Exercise induces new cardiomyocyte generation in the adult mammalian heart

Exercise enhances ¹⁵N-thymidine-positive cells in the heart

We subjected young adult C57Bl/6 mice (2 months old) to 8 weeks of monitored voluntary wheel running (averaging 5.57 ± 0.63 km/day, mean ± s.e.m.) with continuous infusion of ¹⁵N-thymidine. Consistent with previous studies^17,20, exercised mice demonstrated physiologic cardiac remodeling characterized by increased cardiac mass (Supplementary Fig. 1a), augmented echocardiographic (Supplementary Fig. 1b), and cellular measures of hypertrophy (Supplementary Fig. 1c, d). Messenger RNA analysis also confirmed an increased alpha/beta myosin heavy chain (MYH6/MYH7) ratio consistent with physiological hypertrophy (Supplementary Fig. 1e). No signs of systolic dysfunction or increased apoptosis were detected in the hearts of exercised mice (Supplementary Fig. 2a, b). As previously described^2,3, we leveraged the sub-organelle resolution of MIMS, that allowed detection of characteristic ultrastructures, such as sarcomeres, to quantify incorporation of ¹⁵N-thymidine within cardiomyocytes (Fig. 1a). We analyzed 1209 cardiomyocytes from four sedentary mice compared with 1130 cardiomyocytes from four exercised mice. Cardiomyocyte nuclei that underwent DNA synthesis during the labeling period were identified by ¹⁵N labeling. These cells displayed a stereotypical pattern of peripheral heterochromatin and a distinct nucleolus (Fig. 1a). We observed a low frequency of cardiomyocyte DNA synthesis in sedentary control mice (1.24% = 0.02% per day), consistent with our previous results³. However, exercised mice had a significantly higher fraction of ¹⁵N-thymidine labeled cardiomyocytes (3.62% = 0.06% per day, p = 0.0003, Fisher’s exact test, odds ratio (OR) = 2.997, confidence interval (CI) 1.65–5.45) (Fig. 1b, c). In conjunction with the increase in cardiomyocyte DNA synthesis, we also detected an increase in ¹⁵N-thymidine labeled non-cardiomyocytes (non-CMs) in exercised hearts compared to sedentary controls (Supplementary Fig. 2c). To determine if this increase in ¹⁵N-thymidine labeled non-CMs was accompanied by an exercise-induced neo-angiogenic response, we analyzed the capillary density by immunohistochemistry and found an increased number of capillaries per cardiomyocyte following exercise training (4.14 ± 0.07 vs. 3.75 ± 0.08 capillaries/cardiomyocyte in exercised vs. sedentary controls, mean ± s.e.m., p < 0.05, two-sided t test) (Supplementary Fig. 2d). These findings are consistent with exercise-induced angiogenesis, which has been previously described^21,22.

Exercise induces cardiomyogenesis

To account for the polyploidy and multinucleation that can be seen in cardiomyocytes^1,3, particularly during hypertrophic growth²³, we performed cardiomyocyte tracing on serial periodic acid Schiff (PAS)-stained sections in both directions to the ¹⁵N-thymidine-labeled cardiomyocyte nuclei to define the number of nuclei contained in each cell (Fig. 2a). We then utilized in situ hybridization targeting the Y-chromosome in adjacent sections to assess ploidy of each ¹⁵N-thymidine labeled cardiomyocyte nucleus (Fig. 2a). Although we identified polyploid (≥4n) and binucleated ¹⁵N-thymidine cardiomyocytes (Fig. 2b), we also observed a higher frequency of diploid/mononucleated ¹⁵N-thymidine-labeled cells in exercised hearts relative to those from sedentary mice (Fig. 2c), consistent with an exercise-mediated increase in cardiomyogenesis (1.15% vs. 0.25%, p = 0.01, Fisher’s exact test, OR = 4.695, Cl 1.44–15.53) (Fig. 2c). In line with previous findings²⁴, mononucleated ¹⁵N-thymidine-labeled cardiomyocytes were significantly smaller than their binucleated counterparts in the same hearts (Fig. 2d). Although rare, we also found ¹⁵N-thymidine labeled cardiomyocytes with a disturbed sarcomere structure and two adjacent nuclei that appeared to have completed cytokinesis at the time of tissue harvest (Supplementary Fig. 2e). Taken together, these results demonstrate that regular exercise over 8 weeks increased cardiomyogenesis. The numbers translate into a 4.6-fold increase in cardiomyogenesis in exercised animals (0.25% per 8 weeks = 0.0045% per day in sedentary vs. 1.15% per 8 weeks = 0.021% per day in exercised animals). These data suggest that the exercised heart is capable of generating new cardiomyocytes at a projected annual rate of 7.5% vs. 1.63% in sedentary conditions.

Exercise after MI increases the area of proliferation

Exercise mitigates adverse remodeling when initiated after myocardial infarction (MI) in animal models^25–27 and humans²⁸, attenuating fibrosis, dilatation, and cardiac dysfunction, although the mechanisms are incompletely understood. We sought to determine whether exercise-induced cardiomyogenesis could contribute to the benefits of exercise after ischemic injury. Young adult mice underwent experimental MI by ligation of the left anterior descending artery followed by voluntary running or sedentary activity, started 24 h after ischemic injury with continuous labeling with ¹⁵N-thymidine for 8 weeks during the exercise regimen. We have previously shown that cardiomyocyte cell cycle activity increases in the peri-infarct region³. In this study, we confirmed this increase in the peri-infarct region in ¹⁵N-thymidine labeled cardiomyocytes in both sedentary (22.8%) and exercised (20.4%) mice, consistent with our previous data, but with no significant difference between the two groups (Fig. 3a–c). Distinct from the peri-infarct region, exercise led to a substantially higher proportion of ¹⁵N-thymidine labeled cardiomyocytes over 8 weeks in the extended border zone region (>400 μm from infarct) (19.1% ¹⁵N-thymidine labeled cardiomyocytes in exercised vs. 5.3% in sedentary mice, p < 0.0001, Fisher’s exact test) (Fig. 3d). When examining the nucleation and ploidy status of ¹⁵N-thymidine labeled cardiomyocytes in this extended border region, we found a higher frequency of diploid/mononucleated ¹⁵N-thymidine-labeled cells in hearts from exercised compared with those from sedentary mice (Fig. 3e), consistent with an exercise-mediated increase in cardiomyogenesis (0.4% vs. 2.7% of total cardiomyocytes, p = 0.004, Fisher’s exact test, OR = 6.931, CI 1.87–30.83) (Fig. 3f). The higher frequency of proliferating cells in the remote area, has also been observed after MI in mice exposed to hypoxia²⁹. Transgenic activation of cardiomyocyte cell cycle activity in animals improves cardiac function after infarction³⁰. Thus, the exercise-induced cardiomyocyte cell cycle activity seen after MI would be expected to favorably influence cardiac remodeling and likely contributes to the well-documented benefits of exercise in this setting^25–28.

Inhibition of miR-222 blocks exercise-induced cardiac growth

To explore the mechanism of the cardiomyogenic exercise response, we studied the role of miR-222, a microRNA (miRNA) that increases in response to exercise in both animal models and humans, and plays an important role in the cardiovascular effects of exercise^18,31. Although our prior work had suggested a role for miR-222 in cardiomyocyte proliferation¹⁸, these studies were based on work in neonatal rather than adult cardiomyocytes, and proliferation markers that could not unambiguously identify mitotic events in vivo¹⁸. First, we confirmed that voluntary wheel running induced miR-222 expression in the adult mouse heart (Fig. 4a). Next, we inhibited miR-222 upregulation by weekly injection of a sequence-specific locked nucleic acid inhibitor (LNA)-anti-miR-222, together with ¹⁵N-thymidine administration during the 8-week voluntary exercise protocol. Exercised mice injected with scrambled LNA-anti-miR (LNA-Ctr) and sedentary mice treated with LNA-anti-miR-222 served as controls. After 8 weeks of voluntary wheel running, LNA-Ctr treated mice developed physiologic cardiac hypertrophy (Fig. 4b). The cardiac hypertrophic response was attenuated in exercised mice treated with LNA-anti-miR-222, confirming that miR-222 is necessary for exercise-induced cardiac growth even over longer duration (8 weeks) than previously examined¹⁸. To determine whether miR-222 also mediates exercise-induced cardiomyogenesis, we compared the frequency of ¹⁵N-thymidine labeled cardiomyocytes in exercised hearts after treatment with either LNA-anti-miR-222 or LNA-Ctr (Fig. 4c, d). Interestingly, miR-222 inhibition fully prevented the increase in ¹⁵N-thymidine incorporation in response to exercise (Fig. 4c, d). miR-222 inhibition also led to a more modest induction in exercise-induced ¹⁵N-thymidine-positive non-CMs (Supplementary Fig. 3).

To further explore the mechanisms by which miR-222 regulates exercise-induced cardiomyogenesis, we performed quantitative gene expression analysis from sedentary and exercised mouse hearts for relevant genes previously validated as direct targets of miR-222¹⁸. Our data demonstrate that after 8 weeks of endurance exercise, HIPK1 gene expression remained significantly inhibited, without a significant change in p27 or HIPK2 expression (Fig. 4e). Moreover, HIPK1 expression increased significantly with miR-222 inhibition (Fig. 4f) (which abolished exercise-induced cardiomyogenesis) while p27 and HIPK2 expression did not. Taken together with our previously published data demonstrating that HIPK1 is a direct target of miR-222 with anti-proliferative effects in cardiomyocytes¹⁸, these data strongly suggest that HIPK1 contributes to miR-222’s modulation of exercise-induced cardiomyogenesis.

We further investigated other pathways recently linked to cardiomyocyte proliferation. For example, we examined a panel of hypoxia-induced genes^32,33, and found an increase in Slc2a (GLUT1) and LDHA expression after 8 weeks of exercise, pointing to a significant metabolic adjustment potentially due to increased Hif1a activity (Fig. 5a). However, these changes were not significantly affected by miR-222 inhibition suggesting they do not contribute to miR-222’s modulation of cardiomyogenesis (Fig. 5b). In contrast, we did find that miR-222 inhibition increased expression of CCNG2 (cyclin G2), a relevant cell cycle regulator, but could not find CCNG2 as a predicted miR-222 target (through Pictar and Targetscan). We went on to specifically check CCNG2 expression in miR-222 overexpressing neonatal rat ventricular cardiomyocytes (NRVMs), and found an increase in CCNG2 expression, strongly suggesting that this is not a direct miR-222 target (Fig. 5c).

Cell cycle regulators that are reportedly downstream effectors of the Hippo pathway, another pathway implicated in cardiomyocyte proliferation, were increased in exercised hearts. While CDK1, CCNB1, and TEAD2, for example, were all upregulated with exercise (Fig. 5d), miR-222 inhibition caused further increase of those Hippo/cell cycle regulators (Fig. 5e). We subsequently ran a pathway analysis (Ingenuity Pathway Analysis, IPA, Qiagen) from a microarray comparing NRVM with precursor-miRNA-mediated miR-222 overexpression to controls (pre-miR-222 and Ctr-pre) and found cell cycle regulators in general, and with subsequent analysis (Fig. 5f), TEAD2 specifically, to be upregulated with miR-222 expression, suggesting that they are not direct targets of miR-222’s effects in this context. Further exploration of the role of the Hippo pathway in endurance exercise although apparently independent of miR-222 may be of interest for future studies.

Mice

All mice were maintained and studied using protocols in accordance with the Guide for the Use and Care of Laboratory Animals and approved by Harvard University (protocol number 16-05-273) and Massachusetts General Hospital (protocol number 2015 N000029) Animal Care and Use Committees. Two-month-old male C57Bl/6 mice were obtained from Charles River.

Running exercise protocol and ¹⁵N-thymidine labeling

Starting at an age of 2 months, osmotic minipumps (Alzet) were implanted subcutaneously, delivering ¹⁵N-thymidine (Cambridge Isotopes) at a rate of 20 μg h⁻¹ and exchanged weekly for 8 weeks. Mice were individually housed in plexiglass cages (36 L × 20 W × 15 H cm) that contained a stainless-steel running wheel (diameter 11.4 cm; Mini-Mitter, Starr Life Science, USA) equipped with a tachometer. Mice ran voluntarily. Mouse activity and cardiac hypertrophy response was controlled for by recording daily running and morphometric analysis. The sedentary control mice were kept in the same cage system lacking running wheels.

Experimental myocardial infarction

Mice were subjected to experimental MI. Surgeries were performed by a single operator with >20 years of experience in the performance of coronary ligation in rodents. In brief, the animals were anesthetized with isoflurane, intubated, and ventilated. After the thoracic cavity was opened, the left anterior descending coronary artery was permanently ligated approximately 2 mm below the left atrial appendage. The chest was closed and an osmotic minipump (Alzet) administrating ¹⁵N-thymidine (Cambridge Isotope Laboratories) at a rate of 20 μg h⁻¹, was implanted subcutaneously and the animal was placed in recovery. The pump was exchanged weekly for 8 weeks.

miR-222 inhibition

LNA-anti-miR injections were performed as previously described¹⁸. Two-month-old C57Bl/6 male mice were subcutaneously injected with 10 mg kg⁻¹ of LNA-modified anti-miR-222 (LNA-anti-miR-222) or scrambled control (LNA-Ctr) reconstituted in saline for three consecutive days after osmotic pump implantation and then weekly after pump exchanges throughout the experiment. LNA-anti-miR oligonucleotides were purchased from Exicon.

Echocardiographic studies

Echocardiography was performed on conscious mice by using a GE Vivid E90 with a L8-18i-D probe. Parasternal long-axis views, short-axis views, and two-dimensional-guided M-mode images of short axis at the papillary muscle level were recorded. The average of at least three measurements was used for every data point from each mouse.

Sample preparation and MIMS data acquisition and analysis

Tissues were fixed with 4% paraformaldehyde (PFA), embedded in LR (London resin) white, sectioned (0.5–1 μm), and mounted on silicon chips. MIMS analyses were conducted with ion microscopy (NanoSIMS 50 and a large-radius NanoSIMS 50L, Cameca) as described previously^2,3. Briefly, a focused Cs⁺ ion beam (size <100 nm) is rastered over the sample surface. The impact of the Cs⁺ ion beam on the sample surface generates the sputtering of multiple secondary ions from the probed nanovolume (corresponding to a pixel in the image). For each nanovolume, the secondary-ion intensities for ¹²C⁻, ¹²C¹⁴N⁻, ¹²C¹⁵N⁻, and ³¹P⁻ were recorded in parallel. The detection of nitrogen requires the use of poly-atomic ions ¹²C¹⁴N⁻ and ¹²C¹⁵N⁻ as a proxy for ¹⁴N and ¹⁵N. A hue-saturation-intensity transformation of the ¹²C¹⁵N⁻/¹²C¹⁴N⁻ ratio image provides a visual representation of regions where the ratio is above natural abundance (natural ¹⁵N/¹⁴N 0.37%), indicative of ¹⁵N-thymidine labeling. Observers blinded to group assignment performed cardiomyocyte counts as well as the following analyses.

Nucleation and cell volume analysis

Serial sections (0.5–1 μm) were cut from LR white embedded heart to cover up to 100 μm of tissue below and above the section that was mounted on the chip. Sections were stained with a modified PAS staining protocol to identify individual cardiomyocytes. Slides were immersed for 2 × 30 min in xylene and rehydrated through graded alcohols, incubated in Periodic acid solution overnight and then in Schiff’s reagent for two nights, washed, and dehydrated and cleared before mounting. Slides were imaged on a NanoZoomer Whole Slide Scanner (Hamamatsu Photonics). Magnified PAS image in Fig. 3a was taken on Airy Scan LSM800 with Axiocam color camera (Zeiss). To analyze nucleation, ¹⁵N-thymidine-positive cardiomyocytes were tracked by locating the cardiomyocyte in every section of the serial above and below the level of the nucleus for as long as it was present. The total number of nuclei was counted for each cell. PAS-stained sections were also used to confirm cells labeled as cardiomyocytes and non-CMs in all cohorts and adjustments have been made for mislabeled cells if necessary. Outlined ¹⁵N-positive cells were also used to calculate total cellular volume.

Fluorescent in situ hybridization

Sections were incubated in 1 M sodium thiocyanate for 10 min at 80 °C, washed in PBS, and treated for 2 min in 0.2% Triton X (in PBS), digested with Proteinase K (50 μg ml⁻¹) for 30 min at 56 °C. Sections were rinsed in 1× PBS, post-fixed in 4% PFA/PBS for 5 min and treated for 10 min with 50% formamide/4× standard saline citrate buffer at 37 °C. Slides were air-dried and biotinylated-labeled chromosome Y paint (IDMF 1057, Empire Genomics) was suspended in hybridization mix, applied to sections, and sealed under glass with rubber cement. Samples were denatured at 69 °C for 3 min. After overnight incubation at 42 °C, slides were washed three times with 2× standard saline citrate buffer, two 10 min washes each with PBS containing 0.1% Tween (PBST), all at room temperature. Samples were blocked with PBST/10% goat serum and incubated for 1 h in streptavidin-conjugated Alexa Fluor 488 (S32354, Invitrogen) before being washed and mounted. Sections were imaged on a LSM 510 Inverted Confocal (Zeiss LSM 510) and analyzed with Zen lite software (Zeiss).

Capillary density and cross-sectional area quantification

Hearts from animals that underwent the above described exercise regimen but that were not used for LR white embedding and MIMS were snap frozen in OCT  (optimal cutting temperature) as previously described¹⁸. Thick sections of 6 μm were cut from the frozen tissue and placed on a glass cover slip, air-dried for 5 min, followed by a brief 10 min fixation with fresh 4% PFA. Sections were immunofluorescently stained with both CD31 (BD Pharmingen) and wheat germ agglutinin (488-oregon green, Thermo Fisher Scientific) to analyze capillary density (capillaries/cardiomyocyte) and cross-sectional area using standard procedures. AlexaFluor (555-texas red) fluorescent dye (Thermo Fisher Scientific) was used as secondary antibody. Images were taken on a Leica SP8 confocal microscope.

Cell death assay by TUNEL staining

Slides were rehydrated through xylene and graded concentrations of ethanol and stained for cell death using the TMR (tetramethylrhodamine)-red TUNEL (terminal deoxynucleotidyl transferase dUTP nick end labelled) Kit (in situ cell death Detection Kit-TMR Red; Roche, Basel, Switzerland) as per manufacturers instruction. Briefly, slides were incubated for 30 min at 37 °C in 20 μg ml⁻¹ proteinase K, washed twice in 1× PBS 5 min at room temperature and incubated at 37 °C in a red fluorescent (TMR-red) TUNEL cell death detection reagent. Cardiomyocytes were identified by simultaneous immunostaining cardiac troponin T (1:200 dilution, ab8295, Abcam, Cambridge, MA). After washing with 1× PBST (1× PBS added with 0.1% Tween), slides were counterstained with DAPI (4′, 6-diamidino-2′-phenylindol, dichloride; Thermo Scientific/Pierce, Rockford, IL, USA) to visualize the nuclei. Finally, slides were washed and mounted and all quantitative analyses were performed. All images were post-processed in ImageJ (NIH) software. Approximately 14,000 troponin-T-positive cardiomyocytes were counted for the TUNEL analysis. LR = London Resin OCT = optimal cutting temperature TMR = Tetramethylrhodamine TUNEL = Terminal deoxynucleotidyl transferase dUTP nick end labelled

RNA isolation and quantitative real-time PCR

RNA was isolated from both tissue and cell samples using Trizol (Qiagen), and phase separation followed by column purification (Zymo Research). RNA of 1 μg was added to reverse transcription reactions (Applied Bioscience) and quantitative PCR was carried out using SYBR-green (Bio-Rad).

Cardiomyocyte isolation and culture and gene expression

Primary NRVMs were prepared from 1- to 2-day-old rats by use of the neonatal cardiomyocyte isolation system (Worthington Biochemical Corp.). Isolated NRVMs were purified by Percoll gradient centrifugation and plated in 60 mm dishes at 1 × 10⁶ cells per well and cultured in DMEM/5% FBS/10% horse serum for 24 h. Before treatment, NRVM were synchronized, and cultured in 0.2% FBS DMEM media. miRNA-222 as well as control precursors (pre-miR) were purchased from Invitrogen. Transfection of pre-miR was carried out using Lipofectamine RNAiMax (Invitrogen) according to the manufacturer’s protocol. Forty-eight hours after transfection, total RNA was isolated from treated NRVM with Trizol (Invitrogen) according to standard protocols and submitted to the Dana-Farber Cancer Institute Molecular Diagnostics Laboratory for microarray assay using Affymetrix Rat Genome 230 v2.0. Microarray data analysis was performed with Affymetrix Microarray Data Analysis tools.

Statistical analysis

Statistical testing was performed using Prism 3.0/7.0 (Graphpad). For echocardiographic and morphometric analyses, as well as gene expression, cross-sectional area, capillary density, and cell death/volume, results are presented as mean ± s.e.m. from n = 3–5 animals per group and were compared by using two-sided t tests (significance level p < 0.05). More than two groups were compared by using one-way analysis of variance with Tukey’s correction for multiple comparisons. Data comparing event rates, i.e., ¹⁵N-thymidine-positive and -negative cardiomyocytes and non-CMs, were compared using Fisher’s exact test based on the total number of cells counted as unit of analysis (significance level p < 0.05) and are presented as percentage of total cell count as well as in a table. Three to four mice per group were used to ensure sufficient cell counts.

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request. Microarray data have been deposited in the Gene Expression Omnibus (GEO accession number: GSE59641).

Competing interests

The authors declare no competing interests.