Quantitative tandem mass tag-based serum proteomics for longitudinal biomarker monitoring in Duchenne muscular dystrophy

Abstract

Background

Duchenne muscular dystrophy (DMD) is an X-linked recessive disorder characterized by progressive and severe muscle degeneration. Motor function tests are commonly used to evaluate treatment efficacy in clinical trials. However, they are subject to interobserver variability and may lack sensitivity for detecting early changes in disease progression. These limitations highlight the need for blood-based biomarkers to monitor disease status and progression. In this study, we used tandem mass tag-based mass spectrometry to quantify proteins in longitudinal serum samples from patients with DMD and to identify proteins associated with motor function performance.

Methods

Serum samples collected at three time points (baseline, 12, and 24 months) were obtained from participants in the FOR-DMD trial (NCT01603407) and processed for multiplexed analysis using TMT 6-plex isobaric tags and LC-MS/MS. Protein intensities were log2-transformed and analyzed using linear mixed-effects models to assess their associations with age and repeated functional outcome measurements, such as the North Star Ambulatory Assessment (NSAA) score, 6-minute walk test (6MWT), rise from supine velocity (RSV), and 10-meter run/walk velocity (10mRWV). P-values were adjusted for multiple comparisons, with FDR < 0.05 considered statistically significant.

Results

Mixed-model analysis identified 22 proteins associated with age and 77 proteins associated with at least 1 functional outcome, including 26 associated with 2 clinical outcomes after FDR correction. Most associations were observed with NSAA (73 proteins), followed by the 6MWT (28 proteins) and RSV (3 proteins). These proteins spanned multiple disease-relevant categories, including muscle-associated proteins, extracellular matrix (ECM), complement and inflammatory pathways, coagulation/hemostasis, carrier proteins, proteolysis, and cell adhesion.

Conclusion

Using longitudinal serum proteome profiles and clinical outcome data, we identified proteins that associate with age and functional outcomes, particularly NSAA and 6MWT, highlighting key molecular pathways in DMD disease progression.

Trial registration

The FOR-DMD clinical trial was registered on ClinicalTrials.gov (registration no. NCT01603407). First submission: 03/04/2012.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12014-026-09613-4.

Background

Duchenne muscular dystrophy (DMD) is a severe muscle-wasting disease caused by loss-of-function variants in the DMD gene that lead to a lack of dystrophin protein expression [1, 2]. The loss of dystrophin disrupts the dystrophin-glycoprotein complex (DGC), leading to membrane instability, calcium influx, mitochondrial death, chronic inflammation, and fibrotic replacement of muscle tissue. This cascade results in muscle weakness, loss of ambulation (by adolescence), and eventual respiratory/cardiac failure [3–7].

Current clinical trials predominantly enroll ambulatory patients in a younger age range and utilize functional motor assessments such as the 6-minute walk test (6MWT), North Star Ambulatory Assessment (NSAA), rise from supine velocity (RSV), and 10-meter run/walk velocity (10mRWV) to monitor disease progression and response to treatment [8–10]. While these functional tests assess muscle strength, they can be burdensome for young boys with DMD, depend on their cooperation, and may be insensitive to early changes in disease direction [11, 12]. Furthermore, these tests cannot be performed by very young (< 3 years old) and older patients who have lost functional abilities.

Blood-accessible biomarkers are emerging as attractive tools for monitoring disease progression and therapeutic response in Duchenne muscular dystrophy, offering a non-invasive alternative to traditional functional assessments [13, 14]. Despite these advances, however, robust serum biomarkers with prognostic value remain elusive. Early attempts to define such biomarkers used affinity-based multiplexing technologies, such as antibody-multiplexing bead technology targeting 118 proteins [15, 16]. More recently, affinity capture using commercially available aptamers (SomaScan) was used to quantify up to 7289 protein targets across two independent cohorts, identifying a subset of candidate biomarkers that could improve the assessment of disease progression [17]. However, these methods can only measure a targeted set of proteins, are susceptible to cross-reactivity, and show altered binding affinity due to protein isoforms and Single Nucleotide Polymorphisms (SNPs) [18, 19]. Mass spectrometry (MS) is another promising tool increasingly used in biomarker discovery and involves direct measurement of protein abundance by quantifying proteotypic peptides with high specificity [20]. Using isobaric tandem mass tags (TMT) or label-free approaches, serum samples can be analyzed at a depth of thousands of proteins without the reliance on pre-selected antibody or aptamer panel targets, thereby expanding the breadth of biomarker discovery [21, 22]. Our lab previously standardized a TMT-based protein quantification workflow, achieving a coefficient of variation (CV) below 20% for 97% of the proteins identified. Serum proteins from young, untreated DMD boys and age-matched healthy controls were quantified, and comparisons revealed that 38 proteins were elevated and 50 were decreased in the DMD group relative to healthy controls. Most elevated proteins were muscle-associated and proinflammatory, while several decreased proteins were associated with extracellular matrix remodeling and cell adhesion [22]. In this study, we characterized the longitudinal trajectories of serum proteins and explored their association with age and functional outcomes in a well-defined cohort of DMD participants enrolled in the FOR-DMD trial (NCT01603407). Our approach aims to address the need to define monitoring biomarker signatures that can serve as secondary outcome measures to support decision-making in future clinical trials and enhance the evaluation of therapeutic response.

Methods

Sample collection

Serum samples were collected from DMD patients enrolled in the FOR-DMD trial (Finding the Optimum Regimen of Corticosteroids for DMD) (ClinicalTrials.gov: NCT01603407; registration date: May 23, 2012). The DMD participants were enrolled under an Institutional Review Board-approved protocol with informed written consent from parents or legal guardians, and all methods were performed in accordance with the International Conference on Harmonisation guidelines for Good Clinical Practice and the World Medical Association Declaration of Helsinki. Blood samples were processed to serum using a standardized protocol and stored in aliquots at − 80 °C in polypropylene cryogenic vials (Thermo Scientific Nalgene). Tandem Mass Tag (TMT) proteomic analysis was performed on 42 serum samples from 14 patients, collected at baseline, 12, and 24 months. All participants were naïve at baseline but subsequently received one of three glucocorticoid regimens (daily prednisone, daily deflazacort, or intermittent prednisone) [10].

Protein depletion

Aliquots containing 10 µL of serum were subjected to immunodepletion of the 14 most abundant serum proteins using Pierce Top 14 spin-depletion columns (Thermo Fisher Scientific, IL, USA), according to the manufacturer’s protocol. Subsequently, samples were centrifuged to recover the depleted flow-through. Then, the total protein concentration of each depleted serum sample was measured using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, IL, USA) following the manufacturer’s protocol, and readings were obtained with the Varioskan LUX multimode reader (Thermo Fisher Scientific, IL, USA).

Protein digestion and TMT labeling

An aliquot from each depleted sample containing 10 µg of total proteins was brought to a final volume of 100 µL with 50 mM triethylammonium bicarbonate (TEAB). Proteins were denatured by adding 25 µL of 0.1% sodium dodecyl sulfate (SDS) and incubating at 40 °C for 15 min. Disulfide bonds were reduced with 7 µL of 200 mM tris (2-carboxyethyl) phosphine (TCEP) at 55 °C for 1 h, and cysteines were alkylated by adding 44 µL of 200 mM iodoacetamide (IAA) at 40 °C for 30 min in the dark. Proteins were then precipitated by adding 880 µL of ice-cold (– 20 °C) acetone, followed by centrifugation at 1,7000 x g (4 °C) for 30 min. The protein pellets were collected by gently removing the supernatant and re-dissolved in 100 µL of 50 mM TEAB buffer. For protein digestion, 4 µL of 0.1 µg/µL Pierce Trypsin/Lys-C Protease, MS Grade (Thermo Fisher Scientific, IL, USA), was added to achieve an enzyme-to-protein ratio of 1:25 (w/w). Samples were then incubated overnight at 37 °C in a ThermoMixer C (Eppendorf AG, Hamburg, Germany) set at 550 rpm. The resulting peptide concentration in each sample was quantified according to the manufacturer’s protocol using Pierce Quantitative Fluorometric Peptide Assay (Thermo Fisher Scientific, IL, USA).

Equal amounts of peptides from each sample were diluted to 100 µL with 50 mM TEAB and labeled with 6-plex TMT isobaric reagents (TMTsixplex kit, Thermo Fisher Scientific) according to the manufacturer’s protocol, with slight modifications. Each TMT reagent (0.8 mg) was reconstituted in 41 µL of anhydrous acetonitrile, and 100 µL of the peptide solution was added to the corresponding vial. Six serum samples from 2 patients were labeled per batch, for a total of 7 batches. Labeling reactions were carried out for 2 h at 27 °C with shaking at 450 rpm, after which 8 µL of 5% hydroxylamine was added to each sample to quench excess TMT reagent. Samples were incubated for an additional 1 h with shaking. For multiplexing, the labeled samples were pooled into a single 2 mL Protein LoBind tube (Eppendorf AG, Hamburg, Germany), dried, and stored at − 80 °C until high-pH reversed-phase fractionation.

High-pH reverse-phase peptide fractionation

The samples were reconstituted in 0.1% trifluoroacetic acid (TFA) and loaded onto C18 spin columns (Pierce High-pH Peptide Fractionation Kit, Thermo Fisher Scientific, IL, USA) for further fractionation, following the manufacturer’s protocol with minor modifications as described as follows: The columns were conditioned with acetonitrile (300 µL, twice), then washed with 0.1% TFA (300 µL, twice), with centrifugation at 1000 x g for 4 min at each step. Each sample was loaded onto the columns and centrifuged twice at 500 x g for 8 min to collect the flow-through (FT). The columns were then washed twice with 200 µL of water by centrifugation at 500 x g for 8 min to collect the wash. Peptides were eluted using 300 µL of elution buffer composed of ACN and 0.1% triethylamine (TEA), followed by centrifugation at 1000 x g for 8 min. Five consecutive fractions were collected using elution buffers containing 7.5%, 12.5%, 17.5%, 50%, and 80% ACN. The FT, wash, and the eluted fractions were dried using a SpeedVac vacuum concentrator and stored at − 80 °C until further analysis by LC-MS/MS.

Liquid chromatography-mass spectrometry analysis

Each peptide fraction was reconstituted in 0.1% formic acid and analyzed using an Ultimate 3000 RSLCnano system (Thermo Fisher Scientific) coupled to a Q-Exactive HF hybrid Orbitrap mass spectrometer (Thermo Fisher Scientific). About 1.2 µg of each peptide sample was loaded onto an Acclaim PepMap trap column (75 μm diameter, 20 mm length, 100 Å pore size, 3 μm particle size; Thermo Fisher Scientific) at 3.5 µL/min with 0.1% formic acid/2% acetonitrile, then eluted on an EASY-Spray Acclaim PepMap RSLC analytical column (75 μm diameter, 50 cm length, 100 Å pore size, 2 μm particle size; Thermo Fisher Scientific). Peptides were separated using a 145 min linear gradient ranging from 5% to 50% acetonitrile in 0.1% formic acid at 300 nL/min. MS/MS analysis was carried out in positive-ion data-dependent acquisition (DDA) mode. MS parameters included an electrospray voltage of ~ 2 kV, a capillary temperature of 350 °C, an automatic gain control (AGC) target of 3 × 10^6, and a maximum injection time of 80 ms. The MS scan range was set to 375–1800 m/z with a resolution of 60,000 at 200 m/z. For MS/MS, the scan resolution was set to 30,000, and the 20 most intense precursors with 2–5 charge states were isolated for high-energy collision-induced dissociation (HCD) fragmentation (5 × 10^5 AGC target, 110 ms maximum injection time). Additionally, dynamic exclusion was set to 45 s to prevent repeated sampling of the same peptides.

Proteome discoverer data processing

Raw mass spectrometry data were processed with Proteome Discoverer v2.2 (Thermo Fisher Scientific) as previously described with a few modifications [22]. Tandem mass spectra were searched against the Homo sapiens UniProt protein database (SwissProt, TaxID = 9606, May 2022, 42182 entries) using Sequest HT. A precursor mass tolerance of ± 10 ppm and a fragment mass tolerance of ± 0.05 Da were used. The following were set as dynamic modifications: carbamidomethylation of cysteine, oxidation of methionine, acetylation of peptide N-termini, TMT 6-plex modification of lysine side chains, and peptide N-terminus. We have specified TMT modifications on the N-terminus peptide and K-residues as dynamic to assess TMT labeling efficiency, which was ~ 97%, similar to previous studies [23, 24]. Carbamidomethylation of cysteine was also set as a dynamic modification to account for incomplete alkylation of cysteine-containing peptides [22]. Peptide-spectrum matches were filtered through a target-decoy strategy to maintain a false discovery rate (FDR) below 1% at both peptide and protein levels. Reporter-ion-based quantification was performed in Proteome Discoverer using the TMT-6plex workflow. Reporter ions (m/z 126, 127, 128, 129, 130, and 131) were quantified with a tolerance of 0.05 Da, with the integration method as the most confident centroid peak. Only unique peptides were used for quantification, and spectra with any missing reporter channels were excluded. Isotopic impurity corrections provided by the manufacturer were applied to reporter intensities, with a minimum signal-to-noise ratio of 1.5 for each channel. Normalization was carried out using the “total peptide amount” normalization mode in the consensus workflow. The overall protein abundance in each sample was calculated by summing the reporter ion abundances of the corresponding peptides in each fraction.

Statistical analysis

For each protein, we used linear mixed-effects models (LMEMs) to evaluate whether changes in protein levels were associated with concurrent changes in clinical performance, using the lme4 (v1.1–37) and lmerTest (v3.1–3) packages in R [25, 26]. Protein intensity values were log₂-transformed, and both protein and clinical outcomes were expressed as within-patient change from baseline. To determine associations with functional scores, an interaction model was fitted that included protein abundance, glucocorticoid treatment (daily or intermittent), and their interaction to account for treatment-specific variation in protein abundance. The model also incorporated age as a fixed effect, along with random effects to account for baseline differences between patients and across batch runs. P-values for the association between protein change and clinical change were determined using the F-test. To assess the effect of age on serum protein levels, LMEM was conducted for each protein using an additive model with age as the covariate, treatment as a fixed effect, and the same random effects; significance was indicated by the p-value of the age coefficient. Additionally, we compared our TMT signals against the 1.5 K SomaScan dataset generated from the same subset of samples [27]. For each overlapping protein, log₂-transformed TMT and SomaScan abundances were expressed as within-patient change from baseline. We then fitted an LMEM with SomaScan abundance as the outcome and TMT abundance as the predictor, including random effects for patient and TMT batch to account for repeated measures and batch effects. All p-values were adjusted across proteins using the Benjamini–Hochberg false discovery rate (FDR) method to control for multiple comparisons [28]. An adjusted FDR < 0.05 was considered statistically significant. Functional annotations such as biological process, cellular component and molecular function were obtained from UniProt database (https://www.uniprot.org).

Results

Longitudinal serum proteome profiling

Proteins from 42 serum samples, collected from 14 DMD patients aged 4 to 10 at three time points (baseline, 12, and 24 months), were quantified using TMT-based LC-MS/MS analysis. Five patients received daily prednisone (0.75 mg/kg/day), five received intermittent prednisone (0.75 mg/kg/day), and four received daily deflazacort (0.90 mg/kg/day). The patient and sample characteristics in this study are provided in Table 1. A total of 823 proteins were identified at an FDR < 1% (Supplemental Table S1). Of these, 256 proteins had 0% missing values across all 42 analyzed samples; 61 had 14% missing values; 77 had 28% missing values; 82 had 42% missing values; and the remaining proteins had more than 42% missing values. Focusing on the 256 proteins with 0% missing values, and after removing depleted proteins and keratin contaminants, 224 proteins were retained for further analysis.

Table 1.

Patient Demographics and Characteristics

Parameters	Baseline	Year 1	Year 2
Samples, n	14	14	14
Age in years mean (range)	5.76 (4.53–7.64)	6.76 (5.57–8.62)	7.76 (6.54–9.62)
Weight in Kg mean (range)	20.47 (16.00–26.40)	24.28 (17.00–35.80)	28.78 (18.70–43.80)
Height in cm mean (range)	110.65 (101.50–118.30)	114.91 (104.80–122.50)	119.79 (109.20–128.00)
BMI mean (range)	16.64 (14.11–18.86)	18.26 (13.63–23.86)	19.89 (13.75–28.03)
Daily Corticosteroid treatment n (%)	Naïve	9 (64.3%)	9 (64.3%)
Intermittent Corticosteroid treatment n (%)	Naïve	5 (35.7%)	5 (35.7%)
NSAA test mean (range)	21.29 (14–30)	25.00 (15–33)	22.93 (2–33)
6MWT test mean (range)	327.21 (189–481)	383.14 (261–500)	378.23 (0–525)
RSV test mean (range)	0.18 (0.07–0.36)	0.26 (0.11–0.36)	0.22 (0.0–0.40)
10mRWV test mean (range)	0.17 (0.12–0.26)	0.21 (0.12–0.34)	0.19 (0.0–0.34)

We first examined changes in the longitudinal trajectories of the 224 identified and quantified serum proteins, while accounting for treatment type. After correcting for multiple testing, 22 proteins showed significant changes in their longitudinal trajectories over time (FDR < 0.05; Supplemental Table S2), including 6 that increased and 16 that decreased (Fig. 1).

Fig. 1.

Volcano plot illustrating the strength and significance of protein associations with age. Each dot corresponds to a protein, plotted by its coefficient on the x-axis and the –log10 FDR on the y-axis. Red dots indicate proteins that positively associated with age (FDR < 0.05)

Figure 2 shows the longitudinal trajectories of exemplar proteins in individual participants. Carnosine dipeptidase 1 (CNDP1), matrix metalloproteinase-9 (MMP9), and myosin heavy chain IIa (MYH2) increased in DMD subjects between 4 and 10 years of age, except in two subjects where MYH2 remained stable over time. In contrast, myomesin 3 (MYOM3), adiponectin (ADIPOQ), and complement factor B (CFB) decreased with age in individual subjects. Among these, CFB and MYOM3 showed a more pronounced decline over time than ADIPOQ within the 4–10-year age range.

Fig. 2.

Longitudinal trajectories of proteins in individual DMD subjects, colored by treatment. The top panel (a-c) depicts proteins that increase with age, whereas the bottom panel (d-f) depicts proteins that decrease with age. Blue: intermittent corticosteroid treatment; red: daily corticosteroid

We then compared our TMT dataset to the 1.5 K SomaScan dataset previously generated from the same subset of samples [27]. Protein abundance values from both platforms were log₂-transformed and expressed as within-patient change from baseline. Their association was then evaluated using linear mixed-effects models. Of the 224 proteins retained in the TMT dataset, 137 overlapped with SomaScan (Supplemental Figure S1A). Among these overlapping proteins, 71 showed significant positive associations after multiple-testing correction (FDR < 0.05; Supplementary Table S2). Examples of proteins showing significant associations between the two platforms included CA1, MYOM3, MMP9, and ALDOA (Supplemental Figure S1B).

Association of serum proteins with functional outcomes

To enable reliable association analyses between the longitudinal trajectories of serum proteins and functional outcomes in DMD subjects, we focused on the 224 proteins with no missing values. Four outcome measures were used in this analysis: NSAA, 6MWT, RSV, and 10mRWV. The individual trajectories of clinical scores are shown in Fig. 3a. LMEM was used to determine whether changes in clinical scores from baseline were associated with changes in protein levels. The model included subject and batch run as random effects, with age and treatment type as fixed effects. A total of 77 proteins were associated with at least one clinical outcome. Out of all the 224 analyzed proteins, 73 showed a significant relationship with the NSAA score, 28 associated with 6MWT, and 3 associated with RSV (Fig. 3b). Among these proteins, 26 are associated with more than one clinical outcome. In patients receiving daily treatment, 25 proteins were positively associated with both NSAA and 6MWT scores. The breakdown of these candidate biomarkers included 5 ECM/cell adhesion proteins (ABI3BP, SPARCL1, CLEC3B, LYVE1, SERPINF1), 5 coagulation/hemostasis proteins (PROC, PROZ, KNG1, APOH, SERPINF2), 3 complement and inflammation proteins (C4BPB, C6, CD14), 1 cytoskeletal/actin-binding protein (GSN), 6 transport/carrier proteins (APOF, APOA2, APOD, AHSG, GC, AFM), and 5 proteins involved in proteolysis and oxidative stress (CNDP1, GPX3, PI16, FETUB, HABP2). Annexin A2 (ANXA2) was negatively associated. Only APOD was associated with all three clinical outcomes. No protein showed a significant association with 10mRWV. The complete list of proteins significantly associated with at least one clinical outcome, along with their gene ontology, is provided in Supplemental Table S3.

Fig. 3.

Association of serum protein with clinical motor function tests. A Longitudinal trajectory of clinical scores for the NSAA, 6MWT (m), RSV (m/s), and 10mRWV (m/s), colored by treatment. Each line represents an individual patient’s score over time. Blue: intermittent corticosteroid treatment; red: daily corticosteroid. B Venn diagram illustrating the number of protein biomarkers associated with 6MWT (blue), NSAA (red), and RSV (orange). C Scatter plot of proteins showing changes in protein levels (y-axis) and NSAA score (x-axis) from baseline, colored by treatment. Blue: intermittent corticosteroid treatment; red: daily corticosteroid

Discussion

Using TMT-based LC-MS/MS, we identified 823 proteins in longitudinal serum samples collected from 14 individuals with DMD at baseline, 12 months, and 24 months. The high proportion of missing values (ranging from 14% to > 42%) highlights the inherent variability of serum proteomics using a data-dependent acquisition (DDA) approach and underscores the need for stringent data filtering. By focusing on proteins quantified across all samples and time points, we obtained a final dataset of 224 proteins. Although reduced in size, this set provided a reliable and interpretable basis for downstream association analyses with changes in clinical outcomes in DMD.

In this study, we applied linear mixed-effects modeling to test associations of the longitudinal trajectories of these 224 proteins with functional outcomes. Due to the limited sample size and the absence of significant differences in clinical scores between daily deflazacort and prednisone regimens [10], two treatment groups (daily and intermittent) were included in the models. We identified 77 serum proteins associated with at least one clinical outcome measure, 26 of which were associated with two or more outcomes, primarily the NSAA and 6MWT. These candidate biomarkers encompassed muscle-related proteins, proteins involved in the ECM remodeling, complement cascade, proteolysis, transport and carrier function, coagulation and hemostasis, inflammation, cell adhesion, and oxidative stress.

ECM and cell adhesion proteins associated with NSAA included MMP9 and TIMP2, while CLEC3B, SERPINF1 and ANXA2 were additionally associated with 6MWT. MMP9 is a member of the MMP family involved in ECM remodeling. MMP-9 expression was elevated in DMD, exacerbating its pathogenesis by promoting ECM degradation and fibrosis [29]. Here, it was also found to be positively associated with age, as previously shown [30], while remaining unchanged in healthy controls [31]. MMP activity is regulated by endogenous tissue inhibitors of metalloproteinases (TIMPs), and this balance is important for tissue integrity [32]. TIMP2 was negatively correlated with dystrophin levels in Becker Muscular Dystrophy (BMD) muscle, suggesting a compensatory response to increased MMP activity in dystrophinopathies [33]. CLEC3B is a plasminogen-binding protein consisting of three noncovalently linked subunits of 181 amino acids each [34]. It was shown to be implicated in myogenesis, as it demonstrated a positive association with myofiber formation during skeletal muscle regeneration after injury in mdx mice, and in differentiating muscle cells in vitro [35]. Furthermore, its circulating levels were associated with the NSAA score in BMD, making it a plausible candidate for monitoring dystrophinopathies characterized by continuous degeneration–regeneration cycles [20]. Another extracellular protein, SERPINF1, is expressed in skeletal and smooth muscle tissues and has also been reported to promote muscle regeneration by stimulating satellite cell proliferation via ERK1/2, Akt, and STAT3 signaling [36]. This regenerative role could be particularly relevant in DMD, where chronic injury progressively exhausts the muscle regenerative capacity [37, 38]. ANXA2 is a calcium-dependent phospholipid-binding protein that contributes to membrane repair in muscular dystrophies as compensation for membrane fragility [39, 40]. DMD muscle biopsies and myotubes showed overexpression of ANXA2 and its presence in the ECM, due to annexin leakage from sarcolemmal damage [41]. Extracellular annexin contributes to fatty replacement of muscle tissue by promoting adipogenic differentiation of fibro-adipogenic progenitors (FAPs) in Limb-girdle muscular dystrophy type 2B (LGMD2B). Thus, a comparable mechanism could be at play in DMD, where persistent membrane damage promotes differentiation of FAPs, leading to fatty infiltration and muscle loss [42].

Muscle-associated proteins included GSN and MYH2. GSN is an actin-binding protein that severs and scavenges actin released into the circulation following muscle cell injury [43]. Previous TMT and SomaScan studies reported reduced circulating gelsolin in DMD versus controls, likely due to accelerated hepatic clearance of gelsolin–actin complexes [22, 44, 45]. Furthermore, it was associated with NSAA in BMD [20], and in this study, we reported significant associations with both NSAA and 6MWT in DMD. MYH2 is a fast-twitch skeletal muscle motor protein that was previously found to be downregulated in mdx muscle, reportedly due to shifts in fiber-type composition during regeneration [46, 47]. Here, MYH2 was associated with NSAA and showed a positive association with age, suggesting ongoing leakage into the circulation due to chronic muscle damage across the 4–10-year age range.

Proteolytic enzymes and proteins involved in oxidative stress included CNDP1 and GPX3, respectively (both associated with NSAA and 6MWT). CNDP1 is an enzyme found in the blood and brain that hydrolyzes carnosine, a dipeptide composed of beta-alanine and histidine. One of its primary roles is buffering intracellular pH, which helps delay acidosis during intense muscle activity, thereby prolonging performance before fatigue sets in [48]. Although the precise physiological role of CNDP1 remains unclear, it may facilitate histidine acquisition by breaking down carnosine obtained from meat into its constituent amino acids [49, 50]. Here, CNDP1 increased positively with age, in agreement with a previous study [16]. Another study showed that CNDP1 serum levels decreased in untreated DMD and subsequently increased following glucocorticoid exposure [44]. However, future work is needed to test how glucocorticoids modulate CNDP1 and its relation to circulating carnosine. GPX3 is an isoform of the GPx family that functions as an antioxidant by detoxifying hydrogen peroxide and organic hydroperoxides. DMD skeletal muscles typically exhibit signs of chronic oxidative stress, including significantly reduced total GSH levels, an elevated GSSG/GSH ratio, and increased GPx and glutathione reductase levels compared with healthy controls [51, 52]. Furthermore, studies have shown elevated GPx activity in dystrophic tissue relative to normal muscle and elevated serum GPx3 in mdx mice compared to wild-type controls, suggesting a compensatory upregulation of the GPx pathway in response to oxidative damage [53, 54].

Transport, coagulation, as well as inflammation and complement cascade signals, were also significantly associated with clinical outcomes. APOD is a transport glycoprotein involved in lipid trafficking and stress responses and has been shown to exhibit an anti-inflammatory effect, as its overexpression in mice decreased plasma interleukin-6 (IL-6) and tumor necrosis factor-α (TNF-α) levels, which, in turn, reduced the overabundant infiltration of T cells [55]. Furthermore, it is involved in managing oxidative stress as mice with impaired ApoD function exhibited elevated peroxide lipid levels in their brains, increasing their susceptibility to oxidative stress [56]. Thus, its role in inflammation and oxidative stress could help elucidate its role in DMD pathogenesis, where it was found to be upregulated in mdx skeletal muscle [47]. We also found CD14 to be significantly associated with both NSAA and RSV, which mediates innate immune signaling through MyD88/TIRAP/TRAF6 pathways to activate NF-κB and downstream cytokine responses [57, 58]. However, further studies are needed to validate its significance in DMD. PROC is a vitamin K-dependent protein secreted by the liver that mediates inflammation through its anticoagulant and anti-inflammatory activities [59, 60]. It is also implicated to be corticosteroid-responsive, as it was significantly altered in pre- vs. post-corticosteroid-treated DMD patients. Therefore, it could serve as a pharmacodynamic biomarker for the anti-inflammatory response to corticosteroid treatment [61]. Finally, complement proteins CFB and CFH were significantly associated with NSAA, and C4BPB and C6 associated with both NSAA and 6MWT. CFB and C1S decreased over time in our DMD cohort, whereas these proteins did not change significantly in healthy controls between ages 1 and 14 years [31]. This suggests that the decline of CFB and C1S in DMD may be associated with disease progression or a response to corticosteroid treatment. However, the effect of corticosteroids cannot be ruled out, as all DMD patients begin corticosteroid therapy shortly after diagnosis [62].

Comparison of the TMT dataset with SomaScan data on the same subset of serum samples [27] showed significant concordance for several biomarkers identified in this study, including MYOM3, CNDP1, MMP9, APOD, GSN, CD14, CLEC3B, and CFB. Although TMT identified and consistently quantified fewer proteins across all samples than the SomaScan platform, it led to the identification of additional candidate biomarkers associated with clinical outcomes, including GPX3, C6 and PROC, for which no aptamers were available in the SomaScan panel. However, some proteins, such as ANXA2 and SERPINF1, showed weak agreement, possibly due to differences in SomaScan aptamer binding affinities for specific protein isoforms [19] and/or to proteins being bound in complexes, in contrast to the TMT workflow, in which proteins are denatured prior to analysis [22]. In addition, our study was limited by the absence of a common reference channel for abundance normalization across TMT sets. Another limitation of the study is the small sample size. DMD is a rare disease, and access to longitudinal serum samples and clinical data is often limited. Nevertheless, validation of these candidate biomarkers in future independent cohorts will be essential to confirm their clinical utility. Furthermore, despite an initial panel of 823 identified proteins, only 224 proteins were consistently quantified across all samples, likely reflecting data-dependent acquisition (DDA) biases toward higher-abundance proteins and reduced sensitivity for lower-abundance candidates compared to data-independent acquisition (DIA) [63], and the limited in-depth fractionation used in this study. As a result, potentially informative proteins present at low concentrations may not have been captured. Despite these constraints, a key strength of this study is the identification of multiple disease-relevant circulating protein biomarkers, several of which have been reported previously in prior muscular dystrophy studies, supporting their potential utility as monitoring biomarkers in DMD.

Conclusion

In this study, we identified 26 circulating proteins that were significantly associated with at least 2 outcomes in DMD, namely NSAA and 6MWT. These proteins spanned multiple biological categories, including muscle-associated proteins, extracellular and proteolytic enzymes, regulators of coagulation and hemostasis, carrier proteins, and proteins involved in inflammation and the complement cascade. Collectively, these serum protein candidates’ signatures may provide a valuable tool for monitoring disease progression in DMD. With further validation, it could also aid in go/no-go decision-making in clinical trials by enabling a more objective assessment of therapeutic response. However, additional studies are required to evaluate the responsiveness of these biomarkers to different treatment modalities in DMD and to establish their utility across diverse contexts of use.

Abbreviations

DMD

Duchenne muscular dystrophy

TMT

Tandem mass tag

FOR-DMD

Finding the optimum regimen for duchenne muscular dystrophy

LMEM

Linear mixed-effects model

NSAA

North star ambulatory assessment

6MWT

Six-minute walk test

RSV

Rise from supine velocity

10mRWV

10-meter run/walk velocity

FDR

False discovery rate

ECM

Extracellular matrix

DGC

Dystrophin-glycoprotein complex

SNPs

Single Nucleotide Polymorphisms

CV

Coefficient of variation

BCA

Bicinchoninic acid

TEAB

Triethylammonium bicarbonate

SDS

Sodium dodecyl sulfate

TCEP

Tris (2-carboxyethyl) phosphine

IAA

Iodoacetamide

ACN

Acetonitrile

TFA

Trifluoroacetic acid

FT

Flow-through

TEA

Triethylamine

DDA

Data-dependent Acquisition

AGC

Automatic gain control

HCD

High-energy collision-induced dissociation

CA1

Carbonic anhydrase 1

MYOM3

Myomesin 3

MMP9

Matrix metalloproteinase-9

ALDOA

Fructose-bisphosphate aldolase A

CNDP1

Carnosine dipeptidase 1

MYH2

myosin heavy chain IIa

ADIPOQ

Adiponectin

CFB

Complement factor B

ABI3BP

Target of nesh-sh3

SPARCL1

Sparc-like protein 1

CLEC3B

Tetranectin

LYVE1

Lymphatic vessel endothelial hyaluronic acid receptor 1

SERPINF1

Pigment epithelium-derived factor

PROC

Vitamin K-dependent protein c

PROZ

Vitamin K-dependent protein z

KNG1

Kininogen-1

APOH

Beta-2-glycoprotein 1

SERPINF2

Alpha-2-antiplasmin

C4BPB

C4b-binding protein beta chain

C6

Complement component c6

CD14

Monocyte differentiation antigen CD14

GSN

Gelsolin

APOF

Apolipoprotein f

APOA2

Apolipoprotein a-ii

APOD

Apolipoprotein d

AHSG

Alpha-2-hs-glycoprotein

GC

Vitamin D-binding protein

AFM

Afamin

GPX3

Glutathione peroxidase 3

PI16

Peptidase inhibitor 16

FETUB

Fetuin-b

HABP2

Hyaluronan-binding protein 2

ANXA2

Annexin A2

TIMP2

metalloproteinase inhibitor 2

TIMPs

Inhibitors of metalloproteinases

FAPs

Fibro-adipogenic progenitors

LGMD2B

Limb-girdle muscular dystrophy type 2B

GPx

Glutathione peroxidase

GSH

Glutathione

GSSG

Glutathione disulfide

IL-6

Interleukin-6

TNF-α

Tumor necrosis factor-α

MYD88

Myeloid differentiation primary response 88

TIRAP

TIR domain-containing adaptor protein

TRAF6

Tumor necrosis factor receptor-associated factor 6

NF-κB

Nuclear factor-κB

CFH

Complement factor H

C1S

Complement c1s subcomponent

DIA

Data-independent acquisition

Author contributions

AN, ER, and DC processed the samples and generated the data presented. Statistical analyses were performed by AN, CD and RT. MG collected the FOR-DMD subject demographics and serum samples for biomarker analysis. FOR-DMD investigators of the Muscle Study Group contributed to clinical data and serum sample collection. AN and YH wrote the manuscript. The final manuscript was reviewed and approved by all authors prior to submission.

Funding

Research reported in this publication was supported by the National Institute Of Neurological Disorders And Stroke of the National Institutes of Health under award number #R61NS119639 (Hathout, Dang, Spitali, Guglieri, Al-Khalili Szigyarto).

Data availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository [64], with the dataset identifier: PXD076311 and DOI: 10.6019/PXD076311. Scripts for data analysis are available on request.

Declarations

Ethics approval and consent to participate

The FOR-DMD study was registered at ClinicalTrials.gov (registration no. NCT01603407). Serum samples and clinical data were collected following approval from the competent ethics committee at each institution in accordance with the International Conference on Harmonisation guidelines for Good Clinical Practice and the World Medical Association Declaration of Helsinki. Central Institutional Review Board (IRB) approval for the original FOR-DMD study was obtained from the University of Rochester Ethics Committee/IRB and the Newcastle Clinical Trial Unit (NCTU) before the distribution of study documents to each investigative site and the enrollment of participants. Informed written consent for participation in biomarker research was obtained from the parents or legal guardians of participants at the time of enrollment in the clinical study.

Competing interests

The authors declare no competing interests.