DMD genotype correlations from the Duchenne Registry: Endogenous exon skipping is a factor in prolonged ambulation for individuals with a defined mutation subtype

Abstract Antisense oligonucleotide (AON)‐mediated exon skipping is an emerging therapeutic for individuals with Duchenne muscular dystrophy (DMD). Skipping of exons adjacent to common exon deletions in DMD using AONs can produce in‐frame transcripts and functional protein. Targeted skipping of DMD exons 8, 44, 45, 50, 51, 52, 53, and 55 is predicted to benefit 47% of affected individuals. We observed a correlation between mutation subgroups and age at loss of ambulation in the Duchenne Registry, a large database of phenotypic and genetic data for DMD (N = 765). Males amenable to exon 44 (N = 74) and exon 8 skipping (N = 18) showed prolonged ambulation compared to other exon skip groups and nonsense mutations (P = 0.035 and P < 0.01, respectively). In particular, exon 45 deletions were associated with prolonged age at loss of ambulation relative to the rest of the exon 44 skip amenable cohort and other DMD mutations. Exon 3–7 deletions also showed prolonged ambulation relative to all other exon 8 skippable mutations. Cultured myotubes from DMD patients with deletions of exons 3–7 or exon 45 showed higher endogenous skipping than other mutations, providing a potential biological rationale for our observations. These results highlight the utility of aggregating phenotypic and genotypic data for rare pediatric diseases to reveal progression differences, identify potentially confounding factors, and probe molecular mechanisms that may affect disease severity.

pediatric muscular dystrophy. Single or multiple exonic deletions and duplications account for 80% of mutations that cause DMD and the allelic disorder Becker muscular dystrophy (BMD; Bladen et al., 2015). Age at loss of ambulation (LOA) is variable in DMD, but typically, steroid-naïve boys lose independent walking ability between 9 and 11 years while those treated with corticosteroids ambulate on average for an additional 2-3 years (Angelini et al., 1994;Griggs et al., 1991Griggs et al., , 1993Henricson et al., 2013;Mendell et al., 1989;Wang et al., 2014). Steroid benefit has been clearly demonstrated in clinical trials, meta-analysis, and multiple natural history studies, and it has been the main drug intervention in DMD to date (Mendell et al., 1989).
However, the effect size for each of these AONs is not clear.
Aside from the "reading frame hypothesis," which demonstrated out-of-frame DMD exonic deletions typically caused DMD while inframe deletions produced the milder BMD (Aartsma-Rus et al., 2006;Koenig et al., 1989), no concrete rules correlating frameshifting deletions in the mutational hotspot region and disease severity have emerged. However, exceptions to the reading frame rule exist: for instance, out-of-frame deletions of exons 3-7 sometimes result in a BMD phenotype, while in-frame deletions of exon 3 can exist in patients with a DMD phenotype (Kesari et al., 2008;Koenig et al., 1989;Tuffery-Giraud et al., 2009). Recently, some DMD individuals with mutations amenable to exon 44 skipping were observed to have a higher rate of revertant fibers and a larger number of trace dystrophin positive fibers relative to muscles from exon 51 amenable DMD boys (Lourbakos et al., 2011). Consistent with this description, several studies have indicated that exon 44 skip amenable DMD boys have an overall better functional outcome with higher age at LOA relative to typical DMD (Bello et al., 2016;Pane et al., 2014;van den Bergen, Ginjaar, Niks, Aartsma-Rus, & Verschuuren, 2014).
We sought evidence of correlation between age at LOA and genetic mutation subgroups in a large cohort of predominantly U.S. DMD patients using data available from the Duchenne Registry, a large patient-powered self-report registry for individuals or families affected by Duchenne and BMD (Rangel, Martin, & Peay, 2012;Wang et al., 2014). We data mined patient-reported parameters of age, ambulatory status, age at LOA, corticosteroid usage, and genetic mutation from 3,383 participants in the registry and corrected for the potential confounding effects of steroid usage. Patients with mutations correctable by skipping exon 8 or 44 ambulated significantly longer than deletion mutations correctable by skipping of exons 45, 50, 51, 52, 53, or 55, exon duplications, or

Data collection
The Duchenne Registry is an online self-report registry (www. duchenneregistry.org) for individuals and families affected by Duchenne and BMD (Rangel et al., 2012). Participants respond to questionnaires for specific topics: diagnosis, muscle function, corticosteroid, cardiac, respiratory, family history, and genetic testing.
Genetic testing reports are reviewed by certified genetic counselors and annotated in a standard format. We downloaded the deidentified October 2016 freeze of the Duchenne Registry dataset which contains responses from 108 countries. Preliminary quality control and filtering were performed as previously described (Wang et al., 2014). Briefly, we filtered the data for males who reported valid and consistent ambulation status, corticosteroid usage/nonusage, and genetic mutation results confirmed by submission of genetic report to the Duchenne Registry genetic counselors. Participants were limited to residents of member nations of the Organization for Economic Cooperation and Development to ensure comparable levels of health care quality and access. For this analysis, genetic testing, corticosteroid, and muscle function modules were assembled per individual resulting in 1,913 complete profiles ( Figure 1). Patients were then stratified into separate groups according to either the predicted exon skip necessary to produce an in-frame transcript (exon 8, 44, 45, 50, 51, 52, 53, or

Collection, isolation, and propagation of dermal fibroblasts and myoblasts
Skin punches were obtained with informed consent from patients of the CDMD under University of California Los Angeles IRB-approved protocol (#11-001087). Isolation of dermal fibroblasts followed published protocols (Karumbayaram et al., 2012). Cells were reprogrammed using a tamoxifen-inducible MyoD overexpression system as previously described (Kendall et al., 2012) to create induced directly reprogrammed myotubes (iDRM). Primary myoblasts were purified from 50-100 mg of tissue obtained by needle muscle biopsy of vastus lateralis. Each core biopsy was dissociated into 1-mm pieces in a 1:1 MIX of dispase (1.5 U/mL)/collagenase (1,000 U/mL). After 20 minutes at 37 • C, chunks were triturated and passed through a 70-m cell strainer. Cell suspension was centrifuged at 1,200 rpm for 4 minutes, and the pellet was resuspended in 10 mL growth media (Nutrient Mixture F-10 HAM with 20% FBS and 1% pen/strep) and plated into a T75 flask. After 1 hour of preplating, the cell suspension was placed into a new T75 flask, which was considered to be enriched in myoblasts. All derived cells are assigned a patient unique study ID to allow inclusion of clinical data including mutation type and clinical progression.
Primary myoblasts were cultivated in growth media as described above but supplemented with 5 ng/mL of bFGF starting at day 3. The media were changed every 3 days until confluence was reached. At confluence, media were exchanged for skeletal muscle differentiation media (Promocell) for 7 days before being harvested in TRIzol for RNA isolation (Ambion).

RNA isolation, PCR, and quantification
Total RNA was isolated using the Purelink RNA mini kit (Ambion). Percentage of exon skipping was calculated as the molar amounts of skipped PCR product skipped PCR product + unskipped PCR product × 100. For cell lines with a deletion of exons 3-7, primers were used as previously described (Fletcher et al., 2012).

Immunofluorescence on cultured myotubes
Primary myoblasts were grown as described above. After induction of differentiation for 7 days, cells were fixed with acetone for 7 minutes at −20 • C. Dystrophin was detected using MANDYS8 (directed against the central rod domain, sc-58754, Santa Cruz Biotechnology) at a dilution of 1:100. Secondary goat anti-mouse (A32731, Life Technologies) was used at a dilution of 1:500. Images were obtained using a Zeiss microscope at a 20× magnification and processed with Axiovision software and/or ImageJ software.

Muscle biopsy immunohistochemistry
Muscle biopsies were flash frozen in liquid nitrogen-cooled isopen- with Permount. Images were obtained as for cultured myotubes. were not significantly different and were merged. All subjects were currently using corticosteroids therapy (e44 skip; N = 74) showed a difference in age at LOA by the log-rank test (P = 0.035), with a striking median age at LOA of 20 years, in contrast to the median age at LOA of 13 for the remainder of the cohort. The large majority of boys amenable to exon 8 targeted therapy (N = 18) were still ambulatory at age 20 (P < 10 −5 ), and most of these boys have exon 3-7 deletion, which is often affiliated with a BMD phenotype (Muntoni et al., 1994). Consistent with previously reported cohort studies of DMD, the largest group consisted of individuals amenable to exon 51 therapy (e51 skip; N = 106). This group has a more severe disease progression with an observed earlier median age at LOA at age 12 (P = 0.04). No other DMD mutation subgroups showed significant differences in age at LOA (Table 1, Supporting Information Figure S1).

Kaplan-
Among mutations amenable to exon 44 skip therapy, 64% were due     Our results are consistent with a model in which endogenous exon skipping in DMD transcripts results in a low level of in-frame mRNA and production of low levels of rescued dystrophin protein, which contributes toward reduction in disease severity as measured by delay in age of LOA in exon 44 or 8 skippable patients.

DISCUSSION
We previously reported the utility of the Duchenne Registry patient registry to observe effects on age at LOA due to single and combination  (Wang et al., 2014). Furthermore, we demonstrated Duchenne Registry participants were typical of the general DMD population in terms of mutation status, mutation type, or age at diagnosis. Thus, this valuable dataset is highly applicable for observing differences in mutation type within DMD based on age at LOA and serving as a reference population demonstrating typical disease course.
The Duchenne Registry records current medication, mutation type, and age at LOA for a large number of participants, allowing us to observe differences in disease severity based on a limited group of mutations potentially amenable to exon skipping. As a web-based platform, the Duchenne Registry has a relatively low barrier of entry for participants, allowing for a large number of participants. In the available data from the Duchenne Registry, we were able to retain a sub- groups (Bello et al., 2016;van den Bergen et al., 2014). Consistent with the results presented here, both strongly support a delay in age at LOA among individuals amenable to targeted skipping of exon 44.
We estimated the effect of possessing a DMD mutation that is amenable to exon 44 skipping to have a hazard ratio (HR) of 0.54.
This indicates that the impact of these mutations on the phenotype is roughly comparable with corticosteroid treatment. HRs for deflazacort and prednisone were 0.31 and 0.62, which is consistent with our previous finding that deflazacort tends to have a stronger effect at delaying age at LOA than prednisone (Wang et al., 2014). This is similar to Bello et al. (2016), who estimated HR for exon 44 skippable mutations to be 0.34. Their data also showed deflazacort to be more beneficial in delaying age at LOA compared to prednisone (0.34 vs. 0.22) and that the genetic effect was similar to corticosteroid treatment. These consistencies support the overall utility of the patient self-report registry model.  (Deburgrave et al., 2007;Kesari et al., 2008). One biological explanation is that more mildly affected patients may have a higher frequency of revertant fibers due to spontaneous skipping of an exon that generates in-frame transcripts and resulting functional dystrophin (Dwianingsih et al., 2014;Prior et al., 1997). Although an increased rate of revertant fibers has been demonstrated in some cases (Lourbakos et al., 2011), the true incidence is not known. We found evidence of endogenous exon skipping in patient-derived iDRMs and myotube cul- which around 10% of DMD mRNA demonstrates endogenous skipping of exon 8, which is a potential molecular mechanism for why this group of patients is more mild.
Evidence of a more severe than typical disease course in the exon 51 amenable deletion group has not been reported elsewhere, and the molecular mechanism is not clear at this point. Kaplan-Meier analysis of individual subgroups within exon 51 amenable mutations revealed that exon 49-50 deletions appeared to be particularly severe with a median age at LOA of 11 years versus 13 years for the rest of the cohort (P = 0.008). However, this result was not significant in the Cox analysis.
Based on these findings, the exon 8, 44, and 51 skippable subgroups may serve as inappropriate natural history controls for most exonskipping trials, but that other mutations amenable to exon 45, 50, 52, 53, 55, and some 51 or nonsense mutations are generally comparable and can reasonably serve as contemporary natural history controls.
Even in instances where restoration of dystrophin is not the intended effect, the underlying genetic mutation may need to be considered to prevent the confounding of genetic effects within the potential therapeutic effect being sought. Thus, more natural history data may be necessary to improve the sensitivity of ongoing clinical trials for DMD in order to appropriately factor DMD mutation effects into clinical trial designs, and exact mutation type in the DMD gene should be considered in data interpretation. It is possible that restriction in clinical trial entry based on mutation type could reduce subject variability and enhance the ability to observe a study drug effect.
Because much of the Duchenne Registry data are based on patient and parent self-reports, there is sometimes concern regarding accuracy relative to natural history studies derived from academic medical centers where participants are recruited, phenotyped and followed longitudinally by expert clinicians. Although care has been taken to rule out inconsistencies and errors in responses (for instance, DMD genetic reports are reviewed and inputted by certified genetic counselors), misunderstandings and errors can occur in the survey questionnaire. We also note that the retrospective nature of this study and its internet-based recruitment methods may also influence the demographics of respondents, which may tend toward more technically literate families. However, the remarkable consistency of the findings reported here and the recently reported support for the observed superiority of deflazacort relative to prednisone (Griggs et al., 2016;Wang et al., 2014), serve to further validate and increase confidence in this important resource.
Using a large patient self-report registry, we have linked subgroups of genetic mutations to a critical and irreversible physiological milestone in DMD. These results indicate further study of exon 3-7 and exon 45 deletion subjects, which both show delayed age at LOA, will likely produce further understanding of the structure and function of aberrantly affected DMD transcripts and naturally occurring mitigating factors. Furthermore, the lack of equivalence in disease progression between different mutation subgroups that is apparent with the large sample set suggests caution when used for clinical trials or as natural history or external contemporary controls. However, the consistency of the age at LOA across multiple mutation types provides a powerful way to determine if long-term administration of new therapeutics is causing deviation from expected disease course. This will become increasingly important as drugs such as Exondys51 gain approval through the accelerated approval pathway based on a reasonably likely to predict clinical benefit standard.