First cases report
Two four-month-old male Labrador retriever littermates owned by a breeder were referred by their veterinarian to the neurology consultation of the Alfort school of veterinary medicine. The puppies presented with gait stiffness, exercise intolerance, and marked palmigradic and plantigradic posture (Fig. 1A). At the clinical examination, normoreflexia and absence of proprioceptive defect were observed. The two puppies had markedly elevated serum creatine kinase (LRMD1: 83000 UI/L; LRMD2: 30000 UI/L), and the electromyogram (EMG) revealed normal nerve conduction velocities, but spontaneous activity in all the tested muscles (mainly complex repetitive discharges, Fig. 1B). Muscle biopsies from the biceps femoris muscle were sampled on both dogs and revealed lesions of active necrosis, regeneration, endomysial fribrosis and few inflammatory foci with calcifications (Fig. 1C). The phenotype of these two dogs, together with the high CK, the myopathic EMG profile and the histological lesions were evocative of a muscular dystrophy process. An immunostaining of the biopsies showed no reactivity with an anti-dystrophin antibody directed against the central rod domain (Dys 1, Fig. 1D). A multiplex western blot was also performed (Fig. 1E), confirming the absence of Dp 427 dystrophin, and the presence at expected size of other proteins implicated in human limb-girdle muscular dystrophies: dysferlin, γ-sarcroglycans, and calpain 3. Another litter was obtained by the breeder from the same parents, and once more two affected males were born, reinforcing the hypothesis of an X-linked transmission. The breeder kindly gifted to our research unit these two affected males (LRMD3 and 4) as well as an unaffected daughter from the same parents possibly carrying the causal mutation, allowing us to establish a LRMD colony in the same facilities as a pre-existing GRMD one.
LRMD colony establishment
A first litter was obtained by crossing the presumed female carrier and one of the affected males (LRMD4), and the subsequent litters were obtained by crossings between the descendant dogs (carrier females x affected males). The inbreeding of the colony was high (ranging from 25% for LRMD5 and 6, to 41% for LRMD11 to 14). Over a period of 6 years, ten litters were obtained (Fig. 2). A total of 14 LRMD dogs (9 males and 5 females) survived the neonatal period and were followed-up.
Main phenotypic features
At birth, the LRMD dogs had a weight comparable to their healthy littermates, but rapidly showed for most of them striking difficulties to suck and needed intensive nursing during their first days to ensure their survival (feeding by oro-gastric gavage, temperature monitoring and stay in incubator if needed). Despite these cares, around 50% of the LRMD newborns died within their 48 first hours (Fig. 3A) following a paroxystic weakness episode associated for most of them by a severe dyspnea. Myoglobinuria could be assessed in some of these puppies, as well as very high serum CK values (> 100000 UI/L) and hyperkalaemia. The rhabdomyololysis was confirmed in these animals, with a selective involvement of some muscles such as the diaphragm, the tongue or the sartorius cranialis (Fig. S1), confirming that this mortinatality syndrome reproduces the neonatal fulminating form well described in the GRMD dog model .
After few days of intensive nursing, the surviving LRMD puppies became able to suck by themselves and only exhibited growth retardation in comparison to their healthy littermates. At around the age of 2 months, the LRMD puppies showed a stiff gait and difficulties to jump over an obstacle. In the subsequent months, they became stiffer, and rapidly developed posture abnormalities resembling those seen in GRMD dogs notably marked pelvic verticalization, palmigrady and plantigrady. After six months of age most of the LRMD dogs were unable to run and used a walking gait. However, most of them remained ambulant, except one dog, (LRMD8) who completely lost ambulation at 6 months of age and was therefore euthanized. His littermate (LRMD9) also developed a severely compromised locomotion.
Respiratory and digestive signs
At a young age (2–3 months) the LRMD dogs began to show signs of dyspnea mainly paradoxical thoraco-abdominal movements. In the following months, the dogs developed moderate-exercise induced polypnea, noisy breathing, and in some cases intermittent cyanosis accompanied by an elevated serum bicarbonate concentration. Oro-pharyngeal dysphagia became a major feature of the LRMD from the age of 4 months, with a prominent aggravation in the following months. This severe dysphagia probably explained by the co-occurence of markedly reduced jaw opening, prominent macroglossia, retraction of the tongue, and weakness of the paryngeal muscles themselves, was observed in all the LRMD dogs, and was a cause for humane euthanasia for half of them, who were even unable to maintain a convenient hydration state. As a complication of this oro-pharyngeal dysphagia, LRMD dogs frequently developed aspiration bronchopneumonias, which could be successfully treated for most of them, but which caused the death of 4 out of the 14 affected dogs. On chest radiographs, the usual abnormalities found in GRMD dogs were seen: hiatal hernias, megaoesophagus, pectus excavatum, pulmonary hyperinflation associated with a diaphragm flattening .
The mean survival of these 14 LRMD dogs was 21.6 months, with a first quartile at 10.8 months and a third quartile at 31.4 months. The LRMD dog with the longest survival (LRMD7) died at the age of 103.5 months (8.6 years), probably following paroxystic cardiac arrhythmias (sudden death without significant findings at necropsy in a dog known to have prominent ventricular arrhythmias). A compared survival analysis was performed between LRMD and GRMD dogs and showed that LRMD dogs tended to have a better survival (log-rank test p = 0.01) despite close curve profiles between both colonies, and very distinct from the one of healthy dogs (Fig. 3B).
Among the studied biopsies, the general trend for a given muscle was a predominance of necrosis-regeneration lesions in dogs younger than 1 year-old. Inflammatory foci and calcifications were observed in some biopsies from 4- and 6-month-old dogs. Later on, the predominant lesions became the endomysial fibrosis and adiposis after two years (Fig. S2). In parallel, the CK values were highly fluctuating but remained roughly strongly elevated during the first year, and decreased to lower values in older LRMD dogs. Among the sampled muscles, the more affected were the diaphragm and the extensor carpi radialis, according to a pathological scoring (pathological index > 60%).
Functional evaluation and comparison with the GRMD model
The overall phenotypic characteristics of the LRMD dogs resemble those observed in GRMD dogs. In order to position the model in comparison to the “reference” GRMD dog model, a quantitative comparative study between both canine muscular dystrophies was performed, using the tools developed to evaluate GRMD dogs.
As a reflect of the disease progression described above, the clinical score rapidly increased during the first months of the studied LRMD dogs, and tended to progress slower or stabilize after 7–8 months. The same type of evolution was seen in the GRMD population (Fig. 4A). However when comparing more in detail both colonies, the LRMD dogs seemed to have a more rapid and homogeneous evolution in the very first months: at the age of 4 months the LRMD dogs had nearly significantly higher scores (p = 0.07) and the coefficient of variation of the clinical score was less in the LRMD colony (12% vs 34% in GRMD dogs). After this point, two subpopulations emerged, with two more severely affected LRMD dogs from a locomotor point of view (LRMD8 and 9), suggesting that as in the GRMD colony, there may be a severe form leading to a loss of ambulation in LRMD dogs . The less affected LRMD dogs seemed to have higher clinical scores than GRMD dogs in the 10–16 months interval. This was confirmed by a comparison between both colonies at an adult age (Fig. 4B), showing that LRMD dogs had significantly higher clinical scores (mean 67.5%, SD 12.1%) than GRMD dogs (mean 48.1%, SD 11.7%; p = 0.005).
The LRMD dogs that were gait-tested using accelerometry exhibited similar profiles as GRMD dogs compared to healthy dogs: decreased speed, stride length and frequency, decreased total power and increased medio-lateral relative power signing the waddling gait of these dogs. Among the three LRMD dogs longitudinally followed-up (Fig. 5), two showed values amongst the more affected GRMD dogs, and a decrease of the gait quality with age, notably with a dramatic increase of the gait waddle. The third dog (LRMD14) had rather preserved ambulation, attesting to the existence of an inter-individual heterogeneity in LRMD dogs. Despite this heterogeneity and the low number of animals followed-up, the total power, a very discriminating variable, was found significantly decreased at all tested ages (p = 0.035 at 4 months of age, p = 0.0001 at 6 months of age, p = 0.002 at 9 months of age), similarly to GRMD dogs (no significant difference between both models). The four LRMD dogs examined at an adult age obtained gait indices values overlapping those usually measured in the GRMD population and significantly different from the healthy one (p = 0.036 for the stride frequency to p < 0.0001 for the total power). No significant difference was found between LRMD and GRMD dogs, though the relative medio-lateral power was nearly significantly increased in LRMD dogs (p = 0.059, mean LRMD = 42% (SD = 10%), mean GRMD = 29% (SD = 13%)). When projecting these four dogs as supplementary individuals on a PCA plane constructed using healthy and GRMD adults as active individuals, the LRMD dogs projected in the GRMD cloud, attesting the gait characteristic similarities between both colonies (Fig. 5).
Two dogs could be muscle force-tested, at 4 and 6 months of age. At both ages, the muscle force appeared decreased compared to healthy dogs, at similar level as for GRMD dogs at 4 months of age, and at a more intermediary level at 6 months of age (Fig. 6). Muscle relaxation impairment is a feature of GRMD dogs, with an incomplete relaxation after a tetanic contraction or a twitch, at varying levels in function of the animals, and with a correlation to the severity of the phenotype. This feature was also found in the LRMD dogs, with an increase of this residual post-tetanic contraction at the 6 months of age.
The four adult LRMD dogs that underwent respiratory tests showed again similar patterns as GRMD dogs. They had a significantly caudally retracted diaphragm as assessed by the angle index (p < 0.001), at similar levels as GRMD dogs (no significant difference between both models) (Fig. 7A). The LRMD diaphragm was hypokinetic with a decreased range of motion (p = 0.005), but slightly less than GRMD dogs (p = 0.02) (Fig. 7B). The flow-volume loops analysis showed the same abnormalities as those described in GRMD dogs including a dramatic flow decrease at the end of expiration, quantified by the EF75/PEF ratio, which was significantly decreased in LRMD dogs (p = 0.0004) at similar levels as GRMD dogs (no significant difference between both models) (Fig. 7C). The PIF/PEF ratio was also dramatically decreased (p < 0.0001), as in GRMD dogs (no significant difference between both models), however the LRMD dogs tended to overlap with the lowest GRMD values regarding this flows ratio (Fig. 7D).
The dog examined at a young age (LRMD4, 5 months) exhibited hyperechoic lesions of the left ventricular free wall, but the measures performed using conventional echocardiography were within normal ranges (shortening fraction 47.5%). However the use of tissue Doppler imaging to analyze the radial motion of the left ventricular free wall revealed slightly decreased endo-epicardial gradient of velocity, a hallmark of early presymptomatic dilated cardiomyopathy reported in GRMD dogs . The dog examined at later stages (LRMD7) showed dilated cardiomyopathy with a marked decrease of the shortening fraction evolving with time (21.5% at 5.5 years of age and 10.3% at 7.5 years of age). This dog also showed frequent ventricular arrhythmias (left and right ventricular extrasystoles), which were suspected to have led to the sudden death of this animal one year after the last echocardiographic examination.
Identification of the causal mutation
RT-PCRs covering the dystrophin cDNA led to the amplification of all parts of the DMD transcript in LRMD dogs, with the exception of the sequence encompassing exons 20 and 21. Indeed, RT-PCRs between exons 10 and 20 and between exons 21 and 26 amplified normal products while no product could be obtained between exons 15 and 22 (Fig. 8A).
A Southern blot of the EcoRI-digested genomic DNA of LRMD dogs and using cDNA probes encompassing exons 18–24, exon 21 or exon 20 revealed abnormal bands (Fig. 8B), confirming at the DNA level a putative remodeled spot, mapped between exons 20 and 21 of the DMD gene.
The whole intron 20 (4.5 kb) was then explored using four overlapping couples of primers (Fig. 8C). The three first pairs covered the 3.2 kb 5’ segment of intron 20 and allowed amplification of amplicons at the expected size. By contrast, the fourth PCR using the F4 - R4 primers yielded no amplicon in LRMD dogs (Fig. 8C), pinpointing a sequence remodeling lying in the 1.6 kb of the 3’ region of intron. New primers designed to approach the putative mutation site allowed to correctly amplify the 640 bp segment covering the 3’ end of the intron, and sequencing of the amplicon confirmed a normal acceptor splice in LRMD dogs. However a 700 bp region (X:27,623,229 to X: 27,622,528) was delineated and remained non-amplifiable even using long-range PCR conditions. Altogether, these molecular data confirmed in LRMD dogs a gross genomic DNA rearrangement involving the nearly 3’ end of intron 20. In addition, when compared to a healthy unrelated Labrador retriever, a slight size difference was seen in the amplicon using the F1 – R1 pair of primers, which appeared longer in dogs from the LRMD colony whatever their clinical status (data not shown). Sanger sequencing showed normal donor splice site in LRMD dogs, but revealed the insertion, 960 bp downstream the 5’ extremity of intron 20, of 36 bp including a 21 bp polyA and the repetition of 15 bp of the adjacent normal sequence. This insertion explained the shifted size of the smallest band revealed on the southern blot using the exon 20 probe (Fig. 8B). We identified this insertion in all animals of the colony, including healthy dogs and thus concluded to a DMD-unrelated polymorphism that segregates in this line of Labradors.
The LRMD-causing mutation involving the 3’ region of intron 20 was then approached by RNA-sequencing to measure DMD transcription levels from both exons and introns using strand-specific libraries prepared from total RNA. This experiment revealed that minus strand transcription from the Dp427m promoter abruptly decreased within the F4 - R4 PCR region between exons 20 and 21 (Fig. 9A). Ectopic transcription was observed in a region normally located 2 Mb farther toward the centromere on the plus strand 100 kb proximal of TMEM47 gene, where this novel transcription continued for ~ 0.3 Mb (X:29,824,000–30,122,000) into a “no man’s land” of the X chromosome. The level and pattern of transcription across this ~ 0.3 Mb region was similar to DMD large introns, suggesting the causal mutation may be a 2.2 Mb inversion disrupting the DMD gene and involving TMEM47 (Fig. 9A).
In light of this characterization, a new PCR experiment was performed intending both to sequence the breakpoint flanking regions, and also to develop a LRMD mutation diagnostic test able to identify carriers, healthy and diseased dogs. A pair of primers (mutF - mutR) was designed flanking the presumed distant breakpoint, and the lack of amplification between them in the LRMD context confirmed the location of a breakpoint between these primers. The use of these primers in combination with primers located flanking the intron 20 breakpoint (F4 – R4) amplified a PCR product in LRMD but not in healthy dogs. The sequencing of these PCR products allowed the precise characterization of the LRMD mutation as a 2.2 Mb inversion of a region encompassing nucleotides 27,622,834 to 29,823,785 and which can be annotated : chrX:g.27,622,834_29,823,788 inv. This inverted region encompassed the entire TMEM47 gene without disrupting it. The sequencing of the two breakpoints revealed a 27,622,834 to 29,823,789 junction on the one hand (PCR F4-MutR), and a 27,622,824 to 29,823,785 junction on the other hand (PCR MutF-R4). This indicated that the inversion is associated with a 9 nt loss in intron 20 and a 3 nt loss in the distant region, probably attesting to the LRMD mutation probably occurred after two double strand breaks 2.2 Mb apart, and repair by non-homologous end joining after inversion of the released fragment. Based on PCR confirmation of the breakpoint from genomic DNA, close inspection of the RNA-seq data revealed several intronic reads that mapped across the 27,622,834 to 29,823,789 breakpoint and several spliced exonic reads that mapped from DMD exon 20 to multiple locations beyond the inversion breakpoint, confirming transcription and splicing from the DMD region. At last, a multiplex PCR was designed to propose a genetic test able to reliably identify LRMD, carrier, and healthy dogs, and confirmed to discriminate the three genotypes (Fig. 9B).
Characterization of the dystrophin expression in muscles
The dystrophin immunostainings performed on LRMD muscle biopsies using an antibody directed against the C-terminal part of the protein (Dys2) showed a faint but undoubted signal with a normal subsarcolemmal localization (Fig. 10A). The amount of Dys2-positive myofibers was evaluated on 32 skeletal muscle biopsies from 7 LRMD dogs at levels ranging from 0.2% (LRMD6, sartorius cranialis muscle) to 44.1% (LRMD1, tibialis cranialis), with a mean of 11.6% (SD = 12.3%). In order to better characterize this protein, monoclonal antibodies cross-reacting with different regions of the protein were used on serial sections. No staining of the Dys2 positive fibers was obtained using any of the antibodies cross-reacting with upstream regions of the dystrophin protein: the staining was negative for the N-terminal part (MANEX 1A), the first repeat of the rod domain (MANEX 1011 C), or for parts of the rod domain located downstream the mutation (Dys1, MANDYS 107 (Fig. S3)) (Fig. 10A). In some rare fibers however, a positive staining for the N-terminal part of the protein was found (Manex 1A + and Manex1011C+) but these fibers were negative for the C-terminal part (Dys2) (Fig. S3).
A western blot was performed using the Dys2 antibody cross-reacting with the C-terminal part of the dystrophin, in order to know the size of the expressed dystrophin. A band was visible in the two biopsies tested, at around 70 kD (Fig. 10B). Given the cross-reactivity of the protein specifically with an antibody targeting the C-terminal part of the protein, and the size of approximately 70 kD, we addressed the hypothesis of an expression of the Dp71 isoform. A RT-PCR using a forward primer designed in the specific first exon of the Dp71 confirmed an expression of this isoform in the three tested muscle biopsies (Fig. 10C) and was consistent with RNA-seq coverage data observed in the Dp71 region (Fig. 9A). No expression was found in the biopsy originating from a GRMD dog, suggesting that this Dp71 expression is not a compensatory mechanism in a context of canine dystrophinopathy, but rather a specificity of the LRMD context.
Analysis of two modifier genes, Jagged1 and LTBP4
In order to asses if the inter-individual heterogeneity observed in the LRMD colony could be related to polymorphisms in known modifier genes, the mutation in the Jagged1 gene promoter leading to ‘escaper’ phenotypes in dogs  was investigated in 5 LRMD dogs with diverse phenotypes, including the mildest and the oldest survivor. The sequence was identical in every tested LRMD dogs. None of these dogs harbored the G > T point mutation described in GRMD escapers.
Further, coding LTBP4 SNPs were analyzed, since polymorphisms in this gene have been described that strongly modulate the mdx mouse and human DMD phenotype [59, 60]. Eleven SNPs were found in the canine LTBP4 coding sequence, among which only 4 were responsible for amino-acid changes (amino acids number 425, 439, 545 and 668). The 5 LRMD dogs tested were all homozygous and identical for each of the SNPs, in particular the 4 leading to amino acid changes: amino acid 425 was a Threonine, 439 and 545 a Proline, and 668 an Alanine. The similar LTBP4 SNPs profile between LRMD dogs shows that this known modifier is, not more than Jagged1, implicated in the inter-individual heterogeneity seen in this dog model. The homozygosity of the tested LRMD dogs relates to the high inbreeding rate of the colony, which probably makes this LRMD model a facilitated context to discover new DMD phenotype modifiers.