Our group previously reported that FOXC1 and FOXC2 expression co-localizes with PROX1 in lymphatic valve-forming cells at E17 and later at P3 (Fatima et al., 2016). However, the expression pattern of FOXC1 in the mesenteric lymphatic collecting vessels and valves in adult mice remains unknown. We first characterized the expression pattern of FOXC1 and FOXC2 in mature valves of 4 week old adult mice to delineate possible differential or cooperative roles during valve maturation and maintenance. Immunostaining of mesentery tissue with FOXC1, FOXC2, and VEGFR3 antibodies identified colocalization of FOXC1 and FOXC2 within the nuclei of intraluminal valve leaflets while FOXC2 expression was more highly enriched in the valve sinuses and surrounding lymphangion compared to FOXC1 (Figure 1). Of note, FOXC1 expression was most highly enriched in cells located at the leading free-edge (Bazigou et al., 2009; Danussi et al., 2013; Bazigou and Makinen, 2013; Sabine et al., 2018) of the intraluminal side of valve leaflets exposed to pulsatile laminar shear stress (LSS) forces during valve opening/closure cycles (Sabine et al., 2016).

Previous work from our group has demonstrated the critical role for FOXC2 in postnatal lymphatic vascular function (Sabine et al., 2015) and we have also shown that EC-specific deletion of Foxc1 during murine embryonic development impairs lymphatic valve maturation (Fatima et al., 2016). However, the temporal regulatory role of FOXC1 transcriptional activity during valve development, maturation, and maintenance is not well understood. To investigate FOXC1 function in the early postnatal lymphatic vasculature and valve regions, we crossed conditional null Foxc1^fl mice (Sasman et al., 2012) with Chd5-Cre^ERT2 mice (Sörensen et al., 2009) to generate tamoxifen-inducible, EC-specific Foxc1 mutant (EC-Foxc1-KO) mice. Tamoxifen was administered from P1-P5 to induce Cre-mediated recombination and we confirmed deletion of Foxc1 via qPCR analysis of isolated CD31-positive cells from hearts of P6 individuals and by immunostaining of the mesenteric lymphatic vasculature with antibodies against FOXC1, FOXC2, and VEGFR-3 (Figure 2—figure supplement 1a,d,e). We next investigated whether lymphatic valve maturation and maintenance was affected in collecting vessels after induction of Cre-mediated recombination postnatally (Figure 2a–f,u,v). Quantification of regions with PROX1-high expression within lymphatic collecting vessels, indicative of valves, showed there was not a significant difference in total valve number in mice with inactivated Foxc1 compared to littermate controls at P6 (Figure 2u). As VE-Cadherin, coded by the Cdh5 gene, is also expressed in the blood vasculature, we immunostained the mesenteric vasculature with CD31 antibody and found no obvious changes associated with loss of Foxc1 (Figure 2a,b). Using EC-specific Foxc1 mutant mice generated from the Tek-Cre strain, our group previously reported that the proportion of lymphatic valves that had formed mature, v-shaped or semilunar bi-leaflet structures at P7 was significantly reduced in EC-specific Foxc1 mutant mice compared to littermate controls (Fatima et al., 2016). To investigate whether this result was recapitulated using the inducible Cdh5-Cre strain, we also performed immunostaining of mesentery tissue for expression of a marker of valve maturation, α9-integrin, which is a receptor that interacts with its extracellular matrix (ECM) protein ligand fibronectin-EIIIA to regulate the formation of the ECM core of valve leaflets (Bazigou et al., 2009) in addition to PROX1/CD31 immunostaining (Figure 2c – f). Quantification of the proportion of lymphatic valves that had formed mature, v-shaped or semilunar bi-leaflet structures showed a significant reduction in EC-Foxc1-KO mice (Figure 2v), thus recapitulating our previously reported results using the Tek-Cre strain. To verify that the phenotype we observed was not mediated by inactivation of Foxc1 within the blood vasculature, we next crossed conditional null Foxc1^fl mice with Prox1-Cre^ERT2 mice (Srinivasan et al., 2007) to generate inducible LEC-specific Foxc1 mutant (LEC-Foxc1-KO) mice and administered tamoxifen postnatally. PROX1 immunostaining showed no significant differences in total valve number of LEC-Foxc1-KO mice (Figure 2—figure supplement 2a,b,e). Additionally, immunostaining of PROX1 and FOXC1 showed enrichment of FOXC1 expression in intraluminal valve leaflets of littermate control mice (Figure 2—figure supplement 2c) whereas mature lymphatic valves were still observed in LEC-Foxc1-KO mice with reduced FOXC1 expression (Figure 2—figure supplement 2d). However, quantification of mature lymphatic valve structures demonstrated that LEC-Foxc1-KO mice had significantly less mature valves (Figure 2—figure supplement 2f), recapitulating the phenotype observed utilizing the Cdh5-Cre^ERT2 strain (Figure 2). While similar phenotypes were observed between both Cdh5-Cre^ERT2- and Tek-Cre- Foxc1 mutants, these data demonstrate that FOXC1 function is maintained during postnatal lymphatic valve maturation and maintenance as Tek-Cre is expressed embryonically as early as E7.5 (Braren et al., 2006).

Previously, our group reported inducible LEC-specific deletion of Foxc2 via Prox1-Cre^ERT2 (Bazigou et al., 2011) at P4 was shown to significantly reduce total valve number in mesenteric lymphatic collecting vessels and a subset of Foxc2 mutants presented chylous ascites and chylothorax beginning as early as 3 days after tamoxifen treatment with eventual fully penetrant mortality (Sabine et al., 2015). To compare the roles of FOXC1 and FOXC2 in our model, we crossed Cdh5-Cre^ERT2 mice with conditional null Foxc2^fl mice to generate EC-specific Foxc2 mutant (EC-Foxc2-KO) mice and administered tamoxifen from P1-P5 with inactivation confirmed by qPCR in ECs isolated from P6 hearts and by immunostaining of the mesenteric lymphatic vasculature with antibodies against FOXC1, FOXC2, and VEGFR-3 (Figure 2—figure supplement 1b,f,g). Interestingly, while FOXC2 was strongly reduced in collecting vessels of EC-Foxc2-KO mice, we also observed a reduction in FOXC1 expression in collecting vessels, and particularly valve regions, compared to littermate controls (Figure 2—figure supplement 1f and g). Although chylous effusion was occasionally observed in mutant individuals at P6 (4/30; 13%), inactivation of Foxc2 beginning at P1 significantly reduced the total number of valves in mesenteric lymphatic collecting vessels (Figure 2g – l, q, r, w, x). However, no obvious changes were observed in the mesenteric blood vasculature of Foxc2 mutants by CD31 immunostaining (Figure 2g,h). Quantitative analysis also determined that the percentage of the total lymphatic valves with mature, intraluminal bi-leaflet structures was significantly reduced with inactivation of Foxc2 and many valve regions were characterized by collecting vessel constriction, indicative of regression of these structures (Figure 2i – l, x). Several of these valve regions were characterized by intraluminal leaflets shortened in length compared to littermate controls, yet a majority of valve regions were characterized by near complete degeneration of intraluminal valves (Figure 2i – l), similar to our group’s observations previously reported in LEC-Foxc2-KO mice (Sabine et al., 2015). Moreover, expression of α9−integrin was reduced in EC-Foxc2-KO mice compared to littermate controls (Figure 2j – l), indicating that the phenotype observed in EC-Foxc2-KO mice is also associated with impaired valve maturation processes. Thus, our results using a different endothelial-specific Cre strain further support the previously reported role for FOXC2 in postnatal lymphatic vascular function and highlight a more prominent role than FOXC1 in valve maintenance.

Our group previously reported overlapping expression patterns of Foxc1 and Foxc2 in the embryonic murine cardiovascular system and key, cooperative functional roles during embryonic lymphangiogenesis (Kume et al., 1998; Winnier et al., 1999; Kume et al., 2001; Seo et al., 2006; Fatima et al., 2016). To assess the requirement and potential cooperative role of both transcription factors in postnatal lymphatic vascular function, we crossed conditional null Foxc1^fl; Foxc2^fl mice (Sasman et al., 2012) with Cdh5-Cre^ERT2 mice to generate EC-specific Foxc1; Foxc2 mutant (EC-Foxc1; Foxc2-DKO) mice and administered tamoxifen from P1 to P5 with inactivation in compound mutant individuals verified by qPCR and immunostaining of the mesenteric lymphatic vasculature with antibodies against FOXC1, FOXC2, and VEGFR-3 (Figure 2—figure supplement 1c,h,i). At P6, a majority of compound EC-Foxc1; Foxc2 mutant mice (48/78, 61.5%) developed severe chylous ascites (Figure 2o,p) and chylothorax (data not shown). PROX1 immunostaining showed that PROX1-high expressing valve regions were nearly absent in compound mutants whereas CD31 immunostaining determined that the blood vasculature appeared normal (Figure 2m,n,y). Compared to littermate controls, lymphatic collecting vessel hierarchy of EC-Foxc1; Foxc2-DKO mutants was markedly abnormal, characterized by high numbers of interconnecting and inter-looping branches resembling a primitive lymphatic vascular plexus prior to valve differentiation, maturation, and remodeling (Figure 2s,t). Furthermore, direct comparison of lymphatic collecting vessel hierarchy of EC-Foxc1; Foxc2-DKO mice with EC-Foxc2-KO mice shows that combined deletion of Foxc1 and Foxc2 results in a synergistic response (Figure 2q–t). Our group previously reported that ectopic sprouting from degenerating valves was observed in LEC-Foxc2 mutant mice (Sabine et al., 2015), thus the markedly abnormal branching pattern observed in EC-Foxc1; Foxc2-DKO mice is likely a result of accelerated valve regression and ectopic sprouting from these regions to compensate drainage from areas of perturbed flow due to the near complete regression of valves. To verify that Foxc1 and Foxc2 inactivation did not simply reduce PROX1 expression in valve regions, we performed additional immunostaining of mesentery tissue for α9-integrin expression. Modest α9-integrin expression was detected in regressing valve leaflets of EC-specific Foxc2 mutants (Figure 2—figure supplement 3a,b,e,f), however expression was absent in collecting vessel branches of EC-Foxc1; Foxc2-KO mutant where valves are typically present (Figure 2—figure supplement 3c,d,g,h). Collectively, we demonstrate that both FOXC1 and FOXC2 are required for postnatal lymphatic vascular function. FOXC2 has a predominant role in maintaining LEC valve identity, however FOXC1 functions in a cooperative manner with FOXC2 to refine valve maturation processes.

Collecting vessels of LEC-specific Foxc2 mutants were previously characterized by the presence of rounded LECs with discontinuous, ‘zig-zag’ like junctions as opposed to elongated cells with linear, ‘zipper-like’ junctions aligned with flow. Furthermore, increased apoptosis and inappropriate LEC proliferation was detected in degenerating valve LECs of Foxc2 mutants, which was associated with abnormal activation of TAZ signaling and impaired regulation of LEC quiescence and survival (Sabine et al., 2015). Linearized LEC junctions in collecting vessels help to minimize loss of lymph during its transport (Baluk et al., 2007) and disruption of LEC barrier integrity in LEC-Foxc2-KO mice contributed to leakage of lymph and the development of chylous ascites and chylothorax (Sabine et al., 2015). To first determine how inactivation of both Foxc1 and Foxc2 affects collecting vessel LEC junctional integrity, we immunostained mesenteric lymphatic vessels for PROX1 and VE-Cadherin expression (Figure 2—figure supplement 4a–h and Videos 1–4). As previously reported, LECs in wild-type mice were elongated, organized and aligned with the direction of flow, and had continuous cell-cell junctions (Figure 2—figure supplement 4a,c,e,g, Video 1, and Video 3) whereas EC-specific inactivation of Foxc2 resulted in a more rounded shape with disrupted endothelial junctions (Figure 2—figure supplement 4b,f, and Video 2). However, inactivation of both Foxc1 and Foxc2 severely impaired LEC organization as cells appeared more spherical in shape and were poorly elongated in alignment with flow (Figure 2—figure supplement 4d,h, and Video 4). Surprisingly, we not only observed the presence of discontinuous cell-cell junctions in compound Foxc1; Foxc2 mutant mice, but the distribution of VE-Cadherin at cell junctions was broader compared to littermate controls and Foxc2 mutants (Figure 2—figure supplement 4e – h, and Video 4), indicative of an increase in overlapping cell junctions. Because the presence of increased overlapping cell-cell junctions was observed in compound Foxc1; Foxc2 mutant mice, but not Foxc2 mutants, this suggests a separate role for FOXC1 and FOXC2 in regulation of LEC intercellular junctions.

Our group previously reported that active caspase 3-positive LECs were detected only in minor instances throughout collecting vessels of control mice, whereas apoptosis in LEC-specific Foxc2 mutants was more frequently detected and limited to degenerating valves (Sabine et al., 2015). To next investigate the extent and severity of apoptosis in EC-Foxc1; Foxc2-DKO mice, we additionally performed immunostaining of mesentery tissue with PROX1 and active caspase-3 antibodies. In contrast to what was previously reported in LEC-Foxc2-KO mice, a significantly greater number of apoptotic cell bodies were detected throughout the lymphatic collecting vessels of EC-Foxc1; Foxc2-DKO mice compared to littermate controls (Figure 2—figure supplement 4i–k). Of note, apoptotic bodies were more frequently observed in closer proximity to branched regions of the lymphatic collecting vessels (Figure 2—figure supplement 4j) potentially indicating areas of valve degeneration.

Together, these data indicate that FOXC1 and FOXC2 have a complementary role in regulating lymphatic collecting vessel function as both transcription factors are required to maintain valve and LEC barrier integrity to properly transport lymph and limit leakage.

Embryonic lymphatic valve formation in mouse embryos is initiated at the time of active lymphatic drainage (Sabine et al., 2012) and valves are primarily found at sites of vessel branching points and bifurcations where perturbed lymph flow is present (Kampmeier, 1928; Sabine et al., 2012). In vitro, OSS forces acting on LECs upregulate the expression of FOXC2 and GATA2 as well as induce the activation of NFATc1 (Kazenwadel et al., 2015; Sweet et al., 2015; Sabine et al., 2012; Sabine et al., 2015). In cultured human arterial ECs, exposure to LSS forces induced expression of both FOXC1 and FOXC2 (Chen et al., 2017). Furthermore, in vivo models of reduced shear stress in the blood vasculature of zebrafish showed reduced expression of foxc1a and foxc1b (orthologs of mammalian FOXC1 and FOXC2, respectively) compared to controls (Chen et al., 2017). To investigate how different shear stress stimuli affect FOXC1 and FOXC2 expression in LECs, we examined cells cultured under static, laminar, or oscillatory flow conditions. As previously observed, OSS strongly increased the expression of FOXC2 and the formation of overlapping cell-cell junctions (Sabine et al., 2012; Sabine et al., 2015). However, FOXC1 expression was slightly reduced compared to cells cultured under static conditions. In contrast, LSS upregulated both FOXC1 and FOXC2 expression as cells became aligned with unidirectional laminar flow (Figure 3). No correlation was observed between FOXC1 and FOXC2 expression under different shear stress conditions (Figure 3—figure supplement 1a–c). Therefore, mechanical stress induced by flow differentially regulates LEC FOXC1 and FOXC2 expression (Figure 3c). Thus, our in vitro data support our in vivo observations of differences in FOXC1 and FOXC2 expression pattern observed in valves due to exposure to both disturbed flow in valve sinuses and pulsatile laminar flow on the intraluminal side of leaflets, likely contributing to high expression of both FOXC1 and FOXC2 in cells located at the free edge of valve leaflets (Figure 1). Furthermore, the reduction of FOXC1 expression observed in collecting vessels of EC-Foxc2-KO mice in vivo (Figure 2—figure supplement 1g) is likely attributable to reduced flow as a result of perturbed barrier integrity and increased leakiness from lymphatic collecting vessels.

OSS acting on LECs induces the formation of thick cortical actin fibers and perinuclear F-actin stress fiber formation (Sabine et al., 2012; Sabine et al., 2015). Similarly, increased F-actin expression was detected in lymphatic valve forming cells in vivo (Sabine et al., 2012). However, FOXC2 knockdown in cultured LECs induced actomyosin contractility, which was potentiated by OSS leading to impaired cell-cell adhesion (Sabine et al., 2015). To investigate potential cytoskeletal signaling changes in the context of loss of FOXC1 with shear stress, LECs were transfected at subconfluency with scramble control or FOXC1 siRNAs, then seeded into fibronectin-coated flow chambers under static, LSS, or OSS conditions. FOXC1 knockdown resulted in impaired adhesion and reduced viability of LECs as the monolayer of cells re-seeded into flow chambers was characterized by several areas devoid of cells and cells were observed floating over the monolayer under static conditions, suggesting cells had been flushed away under LSS and OSS (Figure 4). Intriguingly, FOXC1 inactivation also resulted in a substantial increase of transverse actin stress fiber formation associated with increased phosphorylation of MLC2 (p-MLC2) (Figure 4a), indicative of increased actomyosin contractility. This was potentiated under both OSS and LSS, with cells also losing their ability to properly orient in the direction of LSS.

Considering that FOXC1 knockdown in LECs impairs cell adhesion, we slightly modified our protocol: LECs were first seeded on fibronectin-coated surfaces, then transfected with scramble control, FOXC1, FOXC2, or combined FOXC1/FOXC2 (50%:50%) siRNAs and kept under static conditions for 2 days. Given that FOXC2 inactivation and exposure to OSS in LECs results in increased actomyosin contractility (Sabine et al., 2015), we investigated whether combined knockdown of FOXC1 and FOXC2 in LECs would result in a synergistic effect. FOXC1-KD LECs were characterized by strongly increased co-localization of p-MLC2 signal with abnormally accumulated transverse F-actin stress fibers compared to control and FOXC2-KD LECs with modest p-MLC2 signal co-localized mostly to the cortical actin ring (Figure 4b). Combined knockdown of FOXC1 and FOXC2 resulted in a similar phenotype to FOXC2-KD LECs (Figure 4b), suggesting a rate-limiting effect of FOXC1 in the control of actomyosin contractility.

As FOXC2 has been shown to be required to maintain continuous cell-cell junctions in the lymphatic endothelium (Sabine et al., 2015) and we observed differences in VE-Cadherin distribution between Foxc2 and compound Foxc1; Foxc2 mutants in vivo (Figure 2—figure supplement 4), we investigated possible differences in VE-Cadherin junctions in FOXC1-KD and FOXC2-KD LECs. Surprisingly, we observed a greater distribution of overlapping VE-Cadherin-positive junctions in FOXC1-KD LECs (Figure 4—figure supplement 1a). In contrast, as described previously, FOXC2-KD LECs were characterized by discontinuous VE-Cadherin junctions (Figure 4b and Figure 4—figure supplement 1a). However, both overlapping and discontinuous VE-Cadherin-positive junctions were observed in FOXC1/FOXC2-KD LECs (Figure 4—figure supplement 1a), thus recapitulating our observations in vivo (Figure 2—figure supplement 4d,h) and indicating separate roles for FOXC1 and FOXC2 regulation of intercellular junctions.

Cell adhesion to ECM is controlled by focal adhesions, which are composed of protein complexes that link the actin cytoskeleton, and especially stress fibers, to the ECM and are critical for translating signals from the ECM environment (Parsons et al., 2010). Because FOXC1 knockdown impaired LEC attachment to fibronectin, we assessed the distribution of two main focal adhesion components, vinculin and paxillin, in FOXC1-KD, FOXC2-KD, and FOXC1/FOXC2-KD LECs to compare the role of FOXC1 and FOXC2 in regulating focal adhesions (Figure 4c and Figure 4—figure supplement 1b). Knockdown of FOXC1 mostly induced formation of vinculin-positive or paxillin-positive focal adhesions at the tip of actin stress fibers (Figure 4c and Figure 4—figure supplement 1b, respectively), whereas knockdown of FOXC2 mostly induced formation of vinculin-positive or paxillin-positive adherens junctions associated with the cortical actin ring (Figure 4c and Figure 4—figure supplement 1b, respectively). Interestingly, combined knockdown of FOXC1 and FOXC2 induced formation of both focal adhesions and adherens junctions in LECs (Figure 4c, Figure 4—figure supplement 1b). Collectively, our data demonstrate that FOXC1 and FOXC2 have separate, yet complementary, roles to regulate LEC cytoskeletal organization and contractility as FOXC1 controls focal adhesions and FOXC2 regulates adherens junctions.

The RhoA-Rho-associated protein kinase (ROCK)-phosphorylated myosin light chain (pMLC) signaling pathway has been well established as a regulator of cytoskeletal contractility mechanisms (Ridley, 2001). Furthermore, perturbed RhoA/ROCK signaling activation has been demonstrated to be a negative regulator of blood endothelial tubulogenesis (Bowers et al., 2016) and barrier function (Spindler et al., 2010). Given the synergistic effect of combined inactivation of Foxc1 and Foxc2 on lymphatic valve maintenance and maturation in vivo and perturbed cytoskeletal organization upon FOXC1 and FOXC2 knockdown in vitro, we hypothesized that FOXC1 and FOXC2 transcription targets may converge to negatively regulate RhoA/ROCK activation in the context of mechanical stress. To identify potential negative regulators of RhoA/ROCK downstream of FOXC1 and FOXC2 transcription, we utilized RNA-seq data previously published from our LEC- Foxc1; Foxc2-DKO mice (Fatima et al., 2016). Recent evidence has shown that a signaling complex, consisting of the planar cell polarity (PCP) protein Prickle1 and the Rho GTPase activating proteins (GAPs) Arhgap21 and Arhgap23, functions to inhibit RhoA on the non-protrusive lateral membrane cortex allowing for dynamic cell morphology necessary for proper migration (Zhang et al., 2016). Our RNA-seq analysis (Fatima et al., 2016) revealed that mRNA expression of Prickle1, Arhgap21, and Arhgap23 was significantly downregulated in dermal LECs isolated from LEC- Foxc1; Foxc2-DKO mouse embryos (Figure 5a). Furthermore, our RNA-seq analysis showed a modest, yet significant reduction of RhoA, but other GAPs associated with regulation of RhoA, endothelial barrier function, and lumen maintenance (van Buul et al., 2014; Barry et al., 2016) were not differentially expressed with the exception of a significant increase in Arhgap18 expression and a modest reduction of Arhgap20 expression (Figure 5—figure supplement 1k). Because of the previous report demonstrating a physical interaction of Prickle1 and Arhgap21/23 complex (Zhang et al., 2016) and our observation that expression of all three genes was significantly downregulated in our RNA-seq analysis, we focused our attention on these potential downstream targets. Therefore, we performed in silico analysis to identify putative FOXC binding sites in the human PRICKLE1, ARHGAP21, and ARHGAP23 loci. First, the Hypergeometric Optimization of Motif EnRichment (HOMER) suite of tools (Heinz et al., 2010) was used to identify regions containing the conserved RYMAAYA FOX consensus binding sequence (Pierrou et al., 1994; De Val et al., 2008; Norrmén et al., 2009) in active areas of transcription identified with the Encyclopedia of DNA Elements (ENCODE) (2012) and the UCSC Human Genome Browser (Kent et al., 2002). Then, we utilized the Evolutionary Conserved Region (ECR) Browser (https://ecrbrowser.dcode.org) (Ovcharenko et al., 2004) tool to identify conserved binding regions between the mouse and human genomes. Several conserved and aligned putative FOX binding regions were identified in regions of the PRICKLE1 (Figure 5b,c), ARHGAP21, and ARHGAP23 (Figure 5—figure supplement 1a,b) loci. To identify direct interaction of FOXC1 and FOXC2 with these putative binding regions, human dermal LECs from juvenile foreskin (HDLECs) were cultured, and ChIP assays were performed using two specific antibodies for FOXC1 and one antibody for FOXC2 for regions identified in the PRICKLE1 (Figure 5d), ARHGAP21, and ARHGAP23 loci (Figure 5—figure supplement 1c,d). We found that binding of both FOXC1 and FOXC2 were significantly enriched in ECR-2, and −6 in PRICKLE1 whereas specific binding of FOXC1 was significantly enriched in ECR-5 and FOXC2 in ECR-1 (Figure 5e–j). Additionally, FOXC1 and FOXC2 binding was significantly enriched in ECR-1–1 of the ARHGAP21 locus and FOXC2 was significantly enriched in ECR-1–2 (Figure 5—figure supplement 1e–g). Finally, only FOXC2 binding was significantly enriched in ECR-1 of the ARHGAP23 locus (Figure 5—figure supplement 1h).

Although both FOXC1 antibodies recognize the same immunogen peptide, we observed notable differences in ChIP signals between the two antibodies. However, significant variations dependent on the antibody manufacturer have been previously noted for antibodies targeted to the same antigen in immunohistochemistry related work (Ramos-Vara, 2005). To validate our ChIP assays, we also performed a negative control experiment using primers to amplify a portion of the promoter region of ICAM1 that was not predicted to bind FOX transcription factors by in silico analysis (Figure 5—figure supplement 1i). ChIP analysis using the antibodies against FOXC1, FOXC2, and IgG for the ICAM1 region from three experimental replicates showed no band signals (Figure 5—figure supplement 1j) validating specificity of FOXC1 and FOXC2 binding to these loci in LECs. Thus, these results demonstrate that FOXC1 and FOXC2 have similar binding capacities for the PRICKLE1 and ARHGAP21 loci whereas FOXC2 shows specificity for the ARHGAP23 locus.

To assess whether Prickle1 is regulated by FOXC1 and FOXC2 in lymphatic valves in vivo, we characterized its expression in lymphatic valves during postnatal development. Prickle1 expression was concentrated on the free edges of PROX1-high valve leaflets located at or near the valve buttress (Sabine et al., 2018) with modest expression detected in other leaflet cells (Figure 5k,m,o, Figure 5—figure supplement 2a–d,g,h, Video 5, and Video 6). Postnatal deletion of Foxc1 reduced Prickle1 expression within leaflet free-edge cells while other valve leaflet LECs retained modest expression (Figure 5l, Figure 5—figure supplement 2e,f,i and Video 7). In contrast, postnatal deletion of Foxc2 or both Foxc1 and Foxc2 resulted in a strong and broad reduction of Prickle1 within the lymphatic vasculature compared to littermate controls (Figure 5m–p). Collectively, these data show that PRICKLE1 is a novel target of FOXC transcription factors in valve forming LECs.

To investigate whether ROCK signaling is activated in the absence of FOXC1 and FOXC2, LECs were transfected with scramble Control, FOXC1, or FOXC2 siRNAs under static conditions and treated with vehicle Control or the ROCK inhibitor Y-27632 (Figure 6). Immunostaining and quantification of F-actin and p-MLC2 demonstrated an increase in the F-actin and p-MLC2 area per cell with loss of FOXC1 and treatment with vehicle. However, inhibition of ROCK reduced F-actin and pMLC2 distribution to a similar level in Control LECs (Figure 6a,b). In contrast, loss of FOXC2 modestly increased F-actin and pMLC2 area per cell, but the increase in relative area per cell was greatly potentiated by OSS. However, similar to FOXC1 knockdown, changes in F-actin and p-MLC2 area were ameliorated by treatment with Y-27632 (Figure 6c,d). Together, this data demonstrate that key targets downstream of FOXC1 and FOXC2 transcription activity function in part to negatively regulate RhoA/ROCK activation in LECs.

As ROCK inhibition rescued impaired cytoskeletal organization in FOXC1- and FOXC2-KD LECs, we assessed whether ROCK inhibition could rescue lymphatic valve degeneration observed in EC-Foxc2-KO and EC-Foxc1; Foxc2-DKO mice as LEC-specific deletion of Foxc1 and Foxc2 did not affect Rock1 and Rock2 expression (Figure 5—figure supplement 1k). Compared to P7 Foxc2 mutants administered DPBS, Foxc2 mutants administered Y-27632 concurrently with tamoxifen from P2 – P5 were able to retain a significantly higher total valve number, similar to totals quantified in littermate control mice injected with DPBS or Y-27632 (Figure 7a–d,i). However, the proportion of mature lymphatic valves present in Foxc2 mutants after ROCK inhibition was significantly lower compared to littermate controls treated with DPBS or Y-27632, but significantly increased compared to Foxc2 mutants administered DPBS (Figure 7j). In contrast, ROCK inhibition was not able to rescue near complete lymphatic valve degeneration in P7 EC-Foxc1; Foxc2-DKO mutants (Figure 7e–h,k).

Although ROCK inhibition was not able to completely rescue the valve phenotypes associated with EC-Foxc2-KO nor EC-Foxc1; Foxc2-DKO mice, we assessed the effect of ROCK inhibition on LEC junctional integrity in collecting vessels of both mutant mouse lines as Y-27632 treatment rescued impaired LEC cytoskeletal organization in vitro. Immunostaining of VE-Cadherin in collecting vessels of P7 EC-Foxc2-KO and EC-Foxc1; Foxc2-DKO mutant mice treated with DPBS identified disrupted LEC cell-cell junctions (Figure 8a,b,e,f) similar to our observations in untreated mutant mice at P6 (Figure 2—figure supplement 4) contributing to loss of lymphatic vascular barrier function. However, ROCK inhibition in both EC-Foxc2-KO and EC-Foxc1; Foxc2-DKO mutant mice was able to in part restore continuous cell-cell ‘zipper-like’ junctions characteristic of lymphatic collecting vessels (Figure 8c,d,g,h), thus resulting in improved junction integrity. Collectively, our in vitro and in vivo data demonstrate that FOXC1 and FOXC2 function in part to negatively regulate perturbed activation of Rho/ROCK signaling in LECs downstream of shear stress-sensing mechanisms.

Because the forkhead DNA binding domain is nearly identical between FOXC1 and FOXC2 and both transcription factors have cooperative roles in early cardiovascular development (Kume, 2009; Lam et al., 2013), we assessed whether FOXC2 could substitute for FOXC1 transcriptional activity in vivo in the context of lymphatic valve development, maintenance, and maturation. To do so, we generated mice that carry a Foxc2 knock-in (Foxc1^c2) allele, in which the Foxc1 coding region has been replaced with the cDNA coding (from the start codon to the stop codon) for Foxc2 (Figure 9a – c). By breeding male and female Foxc1^c2/+ mice, we were able to generate litters consisting of Foxc1^+/+, Foxc1^c2/+, and Foxc1^c2/c2 mice for analysis. To verify the absence of FOXC1 expression within the lymphatic vessels of this model, we first performed co-immunostaining of mesentery tissue from P12 Foxc1^+/+ and littermate Foxc1^c2/c2 mice for FOXC1 and VEGFR-3 (Figure 9d and e), which confirmed the complete reduction of FOXC1 expression in lymphatic collecting vessels in homozygous Foxc2 knock-in mutant mice (Figure 9e). To further characterize changes in expression of FOXC1 and FOXC2 in both the blood and lymphatic vasculature, we also performed co-immunostaining for FOXC1, FOXC2, and VE-Cadherin in mesentery tissue of littermate P6 Foxc1^+/+ controls and Foxc1^c2/+ and Foxc1^c2/c2 mutant mice (Figure 9f – k). Within lymphatic valves of mesentery collecting vessels, nuclear FOXC1 expression was strongly reduced in Foxc1^c2/+ mutants compared to Foxc1^+/+ controls. Furthermore, Foxc1^c2/c2 mutants were absent of nuclear FOXC1 expression within LECs, although a non-specific, positive signal was detected in the membrane of valve leaflets of both Foxc1^c2/+ and Foxc1^c2/c2 mutant mice (Figure 9f – h). Additionally, co-immunostaining of FOXC1, FOXC2, and VE-Cadherin demonstrated loss of FOXC1 expression in the blood vasculature and smooth muscle cells associated with mesenteric arteries and subsequent gain of FOXC2 expression in Foxc1^c2/+ and Foxc1^c2/c2 mutants with no obvious changes observed in the blood vasculature morphology (Figure 9i – k). To investigate potential differences in valve morphogenesis, we also performed immunostaining of mesentery tissue for PROX1 and CD31 (Figure 9l – n). Quantification of PROX1-high valves revealed that total lymphatic valve number was not significantly changed, nor was valve maturation (Figure 9o and p). Collectively, these data demonstrate that FOXC2 is able to functionally substitute for FOXC1 transcription activity during lymphatic valve development and maturation.

Foxc1^fl/fl, Foxc2^fl/fl, Foxc1^fl/fl; Foxc2^fl/fl, Cdh5-Cre^ERT2, and Prox1-Cre^ERT2 mice were described previously (Sasman et al., 2012; Sörensen et al., 2009; Srinivasan et al., 2007). Tamoxifen dissolved in corn oil was orally administered (75 or 100 μg per neonate) from P1 to P5 using techniques as previously described (Butchbach et al., 2007) with slight modifications. In some experiments, mice were injected subcutaneously with DPBS (10 μL/g) or Y-27632 (Fisher, #12-541-0, 15 μg/g) 1 to 2 hr after administration of Tamoxifen from P2 to P5.

Generation of Foxc1^c2/c2 mice is summarized in Figure 9. Using overlapping Foxc1 genomic clones isolated from a 129/SvJ mouse genomic library (Kume et al., 1998), the targeting vector consists of a 4.5 kb 5’ homology region and 4.1 kb 3’ homology region. The Foxc1 coding region was replaced with the cDNA coding (from the start codon to the stop codon) for Foxc2 by using restriction enzyme digestion and PCR cloning. To facilitate the elimination of the selectable marker gene Neo^r in mice, we introduced the ACN cassette flanked by two flox sites (Bunting et al., 1999) in the targeting vector, which was inserted in the 3’ untranslated region of Foxc1. The ACN cassette codes for Cre driven by the testis-specific promoter of the angiotensin-converting enzyme gene and for Neo^r driven by RNA polymerase II; thus the entire cassette is deleted as it passes through the male germ line of mice (Bunting et al., 1999). For negative selection, a PGK-TK cassette was placed at the end of the 3’ homology region. The resultant targeting vector was confirmed by sequencing.

SalI-linearized targeting vector (100 μg) was then electroporated into TL1 ES cells (129S6) as described (Kume et al., 1998). The cells were selected with G418 and gancyclovir. Double-resistant ES cell colonies were screened by Southern blot using 5’ and 3’ probes. One targeted clone (7F11) was injected into host (C57BL/6) blastocysts and produced germline chimeras. Targeted ES cells were screened by Southern blot using 5’ and 3’ probes following BamHI digestion (Figure 9b). Chimeras were mated with Black Swiss (Taconic) females and maintained by interbreeding on a mixed (129 x BlackSwiss) genetic background. After germline transmission, F1 heterozygous knock-in mutant (Foxc1^c2/+) mice (Figure 9c) were then intercrossed to generate Foxc1^c2/c2 mice. Offspring were genotyped either by Southern blot using the 3’ probe or PCR using two specific primers: 5’-CAGCGTTTTCTGCAAAACATA-3’ and 5’-GGACACATAGGCTGATCTCCA-3’. Weaned mice for continued animal colony maintenance from several breeding pairs were generated at near Mendelian ratios for Foxc1^+/+ (77/284, 27.113%), Foxc1^c2/+ (125/284, 44.014%), and Foxc1^c2/c2 (82/284, 28.873%) individuals.

Genotyping of mice for use in analysis was performed by Transnetyx Inc (Cordova, TN) using real-time PCR.

Dissection, immunostaining, and imaging of the lymphatic vasculature in mesentery tissue from 4 week old C57Bl6 mice for whole-mounts was performed as previously described (Sabine et al., 2018). Briefly, dissected mesentery tissue was fixed with 4% PFA, washed with PBS, followed by incubation with 10–20% sucrose solutions, then washed again in PBS before incubation with blocking buffer containing 0.5% BSA, 5% donkey serum, and 0.5% Triton X-100 in PBS, immunostained with primary antibodies at room temperature for 24 hr, washed with 0.5% Triton X-100 for 24 hr, followed with incubation with secondary antibodies for another 24 hr, washed again, rinsed with PBS and cleared for 2 days in 88% Histodenz (Sigma, D2158) in RIMS buffer, and eventually mounted in fresh Histodenz solution.

The lymphatic collecting vessel vasculature of neonates was analyzed by whole mount immunostaining of mesentery tissue harvested from pups at the indicated time. Briefly, mesentery tissue was dissected from the intestinal tract, laid out in plastic dish and left until it was firmly attached. Following fixation with 2% PFA in PBS, tissues were washed with PBS, then permeabilized and blocked in PBS solution containing 0.5% BSA, 5% serum, 0.3% Triton X-100, and 0.1% Sodium Azide. Tissues permeabilized/incubated with blocking buffer were then incubated with primary antibodies listed in Table 1 overnight followed by washes prior to incubation with Alexa Fluor conjugated secondary antibodies and dyes listed in Table 1 overnight. After subsequent washes, samples were post-fixed in 4% PFA, washed in PBS, and flat mounted on slides in mounting medium. Whole-mount staining images are shown with the same orientation, i.e., with flow in the direction of bottom-to-top.

Imaging was performed using a Zeiss AxioVision fluorescence microscope and Zeiss Axiovision software, Zeiss LSM-510 Meta, LSM-800 and LSM 880 confocal microscopes and Zeiss Zen Blue acquisition software, or using a Nikon A1 Confocal Laser Microscope System NIS-Elements software. Images were processed with Imaris and Adobe Photoshop software. Imaris colocalization function was used to produce pictures showing p-MLC2 (Figure 4b), vinculin (Figure 4c) or paxillin (Figure 4—figure supplement 1b) staining that is associated with F-actin staining.

Hearts from neonatal mice were digested in collagenase Type I solution (2 mg/mL) for 40 min at 37°C with gentle agitation. Cells were then filtered through a 70 μm cell strainer and the pellet was resuspended in Buffer 1 (PBS, 0.1% BSA, 2 mM EDTA, pH 7.4). The cell suspension was then incubated with magnetic Dynabeads (Invitrogen) pre-coated with CD31 antibody to isolate the endothelial cell population. After several washes with Buffer 1, RNA was extracted from endothelial cells using RNA STAT solution (Tel-Test) followed by phenol-chloroform treatment. Extracted RNA was subjected to DNAse I treatment and concentration was determined using a NanoDrop machine (Thermo Scientific). cDNA was synthesized using an iScript reverse transcriptase kit (Bio-Rad) according to the manufacturer’s instructions. Triplicates of cDNA samples for qPCR analysis were performed using a Fast qPCR machine (Applied Biosystems), Fast SYBR reaction mix (Applied Biosystems), and gene specific primer sets. Peptidylprolyl isomerase A (Ppia) or 18S was used as an internal standard for mRNA expression. Primer sequences are provided in Table 2.

Culture of human intestinal LECs was performed as described previously (Sabine et al., 2015). LECs were fixed using 4% PFA in PBS, then permeabilized with 0.1% Triton X-100 followed by blocking with 5% donkey serum and incubation with primary and secondary antibodies listed in Table 1. Knockdown experiments were performed by transfecting cells with 40 nM siRNA using Lipofectamine RNAiMAX (Invitrogen). siRNAs are listed in Table 3. Knockdown efficiency was confirmed by immunostaining (Figure 3—figure supplement 1d,e). LECs were analyzed after 48 hr, or used 24 hr post-transfection for flow experiments. For rescue experiments, transfected LECs were treated with vehicle control or 10 µM Y-27632 (StemCell, 72304) diluted in PBS every 12 hr with analysis at 48 hr. No cell lines from commonly misidentified cell lines were used. Primary lymphatic endothelial cells were isolated and cultured as described in Norrmén et al., 2009. Cell identity was confirmed by staining for lymphatic endothelial cell marker PROX1. Cells are isolated and cultured in the presence of 30 µg/ml Gentamicin and 15 ng/ml Amphotericin B, cultured for maximally 10–12 passages and tested negative for mycoplasma contamination by DNA staining.

Flow experiments with cultured LECs were performed as described previously (Sabine et al., 2015). Briefly, LECs were seeded at confluence on slides (μ-Slide I^0.8 Luer; Ibidi) coated with 40 µg/ml human fibronectin, cultured for 24 hr and then subjected to LSS (4 dyn/cm²), OSS (4 dyn/cm² and flow direction change every 4 s) using Ibidi Pump system, or kept in static conditions for an additional 24 hr prior to fixation, immunostaining, and mounting using Ibidi Mounting Medium (ibidi GmbH).

Putative FOX-binding sites were determined first by using the Hypergeometric Optimization of Motif EnRichment (HOMER) (Heinz et al., 2010) suite of tools to scan the entire Genome Reference Consortium Human Build 37 (GRCh37 or hg19) genome corresponding to the conserved RYMAAYA FOX transcription factor binding motif. The output file was then uploaded to the UCSC genome browser (Kent et al., 2002) to identify putative binding sites corresponding to transcriptionally active areas as indicated by histone modification, DNAse sensitivity, and additional transcription factor chromatin immunoprecipitation data as per work reported and summarized on the Encyclopedia of DNA Elements (ENCODE; https://genome.ucsc.edu/ENCODE/)(2012). Putative sites in the human genome were then searched against the mm10 mouse genome using the ECR Browser (https://ecrbrowser.dcode.org) (Ovcharenko et al., 2004) and rVista 2.0 tools to identify conserved and aligned putative binding sites between mouse and human sequences.

Human dermal lymphatic endothelial Cells (HDLECs) from juvenile foreskin (Promo Cell, #C12216) (https://www.promocell.com/product/human-dermal-lymphatic-endothelial-cells-hdlec/) were cultured and used according to the manufacturer's protocol. The cells were cross-linked with 1% formaldehyde, followed by sonication. The sheared chromatin was immunoprecipitated with dynabeads (Invitrogen, #10004D) conjugated with anti-FOXC2 antibody (Abcam, ab5060), anti-FOXC1 antibodies (Origene, TA302875 and Abcam, ab5079), or control IgG (Thermo Fisher Scientific, # 02–6202). DNA extraction and PCR were performed as previously described (Fatima et al., 2016) with primers listed in Table 4 targeting identified evolutionary conserved regions (ECRs) containing putative binding site sequences shown underlined listed below. Images were acquired with a ChemiDoc Touch Imaging System (Bio-Rad) and band intensities were analyzed with Image Lab software.

To identify lymphatic valves, mesentery tissue was stained with PROX1 antibody and areas of high expression were quantified as mature (visible leaflets) or immature (no visible leaflets) on four lymphatic collecting vessels per individual. Total number of valves were determined and percent of mature valves was normalized to the total counted per individual. For assessment of apoptosis in lymphatic collecting vessels, mesentery tissue was immunostained with PROX1 and active caspase-3 antibody and the percentage of PROX1/caspase 3-positive LECs was quantified from 20X high-power fields generated from confocal z-stacks using Fiji software to determine the total number of LECs per field using thresholded PROX1 immunostaining. Quantification was completed from three biological replicates.

For in vitro quantifications, relative nuclear intensity levels were measured using Fiji software. Nuclei were considered as regions of interest from thresholded Hoechst staining pictures. FOXC1 or FOXC2 intensity was then measured in each nuclei using RawIntDen function. For quantification of F-actin and p-MLC2 area per cell, LECs were fist manually segmented and defined as individual regions of interest, then F-actin (or p-MLC2) staining was thresholded using similar parameters for all pictures and F-actin⁺ (or p-MLC2⁺) area was measured per each region of interest. Quantification was completed from three independent experiments.

Statistical analysis for in vivo experiments was performed using GraphPad Prism v5 or v8. P values were obtained by performing a 2-tailed Student’s t test or one-way ANOVA and Tukey’s test. Data are presented as mean ± standard deviation of representative experiments from at least three biological replicates. For quantification of apoptosis in LECs of compound Foxc1; Foxc2 mutant mice, the ROUT method with Q set to 1% was used to identify outliers in the data set that were then excluded from statistical analysis using student’s t-test. P values less than 0.05 were considered statistically significant. Statistical analysis for in vitro experiments was performed using GraphPad Prism v8 to perform mixed-effects analysis and calculation of correlation coefficients. P values less than 0.05 were considered statistically significant for mixed-effects analysis.

All procedures and animal studies were approved by Northwestern University’s IACUC or by the Animal Ethics Committee of Vaud, Switzerland.