (FC) changes in the temporal, frontal, salience, default mode, somatosensory, dorsal attention, and visual networks, independent of bout period. Compared to out-of-bout scans, in-bout scans showed altered FC in the frontal and dorsal attention networks. Lower frontal network FC correlated with longer duration of CH.

Conclusions

The present findings suggest that episodic CH with dynamic bout period shifts may involve bout-associated FC changes in multiple discrete cortical areas within networks outside traditional pain processing areas. Dynamic changes in FC in frontal and dorsal attention networks between bout periods could be important for understanding episodic CH pathophysiology.

Keywords

Cluster headache functional connectivity hypothalamus Bout period independent component analysis

Introduction

Cluster headache (CH) is a primary headache disorder characterized by recurrent and severe unilateral periorbital pain and autonomic symptoms. In episodic CH, patients experience high-frequency attacks during in-bout periods. Between in-bout periods, patients remain asymptomatic for months to years (out-of-bout periods) (1). The clinical picture of CH, with its circadian rhythmicity and cranial autonomic features, supports the hypothesis that the disorder may involve the hypothalamus (2). Furthermore, previous functional imaging studies have demonstrated hypothalamic activation during acute CH attacks (3,4). Attempts have been made to treat intractable CH with deep-brain stimulation (DBS) of the hypothalamus (5,6), although without complete success. This suggests that other brain structures or networks are involved in CH.

Several brain regions in the so-called “pain matrix,” including the anterior cingulate cortex (ACC), posterior thalamus, prefrontal cortex, and insular cortex, have been implicated in CH pathogenesis, especially during acute attacks (7,8). In fact, altered metabolism in the aforementioned areas has been observed in patients with CH (7). We have previously reported dynamic changes between bout periods in frontal pain modulation areas (9,10). Functional abnormalities in occipital areas and the cerebellum have also been reported in patients with CH (11,12). Additionally, we previously noted dynamic bout-related changes in functional connectivity (FC) between the hypothalamus and cerebellar, frontal, and occipital areas (13). Thus, CH pathophysiology may involve multiple brain networks that include not only the hypothalamus, but also central pain modulation networks and non-traditional pain-processing regions.

Resting-state functional MRI (RS-fMRI) is used to examine the temporal correlations of blood oxygen level-dependent (BOLD) fluctuations between brain regions at rest, which is thought to reflect resting neuronal activity and is often related to FC (14). This MR technique can elucidate extensive brain networks, and investigate how resting-state network (RSN) interactions change in neuropsychiatric illness (14). Most published RS-fMRI studies assessing CH patients used seed-based FC analysis focused on the hypothalamus. These previous studies described abnormal FC only between the hypothalamus and pain-related brain areas in patients with episodic CH, or during spontaneous CH attacks (11 –13,15). Information on alterations in whole-brain intrinsic functional networks between bout periods is still lacking. Given the episodic nature of CH, and the fact that its pathophysiology appears to involve not only the hypothalamus but also other large-scale functional networks, identifying the networks with altered FC in relation to bout status might provide a chance to explain both the disease pathophysiology and the observed bout period shifts.

This exploratory study assessed the changes in distributed intrinsic FC networks and their relationship with bout status in CH, to determine whether functional alterations in brain networks are implicated in CH pathophysiology and associated with shifts between bout periods. We evaluated CH-related, whole-brain large scale RSNs via RS-fMRI scan, together with independent component analysis, to test the following hypotheses: (1) FC changes in RSNs occur outside traditional pain processing areas; (2) FC in RSNs differs between in-bout and out-of-bout periods; and (3) FC changes in RSNs are associated with clinical observations.

Materials and methods

Participants

This study was approved by the Institutional Review Board of Taipei Veterans General Hospital (TVGH), Taiwan, and written informed consent was obtained for all participants prior to enrollment. Seventeen patients were recruited from the Headache Clinic at Taipei Veterans General Hospital (TVGH). These patients met the diagnostic criteria for episodic CH, based on the third edition of the International Classification of Headache Disorders (ICHD-3 -β) (1). None of the patients had other concomitant primary (e.g. migraine) or secondary headache disorders. The control group comprised 18 sex, age, and handedness matched healthy volunteers recruited through community advertisements or hospital subject pools. Exclusionary criteria for controls included a family history of CH or migraine, prior diagnosis of a primary or secondary headache disorder, and any chronic pain condition. Infrequent episodic tension-type headache occurring no more than one day per month was acceptable for inclusion as a control. In this study, all patients and control subjects were scanned for the first time specifically for the purpose of this study. Patients were first scanned during an in-bout period that had been ongoing for at least one week at the time of the scan, and then re-scanned during an out-of-bout period, with a mean interval between scans of 7.8 months (range, 6–11 months). No active acute attacks occurred during scanning sessions, and no medications were taken for at least 24 hours before the scans. The participants in this study included some but not all of the subjects from our previous study (13). The statistical analysis of demographic data was calculated using SPSS software version 20 (SPSS, Chicago, IL, USA) with 2-sample Student’s test and Pearson’s chi-square test.

MRI acquisition

MR images were collected at TVGH on a 3T GE Discovery-750 MRI scanner with an eight-channel phase array head coil and cushions to reduce participant motion during image acquisition. A whole-brain T1-weighted scan was acquired for anatomical reference and voxel-wise gray matter volume estimation, using a three-dimensional inversion recovery prepared fast spoiled gradient recalled sequence with the following parameters: repetition time (TR)/ echo time (TE)/ inversion time = 9.2/3.7/450 ms, flip angle = 12°, number of excitations (NEX) = 1, matrix size = 256 × 256, FOV = 256 × 256 mm, voxel size = 1 × 1 × 1 mm³ and 172 axial slices without inter-slice gap. For the RS-fMRI scan, a whole-brain T2*-weighted gradient-echo planar sequence with the following parameters was used: TR/TE = 2500/30 ms, flip angle = 90°, NEX = 1, matrix size = 64 × 64, FOV = 222 × 222 mm, voxel size = 3.47 × 3.47 × 3.5 mm³, 43 interleaved axial slices without inter-slice gap and 200 image volumes. Participants were instructed to relax with their eyes closed, not to fall asleep, and not to think about anything in particular during RS-fMRI acquisition. We recorded no objective measures of subject vigilance information (e.g. heart and respiratory rate) during the imaging session.

Data preprocessing

The RS-fMRI dataset was preprocessed using the FMRIB Software Library (FSL v5.0.8, Functional MRI of the Brain Software Library; http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/; (16)). The following preprocessing pipeline was applied: (1) removal of the first ten volumes, (2) slice timing correction, (3) head motion correction, (4) non-brain tissue removal, (5) spatial smoothing with a 5 mm Gaussian kernel, and (6) grand–mean intensity normalization. A high-pass temporal filter was not applied in the current study due to recent research suggesting that broadband spectral characteristics were observed in large scale resting state networks (17,18). Single-subject independent component analysis (ICA) with automatic dimensionality estimation was applied to the preprocessed RS-fMRI dataset using MELODIC (Multivariate Exploratory Linear Optimized Decomposition into Independent Components (19)). Single-subject components were further classified into artifact or unlikely artifact components using Spatially Organized Component Klassificator (SOCK) based on the characteristics of spatial pattern, time and frequency domain information of each IC (20). We used the default settings of the SOCK toolbox to identify noise ICs in our dataset. To further validate whether SOCK successfully identified the spurious components, an experienced rater blinded to the participants’ status was also trained to identify artificial components based on previous ICA studies (21,22) and performed a manual visual inspection. After artificial component identification, FMRIB’s ICA-based Xnoiseifier (23) was used to aggressively regress the full space of the artificial components and the 24 motion-estimation confound time series (the time series of each individual’s six rigid-body parameter, their temporal derivatives, and the squares of all 12 resulting regressors) out of the preprocessed RS-fMRI dataset. Because the imaging parameters of the training dataset provided by the FSL-FIX tool were not the same as ours, we used the FSL-FIX tool only to remove the effect of self-identified noise components and motion-related artifacts. Since CH is strictly lateralized, we flipped the denoised RS-fMRI dataset and corresponding T1-weighted scan for seven patients with left-side attacks along the mid-sagittal plane, thus enabling us to treat the headache attacks as being on the same side (the right side being symptomatic). This technique was common in previous CH brain imaging studies (7,9,10,13). Furthermore, in order to reduce potential bias due to the brain-flipping step in patients with left-sided CH, we also flipped the corresponding imaging data of seven sex, age, and handedness matched healthy control subjects. After completing the above procedures, individual RS-fMRI datasets were spatially normalized to the Montreal Neurological Institute (MNI) space template using a two-stage approach (13). The image resolution of the MNI space template is 2 mm isotropic. These cleaned, preprocessed RS-fMRI datasets were served as input for further group-level ICA. Because head motion may influence the FC profile of the RS-fMRI dataset, attention was paid to potential motion-related issues. The detailed analytic protocol for motion-related issues is provided in the supplementary methods.

Voxel-wise gray matter volume estimation

Based on our more detailed, previously published methodology (10), an image-based, voxel-wise modulated GMV regressor was used to adjust for the effect of GMV differences when subsequently comparing FC. The voxel-wise gray matter volume estimation was carried out using the steps for the voxel-based morphometry (VBM) approach included in the VBM8 toolbox (http://dbm.neuro.uni-jena.de) (24), Statistical Parametric Mapping software (SPM8, Wellcome Institute of Neurology, University College London, UK, http://www.fil.ion.ucl.ac.uk/spm/), and MATLAB R2010a (Mathworks, Natick, MA) with default settings.

Resting state network construction and with-network comparisons

To obtain group-level RSNs, temporal concatenation ICA was carried out for the entire group using MELODIC with automatic dimensionality estimation. This procedure identified 56 spatial ICs, and all ICs were standardized into a mixture-modeling Z-score map by dividing each component’s weight into the standard deviation of the background noise, with a Z-score > 4 as the threshold for representative RSNs. We visually identified 10 artificial components (residual movements, white matter, and ventricular components) from the 56 group level ICs, so 46 RSNs served as networks-of-interest for within-network analysis. Dual regression was then used to investigate voxel-wised FC changes of RSNs between CH and control groups and bout stages (25). Briefly, the full set of unthresholded group spatial ICs were first used as spatial regressors in a general linear model to derive a participant-specific time course for each component. The time course matrices of each component were then regressed against each participant’s preprocessed dataset, to estimate the participant-specific spatial maps in terms of voxel-wise z-scores. To identify cross-sectional and longitudinal FC changes within each RSN, nonparametric statistical models with 10,000 permutations were carried out using the FSL RANDOMISE tool. An analysis of covariance (ANCOVA) with the corresponding GMV image as a voxel-dependent nuisance variable, and age, sex, and mean frame-wise displacement as non-image based nuisance variables, were used to investigate the cross-sectional FC group differences within RSNs (in-bout/out-of-bout versus controls). To further identify the longitudinal FC changes between in-bout and out-of-bout scans within each RSN, we utilized an ANCOVA model with the FC differences between in-bout and out-of-bout as the dependent variable, the corresponding GMV differences as a voxel-dependent nuisance variable, and the differences of mean frame displacement as a non-image based nuisance variable. The threshold of significance was at p < 0.05 with family-wise error (FWE) corrected for multiple comparisons using threshold-free cluster enhancement (26). Due to the exploratory design of the study, we limited multiple comparison correction to correcting within each RSN, and did not correct across the 46 RSNs. Clusters with significant within-group and between-group effects were further extracted for correlation analysis with clinical observations. The Shapiro–Wilk test was used, and revealed that the mean FC of each cluster with significant statistical difference fulfilled the parametric data assumption. Subsequently, partial Pearson correlation analyses, with age, sex and mean frame-wise frame displacement serving as nuisance variables, were used to examine the relationship between regional FC and the times of bouts, frequency of attacks per day, duration of illness, and frequency of bouts per year within the patient group. Because the clinical correlation analysis was primarily exploratory, we did not correct for multiple comparisons (p < 0.05).

Results

Demographics and RS-fMRI motion profiles

There were no significant differences in sex (p = .17), age (p = .88), or handedness (p = .49) between CH and control groups. Of the 17 patients with CH, 10 had strictly right-sided attacks, and seven had strictly left-sided attacks. The mean duration to scanning after the last attack during an in-bout period was 17.2 hours (range of 10–23 hours). However, the mean duration to the next attack after scanning was not recorded in this study. The patients’ demographic and clinical information is summarized in Table 1. For the RS-fMRI time series, no subjects had a maximum rotation that was ≥ 1.5° or displacement that was ≥ 1.5 mm in any directions. The CH and control groups showed no significant differences in the mean frame displacement (p = .84).

Table 1.

Demographic variables and clinical characteristics of the study participants.

	CH patients (n = 17)
Demographic variables	In-bout	Out-of-bout	HC (n = 18)	p value
Age (years)	34.2 ± 8.3	34.2 ± 8.3	33.2 ± 5.2	0.88
Sex (male/female)	15/2	15/2	10/8	0.17
Handedness (right/left)	16/1	16/1	18/0	0.49
Lateralization (right/left)	10/7	10/7	–	–
Frame displacement (mm)	0.084 ± 0.045	0.076 ± 0.024	0.082 ± 0.048	0.84
Duration of illness (years)	8.6 ± 5.8		–	–
Frequency of bouts per year	1 ± 0.6		–	–
The duration of bout period (days)	21.2 ± 17.6		–	–
The frequency of attacks per day	1.2 ± 0.4		–	–
The mean duration to scanning after last attack (hours)	17.2 ± 5.4		–	–
BDI	9.3 ± 8.2		–	–
A-score	8.2 ± 4.6		–	–
D-score	4.4 ± 3.8		–	–

Notes: Two sample t-tests and Pearson’s chi-square tests were used to test between-group differences in continuous data (age and frame displacement) and categorical data (sex and handedness), respectively. Abbreviations: A-score, anxiety score of The Hospital Anxiety and Depression Scale; BDI, Beck depression inventory score; CH, cluster headache; HC, healthy controls; D-score, depression score of The Hospital Anxiety and Depression Scale.

FC differences between CH and control groups within large-scale RSNs

Forty-six components with potential functional relevance were identified in the full sample, and then grouped into 10 categories based on anatomical location: RSNs of visual, sensorimotor, default mode, frontal, dorsal attention, parietal, salience, cerebellar, temporal, and subcortical categories (Figure 1). Among these, seven categories showed significant changes in intrinsic FC in patients with CH during both in-bout and out-of-bout periods compared to controls (P_{FWE-corrected} < .05; Figure 2, Table 2). Patients with CH exhibited significant changes in cortical FC in the left supramarginal gyrus, left frontal orbital gyrus, right insula, left precuneus, left middle occipital gyrus, right frontal operculum cortex, right inferior parietal lobule, left inferior parietal lobule, left inferior temporal gyrus, right cingulate gyrus, left superior temporal gyrus, left superior frontal gyrus, left inferior frontal gyrus, left medial frontal gyrus, right middle frontal gyrus, and right temporal pole (IC 8, 10, 12, 13, 14, 16, 19, 21, 22, 23 and 36). Furthermore, patients with CH exhibited significant changes in FC in the left precuneus within the default-mode network (DMN; IC 2) and in the right and left insula within the salience network (IC 30 and 40).

Figure 1.

The spatial distribution of 46 large-scale resting state networks (RSNs).

Figure 2.

Aberrant functional connectivity of large-scale RSNs in the CH group.

Table 2.

Anatomical regions with significant functional connectivity differences among CH patients compared to healthy controls.

		MNI coordinates							Effect size
IC	RSN	x	y	z	Cluster size	Anatomical region	Maximum T value	p value	(partial eta squared)
HC < in-bout CH patients
2	Default mode	−4	−60	16	77	L. Precuneus	3.9	0.029	0.479
8	Temporal	−60	−48	16	181	L. Supramarginal Gyrus	3.92	0.008	0.428
8	Temporal	−48	26	−12	86	L. Frontal Orbital Gyrus	4.44	0.048	0.467
10	Temporal	40	20	−4	171	R. Insula	4	0.008	0.394
13	Frontal	−48	−76	24	446	L. Precuneus	4.79	0.002	0.473
13	Frontal	−14	−84	50	94	L. Precuneus	3.98	0.039	0.397
16	Visual	30	−84	16	96	R. Middle Occipital Gyrus	4.41	0.024	0.471
21	Dorsal attention	50	12	−2	580	R. Frontal Operculum Cortex	4.77	0.001	0.489
21	Dorsal attention	36	−50	34	232	R. Inferior Parietal Lobule	4.81	0.004	0.482
23	Dorsal attention	−64	−26	18	263	L. Inferior Parietal Lobule	5.43	0.002	0.479
23	Dorsal attention	−50	−68	−6	143	L. Inferior Temporal Gyrus	3.85	0.012	0.46
23	Dorsal attention	−40	−44	40	134	L. Inferior Parietal Lobule	4.37	0.015	0.471
40	Salience	44	−16	−6	47	R. Insula	5.03	0.009	0.405
HC > in-bout CH patients
30	Salience	−34	14	0	65	L. Insula	3.86	0.028	0.427
36	Frontal	8	16	32	62	R. Cingulate Gyrus	4.06	0.046	0.397
HC < out-of-bout CH patients
8	Temporal	−56	−58	10	249	L. Superior Temporal Gyrus	5.55	0.003	0.41
10	Temporal	46	16	−4	102	R. Frontal Operculum Cortex	4.5	0.029	0.386
12	Visual	−30	−84	−6	99	L. Middle Occipital Gyrus	3.45	0.035	0.256
13	Frontal	−48	−76	30	458	L. Precuneus	4.93	0.001	0.439
13	Frontal	−32	22	38	231	L. Superior Frontal Gyrus	4.52	0.006	0.422
14	Somatosensory	−50	8	16	96	L. Inferior Frontal Gyrus	4.75	0.015	0.484
19	Frontal	−16	42	48	89	L. Superior Frontal Gyrus	5.54	0.027	0.334
19	Frontal	−6	54	20	66	L. Medial Frontal Gyrus	4.54	0.047	0.347
21	Dorsal attention	46	42	4	358	R. Middle Frontal Gyrus	5.56	0.001	0.478
21	Dorsal attention	54	14	−4	238	R. Temporal Pole	4.82	0.002	0.507
21	Dorsal attention	38	−54	36	219	R. Inferior Parietal Lobule	4.76	0.003	0.548
22	Frontal	−6	42	30	78	L. Medial Frontal Gyrus	4.03	0.038	0.461
22	Frontal	−36	22	32	78	L. Precentral Gyrus	4.42	0.038	0.463
23	Dorsal attention	−54	−32	24	146	L. Inferior Parietal Lobule	4.95	0.010	0.452
23	Dorsal attention	−48	10	18	83	L. Inferior Frontal Gyrus	3.97	0.040	0.443
40	Salience	44	−20	−10	120	R. Insula	5.06	0.001	0.449

Notes: Results are described in terms of corresponding large scale RSN, MNI coordinates, cluster size (number of voxels of the corresponded cluster) and anatomical regions. The statistical criteria are set as cluster level FWE corrected p value < 0.05. The T-value and P-value of the voxel with the strongest group effect in a given cluster are also listed. Abbreviations: CH, cluster headache; FWE, family-wise error; HC, healthy controls; IC, independent component; L, left; MNI, Montreal Neurological Institute; R, right; RSN, resting state network.

FC differences between in-bout and out-of-bout periods in the CH group

Compared to out-of-bout scans, in-bout scans exhibited a significant change in FC in the right inferior frontal gyrus and left postcentral gyrus (IC 22 and 23; Figure 3; Table 3).

Figure 3.

Longitudinal FC changes between in-bout and out-of-bout periods among patients with CH.

Table 3.

Anatomical regions with significant functional connectivity differences in CH patients between in-bout and out-of-bout periods.

		MNI coordinates			Cluster		Maximum		Effect size
IC	RSN	x	y	z	size	Anatomical region	T value	p value	(Cohen’s d)
In-bout CH patients > out-of-bout CH patients
23	Dorsal attention	−56	−18	16	89	L. Postcentral Gyrus	4.83	0.034	1.29
In-bout CH patients < out-of-bout CH patients
22	Frontal	50	40	−16	112	R. Inferior Frontal Gyrus	4.99	0.017	1.33

Notes: Results are described in terms of corresponding large scale RSN, MNI coordinates, cluster size (number of voxels of the corresponded cluster) and anatomical regions. The statistical criteria are set as cluster level FWE corrected p value < 0.05. The T-value and P-value of the voxel with the strongest longitudinal effect in a given cluster is also listed. Abbreviations: CH, cluster headache; FWE, family-wise error; IC, independent component; L, left; MNI, Montreal Neurological Institute; R, right; RSN, resting state network.

Association between clinical variables and FC in the CH group

Disease duration was significantly and negatively correlated with a reduction in mean FC in the right cingulate gyrus (r = −0.719, p = 0.003) of the frontal network (IC 36) during the in-bout period.

Discussion

In this study, changes in RSNs with FC differences among patients with CH relative to controls were observed in temporal, frontal, salience, default mode, somatosensory, dorsal attention, and visual networks. Longitudinally, we observed reduced FC in the frontal and dorsal attention networks between in-bout and out-of-bout periods. Notably, regions showing bout-associated FC changes in RSNs were primarily outside the traditional pain processing networks. Lower FC in the frontal network was related to a longer CH illness duration.

Few studies have investigated intrinsic functional fluctuations among patients with CH. Rocca et al. (11) discovered altered intrinsic fluctuations in sensorimotor and primary visual networks in patients with CH in the out-of-bout period. Abnormal hypothalamic FC has been reported in areas associated with pain processing and visual networks both during and between CH attacks (12), while the results of our previous study indicated bout-dependent dynamic hypothalamic FC changes in frontal, occipital, and cerebellar areas (13). Additionally, decreased functional coactivation was detected between the hypothalamus and salience network areas in patients with CH (15). Although these previous studies found changes in intrinsic network FC in patients with CH, none studied alterations in whole-brain large-scale RSNs, or the possible contribution to shifts between bout periods. Here, we observed dynamic FC changes between in-bout and out-of-bout periods in the frontal and dorsal attention networks. Collectively, CH pathophysiology appears to involve multiple brain networks, including areas beyond traditional pain-related networks.

Neuroimaging studies have revealed activation in the ascending pain processing and descending pain modulation systems during the experience of acute pain. The basal ganglia (BG), cerebellum, hippocampus, and frontal, parietal, and temporal cortical areas are also active during acute pain (27). In this study, bout-associated FC changes in the superior temporal gyrus were identified in patients with CH. The superior temporal gyrus is within the temporal cortex, a traditional component of pain-processing networks (27). Anatomic connectivity between the hypothalamus and temporal area has been demonstrated in primates (28), which explains why hypothalamic DBS produces functional changes in the temporal cortex in patients with CH (29). Hence, RSNs with bout-associated FC changes in the temporal area could be associated with nociceptive processing.

Dysfunction in the descending pain circuitry facilitates primary headache disorders by disinhibiting or facilitating nociceptive processing (30). Here, we observed FC differences between CH and control groups in frontal and dorsal attention areas, which modulate the cognitive and affective dimensions of pain processing (30). We have also observed bout-associated GMV and white matter microstructure changes in frontal and limbic pain modulation areas (9,10), as well as hypothalamic FC changes in frontal areas in relation to bout status (13). Furthermore, the dorsal attention network also provides a direct corticocortical pathway for the attentional modulation of pain (31). Here, dynamic FC changes in the frontal network and dorsal attention networks between in-bout and out-of-bout periods were identified. We also found that lower FC in the frontal network was associated with longer disease duration of CH. Collectively, these data suggest that these dynamic alterations in frontal pain modulation networks may relate to shifts between bout periods and disease progression, and provide sites for neuromodulatory therapies for CH. Further research is required to explore this possibility.

Consistent with prior studies assessing acute CH attacks (8,12), we detected CH-related FC differences in the insular area associated with the salience network. The salience network, which mainly includes the dorsal ACC, anterior insula, and frontal operculum, is related to the processing of affective information and attentional modulation, antinociception, and autonomic reactions to painful stimuli (32). Altered brain metabolism and intrinsic FC in the insula and salience network-related areas has been demonstrated in CH patients, even during headache remission (7,11,15). Furthermore, the insular area may be associated with central autonomic regulation (33). Therefore, our findings are consistent with the notion that salience network FC changes in CH may be related to a disruption of cognitive and emotional pain modulation and autonomic responses.

We also detected bout-related FC differences in the precuneus, part of the DMN. The DMN includes the posterior cingulate, precuneus, medial temporal, lateral temporoparietal, and medial frontal areas, and is thought to mediate cognitive processes, influence behavior in response to stressful experiences, and facilitate coping strategies for promoting adaptation (34). Previous RS-fMRI studies have demonstrated disrupted DMN connectivity in several pain conditions, including migraine (35), suggesting the presence of maladaptive brain responses to repeated stress and pain. Furthermore, altered regional homogeneity in CH patients has been detected in the posterior cingulate and other DMN areas during pain processing (8). Therefore, our present findings are consistent with previous observations of chronic pain and primary headache disorders, implicating maladaptive brain responses as part of CH pathophysiology.

Observed FC alterations outside typical pain processing areas included the middle occipital gyrus (the visual network). However, these occipital areas participate in visuospatial perception and localization (36). Altered GMV has been reported in the visual processing areas of patients with CH (37). Furthermore, previous studies have revealed altered visual network fluctuations (11) and hypothalamic FC in occipital areas (12,13). These might reflect visual processing disturbances reported by patients with CH, and could be associated with clinical symptoms related to photophobia during acute attacks.

In this study, ICA did not show any functional network differences involving the hypothalamus. These results contradicted our previous study, in which we found alterations in hypothalamic functional connectivities in patients with CH (13). The use of different image processing procedures (ICA versus a seed-based approach) may partly explain this discrepancy. Furthermore, group level ICA can sensitively detect consistent large-scale functional networks across study participants, but lacks such sensitivity to identify functional networks that are associated with small deep subcortical nuclei, such as the hypothalamus. Further studies are needed to integrate these different imaging processing methodologies to better understand the changes in the brain and CH pathophysiology.

The present study has several noteworthy limitations. First, our sample size was not very large. CH is a rare headache disorder, making patient recruitment difficult. Second, since patients with episodic CH were asymptomatic during the out-of-bout period and did not need to come back to the clinic, it was more difficult to recruit patients during this period than during the in-bout period. Furthermore, most episodic CH patients had just one or two annual bout periods (2). Therefore, the average time interval between two MRI sessions in this study might be up to eight months. However, this relatively long period between the two sessions of MRI may potentially confound the results. Third, in this study, we included fewer female participants in the patient group; therefore, the sex-matching was less successful. Although we had included sex as a covariate in the statistical analysis, we cannot guarantee that the results are not confounded by gender differences. Therefore, we cannot fully generalize these results to female patients with CH. Fourth, in this study, a statistical threshold of p < .001, as corrected for multiple comparisons with Bonferroni’s correction across the 46 RSNs, was required for rejecting the null hypothesis (i.e., no differences in functional connectivity of each RSN). Consequently, no voxels survived this conservative correction across all RSNs, and no firm conclusions could be drawn. Nevertheless, as an exploratory study, our results may shed some light on the hypothesis that CH involves aberrant functional connectivity in multiple networks, and may yet advance our knowledge of large-scale network pathophysiological changes related to CH. Finally, it should also be stressed that while ICA identifies functional interactions, it does not provide information regarding causality. Future causality analyses may provide insight into the connection and causality between the hypothalamus and other brain areas implicated in CH.

To the best of our knowledge, this is the first longitudinal study to demonstrate functional alterations in whole brain large-scale RSNs in relation to bout periods. We observed abnormalities in FC within temporal, frontal, salience, default mode, somatosensory, dorsal attention, and visual networks among patients with CH. Dynamic FC changes between in-bout and out-of-bout periods in the frontal network were strongly related to CH duration. These findings advance our understanding of network functionality related to episodic CH with dynamic bout period shifts. Importantly, these functional changes affected intrinsic FC throughout multiple brain networks, especially those outside the traditional pain-processing networks.

Footnotes

Article highlights

Resting state functional MRI was used to assess the changes in intrinsic brain networks in CH patients.

Abnormalities in functional connectivity (FC) within various cortical networks with dynamic changes between bout periods were found in CH patients.

Episodic CH with dynamic bout period shifts may involve FC changes in discrete cortical areas within networks outside traditional pain processing areas.

Acknowledgements

The authors would like to acknowledge MRI support from the MRI Core Laboratory of National Yang-Ming University, Taiwan.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported in part by grants from the Ministry of Science and Technology of Taiwan (104-2314-B-010-015-MY2, 105-2314-B-016-004-, 104-2218-E-010-007-MY3, and 103-2321-B-010-017-); Taiwan National Science Council [98-2314-B-010-019-MY2, NSC 100-2628-E-010-002-MY3]; Taipei Veterans General Hospital [VGHUST101-G7-1-1, V101C-106, V101E7-003, VGHUST104-G7-1-1, V104C-082, V104E9-001]; Ministry of Science and Technology support for the Center for Dynamical Biomarkers and Translational Medicine, National Central University, Taiwan (MOST 103-2911-I-008-001); Academia Sinica (Grant No. IBMS-CRC103-P04); NSC support for the Centre for Dynamical Biomarkers and Translational Medicine, National Central University, Taiwan [NSC100-2911-I-008-001]; Brain Research Center, National Yang-Ming University, Ministry of Health and Welfare (MOHW104-TDU-B-211-113-003); Tri-Service General Hospital [TSGH-C101-159]; and a grant from Ministry of Education, Aim for the Top University Plan.

References

Headache Classification Committee of the International Headache Society (IHS). The International Classification of Headache Disorders, 3rd edition (beta version). Cephalalgia 2013; 33: 629–808.

Pringsheim

. Cluster headache: evidence for a disorder of circadian rhythm and hypothalamic function. Can J Neurol Sci 2002; 29: 33–40.

Morelli

Pesaresi

Cafforio

et al.

Functional magnetic resonance imaging in episodic cluster headache. J Headache Pain 2009; 10: 11–14.

May

Bahra

Buchel

et al.

Hypothalamic activation in cluster headache attacks [see comments]. Lancet 1998; 352: 275–278.

May

. Hypothalamic deep-brain stimulation: target and potential mechanism for the treatment of cluster headache. Cephalalgia 2008; 28: 799–803.

Leone

Franzini

Proietti Cecchini

et al.

Success, failure, and putative mechanisms in hypothalamic stimulation for drug-resistant chronic cluster headache. Pain 2013; 154: 89–94.

Sprenger

Ruether

Boecker

et al.

Altered metabolism in frontal brain circuits in cluster headache. Cephalalgia 2007; 27: 1033–1042.

Qiu

Liu

et al.

Altered regional homogeneity in spontaneous cluster headache attacks: a resting-state functional magnetic resonance imaging study. Chin Med J 2012; 125: 705–709.

Chou

Yang

Fuh

et al.

Altered white matter microstructural connectivity in cluster headaches: A longitudinal diffusion tensor imaging study. Cephalalgia 2014; 34: 1040–1052.

10.

Yang

Chou

Fuh

et al.

Altered gray matter volume in the frontal pain modulation network in patients with cluster headache. Pain 2013; 154: 801–807.

11.

Rocca

Valsasina

Absinta

et al.

Central nervous system dysregulation extends beyond the pain-matrix network in cluster headache. Cephalalgia 2010; 30: 1383–1391.

12.

Qiu

Wang

et al.

Abnormal brain functional connectivity of the hypothalamus in cluster headaches. PLoS One 2013; 8: e57896–e57896.

13.

Yang

Chou

Fuh

et al.

Altered hypothalamic functional connectivity in cluster headache: a longitudinal resting-state functional MRI study. J Neurol Neurosurg Psychiatry 2015; 86: 437–445.

14.

Greicius

. Resting-state functional connectivity in neuropsychiatric disorders. Curr Opin Neurol 2008; 21: 424–430.

15.

Qiu

Tian

Wang

et al.

Abnormal coactivation of the hypothalamus and salience network in patients with cluster headache. Neurology 2015; 84: 1402–1408.

16.

Jenkinson

Beckmann

Behrens

et al.

FSL. Neuroimage 2012; 62: 782–790.

17.

Boubela

Kalcher

Huf

et al.

Beyond noise: Using temporal ICA to extract meaningful information from high-frequency fMRI signal fluctuations during rest. Front Hum Neurosci 2013; 7: 168–168.

18.

Niazy

Xie

Miller

et al.

Spectral characteristics of resting state networks. Prog Brain Res 2011; 193: 259–276.

19.

Beckmann

Smith

. Probabilistic independent component analysis for functional magnetic resonance imaging. IEEE Trans Med Imaging 2004; 23: 137–152.

20.

Bhaganagarapu

Jackson

Abbott

. An automated method for identifying artifact in independent component analysis of resting-state FMRI. Front Hum Neurosci 2013; 7: 343–343.

21.

Kelly

Alexopoulos

Wang

et al.

Visual inspection of independent components: Defining a procedure for artifact removal from fMRI data. J Neurosci Methods 2010; 189: 233–245.

22.

Kiviniemi

Starck

Remes

et al.

Functional segmentation of the brain cortex using high model order group PICA. Hum Brain Mapp 2009; 30: 3865–3886.

23.

Salimi-Khorshidi

Douaud

Beckmann

et al.

Automatic denoising of functional MRI data: combining independent component analysis and hierarchical fusion of classifiers. Neuroimage 2014; 90: 449–468.

24.

Ashburner

Friston

. Voxel-based morphometry – the methods. Neuroimage 2000; 11: 805–821.

25.

Filippini

MacIntosh

Hough

et al.

Distinct patterns of brain activity in young carriers of the APOE-epsilon4 allele. Proc Natl Acad Sci USA 2009; 106: 7209–7214.

26.

Smith

Nichols

. Threshold-free cluster enhancement: Addressing problems of smoothing, threshold dependence and localisation in cluster inference. Neuroimage 2009; 44: 83–98.

27.

Apkarian

Bushnell

Treede

R-D

et al.

Human brain mechanisms of pain perception and regulation in health and disease. Eur J Pain 2005; 9: 463–484.

28.

Steele

Weller

. Subcortical connections of subdivisions of inferior temporal cortex in squirrel monkeys. Vis Neurosci 1993; 10: 563–583.

29.

May

Leone

Boecker

et al.

Hypothalamic deep brain stimulation in positron emission tomography. J Neurosci 2006; 26: 3589–3593.

30.

Holland

. Modulation of trigeminovascular processing: Novel insights into primary headache disorders. Cephalalgia 2009; 29: 1–6.

31.

Bushnell

Čeko

Low

. Cognitive and emotional control of pain and its disruption in chronic pain. Nat Rev Neurosci 2013; 14: 502–511.

32.

Goulden

Khusnulina

Davis

et al.

The salience network is responsible for switching between the default mode network and the central executive network: Replication from DCM. Neuroimage 2014; 99: 180–190.

33.

Beissner

Meissner

Bär

K-J

et al.

The autonomic brain: an activation likelihood estimation meta-analysis for central processing of autonomic function. J Neurosci 2013; 33: 10503–10511.

34.

Buckner

Andrews-Hanna

Schacter

. The brain’s default network: anatomy, function, and relevance to disease. Ann N Y Acad Sci 2008; 1124: 1–38.

35.

Tessitore

Russo

Giordano

et al.

Disrupted default mode network connectivity in migraine without aura. J Headache Pain 2013; 14: 89–89.

36.

Wandell

Dumoulin

Brewer

. Visual field maps in human cortex. Neuron 2007; 56: 366–383.

37.

Absinta

Rocca

Colombo

et al.

Selective decreased grey matter volume of the pain-matrix network in cluster headache. Cephalalgia 2012; 32: 109–115.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.09 MB