Количественный T1 чувствителен к кортикальной ремиилизации при множестве склерозов: посмертное исследование МРТ.

INTRODUCTION

Multiple sclerosis (MS) is a chronic autoimmune and degenerative disease of the brain and spinal cord that affects almost 3 million people in the world [ 1 ]. MS is primarily characterized by sharply delimited areas of myelin loss, which are associated with varying degrees of axonal reduction and reactive gliosis [ 2 ], both in white matter and in the cortex [ 3 ].

The central nervous system has the capacity to spontaneously regenerate myelin in at least some of these focal lesions, in a process called remyelination (RM), where oligodendrocyte progenitor cells (OPC) [ 4 ] and probably also surviving oligodendrocytes produce new myelin sheaths covering the denuded axons [ 5 ]. Typical features of rebuilt myelin sheaths are their reduced thickness and shortened length of the new internodes [ 6 ].

Studies using light and electron microscopy in autoptic brain tissue of people with MS (pwMS) have highlighted that RM is present and even more extensive in cortical gray than in white matter [ 7 , 8 ]. However, detecting de‐ and remyelination in the cerebral cortex of living pwMS is challenging, especially due to the lack of contrast sensitivity of conventional MRI protocols, as well as the relatively small size of areas of focal myelin damage and regeneration [ 9 ]. Because of this, there are few investigations exploring imaging of RM in the cortex of pwMS. Lazzarotto et al. recently addressed this topic using a semiquantitative MRI measure—magnetization transfer ratio (MTR)—at 3T [ 10 ]. Besides, Barletta et al. combined qT1 and T2* at 7T MRI and showed that the derived “myelin” maps were sensitive to cortical microstructural abnormalities in early MS [ 11 ]. Nevertheless, pathological studies validating imaging biomarkers of cortical RM are scarce.

The identification and validation of such biomarkers is urgently needed. Indeed, assessing the presence of cortical repair processes in pwMS might be even more relevant than in the white matter, since the burden of cortical demyelination (DM) has been shown to correlate to a greater extent with cognitive and physical disability than white matter lesions (WML) [ 12 ]. In line, Lazzarotto et al. reported a reduced risk of disability progression in a subgroup of patients with detectable cortical RM [ 10 ].

Over the past years, several small molecules targeting OPC proliferation and differentiation into oligodendrocytes thus promoting RM have been identified and studied [ 13 ]. Some of them, such as clemastine, penetrate the blood–brain barrier and might act in the cortex, as demonstrated in animal studies [ 14 ]. Thus, fostering cortical RM should be considered a feasible, although challenging target, which can only be successfully pursued by developing reliable in vivo biomarkers. Quantitative MRI (qMRI) provides biophysical measures of tissue microstructural properties [ 15 ]. Accordingly, it might be a suitable tool to explore this target.

In this postmortem MRI‐histology study, we aimed at evaluating whether patterns of qMRI parameters are sensitive to RM in the cortex of brains of pwMS.

METHODS

This study was approved by the ethical review committee of the University Medical Center Göttingen. We obtained six whole brains from the German MS Brain Bank of the Competence Network Multiple Sclerosis (KKNMS). Three patients had relapsing–remitting MS (RRMS, ages 51, 66, and 70 years, EDSS, respectively, 4/2.5/4) and three patients had received a diagnosis of secondary progressive MS (SPMS, ages 58, 65, and 69 years, EDSS, respectively, 8/7.5/8). Clinical information for the included patients is depicted in Table  1 . The autopsies were performed within 24 h after death; subsequently, the brains were immediately immersed in 10% buffered formalin.

TABLE 1 Characteristics of patients studied. Patient Disease course Disease duration Last EDSS 1 SPMS 14 years 7.5 2 RRMS 8 years 4 3 SPMS 23 years 8 4 RRMS 17 years 2.5 5 SPMS 39 years 8 6 RRMS 36 years 4 Mean age at death 63 years (min. 51 years, max 70 years) Sex 4 females|2 males Abbreviations: EDSS, Expanded Disability Status Scale; RRMS, relapsing–remitting MS; SPMS, secondary progressive MS.

Characteristics of patients studied.

Abbreviations: EDSS, Expanded Disability Status Scale; RRMS, relapsing–remitting MS; SPMS, secondary progressive MS.

The fixed whole brains were imaged on a clinical whole‐body 3T MRI system (Magnetom Prisma, Siemens Healthineers, Erlangen, Germany) 3–12 months after initiation of the fixation process, to permit a uniform penetration of the fixative into the entire brain. About 3 weeks before the start of image acquisition, we positioned the brain into a 3D printed MRI‐compatible holder that was shaped to fit a 20‐channel head and neck coil and that was filled with Fomblin® perfluoropolyether (Solvay Specialty Polymers USA, LLC, West Deptford, NJ, USA), a liquid‐phase proton‐free fluorocarbon. As an additional step prior to imaging, we removed air bubbles from the holder using a vacuum pump, to avoid sources of undesired susceptibility artifacts. Since both postmortem tissue changes and chemical fixation influence the MRI properties of the brain, we adapted the MRI protocols to these ex vivo conditions [ 16 ]. For MRI, we employed the built‐in body coil for radiofrequency (RF) transmission; for RF reception, we used the standard 20‐channel phased‐array head and neck coil supplied by the manufacturer.

Ex vivo MRI protocol

We acquired the following MR sequences: (i) Magnetization Prepared Two Rapid Acquisition Gradient Echo (MP2RAGE, 0.67 mm isotropic, TR = 5 s, TE = 1.78 ms, TI1 = 194 ms, and TI2 = 2500 ms) to obtain quantitative T1 maps (qT1) [ 17 ]; (ii) Fast Acquisition with Spiral Trajectory and adiabatic T2prep (FAST‐T2, 1 mm isotropic, spiral TR/TE = 9.3 ms/0.94 ms, six T2prep times of [0, 7.5, 17.5, 67.5, 147.5, 307.5] ms) to assess myelin water fraction (MWF), using the cut‐off of 20 ms to separate the short myelin‐water associated T2 components, as proposed by Nguyen et al. [ 18 ]; (iii) segmented 3D‐Echo Planar Imaging (3D‐EPI, 0.33 mm isotropic, TR = 65 ms, TE = 35 ms, ETL = 13, bandwidth 394 Hz/pixel) to reconstruct quantitative susceptibility mapping (QSM) [ 19 ]; (iv) proton density weighted (TR = 25 ms, flip angle 5°) and MT prepared (TR = 25 ms, flip angle 5°) RF‐spoiled 3D‐GRE of identical geometry (0.57 mm isotropic) to allow MTR map calculation; MTR maps were calculated using the equation ( M
0  −  M
s / M
0 ; M
0  = signal intensity obtained during non‐saturated acquisition, M
s  = signal intensity obtained during saturated acquisition). (v) Multi‐shell, high‐angular‐resolution diffusion weighted imaging: Brain #1, #3, #4, #5 and #6: resolution 1.3 mm isotropic, b ‐value 0/1350/2650/4000 s/mm 2 , TE = 80.0 ms; δ = 22.3 ms; Δ = 36.3 ms; Brain #2: resolution 1.4 mm isotropic, b ‐value = 0/1400/2000/4000 s/mm 2 : TE = 93.0 ms; δ = 28.9 ms; Δ = 42.9 ms. The diffusion weighted images were denoised [ 20 ], and radial diffusivity (RD) maps were computed using a diffusion tensor model.

Brain cutting and sectioning

After MRI, the brains were cut coronally by using personalized 3D‐printed cutting boxes, a state‐of‐the‐art method previously described [ 21 ]. This approach allowed for excellent registration between the MRI slices and the tissue slabs for histology analysis. We used the 3D‐EPI as reference, since it has the highest resolution and provides an accurate depiction of anatomic and pathologic details. Block dissection was based on the identification of cortical lesions in MRI. The block size was approximately 4.0 × 2.8 × 0.5 cm 3 .

Immunohistochemistry

Dissected tissue blocks were fixed in 4% paraformaldehyde, embedded in paraffin, and cut into 4 μm sections, which were subsequently stained immunohistochemically using an avidin‐biotin technique. Myelination was assessed by immunostaining using antibodies against myelin basic protein (anti‐MBP; Dako, Glostrup, Denmark), myelin‐associated glycoprotein (anti‐MAG; Abcam, Cambridge, UK), and brain‐enriched myelin‐associated protein 1 (BCAS1; Santa‐Cruz, Dallas, TX, USA). Immunohistochemistry (IHC) for MHC II antigens (HLA‐DR/DP/DQ; CR3/43; Abcam) was performed to detect activated microglia, indicative of ongoing cortical immune activation. In double labeling experiments, myelin proteins were developed with 3, 3'‐diaminobenzidine (DAB) as the chromogen, and MHC II with Fast Blue, using alkaline phosphatase‐coupled secondary antibodies. IHC for neurofilament proteins (NF200, Sigma, St. Louis, Missouri, USA) was used to confirm the presence of axons in demyelinated lesions. After counterstaining with hematoxylin and coverslipping, the sections were converted to high‐quality digital images with a 20× objective magnification (Olympus/Evident Slideview VS200). The whole slide digital images were then uploaded to an open microscopy OMERO server (version 5.6.3) for distributed collaboration and morphometry. For fluorescent double IHC, Alexa 488‐ and Cy3‐coupled secondary antibodies (Cy3; Jackson, ImmunoResearch Laboratories, Bar Harbor, Maine, USA and Alexa 488; Dianova, Hamburg, Germany) were utilized.

Identification of cortical lesions

Tissue sections stained by MBP IHC were screened by two board‐certified neuropathologists (EB and CS), and the following regions of interest (ROI) were identified and manually segmented as polygons in OMERO: areas of cortical DM, characterized by complete loss of myelin in MBP IHC; regions of cortical RM, characterized by irregularly arranged MBP positive myelin sheaths which often stain pale and sometimes form patches in fully demyelinated cortical areas. The remyelinated areas frequently show a lower density of myelin sheaths and do not fully achieve the characteristic architecture of myelinated fibers in normal cortex (see Figure  1 ); regions of normal appearing cerebral cortex (NAC), in proximity to the DM and RM areas.

areas of cortical DM, characterized by complete loss of myelin in MBP IHC;

regions of cortical RM, characterized by irregularly arranged MBP positive myelin sheaths which often stain pale and sometimes form patches in fully demyelinated cortical areas. The remyelinated areas frequently show a lower density of myelin sheaths and do not fully achieve the characteristic architecture of myelinated fibers in normal cortex (see Figure  1 );

regions of normal appearing cerebral cortex (NAC), in proximity to the DM and RM areas.

MAG and BCAS1 IHC were used to confirm results obtained by MBP IHC. In addition, MAG and BCAS1 IHC were used to identify actively myelinating oligodendrocytes in cortical lesions. The remyelinated lesions chosen for this study did not harbor relevant numbers of MAG or BCAS1 positive oligodendrocytes, thus pointing toward later stage established RM, that is, cortical shadow plaques.

Histological characteristics of de‐ and remyelinated cortical lesions. Well delineated cortical remyelination in patient 5. (A)–(D) Immunohistochemistry for myelin proteins depicted at low magnification demonstrates an area of less intense myelin staining in the upper cortical layers at the depth of a sulcus (A: MBP (brown)/MHC II (blue); B: BCAS1 (brown), C: MAG (brown), D: MBP, green/NF, red). The remyelinated cortical area is segmented by a yellow line. Panels (E)–(H) demonstrate the distinct boundary between the thinly and irregularly myelinated upper cortical layers and the deeper cortical layers with well preserved and regularly structured myelin. (I)–(L): High magnification images of regularly myelinated normal‐appearing cortex (I: MBP; J: BCAS1; K: MAG, L: MBP, green/NF, red) and (M)–(P) remyelinated cortex (M: MBP; N: BCAS1; O: MAG, P: MBP, green/NF, red). Magnified cortical regions are indicated in (A)–(D) and (E)–(H). No evidence of microglia activation, monocyte invasion, or myelin phagocytosis (A, E, I, M: MHC II in blue). No process‐bearing oligodendrocytes as evidence for ongoing myelin formation are observed in BCAS1 (J, N) and MAG IHC (K, O). Panels (A2)–(D2) depict a representative example of a demyelinated cortical lesion. The demyelinated cortical area at the depth of a sulcus is shown at low magnification with MBP (brown)/MHC II (blue) double IHC (A2). The demyelinated area is segmented by a yellow line. Panels (B2), (C2), and (D2) show higher magnification images of demyelinated (C2) and normal‐appearing cortex (D2). Magnified cortical regions are indicated in (A2) and (B2). (Scale A–H and A2, B2 1000 μm; scale I–P and C2, D2 100 μm; stainings: BCAS1, brain‐enriched myelin‐associated protein 1; MAG, myelin‐associated glycoprotein, MBP, myelin basic protein; NF, neurofilament; MHC II (CR3/43) for activated microglia. Only sparse microglia are present, as shown by the black arrows (M and D2)).

To measure myelin content, the whole slide digital images and the segmentation polygons were imported into QuPath software (0.4.2 version), an open source software for digital pathology image analysis. The estimate of myelin content in MBP IHC in the ROIs mentioned above was calculated using color deconvolution of the DAB signal and quantified as optical density (OD) units.

MRI analysis

The areas corresponding to demyelinated, remyelinated, and normal‐appearing cortex were manually identified and segmented on 3D‐EPI images utilizing ITK‐SNAP 3.6.037 [ 22 ]. The use of the cutting box enabled a precise alignment of the brain slabs to the MR images. For ROI segmentation on MRI, we put the digital histopathology whole slide scans next to the MR images. We then manually segmented the ROI in the same form and position as the segmented polygons on the digital histology slide. In most cases, cortical DM and RM areas identified and segmented by the pathologists on digital slides presented a comparable and well‐delimited contour on the reference high‐resolution sequence 3D‐EPI, facilitating the segmentation process. The segmentations were obtained ensuring that the adjacent normal‐appearing cortex (NAC) had a similar anatomical location to the corresponding DM/RM areas, in order to avoid anatomical bias due to varying MRI properties of different cortical layers. The segmentation process was performed by two experienced neurologists (RG and AC), followed by consensus review. The qMRI images (qT1, MWF, MTR, QSM, and RD) were automatically registered to the slab‐matched 3D‐EPI images using Elastix [ 23 ]. Based on the segmentation mask created in the 3D‐EPI space, average qMRI intensity values were calculated for each ROI using an in‐house Python script. In Figure  2 , we provide an example of CL and NAC segmentation.

Remyelinated cortical lesion visualized by MRI 3D‐EPI. (1). MBP IHC of a tissue block of patient 6 revealed a remyelinated cortical lesion (RM) characterized by a reduced density of thinly myelinated fibers (A, B). Based on histopathological findings, the lesion and an adjacent area of normal‐appearing cortex were identified and segmented on 3D‐EPI (RM lesion in green and the corresponding NAC in blue, (2). The remyelinated cortical area is outlined by a yellow line (A, B), while the region of normal‐appearing cortex (NAC) is marked by a black square (A, C). A slightly blurred border separates the well‐preserved myelin from the remyelinated cortical lesion (D, corresponding white squares in A, B). High magnification images of the remyelinated area (F; corresponding small black squares in B, D) and the regularly myelinated cortex (G, corresponding small white squares in C, E). No relevant microglia activation was detected. (Scale bars: A–C 1000 μm; D–G 100 μm; staining: MBP, myelin basic protein (brown), microglia staining: MHC II (CR3/43), blue.)

Statistical analysis

The main objective of this study was to compare multiple myelin‐sensitive qMRI measures in DM, RM, and NAC, and to identify the sequence(s) most suited to detect cortical RM.

We first compared mean qMRI measures in DM and RM cortical lesions with those of the adjacent NAC by Wilcoxon matched pairs tests. We then compared the qMRI metrics of RM and DM directly, using a Mann–Whitney test. To account for potentially confounding factors such as fixation and anatomical differences, we normalized the RM and DM values by subtracting the value of the corresponding NAC (∆DM = |DM – NAC|; ∆RM = |RM – NAC|). We then used mixed‐effect models (qMRI measure = dependent variable, ROI type = independent variable) including block nested in patients as a random effect accounting for clustering within data to confirm the results obtained using univariate analyses. The results of the mixed models are presented as marginal effects.

Last, we performed Spearman's rank‐order correlation between the data obtained from quantitative histology (quantification of myelin) and the corresponding qMRI metrics in all ROIs.

We performed Bonferroni correction for multiple comparisons for each test. Statistical analysis was performed using GraphPad Prism 9.5.1 and R 4.3.1.

RESULTS

From 6 postmortem MS brains, we obtained a total of 129 tissue blocks containing cerebral cortex for histopathological analysis.

Evaluation of IHC for MBP permitted the identification of n  = 25 areas of focal cortical DM (22 subpial und 3 leukocortical CL), n  = 25 areas of focal cortical RM (23 subpial, 1 leukocortical and 1 intracortical lesions), and n  = 45 areas of adjacent normal appearing cerebral cortex (NAC) in 43 tissue blocks. A total of 5 DM and RM areas shared the same NAC. From brain n. 3 we could not consider any lesions due to the widespread cortical DM and lack of sufficiently large NAC. The median surface size of demyelinated cortical lesion areas segmented on 3D‐EPI (0.33 mm isotropic) was 5.12 mm 2 (minimum 2.29 mm 2 , maximum 20.15 mm 2 , range 17.86 mm 2 ), and the median surface size of remyelinated cortical lesion areas was 8.39 mm 2 (minimum 3.05 mm 2 , maximum 22.98 mm 2 , and range 19.93 mm 2 ). The Mann–Whitney test did not reveal a significant difference between the average size of cortical DM and RM lesions ( p  = 0.102).

MWF and MTR values were not different in DM and RM compared to the NAC (DM vs. NAC: mean MWF 8.97% ± 2.79% vs. 8.34% ± 2.6%; mean MTR 0.17 ± 0.03 vs. 0.16 ± 0.03; RM vs. NAC: mean MWF 8.72% ± 1.94% vs. 6.98% ± 2.86%; mean MTR 0.15 ± 0.02 vs. 0.15 ± 0.03). Conversely, in comparison to the adjacent NAC, mean qT1 and RD were higher and mean QSM was lower in DM areas (DM vs. NAC: qT1 372.5 ± 87.13 ms vs. 316.3 ± 85.59 ms, p  < 0.001, RD 25.2e‐5 ± 4.948e‐5 mm 2 /s vs. 21.41e‐5 ± 5.41e‐5 mm 2 /s, p  = 0.001, QSM −8.22 ± 14.94 ppb vs. 7.49 ± 11.66 ppb, p  < 0.001). Similarly, QSM was also lower and RD higher in RM areas (RM vs. NAC: QSM 0.94 ± 9.22 ppb vs. 12.59 ± 14.53 ppb, p  < 0.01, RD 28.16e‐5 ± 6.4e‐5 mm 2 /s vs. 26.25e‐5 ± 5.33e‐5 mm 2 /s, p  = 0.048). RM qT1 did not significantly differ from NAC (qT1 323 ± 42.45 ms vs. 300.1 ± 52.64 ms, p  = 0.074) (Figure  3 ). The qMRI measures in cortical DM and RM areas compared to the corresponding NAC are summarized in Table  2 .

Quantitative MRI metrics in cortical de‐ and remyelinated lesions. On the left, violin plots show the raw data of the five quantitative MRI metrics in remyelinated cortical lesions (RM), demyelinated cortical lesions (DM) and in the matched adjacent normal‐appearing cortex (NAC). In the middle, line plots connect cortical lesions with the corresponding NAC, colored by patient. On the right, the point estimate with confidence interval of the Wilcoxon rank signed test shows a significant increase of T1 in DM, a significant reduction of QSM in both RM and DM, and a significant increase of RD in RM and DM.

TABLE 2 Quantitative MRI. Quantitative MRI (mean ± SD) Region of interest DM NAC (DM) RM NAC (RM) MTR (ratio) 0.17 ± 0.03 0.16 ± 0.03 0.15 ± 0.02 0.15 ± 0.03 MWF (%) 8.97 ± 2.79 8.34 ± 2.6 8.72 ± 1.94 6.98 ± 2.86 qT1 (ms) 372.5 ± 87.13 316.3 ± 85.59 323 ± 42.45 300.1 ± 52.64 QSM (ppb) −8.22 ± 14.94 7.49 ± 11.66 0.94 ± 9.22 12.59 ± 14.53 RD (mm 2 /s) 25.2e‐5 ± 4.948e‐5 21.41e‐5 ± 5.41e‐5 28.16e‐5 ± 6.4e‐5 26.25e‐5 ± 5.33e‐5

Quantitative MRI.

RM areas thus showed values of qT1 comparable to those of NAC, but this was not the case for QSM and RD (∆qT1 DM > ∆qT1 RM, p  = 0.045; ∆ QSM DM vs. ∆ QSM RM, p  = 0.43; RD DM vs. ∆ RD RM, p  = 0.145). This analysis was conducted only for the qMRI metrics showing a significant change in demyelinated and remyelinated cortical lesion areas compared to the NAC (Figure  4 ).

Extent of change of demyelinated and remyelinated cortical lesions in relation to normal‐appearing cortex in qT1, QSM, and RD. Graphs present the direct comparison between the extent of change in demyelinated cortical areas (DM) and remyelinated cortical areas (RM) with respect to the normal‐appearing adjacent cortex (NAC) of the MRI parameters sensitive to cortical damage in our experimental setting. The change is expressed as delta (∆RM = value RM – value NAC, ∆DM = value DM – value NAC).

The linear mixed models confirmed the sensitivity of qT1 to RM (estimated difference between RM and DM: 37.09 ms, p  = 0.04). QSM and RD did not show any significant difference between RM and DM (estimated between RM and DM: 7.6 ppb and 0.91e‐5 mm 2 /s, p  = 0.19 and p  = 0.57, respectively) (Figure  5 ).

Mixed effect models comparison of qT1, QSM, and RD in demyelinated lesions, remyelinated lesions, and normal appearing cortex. Results of linear mixed effect models taking clustering within patient and tissue blocks into account. In comparison to NAC, qT1 is significantly longer only in DM, not in RM. As regards QSM and RD, both parameters differ in DM and RM from NAC.

Concerning the correlation of MRI measures with quantitative histology, our analysis showed that only QSM and qT1 correlated with cortical myelin content in lesions and NAC (QSM rho = 0.29, p  = 0.02, qT1 rho = −0.28, p  = 0.03) (Figure  6 ).

Correlation of qMRI metrics and myelin content determined by immunohistochemistry. Spearman correlation between MRI metrics and mean content of myelin basic protein (MBP) immunohistochemistry as expression of the myelin content. Only qT1 and QSM show a correlation with myelin content.

In Figure  7 , we provide exemplary visual examples of the behavior of the different MR parameters in prototypical de‐ and remyelinated cortical lesions.

Visual correlation of de‐ and remyelinated cortical lesions with qMRI metrics. Representative examples of a prototypical remyelinated cortical lesion (RM, left, patient 5) and a prototypical demyelinated cortical lesion (DM, right, patient 6) and their corresponding quantitative MRI (MTR, magnetization transfer ratio; MWF, myelin water fraction; qT1, quantitative T1; QSM, quantitative susceptibility mapping; RD, radial diffusivity). MBP immunohistochemistry (brown); scale bars 1000 μm.

DISCUSSION

In the present postmortem MRI study with immunohistochemical correlation, we examined whether we could detect cortical remyelinated lesions using quantitative MRI. We found that in our experimental setting, qT1 was more sensitive to RM in the cerebral cortex of pwMS than other qMRI metrics such as QSM, RD, MTR, and MWF. Also, our data indicate that QSM and RD are the most sensitive qMRI metrics to detect demyelinated lesions in the cortex of pwMS.

Previous histological studies showed that demyelinated axons in the cortex of pwMS can be remyelinated [ 7 , 8 , 24 ]. In addition, it is well known that damage to the cerebral cortex has a more significant impact on clinical outcomes than WM pathology [ 25 ]. Consequently, cortical RM has been proposed as a potential outcome measure for clinical trials testing remyelinating/neuroprotective treatments [ 26 ]. However, while several imaging studies have focused on the RM of WML, there are few MRI investigations of RM in the cerebral cortex of pwMS. In this postmortem study, we address this research gap by exploiting the potential of qMRI in detecting areas of myelin loss and RM in the cerebral cortex of pwMS. We selected five qMRI parameters previously shown to correlate with myelin content in the central nervous system in pwMS [ 27 ] and assessed whether they were sensitive to cortical RM.

In the present study, MTR and MWF did not appear to be affected by cortical myelin loss, while they proved to be very sensitive to WM DM in the same experimental setting [ 27 ]. This may have various reasons: first, the cortex has a lower myelin content than the white matter [ 28 ] and, consequently, the generated signal is also weaker. Second, postmortem conditions, including the related tissue dehydration and fixation, impact the MR signal generation in gray matter and white matter differently [ 29 , 30 ]. Last, partial volume effects favored by the smaller size of cortical lesions might also have played a role. Our findings on MTR seem in contrast to previous postmortem studies investigating MTR in the cortex of pwMS, where sensitivity to myelin content was demonstrated [ 31 , 32 , 33 ]. However, two of these studies used high‐field MRI (7T and 9.4T) and the one utilizing 3T MRI showed only marginal MTR decrease in demyelinated cortical areas (~ 5%). Moreover, the fixation time of the specimens that were scanned in two of these studies was significantly longer than in our work (504 and 746 days). Another study comparing MTR, T1, and T2 relaxation times in the cortex of postmortem fixed brains found significantly altered T1 and T2 but not MTR in CL compared to the NAC [ 34 ]. In the latter study, the fixation time was limited to only a few weeks. Therefore, it cannot be ruled out that fixation time influences the differences in MTR between CL and NAC. Furthermore, previous studies showed that iron depletion was associated with increased MTR, especially in the cortex. While this effect could be hidden in the white matter, where the myelin content is higher, it might be more pronounced in the cortex, potentially limiting the MTR decrease associated with DM [ 30 ]. In this investigation, qT1, QSM, and RD proved to be sensitive to focal cortical alterations: in fact, we found a significant increase in qT1 and RD and a significant reduction in QSM in demyelinated cortical lesions compared to the adjacent NAC. These results are in line with the existing literature showing qT1 and QSM sensitivity to cortical myelin [ 35 , 36 ]. Moreover, histology showed a reduced iron density in cortical lesions compared to the adjacent normal appearing gray matter, likely related to a depletion of ferritin‐rich oligodendrocytes and myelin sheaths, which can additionally affect QSM [ 37 ]. Regarding diffusivity, evidence of anisotropy oriented perpendicular to the pial surface was found in the cerebral cortex of fixed postmortem brains [ 38 ]. Previous DTI studies of the cortex of pwMS in vivo and ex vivo showed an increase in fractional anisotropy (FA) in cortical lesions [ 39 ], but none so far investigated RD, as we did in the present study.

As expected, both qT1 and QSM correlated with quantitative immunohistological measures of myelin content in our dataset. The degree of correlation was definitively lower than that found in the white matter, probably reflecting the generally lower myelin content in the cortex as well as partial volume effects. Surprisingly, we did not find a correlation between RD and myelin content, although RD was increased in areas of focal cortical DM. The most likely explanation is the lower spatial resolution of RD maps compared to other maps. On the other hand, it is also possible that axonal loss plays a greater role in influencing diffusion metrics in the cortex [ 39 ]. Alternatively, it is also plausible that other microstructural alterations occurring in CL, such as the distribution of neurons and dendrites, can have an influence [ 40 ].

We then estimated the influence of RM on MRI by comparing the difference of microstructural alterations in cortical DM and RM areas as measured by multiparametric qMRI: while qT1 in remyelinated areas was comparable to that of NAC, QSM was significantly decreased and RD increased in RM. These data suggest that qT1 is potentially more sensitive to the myelin content in the cortex of pwMS than the other metrics. In addition, or alternatively, QSM and RD might to a larger extent be related to altered properties of the newly built myelin sheaths or to microstructural alterations other than DM.

T1 mapping with MP2RAGE already proved to differentiate remyelinated from demyelinated WML in pwMS [ 41 ]. The unaltered qT1 of RM and the correlation between qT1 and MBP staining intensity are in line with previous studies showing a solid relationship between qT1 and the cortical myeloarchitecture of healthy brains [ 42 ]. Less clear is why QSM in RM is significantly lower than in the adjacent NAC, although we detected a correlation between QSM and myelin content. Since oligodendrocytes have the highest iron concentration of all brain cells [ 43 ], this finding might suggest fewer oligodendrocytes in RM or a lower ferritin content in the newly generated myelin sheaths that are also thinner and present at lower density. In fact, previous studies have shown that the amount of iron in oligodendrocytes is increased during active RM; in contrast, it is equal to or lower than that of the adjacent NAC when this repair process is completed [ 44 ]. In the present study, we focused on remyelinated cortical lesions with only negligible ongoing RM, as evidenced by only scarce BCAS1‐positive actively myelinating oligodendrocytes.

Regarding RD, the increase observed in cortical RM areas might be due to the influence of processes independent of RM, such as axonal loss and/or remodeling in neuronal dendritic arborization, as pointed out above.

This study presents some limitations. First, to identify and to define the boundaries of cortical RM on histology is in part challenging, since remyelinated areas do not always have clearly defined borders and frequently show a patchy pattern. Also, cortical remyelinated areas may show a variable density of remyelinated fibers, thus lending inherent variability to MR detection. To reduce the impact of RM stage on MR parameters, we excluded actively remyelinating cortical lesions by BCAS1 IHC. Also, none of the DM or RM lesions studied here contained abundant MHC II positive microglia. Another important aspect is the fixation procedure, which may alter the physical properties of the tissue and have divergent effects on different sequences. Nonetheless, previous studies showed that formaldehyde fixation only mildly affects MTR [ 45 ] and MWF [ 46 ]. Besides, the fixation‐related T1 reduction was shown to be more pronounced in GM [ 47 ]. Based on these previous observations showing that a reduction of cortical T1 signal in fixed tissue in all likelihood exceeds that of the other parameters, we do not consider the observed results to be significantly driven by the fixation process. Concerning the absence of correlation between MWF and MTR with myelin content in the present study, it is possible that these sequences are more prone to fixation effects in the cortex than in the white matter, and that the partial volume effects in the small cortical lesions may attenuate the results. Furthermore, the quantitative maps that we investigated had different spatial resolutions and therefore were influenced by partial volume effects to different degrees, an aspect that could have definitely affected their sensitivity to RM. Myelin content varies according to cortical layer and cortical region [ 48 ]. Accordingly, we chose an adjacent area of NAC for each lesion and segmented it in the same cortical layer as the corresponding lesion. This normalization process should have strongly reduced the anatomical bias. Last, the relatively low number of postmortem brains studied here does not allow adjusting for covariates (such as MS type and age), although these factors might influence our endpoints.

In summary, our data suggest that—among the qMRI measures evaluated in this study—qT1 is most sensitive to RM in the cortex. On the other hand, QSM and RD were affected by any cortical abnormalities. Accordingly, merging the information provided by qT1 with that provided by QSM and RD might help to disentangle mechanisms of damage and regeneration in the cerebral cortex of pwMS. Prospective in vivo studies are needed to further explore the potential of qMRI for the detection of cortical RM. In particular, our study suggests that qT1 should be included in the longitudinal assessment of regional and global microstructural characteristics of the cortex in pwMS.

AUTHOR CONTRIBUTIONS

RG, CS and CG contributed to the study concept and design. MW fine‐tuned the post‐mortem MRI protocols. EB and CS performed histological evaluation. RG, MW, AC, PL, MB, LM, PD collected and analyzed MRI data. SS, RG and CG contributed to statistical analysis. RG and CG wrote the first draft. MW, EB, SS, AC, PL, MB, LM, JF, PD, GN, LK, WB, CS and CG critically reviewed the manuscript.

FUNDING INFORMATION

This work was funded by the Swiss National Science Fund PP00P3_176984 and PP00P3_206151 and supported by the German Ministry of Education (BMBF; KKNMS German competence network for multiple sclerosis). This project was also supported by the Deutsche Forschungsgemeinschaft (DFG) transregional collaborative research center (CRC) TRR 274/1 and 274/2 “Checkpoints of CNS recovery,” Project ID 408885537 B01, STA 1389/5‐1, the DFG under Germany's Excellence Strategy (EXC 2067/1‐390729940), the Gemeinnützige Hertie Foundation, the Deutsche Multiple Sklerose Gesellschaft (DMSG), and the National MS Society (USA), to CS. JF was supported by the clinician scientist programs of the TRR 274/1 and the UMG “Cellular dynamics in pathogenesis and therapy. GN was funded by the Intramural Research Program at NINDS, NIH.”

CONFLICT OF INTEREST STATEMENT

The authors do not report any conflicts of interest relevant to the manuscript.

ETHICS STATEMENT

This study was approved by the ethical review committee of the University Medical Center Göttingen.