In vivo dissection of Rhoa function in vascular development using zebrafish

Abstract

The small monomeric GTPase RHOA acts as a master regulator of signal transduction cascades by activating effectors of cellular signaling, including the Rho-associated protein kinases ROCK1/2. Previous in vitro cell culture studies suggest that RHOA can regulate many critical aspects of vascular endothelial cell (EC) biology, including focal adhesion, stress fiber formation, and angiogenesis. However, the specific in vivo roles of RHOA during vascular development and homeostasis are still not well understood. In this study we examine the in vivo functions of RHOA in regulating vascular development and integrity in zebrafish. We use zebrafish RHOA-ortholog (rhoaa) mutants, transgenic embryos expressing wild type, dominant-negative, or constitutively active forms of rhoaa in ECs, pharmacological inhibitors of RHOA and ROCK1/2, and Rock1 and Rock2a/b dgRNP-injected zebrafish embryos to study the in vivo consequences of RHOA gain- and loss-of-function in the vascular endothelium. Our findings document roles for RHOA in vascular integrity, developmental angiogenesis, and vascular morphogenesis in vivo, showing that either too much or too little RHOA activity leads to vascular dysfunction.

Introduction

An endothelial cell (EC) monolayer forms the innermost layer of all blood vessels. This layer acts as a physical barrier to maintain blood vessel integrity and selective permeability, regulating the exchange of oxygen, solutes, nutrients, and immune cells between the blood and tissues. Under normal physiologic conditions, the EC barrier is maintained throughout the development of both embryonic and adult vascular networks, which form predominantly by angiogenesis – the “sprouting” of new vessels from pre-existing ones. Within the brain, disruption of the EC barrier leads to inflammation, edema, or hemorrhagic stroke, producing a spectrum of disease that includes neurologic dysfunction, disability, cognitive impairment, or death [1]. The formation, maintenance, and disassembly of EC barrier junctions are therefore highly regulated processes under strict spatiotemporal control [2].

There are two principal types of barrier-forming EC junctions, tight junctions and adherens junctions [2,3]. Tight junctions are continuous inter-EC junctions that limit paracellular permeability by controlling the exchange of cells, fluids, and solutes between the vessel lumen and surrounding tissues. Adherens junctions serve as anchors that maintain stable cellular associations under mechanical stress, or dynamic adhesive connections that facilitate cellular movement (e.g., during angiogenic blood vessel growth or remodeling). Both junction types are tethered to the actin cytoskeleton, and their assembly, maintenance and remodeling are modulated by actin-myosin contractile forces [4]. Although the intracellular signaling pathways that control EC cytoskeletal remodeling and junctional adhesion are still an active area of investigation, previous studies have implicated the Ras Homology (Rho)-family member RHOA in these processes (reviewed in [4–6,2,7]).

RHOA is a small, monomeric GTPase that transduces stimuli from hormones, growth factors, cytokines, and transmembrane proteins to downstream effectors of cellular signaling. RHOA acts as a molecular switch, cycling between a GDP-bound inactive state and a GTP-bound active state [8,9]. In response to upstream cellular signals, guanine nucleotide exchange factors (GEFs) switch RHOA to its active, GTP-bound state [10]. Active RHOA can then activate effectors including the Rho-associated protein kinases ROCK1 and ROCK2 [11]. GTPase-activating proteins (GAPs) inactivate RHOA by restoring it to its inactive, GDP-bound state [10]. Although previous studies suggest RHOA may regulate many critical aspects of vascular EC biology including cellular migration, proliferation, actin stress fiber formation, focal adhesion, and adherens junction dynamics [7,12–16], the effects of RHOA on cytoskeletal dynamics and intercellular junction formation appear to be context-specific. Junction-destabilizing factors such as VEGF or thrombin [17–20], and junction-stabilizing factors such as Angiopoietin 1 or Sphingosine-1-phosphate [21–24] both appear to require RHOA and ROCK.

Due to the early embryonic lethality of Rhoa-knockout mice [25] most of the functional characterization of RHOA in the vascular endothelium to date has been performed in cell culture, by overexpressing dominant negative or constitutively active forms of RHOA or by treating ECs in vitro with exogenous factors that modulate their growth and development [4]. ECs in culture are also typically grown under static growth conditions in the absence of other cell types. ECs in vivo are subject to continuous or pulsatile flow, producing laminar shear stress and cyclic stretch — forces that have been shown to differentially impact RHOA subcellular localization, signaling, and its effects on cytoskeletal remodeling [26,27]. ECs in vivo also interact with numerous other cell types that influence their behavior, including vascular smooth muscle cells, pericytes, macrophages, astrocytes, and neurons [28]. Consequently, although in vitro cell culture studies have been useful for identifying the cellular activities of RHOA in ECs, the in vivo biological significance of these studies remains unclear.

Studies of human patients and disease models with cerebral cavernous malformations (CCMs) have provided evidence that RHOA may play a role in regulating vascular integrity in vivo. CCMs are clusters of abnormally dilated blood vessels that exhibit vascular hyperpermeability and are extremely prone to rupture. Familial cases of CCM are associated with mutations in CCM complex genes (CCM1, 2, or 3) [29]. The ECs of murine CCM models exhibit elevated RHOA and ROCK1/2 activity, along with a concomitant increase in actin stress fiber formation and reduced junctional integrity [30–32]. Treatment of murine CCM models with ROCK1/2 inhibitor decreases their hemorrhage and lesion burden [33,31,34,35]. CCM studies therefore suggest that RHOA signaling may negatively regulate vascular integrity in vivo. In support of this hypothesis, another study demonstrated that loss of EC RHOA in postnatal Cdh5-Cre^ERT2; Rhoa^f/f mice is sufficient to block vascular leakage following exogenous histamine but not VEGF-treatment [36].

Other in vitro studies have also suggested roles for RHOA and ROCK1/2 in angiogenesis. RHOA and ROCK1/2 are required for VEGF-stimulated capillary formation from ECs in both in vitro cell culture and in mouse ex vivo retinal explants [37,38,15]. However, studies examining in vivo requirements for RHOA in developmental angiogenesis have produced conflicting reports. In one study, EC-specific deletion of Rhoa in Tie2-Cre; Rhoa^f/f mice generated an embryonic lethal phenotype with vascular hypoplasia, yolk sac vascular remodeling defects, and dilated blood vessels, suggesting that EC-specific RHOA is required for proper vascular development and embryonic survival [39]. In contrast, a more recent study by a different group failed to observe any gross vascular abnormalities in their Tie2-Cre; Rhoa^f/f EC-specific RHOA deficiency mouse model [40]. However, viable Tie2-Cre; Rhoa^f/f pups were obtained at sub-Mendelian ratios in this study. It is therefore possible that EC RHOA deficiency produced an incompletely-penetrant, early embryonic-lethal phenotype in this model. Overall, existing studies of EC-specific RHOA function have provided limited, contradictory information on the requirement of EC RHOA in regulating embryonic vascular development in vivo.

Zebrafish provide a powerful model organism for studying embryonic vascular development and homeostasis. Forward and reverse genetic approaches in zebrafish have identified molecular pathways and genes that regulate vascular integrity and development and have clear orthologs in mammals [41]. Embryos develop externally, making them highly amenable to genetic manipulation. Optical transparency and transgenic strains permit visualization of hemorrhage and vascular development in vivo, in real-time through time-lapse microscopy [42,43]. The small size of zebrafish embryos allows them to survive and continue their development for many days in the absence of circulatory function, facilitating assessment of the specificity of vascular defects.

In this study, we use zebrafish as a model organism to assess the in vivo functions of Rhoa and its effectors Rock1/2 in regulating vascular development and integrity. We examine zebrafish with mutations in the zebrafish RHOA-ortholog rhoaa, transgenic animals with EC-specific expression of wild type or mutant forms of rhoaa, animals treated with a pharmacological inhibitor of Rock1/2, and Rock1 and Rock2a/b dgRNP-injected zebrafish embryos using high-resolution optical imaging to study the consequences of Rhoa gain- and loss-of function in the vascular endothelium. Our results suggest that cranial vessel integrity in developing zebrafish is highly sensitive to either increased or decreased Rhoa gene dosage and activity, and that Rhoa and Rock are required for proper embryonic blood vessel growth and patterning in vivo.

Methods

Animal care, pharmacological treatments, and morpholino injections

Fish were housed in a large zebrafish dedicated recirculating aquaculture facility (4 separate 22,000L systems) in 6L and 1.8L tanks. Fry were fed rotifers and adults were fed Gemma Micro 300 (Skretting) once per day. Water quality parameters were routinely measured, and appropriate measures were taken to maintain water quality stability (water quality data available upon request). Zebrafish husbandry and research protocols were reviewed and approved by the NICHD Animal Care and Use Committee, in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The facility is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). EK strain zebrafish were used for all experiments unless otherwise noted. Embryos were grown at 28.5°C in embryo media (EM) or methylene blue-containing fish water (2 mL of 0.1% methylene blue per 1L media) and staged according to standardized morphological criteria [44]. EM was supplemented with 0.006% 1-phenyl 2-thiourea (PTU) (Sigma), to prevent pigment formation in post-24 hours post fertilization (hpf) embryos.

Rockout (Rho Kinase Inhibitor III; Santa Cruz Biotechnology) was used to inhibit Rock1/2 activity [45]. Rockout was dissolved in Dimethyl sulfoxide (DMSO) to generate a stock concentration of 30 mM and then diluted to a working concentration in EM + PTU. Embryos were treated from 24 hpf onward with 25 μM or 50 μM Rockout or equivalent dilutions of DMSO (solvent controls) in EM + PTU. All embryos were raised at 28.5°C and imaged or assessed for phenotypes at 52 hpf.

Rhosin (Millipore) was dissolved in Dimethyl sulfoxide (DMSO) to generate a stock concentration of 23.2 mM (10 mg/mL) and then diluted to a working concentration in EM + PTU. Embryos were treated from 24 hpf onward with 100 μM Rhosin or equivalent dilutions of DMSO (solvent controls) in EM + PTU. All inhibitor-treated embryos were raised at 28.5°C and imaged or assessed for phenotypes at 52 hpf.

Rhoaa-depleted embryos were generated by injecting one-cell stage embryos with a splice-blocking morpholino (Gene Tools) targeting rhoaa: rhoaa-MO, 5’- CGCAGCCATCTGTCACACAAAATGC -3’ (5 ng). The rhoaa MO also targets the rhoaa start codon (underlined) and may block mRNA translation. “Silent heart” embryos were generated by injecting one-cell stage embryos with a previously described translation-blocking morpholino (Gene Tools) targeting cardiac troponin T type 2a (tnnt2a): tnnt2a-MO, 5’- CATGTTTGCTCTGATCTGACACGCA -3’ (4 ng) [46]).

Whole mount in situ hybridization

Examination of gene expression by whole mount in situ hybridization was performed essentially as previously described [47,48], with the following modifications: Riboprobes were designed to hybridize primarily to the 3’UTR of each gene to minimize cross-reactivity between zebrafish orthologs with coding sequence similarities (see Table S1 for primer sequences used to amplify riboprobes). Prior to mRNA in situ hybridization analyses, embryos were fixed in 4% paraformaldehyde (PFA)/phosphate-buffered saline (PBS) overnight at 4°C or 4–5 hours at room temperature (RT) with gentle agitation on a rotating platform. Embryos were permeabilized in 10 μg/ml proteinase K for 5 minutes (24–32 hpf embryos), or 30 minutes (48 – 52 hpf embryos) at RT.

Whole mount TUNEL staining

All washes and incubations were performed with gentle rocking on a rotating platform, unless otherwise indicated. TdT-mediated dUTP Nick-End Labelling (TUNEL) staining was performed using a Fluorescein In Situ Cell Death Detection Kit (Roche) according to manufacturer’s specifications. Briefly, 24 hpf zebrafish embryos were fixed in 4% PFA/1X PBS overnight at 4°C with gentle rocking on a rotating platform. Embryos were permeabilized in 100% MeOH overnight at −20°C, and sequentially rehydrated 10 min each in 75/25%, 50/50%, 25/75%; (MeOH/1X PBS) washes and transferred into a 96-well plate. Embryos were incubated in 50 μl of TUNEL reaction mixture for 2h at 37°C in the dark and washed 3 × 20 min in 1X PBS at RT.

Zebrafish protein lysate preparation

Zebrafish crude protein lysates were prepared by adding 1 mL cold deyolking buffer (110 mM NaCL, 5.5 mM KCl, 10 mM Tris pH 8.5, 2.7 mM CaCl2) containing cOmplete Mini Protease Inhibitor Cocktail (Roche), and PhosSTOP EASYpack Phosphatase Inhibitor (Roche) in samples collected for phosphoprotein analyses, to 20 – 30 embryos. Yolk was disrupted by gentle pipetting followed by 30 sec of 1100 rpm vortex. Tissue was pelleted by centrifuging 1 min at 3,000 rpm and supernatant was discarded. 0.7 mL of cold wash buffer (110 mM NaCl, 5.5 mM KCl, 10 mM Tris pH 8.5, 2.7 mM CaCl2) containing protease inhibitor was added. Tissues were subjected to Dounce homogenization, centrifuged 5 min at 5,000 rpm and the supernatant discarded. 5 μl of 2X LDS Sample Buffer containing 5 % 2-Mercaptoethanol per embryo was added to each sample. Lysates were vortexed, incubated 10 min at 85°C, then stored at −20°C. Immediately prior to analysis, a 1:2 dilution of lysate to 1X LDS Sample Buffer was prepared and heated 10 min at 85°C, and 15 μl of diluted lysate was analyzed by Western analysis.

Transgenic zebrafish lines

Published transgenic fish lines used in experiments include Tg(gata1:DsRed)^sd2Tg [49], Tg(kdrl:EGFP)^s843 [50], Tg(fli1:EGFP)^y1 [51], Tg(fli1:nEGFP)^y7 [52], Tg(egfl7:GAL4FF)^{gSAlzGFFD478A} [53,54], and Tg(UAS:EGFP)^nkuasgfp1aTg [55].

All cloning was performed using SLiCE cloning techniques unless otherwise noted [56]. See Table S1 for primer sequences used for SLiCE cloning. The p14xUAS-Tom-2A plasmid was made by cloning the Tomato fluorescent protein (Tom) and P2A viral cleavage peptide (2A) [57,58] sequences into the multiple cloning site (MCS) of pT1ump-14xUAS-MCS-POUT (pT1UMP), which includes an E1b basal promoter, the Ocean pout antifreeze protein polyA, and is flanked by Tol1 arms [59]. Select rhoaa coding sequences were PCR-amplified from wild type or rhoaa-mutant (R17, Δ1) zebrafish cDNAs and cloned into pCRII-TOPO (Invitrogen), according to manufacturer’s specifications. To make p14XUAS-Tom-2A-rhoaa, p14XUAS-Tom-2A-rhoaaR17, and p14XUAS-Tom-2A-rhoaaΔ1, rhoaa coding sequences were PCR-amplified from wild type or rhoaa-mutant / pCRII-TOPO template and cloned into p14xUAS-Tom-2A. p14XUAS-Tom-2A-rhoaaN19 and p14XUAS-Tom-2A-rhoaaV14 were made by site-directed mutagenesis of p14XUAS-Tom-2A-rhoaa. G0 Tg(UAS-Tom-2A), Tg(UAS-Tom-2A-rhoaa), Tg(UAS-Tom-2A-rhoaaR17), Tg(UAS-Tom-2A-rhoaaΔ1), Tg(UAS-Tom-2A-rhoaaN19), and Tg(UAS-Tom-2A-rhoaaV14) transgenic zebrafish embryos were each generated by injecting individual p14xUAS-Tom-2A or p14xUAS-Tom-2A-rhoaa isoforms into zebrafish embryos along with zebrafish-optimized Tol1 transposase RNA [60].

Mutant zebrafish lines and genotyping analyses

The y172 (aka R17, rhoaa^R17, or rhoaa^y172) mutant was identified in an F2 ENU mutagenesis screen for dominant mutants displaying a hemorrhage phenotype at 2–4 dpf in the Tg(fli:EGFP)^y1 background (Fig. S1). Bulked segregant analysis and fine genetic mapping of y172 were carried out as described previously [61] and linked y172 to a 3.148 MB region on Chromosome 8, in between determined SSLP markers ZC221E5–6 and ZK83F4–7 (Fig. S1). Whole genome sequencing (Illumina HiSeq)-based identification of deleterious mutations within the critical mapping region revealed a p.G17R mutation in rhoaa. The rhoaa^y172 or rhoaa^R17 mutants were genotyped using derived cleaved amplified polymorphic sequence (dCAPS) analysis [62–64]. Briefly, genomic DNA was isolated from adults or embryos as previously described [65]. PCR was used to introduce an XbaI restriction enzyme site in R17-mutant, but not WT rhoaa PCR product, using rhoaaY172Geno-F and rhoaaY172Geno-R primers (Table S1). PCR products were cleaved with XbaI, generating 194 bp WT and/or 169 bp R17-mutant products, analyzed by gel electrophoresis.

rhoaa^Δ1 mutants were generated using CRISPR/Cas9 technology as previously described [66], using the single-guide RNA (gRNA) sequence ATCGTTGGAGATGGAGCCTGTGG, which was selected using CHOPCHOP (https://chopchop.cbu.uib.no) [67] to target the Rhoaa GTP/GDP binding domain.

Mutagenesis of rock1, rock2a, and rock2b genes was performed using the high-efficiency dgRNP-based CRISPR/Cas9 system [68,69]. The following gRNAs were selected using CHOPCHOP [67] to target the N-terminal kinase domains of each rock gene: rock1, GTGGGCACGCTTCTACACTGCGG; rock2a, AGCCCATGGGTCGTACAGGTAGG; rock2b, GAGCCCTTGGATAGTTCAGGTGG. Target-specific Alt-R^® crRNA, universal Alt-R tracrRNA, Alt-R S.p. Cas9 nuclease, v.3, and Duplex buffer were purchased from Integrated DNA Technologies (IDT). crRNA/tracrRNA duplex preparation, Cas9 RNP complex (dgRNP) formation, injections were performed as previously described [68]. 1 nl (5 μM) of dgRNP complex per gene target was injected into an embryo.

dgRNP-injected embryos, and rhoaa^Δ1 mutant embryos and adults were genotyped as previously described [70]. Genomic DNA was extracted using the REDExtract-N-Amp Tissue PCR Kit (Sigma): 50 μl of extraction solution and 12.5 μl of tissue preparation solution was added to each embryo. Samples were mixed by vortex, incubated 10 min at RT, vortexed, and heated to 95°C for 4 min. 50 μl of neutralization solution B was added, samples were vortexed, and stored at −80°C or used immediately. Solution volumes were doubled for DNA extraction from adult tail clips. For each assay, specific tailed (tail-F and Pigtail-R) genotyping primers (Table S1) were designed to amplify a fragment spanning the CRISPR binding site. A 10 μl total volume AmpliTaq-Gold (Life Technologies) PCR reaction was set up for each sample, containing 1 μl 10X PCR Gold Buffer, 0.5 μl 25 mM MgCl2, 1 μl 0.5 mM tail-F primer, 1 μl 1 mM Pigtail-R primer, 0.2 μl 10 mM FAM-M13 primer, 0.1 μl 10 mM each dNTP Master Mix, 0.1 μl TaqGold polymerase, 1 μl of a 1:10 dilution of genomic DNA in water, and 5.1 μl water. PCR conditions were as follows: 10 min denaturation at 95°C; 35 cycles of 95°C for 30 sec, 58°C for 30 sec, and 72°C for 30 sec; and 10 min final extension at 72°C. 2 μl of PCR product was combined with 0.2 μl ROX400HD (Life Technologies) and 9.8 μl HiDi Formamide, and the solution was denatured at 95°C for 5 min and run on either a ABI3730 or ABI3130XL Genetic Analyzer (Applied Biosystems). Fragment analysis was performed using GeneMapper (Life Technologies) or Microsatellite Analysis Application (MSA; Thermo Fisher Connect) software.

Lentivirus production

Lentiviral constructs for use in endothelial cell culture were generated using SLiCE cloning techniques unless otherwise noted [56]. See Table S1 for primer sequences. Constructs with the CMV promoter were used because the PGK promoter has previously been shown to exhibit reduced or absent activity in endothelial cells, including HUVEC [71,72]. The sCMV promoter was amplified from pCS2+ [73] by PCR and cloned into the pCDH-EF1α-MCS-(PGK-GFP-T2A-PURO) bidirectional promoter lentiviral vector (System Biosciences), replacing the PGK promoter, to generate pCDH-EF1α-MCS-(sCMV-GFP-T2A-PURO). Wild type or mutant (R17, N19, V14, Δ1) rhoaa coding sequences were PCR-amplified from p14XUAS-Tom-rhoaa constructs and cloned into the MCS of pCDH-EF1α-MCS-(sCMV-GFP-T2A-PURO) to generate pCDH-rhoaa (WT, R17, N19, V14, Δ1). Tomato (Tom) fluorescent protein sequences were then PCR-amplified and cloned into the MCS of pCDH-EF1α-MCS-(sCMV-GFP-T2A-PURO) or into pCDH-rhoaa to generate pCDH-Tom control or pCDH-Tom-rhoaa constructs. Notably, previous studies indicate that N-terminal epitope or fluorophore-tagging does not significantly impact RHOA subcellular localization or activity in endothelial cells [74,75].

Lentivirus was made from pCDH-Tom and pCDH-Tom-rhoaa (WT, R17, Δ1, N19, or V14) using the ViraPower Lentiviral Expression System (Invitrogen): 9 μg ViraPower Packaging Mix and 3 μg plasmid were added to 1.5 mL Opti-MEM. 36 μl Lipofectamine 2000 was added to 1.5 mL Opti-MEM. Solutions were gently mixed, incubated 5 minutes (min) at room temperature (RT), then combined, gently mixed, and incubated 20 min at RT. The resulting solution was added dropwise to a T75 flask of 90% confluent HEK293T cells grown in 10% FBS in DMEM media (Gibco) and gently mixed by tilting. Cells were incubated at 37 °C in a humidified 5% CO2 incubator, and inspected for syncytia formation and Tomato/GFP-fluorescence. Cell media was replaced daily. Lentivirus-containing media was collected at both 2 and 3 days post-transfection: Cell debris was pelleted by centrifugation 15 min at 3,000 rpm and discarded. Virus-containing media was filtered through a Millex-HA 0.45 μm filter and stored at −80 °C until use.

Endothelial cell culture, infection, and VEGF-treatment

Passage 1–6 human umbilical vein endothelial cells from pooled donors (HUVEC, Lonza) were cultured in 0.003% Endothelial cell growth supplement (Millipore), 0.01% Heparin, 20% FBS, 2.5 ug/mL Amphotericin B and 0.1X penicillin/streptomycin in M199 media (Gibco) (EC growth media) on 1 mg/ml gelatin / 1X phosphate buffered saline (PBS)-coated tissue culture flasks at 37 °C in a humidified 5% CO2 incubator, as previously described [76].

Stable HUVEC lines expressing Tomato, Tomato-rhoaa, Tomato-rhoaa^R17, Tomato-rhoaa^Δ1, or Tomato-rhoaa^N19 from integrated lentiviral vectors were generated by infecting HUVEC with the appropriate Lentivirus: 2.5 μl 6 mg/mL Polybrene was mixed with 500 μl of HUVEC growth media. Media was removed from 90% confluent HUVEC grown in 6-well tissue culture plates, and Polybrene mixture was added dropwise to cells. 1 mL of Lentivirus-containing media was added to culture, and HUVEC were grown at 37 °C until greater than 30% expressed GFP. Transgene-expressing HUVEC were selected by Puromycin treatment. We were unable to recover stable lines of HUVEC expressing Tomato-rhoaa^V14 due to non-viability of cells.

For immunohistochemistry (IHC) and immunostaining, HUVEC were transferred to 1 mg/ml gelatin /PBS-coated Lab-Tek chamber slides (Thermo Fisher) and allowed to attach overnight for anti-Vinculin IHC and Phalloidin-staining or grown until confluent for anti-Cdh5 IHC, at 37°C.

For VEGF-treatments, HUVEC were seeded in EC growth media, then serum starved in basal M199 media for 4 hours at 37°C. Recombinant Human VEGF 165 protein (R&D Systems, 293VE/CF) was dissolved in 1 mg/mL BSA/PBS at a stock of 50 μg/mL, and supplied to HUVEC at a dose of 50 ng/ml in basal M199 media for 30 minutes (active GTP-RAC1 enzyme-linked immunosorbent assay [ELISA assay]), or 1 hour (anti-Vinculin IHC and Phalloidin staining). Lysates were then collected and analyzed, or cells were fixed for IHC and staining using the methods described below.

Active GTP-RAC1 enzyme-linked immunosorbent (ELISA) assay

A Colorimetric G-LSA activity assay kit (Cytoskeleton, BK128) was used to quantitatively assess GTP-bound RAC1 levels in stable lines of HUVEC, according to manufacturer’s specifications [77]. VEGF-stimulated confluent HUVEC grown in gelatin-coated T75 flasks were washed with ice-cold PBS, then lysed in ice-cold Lysis Buffer containing protease inhibitor. Cell lysates were clarified by centrifugation at 10,000 x g for 1 min at 4°C, aliquoted, snap frozen in liquid nitrogen and stored at −80°C. 20 μl of never-frozen lysate was used in the Precision Red Advanced Protein Assay, to determine protein concentration. Snap-frozen lysates were quickly thawed in a RT water bath, and sample protein concentrations were equilibrated to the lowest concentration sample in each replicate using ice-cold Lysis Buffer containing protease inhibitor. GTP-bound RAC1 levels were determined for sample HUVEC cell lysates, a Lysis Buffer-only blank negative control, and a provided GTP-bound recombinant RAC1 protein positive control. Absorption of the ELISA wells at 490 nm was measured using a Synergy H1 microplate reader (BioTek) with Gen5 software (BioTek). Data is reported as percent average absorbance from two independent experiments ± standard deviation.

HUVEC Immunohistochemistry and staining

All washes and incubations were performed with gentle agitation on a rotating platform at RT, unless otherwise indicated. For anti-Cdh5 and anti-Vinculin IHC, HUVEC were fixed for 5 min in 100% methanol at RT. Cells were washed 4 × 5 min in tris-buffered saline (TBS; 136 mM NaCl, 25 mM Tris, pH 7.4), then 15 min in tris-glycine buffer (0.025M Tris, 0.192M glycine, pH 8.3). Cells were permeabilized 30 min in 0.1% Saponin/TBS, washed 4 × 5 min in TBS, and blocked 1 hour in TBS + 0.1% Tween-20 (TBST) + 2.5 % bovine serum albumin (BSA) + 5% sheep serum, then incubated in primary antibody diluted (1:500 anti-VE Cadherin antibody, abcam ab33168 or 1:500 anti-Vinculin antibody clone hVIN-1, Millipore V9131) in blocking solution, overnight at 4 °C. Cells were washed 4 × 5 min in TBST, incubated 1 hour in a 1:1000 dilution of Alexa Fluor 488- or 546-conjugated secondary antibody (Thermo Fisher) in blocking solution, then incubated in a 1:2000 dilution of Hoechst 33342/TBS (Invitrogen) for 20 min. Cells were washed 4 × 5 min in TBST, mounted in Fluorescent Mounting Medium (Dako), and stored at 4°C in the dark.

For Phalloidin staining, HUVEC were fixed in 2% paraformaldehyde (PFA) / PBS overnight at 4 °C. Cells were washed 4 × 5 min in PBS, then 15 min in tris-glycine buffer. Cells were permeabilized 30 min in 0.1% TritonX100/PBS and incubated in a solution containing a 1:2000 dilution of Hoechst 33342 (Invitrogen) + a 1:100 dilution of Texas Red-633 Phalloidin (Thermo Fisher) in PBS for 20 min. Cells were washed 4 × 5 min in PBS, mounted in Fluorescent Mounting Medium (Dako), and stored at 4°C in the dark.

In vitro protein synthesis and active GTP-RHOA pulldown

Select rhoaa isoform coding sequences with polyAs were PCR-amplified from WT or mutant p14XUAS-Tom-2A-rhoaa constructs, using primers that add the canonical Kozak sequence GCCACC immediately upstream of the ATG start codon (rhoaaCDSkozak-F, rhoaaCDSpolyA-R; Table S1), and cloned into pCRII-TOPO (Invitrogen), according to manufacturer’s specifications. Rhoaa constructs were transcribed and translated from 1 μg of plasmid DNA (50 μl total reaction volume) using the TNT SP6 Coupled Rabbit Reticulocyte Lysate System (Promega), according to manufacturer’s specifications. An equivalent volume “water-only” control was used to assess endogenous GTP-bound and total-RHOA protein levels present in the lysate. This in vitro transcription/translation system was selected to minimize RHOA isoform-specific transcriptional or translational effects and protein degradation by intracellular proteases. 2 μl of crude protein lysate was added to 38 μl 2X LDS Sample Buffer containing 2.5% 2-Mercaptoethanol, vortexed, incubated 10 min at 85°C, then stored at −20°C (Input). RHOA proteins were loaded with GTP, and active GTP-bound RHOA pulldowns were conducted using the bead pull-down format RhoA Activation Assay Biochem Kit (Cytoskeleton), essentially according to manufacturer’s specifications. Briefly, 20 μl of crude protein lysate was added to 980 μl protease-inhibitor containing cold lysis buffer + 70 μl loading buffer + 10 μl water. (TNT Reticulocyte Lysate contains in-excess amounts of GTP for loading). The solution was mixed by pipetting and incubated 15 min on a nutating mixer at room temperature (RT). 100 μl of STOP Buffer was added, and active GTP-bound RHOA was collected by pulldown using 15 μl (50 μg) Rhotekin-RBD beads, according to manufacturer’s specifications. Following pulldown, 20 μl of 2X LDS Sample Buffer containing 5% 2-Mercaptoethanol was added to the beads. This mix was vortexed, incubated 10 min at 85°C, then stored at −20°C (Pulldown). Immediately prior to analysis, a 2:3 dilution of Input to Wash Buffer with protease inhibitor was prepared. Input/Wash Buffer and Pulldown samples were re-heated 10 min at 85°C, and 10 ul - 15 ul of each sample was analyzed by Western analysis.

Western analysis

Protein lysate was run out on NuPAGE Novex Bis-Tris gels using the XCell Sure Lock Mini-Cell system (Invitrogen) at 140 V, according to manufacturer’s specifications. Proteins were transferred to methanol-treated PVDF membranes using a Mini Trans-Blot Cell wet-transfer apparatus (BIO-RAD), according to manufacturer’s specifications at 400 mA for 1.5 hours. All membrane washes and incubations were performed with gentle rocking at room temperature (RT) unless otherwise specified. Following transfer, membranes were washed 5 min in Tris-buffered saline with 0.1% Tween-20 (TBST). For western blot analysis of RHOA-only, membranes were allowed dry completely, rehydrated in methanol, and rewashed with TBST. Membranes were blocked at least 1 hr in 5% skim milk (RHOA, MLC2, α-Tubulin) or 5% bovine serum albumin (T18/S19 pp-MLC2) in TBST. Membranes were incubated in primary antibody diluted in TBST (1:500 anti-RHOA antibody, Cytoskeleton ARH04) or diluted in block (1:30,000 anti-α-Tubulin antibody, Sigma T6199; 1:1000 anti-T18/S19 pp-MLC2 antibody, Cytoskeleton #3674; 1:4000 anti-MLC2 antibody, Abcam ab48003) overnight at 4 °C. After 5 × 5 min TBST washes, membranes were incubated in 1:2000 digital HRP-conjugated secondary antibody (Kindle Bioscience) in TBST. After TBST washes, a 1-Shot Digital ECL reaction (Kindle Bioscience) was performed, and images were acquired using a KwikQuant Imager (Kindle Bioscience). For phosphoprotein analyses only, 20 mM sodium fluoride and 2 mM sodium orthovanadate (phosphatase inhibitors) were added to blocking, primary, and secondary antibody solutions to improve signal [78]. Blots probed with multiple antibodies were stripped between analyses: Membranes were washed 5 min in Tris-buffered saline (TBS), incubated 20 minutes in Restore Western Blot Stripping Buffer (Thermo Scientific), and washed 4 × 5 min in TBST prior to blocking. Quantification of relative band density was performed using ImageJ software [79]. Data is reported as the percent average density from a minimum of 3 blots from independent experiments ± standard deviation.

Microscopy, imaging, and quantitative image analyses

Fluorescent images of HUVEC were acquired using a Zeiss LSM 880 confocal microscope with Fast Airyscan, using either a 40X or 63X objective with optical zoom. All images from a single experimental replicate were acquired using the same settings.

Whole-mount in situ hybridization images and transmitted-light images of live zebrafish embryos were collected using a Leica M205 stereo microscope and LAS V4.7 or LAS X software, with extended depth of focus z-stack image processing. Fluorescent images of zebrafish embryos were collected using a Nikon Yokogawa CSU-W1 spinning disk confocal microscope. Transmitted light movies of live zebrafish embryos were collected using a Leica M205 stereo microscope and LAS V4.7 software, or a Nikon Yokogawa CSU-W1 spinning disk confocal microscope. Prior to imaging, live embryos were immobilized in EM containing buffered MS-222 and embedded in 0.8% low melting point agarose in EM. Zebrafish embryos were genotyped after imaging and data analysis.

All fluorescent images were processed using Fiji/ImageJ [79] or Imaris (Oxford Instruments) and all figures were assembled using Powerpoint (Microsoft) and/or Photoshop (Adobe). Copies of original transmitted light images of live zebrafish embryos were globally adjusted for red color saturation to enhance red blood cell visibility equally across matched (control and treatment) samples. Movies were assembled using Imaris (Oxford Instruments), Premiere Pro (Adobe), and After Effects (Adobe).

All quantitative analyses were performed using Fiji/ImageJ. To quantify Vinculin-positive adhesion complexes in HUVEC, a maximum intensity z-projection was generated, and duplicated. A Gaussian Blur filter was applied to one copy. This blurred image and the “Image Calculator, Difference” function was applied to the original z-projection to enhance visibility and detection of adhesion complexes. The “Find Maxima” function was used to count the number of Vinculin-positive adhesion complex clusters in each cell. The “Maxima Within Tolerance” output was used to create a mask that was then superimposed upon the original image to verify automated cluster counts.

To quantify Phalloidin-stained F-actin stress fibers in HUVEC, maximum-intensity Z-projections were generated for each image, then sequential lines were drawn across stress fibers within a cell. The “Analyze Plot Profile” function was used to generate an intensity profile for each drawn line. The “Analysis, Find Peaks” function of the “BAR” plugin [80] was used to label and count fluorescent peaks above baseline, with each peak corresponding to a single stress fiber.

Zebrafish ISV widths were measured in maximum-intensity Z-projections using the “Analyze, Measure” function to measure the widest point of each vessel. Only non-detached, lumenized ISVs along the yolk sac extension were counted. Blinded measurements were obtained using only the EGFP (green) channel.

Statistical analyses

For each zebrafish experiment, data from two or more independent biological replicates with separate cohorts of embryos were analyzed for the main effects of treatment.

For analyses of hemorrhage in rhoaa^R17 mutant embryos, homogeneity across replicates was determined using the G-test of independence, and homogenous datasets (heterogeneity G-value ≥ 0.05) were combined for statistical analysis. Heterogeneous replicate datasets were analyzed separately and combined only when statistical analyses yielded identical results for each replicate dataset. Significant differences among treatments were determined using two-tailed Fisher’s exact tests on cumulative raw counts, with Bonferroni correction applied to multiple comparisons (alpha = 0.05).

For analyses of wild type versus rhoaa^Δ1/+ incross progeny, and Rhosin-treated embryos with a specific phenotype were categorized (wild type, hemorrhaging, no circulation; wild type CtAs, reduced CtAs, overbranching CtAs) and counted. Homogeneity across replicates was determined using the G-test of independence, and homogenous datasets (heterogeneity G-value ≥ 0.05) were combined for statistical analysis. Heterogeneous replicate datasets were analyzed separately and combined only when statistical analyses yielded identical results for each replicate dataset. Significant differences among treatments were determined using Fisher-Freeman-Halton exact tests of independence (alpha = 0.05) with post hoc two-tailed Fisher’s exact tests with Bonferroni correction applied to multiple comparisons (alpha = 0.05), on raw counts.

For EC-specific transgenic Tomato or rhoaa isoform overexpression experiments, morpholino and dgRNP injection experiments, and Rockout treatment experiments, significant phenotypic differences among groups or treatments across biological replicates were determined using the Cochran-Mantel-Haenszel test for repeated 2×2 tests of independence (alpha = 0.05).

For measurement (blood vessel widths), feature count (adhesion complex clusters, F-actin stress fibers), and relative quantification (western blot densitometry, ELISA) analyses, significant phenotypic differences among groups were calculated using one-way ANOVA with post hoc Tukey test for multiple comparisons (alpha = 0.05).

Results

Zebrafish rhoaa mutants with distinct cranial vascular integrity and patterning defects

We identified a novel mutant (y172) in a forward-genetic ENU mutagenesis screen for dominant or partially-dominant zebrafish mutants with intracranial hemorrhage (Fig. S1a,b). Genetic fine mapping and genome sequencing suggested that the causative defect in the y172 mutant is a Glycine to Arginine substitution at amino acid 17 in the evolutionarily conserved GTP/GDP-binding domain[81,82] of rhoaa, a zebrafish ortholog of mammalian RHOA (Fig. 1a; Fig. S1c–f). We therefore subsequently refer to y172 mutants as rhoaa^R17 or R17 mutants. Whole-mount in situ hybridization analyses show that rhoaa is broadly expressed in wild type embryos at 26 hours post fertilization (hpf), with particularly strong expression in the head (Fig. 1b), becoming almost exclusively expressed in the head by 48 hpf (Fig. 1c,d). Starting at approximately 48 hpf, both R17/+ heterozygous and R17/R17 homozygous mutant embryos begin to exhibit incompletely penetrant cranial hemorrhage (Fig. 1e–g,n) with variable expressivity ranging from small hematomas barely visible under a light microscope to large hemorrhages that fill one or more brain ventricles (Fig. S1b). To determine if R17 mutants have altered blood vessel morphology or patterning, we crossed them to a Tg(kdrl:EGFP)^s843, Tg(gata1:DsRed)^sd2Tg double transgenic reporter line with EGFP-labeled endothelial cells (ECs) and DsRed-labeled primitive blood cells [50,49]. By 52 hpf, both R17/+ and R17/R17 embryos develop vasculogenic primary cranial blood vessels, including the lateral dorsal aortae (LDA; not shown) and primordial hindbrain channels (PHBC), as well as angiogenic cranial central arteries (CtA), although CtA growth is sometimes slightly delayed in R17/R17 animals (Fig. 1h–m; See Fig. S2 for vessel reference diagram [83]). R17/+ and R17/R17 embryos also exhibit normal development of trunk blood vessels, including the dorsal aorta (DA), cardinal vein (CV), and intersegmental vessels (ISVs) at 52 hpf, although dorsal longitudinal anastomotic vessel (DLAV) plexus formation is sometimes delayed (Fig. S2b; Fig. S3a–c).

Fig. 1. rhoaa mutant zebrafish embryos exhibit vascular integrity and patterning defects.

a Amino acid sequences of the GTP/GDP-binding domains of wild type (WT) Rhoaa, known dominant negative (DN) Rhoaa^N19, known constitutively active (CA) Rhoaa^V14, Rhoaa^R17, and Rhoaa^Δ1. b Whole-mount in situ hybridization analysis of rhoaa gene expression in a 26 hours post fertilization (hpf) wild type (WT) zebrafish embryo shown in lateral view, anterior to left. c Whole-mount in situ hybridization analysis of rhoaa gene expression in a 48 hpf WT zebrafish embryo shown in lateral view, anterior to left. d Magnified view of head in C. e-g Stereoscope transmitted light images of 52 hpf Tg(kdrl:EGFP), Tg(gata1:DsRed); rhoaa^R17/+ incross progeny heads shown in lateral view, anterior to left. Arrowheads indicate hemorrhage. h-j Confocal images of the vasculature (kdrl:EGFP; green) and primitive erythrocytes (gata1:DsRed; magenta) in the same animals as in e-g. Arrowheads indicate hemorrhage. k, l, m Zoomed-in view of central arteries (CtAs; boxed areas in h-j). n Quantitation of the percentage of rhoaa^R17/+ incross progeny with indicated genotypes with hemorrhage by 52 hpf. *P < 0.0003 compared to WT siblings, by Fisher’s exact test with Bonferroni correction. o-q Stereoscope transmitted light images of 52 hpf Tg(kdrl:EGFP), Tg(gata1:DsRed); rhoaa^Δ3/+ incross progeny. Arrowhead indicates hemorrhage. r-t Confocal images of the vasculature (kdrl:EGFP; green) and primitive erythrocytes (gata1:DsRed; magenta) of the same animals as in o-q. Arrows indicate overbranching CtAs. Arrowheads indicate hemorrhage. u-w Zoomed-in view of CtAs (boxed area in r-t). x, y Quantitation of the percentage of rhoaa^Δ1/+ incross progeny of the indicated genotypes that (x) have hemorrhage or lack circulation at 52 hpf, and (y) have reduced sprouting or overbranching of CtAs. *P < 0.0003, compared to other genotypes, by Fisher’s exact test with Bonferroni correction. Mean ± SD is shown for each graph.

The rhoaa gene is one of five highly-conserved zebrafish RHOA orthologs (Fig. S4a,b) exhibiting highly overlapping developmental expression patterns (Fig. S4c–h). The strong possibility of compensation by other zebrafish rhoa orthologs and the dominant nature of the R17 mutation suggest that R17 may be acting as either a dominant negative (DN) or constitutively active (CA) allele of rhoaa. In support of this hypothesis, previous studies have shown that mutating the RHOA GTP/GDP-binding domain yields both DN and CA RHOA alleles. For example, RHOA^V14 is a CA version of human RHOA that is unable to hydrolyze GTP, while RHOA^N19 is a DN version of human RHOA that preferentially remains bound to GDP [81,84] (Fig. 1a). To help further explore the effects of and validate the R17 mutation, we used CRISPR/Cas9 mutagenesis [66] to generate an additional zebrafish mutant, A15_C16delinsD, with a small, in-frame 3-base pair deletion and amino acid substitution in the Rhoaa GTP/GDP-binding domain, (designated rhoaa^Δ1 or Δ1; Fig. 1a). We found that the R17 and Δ1 mutants exhibit partially overlapping but distinct phenotypes (Fig. 1, Fig. S3).

Like R17/+ heterozygotes, Δ1/+ heterozygous embryos exhibit wild type morphology with incompletely penetrant cranial hemorrhage beginning at 48 hpf (Fig. 1o,p) and normal cranial and trunk blood vessel patterning at 52 hpf (Fig. 1s,v; Fig. S3e). However, unlike R17/R17 animals, most Δ1/Δ1 homozygotes lack blood circulation and exhibit heart edema, despite having a weak heartbeat (Movie S1; Fig. 1q). They also have smaller heads with atypical morphology (Fig. 1q). Although the vasculogenic primary cranial vessels (LDA, PHBC) are normally patterned in Δ1/Δ1 homozygous mutants, they exhibit ectopic growth and branching of the angiogenic central arteries (CtA) and their vessels are poorly dilated (Fig. 1q,t,w). Overall trunk blood vessel patterning is normal in 52 hpf Δ1/Δ1 homozygotes, although these vessels (especially the ISVs) also fail to fully dilate (Fig. S3f). The impaired blood vessel dilation phenotype of Δ1/Δ1 homozygotes may be due to a lack of blood circulation, as we also observe it (but not CtA overbranching) in cardiac troponin T type 2a (tnnt2a) morpholino antisense oligonucleotide (MO)-injected “silent heart” zebrafish embryos that lack a heartbeat and blood flow (Fig. S5; [46]). Together, these data suggest that while R17/+ and Δ1/+ heterozygotes both exhibit impaired cranial vascular integrity, Δ1/Δ1 homozygotes have additional vascular defects not found in R17/R17 homozygotes.

The difference between R17/R17 and Δ1/Δ1 homozygotes does not appear to be the result of differences in the relative amounts of Rhoa proteins, since both show very similar (approximately 50%) reductions in total Rhoa, as measured by western blotting of whole-embryo zebrafish protein lysates collected at 50 hpf (Fig. S6a,b). These results suggest that the Rhoaa^R17 and Rhoaa^Δ1 proteins may instead possess distinct activities.

To examine whether defects in Rhoaa alone result in vascular integrity defects, we injected Tg(egfl7:GAL4ff), Tg(UAS:EGFP) double-transgenic zebrafish embryos with vascular EC-specific EGFP reporter expression [53,54] with a MO targeting rhoaa (Fig. S6c–j). Compared to uninjected sibling controls, rhoaa morphant zebrafish embryos exhibit weakly penetrant cranial hemorrhage beginning at 48 hpf (Fig. S6f,i), suggesting that reducing Rhoaa alone results in mild impairment of cranial vascular integrity in vivo. rhoaa morphant embryos exhibit lower hemorrhage frequency than either R17/+ heterozygotes or R17/R17 homozygotes and, unlike Δ1/Δ1 homozygotes, do not exhibit blood vessel patterning or circulation defects (Fig. S6d,e,g,h,j). Together, these results suggest that rhoaa^R17 and rhoaa^Δ1 are likely at least partially dominant-negative alleles of rhoaa.

Rhoaa alleles have differential effects on adherens junctions, cell adhesions, and the actin cytoskeleton in cultured primary ECs in vitro

Vascular permeability is modulated primarily by the integrity of EC-EC adherens and tight junction complexes [2]. Adherens junctions consist of transmembrane cadherins, such as Cdh5/VE-Cadherin. Previous studies using primary human umbilical vein ECs (HUVEC) in vitro have shown that increased RHOA activity disrupts adherens junctions [6,85], while DN suppression of RHOA signaling does not alter the distribution of adherens junction proteins in unstimulated cells [86]. To examine the effects of expressing different mutant rhoaa proteins on EC-EC junctions in vitro, we used lentiviral transduction to generate stable HUVEC lines expressing either wild type rhoaa, known DN rhoaa^N19, rhoaa^R17, or rhoaa^Δ1. We also attempted to generate stable HUVEC lines expressing known CA rhoaa^V14, but cells transduced with this allele are not viable (data not shown). We examined adherens junction morphology in confluent monolayers of stable rhoaa isoform-expressing HUVEC lines using immunohistochemical staining with an anti-Cdh5 antibody (Fig. 2a–j).

Fig. 2. Rhoaa isoforms exhibit distinct activities in vitro.

a-e Anti-CDH5 immunostaining of adherens junctions (green) in HUVEC stably transduced with lentiviral constructs expressing wild type (WT) or mutant zebrafish rhoaa isoforms. Nuclei labeled with Hoescht 33342 (blue). f-j Zoomed-in view of boxed areas shown in panels on left. k-o Fluorescent phalloidin staining of F-actin (magenta) in HUVEC expressing wild type (WT) or mutant zebrafish rhoaa isoforms. Nuclei labeled with Hoescht 33342 (blue). p-t Anti-Vinculin immunostaining of adhesions (green) in HUVEC stably transfected with wild type (WT) or mutant zebrafish rhoaa isoforms. Nuclei labeled with Hoescht 33342 (blue). u Quantitation of the number of phalloidin-stained F-actin stress fibers per cell. *P < 0.01, compared to WT and Control samples, by one-way ANOVA with post hoc Tukey test. Mean ± SD is shown for each graph. v Quantitation of the number of Vinculin-positive adhesion clusters per cell. **P < 0.01 compared to all other samples, and *P < 0.0001, compared to WT and Control samples, by one-way ANOVA with post hoc Tukey test. Mean ± SD is shown for each graph.

Compared to “empty-vector” transduced controls (Fig. 2a,f), HUVEC expressing WT rhoaa exhibit jagged, highly discontinuous junctional Cdh5 staining (Fig. 2b,g), as previously reported for ECs with excess RHOA pathway activation [18]. HUVEC expressing rhoaa^R17 also exhibit slightly jagged junctional staining (Fig. 2c,h), although in this case, the staining is much less discontinuous than in HUVEC expressing WT rhoaa. In contrast, HUVEC expressing either the known DN rhoaa^N19 (Fig. 2d,i) or rhoaa^Δ1 (Fig. 2e,j) exhibit thin, mostly continuous junctional Cdh5 staining.

Stress fibers consist of filamentous actin (F-actin) bundles that form in response to cellular tension and terminate in Vinculin-containing multiprotein adhesion complexes [87–90]. Adhesion complexes link the actin cytoskeleton to the extracellular matrix (i.e., focal adhesions) or connect cells through adherens junctions [91,88,90,92,93]. They serve as an anchoring force that can help to maintain or disrupt the EC barrier [94,88] and modulate signals that regulate cell proliferation, migration, and survival [88,94,95]. Previous studies have shown that treating ECs in culture with vascular permeability-inducing factors such as VEGF and thrombin induces stress fiber and adhesion complex formation, and that these phenotypes are RHOA- and ROCK1/2-dependent [19,20,96,88,15,91,97]. We examined the ability of subconfluent, stable rhoaa isoform-expressing HUVEC lines to form F-actin stress fibers and adhesion complexes in response to VEGF stimulation, using phalloidin staining (Fig. 2k–o,u) and immunohistochemical staining with an anti-Vinculin antibody (Fig. 2p–t,v), respectively.

Like controls (Fig. 2k,u), HUVEC expressing ectopic WT rhoaa exhibit F-actin stress fibers (Fig. 2l,u). They also exhibit increased numbers of Vinculin-positive cell adhesions (Fig. 2q,v), as previously reported for ECs with excess RHOA activity [98,99]. Compared to controls, HUVEC expressing known DN rhoaa^N19, rhoaa^R17, and rhoaa^Δ1 exhibit disorganized F-actin, or reduced numbers of F-actin stress fibers (Fig. 2m–o,u) and cell adhesions (Fig. 2r–t,v), as previously reported for ECs following RHOA pathway inhibition [19,20,96,88]. All together, these results suggest that the Rhoaa^R17 protein may have weaker DN activity, while Rhoaa^Δ1 may act as a somewhat stronger DN allele of Rhoaa.

Rhoaa alleles exhibit reduced GTP-binding affinity in vitro

Previously characterized DN isoforms of RHOA such as RHOA^N19 exhibit reduced affinity for GTP and compete with endogenous GTPases for GEF-binding [100–103,84,104]. To assess GTP-binding of our mutant rhoaa alleles, we conducted active-Rhoa pulldown assays with in vitro-synthesized WT and mutant Rhoaa protein isoforms. Like the known DN Rhoaa^N19 mutant, Rhoaa^R17 and Rhoaa^Δ1 exhibit significantly reduced GTP-binding compared to WT Rhoaa (Fig. S7a,b). Although previous studies have reported antagonistic crosstalk between RHOA and RAC1 [2,105–107], we did not detect any significant change in RAC1 activation in VEGF-stimulated HUVEC stably expressing different rhoaa isoforms (Fig. S7c). “Empty vector” control HUVEC and HUVEC expressing ectopic WT rhoaa, rhoaa^R17, rhoaa^N19, or Rhoaa^Δ1 all exhibit equivalent levels of active GTP-bound RAC1 detected using an ELISA assay (Fig. S7c).

Zebrafish Rhoaa mutants exhibit reduced Mlc2 activity in vivo

RHOA/ROCK pathway-dependent di-phosphorylation of the regulatory Myosin Light Chain II (MLC2) at T18 and S19 has previously been shown to contribute to EC actin filament rearrangement, adhesion complex formation, and barrier enhancement or permeability in response to distinct vasoactive stimuli [108–113,24,77,114,40,115]. We examined Mlc2 phosphorylation in our zebrafish mutants by western blotting of whole-embryo zebrafish protein lysates collected at 50 hpf (Fig. S7d,e). Compared to WT embryos, R17/R17 and Δ1/Δ1 homozygotes each exhibit reduced T18/S19 Mlc2 di-phosphorylation. Unlike R17/R17 homozygotes, Δ1/Δ1 homozygotes also exhibit a highly reproducible, marked reduction in mono- and unphosphorylated Mlc2 levels (Fig. S7d,e), further suggesting that Rhoaa^R17 and Rhoaa^Δ1 possess distinct activities in vivo.

Vascular defects are associated with either increased and decreased Rhoaa levels or activity in the zebrafish vascular endothelium in vivo

To explore the in vivo consequences of endothelial-specific increases or decreases in Rhoa activity, we generated a series of germline transgenic zebrafish for UAS-driven co-expression of Tomato fluorescent protein and either wild type rhoaa, rhoaa^R17, rhoaa^Δ1, DN rhoaa^N19, or CA rhoaa^V14. We crossed Tg(UAS:Tomato-2A-rhoaa) transgenics to Tg(egfl7:GAL4ff); Tg(UAS:EGFP) double-transgenic zebrafish to drive vascular endothelial-specific expression of Gal4, and the UAS:EGFP and UAS:Tomato-2A-Rhoaa reporters (Fig. 3a).

Fig. 3. Increased endothelial cell Rhoa activity induces vascular integrity and patterning defects in vivo.

a Schematics of transgenics used for conditional wild type or mutant rhoaa expression in zebrafish endothelial cells. b, f, j Stereoscope transmitted light images of 52 hpf progeny derived from a cross between Tg(egfl7:GAL4ff), Tg(UAS:EGFP) and either Tg(UAS:Tomato-2A-rhoaa) (b, f) or Tg(UAS:Tomato-2A-rhoaa^V14) (j) fish. Panel b shows a non-transgene expressing control sibling of the Tomato-2A-rhoaa transgene-expressing animal in panel f. Black arrowhead in f indicates hemorrhage. c, d, g, h, k, l Tomato (c, g, k) and Tomato/EGFP (d, h, l) confocal images of cranial endothelial cells in the same embryos as in panels b, f, j. Arrows indicate reduced CtA sprouting. e, i, m Tomato/EGFP confocal images of trunk endothelial cells in 52 hpf progeny derived from a cross between Tg(egfl7:GAL4ff), Tg(UAS:EGFP) and either Tg(UAS:Tomato-2A-rhoaa) (e, i) or Tg(UAS:Tomato-2A-rhoaa^V14) (m) fish. White arrowheads indicate ISV sprouting defects or impaired vessel dilation. n-p Magnified Tomato/EGFP confocal images of trunk vasculature in the boxed areas in panels e, i, m. q-t Quantitation of the percentage of 52 hpf Tomato-positive WT Rhoaa-expressing or Tomato-positive Rhoaa^V14-expressing (light grey columns) animals or their Tomato-negative siblings (dark grey columns) with specified phenotype. Phenotypes shown in the graphs are cranial hemorrhage (q), lack of circulation (r), reduced cranial CtA sprouting (s), or defective trunk ISV sprouting (t). *P < 0.0001 compared to Tomato-negative sibling controls, by Cochran-Mantel-Haenszel test. Mean ± SD is shown for each graph.

EC-specific expression of Tomato alone has no effect on vascular integrity or development in zebrafish embryos (Fig. S8). Compared to their control Tomato-negative siblings (Fig. 3b–e,n,q–t), embryos expressing ectopic WT rhoaa in ECs exhibit incompletely penetrant cranial hemorrhage (Fig. 3f,q) or lack of circulation (Fig. 3r), reduced growth of cranial CtAs (Fig. 3g,h,s), impaired growth of trunk ISVs (Fig. 3i,o,t), and vessel dilation defects (Fig. 3g–i). Embryos expressing CA rhoaa^V14 in ECs also sometimes exhibit cranial hemorrhage (Fig. 3q) and have much more penetrant and severe circulation, CtA/ISV angiogenesis, and blood vessel dilation defects (Fig. 3j–m,p,r–t, Movie S2) than either WT-rhoaa expressing animals or their control Tomato-negative siblings (Fig. 3r–t, Fig. S9k–n,s). Unlike WT-rhoaa expressing ECs, CA rhoaa^V14-expressing ECs frequently have a rounded, punctate morphology (Fig. 3k–m,p, Movie S3) and exhibit TUNEL staining (Fig. S10d–f). Timelapse microscopy of Tg(egfl7:GAL4ff), Tg(UAS:Tomato-2A-rhoaa^V14), Tg(fli1:nEGFP)^y1 transgenic embryos with EGFP-labeled EC nuclei shows that ECs expressing CA rhoaa^V14 commonly undergo nuclear and cell fragmentation (Movie S4). Combined, these results suggest that increased Rhoaa dosage and/or activity in zebrafish ECs produces both vascular integrity and vascular growth defects, with high levels of Rhoa activation (in CA rhoaa^V14-expressing ECs) promoting EC death.

To study the consequences of decreased Rhoaa activity in the vascular endothelium in vivo, we examined transgenic zebrafish embryos with EC-specific expression of known dominant-negative DN rhoaa^N19. Compared to Tomato-negative sibling controls (Fig. 4a–d,q,u,v) embryos expressing DN rhoaa^N19 in ECs exhibit severe cranial hemorrhage (Fig. 4e, u) and reduced CtA vessel growth at 52 hpf (Fig. 4f,g,v). Unlike cranial CtAs expressing EC-specific DN rhoaa^N19, most trunk ISVs expressing this isoform grow well and successfully reach the DLAV by 52 hpf, although a subset of ISVs are abnormally dilated (Fig. 4w) and/or detach from the DA, CV, or DLAV (Fig. 4h, r, Movie S5). Like embryos expressing DN rhoaa^N19 in the vascular endothelium, embryos with EC-specific rhoaa^Δ1 expression exhibit severe cranial hemorrhage (Fig. 4i–u), reduced CtA vessel growth (Fig. 4j,k,v), and abnormally dilated and/or detached ISVs (Fig. 4l,s,x) compared to their control Tomato-negative siblings (Fig. S11k–n,s, Fig. 4u,v,x), supporting the idea that Rhoaa^Δ1 acts as a DN Rhoaa protein in vivo.

Fig. 4. Endothelial cell specific expression of dominant negative rhoaa induces cranial vessel integrity, patterning, and trunk vessel dilation defects in vivo.

a, e, i, m Stereoscope transmitted light images of 52 hpf progeny derived from a cross between Tg(egfl7:GAL4ff), Tg(UAS:EGFP) and either Tg(UAS:Tomato-2A-rhoaa^N19) (a, e), Tg(UAS:Tomato-2A-rhoaa^Δ3) (i), or Tg(UAS:Tomato-2A-rhoaa^R17) (m) fish. Panel A shows a non-transgene expressing control sibling of the Tomato-2A-rhoaa^N19 transgenen expressing embryo in panel e. Black arrowheads indicate hemorrhage. b, c, f, g, j, k, n, o Tomato (b, f, j, n) and Tomato/EGFP (c, g, k, o) confocal images of cranial endothelial cells in the same embryos as in panels a, e, i, m. Arrows indicate reduced CtA sprouting. d, h, l, p Tomato/EGFP confocal images of trunk endothelial cells in 52 hpf progeny derived from a cross between Tg(egfl7:GAL4ff), Tg(UAS:EGFP) and either Tg(UAS:Tomato-2A-rhoaa^N19) (d, h), Tg(UAS:Tomato-2A-rhoaa^Δ1) (l), or Tg(UAS:Tomato-2A-rhoaa^R17) (p) fish. White arrowheads indicate ISV dilation, detachment, or growth defects. (q, r, s, t) Magnified Tomato/EGFP confocal images of trunk vasculature in the boxed areas in panels d, h, l, p. u, v Quantitation of the percentage of Tomato-positive embryos expressing the indicated rhoaa isoform (light grey columns) or their Tomato-negative siblings (dark grey columns) with cranial hemorrhage (u) or reduced cranial CtA sprouting (v) at 52 hpf. *P < 0.0001 compared to Tomato-negative sibling controls, by Cochran-Mantel-Haenszel test. w, x, y Quantitation of ISV width in the mid-trunk of non-transgene expressing control siblings (sibling), Tomato-negative ISVs within Tomato/rhoaa transgene-expressing embryos (Tom−), and Tomato/rhoaa-positive ISVs (Tom+). rhoaa isoform indicated on top of graph. * P < 0.01) (N19), * P < 0.001 (Δ1), *P < 0.01 (R17) compared to all other samples, by one-way ANOVA with post hoc Tukey test. Mean ± SD is shown for each graph.

In contrast, compared to Tomato-negative sibling controls (Fig. 4u,v,y, Fig. S11u–x,c’) zebrafish embryos with transgene-driven EC-specific expression of the Rhoaa^R17 protein show relatively low levels of cranial hemorrhage (Fig. 4m,u) and no reduction in CtA growth (Fig. 4n,o,v). A small subset of these embryos do exhibit abnormally dilated and/or detached trunk ISVs (Fig. 4p,t,y), but this phenotype is not as severe or penetrant as seen in DN rhoaa^N19 (Fig. 4h,r,w) or DN rhoaa^Δ1 (Fig. 4l,s,x) expressing animals. These results suggest that Rhoaa^R17 may be a weaker dominant negative isoform of Rhoaa compared to DN Rhoaa^N19 or Rhoaa^Δ1.

Pharmacological inhibition of Rhoa recapitulates EC-specific DN rhoaa overexpression

DN RHO mutants with preferential GDP- over GTP-binding affinities (such as RHOA^N19) are thought to function by competitively binding GEFs, restricting their availability to activate other RHO GTPases [100–103,84,104]. Some EC-expressed RHOA GEFs have also been shown to target other RHO-GTPase family members, including CDC42 and RAC1 [116]. We therefore wanted to determine if the vascular integrity and patterning defects that we observe in rhoaa^Δ1/Δ1 mutants, or embryos with EC-specific expression of DN rhoaa^N19 or rhoaa^Δ1 are indeed due to loss of Rhoa-specific activity.

Unlike mammals, zebrafish do not have direct RHOB or RHOC homologs [117]. RHOB arose more recently in evolution by retrotranscription and insertion, and detailed analyses of gene structure/splicing, and sequence conservation suggest that zebrafish have five RHOA orthologs and no actual RHOC homolog (Fig. S4a,b, [117]). To further test the effects of Rhoa-specific inhibition on vascular development and integrity, we treated zebrafish embryos with Rhosin, a pharmacological inhibitor of RHOA/B/C activity previously used in zebrafish to specifically abrogate Rhoa function [118]. A treatment interval of 24 to 52 hpf was selected to bypass earlier epiboly and gastrulation defects caused by Rhoa and/or Rock inhibition [119]. Overall, we found that, like EC-specific overexpression of DN rhoaa^N19 or rhoaa^Δ1 (Fig. 4), Rhosin treatment induces cranial hemorrhage, reduced CtA growth, and trunk ISV detachment in zebrafish embryos (Fig. S12d–g). These results confirm that the vascular defects that we observe in embryos with EC-specific DN rhoaa overexpression are likely due to a loss of Rhoa-ortholog function, and not to reduced activation of other Rho-family GTPases.

Rock function is required for cranial angiogenesis and maintenance of vascular integrity in vivo

The Rho-associated protein kinases ROCK1 and ROCK2 are some of the best characterized RHOA effectors [11]. Whole-mount in situ hybridization for zebrafish ROCK orthologs rock1 and rock2a revealed enriched expression in the head at 52 hpf, in nearly identical patterns to rhoaa (Fig. 5a–d), making them attractive candidates to mediate RHOA signaling in regulating cranial blood vessel growth and integrity. To determine whether loss of Rock1/2 activity results in comparable vascular defects to rhoaa^Δ1/Δ1 mutants or animals with EC-specific expression of rhoaa^Δ1, we treated zebrafish from 24 – 52 hpf with Rockout, a cell-permeable pharmacological inhibitor of Rock1/2 kinase activity previously used in zebrafish [120,121,45]. Rockout-treated animals exhibit dose-dependent, highly penetrant cranial hemorrhage and CtA growth defects at 52 hpf (Fig. 5e–n).

Fig. 5. Pharmacological inhibition of Rhoa effectors Rock1/2 produces cranial hemorrhage and cranial angiogenesis defects.

a-d Whole-mount in situ hybridization analyses of 48 hpf wild type zebrafish embryos probed for rock1 (a, b) or rock2a (c, d) gene expression. (a, c) Lateral views, anterior to left. (b, d) Higher magnification dorsal views of the heads shown in panels a, c. e, g, I, k Stereoscope transmitted light images of DMSO-treated control sibling (e, i) or 25 μM (g) or 50 μM (k) Rock1/2 inhibitor (Rockout)-treated 50 hpf Tg(kdrl:EGFP), Tg(gata1a:DsRed) 52 hpf embryos. Black arrowheads indicate hemorrhage. f, h, j, l Confocal images of cranial vessels (EGFP-positive, green) and blood cells (DsRed-positive, magenta) in the same embryos shown in panels e, g, I, k. Arrows indicate CtA defects, white arrowheads indicate hemorrhage. m, n Quantitation of the percentage of 52 hpf Rockout-treated (light grey columns) or DMSO-treated (dark grey columns) sibling embryos with cranial hemorrhage (m) or cranial CtA defects (n). *P < 0.0001 compared to DMSO-treated controls, by Cochran-Mantel-Haenszel test. o, q, s, u Stereoscope transmitted light images of uninjected control sibling (o, s) or F0 rock1 dgRNP-(q) or rock2a/b dgRNP- (u) injected Tg(egfl7:GAL4ff), Tg(UAS:EGFP) 52 hpf embryos. Arrowheads indicate hemorrhage. p, r, t, v Confocal images of cranial vessels (EGFP-positive, green) in the same embryos shown in panels o, q, s, u. Arrows indicate CtA defects. w, x, Quantitation of the percentage of 52 hpf rock1 (w) or rock2a/b (x) dgRNP-injected (light grey columns) or uninjected (dark grey columns) sibling embryos with cranial hemorrhage, no circulation, or cranial CtA defects. * P < 0.001 compared to uninjected sibling controls, by Cochran-Mantel-Haenszel test. Mean ± SD is shown for each graph.

ROCK1 and ROCK2 share only 65% identity overall [122]. ROCK1 and ROCK2 knockout mice exhibit different causes of embryonic lethality [123,124], and studies have identified distinct functions and targets for ROCK1 and ROCK2 in certain cellular contexts [125–127]. For example, in vitro studies of cultured ECs have shown that ROCK1 knockdown leads to increased cortical actomyosin staining, thicker adherens junctions, and impaired focal adhesion complex maturation, while ROCK2 knockdown leads to increased stress fiber formation, adherens junction thinning and reduced VEGF-stimulated cord formation [15,128].

We used high-efficiency dgRNP-based CRISPR/Cas9 mutagenesis [68,69] to examine contributions of Rock1 versus Rock2 to zebrafish vascular integrity and development. We targeted both rock2 paralogs (a and b) at once to avoid possible phenotypic masking due to genetic compensation [129–131]. DNA fragment analysis revealed that each of our dgRNPs is highly effective at mutating their respective targets (the N-terminal kinase domains of rock1, rock2a, or rock2b) in F0 injected embryos (Fig. S13). Compared to uninjected control siblings, F0 embryos injected with dgRNPs targeting rock2a/b, but not rock1 exhibit significantly reduced CtA growth and increased cranial hemorrhage and lack of circulation phenotypes at 52 hpf (Fig. 5o–x). These results suggest that Rock2 is required for cranial angiogenesis, circulation, and maintenance of vascular integrity in developing zebrafish.

Discussion

In this study, we have examined the function of the small, monomeric GTPase Rhoa in regulating vascular development and integrity in vivo. We used zebrafish RHOA-ortholog (rhoaa) mutants, transgenic embryos expressing wild type, dominant-negative, or constitutively active forms of rhoaa in ECs, pharmacological inhibitors of Rhoa and its effectors Rock1/2, and Rock1 and Rock2a/b dgRNP-injected zebrafish embryos to study the in vivo developmental consequences of Rhoa gain- and loss-of-function in the vascular endothelium. Our findings document roles for Rhoa function in vascular integrity, developmental angiogenesis, and vascular morphogenesis in vivo.

Identification and characterization of novel zebrafish mutant alleles of Rhoa

We isolated and characterized two mutants carrying novel alleles of the zebrafish RHOA-ortholog rhoaa (rhoaa^Δ1 and rhoaa^R17; Fig. 1). Both mutants display partially dominant cranial hemorrhage phenotypes as heterozygotes. However, while rhoaa^Δ1/Δ1 homozygotes fail to develop circulation and exhibit both cranial vascular patterning and morphological defects, rhoaa^R17/R17 homozygotes exhibit wild type circulation, vascular patterning, and morphology. rhoaa^R71/R17 homozygotes and rhoaa^Δ1/Δ1 homozygotes both exhibit reduced T18/S19 di-phosphorylation of the RHOA/ROCK-pathway target Mlc2 (Fig. S7d,e, [112,113]), but rhoaa^Δ1/Δ1 homozygotes also exhibit somewhat reduced total levels of Mlc2 protein (Fig. S7d,e). rhoaa^Δ1 and rhoaa^R17 mutants have comparable levels of total Rhoa protein, suggesting that their phenotypic differences reflect differences in Rhoaa isoform activities, not protein levels. This idea is supported by the differential phenotypes of ECs expressing rhoaa^R17 or rhoaa^Δ1, both in vitro in primary human ECs (Fig. 2) and in vivo in zebrafish embryos (Fig. 4).

Human umbilical vein endothelial cells (HUVEC) expressing either the known dominant-negative (DN) rhoaa^N19 or rhoaa^Δ1 show essentially identical linear adherens junction morphology and VEGF-stimulated F-actin stress fiber and adhesion complex formation, while HUVEC expressing rhoaa^R17 show slightly more jagged adherens junction morphology (Fig. 2). Similarly, zebrafish treated with a pharmacological inhibitor of global Rhoa activity (Rhosin), or with transgene-driven EC-specific expression of either DN rhoaa^N19 or rhoaa^Δ1, exhibit severe cranial hemorrhage, a strong reduction in angiogenic cranial vessel growth, and detachment of trunk vessels (Fig. 4, Fig. S12), while rhoaa^R17-expressing zebrafish embryos display far less penetrant cranial hemorrhage, and no reduction in cranial vessel growth (Fig. 4). Consistent with previous analyses of DN RHOA proteins such as DN Rhoaa^N19, the Rhoaa^Δ1 and Rhoaa^R17 proteins also exhibit reduced GTP-binding (Fig. S7a,b, [84,101,102,104]). Together, these findings suggest that rhoaa^Δ1 represents a dominant-negative allele of RhoA while rhoaa^R17 is likely a weaker dominant-negative allele.

In contrast to the results with these other mutants, the punctate morphologies, TUNEL-positive staining, and nuclear fragmentation observed in ECs expressing the known constitutive-active rhoaa^V14 mutant (Fig. 3; Fig. S10, Movie S3, Movie S4) suggest that this mutant may be incompatible with endothelial cell viability when expressed in either HUVEC in vitro or in zebrafish endothelial cells in vivo.

Rhoa and vascular integrity

Previous in vitro studies have implicated RHOA/ROCK function in EC barrier disruption in response to permeability inducing factors such as vascular endothelial growth factor, thrombin, histamine, and tumor necrosis factor alpha [132–134,20,135,136,15,36,93,88,128]. RHOA/ROCK have also been shown to mediate EC barrier enhancement in response to angiopoietin 1 and sphingosine-1-phosphate [23,77,24,22,137], and RHOA activation has been observed at sites of spontaneous inter-EC gap closure, but not formation [138].

Our zebrafish studies suggest that the in vivo developing cranial vasculature is exquisitely sensitive to either increased or decreased Rhoa dosage or activity (Fig. 6). Cranial hemorrhage occurs in RHOA inhibitor-treated embryos, heterozygotes of both the stronger DN rhoaa^Δ1 mutant and the weaker rhoaa^R17 mutant (Fig. 1), as well as in transgenic embryos with EC-specific expression of rhoaa^Δ1 or rhoaa^R17 (Fig. 4). We also observe a small but significant level of hemorrhage in embryos with specific MO knockdown of only Rhoaa (Fig. S6). Embryos treated with the ROCK1/2 inhibitor Rockout (Fig. 5) also hemorrhage, indicating that decreased RHOA/ROCK function leads to disruption of vascular integrity in vivo.

Fig. 6. Summary of embryonic vascular defects generated by increased or decreased Rhoa signaling.

Schematic diagrams illustrating the range of different vascular defects found in 52 hpf zebrafish embryos with either increased, wild type, or decreased endothelial cell levels of Rhoa activity. Embryos drawn in lateral view, anterior to the left, with blood vessels in black and cranial hemorrhage in red.

Transgene-driven expression of ectopic wild type rhoaa in ECs also leads to cranial hemorrhage in zebrafish (Fig. 3) and causes adherens junction disruption in HUVEC (Fig. 2), suggesting that increased RHOA/ROCK function can also disrupt vascular integrity. Our combined results therefore suggest that a proper level of Rhoa/Rock signaling is required to maintain cranial vascular integrity in vivo under normal physiologic conditions, with either too much or too little resulting in disruption of barrier function (Fig. 6).

It remains unclear why the barrier-disrupting effects of increased or decreased RHOA/ROCK activity are restricted to the cranial vessels of zebrafish. In rhoaa mutants, EC-specific rhoaa transgene-expressing embryos, and Rock-inhibitor treated embryos, hemorrhage is observed in cranial, but not trunk vasculature beginning at 48 hpf. The trunk vasculature, does, however, exhibit other phenotypes such as impaired angiogenesis, vessel detachment, or vessel dilation (e.g., Fig. 4r–t, w–y). It is not clear what intrinsic molecular or physiological properties make the cranial EC barrier more sensitive to increased or decreased Rhoa activity and/or more prone to rupture than the trunk EC barrier at this developmental time point. Zebrafish embryos undergo a significant increase in peak blood pressure between 40 hpf and 60 hpf [139], a time when cranial vessels are still undergoing significant growth and remodeling, which could impact junctional adhesion. Trunk vessel angiogenesis (ISV growth to the DLAV) is largely complete by 40 hpf [83], so the differential sensitivity of cranial versus trunk vessels at this time could reflect differences in the timing of vessel growth, development, and maturation. Alternatively, there may be intrinsic differences in cranial versus trunk ECs that contribute to their differential sensitivity to gain or loss of Rhoa/Rock activity.

Interestingly, rhoaa mutants (rhoaa^R17/+, rhoaa^R17/R17, and rhoaa^Δ1/+ embryos) typically exhibit recovery of the EC barrier and clearance of pooled erythrocytes from the brain by 3 dpf (data not shown). This is approximately when EC barrier-protective pericytes are recruited to cranial blood vessels and the EC barrier begins to adopt a less permeable blood brain barrier (BBB) identity, characterized by increased numbers of junction proteins and reduced non-specific transcytosis events [140–144]. Developing cranial vessels may therefore be particularly sensitive to Rhoa levels only prior to pericyte recruitment and/or BBB development. Although active growth of cranial vessels also occurs at these early stages, and as discussed below increased or decreased Rhoa signaling can also disrupt cranial vessel angiogenesis, aberrant Rhoa signaling clearly can disrupt endothelial barrier function independently of its effects on angiogenesis, since we observe robust hemorrhage phenotypes in rhoaa^R17 and rhoaa^Δ1 mutant animals despite their apparently normal cranial vessel growth and patterning. Furthermore, zebrafish mutants such as plxnd1 with aberrant angiogenesis do not have increased hemorrhage [145], suggesting that aberrant or excess angiogenesis alone is not sufficient to induce barrier dysfunction.

Rhoa and developmental angiogenesis

Previous in vitro studies have shown that RHOA/ROCK is required for EC migration, proliferation, and angiogenic tip cell protrusion in response to exogenous stimulation by S1P, VEGF, and/or flow [19,132,40]. We find that strong suppression of Rhoa function in transgenic zebrafish with EC-specific expression of dominant-negative rhoaa^N19 or rhoaa^Δ1 or animals treated with ROCK1/2 inhibitor results in dramatically reduced cranial vessel angiogenesis, but has a limited effect on trunk vessel angiogenesis (Fig. 4, Fig. 5). The presumably weaker suppression of Rhoa function in rhoaa^R17 mutants, while sufficient to disrupt cranial vascular integrity (as is evident from the cranial hemorrhage in these embryos), is insufficient to disrupt cranial vessel angiogenesis. Interestingly, reduced cranial vessel angiogenesis is also noted in transgenic embryos expressing ectopic wild type rhoaa or constitutively-active rhoaa^V14 in ECs (Fig. 3), although rhoaa^V14 expression is causing EC death independently from its inhibitory effects on angiogenesis (Movie S3, Movie S4, Fig. S10). Together, these results suggest that, as for vascular integrity, cranial developmental angiogenesis is disrupted by too little or too much Rhoa activity.

The reasons for the more severe angiogenic defects in cranial vs. trunk vessels are unclear, although the later growth of cranial vessels compared to trunk intersegmental vessels could be a factor. Alternatively, the differential sensitivity of cranial vessels to angiogenesis defects in response to altered RHOA/ROCK could reflect intrinsic molecular differences between ECs in these different vascular beds.

Rhoa and vascular morphogenesis

Zebrafish with altered vascular endothelial Rhoa function also exhibit defects in vascular morphogenesis and blood vessel dilation. A previous report showed that HUVEC with excess RHOA activity have impaired vacuole formation and fail to form “lumens” in in vitro cell culture tubulogenesis assays [14]. EC-specific Rhoa knockout mice exhibit increased blood vessel lumen diameter, and EC-specific Rhoa knockout can partially rescue lumen formation in Ras interacting protein 1 (Rasip1) knockout mouse embryos, which normally exhibit increased RHOA/ROCK activity [39], supporting the idea that RHOA functions to inhibit or limit vessel lumenogenesis in vivo.

Our results are generally in line with this previously described presumptive activity for RHOA. In transgenic zebrafish with EC-specific expression of ectopic wild type rhoaa, vessel segments strongly expressing the transgene (as indicated by robust Tomato reporter fluorescence) are frequently narrow and undilated (Movie S2, Movie S3, Fig. 3). In contrast, vessel segments in transgenic zebrafish with strong mosaic EC-specific expression of dominant-negative rhoaa^N19, rhoaa^Δ1, or rhoaa^R17 isoforms often appear excessively dilated compared to vessel segments in the same animals with less, or no transgene expression (Fig. 4). Interestingly, zebrafish embryonic trunk vessels expressing the dominant-negative isoforms of rhoaa also undergo sporadic detachment (Fig. 4). This phenotype is not accompanied by hemorrhage, suggesting that it is not simply the result of generalized EC-barrier breakdown. Future studies will be needed to define the molecular basis for this phenotype.

Specificity of DN Rhoaa isoforms for Rhoa function

DN RHO mutants with preferential GDP- over GTP-binding affinities (such as RHOA^N19) are thought to competitively bind GEFs, restricting their availability to activate other RHO proteins [84,101,102,104]. Although in vitro studies have shown that some RHOA GEFs are capable of interacting with other RHO-family members when expressed in ECs, including CDC42 and RAC1 [116], our findings suggest that the phenotypes of zebrafish DN-RHOA mutants are unlikely to be a result of effects on other RHO family members. The vascular phenotypes of EC-specific Cdc42-knockout mice (occluded vessel lumens, embryo-wide severe hemorrhage) [146,39], and EC-specific Rac1-knockout mice (cardiac defects, trunk vessel angiogenesis defects) [147] do not resemble the phenotypes we observe in zebrafish embryos expressing EC-specific DN rhoaa^N19 or rhoaa^Δ1 (Fig. 4). Instead, these embryos exhibit blood vessel dilation and hypoplasia, phenotypes that do resemble previously reported vascular defects of EC-specific RHOA knockout mice [39]. Furthermore, while zebrafish have five separate RHOA orthologs, they lack direct RHOB and RHOC orthologs [117], and treatment with a RHOA/B/C pharmacological inhibitor (targeting only Rhoa proteins in zebrafish) elicits the same cranial angiogenesis, hemorrhage, and trunk blood vessel defects observed in zebrafish with EC-specific expression of DN rhoaa^N19 (Fig. S12). Furthermore, MO-based targeting of Rhoaa results in a very weakly penetrant cranial hemorrhage phenotype with normal vascular patterning (Fig. S6) that resembles the phenotype observed in our “weaker” DN rhoaa allele (R17) mutants (Fig. 1), suggesting that R17 mutants similarly have reduced Rhoa function. Together, these data suggest that the vascular phenotypes of DN rhoaa isoforms are a result of their effects on the function of Rhoa, not other RHO family members, although additional studies will be needed to further explore the parallels and potential differences between the roles of RHO family members in zebrafish and mammals.

Rock1/2 and other Rhoa effectors regulating vascular integrity and development

Previous in vitro studies have established potentially distinct roles for ROCK1 versus ROCK2 in regulating cytoskeletal dynamics and adhesion [128,148]. For example, a previous report examining HUVEC treated with siRNAs targeting ROCK1 and ROCK2 found that ROCK1 activation promotes F-actin stress fiber and focal adhesion formation, while ROCK2 activity promotes cortical actin fiber assembly and adherens junction thickening [128]. Consistent with its role in modulating the actin cytoskeleton and inter-EC junctions, this study also determined that ROCK1, but not ROCK2 depletion was sufficient to restore the EC-barrier of CCM1- or CCM2-depleted HUVEC [128]. We demonstrate that zebrafish embryos injected with dgRNPs targeting rock2a/b but not rock1 exhibit cranial hemorrhage, circulation defects, and impaired cranial central artery development (Fig. 5o–x), suggesting that Rock2 may play similar roles in regulating the EC-barrier in vivo. Further support for this idea comes from a recently published study showing that expressing dominant negative rock2 in zebrafish ECs induces cranial hemorrhage and circulation defects [149]. However, our study is the first to show that Rock2 is specifically involved in maintaining cranial vascular integrity and angiogenesis in vivo, as dominant negative ROCK2 constructs also inhibit ROCK1 function and potentially other factors [150]. Heart defects in Rock2-deficient zebrafish models have been described previously [128,151] and are likely responsible for the circulation phenotype that we observe in rock2a/b dgRNP-injected animals (Fig. 5x).

As noted in our introduction, a variety of studies suggest that RHOA can play different context-dependent and spatiotemporally regulated roles [16,15], but the molecular basis for these differential roles remains unclear. It could involve different subcellular RHOA pools acting locally on cell-cell junctions or integrin-based adhesions, or crosstalk with other GTPases (e.g., RAC1) [2,105–107]. However, the differential effects of RHOA signaling ultimately require engagement with different sets of downstream effectors. Only a limited number of direct and downstream targets of RHOA and ROCK in ECs have been previously identified (reviewed in [152,11]). Different RHOA GEFs and GAPs have been shown to regulate EC morphogenesis and adhesion [116]. Known targets of RHOA/ROCK signaling include the ERM-family proteins Ezrin, Radixin, and Moesin, and their specific kinases SLK and STK10. ERM proteins serve as membrane-cytoskeletal linkers and have previously been shown to regulate EC morphogenesis and adhesion in vitro [153–155]. Of note, ERM proteins have also been shown to suppress RHOA activity in some cellular contexts [156]. Although these are attractive genes to further explore, it is likely that additional targets will need to be uncovered to explain how RHOA signaling regulates its multitude of cellular activities.

Rhoa function in the vasculature

Together, our findings highlight roles for Rhoa function in vascular integrity, angiogenesis, and vascular morphogenesis during development (Fig. 6). Too much or too little Rhoa function impairs cranial vascular integrity and cranial developmental angiogenesis. Cranial vascular integrity appears to be more sensitive to altered Rhoa gene dosage or activity than cranial angiogenesis, which apparently tolerates more substantial deviations from normal Rhoa levels or activity. Our zebrafish results largely support previous reports that RHOA functions to inhibit or limit vessel lumenogenesis [39]. Suppressing Rhoa function in zebrafish endothelium leads to vessel dilation, while excess EC Rhoa results in undilated vessels. In addition to providing important new information on the in vivo roles of Rhoa in the vascular endothelium, our study highlights the power of the zebrafish as a model for genetic and experimental dissection of vascular biology in an intact animal.

Supplementary Material

Fig. S1–S13

Table S1

Movies S1–S5

Abbreviations and Acronyms:

EC

endothelial cell

RHOA

ras homolog gene family, member A

ROCK

Rho-associated protein kinase

CCM

cerebral cavernous malformation

hpf

hours post fertilization

ENU

N-ethyl-N-nitrosourea

SSLP

Simple sequence length polymorphism

LDA

lateral dorsal aortae

PHBC

primordial hindbrain channel

CtA

cranial central artery

DA

dorsal aorta

CV

cardinal vein

ISV

intersegmental vessel

DN

dominant negative

CA

constitutively active

BBB

blood brain barrier

HUVEC

human umbilical vein endothelial cells

GFP

green fluorescent protein

EGFP

enhanced green fluorescent protein

UAS

upstream activating sequence

2A

P2A viral cleavage peptide

Tg

transgenic

Availability of data and material

All data and materials reported in this manuscript are available from the Zebrafish International Resource Center (https://zebrafish.org) or upon request by contacting the corresponding author.