DIR-visible grey matter lesions and atrophy in multiple sclerosis: partners in crime?

Background The extent and clinical relevance of grey matter (GM) pathology in multiple sclerosis (MS) are increasingly recognised. GM pathology may present as focal lesions, which can be visualised using double inversion recovery (DIR) MRI, or as diffuse pathology, which can manifest as atrophy. It is, however, unclear whether the diffuse atrophy centres on focal lesions. This study aimed to determine if GM lesions and GM atrophy colocalise, and to assess their independent relationship with motor and cognitive deficits in MS. Methods Eighty people with MS and 30 healthy controls underwent brain volumetric T1-weighted and DIR MRI at 3 T, and had a comprehensive neurological and cognitive assessment. Probability mapping of GM lesions marked on the DIR scans and voxel- based morphometry (assessing GM atrophy) were carried out. The associations of GM lesion load and GM volume with clinical scores were tested. Results DIR-visible GM lesions were most commonly found in the right cerebellum and most apparent in patients with primary progressive MS. Deep GM structures appeared largely free from lesions, but showed considerable atrophy, particularly in the thalamus, caudate, pallidum and putamen, and this was most apparent in secondary progressive patients with MS. Very little co-localisation of GM atrophy and lesions was seen, and this was generally confined to the cerebellum and postcentral gyrus. In both regions, GM lesions and volume independently correlated with physical disability and cognitive performance. Conclusions DIR-detectable GM lesions and GM atrophy do not significantly overlap in the brain but, when they do, they independently contribute to clinical disability.

SPMS group having a higher EDSS than PPMS group, and people with PPMS having higher EDSS scores than those with RRMS (both p<0.05). SPMS patients had worse executive function than PPMS, and PPMS worse than RRMS (p<0.05 and p<0.01, respectively).

GM atrophy
Subgroup analyses revealed that deep GM volume loss was present in all MS groups compared to controls, with additional temporal lobe volume loss in the PPMS group, and volume loss of GM in the occipital lobe, amygdala and hippocampus in the SPMS group (Supplemental table 1 and Figure 2a). Voxel-wise comparisons between patient groups yielded no significant results, though ROI analyses revealed significant subgroup differences in insular and thalamic volumes. SPMS patients had a significantly reduced insular volume compared to both healthy controls and PPMS patients (both p<0.01, Bonferroni corrected for multiple comparisons). Thalamic volume was also reduced in SPMS compared to both healthy controls and PPMS patients (p<0.001 and p<0.05, respectively).
Group wise whole brain analyses at p<0.001 (figure 2b) revealed that PPMS patients had a significantly higher lesion probability than controls in the left supplementary motor area (0.35cm 3 ), right medial frontal lobe (0.05cm 3 ), right postcentral gyrus (0.03cm 3 ), right supramarginal gyrus (0.09cm 3 ), and the right cerebellum (0.38cm 3 ). The RRMS subgroup did not show any significant lesional clusters compared to controls. The SPMS group showed consistent clustering of lesions in the supplementary motor area (0.05cm 3 ) compared to controls. All MS groups had significantly more lesions in every ROI than healthy volunteers (all p<0.001), and were comparable among MS groups for all ROIs, with the exception of the cerebellum, in which those with SPMS showed more lesions than those with RRMS (p<0.05, Bonferroni corrected).

Co-localisation of GM atrophy and lesions
When we examined the co-localisation between volume loss and lesion load across MS subtypes (supplemental table 2 and Figure 2c), the PPMS group showed spatial overlap between the two pathological abnormalities throughout the cerebellar and cerebral cortex.
While RRMS patients showed co-localisation of these forms of pathology mainly in the cerebellum, SPMS patients showed very little co-localisation, mainly in the post-central gyrus. The least co-localisation was found in the SPMS patients (0.01cm 3 ), whilst the RRMS patients (0.18cm 3 ) were in between the PPMS (0.30cm 3 ) and SPMS groups in terms of number of regions showing both forms of MR changes. None of the groups showed colocalisation in the deep GM (where significant atrophy was found in the absence of GM lesions). ROI-wise regression analyses in the RRMS group showed a significant association between cerebellar lesion load and volume loss (B=-33.46, p<0.001).

Supplemental table 2. Areas in patients showing a significant association between smaller GM volume and increased lesion load.
At p<0.01 uncorrected, co-localisation of both forms of pathology was largest in patients with PPMS, throughout virtually the entire brain with the exception of deep GM. Particularly the cerebellum in RRMS patients showed a great association between these forms of pathology. Overall, lesion-atrophy clusters were small when compared with the volume of the brain.