A novel deep proteomic approach in human skeletal muscle unveils distinct molecular signatures affected by aging and resistance training

MA versus Y phenotypes, and MA responses to eight weeks of resistance training

MA were significantly older than Y participants (Y: 22 ± 2 years old, MA: 56 ± 8 years old; p<0.001). However, compared to Y participants, pre-intervention MA participant body mass (Y: 69.9 ± 14.3 kg, MA: 76.6 ± 14.7 kg; p=0.374), percent body fat (Y: 35.5 ± 4.6 %, MA: 27.6 ± 6.6 %; p=0.071), and VL mCSA (Y: 19.0 ± 2.7 cm², MA: 18.4 ± 5.2 cm²; p=0.818) were not significantly different.

In MA participants, the eight-week training protocol non-significantly increased VL mCSA (PRE: 18.4 ± 5.2 cm², POST: 19.6 ± 4.2 cm²; p=0.066) and significantly increased isokinetic knee extensor strength at 60 degrees/s (PRE: 141 ± 79 N•m, POST: 153 ± 71 N•m; p=0.048).

Characteristics of proteins identified in the MyoF and non-MyoF fractions

A total of 6,445 non-MyoF proteins and 4,421 MyoF proteins were identified in at least one participant (Figure 1A). Across all participants, the number of non-MyoF proteins detected averaged to be 5,645 ± 266 (range: 4,888–5,987) and the number of MyoF proteins detected averaged to be 2,611 ± 326 (range: 1,944–3,101). Notably, there was a numerically lower number of MyoF proteins detected in MA (pre-intervention) versus Y participants that approached statistical significance (p=0.066), and a non-significant greater number of non-MyoF proteins detected in MA (pre-intervention) versus Y participants (p=0.112; Figure 1C).

Figure 1. MyoF and non-MyoF protein characteristics. Legend: Data presented for MA (pre-intervention) and Y include the total number of proteins identified in each fraction (A), the top 5 protein classifications from each fraction (B), the number of MyoF and non-MyoF proteins detected within and between age cohorts (C), the top 15 highly enriched MyoF proteins (D), and the top 15 highly enriched non-MyoF proteins (E). Data in panels c and d are presented as means with standard deviation bars, and y-axes were scaled as log₁₀ for improved visualization. Symbols: *, indicates multiple histone isoforms were congregated into these two targets based on sequence similarities. Protein names for gene symbols in panel c: ACTA1, Actin Alpha 1, Skeletal Muscle; TNNT1, Troponin T1; MYH13/7/2/8/4/1, myosin heavy chain isoforms 13/7/2/8/4/1; TTN_iso12, titin, isoform 12; MYL3/1/, myosin light chain isoforms 3/1; H2BC12, Histone H2B type 1-K/C/E/F/G/I/type F-S; H2AC20, Histone H2A type 2-A/C. Protein names for gene symbols in panel d: HBA2, Hemoglobin subunit alpha; HBB, Hemoglobin subunit beta; VDAC1, Voltage-dependent anion-selective channel protein 1; FHL1, Four and a half LIM domains protein 1; LDHA, L-lactate dehydrogenase A chain; MB, myoglobin; RTN2, Isoform RTN2-C of Reticulon-2; VDAC2, Voltage-dependent anion-selective channel protein 2; ATP2A2, Sarcoplasmic/endoplasmic reticulum calcium ATPase 2; ATP5MG, ATP synthase subunit g, mitochondrial; HBD, Hemoglobin subunit delta; VDAC3, Voltage-dependent anion-selective channel protein 3; IDH2, Isocitrate dehydrogenase [NADP], mitochondrial. Other note: CONP00761/15636 are non-annotated proteins found in both fractions.

A total of 4,228 proteins overlapped in both fractions yielding 2,217 unique non-MyoF proteins, 193 unique MyoF proteins, and 6,638 unique proteins identified. Using the PANTHER Classification System classifications, the top five protein classes of MyoF proteins, non-MyoF proteins, and proteins in both fractions are presented in Figure 1B. The top 15 enriched MyoF and non-MyoF proteins in MA (pre-intervention) and Y are presented in Figure 1D, 1E. None of the 15 MyoF or non-MyoF proteins met the p<0.01 significance criteria between age cohorts.

MYH isoform peptide identification information

Myosin heavy chain isoforms have been intensely studied in human skeletal muscle for fiber typing purposes and prominent isoforms include the slow-twitch type I isoform (encoded by the MYH7 gene) as well as the fast-twitch IIA (encoded by the MYH2 gene) and IIX (encoded by the MYH1 gene) isoforms [28]. However, other MYH isoforms were highly enriched in the MyoF fraction according to data presented in Figure 1C. Because of this, we opted to provide the peptide sequences used for detecting some of these isoforms in Table 1 below.

Additional MyoF and non-MyoF protein characteristics

Aside from using PANTHER protein classifications, we wanted to present the depth of proteins detected based on processes relevant to muscle biology. Through manual interrogation we found that both fractions contained proteins found in nuclei that regulate chromatin structure and gene expression (Figure 2A), mitochondrial proteins related to oxidative phosphorylation and mitochondrial protein synthesis (Figure 2B), proteolysis-related proteins (Figure 2C), proteins localized to the sarcolemma and solute carrier proteins (Figure 2D), exercise-relevant phosphosignaling proteins (Figure 2E), and proteins associated with protein synthesis (Figure 2F).

Figure 2. Additional protein classifications. Legend: Each fraction was found to contain nuclear proteins that regulate chromatin structure and gene expression (A), mitochondrial proteins related to oxidative phosphorylation and mitochondrial protein synthesis (B), proteolysis-related proteins (C), proteins localized to the sarcolemma and solute carrier proteins (D), exercise-relevant phosphosignaling proteins (E), and proteins associated with protein synthesis (F). Abbreviations: Mito., mitochondrial; Ub, ubiquitin; MAPKs, mitogen of activated protein kinases; mTORC1, mechanistic target of rapamycin complex 1; AMPK, AMP-activated protein kinase; CAMK, Ca²⁺/calmodulin-dependent protein kinase; Euk., eukaryotic.

Also notable are the number of proteins exclusively expressed in either the MyoF fraction (n=141) or non-MyoF fraction (n=2,091). Aside from the non-MyoF fraction containing ~15-fold more unique proteins, the top 5 protein classifications of proteins exclusive to each fraction differed. In this regard, the top protein classes for proteins only found in the MyoF fraction included: gene-specific transcriptional regulator (PC00264) (n=15), chromatin/chromatin-binding, or -regulatory protein (PC00077) (n=14), RNA metabolism protein (PC00031) (n=11), DNA metabolism protein (PC00009) (n=5), transmembrane signal receptor (PC00197) (n=3), and metabolite interconversion enzyme (PC00262) (n=3). The top protein classes for proteins only found in the MyoF fraction included: metabolite interconversion enzyme (PC00262) (n=307), protein modifying enzyme (PC00260) (n=237), protein-binding activity modulator (PC00095) (n=116), scaffold/adaptor protein (PC00226) (n=101), and RNA metabolism protein (PC00031) (n=101).

MyoF and non-MyoF protein differences in MA (pre-intervention) versus Y

In MA (pre-intervention) and Y participants, 112 of the 4,421 MyoF proteins met the p<0.01 significance threshold equating to ~2.5% of the MyoF proteome being affected with aging. 111 of these 112 MyoF proteins were significantly greater in Y versus MA participants (see Table 2 for the top 15 proteins), and only one MyoF protein was significantly greater in the MA versus Y participants (TMPO, Lamina-associated polypeptide 2 isoforms beta/gamma).

Bioinformatics of the 111 MyoF proteins that were significantly greater in the Y versus MA participants indicated that no biological processes were predicted to be affected between age cohorts.

When performing the MA (pre-intervention) and Y participant comparisons for non-MyoF proteins, 543 proteins met the p<0.01 significance threshold equating to ~8.4% of the non-MyoF proteome being affected with aging. 96 of these 543 non-MyoF proteins were significantly greater in Y versus MA participants, and 447 non-MyoF proteins were significantly greater in MA versus Y participants. Table 3 contains the top 15 proteins that showed differential abundances between Y versus MA participants.

Bioinformatics of the non-MyoF proteins that were significantly different between the Y versus MA participants are presented in Table 4 below. Notably, more pathways were predicted to be upregulated in MA versus Y participants.

Alternative proteins isoforms in the MyoF and non-MyoF fractions of MA and Y

The enhanced depth of detection provided by proteomics revealed the presence of numerous isoforms in both protein fractions; specifically, there were 175 isoforms for 82 MyoF proteins and 375 isoforms for 173 non-MyoF proteins. The MyoF proteins with the most isoforms included titin (TTN), myosin-binding protein C (MYBPC), and MICOS complex subunit MIC60 (IMMT); each of these targets had four isoforms detected. Given the vast research interest in titin [29, 30], associated isoform data are plotted in Figure 3A; notably no significant aging or training effects were noted (p>0.01 for all comparisons). The non-MyoF proteins with the most isoforms included Gelsolin (GSN, four isoforms), IMMT (4 isoforms), and Reticulon-4 (4 isoforms); again, no significant age effects were noted for these targets (data not plotted).

Figure 3. MyoF and non-MyoF alternative protein isoform differences detected with proteomics. Legend: Data presented for Y and MA (pre- and post-intervention) include the identified titin isoforms in the MyoF fraction (A), alternative MyoF protein isoforms affected by aging (B), and alternative non-MyoF protein isoforms affected by aging (C), and small nuclear ribonucleoproteins that make up spliceosomes between cohorts (D). Data are presented as mean ± standard deviations for individual protein spectra values (normalized to total run spectra values) and y-axes were scaled as log₁₀ for improved visualization. Symbols: #, indicates lower in MA versus Y at one or both time points (p<0.01); *, indicates greater in MA versus Y at one or both time points (p<0.01); Φ, indicates greater in MA versus Y at one or both time points for panel d only (p<0.05). Notes: (ND), indicates that the isoform number was not provided from the Uniprot’s Homo Sapiens reference database (UP000005640_9606).

We plotted significantly different alternative protein isoform abundances between Y and MA in both protein fractions given that bioinformatics on the non-MyoF fraction indicated “regulation of RNA splicing” (GO:0043484) was predicted to be upregulated in the older cohort (Figure 3B, 3C). MA (pre-intervention) versus Y comparisons indicated that only three alternative MyoF protein isoforms were different between age groups (all higher in Y, p<0.01). However, 14 alternative non-MyoF protein isoforms were significantly different between age groups (11 higher in MA, p<0.01), and while not depicted in Figure 3C, 32 additional alternative non-MyoF protein isoforms were numerically different between age groups (25 higher in MA, p<0.05). Further, when examining the abundances of small nuclear ribonucleoproteins belonging to spliceosome complexes in the non-MyoF fraction, three reached the p<0.01 significance threshold as being more enriched in MA (pre-intervention) versus Y participants (SNRP200, SNRPE, SNRPF), and several others were numerically greater in MA participants (SNRP40/70/A/A1/B/C/D2/D3, p<0.05; Figure 3D). Notably, training did not alter the expression of any SNRP in Figure 3D (p>0.05 for all), and most of these proteins were not detected in the MyoF fraction.

MyoF and non-MyoF protein differences prior to and following resistance training in MA participants

In MA, knee extensor resistance training significantly altered 13 MyoF proteins (11 upregulated and two downregulated, p<0.01; Figure 4A), and 64 non-MyoF proteins (56 upregulated and eight downregulated, p<0.01; Figure 4B, 4C). These alterations represented ~0.3% of the MyoF proteome and ~1.0% of the non-MyoF proteome. Bioinformatics within each fraction were attempted, albeit no pathways were predicted to be significantly affected.

Figure 4. MyoF and non-MyoF proteins altered with resistance training in MA participants. Legend: Data presented for MA prior to and following eight weeks of knee extensor training include proteins in the MyoF fraction (11 up-regulated, 2 down-regulated; (A), the top 15 up-regulated proteins in the non-MyoF fraction (B), and all 8 down-regulated proteins in the non-MyoF fraction (C). Data are presented as mean ± standard deviations for individual protein spectra values (normalized to total run spectra values), and y-axes were scaled as log₁₀ for improved visualization. Notes: No biological processes were predicted to be affected with training based on these alterations.

Proteolysis targets manually interrogated in both fractions

Based on bioinformatics indicating that proteostasis was predicted to be altered with aging (Table 3), we manually interrogated proteolysis-related protein targets (i.e., calpain-1/2, and the summed spectra of 26S proteasome subunits). Figure 5 contains these targets including the 26S proteasome subunits detected in the MyoF and non-MyoF fractions (panels a/d), calpain-1 (panels b/e) and calpain-2 (panels c/f). The summed spectra of the 34 detected 26S proteasome subunits in the MyoF fraction was significantly lower in MA at both time points versus Y participants. Both calpains were also numerically lower in the MyoF fraction of MA at both time points versus Y participants (CAPN2 was significant in MA pre-intervention versus Y, p<0.01), and training did not significantly affect either protein. In the non-MyoF fraction, the summed spectra of the 42 detected 26S proteasome subunits was not significantly different between Y versus MA participants at either time point. However, both calpains were higher in the non-MyoF fraction of MA at both time points versus Y participants (CAPN1 was significant in MA pre-intervention versus Y, p<0.01), and training did not significantly affect either protein.

Figure 5. MyoF and non-MyoF proteasome and calpain proteins. Legend: Data presented for Y and MA prior to and following eight weeks of knee extensor training include proteasome subunits and calpains 1/2 in the MyoF fraction (A–C) and non-MyoF fraction (D–F). Data are presented as mean ± standard deviations for individual protein spectra values (normalized to total run spectra values). Symbols: #, indicates lower in MA versus Y at one or both time points (p<0.01); *, indicates greater in MA versus Y at one or both time points (p<0.01). Notes: Certain p-values not meeting the significance criteria for these data were presented due to the visual differences observed between cohorts.

Ethical approval and study design

Muscle specimens were obtained from two studies whereby approval was obtained from the Auburn University Institutional Review Board. The first protocol in untrained MA participants (approved protocol #21-461 MR 2110) involved investigating the effects of a dietary supplement (312 mg of combined Wasabia japonica extract, theacrine, and copper (I) niacin chelate) versus a placebo on potential blood marker responses over an eight-week period. A unilateral leg resistance training (two days/week) protocol was implemented to perform non-supplementation secondary analyses as presented herein. The six MA participants included in the current study were in the placebo group; thus, no confounding effects of dietary supplementation were expected. Additionally, both the pre-intervention and post-intervention biopsies were collected the same time of day following at least a four-hour fast, and the post-intervention muscle biopsy was collected 72 hours following the last training bout. Y participant muscle tissue was banked from a prior study examining how ten weeks of daily peanut protein supplementation affected resistance training outcomes in untrained individuals (approved protocol #19–249 MR 1907) [51]. Notably, muscle tissue from these participants was collected in the basal state prior to the intervention. Hence, again, there were no potential confounding effects of supplementation. Study procedures for both projects were in accordance with the most recent revisions of the Declaration of Helsinki except for the MA study not being pre-registered as a clinical trial.

Study design and training paradigm in MA participants

Knee extensor resistance training

The resistance training intervention consisted of supervised unilateral leg extensions (two days/week for eight weeks), and the intervention was preceded and followed by strength and VL muscle assessments (described in later paragraphs). All MA participants trained their right legs whereby each training session consisted of five sets of 12 repetitions. The beginning training load was established at ~40% of the participants’ three-repetition maximum (3RM). After each set, participants verbally articulated their perceived repetitions in reserve (RIR) [52], and training load was adjusted accordingly. RIR values of 0-2 after a set resulted in no training load change in each session. RIR values of 3-5 for consecutive sets resulted in the training load being increased by 5-10%. For RIR values ≥6 after one set, the training load was increased by 10-20%. If the weight could not be performed with full range of motion, or the participant could not complete 12 repetitions for a given set, the training load was decreased accordingly.

Strength testing

The first and last workout of the eight-week training paradigm consisted of maximal leg extensor-flexion torque assessments using isokinetic dynamometry (Biodex System 4; Biodex Medical Systems, Inc., Shirley, NY, USA) and 3RM leg extensor strength testing. Prior to dynamometer testing, the participant’s lateral epicondyle was aligned with the axis of the dynamometer’s lever arm, and the hip was positioned at 90°. The participant’s shoulders, hips, and leg were strapped and secured for isolation during testing. Following three warm-up trials at a submaximal effort, participants completed five maximal voluntary isokinetic knee extension and flexion actions at 60 degrees/second. Participants were provided verbal encouragement during each contraction. The isokinetic contraction resulting in the greatest peak torque value was used for analyses. Approximately five minutes following isokinetic dynamometry testing, participants performed 3RM strength testing using a free-weight apparatus. Prior to testing, participants were given a warm-up load and instructed to complete 10 repetitions. After participants recorded their RIR for the warmup set, the weight was adjusted accordingly for another warm-up set of five repetitions. RIR was recorded again to determine the participants starting load for a 3RM attempt. The load was incrementally increased 5-10% per 3RM attempt until 3RM testing concluded, indicated by failure of full range of motion on any of the repetitions, or if RIR recorded was 0. Participants were allowed a full three minutes of recovery between attempts. The isokinetic dynamometry and 3RM testing described was similar for both the first and final workout.

Testing sessions in MA participants

Urine specific gravity testing for hydration

Participants performed a testing battery prior to the start of training (PRE) and 3-5 days following the last resistance training workout (POST). Participants arrived for testing at a minimum of 4 hours fasted and well hydrated. Upon arrival participants submitted a urine sample (~5 mL) for urine specific gravity assessment (USG). Measurements were performed using a handheld refractometer (ATAGO; Bellevue, WA, USA). USG levels in all participants were ≤ 1.020, indicating sufficient hydration [53].

Body composition testing

Body composition was assessed using multi-frequency bioelectrical impedance analysis (InBody 520, Biospace, Inc., Seoul, Korea). From the scan, body fat percentage was recorded. Previously determined test-retest reliability yielded an intraclass correlation coefficient (ICC_3,1) of 0.99, standard error of the measurement (SEM) of 0.87%, and minimal difference (MD) of 1.71% for body fat percentage.

Ultrasonography assessment for muscle morphology

A detailed description of VL assessments using ultrasonography has been published previously by our laboratory [54, 55]. Briefly, real-time B-mode ultrasonography (NextGen LOGIQe R8, GE Healthcare; Chicago, IL, USA) using a multifrequency linear-array transducer (L4-12T, 4–12 MHz, GE Healthcare) was used to capture VL muscle cross-sectional area (mCSA). Prior to scans, the mid-thigh location was determined by measuring the total distance from the mid-inguinal crease in a straight line to the proximal patella, with the knee and hip flexed at 90°, a mark was made using a permanent marker at 50% of the total length. From that location, a permanent marker was used transversely to mark the mid-belly of the VL. This marking is where all pre-intervention ultrasound images were taken as well as the muscle biopsy (described below). All post-intervention images were taken at the pre-intervention biopsy scar to ensure location consistency between scans. During mCSA scans, a flexible, semirigid pad was placed around the thigh and secured with an adjustable strap to allow the probe to move in the transverse plane. Using the panoramic function of the device (LogicView, GE Healthcare), images were captured starting at the lateral aspect of the VL and moving medially until rectus femoris was visualized, crossing the marked location. All ultrasound settings were held constant across participants and laboratory visits (frequency: 10 MHz, gain: 50 dB, dynamic range: 75), and scan depth was noted and held constant across time points per participant. Images were downloaded and analyzed offline using ImageJ software (National Institutes of Health, Bethesda, MD, USA). All ultrasound images were captured and analyzed by the same investigators at each timepoint. Previously determined test-retest reliability on 10 participants measured twice within 24 hours (where BAR captured images and JSG analyzed images) yielded an intraclass correlation of 0.99 and standard error of measurement of 0.60 cm².

Collection of muscle tissue

Muscle biopsies from all participants were obtained from the mid-belly of the right VL, and sampling time of day was standardized for MA participants at pre and post resistance training intervention. Lidocaine (1%, 1.0 mL) was injected subcutaneously above the skeletal muscle fascia at the previously marked location. After five minutes of allowing the anesthetic to take effect, a small pilot incision was made using a sterile Surgical Blade No. 11 (AD Surgical; Sunnyvale, CA, USA), and the 5-gauge biopsy needle was inserted into the pilot incision ~1 cm below the fascia. Approximately 30-50 mg of skeletal muscle was removed using a double chop method and applied suction. Following biopsies, tissue was rapidly teased of blood and connective tissue, placed in pre-labeled foils, flash frozen in liquid nitrogen, and subsequently stored at −80° C until processing described below.

MyoF and non-MyoF protein fractionation

The MyoF and non-MyoF protein fractions were isolated per methods published by our laboratory and others [23, 56]. On the day of homogenization, muscle tissue was powdered on a liquid nitrogen-cooled ceramic mortar and pestle. Approximately 30 mg of tissue was homogenized using tight-fitting pestles in 500 μL of 25 mM Tris, pH 7.2, 0.5% Triton X-100, with added protease inhibitors (Promega, cat# G6521; Madison, WI, USA). Samples were centrifuged at 1,500 g for 10 minutes at 4° C, supernatants (non-MyoF fraction) were transferred to new 1.7 mL tubes, and tubes were stored at -80°C until shipment on dry ice to Seer, Inc. Remaining MyoF pellets were kept on ice and thoroughly aspirated with micro-pipet tips to remove residual supernatant. Thereafter, 300 μL of solubilization buffer was added which contained 20 mM Tris-HCl, pH 7.2, 100 mM KCl, 20% glycerol, 1 mM DTT, 50 mM spermidine with added protease inhibitors (Promega, cat# G6521). Samples were then homogenized using tight-fitting pestles and stored at -80°C until shipment on dry ice to Seer, Inc.

MyoF and non-MyoF proteograph assays and proteomics

Proteograph assay

Proteomics analysis was performed at Seer, Inc. (Redwood City, CA, USA). For each sample, 250 μL of received sample was subjected to the Seer Proteograph Assay protocol. After loading samples onto the SP100 Automation Instrument, protein corona formation and processing was initiated to generate desalted purified peptides for protein identification using Reversed Phase (RP) LC-MS. To form the protein corona, Seer’s proprietary NPs were mixed with the samples and incubated at 37° C for 1 hour. Unbound proteins were removed prior to downstream wash, reduction, alkylation, and protein digestion steps which were performed according to Seer’s Proteograph Assay protocol [25].

LC-MS configuration

Peptides obtained from each of the five NP mixtures were separately reconstituted according in a solution of 0.1% formic acid and 3% acetonitrile [57] spiked with 5 fmol μL PepCalMix from SCIEX (Framingham, MA, USA). Reconstitution volumes varied by NP types to allow for constant peptide quantity for MS injection between samples regardless of starting volume (240 ng: NP1, 400 ng: NP2, 360 ng: NP3, 120 ng: NP4, and 320 ng: NP5). 4 μL of each sample were analyzed with a Ultimate3000 RLSCnano LC system coupled with a Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific; Waltham, MA, USA). Peptides were loaded on an Acclaim PepMap 100 C18 (0.3 mm ID × 5 mm) trap column and then separated on a 50 cm μPAC analytical column (PharmaFluidics, Zwijnaarde, Belgium) at a flow rate of 1 μL/min using a gradient of 5–25% solvent B (100% ACN) mixed into solvent A (100% water) over 26 minutes. The mass spectrometer was operated in Data Independent Acquisition (DIA) mode using 10 m/z isolation windows from 380-1200 m/z and 3-second cycle time. MS1 scans were acquired at 60k resolution and MS2 at 30k resolution.

Data processing

DIA LC-MS data were processed using Proteograph Analysis Suite (PAS) v2.1 (Seer, Inc) using the DIA-NN search engine (version 1.8.1) in library-free mode searching MS/MS spectra against an in silico predicted library based on Uniprot’s Homo Sapiens reference database (UP000005640_9606, download December 9, 2022). Library-free search parameters included trypsin digestion allowing for one missed cleavage, N-terminal methionine excision, fixed modification of cysteine carbamidomethylation, peptide length of 7-30 amino acids, precursor range of 300-1800 m/z, and fragment ion range of 200-1800 m/z. Heuristic protein inference was enabled, MS1 and MS2 mass accuracy was set to 10 ppm. Precursor FDR was set to 0.01, and PG q-value was set to 0.01. Quantification was performed on summed abundances of all unique peptides considering only precursors passing the q-value cutoff. PAS summarizes all NP values for a single protein into a single quantitative value. Specifically, a single protein may have been measured up to five times, once for each nanoparticle. To derive the single measurement value, PAS uses a maximum representation approach, whereby the single quantification value for a particular peptide or protein group represents the quantitation value of the NP which most frequently has measured any given proteins across all samples.

The relative abundances of protein targets were obtained by normalizing raw spectra values for each identified protein to total spectra within-subject. After normalization, undetected protein abundance values were set at zero. Protein values are presented as spectra-normalized values in all figures and results. MyoF and non-MyoF proteome data can be found in Supplementary Tables 1, 2.

Statistics and bioinformatics

Data processing and statistical analysis were performed using Microsoft Excel for Microsoft 365 (Redmond, WA, USA) and GraphPad Prism version 9.2.0 (San Diego, CA, USA). Independent samples t-tests were used for Y versus MA (pre-intervention) to determine age effects, and dependent samples t-tests were used to determine training effects in MA. All data in tables and figures are presented as mean ± standard deviation (SD) values. Training phenotypes were considered significantly different at p<0.05, although approaching values (i.e., p<0.100) were discussed as “numerical” changes due to limited n-sizes. Conversely, significant aging and training effects for protein targets were established as p<0.01 for enhanced stringency given the high number of identified proteins, although again approaching values (i.e., p<0.05) were discussed in certain circumstances due to limited n-sizes.

Bioinformatics was performed using PANTHER v17.0 [58, 59]. First, protein lists from each fraction were characterized using the functional classification tool. Next, overrepresentation tests of PANTHER GO-Slim biological processes were performed between Y and MA participants and in MA participants from pre-to-post training. Parameters for statistical overrepresentation tests included the following: i) entered proteins had to meet the aforementioned p<0.01 significance threshold, ii) protein lists were entered separately based on being up- or downregulated to generate a list of biological processes that were predicted to be directionally affected, and iii) Fisher tests with Bonferroni adjusted p<0.05 values were used as significance thresholds.