The Hippo pathway regulates density-dependent proliferation of iPSC-derived cardiac myocytes

Inducing cardiac myocytes to proliferate is considered a potential therapy to target heart disease, however, modulating cardiac myocyte proliferation has proven to be a technical challenge. The Hippo pathway is a kinase signaling cascade that regulates cell proliferation during the growth of the heart. Inhibition of the Hippo pathway increases the activation of the transcription factors YAP/TAZ, which translocate to the nucleus and upregulate transcription of pro-proliferative genes. The Hippo pathway regulates the proliferation of cancer cells, pluripotent stem cells, and epithelial cells through a cell–cell contact-dependent manner, however, it is unclear if cell density-dependent cell proliferation is a consistent feature in cardiac myocytes. Here, we used cultured human iPSC-derived cardiac myocytes (hiCMs) as a model system to investigate this concept. hiCMs have a comparable transcriptome to the immature cardiac myocytes that proliferate during heart development in vivo. Our data indicate that a dense syncytium of hiCMs can regain cell cycle activity and YAP expression and activity when plated sparsely or when density is reduced through wounding. We found that combining two small molecules, XMU-MP-1 and S1P, increased YAP activity and further enhanced proliferation of low-density hiCMs. Importantly, these compounds had no effect on hiCMs within a dense syncytium. These data add to a growing body of literature that link Hippo pathway regulation with cardiac myocyte proliferation and demonstrate that regulation is restricted to cells with reduced contact inhibition.

www.nature.com/scientificreports/ flow cytometry to measure the expression of three cell cycle markers: BrdU, YAP, and cyclin D1. Furthermore, our previous work showed that human iPSC-derived CMs (hiCMs) are more proliferative when plated sparsely than when in a monolayer based on cell counting using high-content microscopy 21 . Whether the Hippo pathway regulates the apparent difference in CM-proliferation capacity in dense and sparse populations remains to be elucidated. This led us to probe whether the Hippo pathway is involved in regulating the proliferative capacity of CMs in various densities or if YAP is activated in these sparse populations by another mechanism. We have chosen hiCMs as our model system to study this phenomenon, as they are transcriptionally immature-similar to fetal or neonatal CMs in vivo-and have a slight basal proliferative capacity 22,23 . Here, we show that reducing hiCM density by either a scratch assay or by sparse plating increases nuclear YAP and CM-proliferation, and that this proliferative capacity can be further enhanced by combinatorial pharmacological perturbation of the Hippo pathway.

Results
Reducing hiCM density in a scratch assay increases proliferative capacity. We first investigated whether a reduction in hiCM density resulted in increased proliferation and/or any changes in Hippo pathway signaling. The hiCMs we used were purchased pre-differentiated and we confirmed the company's claim of a lack of fibroblast contamination (see "Materials and methods", Figure S1B) and slight proliferative capacity ( Figure S1A,D,E) 21 . The lack of non-muscle cell types is key for interpretation of the data sets measuring proliferation. In addition, hiCMs can be cultured as a dense monolayer in which every cell has the potential to be experiencing contact inhibition. One well-established method to relieve contact inhibition in culture is to scratch a confluent monolayer of cells 24 . Therefore, we used a micropipette tip to scratch a confluent monolayer of hiCMs (Fig. 1A,B). We hypothesized that, as in epithelial cells, YAP would translocate to the nucleus of hiCMs near the scratch, as these hiCMs should have reduced contact inhibition compared to those far from the scratch. To test this, we localized YAP using immunofluorescence 48 h post-scratch. We noticed that hiCMs near the edge of the scratch appeared to have primarily nuclear YAP (Fig. 1C). A proxy for YAP activity is to measure the nuclear accumulation of YAP 25 . Therefore, we acquired higher magnification images to measure the ratio of nuclear YAP to total YAP. We used β-catenin to mark adherens junctions between myocytes in order to delineate separate cells, and measured nuclear YAP and total YAP using fluorescence localization. We found that hiCMs near the scratch had a higher nuclear YAP to total YAP ratio than either hiCMs farther from the scratch on the same plate or hiCMs in an unscratched monolayer (Fig. 1C,D). Interestingly, this appeared to be an exponentially decaying relationship, where the nuclear YAP to total YAP ratio decreased rapidly as a function of distance from the scratch.
We next wanted to know if the increase in nuclear YAP correlated with an increase in hiCMs entering the cell cycle. One standard way to identify cycling cells is to localize Ki67 using immunofluorescence, as Ki67 is localized in the nuclei of cycling cells 26,27 . We found that in a monolayer of hiCMs, Ki67-positive cells were evenly distributed across the well and were found in actinin2-and troponinT-positive cells (Fig. 1E,F, Figure S1D,E). In contrast, for the scratched cells, there was a skewed distribution, in which more Ki67-positive nuclei were near the scratch (Fig. 1E,F). We further used the Fucci probe as evidence that cells near a scratch are cycling (Fig. 1G). Taken together, these data indicate that the relief of contact inhibition by a scratch assay increases the amount of YAP in the nuclei of hiCMs and promotes their entry into the cell cycle. However, it is worth noting that a scratch assay is inherently asymmetric as only hiCMs near a scratch respond, limiting measurements of responses on a population level. This further suggests that increase in cell proliferation is not entirely a paracrine mechanism. If a scratch induced cardiac myocyte proliferation in a nonautonomous fashion, all cells would undergo a similar increase in proliferation. While we cannot make the conclusion that in this system, the increase in nuclear YAP/total YAP ratio is directly causative of the increase in Ki67-positive nuclei near the scratch, we can conclude that the increase in proliferative capacity seen in a scratched monolayer of cardiac myocytes is at least in part cell autonomous.
Reducing hiCM density by sparse plating increases proliferative capacity. We next wanted to identify an assay that could model a uniform reduction in contact inhibition across an entire population of hiCMs. A previous study used flow cytometry and Western blotting to show that embryonic stem cell-derived cardiac myocytes begin cycling when they are sparsely plated 20 . Thus, we plated hiCMs as a monolayer or plated them sparsely. Not surprisingly, a monolayer of hiCMs (~ 1088 hiCMs/mm 2 ) have a low nuclear YAP/total YAP ratio ( Fig. 2A,B). On the other hand, when cells are plated at an eightfold dilution (~ 136 hiCMs/mm 2 ), the nuclear YAP/total YAP ratio increases almost twofold ( Fig. 2A,B).
This increase in nuclear YAP in sparsely-plated hiCMs suggested an increase in proliferative capacity. To test this, we asked if sparse plating caused entry into the cell cycle using Ki67 localization. We localized Ki67 in dense or sparse hiCMs and found an increase in Ki67-positive nuclei, indicating a near doubling in cycling hiCMs (Fig. 2C,D). We similarly used the Fucci probe system and found that the majority of dense hiCMs were not cycling, whereas sparse hiCMs had increased cycling cells (Fig. 2E). To directly test if sparse-plating induced proliferation in hiCMs, we counted the number of hiCMs over time. We found that while dense control hiCMs essentially remained the same over 48 h, there was a ~ 5% increase in the number of sparsely-plated hiCMs (Fig. 2F). Taken together, our data indicates that sparse plating induces hiCMs to enter the cell cycle and also divide. This led us to explore the underlying cause of sparsely-plated hiCMs to enter the cell cycle.
It has been postulated that CM division is preceded by dedifferentiation into a more immature phenotype 28,29 . Thus, we employed RNASeq analysis to examine whether the proliferating sparsely-plated hiCMs were as mature as hiCMs in a monolayer. We found CM-maturity genes (Table 1)  www.nature.com/scientificreports/  www.nature.com/scientificreports/  www.nature.com/scientificreports/ ( Fig. 2G) 22 . However, expression of CM-identity genes α-actinin 2 and troponinT did not change ( Fig. 2G; Table 3). Thus, sparsely-plated hiCMs are less mature than densely-plated hiCMs, but they remain CMs. We next wanted to test if the increase in nuclear YAP ratio in sparsely-plated hiCMs was correlated with higher YAP target gene expression. Thus, we curated a list of YAP target genes from the published literature [30][31][32] . We then compared the expression of these genes between densely-plated and sparsely-plated hiCMs. 32% of YAP target genes had upregulated expression in sparsely-plated hiCMs compared to dense, while 68% did not change, and 0% had decreased expression. In conclusion, plating hiCMs sparsely increases YAP nuclear localization, YAP target gene expression, and ultimately, hiCM proliferation compared to densely-plated hiCMs.

Pharmacological perturbation of the Hippo pathway affects cardiac myocyte proliferation.
A loss of contact inhibition not only increased hiCM proliferation but also the nuclear YAP/total YAP ratio, consistent with the hypothesis that the Hippo pathway is involved in density-dependent hiCM proliferation. We next asked if small molecule inhibition of the Hippo pathway would further increase the proliferative potential of hiCMs plated at a low density. We turned to an MST1/2 inhibitor, XMU-MP-1, and a bioactive lipid that activates YAP, sphingosine-1-phosphate (S1P) 23,33,34 (Fig. 3A). Both of these compounds have been shown to upregulate YAP target genes in human embryonic stem cells, mouse liver cells and human hepatoma cells 30,33 . Furthermore, S1P increases Ki67 in densely plated hiCMs but not cell count when combined with lysophosphatidic acid 23 .
We used RNASeq analysis to test if XMU-MP-1 or S1P upregulate YAP/TAZ target genes in hiCMs, and found that YAP/TAZ target genes were indeed significantly upregulated beyond that of sparse-plating alone ( Fig. 3B; Table 1). This increase in YAP target gene expression suggests that the compound-treated hiCMs are less mature than untreated sparsely-plated hiCMs. We found that S1P treatment did not lower the expression of maturity genes compared to untreated hiCMs, whereas 4.2% of maturity genes were expressed lower after XMU-MP-1 treatment ( Fig. 3C; Table 2). We then confirmed that the hiCMs were still expressing CM specific markers Table 1. RNASeq results of YAP target gene expression. FC fold change. The comparisons are made between the first treatment (outside of the parentheses, listed first) and the second (Inside the parentheses). A fold change greater than 1 indicates in increase in gene expression in the first treatment type compared to the second, and a fold change greater than 1 indicates a decrease in gene expression in the first treatment type compared to the second. p-values were obtained as in the methods. Bold values indicate significant differences www.nature.com/scientificreports/ ( Fig. 3B; Table 1). Taken together, our data indicates that XMU-MP-1 or S1P treatment causes an upregulation of YAP/TAZ target genes with minor effects on maturity of hiCMs. Despite an upregulation in YAP/TAZ target genes, we found that neither XMU-MP-1 nor S1P alone increased proliferation by hiCM count (Fig. 3D). The result that XMU-MP-1 or S1P failed to induce proliferation was not www.nature.com/scientificreports/ surprising, as previous studies have postulated that combinations of multiple genetic or pharmacological factors are required to induce cardiac myocyte proliferation both in vitro and in vivo 35,36 . Therefore, we performed a combinatorial screen to determine if combining XMU-MP-1 and S1P induced proliferation of hiCMs. We started by testing which combination of concentrations was most effective to induce hiCMs to reenter the cell cycle using BrdU incorporation (Fig. 4A). We found that adding 0.1 µM of each compound to hiCMs significantly increased the percent of BrdU-positive hiCMs when sparsely-plated (Fig. 4B). We then found that this combination significantly increased hiCM-proliferation by counting nuclei over time (Fig. 4C). We found that using 0.1 µM of both XMU-MP-1 and S1P also decreased phospho-YAP, which indicated a reduction in the inactivation of YAP (Fig. 4D-F).
It is important to confirm that these proliferative hiCMs are functionally beating upon treatment of XMU-MP-1 and S1P. Therefore, we measured the percentage of beating hiCMs upon sparse plating and treatment of XMU-MP-1, S1P, or both, and found no significant difference (Fig. 4G). Further, it is important to consider that both cytokinetic cell division and binucleation are potential outcomes of increasing overall proliferation of cells. We measured the percentage of binucleated hiCMs upon sparse plating and treatment of XMU-MP-1, S1P, or both, and found no significant difference (Fig. 4H). In this case, it is presumed that proliferation appears to be induced overall, but the balance between the potential outcomes of cell division and binucleation are not  www.nature.com/scientificreports/ changed, and both outcomes are occurring. These results agree with past research inducing hiCMs to divide, showing both binucleation and cell division as outcomes 21 .
Finally, to further confirm that XMU-MP-1 and S1P were acting through the Hippo pathway, we turned to Verteporfin, which inhibits the interaction of YAP with transcription factor TEAD 37 (Fig. 4I). For this experiment we chose 0.01 µM of Verteporfin, which is the highest concentration that did not result in the death of hiCMs (Fig. 4J). We found that Verteporfin masked the effect of the combination of XMU-MP-1 and S1P combination on proliferation (Fig. 4K) without affecting the percentage of beating hiCMs (Fig. 4G). This result further suggests that the increase in proliferation of hiCMs with XMU-MP-1 combined with S1P is through the Hippo pathway.

Discussion
Taken together, we show that the density at which hiCMs are plated affects proliferative capacity, and that this proliferative capacity is regulated at least in part by the Hippo pathway. Further, we modulate the proliferative capacity by scratching a monolayer of cells to induce cells at the scratch periphery to divide, and induce division of sparsely-plated hiCMs by dual inhibition of the Hippo pathway. These experiments were completed using hiCMs, a model system of developmentally immature cells that both divide and binucleate, providing an ideal model for studying cardiomyocyte division. This is an alternative approach to using isolated primary cardiomyocytes from adult rats or mice, which are unsuitable to the experiments we use here. While primary cardiomyocytes are powerful tools, creating dense monolayer is technically challenging. In addition, primary cardiomyocytes only last a few days in culture and are more representative of modeling cell death 38 . Therefore, it is difficult to do contact inhibition studies and proliferative studies in these cell types. As we show here, hiCMs are well suited for these types of studies 21 .
It is not surprising that XMU-MP-1 itself did not increase CM proliferation, as a recent study found that XMU-MP-1 in an intact heart did not increase Ki67 staining 39 . Based on Ki67 staining, recent data has also suggested that inhibiting the Hippo pathway with a single compound in combination with perturbation of another cell cycle pathway (i.e., the Wnt pathway) could be an effective approach to increasing CM proliferation 35 . Here, we show that it is possible to have a combinatorial effect on CM proliferation by only targeting the Hippo pathway. Modulation of the Hippo pathway at two different levels was required to get a proliferative response. This could be part of a new direction for combinatorial therapy, where multiple pathways are each modulated at multiple levels.
There are several other pathways that drive CM proliferation. A recent study suggests the neuregulin receptor, ERBB2, activates YAP in an ERK-dependent manner 40 . The role of ERK signaling in cardiomyocyte proliferation and its crosstalk with the Hippo pathway would constitute an interesting future study. Furthermore, the signaling role that cell-cell adhesions and junctions themselves play in the cardiac regenerative response seems to merge on these pathways. Regenerating cardiac myocytes undergo junction dissolution, and it has been shown that displaced alpha-catenin activates YAP in epidermal cells 15 . Finally, increasing YAP and TAZ activity is likely to have benefits in the heart other than increasing cardiac myocyte proliferation. For example, YAP and TAZ in the epicardium was shown to induce recruitment of T-regulatory cells to the infarcted myocardium. Thus, YAP and TAZ may have the ability to regulate the adaptive immune response and decreasing post-infarct inflammation and myocardial fibrosis 41 .
Finally, our work also has implications in the context of a myocardial infarction. In the developed heart, a myocardial infarction results in massive CM death and the remaining CMs do not proliferate 42 . After a myocardial infarct, there is a reduction in cell density at the infarct zone with an increase in cell cycling shown by Ki67 fluorescence 43 . However, it has been shown that the proliferative cells in the infarct region appear to mostly undergo polyploidization and endomitosis, indicating that there is a reduction in contact inhibition but there is also still a barrier to true cytokinetic proliferation in this region. Interestingly, YAP is primarily nuclear in the border zone 21 days post-cardiac apex resection, but only in Salv CKO animals 44 . If an ideal treatment for a cardiac infarction existed, it would only induce proliferation at the infarct site where cell-cell contact is reduced. Global www.nature.com/scientificreports/ induction of proliferation could induce undesired side effects (e.g. tumor formation). As such, targeting pathways which respond preferentially when contact inhibition is lost could be a useful therapeutic direction. Here, we show that targeting the Hippo pathway can increase proliferation in the context of a loss of contact inhibition.

Reagent or resource Source Identifier
Antibodies www.nature.com/scientificreports/ field potential duration and beating rate, and viability by trypan blue exclusion. We further characterized these cells in Neininger et al., 2019, in which we determined that fewer than 1 in 50,000 cells were actinin2-negative 21 .
XMU-MP-1 (S8334, Selleck Chemicals, Houston, TX) and Verteporfin (SML0534, Sigma Aldrich, St. Louis, MO) were reconstituted to 10 mM in DMSO. S1P (B6707, Apex Bio, Houston, TX) was reconstituted to 10 mM in methanol. When adding drugs to the cells, they were first diluted 1:10 in DMSO then to the appropriate concentration in cardiac myocyte maintenance medium. NucLight Rapid Red Reagent (4717, Essen BioScience, Ann Arbor, MI) was used at a 1:4000 dilution in cell culture media.
Fixation and immunostaining. Cells were fixed with 4% paraformaldehyde (PFA, 15710, Electron Microscopy Sciences, Hatfield, PA) diluted form 16% in PBS at room temperature for 20 min, then extracted for 5 min with 1% Triton X-100 (BP151100, Fischer Scientific, Suwanee, GA) and 4% PFA in PBS. Cells were washed three times in 1 × PBS. After fixation, the following labeling procedures were used: for immunofluorescence experiments, cells were blocked in 10% bovine serum albumin (BSA) in PBS for 20 min. Primary antibodies were diluted in 10% BSA. All primary antibodies were used at 1:200 for 1 h and 45 min at room temperature. Secondary antibodies were diluted in 10% BSA at 1:100 and centrifuged at 13,000 rpm for 2 min before use at room temperature for 1 h.
BrdU (ab142567, abcam, Cambridge, UK) was reconstituted to 10 mM in water, then to 10 µM in cardiomyocyte maintenance medium and filtered through a 0.22 µm syringe filter. Cells were incubated with each compound and BrdU for 24 h, then fixed and permeabilized as above. Then, to hydrolyze DNA, cells were incubated with 1 M HCl for an hour at room temperature, then neutralized with 0.1 M sodium borate (pH 8.5) for 20 min at room temperature. Next, cells were washed with PBS and immunostaining continued as usual.
Scratch assay. Scratches in 10 mm dishes were done by hand with a 10 µL pipette tip. Scratches in 96-well cell culture polystyrene plates were done using a 200 µL pipette tip and a custom plate lid produced by A.C.N. with a slot designed to guide the pipette tip straight across the well. Media was replaced with media containing NucLight Rapid Red Reagent immediately after scratching and every two days afterward until fixation at specified time points.

Statistics and quantification.
Total YAP was measured in FIJI (ImageJ, NIH, Bethesda, MD) by drawing a polygonal ROI around a cell using a 20 × IncuCyte image stained for B-catenin to mark cell boundaries, and measuring average fluorescence intensity. Nuclear YAP was measured in FIJI by drawing a Bezier ROI around a nucleus using a 20 × IncuCyte image stained for YAP and measuring average fluorescence intensity. Fit of nuclear YAP/total YAP graph in Fig. 1D was determined using the curve fitting toolbox in MatLab (MathWorks, Natick, MA).
Nuclear fold change was determined by thresholding live-cell whole-well 4 × stitches with NucLight Rapid Red Reagent, a live-cell nuclear marker. Thresholding was done using the IncuCyte software (Essen Biosciences, Ann Arbor, MI) and a TopHat Background subtraction using a 20 µM rolling ball. Images were acquired and nuclei counted every hour for specified time points. Nuclei count at the final time point was divided by nuclei count at the first time point and normalized to the average nuclear fold change of cells in control conditions on the same plate. www.nature.com/scientificreports/ Statistical significance of Fig. 2D,F were determined by unpaired two-tailed Student's t tests performed in Excel. Statistical significance of Figs. 2B and 4C were determined by paired two-tailed Student's t tests performed in Excel. Statistical significance of Figs. 3D, 4B,E-G,J were determined by one-way ANOVA with a Dunnett's post-hoc test correcting for multiple testing when applicable performed in GraphPad Prism. Each experiment was performed a minimum of 3 times and the mean of the independent experiments and standard error of the mean (SEM) are displayed. RNA sequencing. Cell pellets of 100,000 cells were prepared and RNA was extracted using the RNeasy Mini Kit (Qiagen, #74104). Stranded mRNA (polyA-selected) library preparation was completed by the VANTAGE core at Vanderbilt University. Sequencing was done with an Illumina NovaSeq6000 (S4) PE150. Reads were mapped to reference genome using STAR and differential analyses were performed using DESeq2. Differentially expressed genes were determined using the criteria of fold change ≥ 2 and FDR ≤ 0.05. YAP target genes and cardiac myocyte maturity genes were determined from literature and are listed below. Datasets will be uploaded to a public database. N = 3 for each condition, in that each cell treatment and RNA preparation was done three separate times using three separate purchased vials of cells.