The Splice Index as a prognostic biomarker of strength and function in myotonic dystrophy type 1

Abstract

BACKGROUNDMyotonic dystrophy type 1 (DM1) is a multisystemic, CTG repeat expansion disorder characterized by a slow, progressive decline in skeletal muscle function. A biomarker correlating RNA mis-splicing, the core pathogenic disease mechanism, and muscle performance is crucial for assessing response to disease-modifying interventions. We evaluated the Myotonic Dystrophy Splice Index (SI), a composite RNA splicing biomarker incorporating 22 disease-specific events, as a potential biomarker of DM1 muscle weakness.METHODSTotal RNA sequencing of tibialis anterior biopsies from 58 DM1 participants and 33 unaffected/disease controls was used to evaluate RNA splicing events across the disease spectrum. Targeted RNA sequencing was used to derive the SI from biopsies collected at baseline (n = 52) or a 3-month (n = 37) follow-up visit along with clinical measures of muscle performance.RESULTSThe SI demonstrated significant associations with measures of muscle strength and ambulation, including ankle dorsiflexion (ADF) strength and 10-meter run/fast walk (Pearson's r = -0.719 and -0.680, respectively). The SI was relatively stable over 3 months (intraclass correlation coefficient [ICC] = 0.863). Latent-class analysis identified 3 DM1 subgroups stratified by baseline SI (SIMild, SIModerate, and SISevere); SIModerate individuals had a significant increase in the SI over 3 months. Multiple linear regression modeling revealed that baseline ADF and SI were predictive of strength at 3 months (adjusted R² = 0.830).CONCLUSIONThe SI is a reliable biomarker that captures associations of RNA mis-splicing with physical strength and mobility and has prognostic utility to predict future function, establishing it as a potential biomarker for assessment of therapeutic target engagement.TRIAL REGISTRATIONClinicalTrials.gov NCT03981575.FUNDINGFDA (7R01FD006071), Myotonic Dystrophy Foundation, Wyck Foundation, Muscular Dystrophy Association, Novartis, Dyne, Avidity, PepGen, Takeda, Sanofi Genzyme, Pfizer, Arthex, and Vertex Pharmaceuticals.

Keywords: Muscle biology; Neuromuscular disease; RNA processing; Skeletal muscle.

Figure 1. DM1 participant cohort displays broad spectrum of RNA splicing dysregulation, as assessed by total RNA-seq.

(A) DM1 participant demographic information, including sample size, age at biopsy, and sex distribution in complete cross-sectional and longitudinal cohort subset. The mean normalized performance on clinical outcome measures for the cross-sectional DM1 cohort are reported along with sample size, mean ± SD, and 95% CI. Additional demographic information of all samples used in this report are provided in Supplemental Table 1. (B) Box-and-whisker plots of outcome measure performance in DM1 cross-sectional cohort, where the lines represents median performance, the bounds of the boxes are the 25th and 75th percentiles, and whiskers extend to maximum and minimum values. (C) Heatmap displaying estimated percentage spliced (Ψ) of top 50 significantly dysregulated skipped exon (SE) events between DM1 versus unaffected adult controls (AdCo) and disease control reference groups (DMD and LGMD) subjected to total RNA-seq (|ΔΨ| ≥ 0.1, FDR ≤ 0.05). Both rows (SE events) and columns (individual samples) were subjected to hierarchical clustering. Sample group and sex are annotated above the heatmap and individual IDs are reported below. (D) Ψ values for specific SE events in all sample groups (n = 22 AdCo, n = 95 DM1, n = 10 LGMD, and n = 1 DMD). Bar represents median. **P < 0.01; ****P < 0.0001. NS, not significant. One-way ANOVA with Tukey’s correction, where DMD was not included in the analysis due to insufficient sample size.

Figure 2. RNA splicing events included in composite Splice Index capture RNA mis-splicing patterns observed across the range of estimated, functional MBNL concentrations in DM1 cohort.

(A–C) Twenty-two RNA splicing events included in the Splice Index (SI) panel demonstrate variable sensitivity to [MBNL] observed across the spectrum of DM1 participants. Dose-response curves of 22 events included in the composite SI. Ψ values derived from total RNA-seq (Supplemental Table 3) were plotted against estimated functional concentrations of MBNL ([MBNL]_inferred). Ψ values from both unaffected adult controls (AdCo, n = 22) and DM1 participants (n = 95) were fit to a 4-parameter dose-response curve and events classified as (A) early, (B) intermediate, and (C) late responder events based on estimated EC₅₀ relative to the observed median of all events. Intermediate responder events were classified based on EC₅₀ values within the interquartile range of 22-event distribution. Early responder were classified based on an estimated EC₅₀ in the 25th percentile, while late responder events had values in the 75th percentile. Curve fit is annotated as the thick solid line and the 95% CI of the fit is shown as the above and below lines. All curve-fit parameters are reported in Supplemental Table 4.

Figure 3. Splice Index correlates strongly with clinical outcome measures of muscle strength and motor function in cross-sectional DM1 cohort.

(A and B) Correlation plots of Splice Index (SI) versus quantitative measures of ankle dorsiflexion (ADF) strength and hand grip strength (HGS), respectively. Individual measures are reported as the percentage of predicted strength compared with unaffected individuals. ADF Pearson’s r = –0.719 [–0.808, –0.597], n = 85 and HGS Pearson’s r = –0.716 [–0.805, –0.595], n = 87. (C) Correlation plot of SI versus 10-meter run/fast walk. Individual measures are reported as speed (meters/second). Pearson’s r = –0.680 [–0.782, –0.543], n = 82. All correlations are reported as Pearson’s r [95% CI] with 2-tailed P values.

Figure 4. Mean Splice Index increases between baseline and 3 months in longitudinal DM1 cohort, with no changes in functional endpoints.

(A and B) Assessment of changes in ankle dorsiflexion (ADF) strength in longitudinal DM1 cohort. (A) Comparison of baseline (BL) and 3-month (3M) ADF demonstrates strong test-retest reliability between time points. Line of agreement (x = y) is displayed and intraclass correlation coefficient (ICC) with 95% CI is reported. (B) Mean ADF strength is unchanged in longitudinal DM1 cohort at BL and 3M (n = 34). (C and D) BL (C) and 3M (D) Splice Index (SI) scores correlate strongly with time point–matched measures of ADF strength (n = 34). BL SI vs. BL ADF Pearson’s r = –0.723 [–0.852, –0.507] and 3M SI vs. 3M ADF Pearson’s r = –0.717 [–0.849, –0.499]. (E and F) Assessment of changes in 10-meter run/fast walk (10MRW) speed in longitudinal DM1 cohort. (E) Comparison of BL and 3M 10MRW speed demonstrates strong test-retest reliability (F) and no mean difference between time points (n = 32). (G and H) BL (G) and 3M (H) SI scores correlate strongly with time point–matched measures of 10MRW speed. BL SI vs. BL 10MRW, n = 34, Pearson’s r = –0.725 [–0.854, –0.513] and 3M SI vs. 3M 10MRW, n = 32, Pearson’s r = –0.700 [–0.843 –0.465]. (I and J) Assessment of changes in SI score over 3M in longitudinal DM1 cohort. (I) SI demonstrates moderate test-retest reliability (J), but a significant increase in mean SI score (n = 35). (K) Bland-Altman plot illustrating agreement between SI scores at BL and 3M in longitudinal subcohort. Dotted lines represent mean of the differences (bias) and 95% limits of agreement (mean of differences ± 1.96 SD). Data in B, F, and J are presented as mean ± SD and were analyzed with a paired, 2-tailed t test. ***P < 0.001. NS, not significant. All correlations reported as Pearson’s r [95% CI] with 2-tailed P values.

Figure 5. Splice Index (SI) demonstrates dynamic shifts over 3 months in select DM1 subcohorts stratified by baseline SI score.

(A) SI score significantly increases between baseline (BL) and 3 months (3M) in SI_Moderate group. DM1 participants in longitudinal cohort were stratified by BL SI score: SI_Mild (0 ≤ BL SI ≤ 0.4, n = 11), SI_Moderate (0.4 < BL SI ≤ 0.75, n = 13), and SI_Severe (0.75 < BL SI ≤ 1.0, n = 11). Data represented as mean ± SD. **P < 0.01 by paired, 2-tailed t test. NS, not significant. Connected scatter plots are also displayed to show shifts in BL and 3M paired SI values for each DM1 individual. (B) Kernel density estimation plot of BL and 3M SI distributions in SI_Mild, SI_Moderate, and SI_Severe groups. (C and D) Kernel density estimation plot of BL and 3M (C) ADF and (D) 10MRW distributions in SI_Mild, SI_Moderate, and SI_Severe groups. ADF is reported as the percentage of predicted strength as compared with unaffected individuals and 10MRW in speed (m/s).

Figure 6. Variable sensitivity of 22 events within Splice Index (SI) panel capture RNA splicing shifts over 3 months in baseline SI–stratified subcohorts.

(A–C) Relative sensitivity of RNA splicing events included in SI panel to levels of free [MBNL] allows for detection of changes in mis-splicing across the spectrum of DM1 splicing dysregulation. Normalized ΔΨ (3M – BL) derived from targeted RNA-seq (Supplemental Table 5) of (A) early, (B) intermediate, and (C) late responder RNA splicing events as defined in Figure 2. Data presented as box-and-whisker plots, where lines represent median ΔΨ, the bounds of the boxes are the 25th and 75th percentiles, and whiskers extend to maximum and minimum values. ΔΨ for SI_Mild, SI_Moderate, and SI_Severe subcohorts are displayed for each event.

Figure 7. Multiple linear regression modeling highlights prognostic utility of the Splice Index (SI) in combination with time point–matched outcomes to predict future function at 3 months.

Combination of baseline (BL) SI and time point–matched measure of ankle dorsiflexion (ADF) strength is predictive of performance at 3 months (3M). Agreement plot of actual versus predicted 3M ADF derived from multiple linear regression model with adjusted R² are displayed (Supplemental Figure 12, Model 1). All quantitative parameters are provided in Supplemental Table 9.