Analysis of extracellular mRNA in human urine reveals splice variant biomarkers of muscular dystrophies

Urine contains extracellular RNA (exRNA) markers of urogenital cancers. However, the capacity of genetic material in urine to identify systemic diseases is unknown. Here we describe exRNA splice products in human urine as a source of biomarkers for the two most common forms of muscular dystrophies, myotonic dystrophy (DM) and Duchenne muscular dystrophy (DMD). Using a training set, RT-PCR, droplet digital PCR, and principal component regression, we identify ten transcripts that are spliced differently in urine exRNA from patients with DM type 1 (DM1) as compared to unaffected or disease controls, form a composite biomarker, and develop a predictive model that is 100% accurate in our independent validation set. Urine also contains mutation-specific DMD mRNAs that confirm exon-skipping activity of the antisense oligonucleotide drug eteplirsen. Our results establish that urine mRNA splice variants can be used to monitor systemic diseases with minimal or no clinical effect on the urinary tract.

control subjects. (a) We used particle imaging and tracking analysis software to measure extracellular particle size and concentration in urine and serum samples from DM1 (N = 12 urine, 10 serum) and UA control (N = 9 urine, 8 serum) subjects. Representative traces of particle size from DM1 and UA subjects in urine (left) and serum (right) are shown. Note that the tracing for vehicle (saline) overlies the x axes. (b) Mean particle size from each individual sample in urine (left) and serum (right). **** P < 0.0001; *** mean difference 51.2, 95% CI of difference 21.9 to 80.6; one-way ANOVA. (c) Optical density spectra of exRNA from urine (N = 36 DM1, 28 UA) and serum (N = 23 DM1, 19 UA) samples using a microvolume spectrophotometer (Nanodrop). Vehicle (water) served as reference. The peak of 270 nm reflects an artifact of residual Trizol that was used to purify the RNA 1 .      Supplementary Figure 12 Alternative splicing in human bladder, urothelial cells, kidney, and skeletal muscle. showed no evidence of an alternative splice event, consistent with previous reports that DMD exon 51 is constitutively spliced 8,9 . Sequencing of PCR products from urine exRNA, urine cells, and muscle (lower).   DM1  14  26  40  40  39  5 -81  DM2  2  2  4  59  59  47 -73  DMD/BMD  0  17  17  20  20  7 -49  UA  22  7 29 n/a n/a n/a Total 38 52 90 n/a n/a n/a

Exon inclusion % (range) Exon inclusion %; mean ± SD (range) Exon inclusion % Various muscles (biopsy + autopsy) Vastus lateralis (biopsy) Tibialis anterior (biopsy)
Commercially  Table 5 Summary of alternative splicing patterns analyzed by RT-PCR in human skeletal muscle tissue from DM1 or UA subjects, as previously reported 2,15 . The Savkur study examined splicing in a mixture of biopsy and autopsy specimens from various muscles. In the Nakamori study, vastus lateralis and tibialis anterior muscle biopsies from DM1 and UA subjects in non-overlapping cohorts were examined by RT-PCR, and a combination of muscle biopsy and autopsy specimens identified hundreds of candidate splicing defects in DM by microarray. Splicing of three transcripts, INSR, CLCN1, and ATP2A1, were analyzed in both the vastus lateralis and tibialis anterior biopsies, and exon inclusion percentages are displayed as mean ± SD, with the range of values from high to low shown in parentheses. The remaining values were adapted from scatter plot graphs and are displayed as a range.
Note that the difference of exon inclusion between groups in the tibialis anterior was greater than the difference between groups in the vastus lateralis, which was attributed to the greater clinical involvement of tibialis anterior than vastus lateralis in DM1. Also note that, for most transcripts, exon inclusion showed a wider dynamic range in the DM1 group than the UA group. In the current study, we used commercially available total RNA isolated from skeletal muscle (which muscle is unknown) to measure splicing of INSR,   Supplementary Fig. 6).

Supplementary Discussion
The potential to evaluate exRNA splicing outcomes as pharmacodynamic biomarkers in urine has the advantage of being non-invasive and can be repeated routinely over the course of treatment to evaluate drug target engagement. Our longitudinal monitoring of urine exRNA splicing patterns in DM1 and UA subjects over two years demonstrates feasibility and reliability of the assay. We also demonstrate that alternative splicing analysis by ddPCR using unique primer probe sets for each splice event offers the potential for greater quantitative accuracy of the larger exon inclusion products, and perhaps can increase throughput in a clinical trial or laboratory-based setting. Due to the need for general anesthesia and the absence of a therapeutic benefit, muscle biopsies generally are avoided in children with DM1.
Consequently, detailed study of splicing outcomes in children with DM1 remains an unmet medical need.
Urine exRNA should enable comprehensive non-invasive investigation of splicing outcomes in children with DM for the first time, facilitate clinical trials in these patients earlier, and perhaps enable convenient titration of dose. The shared pathogenic mechanism of alternative splicing misregulation in DM1 and myotonic dystrophy type 2 (DM2) 2 support our results that splicing in urine exRNA also may be useful for monitoring disease activity in DM2 patients.
Our finding that urine exRNA splicing outcomes of some transcripts demonstrate an early transition in asymptomatic patients, while other transcripts correlate with DM1 symptoms, is consistent with a previous study that found splicing of several transcripts in muscle biopsies showed a similar pattern 2 . Curiously, though, for three transcripts these properties appear to show the opposite pattern in urine exRNA as compared to tibialis anterior muscle biopsies in the previous study. For example, splicing of INSR in urine exRNA shows a wide dynamic range and appears related to symptom severity, but in tibialis anterior muscle biopsies showed a narrow dynamic range and an early transition in individuals with minimal symptoms. By contrast, splicing of MBNL2 and MBNL1 shows a narrow dynamic range and early transition in urine exRNA of asymptomatic individuals, but demonstrated a wide dynamic range and graded changes related to weakness of the tibialis anterior muscles biopsied in the previous study. This further argues against muscle tissue being the primary source of splice products found in urine exRNA.
The reason that splicing outcomes are such powerful biomarkers of DM1 has to do with the disease mechanism involving mis-regulated alternative splicing 16 , together with ratiometric measurements of exon inclusion/exclusion being inherently more sensitive than unidirectional changes that are typical of most biomarkers.
Exosomes and microparticles/microvesicles (MVs) are types of EVs distinguished by both size and origin.
Exosomes (30 -100 nm diameter) are released from multivesicular bodies when they fuse with the plasma membrane, while MVs (100 -1000 nm diameter) form by exocytic budding 3 . The mean size of EVs in our study suggests that the majority in urine were MVs, while most EVs in our serum samples were exosomes, suggesting the possibility that the mRNA splice products we are measuring are more prevalent in MVs than exosomes. Indeed, a recent study found that MVs contain a more accurate peripheral representation of glioblastoma cellular mRNA content than do exosomes 17 . Here we find that splice patterns of several transcripts in urine exRNA are more similar to those in kidney tissue and urothelial cells as compared to muscle tissue, suggesting that the exRNA found in urine may represent a pool from multiple different cell types along this urinary route. Our results showing ASO knockdown of Dmpk expression in mouse urinary tract tissues are consistent with earlier pre-clinical studies that demonstrated therapeutic ASO activity by inducing target knockdown and exon skipping in kidney tissue of mice and non-human primates 7,18 , and together suggest that ASOs could have similar effects in human kidney and other tissues lining the urinary tract that release exRNA into the urine.
Our finding that inclusion of MAP3K4 exon 17 was different between males and females within both the DM1 and UA groups was unexpected, as was the significant correlation of exon inclusion with the CTG repeat length in DM1 males. The biological effect of a shift in the exon 17 inclusion percentage is unknown 19 , although it is interesting to note that the sequence of this alternative exon includes ten CUG repeats, and that Map3k4-dependent signaling has been implicated in sex determination and testis development in mice 20 .
In the clinical trial that led to accelerated approval of eteplirsen, immunofluorescence analysis of muscle biopsies demonstrated an increase in the number of muscle fibers expressing detectable dystrophin protein from a pre-treatment baseline of 1.1% to 17.4% after 180 weeks of treatment 21 . Exon skipping at the mRNA level was examined in these muscle biopsies using qualitative end-point RT-PCR and sequencing of PCR products; therefore, precise quantification of exon skipping activity in these biopsies is unavailable. In our study, it was surprising that exon 51 was mostly skipped in urine exRNA and urine cells of both individuals who are being treated with eteplirsen, and suggests that eteplirsen has greater uptake in cells and tissues lining the urinary tract than in deltoid or biceps muscle tissue, which were the two muscles that were examined in the clinical trials that led to accelerated approval of eteplirsen. This may be explained by pharmacokinetic properties of uncharged phosphorodiamidate morpholino ASOs like eteplirsen, which are excreted rapidly after systemic delivery, with little or negligible muscle tissue uptake 22 .
The controversy surrounding the FDA accelerated approval of eteplirsen 21 indicates that realization of the full potential of exon skipping for DMD will require the development of new ASOs that outperform eteplirsen. For novel ASOs that engage their mRNA target in muscle and urinary tract tissues with similar efficacy, such as conjugated morpholinos 18 or ASOs that have phosphorothioate backbones ( Supplementary Fig. 14), splicing measurements in urine RNA have the potential to correlate with drug activity in muscle tissue. Similarly, the absence of detectable pharmacological activity in urine RNA also could serve as an early indicator that the dose may be too low, or that the candidate drug eventually will fail, and thereby could help save valuable resources. These are important and potentially overlooked advantages of non-invasive monitoring of drug target engagement.
Collectively, our demonstration of disease-specific splice variants in urine exRNA suggest the value of biofluids as a means to identify novel splice variants that may help correlate genotype with phenotype for a number of diseases for which non-invasive biomarkers are unavailable. For example, in patients with Hutchinson-Gilford progeria syndrome (HGPS), point mutations in the LMNA gene activate a weak splice site in exon 11 that shortens the transcript and produces a truncated progerin protein 23 . ASOs that reduce use of this weak splice site are being evaluated as strategy to treat HGPS 24 . The presence of LMNA exon 11 in urine ( Supplementary Fig. 2) suggests the potential use of exRNA to monitor drug effects in these patients as well. Our findings also support the study of exRNA from urine, serum, or CSF as a biomarker replacement for tissue biopsies in other diseases with altered mRNA splicing, including limb girdle muscular dystrophy type 1B, spinal muscular atrophy, and amyotrophic lateral sclerosis [25][26][27] .