Cardiac diseases are among the main causes of death worldwide. Despite significant advances in the development of cardiac assist devices, the only causal treatment for severe heart failure remains heart transplantation. However, this option is strongly limited by the low number of donor hearts, and, therefore, new therapeutic approaches for regenerating lost myocardium are needed. Recent efforts in the field are directed at identifying proliferation-inducing substances to induce regrowth of the myocardium.¹

Editorial, see p 1012

In This Issue, see p 1009

Meet the First Author, see p 1010

One of the major limitations of this concept are suitable screening assays because adult mammalian cardiomyocytes cannot be cultured for longer time periods without losing their characteristic cell biological and physiological features.² Therefore, preferentially immature embryonic, neonatal, or embryonic stem/induced pluripotent stem cell-derived cardiomyocytes are utilized for in vitro screening purposes.^3,4 However, these systems do not take into account cell cycle variants, which take place during postnatal heart growth and in adult cardiomyocytes, and do not result in cell division. As the hypertrophic growth phase begins, cardiomyocytes will either become multinucleated by acytokinetic mitosis or polyploid by endoreduplication, a process characterized by the lack of karyokinesis and cytokinesis. In mice ≈90%^5,6 and in humans ≈25% of the adult ventricular cardiomyocytes become binucleated,^7,8 whereas in mice ≈45%⁹ and in humans, 66% of the ventricular cardiomyocyte nuclei are polyploid.⁷

Most studies describing myocardial regeneration by the proliferation of cardiomyocytes use Aurora B-kinase as a gold standard marker for authentic cell division. However, some earlier reports have provided hints that this read-out needs to be treated with caution because of its potential lack of specificity to distinguish proliferation from binucleation.¹⁰ This could be 1 reason for the big discrepancy in numbers of the regenerative capacity of cardiomyocytes after cardiac lesion, as the number of dividing cardiomyocytes after cardiac lesion varies between 0% to 0.0083%^11,12 and 3.2% of border zone cardiomyocytes.¹³ Also, the correct assignment of proliferation signals to cardiomyocytes and especially cardiomyocyte nuclei is an error source leading potentially to a strong overestimation of cardiomyocyte cell cycle activity and proliferation.¹⁴ This prompted us to generate transgenic CAG (chicken β-actin promoter with cytomegalovirus enhancer)-eGFP (enhanced green fluorescent protein)-anillin mice to identify cell cycle activity and cell cycle variants in cardiomyocytes in vitro¹¹ for establishing a screening system for proliferation-inducing substances. Earlier studies have clearly shown that localization of anillin, a scaffolding protein of the contractile ring, could be helpful for distinguishing atypical cell cycle activity from cell division.^10,15 However, detailed cell biological insight into this process still lacks in the heart, and we have, therefore, generated a novel transgenic mouse line, which expresses eGFP-anillin under control of the cardiomyocyte-specific Myh6 (myosin, heavy chain 6) promoter. These mice were crossed to Myh6-H2BmCh (histone 2B monomer Cherry) mice⁵ enabling the unequivocal identification of cardiomyocytes based on mCh fluorescence and of the cell cycle status, as indicated by eGFP-anillin expression and its subcellular localization. eGFP-anillin is localized in the nucleus during G1/S/G2 phases, and it translocates to the cytoplasm after nuclear dissolution in M-phase and resides in the contractile ring and midbody during cytokinesis.¹⁶ Thus, the eGFP-anillin system enables visualization of the M-phase with high spatiotemporal resolution.¹¹

Herein, we demonstrate in vitro, as well as in vivo that an Aurora B-kinase⁺ midbody is visible in cardiomyocytes undergoing binucleation, but that its position and the distance between the daughter nuclei are reliable criteria indicative for either binucleation or authentic cell division. This new method enables to identify cardiomyocyte proliferation, whereas common and established proliferation markers are not suitable for this purpose.

Methods

The authors declare that all supporting data are available within the article and its Online Data Supplement files. Transgenic mouse lines, materials, and further information are available upon personal request at the Institute of Physiology I, University of Bonn, Germany.

Histology and Immunofluorescence Stainings

Cells were fixed with 4% paraformaldehyde in PBS. Adult mouse hearts (8–10 weeks old, male and female) were perfused with 10 mL of 4% paraformaldehyde at room temperature and fixated overnight at 4°C. At embryonic day (E16.5, E18.5) or postnatal day (P2, P5, P7) hearts were harvested and immersion-fixed in 4% paraformaldehyde overnight at 4°C. All tissues and embryoid bodies were incubated in 20% sucrose in PBS before cryopreservation in Tissue Tek O.C.T. compound (Sakura Finetek Europe B.V.). Sectioning of 10 µm cryoslices was performed with a cryotome cardiomyocyte 3050S (Leica).

Fixated cells and tissue slices were stained for the following markers (in 0.2% Triton X in PBS, supplemented with 5% donkey serum; 2 hours at room temperature): α-actinin (1:400, A7811; Sigma-Aldrich), PCM-1 (pericentriolar material 1, 1:100, No. 5213; Cell Signaling), Ki-67 (Kiel-67, 1:400; kindly provided by J. Gerdes, Kiel), and Aurora B-kinase (1:400, ab2254; Abcam). For staining of Aurora B-kinase and Ki-67 on cryosections, antigen retrieval was performed in citrate buffer pH 6.0 at 94°C for 20 minutes.

Primary antibodies were visualized by secondary antibodies conjugated to Cyanine (Cy) 5 (1:400, No. 705-175-147; Jackson ImmunoResearch) diluted in 1 µg/mL Hoechst 33342 (nuclei staining) at room temperature for 1 hour. Fluorescein or rhodamin coupled WGA (wheat germ agglutinin) lectin (No. FL-1021, No. RL-1022; Vector Laboratories) was diluted 1:100 and stained at room temperature for 1 hour. As negative controls for immunofluorescence imaging, specimens were incubated exclusively with the respective secondary antibody.

Immunostainings were documented with an inverted fluorescence microscope (Axiovert 200; Carl Zeiss MicroImaging, Inc) equipped with a slider module (ApoTome; Carl Zeiss MicroImaging, Inc), using filters for 4’,6-diamidino-2-phenylindole, GFP, Cy3, and Cy5, ×25, ×40 differential interference contrast plan apochromat oil objectives, an ebx 75 light source, and an AxioCam MRm digital camera. Pictures were generated with the ZEN V8.1 software (Zeiss). For the quantification of binuclear cardiomyocytes, cardiomyocyte borders were marked by staining for α-actinin. A binucleated cardiomyocyte was defined as an α-actinin⁺ cell with 2 nuclei in close proximity.

Generation of Myh6-eGFP-Anillin Transgenic Mice

G4 hybrid embryonic stem cells¹⁷ were electroporated in the presence of linearized Myh6-eGFP-anillin plasmid, followed by selection for neomycin-resistant cells. Resistant colonies were isolated, propagated, differentiated to embryoid bodies and analyzed for eGFP-anillin expression. Transgenic embryonic stem cell clones were screened for a proper karyotype (40 chromosomes) and aggregated with diploid morula stage CD-1 embryos as described previously.¹¹ The obtained chimeras provided germ-line transmission, and the transgenic progeny was viable, fertile, and had a normal lifespan. Myh6-eGFP-anillin mice were bred to Myh6-H2BmCh-mice⁵ for generation of double transgenic mice.

Animal Procedures

All animal experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals published by the National Institutes of Health (eighth edition, revised 2011) and were approved by the the local ethics review board (Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen, Germany). Double transgenic mice used in this study (Myh6-eGFP-anillin/Myh6-H2BmCh-mice or CAG-eGFP-anillin/Myh6-H2BmCh-mice) were kept on a mixed genetic background (CD-1×C57BL/6Ncr×129S6/SvEvTac).

Acute Ventricular Slice and Atria Preparation

Two- and 5-day old male and female mice were decapitated and hearts quickly removed and chilled for 1 minute in ice-cold cutting solution (Iscove’s Modified Dulbecco’s Medium supplemented with 30 mmol/L 2,3-butanedione monoxime, 0.125 mmol/L nonessential amino acids, 125 μg/mL streptomycin, and 125 U/mL penicillin) saturated with carbogen (95% O₂ and 5% CO₂). Three hundred µm thick slices were prepared in ice-cold cutting solution using a vibratom (VT 1200S; Leica) starting from the apex. For preparation of atria, hearts were dissected and atria isolated by cutting the atrioventricular junction with surgical scissors.

Imaging of Acute Tissue Preparations

Immediately after preparation, slices or atria were transferred into a recording chamber and constantly superfused with gassed recording solution (Iscove’s Modified Dulbecco’s Medium supplemented with 2% FCS, 30 mmol/L 2,3-butanedione monoxime, 0.125 mmol/L nonessential amino acids, 125 μg/mL streptomycin, and 125 U/mL penicillin). The temperature was kept at 35°C. We used a custom-built 2-photon microscope driven with a Ti:sapphire Laser (Chameleon Vision-S; Coherent) running at 950 nm for simultaneous GFP and mCh excitation. To separate the GFP and mCh signals, the emission light was truncated by optical bandpass filters (525/25 and 620/60). The setup was controlled by ScanImage.¹⁸ Image stacks were acquired using a water immersion objective (UMPLFLN, 20×/0.5 W; Olympus). Each stack consisted of several sections with 1024×1024 pixels (0.169 μm/pixel) and 2 μm z-spacing. For time-lapse imaging, the interstack interval was set to 10 minutes.

Dissociation of Postnatal Mouse Hearts and MicroRNA/Small Interfering RNA Transfection

For transfection of postnatal cardiomyocytes, ventricles of P3 Myh6-H2BmCh/CAG-eGFP-anillin double transgenic hearts were dissociated using the Neonatal heart dissociation kit (Miltenyi Biotec). For transfection, 5 µL of 500 nmol/L stock solutions of hsa-microRNA (miR/miRNA)-199a-3p or miRNA negative control No. 1 (containing a random miRNA sequence; Ambion, Life Technologies) were incubated with 0.2 µL Lipofectamine RNAi Max (Invitrogen) in 14.8 µL OPTI-MEM on 96-well µ-clear microtiter plates (Greiner) coated with 0.001% fibronectin for 20 minutes at room temperature. Afterward, 10 000 dissociated P3 cardiomyocytes were added in a volume of 60 µL culture medium (final concentration of miRNAs: 31.25 nmol/L). Medium was changed after 48 hours, and cells were further incubated for 24 hours. For siRNA transfection, P3 cardiomyocytes from Myh6-H2BmCh/CAG-eGFP-anillin mice were transfected as described for miRNAs, but with small interfering RNA (siRNA) specific for p21 (final concentration 50 nmol/L; Ambion/Thermo Fischer Scientific). Medium was changed after 48 hours, and cells were further incubated for 24 hours and fixed.

Knockdown Efficiency Analysis by Quantitative Polymerase Chain Reaction

Knockdown efficiency was measured by quantitative polymerase chain reaction with a TaqMan probe specific for p21 (FAM-MGB Applied Biosystems/Life Technologies) normalized to 18S-RNA VIC-MGB (Applied Biosystems/Thermo Fischer Scientific). Expression in cardiomyocytes treated with scramble control siRNA was set to 100%.

Video Microscopy of Isolated Primary Cardiomyocytes

Videos from primary cultured cardiomyocytes isolated from eGFP-anillin/H2BmCh compound transgenic mice were acquired with an inverted confocal laser scanning microscope (Nikon Eclipse Ti) equipped with a ×40/1.15 numerical aperture water dipping objective (ApoLWD 40× WI SDIC N2). Pictures were taken every 5 minutes for 24 hours. Excitation wavelength for eGFP and H2BmCh were 488 and 543 nm, respectively.

Image Analysis

Z-stacks were analyzed using Fiji.¹⁹ Resolution of pictures in Figure 2F was interpolated to 525×533 pixels from an original resolution of 292×230 pixels by using an antialiasing algorithm. Pictures in Figure 4D and 4E were noise reduced using CANDLE (Collaborative Approach for Enhanced Denoising Under Low-Light Excitation).²⁰

Statistical Analysis

Data are depicted as mean±SEM. Statistical significance was determined by Student unpaired t test or 1-way ANOVA with Bonferroni post hoc test, as indicated in the figure legends. P values <0.05 were considered statistically significant. Normal distribution of data was determined by the Kolmogorov-Smirnov test.

Results

miR-199 Treatment Increases Irregularly Positioned Midbodies in Cardiomyocytes

As a screening system for proliferation-inducing substances in postnatal cardiomyocytes, cardiomyocytes were isolated from P3 hearts of double transgenic Myh6-H2BmCh/CAG-eGFP-anillin mice. In this mouse line, nuclei of cardiomyocytes can be identified by mCh fluorescence, and cell cycle activity can be visualized by eGFP-anillin expression as reported earlier.^5,21 As a positive control for enhanced proliferation, cardiomyocytes were treated with miR-199, which has been reported to induce cell cycle activity.^3,5 Division of cardiomyocytes could be observed and was accompanied by eGFP-anillin⁺ midbodies located in a virtual line between the 2 daughter nuclei (Figure 1A). All midbodies visualized by eGFP-anillin could be verified by staining against Aurora B-kinase (Figure 1A). Of note, Aurora B-kinase does not stain the stem body directly, but its linings, marking a bipartite structure (Figure 1A, arrow).

• Download figure
• Download PowerPoint

Surprisingly, besides authentic cell division, as evidenced by video microscopy,⁵ a high rate of binucleated cardiomyocytes was observed on addition of miR-199 (9.83±2.23% in controls versus 17.74±2.30% after miR-199 treatment, n=4 independent experiments, P=0.048⁵).

To visualize the process of binucleation in vitro, we performed video microscopy of miR-199 treated P3 cardiomyocytes from Myh6-H2BmCh/CAG-eGFP-anillin mice. In binucleating cardiomyocytes, we could distinctly observe dissolution of the nuclear membrane, cytoplasmatic location of eGFP-anillin, the formation of a contractile ring, its contraction, and formation of a midbody (Figure 1B; Online Movie I). However, the midbody did not remain in the middle between the 2 daughter nuclei but moved to 1 side of the division plane and started to fragment and eventually dissolved. A separation of the 2 daughter cells (abscission) did not take place resulting in the formation of a single cell. Unexpectedly, in these binucleating cardiomyocytes, as visualized by eGFP-anillin, midbodies were not found to be located symmetrically between the daughter nuclei but rather randomly distributed in the cytoplasm and of fragmented appearance (Figure 1C). Because of this interesting observation, we decided to investigate the formation and dissolution of these midbodies further. Importantly, these irregular midbodies, as identified by eGFP-anillin, did stain consistently for Aurora B-kinase (Figure 1C), thereby, putting into question, whether this marker provides unequivocal proof for cell division. Aurora B-kinase staining still displayed its bipartite structure as in regular, symmetrical midbodies (Figure 1A and 1C). Asymmetrical, irregular midbodies could also be observed in scramble siRNA-treated controls (Figure 1D), indicating that their appearance is a general phenomenon rather than being induced by application of miR-199. Irregular midbodies made up less than half of all midbodies in miR-199 treated cardiomyocytes (6.99±2.52‰ regular and 5.23±1.43‰ irregular midbodies in relation to total number of cardiomyocytes; Figure 1E [n≥3]) and approximately one-third in controls (0.66±0.66‰ midbodies and 0.24±0.24‰ irregular midbodies in relation to total number of cardiomyocytes; Figure 1E [n≥3]). Midbodies could also be identified by Aurora B-kinase staining. Quantification of their localization with respect to regular/irregular positioning revealed that ≈70% of all midbodies in miR-199 treated cardiomyocytes were irregular (1.24±0.63‰ regular; 3.29±1.03‰ irregular midbodies in relation to total number of cardiomyocytes) and ≈80% in controls (0.24±0.24‰ regular; 0.94±0.66‰ irregular midbodies; Figure 1F [n≥3]). In all cases observed (100%), midbodies identified by eGFP-anillin stained positively for Aurora B-kinase and vice versa.

Downregulation of p21 Increases Binucleation in Postnatal Cardiomyocytes

To assess, if increased binucleation is a feature specific to miR-199 treatment, we aimed to increase cell cycle activity in cardiomyocytes by downregulating the cyclin-dependent kinase inhibitor p21, as has been shown before.²² As expected, knockdown of p21 in P3 primary cardiomyocytes from CAG-eGFP-anillin/H2BmCh double transgenic mice (p21 mRNA expression was reduced to 5.6±1.9% of scrambled control, n=3) led to a strong increase in eGFP-anillin⁺ cardiomyocytes (2.05±0.52% in controls; 17.59±3.76% after p21 knockdown; n=3; P=0.015; Figure 2A and 2B). Similar to our observations on treatment with miR-199, irregular midbodies were evident (Figure 2C). Quantitation revealed 3.3±2.3‰ midbodies and 4.3±2.6‰ irregular midbodies in relation to the total number of cardiomyocytes (Figure 2D; n=3 independent experiments). Accordingly, quantification of midbodies identified by Aurora B-kinase staining (Figure 2E, regular midbody; Figure 2F, irregular midbody) displayed the same distribution pattern, as eGFP-anillin midbodies (Figure 2G, 1.51±0.78‰ midbodies and 3.17±0.93‰ irregular midbodies after p21 knockdown and 0.52±0.52‰ midbodies and 0.32±0.32‰ irregular midbodies in controls). These data implicated that the majority of cell cycle active cardiomyocytes undergo binucleation at this time point. The number of binuclear cardiomyocytes was significantly (P=0.0004) increased after 3 days of incubation in p21 knockdown cultures from 13.41±0.94% in scrambled siRNA controls to 23.89±0.24% in treated cultures (Figure 2H). This underscored a direct correlation between a higher fraction of irregular midbodies and an increased rate of binucleation. Importantly, the process of binucleation resembled the one after miR-199 treatment, namely (1) dissolution of the nuclear membrane, (2) cytoplasmatic location of eGFP-anillin, (3) formation and contraction of a contractile ring, and (4) formation of a midbody with asymmetrical, irregular localization. After dissolution of the midbody, the cardiomyocyte did not divide, and the 2 daughter nuclei stayed together in close proximity (Figure 2I).

In summary, the process of binucleation in vitro shares most features with that of authentic cell division, such as formation and contraction of the contractile ring and formation of a midbody. However, a striking difference was the lack of abscission and the lateral location of the midbody.

Different Location of eGFP-Anillin in Binucleation Events in Myh6-eGFP-Anillin Transgenic Mice

The in vitro findings prompted us to investigate the potential occurrence of irregular, asymmetrical midbodies in cardiomyocytes in mouse hearts during early postnatal development in vivo. During the first 2 weeks of life, ≈90% of left ventricular cardiomyocytes become binucleated.⁵ To be able to live-monitor and compare the processes of binucleation and cytokinesis in cardiomyocytes, a new transgenic mouse model was generated with cardiomyocyte-specific eGFP-anillin expression by using the Myh6 promoter (Online Figure IA). We also hoped that this model would allow us to assign the eGFP-anillin signals unambiguously to cardiomyocytes. Mouse ES-cells were stably transfected with the expression construct, and positive clones were differentiated into embryoid bodies. Only cardiomyocytes in these embryoid bodies displayed the typical eGFP-anillin localization, highlighting the specificity of our transgene (Online Figure IC and ID). Mice were generated from these embryonic stem cells by complementation with diploid embryos. At embryonic day 16.5, expression was restricted to atria (Online Figure IE), as expected from the expression pattern of the Myh6 promoter. Staining for the cell cycle marker Ki-67 at this stage revealed a robust overlap with eGFP-anillin⁺ cardiomyocyte nuclei (Online Figure IF) demonstrating the specificity of the transgene for cell cycle active cardiomyocytes. Specificity for cardiomyocytes was further verified by staining for α-actinin in atria (Online Figure IG), and all eGFP-anillin⁺ nuclei were also positive for Ki-67 (Online Figure IH). The first ventricular cardiomyocyte nuclei positive for eGFP-anillin appeared at embryonic day 18.5 when there is a switch from Myh7 to Myh6 promoter activity (Online Figure II).

At P2, eGFP-anillin expression was visible in both atrial and ventricular cardiomyocytes (Figure 3A). At P5, when much binucleation takes place, eGFP-anillin expression was still abundant in ventricles, as well as atria, indicating cell cycle activity (Figure 3B). Quantification of total eGFP-anillin signals revealed that 5.99±0.84% (n=4) of all cardiomyocytes were positive for eGFP-anillin at P2, a rate which declined to 3.63±0.89% at P5 (n=3) and dropped to 0.97±0.14% (n=3) at P7, indicating further maturation and reduction of the rate of binucleation (Figure 3C).

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

Although at P2 most of the signals were in the nuclei, midbodies could be observed with a symmetrical position in between the daughter cells of dividing cardiomyocytes (Figure 3D). This underscored the occurrence of cell division, which takes place at this early postnatal stage but ceases shortly after.^6,9 Interestingly, ≈60% of the midbodies were found to be preferentially located asymmetrically at 1 side of the cell cycle active cardiomyocyte (Figure 3D, right picture, Figure 3E), similar as observed in our cell culture experiments (Figures 1B, 1C, 1E, 2C, and 2F), indicating binucleation. Both nuclei of these cardiomyocytes displayed uncondensed chromosomes and were positioned in close proximity to each other (Figure 3D).

In contrast to P2, at P5, the majority of the midbodies were found to be located asymmetrically at 1 side of the cell cycle active cardiomyocyte (Figure 3E and 3F) with both daughter nuclei in close proximity to each other (Figure 3F). At P7, ventricles from Myh6-eGFP-anillin mice displayed mainly irregular midbodies (Figure 3E and 3G). This was not surprising because, between P0 and P7, ≈70% of the cardiomyocytes become binucleated.⁵

In atria of P5 hearts, midbodies were located symmetrically between cardiomyocytes, as would be expected for dividing cardiomyocytes (Figure 3H). This result was anticipated, as only ≈10% of all atrial cardiomyocytes become binucleated.⁵

In summary, these data suggested that the position of the midbody is a highly reliable predictor for the fate of cell cycle active cardiomyocytes, thus confirming our observations made in vitro.

Live Imaging of Binucleation in Acute Heart Slices

To monitor and analyze the spatiotemporal process of cardiomyocyte binucleation in detail, we performed 2-photon time-lapse microscopy on 300-µm thick acute heart slices of P2 and P5 Myh6-H2BmCh/Myh6-eGFP-anillin double transgenic mice (Online Figure IB). The Myh6-H2BmCh transgenic line was chosen because it enables straightforward and unequivocal identification of cardiomyocyte nuclei. Between P0 and P10, >70% of ventricular cardiomyocytes undergo binucleation, whereas authentic cell division ceases.^5,9 Therefore, we imaged acute heart sections from P2 mice. As shown above (Figure 3D and 3E), at this time point, both cell division and binucleation were taking place in ventricular cardiomyocytes.^6,9 Time-lapse imaging of dividing cardiomyocytes (5 out of 7 cells observed) revealed the formation of a contractile ring, its contraction and the formation of a regular midbody, which was centrally positioned between the daughter nuclei (Figure 4A, 40 minutes, open arrowhead; Online Movies II and III). The daughter nuclei were clearly separated from each other after karyokinesis (Figure 4A, 40 minutes, 120 min, arrowheads; Online Movies II and III). In contrast, in binucleating cardiomyocytes (2 out of 7 cells observed), a contractile ring was formed, but the midbody moved laterally away from the midline to the cell cortex to an asymmetrical position (Figure 4B, 70 minutes, open arrowhead; Online movie IV). Interestingly, the daughter nuclei were close together (Figure 4B, 10 minutes, 70 minutes, arrowheads) and almost touched each other (Figure 4B, 120 minutes, arrowheads). Quantification of the distance between the daughter nuclei revealed a significantly shorter interval in binucleating cardiomyocytes compared with dividing cardiomyocytes (binucleating cardiomyocytes, 2.3±0.2 µm; n=2; dividing cardiomyocytes, 7.3±0.6 µm; n=5; P=3.4×10^–15; Figure 4C; Online Figure IIC). To corroborate these differences between dividing and binucleating cardiomyocytes we live-imaged acute heart sections from P5 mice, as the vast majority of midbodies at this time point were irregularly positioned (Figure 3E). Accordingly, at P5 the majority of eGFP-anillin⁺ cardiomyocytes in the ventricles are supposed to become binucleated. Time-lapse videos of ventricular cardiomyocytes in acute heart sections demonstrated that eGFP-anillin translocated to the cytoplasm after nuclear membrane dissolution (Figure 4D; 0 minute). The formation of the contractile ring did not take place in most cases observed (7 out of 8; n=8). Instead, eGFP-anillin⁺ structures appeared at the cleavage plane (Online Figure IIA, 90 minutes, arrow; Online Movies V and VI) and condensed to 1 or 2 midbody-like eGFP-anillin⁺ structures. These were located asymmetrically in between the daughter nuclei (Figure 4D, 90 minutes, open arrowhead; Online Figure IIA, 110 minutes, arrows). In all observed cardiomyocytes, the incomplete ring did not contract, but instead, the condensed eGFP-anillin⁺ structures started to fragmentize and disappeared eventually (Figure 4D, 240 minutes; Online Figure IIA, 260–500 minutes; Online Movies V and VI). Interestingly, the 2 daughter nuclei of the binucleating cardiomyocytes were in close proximity or even in contact to each other (Figure 4D, 240 minutes; Online Figure IVA, 360, 500 minutes; Online Movies V and VI), as also observed in sections from P5 Myh6-eGFP-anillin mice (Figure 3F). The mean distance between the daughter nuclei was 1.6±0.6 µm (n=6; Figure 4E). Although the midbody was visible, the distance between the daughter nuclei in every single cardiomyocyte observed was always <5 µm (Figure 4E; Online Figure IID).

During postnatal development, atrial cardiomyocytes display more and longer lasting cell cycle activity than ventricular cardiomyocytes,²³ and only around 8% of them become binuclear in mice.⁵ This implicates that in these cells preferentially cell division takes place. Therefore, we live-imaged cardiomyocytes from acute atria from P5 Myh6-H2BmCh/Myh6-eGFP-anillin mice (Online Movies VII and VIII). In fact, 6 out of 7 atrial cardiomyocytes displayed contraction of the contractile ring (Online Figure IIB, 50 and 60 minutes, arrowheads) and formation of a regular central midbody was visible (Figure 4F, 90 minutes, open arrowhead; Online Figure IIB, 100 minutes, arrowhead). Furthermore, after contraction of the ring, the 2 daughter nuclei remained clearly separated (Figure 4F, 240 minutes, arrowheads), in contrast to P2 and P5 ventricular cardiomyocytes undergoing binucleation. This was especially evident after live imaging of acute atria (Online movies VII and VIII, n=6 events from n=2 mice). However, we were also able to monitor 1 binucleating atrial cardiomyocyte, which was expected, as 8% will eventually become binuclear in adult mice.⁵ Quantitation of the distance between daughter nuclei revealed a striking difference between cardiomyocytes with regular and irregular midbodies (mean distance between daughter nuclei with regular midbody, 7.5±0 µm; n=6; irregular midbody, 1.8 µm; n=1; Figure 4G; Online Figure IIE), as in P2 cardiomyocytes (Figure 4C).

To establish criteria for discrimination of binucleation from authentic cell division, we compared the mean distance of daughter nuclei at the time points when the midbody was visible. It was notable that midbodies appear for a shorter time span P2 ventricular cardiomyocytes compared with P5 ventricular or atrial cardiomyocytes (midbody appearance P2 ventricle, 57±13 minutes, n=7; P5 ventricle, 238±85 minutes, n=5; P5 atrium, 188±39 minutes, n=6). The mean distance between the daughter nuclei of ventricular cardiomyocytes at P5 with irregular midbodies was 1.2±0.5 µm (n=6; Figure 4H). This turned out to be significantly smaller compared with the mean distance between the daughter nuclei of atrial cardiomyocytes at P5 displaying a regular midbody (8.0±0.5 µm; n=6; Figure 4H; P=4.13×10^–6) or compared with the mean distance between the daughter nuclei of ventricular cardiomyocytes at P2 displaying a regular midbody (9.1±0.7 µm; n=5; Figure 4H; P=8.36×10^–6). The mean distance of the daughter nuclei of atrial cardiomyocytes at P5 and ventricular cardiomyocytes at P2 with regular midbodies was not significantly different (Figure 4H; P=0.633). Likewise, the mean distances of the daughter nuclei of ventricular (P2) and atrial cardiomyocytes with irregular midbodies were similar to the mean distance between the daughter nuclei of ventricular cardiomyocytes at P5 with irregular midbodies (atrium P5, 0.5 µm; n=1; ventricle P2, 3.3±0.4 µm; n=2). In summary, the main differences between binucleation and authentic cell division were (1) the contraction of the ring, (2) the distance between the daughter nuclei, and (3) the position of the midbody.

Aurora B-Kinase Marks Both Binucleating and Dividing Cardiomyocytes

Our cell culture experiments and live imaging data of postnatal ventricles revealed that the process of binucleation in cardiomyocytes eventually led to the formation of an irregular midbody or midbody-like structure. Therefore, we wondered if these binucleating cardiomyocytes would also stain positive for Aurora B-kinase, a marker widely used to prove cell division. Staining of P5 cardiac sections revealed Aurora B-kinase⁺ cardiomyocytes, which were identified by perinuclear PCM-1 staining (Figure 5A, arrows). In analogy to the eGFP-anillin⁺ midbody, the Aurora B-kinase⁺ signal was located irregularly at 1 side of the cardiomyocyte. Again, Aurora B-kinase stained the region adjacent to the stem body which was marked by eGFP-anillin in sections from transgenic mice (Figure 3D). In addition, both nuclei were in close proximity and positive for Ki-67 indicating cell cycle activity (Figure 5B). In some Aurora B-kinase⁺ cardiomyocytes, the 2 nuclei were so close together that they appeared to be a single nucleus (Figure 5B). This could explain the irregular location of the midbody, as there appears to be a spatial restriction between the nuclei. In clear contrast, in atria, Aurora B-kinase staining was located in between the obviously separated nuclei (Figure 5C) and membrane staining with WGA underscored the formation of 2 daughter cardiomyocytes. Quantification of regular/irregular midbodies, as identified by staining for Aurora B-kinase, revealed a pattern similar to that of eGFP-anillin⁺ midbodies (Figure 5D). Therefore, Aurora B-kinase consistently marks regular and irregular midbodies and is an appropriate indicator of binucleating and dividing cardiomyocytes.

• Download figure
• Download PowerPoint

These data confirmed our in vitro findings and demonstrated that staining for Aurora B-kinase could not reliably distinguish between authentic cell division and cell cycle variants, such as binucleation.

Discussion

The identification of authentic cell division in cardiomyocytes is of crucial importance for current strategies of myocardial regeneration. For this purpose, it is mandatory to be able to discriminate between authentic cardiomyocyte division and incomplete cell cycle progression. Typically, staining for Aurora B-kinase is used as the ultimate experimental proof of cardiomyocyte division. We have investigated herein the process of binucleation and authentic cell division in cardiomyocytes in detail because of the relevance of the identification of proliferation-inducing substance in cardiomyocytes for cardiac regeneration and repair and the assessment of current readout systems.

To induce cell cycle activity in isolated cardiomyocytes, we used a published miRNA from a bigger screen for proliferation-inducing miRNAs.³ Treatment of P3 cardiomyocytes from eGFP-anillin/H2BmCh transgenic mice with miR-199 led to an increase in the number of midbodies compared with the controls, as expected from earlier reports.^3,5 However, closer analysis provided the unexpected result that almost 50% of these midbodies had an irregular localization and fragmented appearance. Time-lapse video microscopy revealed that these irregular midbodies appeared exclusively in binucleating cardiomyocytes. We also noticed the appearance of asymmetrical midbodies, which were not lying on a virtual line between the daughter nuclei. Therefore, we used eGFP-anillin/H2BmCh compound transgenic mice to identify cardiomyocytes and to track cell cycle activity, cytokinetic events, and midbodies. Both aspects are critical for being able to investigate potential differences between cardiomyocyte binucleation and cell division. Interestingly, irregular and regular midbodies visualized by eGFP-anillin could be stained with Aurora B-kinase, proving that this widely used marker lacks its proclaimed specificity as an unequivocal indicator of cell division in cardiomyocytes. Accordingly, the number of binucleated cardiomyocytes increased significantly after miR-199 treatment or knockdown of the cyclin-dependent kinase inhibitor p21, as did the formation of irregular midbodies.

Since observations in single cells in cell culture are not necessarily indicative of cell biological mechanisms occurring in intact tissue and in vivo, we have generated a new transgenic mouse model. For expressing the cell cycle indicator eGFP-anillin specifically in cardiomyocytes, we utilized the Myh6 promoter enabling to monitor authentic cell division and binucleation of cardiomyocytes in the heart ex vivo and in vivo. For unequivocal visualization of cardiomyocyte nuclei, we crossed these mice with Myh6-H2BmCh transgenic mice and established a model for visualizing cardiomyocyte nuclei and cell cycle activity simultaneously. Analysis of P2 sections revealed 2 types of midbody location: a symmetrical, regular positioning on the virtual line connecting the daughter nuclei and an asymmetrical, irregular positioning with a deviation from this line. Moreover, in cardiomyocytes with irregular midbodies, the distance between the daughter nuclei was smaller than in cardiomyocytes harboring a regular midbody. In P5 sections, an asymmetrical, irregular location of the midbody was evident in the vast majority of the ventricular cardiomyocytes analyzed, whereas in P5 atria the midbodies were located symmetrically between the daughter nuclei. Also, the distance between the daughter nuclei was significantly smaller compared with the atria. This phenomenon has not been reported to date, and its underlying cause is unclear.

One possibility is that increased formation of chromosomal bridges leads to an attachment of daughter nuclei, thereby, providing a steric hindrance during anaphase which could contribute to binucleation. Such bridges have been observed in vitro¹⁵ and in vivo and are most likely caused by erosion of telomeres in the first week after birth.²⁴ It is not clear, whether the contractile ring can break such a chromosomal bridge and abscission can take place unhindered. Interestingly, telomere shortening has been shown to induce p21 expression, thereby, facilitating cell cycle arrest.²⁴

The last and definitive insight into the fundamental differences between authentic cell division and binucleation came from our dual-photon confocal time-lapse videos of acute thick sections of P2, as well as P5 ventricles and whole-mount atria. Videos from P2 sections showed cardiomyocytes undergoing authentic cell division, including karyokinesis with large distance between the daughter nuclei, formation, and contraction of a contractile ring and formation of a regular midbody, which eventually dissolved. In P5, ventricular and atrial cardiomyocytes a contractile ring forms, but only in dividing cells karyokinesis takes place every time. In binucleating cardiomyocytes, none of the contractile rings contracted (n=8), and the daughter nuclei in these cardiomyocytes did not separate as far as during cell division. This is in accordance with previous studies, which monitored binucleation in cardiomyocytes in vitro and observed karyokinesis, but no cytokinesis.^25,26 Importantly, during binucleation, as well as cell division a midbody or midbody-like structure is formed, which can be stained by antibodies against Aurora B-kinase (Figure 5A–5C). This was evident after quantification of midbodies from P2, P5, and P7 ventricular sections which were identified by Aurora B-kinase staining. Although the frequencies of regular and irregular midbodies were equally distributed at P2, the majority of the midbodies were irregularly positioned at P5 and P7, a stage when binucleation is becoming predominant.

Thus, our data demonstrate that Aurora B-kinase does not indicate authentic cell division per se, as it also marks midbodies in binucleating cardiomyocytes. However, we found that the localization of the midbodies is indicative for the underlying cell cycle event: Localization of the midbody symmetrically in the middle of the line between the daughter nuclei is predictive for cell division, whereas an asymmetrical localization divergent from this line predicts a binucleation event. A second and even stronger criterion is the distance between the 2 daughter nuclei because nuclei in binucleating cardiomyocytes are significantly closer to each other. Sometimes the nuclei seemed to touch each other, and discrimination from single nuclei was only possible by immunofluorescence staining (Figure 5B). Interestingly, in adult cardiomyocytes of mice, nuclei are always localized far away from each other,²⁷ which suggests a rearrangement of nuclei after binucleation. Binucleation in cardiomyocytes has been described to take place by regression of the cleavage furrow.²⁸ The midbody was reported to form normally in such a scenario and, therefore, to provide no indication, whether a cell will divide or become binucleated.²⁹ Herein, we show that both the position of the midbody and the distance between the daughter nuclei are discriminatory parameters. Using these indicators enables us to predict, which cardiomyocytes will divide and which will undergo binucleation. For heart sections, we recommend staining for Aurora B-kinase for visualization of the midbody, in combination with WGA for visualization of cardiomyocyte cell membranes and either Ki-67 or PCM-1 to identify the daughter nuclei. As a criterion for binucleation, we suggest a distance of ≥5 µm between the daughter nuclei in cardiomyocytes with Aurora B-kinase⁺ midbody. For the second indicator, the position of the midbody, we recommend to draw a virtual midline between the daughter nuclei and to evaluate the vertical deviation of the midbody. A deviation of >1 µm is a hint for irregular positioning. Both criteria need to apply to distinguish a binucleating from a dividing cardiomyocyte. Our earlier data clearly demonstrate that during embryonic heart development, our criteria for authentic cardiomyocyte division also hold true because we could never observe atypical midbody locations or daughter nuclei in close distance.¹¹ Can, likewise, other factors help distinguishing binucleation from authentic cell division? Loss of sarcomere structure and localization of its components to the cell cortex have been used as an indicator for cell division in postnatal mouse hearts.³⁰ However, this has been found in binucleating cardiomyocytes as well.³¹ In sections from hearts from P2, P5, and P7 Myh6-eGFP-anillin/Myh6-H2BmCh mice sarcomeric structure was disassembled in M-phase. As these cardiomyocytes undergo binucleation, disassembly of sarcomeres is not a valid criterion to distinguish binucleation from cell division. Do our criteria hold up for other species? In rats, postnatal heart growth is very similar to that of mice, and cell cycle variants are identical.³² Therefore, our criteria will most likely apply to rats and probably for all rodent species, which have a high amount of binucleated cardiomyocytes in their ventricles.³³ In larger animal models, such as pigs or sheep binucleation and polyploidization events have been described, too.^34–36 In both large animal models, the same cell cycle variants take place as in rodents which emphasizes the conservation of these fundamental cell biological processes. Although the analysis of cardiomyocyte binucleation events in pig and sheep is beyond the scope of this study, we think that there is a good chance that our criteria will apply for these cell cycle variants also in large animal models.In human hearts, the percentage of binucleated cardiomyocytes is lower than in mice, although the degree of cardiomyocyte polyploidization is significantly higher.^7,8 In human left ventricles, a degree of ≈30% of binuclear cells is established in the first week after birth,³⁷ and this fraction does not change during life.^7,8 Accordingly, binucleation events are not expected postnatally and during cardiac repair. However, average cardiomyocyte nuclear ploidy increases with age,^7,8 as well as after myocardial infarction in the border zone.^38,39 Thus, in humans, the main challenge consists of discriminating endoreplication from authentic cell division. As there are strong similarities between man and mice regarding cell cycle variants, it would be surprising if the process of binucleation in human cardiomyocytes would differ significantly from that in mice, but this important aspect needs to be investigated more in detail in future studies.

In summary, we have shown how important a detailed analysis of the cardiomyocyte cell cycle is to distinguish cell cycle variants from authentic cell division. Markers, such as Ki-67, PCNA (proliferating-cell-nuclear-antigen), pHH3 (phosphorylated histone H3), Aurora B-kinase, which are widely used, are not reliable in the postnatal heart or need to be carefully investigated to estimate the degree of cardiomyocyte proliferation versus hypertrophy.³¹ This has far-reaching implications for studies investigating cardiomyocyte turnover and regeneration after injury and is 1 possible reason for the discrepancies in reports about the extent of cardiomyocyte renewal. To avoid misinterpretation of cardiomyocyte cell cycle activity, we propose as a new method the visualization of the midbody and considering its position, as well as the distance between the daughter nuclei. As cardiomyocyte nuclei identification without an explicit nuclear marker is challenging,^5,14,40 either specific antibodies against cardiomyocyte nuclear components, such as PCM-1⁴¹ or transgenic systems, should be used. Taking these measures into account is crucial for evaluation of the effect of new therapeutic strategies, which aim at inducing proliferation in cardiomyocytes after the loss of contractile myocardium.

Acknowledgments

We thank Andras Nagy (Toronto, Canada) and Marina Gertsenstein (Toronto, Canada) for providing the G4 mouse embryonic stem cell line. We are grateful to Johannes Gerdes (Kiel, Germany) for providing an anti-Ki-67 rabbit serum. We would like to also acknowledge Jürgen Schmidt (Bonn, Germany), Patricia Freitag (Bonn, Germany), and Sabine Grünberg (Bonn, Germany) for technical assistance.

Footnotes