Three‐year quantitative magnetic resonance imaging and phosphorus magnetic resonance spectroscopy study in lower limb muscle in dysferlinopathy

Abstract Background Natural history studies in neuromuscular disorders are vital to understand the disease evolution and to find sensitive outcome measures. We performed a longitudinal assessment of quantitative magnetic resonance imaging (MRI) and phosphorus magnetic resonance spectroscopy (31P MRS) outcome measures and evaluated their relationship with function in lower limb skeletal muscle of dysferlinopathy patients. Methods Quantitative MRI/31P MRS data were obtained at 3 T in two different sites in 54 patients and 12 controls, at baseline, and three annual follow‐up visits. Fat fraction (FF), contractile cross‐sectional area (cCSA), and muscle water T2 in both global leg and thigh segments and individual muscles and 31P MRS indices in the anterior leg compartment were assessed. Analysis included comparisons between patients and controls, assessments of annual changes using a linear mixed model, standardized response means (SRM), and correlations between MRI and 31P MRS markers and functional markers. Results Posterior muscles in thigh and leg showed the highest FF values. FF at baseline was highly heterogeneous across patients. In ambulant patients, median annual increases in global thigh and leg segment FF values were 4.1% and 3.0%, respectively (P < 0.001). After 3 years, global thigh and leg FF increases were 9.6% and 8.4%, respectively (P < 0.001). SRM values for global thigh FF were over 0.8 for all years. Vastus lateralis muscle showed the highest SRM values across all time points. cCSA decreased significantly after 3 years with median values of 11.0% and 12.8% in global thigh and global leg, respectively (P < 0.001). Water T2 values in ambulant patients were significantly increased, as compared with control values (P < 0.001). The highest water T2 values were found in the anterior part of thigh and leg. Almost all 31P MRS indices were significantly different in patients as compared with controls (P < 0.006), except for pHw, and remained, similar as to water T2, abnormal for the whole study duration. Global thigh water T2 at baseline was significantly correlated to the change in FF after 3 years (ρ = 0.52, P < 0.001). There was also a significant relationship between the change in functional score and change in FF after 3 years in ambulant patients (ρ = −0.55, P = 0.010). Conclusions This multi‐centre study has shown that quantitative MRI/31P MRS measurements in a heterogeneous group of dysferlinopathy patients can measure significant changes over the course of 3 years. These data can be used as reference values in view of future clinical trials in dysferlinopathy or comparisons with quantitative MRI/S data obtained in other limb‐girdle muscular dystrophy subtypes.


Introduction
Dysferlinopathy is an autosomal recessive neuromuscular disorder caused by mutations in the dysferlin gene, DYSF, leading to a deficiency of functional dysferlin, a protein that is highly expressed in muscle and that is essential in membrane repair. 1 Dysferlinopathy is characterized by an active inflammatory and degenerative process leading ultimately to muscle fibre necrosis and muscle replacement by fibrous and fatty tissue. 2 Dysferlinopathy patients have been labelled with different phenotypes, most importantly limb-girdle muscular dystrophy type R2 (LGMD R2 [formerly LGMD 2B]), and Miyoshi myopathy (MM). 3 Clinical symptoms usually start between late teenage years and 30 years and are associated with elevated serum creatine kinase (CK) levels. 1,3 The rate of disease progression is variable and is generally faster when the disease begins during early teenage years. 4 Whereas skeletal muscle is highly affected, impacting daily life activity, involvement of cardiac muscle is uncommon in dysferlinopathy. 4 In the last 10 years, magnetic resonance imaging (MRI) has been used as an outcome measure in several studies of dysferlinopathy patients. [5][6][7][8][9] The largest of these studies was the Clinical Outcome Study (COS) for dysferlinopathy initiated by the Jain Foundation, which involved a worldwide multi-centre natural history study in over 200 patients including clinical, biochemical, functional, strength, and imaging evaluations. 8,10 All these aforementioned studies described the use of T 1 -weighted and/or T 2 -weighted qualitative MRI. However, quantitative MRI has been increasingly applied in studies of neuromuscular diseases, 11 including Duchenne muscular dystrophy (DMD), [12][13][14][15] Becker muscular dystrophy, 16,17 facioscapulohumeral dystrophy (FSHD), 18 LGMD R9, 19 inclusion body myositis, 20 GNE myopathy, 21 and late-onset Pompe disease. 22,23 The objective of this study was to analyse the quantitative MRI and phosphorus magnetic resonance spectroscopy ( 31 P MRS) data, acquired in the patients from the Jain COS study, both cross-sectionally and longitudinally. Additionally, correlations between MRI and 31 P MRS parameters and functional and strength data were investigated.

Study set-up and subjects
Quantitative MRI and 31 P MRS data were obtained annually (between November 2012 and November 2017) for 3 years in 54 patients from two sites: Newcastle, UK (n = 42), and Paris, France (n = 12). These 54 subjects were a sub-cohort of the 201 patients enrolled across 15 sites worldwide in the Jain COS for dysferlinopathy (Supporting Information, Table S1). 8 Patients were included in the global study based on genetic confirmation for dysferlinopathy. 10 Ambulatory status was defined as the ability to walk 10 m with or without walking aids or orthotics. 8,10 None of the patients were under corticosteroid treatment since at least 6 months before the start of the study. Data were also obtained in 12 age-matched and sex-matched healthy subjects. Results on the 1-year fat fraction (FF) progression in the 12 patients from Paris have previously been reported. 24 Written informed consent was obtained from patients prior to inclusion, in accordance with the 1964 Declaration of Helsinki and its later amendments. The study registration number was NCT01676077. Healthy control subjects were scanned as part of a methodology MRI/S protocol approved by local ethics committees (Newcastle: United Kingdom National Research Ethics Committee 1417/375/2017; Paris: CPP-Ile de France VI-Groupe Hospitalier Pitié-Salpêtrière, ID RCB: 2012-A01689-34). Informed consent was obtained from all controls.

Quantitative magnetic resonance imaging and phosphorus magnetic resonance spectroscopy data acquisition
Data were acquired on one of two 3 T MRI clinical scanners: an Achieva (Philips Healthcare, Amsterdam, Netherlands) in Newcastle and Trio/Prisma (Siemens Healthineers, Erlangen, Germany) in Paris. For MRI, the local system's body RF coil for signal transmission and surface-array coils for signal transmission were used. For 31 P MRS, a flexible 1 H/ 31 P surface RF coil was employed (Newcastle: 14 cm, Philips; Paris: 11 cm, Rapid Biomedical GmbH, Rimpar, Germany). Patients were positioned feet-first supine and all MRI sequences were centred at one-third of the femur from the superior border of the patella and at the widest part of the calf.
Quantitative water-fat imaging was performed using a 2D (Newcastle) or 3D (Paris) gradient echo 3-point Dixon sequence. In Newcastle, Dixon scans were acquired in five 10 mm slices with a 20 mm interslice gap, covering a volume of 130 mm with an in-plane resolution of 1.5 × 1.5 mm 2 , a flip angle of 8°, and with repetition time (TR)/echo times (TEs) = 10/3.45-4.60-5.75 ms. In Paris, the Dixon sequence was acquired in a 3D volume of 5 mm slices covering at least 170 mm with an in-plane resolution of 1.0 × 1.0 mm 2 , a flip angle of 3°, and with TR/TEs = 10/2.75-3.95-5.15 ms. For water T 2 mapping, a 2D multi-spin-echo (MSE) sequence was employed covering the 130 mm 2D Dixon volume in Newcastle (in-plane resolution = 1.0 × 1.0 mm 2 ) and corresponding to the 3D Dixon volume in Paris (in-plane resolution = 1.4 × 1.4 mm 2 ), with 17 equidistant echoes (TE 1 /ΔTE = 9.9 ms in Newcastle; TE 1 /ΔTE = 9.5 ms in Paris), a TR of 3000 ms, a slice thickness of 10 mm, and a slice gap of 30 mm. 25 For calculating the transmit field (B 1 + ) in each voxel, a B 1 map sequence was run covering the same slices as the MSE sequence. 26 31 P MRS data were obtained in the anterior part of the right leg, unless the muscles were completely replaced by fat, from a non-localized free-induction decay of 64 averages (a TR of 4000 ms, a bandwidth of 3000 Hz, 2048 data points). Total acquisition time for quantitative MRI and 31 P MRS was approximately 40 min.
Magnetic resonance imaging and phosphorus magnetic resonance spectroscopy data processing All quantitative MRI data were processed using in-house written Matlab (MathWorks, Natick, MA, USA) or Python (www. python.org) code (Newcastle, UK, and Paris, France, respectively).
To determine FF and cross-sectional area (CSA), the boundaries of the ROIs were drawn (by I. W.) following individual muscle delineation, avoiding inclusion of other muscles, subcutaneous and intermuscular fat, tendons, and major blood vessels. To assess water T 2 , ROIs delineated the interior of the muscle, avoiding visible fasciae and blood vessels (by J. L. L.). 25 A third observer (R. F. T.) compared the two different segmentations and instructed the two observers to correct where necessary.
Fat fraction and CSA values were computed using the Dixon images (co-registered to the five MSE image slices). Dixon data were reconstructed using a six-component lipid model and considering a single T 2 * decay. 27,28 FF was calculated as SI (fat)/((SI)fat + (SI)water)*100 (with SI = signal intensity). FF maps that included subcutaneous and bone FF values < 95%, or partial fat-water swaps, were excluded for analysis. Before extracting FF and CSA values from the ROIs, FF maps were resized because of the difference in voxel size between Dixon and MSE images. When necessary, ROIs applied to the FF maps were trimmed or shifted due to overlapping subcutaneous fat or patient movement between Dixon and MSE acquisitions, respectively.
Using the MSE images, quantitative water T 2 maps were reconstructed based on a tri-exponential fitting procedure. 25 Only pixels where B 1 + values were between 80% and 120% of the prescribed flip angle were accepted for analysis. ROIs that included <10 pixels were excluded for analysis.
Contractile cross-sectional area (cCSA), defined as lean muscle CSA (expressed in cm 2 ) corresponding to the muscle volume fraction containing the contractile apparatus, was calculated using the FF and CSA values (cCSA = CSA* (1 À (0.01*FF))). 12 We evaluated the global segment for cCSA, which corresponds to the sum of the individual muscle cCSA values, per segment. For the global values for thigh and leg FF and water T 2 , a weighted mean of the values as determined in the individual muscles was calculated. 24 To evaluate disease progression, the annual and 3-year changes in FF, ΔFF% (expressed in %, absolute difference), and cCSA, ΔcCSA rel (expressed in %, relative difference compared with the precedent cCSA value), were evaluated. 21 All MRI outcome measures are reported as mean values of all pixels in the ROI averaged over the five slices.
Phosphorus magnetic resonance spectroscopy data were processed as previously described, 13,21 using AMARES in jMRUI 29 and Topspin (Bruker Medical GMbG, Ettlingen, Germany) to calculate P i,tot /PCr (total inorganic phosphate over phosphocreatine), P i,b /P i,tot (alkaline P i over P i,tot ), P i,tot /γATP (P i,tot over adenosine triphosphate), PCr/γATP, PDE/γATP (phosphodiesters over γATP), and PME/γATP (phosphomonoesters over γATP) ratios, as well as values of pH w (i.e. weighted pH, based on the relative weights of cytosolic P i , P i,a , and alkaline P i , P i,b , resonances) and the intramuscular magnesium concentration, [Mg 2+ ]. Only 31 P MRS data with sufficient signal-to-noise ratio or SNR (i.e. >10 for PCr resonance) were accepted for final analysis.
Inter-site variability between the Paris and Newcastle sites, assessed in four healthy control subjects scanned at both locations, was 1.2% (absolute difference) for FF, 1.0% (relative difference for cCSA), 0.5 ms for water T 2 , and 0.03 units for P i,tot /PCr and PDE/γATP. No inter-site corrections were applied for any of the outcome measures.

Functional, strength, and creatine kinase assessments
The dysferlinopathy-adapted 29-item scale North Star Assessment for LGMD-type Dystrophies (NSAD) was used. 4 For correlations with 31 P MRS, ankle dorsiflexion was assessed by manual muscle testing (MMT). In 90% of cases, functional/ clinical assessments were performed within 24 h of the MRI exam. The remaining 10% was within a month of the MRI exam. Functional tests were performed following the MRI/ 31 P MRS, minimizing any bias induced by muscle strain on the MRI and 31 P MRS outcomes. The frequency of physical activity reported between ages 10 and 18 years 30 and CK concentration were also analysed.

Statistical analysis
Statistical analyses were conducted using SPSS software Version 22 (SPSS, Chicago, IL, USA). The Mann-Whitney tests were performed for comparing demographic data between groups and to assess differences in MRI and 31 P MRS outcomes between controls and patients, with significance level set at P < 0.05.
We also analysed the data using a multilevel linear mixed model (LMM). A first analysis was performed for the MRI parameters to investigate left-right asymmetry, with side as a main variable, adding segment/muscle, group (ambulant vs. non-ambulant), and site as additional factors, and years since symptom onset and body mass index (BMI), for FF and cCSA, or FF for water T 2 , as continuous predictors to account for these confounding effects. The full analysis also included time point as a main variable and the interaction between time point and group. Random between-subject variation was also accounted for by including a random intercept in the model, which permitted interpretation of the variance partition coefficient (VPC). For 31 P MRS, a similar model was used with the predictors being years since symptom onset and FF. The significance level was set at P < 0.05. Bonferroni corrections were applied for multiple comparisons for each of the fixed effects.
The Spearman rank correlation test was used to explore additional relationship between MRI and 31 P MRS variables, and between MRI or 31 P MRS and functional and strength variables (P < 0.05).
The significance level was corrected for comparison of multiple outcome measures (FF, cCSA, and water T 2 for MRI: P = 0.05/3 = 0.017; eight 31 P MRS indices: P = 0.05/ 8 = 0.006). Additional information on the LMM analysis can be found in the Supporting Information.
For FF and cCSA, standardized response means (SRM) were calculated to assess the sensitivity to change over time (SRM ≥ 0.8 is considered highly sensitive to change). 31 For water T 2 and 31 P MRS indices, both SRM and standardized difference means (SDM) 21 were assessed.

Results
Data overview and demographic differences Figure 1A depicts a flow chart of the data sets acquired in the patients across visits. Figure 1B summarizes the main demographic differences between controls and patients. From the 37 ambulant patients at baseline, 30 were still ambulant at the end of the study. Between LGMD R2/2B and MM patients, and between the two sites, there were no significant differences for age (P = 0.56 and P = 0.77, respectively), number of years since symptom onset (P = 0.64 and P = 0.26, respectively), sex (P = 0.70 and P = 0.66, respectively), BMI, and the ambulant/non-ambulant patient ratio (P = 0.49 and P = 0.59, respectively). Only a significant difference in the LGMD R2/ MM patient ratio was found between both sites (P < 0.001). Table S1 summarizes the demographic data for all participants.

Baseline fat fraction and contractile cross-sectional area
Details about significant differences in leg and thigh FF and cCSA between patients and controls can be found in Figures 2 and 3A-3D. No significant differences were found between LGMD R2 and MM phenotypes for global thigh (P = 0.35) and leg (P = 0.71) FF and global thigh (P = 0.96) and leg (P = 0.37) cCSA.
No significant differences were observed between left and right global FF (global segments: P = 0.957, individual muscles: P = 0.886) and global cCSA (P = 0.035) in patients. All further analyses were performed on the mean of left-right FF and cCSA values. Highly significant differences were obvious between ambulant and non-ambulant patients, both for global thigh and leg FF and cCSA (Table 1) and for individual muscle FF (P < 0.001). While 91% of all ambulant patients demonstrated a global thigh FF ≤ 60% at baseline, 78% of all non-ambulant patients had a global thigh FF > 60%. Concerning the global segments, expected significant differences were observed for cCSA, but not for FF. Significant differences were evident between individual muscle FF values ( Figure 2): while thigh and leg posterior muscles showed the highest amount of FF, gracilis, sartorius, tibialis anterior/posterior muscles demonstrated the lowest FF values. No differences between sites were observed for global FF and cCSA and individual FF (P = 0.265). There was a significant effect of the number of years since symptom onset on the extent of muscle FF.
More data can be found in Tables S2 (for global segment FF and cCSA), S3 and S4 (for individual muscle FF), and S5 (for LMM for individual muscle FF).

Fat fraction and contractile cross-sectional area changes over time
In ambulant patients, median increases in global FF of 9.6% and 8.4% and median decreases of global cCSA of 11.0% and 12.8% were observed after 3 years, in thigh and leg, respectively ( Figure 3A-3D). Figure 3E-3H illustrates the individual patient trajectories for global FF and cCSA in thigh and leg. Besides the significant changes over time for global FF and cCSA, significant interactions were observed between time and ambulation status, reflecting the differential trajectories between both groups ( Table 1). The heterogeneity in disease progression between patients is also illustrated by

Water T 2
Global water T 2 values were significantly increased in patients as compared with controls ( Figure 5A and 5B). No significant differences were found between LGMD R2 and MM phenotypes for global thigh (P = 0.73) and leg (P = 0.72) water T 2 values. BL (x = 12), or Y2 (x = 5). Additionally, 31 P MRS data were not acquired in those muscles that were highly or completely replaced by fat (x = 36), based on visual assessment of the images. Following processing of the image data, 9% of the acquired MRI data were omitted for final analysis due to failed image reconstructions (x = 18 of total of 190 data sets). More than half of the 31 P MRS data were omitted for final analysis due to low SNR, which in most of the cases was due to high FF and/or low residual muscle in the anterior part of the leg. A median FF of 8% corresponded to the 31 P MRS data that were kept for final analysis, in contrast to the median FF of 43% for those data sets that were omitted. 'Validated' means 'included for final analysis' following quality control (image series complete, artefact-free, successful image reconstruction, sufficient signal-to-noise ratio). BL, baseline; n, number of patients; x, number of exams; Y1, Year 1; Y2, Year 2; Y3, Year 3. No significant left-right differences were found in patients for water T 2 values (global segments: P = 0.534, individual muscles: P = 0.633). All consequent analyses were performed on the left-right mean water T 2 values. No significant differences in water T 2 were found between ambulant and non-ambulant patients, both for global ( Figure  5A-B, Table 1, Table S6) and individual muscle values (P = 0.796, Figures S7-S8). Unlike global thigh and global leg water T 2 (P = 0.724), significant differences were evident between individual muscle water T 2 values, with the highest values in the anterior compartments of thigh and leg. The analysis did not demonstrate a significant effect of site on water T 2 . However, when performing an additional analysis on the baseline data only, a difference between both sites became apparent (P = 0.006 and P = 0.002, for the global and individual muscle values, respectively) but water T 2 from both sites remained significantly increased as compared with controls. Additionally, FF, as expected, impacted water T 2 , although only observed in individual muscles.
No changes over time in global thigh, global leg, and individual muscle water T 2 were demonstrated ( Figure 5A and 5B, Table 1, Tables S6-S9), but a significant interaction factor of time and group illustrated that differences appeared over time between ambulant and non-ambulant patients ( Table 1).
The SRM values for water T 2 were low, but the SDM values were systematically high in ambulant patients.
More data can be found in Tables S6 (for global segment  water T 2 ), S7 and S8 (for individual muscle water T 2 ), and S9 (for LMM for individual muscle water T 2 ).

Phosphorus magnetic resonance spectroscopy
Almost all 31 P MRS indices were significantly abnormal in patients as compared with healthy control values ( Figure  5C-5I).
No significant effect of site was observed for 31 P MRS indices (P-values between 0.049 for PCr/γATP and 0.940 for PME/γATP). There were, however, significant effects of FF on pH w , P i,b /P i,tot , PDE/γATP, PME/γATP, and [Mg 2+ ] (P ≤ 0.005), or of years since symptom onset on P i,tot /PCr, PCr/γATP, and P i,b /P i,tot (P ≤ 0.005).
Except for P i,tot /PCr and P i,tot /γATP, all other 31 P MRS indices did not change significantly over time and remained abnormal during the whole study duration ( Figure 5C-5I). As for water T 2 , the SRM values for 31 P MRS indices were low, but the SDM values were systematically high. Three-year muscle MRI and 31P MRS in dysferlinopathy Figure 6 shows the quantitative FF and water T 2 maps as well as the 31 P MR spectra in a patient at all four time points and a healthy volunteer.
More data can be found in Tables S10 (for 31 P MRS indices) and S11 (for LMM for 31 P MRS indices).

Relationships between magnetic resonance imaging and spectroscopy variables
Global thigh water T 2 was significantly correlated to ΔFF after 3 years (ρ = 0.52, P < 0.001), although less for ΔcCSA (ρ = À0.37, P = 0.002) ( Figure 7A). In individual muscles, significant correlations between mean water T 2 and ΔFF were observed for the vasti muscles and biceps femoris long head, gracilis, soleus, and extensor digitorum muscles (ρ = 0.50-0.60, P < 0.001). We found that global thigh water T 2 was correlated with the CK concentration (ρ = 0.48, P = 0.002). Water T 2 values were also significantly correlated with 31 P MRS indices such as P i,tot /PCr (ρ = 0.69, P = 0.001) and PCr/γATP (ρ = À0.63, P = 0.004). From an exploratory perspective, we illustrated in Figure 7B the relationship between P i,tot /PCr, time since onset of symptoms, water T 2 , CK levels, and physical activity regime in adolescence. In Tables S12 and S13, the correlation analyses are summarized.

Relationships between magnetic resonance imaging/spectroscopy and functional strength variables
Muscle function as measured by the NSAD decreased steadily with increasing FF and decreasing cCSA ( Figure 7C and 7D). A significant relationship existed between the change in total  Tables S2-S4 for longitudinal FF and cCSA, and in Table 1 and Table S5 for the LMM analyses. *P < 0.008, **P < 0.001 (between visits); # P < 0.008, ## P < 0.001 (between controls and patients); § P < 0.008, § § NSAD score and FF after 3 years in ambulant patients ( Figure  7E and 7F).

Discussion
The presented 3-year quantitative MRI and 31 P MRS study has demonstrated cumulative muscle degradation in legs and thighs, as measured by the yearly and overall increases in FF and decreases in cCSA. Furthermore, muscle water T 2 values and various 31 P MRS indices have proven to be highly and persistently abnormal and were correlated to disease progression as evaluated by an increase in FF.

Study demographics, extent of disease, and disease progression
The investigated study cohort adequately represented the population in the full Jain COS for dysferlinopathy, as described in the T 1 -weighted MRI study. 8 As expected, in line with earlier publications, [5][6][7][8]32 no differences were shown between LGMD R2 and MM phenotypes. The quantitative MRI findings revealed a strong proximo-distal and posterior dominant pattern of muscle fat replacement in 80% of the examined patients, which is slightly less than the 88% reported in Angelini et al. 33 but much higher than the 56% in Jin et al. 7 Upper limb, shoulder, trunk, and pelvis muscles (not investigated in the current study) are also known to be affected, 6,8,9 but it is known that this follows the early posterior thigh and Three-year muscle MRI and 31P MRS in dysferlinopathy leg compartment involvement. Muscle fat replacement was heterogeneous across the patient population, and the pronounced involvement of medio-posterior muscles such as semimembranosus, adductor magnus, and gastrocnemius medialis, as compared with anterior muscles, with an absence of a significant left-right asymmetry confirm the results of previous studies using T 1 -weighted MRI. [5][6][7][8][9]33 Muscle involvement in dysferlinopathy is similar to observations in   Tables S6-S8 for longitudinal water T 2 and Table S10 for 31 P MRS, and and Tables S9 and S11 for the LMM analyses. *P < 0.008 (water T 2 )/0.006 ( 31 P MRS), **P < 0.001 (between visits); # P < 0.008/0.006, ## P < 0.001 (between controls and patients); § P < 0.008/0.006, § § P < 0.001 (between ambulant and non-ambulant patients other forms of LGMD. In LGMD R1/2A 34 and LGMD R9/2I, 19 however, anterior muscles are less impacted in LGMD R9/2I, whereas, in LGMD R12/2L, 35 a higher level of asymmetry might be present. The annual increase in global FF of 3-4%, and concomitant decrease in cCSA, as found in both thighs and legs of ambulant patients, reflect a considerably faster rate of disease progression compared with the ≈1% increase in muscle fat replacement found after 1 year in LGMD R9 (formerly LGMD2I). 19 Despite the significant annual disease progression in dysferlinopathy, there is a large inter-individual variability, which is more pronounced than in LGMD R9. 24 Nevertheless, the knowledge of the annual expected FF increase will enable us to establish clinically meaningful outcome measures for future therapeutic trials. Using global values for FF is a valid approach for evaluating disease progression compared with in-dividual ROIs, as demonstrated by the high SRM values. 17,19,23,24 Disease activity Skeletal muscle water T 2 is another quantitative MRI outcome measure, one that is non-specific, yet sensitive to mechanisms such as inflammation and oedema, and reflects the disease activity. 11 Water T 2 was, in general, significantly elevated in patients, and the highest values were found in the anterior parts of thigh and leg, which were, on average, less replaced by adipose tissue. This confirms findings of earlier studies where hyperintensities found on T 2 -weighted images and classified as oedema preceded muscle fat replacement in muscles that were in the early stages of the Three-year muscle MRI and 31P MRS in dysferlinopathy disease. 5-7 Indeed, we found that patients with higher water T 2 demonstrated faster disease progression as muscle fat replacement increases were more pronounced, a relationship that was established in other myopathies with both a known inflammatory component such as dermatomyositis, 36 inclusion body myositis, 20 FSHD, 18 and DMD, 13,14 but also in late-onset Pompe disease 22 and GNE myopathy. 21 In muscles such as gracilis and sartorius, where FF was low and disease progression was slow, water T 2 was found to be in normal ranges. In a study in DMD, it was demonstrated that water T 2 was significantly increased even before any signs of muscle fat replacement. 14 In the current study, the water T 2 was found to be systematically increased in both thigh and leg of patients over the course of the entire study, especially in ambulant patients, indicating that the disease activity mechanisms are persistent throughout the disease duration and might endure as long as 18 years, as reported earlier. 6 The different 31 P MRS indices were also found to be continuously abnormal over the 3 years. This included a high level of PDE/γATP, associated with an increased phospholipid . (E) Relationship between change in global thigh FF and the total NSAD after turnover, and an elevated alkaline pH w , reflecting an increased intracellular pH or increased interstitial P i,tot pool, 13 both indices pointing towards a disturbance at the level of the sarcolemma, and found earlier in DMD, 13,14 Becker muscular dystrophy, 16 FSHD, 18 and GNE myopathy. 21 The 31 P MRS data were generally obtained in patients with very low muscle fat replacement, revealing that changes in disease activity can be measured before the occurrence of any macroscopic muscle destruction, as demonstrated in DMD. 14 We also found that the shorter the disease duration, the higher the level of P i,tot /PCr was, which is an index that reflects the degree of muscle energy wasting. Interestingly, the patients with the highest P i,tot /PCr also seemed to have a history of a more vigorous physical activity regime, 30 as compared with patients with lower P i,tot /PCr, which might be an indication that there is a potential negative effect of physical exercise in dysferlinopathy patients. 33 As exercise is usually found to be beneficial in neuromuscular disorders, 37 this might also have implications for future treatment of these patients, including physical therapy. The persistent abnormal values for water T 2 and most 31 P MRS indices in dysferlinopathy render their corresponding SRM values indicative of a low sensitivity to change. However, the fact that these MRI and 31 P MRS variables are all significantly different compared with normal muscle (i.e. high SDM 21 ) makes them very useful biomarkers for assessing the impact of a potential therapeutic intervention, especially in the early stages of the disease.

Relationship with function and strength
Similar to the earlier larger-scale dysferlinopathy studies, 5-7,9 a strong relationship was observed between muscle fat replacement, disease duration, and function. However, the Jain COS study was the first study where an LGMD-specific scale of motor performance was used. 4 The strong correlation found between total NSAD and global thigh FF, especially in the earlier stages of the disease, as measured with FF values between 0% and 40%, is important in the search for clinically meaningful outcome measures. In contrast to the 6 min walk distance where a sudden and very strong decline was observed only after FF exceeded 60%, 15 the NSAD is highly sensitive to changes in FF in the earlier disease progression. When the investigated endpoint is the loss of ambulation, the complementarity of the concomitant investigations of total NSAD and quantitative MRI proves its worth, especially in light of potential treatments.

Methodological issues
First, the major methodological issue is the multi-centre aspect of this study. The use of different MRI clinical systems in-troduced a certain degree of variability, although the differences were found to be negligible between Newcastle and Paris, at least in healthy controls. In patients, however, we observed significant differences in water T 2 between sites, but at both sites, values were significantly abnormal as compared with control values. This issue underlines, besides the obvious necessity to harmonize MRI protocols between sites in multi-centre studies, the fact that water T 2 results always need to be interpreted in their proper context. Moreover, as a quantitative MRI outcome measure for disease activity, water T 2 , with its potential to predict a subsequent change in muscle fat replacement, is obviously more relevant in the early stages of the disease. Second, water T 2 values were determined using the tri-exponential fitting procedure. 25 Similarly, as observed in previous studies reporting on water T 2 in fatty replaced muscle, 25,38-40 the water T 2 derived from the tri-exponential fitting has been shown to be abnormally low at high FF values. Although this result agrees with what has been observed when using the 'gold standard' MRS-based method, 40 further research is still needed to explain this observation. More recently, water T 2 mapping procedures based on the extended phase graph (EPG) algorithm, which avoid the need to acquire an additional B 1 + map, have demonstrated their use in fatty replaced muscle. 38,39 Although these EPG methods are promising, in the present work, we have chosen to use the tri-exponential fitting procedure because, besides being a validated method, it is the technique employed in most of our earlier published works involving clinical applications (some of them cited in the current work 13,21,22 ), thus allowing for a coherent comparison between these studies.

Conclusions
This study provided quantitative natural history MRI and 31 P MRS data in a moderately large group of dysferlinopathy patients with a heterogeneous disease state. Although challenging when organized in a multi-centre set-up, the investigated quantitative outcome measures were shown to be sensitive to annual changes (in case of FF) or were persistently abnormal and correlated to subsequent changes in FF (in case of water T 2 and 31 P MRS). Evaluating a global segment instead of all individual muscles separately has not only obvious advantages in processing and analysis but also proved to be as valuable and as sensitive (as individual muscles) in evaluating disease progression and activity in dysferlinopathy. Quantitative MRI/S outcomes, identifying changes in muscle structure, pathophysiology, and metabolism, were shown to change over time. Adding functional tests, such as the total NSAD score, can link these changes to clinically meaningful endpoints. The quantitative data acquired in this study can be used as reference values for future clinical trials in dysferlinopathy and to compare with similar quantitative MRI/S data in other LGMD subtypes. New longitudinal quantitative MRI/ 31 P MRS studies might be anticipated in other anatomical targets, such as trunk, pelvis, and/or upper limb muscles, as we know they are affected further downstream in the disease progression of dysferlinopathy. At the time of publication, the Jain COS2 study is ongoing, which includes quantitative MRI in lower and upper limb.