Modeling of Dysferlinopathy (LGMDR2) Progression

Introduction

Dysferlinopathy is a form of limb-girdle muscular dystrophy (LGMDR2) produced by pathogenic variants in the DYSF gene that encodes the protein dysferlin.¹ It is inherited in an autosomal recessive manner, and its main symptom is muscle weakness. Dysferlin is located at the sarcolemma of the muscle fibers and at the T tubules, playing a key role in the repair process of the membrane after injury^2,3 and in cell calcium handling.^4,5 Its absence or malfunction leads to a defective sealing of the muscle fiber, triggering a cascade of molecular consequences eventually leading to muscle fiber death.⁶ Muscle fibers become progressively replaced by connective and fat tissue, which do not have any contractile properties, leading to irreversible muscle weakness and wasting. Although muscle atrophy and weakness inevitably progress over time, the rate and severity at which this happens is variable, even in patients with the same underlying pathogenic variants belonging to the same family.^7,8

The Dysferlinopathy Clinical Outcome Study 1 (COS 1) is an international, natural history study, funded by the Jain Foundation, through which longitudinal clinical, genetic, and imaging data were collected in 15 sites across Spain, Germany, Italy, France, the United States, Japan, Australia, and the United Kingdom.⁷ Several conclusions were drawn from the data analysis regarding disease progression: Rate of disease progression did not differ according to the initial phenotype⁹; a steeper decline in function was seen during the first 10 years of disease¹⁰; younger age at onset was associated with worse function and faster disease progression¹⁰; Asian patients performed worse than White patients in timed tests across the disease course¹⁰; intense physical activity during teenage years, which can lead to higher rates of muscle injury, was associated not only with earlier symptom onset and earlier need for a wheelchair full-time¹¹ but also with more severe muscle replacement by fat in MRI¹²; and higher baseline water T₂ values were linked to greater functional decline during the next 3 years of progression.¹³ However, some conflicting results were more difficult to integrate, for example those regarding sex differences. Although women scored better in the North Star Assessment for Dysferlinopathy (NSAD) at baseline,⁹ they simultaneously showed higher fat content in several muscles measured through MRI^13,14 Although current findings offer insight into certain aspects of disease progression, contradictions and persistent knowledge gaps highlight the complexity of LGMDR2 and the need for further research to fully understand its natural history.

MRI is increasingly being implemented as an outcome measure in clinical trials, both natural history and interventional studies, because it not only correlates well with clinical changes,^15-18 but it can also overcome some of the limitations clinical assessments face, such as variability because of external uncontrollable factors (from the patient and evaluator), inability to measure loss of function in presymptomatic patients,^19,20 and less sensitivity in detecting changes over a short period of time.²¹ A partial longitudinal analysis of COS 1 MRI data involving 54 patients from 2 sites and 12 healthy controls was published in 2022,¹⁵ assessing fat fraction (FF), contractile cross sectional area, muscle water T_2, and phosphorus MR spectroscopy in the lower limbs, providing broader multidimensional information on MRI findings of disease progression. To describe disease trajectories in a more accurate and thorough way, not only clinical changes should be considered but also changes in MRI-based variables. Moreover, with novel potential disease-modifying therapies being closer to the human trials stage, a clearer understanding of the natural history of muscular dystrophies and of how different outcome measures correlate and influence each other over time is essential.

Our main aim was to characterize natural disease severity and progression in patients with LGMDR2 by modeling quantitative MRI FF measurements of the skeletal muscles of the lower limbs obtained longitudinally, include clinical variables into the models to improve their fit to the FF data, and assess how these variables influence the muscle replacement by fat process over time.

Methods

Results

In this study, 109 patients were included. Fifty-seven were female, and on average, each patient had 3.27 MRI scans done over a 3-year period in 1 of 7 participating centers (eTable 2). A summary of the cohort's demographics and mean right-to-left FF values for each muscle and segment at each visit can be found in Table 1. A little over 3 quarters (77.8%) of patients were ambulant at baseline, and 1 patient only used noninvasive ventilation.

The paired samples t test showed no difference in muscle replacement by fat for each of the assessed muscles between right and left at baseline (p > 0.05). Therefore, we decided to calculate the mean FF value for each muscle using the right and left data when available and use those values for the analysis.

Figure 1A shows the average trajectory of the thigh and lower leg while Figure 1, B and C show the average trajectory of each muscle (B, lower leg and C, thigh). eFigure 1 shows the average trajectory on top of the patients' observed FF values for each individual muscle.

In teenage years and early adulthood, lower leg muscles show higher levels of muscle replacement by fat than thigh muscles; however, the rate of disease progression for thigh muscles is steeper; therefore, after the third decade of life, thigh muscles progressively become more involved than lower leg muscles (Figure 1A).

Two distinct muscle groups arise after analyzing FF trajectory in the lower legs (Figure 1B), the anterior compartment muscles plus the TP, responsible for plantar dorsiflexion, extension of the toes and foot inversion, and the posterior and medial compartment muscles, responsible for plantar flexion and foot eversion, respectively, which show more severe muscle replacement by fat. Something similar can be seen in the thigh muscles (Figure 1C), as the sartorius and gracilis muscles have a different trajectory showing lower FF values as the disease progresses compared with the other thigh muscles, resembling the trajectory of the lower leg anterior compartment. The femoral quadriceps muscles, responsible for knee extension, were less involved than the hamstring and AM muscles, responsible for knee flexion and adduction, respectively. To complement Figure 1, Table 2 provides the corresponding quantitative data, including the mean annual rate of progression and the expected FF value at 40 years of age for each muscle, as described by our models (eTable 3). Table 2 serves as a reference for understanding and comparing the progression rates and trajectories between muscles. In summary, muscles can be grouped into 3 categories: (1) spared muscles: the anterior lower leg muscles, the TP, the gracilis, and the sartorius; (2) severely involved muscles in early disease stages: the posterior and medial lower leg muscles; and (3) muscles increasingly involved with disease progression: the thigh muscles (except for the gracilis and sartorius). This group can be further divided into 2 subgroups based on their trajectories: the anterior compartment and the posterior compartment plus the AM. Overall, muscles that participate in the same joint movement seem to progress similarly over time.

We plotted each patient's modeled trajectory against the observed data obtained in the study for thigh (eFigure 2A) and lower leg muscles (eFigure 2B). Using the average FF trajectory for each muscle, we created an animated image of the thigh and lower leg showing the mean progression of muscle replacement by fat as patient's age (eFigure 3). Moreover, Figure 2 shows the change in the FF maps reconstructed from the Dixon images for 3 patients with different severities of muscle replacement by fat.

Tables 3 and 4 summarize the characteristics of the 3 built models for thigh and lower leg (simple, including baseline variables and including longitudinal variables). Moreover, eTable 4 shows intermediate results in the building of each model and how the addition or removal of variables affected the χ² likelihood ratio test. The thigh baseline model incorporated age at MRI scan, age at symptom onset, baseline NSAD score, and baseline 6MWT velocity, whereas the longitudinal thigh model included age at MRI scan, age at symptom onset, baseline NSAD score, and changes over the first year in 6MWT velocity. The lower leg baseline model included age at MRI scan, age at symptom onset, sex, teen exercise intensity, baseline NSAD score, and baseline 6MWT velocity and the longitudinal model included age at MRI scan, age at symptom onset, sex, teen exercise intensity, baseline NSAD score, and changes over the first year in 6MWT velocity. According to the model, women had on average, an absolute value of FF 8.8% higher than men in the lower legs, with a maximum difference of 12.4% at age 45 years. Neither the addition of ethnicity, residual dysferlin expression, or predicted phenotype based on the genetic variants improved the models' fit to the data.

Discussion

We have analyzed longitudinal thigh and lower leg Dixon-based FF data from patients with LGMDR2 acquired annually over a 3-year period. We have modeled individual muscles' trajectories over time and created multivariate models for the thigh and lower leg, assessing how different variables influence FF progression. These results allowed us to make the following conclusions: (1) Muscles that participate in generating the same joint movement have similar FF trajectories as the disease progresses, suggesting that there could be biomechanical characteristics determining FF progression in dysferlinopathy; (2) female sex is linked to higher FF values in the lower legs but not in the thighs; (3) the younger the age at symptom onset, the higher the predicted FF values; and (4) the baseline NSAD score strongly contributes to explaining FF values over time.

Our findings suggest that muscles involved in producing the same movements progress with a similar trajectory, probably because of their exposure to comparable mechanical loads and tension over time. Given that dysferlin plays a critical role in repairing muscle membrane damage, the failure of this repair mechanism in individuals with dysferlinopathy may lead to synchronized degeneration patterns.^26,27 This supports the idea that the cumulative mechanical stress these muscles endure throughout life might contribute to the observed trajectory of muscle degeneration.

Modeling of lower leg FF values differed between male and female participants while this was not the case for the thigh model. There were no statistically significant differences in age at assessment, age at symptom onset, or disease duration between men and women included in this study in the whole cohort. The available data on differences between male and female patients with dysferlinopathy are scarce and controversial. On the one hand, semiquantitative MRI-based analysis showed that female patients had more severe muscle replacement by fat in some muscles of the thigh and lower leg than their male counterparts.¹⁴ On the other hand, women were found to perform better in muscle function tests (NSAD) while showing similar rates of progression as men.¹⁰ This discrepancy could be partly explained by the fact that the NSAD score better reflects the function of proximal rather than distal lower limb muscles. To our knowledge, there is no published data on molecular or genetic differences between sexes on the skeletal muscle structure of patients with dysferlinopathy. Nevertheless, differences between sexes and muscle lipid accumulation have been described in Bla/J dysferlin-deficient mice, with females showing higher lipid content and upregulation of genes linked to lipid metabolism, especially in the quadriceps and gastrocnemius muscles.²⁸ Moreover, it has been suggested that fast type II fibers are more prone to muscle damage than slow type I fibers in muscular dystrophies²⁹; however, in LGMDR2 mice models and humans, type I myofibers exhibit a higher content of lipid droplets.^30,31 Although skeletal muscle fiber type composition is similar between men and women, it has been described that the ratio of type II/I fibers is greater in men than in women.³² How meaningful these differences are and whether they might play a role in the progression of dysferlinopathy is currently unknown, highlighting the need for further research on biological modifiers that can influence disease progression and response to therapies.

Intense exercise during teenage years in people with LGMDR2 has been linked to earlier symptom onset, and part-time as well as full-time wheelchair use.^11,12 In the multivariate models, teenage exercise levels had a fixed effects value above statistical significance for both thigh and lower leg, but its removal from the models worsened the fit only to the lower leg data. A previously published work¹² looking specifically at the effect teenage exercise could have on muscle replacement by fat measured through T₁-weighted images using Mercuri score concluded only some pelvic muscles showed significantly higher fat content in the higher exercise levels group and no differences were seen between groups in the thigh or lower leg muscles. We agree with their hypothesis in that this lack of difference could be because of the thigh and lower leg muscles having an earlier involvement than the pelvic ones in the natural history of the disease and so, as patients in this cohort have been symptomatic for 16 years on average, there could be an initial difference in earlier stages that we were unable to capture.

Age at symptom onset had a negative effect on FF trajectory; therefore, the older the age at symptom onset, the lower the predicted FF value at a given time point, and this variable remained statistically significant in the 4 multivariate models, highlighting its weight in influencing FF. Numerous studies have shown that longer disease duration is associated with greater disease severity.^7,10,14,33 However, the relation between age at symptom onset and disease duration has proven to be more complex, with fewer evidence supporting a more severe phenotype^8,10 and faster progression¹⁰ in patients with an earlier onset of symptoms. Our findings support the potential importance of age at symptom onset, along with disease duration, in shaping disease trajectory, and therefore, both parameters should be considered in the analysis of patients with LGMDR2. The reason why some patients present with symptoms earlier than others still remains an unanswered question, but we think that this is due to the interaction of multiple factors rather than solely the underlying pathogenic variant, a hypothesis that is also supported by our results because the inclusion of genetic data in our models did not reveal any clinically significant differences between groups.

Baseline NSAD score had a negative effect on FF modeling. Those patients with lower baseline NSAD scores showed higher FF values and were therefore more severely affected. The baseline NSAD score remained statistically significant in all models even after including longitudinal variables, suggesting that it has a strong and persistent power in explaining FF values. This finding is consistent with previous studies reporting a strong correlation between clinical severity and MRI-based measures in LGMDR2.¹⁵

The inclusion of clinical variables in the multivariate models provided valuable insight into the interactions between these factors and FF values. Furthermore, the χ² likelihood ratio test indicated a significant improvement in model fit when moving from bivariate to multivariate models, highlighting the contribution of additional variables to the overall explanatory power of the model. By contrast, despite this improvement in overall data fit, the measures of dispersion did not improve, meaning that the average residual values (i.e., the difference between the observed FF value and the 1 predicted by the model for every patient at each time point) were not overall reduced after including additional variables. This suggests that, although the new variables improved the fit of the models at the group level, they were not able to improve capturing individual level variability in FF. A larger sample size, more longitudinal data points per patient, and probably more complex modeling approaches like nonlinear or machine learning based could potentially clarify this limitation. That said, we are certain this is not purely related to the design of the trial and statistical analysis because we think other variables, such as health comorbidities, medication, not only intensity but also type of exercise carried out as teenagers and in adult life, nutrition and diet, hormonal, and other genetic and epigenetic variants impacting muscle and metabolism, are also influencing disease severity and progression.

Although these models reflect that most patients will show an increase in skeletal muscle fat replacement as they age, the rate at which this is expected to happen varies considerably between patients (eFigure 2). Some patients' modeled trajectories have a negative slope, meaning that according to the model their muscle pathology should improve over time. This is mainly seen in patients who are relatively stable in the progression, usually because they are either very severely or mildly affected, and small fluctuations in FF measurements that reflect a supposed decrease in fat content lead the model into interpreting that the patient is improving. This limitation could be overcome by following patients for longer periods of time and smoothing these yearly fluctuations in the overall trajectory.

Other limitations we faced in this study were those related to the multicenter nature of COS 1. There is variability inherent to the fact that different MRI systems were used at different sites (eTable 1). The same principle is applicable to clinical outcome measures because although all health care professionals involved were trained in how to assess patients correctly and objectively, interobserver variability is a well-recognized bias and one of the most frequent problems large scale studies like COS face. Moreover, although COS is the largest natural history study in dysferlinopathy, a 3-year follow-up is still a short period of time in the natural history of this non–life-limiting rare disease, only offering a snapshot of a long and complex disease progression. The youngest included patient was in its late teens; therefore, the absence of FF data from younger patients means that modeling of FF in patients younger than 15 years was solely based on what we would expect following the muscle fat replacement process in adult life, which could be biased. Finally, this model was built using a single cohort, and its predictive accuracy has not yet been evaluated using independent data. Future validation with an external cohort will be essential to assess generalizability and reliability in capturing FF progression among a broader population. After the completion of COS 1, the Jain Foundation set up COS 2 (NCT01676077), another natural history study with the aim of gaining further understanding of some aspects where COS 1 has fallen short of, focusing on developing standards of care, getting closer to clinical trial readiness, and validating the developed outcome measures from COS 1. This would be an interesting cohort to validate these results with.

In conclusion, this study offers models to better understand and describe progression of muscle FF of the thighs and lower legs of patients with dysferlinopathy using clinical and radiologic data. The analysis of multivariate models provided relevant information on the way different factors shape muscle degeneration. The results presented here contribute to the body of information that can be used for the design of clinical trials and the analysis of results. Based on these data, we suggest that, in the design of future placebo-controlled clinical trials, not only the patient's age is considered for randomization but also age at symptom onset, sex, and baseline NSAD score, meaning the ideal pair of patients would be 2 people of the same sex and age who started showing symptoms at a similar moment in life and have similar disease severity. Moreover, although the effect of physical activity early on in life on posterior disease progression is still not completely understood, some evidence points toward an association between it and faster decline and disease severity and should therefore be assessed during screening and, ideally, included in the randomization process, or at least controlled for during final data analysis.

Author Contributions

C.F. Bolano-Diaz: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. H. Reyngoudt: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design. I.J. Wilson: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. M.K. James: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. F.E. Smith: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design. E. Caldas de Almeida Araujo: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design. H. Gordish-Dressman: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. H. Hilsden: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. L.E. Rufibach: drafting/revision of the manuscript for content, including medical writing for content; study concept or design. D. Wallace: major role in the acquisition of data. L. Ward: major role in the acquisition of data. R. Stramare: major role in the acquisition of data. A. Rampado: major role in the acquisition of data. M. Smith: major role in the acquisition of data. J-M. Boisserie: major role in the acquisition of data. J. Le Louer: major role in the acquisition of data. S. Foster: major role in the acquisition of data. A. Peduto: major role in the acquisition of data. N. Sato: major role in the acquisition of data. T. Tamaru: major role in the acquisition of data. A.M. Sawyer: major role in the acquisition of data. J.W. Day: major role in the acquisition of data. K.J. Jones: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. M.C. Walter: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. T. Stojkovic: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. M. Mori-Yoshimura: major role in the acquisition of data. J.R. Mendell: major role in the acquisition of data. E. Pegoraro: major role in the acquisition of data. V. Straub: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design. A.M. Blamire: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; Study concept or design. P. Carlier: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design. J. Diaz-Manera: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data.

Study Funding

Funds for this study were provided by the Jain Foundation. The John Walton Centre Muscular Dystrophy Research Centre is part of the MRC Centre for Neuromuscular Diseases (Grant number MR/K000608/1) and the National Institute for Health and Care Research (NIHR) Newcastle Biomedical Research Centre (BRC). The NIHR Newcastle BRC is a partnership between Newcastle Hospitals NHS Foundation Trust and Newcastle University funded by the NIHR. This paper presents independent research supported by the NIHR Newcastle BRC. Volker Straub is supported by the NIHR Newcastle BRC. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.

Disclosure

M. Mori-Yoshimura is deceased; to the best of our knowledge, she did not have any relevant disclosures to make. Go to Neurology.org/NG for full disclosures.