The influence of HLA-DRB1*15 on motor cortical pathology in multiple sclerosis.

AIM: Multiple sclerosis (MS) is a common and heterogeneous CNS inflammatory demyelinating disease. The HLA-DRB1 locus may influence clinical outcome. MS cortical pathology is frequent and correlates with measures of clinical disability, including motoric dysfunction that is a predominant feature of disease progression. The influence of HLA-DRB1*15 on motor cortical pathology is unknown. METHODS: A pathologically confirmed age- and sex-matched HLA-DRB1*15+ (n = 21) and HLA-DRB1*15- (n = 26) MS post-mortem cohort was used for detailed pathologic analyses. For each case, adjacent sections of motor cortex were stained for myelin and inflammation, to evaluate the extent and distribution of motor cortical pathology. A subset of MS cases (n = 42) had spinal cord (SC) pathologic outcome data available for comparison. RESULTS: Motor cortical demyelination was more pronounced in younger cases (r = -0.337, P < 0.05), with MS cases carrying the HLA-DRB1*15 allele driving this effect (r = -0.612, P < 0.01). HLA-DRB1*15+ MS cases had more severe motor cortical parenchymal (P < 0.05), perivascular (P < 0.05) and meningeal (P < 0.05) T-cell inflammation compared to HLA-DRB1*15- cases. HLA-DRB1*15 status significantly influenced the extent of motor cortical microglial burden in both NAGM (P < 0.0001) and lesions (P < 0.01) in MS cases. Relationships between the extent of motor cortical and SC pathology were limited, but when present were primarily driven by HLA-DRB1*15+ cases. CONCLUSION: HLA-DRB1*15 status has a significant association with the extent of inflammation in the MS motor cortex, the extent of demyelination in younger MS cases, and influences relationships between motor cortical and SC pathology.

Casting light on multiple sclerosis heterogeneity: the role of HLA-DRB1 on spinal cord pathology.

Clinical heterogeneity in multiple sclerosis is the rule. Evidence suggests that HLA-DRB1*15 may play a role in clinical outcome. Spinal cord pathology is common and contributes significantly to disability in the disease. The influence of HLA-DRB1*15 on multiple sclerosis spinal cord pathology is unknown. A post-mortem cohort of pathologically confirmed cases with multiple sclerosis (n = 108, 34 males) with fresh frozen material available for genetic analyses and fixed material for pathology was used. HLA-DRB1 alleles were genotyped to select a subset of age- and sex-matched HLA-DRB1*15-positive (n = 21) and negative (n = 26) cases for detailed pathological analyses. For each case, transverse sections from three spinal cord levels (cervical, thoracic and lumbar) were stained for myelin, axons and inflammation. The influence of HLA-DRB1*15 on pathological outcome measures was evaluated. Carriage of HLA-DRB1*15 significantly increased the extent of demyelination (global measure 15+: 23.7% versus 15-: 12.16%, P = 0.004), parenchymal (cervical, P < 0.01; thoracic, P < 0.05; lumbar, P < 0.01) and lesional inflammation (border, P = 0.001; periplaque white matter, P < 0.05) in the multiple sclerosis spinal cord. HLA-DRB1*15 influenced demyelination through controlling the extent of parenchymal inflammation. Meningeal inflammation correlated significantly with small fibre axonal loss in the lumbar spinal cord (r = -0.832, P = 0.003) only in HLA-DRB1*15-positive cases. HLA-DRB1*15 significantly influences pathology in the multiple sclerosis spinal cord. This study casts light on the role of HLA-DRB1*15 in disease outcome and highlights the powerful approach of using microscopic pathology to clarify the way in which genes and clinical phenotypes of neurological diseases are linked.

Deep Metric Representation Learning for Clinical Resting State fMRI.

With growing size of resting state fMRI datasets and advances in deep learning methods, there are ever increasing opportunities to leverage progress in deep learning to solve challenging tasks in neuroimaging. In this work, we build upon recent advances in deep metric learning, to learn embeddings of rs-fMRI data, which can then be potentially used for several downstream tasks. We propose an efficient training method for our model and compare our method with other widely used models. Our experimental results indicate that deep metric learning can be used as an additional refinement step to learn representations of fMRI data, that significantly improves performance on downstream modeling tasks.