YAP1 and TAZ negatively control bone angiogenesis by limiting hypoxia-inducible factor signaling in endothelial cells

Blood vessels are integrated into different organ environments with distinct properties and physiology (Augustin and Koh, 2017). A striking example of organ-specific specialization is the bone vasculature where certain molecular signals yield the opposite effect as in other tissues (Glomski et al., 2011; Kusumbe et al., 2014; Ramasamy et al., 2014). Here, we show that the transcriptional coregulators Yap1 and Taz, components of the Hippo pathway, suppress vascular growth in the hypoxic microenvironment of bone, in contrast to their pro-angiogenic role in other organs. Likewise, the kinase Lats2, which limits Yap1/Taz activity, is essential for bone angiogenesis but dispensable in organs with lower levels of hypoxia. With mouse genetics, RNA sequencing, biochemistry, and cell culture experiments, we show that Yap1/Taz constrain hypoxia-inducible factor 1α (HIF1α) target gene expression in vivo and in vitro. We propose that crosstalk between Yap1/Taz and HIF1α controls angiogenesis depending on the level of tissue hypoxia, resulting in organ-specific biological responses.


Introduction
The skeletal system is surprisingly dynamic and undergoes lifelong remodeling even after the completion of developmental growth. The local vasculature plays central roles in embryonic and postnatal bone development but also later in homeostasis and in repair processes (Gerber and Ferrara, 2000;Maes and Clemens, 2014;Zelzer and Olsen, 2005). Endochondral ossification and, in particular, the formation of ossification centers in the mouse embryo requires the ingrowth of blood vessels into aggregates of hypertrophic chondrocytes that express pro-angiogenic signals such as vascular endothelial growth factor A (VEGF-A) (Eshkar-Oren et al., 2009;Maes et al., 2010). The invasion of blood vessels also facilitates the entry of osteoblast precursors into fractured bone and, accordingly, angiogenesis is essential for bone repair and regeneration (Maes et al., 2010;Stegen et al., 2015). It was also shown that capillaries in bone are regionally and molecularly specialized. With the help of advanced bone sample processing, staining and imaging protocols (Kusumbe et al., 2015), we found initially two distinct subpopulations of bone endothelial cells (ECs) that can be distinguished with the cell surface molecules CD31/Pecam1 and endomucin (Emcn) . Columnar vessels in the metaphysis and the capillaries of the endosteum, a connective tissue layer lining the inner surface of compact bone, are associated with osteoprogenitor cells expressing the transcription factor Osterix, express high levels of CD31 and Emcn (CD31 hi Emcn hi ), and have therefore been named type H. In contrast, the highly branched sinusoidal capillary network of the diaphysis shows comparably lower expression of the two markers (CD31 lo Emcn lo or type L) and is predominantly associated with reticular mesenchymal cells and hematopoietic cells Ramasamy et al., 2016b). A further EC subpopulation with high CD31 expression -termed type E -is abundant in embryonic and early postnatal long bones but gradually disappears in adolescent mice (Langen et al., 2017). Indicating that they are source of signals controlling the behavior of perivascular mesenchymal cells, type H and type E but not type L ECs shift the differentiation of co-cultured primitive mesenchymal cells towards the osteoblast lineage (Langen et al., 2017). Interestingly, CD31 hi Emcn hi ECs are strongly reduced in ovariectomized mice, a model of human postmenopausal osteoporosis, as well as in aging mice, which show an agedependent decline in bone mineral density similar to human subjects Xie et al., 2014). Notch and hypoxia-inducible factor (HIF) signaling in ECs promote vessel growth in bone and help to maintain a functional type H EC population. Activation of these pathways in the ECs of aged mice induces the reappearance of CD31 hi Emcn hi together with perivascular Osterixpositive cells and is sufficient to increase the formation of mineralized bone in the metaphysis Ramasamy et al., 2016b). Together, these reports indicate a strong coupling of EC behavior and osteogenesis in the skeletal system, which presumably involves a network of interdependent molecular interactions such as growth factors and matrix molecules.
Yes-associated protein 1 (Yap1) and WW domain containing transcription regulator 1 (WWTR1), which is more widely known as Taz, are closely related transcriptional coactivators and key components of the Hippo pathway, an evolutionarily conserved signaling cascade controlling cell proliferation, differentiation and organ size (Dong et al., 2007;Huang et al., 2005). In mammals, the core module of this pathway consists of the serine/threonine kinases Stk3 (Mst2) and Stk4 (Mst1), which are orthologues of Drosphila Hippo, the large tumor suppressor homolog 1/2 (Lats1/2) kinases, and their interaction partners Salvador (Sav1) and MOB kinase activator 1A/B (MOB1A/B) (Piccolo et al., 2014;Yu and Guan, 2013;Zhao et al., 2011). Activation of Hippo signaling leads to exclusion of Yap1/Taz from the nucleus and promotes the proteolytic degradation of these proteins. In the cell nucleus, Yap1/Taz interact with transcription factors, such as the TEA domain family members Tead1-4, and thereby regulate gene expression and promote growth processes (Yu and Guan, 2013). Accordingly, enhanced expression and nuclear localization of Yap1/Taz were observed in multiple human cancers (Moroishi et al., 2015) and inactivation of upstream Hippo signaling leads to tumor formation (Lu et al., 2010).
Here, we show that Yap1/Taz are critical regulators of vessel growth in an organ-dependent fashion. While inducible inactivation of the two genes in postnatal ECs results in the expected reduction of angiogenesis in the retinal vasculature, vessel growth and the abundance of CD31 hi Emcn hi ECs are significantly increased in long bone. This also leads to an increase in metaphyseal bone formation, while EC-specific overexpression of a stabilized version of Yap1 or inactivation of the upstream kinase Lats2 impair angiogenesis. Furthermore, we show that Yap1/Taz negatively regulate the activity of the HIF pathway and thereby limit the expression of endothelial genes associated with vessel growth.

Hypoxia and Yap1/Taz expression in bone endothelium
Different organs exhibit substantial variation in fundamental parameters such as tissue oxygenation ( Figure 1A,B; Figure 1-figure supplement 1A), which is frequently enhanced in response to injury or in disease conditions (De Santis and Singer, 2015;Samaja, 1988). These differences are likely to reflect organ-specific features such as local oxygen consumption, blood flow, vascular architecture, and vessel diameter (Figure 1-figure supplement 1B,C). In comparison to other organs, such as brain, heart, lung, liver, spleen, or kidney, postnatal long bone contains extensive hypoxic areas. But even within long bone, pimonidazole (hypoxyprobe) staining or immunohistochemical analysis of the oxygen-controlled transcription factors hypoxia-inducible factor 1a (HIF-1a) and 2a (HIF-2a) uncover striking regional differences in oxygenation. The metaphyseal region in proximity of the growth plate is less hypoxic than the secondary ossification center (SOC) or the bone marrow in the diaphysis  Figure 1. Regional differences in hypoxia and Yap1/Taz expression in bone. (A) Tile scan maximum intensity projection of P21 femur with Pimonidazole (green) and DAPI (blue) staining. (B) Quantification of Pimonidazole staining intensity (artificial units, a.u.) in different organs. (C, D) Regional differences in Pimonidazole (C) and HIF2a (D) staining levels in metaphysis (mp) and diaphysis (dp). (E) Principal component analysis of RNA sequencing data using most variable genes across the samples. The first principal component (PC1) explains 63% of all variance; and PC2 13% of the variance between  Figure 1-source data 1). As the consequences of such regional differences are not understood, we isolated endothelial cells (ECs) expressing membrane-anchored mTomato fluorescent protein and nuclear H2B-GFP from Cdh5-mT/nG transgenic reporter femoral metaphysis and diaphyseal bone marrow cavity by fluorescence-activated cell sorting (FACS) at high purity for RNA sequencing ( In line with the regional differences in oxygenation, which are likely to limit the ability to metabolize fatty acids, genes belonging to the glycolytic pathway are higher in bmECs than in mpECs (Figure 1-figure supplement 2I and Figure 1-figure supplement 1-source data 1). Our analysis also uncovered that target genes of Hippo signaling, an evolutionarily conserved pathway that promotes cell proliferation and organ growth, are upregulated in bmECs ( Figure 1G). In sections of P21 femur, immunostaining of Yes-associated protein 1 (Yap1), a transcriptional coregulatory in the Hippo signaling cascade, is more prominently visible in BM sinusoidal ECs, whereas expression in the metaphysis predominates in perivascular cells ( Figure 1H; Figure 1-figure supplement 3A). Immunosignals for the Yap1-related protein Taz mainly decorate non-endothelial cells in the metaphysis and arterial ECs ( Figure 1H; Figure 1-figure supplement 3B). Given that proteolytic degradation limits the biological activity of Yap1/Taz (Piccolo et al., 2014;Yu and Guan, 2013;Zhao et al., 2011), we examined whether acute inhibition of proteasome-dependent degradation would highlight sites where Yap1/Taz levels are actively suppressed. Treatment of 3-week-old mice with the proteasome inhibitor MG132 for 3 hr profoundly increased endothelial and perivascular Yap1 and Taz protein signals in the metaphysis but had relatively limited effects in the diaphysis ( Figure 1I; Figure 1-figure supplement 3C). Further arguing for rapid degradation of Yap1/Taz in the metaphysis, strong phospho-Yap1 S127 immunosignals can be seen in metaphyseal capillaries and associated perivascular cells, which is accompanied by high expression of Lats2 protein in metaphyseal but not diaphyseal vessels ( Figure 1-figure supplement 3D). In contrast, Yap1, Wwtr1, Lats1, and Lats2 transcripts are not significantly different between EC subsets (Figure 1-figure supplement 3E). In Cdh5-mT/ nG reporter mice, Yap1 and Taz immunostaining is prominent in the H2B-GFP+ nuclei of bmECs or in cultured bone ECs in vitro, whereas comparably little signal can be seen in mpECs ( Figure 1J,K; Figure 1-figure supplement 3F). Together, these data indicate that Hippo signaling is active in ECs of the metaphysis leading to rapid degradation of Yap1/Taz in these cells. Arrowheads highlight expression in Emcn+ (red) ECs. Nuclei, DAPI (blue). (I) Immunostaining of Yap1 and Taz in the control (vehicle) and MG132 proteasome inhibitor-treated femoral metaphysis. (J) Nuclear localization (arrowheads) of Yap1 (green) in H2B-GFP+ EC nuclei (shown in red) in 3-weekold Cdh5-mTnG femoral metaphysis (mp) and bone marrow (bm). Higher magnification image shows strong Yap1 and Taz nuclear signals bmECs. (K) Mean intensity (a.u.) of Yap1 and Taz nuclear localization signals in bm and mp ECs. (n = 4; 48 cells in total; data are presented as mean ±sem, P values, two-tailed unpaired t-test). The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Source data for Figure 1B,C,D,F,G,K.      Function of Yap1/Taz in organ-specific angiogenesis Next, we used tamoxifen inducible Cdh5-Cre-ERT2 mice, which allow efficient genetic experiments in all bone ECs in vivo Langen et al., 2017), to generate EC-specific Yap1 and Taz double loss-of-function mutants (Yap1/Taz iDEC ). Following administration of tamoxifen from postnatal day (P) 6-8, Yap1/Taz iDEC long bones at P21 display a much denser vessel network with more endothelial buds and column, a feature of actively growing bone vessels (Ramasamy et al., 2016a), relative to littermate controls ( A second administration regime, in which tamoxifen is given from P1-3, is incompatible with survival of Yap1/Taz iDEC animals until P21 and therefore these mutants were analyzed at P18. This showed that the density of Yap1/Taz iDEC bone vessels is strongly increased at P18, which is particularly evident at the interface (transition zone) between the metaphyseal and diaphyseal capillary networks (Langen et al., 2017), and sinusoidal marrow vessels are enlarged (  Ramasamy et al., 2016b), are also increased in Yap1/Taz iDEC long bone ( Figure 2A,B,E,F). The analysis of EC subpopulations by flow cytometry (Figure 2-figure supplement 2D) and immunostaining revealed that the increase is more pronounced for Yap1/Taz iDEC total and type H ECs, but also the fraction of type L ECs is significantly larger than in littermate controls ( Figure 2E; Figure  Further arguing that Yap1/Taz play an unusual role as negative regulators of bone angiogenesis, Cdh5-CreERT2-controlled and thereby inducible overexpression of a stabilized version of Yap1 (Yap1 S112A ) in ECs (Yap1-KI iEC ) (Figure 2-figure supplement 2A-C) impairs growth of the bone vasculature. In P21 Yap1-KI iEC femur, distal vessel buds and arches, metaphyseal type H vessel columns, and EC proliferation are reduced ( Figure 2G-L). To study the upstream regulation of Yap1/ Taz activity in bone ECs, we generated tamoxifen-inducible EC-specific Lats2 iDEC mutants by combining Cdh5-Cre-ERT2 transgenic mice and animals carrying loxP-flanked Lats2 alleles (Lu et al., 2010). The absence of Lats2 in ECs causes a striking reduction of P21 femur and tibia length and weight ( Figure  The murine retina is a widely used model of postnatal sprouting angiogenesis, in which the growth of an initially two-dimensional superficial vessel plexus can be easily monitored from P1 to P7 (Gerhardt et al., 2003;Pitulescu et al., 2010). Immunostaining of Yap1 and Taz in P6 retina confirmed expression of both proteins in the vascular endothelium ( . Metaphysis (mp), transition zone (tz), diaphysis (dp), and growth plate (gp) are indicated. Note the increased number of Yap1/Taz iDEC vessel buds and columns (B) compared to littermate control (n = 6, data are presented as mean ±sem. P values, two-tailed unpaired t-test). (C, D) Representative confocal image of Emcn+ (red) proliferating (EdU, green) mpECs. Nuclei, DAPI (blue) (C). Quantification of EdU + Emcn + ECs in Yap1/Taz iDEC and control metaphysis (D), (control n = 6 and Yap1/Taz iDEC n = 7, data are presented as mean ±sem. P values, two-tailed unpaired t-test). (E, F) Maximum intensity projections of Emcn hi (red) CD31 hi (green) vessels in the P21 Yap1/Taz iDEC and control femur (E). Metaphyseal column length is significantly increased in Yap1/Taz iDEC mutant compared to control femur (F) (n = 6 data are presented as mean ±sem. P values, twotailed unpaired t-test). (G, H) Representative confocal images of control and Yap1-KI iEC femur. Emcn+ (red) ECs and nuclei (DAPI, blue) are stained (G). Vessel buds and columns are reduced in the Yap1-KI iEC metaphysis relative to littermate control (H) (n = 5, data are presented as mean ±sem. P values, two-tailed unpaired t-test). (I, J) Maximum intensity projection of Emcn hi (red) CD31 hi (green) vessels in P21 Yap1-KI iEC and control femur. The vasculature of the metaphysis (mp) (arrows; dashed lines), the transition zone (tz) connecting the mp to the diaphysis (dp) and arteries (arrowheads) are reduced in Yap1 gain-of-function femur (I). The length of the Yap1-KI iEC Emcn hi CD31 hi vessel columns in femur is significantly reduced (J) (control n = 4 data are presented as mean ±sem. P values, two-tailed unpaired t-test). (K, L) Representative confocal image of proliferating ECs (Emcn, red; EdU, green) in femoral metaphysis. Nuclei, DAPI (blue) (K). Quantification of EdU + Emcn + ECs in Yap1-KI iEC and control metaphysis (L) (control n = 4 data are presented as mean ±sem. P values, two-tailed unpaired t-test).
The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. Source data for Figure 2B,D,F,H,J,L.    Figure 1J,K), the loss of Ctgf and Cyr61 expression and the upregulation of Vegfa and Angptl4 are more pronounced in freshly sorted Yap1/Taz iDEC type L ECs relative to type H ECs. Conversely, Ctgf and Cyr61 transcripts are more strongly increased and Vegfa and Angptl4 reduced in Lats2 iDEC type L ECs relative to type H ECs ( Figure 4H,I).
To understand the underlying mechanism for the surprising function of Yap1/Taz in bone ECs, we used an independent approach in cultured human umbilical vein endothelial cells (HUVEC). HUVECs grown under hypoxic (1% O 2 ) conditions strongly upregulate VEGFA and ANGPTL4 expression relative to normoxia (21% O 2 ). Hypoxia also leads to strong upregulation of the Yap1/Taz target genes CTGF and CYR61 ( Figure 5A) and, consistently, Yap1/Taz nuclear localization is increased under these conditions ( Figure 5B). In vitro siRNA-mediated knockdown (KD) of human YAP1 and TAZ efficiently reduces their expression at the transcript and protein level (  stained with Emcn+ (red) vasculature, and nuclei in DAPI (blue) (D). Vessel buds and columns are strongly reduced in Lats2 iDEC mutants compared to littermate controls (E). (n = 6, data are presented as mean ±sem. P values, two-tailed unpaired t-test). (F) Lats2 iDEC distal vessel buds and arches switch to a tip-like morphology (arrow). The Cdh5-mTnG reporter (red and green) visualizes nuclear fragmentation (arrowheads) in Lats2 iDEC ECs but not in control. Taz immunosignal is strongly increased in Lats2 iDEC mutants. (G) Yap1 (green) immunosignal is enhanced in Lats2 iDEC metaphyseal vessels (Emcn, red). (H) Confocal image of control and Lats2 iDEC femoral metaphysis stained for Emcn (red), CD31 (green), and nuclei (DAPI, blue). (I, J) Representative confocal image of proliferating (Emcn+, red; EdU+, green) ECs in metaphysis. Nuclei (DAPI, blue) (I). Quantification of EdU + Emcn + ECs in Lats2 iDEC and control metaphysis (J) (control n = 4 and Lats2 iDEC n = 4 data are presented as mean ±sem. P values, two-tailed unpaired t-test). (K, L) Apoptotic Emcn+ (red) and active caspase-3+ (aCasp-3, green) ECs in metaphysis. Nuclei (DAPI, blue) (K). Quantification of Emcn + aCasp-3 + ECs (L) (control n = 4 and Lats2 iDEC n = 4 data are presented as mean ±sem. P values, two-tailed unpaired t-test). (M) Representative confocal image of Lats2 iDEC Cdh5-mTnG femur compared to littermate control. The online version of this article includes the following source data and figure supplement(s) for figure 3: Source data 1. Source data for Figure 3A,C,E,J,L.         Figure 5D). To test whether the siYAP/TAZ-induced upregulation of VEGFA requires HIF1a, we used siRNA to interfere with HIF1A expression in HUVECs. Indeed, the HIF1a target genes VEGFA, ANPTL4, IGFBP2 and XBP1 are no longer upregulated after triple knockdown of HIF1A together with YAP1/TAZ ( Figure 5E; Figure 5-figure supplement 1C). Moreover, ChIP experiments show that the binding of HIF1a to a hypoxia response element in the VEGFA gene (position À947) is significantly increased after YAP1/TAZ knockdown ( Figure 5F). Highlighting the crosstalk between Yap1/Taz and HIF1a further, EC-specific inactivation of the Hif1a gene impairs the expression of Vegfa and Angptl4 both in type H and type L ECs in vivo, but it also reduces the expression of Ctgf and Cyr61 in type L cells ( Figure 5-figure supplement 1D). HIF is stabilized under hypoxic conditions, translocates to the nucleus and regulates gene expression to stimulate angiogenesis and other processes (Pugh and Ratcliffe, 2003). In contrast, HIF a subunits are modified by prolyl-hydroxylases and thereby marked for rapid proteolytic degradation in normoxia (Aragonés et al., 2009;Safran and Kaelin, 2003). Cdh5-CreERT2-controlled and thereby EC-specific expression of a stabilized HIF1a mutant (Hif1AdPA iEC ) lacking critical prolylhydroxylation sites (Kim et al., 2006) increases the abundance of metaphyseal type H capillaries, of columns and buds in proximity of the growth plate, and of arteries ( Figure 6A-C), which is consistent with previously reported findings on the role of HIF signaling in bone ECs . Supporting the crosstalk between Yap1/Taz and HIF-1a in vivo, the combination of Yap1/Taz i-DEC and Hif1a iDEC loss-of-function alleles normalizes the increase in bone vascularization seen in Yap1/Taz iDEC samples ( Figure 6D,E).

Endothelial Hippo signaling controls osteogenesis
Based on previous studies revealing that bone angiogenesis and osteogenesis are coupled, we next investigated whether the changes in the Yap1/Taz iDEC femoral vasculature results in altered bone formation. P21 Yap1/Taz iDEC mutant mice display slightly lower body weights than control littermates, whereas the length and weight of their femurs and tibias are not changed significantly ( Figure Figure 7H,I). Trabecular spacing, a parameter reflecting the separation of Tb elements, is reduced in Yap1/Taz iDEC samples. The number of osteoclasts, identified by antibodies directed against a subunit of vacuolar ATPase (ATP6V1B1/B2), is increased in mutants (Figure 7-figure supplement 1B) arguing that it is unlikely that the increase in trabecular bone is caused by impaired bone degradation.
Consistent with the reduced vascular growth after EC-specific expression of stabilized Yap1 S112A , bone formation is also reduced in Yap1-KI iEC femurs. Body weight as well as femoral and tibial length and weight is reduced in P21 Yap1-KI iEC mutants (Figure 7-figure supplement 1C-E). Expression of Yap1 S112A also leads to the reduction of Osterix+ osteoprogenitor cells and of osteopontin immunostaining ( Figure 7F,G). The increase in Osterix+ cells seen after the loss of endothelial Yap1/Taz is restored to control level by the simultaneous and EC-specific inactivation of Hif1a (Figure 7-figure supplement 1F).    Source data 1. Source data for Figure 5A,C,E,F.

Discussion
Numerous studies have highlighted crucial roles of Hippo signaling in organ growth and size control, which also includes the functional characterization of this pathway in the growing vasculature. It was already shown that the loss of Yap1/Taz in ECs impairs endothelial proliferation and sprouting in the embryo and postnatal retina, whereas genetic gain-of-function experiments lead to endothelial hypersprouting and vascular hyperplasia Neto et al., 2018;Sakabe et al., 2017;Wang et al., 2017). In ECs of the lymphatic system, Taz has been linked to maladaptive effects such as the loss of quiescence, aberrant entry into the cell cycle and, ultimately, cell death in response to disturbed flow (Sabine et al., 2015). Likewise, while protective laminar flow suppresses Yap1/Taz activity in cultured ECs, exposure to disturbed flow led to the induction of Yap1/Taz-dependent proliferative and proinflammatory responses (Wang et al., 2016). In the zebrafish embryo, blood flow was also identified as one of the upstream factors controlling the nuclear import of Yap1 in a Hippoindependent fashion (Nakajima et al., 2017). Another study has proposed that endothelial Yap1/ Taz activity controls the expression of bone morphogenetic protein 4 (BMP4) and thereby transiently promotes intramembraneous ossification in the head of zebrafish embryos, while angiogenesis was unaffected (Uemura et al., 2016). These reports already suggest that the functional roles of Yap1/ Taz in the vascular system are diverse and context-dependent. It is, nevertheless, striking that the EC-specific inactivation of the two genes led to opposite outcomes in the retina and in the skeletal system. This is reminiscent of the role of endothelial Notch signaling, which is a potent suppressor of angiogenesis in retina, brain and in tumors (Benedito et al., 2012;Ridgway et al., 2006) but promotes type H formation, vessel growth and osteogenesis in long bone .
While the exact cause of these disparate functional roles of the two pathways requires further investigation, local differences in blood flow and tissue oxygenation are likely to be relevant. Artery caliber, flow rates and calculated endothelial shear stress are comparably small inside bones (Ramasamy et al., 2016b) relative to other tissues so that the impact of hemodynamic parameters might be more limited in bone ECs. Nevertheless, the observed nuclear localization of Yap1 in the sinusoidal endothelium, which has the lowest rates of flow Ramasamy et al., 2016b), is still consistent with the previously reported flow-controlled regulation of Yap1/Taz. The levels of tissue oxygenation also vary largely between different organs. While hypoxia is undetectable in the healthy retina and brain (Kisler et al., 2017;Mezu-Ndubuisi et al., 2013), the interior of bone is highly hypoxic. Live imaging in the adult murine calvarium showed the lowest local oxygen tension around the sinusoidal vasculature, whereas the endosteum is less hypoxic due to the presence of small arterioles (Spencer et al., 2014). In postnatal long bone, we observed that arterioles terminate in type H capillaries of the metaphysis and endosteum but not in sinusoidal (type L) vessels. Accordingly, multiple markers of hypoxia were absent from the metaphysis but highly abundant throughout the diaphysis . Despite of these differences, we also found that the HIF pathway plays crucial roles in bone angiogenesis and the specification of type H ECs Kusumbe et al., 2016). The current study offers important clues that might help to explain the previous findings. Type H ECs in the metaphysis show low steady state levels of Yap1/Taz consistent with the rapid, Hippo-dependent degradation of these proteins. This might allow the necessary level of endothelial HIF signaling required for local EC proliferation and vessel growth in the metaphysis. In contrast, higher levels of nuclear Yap1/Taz in the ECs of the diaphyseal vasculature might suppress excessive activation of the HIF pathway even though the local microenvironment is highly hypoxic. This model is consistent with the observed upregulation of HIF-controlled genes in Yap1/Taz iDEC bone ECs and can also explain why the same mechanism does not apply to highly oxygenated tissues, such as retina, where HIF1a is unstable and rapidly degraded. It was proposed that Yap1 physically interacts with HIF1a with in cultured cancer cell lines (Zhang et al., 2018), whereas HIF2a was shown was shown to enhance Yap1 activity without direct interaction of the two proteins (Jia et al., 2019). Furthermore, it was shown that hypoxia increases the stability of Yap1 by inducing the degradation of Lats2 through the E3 ubiquitin ligase SIAH2 (Ma et al., 2015). The same study also proposed that HIF1a and Yap1 can interact directly, leading to stabilization of   Nuclei, DAPI (blue). Arrowheads mark vessels in transition zone (D). Quantitative analysis of length and number of vessel column (E) (n = 6; data are presented as mean ±sem, P values, two-tailed unpaired t-test).
The online version of this article includes the following source data for figure 6: Source data 1. Source data for Figure 6B,E. HIF1a in cells cultured under hypoxic conditions. While these findings would not explain the observed repression of HIF1a activity in bone ECs, it was also reported that Yap1/Taz can act as transcriptional co-repressors and inhibit the expression of hypoxia-induced genes (Kim et al., 2015). While the exact mechanism requires further investigation, it is increasingly evident that the activity of the two pathways is linked und therefore influenced by differences in oxygenation in healthy organs but, potentially, also in response to pathological processes involving tissue ischemia.
Apart from showing that Yap1/Taz can act as negative regulators of growth processes in the bone, our study highlights the stringent coupling of angiogenesis and osteogenesis in the skeletal system. Our findings also raise the interesting question whether drugs acting on the Hippo pathway, such as Verteporfin, which is a small molecule inhibitor of Yap1-TEAD complex formation and of Yap1/Taz-mediated cell proliferation (Kimura et al., 2016;Liu-Chittenden et al., 2012), might be therapeutically useful to increase bone mineral density. This could be relevant in the context of aging-related loss of mineralized bone or in osteoporosis, which leads to bone weakness, increased risk of fracturing, loss of mobility and chronic pain. Future studies will have to explore this important topic, which is likely to require the development of targeted therapeutic strategies to avoid adverse effects in the many different cell types and organs utilizing the Hippo pathway.

Materials and methods
The Key Resources Table (Supplementary file 1) provides a list with the mouse strains, cell line, antibodies, reagents, kits and software used for this study.
To generate EC-specific Cdh5-mT/nG transgenic reporter mice expressing membrane-anchored tomato protein and nuclear green fluorescent protein, a cassette consisting of membrane-tagged tdTomato (Addgene plasmid #17787), 2A peptide, AU1 tag and H2B-EGFP (Addgene plasmid #11680) followed by a polyadenylation signal sequence and a FRT-flanked ampicillin resistance cassette were introduced by recombineering into the start codon of a large Cdh5 genomic fragment in PAC clone 353-G15. After Flp-mediated excision of the ampicillin resistance cassette in bacteria, positive clones were validated by PCR analysis and used in circular form for pronuclear injection into fertilized mouse oocytes. Founders were screened by PCR and immunostaining with Endomucin Yap1/Taz iDEC and control metaphysis. Graph on the right shows quantitative analysis of Osterix+ (Osx+) cells (control n = 6 and Yap1/Taz iDEC n = 6; data are presented as mean ±sem, P values, two-tailed unpaired t-test). (E) Confocal images showing bone vessels (Emcn), and the bone matrix protein Osteopontin in Yap1/Taz iDEC vs control femur. (F) Representative confocal images of Osterix+ cells (green) in relation to Emcn+ ECs (red) in the P21 control and Yap1-KI iEC femoral metaphysis. Graph on the right shows significant reduction of Osterix+ cells in Yap1-KI iEC mutants (n = 6; data are presented as mean ±sem, P values, two-tailed unpaired t-test). (G) Decreased bone matrix protein Osteopontin (Opn, green) deposition in P21 Yap1-KI iEC femur relative to control. ECs, Emcn (red). (H, I) Representative mCT images of trabecular bone in P21 control and Yap1/Taz iDEC femur (H). Quantitative analysis of trabecular volume (BV/TV, bone volume/total volume) trabecular (Tb.) number, Tb. thickness, and Tb. separation (I). (n = 5; data are presented as mean ±sem, P values, two-tailed unpaired t-test). The online version of this article includes the following source data and figure supplement(s) for figure 7: Source data 1. Source data for Figure 7A,C,D,F,I. (Emcn) in bone. Genotypes of mice were determined by PCR. Cdh5-mT/nG transgenic reporter mice were introduced into the Yap1/Taz iDEC double and Lats2 iDEC mutant background and the corresponding controls.
To generate constitutive-active Yap knock-in (Yap-KI) mice, we mutated Yap1 amino acid 112 serine to alanine (Yap1 S112A ), as this phosphorylation site is responsible for the translocation of Yap1 from nucleus to cytosol. The Yap1 S112A cDNA was inserted into a CAG-STOP-eGFP-ROSA26TV (Addgene plasmid #15912) vector and recombined into bacterial artificial chromosome (BAC) clone containing the murine Rosa26 locus. Linear recombined clones were injected into fertilized mouse oocytes. Founders were screened by PCR and EC-specific Yap1 nuclear localization was conformed in retina.
All animal experiments were performed according to the institutional guidelines and laws, approved by local animal ethical committee and were conducted at the University of Mü nster and the Max Planck Institute for Molecular Biomedicine with permissions (84-02.04.2015.A185, 84-02.04.2016.A160, 81-02.04.2017.A238) granted by the Landesamt fü r Natur, Umwelt und Verbraucherschutz (LANUV) of North Rhine-Westphalia. Animals were combined in groups for experiments irrespective of their sex.

Tamoxifen-inducible Cre-mediated recombination
Pups received daily intraperitoneal injections (IP) of 50 mg or 100 mg of tamoxifen (Pitulescu et al., 2010) from postnatal day 1 (P1) to P3 or from P6 to P8. Tamoxifen (Sigma, Cat#T5648) stocks were prepared by dissolving 20 mg in 500 ml of ethanol and vortexing for 10mins before an equal volume of Kolliphor EL (Sigma, Cat#C5135) was added. 1 mg aliquots were stored at À20˚C and dissolved in the required volume of PBS prior to injection.

Bone sample preparation
Mice were sacrificed and long bones (femur and tibia) were harvested and fixed immediately in icecold 2% paraformaldehyde (PFA) for 6 to 8 hr under gentle agitation. Bones were decalcified in 0.5M EDTA for 16 to 24 hr at 4˚C under agitation, which was followed by overnight incubation in sucrose solution (20% sucrose, 2%PVP) and mounted in bone mounting medium (8% galatine, 2% PVP). Samples were stored overnight at À80˚C. 60-100 mm-thick cryosections were prepared for immunofluorescence staining (Kusumbe et al., 2015).

Pharmacological treatments
Three-week-old wild type mice were injected intraperitoneally either with 50 mg/g MG132 (Millipore, Cat# 474790) or DMSO only (vehicle control). Three hours after injection, mice were sacrificed and femurs were dissected and processed for immunostaining as described below. For labeling of hypoxic cells, mice were intraperitoneally injected with 60 mg/kg Pimonidazole (Hypoxyprobe Inc) 2 hr before analysis.

Proliferation assay in vivo
For the analysis of proliferating cells in bone in vivo, mice received an intraperitoneal injection of 300 mg of EdU for 3 hr before analysis. For retina, P6 pups received 100 mg of EdU for 2 hr. Detection of proliferating cells in fixed bone sections and whole mount retina was achieved by staining with Click-iT-EdU Alexa-647 imaging kit (Invitrogen, Cat# C10340) according to the manufacturer's instructions.
For whole-mount retina staining, P6 eyes were removed and fixed in ice cold 4% PFA for 2 hr. After fixation, retinas were dissected and processed as described previously (Pitulescu et al., 2010). After two washes with ice-cold PBS, samples were incubated in blocking buffer (1%BSA, 1% Triton X-100, 3% heat inactivated donkey serum in PBS) for 1 hr on rotating shaker. Next, blocking buffer was replaced by Pblec buffer (1 mM CaCl 2 , 1 mM MgCl 2 , 0.1 mM MnCl 2 , 0.1% Triton X-100 in PBS). Isolectin-B4 (IB4; Vector, Cat#Ab200839) or rabbit monoclonal anti-ERG (Abcam, Cat# ab22552) were diluted in Pblec buffer and each retina was incubated in 100 ml of solution overnight at 4˚C. Next, samples were washed five times in incubation buffer (diluted blocking buffer 1:1 in PBS) and incubated with the appropriate Alexa Fluor488, 546 and 594-conjugated secondary antibodies for 2 hr at RT. Later, retinas were washed five times with ice-cold PBS and mounted under a stereomicroscope.

Fluorescence-activated cell sorting (FACS)
Single cell suspensions were prepared from femur and tibia as described (Langen et al., 2017). CD31-APC (Goat polyclonal anti-CD31 (APC-conjugated; R and D, Cat# AFB3628A) coupled and Emcn primary antibodies were added to the single cell suspension and incubated for 45mins on ice. Samples were washed two to three times with blocking solution. Secondary anti rat-PE antibody and DAPI (Sigma, Cat#D9542) were added and the incubation continued for 45mins on ice. Next, samples were washed 2-3 times with blocking solution and resuspended using ice-cold PBS. Cell sorting was performed on a FACS Aria II cell sorter (BD Bioscience). Dead cells and debris were excluded by FSC, SSC and DAPI positive signal. Sorted bone ECs were collected in RTL buffer for RNA isolation. FACS data were analysed with FlowJo Software (FLOWJO, LLC).
Primer sequences for qPCR analysis of Yap1 and Wwtr1 expression in freshly isolated bone ECs (Figure 2-figure supplement 1B) are provided in the Key Resources Table (Supplementary file 1).

Immunoprecipitations and immunoblotting
For immunoprecipitations, HUVEC cells were lysed in radioimmunoprecipitation (RIPA) lysis buffer (150 mM NaCl, 25 mM Tris-HCl, 1 ml of 0.5 M EDTA pH 8.0, sodium dodecyl sulfate (SDS), 1% sodium deoxycholate and 0.25% Triton X-100) containing protease inhibitor (cOmplete, Roche) and phosphatase inhibitor tablets (PhosSTOP, Roche). Lysates were centrifuged at 4˚C for 30 min at 20,000 g, and aliquots were set aside for direct input blot analysis. For the immunoprecipitation, the remaining lysates were pre-cleared for 1 hr at 4˚C with Dynabeads protein-G. Pre-cleared lysates were incubated with primary antibody for 2 hr at 4˚C, after which the beads were added and incubated for an additional 2 hr at 4˚C. Immunocomplexes were washed five times with lysis buffer (without SDS and deoxycholate) and analysed by SDS-PAGE.
For immunoblotting, cell lysates were boiled in Laemmli buffer for 5mins and then used for immunoblotting. Proteins were separated in 6 or 8% SDS -PAGE, then transferred onto nitrocellulose membranes, followed by blocking for 1 hr and overnight incubation with primary antibodies. After washing, membranes were incubated with horseradish peroxidase-conjugated secondary antibodies for 1 hr. Antibody binding was visualized by enhanced chemiluminescence reagent (Millipore) using Fuji medical X-ray films.

Chromatin immunoprecipitation (ChiP)
Chromatin immunoprecipitation was performed as described previously (Sivaraj et al., 2013). In brief, siRNA siControl and siYAP1/WWTR1 transfected HUVEC were grown in 1% O 2 for 24 hr, then cross-linked with 1% formaldehyde for 10 min at room temperature. Samples were then sonicated into 200-700 bp fragments using a Branson digital sonifier and chromatin was immunoprecipitated with 5 mg of anti-rabbit IgG or anti-HIF1a, followed by reverse cross-linking. The recovered DNA was purified using the Qiagen DNA isolation kit and DNA was analyzed using qPCR. The VEGFA promoter HRE sequence was amplified using 5'-GCCAGACTCCACAGTGCATA-3'and 5'-CTGA-GAACGGGAAGCTGTGT-3' primer pair. The DNA recovered from chromatin that was not immunoprecipitated was used as input.

RNA sequencing and data analysis
Three-week-old bone ECs were sorted from Cdh5-mTnG metaphysis (mpECs) and diaphysis (bmECs) or Yap1/Taz iDEC and Lats2 iDEC mutant bone with their respective littermate controls. RNA was isolated using the RNeasy Plus Micro Kit (QIAGEN, Cat# 74134) according to the manufacturer's instructions. RNA quality was checked using a 2100 BioAnalyzer (Agilent). 100 ng of RNA were used for preparation of sequencing libraries with the TruSeq Stranded Total RNA Library Prep Kit (Illumina) according to the manufacturer's instructions. Libraries were validated using a BioAnalyzer, quantified by qPCR and Qubit Fluorometric Quantitation (Thermo Fisher Scientific, Cat#Q32851). Libraries were diluted to a final concentration of 15pM for sequencing. The MiSeq Reagent Kit v3 (Illumina, Cat#MS-102-3001) was used for sequencing with a MiSeq (Illumina). Biological triplicates were used.
RNA-seq data analysis was performed as described previously (Langen et al., 2017) with some modifications. The quality assessment of raw sequence data was performed using FastQC (Version: FastQC 0.11.3). Paired-end sequence reads were mapped to the mm10 mouse genome assembly (GRCm38) using TopHat-2 (Version: tophat-2.0.13). The mouse genome was downloaded from the iGenome portal. HTSeq was used to count the aligned reads on a per gene basis (Version: HTSeq-0.6.1).
The count data were normalized using the Variance Stabilizing Transformation (VST) function from the DESeq2 package. Principal Component Analysis (PCA) was performed on transformed read counts using the variable genes to assess the overall similarity between the samples. Differential gene expression analysis between control and mutants were performed using DESeq2. Differentially expressed genes were selected using a FRD-adjusted p-value cut-off <0.01 and an absolute log 2 fold change >0.5. Gene symbols were annotated using biomart (BioConductor version 3.1). Gene ontology analysis and cellular signaling pathways were performed with the Enrichr online tool (http://amp.pharm.mssm.edu/Enrichr/). Heat maps were generated with http://heatmapper.ca/.
RNA-sequencing data of control and mutant bone ECs were uploaded to the Gene Expression Omnibus (GEO) under the accession number GSE102181.

Quantification and statistical analysis
Immunostained bone sections were imaged with a Leica SP8 confocal microscope and the following settings: 1024 Â 1024; 200 speed, low magnification image z stack-4 mm; high magnification stack: 2 mm. Images were analysed, quantified and processed using Volocity (Perkin Elmer), Adobe CS6 Photoshop and Illustrator software.
For quantification of vascular alterations in bone, the number of vessel buds or columns and EC proliferation were quantified manually using Volocity software (Perkin Elmer). Quantitation is based on 3-5 images of bone sections per animal and average values per animals were combined in the graphs.
Statistical analysis was performed using GraphPad Prism software or the R statistical environment (http://r-project.org). All data are presented as mean ± s.e.m. unless indicated otherwise. Unpaired two tailed student t-tests were used to determine statistical significance. p<0.05 was considered significant unless stated otherwise. Sample number was chosen based on experience from previous experiments. Reproducibility was ensured by several independent experiments. No animals were excluded from analysis.