Type II SMA muscle has similar levels of SMN RNA and protein to those of scoliotic controls. To characterize the muscle of SMA type II patients, we collected a cohort of paravertebral muscle samples from patients undergoing orthopedic surgery for scoliosis. Our cohort of samples consisted of 8 type II (or type I bis) SMA patients and 7 non-SMA control individuals, also undergoing the same surgical correction for scoliosis (Figure 1A and Table 1). Muscle tissue was obtained from the surgical discard created during the surgery. To the extent possible, the cohort was sex (9 females, 6 males) and age matched, although on average, the SMA cohort is 4 years younger (mean 12.75 ± 2 years) than the control cohort (16.14 ± 1.3 years) (Figure 1B). All SMA patients received intrathecal nusinersen injections, and a subset had a clinical history of risdiplam treatment (Table 1), although the full clinical history of each sample was not always available.

SMN RNA and protein levels are comparable between SMA type II and control paravertebral muscle. (A) Diagram of the cohort with the sex balance for each group. (B) Age, in years, for the SMA (n = 8) and control (n = 7) groups. Each data point represents 1 patient. Groups were compared with a 2-sided Student’s t test. **P < 0.001. (C and D) Relative quantification (qPCR) of the copy number of SMN1 (C) or SMN2 (D) in gDNA (n = 5 controls, n = 8 SMA). Relative quantification was calculated with RPPH1 as a loading control. Each data point represents the average of 2 technical replicates of the measurement per patient. (E and F) Reads, represented as transcripts per million (TPM) mapping to the SMN full-length transcript (E), and the SMNΔ7 transcript missing exon 7 (F). The origin of the full-length transcript, either the SMN1 or SMN2 locus, is designated below each pair of violin plots. Each data point represents 1 sample: n = 7 for controls, n = 8 for SMA. Adjusted P values are derived from an ordinary 1-way ANOVA with Dunnett’s multiple-hypothesis testing. **Padj < 0.001, ***Padj < 0.0001, ****Padj < 0.00001. (G) Western blot for the approximately 37 kDa SMN protein and approximately 117 kDa vinculin protein. Each patient is labeled on the top. (H) Quantification of the blot in G. Each data point represents the normalized SMN/vinculin value for each sample (n = 6 controls, n = 6 SMA samples). The mean of the group was compared with a 2-sided Student’s t test. NS, not significant (P = 0.8). (I) Correlation between the TPM values of the full-length (FL) SMN transcripts in each sample in E and F compared to the normalized protein quantification in H (n = 6 controls in black, n = 6 SMA in pink). Pearson’s R2 value and it’s associated P value are reported. The solid line represents the simple linear regression and the dashed lines represent the 95% confidence internal.

SMA paravertebral muscle cohort

As the major genetic modifier of SMA is known to be the SMN2 gene (20), we sought to determine the relative copy number of SMN2 in our control versus patient samples. The predicted copy number of SMN2 in type II patients is between 3 and 4 (21). However, absolute copy number detection via traditional PCR or qPCR methods can be difficult without digital PCR methods (22). Therefore, we decided to quantify the number of SMN1 and SMN2 copies relative to our control samples. As expected, in the control samples, we could easily measure SMN1 copies, but detected none in the SMA samples (Figure 1C). By contrast, SMN2 copies were present in the SMA type II samples in variable relative amounts, with 3 out of 8 patients having about the same levels as control samples (Figure 1D), while most samples (5 out of 8) had more, suggesting a gene conversion event with SMN1. However, it is important to note that SMN2 copy number naturally varies in the population.

We performed RNA-seq of the same muscle samples to characterize their transcriptome. We began by analyzing the reads mapping to either the full-length or SMNΔ7 (6) transcripts. Using a variety of SNPs in the introns and UTR regions of the 2 SMN copies (23, 24), we were able to annotate transcripts derived from either the SMN1 or SMN2 copy. As expected, when comparing the full-length canonical SMN transcripts, we observed a dramatic decrease between control and SMA samples, where few to no reads were mapping to SMN1, consistent with the SMA diagnosis and our gDNA qPCR (Figure 1E). In control samples, we observed few full-length SMN reads mapping to SMN2, with the plurality of SMN2 mapping reads resulting in the SMNΔ7 transcript (Figure 1, E and F). However, in the SMA patients, we observed many more full-length than SMNΔ7 transcripts deriving from the SMN2 locus (Figure 1, E and F). We hypothesize that this may be due to treatment with risdiplam, but this cannot be concluded without pre- and posttreatment samples. To validate the SMN mapping approach, we used the same pipeline to map reads from a previous RNA-seq study that profiled the biceps muscle of Duchenne muscular dystrophy (DMD) and SMA type I patients (25, 26). We observed few to no reads coming from the SMN1 locus in SMA type I patients and an increase in SMNΔ7 transcript from the SMN2 locus (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.180992DS1). However, in these untreated samples, there was no significant increase in full-length splicing from the SMN2 allele (Supplemental Figure 1A), further leading us to hypothesize this may be an effect of the risdiplam treatment in our patients. Additionally, a higher SMN2 copy number may be responsible for the increased number of full-length SMN2 transcripts, but we observed no correlation between these 2 parameters (Supplemental Figure 1B). This is in line with previous findings in the spinal cord that SMN1 and SMN2 copy number correlate poorly with protein expression (17).

No correlation was observed between the amount of SMN2 canonical transcript and the amount of SMNΔ7 transcripts (Supplemental Figure 1C) — rather, 3 out of 8 SMA samples had high SMNΔ7 levels, irrespective of the amount of SMN2 full-length transcript. We also tested the level of RNA expression with age, as SMN levels have been shown to decrease in aging (17); however, within our limited age range, no decrease was observed (Supplemental Figure 1D). Previous studies have suggested that the amount of SMN full-length transcript is also dependent on the levels of 2 splicing factors, hnRNPA1 and hnRNPA2 (27, 28). Our RNA-seq data confirm that global expression levels of HNRNPA1 and HNRNPA2B1 correlate with increased levels of total SMNΔ7 observed (Supplemental Figure 1E). Consistent with the role of SMN in splicing, we observed a higher degree of SMN transcripts with retained introns in the SMA samples compared with controls (Supplemental Figure 1F).

We next determined the level of the full-length, approximately 37 kDa SMN protein in the control and SMA samples using Western blotting. On average, SMA patients had similar levels of SMN protein to the scoliotic controls (Figure 1, G and H). We next correlated the RNA and protein levels and observed no correlation between the two (Figure 1I). Collectively these data allow us to conclude that in this small cohort, despite some variation in copy number and RNA levels, at adolescence the full-length SMN protein levels were similar in the SMA and control groups. This lack of observable difference may be due to the low levels of SMN protein that are observed after birth (17).

SMA muscles have altered myofiber size, with the presence of multiple internalized nuclei in a single fiber. We next turned our attention to characterizing the muscle histologically. We performed H&E staining for the samples with well-defined muscle fiber architecture (5 out of 8 SMA patients). H&E staining demonstrated several classic signs of SMA muscle histology, including the presence of hypertrophic fibers in 2 samples, SMA4 and SMA8 (Figure 2A), and the clusters of small fibers suggestive of denervation in samples SMA2 and SMA6 (Supplemental Figure 2A). The concurrent presence of both atrophic and hypertrophic fibers has been previously noted in SMA type III muscle (29). We quantified the area of each myofiber and plotted the distribution of sizes for the control samples and SMA samples (Figure 2B). While the control samples had myofibers with relatively homogeneous areas, the range of the area measurement (Figure 2C) as well as the maximum area was increased in the SMA samples (Figure 2D). Moreover, in addition to the within-sample variation in muscle fiber size in the SMA patients, we also observed a patient-to-patient variability in fiber size that was less prominent among controls (Supplemental Figure 2B).

SMA type II paravertebral muscle after treatment is characterized by abnormal myofiber size distribution. (A) H&E staining of muscle tissues from control and SMA patient samples. Representative images (3–5 images were acquired per sample) are shown for each group. Original magnification, ×200. Scale bars: 50 μm. (B) Histogram of the fiber area. The frequency distribution represents the percentage of all fibers per group that falls within the bin area. Fiber area was measured in arbitrary units. n = 6 controls and n = 5 SMA, with multiple fiber measurements per sample, taken from 1 representative image per sample. (C) The range (highest to lowest) of fiber area measurements per patient. Each data point represents 1 image of 1 patient sample (n = 6 controls, n = 5 SMA). Group averages were compared using a 2-sided Student’s t test. *P < 0.05. (D) The maximum fiber area measured for each patient from 1 representative image. Each data point represents 1 patient (n = 6 controls and n = 5 SMA). Group averages were compared using a 2-sided Student’s t test. *P < 0.05. (E) Representative cropped and enlarged myofibers from images showing myofibers with multiple internalized nuclei from the 2 patients where we observed this phenomenon. Internalized nuclei are designated by white arrows. (F and G) Quantification of single centralized (F) or multiple internalized nuclei (G) in SMA (n = 5) and controls (n = 6). Each data point represents the percentage of such nuclei in the fibers counted in 1 image from each patient sample. Group averages were compared using a 2-sided Student’s t test. NS, not significant (P > 0.05).

Next, we assessed the percentage of the myofibers with internalized nuclei, a morphological feature common to several myopathies (30). Centralization of nuclei occurs as a response to injury during satellite cell fusion in fiber regeneration. Thus, observing centralized/internalized nuclei in a myopathy can imply that the tissue is attempting to repair itself. We first scored the slides for fibers with internalized nuclei and observed an increase in some, but not all, SMA samples. In some samples, particularly in SMA2, -4, and -8, we observed several myofibers with multiple internalized nuclei (Figure 2, E and F). However, this was not present in all samples. Within a single tissue section, we could observe both fibers with and without these multiple internalized nuclei in the same cluster, and in general, these multicentralized nuclei were found in the largest fibers.

Despite the limitations of the sampling method, namely, that we are viewing a small portion of a large muscle, this histological analysis showed that in SMA samples, there is a disorganization of the muscle fiber architecture, with a high range of variability in fiber size both within a single patient and between patients.

RNA-seq of SMA paravertebral muscle reveals changes in calcium regulation and oxidative phosphorylation. To obtain a broad molecular picture of the state of the SMA muscle, we performed RNA-seq of each sample (Figure 3A). Principal component analysis highlighted the heterogeneity in the samples, reflecting what we observed in our histological characterization, with approximately 44% of the variation being explained by PC1 that captured the diagnosis (i.e., SMA vs. control) (Figure 3B). Differential expression analysis found 166 downregulated genes and 396 upregulated genes (Figure 3C and Supplemental Table 1). Pathway analysis showed an enrichment for mitochondrial processes such as oxidative phosphorylation (OXPHOS) and the citric acid cycle (TCA) among the downregulated genes, and an enrichment of calcium signaling and P53 target pathways among the upregulated genes (Figure 3D). We next sought to determine whether we could observe any sex-specific effects in our samples, as within the SMA cohort we had an equal number of both sexes (XX, XY, n = 4). We also performed this comparison with the control samples. However, only the classic X- or Y-linked transcripts were differentially expressed, including XIST and TSIX upregulated in the female SMA muscle and TXLNGY and DDX3Y upregulated in the male samples (Supplemental Table 2).

Transcriptional characterization of treated SMA and control muscle samples. (A) Diagram of the RNA-seq library generation process used for the samples. (B) Principal component analysis (PCA) plot obtained using the top 2,000 variably expressed genes in each muscle RNA-seq library. Each data point represents the cDNA library of a single sample. The variance explained for each principal component (PC) is plotted on the axes. (C) Heatmap of the genes that are differentially expressed (DEGs) between SMA and control samples. Each column represents the DEGs from 1 sample, which have been hierarchically clustered. Diagnosis and sex are designated by the colored bars on the top. Transcripts on the heatmap are presented as z-scored transcripts per million (TPM) counts. DEGs include 166 downregulated and 396 upregulated genes comparing SMA- versus control-derived samples using DESeq2 using the parameters log2(fold change) > 0.5, Padj < 0.05, and standard error estimate for the log(fold change) standard error (lfcSE) of 1. Genes, fold change, and the Padj values for all DEGs can be found in Supplemental Table 1. (D) GO terms associated with the upregulated and downregulated genes. The terms are plotted according to their Padj value of enrichment. (E) TPM of 3 well-known SMA modifier genes, NCALD, NAIP, and PLS3. Each data point represents a single patient. The means of each group were compared with a 2-sided Student’s t test. *P < 0.05. NS, not significant (P > 0.05).

As is the case for complex diseases, rare monogenic diseases can also be modulated by further genetic factors, giving rise to variability in the phenotypes presented (31). Several genetic modifiers have been described for SMA, and a handful of these have been validated in human samples (32–34). We observed an increase in NCALD expression in the SMA samples compared with controls, despite a broad range of its expression (Figure 3E). Previous studies have reported that a decrease in NCALD expression was associated with milder SMA phenotype (35), suggesting that the increased expression of NCALD is a detrimental compensatory mechanism to the loss of SMN. By contrast, we did not observe an increase in NAIP or PLS3 levels (Figure 3E) (36).

MicroRNAs (miRNAs) have also been shown to have an important role in shaping the transcriptome. Many miRNAs have well-described roles in muscle development, and are aptly named myoMiRs (37–39). Perhaps the best established of these is the role of miR-206 in skeletal muscle maturation (40, 41). Furthermore, many studies in SMA have highlighted the essential role of changes of miRNAs to the SMA phenotype, and these miRNAs have also been proposed as biomarkers of nusinersen response and disease progression (38, 42–47). We used miRNA amplification and probe sets to measure the levels of 3 important myoMiRs that have been implicated in SMA and in muscle gene regulation: miR-24, miR-1, and miR-206. No difference was observed in the levels of miR-24 and miR-206. However, the levels of miR-1 were significantly decreased in the SMA samples (Supplemental Figure 2C). This is in line with a recent study that found that SMN, in combination with MYOD, can regulate the expression of miR-1 (42). However, we observed this decrease despite detecting similar levels of SMN protein (Figure 1H), suggesting additional levels of regulation. Of miR-1’s 1,349 target genes predicted by miRDB (see Methods), 52 were among our upregulated genes, including SLC24A2, SYT1, and SHANK2, which are involved in synaptic function (Supplemental Figure 2D and Supplemental Table 3).

Collectively, we established the landscape of altered transcripts in type II SMA muscles. Furthermore, the transcriptome of the SMA patients was more heterogeneous than that of the control individuals despite deriving from similar surgical conditions.

SMA type II muscle shows histological and transcriptional hallmarks of denervation. We hypothesized that the transcriptional heterogeneity could derive from the onset of symptoms and the timeline of treatment, as each patient might have a different degree of denervation. However, many of the clinical metadata about these samples was unavailable. Therefore, to better understand the dynamics at play, we referenced a recent atlas of mouse muscle denervation (48), which had scored the genes most characteristic of denervation in type I fibers. Using a set of 14 genes (Figure 4A), we developed a z-scored denervation grade based on the expression of these markers (see Methods) and scored each of the control and SMA samples. The homogeneous scoring of the control samples compared to the variability in the SMA samples mirrored what we saw in the whole transcriptome (Figure 4B). We could also observe this same trend in the histology, where SMA patients with a score closer to zero, such as SMA8, showed a better formed fiber structure by H&E, similar to what was seen in the controls (Figure 4C).

SMA type II muscle shows histological and transcriptional hallmarks of denervation. (A) Schematic of how the denervation score was calculated for each sample, using the expression (in TPM) of the highlighted genes. MF, myofiber. (B) The denervation score for each sample represented on a number line. (C) H&E images of selected control and SMA samples. (D) Cytochrome oxidase activity staining. Fibers with dark brown staining represent areas with high enzymatic activity, generally associated with type I fibers. Example fibers are highlighted by black arrows for those with dark staining (type I) and light staining (type II). (E) Sirius red–stained images of control and SMA muscle samples. Red staining occurs predominantly on areas with large amounts of collagen I, i.e., fibrotic areas. Scale bars: 100 μm (C–E).

Denervation can also affect fiber type in the muscle or cause a shift between fiber types. Previous studies on SMA patients have found a loss of type II glycolytic fibers (49), which mirrors what is seen in mouse models of SMA (11, 50, 51). However, in all these studies, muscle with a higher starting fraction of type II fibers were investigated. By contrast, studies have shown that in paravertebral muscle, 74% of the fibers are type I (slow) in the superficial and deep thoracic regions and that in the lumbar region, and 57% of the fibers are type I in the superficial muscles versus 63% type I in the deep muscles (52).

To assess changes in fiber type in our cohort, we employed 2 methods. Due to the quality of the muscle sections, we were not able to perform classic antibody-based IHC or immunofluorescence methods to detect fiber types. Therefore, we utilized a bulk RNA-seq deconvolution method, where bulk RNA-seq data can be used to predict the fiber type in the original tissue sample, based on the expression profiles of type I (slow) and type II (fast) muscle fibers derived from single-cell sequencing profiles (53). In accordance with the previous findings, fiber type deconvolution showed that in control samples, we had an average of 66% type I fibers (Supplemental Figure 2E). Similarly, SMA samples had an average of 69% type I, suggesting that RNA expression of overall fiber type markers was not highly affected. Indeed, plotting the changes in myosin expression showed a global decrease in abundance compared with controls, but not in type (Supplemental Figure 2F). Next, we used cytochrome oxidase staining (54). Cytochrome c oxidase is an enzyme that is found in the electron transport chain (ETC) in mitochondria. Intense dark brown is associated with type I fibers due to their large number of mitochondria. In control samples, most fibers stained dark brown, indicative of the expected predominance of type I fibers (Figure 4D). In the SMA samples, however, we observed a wide heterogeneity of staining patterns, with some that were highly similar to controls (patients 4 and 8), and other slides with no staining (SMA3, -5, and -7) (Figure 4D). Overall, this generally followed the trend of the denervation scoring, although an important caveat is that storage conditions can affect enzymatic activity.

As denervation is associated with muscle fibrosis (55), we sought to assess the degree of muscle fibrosis using Sirus red staining. In 3 of the samples with little to no fiber structure, we noticed especially prominent Sirius red staining (Figure 4E; SMA3, -5, and -7), while others had staining very similar to controls (Figure 4E, SMA8). We quantified and scored the fibrosis staining (Table 2) and found a correlation between the denervation and fibrosis scores (Supplemental Figure 2G), but not with the amount of full-length SMN transcript (Supplemental Figure 2H).

Summarized features of the SMA paravertebral muscle samples

In summary, we validated a transcriptional signature that correlated with classic histological signs of denervation, including changes in fiber size and increase in fibrosis, and saw that these characteristics were variable among our 8 samples, although the clinical reasons for this remain unclear.

Mitochondrial ETC complex expression and mtDNA copy number are altered in type II SMA muscle. The major pathways we observed downregulated in the SMA muscles samples were related to mitochondrial metabolic capacity, and particularly OXPHOS (Figure 3D). Using the MitoCarta gene list (56), we compared our differentially expressed genes and found 42 genes overlapping (Supplemental Figure 3A), with many of these related to proteins of the ETC that are responsible for OXPHOS (Supplemental Figure 3, B and C, and Supplemental Table 4). As we had previously observed that our small cohort is heterogeneous, we decide to calculate a mitochondrial OXPHOS score using a panel of 13 genes for each sample to better capture sample-to-sample variation (Figure 5A). As more expression of OXPHOS markers is associated with healthier muscle, given the large number of mitochondria present in type I fibers, SMA samples generally had negative z scores (Figure 5B). Although we hypothesized a negative correlation between the denervation score and the mito-score, samples with lowest mito-scores, for example SMA8 (score –1.7), did not always have the highest denervation scores (score–0.4) or the highest fibrosis scores (score 0.02) (Supplemental Figure 3, D and E, and Table 2). These discrepancies can highlight complexities in the biology, or difficulties in the sampling.

Mitochondrial electron transport chain (ETC) complex expression and mtDNA copy number are altered in type II SMA muscle. (A) Schematic of how the mito-score was calculated for each sample, using the expression (in TPM) of the highlighted oxidative phosphorylation (OXPHOS) genes. (B) The mito-score for each sample represented on a number line. (C) Western blots of control (n = 6) and SMA (n = 6) muscle samples incubated with the anti-OXPHOS antibody cocktail and vinculin as the housekeeping gene. The dotted line represents the 2 individual Western blots, imaged at the same time. Each band represents an ETC complex, which is labeled. Complex IV was not detected in either blot. Complex I was not detected in the blot on the right. The proteins designating each complex are as follows: complex I, NDUFB8; complex II, SDHB; complex III, UQCRC2; complex IV, MTCO1; and complex V, ATP5A. (D–G) Quantification of the blots in C. Each data point represents the normalized complex/vinculin value for each sample (n = 6 controls, n = 8 SMA samples). (D) Complex I, where n = 3 control and n = 4 SMA samples. (E) Complex II. (F) Complex III. (G) Complex V. The means of each group were compared with a 2-sided Student’s t test. *P < 0.05. NS, not significant (P > 0.05). (H) Relative quantification (qPCR) of the mtDNA copy number normalized to the numbers of gDNA copies using the β2-microglobulin gene in control (n = 5) and SMA (n = 8) paravertebral muscle gDNA samples. Each data point represents 1 sample. The means of the groups were compared with a 2-sided Student’s t test. *P < 0.01. (I) Same as in H, but from gDNA derived from the PBMCs of type III SMA patients. Each data point represents 1 sample (n = 22 controls, n = 14 SMA). Groups were compared with a 2-sided Student’s t test. ***P < 0.001.

To better assess the loss of the OXPHOS-complex members, we performed Western blotting on our patient samples with an antibody cocktail that can recognize 1 member of each complex (Figure 5C). Likely due to its hydrophobic nature, we were consistently unable to detect complex IV (mitochondria-encoded protein MTCO1) in any of our samples. Quantification showed a decrease in complex I, II, and V expression in the SMA samples compared with controls (Figure 5, D–G).

Mitochondria are also associated with the metabolism of cholesterol, and accumulation of cholesterol in the mitochondria of the liver has been shown to impair OXPHOS (57). Metabolic dysfunction and abnormal fatty acid signaling have been previously described in SMA (58, 59), including the accumulation of lipids and cholesterol in some SMA mouse models (59). In line with these observations, in our cohort, we observed that the low-density lipoprotein (LDL) receptor–encoding gene, LDLR, was upregulated in SMA samples (Supplemental Figure 3F). This receptor binds to LDLs, which are the primary carriers of cholesterol in the blood. SMA patients, across all subtypes, are known to have signs of dyslipidemia, including increased blood LDL levels, and total cholesterol levels in the plasma and other tissues have been observed in some severe mouse models of SMA (59). More recently, abnormal cholesterol accumulation was found in the dystrophic muscles of DMD patients and mouse models (60), and more studies are uncovering the role of skeletal muscle and fiber type as a global modulator of cholesterol and other lipids (61). Therefore, we decided to assess the total and free cholesterol using the Cholesterol Ester-Glo bioluminescence assay system from the lysates derived from the same muscle samples. However, no differences were observed between total, free, or esterified cholesterol (Supplemental Figure 3G).

We next sought to test the number of mitochondrial DNA (mtDNA) copies in the SMA muscle samples. mtDNA copy number is used as a surrogate measure for the number of mitochondria, when such measurements are not possible. Using a previously established protocol for qPCR quantification of mtDNA copy number (62), we measured mtDNA copy number from the extracted DNA of the same samples used for sequencing. We observed that, globally, mtDNA copy number was significantly lower in SMA muscle than in control samples (Figure 5H). Furthermore, the amount of full-length SMN transcript was correlated to the mito-score (Supplemental Figure 3H). To test whether this was a phenomenon that was more broadly true in SMA pathophysiology, we performed the same measure on DNA extracted from the circulating PBMCs of type III patients, all of whom had been treated with nusinersen and risdiplam, and likewise noted a significant decrease in mtDNA copy number (Figure 5I). Collectively, these results suggest that mitochondrial loss is a property of SMA tissues.

This led us to question whether mitochondrial regulation was a byproduct of the loss of SMN. To test this, we utilized immortalized myoblasts derived from SMA patients (type I and II) or healthy controls (Supplemental Figure 3I). However, we did not see a difference in mtDNA copy number, despite a difference in SMN expression (Supplemental Figure 3J). This is true both among the SMA myoblasts, but also among different control cell lines that had different expression levels of SMN1/2 (Supplemental Figure 3I). To test whether this discrepancy with the in vivo muscle samples was due to the immaturity of the myoblasts, we performed a serum-free differentiation for 7 days to generate myotubes. It was been previously described that myoblasts increase their mtDNA copy number and number of mitochondria as part of the differentiation processes (63), and we could recapitulate this finding in our control myoblasts (Supplemental Figure 3J). The SMA myoblasts were also able to increase their mtDNA copy number to the same degree as the controls (Supplemental Figure 3J). The results mirror what was previously seen in myoblasts differentiated from SMA ESC lines, where copy number did not change, although ATP production capacity was altered (64). Taken together, these data suggest that mtDNA copy number changes are not an intrinsic feature of SMN loss, but rather a downstream consequence of muscle damage, due to either the denervation or changes that occurred to muscle maturation.