Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

Functional Connectivity Alterations Reveal Complex Mechanisms Based on Clinical and Radiological Status in Mild Relapsing Remitting Multiple Sclerosis

Functional Connectivity Alterations Reveal Complex Mechanisms Based on Clinical and Radiological... ORIGINAL RESEARCH published: 20 August 2018 doi: 10.3389/fneur.2018.00690 Functional Connectivity Alterations Reveal Complex Mechanisms Based on Clinical and Radiological Status in Mild Relapsing Remitting Multiple Sclerosis 1,2 † 3,4† 3,4,5 Gloria Castellazzi * , Laetitia Debernard , Tracy R. Melzer , 3,5,6 7,8 1,3,4 John C. Dalrymple-Alford , Egidio D’Angelo , David H. Miller , 1,7,9‡ 3,4,10‡ Claudia A. M. Gandini Wheeler-Kingshott and Deborah F. Mason Edited by: Fabienne Brilot, NMR Research Unit, Department of Neuroinflammation, Queen Square MS Centre, UCL Institute of Neurology, London, University of Sydney, Australia 2 3 United Kingdom, Department of Electrical, Computer and Biomedical Engineering, University of Pavia, Pavia, Italy, New Reviewed by: Zealand Brain Research Institute, Christchurch, New Zealand, Department of Medicine, University of Otago, Christchurch, 5 6 Moussa Antoine Chalah, New Zealand, Brain Research New Zealand, Auckland, New Zealand, Department of Psychology, University of Canterbury, 7 8 Hôpitaux Universitaires Henri Mondor, Christchurch, New Zealand, Department of Brain and Behavioral Sciences, University of Pavia, Pavia, Italy, Brain France Connectivity Center, IRCCS Mondino Foundation, Pavia, Italy, Brain MRI 3T Center, IRCCS Mondino Foundation, Pavia, Friedemann Paul, Italy, Department of Neurology, Christchurch Hospital, Christchurch, New Zealand Charité Universitätsmedizin Berlin, Germany Resting state functional MRI (rs-fMRI) has provided important insights into functional *Correspondence: reorganization in subjects with Multiple Sclerosis (MS) at different stage of disease. Gloria Castellazzi gloria.castellazzi@unipv.it In this cross-sectional study we first assessed, by means of rs-fMRI, the impact of overall T2 lesion load (T2LL) and MS severity score (MSSS) on resting state networks These authors have contributed equally to this work as first co-authors (RSNs) in 62 relapsing remitting MS (RRMS) patients with mild disability (MSSS < 3). These authors have contributed Independent Component Analysis (ICA) followed by dual regression analysis confirmed equally to this work as last co-authors functional connectivity (FC) alterations of many RSNs in RRMS subjects compared to healthy controls. The anterior default mode network (DMNa) and the superior precuneus Specialty section: This article was submitted to network (PNsup) showed the largest areas of decreased FC, while the sensory motor Multiple Sclerosis and networks area M1 (SMNm1) and the medial visual network (MVN) showed the largest Neuroimmunology, a section of the journal areas of increased FC. In order to better understand the nature of these alterations Frontiers in Neurology as well as the mechanisms of functional alterations in MS we proposed a method, Received: 24 January 2018 based on linear regression, that takes into account FC changes and their correlation with Accepted: 30 July 2018 T2LL and MSSS. Depending on the sign of the correlation between FC and T2LL, and Published: 20 August 2018 furthermore the sign of the correlation with MSSS, we suggested the following possible Citation: Castellazzi G, Debernard L, Melzer TR, underlying mechanisms to interpret altered FC: (1) FC reduction driven by MS lesions, (2) Dalrymple-Alford JC, D’Angelo E, “true” functional compensatory mechanism, (3a) functional compensation attempt, (3b) Miller DH, Gandini Wheeler-Kingshott CAM and “false” functional compensation, (4a) neurodegeneration, (4b) pre-symptomatic condition Mason DF (2018) Functional (damage precedes MS clinical manifestation). Our data shows areas satisfying 4 of these Connectivity Alterations Reveal 6 conditions (i.e., 1,2,3b,4b), supporting the suggestion that increased FC has a complex Complex Mechanisms Based on Clinical and Radiological Status in Mild nature that may exceed the simplistic assumption of an underlying compensatory Relapsing Remitting Multiple mechanism attempting to limit the brain damage caused by MS progression. Exploring Sclerosis. Front. Neurol. 9:690. doi: 10.3389/fneur.2018.00690 differences between RRMS subjects with short disease duration (MS ) and RRMS with short Frontiers in Neurology | www.frontiersin.org 1 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS similar disability but longer disease duration (MS ), we found that MS and MS long short long were characterized by clearly distinct pattern of FC, involving predominantly sensory and cognitive networks respectively. Overall, these results suggest that the analysis of FC alterations in multiple large-scale networks in relation to radiological (T2LL) and clinical (MSSS, disease duration) status may provide new insights into the pathophysiology of relapse onset MS evolution. Keywords: relapsing remitting multiple sclerosis, resting state fMRI, functional connectivity, functional impairment, resting state networks INTRODUCTION MS has also been reported in task-free conditions, that is, resting state functional MRI (rs-fMRI). One longitudinal rs- Multiple Sclerosis (MS) is a chronic disease characterized by fMRI study reported that increased FC was detected after the the presence of multifocal inflammatory demyelinated plaques advent of new lesions, which was interpreted as an attempt distributed over space and time within the central nervous to compensate for tissue damage (19). It remains to be system (CNS) (1, 2). The course of the MS disease is highly verified whether such a functional reorganization leads to a varied and unpredictable. The clinical measurement of disease preservation of wellbeing. For example, an increased FC in progression in terms of the rate at which disability accumulates in clinically isolated syndrome (CIS) patients without conventional an individual is challenging. Magnetic Resonance Imaging (MRI) lesions has been suggested as a risk factor for MS (20). has contributed significantly not only to diagnosis, by depicting Interestingly, recent studies in Relapsing Remitting MS (RRMS) white matter demyelinating lesions, but also to the study of have reported a positive correlation between increased FC mechanisms of disease and of functional alterations. in thalamic or in fronto-parietal regions and fatigue scores, In a 20-year follow-up MRI study of lesion load and disability, suggesting that increased FC might be a maladaptative process Fisniku et al. (3) showed that a concurrent change in white (21, 22). Other studies have reported evidence of positive matter lesion load on T2-weighted scans and expanded disability correlation between areas of increased FC and structural status scale (EDSS) scores in the first 5 years of the disease damage (23) or have found an association between increased is indicative of long-term disability. Increasing brain lesion functional connectivity in distinct systems involving attention load and brain atrophy have also been found to correlate with and cognitive control with decreased cognitive ability at early the progression of cognitive impairment in MS (4). Indeed, stages of MS (24), challenging the concept of functional changes in brain gray matter—rather than the white matter— compensation in MS. Nevertheless, a recent study of Rocca have been shown to predict long-term physical disability and et al. showed that also the reverse condition is possible, reporting the evidence of reduced FC correlated with better cognitive impairment in a number of studies (5–8). A review neuropsychological performance in a large cohort of MS subjects by Langer-Gould et al, though, identified sphincter symptoms (25), furtherly questioning the interpretation of altered FC in as the most robust predictor of long-term physical disability (9). MS. More recently, deep gray matter alterations and in particular An understanding of brain function in MS may be better thalamic atrophy have gained increasing relevance in the study of served by looking across the many functional networks in the MS. For example, a study on subjects with radiologically isolated brain, as the diffuse brain injuries present in MS are best syndrome (RIS) has provided evidence that thalamic atrophy revealed when co-varying fluctuations of the blood-oxygen-level- may precede clinical manifestations of CNS demyelination, therefore suggesting the thalamus may be a key region to check dependent (BOLD) signals are identified across widely dispersed for early signs of neurodegeneration in MS (10). Furthermore, neural structures (26). These networks are most readily evident thalamic atrophy has also been found to correlate with cognitive during periods of minimal cognitive demand, that is, when rs- decline and disability, suggesting that thalamic volume may be fMRI is used to reveal resting state networks (RSNs). These RSNs a clinically relevant biomarker to assess the neurodegenerative engage distinct brain regions that exhibit unique spontaneous disease process in MS (11, 12). patterns of low-frequency (around 0.01–0.1 Hz) synchronisations From a functional point of view, studies using task- and by inference functional connectivity (FC) (26, 27). Looking related functional MRI (fMRI) have often demonstrated greater at resting state is particularly suited for disorders such as MS responses in cortical areas, particularly in early stage MS patients, in which individuals may show cognitive impairments. For when compared with healthy controls. These differences are example, the default mode network (DMN) is a RSN that has generally interpreted as evidence of compensatory mechanisms particular relevance as a surrogate marker for early dementia to ameliorate cognitive or sensorimotor deficits in the initial (28, 29). Examination of rs-fMRI has provided important insights stages of the disease (13–17). Together with altered functional into the functional reorganization of the brain in subjects with connectivity between brain regions during cognitive tasks, early relapsing MS (3–5 years disease duration) (30) as well as in such effects imply the use of brain reserve to limit cognitive MS subjects at more advanced disease stage (31) or with longer impairment (18). Increased functional connectivity (FC) in disease duration (32). Frontiers in Neurology | www.frontiersin.org 2 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS In this study, we used an advanced rs-fMRI approach to focus domains: executive function (letter fluency, category fluency, on network changes associated with radiological and clinical Stroop interference) (42), memory (episodic learning and recall scores. First of all, we performed a traditional analysis to see were assessed (both visual) with the Brief Visual Memory Test, which RSNs are affected by RRMS in a cohort of patients with BVMT) (43), attention and working memory [Stroop colors, mild impairments. Then, given that MS can be described as a word reading, Symbol Digit Modality Test (SDMT), Paced multisystem disconnection syndrome (33), this work performed Auditory Serial Addition Test (PASAT)] (44), and visuospatial a comprehensive advanced analysis of the functional status of function [Judgment of Line Orientation (45), Rey Complex the principal large-scale RSNs focusing on identifying patterns of Figure copy (46)]. All patients were also assessed using the RSN FC impairment that discriminate mild RRMS from healthy MS Functional Composite (MSFC) test (47). MSFC score was subjects. To better understand the nature of the detected FC calculated from three components: (i) the average scores from alterations we formulated a priori hypotheses of mechanisms the four trials on the 9-HPT, (ii) the average scores of two 25- based on FC correlations with radiological and clinical metrics. Foot Timed Walk trials and (iii) the number correct from the We also compared RRMS subjects with short disease duration PASAT-3. Raw test scores were converted to z-scores using age- (MS or early MS) with those with longer disease duration adjusted and gender-adjusted normative data for each test and short (MS or established MS) to assess the impact of disease then averaged for each domain. long duration on FC. MRI Acquisition MATERIALS AND METHODS All scans were acquired in a single session on a 3T General Subjects Electric Signa HDxt MR scanner (General Electric Medical MRI acquisitions were performed on 91 subjects. Based on Systems, Milwaukee, WI) with head coil. the McDonald criteria (34) 62 subjects with RRMS (age 38.58 All subjects underwent MRI examination that included: ± 8.25, MSSS = 2.89 ± 1.87) were recruited for the study - rs-fMRI: T2 Gradient Echo (GRE), echo planar imaging from the Christchurch Hospital (Christchurch, New Zealand). (EPI) sequence (TR/TE = 2500/35ms; voxel size = 3.75 × 3.75 The twenty-nine healthy controls (HC) aged 34.45 ± 10.17 × 4 mm , FOV = 240 mm, 37 slices, 240 volumes, acquisition years had no previous history of neurological disorders. All time = 10:10 min). During fMRI acquisition subjects were MS patients had been relapse free and clinically stable for at asked to keep their eyes open while fixating on a cross; this least 1 month before study entry and 10 were receiving disease method may improve reliability relative to “eyes closed” (48). modifying medications. Neurological, neuropsychological and - T1 volumetric imaging (for anatomical reference): 3D T1- MRI assessments were scheduled over 1 month in 3 visits. weighted inversion-prepared spoiled gradient recalled-echo Neurological findings not attributable to MS and psychiatric acquisition (IR-SPGR): TR/TE = 2.8/6.6 ms, TI = 400 ms; symptoms (e.g., cerebrovascular disease, tumors, brain surgery, flip angle = 15 , acquisition matrix = 256 × 256 × 180; depressive disorder as measured by Beck Depression Inventory reconstruction matrix = 512 × 512 × 180 FOV = 240 mm; (BDI) with BDI > 19 cut-off) were defined as exclusion criteria. voxel size = 0.48 × 0.48 × 1 mm , 180 slices) was acquired for The RRMS group was also subdivided (labeled MS and short anatomical reference. MS ) based on their disease duration (35). The MS group long short comprised 36 subjects with early RRMS (defined as≤5 years from Conventional MRI sequences were also acquired for lesion symptom onset, aged 37.34 ± 8.82, MSSS = 2.95 ± 1.99). The detection: MS group included 26 subjects with a more established RRMS long - T2 Flair Spin-Echo (SE): TE/TR = 11/500 ms, TI = 2250 ms, disease duration (between 5 and 15 years from symptom onset, FOV = 220 mm, voxel size = 0.43 × 0.43 × 3 mm . aged 40.62 ± 7.26, MSSS = 2.86 ± 1.79). All subjects received - T2 Propeller: SE, TE/TR = 98/3700 ms, FOV = 220 mm, voxel an MRI scan and clinical assessment by a multidisciplinary team size = 0.43 × 0.43 × 3 mm . at the New Zealand Brain Research Institute (NZBRI). The study - T1 SE: TE/TR = 12/500 ms, FOV = 220 mm, voxel size = 0.43 was approved by the Lower South regional ethics committee of × 0.43 × 3 mm . New Zealand and written informed consent was provided by all participants. Structural MRI Analysis Clinical-Neurological Assessment All patients underwent clinical assessment, including relapse Lesion Load Evaluation and Lesion Filling history, Expanded Disability Status score (EDSS) (36), and For each subject, MS lesions were manually outlined using Modified Fatigue Impact Scale (MFIS) (37). MS severity score Jim software (Jim 4.0 Xinapse System Leicester, UK) on T2 (MSSS) (38) was calculated for all patients. Patients were assessed Flair images to quantify T2 lesion load (49). MS lesions were for depression using the Beck Depression Inventory (BDI-II) also manually outlined on 3D T1-weighted (3D T1) images (39), while subjects’ premorbid IQ was estimated with the and filled using an automatic lesion filling program (LEAP) Wechsler Test of Adult Reading (WTAR) (40). All participants (50) before performing tissue segmentation procedures in order performed the Montreal Cognitive Assessment (MoCA) (41) to limit potential gray matter (GM) and white matter (WM) and 11 standard neuropsychological tests covering four cognitive misclassification due to signal abnormalities in the lesion tissue. Frontiers in Neurology | www.frontiersin.org 3 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS Tissue Segmentation Analysis as noise while others as RSNs, based on previous literature (53– For each subject, GM and WM volumes as well as the 56). Group-ICA decomposes data into spatial maps that are the total intracranial volume were obtained performing tissue ICs relative to the total processed dataset (i.e., the enrolled 91 segmentation on 3D T1 images using SPM8 (Statistical subjects), or the multi-subject ICA components. At group level, Parametric Mapping, Wellcome Department of Imaging the IC maps are the same for each subject and are used as inputs Neuroscience Group, London, UK). For each subject, after for the subsequent dual regression analysis in order to calculate the statistical inference among groups. lesion filling, 3D T1 images were intensity bias corrected, tissue classified and registered using linear and non-linear transformations (DARTEL) within a unified model (51). The Between Group RSNs Comparison and Global resulted images were then segmented into GM, WM, and Alterations Ranking cerebrospinal fluid (CSF) using the customized priors, masked A non-parametric permutation test, referred to as “dual to remove non-brain tissue voxels, modulated, and finally regression” (28, 57, 58), was then applied to compare group- smoothed with a 10 mm Gaussian kernel (49). For the purposes specific FC maps for each IC map. First, this analysis tested the of the study, GM volume was calculated in subject space and statistical differences between HC and MS using two comparisons divided by the total intracranial volume—defined as the sum of or contrasts (MS < HC and MS > HC). We then investigated the GM, WM, and CSF segments—in order to obtain a normalized presence of significant differences in RSN FC between MS short GM volume index. and MS subjects, by directly testing the MS subgroups with long two further contrasts: MS >MS and MS < MS . short long short long In this study, each dual regression analysis was carried out on rs-fMRI Analysis the total ICs using age, gender, education level and GM ratio For each subject, rs-fMRI images were analyzed using the as additional covariates included in the general linear model Independent Component Analysis (ICA) first at single- (GLM). The statistical inference at group level was performed subject pre-processing level (single-ICA) to reliably separate using 5000 permutations. The resulting statistical maps were signal from noise, using the ICA-based Xnoiseifier (FIX) family-wise error (FWE) corrected for multiple comparisons, tool (52) as implemented in FSL (FMRIB Software Library, implementing threshold-free cluster enhancement (TFCE) (59) version 5.0.9). ICA was then applied at group-level (group- using a significance threshold of at least p ≤ 0.05. After that, the ICA) on pre-processed rs-fMRI data using the Multivariate final statistical maps were saved as tstatFC maps. Exploratory Linear Optimized Decomposition into Independent In order to study the FC changes within each RSN and to Components (MELODIC) method in order to characterize the establish a ranking of the networks in terms of their alterations, RSNs (53). for each considered contrast we calculated a global parameter, referred to as global FC or gFC (29) which takes into account Data Pre-processing both the extension of the clusters and the magnitude of the Individual subjects’ pre-processing was performed using FSL FC changes. For each contrast, we used the gFC index only to tools and consisted in motion correction, brain extraction, produce a bar plot that ranked and compared the RSNs in terms spatial smoothing using a Gaussian kernel of full-width-at- of their functional alteration (i.e., decreasing/increasing gFC- half-maximum (FWHM) of 5 mm, and high pass temporal values), taking into account both the magnitude and the spatial filtering equivalent to 150 s (0.007 Hz). Individual rs-fMRI extent of their FC changes. volumes were than registered to the corresponding structural 3D T1 scan using FMRIB’s Linear Image Registration Tool RSNs Correlations With Lesion Load and MSSS (FLIRT) and subsequently to standard space (MNI152) using FC changes have been then correlated with the radiological T2LL FMRIB’s Nonlinear Image Registration Tool (FNIRT) with score and subsequently with the clinical MSSS index, which default options. Then, each 4-dimensional rs-fMRI dataset are clinically relevant for MS diagnosis. For this analysis, we entered single-subject spatial-ICA (single-ICA) decomposition started using the voxels surviving the FWE-corrected threshold using MELODIC, with an automatic estimation of the number of (p ≤ 0.05) in the resulting tstatFC maps (each of MS < HC independent components (ICs), which resulted in spatial maps, and MS > HC). We then used these masks to run a second each with an associated time course. Model order was estimated permutation analysis using T2 lesion load (T2LL) as the using the Laplace approximation to the Bayesian evidence for explanatory variable of interest in the design matrix of the GLM a probabilistic principal component model. For each subject, (60). The new resulting tstat maps were saved as tstatFC T2LL single-ICA results were finally processed with the FIX algorithm and included only those RSN voxels that were both FC altered to clean rs-fMRI data from noisy and artefactual components. and significantly (FWE-corrected p ≤ 0.05) correlated to T2LL. We considered non-null areas within the tstatFC maps T2LL RSNs Identification to calculate parameter estimates, as expressed by Z-values in Pre-processed functional data, containing 240 time points individual masked rs-fMRI images, to obtain a numerical value (volumes) for each subject, were temporally concatenated across of the strength of RSNs temporal coherence. subjects to create a single 4-dimensional data set to run the In order to assess whether the alterations in the tstatFC T2LL group-ICA analysis via MELODIC, with an automatic estimation maps might correlate with MSSS, we used the tstatFC maps T2LL of the number of ICs. At this level, some of the ICs were identified as masks to run a third permutation analysis with MSSS as Frontiers in Neurology | www.frontiersin.org 4 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS explanatory variable of interest. The resulting statistical maps tstatFC maps (see section RSNs Correlations With T2LL_MSSS were saved as tstatFC and included only the areas Lesion Load and MSSS for details). Results were Bonferroni T2LL_MSSS within RSNs that were FC altered, significantly correlated (FWE- corrected for multiple comparisons and a statistical threshold of corrected p ≤ 0.05) to T2LL and MSSS. p ≤ 0.05 was considered significant. Pearson’s correlation analysis was carried out using SPSS to obtain a numerical value of the correlation strength between RSN RESULTS FC change and T2LL for the non-null areas in the tstatFC T2LL Clinical and Neurological Characteristics maps and between RSN FC change + T2LL and MSSS in the tstatFC maps. The demographic and clinical scores for the HC, MS, and MS T2LL_MSSS short and MS subgroups are provided in Table 1. Except for MoCA, long Mechanisms of FC Alterations both MS groups performed worse than HC on all clinical and In order to facilitate the discussion on the mechanisms of FC neuropsychological measures. A significant difference was found alterations in RRMS we introduced a priori a method of analysis for age between HC and MS (p = 0.05) and in particular between to help interpreting possible scenarios, as outlined in Table 2 and HC and MS subjects (p = 0.041), with MS group older long long in the flowchart diagram of Figure S1 in Supplementary Material. than HC. Significant differences were found in disease duration It is known that FC can be found both increased or decreased in (p < 0.001) and EDSS score (p = 0.008) when comparing MS short MS compared to HC (61), but interpretation of such changes is and MS groups, with higher EDSS scores observed in MS . long long debated. By analysing possible correlations between FC changes The mean MSFC score was significantly reduced in both MS short and T2LL it may be possible to hypothesize mechanisms of and MS compared with HC (Mann-Whitney test, MS : long short such changes. Specifically, we looked for correlations between FC p = 0.003; MS : p = 0.012), but no significant differences were long and T2LL and searched the data for four possible scenarios: (1) observed in MSFC between MS and MS . Both MS groups short long increased FC correlating with lower T2LL; (2) decreased FC and also had significantly higher BDI scores (Mann-Whitney test, higher T2LL; (3) increased FC and higher T2LL; (4) decreased FC MS : p = 0.008; MS : p = 0.003 than HC. MFIS, MSSS, and short long and lower T2LL. While the first two scenarios are straightforward T2 lesion load (i.e., T2LL) were not significantly different between (see discussion), interpretation of scenarios where T2LL and MS and MS (measures not relevant for HC). short long FC go in the same direction are less intuitive. For this reason, we performed a further correlation analysis with the MSSS and RSNs Identification defined the following four further scenarios: (3a) increased FC, ICA processing on rs-fMRI images resulted in 35 independent higher T2LL and lower MSSS; (3b) increased FC, higher T2LL, components, 18 of which were classified as RSNs based on and higher MSSS; (4a) decreased FC, lower T2LL, lower MSSS; their frequency spectra and spatial patterns (29, 53, 54, 56). (4b) decreased FC, lower T2LL, higher MSSS. The remaining 17 components probably reflected artifacts like Areas identified as having different FC-values between MS movement, physiological noise or cerebro-spinal fluid (CSF) short and MS were also classified in comparison with the above partial volume effects (62). long table of differences between the entire MS cohort and HC. The identified 18 RSNs were: medial visual network (MVN), lateral visual network (LVN), precuneus network (PN), superior precuneus network (PNsup), sensory motor networks area Non-imaging Statistics M1 (SMNm1), and area S2 (SMNs2), auditory network (AN), Statistical analyses were carried out using SPSS (version executive control network (ECN), default mode network (DMN), 21.0; SPSS, Chicago, IL, USA). Demographic, behavioral and anterior default mode network (DMNa), frontal cortex network radiological differences between groups were assessed with (FCN), language networks (LN) anterior (a) and posterior (p), different tests depending on the typology of the variables right (R) and left (L) ventral attention networks (VAN), salience (binary, normally or non-normally distributed). Specifically, network (SN), task positive network (TPN) and cerebellar χ -test was performed to compare frequency distributions network (CBLN). The cortical regions associated with identified of gender in the three groups. One-way analysis of variance RSNs are provided as Supplementary Material (Figure S2). (ANOVA) with Bonferroni correction was used to assess statistical differences among groups (HC and MS; HC, MS short MS vs. HC: RSNs Comparison and Ranking and MS ) in age. Non-parametric Kruskal-Wallis test was long applied to test differences among the groups in education level, of the RSN Alterations clinical indices (WTAR, BDI and MSFC, see section Clinical- The analysis of FC within the 18 identified RSNs revealed that 16 Neurological Assessment for details) and neuropsychological networks, including MVN, LVN, PN, PNsup, SMNm1, AN, ECN, scores (MoCA, PASAT, and attention, memory, executive, DMN, DMNa, FCN, LNa, LNp, LVAN, SN, TPN, and CBLN, visuospatial cognitive domains). Non-parametric Mann- were functionally impaired in MS compared to HC. Only RVAN Whitney U-test was performed to test differences between and SMNs2 did not show any significant FC impairment when MS and MS groups in EDSS, MSSS, MFIS, disease comparing MS to HC. short long duration and lesion load (T2LL). A Pearson’s correlations When looking at the global profile of FC impairments analysis was performed to assess the association between the RSN resulting from the group analysis, large (more than 1000 voxels) FC change and T2LL for the non-null areas in the tstatFC significantly FC reduced (p < 0.01) areas in MS compared to HC T2LL maps and between RSN FC change + T2LL and MSSS in the (i.e., MS < HC contrast) were observed in the frontal cortex, Frontiers in Neurology | www.frontiersin.org 5 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS TABLE 1 | Demographic and clinical characteristics. HC (n = 29) MS (n = 62) MS (n = 36) MS (n = 26) p short long Gender (female/male) 21/8 47/15 30/6 17/9 > 0.2 a,c Age (years) 34.45 ± 10.17 38.58 ± 8.25 37.34 ± 8.82 40.62 ± 7.26 < 0.05 Education (years) 13.62 ± 2.19 13.03 ± 2.53 13.43 ± 2.73 12.38 ± 2.17 > 0.05 WTAR 107.62 ± 7.08 104.65 ± 9.40 104.29 ± 9.90 105.15 ± 9.05 > 0.1 Disease duration (years) n.a. 5.27 ± 4.08 2.29 ± 1.22 9.46 ± 2.80 < 0.001 EDSS n.a. 1.77 ± 1.18 1.42 ± 0.90 2.23 ± 1.40 0.008 MSSS n.a. 2.89 ± 1.87 2.95 ± 1.99 2.86 ± 1.79 > 0.05 MFIS n.a. 7.45 ± 4.66 7.69 ± 4.95 7.15 ± 4.53 > 0.1 a,b,c BDI 4.45 ± 5.24 8.32 ± 6.10 8.29 ± 6.90 8.23 ± 5.20 < 0.05 PASAT (z score) 0.21 ± 1.05 0.62 ± 1.22 0.56 ± 1.07 0.70 ± 1.44 > 0.1 SDMT (z score) 0.29 ± 1.03 0.12 ± 0.91 0.04 ± 0.92 0.21 ± 0.93 > 0.2 a,b,c MSFC 0.36 ± 0.46 0.11 ± 0.70 0.04 ± 0.58 0.18 ± 0.86 < 0.05 MoCA 28.68 ± 1.51 28.29 ± 1.76 28.17 ± 1.82 28.46 ± 1.72 > 0.2 Executive (z score) 0.80 ± 0.54 0.39 ± 0.77 0.46 ± 0.76 0.28 ± 0.81 > 0.05 Memory (z score) 0.71 ± 0.68 0.47 ± 0.79 0.43 ± 0.84 0.50 ± 0.74 > 0.3 Attention (z score) 0.02 ± 0.71 0.32 ± 0.71 0.29 ± 0.76 0.36 ± 0.66 > 0.1 Visuospatial (z score) 0.27 ± 0.55 0.04 ± 0.59 0.13 ± 0.52 0.09 ± 0.68 > 0.05 Composite z score 0.44 ± 0.49 0.18 ± 0.53 0.22 ± 0.53 0.12 ± 0.68 > 0.05 T2 lesion load (mL) n.a. 16.63 ± 22.23 16.49 ± 23.77 16.82 ± 20.84 > 0.1 WTAR, Wechsler Test of Adult Reading; EDSS, Expanded Disability Status score; MSSS, MS severity score; MFIS, Modified Fatigue Impact Scale; BDI, Beck Depression Inventory; PASAT, Paced Auditory Serial Addition Test; SDMT, Symbol Digit Modality Test; MSFC, MS Functional Composite; MoCA, Montreal Cognitive Assessment. Mean and SD are reported. A chi-square test was used to test difference in gender, whereas one-way ANOVA test was used to test difference in age. Non-parametric Kruskal-Wallis and Mann-Whitney tests were used to test all the other measures. Significant findings are shown in bold. Significant difference between HC and MS. Significant difference between HC and MS short. Significant difference between HC and MS long. Significant difference between MS and MS short long. TABLE 2 | Proposed analysis and hypothesis of mechanisms of functional connectivity (FC) alterations in MS. Scenario Analysis Hypothesis of mechanism FC T2LL MSSS Scenario 1 FC reductions driven by MS lesions Scenario 2 True functional compensation Scenario 3a Functional compensation attempt Scenario 3b False functional compensation Scenario 4a Neurodegeneration (reduced FC not due to MS lesions) Scenario 4b Pre-symptomatic condition (damage precedes clinical manifestation of MS) Multiple-scenarios have been hypothesized to interpret the role of FC changes within the resting state networks (RSNs) of MS subjects. Each proposed mechanism describes a specific relation between FC changes, overall lesion load (T2LL) and MS severity score (MSSS). mainly involving the medial frontal gyrus of DMNa, and the cingulum, right fusiform gyrus and the most anterior part of precuneus area of PNsup and TPN (Figure 1). Decreased FC the precuneus, mainly involving SMNm1 (Figure 1). Extended areas were also found in the anterior cingulate cortex (BA10) areas of increased FC were also found in the inferior and middle and the fusiform gyrus (BA19) involving SN. Coherent results occipital gyri of MVN. Further areas of increased FC were were found when looking at the order ranking of RSNs according detected in the left middle temporal gyrus, mainly involving to the gFC index, the DMNa and PNsup networks showing the LVAN, as well as in both left and right insula areas and in largest and most severe FC reductions in MS (Figure 1). the frontal middle gyrus of ECN. Coherent results were found Furthermore, large areas of significantly increased FC even when considering the gFC network ranking that highlighted (p < 0.01) were observed in MS group compared with HC SMNm1 and MVN as the most affected networks for the (i.e., MS > HC) in the right supplementary motor area, MS > HC contrast (Figure 1). Frontiers in Neurology | www.frontiersin.org 6 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS FIGURE 1 | Altered FC in RSNs of MS vs. HC. On the left: in blue, brain areas showing significantly reduced FC (p ≤ 0.01 FWE-corrected) within the RSNs in MS compared to HC (i.e., MS < HC). The blue bar plot on the bottom shows, for MS < HC, the ranking of the RSNs according to their gFC alteration: DMNa and PNsup (highlighted with an asterisk mark in the bar plot) resulted the networks with the largest FC reductions in MS. On the right: global map showing on top, in red, the RSN voxels that resulted to have a significantly increased FC (p ≤ 0.01 FWE-corrected) in MS vs HC (i.e., MS > HC). The details of each RSN alteration for MS > HC are reported in the red bar plot on the bottom right: SMNm1, MVN (highlighted with an asterisk mark in the bar plot) resulted as the top-ranked altered networks. Correlations Between FC Changes, T2LL anterior cingulate cortex, BA10), LVAN (BA40), LNa (left precentral gyrus, BA44) and FCN (BA11). Of these areas, and MSSS those in MVN, ECN, LNa and FCN showed also positive When comparing MS to HC, results show that there are at least correlations with MSSS (Table 2: Scenario 3b). 4 possible combinations of correlations between T2LL and FC 4) Low FC and low T2LL (Table 2: Scenario 4a or b): and MSSS in a number of areas of the brain (see Figure 2 for a Areas of reduced FC in MS compared to HC were visual description of these findings). These can be also linked to found to correlate positively with T2LL in DMNa scenarios depicted in Table 2: (right superior and medial frontal gyri) and TPN (right 1) Low FC and high T2LL (Table 2: Scenario 1): Areas of reduced precentral gyrus, BA6). These areas were also found FC in MS compared to HC were found to correlate negatively to positively correlate with MSSS (Table 2: Scenario with T2LL in MVN (posterior cerebellar declive), PN (right 4b). posterior cingulate cortex, BA19), CBLN (left cerebellar lobule VI), LVN (left cerebellar Crus I), SN (right inferior and medial MS vs. MS : RSNs Comparison and temporal gyrus, BA37), and TPN (right cerebellar Crus I). short long 2) High FC and low T2LL (Table 2: Scenario 2): Areas of Ranking of the RSN Alterations increased FC in MS compared to HC were found to correlate Direct comparison of the MS and MS groups revealed short long negatively with T2LL in areas of MVN (left calcarine and several areas of significantly greater FC (p < 0.01) in MS short cuneus, BA30), SMNm1 (left precuneus, BA7), ECN (left (i.e., MS > MS ) both in left and right parietal areas short long superior and medial frontal gyrus), LVAN (left angular gyrus), of the supramarginal gyrus, right precuneus, thalamus, and and LVN (right superior occipital gyrus and cuneus, BA18, posterior cingulate cortex, mainly involving the TPN, LVN, BA19). and RVAN (Figure 3). The gFC network ranking reported 3) High FC and high T2LL (Table 2: Scenario 3a or b): Areas of coherent results, showing TPN, LVN, and RVAN as the top- increased FC in MS compared to HC were found to correlate ranked RSNs with a different gFC for the MS > MS short long positively with T2LL in MVN (left superior occipital gyrus contrast. Overlapping these areas onto the maps of alterations and cuneus, BA19), PNsup (right precuneus), AN, ECN (left corrected for T2LL and MSSS from the whole MS group Frontiers in Neurology | www.frontiersin.org 7 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS FIGURE 2 | Global maps of FC alterations which correlate with the overall lesion load (T2LL) and MSSS. Findings have been interpreted according to the multiple-scenario hypothesis presented in Table 2: four of the six proposed mechanism have been identified and represented as a map: (1) reduced FC driven by lesion (magenta voxels, corresponding to reduced FC areas in MS that negatively correlated at p ≤ 0.05 FWE-corrected with T2LL); (2) true functional compensation (blue voxels, corresponding to increased FC areas in MS that negatively correlated with T2LL); (3) false functional compensation (red voxels, corresponding to increased FC areas in MS that positively correlated with T2LL and MSSS); (4) pre-symptomatic condition (green voxels, corresponding to decreased FC areas in MS that positively correlated with T2LL and MSSS). compared to HC (Figure 2), 4.36% of the greater FC in precuneus and in the superior frontal gyrus. Only 0.3 and 0.1% MS > MS corresponds to regions interpreted as true of the altered regions in MS > MS overlap respectively short long short long functional compensation areas (Table 2: Scenario 2) in the with reduced FC driven by lesions (Table 2: Scenario 1) in the Frontiers in Neurology | www.frontiersin.org 8 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS anterior cingulum and with areas interpreted as evidence of cognitive reserve and the ability of one’s brain to adapt and delay pre-symptomatic condition (Table 2: Scenario 4b) in the medial cognitive decline (18, 71). Nevertheless, the hypothesis that an frontal gyrus. increased FC can be considered as evidence of brain functional When considering the MS < MS contrast, areas of reorganization processes (either beneficial or maladaptive) is still short long significantly reduced FC (p < 0.01) in MS compared to to be established (72, 73). short MS were observed in the left middle cingulum and the In order to help fostering novel discussions on this topic we long right precuneus and fusiform gyrus (BA37) of PNsup, in the propose to analyse FC changes in relation to other pathological right inferior occipital gyrus (BA19) of LNp, in the middle markers. Given the specificity of demyelinating lesions to MS, frontal gyrus of ECN as well as in both right and left middle their diagnostic role and their long term predicted value, we temporal gyrus of AN (Figure 3). PNsup, LNp, followed by ECN believe that T2LL is an important factor to be studied in and AN, also resulted as the most gFC altered networks in association with FC changes, at least in the first instance. MS < MS (Figure 3). Moreover, when overlapping these Furthermore, a clinical score like the MSSS can introduce short long areas to areas of alterations corrected for T2LL and MSSS at evaluation of disease severity that encompasses both EDSS and whole group level (Figure 2), 1.2% corresponds to regions of true disease duration. Other factors could be equally considered in functional compensation (Table 2: Scenario 2) in the cingulate alternative models to the one that we proposed, such as thalamic gyrus, while 0.5% overlaps with areas of reduced FC driven by atrophy form longitudinal data, which has been suggested as a lesions (Table 2: Scenario 1) in the inferior temporal gyrus. Only potentially relevant biomarker to assess the neurodegenerative 0.2 and 0.02% of the altered regions in MS < MS overlap disease process in MS (12). Other specific clinical aspects could short long respectively with pre-symptomatic condition (Table 2: Scenario also be included in the model, such as fatigue scores, cognitive 4b) in the superior frontal gyrus and with false functional tests or even non-conventional MRI biomarkers such as iron compensation (Table 2: Scenario 3b) in the inferior frontal accumulation (21, 74, 75). Given the cross-sectional nature of our gyrus. data, here we included gray matter (GM) volume (as opposed to atrophy) as a covariate in the statistical comparison of FC maps between groups. More specifically, to better understand the nature of DISCUSSION increased/decreased FC findings in MS in the present paper we In the current study, we used ICA and dual regression techniques investigated areas that are functionally altered and modulated to investigate whether and how functional connectivity within by the overall lesion load (T2LL), known to be predictive the RSNs is affected by the disease in a mild cohort of RRMS of long-term disability (76). Moreover, within these areas, we subjects. Results confirm a widespread functional alteration, questioned the relevance of compensatory mechanisms through expressed as both areas of decreased and increased FC, of further associations with disease severity using MSSS. In the almost all the RSNs, compared to HC. This result supports the first instance, this specific correlation study has been carried out interpretation of MS as a multisystem disconnection syndrome considering the RRMS group as a whole, independently of disease (33). Compared to HC, RRMS subjects show decreased FC in duration. Results of this exploratory approach suggest indeed two RSNs: DMNa and PNsup, both of which are cognitive that the interpretation of FC decreases or increases as result of networks involved in high cognitive functioning such as working either neural disruption or compensatory brain plasticity may memory, memory retrieval, and future-oriented thinking (63). be an oversimplification as the scenarios presented by the data FC reductions within the DMN and precuneus in RRMS patients are indeed several. Associations with clinical scores of disease has been reported previously (64, 65) and ascribed to factors, such severity, as represented by the MSSS have been used here to as brain hypometabolism and hypoperfusion. The mechanisms help identifying possible hypothesis of FC changes in MS that of FC alterations are still debated and future multi-modal and we have summarized in Table 2. We suggest that areas with longitudinal studies should aim to pin down the origin of such reduced FC and greater T2LL are considered as regions of true damage. “FC reductions driven by MS lesions”, while network areas with Our results highlight that, compared to HC, RRMS subjects increased FC but lower T2LL are considered as regions of “true have increased FC involving two different RSNs: SMNm1, which functional compensation”. The MSSS was not investigated in play a role in motor-control functioning, and MVN, which these areas because it would have not changed our proposed is involved in visual and language functions (66). Evidence interpretation of mechanism, based mainly on the predictive of increased FC within the RSNs of MS patients’ brain has value of T2LL for long-term disability. There are counterintuitive now been observed in multiple studies with and without the scenarios, though, where the increased FC correlates with higher presence of conventional lesions (19, 20, 67, 68). Taken together T2LL and others where a reduced FC is associated with a lower these studies support the hypothesis that increased FC may T2LL. These correlations are difficult to interpret; therefore, be a beneficial compensatory mechanism occurring at least at we propose that the sign of the correlation between FC and early stages of MS (69) which is lost in more advanced disease MSSS scores can discriminate whether FC alterations (both (31). Interestingly, the same pattern of increased FC linked to increased or decreased), also positively correlated with T2LL, white matter (WM) damage, followed by a subsequent global are compensatory or maladaptive. We propose to consider as a FC reduction has been demonstrated using an empirical model “functional compensation attempt” the mechanism driving areas by Tewarie et al. (70). FC increase has also been linked to where an increased FC is associated with a greater T2LL in Frontiers in Neurology | www.frontiersin.org 9 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS FIGURE 3 | Areas with altered FC in RSNs of MS compared to MS patients. On the left: magenta voxels show the areas of significantly greater FC (p ≤ 0.01, short long FWE-corrected) in the MS > MS contrast. On the right: aquamarine voxels represent the global map of the areas found with significantly lower FC (p ≤ 0.01, short long FWE-corrected) in MS compared to MS (i.e., MS < MS ). Interestingly, the direct comparison of MS and MS highlighted a distinct pattern of short long short long short long FC differences. Below the brain representations of areas of differences, bar plots show, for each considered contrast, the ranking of the RSNs according to their gFC parameter. In each bar plot we colored in red the top-ranked networks for MS > MS and in blue the top-ranked networks for the MS < MS contrast. short long short long Note that the top-ranked networks (marked with an asterisk) in one contrast (e.g., MS > MS ) are also some of the bottom-ranked network in the opposite short long contrast (MS < MS ). short long patients with a low MSSS. In other words, despite the greater Advanced microstructural and metabolic imaging, together with damage represented by a greater T2LL, patients presenting a longitudinal study design, could add value to the proposed areas satisfying this scenario are actually doing well in terms mechanistic interpretation and demonstrate its validity. of their MSSS. When areas of increased FC and greater T2LL Searching for areas satisfying the proposed scenarios, our data correlate with greater MSSS, instead, we argue that this can be shows that only four combinations are present in this cohort (see considered as an indication of “false functional compensation” Figure 2 and also Table S1 reported as Supplementary Material): because the increased FC is associated with worse focal pathology - FC reductions driven by MS lesions: in the inferior temporal (T2LL) and a worse clinical score (MSSS). Areas showing a gyrus and in the cerebellum; decreased FC, associated with both a lower T2LL and MSSS, - pre-symptomatic condition: in the frontal lobe (BA6 and BA9); can be considered as areas where the functional damage may - true functional compensation: in the cuneus, precuneus and in result from a “pre-symptomatic condition”, i.e., the functional the superior frontal gyrus; damage (reduced FC) may precede the clinical manifestation - false functional compensation: in the cuneus and in the middle of MS (low MSSS) and is not driven by focal damage (low and superior frontal gyrus (in particular BA10-11). T2LL). In this context, whether this scenario of reduced FC can be considered a compensatory attempt is debatable, but Results do not show areas satisfying the condition of plausible. An interesting argument could be to interpret these neurodegeneration nor the condition of functional compensation changes in terms of a reduced brain functional reserve (18). attempt. On the contrary, a combination of reduced FC and T2LL It is very interesting to note that according to the proposed associated with greater MSSS could be considered as evidence interpretation of FC changes, areas of reduced FC in the of damage due to neurodegeneration, i.e., the reduced FC is cerebellum and in the temporal lobe satisfy the hypothesis of not caused by MS lesions (low T2LL), but may result from the scenario 1 and may reflect FC reductions driven by MS lesions, presence of a coexistent non-focal neurodegenerative alteration while decreased FC in the frontal areas (see Figure 4) satisfies resulting in higher clinical impairment as shown by the MSSS. the criteria for scenario 4b and may reflect the presence of a Frontiers in Neurology | www.frontiersin.org 10 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS FIGURE 4 | Magenta voxels: areas satisfying the condition of reduced FC driven by lesion (Table 2: Scenario 1), mainly located in the cerebellum (crus I and lobule VI) and in the temporal areas (middle and inferior temporal gyri). Green voxels: areas satisfying the criteria for the pre-symptomatic condition (Table 2: Scenario 4b), mainly located in the frontal lobe (superior frontal gyrus). pre-symptomatic pathological condition (i.e., functional damage to consider for future studies is the heterogeneity of the prior to clinical manifestation of MS, i.e., T2LL and MSSS). underlying T2 lesion pathology. This aspect of lesions is Given the connectivity between the cerebellum and the frontal currently under investigation in several studies (80–82), using lobe (77) and under the assumption of the validity of the more specific sequences since these biophysical differences proposed multiple-scenario scheme (Table 2), one may wonder cannot be fully characterized by means of standard clinical MR whether this pre-symptomatic reduced FC may be driven by scans. Dedicated sequences would help characterizing not only cerebellar alterations or by thalamic atrophy, both known to persistent black holes, but also different aspects of demyelination, be relevant in MS (12, 75, 78, 79). Future longitudinal studies inflammation, and microstructure alterations of lesions and to may be able to answer this intriguing question. Interestingly, correlate them with altered FC. areas of increased FC can be found in the cuneus, precuneus, Given that this was only a cross-sectional study, to investigate and superior frontal gyrus. Part of these areas satisfy the the possible presence of FC evolution patterns, we assessed the criteria for scenario 2, which is associated to the proposed effect of disease duration on the FC alterations in the same mild hypothesis of true compensatory mechanism, while another RRMS cohort. Moreover, we believe that disease duration is often non-overlapping part of them satisfy the criteria for scenario overseen to give more attention to other aspects of MS such as 3b, which is associated to the proposed hypothesis of false disability, but from the results of this study it is clear that the compensation. Unfortunately, the present cross-sectional study length of the disease affects patterns of functional alterations. cannot inform as to the evolution of such alterations and In turns, understanding the mechanisms of these patterns could thereby determine the consequential nature of the findings help understanding disease evolution. Dividing the MS cohort by (e.g., did false compensation areas previously respond as true disease duration provided an opportunity to study and compare compensation?) or whether these are independent mechanisms rs-fMRI patterns in two clinical subgroups: (i) a subgroup in of action of the pathology. However, these findings support the the early stage of relapsing remitting MS (disease duration <5 suggestion that increased FC has a complex nature that may years, MS ) and (ii) a subgroup at a later stage who have a short exceed the simplistic assumption of an underlying compensatory mild relapsing remitting form of MS (disease duration >5 years, mechanism attempting to limit the brain damage caused by MS MS ). long evolution, exploiting or exhausting the brain functional reserve. Our results show that the widespread FC alterations within Indeed, these considerations suggest that we cannot exclude an the RSNs differentially characterized RRMS patients depending increased FC in MS may even represent a maladaptive response on their disease duration. More importantly, this RRMS subjects to the brain functional-structural deterioration. Another aspect with shorter (MS ) and longer (MS ) disease duration are short long Frontiers in Neurology | www.frontiersin.org 11 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS differentially characterized by patterns of FC alterations affecting in future studies to investigate whether the RNS FC changes different large-scale networks. Specifically, results show that between MS to HC might be influenced by their differences MS present reduced FC compared to MS subjects in a in BDI and MSFC. From a study design point of view, the short long large portion of the fronto-temporo-parietal cortex involving present work is a cross-sectional investigation and although the prevalently cognitive RSNs (in particular PNsup, LNp and alterations within the RSNs indicate a dysfunction of the system, ECN, see Figure 3). On the other hand, MS subjects influenced by focal damage as well as by disease duration, short show greater FC compared to MS in more sensory areas their implication for MS prognosis will require appropriate long (primary somatosensory of TPN and part of the visual areas of longitudinal data. In order to test the validity of the proposed LVN) mainly located in the parieto-occipital cortex (Figure 3). interpretation of FC changes in this mild cohort of MS subjects, Notably, the involvement of both sensory and cognitive RSNs future studies should learn from the results and consider not in the two groups appear almost complementary (i.e., the top- only a longitudinal design, but also a multi-modal approach. The ranked altered RSNs in one contrast—e.g., MS > MS – present study also investigated mechanisms of FC changes in a short long appear as the bottom-ranked altered networks in the opposite mild cohort of RRMS. An interesting question would be to assess contrast—e.g., MS < MS ), suggesting that specific a larger cohort of patients composed of different MS phenotypes, short long temporal dynamics may characterize MS evolution involving including progressive patients, to see whether our findings could neuroplasticity processes and mechanisms exploiting the brain be linked to brain reserve against physical disability as suggested functional reserve (83). Furthermore, the areas of greater FC in Sumowski et al. (84). in MS subjects (compared to MS ) show the largest short long overlap with the map of true functional compensation identified CONCLUSIONS according to the criteria of the multiple-scenario hypothesis (see Table 2). Noteworthy is that the long-term course of relapse- This exploratory study investigates for the first time a voxel- onset MS is variable, and in MS , it is reasonable to anticipate wise correlation between FC and focal damage (T2LL) followed short that some will have a favorable evolution (i.e., becoming like by a further voxel-wise correlation with a clinical score (MSSS). MS subjects) while others will accumulate disability from This can be considered as a basic model on which to build long future relapses or the development of secondary progression. further analysis, for example using longitudinal measures of However, whether marked early FC abnormalities are able to local atrophy (e.g., in the thalamus) or to include specific predict a less favorable disease progression is unclear, and can clinical or neuropsychological scores. Furthermore, this study only be addressed in a prospective longitudinal study. addresses also the impact of disease duration on FC changes. Some considerations need to be addressed with respect to the As a whole, RSN FC analysis shows that functional alterations study’s limitations. From a technical point of view, in order to in MS at a network level cannot be simply described in reduce structured noise artifacts arising from head motion and terms of compensatory mechanisms or of loss of function. The physiological processes, rs-fMRI data used in this study were analysis of FC changes in relation to overall T2 lesion load treated with the ICA-FIX algorithm which operates a robust but and MSSS suggests that the interpretation of FC alterations non-aggressive denoising of resting state signals (52). A further within RSNs is complex, and may include mechanisms, which possible limitation is that the acquisition of B0-fieldmaps was involve but are not limited to true functional compensations. not included in the MRI protocol for this study. Therefore, Of particular interest are the predominant correlations of FC rs-fMRI images have not been corrected for B0-inhomogeneities reductions and T2LL in the cerebellum and the finding satisfying during the pre-processing step. Moreover, in this study the MS the proposed hypothesis of pre-symptomatic alterations in the group was found significantly older than the HC group (see frontal lobe, both worth further investigations. Our findings Table 1), although the distributions of age in the two groups were show also that FC alterations in MS are influenced by disease strongly overlapping (see Figure S3 in Supplementary Material). duration. Indeed, RRMS with shorter and longer disease duration Given this difference, age was added as additional covariate are characterized by distinct patterns of FC alterations with a in the GLM model (see Materials and Methods at section differential involvement of sensory and cognitive RSNs. Despite Between Group RSNs Comparison and Global Alterations the limitations of a cross-sectional design, this study suggests that Ranking). Furthermore, to exclude whether the observed group novel approaches to study FC alterations in multiple large-scale differences in FC is due to age we run a further dual regression networks may provide new insights in the pathophysiology that analysis using age as explanatory variable of interest, which underlies the evolution of relapse onset MS. Further longitudinal means that the contrast vector is non-null (i.e., 1/−1 to test studies are needed to confirm our hypothesis of the mechanisms for positive/negative effect of age on FC) for the age column that drive FC changes in RRMS and to assess whether FC findings and null elsewhere in the GLM design matrix. This verification are able to predict the future course of the disease. analysis resulted in no voxels surviving the significant threshold of p = 0.05 FWE-corrected, indicating that age does not affect AUTHOR CONTRIBUTIONS the FC results we obtained in this study. Moreover, in this study significant differences were found in BDI and also in MSFC when GC, LD, CW-K, and DFM conceptualized the study. GC designed comparing MS to HC. Interestingly, no significant difference and performed the analyses with support LD. TM and DFM between groups was observed in the PASAT test, which is also acquired all MRI data. LD, JD-A, and DFM enrolled patients and the third component of MSFC. Therefore, it would be worth acquired all the clinical and neuro-radiological data helping for Frontiers in Neurology | www.frontiersin.org 12 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS data interpretation. DHM, CW-K, ED, JD-A, and DFM provided Health (RF-INM-2008-114341, RF-2009-1475845 and RC2014- support and guidance with data interpretation with the clinical 2017) to ED and to CW-K (RC2014-2017) and of Engineering contribution of all physicians. GC, CW-K, and DHM wrote the and Physical Sciences Research Council (EPSRC) to CW-K manuscript, with comments from all other authors. (EP/I027084/1). FUNDING ACKNOWLEDGMENTS This work was supported by the New Zealand Brain Research We thank the patients, their families, all healthy volunteers for Institute, the Canterbury Medical Research Foundation, the making this research possible. We also thank the ECTRIMS UK MS society and UCL-UCLH Biomedical Research Centre and Postdoctoral Research Fellowship Program for its support to GC. the National Institute for Health Research, University College London, UK. ECTRIMS and the Multiple Sclerosis International SUPPLEMENTARY MATERIAL Federation (MSIF) supported the work of GC with funding (ECTRIMS Postdoctoral Research Fellowship Program, MSIF Du The Supplementary Material for this article can be found Pré grant). Further support came from grants of European Union online at: https://www.frontiersin.org/articles/10.3389/fneur. (Human Brain Project; HBP-604102), the Italian Ministry of 2018.00690/full#supplementary-material REFERENCES cognitive task at the earliest stage of MS. Hum Brain Mapp. (2003) 20:51–8. doi: 10.1002/hbm.10128 1. Compston A, Coles A. Multiple sclerosis. Lancet (2002) 359:P1221–31. 14. Ksiazek-Winiarek DJ, Szpakowski P, Glabinski A. Neural plasticity in multiple doi: 10.1016/S0140-6736(02)08220-X sclerosis: the functional and molecular background. Neural Plast. (2015) 2. Van Schependom J, Gielen J, Laton J, D’hooghe MB, De Keyser J, and 2015:307175. doi: 10.1155/2015/307175 Nagels G. Graph theoretical analysis indicates cognitive impairment in 15. López-Góngora M, Escartín A, Martínez-Horta S, Fernández-Bobadilla R, MS stems from neural disconnection. Neuroimage Clin. (2014) 4:403–10. Querol L, Romero S, et al. Neurophysiological evidence of compensatory brain doi: 10.1016/j.nicl.2014.01.012 mechanisms in early-stage multiple sclerosis. PLoS ONE (2015) 10:e0136786. 3. Fisniku LK, Brex PA, Altmann DR, Miszkiel KA, Benton CE, Lanyon doi: 10.1371/journal.pone.0136786 R, et al. Disability and T2 MRI lesions: a 20-year follow-up of 16. Pantano P, Mainero C, Caramia F. Functional brain reorganization in multiple patients with relapse onset of multiple sclerosis. Brain (2008) 131:808–17. sclerosis: evidence from fMRI studies. J Neuroimaging (2006) 16:104–14. doi: 10.1093/brain/awm329 doi: 10.1111/j.1552-6569.2006.00029.x 4. Amato MP, Zipoli V, Portaccio E. Multiple sclerosis-related cognitive changes: 17. Sweet LH, Rao SM, Primeau M, Durgerian S, Cohen RA. Functional magnetic a review of cross-sectional and longitudinal studies. J Neurol Sci. (2006) resonance imaging response to increased verbal working memory demands 245:41–6. doi: 10.1016/j.jns.2005.08.019 among patients with multiple sclerosis. Hum Brain Mapp. (2006) 27:28–36. 5. Bischof A, Caverzasi E, Cordano C, Hauser SL, Henry RG. Advances doi: 10.1002/hbm.20163 in imaging multiple sclerosis. Semin Neurol. (2017) 37:538–45. 18. Cader S, Cifelli A, Abu-Omar Y, Palace J, Matthews PM. Reduced doi: 10.1055/s-0037-1607278 brain functional reserve and altered functional connectivity in patients 6. Calabrese M, Poretto V, Favaretto A, Alessio S, Bernardi V, Romualdi C, with multiple sclerosis. Brain (2006) 129:527–37. doi: 10.1093/brain/ et al. Cortical lesion load associates with progression of disability in multiple awh670 sclerosis. Brain (2012) 135:2952–61. doi: 10.1093/brain/aws246 19. Droby A, Yuen KS, Muthuraman M, Reitz SC, Fleischer V, Klein J, et al. 7. Eshaghi A, Prados F, Brownlee WJ, Altmann DR, Tur C, Cardoso MJ, et al. Changes in brain functional connectivity patterns are driven by an individual Deep gray matter volume loss drives disability worsening in multiple sclerosis. lesion in MS: a resting-state fMRI study. Brain Imaging Behav. (2015) Ann Neurol. (2018) 83:210–22. doi: 10.1002/ana.25145 10:1117–26. doi: 10.1007/s11682-015-9476-3 8. Filippi M, Preziosa P, Copetti M, Riccitelli G, Horsfield MA, Martinelli V, et al. 20. Liu Y, Dai Z, Duan Y, Huang J, Ren Z, Liu Z, et al. Whole brain functional Gray matter damage predicts the accumulation of disability 13 years later in connectivity in clinically isolated syndrome without conventional brain MRI MS. Neurology (2013) 81:1759–67. doi: 10.1212/01.wnl.0000435551.90824.d0 lesions. Eur Radiol. (2016) 26:2982–91. doi: 10.1007/s00330-015-4147-8 9. Langer-Gould A, Popat RA, Huang SM, Cobb K, Fontoura P, Gould MK, 21. Hidalgo de la Cruz M, d’Ambrosio A, Valsasina P, Pagani E, Colombo B, et al. Clinical and demographic predictors of long-term disability in patients Rodegher M, et al. Abnormal functional connectivity of thalamic sub-regions with relapsing-remitting multiple sclerosis: a systematic review. Arch Neurol. contributes to fatigue in multiple sclerosis. Mult Scler. (2017) 24:1183–95. (2006) 63:1686–91. doi: 10.1001/archneur.63.12.1686 doi: 10.1177/1352458517717807 10. Azevedo CJ, Overton E, Khadka S, Buckley J, Liu S, Sampat 22. Jaeger S, Paul F, Scheel M, Brandt A, Heine J, Pach D, et al. Multiple M, et al. Early CNS neurodegeneration in radiologically isolated sclerosis-related fatigue: altered resting-state functional connectivity of the syndrome. Neurol Neuroimmunol Neuroinflamm. (2015) 2:e102. ventral striatum and dorsolateral prefrontal cortex. Mult Scler. (2018). doi: 10.1212/NXI.0000000000000102 doi: 10.1177/1352458518758911. [Epub ahead of print]. 11. Papathanasiou A, Messinis L, Zampakis P, Panagiotakis G, Gourzis P, 23. Roosendaal SD, Hulst HE, Vrenken H, Feenstra HE, Castelijns JA, Pouwels Georgiou V, et al. Thalamic atrophy predicts cognitive impairment in PJ, et al. Structural and functional hippocampal changes in multiple relapsing remitting multiple sclerosis. Effect on instrumental activities of sclerosis patients with intact memory function. Radiology (2010) 255:595–604. daily living and employment status. J Neurol Sci. (2015) 358:236–42. doi: 10.1148/radiol.10091433 doi: 10.1016/j.jns.2015.09.001 24. Hawellek DJ, Hipp JF, Lewis CM, Corbetta M, Engel AK. Increased 12. Schoonheim MM, Ciccarelli O. The value of including thalamic atrophy as functional connectivity indicates the severity of cognitive impairment a clinical trial endpoint in multiple sclerosis. Neurology (2018) 90:677–8. in multiple sclerosis. Proc Natl Acad Sci USA. (2011) 108:19066–71. doi: 10.1212/WNL.0000000000005279 doi: 10.1073/pnas.1110024108 13. Audoin B, Ibarrola D, Ranjeva JP, Confort-Gouny S, Malikova I, Ali-Chérif 25. Rocca MA, Valsasina P, Leavitt VM, Rodegher M, Radaelli M, Riccitelli GC, A, et al. Compensatory cortical activation observed by fMRI during a et al. Functional network connectivity abnormalities in multiple sclerosis: Frontiers in Neurology | www.frontiersin.org 13 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS Correlations with disability and cognitive impairment. Mult Scler. (2018) 46. Meyers JE, Meyers KR. Rey Complex Figure Test and Recognition Trial 24:459–71. doi: 10.1177/1352458517699875 Professional Manual. Psychological Assessment Resources. Odessa, FL (1995). 26. Fox MD, Raichle ME. Spontaneous fluctuations in brain activity observed with 47. Uitdehaag BM, Adèr HJ, Roosma TJ, de Groot V, Kalkers NF, Polman functional magnetic resonance imaging. Nat Rev Neurosci. (2007) 8:700–11. CH. Multiple sclerosis functional composite: impact of reference doi: 10.1038/nrn2201 population and interpretation of changes. Mult Scler. (2002) 8:366–71. 27. Biswal B, Yetkin FZ, Haughton VM, Hyde JS. Functional connectivity in the doi: 10.1191/1352458502ms835oa motor cortex of resting human brain using echo-planar MRI. Magn Reson 48. Patriat R, Molloy EK, Meier TB, Kirk GR, Nair VA, Meyerand ME, et al. The Med. (1995) 34:537–41. doi: 10.1002/mrm.1910340409 effect of resting condition on resting-state fMRI reliability and consistency: a 28. Binnewijzend MA, Schoonheim MM, Sanz-Arigita E, Wink AM, van der comparison between resting with eyes open, closed, and fixated. Neuroimage Flier WM, Tolboom N, et al. Resting-state fMRI changes in Alzheimer’s (2013) 78:463–73. doi: 10.1016/j.neuroimage.2013.04.013 disease and mild cognitive impairment. Neurobiol Aging (2012) 33:2018–28. 49. Debernard L, Melzer TR, Van Stockum S, Graham C, Wheeler-Kingshott doi: 10.1016/j.neurobiolaging.2011.07.003 CA, Dalrymple-Alford JC, et al. Reduced grey matter perfusion without 29. Castellazzi G, Palesi F, Casali S, Vitali P, Sinforiani E, Wheeler-Kingshott CA, volume loss in early relapsing-remitting multiple sclerosis. J Neurol Neurosurg et al. A comprehensive assessment of resting state networks: bidirectional Psychiatr. (2014) 85:544–51. doi: 10.1136/jnnp-2013-305612 modification of functional integrity in cerebro-cerebellar networks in 50. Chard DT, Jackson JS, Miller DH, Wheeler-Kingshott CA. Reducing the dementia. Front Neurosci. (2014) 8:223. doi: 10.3389/fnins.2014.00223 impact of white matter lesions on automated measures of brain gray 30. Louapre C, Perlbarg V, García-Lorenzo D, Urbanski M, Benali H, Assouad and white matter volumes. J Magn Reson Imaging (2010) 32:223–8. R, et al. Brain networks disconnection in early multiple sclerosis cognitive doi: 10.1002/jmri.22214 deficits: an anatomofunctional study. Hum Brain Mapp. (2014) 35:4706–17. 51. Ashburner J, Friston KJ. Unified segmentation. Neuroimage (2005) 26:839–51. doi: 10.1002/hbm.22505 doi: 10.1016/j.neuroimage.2005.02.018 31. Dogonowski AM, Siebner HR, Soelberg Sørensen P, Paulson OB, Dyrby 52. Salimi-Khorshidi G, Douaud G, Beckmann CF, Glasser MF, Griffanti L, Smith TB, Blinkenberg M, et al. Resting-state connectivity of pre-motor cortex SM. Automatic denoising of functional MRI data: combining independent reflects disability in multiple sclerosis. Acta Neurol Scand. (2013) 128:328–35. component analysis and hierarchical fusion of classifiers. Neuroimage (2014) doi: 10.1111/ane.12121 90:449–68. doi: 10.1016/j.neuroimage.2013.11.046 32. Faivre A, Robinet E, Guye M, Rousseau C, Maarouf A, Le Troter A, et al. 53. Beckmann CF, DeLuca M, Devlin JT, Smith SM. Investigations into resting- Depletion of brain functional connectivity enhancement leads to disability state connectivity using independent component analysis. Philos Trans R Soc progression in multiple sclerosis: a longitudinal resting-state fMRI study. Mult Lond B Biol Sci. (2005) 360:1001–13. doi: 10.1098/rstb.2005.1634 Scler. (2016) 22:1695–708. doi: 10.1177/1352458516628657 54. Cole DM, Smith SM, Beckmann CF. Advances and pitfalls in the analysis 33. Filippi, M., van den Heuvel, M. P., Fornito, A., He, Y., Hulshoff Pol, and interpretation of resting-state FMRI data. Front Syst Neurosci. (2010) 4:8. H. E., Agosta, F., et al. (2013). Assessment of system dysfunction in doi: 10.3389/fnsys.2010.00008 the brain through MRI-based connectomics. Lancet Neurol. 12:1189–99. 55. Damoiseaux JS, Rombouts SA, Barkhof F, Scheltens P, Stam CJ, Smith SM, doi: 10.1016/S1474-4422(13)70144-3 et al. Consistent resting-state networks across healthy subjects. Proc Natl Acad 34. Polman CH, Reingold SC, Banwell B, Clanet M, Cohen JA, Filippi M, et al. Sci USA. (2006) 103:13848–53. doi: 10.1073/pnas.0601417103 Diagnostic criteria for multiple sclerosis: 2010 revisions to the McDonald 56. Smith SM, Fox PT, Miller KL, Glahn DC, Fox PM, Mackay CE, criteria. Ann Neurol. (2011) 69:292–302. doi: 10.1002/ana.22366 et al. Correspondence of the brain’s functional architecture during 35. Bozin I, Ge Y, Kuchling J, Dusek P, Chawla S, Harms L, et al. activation and rest. Proc Natl Acad Sci USA. (2009) 106:13040–5. Magnetic resonance phase alterations in multiple sclerosis patients doi: 10.1073/pnas.0905267106 with short and long disease duration. PLoS ONE (2015) 10:e0128386. 57. Beckmann CF, Mackay CE, Filippini N, and Smith SM. Group comparison doi: 10.1371/journal.pone.0128386 of resting-state FMRI data using multi-subject ICA and dual regression. 36. Kurtzke JF. Rating neurologic impairment in multiple sclerosis: an Neuroimage (2009) 47(Suppl. 1):S148. doi: 10.1016/S1053-8119(09)71511-3 expanded disability status scale (EDSS). Neurology (1983) 33:1444–52. 58. Filippini N, MacIntosh BJ, Hough MG, Goodwin GM, Frisoni GB, doi: 10.1212/WNL.33.11.1444 Smith SM, et al. Distinct patterns of brain activity in young carriers of 37. Mathiowetz V. Test–retest reliability and convergent validity of the Fatigue the APOE-epsilon4 allele. Proc Natl Acad Sci USA. (2009) 106:7209–14. Impact Scale for persons with multiple sclerosis. Am J Occup Ther. (2003) doi: 10.1073/pnas.0811879106 57:389–95. doi: 10.5014/ajot.57.4.389 59. Smith SM, Nichols TE. Threshold-free cluster enhancement: 38. Roxburgh RH, Seaman SR, Masterman T, Hensiek AE, Sawcer SJ, addressing problems of smoothing, threshold dependence and Vukusic S, et al. Multiple Sclerosis Severity Score: using disability and localisation in cluster inference. Neuroimage (2009) 44:83–98. disease duration to rate disease severity. Neurology (2005) 64:1144–51. doi: 10.1016/j.neuroimage.2008.03.061 doi: 10.1212/01.WNL.0000156155.19270.F8 60. Sbardella E, Tona F, Petsas N, Upadhyay N, Piattella MC, Filippini N, et al. 39. Beck A, Steer R, Brown G. Manual for the Beck Depression Inventory-II. Functional connectivity changes and their relationship with clinical disability San Antonio, TX: Psychological Corporation (1996). and white matter integrity in patients with relapsing-remitting multiple 40. Blair JR, Spreen O. Predicting premorbid IQ: a revision of the sclerosis. Mult Scler. (2015) 21:1681–92. doi: 10.1177/1352458514568826 National Adult Reading Test. Clin Neuropsychol. (1989) 3:129–36. 61. Sbardella E, Petsas N, Tona F, Pantano P. Resting-state fMRI in MS: general doi: 10.1080/13854048908403285 concepts and brief overview of its application. Biomed Res Int. (2015) 41. Dagenais E, Rouleau I, Demers M, Jobin C, Roger É, Chamelian L, 2015:212693. doi: 10.1155/2015/212693 et al. Value of the MoCA Test as a Screening Instrument in Multiple 62. Gour N, Felician O, Didic M, Koric L, Gueriot C, Chanoine V, et al. Functional Sclerosis. Can J Neurol Sci. (2013) 40:410–5. doi: 10.1017/S03171671000 connectivity changes differ in early and late-onset alzheimer’s disease. Hum 14384 Brain Mapp. (2014) 35:2978–94. doi: 10.1002/hbm.22379 42. Delis D. Delis-Kaplan Executive Functioning System: Examiner’s Manual. San 63. Xu X, Yuan H, Lei X. Activation and connectivity within the default mode Antonio, TX: The Psychological Corporation (2001). network contribute independently to future-oriented thought. Sci Rep. (2016) 43. Benedict R. Brief Visuospatial Memory Test-Revised Professional Manual. 6:21001. doi: 10.1038/srep21001 Odessa, FL: Psychological Assessment Resources (PAR), Inc. (1997). 64. Bonavita S, Gallo A, Sacco R, Corte MD, Bisecco A, Docimo R, et al. 44. Rao S. A Manual For The Brief Repeatable Battery Of Neuropsychological Tests Distributed changes in default-mode resting-state connectivity in multiple In Multiple Sclerosis. Milwaukee, WI: Medical College of Wisconsin (1990). sclerosis. Mult Scler. (2011) 17:411–22. doi: 10.1177/1352458510394609 45. Benton AL, Sivan AB, Hamsher K, Varney NR, Spreen O. Contributions to 65. Rocca MA, Valsasina P, Martinelli V, Misci P, Falini A, Comi G, et al. Large- neuropsychological assessment. In: Benton AL, editor. A Clinical Manual. scale neuronal network dysfunction in relapsing-remitting multiple sclerosis. New York, NY: Oxford University Press (1994). Neurology (2012) 79:1449–57. doi: 10.1212/WNL.0b013e31826d5f10 Frontiers in Neurology | www.frontiersin.org 14 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS 66. Smitha KA, Akhil Raja K, Arun KM, Rajesh PG, Thomas B, Kapilamoorthy 78. Parmar K, Stadelmann C, Rocca MA, Langdon D, D’Angelo E, D’Souza M, TR, et al. Resting state fMRI: a review on methods in resting state et al. The role of the cerebellum in multiple sclerosis-150 years after Charcot. connectivity analysis and resting state networks. Neuroradiol J. (2017) 30:305– Neurosci Biobehav Rev. (2018) 89:85–98. doi: 10.1016/j.neubiorev.2018. 17. doi: 10.1177/1971400917697342 02.012 67. Basile B, Castelli M, Monteleone F, Nocentini U, Caltagirone C, Centonze 79. Wilkins A. Cerebellar dysfunction in multiple sclerosis. Front Neurol. (2017) D, et al. Functional connectivity changes within specific networks parallel 8:312. doi: 10.3389/fneur.2017.00312 the clinical evolution of multiple sclerosis. Mult Scler. (2014) 20:1050–7. 80. Castellaro M, Magliozzi R, Palombit A, Pitteri M, Silvestri E, Camera V, doi: 10.1177/1352458513515082 et al. Heterogeneity of cortical lesion susceptibility mapping in multiple 68. Liu Y, Liang P, Duan Y, Huang J, Ren Z, Jia X, et al. Altered thalamic sclerosis. AJNR Am J Neuroradiol. (2017) 38:1087–95. doi: 10.3174/ajnr. functional connectivity in multiple sclerosis. Eur J Radiol. (2015) 84:703–8. A5150 doi: 10.1016/j.ejrad.2015.01.001 81. Faizy TD, Thaler C, Kumar D, Sedlacik J, Broocks G, Grosser M, et al. 69. Faivre A, Rico A, Zaaraoui W, Crespy L, Reuter F, Wybrecht D, Heterogeneity of multiple sclerosis lesions in multislice myelin water imaging. et al. Assessing brain connectivity at rest is clinically relevant in early PLoS ONE (2016) 11:e0151496. doi: 10.1371/journal.pone.0151496 multiple sclerosis. Mult Scler. (2012) 18:1251–8. doi: 10.1177/13524585114 82. Yiannakas MC, Tozer DJ, Schmierer K, Chard DT, Anderson VM, Altmann 35930 DR, et al. ADvanced IMage Algebra (ADIMA): a novel method for depicting 70. Tewarie P, Steenwijk MD, Brookes MJ, Uitdehaag BMJ, Geurts JJG, Stam multiple sclerosis lesion heterogeneity, as demonstrated by quantitative MRI. CJ, et al. Explaining the heterogeneity of functional connectivity findings in Mult Scler. (2013) 19:732–41. doi: 10.1177/1352458512462074 multiple sclerosis: an empirically informed modeling study. Hum. Brain 83. Tomassini V, Matthews PM, Thompson AJ, Fuglø D, Geurts JJ, Johansen-Berg Mapp. (2018) 39:2541–8. doi: 10.1002/hbm.24020 H, et al. Neuroplasticity and functional recovery in multiple sclerosis. Nat Rev 71. Franzmeier N, Göttler J, Grimmer T, Drzezga A, Áraque-Caballero MA, Neurol. (2012) 8:635–46. doi: 10.1038/nrneurol.2012.179 Simon-Vermot L, et al. Resting-state connectivity of the left frontal 84. Sumowski JF, Rocca MA, Leavitt VM, Meani A, Mesaros S, Drulovic J, et al. cortex to the default mode and dorsal attention network supports Brain reserve against physical disability progression over 5 years in multiple reserve in mild cognitive impairment. Front Aging Neurosci. (2017) 9:264. sclerosis. Neurology (2016) 86:2006–9. doi: 10.1212/WNL.0000000000002702 doi: 10.3389/fnagi.2017.00264 72. Schoonheim MM, Meijer KA, Geurts JJ. Network collapse and Conflict of Interest Statement: DHM received honoraria through payments cognitive impairment in multiple sclerosis. Front Neurol. (2015) 6:82. to his employer, UCL Institute of Neurology, for Advisory Committee and/or doi: 10.3389/fneur.2015.00082 Consultancy advice in multiple sclerosis studies from Biogen Idec, Novartis, 73. Tona F, De Giglio L, Petsas N, Sbardella E, Prosperini L, Upadhyay Mitsubishi Pharma Europe and Bayer Schering Pharma. DHM also received N, et al. Role of cerebellar dentate functional connectivity in balance compensation through payments to his employer for performing central MRI deficits in patients with multiple sclerosis. Radiology (2018) 287:267–75. analysis of multiple sclerosis trials from Biogen Idec, Novartis and Apitope. CW-K doi: 10.1148/radiol.2017170311 received research funding from EPSRC, UK MS Society, Horizon2020, Biogen 74. Londoño AC, Mora CA. Nonconventional MRI biomarkers for in vivo Idec, Novartis, Wings for Life, Spinal Research and CHNF. DFM has received monitoring of pathogenesis in multiple sclerosis. Neurol Neuroimmunol honoraria and travel grants from Biogen Idec, Novartis and TEVA. Neuroinflamm. (2014) 1:e45. doi: 10.1212/NXI.0000000000000045 75. Tona F, Petsas N, Sbardella E, Prosperini L, Carmellini M, Pozzilli The remaining authors declare that the research was conducted in the absence of C, et al. Multiple sclerosis: altered thalamic resting-state functional any commercial or financial relationships that could be construed as a potential connectivity and its effect on cognitive function. Radiology (2014) 271:814–21. conflict of interest. doi: 10.1148/radiol.14131688 76. Popescu V, Agosta F, Hulst HE, Sluimer IC, Knol DL, Sormani MP, Copyright © 2018 Castellazzi, Debernard, Melzer, Dalrymple-Alford, D’Angelo, et al. Brain atrophy and lesion load predict long term disability in Miller, Gandini Wheeler-Kingshott and Mason. This is an open-access article multiple sclerosis. J Neurol Neurosurg Psychiatr. (2013) 84:1082–91. distributed under the terms of the Creative Commons Attribution License (CC BY). doi: 10.1136/jnnp-2012-304094 The use, distribution or reproduction in other forums is permitted, provided the 77. Palesi F, Tournier JD, Calamante F, Muhlert N, Castellazzi G, Chard D, original author(s) and the copyright owner(s) are credited and that the original et al. Contralateral cerebello-thalamo-cortical pathways with prominent publication in this journal is cited, in accordance with accepted academic practice. involvement of associative areas in humans in vivo. Brain Struct Funct. (2015) No use, distribution or reproduction is permitted which does not comply with these 220:3369–84. doi: 10.1007/s00429-014-0861-2 terms. Frontiers in Neurology | www.frontiersin.org 15 August 2018 | Volume 9 | Article 690 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Frontiers in Neurology Pubmed Central

Functional Connectivity Alterations Reveal Complex Mechanisms Based on Clinical and Radiological Status in Mild Relapsing Remitting Multiple Sclerosis

Loading next page...
 
/lp/pubmed-central/functional-connectivity-alterations-reveal-complex-mechanisms-based-on-kCX1JlvKhm

References (87)

Publisher
Pubmed Central
Copyright
Copyright © 2018 Castellazzi, Debernard, Melzer, Dalrymple-Alford, D'Angelo, Miller, Gandini Wheeler-Kingshott and Mason.
ISSN
1664-2295
eISSN
1664-2295
DOI
10.3389/fneur.2018.00690
Publisher site
See Article on Publisher Site

Abstract

ORIGINAL RESEARCH published: 20 August 2018 doi: 10.3389/fneur.2018.00690 Functional Connectivity Alterations Reveal Complex Mechanisms Based on Clinical and Radiological Status in Mild Relapsing Remitting Multiple Sclerosis 1,2 † 3,4† 3,4,5 Gloria Castellazzi * , Laetitia Debernard , Tracy R. Melzer , 3,5,6 7,8 1,3,4 John C. Dalrymple-Alford , Egidio D’Angelo , David H. Miller , 1,7,9‡ 3,4,10‡ Claudia A. M. Gandini Wheeler-Kingshott and Deborah F. Mason Edited by: Fabienne Brilot, NMR Research Unit, Department of Neuroinflammation, Queen Square MS Centre, UCL Institute of Neurology, London, University of Sydney, Australia 2 3 United Kingdom, Department of Electrical, Computer and Biomedical Engineering, University of Pavia, Pavia, Italy, New Reviewed by: Zealand Brain Research Institute, Christchurch, New Zealand, Department of Medicine, University of Otago, Christchurch, 5 6 Moussa Antoine Chalah, New Zealand, Brain Research New Zealand, Auckland, New Zealand, Department of Psychology, University of Canterbury, 7 8 Hôpitaux Universitaires Henri Mondor, Christchurch, New Zealand, Department of Brain and Behavioral Sciences, University of Pavia, Pavia, Italy, Brain France Connectivity Center, IRCCS Mondino Foundation, Pavia, Italy, Brain MRI 3T Center, IRCCS Mondino Foundation, Pavia, Friedemann Paul, Italy, Department of Neurology, Christchurch Hospital, Christchurch, New Zealand Charité Universitätsmedizin Berlin, Germany Resting state functional MRI (rs-fMRI) has provided important insights into functional *Correspondence: reorganization in subjects with Multiple Sclerosis (MS) at different stage of disease. Gloria Castellazzi gloria.castellazzi@unipv.it In this cross-sectional study we first assessed, by means of rs-fMRI, the impact of overall T2 lesion load (T2LL) and MS severity score (MSSS) on resting state networks These authors have contributed equally to this work as first co-authors (RSNs) in 62 relapsing remitting MS (RRMS) patients with mild disability (MSSS < 3). These authors have contributed Independent Component Analysis (ICA) followed by dual regression analysis confirmed equally to this work as last co-authors functional connectivity (FC) alterations of many RSNs in RRMS subjects compared to healthy controls. The anterior default mode network (DMNa) and the superior precuneus Specialty section: This article was submitted to network (PNsup) showed the largest areas of decreased FC, while the sensory motor Multiple Sclerosis and networks area M1 (SMNm1) and the medial visual network (MVN) showed the largest Neuroimmunology, a section of the journal areas of increased FC. In order to better understand the nature of these alterations Frontiers in Neurology as well as the mechanisms of functional alterations in MS we proposed a method, Received: 24 January 2018 based on linear regression, that takes into account FC changes and their correlation with Accepted: 30 July 2018 T2LL and MSSS. Depending on the sign of the correlation between FC and T2LL, and Published: 20 August 2018 furthermore the sign of the correlation with MSSS, we suggested the following possible Citation: Castellazzi G, Debernard L, Melzer TR, underlying mechanisms to interpret altered FC: (1) FC reduction driven by MS lesions, (2) Dalrymple-Alford JC, D’Angelo E, “true” functional compensatory mechanism, (3a) functional compensation attempt, (3b) Miller DH, Gandini Wheeler-Kingshott CAM and “false” functional compensation, (4a) neurodegeneration, (4b) pre-symptomatic condition Mason DF (2018) Functional (damage precedes MS clinical manifestation). Our data shows areas satisfying 4 of these Connectivity Alterations Reveal 6 conditions (i.e., 1,2,3b,4b), supporting the suggestion that increased FC has a complex Complex Mechanisms Based on Clinical and Radiological Status in Mild nature that may exceed the simplistic assumption of an underlying compensatory Relapsing Remitting Multiple mechanism attempting to limit the brain damage caused by MS progression. Exploring Sclerosis. Front. Neurol. 9:690. doi: 10.3389/fneur.2018.00690 differences between RRMS subjects with short disease duration (MS ) and RRMS with short Frontiers in Neurology | www.frontiersin.org 1 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS similar disability but longer disease duration (MS ), we found that MS and MS long short long were characterized by clearly distinct pattern of FC, involving predominantly sensory and cognitive networks respectively. Overall, these results suggest that the analysis of FC alterations in multiple large-scale networks in relation to radiological (T2LL) and clinical (MSSS, disease duration) status may provide new insights into the pathophysiology of relapse onset MS evolution. Keywords: relapsing remitting multiple sclerosis, resting state fMRI, functional connectivity, functional impairment, resting state networks INTRODUCTION MS has also been reported in task-free conditions, that is, resting state functional MRI (rs-fMRI). One longitudinal rs- Multiple Sclerosis (MS) is a chronic disease characterized by fMRI study reported that increased FC was detected after the the presence of multifocal inflammatory demyelinated plaques advent of new lesions, which was interpreted as an attempt distributed over space and time within the central nervous to compensate for tissue damage (19). It remains to be system (CNS) (1, 2). The course of the MS disease is highly verified whether such a functional reorganization leads to a varied and unpredictable. The clinical measurement of disease preservation of wellbeing. For example, an increased FC in progression in terms of the rate at which disability accumulates in clinically isolated syndrome (CIS) patients without conventional an individual is challenging. Magnetic Resonance Imaging (MRI) lesions has been suggested as a risk factor for MS (20). has contributed significantly not only to diagnosis, by depicting Interestingly, recent studies in Relapsing Remitting MS (RRMS) white matter demyelinating lesions, but also to the study of have reported a positive correlation between increased FC mechanisms of disease and of functional alterations. in thalamic or in fronto-parietal regions and fatigue scores, In a 20-year follow-up MRI study of lesion load and disability, suggesting that increased FC might be a maladaptative process Fisniku et al. (3) showed that a concurrent change in white (21, 22). Other studies have reported evidence of positive matter lesion load on T2-weighted scans and expanded disability correlation between areas of increased FC and structural status scale (EDSS) scores in the first 5 years of the disease damage (23) or have found an association between increased is indicative of long-term disability. Increasing brain lesion functional connectivity in distinct systems involving attention load and brain atrophy have also been found to correlate with and cognitive control with decreased cognitive ability at early the progression of cognitive impairment in MS (4). Indeed, stages of MS (24), challenging the concept of functional changes in brain gray matter—rather than the white matter— compensation in MS. Nevertheless, a recent study of Rocca have been shown to predict long-term physical disability and et al. showed that also the reverse condition is possible, reporting the evidence of reduced FC correlated with better cognitive impairment in a number of studies (5–8). A review neuropsychological performance in a large cohort of MS subjects by Langer-Gould et al, though, identified sphincter symptoms (25), furtherly questioning the interpretation of altered FC in as the most robust predictor of long-term physical disability (9). MS. More recently, deep gray matter alterations and in particular An understanding of brain function in MS may be better thalamic atrophy have gained increasing relevance in the study of served by looking across the many functional networks in the MS. For example, a study on subjects with radiologically isolated brain, as the diffuse brain injuries present in MS are best syndrome (RIS) has provided evidence that thalamic atrophy revealed when co-varying fluctuations of the blood-oxygen-level- may precede clinical manifestations of CNS demyelination, therefore suggesting the thalamus may be a key region to check dependent (BOLD) signals are identified across widely dispersed for early signs of neurodegeneration in MS (10). Furthermore, neural structures (26). These networks are most readily evident thalamic atrophy has also been found to correlate with cognitive during periods of minimal cognitive demand, that is, when rs- decline and disability, suggesting that thalamic volume may be fMRI is used to reveal resting state networks (RSNs). These RSNs a clinically relevant biomarker to assess the neurodegenerative engage distinct brain regions that exhibit unique spontaneous disease process in MS (11, 12). patterns of low-frequency (around 0.01–0.1 Hz) synchronisations From a functional point of view, studies using task- and by inference functional connectivity (FC) (26, 27). Looking related functional MRI (fMRI) have often demonstrated greater at resting state is particularly suited for disorders such as MS responses in cortical areas, particularly in early stage MS patients, in which individuals may show cognitive impairments. For when compared with healthy controls. These differences are example, the default mode network (DMN) is a RSN that has generally interpreted as evidence of compensatory mechanisms particular relevance as a surrogate marker for early dementia to ameliorate cognitive or sensorimotor deficits in the initial (28, 29). Examination of rs-fMRI has provided important insights stages of the disease (13–17). Together with altered functional into the functional reorganization of the brain in subjects with connectivity between brain regions during cognitive tasks, early relapsing MS (3–5 years disease duration) (30) as well as in such effects imply the use of brain reserve to limit cognitive MS subjects at more advanced disease stage (31) or with longer impairment (18). Increased functional connectivity (FC) in disease duration (32). Frontiers in Neurology | www.frontiersin.org 2 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS In this study, we used an advanced rs-fMRI approach to focus domains: executive function (letter fluency, category fluency, on network changes associated with radiological and clinical Stroop interference) (42), memory (episodic learning and recall scores. First of all, we performed a traditional analysis to see were assessed (both visual) with the Brief Visual Memory Test, which RSNs are affected by RRMS in a cohort of patients with BVMT) (43), attention and working memory [Stroop colors, mild impairments. Then, given that MS can be described as a word reading, Symbol Digit Modality Test (SDMT), Paced multisystem disconnection syndrome (33), this work performed Auditory Serial Addition Test (PASAT)] (44), and visuospatial a comprehensive advanced analysis of the functional status of function [Judgment of Line Orientation (45), Rey Complex the principal large-scale RSNs focusing on identifying patterns of Figure copy (46)]. All patients were also assessed using the RSN FC impairment that discriminate mild RRMS from healthy MS Functional Composite (MSFC) test (47). MSFC score was subjects. To better understand the nature of the detected FC calculated from three components: (i) the average scores from alterations we formulated a priori hypotheses of mechanisms the four trials on the 9-HPT, (ii) the average scores of two 25- based on FC correlations with radiological and clinical metrics. Foot Timed Walk trials and (iii) the number correct from the We also compared RRMS subjects with short disease duration PASAT-3. Raw test scores were converted to z-scores using age- (MS or early MS) with those with longer disease duration adjusted and gender-adjusted normative data for each test and short (MS or established MS) to assess the impact of disease then averaged for each domain. long duration on FC. MRI Acquisition MATERIALS AND METHODS All scans were acquired in a single session on a 3T General Subjects Electric Signa HDxt MR scanner (General Electric Medical MRI acquisitions were performed on 91 subjects. Based on Systems, Milwaukee, WI) with head coil. the McDonald criteria (34) 62 subjects with RRMS (age 38.58 All subjects underwent MRI examination that included: ± 8.25, MSSS = 2.89 ± 1.87) were recruited for the study - rs-fMRI: T2 Gradient Echo (GRE), echo planar imaging from the Christchurch Hospital (Christchurch, New Zealand). (EPI) sequence (TR/TE = 2500/35ms; voxel size = 3.75 × 3.75 The twenty-nine healthy controls (HC) aged 34.45 ± 10.17 × 4 mm , FOV = 240 mm, 37 slices, 240 volumes, acquisition years had no previous history of neurological disorders. All time = 10:10 min). During fMRI acquisition subjects were MS patients had been relapse free and clinically stable for at asked to keep their eyes open while fixating on a cross; this least 1 month before study entry and 10 were receiving disease method may improve reliability relative to “eyes closed” (48). modifying medications. Neurological, neuropsychological and - T1 volumetric imaging (for anatomical reference): 3D T1- MRI assessments were scheduled over 1 month in 3 visits. weighted inversion-prepared spoiled gradient recalled-echo Neurological findings not attributable to MS and psychiatric acquisition (IR-SPGR): TR/TE = 2.8/6.6 ms, TI = 400 ms; symptoms (e.g., cerebrovascular disease, tumors, brain surgery, flip angle = 15 , acquisition matrix = 256 × 256 × 180; depressive disorder as measured by Beck Depression Inventory reconstruction matrix = 512 × 512 × 180 FOV = 240 mm; (BDI) with BDI > 19 cut-off) were defined as exclusion criteria. voxel size = 0.48 × 0.48 × 1 mm , 180 slices) was acquired for The RRMS group was also subdivided (labeled MS and short anatomical reference. MS ) based on their disease duration (35). The MS group long short comprised 36 subjects with early RRMS (defined as≤5 years from Conventional MRI sequences were also acquired for lesion symptom onset, aged 37.34 ± 8.82, MSSS = 2.95 ± 1.99). The detection: MS group included 26 subjects with a more established RRMS long - T2 Flair Spin-Echo (SE): TE/TR = 11/500 ms, TI = 2250 ms, disease duration (between 5 and 15 years from symptom onset, FOV = 220 mm, voxel size = 0.43 × 0.43 × 3 mm . aged 40.62 ± 7.26, MSSS = 2.86 ± 1.79). All subjects received - T2 Propeller: SE, TE/TR = 98/3700 ms, FOV = 220 mm, voxel an MRI scan and clinical assessment by a multidisciplinary team size = 0.43 × 0.43 × 3 mm . at the New Zealand Brain Research Institute (NZBRI). The study - T1 SE: TE/TR = 12/500 ms, FOV = 220 mm, voxel size = 0.43 was approved by the Lower South regional ethics committee of × 0.43 × 3 mm . New Zealand and written informed consent was provided by all participants. Structural MRI Analysis Clinical-Neurological Assessment All patients underwent clinical assessment, including relapse Lesion Load Evaluation and Lesion Filling history, Expanded Disability Status score (EDSS) (36), and For each subject, MS lesions were manually outlined using Modified Fatigue Impact Scale (MFIS) (37). MS severity score Jim software (Jim 4.0 Xinapse System Leicester, UK) on T2 (MSSS) (38) was calculated for all patients. Patients were assessed Flair images to quantify T2 lesion load (49). MS lesions were for depression using the Beck Depression Inventory (BDI-II) also manually outlined on 3D T1-weighted (3D T1) images (39), while subjects’ premorbid IQ was estimated with the and filled using an automatic lesion filling program (LEAP) Wechsler Test of Adult Reading (WTAR) (40). All participants (50) before performing tissue segmentation procedures in order performed the Montreal Cognitive Assessment (MoCA) (41) to limit potential gray matter (GM) and white matter (WM) and 11 standard neuropsychological tests covering four cognitive misclassification due to signal abnormalities in the lesion tissue. Frontiers in Neurology | www.frontiersin.org 3 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS Tissue Segmentation Analysis as noise while others as RSNs, based on previous literature (53– For each subject, GM and WM volumes as well as the 56). Group-ICA decomposes data into spatial maps that are the total intracranial volume were obtained performing tissue ICs relative to the total processed dataset (i.e., the enrolled 91 segmentation on 3D T1 images using SPM8 (Statistical subjects), or the multi-subject ICA components. At group level, Parametric Mapping, Wellcome Department of Imaging the IC maps are the same for each subject and are used as inputs Neuroscience Group, London, UK). For each subject, after for the subsequent dual regression analysis in order to calculate the statistical inference among groups. lesion filling, 3D T1 images were intensity bias corrected, tissue classified and registered using linear and non-linear transformations (DARTEL) within a unified model (51). The Between Group RSNs Comparison and Global resulted images were then segmented into GM, WM, and Alterations Ranking cerebrospinal fluid (CSF) using the customized priors, masked A non-parametric permutation test, referred to as “dual to remove non-brain tissue voxels, modulated, and finally regression” (28, 57, 58), was then applied to compare group- smoothed with a 10 mm Gaussian kernel (49). For the purposes specific FC maps for each IC map. First, this analysis tested the of the study, GM volume was calculated in subject space and statistical differences between HC and MS using two comparisons divided by the total intracranial volume—defined as the sum of or contrasts (MS < HC and MS > HC). We then investigated the GM, WM, and CSF segments—in order to obtain a normalized presence of significant differences in RSN FC between MS short GM volume index. and MS subjects, by directly testing the MS subgroups with long two further contrasts: MS >MS and MS < MS . short long short long In this study, each dual regression analysis was carried out on rs-fMRI Analysis the total ICs using age, gender, education level and GM ratio For each subject, rs-fMRI images were analyzed using the as additional covariates included in the general linear model Independent Component Analysis (ICA) first at single- (GLM). The statistical inference at group level was performed subject pre-processing level (single-ICA) to reliably separate using 5000 permutations. The resulting statistical maps were signal from noise, using the ICA-based Xnoiseifier (FIX) family-wise error (FWE) corrected for multiple comparisons, tool (52) as implemented in FSL (FMRIB Software Library, implementing threshold-free cluster enhancement (TFCE) (59) version 5.0.9). ICA was then applied at group-level (group- using a significance threshold of at least p ≤ 0.05. After that, the ICA) on pre-processed rs-fMRI data using the Multivariate final statistical maps were saved as tstatFC maps. Exploratory Linear Optimized Decomposition into Independent In order to study the FC changes within each RSN and to Components (MELODIC) method in order to characterize the establish a ranking of the networks in terms of their alterations, RSNs (53). for each considered contrast we calculated a global parameter, referred to as global FC or gFC (29) which takes into account Data Pre-processing both the extension of the clusters and the magnitude of the Individual subjects’ pre-processing was performed using FSL FC changes. For each contrast, we used the gFC index only to tools and consisted in motion correction, brain extraction, produce a bar plot that ranked and compared the RSNs in terms spatial smoothing using a Gaussian kernel of full-width-at- of their functional alteration (i.e., decreasing/increasing gFC- half-maximum (FWHM) of 5 mm, and high pass temporal values), taking into account both the magnitude and the spatial filtering equivalent to 150 s (0.007 Hz). Individual rs-fMRI extent of their FC changes. volumes were than registered to the corresponding structural 3D T1 scan using FMRIB’s Linear Image Registration Tool RSNs Correlations With Lesion Load and MSSS (FLIRT) and subsequently to standard space (MNI152) using FC changes have been then correlated with the radiological T2LL FMRIB’s Nonlinear Image Registration Tool (FNIRT) with score and subsequently with the clinical MSSS index, which default options. Then, each 4-dimensional rs-fMRI dataset are clinically relevant for MS diagnosis. For this analysis, we entered single-subject spatial-ICA (single-ICA) decomposition started using the voxels surviving the FWE-corrected threshold using MELODIC, with an automatic estimation of the number of (p ≤ 0.05) in the resulting tstatFC maps (each of MS < HC independent components (ICs), which resulted in spatial maps, and MS > HC). We then used these masks to run a second each with an associated time course. Model order was estimated permutation analysis using T2 lesion load (T2LL) as the using the Laplace approximation to the Bayesian evidence for explanatory variable of interest in the design matrix of the GLM a probabilistic principal component model. For each subject, (60). The new resulting tstat maps were saved as tstatFC T2LL single-ICA results were finally processed with the FIX algorithm and included only those RSN voxels that were both FC altered to clean rs-fMRI data from noisy and artefactual components. and significantly (FWE-corrected p ≤ 0.05) correlated to T2LL. We considered non-null areas within the tstatFC maps T2LL RSNs Identification to calculate parameter estimates, as expressed by Z-values in Pre-processed functional data, containing 240 time points individual masked rs-fMRI images, to obtain a numerical value (volumes) for each subject, were temporally concatenated across of the strength of RSNs temporal coherence. subjects to create a single 4-dimensional data set to run the In order to assess whether the alterations in the tstatFC T2LL group-ICA analysis via MELODIC, with an automatic estimation maps might correlate with MSSS, we used the tstatFC maps T2LL of the number of ICs. At this level, some of the ICs were identified as masks to run a third permutation analysis with MSSS as Frontiers in Neurology | www.frontiersin.org 4 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS explanatory variable of interest. The resulting statistical maps tstatFC maps (see section RSNs Correlations With T2LL_MSSS were saved as tstatFC and included only the areas Lesion Load and MSSS for details). Results were Bonferroni T2LL_MSSS within RSNs that were FC altered, significantly correlated (FWE- corrected for multiple comparisons and a statistical threshold of corrected p ≤ 0.05) to T2LL and MSSS. p ≤ 0.05 was considered significant. Pearson’s correlation analysis was carried out using SPSS to obtain a numerical value of the correlation strength between RSN RESULTS FC change and T2LL for the non-null areas in the tstatFC T2LL Clinical and Neurological Characteristics maps and between RSN FC change + T2LL and MSSS in the tstatFC maps. The demographic and clinical scores for the HC, MS, and MS T2LL_MSSS short and MS subgroups are provided in Table 1. Except for MoCA, long Mechanisms of FC Alterations both MS groups performed worse than HC on all clinical and In order to facilitate the discussion on the mechanisms of FC neuropsychological measures. A significant difference was found alterations in RRMS we introduced a priori a method of analysis for age between HC and MS (p = 0.05) and in particular between to help interpreting possible scenarios, as outlined in Table 2 and HC and MS subjects (p = 0.041), with MS group older long long in the flowchart diagram of Figure S1 in Supplementary Material. than HC. Significant differences were found in disease duration It is known that FC can be found both increased or decreased in (p < 0.001) and EDSS score (p = 0.008) when comparing MS short MS compared to HC (61), but interpretation of such changes is and MS groups, with higher EDSS scores observed in MS . long long debated. By analysing possible correlations between FC changes The mean MSFC score was significantly reduced in both MS short and T2LL it may be possible to hypothesize mechanisms of and MS compared with HC (Mann-Whitney test, MS : long short such changes. Specifically, we looked for correlations between FC p = 0.003; MS : p = 0.012), but no significant differences were long and T2LL and searched the data for four possible scenarios: (1) observed in MSFC between MS and MS . Both MS groups short long increased FC correlating with lower T2LL; (2) decreased FC and also had significantly higher BDI scores (Mann-Whitney test, higher T2LL; (3) increased FC and higher T2LL; (4) decreased FC MS : p = 0.008; MS : p = 0.003 than HC. MFIS, MSSS, and short long and lower T2LL. While the first two scenarios are straightforward T2 lesion load (i.e., T2LL) were not significantly different between (see discussion), interpretation of scenarios where T2LL and MS and MS (measures not relevant for HC). short long FC go in the same direction are less intuitive. For this reason, we performed a further correlation analysis with the MSSS and RSNs Identification defined the following four further scenarios: (3a) increased FC, ICA processing on rs-fMRI images resulted in 35 independent higher T2LL and lower MSSS; (3b) increased FC, higher T2LL, components, 18 of which were classified as RSNs based on and higher MSSS; (4a) decreased FC, lower T2LL, lower MSSS; their frequency spectra and spatial patterns (29, 53, 54, 56). (4b) decreased FC, lower T2LL, higher MSSS. The remaining 17 components probably reflected artifacts like Areas identified as having different FC-values between MS movement, physiological noise or cerebro-spinal fluid (CSF) short and MS were also classified in comparison with the above partial volume effects (62). long table of differences between the entire MS cohort and HC. The identified 18 RSNs were: medial visual network (MVN), lateral visual network (LVN), precuneus network (PN), superior precuneus network (PNsup), sensory motor networks area Non-imaging Statistics M1 (SMNm1), and area S2 (SMNs2), auditory network (AN), Statistical analyses were carried out using SPSS (version executive control network (ECN), default mode network (DMN), 21.0; SPSS, Chicago, IL, USA). Demographic, behavioral and anterior default mode network (DMNa), frontal cortex network radiological differences between groups were assessed with (FCN), language networks (LN) anterior (a) and posterior (p), different tests depending on the typology of the variables right (R) and left (L) ventral attention networks (VAN), salience (binary, normally or non-normally distributed). Specifically, network (SN), task positive network (TPN) and cerebellar χ -test was performed to compare frequency distributions network (CBLN). The cortical regions associated with identified of gender in the three groups. One-way analysis of variance RSNs are provided as Supplementary Material (Figure S2). (ANOVA) with Bonferroni correction was used to assess statistical differences among groups (HC and MS; HC, MS short MS vs. HC: RSNs Comparison and Ranking and MS ) in age. Non-parametric Kruskal-Wallis test was long applied to test differences among the groups in education level, of the RSN Alterations clinical indices (WTAR, BDI and MSFC, see section Clinical- The analysis of FC within the 18 identified RSNs revealed that 16 Neurological Assessment for details) and neuropsychological networks, including MVN, LVN, PN, PNsup, SMNm1, AN, ECN, scores (MoCA, PASAT, and attention, memory, executive, DMN, DMNa, FCN, LNa, LNp, LVAN, SN, TPN, and CBLN, visuospatial cognitive domains). Non-parametric Mann- were functionally impaired in MS compared to HC. Only RVAN Whitney U-test was performed to test differences between and SMNs2 did not show any significant FC impairment when MS and MS groups in EDSS, MSSS, MFIS, disease comparing MS to HC. short long duration and lesion load (T2LL). A Pearson’s correlations When looking at the global profile of FC impairments analysis was performed to assess the association between the RSN resulting from the group analysis, large (more than 1000 voxels) FC change and T2LL for the non-null areas in the tstatFC significantly FC reduced (p < 0.01) areas in MS compared to HC T2LL maps and between RSN FC change + T2LL and MSSS in the (i.e., MS < HC contrast) were observed in the frontal cortex, Frontiers in Neurology | www.frontiersin.org 5 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS TABLE 1 | Demographic and clinical characteristics. HC (n = 29) MS (n = 62) MS (n = 36) MS (n = 26) p short long Gender (female/male) 21/8 47/15 30/6 17/9 > 0.2 a,c Age (years) 34.45 ± 10.17 38.58 ± 8.25 37.34 ± 8.82 40.62 ± 7.26 < 0.05 Education (years) 13.62 ± 2.19 13.03 ± 2.53 13.43 ± 2.73 12.38 ± 2.17 > 0.05 WTAR 107.62 ± 7.08 104.65 ± 9.40 104.29 ± 9.90 105.15 ± 9.05 > 0.1 Disease duration (years) n.a. 5.27 ± 4.08 2.29 ± 1.22 9.46 ± 2.80 < 0.001 EDSS n.a. 1.77 ± 1.18 1.42 ± 0.90 2.23 ± 1.40 0.008 MSSS n.a. 2.89 ± 1.87 2.95 ± 1.99 2.86 ± 1.79 > 0.05 MFIS n.a. 7.45 ± 4.66 7.69 ± 4.95 7.15 ± 4.53 > 0.1 a,b,c BDI 4.45 ± 5.24 8.32 ± 6.10 8.29 ± 6.90 8.23 ± 5.20 < 0.05 PASAT (z score) 0.21 ± 1.05 0.62 ± 1.22 0.56 ± 1.07 0.70 ± 1.44 > 0.1 SDMT (z score) 0.29 ± 1.03 0.12 ± 0.91 0.04 ± 0.92 0.21 ± 0.93 > 0.2 a,b,c MSFC 0.36 ± 0.46 0.11 ± 0.70 0.04 ± 0.58 0.18 ± 0.86 < 0.05 MoCA 28.68 ± 1.51 28.29 ± 1.76 28.17 ± 1.82 28.46 ± 1.72 > 0.2 Executive (z score) 0.80 ± 0.54 0.39 ± 0.77 0.46 ± 0.76 0.28 ± 0.81 > 0.05 Memory (z score) 0.71 ± 0.68 0.47 ± 0.79 0.43 ± 0.84 0.50 ± 0.74 > 0.3 Attention (z score) 0.02 ± 0.71 0.32 ± 0.71 0.29 ± 0.76 0.36 ± 0.66 > 0.1 Visuospatial (z score) 0.27 ± 0.55 0.04 ± 0.59 0.13 ± 0.52 0.09 ± 0.68 > 0.05 Composite z score 0.44 ± 0.49 0.18 ± 0.53 0.22 ± 0.53 0.12 ± 0.68 > 0.05 T2 lesion load (mL) n.a. 16.63 ± 22.23 16.49 ± 23.77 16.82 ± 20.84 > 0.1 WTAR, Wechsler Test of Adult Reading; EDSS, Expanded Disability Status score; MSSS, MS severity score; MFIS, Modified Fatigue Impact Scale; BDI, Beck Depression Inventory; PASAT, Paced Auditory Serial Addition Test; SDMT, Symbol Digit Modality Test; MSFC, MS Functional Composite; MoCA, Montreal Cognitive Assessment. Mean and SD are reported. A chi-square test was used to test difference in gender, whereas one-way ANOVA test was used to test difference in age. Non-parametric Kruskal-Wallis and Mann-Whitney tests were used to test all the other measures. Significant findings are shown in bold. Significant difference between HC and MS. Significant difference between HC and MS short. Significant difference between HC and MS long. Significant difference between MS and MS short long. TABLE 2 | Proposed analysis and hypothesis of mechanisms of functional connectivity (FC) alterations in MS. Scenario Analysis Hypothesis of mechanism FC T2LL MSSS Scenario 1 FC reductions driven by MS lesions Scenario 2 True functional compensation Scenario 3a Functional compensation attempt Scenario 3b False functional compensation Scenario 4a Neurodegeneration (reduced FC not due to MS lesions) Scenario 4b Pre-symptomatic condition (damage precedes clinical manifestation of MS) Multiple-scenarios have been hypothesized to interpret the role of FC changes within the resting state networks (RSNs) of MS subjects. Each proposed mechanism describes a specific relation between FC changes, overall lesion load (T2LL) and MS severity score (MSSS). mainly involving the medial frontal gyrus of DMNa, and the cingulum, right fusiform gyrus and the most anterior part of precuneus area of PNsup and TPN (Figure 1). Decreased FC the precuneus, mainly involving SMNm1 (Figure 1). Extended areas were also found in the anterior cingulate cortex (BA10) areas of increased FC were also found in the inferior and middle and the fusiform gyrus (BA19) involving SN. Coherent results occipital gyri of MVN. Further areas of increased FC were were found when looking at the order ranking of RSNs according detected in the left middle temporal gyrus, mainly involving to the gFC index, the DMNa and PNsup networks showing the LVAN, as well as in both left and right insula areas and in largest and most severe FC reductions in MS (Figure 1). the frontal middle gyrus of ECN. Coherent results were found Furthermore, large areas of significantly increased FC even when considering the gFC network ranking that highlighted (p < 0.01) were observed in MS group compared with HC SMNm1 and MVN as the most affected networks for the (i.e., MS > HC) in the right supplementary motor area, MS > HC contrast (Figure 1). Frontiers in Neurology | www.frontiersin.org 6 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS FIGURE 1 | Altered FC in RSNs of MS vs. HC. On the left: in blue, brain areas showing significantly reduced FC (p ≤ 0.01 FWE-corrected) within the RSNs in MS compared to HC (i.e., MS < HC). The blue bar plot on the bottom shows, for MS < HC, the ranking of the RSNs according to their gFC alteration: DMNa and PNsup (highlighted with an asterisk mark in the bar plot) resulted the networks with the largest FC reductions in MS. On the right: global map showing on top, in red, the RSN voxels that resulted to have a significantly increased FC (p ≤ 0.01 FWE-corrected) in MS vs HC (i.e., MS > HC). The details of each RSN alteration for MS > HC are reported in the red bar plot on the bottom right: SMNm1, MVN (highlighted with an asterisk mark in the bar plot) resulted as the top-ranked altered networks. Correlations Between FC Changes, T2LL anterior cingulate cortex, BA10), LVAN (BA40), LNa (left precentral gyrus, BA44) and FCN (BA11). Of these areas, and MSSS those in MVN, ECN, LNa and FCN showed also positive When comparing MS to HC, results show that there are at least correlations with MSSS (Table 2: Scenario 3b). 4 possible combinations of correlations between T2LL and FC 4) Low FC and low T2LL (Table 2: Scenario 4a or b): and MSSS in a number of areas of the brain (see Figure 2 for a Areas of reduced FC in MS compared to HC were visual description of these findings). These can be also linked to found to correlate positively with T2LL in DMNa scenarios depicted in Table 2: (right superior and medial frontal gyri) and TPN (right 1) Low FC and high T2LL (Table 2: Scenario 1): Areas of reduced precentral gyrus, BA6). These areas were also found FC in MS compared to HC were found to correlate negatively to positively correlate with MSSS (Table 2: Scenario with T2LL in MVN (posterior cerebellar declive), PN (right 4b). posterior cingulate cortex, BA19), CBLN (left cerebellar lobule VI), LVN (left cerebellar Crus I), SN (right inferior and medial MS vs. MS : RSNs Comparison and temporal gyrus, BA37), and TPN (right cerebellar Crus I). short long 2) High FC and low T2LL (Table 2: Scenario 2): Areas of Ranking of the RSN Alterations increased FC in MS compared to HC were found to correlate Direct comparison of the MS and MS groups revealed short long negatively with T2LL in areas of MVN (left calcarine and several areas of significantly greater FC (p < 0.01) in MS short cuneus, BA30), SMNm1 (left precuneus, BA7), ECN (left (i.e., MS > MS ) both in left and right parietal areas short long superior and medial frontal gyrus), LVAN (left angular gyrus), of the supramarginal gyrus, right precuneus, thalamus, and and LVN (right superior occipital gyrus and cuneus, BA18, posterior cingulate cortex, mainly involving the TPN, LVN, BA19). and RVAN (Figure 3). The gFC network ranking reported 3) High FC and high T2LL (Table 2: Scenario 3a or b): Areas of coherent results, showing TPN, LVN, and RVAN as the top- increased FC in MS compared to HC were found to correlate ranked RSNs with a different gFC for the MS > MS short long positively with T2LL in MVN (left superior occipital gyrus contrast. Overlapping these areas onto the maps of alterations and cuneus, BA19), PNsup (right precuneus), AN, ECN (left corrected for T2LL and MSSS from the whole MS group Frontiers in Neurology | www.frontiersin.org 7 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS FIGURE 2 | Global maps of FC alterations which correlate with the overall lesion load (T2LL) and MSSS. Findings have been interpreted according to the multiple-scenario hypothesis presented in Table 2: four of the six proposed mechanism have been identified and represented as a map: (1) reduced FC driven by lesion (magenta voxels, corresponding to reduced FC areas in MS that negatively correlated at p ≤ 0.05 FWE-corrected with T2LL); (2) true functional compensation (blue voxels, corresponding to increased FC areas in MS that negatively correlated with T2LL); (3) false functional compensation (red voxels, corresponding to increased FC areas in MS that positively correlated with T2LL and MSSS); (4) pre-symptomatic condition (green voxels, corresponding to decreased FC areas in MS that positively correlated with T2LL and MSSS). compared to HC (Figure 2), 4.36% of the greater FC in precuneus and in the superior frontal gyrus. Only 0.3 and 0.1% MS > MS corresponds to regions interpreted as true of the altered regions in MS > MS overlap respectively short long short long functional compensation areas (Table 2: Scenario 2) in the with reduced FC driven by lesions (Table 2: Scenario 1) in the Frontiers in Neurology | www.frontiersin.org 8 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS anterior cingulum and with areas interpreted as evidence of cognitive reserve and the ability of one’s brain to adapt and delay pre-symptomatic condition (Table 2: Scenario 4b) in the medial cognitive decline (18, 71). Nevertheless, the hypothesis that an frontal gyrus. increased FC can be considered as evidence of brain functional When considering the MS < MS contrast, areas of reorganization processes (either beneficial or maladaptive) is still short long significantly reduced FC (p < 0.01) in MS compared to to be established (72, 73). short MS were observed in the left middle cingulum and the In order to help fostering novel discussions on this topic we long right precuneus and fusiform gyrus (BA37) of PNsup, in the propose to analyse FC changes in relation to other pathological right inferior occipital gyrus (BA19) of LNp, in the middle markers. Given the specificity of demyelinating lesions to MS, frontal gyrus of ECN as well as in both right and left middle their diagnostic role and their long term predicted value, we temporal gyrus of AN (Figure 3). PNsup, LNp, followed by ECN believe that T2LL is an important factor to be studied in and AN, also resulted as the most gFC altered networks in association with FC changes, at least in the first instance. MS < MS (Figure 3). Moreover, when overlapping these Furthermore, a clinical score like the MSSS can introduce short long areas to areas of alterations corrected for T2LL and MSSS at evaluation of disease severity that encompasses both EDSS and whole group level (Figure 2), 1.2% corresponds to regions of true disease duration. Other factors could be equally considered in functional compensation (Table 2: Scenario 2) in the cingulate alternative models to the one that we proposed, such as thalamic gyrus, while 0.5% overlaps with areas of reduced FC driven by atrophy form longitudinal data, which has been suggested as a lesions (Table 2: Scenario 1) in the inferior temporal gyrus. Only potentially relevant biomarker to assess the neurodegenerative 0.2 and 0.02% of the altered regions in MS < MS overlap disease process in MS (12). Other specific clinical aspects could short long respectively with pre-symptomatic condition (Table 2: Scenario also be included in the model, such as fatigue scores, cognitive 4b) in the superior frontal gyrus and with false functional tests or even non-conventional MRI biomarkers such as iron compensation (Table 2: Scenario 3b) in the inferior frontal accumulation (21, 74, 75). Given the cross-sectional nature of our gyrus. data, here we included gray matter (GM) volume (as opposed to atrophy) as a covariate in the statistical comparison of FC maps between groups. More specifically, to better understand the nature of DISCUSSION increased/decreased FC findings in MS in the present paper we In the current study, we used ICA and dual regression techniques investigated areas that are functionally altered and modulated to investigate whether and how functional connectivity within by the overall lesion load (T2LL), known to be predictive the RSNs is affected by the disease in a mild cohort of RRMS of long-term disability (76). Moreover, within these areas, we subjects. Results confirm a widespread functional alteration, questioned the relevance of compensatory mechanisms through expressed as both areas of decreased and increased FC, of further associations with disease severity using MSSS. In the almost all the RSNs, compared to HC. This result supports the first instance, this specific correlation study has been carried out interpretation of MS as a multisystem disconnection syndrome considering the RRMS group as a whole, independently of disease (33). Compared to HC, RRMS subjects show decreased FC in duration. Results of this exploratory approach suggest indeed two RSNs: DMNa and PNsup, both of which are cognitive that the interpretation of FC decreases or increases as result of networks involved in high cognitive functioning such as working either neural disruption or compensatory brain plasticity may memory, memory retrieval, and future-oriented thinking (63). be an oversimplification as the scenarios presented by the data FC reductions within the DMN and precuneus in RRMS patients are indeed several. Associations with clinical scores of disease has been reported previously (64, 65) and ascribed to factors, such severity, as represented by the MSSS have been used here to as brain hypometabolism and hypoperfusion. The mechanisms help identifying possible hypothesis of FC changes in MS that of FC alterations are still debated and future multi-modal and we have summarized in Table 2. We suggest that areas with longitudinal studies should aim to pin down the origin of such reduced FC and greater T2LL are considered as regions of true damage. “FC reductions driven by MS lesions”, while network areas with Our results highlight that, compared to HC, RRMS subjects increased FC but lower T2LL are considered as regions of “true have increased FC involving two different RSNs: SMNm1, which functional compensation”. The MSSS was not investigated in play a role in motor-control functioning, and MVN, which these areas because it would have not changed our proposed is involved in visual and language functions (66). Evidence interpretation of mechanism, based mainly on the predictive of increased FC within the RSNs of MS patients’ brain has value of T2LL for long-term disability. There are counterintuitive now been observed in multiple studies with and without the scenarios, though, where the increased FC correlates with higher presence of conventional lesions (19, 20, 67, 68). Taken together T2LL and others where a reduced FC is associated with a lower these studies support the hypothesis that increased FC may T2LL. These correlations are difficult to interpret; therefore, be a beneficial compensatory mechanism occurring at least at we propose that the sign of the correlation between FC and early stages of MS (69) which is lost in more advanced disease MSSS scores can discriminate whether FC alterations (both (31). Interestingly, the same pattern of increased FC linked to increased or decreased), also positively correlated with T2LL, white matter (WM) damage, followed by a subsequent global are compensatory or maladaptive. We propose to consider as a FC reduction has been demonstrated using an empirical model “functional compensation attempt” the mechanism driving areas by Tewarie et al. (70). FC increase has also been linked to where an increased FC is associated with a greater T2LL in Frontiers in Neurology | www.frontiersin.org 9 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS FIGURE 3 | Areas with altered FC in RSNs of MS compared to MS patients. On the left: magenta voxels show the areas of significantly greater FC (p ≤ 0.01, short long FWE-corrected) in the MS > MS contrast. On the right: aquamarine voxels represent the global map of the areas found with significantly lower FC (p ≤ 0.01, short long FWE-corrected) in MS compared to MS (i.e., MS < MS ). Interestingly, the direct comparison of MS and MS highlighted a distinct pattern of short long short long short long FC differences. Below the brain representations of areas of differences, bar plots show, for each considered contrast, the ranking of the RSNs according to their gFC parameter. In each bar plot we colored in red the top-ranked networks for MS > MS and in blue the top-ranked networks for the MS < MS contrast. short long short long Note that the top-ranked networks (marked with an asterisk) in one contrast (e.g., MS > MS ) are also some of the bottom-ranked network in the opposite short long contrast (MS < MS ). short long patients with a low MSSS. In other words, despite the greater Advanced microstructural and metabolic imaging, together with damage represented by a greater T2LL, patients presenting a longitudinal study design, could add value to the proposed areas satisfying this scenario are actually doing well in terms mechanistic interpretation and demonstrate its validity. of their MSSS. When areas of increased FC and greater T2LL Searching for areas satisfying the proposed scenarios, our data correlate with greater MSSS, instead, we argue that this can be shows that only four combinations are present in this cohort (see considered as an indication of “false functional compensation” Figure 2 and also Table S1 reported as Supplementary Material): because the increased FC is associated with worse focal pathology - FC reductions driven by MS lesions: in the inferior temporal (T2LL) and a worse clinical score (MSSS). Areas showing a gyrus and in the cerebellum; decreased FC, associated with both a lower T2LL and MSSS, - pre-symptomatic condition: in the frontal lobe (BA6 and BA9); can be considered as areas where the functional damage may - true functional compensation: in the cuneus, precuneus and in result from a “pre-symptomatic condition”, i.e., the functional the superior frontal gyrus; damage (reduced FC) may precede the clinical manifestation - false functional compensation: in the cuneus and in the middle of MS (low MSSS) and is not driven by focal damage (low and superior frontal gyrus (in particular BA10-11). T2LL). In this context, whether this scenario of reduced FC can be considered a compensatory attempt is debatable, but Results do not show areas satisfying the condition of plausible. An interesting argument could be to interpret these neurodegeneration nor the condition of functional compensation changes in terms of a reduced brain functional reserve (18). attempt. On the contrary, a combination of reduced FC and T2LL It is very interesting to note that according to the proposed associated with greater MSSS could be considered as evidence interpretation of FC changes, areas of reduced FC in the of damage due to neurodegeneration, i.e., the reduced FC is cerebellum and in the temporal lobe satisfy the hypothesis of not caused by MS lesions (low T2LL), but may result from the scenario 1 and may reflect FC reductions driven by MS lesions, presence of a coexistent non-focal neurodegenerative alteration while decreased FC in the frontal areas (see Figure 4) satisfies resulting in higher clinical impairment as shown by the MSSS. the criteria for scenario 4b and may reflect the presence of a Frontiers in Neurology | www.frontiersin.org 10 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS FIGURE 4 | Magenta voxels: areas satisfying the condition of reduced FC driven by lesion (Table 2: Scenario 1), mainly located in the cerebellum (crus I and lobule VI) and in the temporal areas (middle and inferior temporal gyri). Green voxels: areas satisfying the criteria for the pre-symptomatic condition (Table 2: Scenario 4b), mainly located in the frontal lobe (superior frontal gyrus). pre-symptomatic pathological condition (i.e., functional damage to consider for future studies is the heterogeneity of the prior to clinical manifestation of MS, i.e., T2LL and MSSS). underlying T2 lesion pathology. This aspect of lesions is Given the connectivity between the cerebellum and the frontal currently under investigation in several studies (80–82), using lobe (77) and under the assumption of the validity of the more specific sequences since these biophysical differences proposed multiple-scenario scheme (Table 2), one may wonder cannot be fully characterized by means of standard clinical MR whether this pre-symptomatic reduced FC may be driven by scans. Dedicated sequences would help characterizing not only cerebellar alterations or by thalamic atrophy, both known to persistent black holes, but also different aspects of demyelination, be relevant in MS (12, 75, 78, 79). Future longitudinal studies inflammation, and microstructure alterations of lesions and to may be able to answer this intriguing question. Interestingly, correlate them with altered FC. areas of increased FC can be found in the cuneus, precuneus, Given that this was only a cross-sectional study, to investigate and superior frontal gyrus. Part of these areas satisfy the the possible presence of FC evolution patterns, we assessed the criteria for scenario 2, which is associated to the proposed effect of disease duration on the FC alterations in the same mild hypothesis of true compensatory mechanism, while another RRMS cohort. Moreover, we believe that disease duration is often non-overlapping part of them satisfy the criteria for scenario overseen to give more attention to other aspects of MS such as 3b, which is associated to the proposed hypothesis of false disability, but from the results of this study it is clear that the compensation. Unfortunately, the present cross-sectional study length of the disease affects patterns of functional alterations. cannot inform as to the evolution of such alterations and In turns, understanding the mechanisms of these patterns could thereby determine the consequential nature of the findings help understanding disease evolution. Dividing the MS cohort by (e.g., did false compensation areas previously respond as true disease duration provided an opportunity to study and compare compensation?) or whether these are independent mechanisms rs-fMRI patterns in two clinical subgroups: (i) a subgroup in of action of the pathology. However, these findings support the the early stage of relapsing remitting MS (disease duration <5 suggestion that increased FC has a complex nature that may years, MS ) and (ii) a subgroup at a later stage who have a short exceed the simplistic assumption of an underlying compensatory mild relapsing remitting form of MS (disease duration >5 years, mechanism attempting to limit the brain damage caused by MS MS ). long evolution, exploiting or exhausting the brain functional reserve. Our results show that the widespread FC alterations within Indeed, these considerations suggest that we cannot exclude an the RSNs differentially characterized RRMS patients depending increased FC in MS may even represent a maladaptive response on their disease duration. More importantly, this RRMS subjects to the brain functional-structural deterioration. Another aspect with shorter (MS ) and longer (MS ) disease duration are short long Frontiers in Neurology | www.frontiersin.org 11 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS differentially characterized by patterns of FC alterations affecting in future studies to investigate whether the RNS FC changes different large-scale networks. Specifically, results show that between MS to HC might be influenced by their differences MS present reduced FC compared to MS subjects in a in BDI and MSFC. From a study design point of view, the short long large portion of the fronto-temporo-parietal cortex involving present work is a cross-sectional investigation and although the prevalently cognitive RSNs (in particular PNsup, LNp and alterations within the RSNs indicate a dysfunction of the system, ECN, see Figure 3). On the other hand, MS subjects influenced by focal damage as well as by disease duration, short show greater FC compared to MS in more sensory areas their implication for MS prognosis will require appropriate long (primary somatosensory of TPN and part of the visual areas of longitudinal data. In order to test the validity of the proposed LVN) mainly located in the parieto-occipital cortex (Figure 3). interpretation of FC changes in this mild cohort of MS subjects, Notably, the involvement of both sensory and cognitive RSNs future studies should learn from the results and consider not in the two groups appear almost complementary (i.e., the top- only a longitudinal design, but also a multi-modal approach. The ranked altered RSNs in one contrast—e.g., MS > MS – present study also investigated mechanisms of FC changes in a short long appear as the bottom-ranked altered networks in the opposite mild cohort of RRMS. An interesting question would be to assess contrast—e.g., MS < MS ), suggesting that specific a larger cohort of patients composed of different MS phenotypes, short long temporal dynamics may characterize MS evolution involving including progressive patients, to see whether our findings could neuroplasticity processes and mechanisms exploiting the brain be linked to brain reserve against physical disability as suggested functional reserve (83). Furthermore, the areas of greater FC in Sumowski et al. (84). in MS subjects (compared to MS ) show the largest short long overlap with the map of true functional compensation identified CONCLUSIONS according to the criteria of the multiple-scenario hypothesis (see Table 2). Noteworthy is that the long-term course of relapse- This exploratory study investigates for the first time a voxel- onset MS is variable, and in MS , it is reasonable to anticipate wise correlation between FC and focal damage (T2LL) followed short that some will have a favorable evolution (i.e., becoming like by a further voxel-wise correlation with a clinical score (MSSS). MS subjects) while others will accumulate disability from This can be considered as a basic model on which to build long future relapses or the development of secondary progression. further analysis, for example using longitudinal measures of However, whether marked early FC abnormalities are able to local atrophy (e.g., in the thalamus) or to include specific predict a less favorable disease progression is unclear, and can clinical or neuropsychological scores. Furthermore, this study only be addressed in a prospective longitudinal study. addresses also the impact of disease duration on FC changes. Some considerations need to be addressed with respect to the As a whole, RSN FC analysis shows that functional alterations study’s limitations. From a technical point of view, in order to in MS at a network level cannot be simply described in reduce structured noise artifacts arising from head motion and terms of compensatory mechanisms or of loss of function. The physiological processes, rs-fMRI data used in this study were analysis of FC changes in relation to overall T2 lesion load treated with the ICA-FIX algorithm which operates a robust but and MSSS suggests that the interpretation of FC alterations non-aggressive denoising of resting state signals (52). A further within RSNs is complex, and may include mechanisms, which possible limitation is that the acquisition of B0-fieldmaps was involve but are not limited to true functional compensations. not included in the MRI protocol for this study. Therefore, Of particular interest are the predominant correlations of FC rs-fMRI images have not been corrected for B0-inhomogeneities reductions and T2LL in the cerebellum and the finding satisfying during the pre-processing step. Moreover, in this study the MS the proposed hypothesis of pre-symptomatic alterations in the group was found significantly older than the HC group (see frontal lobe, both worth further investigations. Our findings Table 1), although the distributions of age in the two groups were show also that FC alterations in MS are influenced by disease strongly overlapping (see Figure S3 in Supplementary Material). duration. Indeed, RRMS with shorter and longer disease duration Given this difference, age was added as additional covariate are characterized by distinct patterns of FC alterations with a in the GLM model (see Materials and Methods at section differential involvement of sensory and cognitive RSNs. Despite Between Group RSNs Comparison and Global Alterations the limitations of a cross-sectional design, this study suggests that Ranking). Furthermore, to exclude whether the observed group novel approaches to study FC alterations in multiple large-scale differences in FC is due to age we run a further dual regression networks may provide new insights in the pathophysiology that analysis using age as explanatory variable of interest, which underlies the evolution of relapse onset MS. Further longitudinal means that the contrast vector is non-null (i.e., 1/−1 to test studies are needed to confirm our hypothesis of the mechanisms for positive/negative effect of age on FC) for the age column that drive FC changes in RRMS and to assess whether FC findings and null elsewhere in the GLM design matrix. This verification are able to predict the future course of the disease. analysis resulted in no voxels surviving the significant threshold of p = 0.05 FWE-corrected, indicating that age does not affect AUTHOR CONTRIBUTIONS the FC results we obtained in this study. Moreover, in this study significant differences were found in BDI and also in MSFC when GC, LD, CW-K, and DFM conceptualized the study. GC designed comparing MS to HC. Interestingly, no significant difference and performed the analyses with support LD. TM and DFM between groups was observed in the PASAT test, which is also acquired all MRI data. LD, JD-A, and DFM enrolled patients and the third component of MSFC. Therefore, it would be worth acquired all the clinical and neuro-radiological data helping for Frontiers in Neurology | www.frontiersin.org 12 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS data interpretation. DHM, CW-K, ED, JD-A, and DFM provided Health (RF-INM-2008-114341, RF-2009-1475845 and RC2014- support and guidance with data interpretation with the clinical 2017) to ED and to CW-K (RC2014-2017) and of Engineering contribution of all physicians. GC, CW-K, and DHM wrote the and Physical Sciences Research Council (EPSRC) to CW-K manuscript, with comments from all other authors. (EP/I027084/1). FUNDING ACKNOWLEDGMENTS This work was supported by the New Zealand Brain Research We thank the patients, their families, all healthy volunteers for Institute, the Canterbury Medical Research Foundation, the making this research possible. We also thank the ECTRIMS UK MS society and UCL-UCLH Biomedical Research Centre and Postdoctoral Research Fellowship Program for its support to GC. the National Institute for Health Research, University College London, UK. ECTRIMS and the Multiple Sclerosis International SUPPLEMENTARY MATERIAL Federation (MSIF) supported the work of GC with funding (ECTRIMS Postdoctoral Research Fellowship Program, MSIF Du The Supplementary Material for this article can be found Pré grant). Further support came from grants of European Union online at: https://www.frontiersin.org/articles/10.3389/fneur. (Human Brain Project; HBP-604102), the Italian Ministry of 2018.00690/full#supplementary-material REFERENCES cognitive task at the earliest stage of MS. Hum Brain Mapp. (2003) 20:51–8. doi: 10.1002/hbm.10128 1. Compston A, Coles A. Multiple sclerosis. Lancet (2002) 359:P1221–31. 14. Ksiazek-Winiarek DJ, Szpakowski P, Glabinski A. Neural plasticity in multiple doi: 10.1016/S0140-6736(02)08220-X sclerosis: the functional and molecular background. Neural Plast. (2015) 2. Van Schependom J, Gielen J, Laton J, D’hooghe MB, De Keyser J, and 2015:307175. doi: 10.1155/2015/307175 Nagels G. Graph theoretical analysis indicates cognitive impairment in 15. López-Góngora M, Escartín A, Martínez-Horta S, Fernández-Bobadilla R, MS stems from neural disconnection. Neuroimage Clin. (2014) 4:403–10. Querol L, Romero S, et al. Neurophysiological evidence of compensatory brain doi: 10.1016/j.nicl.2014.01.012 mechanisms in early-stage multiple sclerosis. PLoS ONE (2015) 10:e0136786. 3. Fisniku LK, Brex PA, Altmann DR, Miszkiel KA, Benton CE, Lanyon doi: 10.1371/journal.pone.0136786 R, et al. Disability and T2 MRI lesions: a 20-year follow-up of 16. Pantano P, Mainero C, Caramia F. Functional brain reorganization in multiple patients with relapse onset of multiple sclerosis. Brain (2008) 131:808–17. sclerosis: evidence from fMRI studies. J Neuroimaging (2006) 16:104–14. doi: 10.1093/brain/awm329 doi: 10.1111/j.1552-6569.2006.00029.x 4. Amato MP, Zipoli V, Portaccio E. Multiple sclerosis-related cognitive changes: 17. Sweet LH, Rao SM, Primeau M, Durgerian S, Cohen RA. Functional magnetic a review of cross-sectional and longitudinal studies. J Neurol Sci. (2006) resonance imaging response to increased verbal working memory demands 245:41–6. doi: 10.1016/j.jns.2005.08.019 among patients with multiple sclerosis. Hum Brain Mapp. (2006) 27:28–36. 5. Bischof A, Caverzasi E, Cordano C, Hauser SL, Henry RG. Advances doi: 10.1002/hbm.20163 in imaging multiple sclerosis. Semin Neurol. (2017) 37:538–45. 18. Cader S, Cifelli A, Abu-Omar Y, Palace J, Matthews PM. Reduced doi: 10.1055/s-0037-1607278 brain functional reserve and altered functional connectivity in patients 6. Calabrese M, Poretto V, Favaretto A, Alessio S, Bernardi V, Romualdi C, with multiple sclerosis. Brain (2006) 129:527–37. doi: 10.1093/brain/ et al. Cortical lesion load associates with progression of disability in multiple awh670 sclerosis. Brain (2012) 135:2952–61. doi: 10.1093/brain/aws246 19. Droby A, Yuen KS, Muthuraman M, Reitz SC, Fleischer V, Klein J, et al. 7. Eshaghi A, Prados F, Brownlee WJ, Altmann DR, Tur C, Cardoso MJ, et al. Changes in brain functional connectivity patterns are driven by an individual Deep gray matter volume loss drives disability worsening in multiple sclerosis. lesion in MS: a resting-state fMRI study. Brain Imaging Behav. (2015) Ann Neurol. (2018) 83:210–22. doi: 10.1002/ana.25145 10:1117–26. doi: 10.1007/s11682-015-9476-3 8. Filippi M, Preziosa P, Copetti M, Riccitelli G, Horsfield MA, Martinelli V, et al. 20. Liu Y, Dai Z, Duan Y, Huang J, Ren Z, Liu Z, et al. Whole brain functional Gray matter damage predicts the accumulation of disability 13 years later in connectivity in clinically isolated syndrome without conventional brain MRI MS. Neurology (2013) 81:1759–67. doi: 10.1212/01.wnl.0000435551.90824.d0 lesions. Eur Radiol. (2016) 26:2982–91. doi: 10.1007/s00330-015-4147-8 9. Langer-Gould A, Popat RA, Huang SM, Cobb K, Fontoura P, Gould MK, 21. Hidalgo de la Cruz M, d’Ambrosio A, Valsasina P, Pagani E, Colombo B, et al. Clinical and demographic predictors of long-term disability in patients Rodegher M, et al. Abnormal functional connectivity of thalamic sub-regions with relapsing-remitting multiple sclerosis: a systematic review. Arch Neurol. contributes to fatigue in multiple sclerosis. Mult Scler. (2017) 24:1183–95. (2006) 63:1686–91. doi: 10.1001/archneur.63.12.1686 doi: 10.1177/1352458517717807 10. Azevedo CJ, Overton E, Khadka S, Buckley J, Liu S, Sampat 22. Jaeger S, Paul F, Scheel M, Brandt A, Heine J, Pach D, et al. Multiple M, et al. Early CNS neurodegeneration in radiologically isolated sclerosis-related fatigue: altered resting-state functional connectivity of the syndrome. Neurol Neuroimmunol Neuroinflamm. (2015) 2:e102. ventral striatum and dorsolateral prefrontal cortex. Mult Scler. (2018). doi: 10.1212/NXI.0000000000000102 doi: 10.1177/1352458518758911. [Epub ahead of print]. 11. Papathanasiou A, Messinis L, Zampakis P, Panagiotakis G, Gourzis P, 23. Roosendaal SD, Hulst HE, Vrenken H, Feenstra HE, Castelijns JA, Pouwels Georgiou V, et al. Thalamic atrophy predicts cognitive impairment in PJ, et al. Structural and functional hippocampal changes in multiple relapsing remitting multiple sclerosis. Effect on instrumental activities of sclerosis patients with intact memory function. Radiology (2010) 255:595–604. daily living and employment status. J Neurol Sci. (2015) 358:236–42. doi: 10.1148/radiol.10091433 doi: 10.1016/j.jns.2015.09.001 24. Hawellek DJ, Hipp JF, Lewis CM, Corbetta M, Engel AK. Increased 12. Schoonheim MM, Ciccarelli O. The value of including thalamic atrophy as functional connectivity indicates the severity of cognitive impairment a clinical trial endpoint in multiple sclerosis. Neurology (2018) 90:677–8. in multiple sclerosis. Proc Natl Acad Sci USA. (2011) 108:19066–71. doi: 10.1212/WNL.0000000000005279 doi: 10.1073/pnas.1110024108 13. Audoin B, Ibarrola D, Ranjeva JP, Confort-Gouny S, Malikova I, Ali-Chérif 25. Rocca MA, Valsasina P, Leavitt VM, Rodegher M, Radaelli M, Riccitelli GC, A, et al. Compensatory cortical activation observed by fMRI during a et al. Functional network connectivity abnormalities in multiple sclerosis: Frontiers in Neurology | www.frontiersin.org 13 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS Correlations with disability and cognitive impairment. Mult Scler. (2018) 46. Meyers JE, Meyers KR. Rey Complex Figure Test and Recognition Trial 24:459–71. doi: 10.1177/1352458517699875 Professional Manual. Psychological Assessment Resources. Odessa, FL (1995). 26. Fox MD, Raichle ME. Spontaneous fluctuations in brain activity observed with 47. Uitdehaag BM, Adèr HJ, Roosma TJ, de Groot V, Kalkers NF, Polman functional magnetic resonance imaging. Nat Rev Neurosci. (2007) 8:700–11. CH. Multiple sclerosis functional composite: impact of reference doi: 10.1038/nrn2201 population and interpretation of changes. Mult Scler. (2002) 8:366–71. 27. Biswal B, Yetkin FZ, Haughton VM, Hyde JS. Functional connectivity in the doi: 10.1191/1352458502ms835oa motor cortex of resting human brain using echo-planar MRI. Magn Reson 48. Patriat R, Molloy EK, Meier TB, Kirk GR, Nair VA, Meyerand ME, et al. The Med. (1995) 34:537–41. doi: 10.1002/mrm.1910340409 effect of resting condition on resting-state fMRI reliability and consistency: a 28. Binnewijzend MA, Schoonheim MM, Sanz-Arigita E, Wink AM, van der comparison between resting with eyes open, closed, and fixated. Neuroimage Flier WM, Tolboom N, et al. Resting-state fMRI changes in Alzheimer’s (2013) 78:463–73. doi: 10.1016/j.neuroimage.2013.04.013 disease and mild cognitive impairment. Neurobiol Aging (2012) 33:2018–28. 49. Debernard L, Melzer TR, Van Stockum S, Graham C, Wheeler-Kingshott doi: 10.1016/j.neurobiolaging.2011.07.003 CA, Dalrymple-Alford JC, et al. Reduced grey matter perfusion without 29. Castellazzi G, Palesi F, Casali S, Vitali P, Sinforiani E, Wheeler-Kingshott CA, volume loss in early relapsing-remitting multiple sclerosis. J Neurol Neurosurg et al. A comprehensive assessment of resting state networks: bidirectional Psychiatr. (2014) 85:544–51. doi: 10.1136/jnnp-2013-305612 modification of functional integrity in cerebro-cerebellar networks in 50. Chard DT, Jackson JS, Miller DH, Wheeler-Kingshott CA. Reducing the dementia. Front Neurosci. (2014) 8:223. doi: 10.3389/fnins.2014.00223 impact of white matter lesions on automated measures of brain gray 30. Louapre C, Perlbarg V, García-Lorenzo D, Urbanski M, Benali H, Assouad and white matter volumes. J Magn Reson Imaging (2010) 32:223–8. R, et al. Brain networks disconnection in early multiple sclerosis cognitive doi: 10.1002/jmri.22214 deficits: an anatomofunctional study. Hum Brain Mapp. (2014) 35:4706–17. 51. Ashburner J, Friston KJ. Unified segmentation. Neuroimage (2005) 26:839–51. doi: 10.1002/hbm.22505 doi: 10.1016/j.neuroimage.2005.02.018 31. Dogonowski AM, Siebner HR, Soelberg Sørensen P, Paulson OB, Dyrby 52. Salimi-Khorshidi G, Douaud G, Beckmann CF, Glasser MF, Griffanti L, Smith TB, Blinkenberg M, et al. Resting-state connectivity of pre-motor cortex SM. Automatic denoising of functional MRI data: combining independent reflects disability in multiple sclerosis. Acta Neurol Scand. (2013) 128:328–35. component analysis and hierarchical fusion of classifiers. Neuroimage (2014) doi: 10.1111/ane.12121 90:449–68. doi: 10.1016/j.neuroimage.2013.11.046 32. Faivre A, Robinet E, Guye M, Rousseau C, Maarouf A, Le Troter A, et al. 53. Beckmann CF, DeLuca M, Devlin JT, Smith SM. Investigations into resting- Depletion of brain functional connectivity enhancement leads to disability state connectivity using independent component analysis. Philos Trans R Soc progression in multiple sclerosis: a longitudinal resting-state fMRI study. Mult Lond B Biol Sci. (2005) 360:1001–13. doi: 10.1098/rstb.2005.1634 Scler. (2016) 22:1695–708. doi: 10.1177/1352458516628657 54. Cole DM, Smith SM, Beckmann CF. Advances and pitfalls in the analysis 33. Filippi, M., van den Heuvel, M. P., Fornito, A., He, Y., Hulshoff Pol, and interpretation of resting-state FMRI data. Front Syst Neurosci. (2010) 4:8. H. E., Agosta, F., et al. (2013). Assessment of system dysfunction in doi: 10.3389/fnsys.2010.00008 the brain through MRI-based connectomics. Lancet Neurol. 12:1189–99. 55. Damoiseaux JS, Rombouts SA, Barkhof F, Scheltens P, Stam CJ, Smith SM, doi: 10.1016/S1474-4422(13)70144-3 et al. Consistent resting-state networks across healthy subjects. Proc Natl Acad 34. Polman CH, Reingold SC, Banwell B, Clanet M, Cohen JA, Filippi M, et al. Sci USA. (2006) 103:13848–53. doi: 10.1073/pnas.0601417103 Diagnostic criteria for multiple sclerosis: 2010 revisions to the McDonald 56. Smith SM, Fox PT, Miller KL, Glahn DC, Fox PM, Mackay CE, criteria. Ann Neurol. (2011) 69:292–302. doi: 10.1002/ana.22366 et al. Correspondence of the brain’s functional architecture during 35. Bozin I, Ge Y, Kuchling J, Dusek P, Chawla S, Harms L, et al. activation and rest. Proc Natl Acad Sci USA. (2009) 106:13040–5. Magnetic resonance phase alterations in multiple sclerosis patients doi: 10.1073/pnas.0905267106 with short and long disease duration. PLoS ONE (2015) 10:e0128386. 57. Beckmann CF, Mackay CE, Filippini N, and Smith SM. Group comparison doi: 10.1371/journal.pone.0128386 of resting-state FMRI data using multi-subject ICA and dual regression. 36. Kurtzke JF. Rating neurologic impairment in multiple sclerosis: an Neuroimage (2009) 47(Suppl. 1):S148. doi: 10.1016/S1053-8119(09)71511-3 expanded disability status scale (EDSS). Neurology (1983) 33:1444–52. 58. Filippini N, MacIntosh BJ, Hough MG, Goodwin GM, Frisoni GB, doi: 10.1212/WNL.33.11.1444 Smith SM, et al. Distinct patterns of brain activity in young carriers of 37. Mathiowetz V. Test–retest reliability and convergent validity of the Fatigue the APOE-epsilon4 allele. Proc Natl Acad Sci USA. (2009) 106:7209–14. Impact Scale for persons with multiple sclerosis. Am J Occup Ther. (2003) doi: 10.1073/pnas.0811879106 57:389–95. doi: 10.5014/ajot.57.4.389 59. Smith SM, Nichols TE. Threshold-free cluster enhancement: 38. Roxburgh RH, Seaman SR, Masterman T, Hensiek AE, Sawcer SJ, addressing problems of smoothing, threshold dependence and Vukusic S, et al. Multiple Sclerosis Severity Score: using disability and localisation in cluster inference. Neuroimage (2009) 44:83–98. disease duration to rate disease severity. Neurology (2005) 64:1144–51. doi: 10.1016/j.neuroimage.2008.03.061 doi: 10.1212/01.WNL.0000156155.19270.F8 60. Sbardella E, Tona F, Petsas N, Upadhyay N, Piattella MC, Filippini N, et al. 39. Beck A, Steer R, Brown G. Manual for the Beck Depression Inventory-II. Functional connectivity changes and their relationship with clinical disability San Antonio, TX: Psychological Corporation (1996). and white matter integrity in patients with relapsing-remitting multiple 40. Blair JR, Spreen O. Predicting premorbid IQ: a revision of the sclerosis. Mult Scler. (2015) 21:1681–92. doi: 10.1177/1352458514568826 National Adult Reading Test. Clin Neuropsychol. (1989) 3:129–36. 61. Sbardella E, Petsas N, Tona F, Pantano P. Resting-state fMRI in MS: general doi: 10.1080/13854048908403285 concepts and brief overview of its application. Biomed Res Int. (2015) 41. Dagenais E, Rouleau I, Demers M, Jobin C, Roger É, Chamelian L, 2015:212693. doi: 10.1155/2015/212693 et al. Value of the MoCA Test as a Screening Instrument in Multiple 62. Gour N, Felician O, Didic M, Koric L, Gueriot C, Chanoine V, et al. Functional Sclerosis. Can J Neurol Sci. (2013) 40:410–5. doi: 10.1017/S03171671000 connectivity changes differ in early and late-onset alzheimer’s disease. Hum 14384 Brain Mapp. (2014) 35:2978–94. doi: 10.1002/hbm.22379 42. Delis D. Delis-Kaplan Executive Functioning System: Examiner’s Manual. San 63. Xu X, Yuan H, Lei X. Activation and connectivity within the default mode Antonio, TX: The Psychological Corporation (2001). network contribute independently to future-oriented thought. Sci Rep. (2016) 43. Benedict R. Brief Visuospatial Memory Test-Revised Professional Manual. 6:21001. doi: 10.1038/srep21001 Odessa, FL: Psychological Assessment Resources (PAR), Inc. (1997). 64. Bonavita S, Gallo A, Sacco R, Corte MD, Bisecco A, Docimo R, et al. 44. Rao S. A Manual For The Brief Repeatable Battery Of Neuropsychological Tests Distributed changes in default-mode resting-state connectivity in multiple In Multiple Sclerosis. Milwaukee, WI: Medical College of Wisconsin (1990). sclerosis. Mult Scler. (2011) 17:411–22. doi: 10.1177/1352458510394609 45. Benton AL, Sivan AB, Hamsher K, Varney NR, Spreen O. Contributions to 65. Rocca MA, Valsasina P, Martinelli V, Misci P, Falini A, Comi G, et al. Large- neuropsychological assessment. In: Benton AL, editor. A Clinical Manual. scale neuronal network dysfunction in relapsing-remitting multiple sclerosis. New York, NY: Oxford University Press (1994). Neurology (2012) 79:1449–57. doi: 10.1212/WNL.0b013e31826d5f10 Frontiers in Neurology | www.frontiersin.org 14 August 2018 | Volume 9 | Article 690 Castellazzi et al. Mechanisms of FC Changes in RRMS 66. Smitha KA, Akhil Raja K, Arun KM, Rajesh PG, Thomas B, Kapilamoorthy 78. Parmar K, Stadelmann C, Rocca MA, Langdon D, D’Angelo E, D’Souza M, TR, et al. Resting state fMRI: a review on methods in resting state et al. The role of the cerebellum in multiple sclerosis-150 years after Charcot. connectivity analysis and resting state networks. Neuroradiol J. (2017) 30:305– Neurosci Biobehav Rev. (2018) 89:85–98. doi: 10.1016/j.neubiorev.2018. 17. doi: 10.1177/1971400917697342 02.012 67. Basile B, Castelli M, Monteleone F, Nocentini U, Caltagirone C, Centonze 79. Wilkins A. Cerebellar dysfunction in multiple sclerosis. Front Neurol. (2017) D, et al. Functional connectivity changes within specific networks parallel 8:312. doi: 10.3389/fneur.2017.00312 the clinical evolution of multiple sclerosis. Mult Scler. (2014) 20:1050–7. 80. Castellaro M, Magliozzi R, Palombit A, Pitteri M, Silvestri E, Camera V, doi: 10.1177/1352458513515082 et al. Heterogeneity of cortical lesion susceptibility mapping in multiple 68. Liu Y, Liang P, Duan Y, Huang J, Ren Z, Jia X, et al. Altered thalamic sclerosis. AJNR Am J Neuroradiol. (2017) 38:1087–95. doi: 10.3174/ajnr. functional connectivity in multiple sclerosis. Eur J Radiol. (2015) 84:703–8. A5150 doi: 10.1016/j.ejrad.2015.01.001 81. Faizy TD, Thaler C, Kumar D, Sedlacik J, Broocks G, Grosser M, et al. 69. Faivre A, Rico A, Zaaraoui W, Crespy L, Reuter F, Wybrecht D, Heterogeneity of multiple sclerosis lesions in multislice myelin water imaging. et al. Assessing brain connectivity at rest is clinically relevant in early PLoS ONE (2016) 11:e0151496. doi: 10.1371/journal.pone.0151496 multiple sclerosis. Mult Scler. (2012) 18:1251–8. doi: 10.1177/13524585114 82. Yiannakas MC, Tozer DJ, Schmierer K, Chard DT, Anderson VM, Altmann 35930 DR, et al. ADvanced IMage Algebra (ADIMA): a novel method for depicting 70. Tewarie P, Steenwijk MD, Brookes MJ, Uitdehaag BMJ, Geurts JJG, Stam multiple sclerosis lesion heterogeneity, as demonstrated by quantitative MRI. CJ, et al. Explaining the heterogeneity of functional connectivity findings in Mult Scler. (2013) 19:732–41. doi: 10.1177/1352458512462074 multiple sclerosis: an empirically informed modeling study. Hum. Brain 83. Tomassini V, Matthews PM, Thompson AJ, Fuglø D, Geurts JJ, Johansen-Berg Mapp. (2018) 39:2541–8. doi: 10.1002/hbm.24020 H, et al. Neuroplasticity and functional recovery in multiple sclerosis. Nat Rev 71. Franzmeier N, Göttler J, Grimmer T, Drzezga A, Áraque-Caballero MA, Neurol. (2012) 8:635–46. doi: 10.1038/nrneurol.2012.179 Simon-Vermot L, et al. Resting-state connectivity of the left frontal 84. Sumowski JF, Rocca MA, Leavitt VM, Meani A, Mesaros S, Drulovic J, et al. cortex to the default mode and dorsal attention network supports Brain reserve against physical disability progression over 5 years in multiple reserve in mild cognitive impairment. Front Aging Neurosci. (2017) 9:264. sclerosis. Neurology (2016) 86:2006–9. doi: 10.1212/WNL.0000000000002702 doi: 10.3389/fnagi.2017.00264 72. Schoonheim MM, Meijer KA, Geurts JJ. Network collapse and Conflict of Interest Statement: DHM received honoraria through payments cognitive impairment in multiple sclerosis. Front Neurol. (2015) 6:82. to his employer, UCL Institute of Neurology, for Advisory Committee and/or doi: 10.3389/fneur.2015.00082 Consultancy advice in multiple sclerosis studies from Biogen Idec, Novartis, 73. Tona F, De Giglio L, Petsas N, Sbardella E, Prosperini L, Upadhyay Mitsubishi Pharma Europe and Bayer Schering Pharma. DHM also received N, et al. Role of cerebellar dentate functional connectivity in balance compensation through payments to his employer for performing central MRI deficits in patients with multiple sclerosis. Radiology (2018) 287:267–75. analysis of multiple sclerosis trials from Biogen Idec, Novartis and Apitope. CW-K doi: 10.1148/radiol.2017170311 received research funding from EPSRC, UK MS Society, Horizon2020, Biogen 74. Londoño AC, Mora CA. Nonconventional MRI biomarkers for in vivo Idec, Novartis, Wings for Life, Spinal Research and CHNF. DFM has received monitoring of pathogenesis in multiple sclerosis. Neurol Neuroimmunol honoraria and travel grants from Biogen Idec, Novartis and TEVA. Neuroinflamm. (2014) 1:e45. doi: 10.1212/NXI.0000000000000045 75. Tona F, Petsas N, Sbardella E, Prosperini L, Carmellini M, Pozzilli The remaining authors declare that the research was conducted in the absence of C, et al. Multiple sclerosis: altered thalamic resting-state functional any commercial or financial relationships that could be construed as a potential connectivity and its effect on cognitive function. Radiology (2014) 271:814–21. conflict of interest. doi: 10.1148/radiol.14131688 76. Popescu V, Agosta F, Hulst HE, Sluimer IC, Knol DL, Sormani MP, Copyright © 2018 Castellazzi, Debernard, Melzer, Dalrymple-Alford, D’Angelo, et al. Brain atrophy and lesion load predict long term disability in Miller, Gandini Wheeler-Kingshott and Mason. This is an open-access article multiple sclerosis. J Neurol Neurosurg Psychiatr. (2013) 84:1082–91. distributed under the terms of the Creative Commons Attribution License (CC BY). doi: 10.1136/jnnp-2012-304094 The use, distribution or reproduction in other forums is permitted, provided the 77. Palesi F, Tournier JD, Calamante F, Muhlert N, Castellazzi G, Chard D, original author(s) and the copyright owner(s) are credited and that the original et al. Contralateral cerebello-thalamo-cortical pathways with prominent publication in this journal is cited, in accordance with accepted academic practice. involvement of associative areas in humans in vivo. Brain Struct Funct. (2015) No use, distribution or reproduction is permitted which does not comply with these 220:3369–84. doi: 10.1007/s00429-014-0861-2 terms. Frontiers in Neurology | www.frontiersin.org 15 August 2018 | Volume 9 | Article 690

Journal

Frontiers in NeurologyPubmed Central

Published: Aug 20, 2018

There are no references for this article.