Habitual aerobic exercise and circulating proteomic patterns in healthy adults: relation to indicators of healthspan

Abstract

Habitual aerobic exercise enhances physiological function and reduces risk of morbidity and mortality throughout life, but the underlying molecular mechanisms are largely unknown. The circulating proteome reflects the intricate network of physiological processes maintaining homeostasis and may provide insight into the molecular transducers of the health benefits of physical activity. In this exploratory study, we assessed the plasma proteome (SOMAscan proteomic assay; 1,129 proteins) of healthy sedentary or aerobic exercise-trained young women and young and older men (n = 47). Using weighted correlation network analysis to identify clusters of highly co-expressed proteins, we characterized 10 distinct plasma proteomic modules (patterns). In healthy young (24 ± 1 yr) men and women, 4 modules were associated with aerobic exercise status and 1 with participant sex. In healthy young and older (64 ± 2 yr) men, 5 modules differed with age, but 2 of these were partially preserved at young adult levels in older men who exercised; among all men, 4 modules were associated with exercise status, including 3 of the 4 identified in young adults. Exercise-linked proteomic patterns were related to pathways involved in wound healing, regulation of apoptosis, glucose-insulin and cellular stress signaling, and inflammation/immune responses. Importantly, several of the exercise-related modules were associated with physiological and clinical indicators of healthspan, including diastolic blood pressure, insulin resistance, maximal aerobic capacity, and vascular endothelial function. Overall, these findings provide initial insight into circulating proteomic patterns modulated by habitual aerobic exercise in healthy young and older adults, the biological processes involved, and their relation to indicators of healthspan.

NEW & NOTEWORTHY This is the first study to assess the relation between plasma proteomic patterns and aerobic exercise status in healthy adults. Weighted correlation network analysis identified 10 distinct proteomic modules, including 5 patterns specific for exercise status. Additionally, 5 modules differed with aging in men, two of which were preserved in older exercising men. Exercise-associated modules included proteins related to inflammation, stress pathways, and immune function and correlated with clinical and physiological indicators of healthspan.

INTRODUCTION

Habitual aerobic exercise is associated with a myriad of health benefits, including improved physiological function and reduced risk of chronic disease and disability; however, the molecular mechanisms that produce these favorable effects of exercise remain uncertain (22, 46, 53). Recently, a charge has been made to identify the “molecular transducers” of the health benefits of physical activity in humans (46). It is hoped that such efforts may uncover novel therapeutic targets for the prevention and treatment of chronic clinical disorders (46, 53). Among healthy adults, regular exercise exerts perhaps its strongest influence in promoting optimal physiological aging and extending healthspan (10, 58), but insights into the molecular underpinnings of these effects in humans are only beginning to emerge (51).

Interrogating networks of molecules with the use of high-throughput analyses may provide important insight into the biological mechanisms influencing physiological function and disease risk (22). Broad-based analysis of protein signatures (proteomics) is a promising approach for identifying potential molecular transducers of the health benefits of exercise, in part because of the deterministic influence of proteins on physiological function (45). Network analysis identifies groups of associated proteins that may have physiological relevance and allows for more sensitive detection of proteomic changes compared with individual proteins while also greatly reducing multiple testing burden (3, 67). Moreover, because exercise produces a strong systemic physiological stimulus, the blood, i.e., the tissue with the greatest direct contact with all of the cells and organs of the body, may yield unique molecular footprints associated with the biological effects of physical activity (46, 53). That blood and its fluid component, plasma, are accessible with minimally invasive procedures in humans is another advantage of this approach in that any putative biomarkers identified could be assessed in larger clinical and epidemiological research settings.

Some information is available on the effects of acute exercise and shorter-term exercise training on the circulating proteome in humans (54) and regarding the influence of habitual exercise on proteins in other tissues, such as skeletal muscle (57) or urine (56). However, many of the health benefits of exercise are likely induced via long-term adaptations to chronic physical training. To our knowledge, the influence of chronic aerobic exercise on the circulating proteome has not been investigated in healthy young adults or with aging. Similarly, we lack any information as to the relation between exercise-associated effects on the circulating proteome and clinical and/or physiological indicators of healthspan.

The goal of this exploratory analysis was to gain initial insight into the association of habitual aerobic exercise with circulating proteomic signatures in healthy adults, including the impact of exercise on age-related proteomic patterns. A secondary aim was to identify relations between exercise-associated proteomic markers and indicators of healthspan. To accomplish this, we assessed the plasma proteome of groups of healthy sedentary and aerobic exercise-trained young men and women and middle-aged/older men using the SOMAscan multiplex assay, a targeted proteomics platform that measured abundance of 1,129 proteins in a highly sensitive manner. We then employed weighted correlation network analysis to identify modules (patterns) of highly correlated proteins and associated them with exercise status, sex, and age. Finally, we assessed proteomic modules associated with exercise for biological pathways of interest and related the modules to multiple indicators of healthspan.

Our analysis revealed 10 proteomic modules, 9 of which were associated with exercise status, biological sex, and/or age. The modules associated with exercise featured altered proteomic patterns related to inflammation, regulation of apoptosis, and stress and immune responses. These “exercise modules” correlated with several physiological and clinical indicators of healthspan, including diastolic blood pressure, insulin resistance, maximal aerobic capacity (V̇o_2max), and vascular endothelial function. These findings provide initial insight into circulating proteomic patterns modulated by habitual aerobic exercise in healthy young and older adults, the physiological pathways involved, and the relation between proteomic patterns and indicators of healthspan.

MATERIALS AND METHODS

Participants

Participant data and plasma samples from our laboratory’s database were used. A total of 47 healthy men and women from Boulder County, Colorado, and the surrounding area were studied. Specifically, 31 young (aged 19–32 yr) inactive (INAC, n = 16) and aerobic exercise-trained (AEX, n = 15) men and women were included to initially assess the effects of habitual aerobic exercise on the circulating proteome of healthy young adults. Sixteen additional healthy older (aged 55–77 yr) inactive (O-INAC, n = 8) and aerobic exercise-trained (O-AEX, n = 8) men were included to assess the effects of habitual aerobic exercise on age-related changes in the circulating proteome compared with young inactive (Y-INAC, n = 8) and aerobic exercise-trained (Y-AEX, n = 8) men from the initial analysis. Because of the limit of <50 total samples eligible for the SOMAscan analysis and the exploratory nature of this study, older women were not included in the aging analysis, and this important population will need to be addressed in future studies.

Inclusion criteria consisted of lack of medication use; non-smoking; body mass index <40 kg/m²; fasted plasma glucose <126 mg/dl; and absence of chronic diseases (including peripheral artery disease: ankle brachial index >0.9) as determined by medical history, physical examination, blood chemistries, and blood pressure and electrocardiogram at rest and during an incremental treadmill exercise test. Premenopausal women were not pregnant, as determined by a pregnancy test, and were not on birth control. Participants were characterized as inactive if they had performed <2 days of exercise/wk for at least the preceding 2 yr and as aerobically exercise trained if they performed vigorous aerobic exercise ≥5 days/wk for 45 min/session for at least the preceding 5 yr. Participants included were of non-Hispanic Caucasian (n = 42), non-Hispanic Asian (n = 4), or Hispanic Caucasian (n = 1) ethnicity and race. All procedures were reviewed and approved by the Institutional Review Board at the University of Colorado Boulder. The nature, risks, and benefits of all study procedures were explained to volunteers, and their written informed consent was obtained before participation.

Measurements

All measurements were made after an overnight >12-h fast from food and caffeine (water allowed) and >24-h refrainment from physical activity and alcohol. All testing on premenopausal women was performed during the early follicular phase of their menstrual cycle to reduce variations in measurements affected by circulating hormones.

Participant Characteristics

Body mass, body mass index, and waist-to-hip ratio were determined with anthropometry (38). Body percent fat was assessed by dual-energy X-ray absorptiometry (GE Lunar Prodigy Advance). Fasting plasma glucose and insulin were determined by reflective spectrophotometry (Ortho Clinical Diagnostics) and radioimmunoassay (Millipore), respectively. Fasting serum blood lipids were measured by standard assays. Plasma oxidized low-density lipoprotein was assessed by ELISA (Mercodia) and serum high-sensitivity C-reactive protein by immunoturbidimetry (Beckman Coulter). Leisure time physical activity was assessed by the Modifiable Activity Questionnaire (47).

Select Healthspan Indicators

The following well-established healthspan indicators were chosen because they reflect functional status while also predicting future risk of clinical disease and mortality (2, 30, 36, 66). Resting systolic and diastolic blood pressures were measured in triplicate over the brachial artery with a semiautomated device (Dinamap XL, Johnson & Johnson). Insulin resistance was estimated with the homeostasis model of insulin resistance (HOMA-IR) formula: [fasting plasma glucose (mg/dl) × fasting plasma insulin (μU/ml)]/405 (41). V̇o_2max was determined during an incremental treadmill exercise test (Balke protocol) with open-circuit spirometry (15). Endothelial function was measured by brachial artery flow-mediated dilation using high-resolution ultrasonography (Power Vision 6000) as described in detail previously (9, 14, 21). In this procedure, the brachial artery diameter change was measured after a 5-min forearm cuff occlusion distal to the probe. Flow-mediated dilation is reported as percent change from baseline diameter.

Proteomics

Plasma samples were sent to SomaLogic Inc. for analysis of circulating proteins with the SOMAscan v3 assay (SomaLogic Inc., Boulder, CO) as previously described (19). Briefly, the SOMAscan v3 assay is a well-established platform composed of single-stranded DNA aptamers with chemically modified side chains (SOMAmers) that have been selected to target specific proteins (1,129 available). Resulting SOMAmer concentration is proportional to the concentration of the corresponding protein in the sample, and the dynamic range of the assay spans over 8 logs (femtomolar to micromolar). The SOMAmer signal is quantified with DNA microarray and expressed as relative fluorescence units. Samples were processed under Good Laboratory Practice (60) for intra-run normalization and calibration and passed all quality control requirements.

Weighted Correlation Network Analysis and other Statistical Analysis

Statistical analysis was performed with R version 3.3.2. Outliers were identified as protein values ≥ 3 standard deviations from the mean and were removed. Protein values from all groups underwent a weighted (gene) correlation network analysis (WGCNA) using the WGCNA R software package to group similarly expressed proteins into modules, which has the additional benefit of reducing multiple testing burden (33). Briefly, the WGCNA package constructs an adjacency matrix based on pairwise correlations raised to a soft-threshold power, resulting in a network that is then analyzed for tightly interconnected groups of proteins referred to as modules (33). Modules are assigned an arbitrary color for identification (e.g., yellow module) and are characterized by a module “eigenprotein” (the first principal component of protein expressions in that module). Each participant’s individual protein levels then contribute to their module score, a weighted average of their expression levels for proteins within that module (33). A soft-threshold power of 7 was used, along with a minimum module size of 5 proteins. Module robustness was investigated by establishing that average module adjacency was greater than that expected by chance. Specifically, the average adjacency of each module was compared with the mean average adjacency of 1,000 random permutations of the data (all modules P < 0.01 correcting for multiple testing) (34).

Protein lists for each module were analyzed for associated Gene Ontology (GO) biological processes using the Cytoscape plug-in ClueGO, providing an interpretation of the general pathways implicated by the proteomic content of each module (4). Because of the exploratory nature of this study, significance for all subsequent analyses was set at an uncorrected α < 0.05. However, with consideration of multiple testing burden, we comment if results (factorial ANOVAs for each WGCNA-generated module, and correlations between modules and healthspan indicators) survive a Benjamini-Hochhberg correction to a false discovery rate of 0.05.

Raw individual protein levels and participant data can be found online in the Supplemental Data File (Supplemental Material for this article is available online at the Journal website). Outlier data points removed from analyses are bolded.

RESULTS

Ten modules were generated with the following corresponding number of proteins assigned to each module: black 19, blue 37, brown 36, green 24, magenta 16, pink 16, purple 14, red 21, turquoise 579, and yellow 36. Remaining uncorrelated proteins were not used for further analysis.

Habitual Aerobic Exercise and Circulating Proteomic Patterns in Young Adults

Participant characteristics and healthspan indicators.

All young participant characteristics were within normal healthy ranges (Table 1). As expected, young AEX adults had lower body fat indices (body mass, body mass index, body fat percent) and circulating triglycerides compared with the young INAC individuals (all P < 0.05). Additionally, young AEX subjects had higher activity levels (leisure time physical activity) and aerobic exercise capacity (V̇o_2max; both P < 0.0001) compared with their inactive counterparts.

Table 1.

Young participant characteristics and healthspan indicators

	INAC	AEX
Participant characteristics
n, men/women	8/8	8/7
Age, yrL	23 ± 1	24 ± 1
Body mass, kg	81 ± 4	67 ± 3*
Body mass index, kg/m2 L	27 ± 1	22 ± 1*
Body fat, %	33 ± 2	17 ± 2*
Waist-to-hip ratio, U	0.81 ± 0.02	0.77 ± 0.01
Fasting glucose, mg/dl	88 ± 2	88 ± 2
Fasting insulin, µU/mlL	10 ± 2	6 ± 1
Total cholesterol, mg/dl	166 ± 8	160 ± 8
High-density lipoprotein, mg/dl	49 ± 3	56 ± 2
Low-density lipoprotein, mg/dl	94 ± 7	89 ± 7
Triglycerides, mg/dl	110 ± 9	73 ± 5*
C-reactive protein, mg/lL	0.6 ± 0.1	0.6 ± 0.2
Oxidized low density lipoprotein, U/l	36 ± 3	35 ± 4
LTPA, MET-hr/wkL	29 ± 6	136 ± 17*
Select indicators of healthspan
Systolic blood pressure, mmHg	110 ± 3	116 ± 3
Diastolic blood pressure, mmHg	64 ± 2	59 ± 2
HOMA-IR, UL	2.1 ± 0.4	1.4 ± 0.2
V̇o2max, ml·kg−1·min−1	36 ± 2	54 ± 2*
Endothelial function, FMD %Δ	8.1 ± 0.6	7.5 ± 0.6

Data are means ± SE. Δ, change; AEX, aerobic exercise trained; FMD, flow-mediated dilation; HOMA-IR, homeostasis model assessment-insulin resistance; INAC, inactive; LTPA, leisure time physical activity; MET, metabolic equivalent; V̇o_2max, maximal aerobic capacity.

^LData log transformed for statistical analysis;

^*P < 0.05 vs. INAC.

Influence of aerobic exercise training status on circulating proteomic patterns.

We sought to determine if aerobic exercise status influenced circulating proteomic patterns (modules) in young adults. Sex has been shown to influence the plasma proteome (29, 59), and INAC versus AEX groups showed differences in body fat percent. Taking this into consideration, the effect of habitual aerobic exercise status was assessed with a factorial ANOVA (sex: men vs. women; exercise status: inactive vs. exercise trained, controlling for body fat percent as a covariate). Of the 10 modules generated by WGCNA, 4 modules were different between the young INAC and AEX men and women (yellow, turquoise, green, and purple modules; P = 0.008, P = 0.004, P = 0.03, and P = 0.04, respectively; yellow and turquoise strictly significant; Fig. 1). Although not the primary focus of this study, one module was different between men and women (blue module; P = 0.007), and the yellow module had an exercise-sex interaction (P = 0.047; Fig. 2). Young AEX men had a higher yellow module score compared with all other groups (all P < 0.05) suggesting that exercise status has a different effect on the yellow module’s circulating proteome pattern in men compared with women.

Fig. 1.

Modules different with exercise status in young inactive (INAC) and aerobic exercise-trained (AEX) men and women. Data are means ± SE; *P < 0.05.

Fig. 2.

Modules associated with sex (A) or exercise-sex interaction (B) in young inactive (INAC) and aerobic exercise-trained (AEX) men (M) and women (W). Data are means ± SE; AEX*S, aerobic exercise-trained status by sex; *P < 0.05.

Habitual Aerobic Exercise and Circulating Proteomic Patterns in Young and Older Men

Participant characteristics and healthspan indicators.

All male participant characteristics were within normal ranges and are reported in Table 2. As expected, the exercising men had lower body mass indices and higher physical activity and aerobic exercise capacity compared with their respective age-matched inactive peers (P < 0.05). The O-INAC group had higher (P < 0.05) LDL-cholesterol or tended to have higher total cholesterol than the other 3 groups.

Table 2.

Male participant characteristics and healthspan indicators: exercise-aging analysis

	Y-INAC	Y-AEX	O-INAC	O-AEX
Participant characteristics
n	8	8	8	8
Age, yrL	23 ± 1	24 ± 2	66 ± 2†#	63 ± 2†#
Body mass, kg	88 ± 6	75 ± 2	84 ± 6	70 ± 2†
Body mass index, kg/m2L	27 ± 2	23 ± 1	26 ± 1	23 ± 1
Body fat, %	26 ± 2	12 ± 2†‡	27 ± 4#¶	16 ± 1
Waist-to-hip ratio, U	0.86 ± 0.03	0.81 ± 0.01	0.94 ± 0.03#	0.86 ± 0.01
Fasting glucose, mg/dl	88 ± 3	90 ± 2	87 ± 3	93 ± 5
Fasting insulin, µU/mlL	11 ± 2	6 ± 1	6 ± 1	4 ± 0.5†
Total cholesterol, mg/dl	160 ± 10	159 ± 15	202 ± 8†#	173 ± 8
High-density lipoprotein, mg/dl	44 ± 3	53 ± 3	50 ± 3	63 ± 5†
Low-density lipoprotein, mg/dl	92 ± 9	92 ± 12	135 ± 6*	94 ± 5
Triglycerides, mg/dl	118 ± 15	69 ± 8†	82 ± 12	83 ± 7
C-reactive protein, mg/lL	0.3 ± 0.05	0.8 ± 0.3	1.6 ± 0.9	0.5 ± 0.1
Oxidized low density lipoprotein, U/l	36 ± 5	33 ± 6	54 ± 4	47 ± 6
LTPA, MET-hr/wkL	31 ± 9	133 ± 24†‡	28 ± 13	78 ± 12†‡
Select indicators of healthspan
Systolic blood pressure, mmHg	115 ± 4	123 ± 4	127 ± 3	118 ± 4
Diastolic blood pressure, mmHg	66 ± 3	60 ± 3‡¶	76 ± 2	73 ± 3
HOMA-IR, UL	2.5 ± 0.6	1.3 ± 0.2	1.3 ± 0.2	0.9 ± 0.1†
V̇o2max, ml·kg−1·min−1	41 ± 2*	59 ± 1*	27 ± 1*	50 ± 1*
Endothelial function, FMD %Δ	7.7 ± 0.9	8.0 ± 0.8	4.3 ± 0.7	7.0 ± 1.4

Data are means ± SE. Δ, change; FMD, flow-mediated dilation; HOMA-IR, homeostasis model assessment-insulin resistance; LTPA, leisure time physical activity; O-AEX, older aerobic exercise trained; O-INAC, older inactive; V̇o_2max, maximal aerobic capacity; Y-AEX, young aerobic exercise trained; Y-INAC, young inactive.

^LData log transformed for statistical analysis;

^†P < 0.05 vs. Y-INAC;

^#P < 0.05 vs. Y-AEX;

^‡P < 0.05 vs. O-INAC;

^¶P < 0.05 vs. O-AEX;

^*P < 0.05 vs. all groups.

Influence of aging and aerobic exercise status on circulating proteomic patterns.

To determine if aging and aerobic exercise training status influenced circulating proteomic patterns, a factorial ANOVA was performed in men (age: young vs. older; exercise: inactive vs. trained, controlling for body fat percent as a covariate). Of the 10 modules generated by WGCNA, 5 modules were different between young and older men (yellow, black, brown, magenta, and red modules; P = 0.001, P = 0.001, P = 0.03, P = 0.03, and P = 0.04, respectively; yellow and black strictly significant; Fig. 3). Four modules were different with training status between the inactive and trained men (purple, yellow, black, and turquoise modules; P = 0.00008, P = 0.003, P = 0.01, and P = 0.03, respectively; purple, yellow, and black strictly significant; Fig. 4). Of note, the purple, turquoise, and yellow modules were also identified as influenced by exercise status in the cohort of young men and women.

Fig. 3.

Modules influenced by age in young and older men. Data are means ± SE; *P < 0.05: young vs. older men. O-AEX, older aerobic exercise-trained men; O-INAC, older inactive men; Y-AEX, young aerobic exercise-trained men; Y-INAC, young inactive men.

Fig. 4.

Modules different with exercise status in men. Data are means ± SE; *P < 0.05: inactive vs. exercise-trained men. O-AEX, older aerobic exercise-trained men; O-INAC, older inactive men; Y-AEX, young aerobic exercise-trained men; Y-INAC, young inactive men.

To determine if habitual aerobic exercise influences circulating protein patterns that are associated with aging, we sought to identify modules in O-AEX men that were preserved at or toward levels of young men. A multiple linear regression was used to identify modules different with age (O-INAC vs. Y-INAC, P < 0.05) but not different in O-AEX men compared with Y-INAC men (i.e., young control group). This analysis yielded 2 age-associated modules (red and magenta) that were significantly different between the Y-INAC versus O-INAC men (both P < 0.05) but not between the Y-INA versus O-AEX men (both P = 0.2) (Fig. 5).

Fig. 5.

Modules associated with age (Y-INAC vs. O-INAC) and partially preserved in older exercise-trained (O-AEX) men. Data are means ± SE; *P < 0.05. ns, not significant; O-INAC, older inactive men; Y-INAC, young inactive men.

Biological Processes Indicated in Protein Patterns

Proteins within each module were analyzed for associated GO biological processes terms using ClueGO to provide insight into the general pathways represented (1, 4, 18). Additionally, individual proteins were ranked according to their intramodular connectivity score (kME) within their respective module. Proteins with a kME closer to 1 indicate higher contribution to the module compared with proteins with a kME closer to 0. Individual proteins with a kME >0.8 up to the first 10 proteins for each module are reported, along with group means of interest, in each respective table as indicated below. The Universal Protein Resource (UniProt), a consortium for protein sequence and functional information, was used to report known individual functions of these highest contributing proteins.

Habitual aerobic exercise influenced modules.

Processes in modules influenced by exercise training status (yellow, turquoise, purple, black, and green modules) are reported in Fig. 6 and include pathways primarily involved in apoptosis, immune system function, response to stress, inflammation, and phosphate metabolism, respectively. The highest ranked individual proteins within these modules are reported in Table 3. The highest contributing proteins in each exercise-related module were α-2-macroglobulin (yellow module), Ras-related C3 botulinum toxin substrate 1 (Rac1; turquoise module), fibronectin (purple module), apolipoprotein E (black module), and dual specificity mitogen-activated protein kinase 3 (green module). Although not part of the primary focus of this study, GO analysis was performed for the blue module (different with sex status), which primarily represented processes related to the inflammatory response (Fig. 7); the top ranked proteins for the blue module can be found in Table 4.

Fig. 6.

GO biological processes implicated by proteins in modules associated with habitual aerobic exercise status. GO, Gene Ontology.

Table 3.

Modules different with aerobic exercise

Module	Protein	UniProt ID	Primary Function	Protein Level, RFU
				INAC	AEX
Yellow	Alpha-2-macroglobulin	P01023	Inhibits proteinases	20,334 ± 4,513	36,067 ± 4,180
	Complement C3b, inactivated	P01024	Activation of immune response; triglyceride synthesis and glucose transport	10,599 ± 2,403	19,385 ± 1,928
	Apolipoprotein B	P04114	Protein constituent of chylomicrons, LDL, and VLDL	45,566 ± 11,332	70,167 ± 9,376
	Complement 3a	P01024	Activation of immune response; triglyceride synthesis and glucose transport	1,562 ± 279	2,112 ± 163
	Complement C4b	P0C0L4	Activation of immune response	1,071 ± 274	1,754 ± 240
	Proprotein convertase subtilisin/kexin type 7	Q16549	Peptidase activity	1,223 ± 123	937 ± 119
	Sonic hedgehog protein	Q15465	Development; biosynthesis, growth, and differentiation	543 ± 58	329 ± 41
	Insulin-like growth factor-binding protein 4	P22692	Cell proliferation and metabolism	6,807 ± 374	6,027 ± 327
	Complement 3	P01024	Activation of immune response; triglyceride synthesis and glucose transport	86,286 ± 5,895	100,039 ± 6,183
	Protein 4.1	P11171	Regulating membrane stability	4,679 ± 790	5,376 ± 460
Turquoise	Ras-related C3 botulinum toxin substrate 1	P63000	GTPase	12,974 ± 2,321	7,162 ± 1,529
	Ribosome maturation protein SBDS	Q9Y3A5	Ribosome assembly; cellular stress resistance	7,723 ± 1,691	3,286 ± 1,133
	Tyrosine-protein kinase CSK	P41240	Cell growth, differentiation, migration, and immune response	7,214 ± 1,686	2,827 ± 1,006
	Methionine aminopeptidase 2	P50579	Protein synthesis	19,556 ± 4,087	7,160 ± 2,936
	Triosephosphate isomerase	P60174	Oxidoreductase activity	8,776 ± 1,236	5,558 ± 833
	Translationally-controlled tumor protein	P13693	Calcium binding and microtubule stabilization	9,238 ± 1,708	5,150 ± 1,120
	Glycogen synthase kinase-3 alpha/beta	P49840	Negative regulator in glucose homeostasis	10,600 ± 2,186	4,759 ± 1,409
	Elongation factor 1-beta	P24534	RNA binding	1,367 ± 195	870 ± 131
	Cytokine receptor common subunit gamma	P31785	Interleukin receptor	2,749 ± 399	1,760 ± 282
	Aflatoxin B1 aldehyde reductase member 2	O43488	Electron carrier activity	2,457 ± 561	878 ± 335
Purple	Fibronectin	P02751	Inhibits tumor growth, angiogenesis, and metastasis	14,863 ± 1,556	8,084 ± 1,082
	D-dimer	P02671	Wound repair and blood coagulation	8,901 ± 415	7,464 ± 423
	Fibronectin 1.3	P02751	Inhibits tumor growth, angiogenesis, and metastasis	1,084 ± 95	670 ± 69
	Fibrinogen gamma chain	P02679	Blood coagulation	42,871 ± 1,903	35,581 ± 1,727
	Fibrinogen	P02671	Blood coagulation	107,145 ± 3,654	95,775 ± 2,966
	Fibronectin 1.4	P02751	Inhibits tumor growth, angiogenesis, and metastasis	40,453 ± 2,990	29,988 ± 1,850
Black	Apolipoprotein E (isoform E4)	P02649	Binding, internalization, and catabolism of lipoprotein particles	44,233 ± 3,775	34,336 ± 3,351
	Apolipoprotein E (isoform E3)			58,933 ± 5,220	49,640 ± 4,388
	Apolipoprotein E			4,576 ± 502	3,628 ± 429
	Apolipoprotein E (isoform E2)			149,140 ± 10,250	130,004 ± 7,750
	Complement C4	P0C0L4	Activation of immune response	80,951 ± 7,475	72,075 ± 5,994
Green	Dual specificity mitogen-activated protein kinase 3	P46734	Activates immune response and metabolic pathways in response to cytokines and environmental stress	2,348 ± 86	2,623 ± 95
	Serine/threonine-protein kinase MRCK beta	Q9Y5S2	Regulates cytoskeletal reorganization	1,299 ± 46	1,410 ± 48
	Interleukin-5	P05113	Immune, differentiation, and signaling activity	2,889 ± 122	3,359 ± 142
	Breast cancer anti-estrogen resistance protein 3	O75815	Regulates metabolic activity, signaling, development	1,083 ± 29	1,214 ± 37
	UMP-CMP kinase	P30085	Pyrimidine nucleotide biosynthesis	4,040 ± 162	4,473 ± 183
	Macrophage scavenger receptor types I and II	P21757	Macromolecule endocytosis; pro-atherosclerotic	922 ± 34	1,016 ± 43
	Serine protease 27	Q9BQR3	Peptidase activity	6,689 ± 371	8,943 ± 588
	Cadherin-15	P55291	Cell adhesion and differentiation	1,933 ± 60	2,013 ± 59

Data are means ± SE. AEX, aerobic exercise trained; INAC, inactive; RFU, relative fluorescence units.

Fig. 7.

GO biological processes implicated by proteins in modules associated with sex (blue) or age (brown). GO, Gene Ontology.

Table 4.

Modules associated with only sex or aging

				Protein Level, RFU
Module	Protein	UniProt ID	Primary Function	Young men	Young women
Module different with sex
Blue	Calcineurin subunit B type 1	P63098	Regulates calcium sensitivity	2,615 ± 192	3,548 ± 205
	cGMP-dependent 3′,5′-cyclic phosphodiesterase	O00408	Regulates mitochondrial cAMP levels and respiration	3,995 ± 258	5,087 ± 277
	Kallikren-14	Q9P0G3	Peptidase activity in reproductive process	3,859 ± 251	5,267 ± 285
	Cadherin-6	P55285	Cell adhesion proteins	492 ± 32	603 ± 34
	Opioid-binding protein/cell adhesion molecule	Q14982	Binds opioids	1,714 ± 88	2,064 ± 86
	Ephrin type-B receptor 4	P54760	Heart morphogenesis and angiogenesis	3,926 ± 230	4,709 ± 246
	Apolipoprotein D	P05090	Lipid transporter	2,904 ± 169	3,676 ± 144
	Dickkopf-like protein 1	Q9UK85	Signal transducer activity	4,165 ± 355	5,568 ± 451
	Basigin	P35613	Mannose binding; spermatogenesis, and development	3,648 ± 235	4,361 ± 247
	Retinol-binding protein 4	P02753	Retinol delivery	531 ± 27	563 ± 27
Module different only with age				Young men	Older men
Brown	Thrombin	P00734	Blood homeostasis, inflammation and wound healing	15,883 ± 4,280	3,979 ± 1,911
	Inter-α-trypsin inhibitor heavy chain H4	Q14624	Inflammatory responses	11,262 ± 1,527	16,103 ± 1,023
	Tyrosine-protein kinase JAK2	O60674	Cell growth/differentiation and innate/adaptive immunity	5,709 ± 1,067	9,589 ± 744
	Plasma serine protease inhibitor	P05154	Proteolytic activity; blood coagulation	35,557 ± 7,842	51,063 ± 4,617
	Bone morphogenetic protein 7	P18075	Calcium regulation and bone formation	1,181 ± 181	763 ± 96
	Aspartate aminotransferase, cytoplasmic	P17174	Amino acid biosynthesis and glutamate regulator	9,363 ± 1,340	6,723 ± 949
	Interleukin-2	P60568	Activation of immune response	2,264 ± 339	1,911 ± 209
	Growth/differentiation factor 2	Q9UK05	Angiogenesis inhibition	1,502 ± 192	1,050 ± 129
	NAD-dependent protein deacetylase sirtuin-2	Q8IXJ6	Intracellular protein regulator	12,610 ± 1,972	7,488 ± 1,427

Data are means ± SE. RFU, relative fluorescence units.

Aging modules and age-related modules partially preserved with exercise.

We identified five modules different with age in men (yellow, black, brown, magenta, and red modules), with 4 of these modules also influenced by exercise status (discussed above). The lone exception was the brown module, which included proteins associated with the regulation of covalent protein modification, as shown in Fig. 7; the highest ranked individual proteins in this module are listed in Table 4.

Additionally, the 2 age-related modules found in the inactive men that were at least partially preserved in older exercise-trained men, i.e., red and magenta, included proteins associated with the inflammatory response and response to cell stress, respectively, as shown in Fig. 8; the highest-ranking individual proteins are listed in Table 5 with the highest contributing protein for the red module being serine/threonine-protein kinase PAK 3 and for the magenta module leukocyte antigen CD97 (CD97).

Fig. 8.

GO biological processes implicated by proteins in modules partially preserved in habitually trained older men to levels of young men. GO, Gene Ontology.

Table 5.

Age-related modules partially preserved with aerobic exercise in men

Module	Protein	UniProt ID	Primary Function	Protein Level, RFU
				Y-INAC	O-INAC	O-AEX
Red	Serine/threonine-protein kinase PAK 3	O75914	Cytoskeleton regulation, cell migration, or cell cycle regulation	758 ± 89	893 ± 40	835 ± 55
	Interleukin-37	Q9NZH6	Suppressor of inflammation and immune responses	739 ± 76	906 ± 31	877 ± 46
	Cyclin-dependent kinase 5	Q00535	Anatomical structure formation; cell development, differentiation, migration, and morphogenesis	973 ± 67	1,058 ± 37	1,023 ± 40
	Tumor necrosis factor receptor superfamily member 3	P36941	Promotes apoptosis	3,051 ± 223	3,619 ± 138	3,542 ± 134
	Kallikrein-12	Q9UKR0	Peptidase activity	3,128 ± 442	3,826 ± 215	3,687 ± 286
	Glutamate carboxypeptidase 2	Q04609	Peptidase activity; metabolic processes	1,326 ± 115	1,491 ± 63	1,421 ± 71
	Tumor necrosis factor receptor superfamily member 11B	O00300	Bone homeostasis; promotes apoptosis	5,199 ± 651	7,535 ± 238	6,920 ± 635
	Lymphotoxin α-2 β-1	P01374	Adaptive immunity; cell signaling, development, and differentiation	1,168 ± 111	1,384 ± 54	1,304 ± 89
	Arginase-1	P05089	Aging; metabolic processes	436 ± 19	498 ± 14	471 ± 10
	Complement C1q subcomponent	P02745	Activation of immune response	24,624 ± 3,660	29,642 ± 1,879	26,825 ± 2,426
Magenta	Leukocyte antigen CD97	P48960	Leukocyte migration	1,514 ± 67	1,569 ± 81	1,524 ± 78
	Oncostatin-M	P13725	Regulates tumor growth, cytokine production, and cell maturation	1,401 ± 74	1,576 ± 80	1,470 ± 63
	Ras GTPase-activating protein 1	P20936	Inhibitory regulator of Ras-cyclic AMP pathway	882 ± 23	999 ± 48	933 ± 28
	Carbonic anhydrase 4	P22748	Carbon dioxide hydration	1,272 ± 55	1,360 ± 50	1,377 ± 72
	C-C motif chemokine 1	P22362	Immune response	2,391 ± 92	2,766 ± 115	2,730 ± 143
	Tropomyosin beta chain	P07951	Muscle contraction	1,565 ± 94	1,834 ± 91	1,647 ± 106
	Carbonic anhydrase-related protein 10	Q9NS85	Development	548 ± 27	639 ± 26	570 ± 27

Data are means ± SE. O-AEX, older aerobic exercise trained; O-INAC, older inactive; RFU, relative fluorescence units; Y-INAC, young inactive.

Association of Proteomic Patterns with Select Healthspan Indicators

To determine if modules were clinically relevant, the 5 modules related to aerobic exercise status in adults and two modules preserved with age in older trained men were correlated with the physiological and clinical indicators of healthspan in the overall group of participants. We found that 5 of the 7 modules related to exercise status (yellow, black, purple, magenta and turquoise) were related to indicators of healthspan, including diastolic blood pressure, insulin resistance, aerobic exercise capacity and vascular endothelial function. The significance and magnitude of each nominally significant correlation (Pearson’s r) is reported in Fig. 9. The strongest correlation observed was between the purple module and aerobic exercise capacity, which also survives strict control of multiple testing burden (r = −0.48 P = 0.0006, Fig. 9C).

Fig. 9.

Modules influenced by exercise or age-related modules preserved with exercise associated with select indicators of healthspan [diastolic blood pressure (A), HOMA-IR (B), V̇o_2max (C), endothelial function (D)] in all participants. FMD, flow-mediated dilation; HOMA-IR, homeostasis model assessment-insulin resistance; V̇o_2max, maximal aerobic capacity.

DISCUSSION

To our knowledge this is the first study to assess circulating plasma proteomic patterns with habitual aerobic exercise and their association to indicators of healthspan in healthy adults. Through WGCNA, we found 10 distinct co-expressed plasma protein modules. Of these 10 modules, a total of 7 were influenced by habitual exercise status (i.e., different in inactive/active adults or preserved in older trained men), including 3 modules that were influenced by aerobic exercise status in both young adults and young and older men. These findings demonstrate that habitual aerobic exercise has a prominent effect on the circulating proteome, regardless of sex or age. Importantly, 5 of the exercise-influenced modules were associated with at least one clinical indicator of healthspan (diastolic blood pressure, insulin resistance, V̇o_2max, or vascular endothelial function), suggesting that the protein patterns captured within these modules may have clinical relevance. Finally, exercise-related modules were associated with biological processes related to the stress response, inflammatory signaling, the immune system, and apoptosis, suggesting a potential role for these and other pathways described in transducing the beneficial impact of regular aerobic exercise on healthspan.

Proteomic Patterns and Markers Associated with Habitual Aerobic Exercise and Healthspan Indicators

Prominent plasma proteins within our exercise-related modules may represent key modulators of physiological function and could be considered for future study (3, 67). Our analysis identified 3 modules (yellow, purple, turquoise) that were associated with habitual aerobic exercise status in both young individuals and the young and older men cohort.

The yellow proteomic module was correlated with diastolic blood pressure and aerobic capacity. The highest-ranking protein within this module was α-2-macroglobulin, an anti-protease that inhibits matrix metalloproteinases (MMPs) (49), the latter being a key regulator of extracellular matrix remodeling (25, 65) and modulating influence on blood pressure (17, 63). α-2-macroglobulin also is involved in improvements in aerobic capacity with exercise training (62). Other key proteins within the yellow module are associated with pathways of apoptosis, complement system, lipid metabolism, and vascular remodeling, suggesting the importance of inflammation and immune function in the physiological benefits of habitual aerobic exercise.

The turquoise module was associated with insulin resistance and aerobic capacity. It is well established that chronic aerobic exercise is associated with clinically significant improvements in insulin resistance (35). Of note, Rac1, triosephosphate isomerase (TPI), and glycogen synthase kinase-3 α/β (GSK-3) are among the highest-ranking proteins identified in the turquoise module and have previously been shown to be involved in glucose transport, glycolysis, and increased insulin resistance, respectively (24, 61, 64).

The highest-ranking proteins of the purple module consisted of fibronectin and other proteins important for wound healing. These proteins have been shown to facilitate atherogenic and thrombogenic processes in the vascular endothelium and smooth muscle of the arterial wall (42, 44). Moreover, fibrinogen is an independent risk factor for cardiovascular disease (26, 37) and is lowered with chronic exercise (12). In the context of the beneficial effects of regular aerobic exercise, these observations are consistent with our finding of a relation between the proteomic pattern of the purple module and healthspan indicators, including vascular endothelial function, diastolic blood pressure, and V̇o_2max.

Among the proteins identified in the exercise-related modules were several previously indicated in the literature as being altered with exercise, including brain-derived neurotrophic factor (BDNF), interleukin-6 (IL-6), secreted protein acidic and rich in cysteine (SPARC), decorin, heat shock proteins 60 (Hsp60) and 70 (Hsp70), and superoxide dismutase (SOD) from the turquoise module and interleukin-8 (IL-8) from the purple module (7, 28, 54). These proteins represent biological pathways involved in neurogenesis, glucose metabolism, muscle hypertrophy, mitochondrial protein transport, protection from thermal or oxidative stress, and angiogenesis (5, 7, 16, 22, 28, 31, 40, 55). All of these processes are known to participate in mediating the broad, systemic health benefits of regular exercise.

We would expect to distinguish proteomic factors relating to cardiorespiratory fitness between trained and untrained individuals, and as discussed above, the yellow, turquoise, and purple modules were associated with V̇o_2max. However, the proteins in these modules primarily included proteins related to apoptosis, immune, and stress pathways rather than exhibiting markers clearly linked to V̇o_2max or its direct determinants, such as cardiac output or blood volume. Still, some proteins in these exercise- or V̇o_2max-correlated modules have been shown to be related to oxygen transport capacity or cardiac contractility and hypertrophy, including hemoglobin, cAMP-specific 3′, 5′-cyclic phosphodiesterase 4D, and GSK-3 (8, 13, 27, 32), although only GSK-3 is among the top 10 ranking proteins within its respective module. The limitations in identifying proteins directly related to V̇o_2max may be due to this particular SomaLogic assay not being designed to target those pathways, but it is also plausible that proteins directly related to determinants of cardiorespiratory fitness may not be as different in habitually trained individuals assessed at rest. Such signals might only be evident in response to acute bouts of exercise or with dynamic changes in fitness over time. A key goal of future research should be to identify novel biomarkers in these physiological states.

Age-Related Proteomic Patterns and Markers Partially Preserved with Exercise

In the present analysis, we identified 2 age-related modules (red and magenta) that were preserved in older exercise-trained men, suggesting that exercise may partially preserve age-related proteomic patterns. Chronic aerobic exercise promotes optimal physiological aging and extends healthspan (6, 20, 53), suggesting that preserved proteomic patterns in older trained adults may help elucidate molecular transducers of these benefits of exercise. The red and magenta modules included proteins associated with inflammatory and cell stress responses, respectively. Although not a high-ranking protein in the red module, chordin-like protein 1, an antagonist of bone morphogenic protein 4 (involved in bone formation), has previously been shown to differ with age using the same SOMAscan proteomics platform (43). Here we found that chordin-like protein 1 was part of the red module, which differed with age but also was preserved in older exercise-trained men.

The magenta module was associated with insulin sensitivity and vascular endothelial function. The 2 highest-ranking proteins in this module are CD97 and oncostatin M. CD97 is a membrane protein on inflammatory cells that plays an important role in intercellular signaling in the immune system through regulation of cell adhesion and migration (52). CD97 is upregulated in individuals with metabolic syndrome and is related to insulin resistance (48). Oncostatin M is a pro-inflammatory cytokine in the IL-6 family and is believed to be pro-atherosclerotic (50) and involved in mediating insulin resistance (11, 23, 39). These findings provide preliminary identification of proteins related to inflammation and stress response that are partially preserved in older aerobic-trained adults, but future studies are needed to further investigate their roles.

Limitations and Future Directions

Because of the combination of rigorous screening required to exclude clinical disease and study only healthy older adults, the deep physiological phenotyping of our participants that required broad and extensive technical demands, and the high per-sample cost of the SOMAscan proteomic analyses, the size of our subject sample was limited. However, the present analyses and results were intended to represent an initial, hypothesis-generating effort to establish preliminary insight as to the feasibility of the model and to identify putative proteomic targets for a future larger-scale study. We believe that we accomplished these goals in the present investigation. We also recognize that the SOMAscan analysis employed here does not survey the complete circulating proteome and that more research is needed to discover additional circulating proteins and the means to assess them.

It should be noted that the present study is but one component of an assessment of the spectrum of adaptations to aerobic exercise, where in the present study we assessed individuals at a relative physiological steady state (at rest) who have been habitually exercising for years. This timepoint we assessed should be taken into context for the circulating proteome markers identified, as they do not reflect changes in the proteome during or immediately after an acute bout of exercise or dynamic changes in fitness going from the untrained to the trained state. Future studies are needed to complete the overall picture of the effects of exercise, including assessment of dynamic changes in the proteome with acute exercise (the hours before, during, and after exercise), exercise interventions (weeks to months training modulating fitness), and how these changes in the proteome are tied to improvements in physiological and clinical measures (e.g., V̇o_2max). In addition, other tissues (e.g., muscle) and omic pathways (e.g., genomics, epigenetics, metabolomics, etc.) should be included to accurately and holistically capture the local and systemic effects of exercise, as blood and proteomics only provides one small piece of a larger picture and may or may not reflect what is happening in tissue.

Finally, although we identified a proteomic pattern influenced by a sex and exercise status interaction in young adults, additional studies will be needed to further investigate potential sex differences in the circulating proteome with habitual exercise and, in particular, differences with age.

Summary and Conclusions

In summary, employing a SOMAscan assay, we have conducted an initial analysis of >1,000 plasma proteomic markers associated with aerobic exercise status in young men and women, as well as in healthy older men. Using weighted correlation network analysis to identify clusters of highly co-expressed proteins, we then identified 10 distinct plasma proteomic modules (patterns) in our participants, identified the individual protein markers most strongly represented in those patterns, and linked the proteomic composition of the modules to their respective biological pathways using the GO reference database. Finally, we correlated proteomic patterns with numerous clinical and physiological indicators of human healthspan.

We were able to characterize 5 specific proteomic patterns associated with aerobic exercise status in adults, as well as 2 modules that were preserved with aging in regularly exercising men. Habitual exercise-associated proteomic patterns were related to biological pathways involved in wound healing, regulation of apoptosis, glucose-insulin and cellular stress signaling, and inflammation/immune responses. Several of the exercise-related proteomic patterns were associated with physiological and clinical indicators of healthspan, including diastolic blood pressure, insulin resistance, V̇o_2max, and vascular endothelial function. Overall, these findings provide initial insight into circulating proteomic patterns modulated by habitual aerobic exercise in healthy young and older adults, the biological processes involved, and the relation between proteomic patterns and clinical and physiological indicators of human healthspan.

GRANTS

This work was supported by NIH awards R37 AG013038, Colorado Clinical and Translational Science Award UL1 TR001082, and T32 AG000279-14S1.

DISCLAIMERS

Contents are the authors’ sole responsibility and do not necessarily represent official NIH views.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.R.S.-P. and D.R.S. conceived and designed research; J.R.S.-P. performed experiments; J.R.S.-P. and K.S.S.-P. analyzed data; J.R.S.-P., K.S.S.-P., M.B.M., C.R.M., and D.R.S. interpreted results of experiments; J.R.S.-P. and K.S.S.-P. prepared figures; J.R.S.-P. and K.S.S.-P. drafted manuscript; J.R.S.-P., K.S.S.-P., M.B.M., C.R.M., and D.R.S. edited and revised manuscript; J.R.S.-P., K.S.S.-P., M.B.M., C.R.M., and D.R.S. approved final version of manuscript.