You are here

Article Text

Article menu

Download PDFPDF

Multiple sclerosis
Research paper
Staging of cortical and deep grey matter functional connectivity changes in multiple sclerosis
  1. Kim A Meijer,
  2. Anand J C Eijlers,
  3. Jeroen J G Geurts,
  4. Menno M Schoonheim
  1. Department of Anatomy and Neurosciences, VUmc MS Center Amsterdam, Amsterdam Neuroscience, VU University Medical Center, Amsterdam, The Netherlands
  1. Correspondence to Kim A Meijer, Department of Anatomy and Neurosciences, VU University Medical Center, 1007 MB Amsterdam, The Netherlands; k.meijer{at}vumc.nl

Abstract

Objective Functional connectivity is known to increase as well as decrease throughout the brain in multiple sclerosis (MS), which could represent different stages of the disease. In addition, functional connectivity changes could follow the atrophy pattern observed with disease progression, that is, moving from the deep grey matter towards the cortex. This study investigated when and where connectivity changes develop and explored their clinical and cognitive relevance across different MS stages.

Methods A cohort of 121 patients with early relapsing–remitting MS (RRMS), 122 with late RRMS and 53 with secondary progressive MS (SPMS) as well as 96 healthy controls underwent MRI and neuropsychological testing. Functional connectivity changes were investigated for (1) within deep grey matter connectivity, (2) connectivity between the deep grey matter and cortex and (3) within-cortex connectivity. A post hoc regional analysis was performed to identify which regions were driving the connectivity changes.

Results Patients with late RRMS and SPMS showed increased connectivity of the deep grey matter, especially of the putamen and palladium, with other deep grey matter structures and with the cortex. Within-cortex connectivity was decreased, especially for temporal, occipital and frontal regions, but only in SPMS relative to early RRMS. Deep grey matter connectivity alterations were related to cognition and disability, whereas within-cortex connectivity was only related to disability.

Conclusion Increased connectivity of the deep grey matter became apparent in late RRMS and further increased in SPMS. The additive effect of cortical network degeneration, which was only seen in SPMS, may explain the sudden clinical deterioration characteristic to this phase of the disease.

https://doi.org/10.1136/jnnp-2017-316329

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Introduction

    Multiple sclerosis (MS) is a chronic neuroinflammatory disease of the central nervous system and is characterised by progressive demyelination and neurodegeneration of the grey and white matter.1 In most patients, the disease starts with a relapsing–remitting (RR) course, which is followed by a secondary progressive (SP) phase.2 The average time to conversion is around 19 years; by then, approximately 70% of patients with RRMS will have developed SPMS.3 SPMS has a particularly unfavourable prognosis as patients experience a progressive worsening of their clinical symptoms.4 It seems that in early RRMS especially deep grey matter (DGM) atrophy occurs,5 while SPMS is characterised by more pronounced grey matter atrophy throughout the brain, now also involving cortical regions.6 7 Unfortunately, disease progression can only be partly explained by the gradual accumulation of structural damage, which highlights the importance of studying brain function in MS.

    Abnormalities in functional connectivity between regions throughout the brain have been reported shortly after diagnosis,5 8 but as the disease progresses towards SPMS, functional connectivity alterations become more prominent.9 10 However, whether functional connectivity changes follow the apparent pattern of atrophy, that is, moving from DGM structures towards the cortex, and how this relates to clinical and cognitive functions, is unclear. As it is recently shown that an individual’s functional brain network closely reflects differences in behaviour,11 distinct functional network profiles could provide essential information on the underlying mechanism of the clinico-cognitive decline in MS. This approach could help us to better understand the sudden clinical deterioration in SPMS.

    In this study, we aimed to assess whether functional connectivity changes follow the apparent pattern of neurodegeneration, that is, moving from DGM structures towards the cortex, and how this relates to cognition. As such, we investigated connectivity changes in within-DGM structures, between DGM and cortex and within-cortex in early RRMS, late RRMS and SPMS.

    Methods

    Subjects

    All subjects were part of the Amsterdam MS cohort5 12; in total, 96 healthy controls (HC; age 45.9±10.5 years) and 296 patients with MS (age 47.1±10.8 years) were included in this study. Average disease duration of the patients with MS was 12.6 (range 4.6–45.9 years). Of the patients with MS, 243 were diagnosed with RRMS and 53 with SPMS. Of the patients with RRMS, 121 were classified as early RRMS and 122 as late RRMS, based on a median split on disease duration (median 9.9 years). All patients were free from clinical relapses for at least 2 months and were not on steroid treatment prior to participation in the study. The Expanding Disease Severity Scale (EDSS) was used to measure overall disability. The institutional medical ethics committee approved the study. Written and informed consent was obtained from all participants.

    Neuropsychological assessment

    As described previously,5 12 neuropsychological examination consisted of an expanded version of the Brief Repeatable Battery of neuropsychological tests. Cognitive test scores were categorised into seven different cognitive domains, including information processing speed (Symbol Digit Modalities Task), executive functioning (Concept Shifting Task), working memory (Memory Comparison Task), verbal memory (Selective Reminding Test), verbal fluency (Word List Generation Test), visuospatial memory (Spatial Recall Test) and attention (Stroop). For each test, z-scores were computed relative to the mean and SD of HC, which were averaged to obtain one composite score for cognitive performance. Patients were defined as cognitively impaired if their score was at least 2 SD below that of the HC on a minimum of two tests.

    Magnetic Resonance Imaging

    MRI scanning was performed on a 3T GE system using an eight-channel head coil. All subjects underwent a 3D T1-weighted inversion-prepared fast spoiled gradient recall sequence (repetition time (TR) 7.8 ms, echo time (TE)3 ms, inversion time (TI) 450 ms, flip angle (FA) 12°, sagittal 1.0 mm sections, 0.94×0.94 mm2 in-plane resolution) for volumetric measurements, and a three-dimensional (3D) fluid-attenuated inversion-recovery sequence (FLAIR, TR 8000 ms, TE 125 ms, TI 2350 ms, sagittal 1.2 mm slices, 0.98×0.98 mm2 in-plane resolution) for lesion detection. Additionally, resting-state functional MRI (fMRI) scans were acquired with whole-brain coverage using 202 volumes, of which the first two were discarded (EPI, TR 2200 ms, TE 35 ms, FA 20°, 3 mm contiguous axial slices, 3.3×3.3 mm2 in-plane resolution).

    Structural brain measures

    An automated lesion segmentation method was used to detect white matter (WM) lesions on FLAIR images13 and to determine lesion volumes. Lesion filling (LEAP; Lesion Automated Preprocessing) was conducted on 3D T1 scans to reduce the impact of WM lesions on brain tissue segmentation.14 Normalised brain volumes, including whole brain (NBV), GM and WM, were then determined using SIENAX. FIRST (FSL5) was used to segment DGM structures and calculate normalised DGM volume (NDGMV). Additionally, normalised cortical volumes (NCGMV) were calculated by subtracting FIRST regions from the SIENAX-based GM mask.

    fMRI preprocessing

    Data preprocessing was carried out using the default pipeline of MELODIC (FSL5), consisting of motion correction, removal of non-brain tissue, spatial smoothing using a 5 mm full-width-at-half-maximum Gaussian kernel and high-pass temporal filtering equivalent to 0.01 Hz. All resting-state fMRI scans were checked for registration errors, artefacts and excessive motion. The individual level of motion was calculated based on the average frame-to-frame motion. The amount of motion was not different between HC and MS (p=0.34), and no subject moved more than 0.3 mm.

    Region of interest atlas

    For each subject, cortical and DGM regions were segmented with a custom-made atlas as reported previously.5 Cortical regions were derived from the standard space AAL atlas (part of MRIcro; see http://www.cabiatl.com/mricro/mricro/template.html), which was registered to the subject’s T1-weighted scans using inverted non-linear registration parameters and nearest-neighbour interpolation. To include only GM, the cortical AAL atlas was masked by the SIENAX-based GM mask. Next, DGM structures were added to this masked cortical AAL atlas. The complete individualised atlas was registered to each subject’s fMRI scan, using an inverted boundary-based registration (BBR) matrix and nearest-neighbour interpolation.

    After registration and GM masking, voxels with a signal intensity in the lowest quartile of the robust range were excluded to reduce the influence of residual echo-planar imaging (EPI) distortions. Only regions of interest (ROIs) from the custom-made atlas with at least 30% voxels remaining after exclusion of less reliable fMRI signal were included. Based on these criteria, 12 ROIs were excluded, compromising bilateral orbitofrontal areas and nucleus accumbens. The final atlas therefore segmented the fMRI sequence into 80 GM regions of which mean time series were obtained. Of this final atlas, regions were assigned to either DGM regions (ie, thalamus, nucleus caudatus, putamen, globus pallidus, hippocampus, amygdala) or cortical regions (the remaining 68 regions).

    Functional connectivity

    The connectivity matrices were created based on Pearson correlation coefficients between time series of all possible pairs of the 80 ROIs. Only positive correlation coefficients were entered in the connectivity matrix while negative correlation coefficients were set to zero given the controversial nature of negative connectivity values.15 16 Relative connectivity scores were computed based on the mean and SD of an individual functional connectivity matrix.12 This step was chosen to be able to look at the ranking of each brain region within its individual brain network and thereby correcting for global intersubject variability in mean connectivity. Based on these z-score connectivity matrices, average scores were computed for (1) within-DGM connectivity, (2) DGM–cortex connectivity and (3) within-cortex connectivity.

    Statistical analyses

    Statistical analyses were performed using SPSS. Variables were checked for normality using Kolmogorov-Smirnov testing and histogram inspection. Differences in descriptive characteristics were assessed using χ2 tests for categorical variables, Mann-Whitney U tests for non-normally distributed variables or general linear models (GLMs) for normally distributed variables. GLM assessed differences in the connectivity measures. All analyses were corrected for sex and education. As this study specifically investigated different RR phases strongly linked to disease duration, we did not correct for age. However, to explore age effects, the HC group was subdivided into three different age groups according to the boundaries of the 25th and 75th age percentile of the three different MS groups. Additionally, each connectivity measure was related to EDSS, the most severely affected cognitive domain, averaged cognition and structural MRI measures using Pearson correlations or Spearman’s rho. Bonferroni corrected p values were reported and p<0.05 was considered as significant.

    For each of the significant connectivity measures, a post hoc regional analysis was performed to identify which regions drove the changes using a GLM with education and sex as covariates. For instance, for DGM–cortex connectivity, each individual DGM structure was assessed on its connectivity with the cortex, for example the averaged connectivity of the thalamus with all cortical AAL regions. For within-cortex connectivity, the cortical AAL areas were averaged into the four major cortical brain lobes to limit multiple comparison issues.

    Results

    Demographics and clinical characteristics

    Demographic, clinical and MRI data are presented in table 1. As expected, age and disease duration were significantly different between all groups. The late RRMS group contained significantly more men than the HC group. EDSS was higher in SPMS relative to early and late RRMS (both p<0.001), and higher in late RRMS than early RRMS (p<0.05).

    View this table:
    Table 1

    Demographic, clinical and MRI characteristics of included subjects

    MRI characteristics

    SPMS showed reductions in all GM volumes compared with all other groups. In addition, late RRMS showed reduced NGMV and NDGMV compared with both early RRMS and HC (all p<0.01). Patients with early RRMS showed reduced NDGMV compared with HC (p<0.01), but no change in NCGMV. Normalised WM volume (NWMV) was reduced in all patient groups relative to HC (p<0.001) and in SPMS compared with early RRMS (p<0.05). Lesion volume was different between all patient groups, indicating higher lesion volumes in more progressive stages (table 1).

    Cognitive profile

    Of our sample, 16% were cognitively impaired in early RRMS, 25% in late RRMS and 49% in SPMS. The averaged cognitive performance was worse in SPMS than in late RRMS (p=0.004), early RRMS (p<0.001) and HC (p<0.001), whereas late RRMS had lower scores than early RRMS (p<0.001) and HC (p<0.001). Early RRMS also performed worse compared with HC (p=0.001). A similar pattern was observed for other cognitive domains, where patients with SPMS scored worst, followed by late RRMS and then early RRMS (table 2). In MS, the most severely affected domain was information processing speed (IPS; mean z-score=−1.16).

    View this table:
    Table 2

    Cognitive profile of MS phenotypes

    Functional connectivity differences

    Group effects were observed for within-DGM connectivity (p<0.001; figure 1), DGM–cortex connectivity (p<0.001; figure 2) and within-cortex connectivity (p=0.01; figure 3). Post hoc analyses revealed that patients with SPMS showed higher within-DGM connectivity compared with patients with late and early RRMS (respectively p=0.002 and p=0.001), higher DGM–cortex connectivity compared with patients with early RRMS and HC (respectively p<0.001 and p<0.05) and lower within-cortex connectivity compared with patients with early RRMS (p<0.01). Additionally, late RRMS showed higher within-DGM and DGM–cortex connectivity than early RRMS (both p<0.05). To evaluate the effects of volumetric differences, we repeated the analysis while including NBV as an additional covariate, after which all our findings remained significant. No differences in connectivity were observed between the different HC age groups.

    Figure 1
    Figure 1

    Increased within-DGM connectivity in both LRRMS and SPMS relative to ERRMS. Positive connectivity z-scores reflect increases in the level of connectivity, whereas negative connectivity z-scores reflect decreases in the level of connectivity relative to healthy controls. Arms with * indicate significant differences and error bars reflect SE of the mean. DGM, deep grey matter; ERRMS, early relapsing–remitting multiple sclerosis; LRRMS, late relapsing–remitting multiple sclerosis; SPMS, secondary progressive multiple sclerosis.

    Figure 2
    Figure 2

    Increased connectivity between DGM and cortex in both LRRMS and SPMS relative to ERRMS. Positive connectivity z-scores reflect increases in the level of connectivity, whereas negative connectivity z-scores reflect decreases in the level of connectivity relative to healthy controls. Arms with * indicate significant differences and error bars reflect SE of the mean. DGM, deep grey matter; ERRMS, early relapsing–remitting multiple sclerosis; LRRMS, late relapsing–remitting multiple sclerosis; SPMS, secondary progressive multiple sclerosis.

    Figure 3
    Figure 3

    Decreased within-cortex connectivity in SPMS relative to ERRMS. Positive connectivity z-scores reflect increases in the level of connectivity, whereas negative connectivity z-scores reflect decreases in the level of connectivity. Arms with * indicate significant differences and error bars reflect SE of the mean. ERRMS, early relapsing–remitting multiple sclerosis; LRRMS, late relapsing–remitting multiple sclerosis; SPMS, secondary progressive multiple sclerosis.

    Correlations between functional, structural and clinical measures

    In MS, worse averaged cognition and IPS were associated with increased within-DGM (respectively r=−0.145 and r=−0.197, both p<0.01) and DGM–cortex connectivity (respectively r=−0.156 and r=−0.198, both p<0.01), as well as with lower NDGMV (respectively r=0.595 and r=0.562, both p<0.001) and lower NCGMV (respectively r=0.528 and r=0.518, both p<0.001). Higher EDSS scores were correlated with higher within-DGM (rho=0.276, p<0.001), higher DGM–cortex (rho=0.261, p<0.001) and lower within-cortex connectivity (rho=−0.219, p<0.001), as well as lower NCGMV (rho=−0.511, p<0.01) and lower NDGMV (rho=−0.466, p<0.001). Increased within-DGM and DGM–cortex connectivity were related to lower NCGMV (r=−0.184; p=0.002 and r=−0.173; p=0.003, respectively), but not with NDGMV.

    Regional analyses

    Increased within-DGM connectivity was driven by all other DGM regions (ranging from p<0.001 to p=0.02), but most strongly by the putamen and hippocampus. Patients with late RRMS and SPMS showed increased connectivity of all separate DGM regions with the remaining DGM regions compared with patients with early RRMS (all p<0.05). Increased DGM–cortex connectivity was driven by putamen–cortex (p<0.001) and pallidum–cortex (p<0.001). In patients with late RRMS, increased putamen–cortex connectivity was observed compared with patients with early RRMS (p=0.003). In SPMS, increased putamen–cortex and pallidum–cortex connectivity was observed relative to early RRMS (both p<0.001) and HC (respectively p=0.04 and p=0.02). All these DGM connectivity measures (ie, putamen–cortex, pallidum–cortex, putamen–DGM and hippocampus–DGM connectivity) were correlated to EDSS (ranging from rho=0.213 to rho=0.278, p<0.01), whereas only putamen–cortex and hippocampus–DGM were correlated to cognition (respectively r=−0.188 and r=−0.171, both p<0.01). Decreased within-cortex connectivity was driven by frontal (p=0.009), temporal (p=0.001) and occipital lobes (p=0.02). For all these measures, decreased connectivity was only observed in SPMS relative to early RRMS (all p<0.05).

    Discussion

    In the present study, we investigated whether the functional connectivity within and between DGM and cortical regions differs across early RRMS, late RRMS and SPMS. As expected, functional connectivity patterns were most abnormal in SPMS. Across the different disease phases, we observed that within-DGM and DGM–cortex connectivity became stronger, whereas within-cortex connectivity reduced. At the structural level, DGM volume reduced stepwise from early RRMS to late RRMS and then to SPMS. These functional and structural changes were related to cognitive and clinical performance.

    Shifts in functional network balance

    Very few studies have investigated functional connectivity abnormalities across MS phenotypes.9 10 17 18 The earliest studies have shown increased connectivity in clinically isolated syndrome17 as well as decreased connectivity in progressive MS9 relative to HC. Increased connectivity was interpreted as a compensatory mechanism observed only in the early phases of the disease, to preserve clinical function. Decreased connectivity was explained as a maladaptive change, caused by accumulating structural damage.19 More recent studies, however, have related both increases and decreases in functional connectivity to worse cognition and disease progression, adding to a body of literature that has become difficult to interpret.20 Because we were especially interested in functional network profiles, we normalised our connectivity measures for individual overall means and SD to be more sensitive to changes in network topology. This allowed us to specifically examine local shifts in the functional brain network. That is, we studied which connections became stronger or weaker within an individual brain network. In doing so, we observed that, while neurodegeneration is already present in early stages of the disease, the functional network is maintained in early RRMS. It was not until late RRMS that connectivity changes became apparent, which became even greater in SPMS. Increased within-DGM and DGM–cortex connectivity were related to worse cognition. In line with current literature,21–23 patients with SPMS had the worst cognitive performance and scored lower on all examined cognitive domains, whereas the patients with early RRMS only scored worse than HC on IPS and working memory. Previous functional connectivity studies often agreed with our findings in that increased connectivity was found to be related to cognitive deficits.20 For example, increased connectivity of the so-called default-mode network,8 10 24 25 thalamus5 26 and hippocampus27 was associated with cognitive deficits in MS; however, these studies did not evaluate different stages of relapse-onset MS. Nevertheless, it is expected that an increase in functional connectivity reflects cognitive deficits and a broader spectrum of disease burden, as supported by the correlation with disability in this study.

    Involvement of cortical and DGM regions in disease progression

    In this study, we specifically separated the DGM and cortex, as previous structural studies have suggested that early MS mostly features DGM involvement, whereas SPMS also features extensive cortical involvement. For instance, in early RRMS, atrophy of the thalamus was observed, together with increased functional connectivity between the thalamus and especially frontal regions as well as the caudate.5 In later RRMS, increased connectivity of the hippocampus was observed with both DGM regions as well as cortical regions, including the posterior and anterior cingulate.27 Reduced connectivity of resting-state networks involving cortical regions was also observed in RRMS relative to HC,28 but was especially severe in SPMS.9 Recently, another study reported similar widespread increased and decreased functional connectivity changes in another cohort of patients with MS, whereby the level of functional connectivity, of especially cortical regions, decreased along the disease course, while DGM regions showed increased connectivity.18 This study showed earlier changes in decreased cortical connectivity than ours, which may be due to differences in the methodological approaches. Nonetheless, it is striking that both studies show a disparity between DGM and cortical changes. Our analyses revealed that connectivity of the DGM was extensively affected, but not until late RRMS, while DGM atrophy was already present at the early RRMS stage. This connectivity change was especially severe for the putamen and pallidum, which worsened in SPMS. Additionally, only in SPMS, frontal, temporal and occipital regions showed a decrease in within-cortex connectivity, together with cortical atrophy, as was expected by aforementioned studies. Corresponding with previous results, especially GM volumes were able to discriminate between late RRMS and SPMS,6 whereas WM volume only differed between early RRMS and SPMS. In addition, only cortical volume was negatively associated with the extent of DGM connectivity, while DGM volume was not, possibly indicating the crucial role for cortical integrity for a normal network balance in MS. However, owing to the cross-sectional character and included global measures of structural damage, future longitudinal research is needed to elucidate the temporal and spatial association between levels of structural damage and functional connectivity changes.

    Our hypothesis of disease progression in relapse-onset MS

    Our findings could provide a framework to conceptualise the underlying mechanisms of functional network changes in different stages of MS. Atrophy of DGM regions is already observed during the early phases of the disease, probably because these structures are centrally located in the structural brain network, thus most vulnerable for lesional damage in their connected tracts, resulting in dying back of these axons and/or Wallerian degeneration. Because of this, connectivitychanges might start with disturbances in the interaction between DGM structures, but do not seem to involve the cortex as much at this stage. In turn, this disturbed within-DGM connectivity could possibly lead to an improper filtering of irrelevant information, as sensory and sensorimotor gating are major roles of DGM regions. This within-DGM gating provides the possibility to filter out and inhibit redundant input to prevent an overload of information to the higher cortical regions of the brain.29 30 This reduced gating could thus lead to the transfer of abnormally high levels of information to other DGM structures and also from these structures to the cortex. In this way, local effects within the DGM regions might subsequently result in stronger, yet maladaptive, connections between the DGM and cortex. This apparent disinhibition seems to be supported by the observed imbalance in the ratio of excitatory to inhibitory activity in MS.31 32 The specific loss of inhibitory signals in the macaque brain induced increased connectivity.33 This imbalance could be further investigated by studies relating GABA and glutamate spectroscopy measures to changes in functional connectivity. Finally, after a long period of increasing structural damage and inefficient network connections, the connectivity between cortical regions decreases in SPMS, a sign of incipient cortical network collapse that may be specific to this disease stage.20 Although our findings might indicate a pattern of network changes related to the conversion to SPMS, future longitudinal studies are now required to confirm our hypothesis.

    Scientific and clinical potential of fMRI-based connectivity measures

    From a scientific point of view, our findings can be of value in understanding how functional connectivity changes may underlie clinical and cognitive deterioration in MS. Although the current functional network field has provided key insights on the mechanism underlying clinico-cognitive impairment in MS, it has also been characterised by mostly heterogeneous findings, showing both increases and decreases throughout the brain.20 As these studies have usually included mixed patient groups, these observations could actually represent different stages of the disease (ie, an initial increase and later decrease in connectivity). Our results show that connectivity changes do not develop until late RRMS, where only increases in connectivity are seen, involving DGM structures. The additive effect of cortical network degeneration, which was only seen in SPMS, may explain the sudden clinical deterioration observed in this phase of the disease. Our findings can be seen as a first step associating distinct functional connectivity profiles to different MS stages, which may lead to a better prediction of who is at risk of developing SPMS, which is currently very difficult. To be able to translate our findings to clinical practice, however, we need to study the sensitivity and specificity of promising functional network biomarkers for treatment response34–37 and prognosis.38 Furthermore, as was recently shown, specific individual functional brain network characteristics have been identified that closely reflect differences in behaviour, which may indicate their potential for clinical studies.11 Future studies are therefore challenged to elaborate on these findings in order to link individual fMRI characteristics to clinical progression.39

    In conclusion, while neurodegeneration was already apparent in early phases of the disease, shifts in the functional network were not observed until the later phases of RRMS and SPMS. Loss of within-cortex connectivity was unique to SPMS. These functional network changes help to understand the clinical and cognitive deterioration along the disease course of MS.

    Acknowledgments

    The authors would like to thank the Dutch MS Research Foundation for supporting this study and the MS Center Amsterdam.

    References

    1. ↵
      2. Hauser SL ,
      3. Oksenberg JR
      . The neurobiology of multiple sclerosis: genes, inflammation, and neurodegeneration. Neuron 2006;52:61–76.doi:10.1016/j.neuron.2006.09.011
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      2. Lublin FD ,
      3. Reingold SC ,
      4. Cohen JA , et al
      . Defining the clinical course of multiple sclerosis: the 2013 revisions. Neurology 2014;83:278–86.doi:10.1212/WNL.0000000000000560
      OpenUrlCrossRefPubMed
    3. ↵
      2. Vukusic S ,
      3. Confavreux C
      . Natural history of multiple sclerosis: risk factors and prognostic indicators. Curr Opin Neurol 2007;20:269–74.doi:10.1097/WCO.0b013e32812583ad
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      2. Rovaris M ,
      3. Confavreux C ,
      4. Furlan R , et al
      . Secondary progressive multiple sclerosis: current knowledge and future challenges. Lancet Neurol 2006;5:343–54.doi:10.1016/S1474-4422(06)70410-0
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      2. Schoonheim MM ,
      3. Hulst HE ,
      4. Brandt RB , et al
      . Thalamus structure and function determine severity of cognitive impairment in multiple sclerosis. Neurology 2015;84:776–83.doi:10.1212/WNL.0000000000001285
      OpenUrlCrossRefPubMed
    6. ↵
      2. Fisher E ,
      3. Lee JC ,
      4. Nakamura K , et al
      . Gray matter atrophy in multiple sclerosis: a longitudinal study. Ann Neurol 2008;64:255–65.doi:10.1002/ana.21436
      OpenUrlCrossRefPubMedWeb of Science
    7. ↵
      2. Dutta R ,
      3. Trapp BD
      . Relapsing and progressive forms of multiple sclerosis: insights from pathology. Curr Opin Neurol 2014;27:271–8.doi:10.1097/WCO.0000000000000094
      OpenUrlCrossRefPubMed
    8. ↵
      2. Faivre A ,
      3. Rico A ,
      4. Zaaraoui W , et al
      . Assessing brain connectivity at rest is clinically relevant in early multiple sclerosis. Mult Scler 2012;18:1251–8.doi:10.1177/1352458511435930
      OpenUrlCrossRefPubMed
    9. ↵
      2. Rocca MA ,
      3. Valsasina P ,
      4. Absinta M , et al
      . Default-mode network dysfunction and cognitive impairment in progressive MS. Neurology 2010;74:1252–9.doi:10.1212/WNL.0b013e3181d9ed91
      OpenUrlCrossRefPubMed
    10. ↵
      2. Basile B ,
      3. Castelli M ,
      4. Monteleone F , et al
      . Functional connectivity changes within specific networks parallel the clinical evolution of multiple sclerosis. Mult Scler 2014;20:1050–7.doi:10.1177/1352458513515082
      OpenUrlCrossRefPubMed
    11. ↵
      2. Finn ES ,
      3. Shen X ,
      4. Scheinost D , et al
      . Functional connectome fingerprinting: identifying individuals using patterns of brain connectivity. Nat Neurosci 2015;18:1664–71.doi:10.1038/nn.4135
      OpenUrlCrossRefPubMed
    12. ↵
      2. Meijer KA ,
      3. Eijlers AJC ,
      4. Douw L , et al
      . Increased connectivity of hub networks and cognitive impairment in multiple sclerosis. Neurology 2017;88:2107–14.doi:10.1212/WNL.0000000000003982
      OpenUrl
    13. ↵
      2. Steenwijk MD ,
      3. Pouwels PJ ,
      4. Daams M , et al
      . Accurate white matter lesion segmentation by k nearest neighbor classification with tissue type priors (kNN-TTPs). Neuroimage Clin 2013;3:462–9.doi:10.1016/j.nicl.2013.10.003
      OpenUrlCrossRefPubMed
    14. ↵
      2. Chard DT ,
      3. Jackson JS ,
      4. Miller DH , et al
      . Reducing the impact of white matter lesions on automated measures of brain gray and white matter volumes. J Magn Reson Imaging 2010;32:223–8.doi:10.1002/jmri.22214
      OpenUrlCrossRefPubMed
    15. ↵
      2. Chai XJ ,
      3. Castañón AN ,
      4. Ongür D , et al
      . Anticorrelations in resting state networks without global signal regression. Neuroimage 2012;59:1420–8.doi:10.1016/j.neuroimage.2011.08.048
      OpenUrlCrossRefPubMedWeb of Science
    16. ↵
      2. Fox MD ,
      3. Snyder AZ ,
      4. Vincent JL , et al
      . The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proc Natl Acad Sci U S A 2005;102:9673–8.doi:10.1073/pnas.0504136102
      OpenUrlAbstract/FREE Full Text
    17. ↵
      2. Roosendaal SD ,
      3. Schoonheim MM ,
      4. Hulst HE , et al
      . Resting state networks change in clinically isolated syndrome. Brain 2010;133:1612–21.doi:10.1093/brain/awq058
      OpenUrlCrossRefPubMedWeb of Science
    18. ↵
      2. Rocca MA ,
      3. Valsasina P ,
      4. Leavitt VM , et al
      . Functional network connectivity abnormalities in multiple sclerosis: Correlations with disability and cognitive impairment. Mult Scler 2017:135245851769987.doi:10.1177/1352458517699875
    19. ↵
      2. Schoonheim MM ,
      3. Geurts JJ ,
      4. Barkhof F
      . The limits of functional reorganization in multiple sclerosis. Neurology 2010;74:1246–7.doi:10.1212/WNL.0b013e3181db9957
      OpenUrlCrossRefPubMed
    20. ↵
      2. Schoonheim MM ,
      3. Meijer KA ,
      4. Geurts JJ
      . Network collapse and cognitive impairment in multiple sclerosis. Front Neurol 2015;6:82.doi:10.3389/fneur.2015.00082
      OpenUrlPubMed
    21. ↵
      2. Denney DR ,
      3. Sworowski LA ,
      4. Lynch SG
      . Cognitive impairment in three subtypes of multiple sclerosis. Arch Clin Neuropsychol 2005;20:967–81.doi:10.1016/j.acn.2005.04.012
      OpenUrlCrossRefPubMedWeb of Science
    22. ↵
      2. Huijbregts SC ,
      3. Kalkers NF ,
      4. de Sonneville LM , et al
      . Cognitive impairment and decline in different MS subtypes. J Neurol Sci 2006;245:187–94.doi:10.1016/j.jns.2005.07.018
      OpenUrlCrossRefPubMedWeb of Science
    23. ↵
      2. Potagas C ,
      3. Giogkaraki E ,
      4. Koutsis G , et al
      . Cognitive impairment in different MS subtypes and clinically isolated syndromes. J Neurol Sci 2008;267:100–6.doi:10.1016/j.jns.2007.10.002
      OpenUrlCrossRefPubMedWeb of Science
    24. ↵
      2. Hawellek DJ ,
      3. Hipp JF ,
      4. Lewis CM , et al
      . Increased functional connectivity indicates the severity of cognitive impairment in multiple sclerosis. Proc Natl Acad Sci U S A 2011;108:19066–71.doi:10.1073/pnas.1110024108
      OpenUrlAbstract/FREE Full Text
    25. ↵
      2. Tewarie P ,
      3. Schoonheim MM ,
      4. Stam CJ , et al
      . Cognitive and clinical dysfunction, altered MEG resting-state networks and thalamic atrophy in multiple sclerosis. PLoS One 2013;8:e69318.doi:10.1371/journal.pone.0069318
    26. ↵
      2. Tona F ,
      3. Petsas N ,
      4. Sbardella E , et al
      . Multiple sclerosis: altered thalamic resting-state functional connectivity and its effect on cognitive function. Radiology 2014;271:814–21.doi:10.1148/radiol.14131688
      OpenUrlCrossRefPubMedWeb of Science
    27. ↵
      2. Hulst HE ,
      3. Schoonheim MM ,
      4. Van Geest Q , et al
      . Memory impairment in multiple sclerosis: relevance of hippocampal activation and hippocampal connectivity. Mult Scler 2015;21:1705–12.doi:10.1177/1352458514567727
      OpenUrlCrossRefPubMed
    28. ↵
      2. Rocca MA ,
      3. Valsasina P ,
      4. Martinelli V , et al
      . Large-scale neuronal network dysfunction in relapsing–remitting multiple sclerosis. Neurology 2012;79:1449–57.doi:10.1212/WNL.0b013e31826d5f10
      OpenUrlCrossRefPubMed
    29. ↵
      2. Kaji R ,
      3. Urushihara R ,
      4. Murase N , et al
      . Abnormal sensory gating in basal ganglia disorders. J Neurol 2005;252 Suppl 4:iv13–16.doi:10.1007/s00415-005-4004-9
      OpenUrl
    30. ↵
      2. Schmidt R ,
      3. Leventhal DK ,
      4. Mallet N , et al
      . Canceling actions involves a race between basal ganglia pathways. Nat Neurosci 2013;16:1118–24.doi:10.1038/nn.3456
      OpenUrlCrossRefPubMed
    31. ↵
      2. Dutta R ,
      3. McDonough J ,
      4. Yin X , et al
      . Mitochondrial dysfunction as a cause of axonal degeneration in multiple sclerosis patients. Ann Neurol 2006;59:478–89.doi:10.1002/ana.20736
      OpenUrlCrossRefPubMedWeb of Science
    32. ↵
      2. Clements RJ ,
      3. McDonough J ,
      4. Freeman EJ
      . Distribution of parvalbumin and calretinin immunoreactive interneurons in motor cortex from multiple sclerosis post-mortem tissue. Exp Brain Res 2008;187:459–65.doi:10.1007/s00221-008-1317-9
      OpenUrlCrossRefPubMedWeb of Science
    33. ↵
      2. van den Heuvel MP ,
      3. Scholtens LH ,
      4. Turk E , et al
      . Multimodal analysis of cortical chemoarchitecture and macroscale fMRI resting-state functional connectivity. Hum Brain Mapp 2016;37:3103–13.doi:10.1002/hbm.23229
      OpenUrlCrossRefPubMed
    34. ↵
      2. Fox MD ,
      3. Buckner RL ,
      4. White MP , et al
      . Efficacy of transcranial magnetic stimulation targets for depression is related to intrinsic functional connectivity with the subgenual cingulate. Biol Psychiatry 2012;72:595–603.doi:10.1016/j.biopsych.2012.04.028
      OpenUrlCrossRefPubMedWeb of Science
    35. ↵
      2. Liston C ,
      3. Chen AC ,
      4. Zebley BD , et al
      . Default mode network mechanisms of transcranial magnetic stimulation in depression. Biol Psychiatry 2014;76:517–26.doi:10.1016/j.biopsych.2014.01.023
      OpenUrlCrossRefPubMedWeb of Science
    36. ↵
      2. Dunlop K ,
      3. Woodside B ,
      4. Olmsted M , et al
      . Reductions in cortico-striatal hyperconnectivity accompany successful treatment of obsessive-compulsive disorder with dorsomedial prefrontal rTMS. Neuropsychopharmacology 2016;41:1395–403.doi:10.1038/npp.2015.292
      OpenUrl
    37. ↵
      2. Drysdale AT ,
      3. Grosenick L ,
      4. Downar J , et al
      . Resting-state connectivity biomarkers define neurophysiological subtypes of depression. Nat Med 2017;23:28–38.doi:10.1038/nm.4246
      OpenUrl
    38. ↵
      2. Carter AR ,
      3. Astafiev SV ,
      4. Lang CE , et al
      . Resting interhemispheric functional magnetic resonance imaging connectivity predicts performance after stroke. Ann Neurol 2010;67:365–75.doi:10.1002/ana.21905
      OpenUrlCrossRefPubMedWeb of Science
    39. ↵
      2. Dubois J ,
      3. Adolphs R
      . Building a science of individual differences from fMRI. Trends Cogn Sci 2016;20:425–43.doi:10.1016/j.tics.2016.03.014
      OpenUrlCrossRefPubMed
    View Abstract

    Footnotes

    • Contributors Drafting/revising the manuscript: all authors. Study concept or design: KAM, JJGG, MMS. Acquisition of data: MMS. Analysis or interpretation of the data: KAM, AJCE, MMS. Statistical analysis: KAM, AJCE, MMS. Study supervision and coordination: JJGG, MMS. Obtaining funding: JJGG, MMS.

    • Funding This study was supported by the Dutch MS Research Foundation, grant numbers 08-650, 13-820 and 14-358e.

    • Competing interests None declared.

    • Ethics approval Ethics Review Board of VU University Medical Center.

    • Provenance and peer review Not commissioned; externally peer reviewed.