By: Jarne Beliën, Stijn Swinnen, Robbe D'hondt, Laia Verdú de Juan, Nina Dedoncker Patrick Matthys, Jan Bauer, Celine Vens, Sinéad Moylett, Bénédicte Dubois
![](https://static.wixstatic.com/media/5446c2_3af711544c3f45188777c3d53a32fd5b~mv2.png/v1/fill/w_980,h_736,al_c,q_90,usm_0.66_1.00_0.01,enc_auto/5446c2_3af711544c3f45188777c3d53a32fd5b~mv2.png)
Multiple sclerosis (MS) is a chronic autoimmune disorder of the central nervous system (CNS) characterized by considerable interindividual heterogeneity in terms of disease onset and severity. Despite continuous scientific advancement in the field, accurately predicting long-term outcomes and effectively stratifying treatment regimens based on uncertain prognosis remain major challenges in MS care. Clinicians thus face an unmet need for disease activity measures, i.e. biomarkers, with proven prognostic value.
In this paper, we employed an interdisciplinary approach to investigate five potential biomarkers – CHIT1, CHI3L1, sTREM2, GPNMB and CCL18 – in the cerebrospinal fluid (CSF) at diagnostic lumbar puncture in a longitudinal cohort of 192 MS patients. In the first part, we used mixed-effects models to identify CHIT1 as the strongest predictor of faster disability progression up to 15 years after diagnosis (Fig. 1).
![](https://static.wixstatic.com/media/5446c2_03e26ef620ca43af827c9e97f496b51a~mv2.png/v1/fill/w_980,h_314,al_c,q_85,usm_0.66_1.00_0.01,enc_auto/5446c2_03e26ef620ca43af827c9e97f496b51a~mv2.png)
A) 3D scatter plot with regression plane of EDSS and CHIT1 concentration (log, pg/ml) at diagnosis over years from diagnosis to the most recent EDSS assessment. (B) Line graph shows 3 predicted EDSS trajectories for 3 CHIT1 concentrations (log, pg/ml) at diagnosis: the mean CHIT1 value in our cohort (3.63), the mean + 1 SD (4.15) and the mean - 1 SD (3.11), based on model 3. Shade indicates the 95% confidence intervals. SD, standard deviation.
This finding was then corroborated using a machine learning approach. In short, we used a regressor chain model to analyze the predictive power of several MS patient characteristics for disability progression, in particular the CSF biomarkers under study. Indeed, after known clinical predictors, we found CSF CHIT1 concentrations at diagnosis to emerge as the most important predictor. Our machine learning model thus confirmed the prognostic superiority of CHIT1 over the other CSF biomarkers.
The computational resources of the VSC enabled us to leverage the vast collection of publicly available single-cell RNA sequencing datasets. Using these publicly available datasets is a time- and resources-effective way to increase the power of analyses and validate the robustness of ones findings. However, without sufficient computing power provided by supercomputing centers such as the VSC, this wealth of information would remain uncharted.
As CHIT1 thus stood out as the strongest predictor of faster disability progression, we performed an integrative analysis of both in-house as well as publicly available CSF and CNS single-cell RNA sequencing datasets (Fig. 2).
![](https://static.wixstatic.com/media/5446c2_d8c22079126b489194e9a26e5fbcc714~mv2.png/v1/fill/w_980,h_370,al_c,q_90,usm_0.66_1.00_0.01,enc_auto/5446c2_d8c22079126b489194e9a26e5fbcc714~mv2.png)
We found CHIT1 to be predominantly expressed by a specific microglia subset located in active MS lesions and enriched for lipid metabolism pathways, which suggests a role in the de- and/or remyelination processes in MS. Furthermore, we found CHIT1 expression to accompany the transition from a homeostatic towards a more activated, disease-associated cell state in these microglia. Neuropathological evaluation in post-mortem tissue from MS patients confirmed CHIT1 production by lipid-laden phagocytes in actively demyelinating lesions, already in early disease stages (Fig. 3).
![](https://static.wixstatic.com/media/5446c2_2cf8ca59a8e04c17870088ff52a04af1~mv2.jpg/v1/fill/w_980,h_811,al_c,q_85,usm_0.66_1.00_0.01,enc_auto/5446c2_2cf8ca59a8e04c17870088ff52a04af1~mv2.jpg)
(C) A magnification on the outer part of an active lesion is shown. The lesion edge is indicated by the dotted line. Demyelination is reflected by loss of staining for PLP. Small CHIT1+ cells are mostly found close to the lesion edge (small arrows). A larger CHIT1+ cell can be appreciated a bit further away from the lesion edge (arrowhead). (D) A further magnification of the same lesion shows a CHIT1+ Iba1+ cell (yellow arrowhead) and the larger CHIT1+ Iba1- cell (white arrowhead). Inside these cells, PLP+ degradation products reveal the phagocytic character of these cells. TMEM119 is absent on the larger CHIT1+ cell (white arrowhead). (E) Both TMEM119 and Iba1 are however present on small CHIT1+ cells (arrowhead). (F) Double staining for CHIT1 and GFAP demonstrates that CHIT1 is absent in astrocytes. DAPI, 4’,6-Diamidino-2-phenylindole.
We concluded that elevated CSF CHIT1 concentrations at diagnosis likely identify MS patients who already at the start of their disease exhibit significant CNS pathology – i.e. more extensive microglial activation – and are therefore at greater risk for faster disability progression. Further validation is now indispensable to consolidate the potential of CHIT1 as a prognostic biomarker in MS clinical practice, where it could be used to inform the potential benefit of different treatments which might significantly improve MS outcome.
Comments