In this prospective cohort study, consecutive HFrEF patients that were referred for a 15-week supervised combined strength and moderate-intensity aerobic training program to the Cardiac Rehabilitation Centre of the Antwerp University Hospital (ET group) were compared to age-matched HFrEF patients receiving usual care without exercise training (UC group). Randomization into a training and non-training group was considered as non-ethical in view of the strong indication for exercise training in HFrEF (Class IA indication) (Ponikowski et al., 2016). Patients were included when they completed at least 30 of the 45 sessions. The study complied with the Declaration of Helsinki and was approved by the ethics committee of the Antwerp University Hospital. Written informed consent was obtained from all participants.

The change in miRNA levels after a 15-week exercise training program was investigated in the ET group and compared the UC group, and baseline miRNA levels were related to the change in VO₂peak. In the ET group only, the relation between baseline miRNA levels and change in cardiopulmonary exercise test (CPET), cardiac and vascular adaptation, and muscle strength was studied, and the effect of an acute exercise bout on the miRNA panel was assessed (Figure 1).

Supervised in-hospital exercise training consisted of combined aerobic and resistance training, 3 sessions/week (58 min/session) for 15 weeks. Aerobic training intensity was set at 90% of heart rate (HR) at the respiratory compensation point (RCP). When RCP was not reached, exercise intensity was calculated using the Karvonen formula [exercise heart rate = rest heart rate + (0.70^∗ heart rate reserve)] (Karvonen et al., 1957). Strength exercise was an important component of the training program, with the focus primarily on gaining strength during the first 8 weeks. Afterward, aerobic training became more prominent (Supplementary Figure 1).

CPET was performed on a treadmill (Medical Jaeger, Würzburg, Germany) with a graded protocol (equivalents of 40 W + 20 W/min or 20 W + 10 W/min) (Beckers et al., 2011), with an identical protocol for the follow-up test (Cardiovit CS-200 Ergo-Spiro, Schiller AG, Baar, Switzerland). Gas exchange measurements and 12-lead electrocardiogram were recorded continuously. Blood pressure was measured every minute. VO₂peak was determined as the mean VO₂peak during the final 30 s of exercise. Percent predicted VO₂peak was calculated using the Jones equation (Jones and Campbell, 1982). The RCP was estimated from the systematic increased ventilatory equivalent for VCO₂ (VE/VCO₂) and the systematic decrease in end tidal partial pressure of CO₂ (PETCO₂) (Whipp et al., 1989; Algul et al., 2017).

Echocardiography was performed on a Vivid E95 cardiac ultrasound using the 4V transducer for 3-D imaging and analyzed on Tomtec Arena). Left ventricular ejection fraction (LVEF), left ventricular mass index (LVMi), left ventricular end diastolic volume (LVEDV), left atrial volume index (LAVi), interventricular septum (IVS) thickness and diastolic parameters (E/A, E/e’) were recorded. In the UC group, LVEF was calculated using Simpson’s monoplane (4 chamber view) method on M5S transducer, AGFA IMPAX Agility 8.1.2, Vivid E95.

Endothelial-dependent vasodilation of the brachial artery was evaluated by flow mediated dilation (FMD) as previously described (ProSound alfa6, Hitachi-Aloka Medical Ltd.) (Van Craenenbroeck et al., 2015c; Mannaerts et al., 2019). FMD was expressed as the percent change in peak vessel diameter from the baseline value [(peak diameter − baseline diameter)/baseline diameter]. Endothelial-independent dilation was calculated accordingly after sublingual administration of nitroglycerine. Arterial stiffness was assessed with carotid-femoral pulse wave velocity (PWV) and pulse wave analysis (PWA) that calculates augmentation index (AIx) and heart rate corrected AIx (AIx75) using SphygmoCor (Atcor Medical), as previously described (Van Craenenbroeck et al., 2015b). All measurements were done in triplicate.

Bioelectrical impedance analysis was performed on an Omron BF306 Body Fat Monitor (Omron Healthcare Co., Ltd., Kyoto) using 2 electrodes (1 handle in each hand) to provide estimates of total lean mass and fat mass.

Whole blood was collected after an overnight fast prior to the CPET in ethylenediaminetetraacetic acid tubes (EDTA). The first 3 ml of blood was discarded to prevent contamination with skin epithelial cells and endothelial cells. Samples were centrifuged within 30 min after collection (1,500 g, 15 min) at room temperature, and plasma was stored at −80°C.

MiR-23a, miR-140, miR-146a, miR-191, and miR-210 were quantified in plasma samples using miRNA RT-qPCR. In brief, plasma samples were thawed on ice and centrifuged for 10 min (4°C, 16,000 g). RNA enriched for small RNAs (including miRNAs) was isolated using the mirVana Paris Kit (Thermo Fisher Scientific). Four hundred microliter 2X Denaturing Solution was added to 400 μl of plasma. RNA was extracted using acid-phenol:chloroform and ethanol. The aliquoted eluate was immediately stored at −20°C. Reverse transcription and preamplification were performed using TaqMan miRNA primers (Thermo Fisher Scientific) and multiplex qPCR was done in a CFX96 thermal cycler (BioRad) as previously described (Van Craenenbroeck et al., 2015a). Raw Cq values were calculated in BioRad CFX manager software v.3.1 using automatic baseline and threshold settings. Cq values that were undetermined or > 35 were removed from the analysis, to minimize statistical confounding by high quantification cycle values. Data were normalized using geNorm and relative miRNA levels were expressed as log(2^–ΔCq*10⁴) (Gevaert et al., 2018).

Data were analyzed using SPSS 26.0 and R version 3.6.0.

Normality of continuous variables was evaluated using Shapiro-Wilk test. Normally distributed data are expressed as mean ± standard deviation (SD), skewed variables as median and range (1st–3rd quartile). Fisher-exact test was used for comparison of categorical variables, independent samples T-test or Mann-Whitney U-test for comparison of continuous variables. To assess changes with 15 weeks of ET or with acute exercise, linear mixed models were fitted using time and group or visit as fixed effects and patient ID as random effect, or paired samples T-test was used as appropriate.

Correlations were assessed using Pearson correlation analysis. Multiple linear regression analyses adjusting for age and baseline LVEF were performed to assess independent determinants of VO₂peak. A two-sided p-value < 0.05 was considered significant.

Twenty-five patients were included in the ET group and 21 patients in the UC group. Baseline patient demographics, clinical, pharmacological, CPET characteristics, and circulating miRNA levels are shown in Table 1.

At baseline, ET and UC were similar with regard to demographics and clinical characteristics, except for BMI, which was higher in UC (p = 0.042). Ischemic cardiomyopathy was more common in ET compared to UC (p = 0.033), and implantable cardioverter defibrillator (ICD) was less common in ET compared to UC (p = 0.022). Pharmacological therapy was comparable between ET and UC. CPET characteristics were similar between groups, except for work economy, which was lower in ET compared to UC group (6.4 vs. 7.3, p = 0.015).

Baseline miRNA expression was similar between groups, except for miR-23a which was higher in patients referred for ET compared to CG (p = 0.043).

At baseline, better heart (LVEF) and kidney (creatinine) function were associated with higher VO₂peak (respectively, r = 0.303, p = 0.043, and r = −0.514, p < 0.001, Supplementary Figure 2). Patients with lower LVEF had higher miR-210 levels (r = −0.321, p = 0.032, Supplementary Figure 2) independent from age (β = −9.455, p = 0.035, 95%C.I. −18.192, −0.717). None of the other baseline miRNA levels were related with LVEF. No significant correlation was found between baseline miRNA levels and baseline VO₂peak.

Changes in aerobic capacity and clinical characteristics after 15 weeks of follow-up are shown in Table 2 and Supplementary Figure 3. Change in VO₂peak was significantly different between the ET and UC group (+ 0.95 vs. − 0.64 ml/kg/min (difference 1.59, 95% CI 0.06, 3.12, p = 0.041). NYHA class, peak load and load at RCP significantly improved in ET. Both ET and UC patients performed a maximal exercise test, evidenced by a high respiratory exchange ratio (RER).

After 15 weeks of follow-up, plasma levels of miR-146a significantly decreased in the ET group, whereas in the UC group plasma levels remained unaltered (p interaction < 0.05, Figure 2 thick black lines). A significant different evolution in expression levels of miR-191 was observed in ET compared to UC (decrease vs. increase, p interaction < 0.05), but within group differences did not reach significance (dotted-dashed lines, Figure 2). None of the other miRNAs had a significant different evolution between the groups (p-interaction > 0.05).

A single exercise bout (CPET) resulted in a rapid and significant decrease in miR-191 levels in untrained HFrEF patients (p = 0.043). Intriguingly, exercise training resulted in a blunted and even reversed response to acute exercise (Figure 3); a non-significant increase (p = 0.120) after training was observed (p-interaction = 0.003). No significant effect on the other plasma-derived miRNAs was observed, but the same trend of reversal of the miRNA response was observed (except for miR-210).

After 15 weeks of follow-up, VO₂peak significantly changed in ET compared to UC.

In the ET group only, changes in CPET, echocardiographic, muscle strength and vascular function parameters were assessed as secondary characteristics of adaptation to training. Following training, peak load, load at RCP, VO₂ at 50% of peak load during CPET1, BMI, LVEF, LVMi, lean mass and strength characteristics significantly improved (see Table 2).

As previously reported, expression levels of circulating miRNAs change with acute or chronic exercise training (Baggish et al., 2011; Denham and Prestes, 2016). In the present study, we observed a significant decrease in relative expression of miR-23a, miR-140, and miR-146a in the ET group with 15 weeks of training. However, the evolution was only significantly different for miR-146a when compared to the UC group. Our findings are in contrast with Baggish et al. (2011) who observed no change in miR-146a levels with 90 days of rowing training. This difference might be related to the population studied i.c. athletes. To date, evidence on the physiological role of circulating miRNA in the adaptation to exercise is scarce, and to the best of our knowledge, virtually non-existent in the response to training in HFrEF patients. Hence, we can only speculate that the differences in circulating miRNA levels after training that we observed, may result from an underlying active and selective miRNA process that is involved in pathways relevant to exercise adaptation in HFrEF patients, rather than reduced passive release of these miRNAs.

In HFrEF patients, capillary density in skeletal muscle is reduced (Duscha et al., 1999). Both miR-23a and miR-146a were previously shown to stimulate angiogenesis (Zhou et al., 2011; Zhu et al., 2016). Therefore, reduced miR-23a and miR-146a levels after 15 weeks of training may reflect a diminished need for angiogenesis since capillary density increases with endurance and resistance training (Ingjer, 1979; Hudlicka et al., 1992; Holloway et al., 2018). Also, a transient increase in miR-23a and miR-146a may be expected during the training program, reflecting the exercise-induced angiogenesis, but this needs to be explored in future experiments.

Furthermore, HFrEF patients often have skeletal muscle wasting, especially with more advanced disease status, and this contributes to typical HF symptoms and signs such as dyspnea and exercise intolerance, which results in lower VO₂peak and load during CPET (Ponikowski et al., 2016). Exercise training improves skeletal muscle mass and function and has beneficial effects on LVEF and LV remodeling in HFrEF patients (Adams et al., 2017; Tucker et al., 2019). An important driver of skeletal muscle wasting is the ubiquitin-proteasome system (Adams et al., 2021). Both miR-23a and miR-140 were shown to protect against skeletal muscle atrophy through inhibiting the ubiquitin-proteasome pathway and Wnt family member 11 expression, respectively (Wada et al., 2011; Liu et al., 2019). Hence, after training, sufficient skeletal muscle hypertrophy may result in lower miR-23a and miR-140 levels. However, this contrasts the finding that baseline miR-140 was inversely correlated with the change in load at RCP and BMI.

In the present study, we observed a differential expression between ET and UC in miR-23a. This could be attributed to the non-randomized study design, where ET patients might have had more skeletal muscle wasting compared to stable sedentary HFrEF controls, as BMI was significantly lower in ET compared to UC. Unfortunately, we do not have strength characteristics of the UC group. After 15 weeks of combined resistance and aerobic training, BMI significantly increased in the ET group, which could be attributed to increases in skeletal muscle mass, as indicated by higher strength characteristics in ER and coinciding increase in lean mass. However, no correlations with strength characteristics, or between (fold change) miR-23a and percent change in strength or lean mass were observed. Regarding the effect on cardiac hypertrophy, both miR-23a, miR-140, and miR-146a mediate cardiac hypertrophy through targeting the ubiquitin-proteasome pathway, GATA binding protein 4 and dihydrolipoyl succinyltransferase, respectively (Wang et al., 2012; Heggermont et al., 2017; Li et al., 2019). In contrast, we observed an inverse correlation between baseline miR-146a and percent change in LVMi in the ET group.

Baseline miR-210 was inversely related to LVEF. Since miR-210 has been related to hypoxia and upregulates VEGF in endothelial cells (Zheng et al., 2018), the inverse relation with LVEF could reflect the reduced oxygen delivery to the periphery that coincides with worsening LVEF and cardiac output in HFrEF (Piepoli et al., 2010).

In addition, miRNA levels can be altered by acute exercise bouts. Previous research in patients with chronic kidney disease showed a rapid downregulation of circulating miR-146a following an acute exercise bout (Van Craenenbroeck et al., 2015a). In patients with heart failure (average LVEF 47.7%), Xu et al. (2016) observed an increase in circulating miR-21, miR-378, and miR-940 with acute exercise. However, in this study no distinction between heart failure with reduced, preserved or mid-range ejection fraction was made. In healthy athletes, miR-146a and miR-222 were shown to be upregulated by acute exercise both before and after a 90-day rowing training, whereas miR-21 and miR-221 were only upregulated by acute exercise before the training period (Baggish et al., 2011). In the present study, at baseline all miRNA tended to decrease following an acute exercise bout, but this was only significant for miR-191. Intriguingly, this response reversed after 15 weeks of ET, which also suggests a selective training-induced effect on the miRNA expression.

MiR-191 has inhibitory effects on angiogenesis in endothelial cells (Gu et al., 2017; Du et al., 2019) and it stimulates myogenesis (Mitchell et al., 2018). As an acute exercise bout in sedentary patients elicits a hypoxic state, this triggers pro-angiogenic mechanisms. The fact that miR-191 has been shown to inhibit angiogenesis therefore could explain the decreased miR-191 levels observed at baseline. However, this needs to be confirmed in in vitro experiments. Regarding the effect on myogenesis, a single exercise bout provokes acute muscle damage after which myogenesis is established, and therefore lower levels of miR-191. After this initial decrease in myogenesis, we speculate to observe a rise in miR-191 and stimulation of myogenesis to repair the damaged skeletal muscle cells and to increase skeletal muscle hypertrophy. However, we only collected blood samples immediately after CPET so this hypothesis needs to be confirmed. In addition, increased angiogenesis and reduced myogenesis due to lower circulating miR-191 levels following an acute exercise bout may be conflicting. This could be explained by the fact that miRNA are tissue and disease specific, and circulating miRNA levels not always reflect intracellular levels (Pigati et al., 2010).

Finally, we hypothesize that with repeated acute exercise bouts (i.e., the effect of a 15-week training program) in HFrEF patients, the triggers for angiogenesis and myogenesis might have faded out due to increased capillarity and skeletal muscle mass, resulting in the opposite change of miRNA expression levels.

More than half of the HFrEF patients who participate in an ET program may not increase their VO₂peak (Bakker et al., 2018) and despite many efforts, a predictive biomarker for VO₂peak response to training is still lacking. In our previous study, we identified several miRNA that were upregulated in patients with an unfavorable VO₂peak response (Witvrouwen et al., 2021). Among these miRNAs, miR-23a, miR-140, miR-146a, miR-191, and miR-210 were involved in pathways relevant for exercise adaptation processes. In the current study, we observed a significant change in VO₂peak in ET compared to UC; however, the increase within ET was not significant, which could be explained by the fact that BMI significantly increased in ET. Consequently, the observed change in VO₂peak in ml/kg/min is underestimated. Furthermore, we confirmed that baseline miR-23a predicts the change in VO₂peak with training, which may reflect the underlying mechanisms of exercise adaptation since miR-23a was shown to stimulate angiogenesis and to protect against skeletal muscle atrophy (Wada et al., 2011; Zhou et al., 2011). However, we observed clear improvements in muscle strength, but no correlations with miR-23a. This could be attributed to the low sample size. Nevertheless, miRNAs could emerge as promising epigenetic biomarkers of training response.

Whereas aerobic training is known to improve endothelial function in stable coronary artery disease and HFrEF patients (Van Craenenbroeck et al., 2010, 2015c), and both aerobic, resistance and combined aerobic/resistance training showed similar improvements in FMD in patients with hypertension or prehypertension (Pedralli et al., 2020), we did not observe significant improvements in vascular function with 15 weeks of ET. This could be attributed to this subgroup analysis lacking statistical power to draw definitive conclusions.

Furthermore, the study can be biased due to the non-randomized design. However, as stated in the methods, randomizing patients to a training and control group would have been unethical in view of the class IA recommendation of ET in HFrEF patients with favorable effects on morbidity, mortality and quality of life (Ponikowski et al., 2016). As findings of this study are hypothesis generating, they should be validated in larger prospective trials and in in vitro experiments. Future pre-clinical studies could investigate and compare the expression levels in tissue (skeletal muscle, endothelial cells) to the observed changes in plasma levels. Hence, the contribution of miRNA to exercise adaptation processes can be examined, as either miRNA post-transcriptionally influence gene expression or they can be an exercise-induced epiphenomenon in these tissues (f.ex. exercise-induced skeletal muscle hypertrophy results in an increased release of miRNAs in the circulation). This will aid in further unraveling of the underlying mechanisms of response to acute and chronic exercise.