Interaction of YAP with the Myb-MuvB (MMB) complex defines a transcriptional program to promote the proliferation of cardiomyocytes

The Hippo signalling pathway and its central effector YAP regulate proliferation of cardiomyocytes and growth of the heart. Using genetic models in mice we show that the increased proliferation of embryonal and postnatal cardiomyocytes due to loss of the Hippo-signaling component SAV1 depends on the Myb-MuvB (MMB) complex. Similarly, proliferation of postnatal cardiomyocytes induced by constitutive active YAP requires MMB. Genome studies revealed that YAP and MMB regulate an overlapping set of cell cycle genes in cardiomyocytes. Protein-protein interaction studies in cell lines and with recombinant proteins showed that YAP binds directly to B-MYB, a subunit of MMB, in a manner dependent on the YAP WW domains and a PPXY motif in B-MYB. Disruption of the interaction by overexpression of the YAP binding domain of B-MYB strongly inhibits the proliferation of cardiomyocytes. Our results point to MMB as a critical downstream effector of YAP in the control of cardiomyocyte proliferation.

Author summary YAP, the major downstream transducer of the Hippo pathway, is a potent inducer of proliferation. Here we show that the Myb-MuvB complex (MMB) mediates cardiomyocyte proliferation by YAP. We find that YAP and MMB regulate an overlapping set of pro-proliferative genes which involves binding of MMB to the promoters of these genes. We also identified a direct interaction between the B-MYB subunit of MMB and YAP. Based on the binding studies, we created a tool called MY-COMP that interferes with the association YAP to B-MYB and strongly inhibits proliferation of cardiomyocytes. Together, our data suggests that the YAP-MMB interaction is essential for division of cardiomyocytes, underscoring the functional relevance of the crosstalk between these two pathways for proper heart development.

Introduction
The Hippo signaling pathway plays fundamental roles in proliferation and organ size control [1]. In mammals, the Hippo cascade is composed of the MST1/2 and LATS1/2 kinases and the adaptor proteins MOB1 and Salvador (SAV1). When the Hippo pathway is active, LATS1/2 phosphorylate the transcriptional coactivators YAP and TAZ, which results either in their cytoplasmic retention by 14-3-3 proteins or SCF-mediated proteasomal degradation [2]. In contrast, when Hippo is inactive, YAP/TAZ enter the nucleus and act as transcriptional coactivators predominantly by binding to DNA through TEAD transcription factors. The Hippo pathway and its downstream effectors YAP and TAZ are involved in cardiac development and have been implicated in heart regeneration after tissue damage [3,4]. Active YAP promotes proliferation of postnatal cardiomyocytes and induces a fetal-like cell state in adult cardiomyocytes [5,6]. Similarly, cardiac-specific deletion of the Hippo kinases LATS1/2 or of the scaffold protein SAV1 results in heart enlargement and increased proliferation of embryonal and postnatal cardiomyocytes due to increased YAP activity [7,8]. Conversely, deletion of YAP suppresses cardiomyocyte proliferation [5,7,9]. Notably, activated YAP or loss of Hippo signaling can extend the neonatal proliferation of cardiomyocytes to postnatal stages, where proliferation is normally very low [5][6][7]. In addition, recent studies report a better outcome after myocardial infarction in mice with hyperactivated YAP [8,[10][11][12]. Together, these studies identify YAP as an important regulator of cardiomyocyte proliferation and cardiac regeneration. However, the detailed mechanisms by which YAP promotes cardiomyocyte proliferation are still unclear [13]. We recently observed that Myb-MuvB complex (MMB) and YAP co-regulate an overlapping set of late cell cycle genes [14]. MMB consist of the evolutionary conserved MuvB core of the five proteins LIN9, LIN52, LIN54, LIN37 and RBBP4 and associated proteins [15,16]. Depending on its interactions with different binding partners, MuvB can either repress or activate transcription. Specifically, when MuvB interacts with the p130 retinoblastoma protein paralog, and with E2F4 and DP1, it forms the DREAM complex, which represses E2F-dependent gene expression in quiescent cells and in early G1 [17][18][19]. Upon cell cycle entry, p130, E2F4 and DP1 dissociate from the complex and the MuvB core associates with the transcription factor B-MYB to form MMB [19][20][21][22]. MMB mainly regulates genes required for mitosis and cytokinesis. Binding of YAP to enhancers promotes the activation of MMB-bound promoters, providing an explanation for the co-activation of late cell cycle genes by YAP and MMB [14].
Whether the ability of YAP to promote proliferation in vivo is mediated by MuvB complexes has not been addressed. To investigate the possible function of MuvB in YAP-mediated cardiomyocyte proliferation we used genetic approaches in mice and biochemical experiments. We find that YAP and MMB interact in the developing heart. In vivo in mice this interaction is essential for increased mitosis of Hippo-deficient cardiomyocytes. Additionally, we demonstrate that YAP driven proliferation of postmitotic cardiomyocytes is dependent on the interaction of YAP with the B-MYB subunit of MMB.

MuvB is required for proliferation and mitosis of Hippo-deficient embryonic cardiomyocytes
To explore the connection between the Hippo-YAP pathway and MuvB in vivo in the heart, we deleted the Hippo pathway member Salvador (Sav1), a scaffold protein that is required for Hippo kinase activity, and the MuvB subunit Lin9 in cardiac precursor cells. Early cardiac specific deletion of Lin9 and Sav1 was achieved by using mouse strains with conditional (floxed) alleles of Sav1 and Lin9 in combination with Nkx2.5-Cre [23]. Phosphorylation of YAP on S127 was reduced in Sav1 deficient hearts, indicating that YAP is hyperactivated (S1A Fig). We first examined the effect of MMB inactivation on proliferation of Sav1 mutated cardiomyocytes. In these experiments, heterozygous Nkx2.5-Cre; Lin9 fl/+ mice served as a control, because these mice developed and survived normally and showed no cardiac phenotype. To assess proliferation, we stained E13.5 heart sections for the cell cycle marker Ki67. Cardiomyocytes were identified by co-staining for cardiac troponin T (cTnT). The proportion of Ki67-positive cardiomyocytes was significantly increased in Nkx2.5-Cre; Sav1 fl/fl hearts compared to control hearts with wildtype Sav1, indicating that loss of SAV1 promoted cardiomyocyte proliferation as has been reported previously [7] (Fig 1A and 1B; S1B Fig). The fraction of cardiomyocytes staining positive for phosphorylated histone H3 (pH3), a marker of mitotic cells, was also increased in Nkx2.5-Cre; Sav1 fl/fl hearts (Fig 1C and 1D, S1C Fig). In contrast, proliferation and mitosis were strongly reduced by inactivation of Lin9 and the increase in proliferation by Sav1 deletion was blocked in Sav1, Lin9 double mutant cardiomyocytes (Fig 1A-1D, S1B and S1C Fig). These genetic experiments suggest that the proliferative phenotype due to the loss of the Hippo pathway member SAV1 is dependent on the MMB subunit LIN9.

Hippo and MMB regulate an overlapping set of genes in embryonic cardiomyocytes
Next, we explored the impact of Lin9 mutation on the transcriptional program of Hippo-deficient cardiomyocytes by performing RNA-seq using RNA isolated from heart ventricles of E13.5 embryos carrying cardiac specific deletions of Sav1 alone, Lin9 alone or Sav1 and Lin9 together. In Sav1 knockout hearts, E2F target genes and other gene sets related to cell cycle regulation were enriched after inactivation of Hippo-signaling (Fig 2A, S2A and S2B Fig). Strikingly, the activation of these gene sets was suppressed by the simultaneous deletion of Lin9 together with Sav1, indicating that the elevated expression of these genes in Sav1 mutant hearts requires the function of MMB (Fig 2A and 2B, S2A and S2B Fig). By RT-qPCR we independently validated that a set of cell cycle target genes were activated in Sav1 mutated hearts, downregulated in Lin9 mutated hearts and remained downregulated in hearts isolated from double mutant embryos ( Fig 2C).
To find out whether cell cycle genes upregulated in Sav1 hearts are direct targets of MMB, we performed chromatin immunoprecipitation experiments followed by high-throughput sequencing (ChIP-seq) ( Fig 2D). We identified 3,357 binding sites for LIN9 in the fetal heart. By comparison with previously reported ChIP-seq data of histone modifications characteristic for active promoters and enhancers, most LIN9 binding sites are located in active promoters while less than 5% of LIN9 sites are found in enhancers or super-enhancers ( Fig 2E, S2C Fig). ChIP-seq of YAP in E16.5 hearts showed that YAP predominantly binds to enhancers and super-enhancers rather than to promoters (Fig 2E), which is consistent with recent genome wide studies in human cell lines [14,[24][25][26]. Consequently, there was little overlap in the binding of YAP and LIN9 (Fig 2F, S2C Fig). Genome browser tracks of LIN9 and YAP-bound promoters and of YAP-bound enhancers illustrating these findings are shown in Fig 2G and  S2D Fig. To identify the direct targets of LIN9 in the heart, we plotted changes in gene expression upon deletion of Lin9 against the density of promoter-bound LIN9. This revealed a correlation between genes that are activated by LIN9 (i.e. that are downregulated after deletion of Lin9) and promoter binding of LIN9 (Fig 2H). In particular, LIN9 strongly bound to the promoters of cell cycle genes and E2F target genes (Fig 2I). While overall LIN9-binding correlated with Error bars indicate SDs. Number of mice analyzed per genotype: Lin9 fl/+ ; Sav1 +/+ n = 5, Lin9 fl/+ ; Sav1 fl/fl n = 4, Lin9 fl/fl ; Sav1 +/+ n = 6 and Lin9 fl/fl ; Sav1 fl/fl n = 5 C) and D) Heart sections from E13.5 mice with the indicated genotypes were stained for the mitosis marker pH3 (red). Cardiomyocytes were identified by staining for cTnT (green). Example microphotographs are shown in (C). Asterisks indicate pH3-positive cardiomyocyte nuclei. Scale bar: 10 μm. The quantification of pH3-positive cells is shown in (D). Number of mice analyzed per genotype: Lin9 fl/+ ; Sav1 +/+ n = 5, Lin9 fl/ + ; Sav1 fl/fl n = 4, Lin9 fl/fl ; Sav1 +/+ n = 6 and Lin9 fl/fl ; Sav1 fl/fl n = 6. Error bars indicate SD. B, D: Student's t-test. � = p<0.05, �� = p<0.01, ��� = p<0.001, ns = not significant.
https://doi.org/10.1371/journal.pgen.1008818.g001 Cell cycle genes upregulated in Sav1 knockout hearts are direct targets of LIN9. A) GSEA comparing the effect of deletion of Sav1 in Nkx-Cre LIN9 wt and LIN9 KO heart ventricles at E13.5. Gene sets related to cell cycle are highlighted in blue. NES: normalized enrichment score. B) Boxplot comparing differences in E2F target gene expression between Nkx2.5-Cre; Sav1 fl/fl (Sav1 KO) and Nkx2.5-Cre; Sav1 +/+ (Sav1 wt) heart ventricles in Lin9 fl/fl (LIN9 KO) or Lin9 fl/+ (LIN9 wt) background. C) Expression of the indicated genes relative to actin and Hprt in heart ventricles of embryos with the expression changes, LIN9 only weakly bound to the promoters of some gene sets that are also downregulated in Lin9 knockout hearts, including genes related to mitochondrial function, oxidative phosphorylation, metabolism, heart muscle contraction and ion channel activity ( Fig  2I, S2E and S2F Fig). Downregulation of these genes is likely an indirect consequence of loss of Lin9. Strikingly, plotting changes in gene expression upon cardiac-specific deletion of Sav1 against LIN9-promoter occupancy showed that promoters of genes activated in Sav1 knockout hearts were bound by LIN9, indicating that they are direct targets of MMB ( Fig 2J). Altogether these data support the idea that YAP activates LIN9-bound cell cycle promoters from distant enhancers as previously shown in human MCF10A cells [14]. In summary, the deletion of Sav1 hyperactivates YAP, which in turn results in induction of cell cycle genes whose promoters are bound by LIN9.

Lin9 is required for proliferation of Hippo-deficient postnatal cardiomyocytes
Deletion of Lin9 in heart progenitor cells by Nkx2.5-Cre resulted in embryonic lethality due to defects in cardiomyocyte proliferation and division resulting in enlarged nuclei and leading to reduced thickness of the compact myocardium of both ventricles (Fig 3A-3D To circumvent the embryonic lethality associated with deleting Lin9 in early cardiac precursors, we used a transgenic mouse line with Cre recombinase driven by the alpha-MHC-promoter, which is active at a later stage during heart development compared to Nkx2.5-Cre [27]. α-MHC-Cre; Lin9 fl/fl mice survived into adulthood without any obvious heart phenotype and without differences in the heart to body weight when compared to heterozygous control animals ( Fig 3E and 3F, S3C Fig). To investigate the efficiency of Cre mediated deletion in the heart, we used a mT/mG reporter strain [28]. Upon Cre-induced recombination of the mT/mG reporter gene, membrane bound Tomato (mT) is removed and the expression of membrane bound EGFP (mG) is activated (Fig 3G, S3D Fig). In control mice that do not harbor any Cre recombinase, no recombination of the mT/mG reporter gene took place and more than 95% of cardiomyocytes exhibited red fluorescence, as expected. After combining Nkx2.5-Cre with the mT/mG reporter strain more than 95% of cardiomyocytes exhibited green fluorescence in 13.5 dpc hearts, indicating efficient recombination. In contrast, in α-MHC-Cre; mT/mG mice, a much smaller proportion of about 30-40% cardiomyocytes were GFP positive at E13.5. The relative low recombination frequency in α-MHC-Cre mice at this developmental time point can explain the lack of an embryonic heart phenotype in α-MHC-Cre; Lin9 fl/fl mice. However, at P10, recombination in the heart of α-MHC-Cre mice was almost 100%. Cardiac specific deletion of Lin9 in α-MHC-Cre mice at P10 was confirmed by genomic PCR and, importantly, indicated genotypes was analyzed by RT-qPCR. D) E16.5 hearts were subjected to ChIP-seq using antibodies specific for LIN9 and YAP. The table shows the number of reads, aligned reads and called peaks. E) Genomic localization of LIN9 and YAP in fetal (E16.5) heart ventricles. Active promoters, enhancers and super-enhancers are defined based on publicly available ChIP-seq data for histone marks. F) Venn diagram showing the overlap between binding by LIN9 and YAP in E16.5 hearts. G) Genome browser tracks illustrating the binding of LIN9 to the Aspm promoter and binding of YAP to the Amotl2 promoter and to an intergenic enhancer. ChIP-seq data for histone modifications are from ENCODE (GSE31039). Additional examples are shown in S2D Fig. H) Bin plot correlating changes in gene expression in Nkx-Cre; Lin9 fl/fl heart ventricles with binding of LIN9 to the promoter. Analyzed was a region 1kb upstream the TSS and input signals were subtracted. 15,642 expressed genes were grouped into 15 bins and the mean of each bin is plotted.  the absence of the LIN9 protein was validated by IP-western (Fig 3H and 3I). Taken together while early cardiac deletion of Lin9 caused a dramatic and embryonal lethal phenotype, postnatal deletion by α-MHC-Cre did not result in a heart phenotype.
In hearts of 10 days old wildtype mice, LIN9 remained associated with chromatin as determined by ChIP-seq (S3E- S3G Fig). The overall chromatin binding pattern of LIN9 was similar in embryonic and postnatal hearts and the shared binding sites reflect high confident LIN9binding sites (S3G- S3J Fig).
To address whether LIN9 is required for mitosis of postnatal cardiomyocytes in Hippo-deficient hearts, we next crossed α-MHC-Cre; Lin9 fl/fl mice to Sav1 fl/fl mice. Under normal conditions, there are almost no dividing cardiomyocytes in the postnatal heart, as expected. The deletion of Sav1 robustly increased the fraction of mitotic cardiomyocytes (Fig 3J and 3K). Strikingly, this phenotype was suppressed when Lin9 was lacking and the fraction of mitotic cardiomyocytes remained low in α-MHC-Cre; Sav1 fl/fl ; Lin9 fl/fl double mutant mice, indicating a role for MMB in mitotic entry of postnatal cardiomyocytes due to Hippo deficiency (Fig 3J  and 3K).
Thus, LIN9 is dispensable in the postnatal heart probably because cardiomyocytes hardly divide at this time point. However, LIN9 becomes necessary for ectopic cardiomyocyte proliferation in the absence of SAV1.

Cardiomyocyte proliferation by activated YAP requires LIN9
To directly test whether increased cardiomyocyte proliferation due to activated YAP depends on MMB, we transduced neonatal cardiomyocytes with an adenovirus expressing a constitutive active version of YAP in which S127, the site of the inactivating phosphorylation by LATS kinases, is mutated to alanine ( Fig 4A). To determine the requirement of MMB for YAP (S127A) induced proliferation, we used cardiomyocytes isolated from mice with a conditional allele of Lin9 (Lin9 fl/fl ) and expressing a hormone inducible CreER recombinase that can be activated by the addition of 4-hydroxytamoxifen (4-OHT) ( Fig 4B). Treatment with 4-OHT resulted in efficient deletion of Lin9 in neonatal Lin9 fl/fl ; CreER cardiomyocytes ( Fig 4C). Expression of YAP(S127A) robustly induced mitotic entry of embryonic E14.5 and postnatal P1 cardiomyocytes, as reported before [5,9] (

YAP physically interacts with MMB in cardiomyocytes
YAP and MMB could independently regulate a similar set of genes required for cell cycle regulation or they could cooperate by binding to each other. To address these possibilities, we next investigated whether YAP and B-MYB interact in the heart. Proximity ligation assays (PLA) showed that YAP indeed interacted with B-MYB in embryonic cardiomyocytes (Fig 5A and  5B). The interaction was specific, as demonstrated by the loss of the PLA signal upon siRNAmediated depletion of B-MYB or YAP (Fig 5A and 5B). YAP also interacted with LIN9, a core subunit of the MuvB complex that is required for binding of B-MYB to MuvB [29]. Binding between YAP and LIN9 was specific as demonstrated by siRNA mediated depletion of LIN9 or YAP (Fig 5A and 5B). Furthermore, the PLA signal was lost when one of the two antibodies was omitted or after genetic deletion of Lin9 in Nkx2.5-Cre; Lin9 fl/fl cardiomyocytes, providing further evidence for an interaction between YAP and MMB in cardiomyocytes (Fig 5C and  5D).

B-MYB interacts with the tandem WW domains of YAP
To gain further insights into the interplay between YAP and B-MYB and to develop tools to interfere with the interaction, we mapped the domains of YAP and B-MYB that are involved in the interaction. In co-immunoprecipitation experiments with HA-B-MYB and a set of truncated flag-tagged YAP constructs, B-MYB interacted with YAP only when the tandem WW domains were present. B-MYB did not interact with the C-terminal transactivation domain or the PDZ-binding motif of YAP (Fig 6A and 6B, S5A Fig). Internal deletion of the tandem WW domains abolished the binding, confirming that these domains are required for the YAP-B-MYB interaction ( Fig 6C). To verify that the YAP WW domains mediate the interaction with B-MYB, we performed pulldown experiments with recombinant GST fused to the N-terminal part of YAP containing the TEAD-binding and WW domains (GST-TEAD-WW1/2) or just the two WW domains (GST-WW1/2) (S5B Fig). HA-B-MYB specifically bound to GST-TEAD-WW1/2 and GST-WW1/2 and but not to GST alone ( Fig 6D). As a control, HA-TEAD4 only bound to GST-TEAD-WW1/2 containing the TEAD-binding domain, but not to GST alone or to GST-WW1/2. HA-tagged EB1, which was used as a negative control, did not bind to any of the GST constructs. Although B-MYB can independently bind to

Overexpression of the YAP binding domain disrupts the B-MYB-YAP interaction and inhibits cardiomyocyte proliferation
Since YAP binds to the N-terminus of B-MYB and since the MMB-interaction domain (MBD) is located in the C-terminal region, the binding of YAP to B-MYB is likely not mediated by binding of YAP to the MuvB core [14,29]. In pulldown experiments with GST-WW1 and with a set of HA-tagged B-MYB deletion mutants we mapped the minimal binding site for YAP to amino acids 80 and 241 of B-MYB (Fig 7A and 7B). For example, B-MYB(2-241) robustly bound to YAP whereas B-MYB(2-79) and B-MYB(242-410) failed to bind to YAP (Fig 7B). The interaction between B-MYB and YAP could be indirect since exogenously expressed HA-B-MYB likely associates with additional cellular proteins. Purified, recombinant histagged B-MYB(2-241) interacted with GST-WW1/2, indicating that the binding between B-MYB and YAP is direct (Fig 7C).
To identify the region in B-MYB that is responsible for binding to the WW domains, we used an overlapping peptide library in μSPOT format that displayed the whole length of the B-MYB protein as 15mer peptides with 2 amino acid offset. The peptide array was incubated with GST-WW1/2 and binding was detected with an anti-GST antibody. These experiments confirmed direct binding of B-MYB to the WW domains of YAP and identified a 15 amino acid peptide containing a PPXY motif as the fragment with the highest YAP binding capacity (S6A and S6B Fig). Proline rich PPXY motifs are known ligands for the WW domain, a protein-interaction domain characterized by two tryptophan-residues separated by 20 to 22 amino acids [30,31]. To directly test whether the PPXY motif is required for binding to YAP, we performed pulldown experiments with a deletion mutant of recombinant B-MYB lacking the PPXY sequence (ΔPPXY). Compared to his-tagged B-MYB(2-241), binding of his-B-MYB (2-241,ΔPPXY) to the WW domains of YAP was strongly reduced (Fig 7C). The PPXY motif was also required for the interaction with YAP in the context of the full-length B-MYB protein in cells (S6C and S6D Fig). Taken together, these results indicate that the YAP interacting region of B-MYB is located between amino acids 80 and 241 of B-MYB and involves a PPXY motif.
Since B-MYB and YAP directly interact, we next asked whether overexpression of the YAP binding domain of B-MYB can interfere with the interaction of B-MYB with YAP due to direct competition for the binding site ( Fig 7D). To address this possibility we created B-MYB(2-241) fused to a HA tag and a nuclear localization signal (HA-NLS-B-MYB-2-241), and named it MY-COMP for MYB-YAP competition. First, we expressed MY-COMP in cells and performed co-immunoprecipitation experiments. In HeLa cells that express MY-COMP, the amount of full-length HA-B-MYB co-precipitating with flag-YAP was strongly reduced when compared to control transfected cells (Fig 7E, S6E Fig). Expression of MY-COMP also interfered with the endogenous YAP and B-MYB interaction as determined by proximity ligation assays (PLA) (Fig 7F, S6F Fig).
Next, we asked whether the interaction between YAP and B-MYB is important for promoting proliferation in neonatal cardiomyocytes. We first tested whether the ability of YAP [S127A] to stimulate neonatal cardiomyocyte proliferation and to promote entry into mitosis depends on the WW domains, which mediate the binding to MMB. To do so, we transduced P1 cardiomyocytes with adenoviral constructs encoding for LacZ or YAP[S127A] (Fig 8A) and measured induction of mitosis and proliferation by phospho-H3 and Ki67 staining. While YAP[S127A] robustly induced mitosis and proliferation, YAP[S127A] lacking the WW domains (YAP[S127A/ΔWW]) failed to increase the fraction of pH3 and Ki67-positive cardiomyocytes (Fig 8B and 8C, S7A and S7B Fig). Similarly, a point mutant of YAPS127A deficient in binding to TEAD (YAP[S127A/S94A]), was also not able to increase proliferation and mitosis, indicating that these functions of YAP are TEAD dependent (Fig 8A-8C, S7A and S7B  Fig).
We next asked whether the interaction between YAP and B-MYB is critical for the high proliferation rate of embryonal cardiomyocytes. To address this question, we infected embryonal cardiomyocytes with an adenovirus expressing MY-COMP coupled to GFP through a T2A self-cleaving peptide. Infected cardiomyocytes were detected by their green fluorescence and by staining for cTnT ( S7C Fig). Strikingly, staining for phospho-H3 showed that expression of MY-COMP strongly suppressed mitosis of embryonal cardiomyocytes (Fig 8D, S7C Fig). Importantly, this effect was diminished by deletion of the PPXY motif in MY-COMP, indicating that the ability to prevent proliferation correlates with the ability to disrupt the YAP-B-MYB interaction. The effect of MY-COMP is not due to interference with the DNA-binding of B-MYB, because expression of MY-COMP with a mutation in the DNA-binding domain that has been shown to prevent the interaction with DNA [32], showed the same phenotype. Taken together these observations are consistent with the notion that the YAP-dependent neonatal cardiomyocyte proliferation is mediated by the interaction of YAP with MMB.

MMB target genes are downregulated in differentiated cells and reactivated by YAP
In another line of evidence, we investigated whether YAP is able to regulate MMB target genes in C2C12 cells. Co-immunoprecipitations confirmed that B-MYB and YAP interact in C2C12 myoblasts (Fig 9A). The MMB target genes CDC20 and TOP2A were expressed at high levels in asynchronously growing C2C12 cells and were downregulated during myogenic differentiation (Fig 9B). Expression of a constitutive active YAP(S127A) partially reverted the downregulation of these target genes in differentiated cells (Fig 9B). We next asked whether the reinduction of cell cycle genes in differentiated C2C12 cells is dependent on the YAP WW domains. C2C12 cells were differentiated into myotubes and then transduced with adenoviral constructs encoding for LacZ or for YAP mutants. The expression of YAP mutants was validated by immunoblotting (Fig 9C). Next, expression of cell cycle genes and YAP target gene was analyzed by RT-qPCR. Importantly, like the induction of proliferation of cardiomyocytes, the upregulation of MMB target genes in differentiated C2C12 cells by YAP was dependent on the WW and the TEAD-binding domains (Fig 9D). Strikingly activation of YAP target genes that are not regulated from distant enhancers but by binding of YAP to the proximal promoter, such as Ctgf and Cyr61 and Amotl2 (see Fig 2G and S2D Fig), did not require the WW domains, indicating that their induction is independent from binding to MMB (Fig 9D).
We next asked how the expression of B-MYB and MMB target genes is regulated during heart development. Hearts were collected at embryonic stage E16.5 and at postnatal days P1 and P10. mRNA was isolated and subjected to RT-qPCR. A panel of YAP-MMB target genes such as Cenpf, Nusap1, Top2a and Birc5 were expressed at high levels in embryonic hearts when cardiomyocytes still divide but strongly declined in P1 and P10 hearts when cardiomyocytes differentiate and exit the cycle (S8A Fig). The downregulation of TOP2A, NUSAP1 and CENPF in adult hearts was also observed on protein level (S8B Fig). Mybl2 also declined at P1 compared to E16.5 and further decreased at P10, a pattern that was also confirmed on protein level (S8A and S8B Fig). Thus, downregulation of MMB target genes correlates with the postnatal cell cycle exit of cardiomyocytes. Conversely, Mybl2 and mitotic MMB target genes were expressed at higher levels in P1 Sav1 knockout hearts when YAP is activated compared to wildtype hearts (S8C Fig). Together these data support the view that MMB contributes to the induction of cell cycle genes in response to Hippo-deficiency or YAP activation.

Discussion
Recent studies have shown that the Hippo-YAP pathway plays essential roles during heart development and heart regeneration [4]. While the deletion of YAP in embryonic heart impairs cardiomyocyte division, the expression of constitutively active YAP promotes cardiomyocyte proliferation [5,9,33,34]. We now show that MMB is required for the ability of YAP to induce cell division in this tissue ( Fig 9E). First, we used a mouse model in which cell cycle genes were induced in embryonic hearts by activation of YAP through the knockout of the Hippo pathway protein Sav1. These genes were not induced when the MuvB subunit Lin9 was deleted together with Sav1. Proliferation of cardiomyocytes is also dependent on both MMB and YAP, since deletion of Lin9 abolishes the enhanced proliferation induced by deletion of Sav1. This requirement is not limited to the embryonic heart but the same dependency on Lin9 for cardiomyocyte proliferation was also observed in postnatal hearts using Lin9 / Sav1 double knockout mice expressing the Cre transgene from the alpha-MHC promoter. Further evidence for this dependency on LIN9 comes from experiments using cultured embryonal or postnatal cardiomyocytes with a conditional Lin9 allele and a hormone inducible Cre recombinase. Adenoviral expression of activated YAP was only able to induce mitosis when Lin9 was present but not after Cre-mediated deletion of Lin9. Although it is possible that YAP and MMB could independently induce a similar set of genes required for cell cycle regulation, our findings that B-MYB and YAP directly interact and that YAP induced cardiomyocyte proliferation depends on its WW domains supports the notion that the synergy is a result of the interaction between YAP and MMB. Furthermore, cardiomyocyte mitosis is inhibited by MY-COMP, a fragment of B-MYB that contains the YAP binding domain of B-MYB. When fused to a nuclear localization signal and expressed in cells, MY-COMP interferes with the association YAP to B-MYB.
Genome binding studies in embryonic hearts demonstrated that LIN9 binds to the promoters of cell cycle genes activated by YAP, indicating that these genes are direct targets of MMB. By ChIP-seq, YAP does not colocalize with LIN9 to the promoters of cell cycle genes but instead binds to enhancers, consistent with recent data from human cancer cell lines where YAP also predominantly binds to distant sites [24][25][26]. This suggests that YAP interacts with MMB-regulated promoters through chromatin looping, resulting in the activation of a subset of cell cycle genes, similar to what we have recently described for human MCF10A cells [14]. Given that YAP not only interacts with MMB but that the MMB subunit B-MYB is also a transcriptional target of YAP, one can speculate that YAP and B-MYB form a positive feedback loop enabling expression of downstream cell cycle genes.
Remarkably, although LIN9 was required for cardiomyocyte cell cycle induction due to activated YAP, LIN9 was completely dispensable for homeostasis of the adult murine heart. This absence of a phenotype in adult cardiac-specific Lin9 knockout mice was surprising, because LIN9 is a key subunit of the DREAM complex that is involved in repression of cell cycle genes in non-dividing cells [15]. The observation that the genetic inactivation of Lin9 in the postnatal heart is not associated with ectopic proliferation of cardiomyocytes suggests that loss of DREAM does not lead to sufficient de-repression of cell cycle genes to induce cardiomyocyte proliferation. It is therefore possible that DREAM does not permanently silence cell cycle genes, but keeps them in a poised state for re-activation by pro-proliferative signals, consistent with our finding that LIN9 remains bound to promoters in postnatal hearts. Also supporting this possibility is finding that cell cycle promoters, as opposed to promoters of genes that regulate cell-cell contacts and the cytoskeleton, are readily accessible in adult murine hearts before introduction of a YAP5SA transgene [6]. Thus, MuvB complexes may have a dual role in cardiomyocyte proliferation, contributing to the inactive, but primed state of cell cycle genes as well as to their YAP-mediated activation. It will be important to investigate whether epigenetic changes at MuvB bound promoters contribute to the loss of the regeneration capacity of adult hearts [35].
YAP not only plays a role in cardiac development but is also a potent oncogene in different cancers. Previous findings have implicated the YAP WW domain in oncogenic transformation [36]. Targeting WW domain mediated interactions by MY-COMP, or by more potent inhibitors based on MY-COMP, may prove a valid strategy to inhibit the oncogenic pro-proliferative functions of YAP. Such inhibitors could serve as selective therapeutic agents in cancers with high levels of YAP.

Ethics statement
All animal studies were carried out in accordance with national and institutional guidelines and were approved by the regional state administration agency for Würzburg (approval number 55.2-2532-2-293). Euthanasia was by cervical dislocation of adults and decapitation of embryos.

Mice
Sav1 tm1.1Dupa /J mice were obtained from Jackson laboratories, in which LoxP sites flank exon 3 of the Sav1 gene [23]. We have previously descripted Lin9 fl mice, in which LoxP sites flank exon 7 of the Lin9 gene [37]. Cardiomyocyte-specific deletion of Sav1 and Lin9 was achieved by crossing mice to Nkx2.5-Cre mice [38]. Postnatal cardiomyocyte-specific deletion of Sav1 and Lin9 was achieved by crossing mice to α-MHC-Cre mice [27]. To obtain Lin9 fl/fl CreER T2 , Lin9 fl mice were crossed with a mouse line ubiquitously expressing CreER T2 transgene [39]. Mice were maintained in a C57/Bl6 background.

Cell culture
HEK293A (Thermo Fisher Scientific) cells, HeLa (ATCC CCL-2, female) cells and C2C12 (ATCC CRL-1772) cells were maintained in DMEM supplemented with 10% FCS (Thermo Fisher Scientific) and 1% penicillin/streptomycin (Thermo Fisher Scientific). Differentiation of C2C12 cells was induced by culturing in DMEM medium with 2% horse serum (Sigma). Primary embryonic and postnatal cardiomyocytes were isolated by enzymatic digestion with the Pierce Primary cardiomyocyte isolation kit (Thermo Fisher Scientific). Cardiomyocytes were enriched by preplating on tissue-culture plastic to remove nonmyocytes. Cardiomyocytes were initially cultured for 48h on fibronectin-coated dishes with 5% horse serum (Sigma) and 1% penicillin/streptomycin (Thermo Fisher Scientific) and cardiomyocyte growth supplement at 37˚C and 5% CO2 to prevent proliferation of nonmyocytes. To induce the deletion of Lin9, conditional cardiomyocytes were treated with 100 nM 4-hydroxytamoxifen (Sigma) for 4 days. Cardiomyocytes were transduced with adenovirus (multiplicity of infection, 25) in serum-free medium for 24 h and cultured for additional 48 h.

RT-qPCR
Total RNA was isolated using peqGOLD TriFast (Peqlab) according to the manufacturer's instructions. RNA was transcribed using 100 units RevertAid reverse transcriptase (Thermo Fisher Scientific). Quantitative real-time PCR reagents were from Thermo Fisher Scientific and real-time PCR was performed using the Mx3000 (Stratagene) detection system. Primer sequences are listed in S1 Table. Expression differences were calculated as described before [20].
Following deparaffinization, antigen retrieval was performed by boiling samples in 10mM sodium citrate buffer (pH6.0) for 10 minutes in a microwave oven. After 30 minutes cool down, samples were blocked with 3% BSA in PBS-T (0.1% Tween-20 (AppliChem)) and incubated with the primary antibodies diluted in PBS-T over night at 4˚C. The following antibodies were used: TroponinT (CT3) (developmental studies hybridoma bank) 1:50, phospho-Histone H3 (Ser10) (Santa Cruz Biotechnology, sc-8656) 1:100, Ki-67 (SP6) (Thermo Scientific, RM-9106) 1:200. After three washing steps with PBS-T, secondary antibodies conjugated to Alexa 488 and 594 (Thermo Fisher Scientific) and Hoechst 33258 (Sigma) were diluted 1:200 in 3% BSA in PBS-T and incubated with the coverslips for 2 hours at room temperature. Finally, slides were washed three times with PBS-T and mounted with Immu-Mount (Thermo Fisher Scientific). Pictures were taken with an inverted Leica DMI 6000B microscope equipped with a Prior Lumen 200 fluorescence light source and a Leica DFC350 FX digital camera.
For morphometric measurements, we measured transverse sections at the level of atrioventricular valves. Wall thickness was calculated as the ventricular compact myocardial thickness divided by its outer circumference. Myocardial area was quantified in each section using Image J as previously described [43].

siRNA transfection of embryonal cardiomyocytes
Double-stranded RNA was purchased from Eurofins. siRNAs were transfected in a final concentration of 30 nM using RNAiMAX (Thermo Fisher Scientific) according to the manufacturer's protocol. siRNA sequences are listed in S1 Table [44,45].

Flow cytometry
Cardiomyocytes were isolated and enriched with the Pierce Primary cardiomyocyte isolation kit (Thermo Fisher Scientific) according to the manufacturer's instructions. Isolated and enriched cardiomyocytes were fixed with 4% PFA in PBS for 5 minutes. Detection was performed using a Beckman Coulter FC 500 cytomics and data were analyzed in CXP analysis 2.2 software. Gating and compensation were based on fluorophore-negative controls. 10.000 cells were analyzed per genotype.

RNA-Seq
For whole transcriptome analysis, total RNA was isolated in triplicates from ventricular heart tissue with the desired genotype. DNA libraries were generated using 1μg RNA with the magnetic mRNA isolation module and NEBNext Ultra II RNA Library Prep Kit for Illumina (New England Biolabs). DNA library was amplified by 7 PCR cycles and quality was analyzed using the fragment analyzer (Advanced Analytical). Libraries were sequenced on the NextSeq 500 platform (Illumina).
μSPOT slides were blocked by incubation with 2% (w/v) BSA in PBS for 60 min. Afterwards, slides were incubated with YAP-WW1/2 or GST protein in 0.1% BSA in PBS for 30 min. Slides were washed 6 times with PBS for 5 min and then incubated with 1:10,000 diluted primary antibody (Anti-GST HRP conjugated (RPN1236V; Sigma Aldrich) in 0.1% BSA in PBS for 30 min, after which the slides were washed 6 times with PBS for 5 min. Peptide binding was detected through chemiluminescent detection with SuperSignal West Femto Maximum Sensitive Substrate (Thermo Scientific) using a c400 imaging system (Azure). Binding intensities were evaluated using FIJI including the Microarray Profile addon (OptiNav). The error range and standard deviation were defined by comparing the intensities of each peptide duplicate on the respective array.

Bioinformatic analysis
After sequencing, bases were called using Illuminas GenerateFASTQ v1.1.0.64 software and sequencing quality was controlled with the FastQC script. For RNA-seq, reads were mapped with TopHat v2.1.0 [47] and BOWTIE2 [48] with default parameters to the murine genome (mm10). Samples were normalized to the sample with the smallest number of mappable reads and a read count matrix for each Ensembl gene (GRCm38.p6) was generated with the sum-marizeOverlaps function in the R package {GenomicFeatures}. Before differential gene expression analysis, non-or weakly expressed genes were removed using the threshold: mean read count per gene over all samples >1. Differentially expressed genes were called with EdgeR and p-values were adjusted for multiple-testing with the Benjamini-Höchberg procedure (FDR: false discovery rate). Gene set enrichment analyses were performed with signal2noise metric, 1000 permutations and a combined gene set database comprising Hallmark and C2 gene sets.
For ChIP-seq, sequenced reads were mapped to the Mus musculus genome mm10 with BOWTIE v1.2 [48] with default parameters and subsequently normalized to the sample with the smallest number of sequenced reads. Peaks were called with MACS14 [49] with maximal 3 duplicates, a p-value cut-off of 1e-5 and the input sample as control. Resulting peaks were annotated to the next transcriptional start of Ensembl genes with the closestBed function from the bedtools suite v2.26.0 [50]. Overlapping peaks were identified with bedtools intersect and a minimal overlap of 1bp. LIN9 occupancy was calculated in a window of +/-1kb around TSSs with bedtools coverage function. Density matrices were generated with deeptools v2.3.5. [51] computeMatrix function at a resolution of 1bp and subsequently used for plotting heat maps with plotHeatmap. Mapped ChIP-seq data for histone marks were taken from the ENCODE portal [52] (https://www.encodeproject.org/) with the following identifiers: ENCFF056JGV, ENCFF295HNV, ENCFF642EEK, ENCFF687BWU. Promoters, enhancers and super-enhancers were defined as described previously [14].
In box plots, the central line reflects the median, the borders of the boxes indicate the first and third quartile and whiskers extend to 1.5 of the interquartile range. Outliers are shown as dots. P-values were calculated with a two-tailed Wilcoxon rank sum test (unpaired samples) or Wilcoxons signed-rank test (paired samples). ChIP-and RNA-sequencing datasets are available at the NCBI's Gene Expression Omnibus [53] under the accession number GEO: GSE137132.  Fig 5A. p-values were calculated using a permutation test with 1000 permutations. "Signal2Noise"was used as a metric to rank genes. ES: enrichment score. B) Heatmap depicting the mRNA expression of LIN9 regulated cell cycle genes in hearts of Nkx2.5-Cre mice with the indicated genotypes as determined by RNA-seq. C) Heat map documenting binding of LIN9 and YAP at LIN9 peaks in promoters or at YAP peaks in enhancers and superenhancers in E16.5 heart ventricles. Read density is plotted in a window of +/-2kb around the peak at a resolution of 2bp. Data for histone modifications are taken from ENCODE. D) Genome browser tracks illustrating the binding of LIN9 to the Mybl2, Anln and Top2a promoter and binding of YAP to the Cyr61 and Ctgf promoter and to an intergenic enhancer on chromosome 1. ChIP-seq data for histone modifications are from ENCODE (GSE31039). E) GSEA comparing expression differences in Nkx2.5-Cre; Lin9 fl/fl (LIN9 KO) and Nkx2.5-Cre; Lin9 fl/+ (LIN9 wt) heart ventricles from E13.5 mice in two biological replicates (each done in triplicate). The C2 MSigDB was spiked with the Hallmark gene sets and a set of LIN9 direct targets genes from [14]. Gene sets related to respiration/TCA cycle ("oxphos") and hematopoietic cells are highlighted in blue and orange, respectively. NES: normalized enrichment score. F) Representative gene sets from the analysis in C. p-values were calculated using a permutation test with 1000 permutations. "Signal2Noise"was used as a metric to rank genes. ES: enrichment score.  See Fig 3G. E) The expression of Lin9 relative to Actin and Hprt was investigated in E16.5, P1 and P10 hearts by RT-qPCR. n = 3 independent replicates. F) The expression of LIN9 in lysates prepared from hearts at the different developmental stages was investigated by immunoblotting. β-actin served as a loading control. G) Heat map documenting binding of LIN9 at LIN9 peaks in promoters called in E16.5 or P10 cardiomyocytes or overlapping peaks. Read density is plotted in a window of +/-2kb around the peak at a resolution of 2bp. Data for histone modifications are taken from ENCODE (GSE31039). H) Venn diagram depicting the common LIN9 peaks in E16.5 and P10 hearts. The number in brackets refers to the peaks located in promoters. I) Plot illustrating the genomic localization of LIN9 in postnatal (P10) heart ventricles as determined by ChIP-seq. J) Histogram showing the absolute distance between overlapping LIN9 peaks called in E16.5 and P10 heart ventricles located in promoters (n = 1,458) at a resolution of 20 bp.