Transient Cell Cycle Induction in Cardiomyocytes to Treat Ischemic Heart Failure

The regenerative capacity of the heart after myocardial infarction (MI) is limited. Our previous study showed that ectopic introduction of Cdk1/CyclinB1 and Cdk4/CyclinD1 complexes (4F) promotes cardiomyocyte proliferation in 15-20% of infected cardiomyocytes in vitro and in vivo and improves cardiac function after MI. Here, we aim to identify the necessary reprogramming stages during the forced cardiomyocyte proliferation with 4F on a single cell basis. Also, we aim to start the rst preclinical testing to introduce 4F gene therapy as a candidate for the treatment of ischemia-induced heart failure. Temporal bulk and single-cell RNAseq and further biochemical validations of mature hiPS-CMs treated with either LacZ or 4F adenoviruses revealed full cell cycle reprogramming in 15% of the cardiomyocyte population after 48 h post-infection with 4F, which was associated with sarcomere disassembly and metabolic reprogramming. Transient overexpression of 4F, specically in cardiomyocytes, was achieved using a polycistronic non-integrating lentivirus (NIL) encoding the 4F; each is driven by a TNNT2 promoter (TNNT2-4F-NIL). TNNT2-4F-NIL or control virus was injected intramyocardially one week after MI in rats or pigs. TNNT2-4F-NIL treated animals showed signicant improvement in left ventricular ejection fraction and scar size compared with the control virus treated animals four weeks post-injection. In conclusion, the present study provides a mechanistic demonstration of the process of forced cardiomyocyte proliferation and advances the clinical feasibility of this approach by minimizing the oncogenic potential of the cell cycle factors using a novel transient and cardiomyocyte-specic viral construct. rat injection, echocardiography; M.S. and K.M.K.: echocardiography and MRI analyses; Y.G., Y.H., and Y.N.: Mouse surgery, modRNA, and viral injection. L.M., P.K.L., and B.G.H.: Metabolic analysis; K.C., and R.T.: Bioinformatics analyses; B.M.A. and J.S.: Electrophysiology analysis; H.R.J., A.S., Z.I., A.M.S., and S.H.: histology and analyses including staining, imaging, and quantications. A.S.E.: MRI imaging quantication; modRNA Design and supervision of rat and pig


Introduction
The mammalian cell cycle involves numerous feedback loops that either permit or prevent cell division depending on the cell type 1,2 . Although fetal myocytes proliferate to achieve cardiac growth and tissuespeci c stem cells undergo cytokinesis postnatally, differentiated cells typically become post-mitotic and permanently exit the cell cycle 3 . As a result, the regenerative capacity of most organs, including the heart, is limited. The ability to control proliferation in the postnatal heart would represent a powerful approach to promote cardiac repair after infarction.
Recently, we took a combinatorial approach to screen for factors and conditions that could recapitulate the fetal state of cardiomyocyte division. We found that ectopic introduction of the Cdk1/CyclinB1 and the Cdk4/CyclinD1 complexes (4F, i.e., four factors) promoted cell division in at least 15% of mouse and human cardiomyocytes in vitro 4 . Rigorous assessment of cardiomyocyte cytokinesis in vivo using the Cre-recombinase dependent Mosaic Analysis with Double Markers (MADM) 5 lineage tracing system revealed similar e ciency in mouse hearts, leading to cardiac regeneration upon delivery of the 4F one week after myocardial infarction 4 . Moreover, in vitro and in vivo results show that myocytes undergo only one round of division after transduction with these cell cycle factors because the overexpression of the 4F in cardiomyocytes is self-limiting through proteasome-mediated degradation of the protein products 4 . Interestingly, a recent study showed that AAV-mediated expression of microRNA-199a in infarcted pig hearts could initially stimulate cardiac repair through induction of cardiomyocyte proliferation; however, subsequent persistent and uncontrolled expression of the microRNA resulted in sudden arrhythmic death of most of the treated pigs 6 . Our approach is currently one of the most robust methods for inducing cardiomyocyte proliferation; however, the timing, dosage, and speci city of this therapy's expression must be tightly controlled. Therefore, there is a need to transiently and speci cally express 4F in cardiomyocytes to induce one cycle of proliferation to avoid any adverse effects in the heart and other tissues if escaped.
Uncertainty regarding the mechanisms underlying the functional improvement observed with cell-based therapies is a signi cant limitation of these approaches and led to skepticism about their clinical applicability (reviewed in 7,8 ). Therefore, prior to starting any translational efforts, we rst sought to gain mechanistic insights during the cell cycle reprogramming at a single cell transcriptomics level in a temporal manner following overexpression of the 4F in cardiomyocytes. We were able to identify the main reprogramming steps needed for the cardiomyocytes to complete the cell cycle. These ndings provide an essential foundation that enables one to ascribe subsequent cardiac functional improvements to the generation of new cardiomyocytes. Then, we provided the rst proof-of-concept evidence for this approach's clinical applicability by minimizing any oncogenic potential of the cell cycle factors using a novel transient and cardiomyocyte-speci c viral construct in rat and pig models of heart failure.

Results
Temporal bulk and single-cell RNAseq of mature hiPS-CMs reveals that cell cycle reprogramming is associated with sarcomere disassembly and metabolic reprogramming during forced cell cycle induction Previously, we reported that 4F induces at least 15% of mouse and human cardiomyocytes in vitro and in vivo to undergo proliferation within the rst 48 h post infection 4 . This high percentage of proliferating cells within the bulk population provided con dence that temporal bulk RNAseq could identify the signi cant transcriptional reprogramming events during the progress of cardiomyocyte proliferation at a transcriptional level. Therefore, to investigate the transcriptional changes during cell cycle progression, we conducted temporal bulk RNAseq on 60-day-old mature hiPS-CMs treated with either LacZ (control) or 4F adenovirus for 24, 48,  1&2). These data suggest that cardiomyocytes need to withdraw from their primary function of contraction during the cell cycle to enter the cell cycle. These transcriptomics changes were functionally con rmed by using time-lapse impedance contractility tracing. We found that cardiomyocyte contractile force declined signi cantly during proliferation (48 h post-infection), which coincided with the appearance of arrhythmic episodes; normal contractile force and rhythm returned to baseline levels 84 h post-viral infection (Fig. 2c). Furthermore, during the G2/M phase, especially during anaphase or cytokinesis, the sarcomeric structures were disrupted, as shown by troponin-T Immunostaining of the sarcomeres (Fig. 2d). The impedance-based force measurements indicate the contractile force generated by the monolayer sheet of cardiomyocytes, which could have a mixed-signal from proliferating and nonproliferating cardiomyocytes. Therefore, we investigated the sarcomeric disassembly implication on the electrophysiological properties and ion current in cardiomyocytes during proliferation. The decrease of the Scn5a transcript suggested a possible change in resulting I Na , but the transient nature of the scn5a decrease along with the knowledge that transcript levels do not necessarily predict levels of functioning protein motivated us to assess I Na . Mean I Na density tended to increase compared to LacZ (Extended data Fig. 1a), but the difference between lacZ and the 4F 48 or 72h time points was not signi cantly different. Cell capacitance was also not different (Extended data Fig. 1b). We also assessed the voltagedependence of steady-state inactivation because there is a development shift of the inactivation midpoint (V 0.5 ) whereby V 0.5 becomes more negative with embryonic heart maturation 9 . As with current density, there was no signi cant difference caused by 4F (Extended data Fig. 1c), though there is a trend towards a more negative V 0.5 (Extended data Fig. 1d). These data suggest that 4F does not necessarily reduce excitability, and there is an unexpected tendency for larger current with more mature properties in cells treated with 4F for 48 and 72 h compared to control cardiomyocytes.
To further investigate in detail the transcriptional modi cations during cardiomyocyte proliferation at a single-cell level, we conducted a temporal single-cell RNAseq of 60-day-old mature hiPS-CMs infected with either LacZ (control) or 4F adenovirus for 24, 48, and 72 h. Gene expression data were collected from 7000 cells/condition as summarized UMAP blots (Fig. 3a). All cells were positive for cardiac markers (TNNT2, TNNC1, and MYH7) (Fig. 3b) Consistent with these transcriptomic changes, hiPS-CMs had markedly lower oxygen consumption and extracellular acidi cation rates 48 h after 4F infection, suggesting lower oxidative phosphorylation and glycolysis in proliferating myocytes (Extended Data Fig. 3a). Furthermore, in 4F-overexpressing hiPS-CMs, stable isotope tracing experiments using 13 C 6 -glucose demonstrate signi cantly higher enrichment of 13 C in intermediates or end products in NAD synthesis (NAD + ), the hexosamine biosynthetic pathway (UDP-HexNAc), phosphatidylethanolamine synthesis (CDP-ethanolamine), and pyrimidine nucleobase biosynthesis (uracil) (Extended Data Fig. 3b). These data indicate that proliferating cardiomyocytes diminish catabolic pathway activity and augment biosynthetic pathway activity before or during cell cycle progression.
Transient expression of the 4F using modi ed RNA showed low infection e ciency The insights gained into 4F-mediated cardiomyocyte cell cycle reprogramming encouraged us to standardize the approach to treat heart failure. Our previous studies using adenoviruses provided proof of principle for the e cacy of our approach 4 ; however, the use of adenovirus is not clinically applicable because of the high prevalence of immune response in humans. From our work with adenovirus, we noticed that most of the cardiomyocyte proliferation events occur within the rst 48-72 h after introducing the virus, after which cardiomyocytes shut down the protein expression of the 4F through activation of the proteasome system 4 . However, this was not the case in other cell types, such as in neurons, where the expression and proliferation capacity last for over seven days (data not shown as it is out of the context of the manuscript). Furthermore, the induction of long-term cell cycle activity in the heart may become oncogenic and have deleterious effects on the heart 6 . Therefore, we tested the transient expression of the 4F using modi ed RNA (modRNA) delivery and its ability to induce cardiomyocyte proliferation in vivo using the recently optimized delivery method for delivering modRNA to the heart using citrate sucrose buffer 10 . modRNAs were injected into the myocardium of C57BL/6 mice in citrate sucrose buffer; then, we assessed the cardiomyocyte expression of nuclear-GFP, CDK1, CCNB, CDK4, and CCND at 48 h after injection (Extended Data Fig. 4a-d). The infection e ciency was very low (<0.01%) (Extended Data Fig.  4d). To obtain proof-of-principle for the e cacy of the transient expression approach in inducing proper cardiomyocyte cytokinesis, we injected the 4F+GFP modRNAs into the heart of MADM lineage-tracing transgenic mice crossed with tamoxifen-induced alpha-MHC-Cre. Knowing that only a few cells would be informative given the small number of recombined cells expected and the low infection e ciency of the modRNA, we identi ed the site of injection by the nuclear GFP expression. We found signi cant induction of phospho-histone H3 (pHH3) in cardiomyocytes at the site of the 4F injection compared to control (Extended Data Fig. 4e), suggesting that the modRNAs can induce cell division in vivo. Also, we observed very few MADM recombination events that led to single-colored cells (10-15/heart) that co-localized with pHH3 positive nuclei, suggesting true cytokinesis in this setting (Extended Data Fig. 4f). These data are suggestive that transient expression of the 4F may be su cient to induce cell division but will require more e cient delivery and multiple injections as the infection e ciency and the spread of the modRNA at the injection site is very limited.
Characterization of cardiomyocyte-speci c non-integrating lentivirus system for gene therapy The low in vivo infection e ciency of modRNA motivated us to develop and optimize an alternate strategy for 4F delivery and transient expression using a non-integrating lentivirus (NIL). NIL is known for its high infection e ciency and transient expression of the encoded protein, which is limited to 2-3 days, followed by a signi cant decline in expression [10][11][12][13] . Several ongoing clinical trials are aiming to treat various diseases with the lentivirus gene therapy approach 14 . This rapid degradation kinetics makes NIL the optimal delivery vehicle for the 4F. Additionally, when using NIL, it is important to target only cardiomyocytes by controlling the 4F expression with the cardiac-speci c troponin-T promoter (TNNT2), which we previously optimized 15 . Polycistronic TNNT2-4F-NIL induced cardiomyocyte proliferation in vitro and in vivo Each cell cycle factor of the 4F was cloned into a lentivirus backbone under TNNT2 promoter (4F separate lentiviruses); also, all 4F were cloned in one polycistronic lentivirus backbone with each factor is driven by a TNNT2 promoter (4F polycistronic lentivirus) (Extended Data Fig. 6a). First, we assessed each cell cycle factor's protein expression four days post-infection in hiPS-CMs using western blot. TNNT2-4Fpolycistronic-NIL was signi cantly more e cient in inducing higher protein expression of all the cell cycle factors compared to the TNNT2-4Fseparate-NIL (Extended Data Fig. 6b-c). Furthermore, TNNT2-4Fpolycistronic-NIL showed 80-100% infection e ciency as assessed by immuno uorescence (Extended Data Fig. 6d-e). Four days post-infection with TNNT2-4Fpolycistronic-NIL, we found that 15-20% of the cardiomyocytes were positive for 5-ethynyl-2´-deoxyuridine (EDU), which marks new DNA synthesis and histone H3 phosphorylation (PHH3), which marks cells in the G2/M phase ( Fig. 4a-b). Furthermore, the total cell number was increased by 20-30% (Fig. 4b). Assessments 10 days post-infection showed that cell number increase and EDU labeling for the divided nuclei persisted; however, the PHH3 was abolished, indicating the transient nature of the cell cycle induction in the cardiomyocyte and the likelihood that TNNT2-4Fpolycistronic-NIL induced only one cycle of proliferation.
To test the e cacy of the TNNT2-4Fpolycistronic-NIL in inducing cardiomyocyte proliferation in vivo, we used a cardiomyocyte cytokinesis lineage-tracing animal model (inducible a-MHC-Cre::MADM-lineagetracing) 4,5,16,17 . In these lineage-tracing mice, cardiomyocytes that undergo cytokinesis produce daughter cells that are either red, green, yellow (red+green), or colorless, based on allelic recombination of uorescent reporters; if the cardiomyocytes fail to divide, they will remain double-colored (i.e., yellow), or colorless if no recombination occurs. Thus, the presence of single-colored red or green cells de nitively indicates cells that have undergone cell division, although dividing cells are underrepresented by singlecolored cells because double-colored (yellow) or colorless cells also could have divided. Animals were subjected to a 60-min occlusion of the left anterior descending artery followed by reperfusion, then oneweek later, TNNT2-4Fpolycistronic-NIL or LacZ-NIL (control) is injected intramyocardially (Fig. 4c).
Tamoxifen injection was carried out as described in 4 , starting 48 h after the virus injection for three days to initiate recombination events. Mice were sacri ced one week after the viral injections, and hearts were sectioned to enumerate the cytokinesis events. All surgeries, imaging, and microscopy analyses were blinded about treatment, and animals were decoded after all data were analyzed. The analysis was done on ten different sections from each heart across the whole myocardium. After intramyocardial injection of TNNT2-4Fpolycistronic-NIL, it showed at least 15% of the recombinant cardiomyocytes were singlecolored, compared to <1% in hearts injected with control virus (Fig. 4d-e).
TNNT2-4Fpolycistronic-NIL improves cardiac function in a rat model of heart failure Before starting in vivo functional studies, we sought to validate that TNNT2-NIL drives cardiomyocytespeci c expression of the 4F in vivo. Therefore, TNNT2-4Fpolycistronic-NIL or GFP-NIL control virus were injected intramyocardially, and the rats were sacri ced one-week post-injection. Western blotting and Immunostaining con rmed the expression of 4F in the rat hearts six days after injection (Extended Data Fig. 7a-d). Furthermore, RNA expressions of the overexpressed human CDK1, CDK4, CCNB, and CCND in rat hearts were only detected in the heart and not in the other organs six days post-viral injection (Extended Data Fig. 8a).
Then we started to test the effects of TNNT2-4Fpolycistronic-NIL on cardiac function following cardiac damage in vivo. Rats were subjected to a 2 h coronary occlusion followed by reperfusion. One week later, TNNT2-4Fpolycistronic-NIL or control TNNT2-GFP-NIL was injected into the peri-infarct region of the heart. Rats were followed for four weeks and then sacri ced, and the cardiac tissue was processed and analyzed ( Fig. 5a). Gross assessment of the heart weight showed that HW/BW was signi cantly decreased in the hearts treated with TNNT2-4Fpolycistronic-NIL compared to the control virus treated hearts (Fig. 5b). TNNT2-4Fpolycistronic-NIL-treated groups exhibited a signi cant increase in left ventricular ejection fraction four weeks post-viral injection compared to the control group, as assessed by blinded echocardiography (Fig. 5c). Consistently with the improvement in cardiac function, histological analyses of the hearts revealed an approximately 30% reduction in the scar size in hearts treated with TNNT2-4Fpolycistronic-NIL compared to control hearts (Fig. 5d). Interestingly, assessment of cell size at the border and remote zones showed a signi cant reduction in the cardiomyocyte cross-sectional area ( Fig. 5e-f). As the virus was injected in the border zone, so the reduction in cardiomyocyte cross-sectional area could be due to the induction of proliferation; however, the reduction in cardiomyocyte crosssectional area at the remote zone could be due to improvement of the overall functionality of the heart and the protection from progression towards dilatation after treatment with TNNT2-4Fpolycistronic-NIL.
Double reporter to permanently label cardiomyocytes activation of aurora kinase B in vivo in large animals To track mitotic events in vivo in large animals, we developed a new reporter system based on the Aurora kinase B (AurKB) promoter region activation. AurKB is one of the central protein kinases that ensure the proper execution and delity of mitosis and is expressed only for a short time during the cytokinesis process, localizing to the central spindle during anaphase and in the midbody during cytokinesis 18 . It has been considered as a putative marker for mitosis in several cell types, including cardiomyocytes [19][20][21] . A recent study demonstrated that AurKB correct positioning to the midbody in cardiomyocytes during mitosis is positively correlated with cytokinesis and that 70% of the neonatal cardiomyocytes that express AurKB undergo complete cytokinesis with correct midbody positioning 20 . To develop this reporter, we used the previously well-characterized 1.8kb promoter region of the human AurKB gene, which is highly conserved between species 22 , and cloned it into a lentivirus to drive the expression of GFP protein (Extended Data Fig. 9a). We will refer to this reporter as the AurKB-GFP reporter throughout the manuscript. We generated and validated this reporter to detect mitosis/cytokinesis events in proliferating cells, e.g., HEK293 cells (Extended Data Fig. 9a). Live cell imaging of hiPS-CMs overexpressing 4F over 96 h showed GFP expression during the M phase, which reaches the maximum intensity during cytokinesis; in contrast, there is no GFP expression observed in lacZ-treated cells (supplementary movie 1&2). In line with a previous report 20 , live imaging of hiPS-CMs treated with 4F indicated that 70% of the GFP expressing cells completed cytokinesis while the remaining 30% were stuck in mitosis without completing cytokinesis (Extended Data Fig. 9b and supplementary movie 1&2). Fixation and Immunostaining demonstrated that 36 h post-infection with 4F, the GFP signal is co-localized with the AurKB protein expression at the midbody during mitosis (Extended Data Fig. 9c-d). AurKB protein expression fades after two days post-infection; however, the GFP protein remained and marked that 20-30% of hiPS-CMs infected with 4F adenovirus, with a decline in the GFP signal after day 4 (Extended Data Fig. 9e-f). EDU nuclei labeling was observed in GFP positive cells (Extended Data Fig. 9e), indicating the S-phase's completion before entering the M phase.
After demonstrating the AurKB promoter region's ability to reliably indicate mitosis (100%) and cytokinesis (70%) events in cardiomyocytes through transient expression of GFP, we developed a permanent marker for mitosis to be used in vivo and in situ. To this end, we developed a double reporter system to track mitotic events in situ and in vivo based on the AurKB promoter described above. In this reporter system, we cloned a Lox-DsRed-Stop-Lox-GFP construct under CAG promoter in lentivirus; in another lentiviral construct, we cloned the Cre encoding protein sequence under the in uence of the AurKB promoter (AurKB-Cre) (Extended Data Fig. 10a). Using this double reporter system, all infected cells will become DsRed positive; when mitosis occurs, Cre recombinase will be expressed and will cut out the DsRed-Stop sequence, turning these cardiomyocytes permanently into GFP positive cardiomyocytes. Therefore, the presence of green cells will be an indication of mitotic events. We rst validated the color switch in normally proliferating cells (HEK293) (Extended Data Fig. 10b). Then, we further validated the e ciency of this reporter system in detecting mitotic events induced by 4F in hiPS-CMs (Extended Data Fig. 10c-d) and pig heart tissue in situ using our recently developed heart slices culture system 23,24 (Extended Data Fig. 10e-f). This reporter system indicates the number of infected cells (total red-and green-labeled cells) and the number of mitotic events (green-labeled cells). Therefore, the quotient of green and red cardiomyocytes provides a quanti cation of mitotic cardiomyocytes.
TNNT2-4Fpolyscistronic-NIL induces cardiomyocyte proliferation, improves cardiac function, and reduces scar size in a porcine model of heart failure As a proof of concept for our approach's e cacy in inducing transient cardiomyocyte proliferation in large animals, we injected TNNT2-4Fpolyscistronic-NIL or control LacZ-NIL into the pig myocardium one week after induction of myocardial infarction with a 90-min coronary occlusion followed by reperfusion (Fig. 6a). The double reporter system was co-injected into the border zone with the therapeutic or control virus to assess the extent of cardiomyocyte proliferation induced by the TNNT2-4Fpolyscistronic-NIL.
Four weeks post-treatment, every pig treated with TNNT2-4Fpolyscistronic-NIL showed signi cant improvement in gross heart failure measures such as HW-BW and LW/BW (Fig. 6b). Furthermore, the cardiac functional parameter, ejection fraction, assessed by blinded echocardiography (Fig. 6c) and MRI (Fig. 6d, Extended Data Fig. 11 and supplementary movies 3-6) demonstrated signi cant improvement in animals treated with TNNT2-4Fpolyscistronic-NIL compared to control virus-treated pigs. Also, TNNT2-4Fpolyscistronic-NIL treated pigs exhibited a 25% reduction in scar size compared to control pigs (Fig.  6e).
TNNT2-4Fpolyscistronic-NIL-treated animals showed that 30% of the total labeled cardiomyocytes with the double reporter system at the injection site are GFP positive, indicating the cardiomyocyte mitotic activation. In contrast, almost no background proliferation was detected in control virus-treated hearts ( Fig. 6f-h), supporting the concept that the improvement in function is due to induction of cardiomyocyte proliferation.

Discussion
Direct induction of the cell cycle using 4F is one of the most robust methods in inducing cardiomyocyte proliferation 25,26 ; however, it is essential to understand the mechanism of action of this potential heart failure gene therapy and to tightly control its timing, dosage, and speci city of expression in cardiomyocytes. Here, we describe the essential processes associated with forced cardiomyocyte proliferation following cell cycle induction, including sarcomeric disassembly and metabolic reprogramming, in a temporal sequence and on a single cell transcriptomic level. Furthermore, we provide the rst proof of concept for this approach's e cacy in improving cardiac function after infarction in a large animal model using a transient and cardiac-speci c gene therapy approach.
Understanding the process of human cardiomyocyte proliferation and the reprogramming steps needed for the cardiomyocytes to complete the process is essential to advance the eld of cardiac regeneration.
Several efforts comparing proliferating fetal cardiomyocytes and adult cardiomyocytes have yielded a certain degree of understanding of the process 25 . However, the highly variable nature of fetal and adult cardiomyocytes limits the ability to elucidate the reprogramming events during proliferation. In the present study, we attempted to identify the essential reprogramming events associated with cell cycle induction by monitoring the same human cardiomyocytes during proliferation on a single-cell transcriptomic level. First, we found that sarcomeric disassembly is an essential step for cardiomyocytes cytokinesis. This nding is consistent with the recent suggestion that proteins responsible for sarcomere assembly, e.g., ephrin-B1, are essential elements for the cell cycle blockade in adult cardiomyocytes as described in a recent preprint 27 .
Furthermore, we demonstrated that proliferating cardiomyocytes shift their metabolism from energy production to biosynthesis. This nding is consistent with the need for new building blocks for cell growth and division. Recent studies suggest that glycolysis 28,29 , glucose oxidation 30 , and the mevalonate pathway 31 in uence myocyte proliferation; our results build upon these previous reports and support the idea that metabolic activity changes may be required for successful myocyte division. Interestingly, a recent paper demonstrated switching the metabolic substrate from fatty acids to glucose induces cardiomyocyte cell-cycle progression 30 . Considering that glucose is a primary biosynthetic substrate in cardiomyocytes 32 , this supports our nding that the switch from catabolic to biosynthetic activities is essential for cell cycle progression in cardiomyocytes.
The advantage of these single-cell RNAseq data is that we reached a time resolution that enabled us to compare two subpopulations of cardiomyocytes, both of which received 4F for 48 hours; one subpopulation was delayed entering mitosis while the other subpopulation was in mitosis. This comparison will impact the eld of cardiac regeneration and has been long-awaited. This comparison was not possible before because there was no approach to induce cardiomyocyte regeneration that reached the e ciency we achieved with the 4F. Furthermore, direct induction of cell cycle is a clean approach to understand the mechanism of cardiomyocyte proliferation, unlike other approaches that have many off-target effects, e.g., manipulations of YAP 33 , a developmental gene that induces dedifferentiation, and use of microRNAs 6 , which have multiple off-target effects.
The insights we gained into the process of cardiomyocyte proliferation and the ability to de ne the 15% cardiomyocyte subpopulation that temporarily undergoes mitosis at 48 h (mitotic subpopulation) and track the reprogramming events in this subpopulation motivated us to perform the next translational steps: that is to provide a proof of concept for the e cacy of this approach in improving cardiac function in animal models of heart failure. Interestingly, a recent study showed that AAV-mediated expression of microRNA-199a in infarcted pig hearts initially stimulates cardiac repair through induction of cardiomyocyte proliferation; however, subsequent persistent and uncontrolled expression of the microRNA resulted in sudden arrhythmic death of most of the treated pigs 6 . Our previous in vitro and in vivo results show that myocytes undergo only one round of division after transduction with cell cycle factors because the overexpression of the 4F in cardiomyocytes is self-limiting through proteasomemediated degradation of the protein products 4 . Therefore, to perform these translational steps, we needed to transiently and speci cally express 4F in cardiomyocytes to induce one cycle of proliferation and avoid any adverse effects in other tissues. Over the past decade, there have been signi cant advances in genetherapy delivery approaches for transient gene expression using either ModRNA 34 or NIL 13 . Although the modi ed RNA approach is a promising virus-free delivery system for transient expression, its poor delivery and speci city to cardiomyocytes limit its applicability. Therefore, NIL was the tool of choice for our animal studies for transient expression with high infection e ciency, as reviewed in 35 . Our data show that a polycistronic TNNT2-4Fpolycictronic-NIL induces 4F expression only in cardiomyocytes both in vitro and in vivo. Furthermore, TNNT2-4Fpolycistronic-NIL induces proliferation markers in hiPS-CMs, cardiomyocyte cytokinesis in vivo in MADM mice, and signi cantly improves cardiac function in both a rat and a pig of heart failure caused by myocardial infarction.
In conclusion, we have provided a mechanistic understanding of the process of forced cardiomyocyte proliferation and advanced the clinical feasibility of the 4F gene therapy approach for heart failure treatment by minimizing the oncogenic potential of the cell cycle factors using a novel transient and cardiomyocyte-speci c viral construct. Further studies are needed to prove effectiveness and safety in more chronic heart failure models in large animals. These studies will pave the way for the rst test of this promising approach in patients with ischemic cardiomyopathy.

Limitations of the study
The use of 60-day mature hiPS-CMs instead of adult primary human cardiomyocytes in the mechanistic studies is a limitation. However, there is a lack of a reliable long-term culture model of adult human cardiomyocytes and the inability to perform single-cell RNAseq on adult cardiomyocytes other than single nuclear RNAseq due to the large size of the adult cardiomyocytes. More importantly, nuclei are at different integrities during the cell cycle stages, which will lead to di culties in isolating mitotic nuclei and will obfuscate the interpretation of any nuclear RNAseq. Therefore, we used the hiPS-CMs, which are highly pure cardiomyocyte cultures obtained from Cellular Dynamics, Inc. These cells are selected after differentiation using an a-MHC-Blastocidin selection cassette. This strategy yields nearly 100% pure iPSC-CMs, as indicated in our TNNT2 immunostaining images and single-cell RNAseq data. For consistency, only cells that mature for at least 60 days were used for experiments. After this time, the cells have matured to a point at which they have minimal proliferation capacity and minimal basal expression of cyclins and Cdks 4 . Thus, the use of hiPS-CMs provided a homogenous cell population as a starting material to track the reprogramming events during cardiomyocyte proliferation.
The data from the AurKB must be interpreted cautiously as mitotic rather than cytokinesis events. As we described here, there is a 30% overestimation in cytokinesis events reported by this reporter. Nevertheless, as we show here, the reporter reliably estimates cell cycle induction and mitotic entrance in cardiomyocytes in vitro, in situ, and in vivo in large animal models with almost no background labeling in the control groups. We preferred to use the CAG promoter for this reporter rather than TNNT2 due to the delayed kinetics of the TNNT2 promoter (Extended Data Fig. 5b-c), which could complicate the experimental design and interpretation of results.
All functional e cacy and speci city of expression have been assessed four weeks after injection in an acute heart failure model where the virus was injected one-week post-I/R. Therefore, an assessment of the safety and functional e cacy of the TNNT2-4Fpolycystronic-NIL for a longer time (4-6 months) will be needed. Besides, there is a need for further studies to assess the e cacy of the TNNT2-4Fpolycystronic-NIL in a more chronic heart failure model where the virus is injected one or two months after I/R.

Methods
Cloning and preparation of integrating and non-integrating lentivirus.  Here, we rst ltered out any genes without at least two samples with a C.P.M. (counts per million) between 0.5 and 5000. C.P.M.s below 0.5 indicate nondetectable gene expression, and C.P.M.s above 5000 are typically only seen in mitochondrial genes. If these high-expression genes were not excluded, their counts would disproportionately affect the normalization. After excluding these genes, we renormalized the remaining ones using "calcNormFactor" in edgeR, then calculated P-values for each gene with differential expression between samples using edgeR's assumed negative-binomial distribution of gene expression. We calculated the false discovery rates (FDRs) for each P-value with the Benjamini-Hochberg method based on the built-in R function "p.adjust". Genomics) with default parameters. The counts" matrices across the samples were aggregated using cell ranger aggr. The resulting les were processed in R using the package Seurat (version 3.1.3) 36 . All cells with at least three detected genes and less than 30% of reads from mitochondrial genes and all detected genes in at least 200 cells were used in the further analyses. The remaining data were normalized using the "LogNormaliz" method. Principal Component Analysis for the subset of the 2000 most variable genes (Seurat function FindVariableFeatures) was then performed on the scaled data. The cells were clustered using the Louvain Algorithm with the resolution parameter value of 0.5 (Seurat function FindClusters) after determining the shared nearest neighbor graph using the rst ten principal components (Seurat function FindNeighbors). The data were visualized using the UMAP algorithm with the rst ten principal components as input (Seurat function RunUMAP). The cells were grouped into ten clusters based on the distribution of expression of the cell cycle genes of interest. Differential analysis between all pairs of clusters was performed using the Wilcoxon rank-sum test to identify the differentially expressed genes (Seurat function FindMarkers). The dimensionality reduction results were reformatted for compatibility with the learn_graph function in the R package monocle3, used for trajectory analysis 37 . This analysis was done for ve groups of cells -24 h unique cluster, 48 h pre-mitotic cells, 48 h mitotic cells, 72 h unique cluster, and 48 h quiescent cells.

Metabolic ux assessment
The bioenergetics of hiPS-CMs were measured using a Seahorse Bioscience XF96e Flux Analyzer. For these experiments, the assay medium consisted of unbuffered phenol red-free DMEM pH 7.4, supplemented with 5 mM glucose, 1 mM glutamine, and 100 µM L-carnitine, 100 nM insulin, and 100 µM BSA-palmitate. Following microplate insertion, the XF96e automated protocol consisted of a 12 min delay followed by baseline oxygen consumption rate (OCR) and extracellular acidi cation rate (ECAR) measurements. All experiments were conducted at 37°C. Data were normalized to the protein content.

Stable isotope-resolved metabolomics (SIRM)
hiPS-CMs were incubated in growth medium containing 5.5 mM 13 C 6 -glucose in 6-well plates for 8 h, after which cell reactions were quenched in cold acetonitrile, and extracted in acetonitrile: water: chloroform (v/v/v, 2:1.5:1), and processed as described previously for metabolite assessments using mass spectropmetry [38][39][40][41][42] . Stable isotope data analyses were performed by obtaining the mass spectrometer .raw les, which are rst converted to .mzML format with msConvert tool, a part of an open-source ProteoWizard suite, described in detail by Adusumilli and Mallick 43 . Isotopologue peak deconvolution and assignments were performed using El-MAVEN. Peaks were assigned using a metabolite library rst generated and veri ed using full-scan MS and MS/MS spectra of unlabeled samples, as described previously 41,42,44 . The library contained metabolite names and corresponding molecular formulae used to generate theoretical m/z values for all possible isotopologues and retention times. The El-MAVEN parameters for compound library matching were as follows: EIC Extraction Window ± 7 ppm; Match Retention Time ± 0.60 min. For 13 C isotopologue peak detection, the software criteria were set as follows: Minimum Isotope-parent correlation 0.5; Isotope is within seven scans of the parent; Abundance threshold 1.0; Maximum Error To Natural Abundance 100%. All assignments were visually inspected and compared with unlabeled samples for reference. Any peak groups assigned in error, e.g., not present or having different retention times than in the unlabeled samples, were deleted, and correct peak assignments were added manually, as described in 42 . Finally, the peak list with corresponding abundances was exported to a comma-separated (CSV) le and uploaded to the Polly work ow to perform natural abundance correction using Polly Isocorrect. Finally, the data were analyzed and plotted with GraphPad Prism 8.0 (GraphPad Software, San Diego, Ca, U.S.A.).

Immunocytochemistry and immunohistochemistry
The hiPS-CMs were xed in 4% formaldehyde for 20 min (Thermos Scienti c Cat#28908).  Table 1 showed a list of primary and secondary antibodies used in this study. Cells/ tissue sections were then washed three times with PBS and stained with DAPI 1µg/ml (Biotium Cat# 40043) to stain the nucleolus blue. For EDU detection, the cells were also treated with 5µM 5-ethyl-2-deoxyuridine (EDU) for the course of the experiment, which will incorporate into the newly synthesized DNA. After xation, permeabilization, and blocking of the cells/ tissue sections, the EDU incorporation was visualized using Click it EDU-Alexa-Flour 647 imaging kit (Thermo Fisher Cat# C10340). For live-cell imaging, the cells were treated with NucBlue live cells stain (Thermo Fisher ) for 20 minutes.
Imaging was conducted for the whole well using the high content imaging instrument, Cytation 1. The percentage of co-localization of PHH3, EDU, GFP, or gene expression and Troponin-T was quanti ed using Gen 5.05 software.

Animal experiments
Animal studies were performed following the University of Louisville animal use guidelines, and the protocols were approved by the Institutional Animal Care and Use Committee (IACUC) and were accredited by the Association for Assessment and Accreditation of Laboratory Animal Care.

MADM mice experiment
For lineage tracing, we used mosaic analysis with double markers (MADM) transgenic mice were developed as prescribed in 4 . All the surgeries were performed as described in 45,46 . Adult (about 12 weeks old) female MADM mice were anesthetized with sodium pentobarbital (60 mg/kg i.p.). After opening the chest through a left thoracotomy, a nontraumatic balloon occluder was implanted around the mid-left anterior descending coronary artery (L.A.D.) using an 8-0 nylon suture. Myocardial infarction was produced by 60-min coronary ischemia, followed by reperfusion (I/R). Rectal temperature was carefully monitored and maintained around 37°C throughout the experiment. Successful performance of coronary occlusion and reperfusion was veri ed by visual inspection and by observing S-T segment elevation and widening of the QRS on the electrocardiogram during ischemia and their resolution after reperfusion.
Seven days after I/R, mice were re-anesthetized with sodium pentobarbital, 60 mg/kg I.P. and the chest reopened through a central thoracotomy. The mice were randomly selected to be injected with 20 ul of TNNT2-4Fpolycistronic-NIL, TNNT2-LacZ-NIL virus intramyocardially using a 30-gauge needle. The injections were made at the border between infarcted and non-infarcted myocardium as two injections 10 µL each (2x10 7 transducing units (T.U.) per mouse heart). Forty-eight hours after injection, mice received

Rat experiments
All surgeries were performed as described in 47,48 . Brie y, Female Fischer 344 (F344) rats at the age ranging from 8-12 weeks were anesthetized with ketamine (37 mg/kg) and xylazine (5 mg/kg), intubated, and ventilated with a rodent respirator. Anesthesia was maintained with 1% iso urane inhalation, and body temperature was kept at 37°C with a heating pad. All rats underwent a 2 h occlusion of the left anterior descending coronary artery, followed by reperfusion. Seven days after MI, echocardiography was performed to ensure the development of MI. All rats in this study had EF drop > 20 points from baseline.
Rats were randomized into two groups (TNNT2-GFP-NIL, TNNT2-4Fpolycistronic-NIL). Rats were reanesthetized with ketamine/xylazine, intubated, and ventilated. The chest was reopened to expose the heart. Viral vectors (1x10 8 T.U. per rat heart in 100 µl PBS) were injected into the left ventricle along the infarct border at ve sites (20 µl/site) using a 30G needle. The rat surgeon was blinded to whether 4F or control non-integrated lentivirus was administered in each animal.
Cardiac function was assessed by serial echocardiography at baseline (before MI), one week after MI (before virus injection), and then four weeks after virus injection. Animals were anesthetized lightly with iso urane, placed on the imaging table in the supine position, and prepared for imaging using the Vevo 2100 Imaging System (Visual Sonics) equipped with a 25-MHz transducer. Parasternal longitudinal axis images were acquired and analyzed by LV trace using the Vevo LAB 3.2.6 to obtain the LV functional parameters, including the end-diastolic and end-systolic area, volume, stroke volume, fractional shortening, and ejection fraction. Imaging and calculations were done by an individual who was blinded to the treatment, and the code was broken after all data acquired.
At the end of the experiments (5 weeks after MI), animals were sacri ced, and their hearts were harvested for histological studies. The frozen hearts were sectioned longitudinally into 400-500 sections 8um thickness (take a section to throw one add two sections per slide), and one slide for every ten slides (20-25 slide per animal) were stained with Standard Masson's Trichrome staining to determine scar size. The stained sections were imaged using the Keyence BZ9000 imaging system (4X magni cation). Image J software was used to measure the scar area (blue) and healthy area (red) on longitudinal sections. Individuals assessing scar area were blinded to the treatment applied in each animal.

Pig experiments
All surgeries were performed as described in 49 . Yorkshire pigs weighing 25-35 kg received 200 mg amiodarone orally daily for seven days pre-operatively. Pigs were premedicated with an intramuscular injection of a solution containing ketamine hydrochloride and xylazine.
Pigs were injected with a dose of buprenorphine S.R. before the procedure. To create myocardial infarction, the right neck's skin was cut to make a small opening, allowing access to the right carotid artery. A 7-8F fast-cath sheath was introduced into the carotid artery. The pig was injected with Heparin (300 units/kg I.V.) to prevent clotting of the sheaths and catheters during the procedure. After intubation, the pig was mechanically ventilated. Anesthesia was maintained with iso urane. Body temperature was monitored continuously with a rectal probe attached to a thermocouple and maintained within physiology range using a veterinary blanket. The pigs were subjected to 5 minutes of stabilization, followed by baseline hemodynamics and echocardiography.
A 6-7F Hockey-stick catheter was guided to the left main coronary artery under uoroscopy as following: the catheter engaged the left main coronary ostium, and an angioplasty-type balloon catheter and guidewire assembly were uoroscopically guided into the L.A.D. Then, the wire was advanced into the distal L.A.D., and an appropriate balloon catheter was telescoped over the wire and positioned above the rst diagonal branch (the entire L.A.D. territory was included for occlusion). The balloon's placement will be veri ed by intracoronary contrast dye injection (Contrast media) and documented by cine angiogram before in ation. The balloon was in ated to occlude the L.A.D. and the L.A.D. occlusion was maintained for 90 minutes to produce myocardial infarction, targeting an infarct size of >50% of the area at risk. In ation and position of the balloon were veri ed by contrast angiogram again at the end of ischemia. If necessary, the balloon will be repositioned, and such "positional re-in ation" will be limited to less than 20 seconds to avoid any preconditioning. Once the balloon is in ated, external de brillator pads were placed on the pig's chest for "hands-free" cardioversion if ventricular brillation occurs using a bipolar de brillator (HP Codemaster XL+) at 300 Joules. After the 90 min ischemic period, the intracoronary balloon was de ated to initiate reperfusion. After the procedure of myocardial infarction, the balloon catheter was withdrawn, and a cine angiogram was taken to document the wide open of the L.A.D. artery.
After withdrawing the Hockey-stick guide catheter, the arterial sheath catheter was removed, and the arterial was repaired by anastomosis. The skin incision was closed in 3 layers using 3-0 Vicryl for internal sutures and 3-0 P.D.S. for the nal subcutaneous layer. The pig was weaned from anesthesia, and the animal was extubated when appropriate and allowed to recover. Animals received antimicrobial therapy cetiofur pre-operatively and every 24 hours for the rst 48 hours postoperatively. Animals were prepped and draped in a routine sterile fashion. 5% dextrose and normal saline was continuously infused during the procedures Seven days after the MI procedure, pigs were re-anesthetized as described above. Pigs were subjected to MRI scans and echocardiography. Anesthesia was maintained with iso urane. Body temperature was monitored continuously with a rectal probe attached to a thermocouple and maintained within physiology range using a veterinary blanket. Animals will be prepped and draped in a routine sterile fashion. A dose of buprenorphine S.R. will be given before the procedure. The chest was opened through the initial skin incision and continued down to the sternum. To avoid a post-surgical pulmonary complication, a midline sternotomy to expose the heart in the intrapleural space without breaking the pleural membrane was performed with extra care to avoid rapture of the pleural membrane intact during the opening of the mediastinum. Then the heart suspended in a pericardial cradle as described in 50,51 .
In the case of the pleural membrane broken, efforts are made to close the tear by re-approximation of the pleural membrane with 6-0 Prolene and to reestablish a negative pressure in the pleural cavity using a withdrawal tube with a purse-string closure. After the chest was opened, the pericardium was cut vertically, and the heart was exposed and suspended in a pericardial cradle. Five intramyocardial injections (200ml each) (3x10 9 T.U. total/pig heart) of viral vectors were performed along the infarct border, and the site of injection was demarcated with a 6-0 Prolene suture. After completing these procedures, warm normal saline (approx. 500mL) was used for ushing the thoracic cavity. This ush was suctioned out before closing the chest. The pericardium was approximated as soon as possible. The sternum was closed with a 20G stainless steel suture and 5 Green Braided P.T.F.E. nonabsorbable surgical sutures. The chest was closed in layers (0 PDS II suture for the muscle and 2-0 PDS II suture subcutaneously), and a single mediastinal tube (18F catheter), 3-way valve, and 60cc syringe will be used to reestablish a negative intrapleural pressure and evacuate any remaining blood or irrigation solution. The chest tube (18F catheter) was removed before skin closing after no visible air leak or blood accumulation, and a purse-string suture (2-0 PDS II) was used to ensure an airtight seal. The skin incision was then glued with Vetbond adhesive. Then the chest was closed, and the inhaled anesthetic was turned off, the animal extubated when appropriate, and allowed to recover. Animals received antimicrobial therapy cetiofur pre-operatively and every 24 h for the rst 48 h postoperatively.
Four weeks after virus injection animal was anesthetized with iso urane as described before to perform the nal cardiac MRI and echocardiography. Body temperature will be monitored and maintained within the physiological range. Arterial blood pressure and surface E.C.G. were monitored continuously. Then the animal was deeply anesthetized with 5% iso urane. A bolus of 3-6 ml/kg of 3M KCl solution will be injected into the left atrium until the heart is arrested. After the cessation of vital signs, the heart will be harvested for postmortem procession.
T.T.C. stain of the pig heart Directly after the heart was harvested, the aorta was perfused with normal saline (500-1000 ml) to ush out vascular blood. The heart was weighed and transversely sliced into 5-6 sections. Heart sections were incubated in 1% T.T.C. at 37 o C for 5 min. Then right ventricular, atriums were removed, and LV sections were weighted. The pictures of LV slices were taken using a professional camera. Images were analyzed using Image-J software. Scar size percentages were calculated.

Statistical Analyses
For all assays, power analyses were performed to choose the group sizes, which will provide >80% power to detect a 10% absolute change in the parameter with a 5% Type I error rate. These power analyses indicated a minimum of 4 experimental replicates per group; therefore, we used a range of 4 -15 experimental replicates per group for each assay. Special statistical consideration for RNAseq was detailed under the methods section. The Kolmogorov-Smirnov (K-S) test for normality was conducted; all data sets showed normal distribution. Then, differences between the two groups were examined for statistical signi cance with unpaired Student t-tests. However, to compare data consisting of more than two groups, we performed one-or Two-way ANOVA tests followed by Bonferroni post hoc multiple comparisons to get the corrected p-value. A value of P<0.05 was regarded as signi cant. Error bars indicate S.D. The person who performed the analysis was blinded to the experimental groups. Two blinded clinical cardiology fellows assessed all cardiac function Echocardiography and MRI analyses, and the data represented are the average of their analyses. Usually, there were no signi cant discrepancies between the two readings.