Clinical improvement of DM1 patients reflected by reversal of disease-induced gene expression in blood

Abstract

Background: Myotonic dystrophy type 1 (DM1) is an incurable multisystem disease caused by a CTG-repeat expansion in the DM1 protein kinase (DMPK) gene. The OPTIMISTIC clinical trial demonstrated positive and heterogenous effects of cognitive behavioral therapy (CBT) on the capacity for activity and social participations in DM1 patients. Through a process of reverse engineering, this study aims to identify druggable molecular biomarkers associated with the clinical improvement in the OPTIMISTIC cohort.

Methods: Based on full blood samples collected during OPTIMISTIC, we performed paired mRNA sequencing for 27 patients before and after the CBT intervention. Linear mixed effect models were used to identify biomarkers associated with the disease-causing CTG expansion and the mean clinical improvement across all clinical outcome measures.

Results: We identified 608 genes for which their expression was significantly associated with the CTG-repeat expansion, as well as 1176 genes significantly associated with the average clinical response towards the intervention. Remarkably, all 97 genes associated with both returned to more normal levels in patients who benefited the most from CBT. This main finding has been replicated based on an external dataset of mRNA data of DM1 patients and controls, singling these genes out as candidate biomarkers for therapy response. Among these candidate genes were DNAJB12, HDAC5, and TRIM8, each belonging to a protein family that is being studied in the context of neurological disorders or muscular dystrophies. Across the different gene sets, gene pathway enrichment analysis revealed disease-relevant impaired signaling in, among others, insulin-, metabolism-, and immune-related pathways. Furthermore, evidence for shared dysregulations with another neuromuscular disease, Duchenne muscular dystrophy, was found, suggesting a partial overlap in blood-based gene dysregulation.

Conclusions: DM1-relevant disease signatures can be identified on a molecular level in peripheral blood, opening new avenues for drug discovery and therapy efficacy assessments.

Keywords: Biomarker; Lifestyle intervention; Myotonic dystrophy type 1; Peripheral blood; RNA-seq; Therapeutic Response.

Fig. 1

Distribution of changes in outcome measures per patient. Per outcome measure, changes between baseline and 10 months of CBT were scaled by the root mean square. Additionally, for some outcome measures, a sign adjustment was performed so that an increased score is always associated with improved health status. All outcome measures are shown per patient, where patients were ordered along the y-axis by their change in DM1-Activ-c scores (purple squares). Boxes enclose the 25th to 75th percentiles, divided by a thick line that represents the mean compound response score. The whiskers represent the lowest/highest value no further than 1.5 times the interquartile range. Quality of life (red): Myotonic Dystrophy Health Index, Individualized Neuromuscular Quality of Life Questionnaire, Adult Social Behavioural Questionnaire, Illness Management Questionnaire, Checklist individual strength — Subscale activity. Physical assessments (blue): Six-Minute Walk Test, BORG Scale, accelometery measures. Fatigue scores (black): Fatigue and Daytime Sleepiness Scale, Checklist Individual Strength — Subscale fatigue, Jacobsen Fatigue Catastrophizing Scale. Cognition and other (gray): Trail Making Test, Stroop Color-Word Interference Test, McGill Pain Questionnaire, Beck Depression Inventory — Fast Screen, Social Support — Discrepancies and Negative Interactions, Apathy Evaluation Scale — Clinical version, Self-Efficacy Scale 28

Fig. 2

Changes in gene expression after cognitive behavioral therapy. A linear mixed effect model was fitted for each gene, estimating the fixed effect of CBT while accounting for random effects of the individual. The p-values for the fixed effect were estimated via Satterthwaite’s degrees of freedom method and FDR corrected. A Volcano plot of significance (−log10 of the nominal p-value) and the effect size for changed expression after 10 months of CBT. Genes for which the effect size of CBT is significant (FDR < 0.05) are visualized in black. B Heatmap of changes in normalized logCPM values between the baseline and the 10-month assessment for the 560 genes significantly associated with the CBT effect size (scores ranging from −3 (dark red) to +3 (dark blue)). Patients and genes were clustered based on the complete linkage method for hierarchical clustering, and values were centered and scaled per gene. For each patient, changes in DM1-Activ-c (delta-DM1-Activ-c) and Six-Minute Walk Test (delta-6MWT) as well as the compound response score were added (scores ranging from −1.6 (dark red) to +2.76 (dark blue)). Delta-DM1-Activ-c and delta-6MWT were scaled by their root mean square. C Expression values (logCPM) at baseline (blue) and after CBT (red) of the four genes with the lowest nominal p-values from panel A including their Pearson correlations

Fig. 3

Gene expression levels associated with CTG-repeat length. For each gene, a mixed effect model was fitted with before/after CBT and CTG-repeat length as fixed effects, while accounting for (random) effects of the individual. The p-values for the fixed effects were estimated via Satterthwaite’s freedom method and FDR corrected. A Volcano plot of significance (−log10 of the nominal p-value) and effect size of the CTG-repeat length (per 100 CTGs) on gene expression. Genes for which the effect of CTG-repeat is significant (FDR < 0.05) are visualized in black. B For the four genes with the lowest nominal p-values from panel A, the gene expression values (logCPM) are plotted against the CTG-repeat length. Blue dots represent baseline expression values, and red dots expression values after CBT. The regression line is fitted over all values, independent of the time point. The Pearson correlation coefficients for the association between CTG-repeat length and gene expression are displayed for each gene

Fig. 4

Gene expression association with compound response scores. For each gene, a mixed effect model was fitted with before/after CBT and compound response scores as fixed effects, while accounting for (random) effects of the individual. Compound response scores were fitted for gene expression values after CBT and set to be zero at baseline; the effect size of this covariate therefore expresses changes in gene expression compared to the baseline values that are associated with clinical response. The p-values for the fixed effects were estimated via Satterthwaite’s freedom method and FDR corrected. A Volcano plot of significance (−log10 of the nominal p-value) and the effect size of the compound response score on gene expression. Genes for which the effect size of compound response is significant (FDR < 0.05) are visualized in black. B For the four genes with the lowest nominal p-values from panel A, the changes in gene expression (delta logCPM after-before CBT) are plotted against the compound response scores, including Pearson correlations

Fig. 5

Clinical improvement is linked to normalization of expression of CTG-repeat-associated genes. Linear mixed effect models were fitted for each expressed gene, with CBT as a fixed effect, patient as a random effect, and either CTG repeat or compound response as the predictor. p-values for the regression coefficients were estimated via Satterthwaite’s degrees of freedom and considered significant for values smaller than 0.05 after FDR correction. Furthermore, differences in gene expression of blood samples from DM1 patients and controls were calculated based on an external study using a Wilcoxon signed-rank test on normalized, log-transformed gene counts. A Venn diagram illustrating the number of significant genes associated with CTG-repeat length and compound response, as well as their overlap (disease-relevant changes). B For all expressed genes, the regression coefficients of the compound response scores are plotted against the regression coefficients of the CTG-repeat lengths, including their Pearson correlation. For illustrative purposes, the regression coefficients of the CTG repeat have been multiplied by 100. Furthermore, the x-axis has been scaled between −0.25 and 0.25, removing 12 outliers from the figure. Colored in purple are the genes for which both regression coefficients were significant (FDR < 0.05). C For all expressed genes, the compound response effect size is plotted against the DM1 effect size based on an external study comparing blood expression profiles from DM1 patients and controls, including their Pearson correlation [25]. For illustrative purposes, the x-axis has been scaled between −1.5 and 1.5, removing 6 outliers from the figure. Colored in purple are the same 97 genes as in panel B