Response of circulating miRNAs to acute exercise: A systematic Review and Meta-Analysis

Kirstin MacGregor, Colin N Moran, Sophie Broome, Patrick J Owen, Séverine Lamon, Danielle Hiam
doi: https://doi.org/10.1101/2022.12.20.22283756

Abstract

Objective Cell-free microRNAs (cf-miRNAs) are secreted from cells and transported via the blood to exert their effect on target tissues. Numerous pathophysiological adaptations, including exercise, alter cf-miRNA levels. The aim of the systematic review was to investigate the cf-miRNA response to an acute bout of exercise and to interpret it using a robust correlated and hierarchical effects (CHE) meta-analysis.

Design The systematic review was registered in PROSPERO (CRD42021256303). A CHE meta-analysis was used to compare the changes in cf-miRNA levels and the influence of exercise modality. An exploratory machine-learning-based approach was used to capture influential moderators.

Data sources Primary studies were retrieved from PubMed and SPORTDiscus (09.03.2022). Relative changes in cf-miRNA expression in response to exercise were computed for each study. The ROBINS-I, GRADE and AMSTAR2 tools were used to assess evidence certainty and risk of bias.

Eligibility criteria Thirty-six studies including an acute exercise intervention in N=880 healthy males and females aged 18-45yrs met the eligibility criteria.

Results Muscle enriched cf-miR-1 (N=320), cf-miR-133a (N=195) and cf-miR-133b (N=132) levels increased 1-2hr (cf-miR1: FC = 2.72, 95% CI= 1.5-4.0; cf-miR133a: FC = 2.10, 95% CI = 1.6-2.6; cf-miR-133b: FC = 2.39, 95% CI = 1.2-3.6) and 24 hr post-exercise (cf-miR1: FC = 2.25, 95% CI= 1.3-3.2; cf-miR133a: FC = 1.81, 95% CI = 1.4-2.2; cf-miR-133b: FC = 1.99, 95% CI = 1.2-2.8).

Conclusion Acute exercise triggers temporal and modality specific responses in cf-miRNAs. levels. Influential moderators included sample size, collection time point, exercise modality, age and the use of various technical quality controls.

What is already known?

  • Exercise acutely alters cell-free miRNA (cf-miRNA) levels in human serum and plasma, but research is poorly reproducible

What are the new findings?

  • Muscle-enriched cf-miRNA levels robustly increase following an acute bout of exercise, with temporal and modality specific responses

  • The implementation of a CHE model, a novel statistical approach within the miRNA field, allowed to identify key methodological factors moderating cf-miRNA levels.

  • Strict implementation of these factors is warranted to improve rigour and reproducibility in this field.

Introduction

Extra-cellular or cell-free microRNAs (cf-miRNAs) are circulating miRNA molecules found in most biological fluids including blood, urine and saliva (1) that mediate protein expression via the regulation of translation (2-4). Cf-miRNAs are secreted from cells (5) and, while our understanding of their role and relevance in circulation is limited, specific cf-miRNAs display high levels of regulation in numerous pathological (6) and physiological conditions, including exercise (7).

Exercise is a potent physiological stressor triggering acute and chronic biological responses across the cardiovascular, musculoskeletal, and endocrine systems. Exercise relies on the activation of oxygen-dependent or –independent metabolic pathways to trigger muscle contraction (8) and results in a range of positive adaptations at the systemic and cellular levels. Depending on the type of exercise performed, long term adaptations include muscle hypertrophy and improvements in cardiovascular output, muscle fatigability and metabolic function. From a molecular perspective, even a single bout of exercise is enough to elicit a rapid physiological response in the muscle (9, 10) and the heart (11), at the gene (9, 10), protein (9, 12) and miRNA level (12, 13).

Upon acute physiological stress such as exercise, miRNAs produced in the muscle and other tissues are selectively or non-selectively secreted into the circulation (5). This might be to exert their effects on recipient cells (14), underpinning a role in cross-tissue communication (14, 15). Plasma and serum constitute readily available biological miRNA pools that underlie the popularity of cf-miRNA studies and the exponential increase of related research outputs (16). Research into cf-miRNAs and exercise is increasingly common but poorly reproducible. In a recent editorial, we highlighted that widespread inconsistency in methodologies may partly explain this variability (16). To avoid misinforming the field, the quality of individual methodological approaches must be considered when integrating and discussing existing findings. The aim of the systematic review was to examine the current literature regarding the cf-miRNA response to an acute bout of exercise, and to interpret it in the light of the known limitations of the field using a robust correlated and hierarchical effects (CHE) meta-analysis.

Exercise training leads to positive adaptations that, in the long term, are essential to reduce the increasing burden of chronic disease (17). Comprehensively mapping the molecular response to exercise, including at the cf-miRNA level, is essential to add to the existing body of knowledge and to inform future exercise and health research.

Methods

This systematic review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020, see supplementary file 1)(18). The protocol was registered in the international prospective register of systematic reviews (PROSPERO) (CRD42021256303). The Assessing the Methodological Quality of Systematic Reviews 2 (AMSTAR2) tool was used to appraise the systematic review (supp file 2).

Search strategy

The electronic databases PubMed and SPORTDiscus were searched by one reviewer (KM) to identify all relevant articles. The detailed search criteria can found in supplementary file 3. Databases were searched from inception to 17 May 2021 and updated on 09 March 2022. Additional articles were identified through hand searching the reference list of all included studies.

Study inclusion and exclusion criteria

The Population, Intervention, Comparison, Outcomes and Study design (PICOS) framework was used to define the study eligibility criteria. Eligible participants were males and females who were: 18-45 yr; <30 kg/m2; and, free from known cardiac disease, metabolic disease or cancer. Studies including an acute exercise intervention of any duration and modality were eligible. Comparisons were made between miRNA expression pre and up to 48 hr post exercise bout. Studies including the primary or secondary outcome of circulating miRNA extracted from serum, plasma, or extra-cellular vesicles were eligible. Experimental studies were considered for analyses if they were original research articles, published in full in a peer reviewed journal and full text available in English. Limiting to English language was previously shown to not meaningfully influence effect estimates (19).

Study selection

Prior to study screening, all duplicate studies were removed. Duplicate studies were identified by conducting an automated title screen using excel and confirmed manually screening the title, publication year and authors. Study selection was conducted independently by two reviewers (KM and DH) and followed a two-phase screening strategy. In phase one, title and abstracts were screened in all studies identified during the electronic database search. Studies that did not meet the inclusion criteria were excluded. In phase two, full text screening was conducted to exclude all remaining articles that did not meet the inclusion criteria. Any disagreements in study eligibility between the two reviewers were discussed and resolved by consensus or by arbitration from a third reviewer (SL or CM).

Data extraction

Data extracted from all eligible studies independently by two reviewers (KM and DH) using an a-priori designed data extraction form. Data extracted included author information, participant characteristics, exercise type, sampling timepoints, and study outcomes. Additionally, methodological data regarding analysis of miRNA expression were extracted and included: method of miRNA expression analysis (i.e. single qPCR, miRNA array or via sequencing); miRNA species assessed; quality control for haemolysis, spike-in control for RNA extraction and reverse transcription and whether miRNA was normalised to spike-in or stable miRNAs (supp file 4 for details). Any disagreements in data extraction between the two reviewers were discussed and resolved by consensus or by arbitration from a third reviewer (SL or CM). Where data were not available in the published manuscript, authors were contacted to obtain relevant information. Authors were emailed twice and given four weeks to respond. One article (20) was subsequently removed as the minimum data required for meta-analysis (mean and SD) could not be retrieved.

Quality assessment

Quality of all included studies were assessed independently by two independent reviewers (KM and DH) using the risk of bias in non-randomised studies of interventions (ROBINS-I) tool (21). Any disagreements between the two reviewers were discussed and resolved by consensus or by arbitration from a third reviewer (SL or CM). The ROBINS-I tool has seven domains: bias due to confounding, bias due to selection of participants, bias in classification of interventions, bias due to deviations from intended interventions, bias due to missing data, bias in measurement of outcomes and bias in selection of the reported results. Detailed information on the criteria used for grading are documented in supp file 5. Each domain was assessed as low, moderate, serious or critical risk of bias. The overall risk of bias for each study was determined to be the same level as the highest risk of bias allocated to an individual domain. Studies were not excluded from the meta-analysis based on risk of bias; however, risk of bias was taken into consideration during reporting of results.

Outlier and influence diagnostics

To identify potential outliers, we identified any primary studies where the 95% confidence interval did not overlap with the pooled estimate 95% confidence interval. To identify influential studies, we calculated and plotted Cook’s distance that examines the change in overall effect size when excluding one study at a time (22) .

Publication Bias

Qualitative evaluation was estimated by funnel plot asymmetry, which suggests the presence of publication bias. Egger’s regression was used to confirm the findings. Certainty of evidence: The GRADEpro GDT software was used to assess certainty of evidence for primary studies (23). This was based on risk of bias, inconsistency, indirectness, imprecision, and publication bias. The grades of evidence are characterised as follows: High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect (24). Detailed information on the criteria used for grading are documented in supp file 6.

Data analyses

All data was analysed in R version 4.1.1 using the add-on packages; metafor (25), clubSandwich (26), metaForest(27), tidyverse (28), pema (29) and robvis (30). For each study, we computed the mean and standard deviation (SD) of the fold-change in miRNA expression, calculated via the delta-Ct method and expressed in arbitrary units, between the baseline and the relevant post-exercise timepoints. We focused on the baseline and three specific timepoints post an acute exercise bout across studies: immediately post-(POST), 1-2 hr post-(1HP) and 24 hr post (24HP). MicroRNAs that had been investigated in five or more studies were selected for further analysis to reduce the risk of type I errors (31). These included miR-1, miR-21, miR-126, miR-133a/b, miR-146, miR-206, miR-208, miR-210, miR-221 and miR-222. Due to insufficient extra-cellular vesicles (ECV) data (N = 4 primary studies) we were unable to conduct a meta-analysis of miRNA levels in ECV and therefore we report meta-analysis findings only for cell-free miRNAs. For each miRNA, a variance-covariance matrix was constructed to approximate sampling errors of the dependent timepoints within each cohort (repeated measures). A correlated and hierarchical effects mixed effects meta-analysis was performed, which included a continuous-time autoregressive structure as the timepoints were not evenly spaced (25, 32). This was followed by cluster-robust estimate using the “sandwich” estimator to account for any misspecification of the model (26). The fixed effects were the fold-change compared to baseline at each time point and the influence of exercise modality (endurance or resistance) on miRNA levels. If there were two cohorts in a study, these were treated as separate cohorts. The random effects were of a nested model structure, where observations within the same cohort were dependent and observations from different cohorts were independent. Finally, to understand whether we needed to account for between-study variation in the meta-analyses, we constrained the between-study variance to zero and ran the likelihood ratio test comparing the model to the full model. Based on the Akaike information criterion (AICc), there was significant variation between cohorts indicating that a three-level model approach was warranted. Heterogeneity at each level of the meta-analysis was assessed via the Cochrane Q-test and I2 to determine the variance attributed to the sampling error, and variability between studies (33).

Exploratory/Meta-regression

Moderator analysis was then conducted to explain the widespread variability in cf-miRNA findings by evaluating the influence of differences in methodological techniques (16) and other sources of variability (sample size and proportion of males). The number of moderators coded (n=8) was large relative to the sample size. Therefore, to minimise the risk of including all moderators in a meta-regression and overfitting the model, we implemented machine-learning algorithm ‘random forests’ that are robust to overfitting. To select the moderators for clustered cross-validation and model tuning, a precursive replication was run and replicated 100 times. Only moderators that achieved positive variable importance in more than 50% of the replications were retained. Moderators were ranked in terms of their importance in predicting the effect size. Lastly, a measure of ‘predictive performance’, or ‘out-of-bag’ Roob2, was calculated for each moderator. Moderators that consistently displayed negative variable importance (i.e., that showed a reduction in predictive performance) were dropped. To ensure that each model converged (indicating the model can make predictions and detect some reliable patterns in the data) the metaforest was replicated 100 times and the distribution of Roob2 was plotted to ensure that the majority of the distribution was above zero. Moderators that improved predictive performance were then entered into a linear meta-regression to understand their association with the effect size. If the model failed to converge and RRoob2 was negative, indicating that the moderators did not improve the fit of the model and had no influence on the effect size, a Bayesian regularised meta-regression was run to ensure the robustness of the findings.

Narrative results

Literature search

The electronic database search retrieved 337 articles, of which 310 articles remained following removal of duplicates. Following title and abstract screening, 68 articles were identified as potentially eligible. Based on a full text assessment, 36 eligible articles were included in the meta-analysis (Figure 1).

Figure 1:
Figure 1: Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines flow chart for literature search and study selection.
Figure 2:
Figure 2: Summary plot for the risk of bias assessment of included studies using the Risk of Bias in Non-Randomised Studies of Interventions (ROBINS-I) tool.

Participant and study characteristics

Participant (N=880) and primary study (N=36) characteristics are described in Table 1.

View this table:
Table 1. Study and participant characteristics of eligible studies

MiRNA analysis characteristics

MiRNA analysis characteristics for primary studies (N=36) are described in Table 2.

View this table:
Table 2. MiRNA analysis characteristics of primary studies

Study quality

Study quality was assessed using the ROBINS-1 tool for assessing risk of bias in non-randomised studies of interventions3 (supp file 8, figure 1). Quality of studies was frequently decreased due to: failure to adjust for baseline confounding factors including age, BMI, sex, menstrual cycle status and fitness; insufficient control for exercise intensity or duration; failure to appropriately control for nutritional intake or exercise 24 hr prior to the intervention; missing data without explanation and lack of reported N values; issues with reporting of miRNA statistical analysis, including absence of multiple testing correction and insufficient reporting of statistical analysis. Out of the 35 studies, two studies had an overall low risk of bias 4,5, eight studies had an overall moderate risk of bias 1,2,6–11, 23 studies had an overall serious risk of bias 12–34 and 3 studies had an overall critical risk of bias 35–37.

Certainty of evidence

There was a low certainty of evidence for all cf-miRNAs and our confidence in the effect estimate was limited. This was mainly due to a high risk of bias, and a high level of imprecision these findings are summarised in supp file 5.

Meta-analysis

Circulating MiRNA-1

Findings for miR-1 are summarized in Table 3 (Nstudies = 14, Nparticipants = 320). There were no differences in miR-1 levels immediately following a bout of acute exercise, but miR-1 levels significantly increased from baseline in the 1-2 hours post-acute exercise and 24 hours post-acute exercise. Overall, miR-1 levels were also higher in response to endurance compared to resistance exercise (Fig 3A). Due to limited studies, we were however unable to delineate the influence of exercise modality on the levels of miR-1 at each time point (interaction).

View this table:
Table 3: Summary of meta-analysis findings
Figure 3.
Figure 3.

Forest Plot (Top): Overall fold change and 95% confidence intervals of miR-1 (3A) and miR-133a (3B) levels after an acute bout of exercise. *Denotes significant difference in miR-1 (3A) and miR-133a (3B) levels compared to baseline p<0.05. Longitudinal Plot (Bottom) of miR-1 (3A) and miR-133a (3B) levels across the timepoints in each of the primary studies coloured by exercise modality. Black line denotes endurance exercise, blue dashed line denotes resistance exercise.

Circulating miRNA-133a

Findings for miR-133a are summarized in Table 3 (Nstudies = 14, Nparticipants = 195). MiR-133a levels increased across all time points. Overall, miR-133a levels were also higher in response to endurance compared to resistance exercise (Fig 3B). Once again due to limited studies, we were unable to delineate the influence of exercise modality on the levels of miR-133a at each time point (interaction).

Circulating miRNA-133b

Findings for miR-133b are summarized in Table 3 (Nstudies = 10, Nparticipants = 132). There were no differences in miR-133b levels immediately following a bout of acute exercise, but miR-133b levels significantly increased from baseline in the 1-2 hours post-acute exercise and 24 hours post-acute exercise (Fig 4A).

Figure 4.
Figure 4.

Forest Plot: Overall fold change and 95% confidence intervals of miR-133b (4A), miR-206 (4B) and miR-146 (4C) levels after an acute bout of exercise. *Denotes significant difference in miR-133b (4A) levels compared to baseline p<0.05.

Circulating miRNA-206

Findings for miR-206 are summarized in Table 3 (Nstudies = 8, Nparticipants =). There were no differences in miR-206 levels immediately following a bout of acute exercise, in the 1-2 hours post-acute exercise or 24 hours post-acute exercise (Fig 4B).

Circulating miRNA-146

Findings for miR-146 are summarized in Table 3 (Nstudies = 9, Nparticipants = 236). There were no differences in miR-146 levels immediately following a bout of acute exercise, in the 1-2 hours post-acute exercise or 24 hours post-acute exercise (Fig 4C).

Circulating miR-21, miR-126, miR-208, miR-210, miR-221 and miR-222

In the miR-21, miR-126, miR-210, miR-221 and miR-222 analyses, the likelihood plot of tau2 (indicative of between study variance) was flat and the random effect was not significant (Table 3). This could indicate either that there was no significant residual heterogeneity between studies, or that the model was over-parameterised. Therefore, all the models were simplified by removing the correlated random effects. Findings for miR-21 (Nstudies = 10, Nparticipants = 108), miR-126 (Nstudies = 7, Nparticipants = 167), miR-208 (Nstudies = 6, Nparticipants = 60), miR-210 (Nstudies = 5, Nparticipants = 49), miR-221 (Nstudies = 8, Nparticipants = 83), and miR-222 (Nstudies = 8, Nparticipants = 128) are summarized in Table 3. There were no significant changes in miR-21, miR-126, miR-208 (Supp Fig 2), miR-210, and miR-222 (Supp Fig 3) levels at any time point after an acute bout of exercise.

Discussion

The primary aim of this systematic review and meta-analysis was to evaluate the influence of an acute exercise bout on cf-miRNA levels in apparently healthy individuals. Eleven individual cf-miRNAs across 36 primary studies (participants n= 880) met our inclusion criteria. A robust CHE meta-analysis determined that cf-miR-1 and cf-miR-133b levels increased 1-2 hr and 24 hr post exercise, and cf-miR-133a levels increased immediately, 1-2 hr and 24 hr post exercise. Levels of cf-miR-1 and cf-miR-133a were greater in response to endurance compared to resistance exercise. Collectively, our findings reveal temporal and modality specific cf-miRNA responses to an acute exercise bout.

Muscle-enriched circulating miRNAs levels are altered with acute exercise

Exercise is one of the most potent physiological stimuli. Muscles contract and relax repeatedly leading to substantial physiological changes. A recent meta-analysis showed that out of the ∼20,000 human protein coding genes (34), almost 3,000 genes change their expression in response to acute exercise (35). In addition, the levels of hormones (36), cytokines (37) and other molecules alter, signalling the change in physiological demand to all parts of the body. The human genome encodes 2,693 mature miRNAs (miRBase v. 22), with at least 10-15% exported out of the cell and found at currently detectable levels in the circulation (38). Cf-miRNAs may be involved in cell-cell communication, influencing gene expression in cells other than the one that produced them (14, 15). Here we show that the levels of cf-miRNAs-1, -133a and -133b robustly increase in response to acute exercise and may be part of this signalling network.

miR-1 is a widely conserved miRNA with high levels of expression primarily in muscle tissue (39) and is part of a group of muscle-enriched miRNAs originally coined as “myomiRs” (40). In skeletal muscle, miR-1 is regulated by mTOR (41), a key regulator of skeletal muscle response to exercise (42) that is equally important for cardiac remodelling following exercise (42). Here we show increases in cf-miR-1 following exercise that may indicate involvement in cardiac or skeletal muscle remodelling, or both, possibly due to modulation of the PI3K/AKT pathway.

miR-133a is also a widely conserved miRNA primarily expressed in muscle tissue (43). The genes that produce miR-133a and miR-1 are clustered together, with the two miRNAs transcribed as a single primary transcript (44). In skeletal muscle, miR-133a stimulates muscle growth by reducing myoblast differentiation, possibly through downregulation of tropomyosin-4, an actin binding component of the cytoskeleton (45). In cardiac tissue, miR-133a is linked to hypertrophy, although so far only as a marker of disease (e.g. hypertrophy (46) or sudden cardiac death (47)). In addition, miR-133a has predicted binding sites in the TCF7, MSI1, and PAX5 genes of the Wnt signalling pathway (48), which is important for both cardiac and skeletal muscle development and maintenance. miR-133b is identical to miR-133a except for a single base at the 3’ end (44). It is also primarily expressed in muscle tissues. Circulating miR-133b is reduced in sarcopenic individuals (49), and miR-133b also has targets in the Wnt signalling pathway (50). Increases in cf-miR-133 and cf-miR-133b following exercise may indicate their involvement in protection or remodelling of cardiac or skeletal muscle tissue response through modulation of tropomyosin-4 or Wnt signalling.

MiR-133b is clustered with another miRNA, miR-206, in a similar fashion to miR-133a and miR-1. All four miRNAs, miR-1, miR-133a, miR-133b and miR-206, increase their levels during the development of human skeletal muscle (51). In fact, the expression patterns of miR-206 are generally similar to miR-1, notwithstanding the differences reported here; although, their sequences differ by four nucleotides (52). Nonetheless, we did not detect a robust response to exercise for miR-206 but found that exercise modality and sampling timepoint were significant variables moderating the levels or miR-1, miR-133a, miR-133b and miR-206. This may be a reflection of the close physical relationship between these four miRNAs.

Over 60% of human protein coding genes are regulated by miRNAs (53), meaning that every signalling pathway most likely contains some elements that are under miRNA regulation. Most miRNAs can bind to multiple mRNAs (54), and each mRNA typically has binding sites for several distinct miRNAs; although, in both cases, the distribution is not uniform (55). This network of interactions allows miRNAs to function in groups to fine tune control of gene expression with a high degree of resolution. The specific miRNAs reported here were chosen due to their investigation in at least five primary studies. However, the cf-miRNAs in which we report changes relating to exercise are all myo-miRs (40), and we may expect them to work together. Their presence in the circulation suggests that they are exported from exercising muscle and may be signalling to other tissues to prepare for physiological changes.

Given the known differences in physiological demand that result from different modes of exercise, it is reasonable to expect that different miRNAs may respond differently to different acute exercise protocols. Significant differences were detected in the overall expression levels of miR-1, miR-133a and miR-221 between exercise modality groups, where endurance exercise induced a greater response to miR-1 and miR-133a and resistance exercise induced a greater response in miR-221. Due to limited primary studies, we were however unable to run an interaction to delineate the influence of exercise modality on the levels of these miRNAs at each timepoint. It will be important to investigate these differences between exercise modalities in future studies to understand how different modes of exercise can influence ci-miRNA levels and therefore the specific pathways that are altered.

Strengths and limitations

This systematic review included 36 primary studies; however, few were considered low risk of bias. Bias due to deviations from intended interventions and co-interventions were judged as moderate, serious or critical in 26 studies. This was primarily attributed to insufficient control of the exercise intervention, including intensity and duration, and nutritional intake. Additionally, most studies failed to appropriately report baseline confounding variables and/or recruit participants with similar baseline confounding variables, including fitness, BMI, sex and age. All these factors are important moderators of miRNA expression (56, 57) and future studies within the field should implement rigorous methodology to minimise risk of bias, increase robustness and reproducibility, and generally facilitate comparability within the literature body. Improving quality, rigour and reproducibility (58) in physiological sciences has recently become a focus of many (59, 60). As we hope to increase reproducibility and transparency in the cf-miRNA field (16), the full R code and the data used in the analysis is available on GitHub (https://github.com/DaniHiam/circulating-miRNAs-and-exercise-).

Instead of the planned subgroup/meta-regression approaches, we ran an exploratory machine-learning-based approach to try to capture relevant moderators that could be influencing effect size. The a priori moderators included aspects of the methodological approach, sample size, and proportion of males, based on literature that had previously highlighted them as common pitfalls in cf-miRNA analysis (16, 61, 62). Sample size was the highest ranked moderator in the analysis of miRNA-1, miRNA-133a and miRNA-133b. This is not overly surprising as small sample size in primary studies is a commonly cited limitation of meta-analyses that leads to increased risk of type I errors (false negatives) and type II errors (false positives) in the overall effect size (63). Along the same lines, implementation of best practise recommendations for miRNA analysis (16, 64), including stringent and rigorous methodological checks, was commonly missing and may explain some of variability in effect sizes. Specifically, the assessment of haemolysis (missing in 69% of studies) and inclusion of a spike-in control to normalise for differences in miRNA input during RNA extraction (missing in 44% of studies) were ranked as influential moderators. Due to the difficultly to accurately determine the concentrations of circulating miRNAs, many studies elect to use equal amounts of total RNA volume as a surrogate for equal amounts of RNA input for RNA extraction(16). This method is only valid if an appropriate spike-in control is used to account for differences cf-miRNA input. Control of haemolysis is another common pitfall that may lead to discrepancies between studies, in exercise research and more broadly (16). For example, erythrocytes can contain miRNAs and should be thoroughly eliminated to reduce the confounding variation (16, 65). Altogether, these differences in the use of quality controls can dramatically increase between-study variability. As previously discussed by our group (16), these methodological checks should become the ‘norm’ in cf-miRNA studies as they will increase reproducibility and help untangling the cf-miRNA response following exercise as well as in multiple pathophysiological conditions.

Consistent with previous reports in the fields of physiology (66) and exercise physiology (67), only 13% of the participants of the 36 selected studies were females, and menstrual status was controlled in only two of these studies. Twenty-seven studies did not include females at all, eight studies included both males and females, and no study focused on females exclusively. The reasons why sex-specific differences in the response to exercise, as well as in other fields of physiology and medicine, have been overlooked historically tend to relate to financial costs and feasibility (68). The lack of female-only and mixed sex studies however, leads to insufficient information to differentiate between the divergent exercise response of males and females (69, 70). Skeletal muscle, the primary tissue allowing locomotion, is one of the main tissues underpinning sex-based differences and has up to 3,000 genes differentially expressed in males and females at baseline (71). A recent study highlighted profound differences in the skeletal muscle methylome of males and females (69), suggesting that these inherent, sex-specific differences may also be reflected at other epigenetic levels, including non-coding RNA expression. Limited literature in other tissues indicates that miRNA expression is sex-specific (72-74). This includes plasma (75), where androgenic and ovarian hormones may modulate the expression of specific miRNA species (75, 76). Our findings must therefore be interpreted in the light of this limitation, and future studies investigating the changes in cf-miRNAs levels should systematically include sex and treat it as a fixed factor or covariate in the analyses.

In conclusion, we implemented a CHE meta-analysis and reported the effect of an acute bout of exercise on common cf-miRNAs in humans. Muscle-enriched cf-miRNA levels robustly increased following an acute bout of exercise with temporal and modality specific responses. By implementing a novel statistical approach within the miRNA field, we were able to identify key methodological factors moderating cf-miR expression following acute exercise.

Data Availability

The full R code and the data used in the analysis is available on GitHub (https://github.com/DaniHiam/circulating-miRNAs-and-exercise-).

https://github.com/DaniHiam/circulating-miRNAs-and-exercise-

References

  1. 1.↵
    Weber JA, Baxter DH, Zhang S, Huang DY, Huang KH, Lee MJ, et al. The microRNA spectrum in 12 body fluids. Clin Chem. 2010;56(56):1733–41.
    OpenUrlAbstract/FREE Full Text
  2. 2.↵
    Guo H, Ingolia NT, Weissman JS, Bartel DP. Mammalian microRNAs predominantly act to decrease target mRNA levels. Nature. 2010;466(466):835–40.
    OpenUrlCrossRefPubMedWeb of Science
  3. 3.
    McGeary SE, Lin KS, Shi CY, Pham TM, Bisaria N, Kelley GM, et al. The biochemical basis of microRNA targeting efficacy. Science. 2019;366(6472).
  4. 4.↵
    Vasudevan S, Tong Y, Steitz JA. Switching from repression to activation: microRNAs can up-regulate translation. Science. 2007;318(318):1931–4.
    OpenUrlAbstract/FREE Full Text
  5. 5.↵
    Arroyo JD, Chevillet JR, Kroh EM, Ruf IK, Pritchard CC, Gibson DF, et al. Argonaute2 complexes carry a population of circulating microRNAs independent of vesicles in human plasma. Proc Natl Acad Sci U S A. 2011;108(108):5003–8.
    OpenUrlAbstract/FREE Full Text
  6. 6.↵
    Lu J, Getz G, Miska EA, Alvarez-Saavedra E, Lamb J, Peck D, et al. MicroRNA expression profiles classify human cancers. Nature. 2005;435(435):834–8.
    OpenUrlCrossRefPubMedWeb of Science
  7. 7.↵
    Russell AP, Lamon S. Exercise, Skeletal Muscle and Circulating microRNAs. Prog Mol Biol Transl Sci. 2015;135:471–96.
    OpenUrlCrossRefPubMed
  8. 8.↵
    Hargreaves M, Spriet LL. Skeletal muscle energy metabolism during exercise. Nat Metab. 2020;2(2):817–28.
    OpenUrl
  9. 9.↵
    Perry CG, Lally J, Holloway GP, Heigenhauser GJ, Bonen A, Spriet LL. Repeated transient mRNA bursts precede increases in transcriptional and mitochondrial proteins during training in human skeletal muscle. J Physiol. 2010;588(Pt 23):4795–810.
    OpenUrlCrossRefPubMed
  10. 10.↵
    Russell AP, Wallace MA, Kalanon M, Zacharewicz E, Della Gatta PA, Garnham A, et al. Striated muscle activator of Rho signalling (STARS) is reduced in ageing human skeletal muscle and targeted by miR-628-5p. Acta Physiol (Oxf). 2017;220(220):263–74.
    OpenUrl
  11. 11.↵
    Ludlow AT, Gratidão L, Ludlow LW, Spangenburg EE, Roth SM. Acute exercise activates p38 MAPK and increases the expression of telomere-protective genes in cardiac muscle. Exp Physiol. 2017;102(102):397–410.
    OpenUrl
  12. 12.↵
    Zacharewicz E, Della Gatta P, Reynolds J, Garnham A, Crowley T, Russell AP, et al. Identification of microRNAs linked to regulators of muscle protein synthesis and regeneration in young and old skeletal muscle. PloS one. 2014;9(9):e114009.
    OpenUrlCrossRefPubMed
  13. 13.↵
    Russell AP, Lamon S, Boon H, Wada S, Guller I, Brown EL, et al. Regulation of miRNAs in human skeletal muscle following acute endurance exercise and short-term endurance training. J Physiol. 2013;591(591):4637–53.
    OpenUrlCrossRefPubMed
  14. 14.↵
    Thomou T, Mori MA, Dreyfuss JM, Konishi M, Sakaguchi M, Wolfrum C, et al. Adipose-derived circulating miRNAs regulate gene expression in other tissues. Nature. 2017;542(542):450–5.
    OpenUrlCrossRefPubMed
  15. 15.↵
    Guescini M, Canonico B, Lucertini F, Maggio S, Annibalini G, Barbieri E, et al. Muscle Releases Alpha-Sarcoglycan Positive Extracellular Vesicles Carrying miRNAs in the Bloodstream. PloS one. 2015;10(10):e0125094.
    OpenUrlCrossRefPubMed
  16. 16.↵
    Hiam D, Lamon S. Circulating microRNAs: let’s not waste the potential. Am J Physiol Cell Physiol. 2020;319(319):C313–C5.
    OpenUrl
  17. 17.↵
    McGee SL, Hargreaves M. Exercise adaptations: molecular mechanisms and potential targets for therapeutic benefit. Nat Rev Endocrinol. 2020;16(16):495–505.
    OpenUrlPubMed
  18. 18.↵
    Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021:n71.
  19. 19.↵
    Morrison A, Polisena J, Husereau D, Moulton K, Clark M, Fiander M, et al. The effect of English-language restriction on systematic review-based meta-analyses: a systematic review of empirical studies. Int J Technol Assess Health Care. 2012;28(28):138–44.
    OpenUrlCrossRefPubMed
  20. 20.↵
    Kuji T, Sugasawa T, Fujita S-i, Ono S, Kawakami Y, Takekoshi K. A Pilot Study of miRNA Expression Profile as a Liquid Biopsy for Full-Marathon Participants. Sports. 2021;9(9):134.
    OpenUrl
  21. 21.↵
    Sterne JAC, Hernán MA, Reeves BC, Savović J, Berkman ND, Viswanathan M, et al. ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions. BMJ. 2016:i4919.
  22. 22.↵
    Viechtbauer W, Cheung MW. Outlier and influence diagnostics for meta-analysis. Res Synth Methods. 2010;1(1):112–25.
    OpenUrlPubMed
  23. 23.↵
    Prime. MUaE. GRADEpro GDT: GRADEpro Guideline Development Tool [Software]. 2022.
  24. 24.↵
    Balshem H, Helfand M, Schünemann HJ, Oxman AD, Kunz R, Brozek J, et al. GRADE guidelines: 3. Rating the quality of evidence. J Clin Epidemiol. 2011;64(64):401–6.
    OpenUrlCrossRefPubMedWeb of Science
  25. 25.↵
    Viechtbauer W. Conducting Meta-Analyses in R with the metafor Package. Journal of Statistical Software. 2010;36(36):1–48.
    OpenUrl
  26. 26.↵
    Pustejovsky JE, Tipton E. Meta-analysis with Robust Variance Estimation: Expanding the Range of Working Models. Prevention Science. 2022;23(23):425–38.
    OpenUrl
  27. 27.↵
    Van Lissa CJ. Metaforest: Exploring Heterogeneity in Meta-analysis Using Random Forests. PsyArXiv. 2017.
  28. 28.↵
    Wickham H, Averick M, Bryan J, Chang W, McGowan LDA, François R, et al. Welcome to the Tidyverse. Journal of Open Source Software. 2019;4(4):1686.
    OpenUrl
  29. 29.↵
    Van Lissa CJ, van Erp S. Select relevant moderators using Bayesian regularized meta-regression. PsyArXiv. 2021.
  30. 30.↵
    McGuinness LA, Higgins JPT. Risk-of-bias VISualization (robvis): An R package and Shiny web app for visualizing risk-of-bias assessments. Research Synthesis Methods. 2021;12(12):55–61.
    OpenUrl
  31. 31.↵
    Van den Noortgate W, López-López JA, Marín-Martínez F, Sánchez-Meca J. Three-level meta-analysis of dependent effect sizes. Behavior research methods. 2013;45(45):576–94.
    OpenUrlCrossRefPubMedWeb of Science
  32. 32.↵
    Ishak KJ, Platt RW, Joseph L, Hanley JA, Caro JJ. Meta-analysis of longitudinal studies. Clinical Trials. 2007;4(4):525–39.
    OpenUrl
  33. 33.↵
    Assink M, Wibbelink CJM, editors. Fitting three-level meta-analytic models in R: A step-by-step tutorial 2016.
  34. 34.↵
    Nurk S, Koren S, Rhie A, Rautiainen M, Bzikadze AV, Mikheenko A, et al. The complete sequence of a human genome. Science. 2022;376(376):44–53.
    OpenUrlCrossRef
  35. 35.↵
    Pillon NJ, Gabriel BM, Dollet L, Smith JA, Puig LS, Botella J, et al. Transcriptomic profiling of skeletal muscle adaptations to exercise and inactivity. Nature communications. 2020;11(11):1–15.
    OpenUrl
  36. 36.↵
    Howlett TA. HORMONAL RESPONSES TO EXERCISE AND TRAINING: A SHORT REVIEW. Clinical endocrinology. 1987;26(26):723–42.
    OpenUrlPubMed
  37. 37.↵
    de Salles BF, Simão R, Fleck SJ, Dias I, Kraemer-Aguiar LG, Bouskela E. Effects of Resistance Training on Cytokines. Int J Sports Med. 2010;31(31):441–50.
    OpenUrlCrossRefPubMed
  38. 38.↵
    Geekiyanage H, Rayatpisheh S, Wohlschlegel JA, Brown R, Ambros V. Extracellular microRNAs in human circulation are associated with miRISC complexes that are accessible to anti-AGO2 antibody and can bind target mimic oligonucleotides. Proceedings of the National Academy of Sciences. 2020;117(117):24213–23.
    OpenUrlAbstract/FREE Full Text
  39. 39.↵
    Safa A, Bahroudi Z, Shoorei H, Majidpoor J, Abak A, Taheri M, et al. miR-1: A comprehensive review of its role in normal development and diverse disorders. Biomedicine & Pharmacotherapy. 2020;132:110903.
    OpenUrl
  40. 40.↵
    Horak M, Novak J, Bienertova-Vasku J. Muscle-specific microRNAs in skeletal muscle development. Developmental Biology. 2016;410(410):1–13.
    OpenUrlCrossRefPubMed
  41. 41.↵
    Sun Y, Ge Y, Drnevich J, Zhao Y, Band M, Chen J. Mammalian target of rapamycin regulates miRNA-1 and follistatin in skeletal myogenesis. Journal of Cell Biology. 2010;189(189):1157–69.
    OpenUrlAbstract/FREE Full Text
  42. 42.↵
    Watson K, Baar K. mTOR and the health benefits of exercise. Seminars in Cell & Developmental Biology. 2014;36:130–9.
    OpenUrlCrossRefPubMed
  43. 43.↵
    Yu H, Lu Y, Li Z, Wang Q. microRNA-133: Expression, Function and Therapeutic Potential in Muscle Diseases and Cancer. Current Drug Targets. 2014;15(15):817–28.
    OpenUrlCrossRefPubMed
  44. 44.↵
    Nohata N, Hanazawa T, Enokida H, Seki N. microRNA-1/133a and microRNA-206/133b clusters: Dysregulation and functional roles in human cancers. Oncotarget. 2012;3(1).
  45. 45.↵
    Caporali S, Calabrese C, Minieri M, Pieri M, Tarantino U, Marini M, et al. The miR-133a, TPM4 and TAp63γ Role in Myocyte Differentiation Microfilament Remodelling and Colon Cancer Progression. Int J Mol Sci. 2021;22(22):9818.
    OpenUrl
  46. 46.↵
    Wen P, Song D, Ye H, Wu X, Jiang L, Tang B, et al. Circulating MiR-133a as a Biomarker Predicts Cardiac Hypertrophy in Chronic Hemodialysis Patients. PloS one. 2014;9(9):e103079.
    OpenUrl
  47. 47.↵
    Yan F, Chen Y, Ye X, Zhang F, Wang S, Zhang L, et al. miR-3113-5p, miR-223-3p, miR-133a-3p, and miR-499a-5p are sensitive biomarkers to diagnose sudden cardiac death. Diagnostic Pathology. 2021;16(16):67.
    OpenUrl
  48. 48.↵
    Hrdličková R, Nehyba J, Bargmann W, Bose HR, Jr.. Multiple Tumor Suppressor microRNAs Regulate Telomerase and TCF7, an Important Transcriptional Regulator of the Wnt Pathway. PloS one. 2014;9(9):e86990.
    OpenUrlCrossRefPubMed
  49. 49.↵
    Iannone F, Montesanto A, Cione E, Crocco P, Caroleo MC, Dato S, et al. Expression Patterns of Muscle-Specific miR-133b and miR-206 Correlate with Nutritional Status and Sarcopenia. Nutrients. 2020;12(12):297.
    OpenUrl
  50. 50.↵
    Xu H, Zhao G, Zhang Y, Jiang H, Wang W, Zhao D, et al. Mesenchymal stem cell-derived exosomal microRNA-133b suppresses glioma progression via Wnt/β-catenin signaling pathway by targeting EZH2. Stem Cell Research & Therapy. 2019;10(10):381.
    OpenUrl
  51. 51.↵
    Koutsoulidou A, Mastroyiannopoulos NP, Furling D, Uney JB, Phylactou LA. Expression of miR-1, miR-133a, miR-133b and miR-206 increases during development of human skeletal muscle. BMC Developmental Biology. 2011;11(11):34.
    OpenUrl
  52. 52.↵
    Chen J-F, Mandel EM, Thomson JM, Wu Q, Callis TE, Hammond SM, et al. The role of microRNA-1 and microRNA-133 in skeletal muscle proliferation and differentiation. Nature genetics. 2006;38(38):228–33.
    OpenUrlCrossRefPubMedWeb of Science
  53. 53.↵
    Friedman RC, Farh KK, Burge CB, Bartel DP. Most mammalian mRNAs are conserved targets of microRNAs. Genome Res. 2009;19(19):92–105.
    OpenUrlAbstract/FREE Full Text
  54. 54.↵
    Selbach M, Schwanhäusser B, Thierfelder N, Fang Z, Khanin R, Rajewsky N. Widespread changes in protein synthesis induced by microRNAs. Nature. 2008;455(455):58–63.
    OpenUrlCrossRefPubMedWeb of Science
  55. 55.↵
    Zhang F, Wang D. The Pattern of microRNA Binding Site Distribution. Genes. 2017;8(8):296.
    OpenUrl
  56. 56.↵
    Baggish AL, Hale A, Weiner RB, Lewis GD, Systrom D, Wang F, et al. Dynamic regulation of circulating microRNA during acute exhaustive exercise and sustained aerobic exercise training: Circulating microRNA in exercise. The Journal of Physiology. 2011;589(589):3983–94.
    OpenUrlCrossRefPubMed
  57. 57.↵
    Banzet S, Chennaoui M, Girard O, Racinais S, Drogou C, Chalabi H, et al. Changes in circulating microRNAs levels with exercise modality. Journal of Applied Physiology. 2013;115(115):1237–44.
    OpenUrlCrossRefPubMed
  58. 58.↵
    Héroux ME, Butler AA, Cashin AG, McCaughey EJ, Affleck AJ, Green MA, et al. Quality Output Checklist and Content Assessment (QuOCCA): a new tool for assessing research quality and reproducibility. BMJ Open. 2022;12(12):e060976.
    OpenUrlAbstract/FREE Full Text
  59. 59.↵
    Munafò MR, Chambers C, Collins A, Fortunato L, Macleod M. The reproducibility debate is an opportunity, not a crisis. BMC Res Notes. 2022;15(15):43.
    OpenUrl
  60. 60.↵
    Munafò MR, Nosek BA, Bishop DVM, Button KS, Chambers CD, Percie du Sert N, et al. A manifesto for reproducible science. Nature Human Behaviour. 2017;1(1):0021.
    OpenUrl
  61. 61.↵
    Kok MGM, de Ronde MWJ, Moerland PD, Ruijter JM, Creemers EE, Pinto-Sietsma SJ. Small sample sizes in high-throughput miRNA screens: A common pitfall for the identification of miRNA biomarkers. Biomol Detect Quantif. 2018;15:1–5.
    OpenUrlPubMed
  62. 62.↵
    de Ronde MWJ, Ruijter JM, Moerland PD, Creemers EE, Pinto-Sietsma SJ. Study Design and qPCR Data Analysis Guidelines for Reliable Circulating miRNA Biomarker Experiments: A Review. Clin Chem. 2018;64(64):1308–18.
    OpenUrlAbstract/FREE Full Text
  63. 63.↵
    Wetterslev J, Thorlund K, Brok J, Gluud C. Estimating required information size by quantifying diversity in random-effects model meta-analyses. BMC Medical Research Methodology. 2009;9(9):86.
    OpenUrl
  64. 64.↵
    Nielsen S, Akerstrom T, Rinnov A, Yfanti C, Scheele C, Pedersen BK, et al. The miRNA plasma signature in response to acute aerobic exercise and endurance training. PloS one. 2014;9(9):e87308.
    OpenUrlCrossRefPubMed
  65. 65.↵
    Listowski MA, Heger E, Bogusławska DM, Machnicka B, Kuliczkowski K, Leluk J, et al. microRNAs: fine tuning of erythropoiesis. Cell Mol Biol Lett. 2013;18(18):34–46.
    OpenUrl
  66. 66.↵
    Garcia-Sifuentes Y, Maney DL. Reporting and misreporting of sex differences in the biological sciences. eLife. 2021;10:e70817.
    OpenUrlCrossRef
  67. 67.↵
    Cowley ES, Olenick AA, McNulty KL, Ross EZ. “Invisible Sportswomen”: The Sex Data Gap in Sport and Exercise Science Research. Women in Sport and Physical Activity Journal. 2021:1–6.
  68. 68.↵
    Lamon S, Knowles O. Why are males still the default subjects in medical research? The Conversation [Internet]. 2021 13.06.2022. Available from: https://theconversation.com/why-are-males-still-the-default-subjects-in-medical-research-167545.
  69. 69.↵
    Landen S, Hiam D, Voisin S, Jacques M, Lamon S, Eynon N. Physiological and molecular sex differences in human skeletal muscle in response to exercise training. J Physiol. 2021.
  70. 70.↵
    Landen S, Voisin S, Craig JM, McGee SL, Lamon S, Eynon N. Genetic and epigenetic sex-specific adaptations to endurance exercise. Epigenetics. 2019;14(14):523–35.
    OpenUrlPubMed
  71. 71.↵
    Oliva M, Muñoz-Aguirre M, Kim-Hellmuth S, Wucher V, Gewirtz ADH, Cotter DJ, et al. The impact of sex on gene expression across human tissues. Science. 2020;369(6509).
  72. 72.↵
    Shen JJ, Wang YF, Yang W. Sex-Interacting mRNA- and miRNA-eQTLs and Their Implications in Gene Expression Regulation and Disease. Frontiers in genetics. 2019;10:313.
    OpenUrl
  73. 73.
    Song R, Ro S, Michaels JD, Park C, McCarrey JR, Yan W. Many X-linked microRNAs escape meiotic sex chromosome inactivation. Nature genetics. 2009;41(41):488–93.
    OpenUrlCrossRefPubMedWeb of Science
  74. 74.↵
    Guo L, Zhang Q, Ma X, Wang J, Liang T. miRNA and mRNA expression analysis reveals potential sex-biased miRNA expression. Sci Rep. 2017;7:39812.
    OpenUrl
  75. 75.↵
    Murri M, Insenser M, Fernández-Durán E, San-Millán JL, Luque-Ramírez M, Escobar-Morreale HF. Non-targeted profiling of circulating microRNAs in women with polycystic ovary syndrome (PCOS): effects of obesity and sex hormones. Metabolism: clinical and experimental. 2018;86:49–60.
    OpenUrl
  76. 76.↵
    Lamon S, Le Carré J, Petito G, Duong HP, Luthi F, Hiam D, et al. The effect of the menstrual cycle on the circulating microRNA pool in human plasma: a pilot study. Human reproduction (Oxford, England). 2022.
Back to top
Posted December 21, 2022.
Download PDF

Supplementary Material

Data/Code
Email
Response of circulating miRNAs to acute exercise: A systematic Review and Meta-Analysis
Kirstin MacGregor, Colin N Moran, Sophie Broome, Patrick J Owen, Séverine Lamon, Danielle Hiam
medRxiv 2022.12.20.22283756; doi: https://doi.org/10.1101/2022.12.20.22283756
Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools

Subject Area