Regional brain atrophy is related to social cognition impairment in multiple sclerosis

ABSTRACT Background: Multiple sclerosis exhibits specific neuropathological phenomena driving to both global and regional brain atrophy. At the clinical level, the disease is related to functional decline in cognitive domains as the working memory, processing speed, and verbal fluency. However, the compromise of social-cognitive abilities has concentrated some interest in recent years despite the available evidence suggesting the risk of disorganization in social life. Recent studies have used the MiniSEA test to assess the compromise of social cognition and have found relevant relationships with memory and executive functions, as well as with the level of global and regional brain atrophy. Objective: The present article aimed to identify structural changes related to socio-cognitive performance in a sample of patients with relapsing-remitting multiple sclerosis. Methods: 68 relapsing-remitting multiple sclerosis Chilean patients and 50 healthy control subjects underwent MRI scans and neuropsychological evaluation including social-cognition tasks. Total brain, white matter, and gray matter volumes were estimated. Also, voxel-based morphometry was applied to evaluate regional structural changes. Results: Patients exhibited lower scores in all neuropsychological tests. Social cognition exhibited a significant decrease in this group mostly related to the declining social perception. Normalized brain volume and white matter volume were significantly decreased when compared to healthy subjects. The regional brain atrophy analysis showed that changes in the insular cortex and medial frontal cortices are significantly related to the variability of social-cognitive performance among patients. Conclusions: In the present study, social cognition was only correlated with the deterioration of verbal fluency, despite the fact that previous studies have reported its link with memory and executive functions. The identification of specific structural correlates supports the comprehension of this phenomenon as an independent source of cognitive disability in these patients.


INTRODUCTION
Multiple sclerosis (MS) has become a significant source of disability and cognitive impairment in young patients 1 attributed to a disconnection syndrome derived from white matter damage 2 .
The patients develop subclinical or measurable cognitive decline since the earliest stages of the disease, with significant changes in cortical recruitment related to cognitive tasks 3 . These phenomena can be understood as the use of cognitive reserve and the start of neural compensation 4 . Interestingly, the existence of mild involvement of socio-cognitive functions early in the MS onset 5 has attracted less interest and could be related to similar mechanisms as those described previously.
Social cognition is defined as the integration of the processes that allow interaction among subjects. It relies on the exchange of social signals that enable us to obtain information about other persons and learn about the environment. Starting from a fundamental phenomenon of attribution of intention, social cognition allows the generation of a shared reality 6 . Some of its most studied domains are theory of mind, the ability to represent the psychological perspective of other subjects 7 , and social perception, the ability to perceive mental states of others based on behavioral signals. There are different approaches to the clinical evaluation of these cognitive domains, allowing to measure whether the subjects can perceive socially relevant information from their environment and to elaborate a coherent mental representation of the cognitive world of others 8 . In this article, the theory of mind can be measured by the faux-pas test (short stories in which intentions must be inferred) and social perception by a series of images in which the patient must identify the emotion that is expressed ( face emotion recognition test, as detailed in the methods section).
There is emerging evidence on the compromise of social cognition in MS and its relatively constant relationship with other cognitive domains 9 . However, the wide variability of the subjects studied in Latin America and the Caribbean has made it difficult to reach a pertinent neuropsychological approach to the local characteristics of the patients, at the same time that the peculiarities of social cognition in this group of patients have not been systematically studied 10 . Recently, a significant relationship with impaired memory and executive functions has been shown, as well as with different measures of global brain atrophy and regional changes in associative visual and executive cortical areas in progressive forms of the disease 11 .
The present study aimed to identify changes in the sociocognitive performance and related structural and clinical findings in a sample of Chilean patients with relapsing-remitting MS.

METHODS
The present investigation was approved by the ethics committee of the medical school of the Pontifical Catholic University of Chile. All subjects gave their informed consent.

Recruitment
Patients with relapsing-remitting MS under clinical follow-up at a single university center in Chile were invited to participate in a neuropsychological evaluation concomitant with their routine scheduled radiological control. Subjects with Mini Mental state examination (MMSE) score lower than 24 points and significant cognitive impairment in previous screenings were excluded as required by local regulatory law for clinical research. Only patients without acute symptoms or active clinical progression were included.
Healthy volunteers recruited through advertisements in university and healthcare settings, with matched education and age, were included to undergo the same neuropsychological examination, and structural/functional images were acquired using the same protocol.
Information about age, achieved educational level, and employment status was registered for both groups. The patients' expanded disability status scale (EDSS) score and disease duration were registered.
We included 68 patients and 50 healthy control subjects who underwent a full neuropsychological evaluation and magnetic resonance imaging. Twenty-six (52%) participants in the healthy control group and 48 (70.6%) in the patients group were women (p=0.039, chi-square test).
The mean age was 37.43 ± 11.2 years in the patients and 37.97 ± 10.76 years in the control group (p-value = 0.765). The mean disease duration was 5.052 ± 3.649 years, and the median EDSS was 1 (range 1-4.5). Most patients (94.3%) had a university or technical career compared to 98% in the control group (p-value = 0.098). Employability level was 91.5% in the patients and 94% in the control subjects, (p-value = 0.731).

Neuropsychological evaluation
Cognitive evaluation was based on the widely reported approach for MS patients 12 . Processing Speed (Symbol Digit Modality Test -SDMT) 13 , Verbal and Visual Episodic Memory, California Verbal Learning Test (CVLT), the Brief Visuospatial Memory Test-Revised (BVMT-R) 14 , Working Memory, Paced Auditory Serial Addition Test (PASAT) 15 , Inhibitory control (Stroop test, Spanish version) 16 , and cognitive shifts with categorical flexibility and lexical fluency, and control of interference (FAS test) 17 were measured. All the cognitive tasks were performed in Spanish by a single trained evaluator as previously detailed 11 . Performance in each of these cognitive tasks was expressed as a z-score (the z-score was calculated using the means and standard deviations derived from the local validation of the test).
Depression, fatigue, and mental status were evaluated by applying the Beck Depression Inventory II (BDI-II) 18 , Fatigue Severity Scale (FSS) 19 , and Mini-Mental Status Examination (MMSE), respectively. Even considering the well-known limitations of MMSE, it was applied to all subjects according to the local regulatory law for clinical research. Social cognition was evaluated using the Mini-Social Cognition and Emotional Assessment (MiniSEA) 20 , a 30 min composite battery, which is the reduced version of the Social Emotional Assessment test 21 . The MiniSEA consists of two different items including a shortened version of the Faux-Pas (FP) and the Face Emotion Recognition to evaluate the theory of mind and social perception, respectively. Facial emotion recognition is considered a relevant and useful approach to evaluate the social perception domain 22 .
The theory of mind section includes ten short stories in which a character inadvertently hurts or offends another. Then, the subject needs to infer the other's mental state by making attributions to their knowledge, beliefs, and emotions. Half of the vignettes are control stories, and the other half includes a principal character who inadvertently offends another. The subject is expected to recognize the situations in which an FP is committed, why the leading subject did it (cognitive theory of mind), and how the victim must have felt (affective theory of mind). The Social Perception item consists of 35 pictures for face affect recognition of basic emotions among a list presented at the bottom of the screen including happiness, sadness, anger, surprise, fear, disgust, and neutrality. A global MiniSEA score and individual performances on social perception and theory of mind task were obtained separately. Given the absence of a locally validated version of MiniSEA, the results are expressed as absolute scores. Social perception and Theory of Mind scores were calculated by expressing the total punctuation as a fraction of 15 points each one. MiniSEA total score was obtained as the sum of both subdomains scores.
The neurocognitive assessment was performed within ten days after the MRI acquisition. The entire neuropsychological examination of each patient took approximately 40 minutes.

Magnetic resonance imaging parameters
Images were acquired by a Philips Ingenia 3 T MRI scanner. T1 weighted 3D images (TR 7.8 ms, TE 3.6 ms, FOV 240 x 240 x 164, flip angle 8°, SENSE factor 2.5, acquisition time 4 min 8 s) and FLAIR (TR 4800 ms, TE 290 ms, FOV 240 x 240 x 164, acquisition time 4 min 33 s) sequences were acquired for the segmentation of brain lesions as detailed below.

Structural study
Structural Image Evaluation using Normalization of Atrophy (SIENAX 23 , part of FSL, Fmrib, Oxford, UK, available at http://www.fmrib.ox.ac.uk/) was used to estimate brain tissue volume, normalized for subject head size. SIENAX starts by extracting brain and skull images from the single wholehead input data. The brain image is then affine-registered to MNI152 space (using the skull image to determine the registration scaling); this is primarily done to obtain the volumetric scaling factor, to be used as a normalization for head size. Next, tissue-type segmentation with partial volume estimation is carried out to calculate the total volume of brain tissue (including separate estimates of volumes of grey matter, white matter, peripheral grey matter, and ventricular CSF).

Voxel-based morphometry (VBM)
Lesions were segmented by the lesion growth algorithm 24 as implemented in the LST toolbox version 1.2.3 (http://www. statisticalmodelling.de/lst.html) for SPM. The algorithm first segments the T1 weighted images into the cerebrospinal fluid, grey matter, and white matter. This information is then combined with the co-registered FLAIR intensities to calculate lesion belief maps. By thresholding these maps with a prechosen initial threshold (k), an initial binary lesion map is obtained which is subsequently grown along voxels that appear hyperintense in the FLAIR.
The outcome, a lesion probability map, in alignment with a T1 weighted 3D image, creates a filled image in native space. Then, volumes based on a voxel-wise comparison of grey and white matter volumes were applied to T1 weighted filled images by using SPM8 (statistical parametric mapping 8; functional imaging laboratory, Wellcome Department of Imaging Neuroscience, Institute of Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm). It was run on Matlab 2015b (Mathworks, Natick, MA, USA). T1 weighted filled images were segmented, replaced to a DARTEL template, and then normalized to MNI space using VBM8 preprocessing in SPM8 25 . A two-sample t-test was selected as a factorial design to perform a regional comparison between control subjects and patients. Multiple regression was chosen as the factorial design for MiniSEA, social perception, and theory of mind scores. Interaction analysis was performed to look for differences in the association of regional brain volume and cognitive performance between groups. Significance was assessed using a family-wise error (FWE) corrected for p-value < 0.05 and cluster size > 20 voxels. The cluster size was selected considering the main size of the regions included in the Harvard-Oxford Atlas to avoid the over-representation of small clusters mostly related to noise 26 .

Statistical analysis
Statistics were performed using Minitab 17 Statistical Software (2010) Computer software, State College, PA, Minitab, Inc. (http://www.minitab.com). Non-parametric tests were selected since that normality of the data was discarded using the Kolmogorov-Smirnov test. Correlation between continuous variables (scores in neuropsychological examinations) was assessed using the Spearman correlation. Differences between groups (age, fatigue, depression, MMSE, and neuropsychological scores) were assessed using the Mann-Withney U test. The interaction of potential confounders, as depression and fatigue scores, with traditional and socio-cognitive domains and the effect of differences in gender composition in the relationship between the domains of social cognition and traditional cognitive domains were assessed by Analysis of Covariance (ANCOVA). The result was considered significant with a p-value < 0.05.
The primary outcomes analyzed through this research were social cognition scores and regional volume as assessed by VBM. Being part of each group (healthy control or patients) was considered as a modifying factor of the effect in the association between clinical findings and MRI data.

RESULTS
No significant difference in mental status (MMSE) was found in comparison with healthy subjects, while significantly higher scores in fatigue and depression questionnaires were found in the patient group (Table 1).
As shown in Table 3, patients exhibited a lower score than the control subjects in the MiniSEA (25.26 vs. 26.106, p = 0.046) and FER test (11.83 vs. 12.82, p = 0.015). The faux-pas score was lower in the patient group, but without statistical significance (13.725 vs. 14.624, p = 0.153).
No significant correlation was found between the performance in neurocognitive tasks and socio-cognition performance, except by verbal fluency and global MiniSEA score in the patient's group (rho = 0.365, p = 0.003).
A significant correlation was not found for the patient group between fatigue and the MiniSEA score (rho = 0.123,    Given the absence of a perfect match in gender composition between the healthy control and patient groups, we also studied the effect of this variable in the observed differences in social cognition. Nevertheless, no significant influence of gender was observed for MiniSEA (F=1.83, p=0.179), Social Perception (F=0.01, p=0.943), or Theory of Mind (F=3.10, p=0.081) scores.
Additionally, when correcting for the effect of differences in gender composition in the relationship between the domains of social cognition and traditional cognitive domains, the analysis of covariance did not show a significant effect of the other cognitive tests on the global score of MiniSEA or the domain of social perception. Only a little higher variability in the theory of mind score was observed due to the effect of performance in the visual memory test (F=5.23, p=0.025).

Structural analysis
Patients exhibited significantly lower whole brain and white matter volumes compared to healthy controls (Table 4). No significant correlation between those global or segmentary volumes and sociocognition performance was found.
When looking for differences in regional atrophy between groups, we found significant clusters of voxels in both hemispheres compromising insula, caudate, cingulate gyrus, and medial frontal gyrus showing decreased values in patients compared to control subjects ( Figure 1).
When the correlation between Regional Volume and Social Cognition Performance was analyzed, specific areas such as the insula, medial frontal cortex, and cingulate cortex were significantly correlated to the cognitive processes of social perception and theory of mind in both patient and control groups (Figure 2).
The interaction analysis showed that the left insula and both medial frontal regions had significantly different levels of association with the theory of mind and social perception scores (Figure 3). Table 5 details the coordinates, laterality, and size of the identified regions.

DISCUSSION
The present study recruited a considerably large sample of patients with recurrent remitting MS and a control group matched by age and educational level, with the aim of studying the deteriorations in the treatment of social cognition and the brain structure changes associated with this phenomenon.  Figure 1. Voxel Based Morphometry. Selected sagittal images show voxels with significant differences between healthy control group (n = 50) and multiple sclerosis patients (n = 68). Brain atrophy observed in multiple sclerosis is not a homogenously distributed phenomenon; specific brain regions concentrate more atrophy. Insula, caudate, cingulate gyrus, and medial frontal gyrus are among the compromised regions also frequently involved in social cognition processes (comparison of control subjects versus patients). Only significant clusters are shown (p<0.05). Warmer colors represent higher T-values. The distribution of the findings in the neuropsychological study requires some attention. The compromise of processing speed, verbal fluency, and verbal and visual memories are compatible with the findings reported in the literature 27 . The relative conservation of performance in tests of working memory and cognitive interference may be related to compensatory changes in brain connectivity, the study of which will require a functional approach. Despite some variations, our neuropsychological findings were in a feasible range of the patients with MS 28 . Possibly, the relative preservation of several explored neurocognitive domains may be related with some characteristics of the sample, as the disease duration and wide use of therapies directed to modify the course of the disease were present in our center.
Despite the relative independence of social cognition from performance on other cognitive tests, there was a significant correlation with the verbal fluency test in the group of patients. There is evidence that sociocognitive functioning and measures of rapid information flow such as verbal fluency are related. The focus of this relationship would be found in the neurodynamics of prefrontal regions 29 . At the same time, it is important to recognize that the findings on both the pattern of socio-cognitive deterioration in MS and its correlation with the structural study do not exactly coincide with what was previously reported, even using the same test battery. A recent study reports a significant relationship of the domains of social cognition with performance in memory and executive functions tests, as well as an association with global brain atrophy and areas of visual association. This differs significantly from our findings, even though Table 5. Voxel Based Morphometry comparison of association between social cognition performance and regional brain volume in control (n=50) versus patients (n=68) (interaction analysis). The left insula and medial frontal region (bilateral) showed significantly different levels of association with the theory of mind and social perception in healthy control subjects and RR-MS patients. Interestingly, those regions exhibit a well-documented influence in socio-cognitive functions. the sample was also Latin American patients (specifically Chilean). This is probably related to different clinical expressions in RR-MS and primary progressive patients. The relative importance of each sub-domain and its structural correlate vary significantly 11 . The above described pattern may have a functional correlate with the findings in the exploration of social cognition. Since this domain showed a compromise based fundamentally on the change in performance in the facial expression recognition test, there is evidence of the relationship of this process with the information processing speed. The relative conservation of the Theory of Mind domain is probably related to the greater complexity of this process and the conservation of the processes of cognitive interference and working memory involved in the performance of these tests. On the other hand, the significant correlation between verbal fluency and the MiniSEA score has a history in the multiple sclerosis literature 30 . Although it would be beyond the sensitivity of our statistical method, we speculate that this process would be mediated by the effect on social perception, since that the FAS test is considered an adequate measure of the integrity of the processes of information flow, flexibility, and control of cognitive interference 17 .

Peak MNI Coordinates
On the other hand, the level of selected depressive symptoms was shown to significantly impact performance in the social cognition test, especially the social perception subdomain. Although some background has indicated the independence of mood and social cognition performance in MS 31,32 , the interference of depression in social cognition has been reported in other clinical settings 31 .
Although in this sample of patients it was possible to rule out the influence of gender on the differences observed in the domain of social cognition, it is important to recognize that gender has been classically associated with socio-cognitive variability. However, many of the differences in this area refer to developmental aspects 33,34 . A future study with a design that could exclude this factor would provide more insight.
This article contributes to clarifying one of the several expressions of social cognition deterioration in MS. Specifically, robust corroboration of the decline in facial expression recognition and its neural correlates has been accomplished. Additionally, our results have the interesting feature of exploring this cognitive process in a considerable number of Latin American patients.

ACKOWLEDGEMENTS
T. Labbe thanks CONICYT (Comisión Nacional de Ciencia y Tecnología) for the doctoral scholarship (21160298) received during the development of this paper. The authors thank CONICYT-PIA, CONICYT, Project number ACT1416 for funding Support. SU thanks funding from the Millennium Science Initiative of the Ministry of Economy, Development and Tourism, grant Nucleus for Cardiovascular Magnetic Resonance; CONICYT, FONDECYT Regular Project Number 1181057.