The clinical course had been secondary progressive in ten subjects and primary progressive in two. In four subjects, the clinical course could not be determined (). The estimated mean EDSS was 8 (SD: 1; median: 8.5; range: 6.5–9). The mean brain weight was 1131 g (SD: 113 g, range: 1000–1310 g). The pH of the CSF was 6.9 (SD: 0.4, range: 6.5–7.5). The mean temperature of the tissue during MRI was 22.3 °C (SD: 2.6 °C, range: 17.5–25.1 °C).
Overview of patients and lesion characteristics in post mortem brain slices of 16 patients with multiple sclerosis (38 white matter lesions, WMLs)
Seventy-five regions of high signal (RHS) thought to be MS WMLs were detected on T2w MRI. Twenty-five/75 RHS had to be discarded from further analysis due to (i) poor in-plane co-registration between MRI and specimen (6), (ii) poor identification of different tissue components on histology (5), (iii) misidentification of a vascular lesion (1) or NAWM (10) as WMLs, and (iv) other technical problems (3).
Of the remaining 50 WMLs, a further 12 had to be discarded because no clear correspondence could be determined between the T2w MRI and b = 0 map (). Hence, 38 histologically confirmed WML, visible on T2w MRI as well as b = 0 images were studied. Between one and six WMLs in each brain slice (mean 2.4 WMLs; SD: 1.4 WMLs) could be used for the analysis. Quantitative MR values were also obtained in two to three regions of NAWM in each slice and then averaged.
Fig. 2 T2-weighted (T2w) MRI, b = 0, mean diffusivity (MD) and fractional anisotropy (FA) maps of post mortem multiple sclerosis brain. On T2w MRI of a coronal brain slice, seven exemplary regions of interest (ROI; marked in orange) were identified (more ...)
Twenty-eight/38 WMLs (73.7%) were demyelinated, three were fully remyelinated (7.9%) and seven partially remyelinated (18.4%) (). Two/38 (2.6%) WMLs were classified as early active, nine (23.7%) as CA, and 27 (71.1%) as CI WMLs. All (fully or partially) rWMLs were CI WMLs.
Twenty-two/38 WMLs (57.9%) were hypo-intense and 16/38 (42.1%) iso-intense on T1w MRI. Fully rWMLs (shadow plaques) represented two/22 (9.1%) T1 hypo-intense WMLs and 1/16 (6.25%) T1 iso-intense WMLs. Partially remyelinated WMLs contributed three/22 (13.6%) T1 hypo-intense WMLs and four/16 (25%) T1 iso-intense WMLs.
Fibrillary gliosis was classified in 35/38 WMLs and observed as being moderate in five/35 (14.3%), and severe in 30/35 (85.7%) WMLs.
Comparison of MS lesions and NAWM
MD, FA, myelin content (Trmyelin), axonal count and gliosis (Trgliosis) all differed significantly between WMLs and NAWM. Results were similar when the analysis was restricted to (partially or fully) rWMLs versus NAWM: rWMLs displayed significantly higher MD, more pronounced gliosis and lower FA, myelin content and axonal count ().
Table 2 Comparison (means, standard deviations, p values) of (i) white matter lesions (WMLs) versus normal-appearing white matter (NAWM), (ii) remyelinated WMLs (rWMLs) versus NAWM and (iii) demyelinated WMLs (dWMLs) versus rWMLs in post mortem brain of patients (more ...)
Correlation of MR diffusion and pathology
Myelin content (Trmyelin) correlated with FA (r = − 0.79, p < 0.001) and MD (r = 0.68, p < 0.001). A similarly strong correlation was detected between axonal count and both FA (r = 0.70, p < 0.001) and MD (r = − 0.66, p < 0.001). A weaker correlation emerged between both MD and gliosis (r = − 0.55, p = 0.002) and FA and gliosis(r = 0.50, p = 0.004) (, ).
Correlationsa between indices assessed in post mortem brain of patients with multiple sclerosis
Fig. 3 Correlation between diffusion indices and quantitative histology in post mortem multiple sclerosis brain (white matter lesions, normal-appearing white matter). The plots illustrate the association of fractional anisotropy (FA) and mean diffusivity (MD) (more ...)
Correlation between neuropathological features
Myelin content (Trmyelin) was strongly associated with axonal count (r = − 0.81, p < 0.001) (, ). Both Trmyelin and axonal count were moderately associated with Trgliosis. More severe gliosis correlated with lower myelin content and a smaller axonal count ().
Correlation between transmittance (Tr) of sections stained for Luxol fast blue (Trmyelin, inversely proportional to myelin content) and axonal count in post mortem multiple sclerosis brain (white matter lesions, normal-appearing white matter).
Regression analysis of confounding correlations
Predicting myelin content
The univariate correlation of Trmyelin with FA (− 0.79, p < 0.001) is reduced to partial correlation − 0.65, p < 0.001 after adjustment for MD. The univariate correlation between Trmyelin and MD (r = 0.68, p < 0.001) is reduced to partial correlation 0.43, p = 0.019 after adjusting for FA.
These results suggest that FA and MD are independently associated with Trmyelin. The correlations remain robust when adjusting for section thickness of the LFB-stained slides.
Predicting axonal count
The univariate correlation between axonal count and FA (0.70, p = 0.001) is substantially reduced and no longer significant when adjusted for Trmyelin (partial r = 0.16; p = 0.399). Albeit reduced, the association between axonal count and FA remains significant after adjustment for Trgliosis (partial r = 0.57, p = 0.001).
The univariate correlation between axonal count and MD is substantially reduced and no longer significant after adjustment for Trmyelin (partial r = − 0.24; p = 0.212). The association between axonal count and MD remains significant after adjustment for Trgliosis (partial r = 0.51, p = 0.005).
Together, these results suggest that the correlations of axonal count with both FA and MD are not independent of the (strong) correlation between axonal count and Trmyelin, but unrelated to the severity of gliosis. In a post hoc analysis box-plots were produced to further illustrate the relationships between FA, myelin content and axonal count ().
Fig. 5 Box-plots of fractional anisotropy (FA) in regions of interest (ROIs) with (A) low myelin content (i.e., high transmittance of sections stained for Luxol fast blue, Trmyelin) and (B) high axonal count. ROIs with low myelin content and high axonal count (more ...) Predicting gliosis
We were unable to assess whether the univariate association between Trgliosis and FA (r = 0.50, p < 0.01) is independent of Trmyelin, as neither association is significant in a model with both. Nor has it been possible to assess whether the association between Trgliosis and MD (r = − 0.55, p < 0.01) is independent of Trmyelin.
Comparison of demyelinated and remyelinated lesions
dWMLs had ~ 60% lower axonal count (p = 0.029) and a trend for higher MD (p = 0.061) and lower FA (p = 0.167) than rWMLs ().
Other histopathological lesions subgroups
No significant differences in MD or FA were seen between CA and CI WMLs or between T1 hypo-intense and iso-intense WMLs. There were too few EA WMLs to investigate this lesion category.
None of the described relationships was substantially affected by estimated EDSS, age, disease duration, time between death and tissue retrieval, or time between death and MRI.
Batch and section thickness
The mean thickness of the histological sections was 9.4 μm (SD: 3.6 μm), and 4.8 μm (SD: 0.7 μm) for LFB- and GFAP-stained samples, respectively. Staining of the samples for LFB and GFAP was performed in nine and six batches, respectively. Trmyelin in WMLs did not vary with batch, but significant batch variation was seen for Trmyelin in NAWM (p < 0.001), and borderline significance variation for Trgliosis in WMLs and NAWM. None of the observed pathology–MR correlations were affected by including section thickness and batch in the regression analysis (correlation coefficients were altered by no more than ± 0.01, ± 0.05 and significant p values by not more than ± 0.001, ± 0.002 when adjusting for batch and thickness, respectively).