Quantifying Remyelination Using χ-Separation in White Matter and Cortical Multiple Sclerosis Lesions

Abstract

Background and objectives: Myelin and iron play essential roles in remyelination processes of multiple sclerosis (MS) lesions. χ-separation, a novel biophysical model applied to multiecho T2*-data and T2-data, estimates the contribution of myelin and iron to the obtained susceptibility signal. We used this method to investigate myelin and iron levels in lesion and nonlesion brain areas in patients with MS and healthy individuals.

Methods: This prospective MS cohort study included patients with MS fulfilling the McDonald Criteria 2017 and healthy individuals, aged 18 years or older, with no other neurologic comorbidities. Participants underwent MRI at baseline and after 2 years, including multiecho GRE-(T2*) and FAST-(T2) sequences. Using χ-separation, we generated myelin-sensitive and iron-sensitive susceptibility maps. White matter lesions (WMLs), cortical lesions (CLs), surrounding normal-appearing white matter (NAWM), and normal-appearing gray matter were segmented on fluid-attenuated inversion recovery and magnetization-prepared 2 rapid gradient echo images, respectively. Cross-sectional group comparisons used Wilcoxon rank-sum tests, longitudinal analyses applied Wilcoxon signed-rank tests. Associations with clinical outcomes (disease phenotype, age, sex, disease duration, disability measured by Expanded Disability Status Scale [EDSS], neurofilament light chain levels, and T2-lesion number and volume) were assessed using linear regression models.

Results: Of 168 patients with MS (median [interquartile range (IQR)] age 47.0 [21.7] years; 101 women; 6,898 WMLs, 775 CLs) and 103 healthy individuals (age 33.0 [10.5] years, 57 women), 108 and 62 were followed for a median of 2 years, respectively (IQR 0.1; 5,030 WMLs, 485 CLs). At baseline, WMLs had lower myelin (median 0.025 [IQR 0.015] parts per million [ppm]) and iron (0.017 [0.015] ppm) than the corresponding NAWM (myelin 0.030 [0.012]; iron 0.019 [0.011] ppm; both p < 0.001). After 2 years, both myelin (0.027 [0.014] ppm) and iron had increased (0.018 [0.015] ppm; both p < 0.001). Younger age (p < 0.001, b = -5.111 × 10^-5), lower disability (p = 0.04, b = -2.352 × 10^-5), and relapsing-remitting phenotype (RRMS, 0.003 [0.01] vs primary progressive 0.002 [IQR 0.01], p < 0.001; vs secondary progressive 0.0004 [IQR 0.01], p < 0.001) at baseline were associated with remyelination. Increment of myelin correlated with clinical improvement measured by EDSS (p = 0.015, b = -6.686 × 10^-4).

Discussion: χ-separation, a novel mathematical model applied to multiecho T2*-images and T2-images shows that young RRMS patients with low disability exhibit higher remyelination capacity, which correlated with clinical disability over a 2-year follow-up.

Figure 1. MRIs and χ-Separation Maps

(A) 3D magnetization-prepared 2 rapid gradient echo. (B) Quantitative susceptibility mapping based on multiecho gradient echo. (C) χ-separation map of diamagnetic sources. Brighter tones of yellow indicate higher values of susceptibility (i.e., higher content of myelin). (D) χ-separation map of paramagnetic sources. Brighter tones of yellow indicate higher values of susceptibility (i.e., higher content of iron).

Figure 2. Imaging Postprocessing and Transformation Pipeline

(A) Individual acquisitions were skull stripped and rigidly coregistered using FSL. (B) Longitudinal acquisitions were coregistered on MP2RAGE, as follows: MP2RAGE were first affinely (skull-stripped) and then nonlinearly (with the skull) registered using ITK Snap (greedy diffeomorphic registration). (C) Lesion masks (i.e., masks for white matter lesions, cortical lesions, normal appearing white matter, and normal appearing gray matter) from baseline were transformed to baseline and follow-up maps, using nearest neighbor interpolation. (D) Regions of interest (i.e., masks for corpus callosum and putamen derived from FreeSurfer) were transformed from baseline to baseline and from follow-up to follow-up, respectively, using nearest neighbor interpolation. FLAIR = fluid-attenuated inversion recovery; FSL = FMRIB software library; MEGRE = multiecho gradient echo; MP2RAGE = magnetization-prepared 2 rapid gradient echo; χ = chi (indicates negative and positive χ-separation maps).

Figure 3. Iron and Myelin in the Referential Regions of Interest of Healthy Individuals and Patients With MS at Baseline and Follow-Up

Comparison of absolute susceptibility values, in (A) healthy individuals vs patients with MS, and (B) baseline vs follow-up. p Values derive from Wilcoxon rank-sum tests and are considered significant if p < 0.0125 (i.e., after Bonferroni correction for 4 comparisons). HC = healthy controls; MS = multiple sclerosis; ns = not significant; ppm = parts per million.

Figure 4. Content of Iron (Blue) and Myelin (Yellow) in Regions of Interest in Patients With MS

Absolute susceptibility values of iron (blue) and myelin (yellow) at baseline are displayed, in white matter and cortical lesions, as well as the perilesional normal appearing white and gray matter. In addition, the corpus callosum (as a referential region of interest with a high content of myelin) and the putamen (as a referential region of interest with a high content of iron) are displayed. CL = cortical lesions; NAGM = normal appearing gray matter; NAWM = normal appearing white matter; ppm = parts per million; WML = white matter lesion.

Figure 5. Longitudinal Changes of Myelin and Iron in White Matter and Cortical Lesions

Susceptibility index for myelin in white matter lesions (A) and cortical lesions (C), as well as the susceptibility index for iron in white matter lesions (B) and cortical lesions (D) between baseline and 2-year follow-up. The susceptibility index describes the content of myelin (or iron, respectively) in the lesion, relative to the myelin (or iron, respectively) in perilesional normal appearing tissue. A susceptibility index of 1 indicates an equal susceptibility indicative of similar content of myelin (or iron, respectively) in lesional and perilesional tissue. One dot gives the median index of 1 patient. Diagonal lines connect the same patient between baseline and follow-up. p Values derive from linear mixed models adjusted for age. The figures indicate a significant increase of both myelin and iron in white matter lesions. Although there was a significant increase of iron in cortical lesions between baseline and follow-up, this finding lost its significance after adjustment for age. CL = cortical lesion; WML = white matter lesion.

Figure 6. Changes in Myelin as Well as Clinical Changes Between Baseline and Follow-Up in White Matter Lesions

In the scatter plot, 1 dot represents 1 patient. The y-axis shows the median change in myelin susceptibility index between baseline and follow-up. The x-axis shows the change of EDSS between baseline and follow-up. The regression analysis indicates that patients with a greater increase of the median susceptibility index have larger clinical improvement. Bar charts show all lesions of selected patients with highest EDSS improvement (top improvers, green dots, left bar charts) and highest EDSS progression (top progressors, red dots, right bar charts), with demographic and clinical information. In the bar charts, red bars show lesions with a decrease of susceptibility on myelin-sensitive maps between baseline and follow-up (indicating demyelination), green bars show lesions with an increase of susceptibility on myelin-sensitive maps between baseline and follow-up (indicating remyelination). BL = baseline; EDSS = expanded disability status scale; FU = follow-up; PPMS = primary progressive multiple sclerosis; RRMS = relapsing-remitting multiple sclerosis; y/o = year old.