Multi-omics profiling in spinal muscular atrophy (SMA): investigating lipid and metabolic alterations through longitudinal CSF analysis of Nusinersen-treated patients

Abstract

Spinal muscular atrophy (SMA) is a rare neuromuscular disease caused by biallelic mutations in the SMN1 gene, leading to progressive muscle weakness due to degeneration of the anterior horn cells. Since 2017, SMA patients can be treated with the anti-sense oligonucleotide Nusinersen, which promotes alternative splicing of the SMN2 gene, by regular intrathecal injections. In this prospective study, we applied metabolomic, lipidomic, and proteomic analysis to examine sequential CSF samples from 13 SMA patients and controls. This multi-omic approach identified over 800 proteins and 400 small molecules including lipids. Multivariate analysis of multi-omic data successfully discriminated between the CSF derived from SMA patients and control subjects. Lipidomic analysis revealed increased levels of cholesteryl esters and lyso-phospholipids, along with reduced levels of cholesterol and phospholipids in the CSF of SMA patients as compared to healthy controls. These data, combined with results from functional assays, led us to conclude that SMA patients exhibit altered levels and function of high-density-lipoprotein (HDL)-like particles in the CSF. Notably, Nusinersen therapy was observed to reverse disease-specific profile changes toward a physiological state, potentially explicable by restoring HDL function.

Keywords: Biomarker; Lipid metabolism; Mass spectrometry; Neuromuscular; Omics.

Fig. 1

Study design. A Intrathecal administration of Nusinersen, dose followed standard treatment protocol (FDA). CSF sampling occurred in parallel with Nusinersen treatment. In total, CSF samples were taken from 13 SMA patients (type I: n = 5, type II: n = 7, type 3: n = 1), including six therapy naïve samples (two SMA type I and four SMA type II) and from 15 age- and sex-matched controls (seven female, six male). Samples on treatment with Nusinersen (T1–T9) were aggregated at timepoint T4

Fig. 2

Lipidomics analysis of CSF from SMA patients and controls (ctrl). A Wilcoxon sum rank test of lipid classes obtained from the CSF of therapy naïve (TN) SMA patients as well as under therapy with Nusinersen (T1–T9). B, C OPLS-DA analysis of lipidomics data obtained from CSF of ctrl and SMA patients (green dots: SMA type I, blue dots: SMA type II, purple dot: SMA type III). B Comparison between ctrl and TN SMA patients. C Comparison between ctrl and SMA patients under therapy (T4)

Fig. 3

A Top 25 lipid species in therapy naïve SMA patients (TN) vs. controls (ctrl) according to molrank in OPLS-DA. A molrank refers to the ranking of molecular features based on their importance in the OPLS-DA and helps to identify which molecular feature contributes most significantly to the separation between different groups in the dataset. Complete list of significantly altered lipid species in TN SMA patients compared to ctrl is provided in a separate table in the supplement section. B Box plots from the 25 top lipids derived from OPLS-DA. Red bar: ctrl group, turquoise bar: SMA group with red dots indicating SMA type I and blue dots indicating SMA type II patients C ROC plot with an Area under curve (AUC) value of 0.89 indicating a good performance of our model. D Confusion matrix generated to evaluate the performance of the MVA model. Out of 20 cases, the algorithm identified 12 true negatives (TN), 4 true positives (TP), two false positives (FP), and two false negatives (FN)

Fig. 4

A Top 25 lipid species important for discrimination in OPLS-DA of CSF from T4 SMA patients versus ctrl. B Univariate analysis using box plots of selected 25 lipid species from discriminant analysis. Red bars indicate ctrl subjects and turquoise bars indicate CSF from SMA patients under therapy (T4) including green dots for SMA type I, blue dots for SMA type II, and purple dots for SMA type III patients, respectively. C ROC plot with an area under curve (AUC) value of 0.99 indicating very good performance of our model. D Confusion matrix: Out of 27 cases, 12 were identified truly negative, eight true positives, two false negatives, and two false positives. E CSF cholesterol efflux capacity (CEC) assay. Following Shapiro–Wilk test for normality testing, data were compared using Welch’s t test. CEC was measured at 2.16 ± 0.07 (SEM)% at baseline, 1.72 ± 0.06 (SEM) % in TN SMA patients, and 1.92 ± 0.05 (SEM) % in T4 SMA patients. *p ≤ 0.05; ***p ≤ 0.001

Fig. 5

Multivariate analysis of metabolomics profiles obtained from CSF of study cohort. A OPLS-DA of therapy-naïve (TN) SMA patients compared to ctrl. B OPLS-DA of SMA patients under Nusinersen therapy (T4) compared to ctrl. C Volcano plot of metabolites identified in the CSF of SMA patients in the therapy-naïve state (TN) as well as under therapy (T1–T7). D OPLS-DA of SMA patients type I versus type II under Nusinersen therapy. E Univariate analysis comparing the CSF of type I and type II SMA patients focusing on the top 25 metabolites identified through OPLS-DA

Fig. 6

Univariate analysis of top 25 compounds identified through metabolomics analysis. A Ranking of the 25 most important metabolites responsible for altered CSF metabolome profile in TN SMA patients compared to ctrl. B These 25 metabolites were selected for univariate analysis using box plots. C ROC plot with an AUC of 0.91. D Confusion matrix: out of 38 cases, 28 cases were identified as true negative, six as true positive, four as false negative, and zero false positive

Fig. 7

Univariate analysis of top 25 compounds identified through metabolomics analysis in the CSF of T4 SMA patients vs. ctrl. A VIP plot of top 25 metabolites according to molrank. B Univariate analysis using box plot. Red bars indicate ctrl and Turquoise bars indicate SMA patients including SMA type I (green dots), SMA type II (blue dots), and SMA type III (purple dot). C ROC plot with a calculated AUC-ROC of 0.96. D Confusion matrix generated predicted 26 cases as true negative (TN), 23 cases as true positive (TP), two false positives (FP), and three false negatives (FN)

Fig. 8

Proteomics analysis of CSF obtained from SMA patients and ctrl. A, B OPLS-DA of proteomics profiles obtained from the CSF of SMA patients in the therapy-naïve state (TN, A) as well as in response to Nusinersen treatment (T4, B)