Сочетание осаждения железа при повреждении MS с способностью к ремиилюции с использованием дисперсионного разделения источников MRI.

Introduction

Multiple sclerosis (MS) is one of the most common neuroinflammatory diseases of the central nervous system (CNS) ( Reich et al., 2018 ). One of the main histological characteristics of MS lesions in the CNS is myelin degeneration (demyelination) and recovery (remyelination), which are relevant to dynamic course of the disease ( Lucchinetti et al., 2000 , Reich et al., 2018 ). Particularly, remyelination can help axons resist neurodegeneration, which is a cause of permanent disability in people with MS. Hence, temporal changes in the myelin content of MS lesions have been recognized as a key marker in understanding the disease course ( Franklin and Simons, 2022 , Rahmanzadeh et al., 2022 , Klotz et al., 2023 ) and elucidating the underlying mechanisms of remyelination ( Franklin and Simons, 2022 , Klotz et al., 2023 ).

Impaired iron homeostasis is another important pathological aspect of MS lesions. Cellular iron is essential for myelin synthesis but also can mediate CNS neurodegeneration ( Hametner et al., 2013 , Stephenson et al., 2014 ). For instance, iron deposition in microglia/macrophages is associated with oxidative stress on neurons ( Gilgun-Sherki et al., 2004 , Hametner et al., 2013 , Absinta et al., 2019 , Absinta et al., 2021 ) as well as polarization towards a pro-inflammatory profile ( Hametner et al., 2013 , Absinta et al., 2019 , Absinta et al., 2021 ). On the other hand, iron depletion is linked to remyelination failure due to the role of iron in myelin production ( Stephenson et al., 2014 , Lee et al., 2019 ). Therefore, studying brain iron and its relevance to myelin changes can be useful for elucidating the role of iron in demyelination and remyelination in MS lesions and disease progression ( Absinta et al., 2019 , Absinta et al., 2021 , Rovira and Pareto, 2024 ).

Histopathology has been recognized as the gold standard tool to study iron and myelin alterations in MS lesions. However, its invasiveness limits its utility for longitudinal monitoring, precluding the tracking of pathological dynamics in MS lesions over time ( Lucchinetti et al., 2000 , Kuhlmann et al., 2017 ). As an alternative, MRI has been a key imaging method for the longitudinal assessment of the CNS in MS ( Brex et al., 2002 , Thompson et al., 2018 , Granziera et al., 2021 , Rahmanzadeh et al., 2022 ). Dedicated efforts have been directed towards developing MRI methods that enhance sensitivity of MRI to iron and/or myelin changes in lesions, such as myelin water imaging and magnetization transfer imaging for myelin, and magnetic susceptibility-sensitive contrasts for iron ( MacKay and Laule, 2016 , Granziera et al., 2021 ). However, these methods have certain limitations related to spatial resolution, quantitative accuracy, and specificity to targeted pathologic substrates ( Walsh et al., 2013 , Granziera et al., 2021 , Rahmanzadeh et al., 2022 ). For example, quantitative susceptibility mapping (QSM) has demonstrated unique sensitivity to iron and myelin but presents challenges in interpretation when both substrates coexist within the same voxel ( Wang and Liu, 2015 ).

Magnetic susceptibility source separation (χ-separation or chi-separation) imaging ( Shin et al., 2021 ) is a novel MRI technique that enables the separate quantification of diamagnetic (χ dia ) and paramagnetic (χ para ) susceptibility signals , corresponding to diamagnetic myelin and paramagnetic iron concentrations in CNS tissues ( Shin et al., 2021 , Kim et al., 2023 ). In MS lesions, this enables the visualization of the myelin alterations (demyelination and remyelination) ( Muller et al., 2024 ), as well as iron-laden microglia/macrophages across the lesion, which visualization has been limited to the lesion boundaries in conventional MRI methods (e.g., paramagnetic rim sign [PRS] ( Absinta et al., 2019 , Absinta et al., 2021 )) due to the confounding effects of myelin.

In this study, we investigated longitudinal changes in myelin and iron signals, and their interplay in MS lesions using χ-separation. Specifically, we identified intra-lesional iron accumulation with χ para hyperintensity (termed hyper-paramagnetic sign [HPS]) and explored a potential of early-stage HPS as a new imaging marker for impaired remyelination capability.

Materials and methods

Study participants

This study was approved by the local institutional review board, and written informed consent was obtained from all participants. Among 60 consecutive participants with MS (pwMS), 50 underwent MRI scans annually—including χ-separation imaging—between August 2016 and October 2023 (38 women, age = 35.1 ± 10.1 years old, follow-up duration = 39.2 ± 21.1 months; see Fig. S1 ). The diagnosis of MS was made by experienced neurologists (W.K. and J.H.L.), using the most up-to-date McDonald criteria available at the time of diagnosis ( Thompson et al., 2018 ). For each participant, demographics and clinical features, sex, and diagnosis date were collected. Additionally, at every MRI scan, Neurostatus Expanded Disability Status Scale (EDSS), age, and treatment status (no treatment, immune-suppressant, or immune-modulatory) were collected.

To explore MS lesion dynamics of myelin and iron from their available earliest stage during the follow-ups, we included lesions that satisfied the following criteria for the analysis: 1) newly-noted lesion (NNL) during follow-ups on T2-weighted fluid-attenuated inversion recovery (FLAIR) image or 2) contrast-enhancing lesions (CEL) in contrast-enhanced T1-weighted image (CET1) ( Chen et al., 2014 , Absinta et al., 2016 ). Throughout the follow-up period, we define the earliest time point at which a lesion meets the criteria as its “early stage.” The χ-separation image acquired at this time served as the baseline measurement for each lesion’s longitudinal analysis. Only participants who underwent at least one additional follow-up χ-separation imaging after baseline were included in the longitudinal analysis ( Fig. S1 ).

MRI data acquisition and postprocessing

We used two MRI protocols at different 3T MRI systems: Protocol A (type I and II) and Protocol B (type I and II), both including T2-FLAIR for lesion segmentation and NNL detection, CET1 for identifying CELs, and MRI contrasts for χ-separation imaging (multi-echo gradient echo [MEGE] and T2-weighted contrasts). The detailed MRI sequence protocols are summarized in Supplementary Materials (Tables S1 and S2) .

For χ-separation imaging ( Shin et al., 2021 ), MEGE data was preprocessed to calculate R2* (= 1/T2*) and local frequency maps, and T2-weighted contrasts were used for estimating R2 (= 1/T2) map (see Supplementary Methods ). The phase images of MEGE data were processed with phase unwrapping using SEGUE ( Karsa and Shmueli, 2019 ), followed by background removal using V-SHARP ( Wu et al., 2012 ), and signal-to-noise ratio-weighted echo combination, generating local frequency maps. The data processing was conducted using in-house methods developed in MATLAB (MATLAB 2022a, MathWorks Inc., Natick, MA, USA). The χ-separation maps, including χ para and χ dia maps, were reconstructed using the local frequency map along with R2′ or R2* as inputs ( Ji et al., 2024 ). The χ-separation reconstruction was performed using publicly available software, Chi-separation toolbox ( https://github.com/SNU-LIST/chi-separation ). More detailed processing is summarized in Supplementary Methods . In this study, χ para and χ dia maps represent absolute susceptibility, where higher intensity indicates a higher concentration of paramagnetic and diamagnetic susceptibility sources, respectively.

To explore the inter-scanner reliability of χ-separation, a region of interest (ROI) analysis was conducted in 10 white matter and 8 deep gray matter regions to compare χ-separation results from two follow-up scans acquired with different hardware systems (see Supplementary Methods for details).

Clinical and radiological evaluations

For each lesion, radiologic characteristics were assessed at the time of detection (as NNLs or CELs) by two experienced investigators (H-G. S. and J.J.) in consensus. The characteristics included the presence of contrast enhancement (CE), PRS, and HPS. The PRS was assessed using QSM, following the consensus criteria ( Bagnato et al., 2024 ). HPS was evaluated in the χ para map, characterized by χ para hyperintensity relative to adjacent normal-appearing white matter at least on two slices where visibility is maximized ( Fig. 1 A). Of note, by definition, HPS includes conventional hyper-paramagnetic signal along the boundary (equivalent to PRS) but also hyper-paramagnetic signal within the lesions, extending the concept of PRS. Hyperintense central veins ( Sati et al., 2016 ) on χ para were carefully assessed and excluded for HPS detection ( Bagnato et al., 2024 ). Additionally, lesional χ-separation contrast relative to surrounding normal-appearing white matter was assessed (e.g., hypo-intensity in χ dia [lesional demyelination] and hyper-intensity in χ para [lesional iron deposition]), as proposed by Kim et al. ( Kim et al., 2023 ).

Presentations of hyper-paramagnetic sign and paramagnetic rim sign (A) and longitudinal outcomes of myelination in MS lesions on χ dia (B). MS lesions are denoted by white arrows in MRI images. A. Examples of MS lesions with different HPS and PRS classifications (first column: T2 FLAIR; second column: χ dia ; third column: χ para ) and schematic diagram of χ para appearances corresponding to the classifications (fourth column). MS lesions consistently exhibit hypointensity in χ dia (demyelination) but display distinct patterns in χ para (paramagnetic iron signal) compared to adjacent normal-appearing white matter. In χ para maps, HPS is characterized by intra-lesional hyperintensity and/or paramagnetic rim, indicating iron deposition throughout lesions, while PRS is represented by a hyperintense paramagnetic rim surrounding the lesion. Note that the central vein signs are also visible in χ para map in the first and second rows. B. Different types of longitudinal outcomes of myelinations in MS lesions on χ dia are shown in chronological order from left to right (time points [TPs] 1–3). Lesions are categorized as remyelinating or demyelinating, and stable category is not displayed. HPS = hyper-paramagnetic sign, PRS = paramagnetic rim sign.

Through visual inspection of MRI follow-ups, we categorized the longitudinal changes of each lesion observed on χ dia maps into three types, based on changes in signal intensity over time: remyelinating (increasing χ dia signal), stable, and demyelinating (decreasing χ dia signal) ( Fig. 1 B). To assess the quantitative characteristics of lesions and their longitudinal changes, region-of-interest analyses were performed on the included lesions across time points. Lesion masks were manually segmented in FLAIR image using ITK-SNAP by an experienced neuroradiologist (J.J.), and lesion volume (in mm 3 ) and averaged χ para and χ dia values (in ppb) were measured per lesion.

Statistical analysis

Descriptive statistics of clinicoradiological features were assessed both on a per-participant and on a per-lesion basis. Also, a group-wise comparison between lesions with different longitudinal outcomes in χ dia (i.e., remyelinating, stable, and demyelinating lesions) was performed. The choice of statistical tests was based on the characteristics of the variables ( Supplementary Methods ). To further investigate the influence of iron deposition on longitudinal myelin changes using quantitative measurements, we also fitted a linear regression model to examine the relationship between the quantitative χ para signal at lesion detection vs. the annual rate of χ dia signal changes (= [χ dia at last observation − χ dia at lesion detection]/observation duration).

An association between early-stage HPS and future remyelination outcome (remyelinated or not) was tested using a generalized linear mixed model (GLMM) with random effects for participants. First, predictors significantly related to future remyelination outcomes (response variable) were identified by performing a univariate GLMM analysis for each lesional clinicoradiological feature at the time of lesion detection (predictor variable), including age, sex, disease duration, EDSS, treatment status, demyelination sign on χ dia , HPS, PRS, CE, lesion volume, and χ dia and χ para values. Subsequently, predictor variables with significant correlations (with threshold P of 0.05) in the univariate GLMM regressions were further analyzed using a multivariate GLMM to evaluate the effects of early-stage HPS on future remyelination outcome after adjusting potential confounders. Another multivariate GLMM was used to calculate the annual change of adjusted χ dia in lesions with and without early HPS, exploring the effects of early-stage HPS on the quantitative myelin evolution. For all multivariate GLMM, age was additionally included to adjust its effects. More details are described in Supplementary Methods .

To explore the clinical implication of HPS, we compared the number of HPS-positive (HPS+ ) and PRS-positive (PRS+ ) lesions at the baseline scan in the whole cohort (i.e., 50 pwMS, regardless of the presence of NNLs or CELs [ Fig. S1 ]), between the groups with and without the confirmed disability improvement (CDI). CDI was evaluated as defined by Cree et al. ( Cree et al., 2021 ). Given that these variables are countable, a Poisson regression was used for the comparison.

All analyses were performed with R statistical software (version 4.2.2, R Foundation; https://www.R-project.org ) and MATLAB (MATLAB 2022a, MathWorks Inc., Natick, MA, USA), and P < 0.05 was considered to indicate statistical significance.

Results

Characteristics of included participants and lesions for longitudinal χ-separation follow-up

We identified 62 MS lesions with early-stage characteristics (31 CELs and 38 NNLs; 7 overlapped) from 23 participants (16 women; 33.0 ± 10.4 y/o; 2.7 ± 2.0 lesions per participant, Fig. S1 ). All participants were with relapsing-remitting multiple sclerosis at their first visit. Detailed demographic information is summarized in Table 1 .

Table 1 Characteristics of participants with early-stage MS lesions. Demographic and Clinical Data (No. = 23) Age, mean ± SD, y 33.0 ± 10.4 Sex, No. (%)  Female 16 (70)  Male 7 (30) Disease duration at lesion detection, median (IQR), m 21 (4.5 – 76) EDSS at lesion detection, median (IQR) 1 (1–2.25) Clinical status at lesion detection, No. (%)  Stable 11 (48)  Active 12 (52) Treatment, No. (%)  Immunomodulatory 11 (48)  No 12 (52) EDSS = Expanded Disability Status Scale.

Characteristics of participants with early-stage MS lesions.

EDSS = Expanded Disability Status Scale.

Early-stage Lesion appearance in χ-separation maps

When the lesions were detected with early-stage features, most of the lesions (58 out of 62, 94 %) appeared hypointense in χ dia (indicating demyelination). At the same time point, approximately half of the lesions (27 of 62, 44 %) were HPS+, indicating relatively frequent iron deposition in the lesions at their early stage, and 11 lesions (18 %) showed paramagnetic rim appearance (i.e., PRS+ lesions). Of note, even though HPS and PRS were assessed separately in different MRI contrasts, all early PRS+ lesions showed early HPS.

Longitudinal outcomes of MS lesion in myelin signal

There were a total of 226 observations from 62 lesions (3.6 ± 1.4 yearly observations per lesion; follow-up duration of 35 ± 14 months). When categorizing the longitudinal outcomes of MS lesions based on χ dia signal, approximately half of the lesions (26 of 62) showed increased χ dia signal over time (indicating remyelination; Table 2 ). Additionally, 14 lesions exhibited stable χ dia , and the rest of 22 lesions showed progressive χ dia signal loss (demyelination).

Table 2 Lesion characteristics according to the types of longitudinal outcomes in myelin signal. Characteristic Remyelinating Stable Progressively demyelinating P value (No. = 26) (No. = 14) (No. = 22) Contrast enhancement at lesion detection, No. (%) 14 (54) 6 (43) 11 (50) 0.80 Early-stage χ dia appearance, No. (%) 0.82  Hypointense χ dia (Demyelination) 25 (96) 13 (93) 20 (91)  Isointense χ dia (Preserved myelination) 1 (3.8) 1 (7.1) 2 (9.1) Early-stage HPS, No. (%) <0.001  HPS+ 2 (7.7) *† 12 (86) † 13 (59) *  HPS− 24 (92) 2 (14) 9 (41) Early-stage PRS, No. (%) 0.20  PRS+ 2 (7.7) 4 (29) 5 (23)  PRS− 24 (92) 10 (71) 17 (77) Early-stage lesion volume, mean ± SD, mm 3 175 ± 157 * 163 ± 110 425 ± 523 * 0.020 Early-stage χ para , mean ± SD, ppb 17 ± 8 † 26 ± 12 † 22 ± 12 0.018 Early-stage χ dia , mean ± SD, ppb 26 ± 7 23 ± 7 22 ± 8 0.65 Final χ dia , mean ± SD, ppb 28 ± 8 *† 20 ± 8 † 15 ± 8 * 0.001 HPS = hyper-paramagnetic sign, PRS = paramagnetic rim sign, ANOVA = Analysis of variance. * significant post-hoc Tukey’s or Fisher’s exact test between remyelinated and progressively demyelinated group (P < 0.05). † significant post-hoc Tukey’s or Fisher’s exact test between remyelinated and stable group (P < 0.05).

Lesion characteristics according to the types of longitudinal outcomes in myelin signal.

HPS = hyper-paramagnetic sign, PRS = paramagnetic rim sign, ANOVA = Analysis of variance.

significant post-hoc Tukey’s or Fisher’s exact test between remyelinated and progressively demyelinated group (P < 0.05).

significant post-hoc Tukey’s or Fisher’s exact test between remyelinated and stable group (P < 0.05).

Factors related to longitudinal outcomes of MS lesion in myelin signal

When investigating clinicoradiological factors related to the longitudinal outcomes of lesions in myelin signal ( Table 2 ), the early-stage HPS was significantly associated with longitudinal myelin outcomes of lesions (P < 0.001), suggesting an association between the absence of early HPS and the future remyelination. In contrast, early-stage PRS was not significantly associated (P = 0.20; see Discussion). Notably, only a small fraction of HPS+ lesions (2 of 27) showed remyelination, whereas the majority of HPS-negative (HPS− ) lesions (24 of 35, 69 %) exhibited a future increase in χ dia , suggests remyelination failure in HPS+ lesions but not in HPS− lesions. Figs. 2 and S2 illustrate these myelin signal changes according to the presence of HPS. Interestingly, among 11 early HPS− lesions without future remyelination, eight lesions (72 %) showed subsequent HPS in the next MRI within a year after their detection ( Fig. S3 ), suggesting an extended time window for early-stage HPS related to remyelination capacity (see Discussion). Additionally, when we analyzed a relationship between early-stage χ para signal vs. the annual rate of χ dia signal changes, a linear regression model reported a negative relationship between the two quantities (slope = −0.10 ppb/(ppb/year) [95 % confidence interval = −0.01 to −0.18 ppb/(ppb/year)], P = 0.02), aligning with our qualitative observation above based on HPS and myelin longitudinal outcomes.

Examples of a HPS + lesion with demyelination (A) and a HPS − lesion with remyelination (B). Each panel displays T2 FLAIR (first row) and myelin signal χ dia (second row) images (TP1-4, chronological order from left to right; A: axial view, B: coronal view). CET1 and paramagnetic iron (χ para ) images of the lesions are displayed in inset in T2 FLAIR and χ dia , respectively (white boxes). A. When the demyelinating lesion is newly noted (TP1), it exhibits hypo-intense χ dia signal (early-stage demyelination) but HPS in χ para (early-stage intralesional iron deposition). This lesion with early HPS shows progressively expanding signal loss in χ dia , indicating demyelination over time. B. The remyelinating lesion in right periventricular area (arrows on B) shows rim-shaped contrast enhancement at TP1 (B, CET1). At this early stage (TP1), the lesion shows diffusive hypo-intense χ dia (early-stage demyelination) and hypo-intense χ para (without iron accumulation). The lesion boundary is denoted by yellow dotted lines. This lesion without early HPS shows shrinking lesion size in both T2 FLAIR and χ dia images and also recovery of myelin signal intensity in χ dia (remyelinating lesion). TP = Time point, CET1 = contrast-enhanced T1 weighted image. HPS = hyper-paramagnetic sign.

GLMM analysis on longitudinal myelin signal changes

A univariate GLMM using binary remyelination outcomes based on longitudinal χ dia as a response variable identified four variables that were significantly associated with future remyelination. They included the absence of early-stage HPS, sex (female), lower early-stage χ para value, and high early-stage EDSS (P < 0.05; Table 3 ). In the multivariate GLMM analysis, only the absence of early HPS and high early-stage EDSS were significantly associated with future remyelination (P < 0.001 and P = 0.004, respectively; Multivariate analysis 1 in Table 3 ). In a separate multivariate GLMM analysis using quantitative χ para value at the early stage of the lesion instead of HPS, both low early-stage χ para signal and high early-stage EDSS associated with future remyelination outcomes, aligning with early HPS (Multivariate analysis 2 in Table 3 ).

Table 3 GLMM analysis results of participants' demographic and clinical characteristics and lesional radiological features on future remyelination outcomes*. Univariate analysis Characteristic Coefficient † 95 % CI, lower 95 % CI, upper P value Age at lesion detection, y −0.035 −0.117 0.047 0.405 Sex (male) −0.86 −0.869 −0.851 <0.001 Disease duration at lesion detection, m 0.006 −0.012 0.024 0.487 EDSS at lesion detection 1.007 0.181 1.833 0.017 No treatment at lesion detection 1.603 −0.138 3.345 0.071 No early-stage χ dia hypointensity −1.766 −5.117 1.586 0.302 Absence of early-stage HPS 3.614 1.383 5.844 0.001 Absence of early-stage PRS 1.375 −0.383 3.132 0.125 Contrast enhancement at lesion detection 0.6 −0.757 1.956 0.386 Early-stage lesion volume, mm 3 −0.002 −0.005 0.001 0.154 Early-stage χ dia , ppb 0.055 −0.041 0.151 0.259 Early-stage χ para , ppb −0.076 −0.143 −0.009 0.025 

 Multivariate Analysis 1 Characteristic Coefficient † 95 % CI, lower 95 % CI, upper P value Age at lesion detection, y −0.072 −0.166 0.022 0.134 Sex (male) −0.687 −2.409 1.035 0.434 EDSS at lesion detection 1.051 0.336 1.766 0.004 Absence of early-stage HPS 3.916 1.953 5.878 <0.001 

 Multivariate Analysis 2 Characteristic Coefficient † 95 % CI, lower 95 % CI, upper P value Age at lesion detection, y −0.06 −0.149 0.029 0.186 Sex (male) −0.562 −2.482 1.358 0.566 EDSS at lesion detection 0.895 0.119 1.672 0.024 Early-stage χ para , ppb −0.089 −0.173 −0.005 0.038 GLMM = generalized linear mixed model, HPS = hyper-paramagnetic sign, PRS = paramagnetic rim sign. * Future successful remyelination outcome = 1, future remyelination failure outcome = 0. † Fixed effect regression coefficient.

GLMM analysis results of participants' demographic and clinical characteristics and lesional radiological features on future remyelination outcomes*.

GLMM = generalized linear mixed model, HPS = hyper-paramagnetic sign, PRS = paramagnetic rim sign.

* Future successful remyelination outcome = 1, future remyelination failure outcome = 0.

Fixed effect regression coefficient.

In another multivariate GLMM regression utilizing 226 yearly measured quantitative χ dia observations of the lesions ( Table S3 ), the annual change of adjusted χ dia was −1.94 ppb/year (demyelinating) in early HPS+ lesions, and 0.66 ppb/year (remyelinating) in early HPS− lesions (P < 0.001; Fig. 3 ). Other variables involved in the analysis are summarized in Table S3 . Of note, an interaction effect was observed between early-stage HPS and follow-up duration, which also closely accords with the results illustrated in Fig. 3 .

Longitudinal changes of adjusted χ dia value (diamagnetic myelin signal) in early HPS+ (A) and early HPS− (B) lesions ( total 226 yearly time points for 62 lesions, 3.6 ± 1.4 observations per lesion). In each panel, black dots connected by a line represent repeated measurements of the same lesion over time, while a blue solid line denotes the linear regression model fit for all lesions in each panel (grey area: 95 % confidence interval). After adjustment of confounding factors using a multivariate GLMM, the estimated annual change of adjusted χ dia is −1.94 ppb/year (demyelinating) in early HPS+ lesions and 0.66 ppb/year (remyelinating) in early HPS− lesions (P < 0.001). HPS = hyper-paramagnetic sign, GLMM = generalized linear mixed model. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Clinical significance of iron deposition in lesions

Among 50 participants included in this study, pwMS with CDI had a lower number of HPS+ lesions at baseline (median [IQR], 1 [1–3])) than those without CDI (4 [0–9]; P < 0.001). The number of PRS+ lesions was also lower in the CDI group compared to non-CDI (1.0 [0–2] in CDI vs. 1 [0–7] in non-CDI; P < 0.001).

Discussion

In this study, we employed an advanced MRI technique, χ-separation, to investigate the longitudinal changes in myelin signal in MS lesions, as well as the effects of early-stage iron signal hyperintensity on the myelin signal evolution over time. We identified various types of longitudinal outcomes of MS lesions in myelin signal and their association with early-stage iron signal hyperintensity, so-called HPS. Our results suggested that early-stage HPS is closely linked to impaired remyelination capacity ( Fig. 4 ) and decreased disability improvement.

Schematic representation of scenarios for myelin changes in MS lesions according to HPS. In early stage of lesions (yellow box), MS lesions could exhibit dichotomous patterns of iron deposition, detected by paramagnetic iron signal in χ para : lesion with normal or less iron deposition in lesions (HPS− lesions; A) and lesion with excessive iron deposition (HPS+ lesions; B). During this early stage, some lesions may display contrast enhancement, possibly preceding iron deposition (orange triangle). After the early stage, the remyelination pathways of these lesions diverge depending on the presence of HPS. HPS− lesions (A) typically demonstrate successful recovery of myelin contents (remyelination), indicated by an increase in diamagnetic myelin signal on χ dia (C). On the other hand, HPS+ lesions (B) exhibits progressive myelin loss, reflected by a decrease in χ dia signal (D). These scenarios illustrate the association between the early-stage HPS in MS lesions and their subsequent remyelination outcome. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Our results suggested that early-stage iron deposition in MS lesion, detected by HPS, was associated with future remyelination failure. Although iron plays conflicting roles in the remyelination process ( Hametner et al., 2013 , Stephenson et al., 2014 , Lee et al., 2019 , Absinta et al., 2021 ), our findings are consistent with the previous studies, which demonstrated decreased remyelination capacity in the lesions with iron deposition, compared to those without ( Hametner et al., 2013 , Walsh et al., 2013 ). Notably, multiple histopathological studies observed a nodular iron accumulation pattern within MS lesions ( Hametner et al., 2013 , Wisnieff et al., 2015 , Absinta et al., 2019 ), which corresponds to hyperintense χ para core in HPS+ lesions. These studies also reported iron accumulation inside of active lesions but not in remyelinated ones ( Hametner et al., 2013 ). This excessive iron can be associated with oligodendrocyte cytotoxicity ( McGahan et al., 2005 ), impeding myelin repair ( Cui et al., 2013 ). A recent study also suggested hypo- or iso -intense QSM (indicative of low iron concentration ( Walsh et al., 2013 , Wisnieff et al., 2015 )) in MS lesions can predict their future remyelination ( Rahmanzadeh et al., 2022 ), agreeing with our observations of remyelination in HPS− lesions. However, it is important to note that interpreting conventional QSM signal as histopathological substrates (iron or myelin) can be challenging due to the joint contribution of both iron and myelin to QSM ( Walsh et al., 2013 , Wisnieff et al., 2015 , Wiggermann et al., 2017 ), while χ-separation can provide more specific information on paramagnetic iron and diamagnetic myelin separately.

In this study, the number of HPS and PRS showed significant correlations with the disability improvement, highlighting their clinical relevance. HPS also demonstrated higher sensitivity than PRS in prediction of future remyelination ( Table 2 ). This suggests that HPS, identifiable in χ-separation, may extend the clinical aspect of PRS for detecting inflammatory activity by allowing the additional assessment of iron levels “inside” MS lesions ( Rovira and Pareto, 2024 ). In conventional MRI, detecting iron accumulation has been largely limited to the lesion's border (i.e., PRS), as myelin also contributes to the contrasts at the lesion core. However, in active white matter lesions, intracellular iron accumulation in inflammatory myeloid cells occurs not only at the lesion boundary but also within the lesions ( Hametner et al., 2013 ). Mehta et al. demonstrated that iron is preferentially taken by pro-inflammatory (M1-polarized) macrophages as well as promotes the expression of macrophage M1 polarization ( Mehta et al., 2013 ). Additionally, a previous study on early active MS lesions suggested macrophage polarization (M1/M2) correlated with iron deposition ( Tham et al., 2021 ). These associations between iron levels and microglia/macrophages activation may indicate overlapping histopathological characteristics between PRS and HPS. Therefore, HPS, with its added sensitivity to intralesional iron accumulation, may serve as a more comprehensive imaging marker, potentially extending PRS. Of note, progressively demyelinating lesions without PRS or HPS may be explained by other factor affecting remyelination capability, such as proliferation and differentiation of oligodendrocyte precursor cells, clearance of myelin debris, and favorable microenvironment ( Franklin and Simons, 2022 ). Hence, a further study considering multiple factors is needed for better understanding the pure effects of HPS on remyelination capability.

Among 35 early HPS− lesions, 11 lesions did not show remyelination during follow-up, but eight of them demonstrated subsequent HPS in the next MRI ( Fig. S3 ). Notably, all the eight lesions showed CE at their early stage. It could be a plausible explanation for this observation that in MS lesions, CE may precede the formation of HPS with temporal overlap ( Fig. 4 ) ( Chen et al., 2014 , Kuhlmann et al., 2017 , Franklin and Simons, 2022 ). Previous studies have reported that early active lesions can exhibit lower iron concentrations than normal-appearing white matter, but iron accumulation occurred inside the active lesions at a pathologically later stage ( Hametner et al., 2013 , Tham et al., 2021 ), which aligns with our observations. It is worth mentioning that due to the nature of observational cohort of this study, follow-up interval was relatively long, which may had led to under-representation of relationship between CE and HPS ( Absinta et al., 2016 ). A future study with more frequent scan intervals could further consolidate the findings.

Although previous studies have reported negative associations between baseline EDSS and brain remyelination capacity ( Bodini et al., 2016 , Muller et al., 2024 ), our results were inconsistent with these reports ( Table 3 ). This discrepancy may stem from differences in study designs; specifically, the earlier studies either did not examine long-term remyelination ( Bodini et al., 2016 ) or did not focus on lesion characteristics from their early stage ( Muller et al., 2024 ). To the best of our knowledge, the relationship between long-term remyelination and early-stage EDSS has not yet been fully investigated. Additionally, considering the frequent relapses and remissions during the early stages of MS ( Confavreux and Vukusic, 2006 , Kappos et al., 2006 , Lublin et al., 2022 ), the distinct demographics of our cohort — characterized by relatively short disease duration and mild disability — may influence EDSS, particularly at the time of lesion detection. When we further analyzed the relationship between longitudinal changes in EDSS and lesional χ dia ( Supplementary Material s), similar to earlier studies ( Muller et al., 2024 ), the previously reported negative associations between EDSS changes and myelin signal changes were reproduced in our data ( Fig. S4 ), though this did not show statistical significance. Future studies with larger cohorts, focusing on long-term myelin changes from lesion onset, may help clarify our findings.

While this study revealed an association between early-stage HPS and remyelination capability, several limitations should be considered. First, the number of participants and lesions was limited due to the strict criteria for assessing early-stage lesional characteristics. Second, although histopathological validity of χ-separation had been explored previously ( Shin et al., 2021 ), this study does not include histological evaluation in MS lesions. Third, differences in MR contrasts may exist between the various MRI protocols and hardware, albeit χ-separation imaging, as a quantitative MRI method, is less dependent on acquisition protocols and hardware, as demonstrated by small inter-system variability of χ-separation results ( Fig. S5 ). Fourth, the early stage investigated in this work was defined within 1-year time window, which corresponds to the earliest detection of new lesions in the current follow-up frequency. Considering the rapid dynamics of iron pathology in MS lesions ( Hametner et al., 2013 , Tham et al., 2021 ), it is possible that in some lesions, the most acute phase of lesions was missed between follow-ups. A future study with frequent follow-up χ-separation imaging can help better understand early-stage dynamics of iron pathology in vivo . Last, we investigated the effects of HPS on qualitatively assessed myelin changes (i.e., longitudinal myelin outcomes). Although quantitative χ dia signal changes demonstrated an HPS dependency consistent with qualitative findings ( Fig. 3 ), further efforts to determine thresholds for quantitatively classifying longitudinal myelin changes can be an important future step.

In conclusion, this study successfully demonstrated longitudinal changes of χ-separation MRI signals with added specificity for iron and myelin from the early stage of MS lesions. Myelin signal measured by χ-separation, χ dia , successfully revealed distinct pathways of myelin dynamics in MS lesions, such as remyelination and demyelination, indicating a potential utility of χ-separation MRI to monitor the efficacy of remyelination therapy ( Reich et al., 2018 ). We also found that excessive iron deposition in early-stage lesions is associated with quantitative and qualitative impairment of future remyelination, suggesting that early HPS may serve as a potential marker of remyelination failure and clinical worsening. The absence of HPS might serve as a potential marker for MS lesions that are more amenable to myelin repair therapeutics intervention.

Declaration of Generative AI and AI-assisted technologies in the writing process

During the preparation of this work the authors used ChatGPT 4.0 (OpenAI) for grammar checks. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the published article.

CRediT authorship contribution statement

Hyeong-Geol Shin: Writing – review & editing, Writing – original draft, Visualization, Software, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Woojun Kim: Writing – review & editing, Supervision, Project administration, Methodology, Investigation. Jung Hwan Lee: Resources, Methodology, Investigation, Data curation. Hyun-soo Lee: Methodology, Data curation. Yoonho Nam: Supervision, Software, Investigation, Formal analysis, Data curation, Conceptualization. Jiwoong Kim: Investigation, Formal analysis. Xu Li: Writing – review & editing, Writing – original draft, Supervision, Formal analysis. Peter C.M. van Zijl: Writing – review & editing, Writing – original draft, Supervision, Conceptualization. Peter A. Calabresi: Writing – review & editing, Writing – original draft, Supervision, Investigation. Jongho Lee: Writing – review & editing, Writing – original draft, Supervision, Investigation. Jinhee Jang: Writing – review & editing, Writing – original draft, Visualization, Supervision, Methodology, Investigation, Formal analysis, Data curation, Conceptualization.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Hyun-Soo Lee is an employee of Siemens Healthineers. Under a license agreement between Philips and the Johns Hopkins University, Dr. van Zijl and Johns Hopkins University are entitled to fees related to an imaging device used in the study discussed in this publication. Dr. Knutsson as Dr. van Zijl’s spouse is therefore also conflicted. Dr. van Zijl also is a paid lecturer for Philips. This arrangement has been reviewed and approved by the Johns Hopkins University in accordance with its conflict-of-interest policies. Dr. Peter A. Calabresi has received consulting fees for serving on scientific advisory boards for Idorsia, Myelin Repair Foundation, Lilly, and Novartis, and is PI on a grant to the Johns Hopkins University from Genentech. Dr. Jongho Lee is an inventor of a χ-separation-related patent owned by Seoul National University. All remaining authors have declared no conflicts of interest.

Acknowledgement

This study was supported by the National Research Foundation of Korea funded by the Korean government (MSIT) (RS-2023-00208409 and RS-2024-00349509).

Footnotes

Supplementary data to this article can be found online at https://doi.org/10.1016/j.nicl.2025.103748 .

Supplementary data

The following are the Supplementary data to this article:

Supplementary Data 1

Data availability

Data will be made available on request.

References

Associated Data

Supplementary Materials

Supplementary Data 1

Data Availability Statement

Data will be made available on request.