microRNA-133a as an indicator of disease progression and treatment response in X-linked myotubular myopathy

Abstract

X-linked myotubular myopathy (XLMTM) is a rare pediatric neuromuscular disease caused by loss-of-function variants in myotubularin (MTM1). With novel therapies entering clinical trials, the discovery of robust biomarkers that reflect disease severity and therapeutic efficacy is critically required. Using high-throughput and directed approaches, we identified a decrease in miR-133a expression as a marker of XLMTM disease in skeletal muscle and plasma of a mouse model of XLMTM (Mtm1 KO). miR-133a is a muscle-enriched non-coding RNA (myomiR) involved in muscle development and function and is implicated in the regulation of the XLMTM modifier gene DNM2. miR-133a has emerged as both a treatment-effect biomarker and therapeutic candidate in other neuromuscular diseases. We demonstrate that miR-133a expression negatively correlates with disease severity in Mtm1 KO mice and is upregulated in response to treatments that improve DNM2 expression and/or significantly rescue XLMTM. Moreover, we show that miR-133a expression in treated Mtm1 KO mice positively correlates with treatment response and was shown to have high discrimination accuracy for XLMTM by linear discriminant analysis (79%-90%) and receiver operating characteristic curve analysis (AUC >0.80). These results support miR-133a as a robust, circulating biomarker that reflects disease severity and treatment response in XLMTM.

Keywords: MT: Non-coding RNAs; biomarker; microRNA; muscle disease; myotubular myopathy; therapy.

Graphical abstract

Figure 1

miR-133a and miR-486 are dysregulated in skeletal muscle of Mtm1 KO mice (A–F) TaqMan quantitative real-time PCR results for miR-133a and miR-486 expression at postnatal day (PND) 21, PND28, and PND36 in 4 hindlimb skeletal muscles (TA, Quad, Gastroc, and Ham) of Mtm1 KO mice relative to WT controls (n = 7–10 mice/group). Quantitative real-time PCR calculations were performed using the 2^−ΔΔCT method, with SnoRNA202 as an internal reference small RNA. All samples are normalized to WT. (G and H) Pearson’s correlation analysis in TA and Quad (G) and Gastroc and (H) of Mtm1 KO mice showing a negative correlation between skeletal muscle expression of miR-133a and PND exclusively in TA and Quad (R = 0.66; R = 0.60; p < 0.01). Quantitative real-time PCR calculations were performed using the 2^−ΔΔCT method, with U6 as an internal reference small RNA. All samples are normalized to WT. For all graphs, each symbol represents an individual sample, and each graph represents mean (SD). For all 2-group comparisons, 2-tailed unpaired Student’s t tests were performed to determine significance. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.0001.

Figure 2

miR-133a expression correlates with myofiber size in skeletal muscle of Mtm1 KO mice (A) Myofiber size in TA, reported as minimum Feret’s diameter (μM) in WT and Mtm1 KO mice from PND21 to PND36. Histogram depicting fiber size distribution at PND36 highlights abundance of smaller myofibers in Mtm1 KO compared to WT at PND36 (n = 10–16 mice/group). (B) Pearson’s correlation analysis in TA of all untreated Mtm1 KO mice indicates a strong negative correlation between myofiber size (μM) and PND. (C) Pearson’s correlation coefficient (R) derived from linear regression of miR-133a expression in WT and Mtm1 KO mice from PND21 to PND36 (n = 7–8 mice/group). (D) Representative images of TA dystrophin identified by immunofluorescence at PND21, PND28, and PND36. Myofiber size is reported by minimum Feret’s diameter (μM) and calculated using a custom Fiji macro written to automatically detect and quantify cross-sectional area of fibers. Magnification 20×. Scale bar, 50 μM. For all graphs, each symbol represents an individual sample, and each graph represents mean (SD). For all 2-group comparisons, 2-tailed unpaired Student’s t tests were performed to determine significance. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.0001.

Figure 3

Relationship between miR-133a expression and myofiber size in skeletal muscle of Mtm1 KO mice (A–C) Myofiber size in (A) Quad, (B) Gastroc, and (C) Ham reported as minimum Feret’s diameter (μM) in WT and Mtm1 KO mice from PND21 to PND36. Histogram depicting fiber size distribution at PND36 in each muscle type highlights abundance of smaller myofibers in Mtm1 KO compared to WT at PND36 (n = 5–7 mice/group). (D–F) Pearson’s correlation analysis in (D) Quad, (E) Gastroc, and (F) Ham of Mtm1 KO mice indicates little to no correlation between myofiber size and PND and demonstrates that myofiber size does not change with disease progression. (G–I) Pearson’s correlation coefficient (R) derived from linear regression of miR-133a and myofiber size in (G) Quad, (H) Gastroc, and (I) Ham of Mtm1 KO mice at PND21, PND28, and PND36 (n = 5–7 mice/group). For all graphs, each symbol represents an individual sample, and each graph represents mean (SD). For all 2-group comparisons, 2-tailed unpaired Student’s t tests were performed to determine significance. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.0001.

Figure 4

miR-133a expression has high discrimination accuracy for XLMTM (A–C) Two-dimensional plot of miR-133a expression values in WT and Mtm1 KO mice that made up the training set at PND21 (A), PND28 (B), and PND36 (C) (n = 20–24 mice/group). Genotype classification was performed by LDA using miR-133a expression and myofiber size in combination and is represented as a line. Each symbol represents individual samples. (D–F) One-dimensional plot of myofiber size (μM) in WT and Mtm1 KO mice that made up the training set at PND21 (D), PND28 (E) PND36 (F) (n = 20–24 mice/group). The miR-133a expression value yielding the greatest genotype separation between WT and Mtm1 KO as determined by LDA is represented as a line. (G–I) One-dimensional plot of myofiber size (μM) in WT and Mtm1 KO mice that made up the training set at PND21 (G), PND28 (H), and PND36 (I) (n = 20–24 mice/group). Value of myofiber size (μM) yielding the greatest genotype separation between WT and Mtm1 KO as determined by LDA is represented as a line. (J–L) ROC curve analysis at PND 21 (J), PND28 (K), and PND36 (L) for different thresholds of miR-133a expression and myofiber size (μM). ROC curve indicates the true positive rate (% Sensitivity) and corresponding false positive rate (100% Specificity) across all skeletal muscle expression values of miR-133a (n = 23–31 mice/genotype). ROC plots show for different expression levels of miR-133a the rate of true positives (indicating miR-133a correctly classifies XLMTM from WT at the defined cutoff value [y axis] and the corresponding rate of false positives (indicating miR-133a expression incorrectly classifies XLMTM from WT [x axis]). The miR-133a expression value that corresponds with a true positive rate of 0.80 (80%) is indicated on each graph.

Figure 5

Circulating miR-133a associates with DSSs in Mtm1 KO mice TaqMan quantitative real-time PCR results for miR-133a and miR-486 expression at PND21 (A and B), PND28 (C and D) and PND36 (E) in plasma of Mtm1 KO mice relative to WT controls (n = 6–18 mice/group). Violin plots representing association between proportion of animals with varying DSSs (DSS 0–34) and circulating −log₁₀mir-133a Cq values at PND 21 (B), PND 28 (D), and PND36 (F) (n = 6–18 mice/group). Pearson’s correlation analysis in plasma showing a negative correlation between (G) circulating miR-133a and (H) miR-486 expression and PND (n = 6–18 mice/group). For all graphs, each symbol represents an individual sample. For all 2-group comparisons, 2-tailed unpaired Student’s t tests were performed to determine significance. For all log-transformed Cq values, t tests were performed on the untransformed Cq values. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.0001.

Figure 6

miR-133a expression is responsive to treatment in Mtm1 KO mice and correlates with phenotypic improvement TaqMan quantitative real-time PCR results for miR-133a expression in TA of WT and Mtm1 KO mice at PND36, treated with MTM1 gene replacement therapy (AT132) (A and B), tamoxifen (C), and valproic acid (VPA) (D) (n = 4–8 mice/group). (B) TaqMan quantitative real-time PCR results for Dnm2 mRNA expression in TA of WT and Mtm1 KO mice treated with MTM1 gene replacement therapy (AT132) (n = 4–8 mice/group). Quantitative real-time PCR calculations were performed using the 2^−ΔΔCT method using SnoRNA202 as an internal reference small RNA for miR-133a and B2m as an internal reference gene for Dnm2. All samples are normalized to WT (n = 4–8 mice/group). (E) One-dimensional LDA plot of miR-133a expression in WT and Mtm1 KO mice at PND36 treated with MTM1 gene replacement therapy (AT132), VPA, and TAM (test set) (n = 4–8 mice/group). miR-133a expression in the training set is represented at PND21, PND28, and PND36 (n = 20–24 mice/group). For each time point, the value of miR-133a expression yielding the greatest genotype separation between WT and Mtm1 KO as determined the LDA training set is represented as a dotted line. (F–H) Correlation matrices for miR-133a expression and various XLMTM disease-relevant parameters in untreated Mtm1 KO mice (F), Mtm1 KO mice treated with TAM or VPA (G), and Mtm1 KO mice treated with MTM1 gene replacement therapy (AT132) (H). Pearson’s correlation coefficient (R) for each correlation is indicated in white, and strength and directionality are indicated by color (see key). BW, body weight; DNM2, Dnm2 mRNA; GS, grip strength; LR, lactated Ringer’s; miR, miR-133a; RE, rears; WG, weight gain. For all graphs, each symbol represents an individual sample, and each graph represents mean (SD). For all 2-group comparisons, 2-tailed unpaired Student’s t tests were performed to determine significance. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.0001.