Resting-state fmri functional connectivity: a new perspective to evaluate pain modulation in migraine?

DOI 10.1007/s10072-015-2145-x NEUROIMAGING OF HEADACHES Resting-state fmri functional connectivity: a new perspective to evaluate pain modulation in migraine? Bruno Colombo Maria Assunta Rocca Roberta Messina Simone Guerrieri Massimo Filippi Ó Springer-Verlag Italia 2015 Abstract Resting-state (RS) functional magnetic resonance imaging (fmri) is a relatively novel tool which explores connectivity between functionally linked, but anatomically separated, brain regions. The use of this technique has allowed the identification, at rest, of the main brain functional networks without requiring subjects to perform specific active tasks. Methodologically, several approaches can be applied for the analysis of RS fmri, including seed-based, independent component analysisbased and/or cluster-based methods. The most consistently described RS network is the so-called default mode network. Using RS fmri, several studies have identified functional connectivity abnormalities in migraine patients, mainly located at the level of the pain-processing network. RS functional connectivity is generally increased in painprocessing network, whereas is decreased in pain modulatory circuits. Significant abnormalities of RS functional connectivity occur also in affective networks, the default mode network and the executive control network. These results provide a strong characterization of migraine as a brain dysfunction affecting intrinsic connectivity of brain networks, possibly reflecting the impact of long lasting pain on brain function. B. Colombo (&) S. Guerrieri Headache Research Unit, Department of Neurology, San Raffaele Scientific Institute, Vita-Salute San Raffaele University, Via Olgettina 60, Milan, Italy e-mail: colombo.bruno@hsr.it M. A. Rocca R. Messina M. Filippi Neuroimaging Research Unit, Division of Neuroscience, Institute of Experimental Neurology, San Raffaele Scientific Institute, Vita-Salute San Raffaele University, Via Olgettina 60, Milan, Italy Keywords Migraine Resting-state functional MRI Pain Functional connectivity Introduction If we consider network as a representation of a complex system made of finite number of nodes (the basic network element) and links (the connection between nodes), we can state that the brain should be defined as the most complex, unique and efficient network system. Neurons are the basic elements of this structural network: this anatomical substrate (approximately 86 billion neurons in the human brain) is able to create a coherent physiological activity, information processing and mental representations. In particular, brain network is composed of different brain regions with specific function, but that are continuously sharing information. In this way, they form a complex integrative network in which a stimulus is analyzed and transported in a continuous stream between structurally and functionally linked brain areas. A neural network is made of structurally and functionally interconnected areas at many levels (microscopic synapses, mesoscale circuitry and macroscale anatomical sites and fiber tracts). Connections can be either anatomical (a structural, physical connections between nodes appearing as fiber tracts) or functional (synchronous neuronal oscillations, considering that physical connection is not mandatory for a functional connection between nodes). The basic brain network connectivity, according to the graph theory of network analysis, has a short mean path length (related to high global efficiency and information transfer), high clustering (associated to robustness to random error), a distribution in agreement with the presence of hubs (nodes with high degree or high centrality) and a modular community

S42 structure (each module contains several densely interconnected nodes, with few connections between nodes in different modules) [1, 2]. Functional connectivity (FC) can be defined as the temporal dependency between spatially remote neurophysiological events [3, 4], and the study of this connectivity is made possible by functional magnetic resonance imaging (fmri). This technique allows describing the relationship between the neuronal activation patterns of anatomically separated brain areas, reflecting the quality of functional intercommunications between regions. Functional magnetic resonance imaging: the beginning of a new era in brain research fmri studies are based on the modification of blood oxygenation level-dependent (BOLD) signal derived from the neural response to an externally controlled task or stimulus. For this reason, BOLD signal is used as a non-invasive and indirect measure of changes in neuronal activity. fmri is a technique able to detect large-scale neural networks at different spatial and temporal resolutions. During active fmri tasks, the signal derived in the on period (the stimulus) is contrasted with that of a control condition, thus measuring a relative signal change caused by the ongoing process. For example, modifications of visual cortex fmri BOLD signal can be detected during visual stimulations. In the past few years, more attention has been focused on the analysis of the baseline state of the brain by measuring FC between brain areas as the level of co-activation of spontaneous fmri time series recorded during rest ( restingstate experiments ). In the resting-state (RS) studies, subjects are placed into the scanner and instructed to relax (with eyes closed or with a fixation point) and not to think of something without falling asleep, while their level of spontaneous brain activity is measured. RS fmri has certain properties that make it particularly advantageous for the study of neurological diseases: the limited behavioral demands of the acquisition procedures (the task is to rest quietly) and the simplicity of the data acquisition. First results of its application demonstrate that during rest, the right and left hemispheric areas of the motor network are not silent, showing a direct correlation with fmri BOLD time series, confirming ongoing information processing and coherent FC between these areas [5]. These results opened a new era in brain research. Because it was considered as a noise signal in task-response studies, this spontaneous aspect of the BOLD signal was until that moment minimized. In fact, noise is the modulation in a measured signal that is not related to the activity of interest. For this reason, minimization of noise through averaging is a simple method to amplify the effect in study. Neurol Sci (2015) 36 (Suppl 1):S41 S45 Nevertheless, there are some motivations to emphasize the interest on this noise signal. First, spontaneous BOLD activity is not random noise, but it is well organized. In particular, the spontaneous low-frequency fluctuations (0.01 0.1 Hz) observed in the BOLD signal have been interpreted to display spatial structure linked to task-related activation [6]. Cardiac and respiratory oscillations have been detected with a complete different frequency pattern ([0.3 Hz). Is still debated if these fluctuations in BOLD signal are due to changes in the brain physiology independent of neuronal function or reflect neuronal baseline activity of the brain when external input or task-directed neuronal activity is not present [7, 8]. The assumption that coherencies in resting fluctuations are the representation of functional RS networks linked to underlying neuronal modulation is consistent with the replication of these coherencies in specific functional brain gray matter areas (i.e., motor network, brain regions involved in language, auditory and visual processing) [9, 10]. The large body of results in this field underlies that during rest the brain network is not idle, but rather shows a large amount of spontaneous activity that is strictly correlated between multiple specific brain areas [11]. Therefore, RS BOLD fluctuations of cortical and subcortical regions reflect spontaneous neuronal activity. The related observed temporal correlation between fmri time series and specific anatomically separated brain areas is the reflex of a level of ongoing FC between brain regions during rest. One important and validated way to study the functional connections of a particular brain area is to correlate the RS series of such an area of interest (a priori choice, investigatordriven region of interest-based approach) against the time series of other regions [12]. The final result is a FC map defining the functional connections of predefined brain region (the so-called seed), which provides robust information about the functional links between the seed and other brain regions. Alternatively, to examine whole-brain connectivity patterns, model-free methods have been developed (without the need of defining an a priori seed region) [13], including principal components analysis (PCA), independent component analysis (ICA) and normalized cut clustering. ICA-based methods (data-driven methods) show a very good level of consistency and can be applied to whole-brain voxel-wise data. It is important to state that ICA, clustering and seed methods show a high level of overlap. The ICA approach enables identification of several networks consisting of spatially independent and temporally correlated areas. Group RS analysis (although using different methods and different acquisition protocols) has consistently identified specific functionally linked sub-networks in resting conditions, the so-called RS networks. They consist of functionally linked brain regions, although anatomically

separated, showing a high level of ongoing FC at rest and include the primary sensorimotor network, the primary visual and extra-striate visual network, right and left lateralized networks consisting of superior parietal and superior frontal regions, the salience network which is an anterior cingulated/frontoinsular system with links to limbic and subcortical autonomic control areas, and the socalled default mode network consisting of posterior cingulated cortex-precuneus/medial temporal/lateral temporoparietal/medial frontal network that is often deactivated during cognitively demanding tasks [14 16]. The default mode network is thought to mediate processes that are important for the RS, and its response to cognitive tasks is indeed unique. Most of these RS networks tend to represent specific functional networks, overlapping brain areas that are known to share a common function (primary motor areas, primary visual regions and parietal frontal networks involved in attention processing) [17]. RS networks may also show an internal specific topology that is strictly organized to their sub-functions [18]. It is conceivable that FC may give help to stabilize functional systems in an active state, improving performance and their reaction time. In the absence of an active and specific task, these networks show evidence of a tight spatial correspondence with the functioning circuits during sensorimotor, emotional and cognitive tasks [19]. The connectivity strength within these networks at rest has been demonstrated to correlate with emotional and cognitive situations, supporting the possibility to assess particular conditions in a disease state [20]. Recent investigations have demonstrated that almost all functionally linked regions of the RS networks are interconnected in a structural way (white matter tracts) [21]. This is in agreement with the concept of a general structural group of RS networks, linking structural and functional connectivity on a whole-brain scale. Resting-state functional connectivity in neurological disorders Analyses of RS time series have described a very efficient organization of functional communication in brain networks, showing that brain is not made of random networks, but is specialized in a complex system with a high level of local and global efficiency. One application of network theory is to provide a possible measure to quantify differences between controls and patient groups in parameters of brain networks derived from fmri (comparison of correlation patterns between groups). Disturbances in the correlation structure of spontaneous activity have been reported in a large number of neurological and psychiatric conditions. Recent studies have suggested a direct link between RS FC patterns and human cognition, and several works have detected possible functional disconnectivity effects in many neurological and psychiatric disorders, including Alzheimer s disease, multiple sclerosis, amyotrophic lateral sclerosis [22 24] and schizophrenia [25]. Most of these studies have been focused on the default mode network, although more recent works have examined the overall organization of functional brain networks. The results of these studies underline the decreased integrity of interconnected brain networks, rather than single brain areas: there is convergent evidence that abnormal topological organization of structural and functional brain networks could be hypothesized as the basic abnormality in degenerative neuropsychiatric pathologies [26]. Resting-state functional MRI connectivity in migraine S43 As non-invasive way to measure intrinsic fluctuations in BOLD signals, RS fmri has attracted considerable attention in migraine patients, to examine potential alterations of baseline intrinsic brain activity caused by long-term migraine attacks. In previous studies focused on patients suffering from chronic pain disturbances (e.g., chronic pain and fibromyalgia), default mode network (DMN) FC resulted disrupted, confirming that pain has a significant impact on brain activity and function, possibly affecting first dynamics of pain perception [27, 28]. In migraine patients, several studies were focused on potential alterations of baseline intrinsic brain activity caused by repetitive attacks. One RS FC study in migraine patients during the painfree state demonstrated alterations of baseline functional interaction within the periaqueductal gray matter (PAG) network, a key region involved in nociceptive processing [29]. This dysfunction could be attributable to the impairment of the descending pain modulatory circuits. In another study (analyzing the temporal homogeneity of regional fmri BOLD signals) a focal functional alteration, significantly related to longer disease duration, was detected in pain-processing areas such as in rostral anterior cingulated cortex, prefrontal cortex and orbitofrontal cortex [30]. Another study of migraine without aura patients investigated whether brain regions with abnormal regional RS properties show dysfunctional connectivity. The results showed that these patients had altered RS spontaneous neuronal activity in pain-processing areas, including the left rostral anterior cingulated cortex, bilateral prefrontal cortex and right thalamus [32]. An investigation of migraine without aura patients found an aberrant RS FC within the salience and executive networks, as well as increased connectivity between the default mode network and the executive network and the insula (a critical region involved in pain processing). The

S44 correlation found between increased RS FC within the insula and the duration of migraine suggested that these abnormalities may be the consequence of a persistent central neural system dysfunction [31]. In a more recent study of migraine without aura patients, the prefrontal and temporal areas of the DMN, showed reduced RS FC [33]. Interestingly, these functional abnormalities were unrelated to detectable structural abnormalities or clinical and neuropsychological features of migraineurs. Since these regions are implicated in sensory-discriminative, integrative and cognitive pain functions in the well-known neurolimbic pain network, these findings open questions about a possible link with maladaptive brain response to repeated stress, specific functional damage and migraine. RS FC analysis has been combined with morphometric analysis to characterize central nervous system abnormalities in medication-overuse headache patients. This particular group of patients showed an altered RS FC in the DMN as well as an increased connectivity between the precuneus and hippocampal temporal areas, which did not have corresponding structural abnormalities. The authors stated that alterations in pain-processing networks might be due to long-lasting pain processes, whereas alterations in DMN could be related to addiction processes [34]. The impaired FC might precede alteration in brain areas morphology. Another recent paper investigated brain RS FC in the interictal phase of migraine with aura and found no difference between patients and controls [35]. To define how early network abnormalities occur in migraine patients, a very recent study explored RS FC and functional interaction among networks in pediatric patients with migraine, as well as their correlation with patients clinical characteristics [36]. RS fmri is particularly advantageous for the study of pediatric patients due to the limited behavioral demands of the acquisition procedure (young participants may in fact have difficulty performing a task paradigm in task-based fmri studies). Compared to pediatric controls, pediatric migraine patients had an increased RS FC of the precuneus of the DMN and the dorsolateral prefrontal cortex of the right working memory network. They also experienced a decreased RS FC of the anterior cingulum of the salience network and the temporo-parietal junction of the left working memory network. Functional network connectivity analysis detected a decreased communication between the visual and the fronto-parietal network in pediatric migraine patients compared to controls. No significant correlation was found between intra- and internetwork RS FC abnormalities and patients clinical characteristics. These data showed that RS FC abnormalities occur in pain-processing networks of pediatric migraine patients. Brain regions involved in cognition were selectively involved, suggesting that abnormalities of cognitive modulation of pain in migraine patients occur from an early stage of the disease. Moreover, the increased connectivity between the visual networks and the fronto-parietal attention network supports an enhanced attention to visual stimuli in these patients. Conclusions All these body of scientific data has to be interpreted with caution because of small sample sizes, patients clinical heterogeneity and sometimes lack of consistent methodological approach. Admittedly, over-interpretation of these results is a risk when subject s motion is present. Recent studies demonstrated that even small amounts of in-scanner subject motion systematically biases measured FC, with motion strongly related to age. Advances in MRI acquisition and post-processing techniques have been shown to mitigate motion artifact adequately [37, 38]. The unsolved question is whether functional changes are the cause or the consequence of a long-term migraine condition. The study of functional network dysfunction in migraine is in its infancy. The effort to harmonize imaging acquisition parameters and clinical phenotyping/genotyping strategies will provide an opportunity for integrating genomic data to understand how genetic spectrum may be associated with abnormalities in functional brain networks that lead to symptoms of migraine. Spontaneous activity in brain function, as observed through fluctuations in the fmri BOLD signal, could provide new insights to disentangle the knots about the interpretation of pain in migraine. Longitudinal studies are necessary to evaluate the specific timing of selective network involvement. Future topics to improve the knowledge of network organization will depend on the understanding of how brain structure influences brain function and vice versa. The integration of structural and functional connectivity into a common framework may lead to a more complete definition of the overall altered brain organization in migraine patients. Conflict of interest The authors certify that there is no actual or potential conflict of interest in relation to this article. References Neurol Sci (2015) 36 (Suppl 1):S41 S45 1. Sporns O, Kötter R (2004) Motifs in brain networks. PLoS Biol 2:1910 1918 2. Hagmann P, Cammoun L, Gigandet X et al (2008) Mapping the structural core of human cerebral cortex. PLoS Biol 6:e159 3. Aertsen AM, Gerstein GL, Habib MK et al (1989) Dynamics of neuronal firing correlation: modulation of effective connectivity. J Neurophysiol 61:900 917 4. Friston KJ, Frith CD, Liddle PF et al (1993) Functional connectivity: the principal-component analysis of large (PET) data sets. J Cereb Blood Flow Metab 13:5 14

5. Biswal B, Yetkin FZ, Haughton VM et al (1995) Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn Reson Med 34:537 541 6. Cordes D, Haughton VM, Arfanakis K et al (2001) Frequencies contributing to functional connectivity in the cerebral cortex in resting-state data. Am J Neuroradiol 22:1326 1333 7. Gusnard DA, Raichle ME (2001) Searching for a baseline: functional imaging and the resting human brain. Nat Rev Neurosci 2:685 694 8. Peltier SJ, Noll DC (2002) T2 dependence of low frequency functional connectivity. NeuroImage 16:985 992 9. Hampson M, Peterson BS, Skudlarski P et al (2002) Detection of functional connectivity using temporal correlations in MR images. Hum Brain Mapp 15:247 262 10. Hampson M, Olson IR, Leung HC et al (2004) Changes in functional connectivity of human MT/V5 with visual motion input. NeuroReport 15:1315 1319 11. Greicius M (2008) Resting-state functional connectivity in neuropsychiatric disorders. Curr Opin Neurol 21:424 430 12. Fox MD, Raichle ME (2007) Spontaneous fluctuations in brain activity observed with functional magnetic resonance imaging. Nat Rev Neurosci 8:700 711 13. Beckmann CF, DeLuca M, Devlin JT et al (2005) Investigations into resting-state connectivity using independent component analysis. Philos Trans R Soc Lond B Biol Sci 360:1001 1013 14. Buckner RL, Andrews-Hanna JR, Schacter DL (2008) The brain s default network: anatomy, function, and relevance to disease. Ann N Y Acad Sci 1124:1 38 15. Seeley WW, Menon V, Schatzberg AF et al (2007) Dissociable intrinsic connectivity networks for salience processing and executive control. J Neurosci 27:2349 2356 16. Fox MD, Snyder AZ, Vincent JL et al (2005) The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proc Natl Acad Sci USA 102:9673 9678 17. De Luca M, Beckmann CF, De Stefano N et al (2006) fmri resting state networks define distinct modes of long-distance interactions in the human brain. Neuroimage 29:1359 1367 18. Damoiseaux JS, Rombouts SA, Barkhof F et al (2006) Consistent resting-state networks across healthy subjects. Proc Natl Acad Sci USA 103:13848 13853 19. Smith SM, Fox PT, Miller KL et al (2009) Correspondence of the brain s functional architecture during activation and rest. Proc Natl Acad Sci USA 106:13040 13045 20. Hampson M, Tokoglu F, Sun Z et al (2006) Connectivity-behavior analysis reveals that functional connectivity between left BA39 and Broca s area varies with reading ability. Neuroimage 31:513 519 21. Van den Heuvel MP, Mandl RCW, Kahn RS et al (2009) Functionally linked resting state networks reflect the underlying structural connectivity architecture of the human brain. Hum Brain Mapp 30:3127 3141 22. Rombouts SA, Barkhof F, Goekoop R et al (2005) Altered resting state networks in mild cognitive impairment and mild Alzheimer s disease: an fmri study. Hum Brain Mapp 26:231 239 S45 23. Lowe MJ, Beall EB, Sakaie KE et al (2008) Resting state sensorimotor functional connectivity in multiple sclerosis inversely correlates with transcallosal motor pathway transverse diffusivity. Hum Brain Mapp 29:818 827 24. Mohammadi B, Kollewe K, Samii A et al (2009) Changes of resting state brain networks in amyotrophic lateral sclerosis. Exp Neurol 217:147 153 25. Liu Y, Liang M, Zhou Y et al (2008) Disrupted small-world networks in schizophrenia. Brain 131:945 26. Seeley WW, Crawford RK, Zhou J et al (2009) Neurodegenerative diseases target large-scale human brain networks. Neuron 62:42 52 27. Baliki MN, Geha PY, Apkarian AV et al (2008) Beyond feeling: chronic pain hurts the brain, disrupting the default-mode network dynamics. J Neurosci 28:1398 1403 28. Napadow V, LaCount L, Park K et al (2010) Intrinsic brain connectivity in fibromyalgia is associated with chronic pain intensity. Arthritis Rheum 62:2545 2555 29. Mainero C, Boshyan J, Hadjikhani N (2011) Altered functional magnetic resonance imaging resting-state connectivity in periaqueductal gray networks in migraine. Ann Neurol 70:838 845 30. Yu D, Yuan K, Zhao L et al (2012) Regional homogeneity abnormalities in patients with interictal migraine without aura: a resting-state study. NMR Biomed 2:806 812 31. Xue T, Yuan K, Zhao L et al (2012) Intrinsic brain network abnormalities in migraines without aura revealed in resting-state fmri. PLoS ONE 7:52927 32. Xue T, Yuan K, Cheng P et al (2013) Alterations of regional spontaneous neuronal activity and corresponding brain circuit changes during resting state in migraine without aura. NMR Biomed 26:1051 1058 33. Tessitore A, Russo A, Giordano A et al (2013) Disrupted default mode network connectivity in migraine without aura. J Headache Pain 14:89 34. Chanraud S, Di Scala G, Dilharreguy B et al (2014) Brain functional connectivity and morphology changes in medicationoveruse headache: clue for dependence-related processes? Cephalalgia 34:605 615 35. Hougaard A, Amin FM, Magon S et al (2015) No abnormalities of intrinsic brain connectivity in the interictal phase of migraine with aura. Eur J Neurol. doi:10.1111/ene.12636 (Epub ahead of print) 36. Messina R, Rocca MA, Colombo B et al (2015) Resting state functional connectivity abnormalities in pediatric patients with migraine. Presented as oral communication, American Academy of Neurology, Washington DC, 2015 37. Satterthwaite TD, Wolf DH, Loughead J et al (2012) Impact of in-scanner head motion on multiple measures of functional connectivity: relevance for studies of neurodevelopment in youth. Neuroimage 60:623 632 38. Satterthwaite TD, Elliott MA, Gerraty RT et al (2013) An improved framework for confound regression and filtering for control of motion artifact in the preprocessing of resting-state functional connectivity data. Neuroimage 64:240 256