YAP and TAZ maintain PROX1 expression in the developing lymphatic and lymphovenous valves in response to VEGF-C signaling

Lymphatic vasculature is an integral part of digestive, immune and circulatory systems of vertebrates. Lymphatic vessels absorb interstitial fluid and return it to blood circulation (Adams and Alitalo, 2007). In addition, specialized lymphatic vessels of the gut known as lacteals absorb digested lipids. Lymphatic vessels regulate immune response by transporting immune cells from tissues to lymph nodes. Lymphatic vessels also resolve inflammation by clearing extravasated fluids and immune cells at the site of inflammation. Defects in lymphatic vasculature can cause lymphedema, a disease in which tissues swell due to excessive fluid accumulation. Currently, we lack approaches to treat lymphedema, which has common comorbidities such as infections, inflammation, obesity and fibrosis (Tammela and Alitalo, 2010). Improved understanding of the mechanisms that regulate lymphatic vascular development could provide innovative opportunities for treating lymphedema and other lymphatic vascular disorders.

Lymphatic vasculature is arranged in a hierarchical manner. Blind-ended lymphatic capillaries collect interstitial fluid, immune cells or digested lipids (simply called as lymph), and transport them to collecting lymphatic vessels. Lymphatic valves (LVs) within the collecting lymphatic vessels regulate the unidirectional flow of lymph and prevent backflow. Finally, lymph travels through the thoracic duct or the right lymphatic duct and returns to blood circulation via two pairs of lymphovenous valves (LVVs) located bilaterally at the intersection of jugular and subclavian veins (Geng et al., 2016; Srinivasan and Oliver, 2011).

Lymphatic vessels, LVs and LVVs are established from lymphatic endothelial cells (LECs) that originate predominantly from the embryonic veins (Pichol-Thievend et al., 2018; Srinivasan et al., 2007). Other sources could make minor contribution to the lymphatic vasculature in a tissue-specific manner (Gancz et al., 2019; Lioux et al., 2020; Martinez-Corral et al., 2015; Maruyama et al., 2019; Stanczuk et al., 2015). LEC progenitors are specified in the embryonic veins by the transcription factor PROX1, which activates the expression of molecules such as the receptor tyrosine kinase VEGFR3 and the glycoprotein podoplanin (Hong et al., 2002; Petrova et al., 2002; Srinivasan et al., 2014; Wigle et al., 2002; Wigle and Oliver, 1999). The potent lymphangiogenic molecule VEGFC associates with VEGFR3 to promote LEC migration from the veins to form the lymph sacs (Karkkainen et al., 2004). A subset of LEC progenitors do not upregulate VEGFR3 expression and remain in the veins to form two pairs of LVVs through which lymph sacs interact with the veins (Geng et al., 2016; Srinivasan and Oliver, 2011).

Lymphatic vessels sprout from the lymph sacs to form the primitive lymphatic plexus in various tissues. Subsequent maturation of the lymphatic plexus results in the formation of lymphatic capillaries and collecting lymphatic vessels (Norrmén et al., 2009). LVs develop within the collecting lymphatic vessels (Bazigou et al., 2009; Norrmén et al., 2009; Petrova et al., 2004). Maturation of lymphatic plexus and the formation of valves are regulated by various signaling molecules such as VEGF-C, neuropilin 2, Wnt/β-catenin, S1PR1, BMP9, ephrin B2, EPH-B4, plexins and semaphorins (Bouvrée et al., 2012; Cha et al., 2016, 2018; Geng et al., 2020; Jurisic et al., 2012; Levet et al., 2013; Liu et al., 2014, 2016; Makinen et al., 2005; Martin-Almedina et al., 2016; Nurmi et al., 2015; Yuan et al., 2002). Shear stress generated by lymph flow also plays a deterministic role in lymphatic vascular morphogenesis (Cha et al., 2016, 2018; Choi et al., 2019, 2017a,b; Geng et al., 2020; Liu et al., 2014; Sabine et al., 2012; Sweet et al., 2015; Wang et al., 2016). Despite these advances, our basic understanding of the mechanisms of lymphatic vascular development remain incomplete. How do the various signaling pathways coordinate with each other to regulate lymphatic vascular development in a precise spatiotemporal manner is not clear.

Hippo signaling controls cell proliferation, survival and differentiation during tissue development, regeneration and homeostasis (Zheng and Pan, 2019). The transcription co-factors YAP (Yes-associated protein) and TAZ (transcriptional co-activator with PDZ-binding motif) are key downstream effectors of the Hippo pathway. When Hippo signaling is ‘off’, non-phosphorylated YAP and TAZ accumulate in the nucleus and interact with transcription factors such as TEAD4, TBX5 and SMAD to induce target genes. Importantly, YAP and TAZ function as gatekeepers of biochemical and mechanical signaling pathways such as Wnt/β-catenin, TGFβ, GPCR, ECM stiffness and shear stress (Piccolo et al., 2014). In the blood vasculature, YAP and TAZ promote retinal angiogenesis and maturation of endothelial cell barrier by regulating VEGF, BMP, Rho GTPase and biomechanical signals (Kim et al., 2017; Neto et al., 2018; Sakabe et al., 2017; Wang et al., 2017). Intriguingly, YAP and TAZ inhibit angiogenesis within bones by negatively regulating HIF pathway (Sivaraj et al., 2020). The co-repressor activity of the YAP/TAZ/TEAD4/NuRD complex is likely responsible for the transcriptional inhibitory function of YAP and TAZ (Kim et al., 2015).

We are starting to understand the roles of YAP and TAZ in the mammalian lymphatic vasculature. TAZ and the ECM protein CTGF, which is a canonical target of YAP and TAZ, are expressed in LVs, but not in non-valvular LECs (Sabine et al., 2015). Conditional deletion of Yap and Taz from mouse LECs disrupts the maturation of the lymphatic vessels and results in the absence of LVs (Cho et al., 2019). Maturation of collecting lymphatic vessels involves the pruning of excessive branches and downregulation of PROX1 and VEGFR3 expression (Norrmén et al., 2009; Petrova et al., 2004). PROX1 fails to undergo downregulation in the absence of YAP and TAZ. Furthermore, VEGF-C inhibits YAP and TAZ activity to promote PROX1 expression in primary human LECs (Cho et al., 2019). The downstream side of LVs experience oscillatory shear stress (OSS), and OSS enhances the expression of the lymphedema-associated transcription factor FOXC2 that is necessary for LV development (Petrova et al., 2004; Sabine et al., 2012). Importantly, FOXC2 inhibits OSS-induced YAP and TAZ activity in primary human LECs (HLECs) (Sabine et al., 2015). Furthermore, stiff ECM could also enhance YAP and TAZ activity in HLECs (Frye et al., 2018). Despite these findings, whether YAP and TAZ are required for the formation of LVs and LVVs or for their maintenance is not known. To address this important issue, we investigated the mechanisms of YAP and TAZ activity during LV and LVV development in mice.

LVV development starts at embryonic day (E) 12.0 in mice (Geng et al., 2016; Srinivasan and Oliver, 2011). PROX1, FOXC2 and GATA2 are upregulated in LVV-ECs, which delaminate from the luminal side of the jugular vein. LVV-ECs rapidly reaggregate and elongate to form mature LVVs at E12.5; after this stage, the changes within the valves are subtle.

We performed immunohistochemistry to investigate the nuclear localization of YAP and TAZ, and the expression of their target molecule CTGF in LVVs starting from E12.0. YAP, TAZ and CTGF displayed low expression in LVV-ECs until E15.5 (Fig. S1A,B). At E15.5, YAP and TAZ were detected in both the cytoplasm and nucleus of LVV-ECs (Fig. S1C). CTGF was observed in the ECM of LVV-ECs (Fig. S1D). By E16.5, YAP and TAZ were predominantly nuclear in LVV-ECs (Fig. S1E), and CTGF remains enriched in LVVs (Fig. S1F). Semi-quantitative measurement of YAP, TAZ and CTGF expression indicates that their expression in LVVs increases significantly between E13.5 and E16.5 (Fig. S1G,H).

YAP and TAZ were similarly expressed in mature mesenteric LVs (Fig. 1). Although YAP and TAZ were not expressed in newly formed LV-ECs at E16.5 (Fig. 1A), they were detectable in LV-ECs at E17.5 (Fig. 1B) and were strongly expressed by E18.5 (Fig. 1C,D). Angiopoietin 2 (ANGPT2), another target of YAP and TAZ (Choi et al., 2015; He et al., 2018; Kim et al., 2017), was also observed in LV-ECs at E18.5 (Fig. 1E). In addition, CTGF was expressed in mature LVs as reported previously (Fig. 1F) (Sabine et al., 2015). Together, these results suggest that the activity of YAP and TAZ gradually increases in developing LVs and LVVs.

To determine the function of YAP and TAZ during lymphangiogenesis, we used Lyve1-Cre to conditionally and constitutively delete YAP/TAZ in the LEC progenitors from E10.5 in mouse embryos (Pham et al., 2010; Xin et al., 2013, 2011). Lyve1-Cre;Yap^flox/flox and Lyve1-Cre;Taz^flox/flox mice were born alive and were phenotypically normal. However, we did not obtain Lyve1-Cre;Yap^flox/flox;Taz^flox/flox (referred to as Yap/Taz^LECKO) pups. We analyzed Yap/Taz^LECKO mouse embryos at various stages. Although E14.5 Yap/Taz^LECKO embryos did not show any gross phenotypes, YAP and TAZ expression was strikingly downregulated in their LECs and LVV-ECs (Fig. S2). E15.5 Yap/Taz^LECKO embryos showed edema with variable penetrance, with most embryos showing no obvious phenotype (Fig. 2A-D).

There are two pairs of bilaterally located LVVs in mammalian embryos (Geng et al., 2016; Srinivasan and Oliver, 2011). We detected four LVVs in all E13.5 and most E15.5 Yap/Taz^LECKO embryos (Fig. 2E-H). The number of LVVs was slightly but significantly reduced in E15.5 Yap/Taz^LECKO embryos compared with littermate controls (Fig. 2I). Strikingly, by E16.5 and E17.5, Yap/Taz^LECKO embryos almost completely lacked LVVs (Fig. 2I-M). Unlike control E17.5 embryos, E17.5 Yap/Taz^LECKO embryos displayed only a few LVV-ECs or small holes where LVVs normally form (Fig. 2M, white arrowhead and higher magnification inset). Venous valves that develop close to LVVs are a part of the blood vasculature, although they share the same molecular profile as LVVs (Bazigou et al., 2011; Geng et al., 2016; Munger et al., 2016, 2013; Srinivasan and Oliver, 2011). Consistent with Lyve1-Cre activity in embryonic veins, E17.5 Yap/Taz^LECKO embryos also lacked venous valves (Fig. 2K,M, yellow arrowheads).

E16.5 Yap/Taz^LECKO embryos lacked observable defects in the lymphatic vessels of the dorsal skin, and LV-forming endothelial cells were observed within their lymphatic vessels of the mutants (Fig. 3A,B, arrows). However, by E18.5, Yap/Taz^LECKO embryos had defective lymphatic vessels that were more dilated, had fewer branch points and did not migrate sufficiently from the lateral edges (Fig. 3E-G). In addition, unlike control embryos (Fig. 3H, arrows), E18.5 Yap/Taz^LECKO embryos no longer had any LVs (Fig. 3G,I).

The mesenteric lymphatic vessels of E18.5 Yap/Taz^LECKO embryos were immature, as suggested by strong expression of LYVE1, VEGFR3 and PROX1 (Fig. S3). LVs were missing from these immature collecting vessels (Fig. S3C-G). However, the guts of Yap/Taz^LECKO embryos were much smaller in size when compared with controls (Fig. S3E,F). This is likely due to the blood vascular defects caused by Lyve1-Cre expression in the blood vessels of the gut, as reported previously (Dellinger et al., 2013; Geng et al., 2020). Consequently, we are unable to conclude whether the defects in the mesenteric lymphatic vessels of Yap/Taz^LECKO embryos are due to YAP and TAZ activity in LECs or to blood vascular endothelial cells. Nevertheless, Lyve1-Cre is specific to dermal lymphatic vasculature (Dellinger et al., 2013; Geng et al., 2020). Hence, based on our findings we conclude that YAP and TAZ are not required for the differentiation of valvular endothelial cells (LVVs and LVs). This conclusion coincides with the lack of YAP and TAZ expression and activity in the newly differentiated valvular endothelial cells. However, LVVs and LVs degenerate in the absence of YAP and TAZ.

The small molecule verteporfin (VP) disrupts the interaction between YAP, TAZ and TEAD, thereby inhibiting the transcriptional activity of YAP and TAZ. Primary human LECs (HLECs) treated with VP for 2 h showed reduced expression of the canonical YAP and TAZ target genes CTGF (CCN2), CYR61 (CCN1) and ANKRD1, as expected (Fig. 4A). To comprehensively identify the transcriptional targets of the TEAD/YAP/TAZ complex, we treated HLECs with 20 µM VP for 2 h, extracted RNA and performed RNA-seq analysis. Based on the Log₂(fold change) >1 that we set for differentially expressed genes, we determined that 794 genes were upregulated and 2161 genes were downregulated by VP [Fig. 4B and see data in the Dryad Digital Repository (Cha et al., 2020) and Table S1]. Gene ontology (GO) analysis showed that VP altered the expression of genes that regulate vascular development (Fig. 4C). Genes that regulate lymphangiogenesis, such as PROX1, FLT4, ANGPT2 and DLL4 were also dramatically downregulated in VP-treated HLECs (Fig. 4D).

PROX1 is the master regulator of lymphatic vascular development and is required for the formation and maintenance of LVs and LVVs (Geng et al., 2016; Johnson et al., 2008; Srinivasan and Oliver, 2011). We verified that VP treatment of HLECs reduces PROX1 expression at the RNA and protein levels (Fig. 5A). Importantly, siRNA-mediated knockdown of YAP and TAZ also reduced PROX1 levels in HLECs (Fig. 5A,B).

The ENCODE database predicted a TEAD4-binding site in the regulatory elements of PROX1, close to the GATA2-binding site that was reported previously (Fig. 5C) (ENCODE Project Consortium, 2012; Kazenwadel et al., 2015). Using a targeted approach, we verified that a TEAD4-binding site (Fig. 5D) is indeed present in the upstream regulatory elements of PROX1, and is conserved among several mammals (Fig. 5E). Furthermore, ChIP-PCR revealed that YAP binds to this site in HLECs (Fig. 5F). Together, these data suggest that YAP/TAZ cooperates with TEAD4 to directly activate PROX1 expression in HLECs. However, the functional significance of this binding site is currently unknown, and it is likely that YAP and TAZ associate with multiple sites in the distal regulatory elements of genes, as described previously (Galli et al., 2015; Stein et al., 2015).

Activation of the Hippo signaling pathway leads to MST1/2-mediated phosphorylation and activation of LATS1/2, which in turn phosphorylate and inactivate YAP/TAZ. Therefore, to enhance YAP/TAZ activity in HLECs, we treated cells with XMU-MP-1, a chemical inhibitor of MST1 and MST2, or knocked down LATS1 and LATS2 (siLATS1/2). Both treatments resulted in the upregulation of the YAP and TAZ target genes CTGF and CYR61, as anticipated (Fig. S4A,B). In contrast, PROX1 expression was downregulated by siLATS1/2 and XMU-MP-1 treatments (Fig. S4A,B), consistent with recent findings of Cho et al. (Cho et al., 2019). Thus, PROX1 expression in HLECs is delicately dependent on the activity of YAP and TAZ.

To test whether YAP and TAZ regulate Prox1 expression in vivo, we bred Prox1-tdTomato transgenic reporter mice into the Yap/Taz^LECKO background. In the Prox1-tdTomato transgenic mice the expression of tdTomato is driven by ∼100 kb regulatory elements of Prox1 (Gong et al., 2003). Thus, the transcriptional regulation of Prox1 could be visualized using Prox1-tdTomato mice. Compared with Prox1-tdTomato littermate controls, Prox1-tdTomato; Yap/Taz^LECKO embryos showed dramatically reduced expression of PROX1 and tdTomato in LVV-ECs (Fig. 6A,B, arrows; Fig. 6C) and in LECs of lymph sacs (Fig. 6A,B, arrowheads; Fig. 6C) at E15.5.

As mentioned previously, PROX1^high; tdTomato^high LVs that were observed in the dermal lymphatic vessels of E18.5 Prox1-td-Tomato embryos (Fig. 6D, arrows) were absent in Prox1-td-Tomato; Yap/Taz^LECKO littermates (Fig. 6E). Moreover, Prox1 and other YAP/TAZ target genes showed reduced expression in LECs isolated from Yap/Taz^LECKO embryos compared with controls (Fig. 6F). We wanted to determine whether the observed downregulation of Prox1 in Yap/Taz^LECKO embryos is due to the absence of LVs or also due to reduced Prox1 expression in LECs. Therefore, we measured the intensities of PROX1 and tdTomato signals in a semi-quantitative manner from E18.5 Prox1-td-Tomato and Prox1-td-Tomato; Yap/Taz^LECKO embryos (Fig. 6G). Expression of tdTomato, but not PROX1, was modestly downregulated in the LECs of Prox1-td-Tomato; Yap/Taz^LECKO embryos. These results suggest that YAP and TAZ are necessary to maintain PROX1 expression primarily in valvular endothelial cells. This conclusion is supported by the fact that CTGF is expressed exclusively in the valves (Fig. 1, Fig. S1 and Sabine et al., 2015). In summary, we infer that Yap and Taz are thus required to maintain valvular endothelial cell identity at least in part by promoting Prox1 expression.

To further investigate the relationship between PROX1, YAP and TAZ, we deleted Yap and Taz in PROX1-expressing cells, including LECs, using Prox1^+/Cre. In these mice, one functional allele of Prox1 is replaced by Cre recombinase, resulting in impaired LVV development (Srinivasan et al., 2010; Srinivasan and Oliver, 2011). E14.5 Prox1^+/Cre;Yap^+/f;Taz^+/f embryos lacked LVVs, as expected, as did E14.5 Prox1^+/Cre;Yap/Taz^LECKO embryos (Fig. 7A-C). The glycoprotein endomucin is expressed in venous endothelial cells, but not in LECs (D'Amico et al., 2009). Both E14.5 wild-type and Prox1^+/Cre;Yap^+/f;Taz^+/f embryos displayed endomucin expression in venous endothelial cells but not in LECs lining the lymph sacs (Fig. 7A,B). In contrast, E14.5 Prox1^+/Cre;Yap/Taz^LECKO embryos expressed endomucin in the lymph sacs (Fig. 7C,C″,D). This observation suggests that a subset of LECs in Prox1^+/Cre;Yap/Taz^LECKO embryos have abnormally acquired a partial blood vascular endothelial cell identity. This phenotype is also observed in embryos that completely lose Prox1 after the specification of LECs (Johnson et al., 2008). Hence, we studied the expression of PROX1 by immunohistochemistry and found that it was indeed downregulated in the LECs of Prox1^+/Cre;Yap/Taz^LECKO embryos compared with their littermates (Fig. 7E-H).

We were unable to obtain Prox1^+/Cre;Yap/Taz^LECKO embryos beyond E14.5, likely due to the expression of Cre in hepatocytes, which rely on YAP and TAZ for their growth (Camargo et al., 2007; Dong et al., 2007). Therefore, we analyzed E17.5 Prox1^+/Cre;Taz^LECHet;Yap^LECKO and Prox1^+/Cre;Yap^LECHet;Taz^LECHet embryos. At E17.5, the lymphatic vessels of wild-type embryos had crossed the dorsal midline and LVs were observed at the lateral edges of the skin (Fig. S5A and data not shown). In contrast, the lymphatic vessels were dilated and did not reach the midline in Prox1^+/Cre;Taz^LECHet;Yap^LECKO and Prox1^+/Cre;Yap^LECHet;Taz^LECHet embryos (Fig. S5B,C). Intriguingly, Prox1^+/Cre;Taz^LECHet;Yap^LECKO embryos developed cystic structures in their lymphatic vessels (Fig. S5C), whereas wild-type and Prox1^+/Cre;Yap^LECHet;Taz^LECHet embryos did not. We propose that Prox1^+/Cre;Taz^LECHet;Yap^LECKO embryos express relatively low levels of PROX1 compared with Prox1^+/Cre;Taz^LECHet;Yap^LECHet embryos, which exacerbates defects in lymphatic vessel morphogenesis. Overall, our data reveal that genetic reduction of Prox1 together with Yap and Taz results in the partial loss of LEC identity and improper morphogenesis of lymphatic vessels.

PROX1 expression is positively regulated by VEGF-C signaling in HLECs (Koltowska et al., 2015; Srinivasan et al., 2014). We hypothesized that VEGF-C/VEGFR3 signaling activates PROX1 expression in LECs through YAP/TAZ. We treated high confluent (∼100%) or low confluent (∼50%) HLECs with 100 ng/ml VEGF-C and investigated the phosphorylation of YAP. We determined that pYAP is strikingly reduced by VEGF-C specifically in low confluent cells (Fig. S6). Thus VEGF-C regulates the phosphorylation of YAP in a cell density-dependent manner.

To test whether VEGF-C regulates Hippo signaling, we treated 50% confluent HLECs with VEGF-C and isolated RNA and protein. We found that the expression of YAP and TAZ target genes was significantly upregulated in VEGF-C-treated HLECs relative to controls (Fig. 8A). Additionally, VEGF-C treatment inhibited the phosphorylation of YAP, as mentioned above (Fig. 8A,B). In contrast, VEGF-A did not increase the expression of YAP and TAZ target genes or reduce the level of pYAP (Fig. 8A,B). Thus, YAP and TAZ activity is enhanced by VEGF-C, but not by VEGF-A in HLECs.

VEGF-C increased the levels of YAP in the nucleus of HLECs (Fig. 8C). Moreover, VP treatment and siRNA-mediated knockdown of YAP and TAZ inhibited the VEGF-C dependent upregulation of PROX1 in HLECs (Fig. 8D). We used 293T cells as a heterologous system to further verify whether VEGF-C and VEGFR3 signaling could enhance YAP and TAZ activity. We infected 293T cells with lentiviruses expressing VEGFR3. Following antibiotic selection, we treated stably infected cells with VEGF-C for 0.5 or 2 h. Western blotting and qRT-PCR confirmed that VEGF-C could indeed downregulate pYAP and upregulate the expression of YAP/TAZ target genes (Fig. 8E). Together, these data suggest that VEGF-C activates YAP and TAZ activity in HLECs and VEGFR3-expressing 293T cells. Furthermore, VEGF-C promotes PROX1 expression in HLECs through YAP and TAZ.

Next, we investigated whether the Hippo pathway regulates VEGF-C induced behaviors in HLECs. To examine HLEC migration, we performed an in vitro wound-healing assay. VEGF-C treatment promoted the migration of HLECs, as anticipated (Fig. S7A,B,E). Importantly, VP treatment significantly inhibited VEGF-C-induced HLEC migration (Fig. S7C-E). Additionally, siRNA-mediated depletion of YAP and TAZ significantly reduced VEGF-C-induced HLEC proliferation (Fig. S7F-J). Thus, YAP and TAZ are necessary for the proper response of HLECs to VEGF-C.

Our in vitro studies indicate that VEGF-C signaling enhances YAP and TAZ activity. Our in vivo results have revealed that YAP and TAZ are required for lymphatic vessel, LVV, LV and venous valve morphogenesis. VEGF-C signaling is necessary for the development and maintenance of lymphatic vessels (Karkkainen et al., 2004, 2001; Nurmi et al., 2015). In addition, VEGFR3 is strongly expressed in LVs (Bazigou et al., 2009; Norrmén et al., 2009). Time-specific global deletion of Vegfc at E14.5 prevented the maturation of collecting lymphatic vessels and the development of LVs within the collecting lymphatic vessels (Nurmi et al., 2015). However, as the lymphatic vessels of these mice were thinner and immature, whether VEGF-C directly or indirectly regulates LV development remains unknown. VEGFR3 is also modestly expressed in LVVs and venous valves (Bazigou et al., 2011; Geng et al., 2016). Saphenous venous reflux is observed in individuals with Milroy's disease carrying mutations in VEGFR3 (Mellor et al., 2010). Yet, it is unclear whether LVV and venous valve development requires VEGF-C and VEGFR3 signaling.

To investigate the relationship between VEGF-C and VEGF3 signaling and Hippo signaling in vivo, we analyzed three mouse models in which VEGF-C signaling was inhibited. We generated Vegfc^+/CreERT2 mice in which we replaced the open reading frame of Vegfc with cDNA coding for CreERT2. Vegfc^+/CreERT2 mice recapitulated the phenotypes of Vegfc^+/− mice, including severe lymphatic vascular hypoplasia in the skin (Fig. S8). In addition, Vegfc^+/CreERT2 mice developed chylous ascites at birth, swollen paws and tail and lymphatic vascular hypoplasia of the heart and diaphragm as reported previously in Vegfc^+/− mice (Karkkainen et al., 2004). Vegfc^+/CreERT2 mice will henceforth be referred to as Vegfc^+/− mice in this study.

LVVs appeared normal in E15.5 Vegfc^+/− embryos (Fig. S9A,B, arrows). Venous valves were also present in E17.5 Vegfc^+/− embryos (Fig. S10A,B, red arrowheads). Similarly, LVVs and venous valves were observed in Vegfr3^+/chy embryos, which express a dominant-negative allele of VEGFR3 and phenocopy Vegfc^+/− mice (Fig. S9C,D, pseudo-colored in magenta and yellow, respectively) (Karkkainen et al., 2001). These results suggest that, despite severe lymphatic vascular defects, LVV and venous valve development are not affected in Vegfc^+/− and Vegfr3^+/chy embryos.

We analyzed LVVs and venous valves in E17.5 Vegfr3^+/EGFP;Vegfc^+/− embryos in which VEGF-C and VEGFR3 signaling is expected to be more than in Vegfc^−/−, but less than in Vegfc^+/− embryos. LVVs and venous valves were observed in wild-type and Vegfr3^+/EGFP embryos, as anticipated (Fig. 9A,B, white arrows and white arrowheads, respectively). Lymph sacs were observed, but LVVs were absent in E17.5 Vegfr3^+/EGFP;Vegfc^+/− embryos (Fig. 9C, red arrows). In contrast, venous valves of Vegfr3^+/EGFP;Vegfc^+/− embryos appeared indistinguishable from control littermates (Fig. 9C, white arrowheads). These results suggest that LVVs are more reliant on VEGF-C signaling for their development compared with venous valves.

We also analyzed E17.5 Vegfr3^chy/chy embryos in which VEGF-C/VEGFR3 signaling is completely abolished (Karkkainen et al., 2001; Zhang et al., 2010). Lymph sacs and LVVs were absent from these embryos (Fig. 9D, asterisk and red arrow, respectively). Importantly, very few PROX1⁺ cells were observed on the walls of veins where venous valves normally form (Fig. 9D, yellow arrowheads). It is possible that the lack of lymph sacs resulted in the absence of LVVs and venous valves. To test this possibility, we generated a new double-transgenic mouse model to abolish VEGF-C and VEGFR3 signaling in a time-specific manner after the development of lymph sacs. In these mice, 3rd generation reverse tetracycline-regulated transactivator (rtTA3) is expressed from the Rosa26 locus and is regulated by CAGG regulatory elements (CMV enhancer with chicken β-actin promoter). In addition, a miRNA-based shRNA targeting Vegfr3 is expressed from tetracycline response element (TetO) that is knocked into the constitutively active Col1a1 locus (Premsrirut et al., 2011). Doxycycline (Dox), when administered in food, will activate rtTA3. Transcriptionally active rtTA3 will bind TetO to activate the expression of shRNA that will knock down Vegfr3. We exposed these ShVegfr3;rtTA3 embryos to Dox from E12.5, after the formation of lymph sacs, and analyzed them at E17.5. The dorsal skin of E17.5 ShVegfr3;rtTA3 embryos was devoid of lymphatic vessels, which is consistent with the phenotypes of Vegfc^+/− and Vegfr3^+/chy embryos (Fig. 9E,F) (Karkkainen et al., 2004, 2001). Lymph sacs were observed in E17.5 ShVegfr3;rtTA3 embryos (Fig. 9G,H). However, they lacked LVVs (Fig. 9H, red arrow) and their venous valves were severely reduced in size (Fig. 9H, yellow arrowhead). The results from Vegfr3^chy/chy and ShVegfr3;rtTA3 embryos indicates that VEGF-C signaling is essential for the development of LVVs and venous valves.

Having determined that VEGF-C signaling is necessary for valve development, we tested whether YAP/TAZ activity is downregulated in Vegfc^+/− embryos. The LV-ECs of E18.5 wild-type embryos had aggregated with each other, giving the LVs a compact and mature appearance (Fig. 10A, arrow). The YAP/TAZ target protein CTGF was expressed in the LVs of wild-type embryos (Fig. 10A, arrow). In contrast, there were fewer LVs in E18.5 Vegfc^+/− embryos (Fig. 10B; data not shown). The LVs that were observed in E18.5 Vegfc^+/− embryos were immature and CTGF was dramatically downregulated in the developing LVs of Vegfc^+/− littermates (Fig. 10C, arrow). LVs further matured in 2-day-old (P2) wild-type pups, and expressed TAZ and the YAP and TAZ target protein ANGPT2 (Fig. 10D,E, arrows). In contrast, the LVs of P2 Vegfc^+/− littermates remained immature and did not express ANGPT2 (Fig. 10F, arrow). TAZ expression was also downregulated in the LVs of P2 Vegfc^+/− pups (Fig. 10G, arrow). YAP and TAZ expression was also downregulated in the mesenteric LVs of P2 Vegfr3^+/chy pups and E18.5 shVegfr3;rtTA3 embryos that were exposed to Dox for 2 days (Fig. S11). We sorted LECs from the mesentery of P2 wild-type and Vegfc^+/− pups, extracted RNA and performed qRT-PCR to estimate the expression of YAP and TAZ target genes. The expression of Prox1 and other YAP and TAZ target genes, Ctgf and Cyr61, was downregulated in the LECs of Vegfc^+/− pups (Fig. 10H). These results suggest that VEGF-C enhances YAP and TAZ activity in LVs.

In summary, VEGF-C and VEGFR3 signaling regulates LV, LVV and venous valve development. YAP and TAZ are essential mediators of VEGF-C signaling during valve development. VEGF-C signaling enhances the transcriptional activity of YAP and TAZ to promote the expression of genes such as PROX1, CTGF and ANGPT2 that are potently expressed in valves. Consequently, in the absence of YAP and TAZ, the development of lymphatic vessels, LV, LVV and venous valves is defective.

Our work has identified several previously unknown mechanisms that operate during lymphatic vascular development. We and others have previously reported a feedback loop between PROX1 and the VEGF-C/VEGFR3 signaling pathway in LECs (Koltowska et al., 2015; Srinivasan et al., 2014). PROX1 directly activates the expression of VEGFR3. In turn, VEGF-C and VEGFR3 signaling maintains the expression of PROX1. The mechanisms by which VEGF-C and VEGFR3 signaling regulates PROX1 expression are not fully understood. Here, we have identified YAP and TAZ as critical mediators of the PROX1 expression in response to VEGF-C signaling. VEGF-C is unable to enhance PROX1 expression in HLECs in the absence of YAP and TAZ. This finding establishes YAP and TAZ as crucial components of the PROX1 and VEGFR3 feedback loop.

PROX1, FOXC2, GATA2 and VEGFR3 are expressed at much higher levels in valvular endothelial cells compared with the rest of the lymphatic vasculature (Bazigou et al., 2009; Geng et al., 2016; Kazenwadel et al., 2015; Norrmén et al., 2009). Previous reports, including our own, have shown that PROX1, FOXC2 and GATA2 are necessary for the development of LVs and LVVs (Geng et al., 2016; Kazenwadel et al., 2015; Mahamud et al., 2019; Norrmén et al., 2009; Petrova et al., 2004). However, whether VEGFR3 is necessary for the development of valves has not been demonstrated, although superficial venous valve reflex is observed in individuals with Milroy's disease, which is characterized by heterozygous inactivating mutations in VEGFR3 (Mellor et al., 2010). In this work we have shown that VEGF-C and VEGFR3 signaling is indeed required for the proper development of LVs, LVVs and venous valves. Furthermore, we show that VEGF-C and VEGFR3 signaling activates YAP and TAZ, which in turn maintains the expression of PROX1 in valvular endothelial cells. These data suggest that the PROX1 and VEGFR3 feedback loop is operational in developing valves. Intriguingly, deletion of Yap and Taz did not affect the differentiation of valvular endothelial cells. Instead, between E16.5 and E18.5, PROX1 expression was downregulated and LVs and LVVs degenerated in the absence of YAP and TAZ. Thus, YAP and TAZ are crucial for the PROX1 and VEGFR3 feedback loop during the maturation stage of valve development.

YAP and TAZ are necessary for the proper functioning of migratory tip cells during angiogenesis (Kim et al., 2017; Neto et al., 2018; Sakabe et al., 2017; Wang et al., 2017). However, lymphangiogenesis happens normally in Yap/Taz^LECKO embryos until E16.5, after which their lymphatic vessels become dramatically dilated. Considering the facts that CTGF expression is enriched in valves, and that the onset of lymphatic vessel defects in Yap/Taz^LECKO embryos coincides with the degeneration of LVs and LVVs, we suggest that YAP and TAZ are playing a prominent role in valvular endothelial cells. Thus, degeneration of LVs and LVVs in Yap/Taz^LECKO embryos prevents lymphatic drainage, resulting in the dilation of lymphatic vessels. Nevertheless, lymphatic vessels of Prox1-heterozygous embryos that lack three out of four Yap/Taz alleles develop abnormal cysts that are not seen in Prox1^+/− embryos that lack LVVs and LVs. Thus, YAP and TAZ are likely also necessary in LECs. Additionally, deletion of Yap and Taz in a Prox1-heterozygous background resulted in the abnormal expression of endomucin in the LECs at E14.5. This phenotype is partially reminiscent of embryos in which Prox1 was deleted after the specification of LECs (Johnson et al., 2008). These results suggest that YAP and TAZ play a modest role in promoting Prox1 expression in LECs.

In contrast to our findings, Cho et al. showed that YAP and TAZ inhibit PROX1 expression in the developing lymphatic vessels (Cho et al., 2019). The differences could be due to the distinct Cre lines that were used. Although we used Lyve1-Cre, Cho et al. used Prox1-CreERT2 (Bazigou et al., 2011). Although Cho et al. observed severe edema in their mutant embryos, we did not observe edema in most of our samples. The more severe phenotype observed by Cho et al. could be due to more potent and/or rapid gene deletion by Prox1-CreERT2. As mentioned above, we suspect that YAP and TAZ play a primary role in maintaining the integrity of LVs and LVVs. Degeneration of LVVs at an earlier stage in the Yap/Taz^LECKO embryos generated by Cho et al. might have affected lymphatic drainage earlier, which in turn would have prevented lymphatic vessel maturation, resulting in sustained expression of PROX1. Deleting Yap and Taz at various developmental time points and in specific compartments of the lymphatic vasculature (LEC progenitors, LECs, LVs, LVVs, tip cells and stalk cells) could test these possibilities and provide better resolution of YAP and TAZ activity.

There are also differences in in vitro data generated by Cho et al. and us. Whereas we have determined that inhibition of YAP and TAZ activity by VP or siRNA inhibits PROX1 expression, Cho et al. draw the opposite conclusion from their results. It is possible that these differences are due to differences in cell lines or culture conditions that were used. Nevertheless, there are important points of congruity between our findings and those of Cho et al. Specifically, we also found that overactivation of YAP and TAZ, via pharmacological inhibition of MST1/2 or RNAi-mediated depletion of LATS1/2 reduces PROX1 expression in confluent HLECs. These results – both loss and overactivation of YAP and TAZ could result in the downregulation of PROX1 expression in the lymphatic vasculature – suggest that a precise level of YAP and TAZ activity regulates PROX1 expression.

Cell-density appears to function as a buffer that regulates YAP and TAZ activity in response to VEGF-C. We found that VEGF-C could reduce the phosphorylation of YAP and promote the expression of YAP and TAZ target genes only in HLECs grown under low-cell density. These later results are consistent with the report by Grimm et al., who elegantly showed that YAP is necessary for VEGF-C-induced proliferation of LECs in zebrafish (Grimm et al., 2019). The mechanisms and significance of this finding is currently unknown.

Many important questions remain regarding the role of YAP and TAZ in lymphatic vascular development. Multiple signaling pathways including shear stress, Wnt/β-catenin signaling and integrin/extracellular matrix signaling regulate YAP/TAZ activity (Azzolin et al., 2014, 2012; Dupont et al., 2011; Frye et al., 2018; Sabine et al., 2015; Zheng and Pan, 2019). These signaling pathways also regulate lymphatic vascular development and valve morphogenesis (Cha et al., 2016, 2018; Frye et al., 2018; Sabine et al., 2012, 2015). It remains to be seen whether these signaling pathways could activate YAP and TAZ in a functionally relevant manner during lymphatic vascular development. In primary human LECs, the transcription factor FOXC2 inhibits abnormal activation of TAZ by oscillatory shear stress (Sabine et al., 2015). Consequently, deletion of FOXC2 from the LVs of postnatal mice results in the abnormal activation of TAZ leading to the loss of quiescence, increased proliferation and apoptosis (Sabine et al., 2015). Whether deletion of Yap and Taz could ameliorate valve defects in mice lacking Foxc2 needs to be tested. Finally, XMU-MP-1 is used pharmacologically to protect the heart against pressure overload (Triastuti et al., 2019). It will be important to determine whether XMU-MP-1 could be repurposed to treat lymphedema when VEGFR3 signaling is compromised, as in the case of Milroy's disease.

Primary antibodies for immunohistochemistry on mouse tissues were as follows: rabbit anti-PROX1 (11-002, Angiobio; 1:500), goat anti-human PROX1 (AF2727, R&D Systems; 1:500), sheep anti-mouse FOXC2 (AF6989, R&D Systems; 1:300), goat anti-mouse VEGRF3 (AF743, R&D Systems; 1:300), rat anti-mouse CD31 (553,370, BD Pharmingen; 1:500), goat anti-human ANGPT2 (AF623, R&D Systems; 1:300), rabbit anti-mouse LYVE-1 (11-034, Angiobio; 1:3000), rabbit anti-human/mouse anti-YAP/TAZ (8418, Cell Signaling; 1:200), rabbit anti-human/mouse TAZ (HPA007415, Sigma; 1:200), rabbit anti-human/mouse CTGF (ab6992, Abcam; 1:200), rat anti-mouse endomucin (14-5851, eBioscience; 1:3000), rabbit anti-ZO-1 (40-2200, Invitrogen; 1:100), rabbit anti-human/mouse pHH3 (06-570, Millipore; 1:300) and goat anti-mouse GATA2 (AF2046, R&D Systems; 1:300).

Primary antibodies for immunocytochemistry: mouse anti-mouse/human YAP (sc-101199, Santa Cruz; 1:100), mouse anti-mouse/human TAZ (560235, BD Pharmingen; 1:100), goat anti-human PROX1 (AF2727, R&D Systems; 1:500), rabbit anti-mouse/human ZO-1 (40-2200, Invitrogen; 1:100) and rabbit anti mouse/human β-catenin (9562, Cell Signaling; 1:200).

Secondary antibodies for immunohistochemistry and immunocytochemistry were as follows: Cy3-conjugated donkey anti-rabbit (711-165-152; 1:500), Cy3-conjugated donkey anti-sheep (711-165-147; 1:500), and Cy5-conjugated donkey anti-rat (712-175-150; 1:500) and Alexa 488-conjugated donkey anti-goat (705-545-147; 1:500) were purchased from Jackson ImmunoResearch Laboratories. Alexa 488-conjugated goat anti-chicken (A-11039; 1:500) and Alexa 488-conjugated donkey anti-rat (A-21208; 1:500) were purchased from Life Technologies.

Primary antibodies for western blotting were as follows: mouse anti- β-actin (A5441, Sigma; 1:100,000), goat anti-human PROX1 (AF2727, R&D Systems; 1:1000), mouse anti-mouse/human YAP (sc-101199, Santa Cruz; 1:1000), mouse anti-mouse/human TAZ (560235, BD Pharmingen; 1:2000), rabbit anti-human/mouse pYAP (4911, Cell Signaling; 1:500), rabbit anti human/mouse Lats1/2 (3477, Cell Signaling; 1:500), rabbit anti-human/mouse pLATS1/2 (8654, Cell Signaling; 1:500), rabbit anti human/mouse pERK1/2 (4376, Cell Signaling; 1:1000), rabbit anti human/mouse ERK1/2 (4695, Cell Signaling; 1:1000), mouse anti-human VEGFR3 (MAB3757, Millipore; 1:1000) and rabbit anti GAPDH (PAB13195, Abnova; 1:1000).

HRP-conjugated secondary antibodies for western blotting were as follows: goat anti-mouse IgG (A4416, Sigma; 1:5000), goat anti-rabbit IgG (GtxRb-003-EHRPX, Immuno Reagent; 1:5000), donkey anti-goat IgG (705-035-003, Jackson ImmunoResearch; 1:5000) and donkey anti-sheep IgG (HAF016, R&D Systems; 1:5000).

We used de-identified primary human lymphatic endothelial cells (HLECs) for experiments. Dr Donwong Choi (Keck School of Medicine, University of Southern California, USA) provided the HLECs (Choi et al., 2019, 2017a,b, 2016). HLECs were grown on gelatin-coated plates or glass slides and were maintained in EBM2 media from Lonza. All experiments were conducted using passage 5-6 cells. HLECs were treated as potential biohazards and were handled according to institutional biosafety regulations.

VEGF-C (9199-vc-025/CF, R&D Systems), verteporfin (SML0534, Sigma Aldrich) and XMU-MP-1 (S8334, Selleck Chemicals) were diluted in manufacturer's recommended solvents.

ChIP assays were performed using the EZ-ChIP kit (MilliporeSigma) according to the manufacturer's instructions. Around 1.0×10⁷ HLECs were used per ChIP. Briefly, HLECs were grown on a culture dish at around 100% confluence. Subsequently, HLECs were fixed in 1% formaldehyde for 10 min at room temperature and glycine at a final concentration of 0.125 M was added for 5 min. Cells were washed with 20 ml of ice-cold PBS twice (10 min each) and harvested. Cells were lysed and sonicated as previously described (Cha et al., 2016, 2018).

Chromatin immunoprecipitation was performed using 0.2 μg of rabbit anti-human YAP (ab52771, Abcam) or 1.0 μg of normal rabbit IgG antibody (sc-2027, Santa Cruz Biotechnology). Following ChIP, PCR or q-PCR was performed using primers flanking the predicted TEAD4-binding site or a control site within the PROX1 promoter (5′-AGCCAGGGAATGAGTACAGG-3′ and 5′-AGGAAGCCTGTGCATTAACAC-3′).

Immunohistochemistry on sections was carried out according to our previously published protocols (Cha et al., 2016, 2018; Geng et al., 2016). Briefly, freshly collected embryos were washed in 1× PBS and fixed in 4% paraformaldehyde (PFA) overnight at 4°C. Subsequently, the embryos were washed three times (10 min each) in ice-cold PBS, incubated in 15% sucrose overnight at 4°C and then in 30% sucrose at 4°C until fully submerged in the solution. Embryos were then cryo-embedded in OCT solution (Sakura, Tokyo, Japan). Cryosections (12 μm) were prepared using a cryotome (Thermo Fisher Scientific, HM525 NX) and immunohistochemistry was performed using the indicated antibodies. E11.5 embryos were sectioned in a transverse orientation and E12.0-E16.5 embryos were sectioned frontally. Several consecutive sections were analyzed to determine the presence or absence of LVVs and VVs.

Whole-mount immunohistochemistry using embryonic skin or guts was performed according to our previous protocol (Cha et al., 2016, 2018). Either whole embryos or isolated guts were washed in 1×PBS and fixed in 1% PFA for 1 h to overnight (depending on the antibody) at 4°C. Subsequently, the dorsal skins were isolated, washed and samples were immunostained using the iDISCO protocol (Renier et al., 2014). Samples were visualized and analyzed as described previously (Cha et al., 2016, 2018). Fluorescent intensities were measured in a semi-quantitative manner using ImageJ.

Cells were fixed in 1% PFA at room temperature for 30 min. Cells were subsequently permeabilized with 0.3% Triton X-100 for 10 min at room temperature, then washed with PBST (PBS+0.1% Triton-X100) and blocked in 0.5% BSA PBST for 1 h at room temperature. Samples were incubated with primary antibodies at 4°C overnight. Samples were then washed with PBST and incubated with secondary antibodies for 2 h at room temperature, and then washed with PBST three times (10 min each), mounted and visualized as previously described (Cha et al., 2016, 2018).

HLECs were seeded at 40-50% confluence on plates. The following day, cells were transfected with equal amounts of siControl (51-01-14-03, Integrated DNA Technologies), siYAP (hs.Ri.YAP1.13.1, hs.Ri.YAP1.13.3) and siTAZ (hs.Ri.TAZ.13.2, hs.Ri.TAZ.13.3) or siLATS1 (hs.Ri.LATS1.13.1, hs.Ri.LATS1.13.3) and siLATS2 (hs.Ri.LATS2.13.2, hs.Ri.LATS2.13.3) using Lipofectamine RNAimax (Thermo Fisher Scientific) according to the manufacturer's instructions. After 2-3 days cells were treated with VEGF-C and harvested with Trizol (Invitrogen) or RIPA buffer for qRT-PCR or western blotting, respectively.

Prox1^+/Cre (Srinivasan et al., 2010), Tg(Prox1-tdTomato) (Gong et al., 2003), Lyve1-Cre (Pham et al., 2010), Vegfr3^+/EGFP (Ichise et al., 2010), Yap^flox and Taz^flox mice have been described previously (Xin et al., 2013, 2011). Vegfc^+/CreERT2 mice were generated by Cyagen by inserting the cDNA for CreERT2 immediately downstream of ATG at the Vegfc locus. Mirimus generated the shVegfr3;rtTA3 mice according to their published protocols (Premsrirut et al., 2011).

Prox1^+/Cre mice were maintained in NMRI background. Other mice were maintained in C57BL6 or C57BL6/NMRI mixed backgrounds. We used both male and female mice for the experiments. All mice were housed and handled according to the institutional IACUC protocols.

PROX1 regulatory elements were analyzed through ENCODE (ENCODE Project Consortium, 2012). For targeted analysis, PROX1 regulatory element sequences obtained from Ensembl were aligned using Clustal Omega, and Homer was used for TEAD4 binding site identification (Heinz et al., 2010; Madeira et al., 2019; Yates et al., 2020).

Total RNA was purified from HLECs treated with VP for 2 h. RNA was subjected to ribosomal RNA depletion followed by Truseq stranded total RNA library preparation according to the manufacturer's instruction (Illumina). The resulting RNA-seq libraries were analyzed on the Illumina HiSeq sequencing platform.

The obtained sequencing reads were mapped with the bowtie2 algorithm using the RefSeq annotations (hg19 genome build) (Langmead and Salzberg, 2012). We used the RNA-seq analysis work flow within the Partek Genomics Suite (Partek Incorporated) for quantitation and statistical analysis (ANOVA) of the transcriptome data. We identified those transcripts that exhibited statistically significant differential expression in the VP-treated samples compared with the control samples. We rank ordered the two lists based on the expression level and magnitude of change. Using these rank-ordered lists, we performed gene ontology (GO) analysis for enriched biological terms (Eden et al., 2009). The differentially expressed genes were analyzed using the functional annotation platform of DAVID (Huang et al., 2009a,b).

Scanning electron microscopy was performed according to our previous protocol (Geng et al., 2016; Geng and Srinivasan, 2018).

For biochemical analysis, n indicates the number of times the experiments were independently performed and for histological analysis n indicates the number of embryos analyzed per genotype. All experiments were performed at least three times or more. Data were presented as mean±s.e.m. GraphPad Prism 7 software was used to perform the statistical analysis. Data were analyzed using the unpaired, two-tailed, Student's t-test. P<0.05 was considered significant.

Cells were harvested with RIPA lysis buffer and western blots were performed using standard protocol. The density of bands was measured by ImageJ and presented as mean±standard deviation (s.d.).

We thank Dr Eric Olson (University of Texas Southwestern) for the Yap^+/f and Taz^+/f mice, Dr Hirotake Ichise (University of Ryukyus) for the Vegfr3^+/EGFP mice, Dr Pierre Chambon for CreERT2 cDNA, Dr Alfonso Lavado (St Jude Children's Research Hospital) for insightful comments, Dr Angela Andersen (Life Science Editors) for editorial assistance, and Mrs Lisa Whitworth and Mr Brent Johnson (Oklahoma State University) for scanning electron microscopy.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.195453.reviewer-comments.pdf