A Proliferative Burst During Preadolescence Establishes the Final Cardiomyocyte Number

SUMMARY

It is widely believed that perinatal cardiomyocyte terminal differentiation blocks cytokinesis, thereby causing binucleation and limiting regenerative repair after injury. This suggests that heart growth should occur entirely by cardiomyocyte hypertrophy during preadolescence when, in mice, cardiac mass increases many-fold over a few weeks. Here we show thata thyroid hormone surge activates the IGF-1/IGF1-R/Akt pathway on postnatal day-15andinitiates a brief but intense proliferative burst of predominantly binuclear cardiomyocytes. This proliferation increases cardiomyocyte numbers by ~40%, causing a major disparity between heart and cardiomyocyte growth. Also, the response to cardiac injury at postnatal day15 is intermediate between that observed at postnatal day-2 and -21, further suggesting persistence of cardiomyocyte proliferative capacity beyond the perinatal period. If replicated in humans, this may allow novel regenerative therapies for heart diseases.

INTRODUCTION

The regulation of heart size during postnatal development is a fundamental process directly relevant to remodeling of the heart in response to congenital heart diseases. Postnatal change in the size of the mammalian heart through replication of cardiac muscle cells (cardiomyocytes, CMs)is thought to be limited by early postnatal terminal differentiation (Soonpaa et al., 1996; Walsh et al., 2010). This notion was proposed as early as 1894 by Bizzozero, who suggested that replication of CMs ceases at birth, and classified mature CMs as elementi perenni, or cells of static populations.

The view that CMs terminally differentiate in the neonatal period (by postnatal day5 (P5)) is supported by evidence that during early preadolescence (from P5 to P14) expression of genes responsible for cell cycle reentry, mitosis, and cytokinesis falls precipitously (Walsh et al., 2010). This change is preceded between P5 and P10 by extensive binucleation of CMs (Soonpaa et al., 1996; Walsh et al., 2010), widely considered a marker of terminal differentiation; by P7, hearts lose regenerative capacity (Porrello et al., 2011). Between P10 and P21 (weaning), estimates of S phase indicate that CMs are quiescent (Soonpaa et al., 1996). Assessments of mitosis in murine CMs at select ages (Walsh et al., 2010) also support the conclusion that replication ceases by P5–P7.

In mammals, the heart grows markedly between the immediate post-neonatal period and puberty. If CMs are terminallydifferentiated, this growth shouldbe due almost entirely to an increase in the volume of constituent CMs,whichthroughout post-neonatal lifeaccount for80% of the volume of the myocardium (Li et al., 1996). In humans a disparity between changes in CM volume and heart volume is readily observed between birth and 20 years (Mollova et al., 2013). This disparity suggests a 3.4-fold increase in the CM population number (CPN). Senyo et al., (2013) have investigated the source of such new CMs in the mouse, the only species in which the timing of terminal differentiation has been evaluated. They report that CMs are added to post-neonatal hearts at a rate of 0.76% per year, and that these new cells are derived from a small fraction (< 0.2%) ofmononuclear CMs that retain proliferative capacity. This rate of CM addition predicts only anegligible (~1.06-fold) increase in CPN between the onset of terminal differentiation (around P5) and puberty (P35).

Here we show a 1.4-fold increase in CPN during the preadolescent period, which occurs as a discrete proliferative burst on P15 initiated by a surge in thyroid hormone (T3). Proliferation is evident from an increase in CPN, the re-expression of mitosis-related genes, readily apparent mitotic figures in mature mono- and binuclear CMs and, consistent with cell division giving rise to smaller progeny, a sharpfall in both mono- and binuclear CM volume. The brevity of this proliferative burst could explain why it has previously gone undetected. This study provides compelling evidence for retention of CM proliferative competence long after the neonatal period, which requires a significant revision of the generally accepted view of CM terminal differentiation.

RESULTS

Disparity Between Heart and CM Growth during Postnatal Development

Between early preadolescence (~P10) and puberty (~P35),mouse body weightnearly doubles each week (Figure 1A). Assuming that circulatory volume increases in direct proportion to body weight, circulatory volume should increase by ~15% each day, with equivalent increases in heart weight (Figures 1B and 1C). Echocardiography shows that between P10 and P35 stroke volume increases 3.5-fold (p< 0.001) (Figure 1D). This is associated with LV chamber remodelinggiving an 86% increase in LV end diastolic dimension (LVEDD) (p< 0.001; Figure 1E), and a 4.6-fold increase in LV volume at diastole (p< 0.001; Figure 1F), without a significant change in LV free wall thickness at diastole (FWd or h) (Figure 1G). This adaptation produces a 52% decrease in the LV h/R_i ratio (where R_i is the internal LV chamber radius; Figure 1H), consistent with eccentric hypertrophy, and maintains the LV weight-to-stroke volume ratio (1.76:1 at P10 versus 1.78:1 at P35). Also between P10 and P35, LVEDD length-to-diameter ratio decreases by 40% (p< 0.001; Figure 1I) indicating increased LV sphericity.At the cellular level, CM length increasesby 46μm(p< 0.001; Figure 1K) between P10 and P35, with minimal (+1μm) change in diameter (p< 0.001; Figure 1L).

Figure 1. Rapid Body Growth Involves CM Elongation and Eccentric LV Remodeling.

(A and B) Concomitant increases in body and heart weights of WT mice during the period from preadolescence (P10) to just after puberty (P35).

(C) Increase in heart size from P10 to P35 is illustrated by hematoxylin and eosin stained coronal heart sections.

(D to L) Hemodynamic and cardiac and CM morphological changes in WT mice from P10 to P35. LVEDD and LVED volume, left ventricular end-diastolic dimension and volume, respectively; LV FWd, LV free wall thickness in diastole.

Data are shown as mean ± SEM. The number of animals,or cells, studied is shown by n. ***p< 0.001.

The increase in ventricular weight between P10 and P35 (21.7 ± 1.43 mg, n = 9 at P10 versus 75.3 ± 5.56 mg, n = 5 at P35; +3.47-fold; p< 0.001) exceeds the increase inventricular CM volume by 1.7-fold, calculated on the basis of a cylindrical model (18.2 ± 0.31 pl [picoliter], n = 511 at P10 versus 36.3 ± 0.78 pl, n = 497; +1.99-fold; p< 0.001). This disparitybetween increases in heart and CM volume suggested an increase in CM numbers during preadolescence.

Extensive CM Proliferation in the Preadolescent Heart

We determined total CM numbers in ventricular myocardiumby enzymatic disaggregation and direct cell counting. Estimates of total CM numbers (summarized in Figure 2A)identify two distinct increases in CPN: an ~40% increase between P1 and P4, and a further~40% increase (~500,000 CMs) between P14 and P18. 22% of the post-P14 CPN increase occurred by ~4:00PM on P15 (P15 afternoon or P15A; p< 0.001; Figure2A), with no further change between P18 and P365. Because variable CM yields amongmice of different ages might confound our count, we calculatedthe CM fraction of myocardial volume-to-CM volume ratio, a yield-independent method for estimating CM numbers(Chaudhry et al., 2004), and found 1.26 ± 0.03 × 10⁶ CMs in both ventricles at P14 (n = 10) versus 2.2 ± 0.06 × 10⁶ at P18 (n = 5) (p< 0.001). Thus both a hemocytometer-based method anda yield-independent methodshowa large increase in CM numbers between P14 and P18.The latter method,because it requires several assumptions, likely overestimates the increase. Collectively, these data indicate that the preadolescent increase in CPN results from a discrete CM proliferative burst at ~P15, rather than continuous low-level CMaddition.

Figure 2. A Period of CM Proliferation During Preadolescent Heart Growth.

(A) Total numbers of cardiomyocytes in both cardiac ventricles of mice (CM population number, CPN). The number of animals studied is shown in square brackets.

(B to G) Enhanced expression of the mitosis-related genes, Ki67, Cyclin B1, polo-like kinase-1 (Plk 1), aurora A, Survivin and anillin on P15. Also note suppressed expression of all these genes in P35 hearts (P< 0.001 versus P13 values). Values were determined using RNA prepared from 5 animals at each time point.

(H to J) An example of flow cytometric analysis of BrdU⁺/cardiac troponin T⁺ (cTnT⁺) CM-derived nuclei obtained from a mouse given a single intraperitoneal injection of BrdU on the night of P14 and then sacrificed on P18.

(K) A representative example of analysis of BrdU uptake (at P14 evening) and ploidy, indicating that 96.4% of nuclei were 2n and 3.6% were 4n (on P18) in this cell preparation from a single heart.

Data are shown as mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001.

To determine the time of onset of mitosis, we measured the expression of several mitosis-promoting genes in the cardiac ventricles daily from P13 to P16 (Figures2B–2G). We found ~5–12-fold increases in mRNA levels (p< 0.05) of all these genes on the morning (~9:00AM) of P15, with levels on P16 falling to nearP13 levels. Thus, CMs are in M-phase as early as 9:00AM on P15. Assuming that the combined duration of S and G₂ phases is ~10 h (for a 24-h cell cycle), S phase could start around 10:00PM on P14. To identify S-phase-timing we gave a single intraperitoneal injection of BrdU [bloodstream half-life ~2h (Kriss and Revesz, 1962)]at ~9:30PM on the night of P14. Mice were sacrificed on P18, and CM isolated and purified to remove non-CMs; their nuclei were then liberated by lysis and analyzed byflow cytometry (an example is shown in Figures 2H–2J). >99% of nuclei from CM-enriched cardiac cells were cTnT-positive, arguing against significant contamination bynon-CM nuclei. Flow cytometry further indicated that 11.3 ± 1.9% (n = 6) of CM-derived (cardiac troponin T (cTnT)-positive) nuclei were BrdU positive (e.g., Figure 2J).BrdU'sshort half-life in the circulation makes it unlikely that all CMs entering S phase on the night of P14 would be labeled by a single intraperitoneal BrdU pulse, but cell/nuclear divisions during the 3-day chase period would increase the number of BrdU-labeled nuclei. Thus, while our findings do not establish the number of nuclei entering S phase on the night of P14, they do show that a new S phase in CMs begins late on P14.

The long chase period also determinedthatthe BrdU-labeled nuclei underwentnuclear division after labeling on the evening of P14. By the end of the chase period the majority of the BrdU-labeled CM nuclei were 2n (96.4 ± 0.48% were BrdU⁺ 2n nuclei and 3.6 ± 0.5% were BrdU⁺ 4n nuclei;n = 5/group) (Figure2K). Moreover, despite this post-P14 DNA synthesis in CMs, ploidy was not increased between P14 and P18 (FigureS1). Together, these findings indicateefficient karyokinesis.

We next identified mitotic CMs by monitoring co-expression of aurora B and the sarcomeric marker protein, α-cardiac muscle actin (SαA), by confocal microscopy. The subcellular localization of aurora B isdependent on cell cycle phase, so it can be used to distinguish between potential outcomes of progression into M phase (Carmena and Earnshaw, 2003). Transverse sections of ventricles of mice sacrificed between 8:00AM and 12:00PM revealed a 36-fold increase in LV CM mitoses between P14 and P15(p< 0.001) followed by an abrupt 5.8-fold decrease between P15 and P16 (p< 0.001) (Figures3A–3E). These changes paralleled changes in expression of mitosis-promoting genes (Figures2B–2G). Nuclear localization of aurora B in most of the mitotic CMs indicated that these cells were in prophase (Figure 3B).

Figure 3. The Proliferative Burst is Temporally Discrete, and Involves Division of Predominantly Subendocardial CMs.

(A to D) Immunohistochemical identification of mitotic CMs (red staining, aurora B⁺ cells) in transverse cut tissue sections showing localization in the subendocardial regions of the left ventricle (LV) of P15 mice, with few aurora B⁺ cells evident in the subepicardial regions of the left ventricle, in the right ventricle (RV) or in LV sections of P14 or P16 hearts.The nuclear localization of aurora B indicates that these CMs are in prophase.

(E and F) Quantitation of aurora B⁺ cells identified (A to D), as seen in various regions of the LV and RV of P14, P15 and P16 hearts.

(G) Quantitation of LV aurora B⁺ P15-CMs showing that ~90% are bi-nucleated. In the adjacent photomicrograph the punctate aurora B staining (red) indicates that these CMs, labeled P, are in prophase. In another CM, labeled A (anaphase), aurora B localization between nuclei pairs is consistent with its localization at centromeres between kinetochores, or at the anaphase spindle midzone (asm). Note also the loss of cross-striations and marginalization of sarcomeric structures in the CM labeled A; blue, DAPI; green, α-MHC.

(H) Approximately equal percentages of P15 1N and 2N CMs are aurora B⁺.

(I) CMs in late telophase/cytokinesis in longitudinal sections of the LV subendocardium, detected using an MKLP1 antibody. Arrowheads indicate XY and XZ reconstruction planes in the enlarged insets.

(J) Frequency of MKLP-1⁺CM-specific events in transverse sections of the LV subendocardium. Data are shown as mean ± SEM. The number of animals studied is shown by n or by values in square brackets. **p< 0.01; ***p< 0.001; gp., group. See also FiguresS2, S3, and S4.

Mitotic CM nuclei were not uniformly distributedthroughout the LV wall, being 2.6-fold more abundant in subendocardialthan in subepicardial myofibers of the P15 LV (p< 0.01; Figures 3B and 3F).In these transverse cross-sections, CMs in the subendocardium are cut along their short axes, while CMs in the subepicardium are cut along their long axes; these latter myofibersare oriented circumferentially in the LV wall (Figure 3B).

We also found that 14.5 ± 2.6% of LV CM nuclei were in mitosis, versus 0.73 ± 0.4% in the RV (Figure 3E) (n = 5/ group; p< 0.001). Figures S2A–S2C illustrate this differential distribution of mitotic CMs in a transverse section of P15 heart, which is in contrast to the broad distribution of mitotic CMs in P2 neonatal heart(FiguresS3A–S3C).

To verify the abundance of mitotic CMs in P15 cardiac ventricles, we purified CMsafter enzymatic disaggregation,and used [.alpha]-sarcomeric myosin heavy chain (α-MHC) and aurora B immunofluorescence to identify CMs inM phase. ~10% of total mitotic P15-CMs were mononuclear (had an aurora B-positive nucleus, or aurora B localized at the anaphase spindle midzone), and ~90% were binucleate (aurora B-positive nuclei, or aurora B localized at the anaphase spindle midzone of 2 pairs of nuclei) (Figure 3G). 34.2 ± 4.4% and 30.8 ± 2.7% of mono- and binuclearCMs, respectively, were in M phase (n = 6) (Figure 3H);thisis ~2-fold higher than inimmunohistochemical analyses (Figure 3E), which have been shown tounderestimate mitoses (Mollova et al., 2013).

In some mitotic binuclear CMs we found sarcomeric structures on the periphery (e.g., Figure 3G shows a binucleate CM in anaphase, labeled “A”). Localized disruption of sarcomeresmay facilitate chromosome movement and cytokinesis, but could also adversely affect LV function. However,cardiac contractility was similar in P14 and P15 hearts (LV fractional shortening: 35.8 ± 3% in P14 hearts versus 35.1 ± 2.3% in P15A) hearts;n = 6/group; p = 0.87) suggesting that the sarcomeric disruption noted in P15 CMs was not sufficient to depress LV function.

Cytokinesis requires the formation and constriction of a mature contractile ring. Aurora B and the centralspindlin component, mitotic kinesin-like protein (MKLP1), make independent contributions to this process (Lewellyn et al., 2011). Because most of the mitotic CMs in P15A hearts were in the LV subendocardium (where myofibers are oriented longitudinally), we examined CM MKLP1 localization in longitudinal sections, where axial views of CMs in late mitosis are more likely. ~4% of CMs contained MKLP1 between pairs of CM nuclei (Figures 3I and 3J). The frequency of these CM MKLP1⁺ events and of mitotic figures (Figures S4A–S4E) supports CM proliferation in P15 hearts.

Modes of Mitosis inMononuclear and Binuclear CMs

Mononuclear CMs that have entered M phase can produce two mononuclear daughter cells via a conventional cell cycle. We calculated the number of mono-, bi-, and multinuclear CMs added to the ventricles between the morning of P14 (9:00AM) and the afternoon of P15 (4:00PM), by multiplying the average CPN (Figure 2A) bythe percentages of CMs that were mononucleate, binucleate, or multinucleate at these times (Figures 4A–4D, Table S1). The most striking change was the addition of 11.5 × 10⁴ mononuclear CMs (2.3-fold increase). For extant mononuclear CMs to be the sole source of this addition, all mononuclearCMspresent at P14 (calculated to be ~8.5 x 10⁴ cells) would need to undergo mitosis on P15. However, the rate of mitosis in mononuclear CMs on P15was only ~34% (Figure 3H);this could account forat most 2.9 × 10⁴ cells (or 25% of the addition). Moreover, mitosis of mononuclear CMs does not explain the origin of the 6 × 10⁴ binuclear, and the 9.4 × 10⁴ multinuclear, CMs added between P14 and P15A (Table S1).

Figure 4. Increase in CM Population Number (CPN) involves Division of Mono- and Binuclear CMs.

(A to C) Percentage of mononuclear (mono-CM), binuclear (bi-CM) and multinuclear (multi-CM) CMs in the hearts of P14, the afternoon of P15 (P15A) and P18 mice.

(D) Number of CMs in cardiac ventricles of P14, P15A and P18 mice. Total numbers are shown in the left panel and the numbers of mono-, bi-, and multinuclear CMs in these hearts is shown in the right panel (these data are from Figure 2A). We calculated the number of mono-, bi-, and multinuclear CMs that are added to the cardiac ventricles by multiplying the average CPN (Figure 2A) with the percentage of CMs that were mono-, bi- or multinuclear at these developmental ages (shown in Figures 4A–4C). In A–D the number of animals studied is shown by the value in square brackets.

(E to H) Volume (E and F), width (G) and length (H) of mono- and binuclear CMs in P14, P15A and P18 mouse hearts. The volume of individual CMs is represented by a red (mono-CM) or green dot (bi-CM), and was from the hearts of 5 mice at each time indicated (E and F). In G and H, histograms represent average widths or length of CMs (n = 69–107 mono-CMs/group and n = 391–512 bi-CMs/group) from 5 mice at each time indicated.

(I) Scheme illustrating potential modes of cell cycling during maturational heart growth.Mononuclear CMs undergo a conventional cell cycle to replicate.Binuclear CMs remain static or undergo cell division. This involves each of the two nuclei undergoing karyokinesis,with cytokinesis taking place between nuclei pairs. This generates two mononuclear CMs at the two poles and a smaller binuclear CM remains at the center. Overall, these processes increasethe number of mononuclear CMs, without greatly changing the number of binuclear CMs. However, between P14 and the afternoon of P15 (P15A), mono- and binuclear CM volume decreases because smaller daughter cells are generated from larger preexisting CMs. Data are shown as mean ± SEM. **p< 0.01; ***p< 0.001. See Figure S1 and Table S1.

Binuclear CMs are thought to be incapable of division, because polyploidy is considered to indicate terminal differentiation. However, Miyaoka et al. (2012)propose that binuclear hepatocytes that have entered M phase assemble all condensed chromosomes from two nuclei and produce two mononuclear daughter cells, so thata binuclear cell is “consumed” with the birth of2 mononuclear cells.Our calculationsdo not support this as the dominant mechanism, because binuclear CM numbers increase by ~6% (or 6 × 10⁴ cells).Nonetheless, the finding that ~90% of all P15 mitotic CMs are binucleate(Figure 3G) suggests a major role for polyploid cells inthe CM hyperplastic burst.Theinitiation ofM phase and karyokinesis at each nucleus of a binuclear CM, followed by cytokinesis between pairs of newly generated daughter nuclei, could yield 2 small mononuclear cells at the poles, with a binuclear cell remaining at the center. The 2 nuclear divisions would each contribute 1 nucleus to the binuclear daughter cell. We refer to this as the “2+1 cell cycle.”In vitro examples of a “2+1 cell cycle” in a binuclear CM have been observed (Engel et al., 2005). Alternatively, mitotic binuclear CMs could undergo acytokinetic mitoses to yield multinuclear CMs, or not proceed beyond prophase (i.e., abortive mitosis).

We considered these possibilities because confocal immunocytochemical studies with aurora B revealed manybinuclear CMs in prophase; others displayed dual karyokinesis with evidence of a spindle midzone between the nuclei in each pair (that is, post-karyokinesis) (e.g., Figure 3Ginset). If a significant number of“2+1 cell cycles”occur, the volume of binuclear cellsshould decreaseprecipitously between P14 and the afternoon of P15. Consistent with this scenario, the average volumes of bi- and mononuclear CMdecreased by 60% (p < 0.001) and 43% (p < 0.001) (Figures 4E–4H), respectively,despite an ~10% increase in heart weight (from 35.6 ± 0.86 mg, n = 10 at P14 to 39.1 ± 1.3 mg at P15, n = 11; p< 0.05).If all mitotic mononuclear CMs at P15 undergo a conventional cell cycle to yield 2 mononuclear daughter cells, we calculate that it would require only ~7.3% of the binuclear CMs to undergo an unconventional “2+1 cell cycle” in order to achieve theobserved increase intotal mononuclear CM numbers, including the 6 × 10⁴ binuclear CMs that are added (presumably through subsequent endocycles in some mononuclear cells) between P14 and P15A.These concepts are schematicallyillustrated in Figure 4I. While some of the 9.4 × 10⁴multinuclear CMs added between P14 and P15 might be intermediates of the “2+1 cell cycle,” most are unlikely to divide further becausetheir numbers decreasebyonly ~7%in the following 3 days (Figure 4D).The finding that binuclear CM size (but not numbers) decreases sharply(by ~60%), as mononuclear CM numbers increase by 2.3-fold between P14 and P15,supports polyploid CM mitoses followed by cell division as the major component of the CM proliferative burst.

c-kit⁺ Cardiac Stem and Progenitor Cells in Preadolescent Hearts

We next evaluated if CM progenitor cells (c-kit⁺ and Nkx-2.5⁺) derived from c-kit⁺ resident stem cells—the most abundant CM-forming stem cell in postnatal hearts (Linke et al., 2005)—might contribute to the mononuclear CM pool. Flow cytometric analysis of the non-CM fraction of P14 ventricles showed 209 ± 52 (n = 5) c-kit⁺/Nkx-2.5^– small interstitial cells (which include c-kit⁺ cardiac- and bone marrow-derived stem cells), and 69 ± 29 (n = 5) c-kit⁺/Nkx-2.5⁺CM progenitors. These numbers are insufficient to account for the addition of~500,000 CMs between P14 and P18.

Thyroid Hormone Triggers Cardiac Growth and Hyperplasia during Early Preadolescence

Despite a highly significant correlation (r² = 0.93; p< 0.0001) between heart and body weight between P2 and P84 (Figure S5), from P11 to P18 heart weight increased more than body weight (Figure S5). This suggests that thepreadolescent increase in heart growthcannot be accounted for by rapid circulatory volume expansion (or body weight) alone.

Immediately after P10, the rate of heart growth exceeded that of body growth (Figure 5A), resulting in a rapid 30% increase in heart-to-body weight ratio between P10 and P17 (p< 0.001). The initial phase of this increase, between P10 and P14, was by eccentric hypertrophy. This involved a 1.56-fold increase in cardiac mass (p < 0.001; Figure 1B), but also increases in LV chamber dimensions (1.24-fold increase in LVEDD; p < 0.001) and volume (2.8-fold increase; p < 0.001) (Table S2), with more prominent increases in CM length than width (Figures S6A–S6C), but without CM hyperplasia (Figure 2A). These changes caused a 22% decrease in the h/Ri ratio (p < 0.05; Table S2), indicating an increase in LV wall stress based on the Law of LaPlace.But ventricular α/β-MHC mRNA ratio also increased ~5-fold (p < 0.01; Figure 5B) and α-MHC levels increased 2.5-fold (p < 0.05; Figure 5C), whileatrial natriuretic factor mRNA levels (a marker of pathological hypertrophy) werenot significantly increased (Figure 5D).

Figure 5. Disparity in the Relationship between Heart and Body Growth, withCM proliferation requires a T3 Surge.

(A) Temporal change in heart-to-body weight ratios of WT mice showing that heart growth exceeds body growth between P10 and P18.

(B and C) Increase in LV α/β-MHC mRNA ratio (B) and in α-MHC expression (Western blot above; quantitation below) (C) from P10 and P14, and blockade of the increase in mRNA ratio by PTU (B).

(D) Expression of ANP mRNA is not significantly changed between P10 and P14.

(E) 3,3’,5-triiodo-L-Thyronine (T3) levels increase markedly between P10 and P12 and remain stable thereafter. The increase in T3 is abrogated by PTU (red dot and green bar).

(F and G) Increase in heart-to-body weight ratio from P10 to P14, P15, and P18, and inhibition of these increases by PTU (F). PTU also prevents the increase in CPN between P10 and P18 (G). (H) Treatment of cultured cardiomyocytes isolated from P14 mice with T3 for 7 days increased the percentage of BrdU⁺ CMs (cTnT⁺).

Data shown are the means ± SEM.The number of individual animals studied is shown by n. *p < 0.05; **p < 0.01; ***p < 0.001. See Figures S5 and S6and Table S2.

This molecular and morphological signature suggests a 3,3',5-triiodo-L-thyronine (T3 or thyroid hormone)-mediated effect, since neither physiological nor pathological cardiac hypertrophy cause large increases in the α/β-MHC mRNA ratio (e.g., Perrino et al., 2006), but T3 excess does (Haddad et al., 2008). Consistent with this hypothesis, serum T3 levels increase5.6-foldbetween P10 and P12 (Figure 5E). To determine if T3 is necessary for the post-P10 cardiac growth, we inhibited T3 biosynthesis with propylthiouracil (PTU). PTU administration from P7 decreased serum T3 levels at P14 by 43% (p< 0.05) (Figure 5E), prevented the increase in the α/β-MHC mRNA ratio (Figure 5B), and reduced heart weight more than body weight, so that by P14–P18 the heart-to-body weight ratios of PTU-treated mice were not significantly different from P10 mice (Figure 5F). This is consistent with a high level of circulating T3 regulating cardiac growth during early preadolescence. The subsequent decrease in heart-to-body weight ratio between P16 and P21 (Figure 5A) was due to a reduced rate of heart growth (relative to the previous 5 days) (Figure 1B), during which body weight continued to rise (Figure1A), but not due to a reduction in serum T3, whichremained high between P12 and puberty (Figure 5E).

Hence, preadolescent growth is characterized by a bidirectional disparity in heart and body weight changes—first, between P10 and P16, when heart growth exceeds that of the body, and subsequently, between P16 and P21, when body growth exceeds that of the heart. Together these findings suggest that between P10 and P14, T3 critically modulates the mode by which body weight/circulatory volume drives heart growth, allowing a more rapid increase in heart size during early preadolescence.

We next evaluated the role of T3 in CM hyperplasia. Blockade of T3 biosynthesis abrogated the developmental increase in the CPN by P18 (Figure 5G).Also, a 7-day exposure of cultured P14 CMs to T3 (70 nM) produced an ~2-foldincrease in DNA synthesis (BrdU incorporation)(p < 0.05; Figure5H).

We then explored a potential link between T3 and IGF-1 in preadolescent hearts. The IGF1-R/Akt pathway in the heartis both pro-survival and pro-proliferative (Dai and Kloner, 2011); IGF-1 is secreted by cardiac fibroblasts and CMs (Horio et al., 2005). Akt is required for physiological growth in response to IGF-1 stimulation (DeBosch et al., 2006).T3 regulatesIGF-1 mRNAin osteoblasts via anIgf1thyroid hormone response element (TRE) (Xing et al., 2012). IGF-1 causes fetal heartCM proliferationby activating the IGF1-R/PI3K/Akt pathway (Sundgren et al., 2003), and postnatal CM numbers are increased in transgenic mice with cardiac-restricted IGF-1 overexpression (Reiss et al., 1996). We found thatIGF-1 mRNA and IGF-1 expression areincreased 2.3-fold (p < 0.05; Figure 6A) and 39-fold (p < 0.05; Figures 6B–6F), respectively, in P15 relative to P10 hearts. Activation of IGF1-R requires phosphorylation that activates phosphatidylinositol-3 kinase (PI3K), which phosphorylates and activatesAkt.PTU suppressedboth phospho-Tyr¹¹⁶¹-IGF1-R/GAPDH(Figure 6G and 6H) and phospho-Ser⁴⁷³-Akt/total Akt levels (Figure 6I) indicating the involvement of T3 in activating the IGF1-R/Akt pathway in P15 ventricles.

Figure 6. Increased Expression of IGF-1 mRNA and IGF-1 in P15 Ventricles Causes IGF1-R and Akt Phosphorylation that is inhibited by PTU.

(A–F) Ventricular IGF-1 mRNA levels and IGF-1 immunostaining (IGF-1, red and DAPI, blue staining, B–E; quantitation, F) at P10, P14, P15 and P16, showing significant increases from P10 to a peak at P15.

(G–I) Immunoblotting (G) of ventricular myocardium obtained from P10 and P15 mice, and quantitation showing significant increases in the P-IGF1-R- (H), and P-Akt-to-total Akt ratios (I)between P10 and P15 and suppression of these increases at P15 by PTU treatment. Total Akt/GAPDH levels did not significantly change between P10 and P15 (I).

(J–K) Cytosolic Akt localization in P10 LV sections (red) (J). In the presence of αMHC staining (green) the cytoplasm appears yellow due to co-localization of Akt with αMHC (K).

(L–M) Nuclear Akt localization in P15 LV sections is evident from the purple staining (white arrows), which results from the co-localization of Akt (red) with DAPI (blue), and isalso shown in the adjacent magnified panel (L);nuclear localization of Akt (purple; white arrowsremains unchanged with staining of cytoplasmically-localized αMHC (green) (M).

(N) Quantitation of Akt⁺ CM nuclei shows few if any cells in P10 right ventricle (RV), in the endocardium (Endo) and epicardium (Epi) of P10 LV, and in P15 RV. In P15 LV, Akt⁺ CM nuclei are readily identified but are 51-fold more abundant in endo- versus epicardium.n = 3–7 animals for each individual evaluation

Data shown are the means ± SEM. In all studies. *p<0.05; ***p<0.001.

Aktphosphorylatesmultiple substrates implicated incell survival and heart growth. In CMsnuclearlocalization ofAktis functionally important.Expression of constitutively activated Akt results in CM hypertrophy (Condorelli et al., 2002), butnuclear overexpression increasesCMnumbers (Rota et al., 2005). Akt was mostly localized to CM nuclei in P15 hearts(Figures 6J–6N); in P10 hearts its localization was cytoplasmic.Consistent with the distribution of mitotic CMs (Figures 3E and 3F), Akt⁺CM nuclei were predominantly in thesubendocardium of the P15 LV wall(Figure 6N) and were not detected in the RV. Collectively, these findings support a role for T3 in triggering the P15 CM hyperplastic burst.

Cardiac Function after Myocardial Infarction in Neonatal and Early Postnatal Hearts

Extensive cardiac regeneration after injury, leading to restoration of the ventricular wall without scar formation, is observed in P1 but not P7 or P14 mice (Porrello et al., 2011, 2013). This is consistent with the prevailing view that CMs permanently exit the cell cycle by the end of the neonatal period (at ~P5). However, prominent CM proliferation at P15, as shown above, might allow cardiac regeneration. We thusevaluated the early regenerative potential of P15 mouse hearts, by subjecting them to myocardial infarction and comparing their response to those of P2 and P21 animals. Figure 7A shows examples of P2- and P15-injured hearts at 7 or 21days post infarction (dpi). Despiteno significant differences in infarct sizes at 1-dpi (Figure 7B), P21 infarct sizes were 6.8-fold greater(p < 0.001) at 7dpi than P2 infarct sizes, while infarct size was intermediate in P15 hearts; infarct size was also 2.4-fold greater (p < 0.01) at 7-dpi in P21 versus P15 hearts (Figure 7B). We next compared levels of DNA synthesis in CMs within the remote (non-ischemic) and border zones,as compared to control regions in age-matched sham-operated mice. We gaveBrdU intraperitoneally at 1- and 3-dpi,and determined the number of BrdU⁺ CMs in LV myocardium at 7-dpiby immunohistochemistry andconfocal microscopy. There was robust cell cycle activity throughout the LV myocardium in both sham-operated and infarcted P2 hearts, but ~10-fold lower levels of cell cycle activity in the border zone of P15 mice (versus that in injured P2 hearts; P < 0.05) (Figure 7C). BrdU⁺ CMs were virtually undetectable in the border zone of infarcted P21 hearts (Figure 7B). The detrimental impact of infarction on LV function (LV fractional shortening and ejection fraction; Figures 7D and 7E, respectively) and LV wall thinning (Figure 7F) was not evident in P2 hearts at 7-dpi, as may be expected with extensive regenerative repair (Figures 7A and6C); it was intermediate at P15, and severe at P21 (Figures7D and 7E). These differences in functional and morphological outcomes at 7-dpi in P2 and P15 mice were independent of the size of the initial infarct (Figure 7B). Thus, in murine hearts, the capacity for cardiac regeneration after P15 is intermediate between that at P2 and at P21.

Figure 7. Reparative Response of P2, P15 and P21 Hearts to Myocardial Infarction.

(A) Representative coronal sections of hearts from P2 and P15 mice at 7- and 21-days post myocardial infarction (MI; dpi) induced by left anterior descending coronary artery ligation. Sections are stained with picrosirius red to delineate scar tissue (arrows), and fast greento delineate viable myocardium.Note the virtual absence of scar tissue in P2-MI mice at 7 and 21dpi, compared to mice whose MI occurred at P15.

(B) Percent infarct size relative to the LVof P2, P15 and P21 mice 1- and 7-dpi determined from 2,3,5-triphenyltetrazolium chloride and picrosirius red with fast greenstained sections, respectively. Note the differences in infarct sizes at 7-dpi despite similar infarct sizes at 1-dpi.

(C) BrdU⁺/α-MHC⁺CMs in the remote (non-ischemic) and border zones of P2, P15 and P21 hearts at 7dpi, showing fewer cells with evidence of DNA synthesis (BrdU incorporation) in P15 hearts compared to P2 hearts, and the absence of such cells in P21 hearts.

(D) LV fractional shortening (FS) of P2, P15 and P21 hearts at 7-dpi or sham-operation, determined from B-mode echocardiographic views at the mid-papillary or apical (peri-infarct) levels. Note, the reduction in FS in P15 hearts and the more marked reduction in FS in P21 hearts. (E) Echocardiographically-determined LV ejection fraction (EF) of P2, P15 and P21 hearts at 7dpi or shamoperation. Consistent with the changes in FS, note the significant reduction in EF in the P21 hearts.

(F) Echocardiographically-determined LV wall thickness of P2, P15 and P21 hearts at 7dpi, expressed as a ratio of wall thickness in the peri-infarct or control regions to that at the midpapillary level. Consistent with the reductions in FS and EF, note the significant reduction in LV wall thickness in the P21 hearts.

Data shown are the means ± SEM. In all studies, n = 3–7 animals for each individual evaluation. *p<0.05; **p<0.01; ***p<0.001.

DISCUSSION

Although CM terminal differentiation has beenextensively studied, when it occurs and what role it plays in early postnatal heart maturation are fundamental questions that remain largely unanswered. Here we show that over a precisely timed interval during the third week of life, ~500,000 CMs are added to the preadolescent mouse heart. This represents an ~40% increase in CPN, with CM numbers remaining static thereafter. These new cells are derived from CMs previously believed to have lost, shortly after birth, their capacity to divide. Further, we find evidence that thyroid hormone triggers the hyperplastic burst at P15.Hence it is likely that terminal differentiation in CMs does not occur before P15: these findings question the 100-year-old concept that CMs terminally differentiate in the neonatal period (Bizzozero, 1894).

The Timing of CM Terminal Differentiation in Postnatal Murine Hearts

Terminal differentiation in mature mammalian CMs means that these cells havepermanently withdrawn from the cell cycle. Mechanisms responsible might include a marked suppression of basal levels of expression of cell cycle genes that promote cell division, and/or increased expression of genes that inhibit the cell cycle (Poolman and Brooks, 1998). When CMs are forced to reenter the cell cycle, as with forced overexpression of S phase cyclins, endoreplicative cell cycles (or endocycles) produce large polyploid cells (Soonpaa et al., 1997). Alternatively, cells that proceed past the G₂/M phase checkpoint undergo acytokinetic mitoses to generate large multinucleate cells (Soonpaa et al., 1997). However, despite prominent cell cycle checkpoints that produce this phenotype, a small number of preexisting CMs in adult hearts can undergo cell division, which over time may replacecells lost by apoptosis (Senyo et al., 2013). These may be a small subset of CMs that have retained replicative competency, or they may reflect a stochastic equilibrium between replicative competence and terminal differentiation that heavily favors the latter. This has been taken to suggest that, as murine CMs exit the neonatal period (immediately after P5), most undergo an endocycle that produces binuclear cells, whichareconsidereda defining manifestation of terminal differentiation (e.g., Walsh et al., 2010).

Here we show thatnumerous CMs reenter the cell cycleon the night of P14; this culminates in cell division during P15.Several lines of evidence suggest that new CMs are derived from preexisting CMs. First, aurora B labeling indicates that ~30% of both mono- and binuclear CMs are in mitosis on the morning of P15, and MKLP1 expression in ~4% of LV subendocardial CMs indicates impending cytokinesis. MKLP1 plays an essential role in midbody formation and completion of cytokinesis(Zhu et al., 2005), and marks imminent cytokinesis. Second, cell counting indicates that 2.6 × 10⁵ CMs are added to the cardiac ventricles between P14 and P15A. Third, consistent with cell division giving rise to small progeny, the average mono- and binuclear CM volumeabruptly decreases by ~40–60% and, consistent with extensive cytokinesis, this is associated witha loss of large mono- and binuclear CMs observed in P14 ventricles. Fourth, addition of ~500,000 CMs to the cardiac ventricles between P14 and P18 occurs with minor changesin polyploidy or acytokinetic mitosis (as assayed by multinucleation). On the morning of P15, ~90% of mitotic CMs are binucleate and, based on nuclear localization of aurora B, most appear to be in prophase. Those that are in anaphasedisplay localized disruption of sarcomeres around their nuclei, but noloss of contractile function (fractional shortening) during the CM replicative burst. Finally, there is a prominent 1-day increase in many mitosis-regulating genes in P15 heart. These considerationsindicate that murine CMs are not terminally differentiated during early preadolescence, and challenge the notion that binuclear CMs cannot divide.

We suppose that the synchronized and temporally restricted nature of the CM proliferative burst in P15 hearts hasinhibited its detection.Soonpaa et al.(1996) found near negligible levels of S phase in P10–P21-CMs using a 2h in vivo³H-thymidine labeling protocol. This approachcould miss a short synchronized proliferative burst in which S-phase starts during the night of P14 and ends before the morning of P15 (S-phase duration in CMs is not known, but it could last ~4 h in a 24h cell cycle). In support of this notion, we found extensive nuclear localization of aurora B, indicating that these CMs have just passed the G₂/M-phase transit point by the morning of P15. Walsh et al.(2010) studied CM mitosis in P14, but not P15 CMs. Relative to P15 hearts, we observed very low levels of CM mitoses in P14 or P16 hearts.

Porrello et al. (2011) report that hearts retain significant cardiac regenerative potential for a brief time after birth. The link between regenerative potential and postnatal cell cycle arrest, as assessed by CM binucleation, was suggested to indicate a brief neonatal “cardiac regenerative window” that closes before P7. We show here that CMs are not terminally differentiated at P15. To askif regenerative repair is possibleat P15, we assessed recovery from myocardial infarction in neonatal and postnatal hearts (P2, P15 and P21). Our findings indicate an intermediate phenotype in hearts injured at P15 (ejection fraction unchanged, but with a 64% depression of fractional shortening at the peri-infarct level), relative to those injured at P2, which showed a high degree of repair after injury (ejection fraction unchanged with no depression of fractional shortening at the peri-infarct level). Hearts injured at P21 (ejection fraction reduced by 40%, with ~90% depression of fractional shortening at the peri-infarct level) showed no evidence of injury-induced CM cell cycle reentry. Thisintermediate cardiac pathology in P15 injured hearts mightbe due tothe level of CM proliferation:BrdU incorporation into CMs in the border zonewas lower than that in P2-injured hearts, possibly because CMs terminally differentiate within the 7dpi follow-up period. Upregulation of the T3-triggered IGF-1/IGF1R/phospho-Akt axis at P15, which is pro-survival(Dai and Kloner, 2011), may alsohaveplayed a role in attenuating cardiac pathology in P15- versus P21-injured hearts. Finally, immune responses to injury influence cardiac regeneration and functional outcomes (Xin et al., 2013), factors that change profoundly as the neonate transitions into preadolescence (Levy, 2007).These complexities would need elucidationto enable efficient regenerative repair in preadolescent hearts.

PreadolescentHeart Growth occurs via a Distinct Form of Physiological Hypertrophy

It is axiomatic that the main purpose of heart growth during preadolescence is to increase stroke volume and cardiac output to accommodate a growth spurt that almost quadruples body weight (and hence circulatory volume) of P10 mice in just 25 days. Our studies indicate that maturational heart growth, while sharing similarities with other forms of cardiac hypertrophy, shows differences that suggest that it is a distinct form of physiological hypertrophy.

Mechanistically, LV volume overload of mitral regurgitation leads to fiber elongation, chamber enlargement, and eccentric hypertrophy (Grossman et al, 1975). Moreover,an increase in LV sphericity and a decrease in h/Ri ratio causean increase in LV end-diastolic wall stress.By contrast, increases in LV chamber volume during endurance exercise are accommodated byelliptical remodeling of the heart (Schiros et al., 2013). This limits the increase in LVEDD and, hence,based on LaPlace's law (T = P × R_i/h, where T is the tension or wall stress, and P is the pressure difference across the LV wall), wall stress and myocardial oxygen demand do not need to increase as much as with an equivalent increase in LV volume due to spherical remodeling.We find that LV chamber sphericity increases,and the h/R_i ratio decreases (by ~50%), duringmaturational heart growth. Thus, LV wall stress must increase progressively during maturational growth. Despite this, stroke volume increases, most likely due to an increase in α-M□□expression and to CM proliferation. Maturational heart growth thuscombines CM proliferation with an increase in α-MHC, which correlates directly with overall cardiac performance (Krenz and Robbins, 2004), thus indicating a distinctform of cardiac remodeling.

Hyperplastic Heart Growth of Preadolescence involves Hemodynamic–endocrine Interplay

We observed that preadolescentCM hyperplasia was abrogated by inhibition of T3 biosynthesis. Moreover, T3 promoted cell cycle reentry of cultured P14 CMs. While these findings are consistent with T3 being both necessary and sufficient for the preadolescent proliferative burst, the fact that the spike in T3 occurs well before proliferation begins, as well as the uneven but non-random distribution of mitotic CMs within the LV wall, suggests a greater level of complexity. T3 also causes CMs to elongate (Pantos et al., 2007), which is expected to increase LVEDD and hence wall stress. We showedthat LVEDD increased markedly between P10 and P14. During isovolumic LV contraction, wall stress is higher in subendocardial than in subepicardial regions (Mirsky, 1973). And, as shown in fetal (Saiki et al., 1997) and early neonatal CMs (Sedmera et al., 2003), wall stress stimulates CM replication. Thus, we speculate that the preadolescent CM proliferative burst involves an orchestrated interplay between a permissive increase in T3 level and a regional increase in wall stress.

Theseunique features of maturational heart growth are likely to be important for maintaining cardiac performance during a critical period of body growth, which is unsupported by maternal/placental nutrient and oxygen supply and involves increasing circulatorypressures. Furthermore, the finding that CM proliferative capacity is sustained well beyond the neonatal period may have important clinical implications. For example, congenital hypoplastic heart syndromes are often treated with palliative surgery and eventually cardiac transplantation. If human CMs, like those of the mouse, continue to divide efficiently during preadolescence, this may provide a basis for stimulating cardiomyogenesis in young patients by the administration of T3. Indeed, a therapeutic benefit of T3 in the postoperative period has been reported in newborn children undergoing operations for complex congenital heart diseases (Chowdhury et al. 2001).

EXPERIMENTAL PROCEDURES

Animals

To minimize size variation only male mice (C57BL/6J) from litters of 6–7 were used, unless otherwise indicated. To avoid potential inaccuracies in the timing of birth, only animals with confirmed birth dates were used. Animals were handled according to Emory University or Victor Chang Cardiac Research Institute Institutional Animal Care and Use Committee Guidelines.

Echocardiography

Echocardiography was performed under 2% isofluorane anesthesia as described (Li et al., 2008). SeeExtended Experimental Procedures.

Ventricular Myocardial Wall Volume

Ventricular volumes were determined from ventricular weights and the known specific gravity of muscle. See Extended Experimental Procedures for details.

Myocardial Infarction and Infarct Size Determination

Mice were anesthetized and subjected to LAD ligation via left thoracotomy. Infarct sizes were determined on triphenyl tetrazolium chloride (1 dpi) or picrosirius red/fast green stained (7 dpi) sections, using Image J software (NIH) or scanning by Power Mosaic microscopy, respectively. See Extended Experimental Procedures.

CM Isolation, Nucleation, Ploidy, and Dimension Analyses

CMs were isolated by Langendorff perfusion and distinguishedfrom non-CMs by cytoplasmic size and phase contrast microscopy.Samples (20 μl; ~150 CMs/aliquot) were loaded in the counting chamber with a wide-bore pipette and counted in triplicateusing a hemocytometer. After purificationCMs werestained with 4',6-Diamidino-2-phenylindole (DAPI)for nucleation counts, which were plotted as percentage of counted CMs using procedures described by us previously (Li et al., 2008).CM dimensions were measured using NIS-Elements 3.0 software (Nikon).See Extended Experimental Procedures.

Immunofluoresence

CMs were fixed and subjected to immunohistochemical evaluation, as described(Li et al., 2008).See Extended Experimental Procedures.

Quantitative RT-PCR and Western Blot Analyses

RNA prepared from snap frozen heart ventricles was subject to qRT-PCR using standard techniques. See Extended Experimental Procedures for details. Proteins extracted from snap frozen hearts were fractionated by SDS-PAGE and subjected to immunoblotting as described (Li et al., 2008). See Extended Experimental Procedures.

CM Culture and Cardiac Stem Cell Analyses

CM isolated from P14 hearts were cultured and subjected to T3 and BrdU treatment, and cardiac stem cells determined by flow cytometer analysis after staining for c-kit and Nkx2-5 as detailed in Extended Experimental Procedures.

Serum T3 Measurement

Serum T3 levels were determined with a T3 ELISA kit (Alpha Diagnostic International).

Statistics

Data are presented as the mean ± standard error of the mean (SEM). Statistical differences of continuous variables were determined by ANOVA followed by post hoc Bonferroni or Dunnet's tests or by Student's t test. GraphPad Prism software was used for statistical analysis. p < 0.05 was considered significant.

• Murine cardiomyocytes continue to proliferate well after the neonatal period

• A proliferative burst in preadolescence establishes the final cardiomyocyte number

• Proliferation is initiated by the thyroid hormone/IGF-1/IGF1-R/Akt pathway

• Persistence of cardiomyocyte proliferation improves outcomes after cardiac injury