Уменьшается объем клаустрома в множестве склерозов и прогнозов инвалидности.

Introduction

Multiple Sclerosis (MS) is an immune‐mediated inflammatory and neurodegenerative disease that affects gray matter (GM) and white matter (WM) in the central nervous system (CNS) [ 1 ]. MS GM damage, including both focal lesions and tissue loss, correlates with disability [ 1 ]. In contrast to WM lesions, damage to GM is not readily visualized on clinical imaging and is often assessed only on a research basis. Biomarkers of GM damage that are quantifiable and reproducible are needed.

The claustrum is a thin, irregularly shaped structure located between the external capsule and the extreme capsule near the insula; its function is enigmatic [ 2 ]. Historically credited as the “seat of the soul”, the claustrum is a deep GM nucleus with widespread connectivity [ 3 ]. The highly connected nature of the claustrum positions it to coordinate cortical networks to support higher‐level functions such as attention and consciousness. The claustrum has also been found to co‐activate with other structures in higher level networks [ 4 ]. Additional evidence suggests the claustrum may be involved with slow‐wave and REM sleep by coordinating neuronal inhibition and excitation [ 5 ]. Claustrum connections with thalamus, striatum, and neocortex suggest that damage or disruption of the claustrum may have broad functional implications.

The claustrum has rarely been studied in MS. Neuropathological examination identified a single active lesion within the claustrum of one MS patient [ 6 ]. Prior work has mostly investigated claustrum pathology in MS in conjunction with the nearby insula or as part of deep brain structures [ 7 , 8 , 9 , 10 , 11 ]. Additionally, these studies often report small sample sizes. Thus, conclusions about a specific role of the claustrum in MS pathology are limited [ 7 , 8 , 9 , 11 ].

In MS, the thalamus is an extensively studied deep structure with a high correlation to disability. We hypothesized that, given its extensive connectivity, damage to the claustrum would also strongly correlate with measures of MS‐related disability. The present imaging study assessed the impact of claustrum volume loss on clinical disability in MS patients.

Method

Subjects

Two cohorts of subjects contributed imaging data. The primary analysis in this study was done using data from 15,793 participants drawn from the Multiple Sclerosis Partners Advancing Technology and Health Solutions (MS PATHS) study, comprised of 10 sites in the United States and Europe, which contributed clinical, performance, and imaging data as part of routine clinical care (Table  1 ). Images from healthy control (HC) participants were also collected using identical imaging protocols. MS PATHS participants consented to share pseudo‐anonymized data with the research sponsor and the network investigators under the supervision of local Institutional Review Boards. Our analysis utilized the first available imaging session. Clinical progression over the study period was assessed using the longitudinal clinical assessments. Participants were members of one of the following groups: HC, Clinically Isolated Syndrome (CIS), Relapsing Remitting MS (RRMS), Progressive Relapsing MS (PRMS), Secondary Progressive MS (SPMS), and Primary Progressive MS (PPMS). Recognizing that PRMS is an older term, this was the definition used at MS PATHS initiation and was carried through.

TABLE 1 Demographics. MS PATHS HC CIS RRMS PRMS SPMS PPMS
N
308 573 9506 1385 2794 1227 Age 42.72 (10.87) 44.72 (12.14) 49.90 (11.75) 52.20 (12.54) 54.59 (11.25) 55.77 (12.44) %Female 71.85 71.21 76.32 65.69 72.23 62.56 Disease duration NA 11.34 (9.43) 17.88 (9.89) 20.06 (10.96) 24.02 (11.15) 18.81 (11.48) PDDS NA 0.80 (1.65) 1.16 (1.60) 2.97 (2.20) 3.64 (2.02) 3.64 (2.28) WST −0.16 (2.93) 0.06 (1.67) −0.64 (2.56) −2.57 (5.38) −3.16 (5.75) −3.35 (5.24) MDT −0.15 (1.76) 0.28 (2.09) −0.01 (2.21) −0.70 (3.30) −0.94 (3.41) −0.78 (3.70) PST −0.59 (1.50) −0.20 (1.19) −0.27 (1.20) −1.12 (1.33) −1.00 (1.26) −1.04 (1.25) CST −0.99 (1.39) −0.62 (2.09) −0.50 (1.26) −1.02 (1.49) −1.02 (1.49) −0.99 (1.54) John L. Trotter Cohort HC RMS PMS
N (71) 21 22 28 Age 56.68 (13.28) 63.09 (9.46) 65.58 (8.18) %Female 59.09 83.33 61.29 Disease duration NA 26.95 (9.49) 26.30 (9.48) EDSS NA 2.41 (1.17) 5.96 (1.49)
Note: Mean (SD) shown. Neurological task results are represented as average z scores. EDSS scores are presented as raw averages. Abbreviations: CST, contrast sensitivity test of visual acuity; MDT, manual dexterity test of upper extremity function; PDDS, patient determined disease steps; PST, processing speed test; WST, Walking speed test.

Demographics.

Note: Mean (SD) shown. Neurological task results are represented as average z scores. EDSS scores are presented as raw averages.

Abbreviations: CST, contrast sensitivity test of visual acuity; MDT, manual dexterity test of upper extremity function; PDDS, patient determined disease steps; PST, processing speed test; WST, Walking speed test.

Data comprising the validation cohort included 21 HC, 28 PMS, and 22 RMS subjects at the John L. Trotter MS Center not included in the MS PATHS cohort. The PMS cohort had no relapses or gadolinium‐enhancing lesions within 24 months preceding enrollment. RMS and HC subjects in the validation cohort were recruited to reflect the age and sex distribution of the non‐relapsing PMS patients. RMS participants had no history of disability worsening unless directly related to an MS attack.

Clinical Assessments

MS PATHS clinical assessments were performed during routine clinical follow‐up. Patients self‐reported their disability status using the Patient Determined Disease Steps (PDDS), which ranges from 0 to 8; a higher score reflects increased disability status. The PDDS highly correlates with the examiner‐determined Expanded Disability Status Scale (EDSS) [ 12 ]. Objective performance testing was completed using the MS Performance Test (MSPT) which is a digital adaptation of several commonly used tests based on the MS Functional Composite [ 13 , 14 , 15 ]. The MSPT included the walking speed test (WST) which is analogous to the 25‐ft walk assessment (25FTW); the processing speed test (PST) which is an electronic adaptation of the Symbol Digit Modality Test; the manual dexterity test (MDT) which is a digital adaptation of the Nine‐hole Peg Test; and the contrast sensitivity test (CST) which is an adaptation of the Sloan Low Contrast Letter Acuity Test. In the validation John L. Trotter MS Center cohort, a full EDSS was performed by a trained neurologist‐examiner.

Imaging Acquisition

MS PATHS imaging parameters have been previously described [ 14 ]. In brief, each subject and visit contributed a 3D T1‐weighted MP‐RAGE and a 3D T2‐weighted FLAIR image. Both sequences were acquired with 1 mm [ 3 ] isotropic voxels with the same field of view. All imaging was performed on 3 T Siemens (Erlangen, Germany) scanners.

Subjects in the John L. Trotter cohort were imaged on a 3 T Trio MRI scanner (Siemens) using a 32‐channel phased‐array radiofrequency head coil. FLAIR images (voxel size 1 × 1 × 3 mm) and MP‐RAGE (voxel size 1 × 1 × 1 mm) were acquired.

Claustrum and Thalamus Segmentation

Claustrum probability maps were generated using a deep learning segmentation tool applied to T1‐weighted images [ 16 ]. Claustrum masks were then created by applying a threshold of 0.4 to the probability masks. Figure  1 shows an example of the claustrum segmentation approach [ 17 ]. The bilateral thalami were segmented using FastSurfer, which is a machine learning adaptation of FreeSurfer [ 18 , 19 ]. Left and right claustrum volumes were highly correlated in the MS PATHS cohort ( r  = 0.86, p  < 0.05). Similarly, left and right thalamic volumes were highly correlated ( r  = 0.94, p  < 10 −16 ). Because there was no statistical difference in left versus right volumes in either claustrum or thalamic measures across subjects, the total (left plus right) claustrum and thalamic volumes were used for each individual in analyses.

Claustrum segmentation. (a) Unaltered brain scan in the axial view, this is the original image from a scan on a MS PATHS participant. (b) Claustrum probability mapping thresholded at 0.4, highlighted in red. (c) Claustrum is highlighted in white after probability determination is set. Left and right claustrum volumes are correlated at r  = 0.82 in the MS PATHS cohort, meaning there are subtle differences between representations of the right and left claustrum in imaging sequences.

Statistical Analysis

RStudio version 2023.06.2 + 561 was used to conduct all analyses. Total claustrum volumes and total thalamic volumes were determined for each individual. MS patient groups versus HC were compared, with MS patients stratified by clinical subtype. To quantify mean volume differences, we fit a linear model for claustrum and thalamic volume separately with group (HC, CIS, RRMS, PRMS, SPMS, PPMS), age, and sex as factors. Additionally, this regression model, and all subsequent models, used claustrum and thalamic volumes corrected for estimated total intracranial volume by regression across the study population. A sensitivity regression analysis was performed to assess the impact of “site” on our primary variables of interest; inclusion of site did not significantly impact results (< 2% variance) and was not considered further. To analyze a potential effect of MS type, we determined Cohen's D values to compare individual patient groups. These analyses were replicated in the John L. Trotter MS Center validation cohort.

The relationships between claustrum or thalamic volume and measures of clinical disability or performance were examined. Corrected claustrum or thalamic volume was modeled as a function of z ‐scored performance measures (WST, MDT, PST, CST), MS subtype (allowing for an interaction term with performance), age, and sex.

Correlation measures were conducted to assess and compare the impact of claustrum volumes and thalamic volumes on PDDS scores and age. Full correlations (i.e., Pearson correlation coefficient) were performed to show the dependency between two variables. Partial correlations were performed to control for the effects of covariates. For example, the partial correlation of claustrum volume and PDDS was controlled for the shared contribution from age and thalamic volume.

The preceding analyses quantified the cross‐sectional relationships between claustrum and thalamic volumes, MS subtypes, and disability measures. Baseline claustrum and thalamic volumes were examined individually for the ability to predict longitudinal clinical worsening. Clinical worsening was defined as sustained disability progression (SDP), an increase in PDDS score by one or more that was sustained until the end of observation. By definition, worsening of disability cannot be confirmed as sustained at the last visit. Median follow‐up time after CDP was 14.6 months; 80% of worsening events had 6 months of follow‐up after CDP. A Cox proportional hazards model was used to examine the relationship between SDP and baseline claustrum volume (expressed as quartiles), baseline thalamic volume (expressed as quartiles), age, and baseline PDDS score. We also performed ROC analysis using SDP as the outcome and claustrum and thalamic volume as predictors. From this ROC model, we extracted an ideal cutoff value, as well as performance characteristics such as area under the curve (AUC), sensitivity, and specificity.

Results

Participant Demographics

The MS PATHS cohort consisted of 15,793 participants who contributed imaging and clinical data (Table  1 ). The clinical task data (WST, MDT, PST, and CST) are represented as z scores corrected for age. The MS cohort was predominantly female. Progressive subtype patients were older than control or relapsing MS patients. Secondary progressive patients had a longer disease duration compared to relapsing or primary progressive patients. PDDS was higher in progressive compared to relapsing MS patients.

The John L. Trotter cohort consisted of 71 participants who were enrolled in imaging studies and followed longitudinally (Table  1 ). By design, the HC and RMS patients were of similar age and sex distribution as the PMS patients. RMS and PMS patients were matched by age and sex, while HC were recruited after the fact and were matched based on availability. PMS patients had a higher EDSS (mean 5.96) than RMS patients (mean 2.41).

Claustrum Volumes Are Reduced in MS and Vary by Subtype

Using MS PATHS patient data, claustrum volumes were assessed in relation to clinical subtype (HC, CIS, RRMS, PRMS, SPMS, and PPMS) after correction for age and sex (Figure  2 ). The distribution of claustrum volumes is shown in Figure  2a . Claustrum volumes were significantly reduced in MS compared to HC (HC vs. all MS groups p  = 9.37 × 10 −5 ) and were reduced more so between progressive subtypes and relapsing subtypes (CIS vs. PMS p  = 5.91 × 10 −6 ; RRMS vs. PMS p =  2.49 × 10 −11 ).

Claustrum and thalamus volumes are reduced in MS. (a) Combination violin plots and box plots of claustrum volume are shown. The horizontal line in the box plots is the median, and the box represents the inner quartile distribution. The violin plots encompass the entire distributions. The HC group had larger claustrum volumes than all MS patient groups ( p  < 10 –5 ). CIS and RRMS groups had larger claustrum volumes than progressive MS groups ( p  < × 10 –6 ; p  < 10 –11 ). (b) Similar plot to (a) for thalamic volume. The HC group had larger thalamus volumes than all MS patient groups ( p  < 1.071 × 10 –4 ). CIS and RRMS groups had larger thalamus volumes than progressive MS groups ( p  < 10 –11 ; p  < 10 –11 ). Regression data regarding figure a are included in Table  S2 . (c) Pairwise group contrasts were extracted from the linear models for claustrum and thalamic volumes. Cohen's D values for the thalamus (red text) and claustrum (blue text). A smaller value means a larger difference between groups. All analyses were corrected for age, sex, and estimated intracranial volume (ETIV).

Overall, a similar pattern was present in the thalamus (Figure  2b ) wherein thalamic volumes were reduced in MS compared to HC (HC vs. all MS groups p  = 1.169 × 10 −6 ) and further reduced in PMS compared to CIS/RRMS subtypes (CIS vs. PMS p  = 10 −11 ; RRMS vs. PMS p  = 10 −11 ).

This pattern is reflected in the effect sizes (Cohen's D ) shown in Figure  2c , where moderate effect sizes differentiate HC from progressive MS, but smaller effect sizes differentiate HC from relapsing MS. Figure  2c above the diagonal shows the Cohen's D values for comparing claustrum volumes between groups; below the diagonal shows the effect sizes comparing thalamic volumes. Both claustrum and thalamic volumes were similarly reduced by MS, particularly in progressive MS subtypes, as reflected in the largely symmetric contrast matrix in Figure  2c . For both the claustrum and the thalamus, MS patients had smaller volumes compared to HC (claustrum mean MS 519.2 vs. HC 734 mm 3 , p  < 0.00001; thalamus mean MS 13,074.2 vs. HC 14,984 mm 3 , p  < 0.00001) (solid box Figure  2c ). CIS patients had claustrum and thalamus volumes intermediate between MS and HC groups (claustrum CIS mean 624 mm 3 ; thalamus CIS mean 13,970 mm 3 ). Relapsing MS subtypes had larger claustrum and thalamic volumes compared to progressive MS subtypes (claustrum relapsing mean 599 mm 3 vs. progressive mean 465 mm 3 , p  < 0.00001; thalamus relapsing mean 13,697 mm 3 vs. progressive mean 12,659 mm 3 , p  < 0.00001). Additional statistical information can be found in Table  S1 .

To examine the replicability of the claustrum finding, a similar analysis was performed in the John L. Trotter MS Center cohort. Consistent with our results in the larger MS PATHS cohort, claustrum volumes were smaller in the MS groups (mean 270.00 mm 3 ) compared to HC (mean 378.00 mm 3 ) (HC vs. RMS p  = 0.130; HC vs. PMS p  = 0.016) (Figure  3 ), with the smallest volumes in the PMS patients (mean 252.00 mm 3 ).

Claustrum volumes were reduced in MS patients in a replication cohort. Replication Cohort ( N  = 71) of patients at the John L. Trotter MS Center and HC volunteers. Claustrum volume was smaller in the MS patient cohort than in the HC cohort ( p  < 0.130). Claustrum volume was not significantly smaller in the progressive MS group compared to the relapsing MS group ( p  < 0.569), but was significantly smaller compared to the HC ( p  < 0.016) group.

Claustrum Volumes Correlate With Measures of Disability in People With MS

In the MS PATHS cohort, smaller claustrum volumes were associated with worse performance on all objective measures of performance (WST, PST, MDT, and CST; Table  2a ). A similar pattern was found with thalamus volume measures and performance measures (Table  2b ). The statistical strength of these relationships was similar for the claustrum and thalamus (Table  2 , Tables  S2 and S3 ). The correlations of volume with test performance were not strongly affected by MS subtype.

TABLE 2 Claustrum and thalamus volumes correlated to a similar degree with subjective clinical performance and objective clinical test results in MS patients. Predictors Claustrum estimates Claustrum Std. Beta Thalamus estimates Thalamus Std. Beta Z pst −34.93 −0.16 19.72 0.01 Z mdt 3.03 0.03 32.13 0.04 Z cst 13.73 0.07 121.34 0.09 Z wst 7.48 0.08 11.35 0.02
Note: WST, PST, MDT, and CST were strongly related to the claustrum and thalamus. Smaller volumes in either brain area were associated with poorer performance. Regression data regarding Figure  3 are presented in Tables  S2 and S3 .

Claustrum and thalamus volumes correlated to a similar degree with subjective clinical performance and objective clinical test results in MS patients.

Note: WST, PST, MDT, and CST were strongly related to the claustrum and thalamus. Smaller volumes in either brain area were associated with poorer performance. Regression data regarding Figure  3 are presented in Tables  S2 and S3 .

Claustrum and Thalamus Volumes Correlate With PDDS Scores to a Similar Degree

The relationships between PDDS score (self‐reported disability, analogous to EDSS), age, claustrum volume, and thalamic volume were examined using full (Figure  4 , below the diagonal) and partial correlation (Figure  4 , above the diagonal) analyses. PDDS showed a strong correlation with all four variables. Full correlation within age, PDDS score, claustrum volume, and thalamic volume suggests strong linear relationships between each variable and PDDS score. A significant fraction of PDDS variability is explained by age, claustrum volume, and thalamic volume. Partial correlation using age, PDDS score, claustrum volume, and thalamic volume showed that smaller claustrum and thalamic volumes each correlated with worse PDDS. The magnitude of the partial correlation was similar for the claustrum and thalamus, demonstrating that both structures similarly and independently correlated with worse PDDS. Individually, there are strong relationships between each variable and PDDS score.

Claustrum and Thalamus volumes are similarly and independently negatively correlated with disability. Full (below diagonal) and partial (above diagonal) correlation results are presented as a matrix. Covariates included in the partial correlation are PDDS score, age, claustrum volume, and thalamus volume.

Claustrum Volumes Predict Future Disability Accrual More Accurately Than Thalamic Volumes

Analyses thus far investigated cross‐sectional relationships between claustrum and thalamic volumes and clinical test results and characteristics. Next, we compared the predictive power of baseline claustrum and thalamic volumes on longitudinal SDP. Claustrum volume, thalamus volume, age, and baseline PDDS were examined as predictors. Smaller claustrum volumes, accounting for thalamic volume, predicted more rapid disability accumulation ( p  = 0.030) (Figure  5a ). After accounting for the claustrum volume, thalamic volumes were not independently associated with SDP, as demonstrated by the stepwise acceleration in SDP with smaller claustrum (Figure  5b ; upper) compared to no clear relationship with thalamic volumes (Figure  5b ; lower). As expected, higher age and worse disability at baseline were also independently associated with SDP.

Claustrum volume predicts disability accumulation better than thalamus volume. (a) Forest plot of claustrum and thalamic volumes as predictor of subsequent SDP, using a Cox Hazard model. (b) Kaplan–Meier survival plots for volumes separated into quartiles, controlled for age and PDDS score. Claustrum volume (upper) predicted MS disease progression (PDDS) over 5 years better than thalamus volume (lower).

We next investigated claustrum and thalamic volumes as continuous variables for predicting SDP using ROC analysis. Claustrum volume outperformed thalamic volume (AUC 0.67 vs. 0.60) as a predictor of SDP. For claustrum and thalamus, the sensitivities were similar, but claustrum had a more favorable specificity for SDP (Table  3 ).

TABLE 3 ROC analysis shows claustrum volume outperforms thalamic volume in the prediction of SDP out to 5 years through both AUC (Area Under the Curve) and specificity. Cutoff (mm 3 ) AUC AUC CI Specificity Sensitivity Claustrum 668 0.67 [0.65. 0.69] 0.55 0.72 Thalamus 13,695 0.60 [0.58. 0.62] 0.45 0.72
Note: Sensitivity was similar for both claustrum and thalamic volumes.

ROC analysis shows claustrum volume outperforms thalamic volume in the prediction of SDP out to 5 years through both AUC (Area Under the Curve) and specificity.

Note: Sensitivity was similar for both claustrum and thalamic volumes.

Discussion

The present study measured the claustrum in a large number (> 15,000 MS and HC) of subjects and showed the claustrum to be smaller in people with MS than HC. Smaller claustrum size in MS was further confirmed in a local MS patient and healthy cohort. The comparative size of the claustrum in people with MS versus healthy people has not been reported prior. Claustrum size was diminished most in those with progressive MS clinical subtypes. We found claustrum and thalamic volume differences between MS patients and HC, and among MS patient clinical subgroups to be similar. This similar pattern of volume loss seen in different clinical subgroups suggests that MS clinical subtypes are not distinct from one another, but rather exist upon a spectrum, with disease processes impacting the claustrum and thalamus similarly. Smaller volumes of the claustrum and thalamus independently correlated with worse clinical performance measures in MS patients. Most importantly, smaller claustrum size, at baseline, predicted worse longitudinal neurological performance out to 5 years with improved specificity compared to thalamic volume.

While the role of the claustrum in humans may not be entirely known, the claustrum is an important structure, as evidenced by its conservation across species. The claustrum has extensive connections with the cortex, both ipsilateral and contralateral. Similar to the thalamus, the claustrum is believed to serve as a hub structure, connecting regions of the brain to one another. The hub‐like nature of the claustrum is supported by the high degree of structural connectivity to multiple areas of cortex and deep gray structures, particularly the putamen. In both rodent models and humans, the claustrum exhibits highly distributed functional connectivity as well. These converging lines of evidence support the hypothesis that damage to the claustrum may result in disproportionately higher levels of disability.

This study replicated the well‐documented association between thalamic volume and disability in MS. Of note, thalamic volume did not independently contribute to predicting disability progression compared to claustrum volume. In models excluding claustrum volume, thalamic volume was associated with disability progression, but more modestly than claustrum volume. ROC analyses showed that claustrum and thalamus were similarly sensitive for predicting disability progression, but claustrum was more specific. This constellation of findings suggests that claustrum volume loss is a more statistically reliable biomarker but emphasizes that damage to both structures is highly important to MS patients.

Thalamic volume is a popular biomarker of MS pathology owing to its strong relationship with disability and the ease with which the thalamus can be segmented. This makes the thalamus‐based biomarkers ideal for clinical use. In this work, we show a modest advantage of the claustrum as a correlate of longitudinal disability, with the caveat that volumetric assessment of the claustrum is computationally more involved. A primary conclusion of this work is that hub‐like GM structures are critically important to MS pathobiology, manifesting as strong correlations with clinical disability.

Recent data indicate the claustrum to be involved in feedforward inhibition in both cortico‐claustral and claustra‐cortical communications [ 20 ]. In particular, the claustrum is important for the suppression of cortical activity related to irrelevant stimuli and thus for focusing attention. In the present study, the processing speed test (PST), which requires focused attention for optimal performance and is frequently affected in MS [ 21 ], was highly correlated with both claustrum and thalamic volumes.

In MS, a gradient of injury has been suggested wherein more GM damage and microglial activation occur near the ependymal or meningeal surface compared to tissue more distant from the cerebrospinal fluid (CSF) [ 22 , 23 ]. It has been proposed that factors within the CSF may mediate this damage; this damage from CSF factors to the thalamus has been strongly implicated by several autopsy studies [ 24 ]. Unlike the thalamus, which lies adjacent to the ependyma of the third ventricle or cortical GM bordering the meninges, the claustrum does not border CSF. Thus, mechanisms of injury to the claustrum in MS are unlikely to be due to toxic factors from CSF. Although the claustrum volume loss observed might be related to MS lesions intrinsic to the claustrum, given the small size of the claustrum versus its extensive connections, it seems more likely that reduced connections outside of the claustrum are the driving force for claustrum atrophy. This hypothesis parallels observations in the thalamus, which shares with the claustrum the key features of being a highly connected deep GM structure impacted by MS. Prior work has shown that extrinsic, rather than intrinsic, pathology drives thalamic atrophy [ 25 ]; we hypothesize that the same mechanism affects the claustrum.

The claustrum, due to its highly connected anatomy, may be significantly involved in many critical common functions. Damage to the claustrum caused by MS may impact essential functions, such as sleep waves and the circadian rhythm. Inhibitory interneurons in the claustrum are important for slow wave sleep and consolidation of learning during sleep [ 5 ]. Claustrum neurons show a greater spike rate during cortical slow wave sleep than during the awake state in mice [ 5 ]. Slow wave sleep is associated with cognitive test performance and is generally reduced in MS, contributing to poor performance [ 26 ]. More studies are needed to understand claustrum injury or disturbed connections in relation to the disturbance of specific sleep states in MS.

This study has a few limitations which may impact the interpretation of the results. We relied on an automated segmentation algorithm for measuring claustrum volume. This was necessary given the thousands of MR scans studied but may have some unrecognized bias in segmenting diseased structures. The observation of stepwise associations of worse claustrum atrophy with worse MS clinical subtypes supports that the automated segmentation performed well. MS PATHS utilized a standard image acquisition protocol, but was implemented on a variety of clinical scanners which may introduce unrecognized bias. The PDDS, a self‐reported measure, could introduce bias related to self‐assessment of disability. In practice, PDDS and EDSS, an objective measure, are highly correlated [ 15 ]. Moreover, the objective measures used in this study (WST, MDT, PST, and CST) demonstrated similar associations with claustrum and thalamic atrophy. We did not assess focal MS lesions in or around the claustrum. Analysis of intrinsic claustrum lesions in relation to MS is the focus of ongoing study. Lastly, compared to the discovery cohort of 15,793, the replication cohort had a limited size of 71, but more detailed clinical information.

This study establishes the importance of the claustrum as a deep GM structure that is affected by MS. Loss of claustrum volume showed strong associations with current and future neurological disability. Given the learning dysfunctions and reduced attention that are common in MS, we speculate that reduced function of the claustrum could contribute to these impairments.

Author Contributions

N.S. contributed to the analysis, drafting, interpretation, reviewing, and final approval of the work. B.J. contributed to the analysis, drafting, and final approval of the work. D.A.Y. contributed to the data acquisition, conception of the imaging software used in the work and its final approval. R.T.N. contributed to the data acquisition and final approval of the work. M.R.B. contributed to the design, drafting, reviewing, final approval, and accountability of the work. A.H.C. contributed to the conception and design, data acquisition, drafting, reviewing, final approval, and accountability of the work.

Conflicts of Interest

N.S. and B.J. have nothing to disclose. R.T.N. has received consulting fees from Alexion Pharmaceuticals, Biogen, Bristol Myers Squibb, Celltrion, Genentech, Genzyme, EMD Serono, Horizon Therapeutics, Immpact‐Bio, Kyverna, Novartis, Sandoz, and TG Therapeutics. M.R.B. has received consulting honoraria from Genentech, Bristol Myers Squibb, Novartis, EMD Serono, TG Therapeutics, Sanofi, and Roche. A.H.C. has received consulting fees from Biogen, Bristol Myers Squibb, Genentech, EMD Serono, Horizon Therapeutics, Janssen, Novartis, Octave, and TG Therapeutics.

Supporting information

Table S1.

Table S2.

Table S3.

Acknowledgements

MS PATHS was supported by Biogen. This study was supported by generous gifts from the Firestone Fund and the Leon and Harriet Felman Fund for Human MS Research. D.A.Y. is supported by NIH/NIA grants RF1 AG077658 and RF1 AG082030. M.R.B. was supported by K23 NS128325 and a generous gift from The Laura Fund. A.H.C. was supported by The Manny & Rosalyn Rosenthal—Dr. John L. Trotter MS Center Chair in Neuroimmunology of the Foundation for Barnes‐Jewish Hospital. We thank Dr. Biao Xiang for his contributions to imaging the validation cohort.

Funding: MS PATHS was supported by Biogen. This work was supported by the National Institutes of Health, K23 NS128325, the Leon and Harriet Felman Human MS Research Fund, the Manny & Rosalyn Rosenthal—Dr. John L. Trotter MS Center Chair in Neuroimmunology of the Foundation for Barnes‐Jewish Hospital, and the Firestone Fund.

Matthew R. Brier and Anne H. Cross contributed equally to this work.

Funding Statement

This work was funded by National Institutes of Health ‐ National Institute of Neurological Disorders and Stroke grant K23 NS128325; Leon and Harriet Felman Human MS Research Fund; The Manny & Rosalyn Rosenthal ‐ Dr. John L. Trotter MS Center Chair in Neuroimmunology of the Foundation for Barnes‐Jewish Hospital; Firestone Fund.

Data Availability Statement

MS PATHS data are available through a data use agreement with Biogen. John L. Trotter data will be made available upon reasonable request and establishment of a data use agreement.

References

Associated Data

Supplementary Materials

Table S1.

Table S2.

Table S3.

Data Availability Statement

MS PATHS data are available through a data use agreement with Biogen. John L. Trotter data will be made available upon reasonable request and establishment of a data use agreement.