Endothelial cell SMAD6 balances Alk1 function to regulate adherens junctions and hepatic vascular development

ABSTRACT BMP signaling is crucial to blood vessel formation and function, but how pathway components regulate vascular development is not well-understood. Here, we find that inhibitory SMAD6 functions in endothelial cells to negatively regulate ALK1-mediated responses, and it is required to prevent vessel dysmorphogenesis and hemorrhage in the embryonic liver vasculature. Reduced Alk1 gene dosage rescued embryonic hepatic hemorrhage and microvascular capillarization induced by Smad6 deletion in endothelial cells in vivo. At the cellular level, co-depletion of Smad6 and Alk1 rescued the destabilized junctions and impaired barrier function of endothelial cells depleted for SMAD6 alone. Mechanistically, blockade of actomyosin contractility or increased PI3K signaling rescued endothelial junction defects induced by SMAD6 loss. Thus, SMAD6 normally modulates ALK1 function in endothelial cells to regulate PI3K signaling and contractility, and SMAD6 loss increases signaling through ALK1 that disrupts endothelial cell junctions. ALK1 loss-of-function also disrupts vascular development and function, indicating that balanced ALK1 signaling is crucial for proper vascular development and identifying ALK1 as a ‘Goldilocks’ pathway in vascular biology that requires a certain signaling amplitude, regulated by SMAD6, to function properly.


INTRODUCTION
Blood vessel formation involves the expansion of a primitive vascular network of endothelial cells into different organs and tissues during embryonic life (Carmeliet, 2000). As vessels remodel and mature under the influence of environmental signals such as blood flow and tissue-specific signaling, larger arteries carry blood away from the heart and veins return blood to the heart. Extensive capillary beds form between arteries and veins, and these capillaries acquire organ-specific properties that support tissue metabolism and function (Augustin and Koh, 2017;Aird, 2007;Rafii et al., 2016).
Among organ-specific developmental programs, the fetal liver is unique in that it receives oxygenated blood from the placenta via the portal vein rather than arteries.
Blood flows through the liver parenchyma via capillaries that over time specialize into sinusoids comprised of liver sinusoidal endothelial cells (LSEC), then returns to the heart via the central vein (Swartley et al., 2016;Lammert et al., 2003). As LSEC differentiate they down-regulate some capillary markers and up-regulate markers not normally expressed by blood capillaries, forming a uniquely discontinuous basement membrane with fenestrations that permits filtration of blood components (Koch et al., 2021;Poisson et al., 2017). In mice, LSEC maturation starts embryonically around E12 and increases through post-natal stages (Matsumoto et al., 2001;Gómez-Salinero et al., 2022). Thus, development of the fetal liver vasculature is unique in several ways but remains poorly understood.
Among numerous regulators of vascular development, the BMP (bone morphogenetic protein) signaling pathway is essential for proper blood vessel formation

Smad6 Functions in Endothelial Cells during Embryogenesis
To begin detailed investigations of Smad6 function during mammalian development, we first generated mice carrying the Smad6-lacZ knock-in allele (Smad6 tm1Glvn ) on the C57BL6J background (N>10 backcross generations). Mice homozygous for this allele are hereafter referred to as Smad6 -/global mutant mice.
. CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 We next asked whether Smad6 function is required cell-autonomously in vascular endothelial cells during mammalian development. A Smad6 floxed allele was generated by inserting LoxP sites around exon 4, which encodes a MH2 protein domain required for function (Supp. Fig. 1B). Mice carrying this allele were bred to an endothelial cell-specific tamoxifen-inducible line, Tg(Cdh5-cre/ERT2)1Rha (hereafter referred to as Cdh5-Cre ERT2 ), and excision was induced at E10.5 via tamoxifen oral gavage (Supp. Fig. 1C). We attempted to analyze pups at birth, but they required harvest via C-section just prior to birth due to maternal distress. None of the 4 pups genotyped as Smad6 fl/fl ;Cdh5-Cre ERT2 (Smad6 iΔEC/iΔEC ) were viable at E20.5 harvest, while 7 non-mutant pups were alive at this time (Supp. Fig. 1D); these findings indicate that Smad6 iΔEC/iΔEC embryos did not survive to birth, similar to Smad6 -/global mutant embryos. Phenotype scoring of whole Smad6 iΔEC/iΔEC mutant embryos at E16.5 recapitulated Smad6 -/global mutant vascular phenotypes, with significant abdominal and jugular hemorrhage compared to littermate controls (Fig. 1B-E), and sporadic edema, paleness, and blood-filled lymphatics (data not shown). These findings indicate that the primary vascular phenotypes associated with Smad6 global loss are largely endothelial cell-specific and show that Smad6 function is required in endothelial cells for vascular integrity in vivo.
The prevalence of abdominal hemorrhage suggested liver involvement, and examination of isolated E16.5 livers revealed significant vessel dilation and hemorrhage in both Smad6 -/global and Smad6 iΔEC/iΔEC livers, with pale regions not seen in controls ( Fig. 1F-G). The phenotype scoring of isolated livers (Supp. Fig2B) was similarly . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 penetrant between the classes of mutant embryos, suggesting that whole embryo abdominal scoring was less sensitive and that Smad6 functions in embryonic liver endothelial cells.
We hypothesized that the residual reduced penetrance of vascular phenotypes in Smad6 iΔEC/iΔEC embryos compared to Smad6 -/global mutants might reflect later or less complete removal of Smad6 from endothelial cells. We used a tamoxifen-inducible global Cre line, UBC-Cre ERT2 , to generate Smad6 iΔ/iΔ E16.5 embryos using the same excision protocol as for Smad6 Δ ECiΔEC embryos and found that the severity of vascular phenotypes mirrored that of the Smad6 -/global mutants (Supp. Fig. 3A-G). Thus, Smad6 deletion mid-gestation did not lead to significant reduced phenotype penetrance, and the Smad6 -/and Smad6 iΔ/iΔ lines were used interchangeably in subsequent experiments. For Smad6 iΔEC/iΔEC embryos, gene excision via PCR analysis of lung lysates at E16.5 revealed a predicted excision band not seen in controls (Supp. Fig.   3H-I). We next analyzed Smad6 liver expression in E16.5 Smad6 iΔEC/iΔEC embryos and found that isolated PECAM1-enriched cell populations had significantly reduced levels of Smad6 RNA that correlated with phenotype severity (Supp Fig 3 J-K), indicating that embryo-dependent partially inefficient excision may be responsible for residual reduced penetrance of the liver vascular phenotype in Smad6 iΔEC/iΔEC embryos.

Smad6 is Expressed in Embryonic Liver Endothelial Cells
The hepatic vascular defects seen in Smad6 iΔEC/iΔEC embryos were interesting, as the embryonic liver vasculature consists of veins and capillaries at this stage, and . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 FINAL 032323 1 0 hepatic artery formation is only detectable just before birth (Swartley et al., 2016). In contrast, robust embryonic Smad6 expression via the lacZ reporter readout was documented in larger arteries and the outflow tract that did not have obvious defects (data not shown) (Wylie et al., 2018;Galvin et al., 2000). To more rigorously examine vascular Smad6 expression, we reanalyzed several published single-cell (sc) RNA seq datasets. scRNA seq data from an EC Atlas of adult mouse tissues (Kalucka et al., 2020) revealed Smad6 expression in endothelial cells of several organs (Supp. Fig 4A-A'); Smad6 expression was substantial in vein and capillary endothelial cells along with arterial expression in liver and lung, while expression was more localized to arterial endothelial cells in the brain (Supp. Fig 4B-D

, B'-D').
Re-analysis of a second dataset from adult mouse brain and lung endothelial cells (Vanlandewijck et al., 2018;He et al., 2018) revealed Smad6 RNA expression in venous and capillary endothelial cells, albeit at a lower prevalence than in arterial cells (Supp. Fig 4E-F). Additionally, Gomez-Salinero et al. re-analyzed the Tabula Muris database for highly expressed liver genes and identified Smad6 expression as enriched in liver endothelial cells (Gómez-Salinero et al., 2022). Taken together, these data indicate that Smad6 is expressed in endothelial cells from all caliber adult mammalian blood vessels, including veins and capillaries.
To assess embryonic Smad6 endothelial cell expression, we re-analyzed the Mouse Organogenesis Cell Atlas (MOCA) dataset (Cao et al., 2019) by extracting endothelial cell information by embryonic stage (E9.5-13.5) and organ ( Fig. 2A-A'). This analysis revealed substantial Smad6 expression in embryonic (E9.5-13.5) arterial, endocardial, and liver endothelial cells ( Fig. 2A'', B). Next, a recently published scRNA . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101/2023 doi: bioRxiv preprint 1 1 seq dataset of mouse liver endothelial cells from E12-P30 (Gómez-Salinero et al., 2022) was re-analyzed with a focus on embryonic stages, revealing Smad6 expression in most liver embryonic endothelial cell clusters from E12-E18, including expression in the larger portal and central veins and fetal sinusoidal endothelial cells (FS1-FS5) (Fig. 2C-C', D). Finally, we assessed Smad6 expression in E16.5 embryonic livers via the lacZ reporter and documented lacZ expression in surface vessels and in endothelial cells lining hepatic vessels (Fig. 2E-F), consistent with the scRNA seq data and with the conclusion that Smad6 is expressed in veins and capillaries of the embryonic liver.

Smad6 Exhibits Epistasis with Alk1 in Embryonic Liver Blood Vessels
Having established that endothelial cell Smad6 function regulates vessel integrity in the embryonic liver, we next asked how SMAD6 exerts its effects, and we hypothesized that some aspect of BMP function is normally regulated by SMAD6. We turned our attention to the BMP Type I receptor Alk1, as signaling through complexes containing this receptor is important for vascular integrity and flow-mediated responses (Baeyens et al., 2016;Tual-Chalot et al., 2014). We hypothesized that SMAD6 antagonizes ALK1 signaling and predicted that genetic reduction of Alk1 would rescue the loss of vascular integrity seen with Smad6 loss. Global Alk1 deletion is lethal at midgestation with impaired vascular development and vessel dilation (Oh et al., 2000), and homozygous endothelial cell deletion of Alk1 is lethal within several days of excision in neonates due to AVMs and pulmonary hemorrhage (Tual-Chalot et al., 2014;Park et al., 2008b). We confirmed that endothelial-specific deletion of Alk1 starting at E10.5 was embryonic lethal at E16.5 and likely earlier, since the mutant embryos were partially . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023

FINAL 032323
1 2 resorbed at this time point (Supp. Fig 5A). We showed that concomitant deletion of Smad6 did not alter Alk1-dependent lethality with this excision protocol (Supp. Fig 5B), so we tested genetic epistasis in embryos with one Alk1 allele and both Smad6 alleles deleted in endothelial cells. Whole embryo examination revealed a trend for increased rescue of total hemorrhage in Smad6 iΔEC/iΔEC ;Alk +/iΔEC embryos relative to Smad6 iΔEC/iΔEC embryos ( Fig. 3A-B), while isolated liver analysis showed highly significant rescue of liver vascular defects in embryos with reduced Alk1 gene dosage (Fig. 3C-D). None of the Smad6 iΔEC/iΔEC ;Alk +/iΔEC livers presented with a severe vascular phenotype, and 77% of embryonic livers with reduced Alk1 gene dosage had no discernable vascular defect.
These results indicate that Smad6 and Alk1 have an epistatic relationship in embryonic liver endothelial cells and suggest that SMAD6 normally restricts Alk1 activity.

Smad6 Maintains Vessel Integrity in the Embryonic Liver via Alk1 Regulation
To further understand Smad6 function and epistasis with Alk1 in embryonic liver vessels, we performed histological and marker analysis on embryonic liver sections. In general, microscopic phenotypes were shared by livers from Smad6 -/global mutant embryos, Smad6 iΔ/iΔ mutant embryos, and Smad6 iΔEC/iΔEC mutant embryos, suggesting that effects of Smad6 loss on liver development result from endothelial cell-selective functions of Smad6. H&E stained E16.5 liver sections revealed areas of hemorrhage and loss of tissue organization in both Smad6 -/-And Smad6 iΔEC/iΔEC mutants not seen in controls (Fig. 4A), amongst areas that appeared relatively normal. This mosaicism was not observed in either the Alk1 +/iΔEC or Smad6 iΔEC/iΔEC ;Alk +/iΔEC liver sections that appeared similar to controls, indicating that reduced gene dosage of Alk1 rescued the regions of hemorrhage and disorganization. The areas of hepatic vascular loss . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101/2023 doi: bioRxiv preprint 1 3 appeared skewed towards peripheral areas of the liver lobes, and staining for the vascular marker PECAM1 (CD31) and the red blood cell marker Ter119 revealed large and dilated peripheral vessels in Smad6 -/-And Smad6 iΔEC/iΔEC mutant livers (Fig. 4B), consistent with the whole liver phenotype.
Staining of adjacent sections for Lyve1 (marker of liver sinusoidal endothelial cells (LSEC)), Vegfr3 (early endothelial cell marker) or PECAM1 (pan-endothelial cell marker) revealed co-incident expression or loss of expression in liver capillaries of Smad6 mutant embryos, indicating avascular areas within the liver parenchyma; these avascular areas were not seen in livers of Smad6 iΔEC/iΔEC ;Alk +/iΔEC embryos with reduced Alk1 gene dosage (Fig. 4C). Quantification revealed that Lyve1+ area was significantly reduced in both Smad6 -/global mutant and Smad6 iΔEC/iΔEC livers, but not in Smad6 iΔEC/iΔEC ;Alk +/iΔEC livers with reduced Alk1 dosage (Fig. 5A-B), confirming that peripheral mosaic avascularity in the liver parenchyma is a hallmark of endothelial Smad6 loss and that reduced Alk1 gene dosage rescues this phenotype.
Examination of avascular areas in the mutant livers revealed significant levels of cleaved caspase3 in regions that border vascularized areas, indicative of parenchymal cell death that was not seen in controls or Smad6 iΔEC/iΔEC ;Alk +/iΔEC mutant livers with reduced Alk1 gene dosage (Fig 5C-D). Thus, endothelial loss of Smad6 function leads to mosaic loss of capillary vessels in the embryonic liver parenchyma accompanied by cell death that is dependent on Alk1 gene dosage, indicating that Smad6 restriction of Alk1 signaling normally regulates vascular integrity during liver embryogenesis.
. CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

Smad6 Promotes Liver Sinusoidal Endothelial Cell Differentiation
Liver sinusoidal endothelial cells (LSEC) are specialized liver capillary vessels with a discontinuous basement membrane and fenestrations. During mid-gestation LSEC begin to form and express markers of classic blood endothelial cells, but over time they acquire LSEC-specific markers and develop a unique phenotypic and functional profile (Geraud et al., 2017). Classic endothelial cell markers such as PECAM1 and CD34 are down-regulated while LSEC markers such as Lyve1 and Stabilin-2 are upregulated (Sugiyama et al., 2010;Poisson et al., 2017;Schledzewski et al., 2011). Reversal of LSEC differentiation leads to loss of fenestration and basement membrane deposition, a process called capillarization, which leads to liver fibrosis and dysfunction in adult livers (Schaffner and Poper, 1963;Xie et al., 2012;DeLeve, 2015).
We asked whether the remaining capillary beds in Smad6 mutant livers exhibited abnormal LSEC differentiation. We found that both Smad6 iΔ/iΔ and Smad6 iΔEC/iΔEC mutant E16.5 livers had excessive Collagen IV deposition in capillary beds that was not seen in controls or in mutant livers with reduced Alk1 gene dosage, accompanied by ectopic staining for the smooth muscle marker αSMA (Fig 6A-B). This excess of basement membrane protein indicates that LSEC of Smad6 mutant livers are more capillarized. Further analysis revealed that the diameter of capillary vessels was significantly increased in both the Smad6 iΔ/iΔ and Smad6 iΔEC/iΔEC E16.5 mutant livers, and this phenotype was partially rescued by reduced gene dosage of Alk1 (Fig. 6C).
To better define the differentiation status of LSEC in embryonic mutant livers lacking endothelial Smad6 function, we enriched for PECAM1 + endothelial cells from . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made RNA levels of the LSEC maturation markers, Stabilin-2 and Lyve1, were significantly decreased in Smad6 iΔEC/iΔEC liver endothelial cells compared to controls (Fig. 6D, E), consistent with the idea that liver vessels lacking Smad6 are more capillarized than stage-matched controls. Expression of Gata4, a transcription factor that regulates LSEC differentiation (Geraud et al., 2017), trended down in Smad6 iΔEC/iΔEC liver endothelial cells although it did not reach significance (Fig. 6F). The PECAM-negative fraction of dissociated livers showed no significant changes in vascular markers (Supp. Fig 6A-C).

Flow Responses
SMAD6 stabilizes endothelial adherens junctions, maintains vascular barrier function, and is required for flow-mediated endothelial cell alignment (Wylie et al., 2018;Ruter et al., 2021). To define the cellular mechanism of SMAD6 function and further explore its relationship to ALK1 signaling in endothelial cells, we depleted RNA levels in primary human endothelial cells (HUVEC) using siRNA knockdown. Absent flow, we confirmed that endothelial cells depleted for Smad6 have adherens junctions that appear destabilized compared to controls. In contrast, endothelial cells depleted for Alk1 have linear junctions that appear stable, and concurrent depletion of Smad6 and Alk1 rescues the destabilized junction morphology seen with Smad6 depletion (Fig. 7A,   insets). Under laminar flow conditions these relationships remained, and concurrent depletion of Smad6 and Alk1 rescued the mis-alignment seen with Smad6 depletion alone ( Fig. 7A-B). We confirmed that endothelial cell junction de-stabilization was . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made phenocopied by addition of BMP9, and junction morphology was normalized with Alk1 depletion with or without Smad6 depletion upon BMP9 exposure (Supp. Fig. 7), indicating that the effects of Alk1 on junctions are downstream of BMP9. We next functionally assessed monolayer integrity using an adapted protocol (Dubrovskyi et al., 2013) that reveals biotin-labeled matrix accessible to streptavidin, and found that labeling was significantly increased over controls in endothelial cells depleted for Smad6 under both static and flow conditions, and rescued to control levels with concurrent depletion of Smad6 and Alk1 (Fig. 7C-E). We further analyzed functional effects of Smad6 and Alk1 on endothelial junctions by measuring electrical resistance across monolayers using Real Time Cell Analysis (RTCA). We confirmed that cells depleted for Smad6 had reduced electrical resistance compared to controls, and this disruption was rescued by concurrent Alk1 depletion (Fig. 7F). These findings are consistent with the idea that SMAD6 regulates endothelial junctions and manages flow responses via negative modulation of BMP9/ALK1-dependent signaling.

SMAD6 Regulates Endothelial Cell Contractility and PI3K Signaling via ALK1
The destabilized junction morphology of EC depleted for Smad6 was reminiscent of hypercontractility, so we hypothesized that SMAD6 regulates endothelial cell contractility. The contractility agonist thrombin induced junction destabilization that phenocopied the junction morphology induced by Smad6 depletion in endothelial cells, while contractility blockade via blebbistatin led to a more linear junction morphology in Smad6 silenced endothelial cells, indicating that Smad6 depletion regulates endothelial cell contractility (Fig. 8A). Alk1 depletion blunted thrombin-induced junction . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ; https://doi.org/10.1101/2023.03.23.534007 doi: bioRxiv preprint 1 7 destabilization independent of Smad6 depletion. Functionally, thrombin treatment significantly increased biotin matrix labeling in all conditions over similarly depleted controls, while contractility blockade rescued the increased matrix labeling seen with Smad6 depletion (Fig. 8B-C). These results indicate that SMAD6 is required to modulate and prevent endothelial cell hypercontractility, and that this effect goes through ALK1 signaling. ALK1 activation via BMP9 inhibits PI3K (PI3 kinase) signaling in endothelial cells (Ola et al., 2016;Alsina-Sanchís et al., 2018;Ola et al., 2018), consistent with our finding that ALK1 regulates endothelial cell contractility. We thus hypothesized that SMAD6 acts to negatively modulate ALK1 activity and maintain appropriate levels of PI3K signaling. Endothelial cell exposure to the PI3K agonist 740Y-P rescued both junction morphology and biotin matrix labeling in Smad6 depleted cells (Fig. 8D-F).
Conversely, inhibition of PI3K signaling via wortmannin treatment induced destabilized junction morphology and increased biotin matrix labeling in control endothelial cells to similar levels as Smad6 depletion, and Alk1 depletion blunted endothelial cell responses to wortmannin (Fig. 8D-F). Thus, our results show that endothelial cell SMAD6 maintains a balance of PI3K signaling through negative modulation of ALK1 to regulate endothelial cell contractility and vessel integrity.
. CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

DISCUSSION
Our findings reveal that SMAD6, a negative regulator of BMP signaling, is required developmentally in endothelial cells for proper blood vessel integrity. Loss of endothelial Smad6 leads to hemorrhage and dilation of veins and capillaries of the embryonic liver, and we identify for the first time ALK1 signaling as an important negative target of Smad6 function in vivo. Mechanistically, Smad6 modulates Alk1 activity to balance endothelial cell contractility that is activated by Alk1, and PI3K signaling that is normally repressed by Alk1. Thus, SMAD6 functions to maintain a balance of ALK1 signaling that in turn sets PI3K signaling levels and contractility in endothelial cells and developing blood vessels; this balance is required for vessel integrity and function ( Figure 9) and identifies vascular ALK1 signaling as a finely tuned pathway regulated by SMAD6.
Despite robust SMAD6 expression in larger arteries during development (Galvin et al., 2000;Wylie et al., 2018), we found that vascular phenotypes resulting from global and endothelial-selective Smad6 deletion were predominant in vessels characterized as veins or capillaries during embryogenesis. Our focused analysis of the embryonic liver, which receives oxygenated blood from the placenta via the umbilical vein (Swartley et al., 2016;Ober and Lemaigre, 2018), revealed significant vessel dilation and hemorrhage that likely contributed to embryonic lethality. Transcriptional profiling showed that most liver endothelial cell clusters had significant Smad6-expressing cells at both embryonic and post-natal stages (Gómez-Salinero et al., 2022). Additionally, Smad6 was identified among the top-enriched genes in liver endothelial cells from the . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Genetic loss of Smad6 in embryonic endothelial cells resulted in a mosaic pattern of vascular loss in the in the liver parenchyma, as areas completely devoid of capillaries were juxtaposed with areas containing capillaries. As predicted, areas of mutant livers lacking capillaries appeared pale and harbored cells with more pyknotic nuclei and elevated apoptosis levels, while vascularized areas had significantly dilated capillaries that appeared intact and patent, although less mature along the LSEC differentiation lineage. Areas of hemorrhage were found within and near the avascular regions, suggesting that hemorrhage preceded capillary loss. Although the severity of the liver phenotype inversely correlated with overall Smad6 RNA levels in enriched endothelial cell populations from the same livers, the mosaic pattern of parenchymal vessel loss was also documented in embryos globally deleted for Smad6, indicating that the mosaicism likely results from some aspect of Smad6 function that is unique to liver vascular development.
The embryonic liver vascular phenotypes -hemorrhage, mosaic capillary loss and vessel dilation -also exhibited a striking distribution to the periphery of the liver lobes, with the central areas being relatively unaffected. This pattern may result from . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ; https://doi.org/10.1101/2023.03.23.534007 doi: bioRxiv preprint the developmental program of the embryonic liver, as liver sinusoids are more proliferative at the periphery (Swartley et al., 2016), and liver lobes are perfused in a peripheral to central wave developmentally (Lorenz et al., 2018). Thus, the requirement for Smad6 function correlates with endothelial proliferation and vascular perfusion in the embryonic liver, consistent with the role of SMAD6 in flow-mediated responses of endothelial cells (Ruter et al., 2021). The concept that blood flow and flow patterns influence the spatial distribution and mosaicism of the endothelial Smad6 loss phenotype is compelling. Based on computational modeling (Bernabeu et al., 2014;Mut et al., 2009;Balogh and Bagchi, 2019;Rani et al., 2006;Hewlin and Tindall, 2023) shear stress is predicted to be highest at the point of entry of afferent vessels into an organ, and in the embryonic liver this is the umbilical/portal circulation that delivers oxygenated blood from the placenta. Thus, this venous circulation can be considered The liver is the site of synthesis of BMP9, which is a primary secreted ligand for . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made the ALK1 arm of the BMP signaling pathway (Larrivée et al., 2012;David et al., 2007a;Miller et al., 2000), and BMP9 expression increases with developmental age through postnatal day15 (Bidart et al., 2012). This relationship led us to hypothesize that Smad6 genetically interacts with Alk1, and we confirmed that BMP9 exposure mimics Smad6  (Ruter et al., 2021). A major feature of the gain-of-function phenotype revealed by SMAD6 loss is hyper-contractility associated with reduced PI3K signaling, which is consistent with the effects of SMAD6 on endothelial junctions (Wylie et al., 2018). Taken together, these findings suggest that endothelial Alk1 signaling is an . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ; https://doi.org/10.1101/2023.03.23.534007 doi: bioRxiv preprint example of a "Goldilocks" pathway that requires a certain signaling amplitude to function properly, similar to the regulation described for cytokine signaling and neural circuits (Graham et al., 2022;Humphries, 2016;Petersen and Berg, 2016). Moreover, Smad6 is a prominent transcriptional target that is upregulated downstream of BMP9/Alk1 signaling, consistent with the idea that negative modulation via SMAD6 is important to counteract positive inputs and promote vessel integrity. Current therapies to mitigate symptoms of HHT2, a disease resulting from genetic loss of the Alk1 arm of BMP signaling, include PI3K blockade (Robert et al., 2020); our work suggests that modulating the Alk1 arm of BMP signaling may restore the required balance in the pathway.

Mice
All animal experiments were approved by the University of North Carolina at Chapel Hill (UNC-CH) Institutional Animal Care and Use Committee. All mice were on a C57BL/6J genetic background, and both male and female embryos were included.
Cre ERT2 negative littermates were used as controls. Embryos were collected at indicated timepoints into PBS on ice, euthanized according to IACUC approved methods, and fixed in 4% paraformaldehyde (PFA) at 4°C for 24-72 hr.
For DNA analysis, embryonic tail snips or lung tissue was incubated in 0.2mg/mL Proteinase K in DirectPCR Lysis Reagent (Viagen Biotech 101-T) at 55°C for 4 hr, followed by enzyme inactivation at 85°C for 45 min. Forward and reverse primers (Smad6 excised F+R) were designed to anneal upstream of the 5' loxP site and downstream of the 3' loxP site (Supp. Fig 1B). See Table 1 for primer details.  Table 1.

MACS Enrichment and qRT-PCR
Results were analyzed using delta-delta-CT methods, and CT values were normalized to GAPDH or B-actin and relative to the WT average.

Semi-Quantitative Phenotype Score
Intact embryos were examined and imaged. For liver phenotype scores, livers were dissected either pre-or post-fixation and imaged. A severity guideline key (Supp .   Fig 2) was created that ranked presentations of each phenotype (jugular hemorrhage, abdominal hemorrhage, liver hemorrhage/paleness). Embryo genotypes were blinded and images scored by a single researcher to maintain internal consistency.

LacZ Staining
LacZ staining was performed as described (Nagy et al., 2007) with modifications as follows: . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ; https://doi.org/10.1101/2023.03.23.534007 doi: bioRxiv preprint Wholemount: Briefly, E16.5 embryos in PBS on ice were dissected and the heart, lungs, liver, and intestines removed. All tissues were incubated in freshly made 0.2% glutaraldehyde (Electron Microscopy Sciences cat no. 16120) + 5mM EGTA + 2mM MgCl 2 in PBS on ice for 30 min, washed in wash buffer (2mM MgCl 2 + 0.02% IGEPAL + 0.01% sodium deoxycholate in 0.1M Sodium Phosphate Buffer pH 7.3) for 3x 15 min at RT, then incubated in freshly made stain solution (5mM potassium ferricyanide + 5mM potassium ferrocyanide + 1mg/mL X-gal (Promega cat no. V3941) in wash buffer) for 8 hr at 37°C with gentle rocking. Tissues were washed in PBS, then incubated in 4% PFA in PBS for 1hr at RT before imaging with a stereomicroscope or embedding in paraffin.
Frozen Sections: E16.5 embryos were euthanized, rinsed in PBS with Mg 2+ and separated above the liver. Embryos were fixed in cold 0.25% glutaraldehyde (Electron Microscopy Sciences cat no. 16120) in PBS, washed 3x 5 min in PBS, and sunk in 30% sucrose in PBS at 4°C for 12hr. The embryo pieces were embedded in OCT, frozen, sectioned at 10µm, and stored at -80°C. Before staining, sections were warmed at room temp for 20min, washed in PBS 5min, fixed in 0.25% glutaraldehyde in PBS 5min at room temp, washed 3x5min in PBS. Slides were incubated in freshly made stain solution (5mM potassium ferricyanide + 5mM potassium ferrocyanide + 1mg/mL X-gal (Promega cat no. V3941) in wash buffer) overnight at 37°C. Sections were post-fixed in 4% PFA for 1hr at RT, washed in PBS 3x5min, counterstained with Nuclear Fast Red (Sigma, N3030) 4 min, and mounted in 80% glycerol.
. CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made   Table 2) and incubated overnight at 4°C. Sections were washed in PBS, re-blocked 20 min at RT, then incubated in secondary antibodies, DAPI, and fluorescently-conjugated primary antibodies (see Table 2). Slides were rinsed in PBS, . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Following fixation in 4% PFA, HUVEC were washed with PBS, permeabilized in 0.1% Triton X-100 (Sigma T8787) at RT for 10 min and blocked at RT for 1 hr in blocking solution (5% NBCS, 2x antibiotic-antimycotic (Gibco), 0.1% sodium azide (Sigma s2002-100G). Cells were incubated in primary antibody (

Single Cell RNA Sequencing Analysis
Mouse Organogenesis Cell Atlas (Cao et al., 2019): Analysis was performed using the R package Seurat. The MOCA dataset contains over 2 million cells and the gene count matrix is over 20GB, so a randomly down-sampled dataset containing 10,000 cells was downloaded for QC check and EC annotation. Dimension reduction results from tdistributed stochastic neighbor embedding (t-SNE) and QC showed that cell clusters did not correlate with total detectable molecules/cell (nCount_RNA) or the number of detectable genes/cell (nFeature_RNA), suggesting that data from MOCA were analyzed properly. Next, cell type annotations in the dataset and expression patterns of panendothelial markers Cdh5 and Pecam1 were plotted. Only clusters annotated as EC and endocardial cells show high levels of Cdh5 and Pecam1, indicating that annotation . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ; https://doi.org/10.1101/2023.03.23.534007 doi: bioRxiv preprint in the data is correct. After confirming that QC was properly performed and EC annotation was correct in MOCA with this subset of data (data not shown), EC were extracted from the original gene count matrix and subjected to t-SNE visualization to show EC clusters labeled by developmental stage (Fig 2A) and inferred embryonic tissue origin (Fig2 A'). The expression of genes of interest was plotted by FeaturePlot ( Fig.2 A'') and VlnPlot. (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ; https://doi.org/10.1101/2023.03.23.534007 doi: bioRxiv preprint 1 TARGETplus Human ACVRL1 siRNA, Dharmacon L-005302-02-0005) using Lipofectamine 3000 (ThermoFisher L3000015) according to manufacturer directions.
HUVEC were transfected at 70% confluence for 24 hr at 37°C, then incubated with fresh EGM-2 for 24 hr. Cells were seeded onto glass chamber slides coated with 5µg/mL fibronectin (Sigma, F2006-2MG) for experiments.
Briefly, 0.1mg/mL fibronectin was incubated with 0.5mM EZ-Link Sulfo-NHS-LC-Biotin (ThermoFisher A39257) for 30 min at RT. Biotinylated fibronectin (0.5µg/mL) was coated onto glass chamber slides for 30 min at RT, then HUVEC were seeded at a density of 7.5x10 4 cells/mm 2 . Following drug treatments or flow experiments, confluent HUVEC were treated with 25µg/mL Streptavidin-488 (Invitrogen S11223) for 3 min at RT then immediately fixed in warm 4% PFA as described above. For quantification, at least three 40x confocal z-stack images/condition/experiment were taken. The . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 streptavidin channel was thresholded in ImageJ and the % labeled area measured, then normalized to the siNT control average for each respective experiment.

Real Time Cell Analysis (RTCA)
An xCELLigence Real-Time Cell Analyzer (RTCA, Acea Biosciences/Roche Applied Science) was used to assess barrier function of HUVEC monolayers. HUVEC were seeded at a density of 60,000cells/well of the E-plate (E-plate 16, Roche Applied Science), then electrical impedance readings acquired every 2 min for 24 hr. Results are reported at 24 hr as the percent change in cell index calculated using the following formula: (Cell Index siRNA -Cell Index NT )/ABS(Cell Index NT ).

Imaging and Analysis
Whole embryo and intact liver images were acquired using a Leica MZ 16 F stereomicroscope and an Olympus DP71 camera. H&E stains were scanned at 20x on an Olympus SLIDEVIEW VS200. Images of fluorescently stained tissue sections were acquired using an Olympus FV300 confocal microscope with Fluoview software or on an Olympus SLIDEVIEW VS200 with OlyVIA software. Images were processed in ImageJ or QuPath software and shown in figures as compressed Z stacks.
Vascularized liver area: Scans of whole liver sections were traced along the outer edge of the DAPI channel to measure total liver area. Traces were then made on the LYVE1+ channel around the vascularized zones. Apoptosis: Scans of whole liver sections were imported into QuPath, and the Cleaved Caspase 3 channel was . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023  In experiments with two groups, Student's two-tailed unpaired t-test was used to determine statistical significance. For % change in RTCA experiments a one sample ttest against a hypothetical mean of "0" was used. One-way ANOVA with Tukey's test to correct for multiple comparisons was used to compare differences between more than two groups. For thrombin, blebbistatin, 740Y-P, and wortmannin experiments, two-way ANOVA with Tukey's multiple comparisons test was used to determine statistical significance.
. CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Table 2. Antibodies *Antibodies validated using secondary-only controls, and specificity was confirmed by staining tissues/structures known to be positive.

ACKNOWLEDGEMENTS:
We thank the UNC Animal Models Core for assistance in generating the Smad6 floxed mouse line. We thank Wendy Salmon (Hooker Imaging Core, UNC) for microscopy support. The UNC Hooker Imaging Core Facility is supported in part by P30 CA016086 Cancer Center Core Support Grant to the UNC Lineberger Comprehensive Cancer Center. We thank Caroline Crater for mouse room support, and Bautch Lab members for critical discussion and feedback. We thank the Center for Gastrointestinal Biology . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made      Smad6 iΔ/iΔ , n=3; Smad6 iΔEC/iΔEC , n=6; Alk1 +/iΔEC , n=5; Smad6 iΔEC/iΔEC ;Alk +/iΔEC , n=6 livers. ***, P<0.001; ****, P<0.0001; ns, not significant. Data are individual points ±SD.
Statistics, one-way ANOVA with Tukey's test to correct for multiple comparisons.
. CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ; https://doi.org/10.1101/2023.03.23.534007 doi: bioRxiv preprint  HUVEC treated with non-targeting (NT), Smad6, and/or Alk1 siRNA were exposed to static conditions or laminar flow (7.5 dyn/cm 2 /72hr). (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023   . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023  Gain-or loss-of-function of Alk1 leads to vascular dysfunction but with distinct phenotypes, and reduced Alk1 gene dosage or depletion partially rescues the gain-offunction phenotypes induced by loss of the negative regulator Smad6. These findings identify ALK1 as a "Goldilocks" pathway in vascular development.
. CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023  . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023  (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  HUVEC treated with non-targeting (NT), Smad6, and/or Alk1 siRNA were cultured on fibronectin to confluence and treated as indicated with BMP9 ligand for 24 hr at 37°C. . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  Far right, areas of abnormal parenchyma in Smad6 mutant liver sections. Yellow outlined areas to left are magnified to right. Scale bar, 20µm. Arrow, hemorrhage; arrowhead, tissue disorganization. (B) Representative immunofluorescence images for PECAM1 (endothelial cells) and Ter119 (red blood cells) at periphery of E 16.5 livers of indicated genotypes. Scale bar, 50µm. (C) Representative images of immunofluorescence of E16.5 liver sections (adjacent sections for each liver) for indicated liver endothelial markers. Yellow dashed line, liver outline; blue dashed line, avascular areas. Scale bar, 500µm.
. CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  HUVEC treated with non-targeting (NT), Smad6, and/or Alk1 siRNA were exposed to static conditions or laminar flow (7.5 dyn/cm 2 /72hr). Data are mean ± SD (each data point = 1 field of view), with n ≥ 3 experimental replicates per condition. *, P<0.05; **, P<0.01; ****, P<0.0001; ns, not significant. Statistics, one-way ANOVA with Tukey's test to correct for multiple comparisons. (F) Quantification of % change in cell index for RTCA measured at 24hr. Normalized to NT siNRA control cell index. Data are mean ± SD (each data point an experimental replicate), with n = 3 replicates per condition. *, P<0.05; ns, not significant. Statistics, one sample t-test comparing conditions to hypothetical mean of 0.
. CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  Gain-or loss-of-function of Alk1 leads to vascular dysfunction but with distinct phenotypes, and reduced Alk1 gene dosage or depletion partially rescues the gain-of-function phenotypes induced by loss of the negative regulator Smad6. These findings identify ALK1 as a "Goldilocks" pathway in vascular development.
. CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 25, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023