Distance running alters peripheral microRNAs implicated in metabolism, fluid balance, and myosin regulation in a sex-specific manner

Abstract

Microribonucleic acids (miRNAs) mediate adaptive responses to exercise and may serve as biomarkers of exercise intensity/capacity. Expression of miRNAs is altered in skeletal muscle, plasma, and saliva following exertion. Women display unique physiologic responses to endurance exercise, and miRNAs respond to pathologic states in sex-specific patterns. However sex-specific miRNA responses to exercise remain unexplored. This study utilized high-throughput RNA sequencing to measure changes in salivary RNA expression among 25 collegiate runners following a single long-distance run. RNA concentrations in pre- and post-run saliva was assessed through alignment and quantification of 4,694 miRNAs and 27,687 mRNAs. Pair-wise Wilcoxon rank-sum test identified miRNAs with significant [false discovery rate (FDR) < 0.05] post-run changes. Associations between miRNA levels and predicted mRNA targets were explored with Pearson correlations. Differences in miRNA patterns between men (n = 13) and women (n = 12) were investigated with two-way analysis of variance. Results revealed 122 salivary miRNAs with post-run changes. The eight miRNAs with the largest changes were miR-3671, miR-5095 (downregulated); and miR-7154-3p, miR-200b-5p, miR-5582-3p, miR-6859-3p, miR-6751-5p, miR-4419a (upregulated). Predicted mRNA targets for these miRNAs represented 15 physiologic processes, including glycerophospholipid metabolism (FDR = 0.042), aldosterone-regulated sodium reabsorption (FDR = 0.049), and arrhythmogenic ventricular cardiomyopathy (FDR = 0.018). Twenty-six miRNA/mRNA pairs had associated changes in post-run levels. Three miRNAs (miR-4675, miR-6745, miR-6746-3p) demonstrated sex-specific responses to exercise. Numerous salivary miRNAs change in response to endurance running and target the expression of genes involved in metabolism, fluid balance, and musculoskeletal adaptations. A subset of miRNAs may differentiate the metabolic response to exercise in men and women.

INTRODUCTION

Microribonucleic acids (miRNAs, also miR) are short, noncoding molecules that regulate the translation of coding messenger RNA (mRNA) into proteins (10). These master regulators are found in cells throughout the human body and control numerous physiologic processes, including cellular metabolism (34), repair mechanisms (19), and signaling. In fact, packaging of miRNAs within extracellular vesicles, or other protein-mediated carriers, allows for transportation through extracellular fluids to distant tissues, where they can dock, cross the cell membrane, and influence targeted protein pathways (8, 9). These characteristics have garnered miRNAs much attention as potential biomarkers in human health and disease (27).

In healthy human adults, miRNAs appear to mediate physiologic adaptations to endurance exercise (7, 9). Several muscle-related miRNAs have been described in adult human tissue, and these miRNAs demonstrate dynamic expression patterns following aerobic exercise (17, 32). Interestingly, muscle-related miRNAs can also be measured in circulating plasma following exercise. For example, a study of plasma-based miRNAs in 10 male rowers identified a subset of miRNAs that underwent dynamic regulation following sustained aerobic activity, and these miRNAs targeted mRNAs involved in skeletal and cardiac muscle contractility (2).

Circulating levels of individual miRNAs may reflect exercise intensity and capacity, correlating with measures such as maximum oxygen uptake (29). Other studies have noted that the direction and magnitude of miRNA changes may depend on the selected exercise activity or the body tissue in question (4, 39). This is illustrated by miR-133a, a muscle-related miRNA preferentially influenced by eccentric exercise regimens (involving increased work during the negative or gravity-driven phase). Though the majority of exercise studies have focused on blood-based miRNA levels and muscle-related targets, a study by Konstantinidou and colleagues (2016) (22) recently examined levels of eight salivary miRNAs and identified two that were altered after stationary bike exercise.

Saliva is one of the richest sources of miRNA in the human body (41). Salivary miRNAs originating from local tissue (such as the tongue and buccal mucosa) likely play a role in digestion, metabolism, and fluid balance. However, the majority of miRNAs in human saliva are concentrated within exosomes (12). Emerging evidence suggests that exosomal miRNA levels in whole saliva may provide important information about physiologic responses occurring in distant tissues (20). For example, studies of adolescents with sports-related concussions show that salivary miRNA profiles reflect the miRNA patterns in cerebrospinal fluid after traumatic brain injury (17). Thus saliva provides a noninvasive, abundant source of miRNAs that may yield information about systemic physiologic changes. These qualities make saliva an attractive biofluid for characterizing the body’s miRNA response to endurance exercise.

Previous studies of the miRNA response to exercise have focused almost exclusively on male participants, despite the fact that women display distinct musculoskeletal (26), cardiovascular (5), and metabolic responses to aerobic training (38). Indeed, sex-specific miRNA responses have been noted during typical human development and pathophysiologic responses to environmental insults (30). Such sex-specific miRNA patterns are thought to occur in response to gonadal steroids, such as dihydrotestosterone and progesterone. Notably, gonadal steroids are dynamically regulated during exercise. In this manner, the dynamic gonadotropin response to exercise may lead to distinct miRNA profiles among men and women athletes.

We hypothesize that levels of salivary miRNAs change in response to endurance running and that a subset of miRNA changes are sex specific. Furthermore, we posit that “altered” miRNAs target a diverse array of gene pathways involved in local metabolism and fluid balance, as well as distal muscle- and brain-related adaptations. In the present study we employ high-throughput RNA sequencing technology to assess levels of both miRNAs and mRNAs in human saliva following endurance running.

METHODS

This study was approved by the Independent Review Board at Marist College. Written informed consent was obtained from all participants.

Participant Information.

An observational cohort design was used to identify changes in salivary RNA among competitive distance runners by comparing pre- and post-run transcript profiles in a pairwise fashion. Participants included 13 men and 12 women collegiate distance runners, ages 18–23 yr, who completed their weekly “long run” on the day of the study. Long run was defined by a run exceeding 55 min and comprising ≥20% of weekly running distance. Participant exclusion criteria included acute illness (e.g., upper respiratory infection or gastrointestinal infection) or active orthopedic injury in the past 7 days. For all participants, participant characteristics were collected or calculated through self-reported surveys, including: sex, race/ethnicity, age (years), body mass index (kg/m²), medications, time since last meal (hours), and dietary restrictions (presence/absence). Participant fitness characteristics, including average distance run per week (km), run distance on the day of saliva collection (km), run duration (minutes), and run pace (minutes/km) were also recorded. Saliva and vital signs (including heart rate, body temperature, and blood pressure) were collected ~10–20 min before the run and again 10–20 min after completion of the run. Whole saliva was collected at these two time-points through active expectoration into Oragene RE-100 saliva collection kits (DNA Genotek, Ottawa, Canada) following oral water rinse.

RNA processing/filtering.

Saliva samples were transported to the State University of New York (SUNY) Upstate Molecular Analysis Core Facility and stored at −20^◦ C before RNA extraction with a standard Trizol technique and RNeasy mini columns (Qiagen, Valencia, CA). The RNA quality and yield for each sample were assessed with an Agilent Bioanalyzer (Agilent Technologies, Santa Clara, CA) before completion of strand-specific library construction according to the Illumina TruSeq Small RNA protocol. High-throughput RNA sequencing was completed on a NextSeq 500 Instrument (Illumina, San Diego, CA) at a targeted depth of 10 million single-end 50 base reads per sample. Adapter trimming, QC analysis, and RNA read alignments were performed in Partek Flow Software (Partek, St. Louis, MO). Reads were aligned to build 38 of the human genome using the Shrimp2 aligner and quantified using the E/M algorithm to the RefSeq transcript (version 82), miRBase precursor microRNA (version 21), or miRBase mature microRNA (version 21) reference databases. Alignment parameters for mature and precursor miRNAs included: 100% window length with local, ungapped alignment and 10 maximum hits per read. Single best mapping was used to align mRNA transcripts and included a 140% match window length, with a mismatch score of −15, gap open score of −33, and a window overlap filter of 90%. The 27,867 transcripts interrogated by RefSeq alignment (including mRNAs, long noncoding RNAs, small nucleolar RNAs, and other noncoding RNA species) were filtered to include only the 8,901 RNAs present in raw counts ≥ 10 in at least 20% of samples. The 4,694 miRNAs interrogated by miRBase alignment (including both mature and precursor miRNAs) were filtered to include only 460 miRNAs: 408 miRNAs were chosen based on their robust salivary expression (present in raw counts of ≥ 10 per sample in ≥ 25% of samples), and 52 miRNAs were included based on previous studies (1, 2, 4, 29, 39) indicating they might be influenced by exercise (n = 27). We also included 25 salivary miRNAs previously identified by our laboratory as potential biomarkers of sports-related concussion. The rationale for including these 25 targets in the present study was to ensure that molecules being explored as sideline indicators of traumatic brain injury were not confounded by the influence of noncontact aerobic exercise (as has been reported for other non-RNA concussion biomarkers) (35). For all RNAs, the raw miRNA and mRNA read counts within each sample were separately quantile normalized, mean-centered, and z-score transformed by dividing the expression level by the standard deviation of each variable before statistical analysis.

Statistical analysis.

The primary outcome of the study was the identification of miRNA features whose concentrations were altered following long run completion. Changes in miRNA concentration post-run were determined for the 25 participants using a paired nonparametric Wilcoxon rank-sum test with Benjamini Hochberg false discovery rate (FDR) correction. The miRNAs with FDR < 0.05 were considered to be significantly altered post-run. A partial least squares discriminant analysis (PLSDA) was used to visualize total miRNA profiles among pre- and post-run samples. Contributions of individual miRNAs to sample localization on the two-dimensional PLSDA were quantified with a variable importance in projection (VIP) score. The 15 miRNAs with the highest VIP score were utilized for hierarchical clustering of pre- and post-run samples with a Pearson distance measure and a complete clustering algorithm. DIANA miRPath (version 3) software was used to interrogate the miRNAs with the most significant post-run changes for predicted mRNA targets. The mRNAs with moderate target prediction evidence (microT-cds threshold >0.90, P < 0.05) were then utilized to identify overrepresented KEGG pathways (FDR < 0.05) using a Fisher’s exact test of conservative EASE score measurement. Functional influences of post-run miRNA changes on local salivary mRNA expression were interrogated with Wilcoxon rank-sum testing. The salivary mRNAs with nominal post-run expression changes (FDR < 0.1) were cross-referenced against predicted mRNA targets of the salivary miRNAs with post-run changes. Potential interactions among the protein products for targeted mRNAs were interrogated in String v10 software. Pearson correlation testing was used to identify miRNA-mRNA pairs with associated changes in post-run expression (R ≥ |0.40|). Pearson testing was also used to investigate participant factors and exercise variables that might influence circulating miRNA levels (R > |0.50|). For correlation testing, permissive P values (FDR < 0.15) were employed to avoid type II errors in which external influences from participant factors or exercise variables on miRNA expression might be overlooked (13, 16, 36). Finally, a two-way ANOVA was used to explore interactions between participant sex (a factor of interest) and collection time (post-run status). The miRNAs influenced (FDR < 0.05) by sex, post-run status, or a sex-run status interaction were reported.

RESULTS

Participants.

The participants were 48% women (12/25) and 92% Caucasian (23/25) (Table 1). Their mean age was 20 yr (± 1.3; range 18–23) and mean body mass index was 20.4 (± 1.7; range 17.8–23.9). The mean time since last meal was 2.8 h (± 3.3; range 0–10), and 20% (5/25) were fasting (no food/drink in past 8 h). Twenty percent (5/25) took a prescription medication, including three women and two men. Medications included selective serotonin reuptake inhibitors (antidepressants; n = 1), oral contraceptives (n = 2), inhaled beta-agonists and corticosteroids (for asthma; n = 2), and a GABA agonist (for anxiety control; n = 1). Twenty-four percent (6/25) reported dietary restrictions. Three participants reported nut-related allergies, one avoided gluten and dairy by choice, and one consumed a vegetarian diet. Average weekly distance run was 89 km/week (±16; range 64–121). On the day of saliva collection, participants ran an average of 18 km (±2.4; range 13–23) at a pace of 4:37/km (±0:22; range 3:59–5:19). The mean run duration was 83 min (±13; range 55–120). Pre-run, mean systolic blood pressure for all participants was 117 mmHg (±7; range 108–136) and diastolic pressure was 71 mmHg (±5; range 60–82). Pre-run heart rate was 64 beats/min (±8; range 44–80), and temperature was 36.7°C (±0.7; range 34.4–37.7). Post-run, mean systolic pressures increased (P = 0.0002) to 126 mmHG (±8; range 110–150), and mean diastolic pressures increased (P = 0.0004) to 78 mmHg (±6; range 66–90). Mean post-run heart rate was elevated (P = 1.2 × 10⁻¹⁰) to 89 beats/min (±10; range 75–111), but mean body temperature (36.4°C ± 0.9) was not changed (P = 0.22). Post-run saliva samples and vital signs were collected, on average, 11 min (±3; range 10–20) after run completion.

Table 1.

Participant characteristics

	All Participants, n = 25
Participant characteristics
Women, n (%)	12 (48)
Caucasian, n (%)	23 (92)
Age, yr, mean (SD; range)	20 (1.3; 18–23)
Body mass index, kg/m2, mean (SD; range)	20.4 (1.7; 17.8–23.9)
Fasting, n (%)	5 (20)
Dietary restrictions, n (%)	6 (24)
Run characteristics
Average weekly distance, km, mean (SD; range)	89 (16; 64–121)
Run distance, km, mean (SD; range)	18 (2.4; 13–23)
Run duration, min, mean (SD; range)	83 (13; 55–120)
Run pace, min/km, mean (SD; range)	4:37 (0:22; 3:59–5:19)
Vital signs
Pre-run heart rate, beats/min, mean (SD; range)	64 (8; 44–80)
Pre-run body temperature, °C, mean (SD; range)	36.7 (0.7; 34.4–37.7)
Pre-run systolic blood pressure, mmHg, mean (SD; range)	117 (7; 108–136)
Pre-run diastolic blood pressure, mmHg, mean (SD; range)	71 (5; 60–82)
Post-run heart rate, beats/min, mean (SD; range)	89 (10; 75–111)
Post-run body temperature, °C, mean (SD; range)	36.4 (0.9; 34.7–37.9)
Post-run systolic blood pressure, mmHg, mean (SD; range)	126 (8; 110–150)
Post-run diastolic blood pressure, mmHg, mean (SD; range)	78 (6; 66–90)
Time of post-run collection, min, mean (SD; range)	11 (3; 10–20)

Total miRNA expression.

There were a total of 1.24 × 10⁷ raw miRNA read counts among the 50 saliva samples. The mean miRNA read count per sample was 2.5 × 10⁵ (± 2.1 × 10⁵; range 1.1 × 10⁴–1.8 × 10⁶). There were 262 miRNAs present (raw counts > 0) in all 50 samples. There were no miRNAs universally present only in one sample group (pre- or postexercise) and universally absent in another. The closest any miRNA came to this expression pattern was pre-miR-206, which was present in 15 saliva samples pre-run, but only 6 samples post-run. Notably, alterations in miR-206 have been observed in previous miRNA exercise studies. Among the remaining 26 “exercise” miRNAs interrogated, 11 were absent in all 50 samples. This was also true for 7/25 (28%) of “concussion” miRNAs.

Running-induced miRNA changes.

There were 122 salivary miRNAs with significant (FDR < 0.05) changes post-run (Fig. 1A). Sixty-three miRNAs were downregulated post-run and 59 were upregulated. There were eight miRNAs with significant (FDR < 0.05) and highly consistent changes (≥23/25 samples) post-run (Fig. 1B). Six of these miRNAs were elevated post-run (miR-7154-3p, miR-200b-5p, miR-5582-3p, miR-6859-3p, miR-6751-5p, and miR-4419a), and two were decreased (miR-3671, miR-5095). There was one miRNA downregulated in all 25 post-run samples (miR-7154-3p, V-stat < 0.0001, FDR = 2.63E-5). One of the 122 miRNAs had been identified in previous exercise studies (pre-miR-206, FDR = 0.035, V-stat = 146), and two miRNAs were identified in concussion studies (miR-200b-3p, V-stat = 242, FDR = 0.035; miR-30e-5p, V-stat = 26, FDR = 0.0036). A two-dimensional PLSDA using total salivary miRNA profiles for each participant resulted in nearly complete separation of pre- and post-run samples, while accounting for 20.7% of the variance in expression (Fig. 2). Hierarchical clustering using the 15 miRNAs with the largest variable projection importance on PLSDA segregated all but 5/25 (20%) of the pre-run samples, and 3/25 (12%) of the post-run samples (Fig. 3). Notably, 5/8 “misclustered” samples clustered with their pre- or post-run counterpart. The 15 miRNAs were largely clustered into 10 that were upregulated post-run and five that were downregulated. The strongest miRNA clustering was observed for miR-6859-2-5p/miR-4253, and miR-552-5p/miR-4417.

Fig. 1.

Salivary microRNAs (miRNAs, miR) demonstrate robust, consistent post-run changes in abundance. A: a volcano plot displays the distribution of miRNA expression across pre- and post-run paired samples. There were 122 miRNAs with significant [false discovery rate (FDR) < 0.05] expression changes post-run. Eight of these miRNAs also displayed consistent pairwise expression changes across participants. B: whisker box-plots are shown for the eight salivary miRNAs with post-run expression changes on paired nonparametric Wilcoxon rank-sum testing (n = 25). Quantile normalized, z-score transformed mean group expression levels are displayed for pre-run samples (dark gray) and post-run samples (light gray).

Fig. 2.

Total salivary miRNA expression differentiates pre- and post-run samples. A two-dimensional partial least squares discriminant analysis using the expression profile of 460 salivary miRNAs (precursor and mature) resulted in nearly complete separation of pre-run (+, 1) and post-run (Δ, 0) samples (n = 25), while accounting for 20.7% of the variance in expression. The 95% confidence interval for sample dispersion is noted by the shaded ovals.

Fig. 3.

Hierarchical clustering of individual salivary RNA profiles pre- and post-run. Expression levels for the 15 salivary miRNAs with the largest variable importance in projection scores on partial least squares discriminant analysis were used to perform hierarchical clustering of individual samples with a Pearson distance metric and a complete clustering algorithm. This approach segregated all but 5/25 of the pre-run samples (red, PRE), and 3/25 (12%) of the post-run samples (green, POST). Note that “misclustered” samples tended to colocalize with associated pre- or post-run saliva from the same individual, belying the contributions of individual variation to salivary miRNA profiles. The individual miRNA pairs most closely clustered were miR-6859-2-5p/miR-4253, and miR-552-5p/miR-4417.

Predicted mRNA targets of altered miRNAs.

To explore the function of salivary miRNA changes, we interrogated the eight miRNAs most significantly altered post-run for predicted gene targets. These eight miRNAs targeted 2,389 mRNAs (microT-cds > 0.90, P < 0.05). The largest number of targeted mRNAs (1,159) was observed for miR-5582-3p. The 2,389 mRNA targets were found to be overrepresented (FDR < 0.05) in 15 KEGG pathways (Table 2A), including pathways related to metabolism (glycerophospholipid metabolism, P = 0.042; 17 genes, 6/8 miRNAs), water balance (aldosterone-regulated sodium reabsorption, P = 0.049, 12 genes, 5/8 miRNAs), and cardiac conduction (arrhythmogenic right ventricular cardiomyopathy, P = 0.018, 12 genes, 4/8 miRNAs). Interestingly, four brain-related pathways were also overrepresented, including Wnt signaling (P = 4.0E-5), morphine addiction (P = 0.0033), GABAergic synapse (P = 0.00016), and prolactin signaling (P = 0.0035).

Table 2.

KEGG pathways targeted by the miRNA clusters changed post-run

A. Salivary miRNAs Most Altered Post-run (n = 8)				B. Sex-related miRNAs Altered Post-run (n = 3)
KEGG Pathway	P Value	Genes, n	miRNAs, n	KEGG Pathway	P Value	Genes, n	miRNAs, n
Signaling pathways regulating pluripotency of stem cells	2.62E-07	35	7	fatty acid biosynthesis	4.93E-32	3	1
TGF-β signaling pathway	2.09E-05	15	4	fatty acid metabolism	1.76E-08	3	1
Wnt signaling pathway	4.03E-05	30	6	glycosphingolipid biosynthesis	0.002168	2	2
GABAergic synapse	0.000165	17	5	circadian entrainment	0.007997	9	2
Morphine addiction	0.003384	18	4
Proteoglycans in cancer	0.003384	40	6
Prolactin signaling pathway	0.003595	17	5
Arrhythmogenic right ventricular cardiomyopathy (ARVC)	0.01787	12	4
Gap junction	0.01787	17	5
Ras signaling pathway	0.01787	36	7
Pathways in cancer	0.01787	66	7
Chronic myeloid leukemia	0.02768	17	5
Glycerophospholipid metabolism	0.04247	17	6
Colorectal cancer	0.04529	14	6
Aldosterone-regulated sodium reabsorption	0.04961	12	5

Overlapping changes in mRNA/miRNA expression.

The full set of 122 miRNAs “altered” post-run (FDR<0.05) targeted a total of 140 mRNA transcripts that were detectable and nominally altered (FDR<0.1) in post-run saliva (Supplemental Table S1). (The online version of this article contains supplemental material.) This represented 53% (140/269) of the total saliva mRNA changes detected post-run and exceeded the number of miRNA-mRNA interactions expected by chance alone (P < 0.0001). Among the 122 miRNAs of interest, 96 (79%) targeted at least one transcript altered in post-run saliva. The miRNA that targeted the largest number of genes with post-run salivary changes was miR-4419a (25 altered mRNA targets). Fifty-four of the 122 miRNAs targeted five or more “altered” transcripts. The transcript targeted by the largest number of altered miRNAs was sestrin 3 (SESN3) (11 miRNAs). Fifty-nine of the 269 transcripts were targeted by three or more miRNAs of interest. These 59 mRNAs demonstrated a significant number of protein-protein interactions (P = 0.021) in String analysis, with 12 edges, and a clustering coefficient of 0.167. The protein domains significantly enriched among these miRNA targets were myosin tail (FDR = 0.024), and myosin NH₂-terminal SH3-like (FDR = 0.024). Among the altered mRNAs and miRNAs with predicted target interactions, 26 demonstrated correlations (R > |0.40|) in concentration change post-run (Table 3). Six of the miRNA/mRNA post-run changes were inversely correlated, and the remaining miRNA/mRNA pairs were positively correlated. The largest number of miRNA/mRNA target correlations were observed for colony stimulating factor 2 receptor beta (CSF2RB) (n = 4) and histone H2A deubiquitinase (MYSM1) (n = 4). The strongest correlation was observed between TAL BHLH Transcription Factor 1 (TAL1) and miR-605-3p (R = 0.71; P = 1.0E-8).

Table 3.

Associations between microRNA changes and target mRNA changes in saliva

mRNA	microRNA	Post-run Δ, mRNA/miRNA	Pearson R (P value)	Target Prediction Strength
ARID1A	miR-9-5p	↑/↓	−0.43 (0.001)	94
CSF2RB	miR-5698	↑/↓	−0.49 (0.0003)	80
	miR-181a-5p	↑/↓	−0.42 (0.002)	68
	miR-1273f	↑/↑	0.48 (0.0004)	69
	miR-4419a	↑/↑	0.51 (0.0001)	71
BRMS1L	miR-3671	↓/↓	0.52 (7.2E-7)	52
G3BP2	miR-4253	↓/↓	0.41 (0.002)	71
FJX1	miR-4419a	↑/↑	0.44 (0.0009)	53
JAK3	miR-6751	↑/↑	0.41 (0.002)	95
KAT2B	miR-181b-5p	↑/↑	0.42 (0.001)	50
	miR-7852-3p	↑/↑	0.44 (0.001)	59
HOXB9	miR-4251	↓/↓	0.40 (0.003)	52
MAP4K5	miR-429	↓/↓	0.49 (0.0002)	99
MYSM1	miR-548ac	↓/↑	−0.45 (0.001)	69
	miR-6737-3p	↓/↑	−0.43 (0.001)	72
	miR-429	↓/↓	0.46 (0.0006)	52
	miR-936	↓/↓	0.45 (0.001)	65
PREX1	miR-1915-3p	↑/↑	0.47 (0.0005)	72
PIAS2	miR-200a-3p	↑/↑	0.51 (3.1E-6)	77
	miR-4418	↑/↑	0.59 (6.1E-8)	58
OLIG1	miR-4488	↑/↑	0.46 (0.0007)	76
SOX8	miR-6124	↑/↑	0.40 (0.003)	75
SORL1	miR-3671	↑/↓	−0.45 (0.0009)	92
TAL1	miR-605-3p	↓/↓	0.71 (1.0E-8)	61
ZNF536	miR-4419a	↑/↑	0.42 (0.002)	73
ZBTB5	miR-302e	↑/↑	0.42 (0.002)	97

Direction of change for each mRNA/microRNA pair are denoted by arrows. For all mRNAs reported, the level of significance for post-run changes is false discovery rate (FDR) ≤ 0.1. For all reported microRNAs, the level of significance for post-run changes is FDR ≤ 0.05. Strength of evidence for predicted mRNA/microRNA interaction is denoted by micro-T-cds score, where 100 indicates the strongest possible, experimentally validated interaction.

Correlations between post-run miRNA changes and participant characteristics.

Participant characteristics and exercise measures were also evaluated for associations with post-run changes in salivary miRNA expression using Pearson (continuous variables) or Spearman Rank (dichotomous variables) testing (Supplemental Table S2). No miRNA changes were associated (R > |0.5|, FDR < 0.15) with time of sample collection post-run. Eight miRNAs were associated with run duration (minutes), but none of these miRNAs were associated with run distance (km) or pace of the run (min/km). There were no associations between post-run miRNA changes and post-run changes in temperature, systolic blood pressure, or diastolic blood pressure. However, there were 24 miRNA levels associated with post-run heart rate changes (beats/min). There were two miRNA changes associated with time since last meal, but no miRNAs associated with presence/absence of dietary restrictions. There were 23 miRNAs with post-run changes associated with female sex. Two miRNA changes were associated with participant age. Of the eight miRNAs with the most robust post-run changes on Wilcoxon testing, two had associations with female sex: miR-5095, and miR-6859-3p. None of the other variables showed significant associations with these eight miRNAs.

Influence of sex on miRNA responses to exercise.

A two-way ANOVA was employed to further examine the relationship between female sex and the salivary miRNA response to exercise. There were 13 miRNAs that differed (FDR < 0.05) between male and female subjects, and two of these miRNAs displayed robust interaction with post-run exercise status (Fig. 4). Both miR-6745 (FDR = 0.00025) and miR-6746-3p (FDR = 0.032) were decreased post-run in female subjects, but not in male subjects. A third miRNA, miR-4675, demonstrated differences among male/female subjects (FDR = 0.034) and pre-/post-run status (FDR = 0.037), but no interaction between sex and post-run status (FDR = 0.94). Hierarchical clustering was used to visualize patterns of salivary miRNA expression among male and female subjects before and after running by using the 18 miRNAs with the most robust interactions between sex and post-run status (Fig. 5). Three clusters of miRNA trends emerged: 1) Five miRNAs that tended to increase post-run only in female subjects; 2) Four miRNAs that tended to decrease post-run only in female subjects; and 3) Nine miRNAs with relatively consistent expression in male participants that demonstrated low patterns of expression in female runners both before and after the run. The three salivary miRNAs with unique post-run alterations in female participants (miR-6745, miR-6746-3p, miR-4675) targeted 265 mRNAs that were overrepresented in four KEGG pathways (Table 2B). Three of the four overrepresented pathways involved biosynthesis (Fatty acid biosynthesis, P = 4.9E-32, 3 genes, 1/3 miRNAs; glycosphingolipid biosynthesis, P = 0.0021, 2 genes, 2/3 miRNAs), or metabolism processes (fatty acid metabolism, P = 1.7E-8, 3 genes, 1/3 miRNAs).

Fig. 4.

Individual salivary miRNAs demonstrate sex-specific responses to exercise. A: levels of three salivary miRNAs were significantly [false discovery rate (FDR) < 0.05] influenced by pre-/post-run status and sex (miR-4675), or run-status/sex interactions (miR-6745 and miR-6746-3p) on a 2-way ANOVA. Whisker box plots of quantile normalized abundance are shown for male (n = 13) and female (n = 12) participants pre- and post-run. B: the Venn diagram indicates the number of salivary miRNAs influenced by sex, run-status, or sex/run-status interactions. C: two-way ANOVA FDR adjusted P values are shown for the 13 salivary miRNAs influenced by participant sex.

Fig. 5.

Visualization of sex-specific salivary miRNA changes across individual samples. A hierarchical clustering approach utilized individual expression levels for the 18 miRNAs influenced by run-status or sex on two-way ANOVA (FDR < 0.05) to visualize patterns of salivary miRNA expression in female (purple; n = 12) and male (green; n = 13) participants pre-run (blue) and post-run (pink). Five miRNAs tended to increase post-run only in female subjects. Four miRNAs tended to decrease post-run only in female subjects. Nine miRNAs demonstrated relatively consistent expression in male participants but displayed low patterns of expression in female runners both before and after the run.

DISCUSSION

Utilizing high-throughput sequencing techniques, this investigation identified a number of miRNAs that are altered in saliva following endurance exercise. The downstream genetic pathways targeted by these miRNAs involve metabolic processes (glycerophospholipid metabolism) and fluid regulation (aldosterone-regulated sodium reabsorption). In addition, circulating salivary miRNAs demonstrated enrichment for systemic targets, such as cardiac conduction (arrhythmogenic right ventricular cardiomyopathy), a finding underscored by the large number of miRNAs associated with post-run changes in heart rate (Supplemental Table S2). This lends support to the idea that salivary miRNA measurement may provide a non-invasive “liquid biopsy” of the systemic epitranscriptome response in humans.

Further support for this idea can be found through interrogation of the individual genes targeted by miRNAs post-run. For example, 11 of the 122 (9%) miRNAs that changed after endurance running targeted expression of SESN3, a transcript that was increased in saliva post-run (Supplemental Table S1). The SESN3 gene encodes a sestrin family protein that is induced by stress and reduces levels of intracellular reactive oxygen species (15). Interestingly, the SESN3 protein product is critical for normal regulation of blood glucose and is dysregulated in insulin resistance and obesity (24). Thus, downregulation of the miRNAs that target SESN3 may permit upregulation of this adaptive response during endurance exercise.

Remarkably, of the 269 mRNAs with post-run salivary changes, 59 (22%) were targeted by three or more miRNAs with post-run changes. Analysis of these 59 mRNAs revealed a significant number of putative interactions between their protein products, suggesting targeted regulation of a transcriptional network. In fact, the KEGG pathway with the most significant enrichment among this group was myosin-related proteins, which may play a role in musculoskeletal adaptations to endurance exercise. Another miRNA-mRNA interaction between miR-605-3p and its gene target, TAL1, appeared uniquely geared toward endurance running adaptation. Levels of miR-605-3p and TAL1 were robustly downregulated after exercise in a highly correlated manner (Table 3). The TAL1 protein product is implicated in erythroid differentiation and stimulated by erythropoietin but may be reduced through miR-605-3p-mediated cleavage (33). The paradoxical direct correlation between TAL1 and miR-605-3p could be explained by a phase shift in expression, wherein downregulation of miR-605-3p permits the translation and exhaustion of TAL1 mRNA reserves.

Surprisingly, the eight miRNAs with the most robust post-run changes also targeted a number of brain-related transcripts involved in processes such as Wnt Signaling, GABAergic synapse, and morphine addiction (Table 2A). This finding is consistent with previous reports that salivary miRNA may provide a window into central nervous system function (16, 17). Indeed, several of the mRNA/miRNA pairs with associated post-run changes are implicated in neuronal-related processes (Table 3). For instance, ARID1A is upregulated post-run and forms an integral part of the neural progenitors-specific chromatin remodeling complex (npBAF) that is required for proliferative control in neural stem cells (25). Decreased levels of miR-9-5p [which regulates npBAF (42) and displays inverse associations with ARID1A post-run] may facilitate the contributions of endurance exercise to hippocampal neurogenesis, potentiation, and memory (40). It is important to note that two of the study participants were taking antidepressant medication, and this may have influenced brain-related miRNAs in a subset of samples. However miRNA profiles for these two participants clustered tightly with peers (Fig. 3), suggesting that their miRNA profiles were minimally influenced by antidepressant medication. Thus, reported changes in brain-related miRNAs are likely the result of exercise-induced effects.

It is notable that four of the eight miRNAs most robustly changed in post-run saliva target 18 mRNAs implicated in morphine addiction (Table 2A). In the context of endurance exercise, these miRNAs may promote an endorphin response that underlies the experience of a “runner’s high” (euphoria associated with prolonged running). This possibility is supported by the observation that eight miRNAs demonstrate significant correlations (R > |0.40|) with run duration (Supplemental Table S2), a critical component for runner’s high ascertainment. In addition, five of the eight miRNAs associated with run duration also target nine morphine-related genes, a greater enrichment than expected by chance alone (FDR = 9.7E-7). Given these intriguing connections, further exploration of this potential mechanism is certainly warranted.

A novel finding of the current study is the unique influence of participant sex on the peripheral miRNA response to exercise. Previous studies have demonstrated that women have distinct cardiovascular (5), metabolic (38), and musculoskeletal (26) responses to exercise training. In skeletal muscle, these differences may be attributed, in part, to disparate transcriptional responses to resistance exercise (31). Sex-specific miRNA expression is known to play an important role in typical development and physiological adaptations to environmental stressors (30). Here, we show that post-run changes in the levels of 23 salivary miRNAs are associated with sex, and three miRNAs display potential sex-exercise interactions on two-way ANOVA (Fig. 3). Together, gene targets for these three miRNAs overrepresent metabolic pathways (Table 2B). In particular, miR-6745 targets three mRNAs (FASN, ACACB, and ACSL4) implicated in fatty acid biosynthesis. Among these targets, FASN has been implicated in the acute exercise response of skeletal (11) and hepatic (18) tissue. Interactions between miR-6745 and FASN may be partially responsible for the unique fatty acid kinetics observed in women during exercise (37) and weight-loss regimens (28). Thus, miR-6745 dynamics may contribute to sex-specific patterns of weight loss and fat redistribution resulting from endurance exercise.

Of the 27 miRNAs interrogated in previous exercise studies (1, 2, 4, 15, 29, 39), 12 were reliably detected in pre- and post-run saliva in the present study. Only one of these, pre-miR-206, was altered in post-run saliva, demonstrating significant downregulation. This deviates from a study of blood miRNAs in 14 male athletes postmarathon that demonstrated upregulation of miR-206 immediately post-run (29). It is possible that systemic increases in miR-206 following endurance exercise may be facilitated by processing and exhaustion of precursor miR-206 stores observed here. Alternatively, miR-206 may target alternative transcripts in blood and saliva, thus undergoing paradoxical shifts in the two biofluid spaces. Notably, the majority of previously described “exercise-related” miRNAs were studied based on their specificity to cardiac or skeletal muscle tissue. Therefore, it is not completely surprising that many (15/27) of these miRNAs are not robustly expressed in saliva. The one previous study of salivary miRNAs identified changes in miR-33a and miR-378a (22). The present study found no changes in miR-33a and did not detect miR-378a in the saliva of the 25 participants. However, salivary levels of miR-378f, which shares an identical seed sequence (UCCUGAC) with miR-378a, were upregulated post-run in 17/25 participants. Like the previous salivary miRNA exercise study, we also found that the miRNAs altered post-run were enriched for lipid metabolism targets. Differences in the individual miRNAs identified between these two studies may be explained by the female composition of our cohort, the use of an RNA-sequencing vs. PCR approach, or differences induced by running vs. cycling exercises. The latter variable may be particularly important given the finding by Banzet and colleagues (2013) (4) that changes in circulating miRNA levels are dependent on exercise modality.

Previously we have reported that salivary miRNAs may hold utility as biomarkers of concussion. Here, we show that of the 25 salivary miRNAs identified in previous concussion studies (17, 20), only two (miR-200b-3p and miR-30e-5p) are altered post-run. This finding highlights the suitability of salivary miRNA as a biomarker for sports-related concussion, where the confounding influence of exercise on salivary miRNA expression must be considered. The two miRNAs with running-induced changes (miR-200b and miR-30e-5p) may not be ideal for sideline concussion diagnosis. However, these two miRNAs may still be suitable as therapeutic biomarkers that can be tracked longitudinally in the weeks following concussion (especially since miR-30e predicts prolonged concussion symptoms). In light of findings that postconcussion exercise regimens may speed symptom recovery (3, 23), it is notable that levels of exercise-induced changes in these two miRNAs oppose the direction of change observed in children with prolonged concussion symptoms. For example, miR-30e is lower in children with prolonged concussion symptoms (compared with those with acute symptom resolution) but is increased by endurance running (20). For this reason, future studies tracking miR-200b-3p and miR-30e-5p in concussed patients undergoing low-impact exercise therapy (23) may provide an objective window into brain recovery.

To our knowledge this is the largest study of miRNA expression in human athletes (n = 25) and the first to examine the entire microtranscriptome (all 4,694 mature and precursor miRNAs) in exercising human participants. Nonetheless, there are several limitations to the current study. Foremost, is the lack of data regarding the menstrual cycle of female participants. Given the influence of circulating gonadotropins on miRNA expression, it is possible that women in the preovulatory phase (when estrogen and luteinizing hormone are elevated) might display disparate miRNA patterns from women in the luteal phase (when progesterone surges) (6). This certainly could have influenced the current data set. Additional analyses of miRNA from muscle or blood samples with the current high-throughput approach would also offer additional information about the aggregate human response to exercise. Unfortunately, such samples were not available for the current study. Future investigations interrogating global miRNA levels alongside physiologic markers (e.g., creatinine kinase, aspartate aminotransferase) and measures of exercise capacity (e.g., maximum oxygen uptake, anaerobic lactate threshold) would also provide valuable contextual data for interpretation. It should be noted that the current design employed a noncontrolled exercise regimen in which participants completed a range of running courses with slight variations in intensity (rather than using a common treadmill-based regimen). Such an approach makes replicability difficult but does provide a realistic setting, in which competitive distance runners typically train. By examining associations between course mileage and run speed we have attempted to control for some of the variation between individual running experiences. Although the current study is also among the first to begin exploring downstream changes in gene targets of exercise-related miRNAs, the filtering parameters (RNAs ≤ 50 base pairs) employed in library generation and sequence alignment may have limited recognition of mRNA targets. This approach eliminates many mRNA candidates with sequences > 50 base pairs that may be influenced by exercise. Thus, the absence of mRNAs of interest from Table 3 may represent type II errors and should not be assumed to represent a lack of mRNA-exercise response. Finally, it should be noted that 5/25 runners were fasting at the time of saliva collection, and this may have impacted salivary RNA profiles. However, previous work by de Jong and colleagues (2011) (21) demonstrates that the salivary proteome is not influenced by nutritional status, and our own investigations of saliva miRNA in healthy and concussed adolescents reveal minimal influence of dietary patterns on RNA concentrations (17).

In conclusion, a compelling number of salivary miRNAs demonstrate changes in concentration following sustained aerobic exercise. The salivary miRNAs with the largest changes target physiologically relevant pathways, including metabolism, fluid regulation, and cardiac conduction. This miRNA response is influenced in a sex-specific manner, particularly for a subset of miRNAs involved in fatty acid biosynthesis. Thus, salivary miRNAs may provide insight into a number of adaptive responses to aerobic training, and represent an easily accessible biomarker in endurance athletes.

GRANTS

This research was supported by a research agreement between Quadrant Biosciences Inc. and the SUNY Upstate Medical University in collaboration with Marist College.

DISCLOSURES

FAM and SDH are co-inventors of preliminary patents for miRNA biomarkers in disorders of the central nervous system that are assigned to the SUNY Upstate and Penn State Research Foundations and licensed to Quadrant Biosciences, Inc. SDH serves as a consultant for Quadrant Biosciences, Inc. These conflicts of interest are currently managed by the Penn State College of Medicine. The other authors have no conflicts to disclose.

AUTHOR CONTRIBUTIONS

S.D.H. and Z.G. conceived and designed research; S.D.H., F.A.M., and Z.G. analyzed data; S.D.H., F.A.M., and Z.G. interpreted results of experiments; S.D.H. and Z.G. prepared figures; S.D.H. drafted manuscript; S.D.H., P.J., F.A.M., O.P., and Z.G. edited and revised manuscript; S.D.H., P.J., F.A.M., O.P., and Z.G. approved final version of manuscript; P.J., F.A.M., O.P., and Z.G. performed experiments.