Cross‐sectional serum metabolomic study of multiple forms of muscular dystrophy

Abstract Muscular dystrophies are characterized by a progressive loss of muscle tissue and/or muscle function. While metabolic alterations have been described in patients’‐derived muscle biopsies, non‐invasive readouts able to describe these alterations are needed in order to objectively monitor muscle condition and response to treatment targeting metabolic abnormalities. We used a metabolomic approach to study metabolites concentration in serum of patients affected by multiple forms of muscular dystrophy such as Duchenne and Becker muscular dystrophies, limb‐girdle muscular dystrophies type 2A and 2B, myotonic dystrophy type 1 and facioscapulohumeral muscular dystrophy. We show that 15 metabolites involved in energy production, amino acid metabolism, testosterone metabolism and response to treatment with glucocorticoids were differentially expressed between healthy controls and Duchenne patients. Five metabolites were also able to discriminate other forms of muscular dystrophy. In particular, creatinine and the creatine/creatinine ratio were significantly associated with Duchenne patients performance as assessed by the 6‐minute walk test and north star ambulatory assessment. The obtained results provide evidence that metabolomics analysis of serum samples can provide useful information regarding muscle condition and response to treatment, such as to glucocorticoids treatment.

markers of metabolic syndrome such as serum levels of leptin, 17 creatine, arginine, branched amino acids and phosphatidylcholine [18][19][20] were reported to be elevated in DMD, leading us to further study metabolites profiles in DMD patients.
The aims of this study were (i) to identify metabolites able to discriminate between patients and controls, and (ii) to test associations between biomarkers levels and clinical performance as measured by the 6-minute walk test (6MWT) or the North Start Ambulatory Assessment (NSAA).
Furthermore, we investigated whether the observed signature could be translated to other MDs, such as the milder allelic variant Becker muscular dystrophy, but also genetically separate myotonic dystrophy type 1 (CUG expansion in the DMPK gene), facioscapulohumeral muscular dystrophy (shortening of a subtelomeric repeat unit enabling the production of the toxic Dux4 retrogene) and limbgirdle muscular dystrophies type 2A and 2B (lack of calpain-3 and dysferlin, respectively).

| Participants
Patient and healthy control samples were obtained from the Naples Human Mutation Gene Biobank (NHMGB), partner of EuroBioBank and of the Telethon Network of Genetic Biobanks. A total of 30 DMD (mean age 9.3 years), 10 Becker muscular dystrophy (BMD, mean age 10.8 years), 10 facioscapulohumeral muscular dystrophy (FSHD, mean age 39.6 years), 10 myotonic dystrophy type 1 (DM1, mean age 43.9 years), 5 limb-girdle muscular dystrophy type 2A (LGMD2A, mean age 32.6 years), 5 LGMD2B (mean age 29.2 years), 10 paediatric controls (mean age 15.4 years) and 12 adult controls (mean age 46.4 years) were included in the study. Blood was drawn in the morning, and all individuals had fasted since the previous day.
Metadata connected to this cohort are shown in Table S1.
The study was approved by the Institutional Review Board.
Informed consent forms for blood drawing were obtained for all participants, at the time of the scheduled follow-up. The investigation was conducted according to the declaration of Helsinki.

| Serum preparation
Venous blood was allowed to clot for 30 minutes in red-capped tubes followed by centrifugation for 10 minutes at 2350 g. Serum was carefully removed, aliquoted and stored at À80°C pending use.

| Data acquisition
Experiments were performed as previously detailed. 21 Briefly, 15 lL of serum samples was used for protein precipitation using methanol containing a mixture of internal standards (Table S2)

| Statistics and pathway analysis
Data analysis was performed after data standardization. In 4 cases, 2 peaks were assigned to the same chemical ID using the same ionization mode (hexanoylglycine, kynurenic acid, N-acetyl-DLtryptophan in both positive and negative modes); given the high correlation in the peak area for these instances, we only retained the measurement with higher intensity. To determine differential representation of metabolites in serum, we used a linear model, and the Bonferroni correction for multiple testing was applied. An adjusted P-value of <.05 was considered significant. Linear models were used to investigate potential associations between metabolites levels with age, 6MWD and NSAA. Pathway analysis of the metabolomics data set was performed with the global test. 23,24 Permutations were used to take into account the correlation between metabolites measurements. Subset's option was used to define pathways obtained from WikiPathways. 25 Weight's option was used to score each metabolite in order to take multiple mappings of 1 metabolite to a measurement into account. For example, a weight of 1/3 was assigned to measurements with 3 metabolite mappings, while a weight of 1 was assigned to measurements with a single metabolite mapping. The Westfall and Young's maxT method was used to correct for multiple testing. All analyses were performed in R using the lm, cor, cor.test and gt functions.

| Data availability
The data sets generated during and/or analysed during this study are available as supplementary data.  Fifteen metabolites were found to be differentially present in the DMD patients sera compared to age-matched healthy controls ( Figure 1A and Table 1). All metabolites showed reduced levels in DMD patients compared to healthy controls except for creatine ( Figure 1B). No or moderate correlations were observed among the 15 metabolites or with creatine kinase activity with the exception of p-coumaric acid correlating with dehydroisoandrosterone 3-sulphate ( Figure S1). Two metabolites showed a significant positive association with age ( Figure S2). While the association between age and dehydroisoandrosterone 3-sulphate was justified by the data distribution, the association with 5a-DHT was mostly driven by a few cases.
To understand and explain the differences between DMD patients and controls, a pathway analysis was performed, which allowed the identification of 23 molecular pathways (Table S5). The metabolism of polyamines was one of the most significant ones with creatine and creatinine anticorrelating ( Figure 3A,B). Pathway analysis highlighted other pathways where metabolites contribute to the pathway in opposite direction. Citric acid and L-aspartic acid were mapped to the alanine and aspartate metabolism pathway, and the ratio between the 2 was elevated in both DMD and BMD patients ( Figure 3C,D).
The creatine/creatinine ratio has recently been reported to correlate with age in DMD patients suggesting an association with the disease progression. 19 In our cohort of DMD patients, the time test data including 6MWD and NSAA were available for 27 of 30 patients (3 were non-ambulant). When we tested the association between the disease severity and the creatine/creatinine ratio, a significant correlation with both 6MWD and NSAA data (R = .76 P < 10 À6 , R = .75 P < 10 À5 , respectively) was observed ( Figure 4A).
The ratio was found to be elevated also in other forms of muscular dystrophy such as LGMD2A and LGMD2B ( Figure 4B). When testing for a possible relationship between metabolites levels and clinical scores in DMD patients, only creatinine showed a significant association with both 6MWT and NSAA data ( Figure 4C).   F-H). DMD is Duchenne muscular dystrophy, BMD is Becker muscular dystrophy, FSHD is facioscapulohumeral muscular dystrophy, DM1 is myotonic dystrophy type 1, LGMD2B and LGMD2A are limb-girdle muscular dystrophies type 2B and 2A. The healthy group is composed by the group of adult healthy people. * indicates significant differences between groups with Bonferroni-adjusted P-value <.05. # indicates differences between DMD and BMD patients with a P-value <.01 SPITALI ET AL.

| 2445
Creatinine, that represents its degradation product and is present in urine and blood, is usually considered a marker of renal function.
As in Duchenne patients, creatine is normally synthesized by the liver but not metabolized in muscles, high creatine and low crea- of what has been observed in dystrophic animal models. 33 Finally, citrulline levels were found to be reduced in DMD patients; interestingly, citrulline is being tested in a single-centre, randomized, placebo-controlled trial in combination with metformin. 34 This therapeutic approach aims to stimulate mitochondrial function and to compensate oxidative stress by increasing the production of nitric oxide (NO). In fact, NO is synthesized from the precursor arginine, which is in turn synthesized from citrulline. 35 It was recently shown that both arginine and citrulline can boost the production of NO in humans. 36 Restoration of citrulline serum levels and NO levels in DMD patients could be used as pharmacodynamics biomarkers to study the effect of this ongoing combination therapy.
While our study is controlled for food intake, other factors such as physical exercise, supplements, stress and time of the day could affect the observed profiles, and they were not investigated in this study. Additional studies in more controlled settings such as clinical trials are required to validate the identified associations.
Despite these limitations our study provides a list of biomarkers useful as potential candidates to evaluate the patients' disease progression in prospective natural history studies and response to therapy in dose-finding studies.