The blood flow-klf6a-tagln2 axis drives vessel pruning in zebrafish by regulating endothelial cell rearrangement and actin cytoskeleton dynamics

Recent studies have focused on capillary pruning in various organs and species. However, the way in which large-diameter vessels are pruned remains unclear. Here we show that pruning of the zebrafish caudal vein (CV) from ventral capillaries of the CV plexus in different transgenic embryos is driven by endothelial cell (EC) rearrangement, which involves EC nucleus migration, junction remodeling, and actin cytoskeleton remodeling. Further observation reveals a growing difference in blood flow velocity between the two vessels in CV pruning in zebrafish embryos. With this model, we identify the critical role of Kruppel-like factor 6a (klf6a) in CV pruning. Disruption of klf6a functioning impairs CV pruning in zebrafish. klf6a is required for EC nucleus migration, junction remodeling, and actin cytoskeleton dynamics in zebrafish embryos. Moreover, actin-related protein transgelin 2 (tagln2) is a direct downstream target of klf6a in CV pruning in zebrafish embryos. Together these results demonstrate that the klf6a-tagln2 axis regulates CV pruning by promoting EC rearrangement.


Author summary
Vascular remodeling is critical for vascular physiology and pathology. The primitive vascular plexus formed by angiogenesis, subsequently undergoes extensive vascular remodeling to establish a functionally and hierarchically branched network of blood vessels. Vascular remodeling mainly consists of vessel pruning and fusion. Endothelial cell rearrangement plays an essential role in vessel pruning, which involves endothelial cell migration and polarity. Dysfunction of flow-induced vascular remodeling will cause arteriovenous malformation and impair reperfusion of ischemia stroke. In this study, we show that the large-diameter vessel of the caudal vein is pruned from ventral capillaries of PLOS Genetics | https://doi.org/10.1371/journal.pgen.1009690 July 28, 2021 1 / 24 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Introduction
Primitive vascular beds are often formed by angiogenesis, which is defined as the sprouting of new blood vessels from existing ones [1]. To adapt to supplies of oxygen and nutrients, the primitive vascular plexus remodels itself and eventually develops into a functionally and hierarchically branched network of blood vessels. During vascular remodeling, a subset of local vessels is pruned away by active regression [2], whereas other microvessels fuse into larger diameter vessels in response to hemodynamic force [3]. Vascular pruning has been described for many vascular beds, such as the vasculature of the mouse retina and venous intersegmental vessels (vISVs) [2], midbrain vasculature [4], eye cranial division of the internal carotid artery (CrDI) [5] and subintestinal veins (SIVs) [6] of the zebrafish. It appears that pruning occurs preferentially in vascular loops with different branches. Flowinduced endothelial cell (EC) rearrangement plays an essential role in vascular pruning [2]. Flowinduced EC rearrangement involves EC migration, flow-induced EC polarity, junction remodeling [7][8][9]. Researchers have used live images to monitor vascular remodeling and simultaneously measure the velocity of blood flow [4,5,10,11]. Changes in blood flow inhibit EC rearrangement and vascular pruning in different vascular beds. Several signaling systems have been reported to regulate these processes. For example, integrin signaling is required for shear stress-induced EC migration [12]. Endoglin inhibits vascular malformation by regulating flow-induced cell migration through vascular endothelial growth factor receptor 2 (VEGFR2) signaling [13]. GTPase ras homolog family member A (RhoA) and ras-related C3 botulinum toxin substrate 1 (Rac1) are essential for shear stress-induced EC polarization [14]. Apelin signaling is involved in flowinduced endothelial polarity in zebrafish and in vitro [15]. Partition-defective 3 (PAR-3) responds to laminar flow to control endothelial polarity and vascular inflammation [16], and dynamic VEcadherin in cell-cell junctions is involved in vessel pruning [2,6]. Phosphatidyl inositol 3-kinase (PI3K) signaling prevents actomyosin contractility to regulate EC rearrangement during vascular development [17]. However, the precise role of blood flow in vessel pruning remains unclear.
In this study we used time-lapse live imaging of several transgenic reporter lines to show that the zebrafish caudal vein (CV) is remodeled from ventral capillaries of the CV plexus (CVP) by vessel pruning. Using this model, we investigated how hemodynamic forces regulate vessel pruning. Our results suggest that the magnitude of the difference in blood flow between two branches can contribute to vascular stabilization/pruning. Moreover, the results of live imaging show that junction remodeling and actin cytoskeleton dynamics are involved in CV pruning. We show that the Kruppel-like factor 6a (klf6a)-actin-related protein transgelin 2 (tagln2) axis drives CV pruning in zebrafish embryos by regulating EC rearrangement. mechanism remain unclear. To address this, we used confocal microscopy to observe CV formation in wild-type (WT) Tg(flk1: EGFP) transgenic live embryos from 32 h post fertilization (hpf) to 72 hpf ( Fig 1A). We found that CV formation is a dynamic process and that the CV is remodeled from the ventral capillaries of the CVP in zebrafish embryos. As shown in Fig 1B, the ventral capillaries of the CVP migrated horizontally and connected with each other at 32 hpf. At 36 hpf, they developed into luminal vessels with vascular loops. As a result of blood flow, these luminal vessels went through a dramatic remodeling process and developed into the CV at 60 hpf. It is interesting that there was a negative correlation between the number of vascular loops (white arrowheads) and CV formation. As shown in Fig 1C, as the number of vascular loops decreased from 8.47 at 36 hpf to 2 at 54 hpf, the ventral capillaries of the CVP eventually rearranged to establish the uniform vessel called the CV at 60 hpf. Together the remodeling of the ventral capillaries of the CVP to establish the CV is accompanied by a decrease in vascular loops in zebrafish embryos.

PLOS GENETICS
The klf6a-tagln2 axis promotes vessel pruning

Vessel pruning drives CV formation in zebrafish embryos
The gradual decrease in the number of vascular loops during CV formation in zebrafish resembles the vessel pruning described in the mouse retina [2]. During the pruning of vascular loops in the mouse retina model, one branch is stable, whereas the other is pruned by flowinduced EC rearrangement. Here we hypothesized that the gradual decrease in vascular loops during CV formation in zebrafish is regulated by vessel pruning. To validate this hypothesis, we used time-lapse live imaging to monitor the remodeling of vascular loops during CV formation in zebrafish embryos of the Tg(fli1a:nEGFP;kdrl:mCherry) line ( Fig 2B). By analyzing these time-lapse imaging sequences (14 videos), we found that stabilization/regression of competitive branches occurred during CV formation in zebrafish and that EC nucleus migration and vessel stenosis were involved in the regression of vessel branches (Fig 2A and 2B). As shown in Fig 2B, the diameters of the two branches were initially comparable at 44 hpf (stage 1 in Fig 2A). Subsequently, the lower branch became stenotic and eventually regressed, whereas the upper one remained stable.
In the lower branch, the EC nucleus (shown in blue) migrated against the direction of the blood flow from 44 to 50 hpf (stage 2 in Fig 2A), followed by a rapid stenosis of the branch at 50 hpf (stage 3 in Fig 2A). Later on, this EC nucleus (shown in blue) continued to move forward from 50 to 54 hpf (stage 4 in Fig 2A), and then it moved around to the upper branch from 54 to 72 hpf (stage 5 in Fig 2A). Meanwhile, another EC nucleus (shown in white) also moved against the blood flow from 52 to 56 hpf (stage 2 in Fig 2A). The lower branch became stenotic once again at 58 hpf (stage 3 in Fig 2A). Subsequently, the EC nucleus (shown in white) divided into two daughters at 66 hpf, and one daughter continued to move forward from 68 to 72 hpf (stage 4 in Fig 2A). At 72 hpf, the lower branch was almost regressed (stage 5 in Fig 2A).
To further observe EC migration in the regressing branch during CV formation in zebrafish embryos, we generated the knock-in (KI) fish KI(Cdh5-mRFP) using CRISPR/Cas9 technology. An mRFP sequence was inserted into the last exon (exon 12) of the cdh5 gene before the stop codon and fused into cdh5-mRFP, which was identified by PCR of its genomic DNA using target site-specific and donor-specific primers and subsequent sequencing analysis (S1A-S1C Fig). To observe the behavior of ECs in branch stenosis, we performed live imaging of Tg(fli1a:nEGFP);KI(Cdh5-mRFP) embryos. We found that both of EC nucleus migration and dynamic changes in cytoplasm contributed to vessel stenosis during CV formation in zebrafish embryos (S1D Fig). The EC nucleus (shown in white) migrated with the blood flow from 48 to 50 hpf. In contrast, the EC nucleus (shown in blue) migrated against the blood flow from 50 to 60 hpf. As the EC nuclei migration, the retraction of cytoplasmic extension can be observed.
EC polarity against the blood flow plays a pivotal role in vessel pruning [2]. To determine whether EC polarity against the blood flow is involved in the remodeling of the ventral capillaries of the CVP into the CV, we generated the transgenic zebrafish embryo line Tg(fli1a: EGFP;fli1a:B4GALT1-mCherry) following a previous report [15]. As shown in S2A Fig, the ECs of arterial intersegmental vessels (aISVs) became polarized at 48 hpf, in which the position of Golgi is relative to the nucleus against the blood flow. We then performed time-lapse live imaging to trace EC polarities within the regressing branch during CV formation in WT zebrafish embryos from 46 to 52 hpf (S2B Fig). We found that, among nine moves, 19 out of 22 venous ECs did not reveal polarities against the blood flow during zebrafish CV formation, which is consistent with the previous study [15].
Taken together, our results demonstrate that CV formation is a process of vessel pruning (hereafter, "CV pruning"). There is a growing difference in blood flow velocity between neighboring vessels in CV pruning Previous researchers presumed that the growing difference in blood flow velocity between neighboring branches determines vessel pruning [2]. To validate this hypothesis, we performed time-lapse live imaging of Tg(flk1:EGFP;gata1:dsRed) embryos to monitor dynamic changes in both blood flow velocity and diameter within neighboring branches of vascular loops in CV pruning ( Fig 2C). We found that the difference in blood flow velocity increased until one branch was pruned. At 48 hpf, the diameter of the upper and lower branches did not differ, and the blood flow velocity was comparable between the two branches ( Fig 2C-2E). Subsequently, the lower branch became stenotic as the diameter decreased and regressed at 73.5 hpf (Fig 2C and 2D). This regression was accompanied by a decrease in blood flow velocity until blood flow ceased at 56 hpf ( Fig 2C and 2E). In contrast, the upper branch became stable and showed a gradual increase in blood flow velocity in this period. These results reveal a growing difference in blood flow velocity between neighboring branches during CV pruning that may trigger vessel pruning.
To further validate the effect of blood flow on CV pruning in zebrafish embryos, we treated Taken together, these results show that reduced blood flow inhibits CV pruning in zebrafish embryos.

Klf6a is responsive to blood flow and is expressed in the CV of zebrafish embryos
The Kruppel-like factor (Klf) family of transcription factors plays important roles in regulating hemodynamic force-mediated cardiovascular homeostasis [21]. Of 24 Klf family genes in zebrafish, only klf6a was enriched in the CVP of zebrafish embryos at 36 hpf by whole-mount in situ hybridization (WISH) [22]. Therefore, we speculated that klf6a could be involved in CV pruning in zebrafish embryos. Consistent with previous reports [22,23], our WISH results showed that klf6a was indeed enriched in the zebrafish CVP region at 36 hpf compared to klf2a, klf2b, klf4, and klf6b (S4A Fig). To determine whether klf6a responds to blood flow, we performed a flow chamber experiment in which we exposed human umbilical vein ECs (HUVECs) to static (0 dyn/cm 2 ) or physiological shear stress (12 dyn/cm 2 ) for 12 h. Our arrows indicate the direction of the blood flow. Colored circles indicate ECs nuclei. 24 time-lapse live imaging were taken. Scale bar: 10 μm. (C) Time-lapse live imaging of Tg(flk1:EGFP;gata1a:dsRed) embryos shows vessel stenosis and the change of blood flow in CV pruning. Arrowheads indicate the lack of blood flow in the regressing vessel. The arrow indicates the direction of the blood flow. 10 time-lapse live imaging were taken. Scale bar: 10 μm. (D) The diameters of two branches at the indicated stages. (E) The velocity of RBCs in two branches at the indicated stages (48 hpf: S1 Video, 51 hpf: S2 Video, 53.5 hpf: S3 Video, 56 hpf: S4 Video, 58.5 hpf: S5 Video). The number of RBCs is calculated for blood flow at each stage: upper branch, 48.5 hpf: n = 5, 51 hpf: n = 10, 53.5 hpf: n = 8, 56 hpf: n = 9, 58.5 hpf: n = 10; lower branch, 48.5 hpf: n = 10, 51 hpf: n = 9, 53.5 hpf: n = 2. https://doi.org/10.1371/journal.pgen.1009690.g002

PLOS GENETICS
The klf6a-tagln2 axis promotes vessel pruning results revealed that flow shear stress (FSS) induces KLF6 expression at both the mRNA (S4B To further validate whether klf6a is expressed in the CV of zebrafish embryos, we generated a KI fish of KI(klf6a-HA-P2A-gal4);Tg(UAS:EGFP) with CRISPR/Cas9 technology. An HA-P2A-gal4 sequence was inserted into the last exon of the klf6a gene before the stop codon and fused into klf6a-HA-P2A-gal4, as described in Fig 3A. Then, the adult F 0 KI(klf6a-HA-P2A-gal4) line was crossed with a Tg(UAS:EGFP) fish. We identified later generations by PCR analyses of their genomic DNA using target site-specific and donor-specific primers and subsequent sequencing analysis (S5B and S5C Fig). Consistent with the expression of klf6a in zebrafish in our WISH results and other reports [22,23], KI(klf6a-HA-P2A-gal4) embryos showed weak but clear expression of EGFP in the vasculature of the CV (Figs 3B and S5A).
These results reveal that klf6a is expressed in the CV of zebrafish embryos and is responsive to hemodynamic force in vitro and in zebrafish embryos.

Klf6a regulates CV pruning in zebrafish embryos
To confirm the role of klf6a in zebrafish CV pruning, we injected klf6a MO into Tg(flk1:EGFP; gata1a:dsRed) embryos at the single-cell stage. We observed that knockdown of klf6a To further confirm the effect of klf6a on CV pruning in zebrafish embryos, we generated a zebrafish mutant of klf6a by CRISPR/Cas9, in which a 7 bp DNA fragment was deleted from exon 2 (S7 Fig). Compared to a sibling, ablation of klf6a increased the number of vascular loops in the CV at 72 hpf (average loops: sibling = 0.43478, klf6 -/-= 1.92308; Fig 3D). These results demonstrated that klf6a was required for CV pruning in zebrafish embryos.

Klf6a regulates EC nucleus migration in CV pruning
To analyze the effect of klf6a on the migratory behavior of ECs in zebrafish CV pruning, we performed time-lapse live imaging of Tg(fli1a:nEGFP;kdrl:mCherry) transgenic embryos to trace the migration of the EC nucleus. As shown in Fig 4A, as the lower branch of the vascular loop in the control morphant became stenotic, the nuclei of two ECs (white and yellow circles) migrated against the blood flow from 50 to 54 hpf and from 54 to 60 hpf, respectively. However, in the klf6a morphant, the EC nucleus (blue circles) migrated against the blood flow from

PLOS GENETICS
The klf6a-tagln2 axis promotes vessel pruning 51 to 56 hpf. The EC nucleus (shown in yellow) migrated in the direction of the blood flow. At 56 hpf, these two nuclei encountered each other, and they remained in the middle of the regressing vessel branch without significant migration from 56 to 59 hpf ( Fig 4A).
Statistical analysis of our results revealed that 79% of nuclei (n = 11/14 ECs) migrated against the blood flow within the regressing branch in the control MO-injected embryos ( Fig  4B). However, klf6a morphant showed caused a 54.8% decrease of nuclei (control: n = 11/14 ECs, klf6a MO: n = 5/21 ECs) migrating against the blood flow, a 23.8% increase of nuclei (control: n = 2/14 ECs, klf6a MO: n = 8/21 ECs) migrating with the blood flow, and a 31% increase of nuclei (control: n = 1/14 ECs, klf6a MO: n = 8/21 ECs) without migration (Fig 4B). Moreover, to evaluate the effect of KLF6 on EC migration in vitro, we transfected cultured HUVECs with siKLF6 and found that knockdown of KLF6 by siRNA significantly inhibited its expression at both the mRNA ( Fig 4C) and protein (Fig 4D) levels. Next, we performed a wound healing assay to evaluate the effect of KLF6 on EC migration in HUVECs. The results revealed that knockdown of KLF6 by siRNA significantly inhibited EC migration in HUVECs (Fig 4E). These results reveal that klf6a regulates EC nuclei migration during vessel pruning.
Therefore, our results demonstrate that klf6a is required for the proper migration of EC nuclei in CV pruning in zebrafish embryos.

Klf6a regulates junction remodeling and rearrangement of actin cytoskeleton in CV pruning
To further delineate the underlying mechanism of klf6a regulation of CV pruning in zebrafish embryos, we evaluated the effect of klf6a on junctions and actin cytoskeleton. We first checked the localization of KLF6 in To determine whether klf6a respond to blood flow to regulate EC rearrangement, we performed flow chamber experiment with HUVECs exposure to static (0 dyn/cm 2 ) or FSS (12 dynes/cm 2 ) for 24 hours. SiCTR-transfected HUVECs showed EC alignment in the direction of the flow in response to FSS. However, knockdown of KLF6 did not affect EC alignment in the direction of the flow, but caused excessive actin fiber stress (S8C Fig). To further investigate the role of klf6a on actin cytoskeleton in vessel pruning in vivo, we followed a previous study to generate the transgenic line Tg(Fliep:Lifeact-EGFP) [24], in which the small F-actin-binding peptide Lifeact was fused with the EGFP protein and was driven by the endothelial Fliep promoter. To monitor junction remodeling and actin cytoskeleton dynamics, we performed time-lapse live imaging of the developing CV in Tg(Fliep:Lifeact-EGFP)/KI(cdh5-RFP) embryos at three stages: the multicellular tube stage, multicellular to unicellular tube transformation, and retraction. At the multicellular tube stage, we observed the synchronous retraction of junction and F-actin as vessel became stenotic in control embryos. As shown in Fig 5A, cdh5-positive AJ (white arrowhead) moved away from its adjacent junction on the right from 0 to 8 h, accompanied by the gradual shrinkage of junctional ring. Factin co-localized with cdh5-positive AJ, and underwent a similar rearrangement as junction at the same time (white and blue arrowheads, Fig 5A). In addition, a gradual polymerization of F-actin can be observed at junction from 3 to 8 h (yellow arrowheads, Fig 5A). At the multicellular to unicellular tube transformation stage, we observed junction remodeling and F-actin polymerization/depolymerization as the branch became stenotic. As shown in S9A Fig,  junction together with F-actin moved away to the right from 0 to 8 h (white arrowhead). In the non-narrow region, F-actin depolymerized from 2 to 5 h (white arrow), whereas F-actin polymerized at junctions in the narrow region (yellow arrowhead). During retraction, the dissociation and retraction of F-actin was clearly visible (S9B Fig). However, klf6a morphants showed compromised junction remodeling and rearrangement of actin cytoskeleton, which resulted in defective branch stenosis during CV pruning ( Fig 5B). As shown in Fig 5B, knockdown of klf6a did not lead to movement or retraction of junction nor the rearrangement of actin rearrangement from 0 to 8 h, although F-actin co-localized with cdh5-positive AJ (white arrow, Fig 5B).
Taken together, these results show that junction remodeling and actin cytoskeleton dynamics are involved in CV pruning in zebrafish and that klf6a regulates both processes.

Tagln2 functions as a direct downstream target of klf6a in zebrafish embryos
The influence of transcription factor klf6a on actin cytoskeleton arrangement raised the possibility that cytoskeleton-associated genes could be involved in klf6a-mediated CV pruning in zebrafish embryos. After analyzing the available RNAseq profiles of Klf6-related transcripts in non-EC types reported by Laitman et al. [25], we found that two actin-related genes were downregulated by Klf6: activity-regulated cytoskeleton-associated protein (Arc) and transgelin 2 (Tagln2). Tagln2 expression is enriched in the vasculature of zebrafish embryos [26]. Therefore, we speculated that tagln2 could be the missing downstream target of klf6a for regulating CV pruning in zebrafish embryos. In support of this assumption, we found that knockdown of KLF6 by siKLF6 decreased TAGLN2 expression at the mRNA and protein levels in vitro ( Fig  6A and 6B). In addition, the klf6a homozygous mutant showed decreased mRNA for tagln2 in the CVP region at 36 hpf, as determined by WISH (Fig 6C).
To confirm whether tagln2 is a direct target of klf6a in zebrafish, we performed a bioinformatics analysis to predict the potential sequence within the zebrafish tagln2 promoter to which transcription factor Klf6a can bind. As shown in Fig 6D, three CACCC sequences to which Klf6a can bind are located in the -1977 to -1643 bp region of the tagln2 promoter. We then performed chromatin immunoprecipitation (ChIP) assay on the zebrafish embryos to validate this. The results revealed that Klf6a can bind to this promoter region within the tagln2 gene but not its exon 1 (Fig 6E).
These results demonstrate that tagln2 is the direct downstream target of klf6a in zebrafish, which suggests that tagln2 could be involved in klf6a-mediated CV pruning.

Tagln2 regulates CV pruning in zebrafish by promoting EC nucleus migration
To determine the role of tagln2 in zebrafish CV pruning, we first validated the efficiency of tagln2 MO using a tagln2-EGFP fusion protein (S10A Fig), the expression of which was clearly inhibited by tagln2 MO. This MO was then used to knockdown the tagln2 gene in embryos to evaluate the effect on CV pruning at 72 hpf (S10B and S10C Fig). Statistical analysis clearly showed that the disruption of tagln2 by MO increased the number of vascular loops in the CV at 72 hpf (average loops: control = 0.6, tagln2 MO = 1.4615; S10C and S10D Fig). To further validate the role of tagln2 in CV pruning, we then generated the tagln2 mutant in zebrafish using CRISPR/cas9 technology, in which a 7 bp fragment was deleted from exon 1 (S11 Fig). Compared to the WT sibling, tagln2 homozygous mutations showed a significant increase in the number of vascular loops at the CV at 72 hpf (average loops: sibiling = 0.44, tagln2 -/-= 1.81818; Fig 7A and 7B).
Next, we analyzed the effect of tagln2 on EC behavior using time-lapse live imaging of Tg(fli1a:nEGFP;kdrl:mCherry) embryos. In control MO-injected embryos, EC nucleus migration against the blood flow (white circle) could be readily observed within regressing vessel branches from 48 to 54 hpf ( Fig 7C). However, in the tagln2 morphant, EC nuclei (white circle) showed no obvious movement from 48 to 58 hpf ( Fig 7C). Knockdown of tagln2 by MO caused a 38% decrease in EC nuclei (control: n = 14/24 ECs, tagln2 MO: n = 3/15 ECs) migration against the blood flow, a 3% increase in EC nuclei (control: n = 4/ 24 ECs, tagln2 MO: n = 3/15 ECs) migration with the blood flow, and a 35% increase in ECs nuclei (control: n = 6/24 ECs, tagln2 MO: n = 9/15 ECs) without migration (Fig 7D). These results strongly support our proposition that tagln2 is required for CV pruning by regulating EC nucleus migration. To evaluate the effect of TAGLN2 on EC migration in vitro, we used siTAGLN2 to silence TAGLN2 in cultured HUVECs and found that TAGLN2 expression was significantly inhibited at both the mRNA ( Fig 7E) and protein (Fig 7D) levels. The results of a wound healing assay performed to evaluate the effect of TAGLN2 on EC migration in HUVECs revealed that knockdown of TAGLN2 by siRNA significantly inhibited EC migration in HUVECs (Fig 7G).
Taken together, our results demonstrated that tagln2 acts as a downstream target of klf6a and regulates CV pruning in zebrafish by promoting EC nucleus migration.

PLOS GENETICS
The klf6a-tagln2 axis promotes vessel pruning

PLOS GENETICS
The klf6a-tagln2 axis promotes vessel pruning

Tagln2 regulates junction remodeling and the arrangement of actin cytoskeleton in CV pruning
To further understand the role of tagln2 in actin cytoskeleton, we performed immunofluorescence staining of confluent HUVECs in vitro with anti-TAGLN2 and found that TAGLN2 colocalized with F-actin at a significant level (S12A Fig). This indicated a potential role of TAGLN2 in actin cytoskeleton arrangement. To evaluate the effect of TAGLN2 on actin cytoskeleton in vitro, we then performed an immunofluorescence assay on cultured HUVECs with anti-VE-cadherin antibody and phalloidin. We observed actin filament bundles closely flanking VE-cadherin-positive AJs in confluent siCTR-transfected HUVECs (S12B Fig). Knockdown of TAGLN2 by siTAGLN2 did not disrupt actin filament bundles closely flanking linear AJs but caused excessive actin fiber stress (S12B Fig). To further investigate the role of tagln2 on actin cytoskeleton in vessel pruning in vivo, we performed time-lapse live imaging of the developing CV in Tg(Fliep:Lifeact-EGFP)/KI (cdh5-RFP) embryos. We found that knockdown of tagln2 by tagln2-MO significantly inhibited junction remodeling and actin cytoskeleton dynamics in zebrafish CV pruning (Fig 8). As shown in Fig 8A, control embryo showed a gradual detachment of junctions (blue arrowheads) from 0 to 9 h, and exhibited an opposite direction of junction movement (white arrowheads) from 0 to 3 h as multicellular tube became stenotic. F-actin formed at junction, and underwent similar rearrangement as junction remodeling (white and blue arrowheads, Fig 8A). However, knockdown of tagln2 resulted in compromised junction remodeling and actin cytoskeleton dynamics during CV pruning (Fig 8B). In tagln2 morphants, the junction moved to the right of adjacent junction, and the two junctions came into contact with each other from 0 to 5 h (white arrowhead). However, F-actin located at junction, we can observe a depolymerization of F-actin at the other junction from 0 to 8 h (white arrows).
Taken together, these findings show that tagln2 regulates junction remodeling and rearrangement of actin cytoskeleton in vessel pruning, thereby promoting CV formation in zebrafish embryos.

Discussion
Vessel pruning has been well described in different vascular beds of zebrafish, such as the hindbrain vasculature [4], CrDI in the eye [5], SIVs [6], and vISVs [2]. In this study, compelling evidence from our live imaging showed that the large-diameter vessel of the CV is pruned from ventral capillaries of the CVP in zebrafish embryos characterized by vessel stabilization/ regression and EC rearrangement within the regressing branch.
Vessel pruning, which is essential for the creation of an ordered and hierarchical vascular network, eventually results in a ramified vasculature [1,7]. Flow-dependent endothelial rearrangement drives vessel pruning in a mouse retina model [2] and in various vascular beds of zebrafish embryos [27]. However, it is unclear how EC rearrangement causes vessel stenosis within the regressing vessel. By tracing EC behavior with time-lapse imaging of Tg(fli1a: nEGFP;kdrl:mCherry) zebrafish embryos, we found that EC nucleus migration against the direction of the blood flow clearly precedes vessel stenosis (Fig 2B). Moreover, we observed junction remodeling and actin cytoskeleton dynamics as vessel stenosis in Tg(Fliep:Lifeact-EGFP)/KI(cdh5-RFP) embryos (Figs 5A, 8A and S9). These results indicate that both EC nucleus migration, junction remodeling and actin cytoskeleton dynamics contribute to vessel stenosis during CV pruning in zebrafish embryos.
Several groups of researchers have performed imaging of blood flow and vascular remodeling in different vascular beds and analyzed the relationship between them. Reductions in blood flow inhibit vessel pruning [4,5,10,11]. However, the precise role of blood flow in vessel pruning remains unclear. By tracing EC behavior with time-lapse imaging of Tg(fli1a:EGFP; gata1:dsRed) embryos, we found that vessel stenosis precedes no perfusion in the regressing vessel ( Fig 2C). Further, the graded difference in blood flow between the two branches increases as one of the branches becomes stenotic (Fig 2C-2E). Reduced blood flow in embryos treated with tricaine and tnnt2a MO resulted in compromised CV pruning (S3 Fig). Differences in pressure between branches can be used to predict the location of sprouting, in which sprouts migrate form from vessels at lower pressure toward vessels at higher pressure [11]. In future studies, it could be useful to predict which branch is stable or pruned, based on flow dynamics.
Actin cytoskeleton plays important roles in angiogenic sprouting [28], aortic lumen expansion [29], and flow-induced vascular dilation and remodeling [30]. In addition, FSS induces actin cytoskeleton remodeling in vitro [31]. Our time-lapse live imaging of Tg(Fliep:Lifeact-EGFP)/KI(cdh5-RFP) embryos showed junctional detachment and retraction in vessel pruning (Figs 5A, 8A and S9), which resembles anastomosis in reverse [32]. In anastomosis of DLAVs, F-actin forms at junctions [24]. In our study, we found that F-actin forms at junction, and go through similar rearrangement as junction remodeling during CV pruning. Furthermore, Factin depolymerization was found in the non-narrow region at multicellular to unicellular tube transformation stage, whereas F-actin polymerization at junctions seems to be preferred in the narrow region as branch stenosis (S9 Fig). These results suggest that actin cytoskeleton is essential for flow-induced vessel pruning.
Our strong results support the important role of klf6a in CV pruning in zebrafish embryos. First, results from WISH (S4 Fig) and the KI(klf6a-HA-P2A-gal4) fish (Figs 3B and S5A) clearly confirm that klf6a is expressed in the ECs of the CV in zebrafish embryos, consistent with a previous report [23]. Second, klf6a is responsive to hemodynamic force in vitro and in vivo (S4C and S4D Fig). Third, a disruption of klf6a function impairs EC nucleus migration against the blood flow within the regressing branch (Fig 4A and 4B) and results in compromised CV pruning (Figs 3C, 3D and S6). Furthermore, klf6a regulates junction remodeling and actin cytoskeleton dynamics in vivo (Fig 5).
Researchers previously identified a crucial role of TAGLN2 in regulating actin cytoskeleton assembly [33][34][35]. Actin-binding protein tagln2 functions as a direct downstream target of klf6a to regulate CV pruning in zebrafish (Figs 6 and 7). First, klf6a regulates tagln2 expression in the zebrafish CVP region and in vitro (Fig 6A and 6C). Second, Klf6a can bind to the tagln2 promoter (Fig 6D and 6E). Third, the tagln2 morphant and mutants show defective CV pruning (Figs 7A, 7B and S10). Moreover, our results show that a disruption in tagln2 results in defective EC nucleus migration, junction remodeling, and actin cytoskeleton dynamics in vivo (Figs 7C,  7D and 8). These results mimic the phenotype observed in the klf6a morphant or mutants.
In conclusion, our results demonstrate that the klf6a-tagln2 axis promotes vessel pruning in CV formation in zebrafish by regulating EC rearrangement and actin cytoskeleton dynamics.

Ethics statement
All zebrafish experimentation was approved by the Ethics Committee of Chongqing University Affiliated Cancer Hospital. The ethical review number: CZLS2021172-A. the adjacent one (white arrowheads), and F-actin undergoes similar rearrangement within it (white arrowheads). However, depolymerization of F-actin is also observed at junctions (white arrowheads). Six vascular loops are taken time-lapse live imaging. Scale bar: 20 μm. https://doi.org/10.1371/journal.pgen.1009690.g008
Tg(fli1a:B4GALT1-mCherry) zebrafish was generated as described previously [15]. Briefly, embryos at the single-cell stage were co-injected with pTol fli1a:B4GALT1-mCherry plasmid (25 ng) and Tol2 transposase mRNA (25 ng). Embryos showing red fluorescence were selected and grown to adulthood. The founder was identified by specific expression of B4GALT1-m-Cherry in the blood vessels of their progeny.
Tg(Fliep:Lifeact-EGFP) zebrafish was generated as described previously [24]. Briefly, embryos at the single-cell stage were co-injected with pTol fliep:Lifeact-EGFP plasmid (25 ng) and Tol2 transposase mRNA (25 ng). Embryos showing green fluorescence were selected and grown to adulthood. The founder was identified by specific expression of Lifeact-EGFP in the blood vessels of their progeny.

Microinjection of MO oligonucleotides
All MO were ordered from Gene Tools LIC. 4 ng antisense oligonucleotide (MO) was injected into embryos at the single-cell stage. The MO sequences are listed in S1 Table. The efficiency of the tagln2 MO was validated by the pcDNA3.1-tagln2-EGFP plasmid, which consists of the sequence complementary to the tagln2 MO.

Generation of the zebrafish klf6a/tagln2 mutant and KI by CRISPR/Cas9
The generation of zebrafish mutants and KI was performed with CRISPR/Cas9 by Nanjing XinJia Medical Technology (China). Briefly, the klf6a short guide RNA (sgRNA) was targeted to exon 2, and the target sequence was GAATTCGGATGCCAGCAGCGAGG. The tagln2 sgRNA was targeted to exon 1, and the target sequence was CACCTCGCGACTCAGACCG-TAGG. SgRNA was synthesized following the manufacturer's protocol (NEB, E2050s). Cas9 mRNA was synthesized with Ambion mMESSAGE mMACHINE mRNA transcription synthesis kits. Cas9 mRNA (600 ng/μL) and klf6a gRNA or tagln2 gRNA (300 ng/μL) were coinjected into WT embryos at the single-cell stage. The embryos were raised to adulthood, and the founder (F 0 ) mutants were identified by genomic PCR, followed by sequencing. To confirm germline-transmitted mutations of both the klf6a and tagln2 genes, we outcrossed adult F 0 founders with WT AB fish to obtain heterozygous mutant zebrafish (F 1 ). Adult F 1 mutants were validated by genomic PCR and sequencing. The same heterozygous mutants of F 1 were in-crossed to generate homozygous mutant embryos (F 2 ) and used for further experimental analyses. The primers used to genotype the klf6a and tagln2 mutants are listed in S2 Table. The KI sgRNA was targeted to the last exon of the zebrafish klf6a. The klf6a reporter donor plasmid klf6a-HA-P2A-gal4 consists of three parts: a left arm, a klf6a-HA-P2A-gal4 coding sequence, and a right arm. The left arm goes from the 5 0 side of the sgRNA to the end of exon 4. The right arm includes the stop codon and 3 0 regulatory elements of klf6a. The sgRNA, cas9 mRNA, and donor plasmid were co-injected into zebrafish embryos at the single-cell stage. The embryos were raised to adulthood, and the founder (F 0 ) transgenic lines were identified by genomic PCR, followed by sequencing. To confirm the germline-transmitted transgenic line, adult F 0 founders were outcrossed with Tg(UAS:EGFP) fish to obtain the F 1 transgenic line, followed by identification with genomic PCR and green fluorescence. The F 1 transgenic line were outcrossed to generate F 2 embryos for further experimental analyses. Target side-primer (F2) and donor-primers-(gal4-R2) were used in PCR to identify the successful KI zebrafish. The primers used to identify KI(klf6a-HA-P2A-gal4) are listed in S2 Table. To generate a KI fish of KI(Cdh5-mRFP) line, the mRFP sequence is inserted into the last exon (exon 12) of zebrafish cdh5 gene before the stop code and fused into cdh5-mRFP. Briefly, cdh5-sgRNA, cas9 mRNA and the donor plasmid were co-injected into zebrafish embryos at the single-cell stage. The embryos were raised to adulthood, and the founder (F 0 ) transgenic lines were identified by genomic PCR of target site-primer (chd5-F4) and donor-specific primer (mRFP-5-R) and subsequent analysis of sequencing. The primers used to identify KI (cdh5-mRFP) are listed in S2 Table.

Tricaine treatment
Tricaine was dissolved in ddH 2 O at a concentration of 4 mg/mL, and the pH was adjusted to 7.2 ± 0.2 for storage. Tricaine was added into the egg water to a final concentration of 0.06 mg/ mL at 30 hpf until imaged at 72 hpf.

Time-lapse live imaging of zebrafish embryos
All confocal imaging was performed with a Leica SP8 confocal microscope at 28.5˚C. Timelapse live imaging of zebrafish embryos was performed as described previously [36]. Briefly, to prevent the formation of pigment, embryos were treated with 0.003% 1-phenyl-2-thiourea (PTU; Sigma) at 24 hpf. For time-lapse live imaging, the zebrafish embryos were immobilized with 1% low-melt agarose (Sigma) without tricaine treatment. Stacks were taken every 2 or 2.5 h with a step size of 1.5 μm.

Calculation of the velocity of RBCs
The velocity of the blood flow was calculated as described previously [41]. Briefly, a Leica SP8 high-speed module was used to record videos of RBCs with the setting 512 × 200 (one stack was completed in 29 ms). We calculated the blood flow velocity in the CV manually using Ima-geJ as follows: For each video, we calculated the movement of three or four different RBCs, taking systole and diastole into account. Only clear and stable RBC movements in the CV area were converted from video into images and analyzed with ImageJ. We converted pixel units into known distance (pixels/um) using the scale bar provided by LAS X and ImageJ for conversion. The travel distance of single RBC was calculated as displacement distance (um) as a function of time (ms). The average blood flow velocity was calculated from the displacement and the time data.

WISH
WISH of zebrafish embryos was performed as described previously [41]. Images were captured by microscope (Zeiss Stemi 2000-C). The probe primers are listed in S2 Table.

ChIP assay
ChIP assay was performed with a Cell Signal Technology (S9004) kit instruction book. As there is no specific antibody of Klf6a in zebrafish for the ChIP assay, we used the Klf6a overexpression system, consistent with a previous report [23]. Klf6a-myc mRNA was injected into zebrafish embryos at the single-cell stage, and then the embryos were collected for ChIP assay at 36 hpf. The klf6a-myc plasmid was a gift from Professor Feng Liu at the Institute of Zoology, CAS of China. The Myc antibody was used to IP the sequence of tagln2 targeting by Klf6a. IgG was used as a negative control. The eluted DNA (precipitated by the Myc or IgG antibody) was assayed by PCR. The primers were designed according to the klf6a-CACCC binding sites. Nonspecific primers were used as negative controls. The primers are summarized in S2 Table. Cell culture, immunofluorescence staining, and Western blotting Primary HUVECs were obtained from Science Research Laboratories and cultured with EC medium (ECM) and 10% fetal bovine serum (FBS). All antibodies used in this study are listed in S3 Table. KLF6 siRNA is ordered from Santa Cruz (sc-38021). The Tagln2 siRNA sequence is 5 0 -CUCUGUGUCCUCCGUUCAUTT-3 0 and 5 0 -AUGAACGGAGGACACAGAGTT-3 0 (Shanghai GenePharma). Immunofluorescence staining and western blotting were performed as described previously [41].

Quantitative RT-PCR analysis
Quantitative real-time PCR (qPCR) was performed with TB Green Fast qPCR Mix (RR430A; Takara). The qPCR primers are listed in S2 Table.

Flow chamber experiment
The flow chamber experiment was performed as described previously [41]. A custom-built flow chamber consisting of two parallel plates made of polymethyl-methacrylate was used to apply uniform shear stress on HUVEC monolayers. The lower plate was flat, and a rectangular channel was engraved on the upper plate with an automatic milling machine. HUVECs were grown to confluence on a patch of polylysine slide coated with fibronectin. The patch was then placed on the lower plate, and the upper plate was mounted to form a sealed channel of parallel-plate geometry.
The actual shear stress (t) applied to the cells can be expressed in terms of volumetric flow rate (Q), medium viscosity (μ; 1.3 × 10 2 Pa ▪ s), and width (w; 25 mm) and height (h; 0.25 mm) of the channel: τ = Qμ/wh 2 . In our setup, flow rates of 15 mL/min were applied to obtain shear stress values of 12 dyn/cm 2 . The flow rate was controlled with a peristaltic roller pump.

Statistical analyses
To quantify defective CV formation, we quantified seven segments of each embryo. All statistical analyses were done with GraphPad Prism 5. The statistical significance of the difference between control and experimental groups was determined with Student's unpaired two-tailed t test or unpaired two-tailed chi-square test. Data are presented as means ± SEM. P > 0.05 was considered non-significant. � P < 0.05, �� P < 0.01, and ��� P < 0.001, as shown in figures and figure legends.   for the klf6a or tagln2 mutant, KI(klf6a-HA-P2A-gal4)