Temporospatial inhibition of Erk signaling is required for lymphatic valve formation

Intraluminal lymphatic valves (LVs) and lymphovenous valves (LVVs) are critical to ensure the unidirectional flow of lymphatic fluid. Morphological abnormalities in these valves always cause lymph or blood reflux, and result in lymphedema. However, the underlying molecular mechanism of valve development remains poorly understood. We here report the implication of Efnb2-Ephb4-Rasa1 regulated Erk signaling axis in lymphatic valve development with identification of two new valve structures. Dynamic monitoring of phospho-Erk activity indicated that Erk signaling is spatiotemporally inhibited in some lymphatic endothelial cells (LECs) during the valve cell specification. Inhibition of Erk signaling via simultaneous depletion of zygotic erk1 and erk2 or treatment with MEK inhibitor selumetinib causes lymphatic vessel hypoplasia and lymphatic valve hyperplasia, suggesting opposite roles of Erk signaling during these two processes. ephb4b mutants, efnb2a;efnb2b or rasa1a;rasa1b double mutants all have defective LVs and LVVs and exhibit blood reflux into lymphatic vessels with an edema phenotype. Importantly, the valve defects in ephb4b or rasa1a;rasa1b mutants are mitigated with high-level gata2 expression in the presence of MEK inhibitors. Therefore, Efnb2-Ephb4 signaling acts to suppress Erk activation in valve-forming cells to promote valve specification upstream of Rasa1. Not only do our findings reveal a molecular mechanism of lymphatic valve formation, but also provide a basis for the treatment of lymphatic disorders.


INTRODUCTION
7][8] The expression levels of the essential transcription factors for valve development, including GATA2, FOXC2, and PROX1, are elevated in valve-forming LECs following the initiation of lymphatic fluid flow. 2,9,10Lymphatic vessels have one-way valves, including intraluminal lymphatic valves (LVs) and lymphovenous valves (LVVs), to ensure the unidirectional flow of lymphatic fluid.1][12] By revealing a bicuspid valve structure similar to that found in mammals, a recent study provided compelling evidence for the existence of LVs and LVVs in zebrafish facial lymphatic vessels (FLVs). 13,14However, it is unknown whether other valve structures exist in the lymphatic system of zebrafish.
6][17] The MAPK/Erk pathway is primarily initiated by GTP-bound Ras activating Raf.Raf phosphorylates Mek, which in turn phosphorylates Erks, the main components of the Ras-Raf-Mek-Erk signaling cascade. 18,191][22] Numerous studies have shown that the Erk pathway is vital to lymphatic development.Zebrafish embryos lacking Vegfr3 fail to initiate lymphatic vessel sprouting or differentiation. 23Transgenic expression of RAF1 S259A , a gain-offunction RAF1 mutant associated with human Noonan syndrome, in the endothelial cells of mouse embryos activates Erk, leading to enhanced commitment of venous ECs to the lymphatic fate and eventual lymphangiectasia. 24Overexpression of Ras in the endothelial cell lineage of mice also causes lymphatic vessel hyperplasia. 25Additionally, treatment with MEK inhibitors alleviates lymphatic anomaly-associated clinical symptoms in a patient carrying an ARAF S214P gain-of-function mutation and prevents increased LEC sprouting in human primary dermal lymphatic endothelial cells (HDLECs) or zebrafish transgenic larvae that overexpress ARAF S214P, . 26Although the involvement of Erk signaling in the formation of lymphatic vessels is well understood, less is known about its function in the generation of lymphatic valves.One report suggested that termination of Vegfr3 signaling by Epsin1/2 in collecting lymphatic trunks is required for normal LV development in mice, but the underlying mechanism has not been investigated.
Here, we find two more types of valve structures in the zebrafish, helping to gain a deeper understanding of the lymphatic system in this vertebrate animal model.Significantly, we disclose a specific regulation of Erk signaling during lymphatic valve development.Although Erk signaling is involved in LEC development, we show that Erk inhibition is necessary for zebrafish valve cell development.Erk activity must be repressed in the FCLV to initiate the valve-forming cell program.Then, further inhibition of Erk signaling by Efnb2-Ephb4 signaling upstream of Rasa1 appears essential for the specification of valve-forming cells.In addition, the dynamic expression pattern of Efnb2 in valve-forming LECs at the vessel bifurcation likely accounts for the restricted repression of Erk activity.

RESULTS
Identification of new valves in the zebrafish lymphatic system Although intact lymphatic vessels were imaged in the zebrafish some time ago, [49][50][51][52] two pairs of bicuspid valves, including one pair of lymphatic valves (LVs) and another pair of lymphovenous valves (LVVs), have only been found in zebrafish facial lymphatic vessels (FLVs) more recently. 13However, whether there exist other specialized valve structures is unknown.
4][55] However, continuous observation in Tg(lyve1b:TopazYFP) transgenic fish, which labels lymphatic and venous vessels, revealed that the FLS did not disappear after 5 dpf, but continued to grow and connected with both the otolithic lymphatic vessel (OLV), the trunk superficial lateral lymphatic vessels (LLs), and lymphatics around kidney (KLs) (Fig. 1a, Supplementary Fig. 1a and Supplementary Movie 1, 2).As the FLS exhibits morphological changes during lymphatic development, we named this structure the remaining FLS (RFLS) after 3 dpf.Given that the FLS/RFLS contributes to FLVs and connects with the CCV, we assumed that there may exist an LVV between the RFLS/FLS and the CCV.By carefully observing Tg(gata2a:EGF-P;lyve1b:DsRed2) transgenic fish, 56 we noted that some cells located at the junction between the RFLS and CCV expressed gata2a:EGFP fluorescent protein from 72 hpf onward and became much more condensed by 77 hpf and 5 dpf (Fig. 1b and Supplementary Fig. 1b).By detecting Prox1a expression, we found that, compared to the moderate level of Prox1a in the LECs, these cells displayed relatively high levels of Prox1a (Prox1a hi ) expression at 6 dpf (Fig. 1c and Supplementary Fig. 1c), which is a characteristic of valve cells. 9,10Using transmission electron microscopy (TEM), we further confirmed the existence of a structure that prevented direct connection of RFLS to CCV (Fig. 1d).When observing the dynamic behaviors of this presumed valve structure, we found that the leaflet motion of RFLS-CCV LVVs was similar to that of FCLV-PHS LVVs discovered previously at the junction between facial collecting lymphatic vessels (FCLVs) and the primary head sinus (PHS) 13 (Supplementary Movie 3).Therefore, we conclude that, in addition to the FCLV-PHS LVVs, another pair of LVVs is present at the junction of the RFLS and CCV, referred to as RFLS-CCV LVVs in the head region (Fig. 1j).
Consistent with the findings in the resting glass catfish (Kryptopterus bicirrhis), 57 we found that blood from the caudal fin indeed flowed through the caudal heart (CH), an interesting structure located under the tail notochord, 57,58 into the caudal vein (CV) in the zebrafish (Supplementary Movie 4).And we could clearly observe that red blood cells labeled with epb41b:mCherry flowed through the tail lymphatic vessels into the CV by passing the CH in some Tg(epb41b:mCherry) fish and this would be much more obvious after stress or handling of fish (Fig. 1e and Supplementary Movie 4).However, to our surprise, as observed in different transgenic backgrounds, the caudal heart structure was fli1a:EGFP, labeling endothelial cells, and lyve1b:TopazYFP positive, but flk:mCherry, labeling blood endothelial cells only, negative.Thus, that structure could be described as a lymphaticlike structure, which connects the tail lymphatic vessels with both the caudal vein (CV) and the trunk lymphatic vessels, and might receive blood from tail structure under stress conditions (Fig. 1f, g and Supplementary Movie 5, 6).By injecting 2000 kDa Dextran-Rhodamine into the tail lymphatic vessels to imitate the lymph, we further observed direct link between the CH and CV (Supplementary Movie 7).We performed retro-orbital injection of 2000 kDa Dextran-Fluorescein to label the circulation and found the existence of caudal heart in adult fish (Supplementary Fig. 1d).The caudal heart of adult fish also uniquely expressed lymphatic marker lyve1b:TopazYFP (or lyve1b:DsRed2), but not blood vascular marker flk:mCherry (or flk:EGFP), indicating CH is a lymphatic-like structure (Supplementary Fig. 1e).Moreover, a number of bicuspid valve structures labeled with gata2a:EGFP were discovered in the CH region, including several LVs in CH and a specific LVV at the junction between the CH and CV, referred to as the CH-LVs and CH-CV LVV, respectively (Fig. 1h and Supplementary Movie 8, 9).We also confirmed the valve structure of CH-CV LVV by transmission election microscopy (Fig. 1i and Supplementary Fig. 1f).Taken together, these data provide the first evidence of valve structures in the trunk region of zebrafish (Fig. 1j).

Specific repression of Erk activity in FCLV during valve cell specification
In Tg(gata2a:EGFP;lyve1b:DsRed2) fish, we observed a specific accumulation of cells with high-level expression of gata2a:EGFP in the FCLV but not the LFL, and then at the valve-forming region around 3 dpf (LVV at 72 hpf and LV at 77 hpf), which develops into the bicuspid valve structure over the next 2 days (Fig. 2a).Prox1a hi expression at different stages validated lymphatic valve development.The FCLV accumulated LECs with Prox1a hi expression about 3 dpf ( ~77 hpf) at valve sites, indicating valve-forming LEC specification (Fig. 2b).
Next, Tg(lyve1b:Kaede) transgenic fish were used to track valveforming LEC emergence and lineage specification in LVs using photoconvertible labeling and in vivo time-lapse photography.Photoconverting LECs in the FCLV and LFL at 3 dpf showed that most FCLV-LV cells at 5 dpf came from the FCLV (Fig. 2c).Although a small proportion of LFL LECs near the FCLV contributed to the valve structure, LFL LECs slightly distant from the FCLV did not differentiate into valve-forming LECs (Fig. 2c), which is consistent with previous reports that valve-forming LECs derive from lymphatic vessels in anatomical proximity to the LV structure. 10,59,60Thus, valve-forming LECs are mostly FCLV-derived at 3 dpf.
Vegfc-Vegfr3-mediated activation of Erk signaling is vital to lymphatic endothelial cell (LEC) specification and migration. 23,24owever, it has not been described how Erk signaling is dynamically controlled throughout lymphatic development, particularly in valve specification.As previously reported, the ERK-KTR (ERK kinase translocation reporter) biosensor could be used to determine Erk signaling activity by measuring the cytoplasmic/ nuclear (C/N) signal ratio. 61Though ERK-KTR system has been used in zebrafish, [62][63][64] we also generated a transgenic line Tg(ef1a:ERKKTR) that expresses the ERK-KTR biosensor ubiquitously to evaluate its responsiveness in our hands.Erk activity was found to be higher in most cells in the embryonic margin (Supplementary Fig. 2a), which is similar with the distribution of p-Erk reported previously. 65When embryos were incubated in the MEK inhibitor selumetinib from sphere stage onwards, this unique pattern of Erk activity disappeared (Supplementary Fig. 2a), shortening the anterior-posterior axis 66 and disrupting the somite boundary formation 63 (Supplementary Fig. 2b).Therefore, the Erk activity can be monitored by the ERK-KTR biosensor.
We created another ERK-KTR biosensor transgenic line, Tg(ga-ta2aECE:ERKKTR), driven by gata2a endothelial core element (ECE), 13 to measure LEC Erk activity at the single cell level with high temporal resolution.By visualizing the ERK-KTR biosensor with nuclear marker Tg(gata2aECE:nls-mCherry), we spatiotemporally evaluated LEC ERK-KTR C/N ratios.Using two distinct statistical approaches, we consistently observed a slight decrease in Erk activity in both FCLV and aLFL LECs from 52 to 72 hpf.However, a distinct reduction in Erk activity was only evident in FCLV LECs from 72 to 77 hpf during valve development, whereas it remained relatively unchanged in aLFL LECs (Fig. 2d, e, Supplementary Fig. 2c-f, and Supplementary Movie 10).We further used selumetinib to inhibit MEK activity from 56 hpf to test the responsiveness of this reporter line in lymphatic vessels.After inhibition, aLFL Erk activity and p-Erk protein level were found to be significantly reduced (Supplementary Fig. 2g-i).Furthermore, we applied time-lapse imaging on Erk reporter fish from 72 to 78 hpf.The results showed a more consistent increase of nuclear fluorescence intensity and corresponding decrease of ERKKTR intensity ratio of cytoplasmic to nuclear (C/N) fluorescence in FCLV LECs than in aLFL LECs (Supplementary Fig. 3a-c, and Supplementary Movie 11), indicating a decrease of Erk activity in FCLV LECs.This observation was supported by p-Erk immunostaining, which revealed that the p-Erk signal was clearly suppressed in the FCLV region compared to the aLFL region and decreased further in FCLV LECs at 77 hpf (Supplementary Fig. 4).We hypothesize that FCLV Erk suppression is necessary for the initiation of gata2amarked valve-forming LEC program and subsequently valve formation.
Depleting rasa1 impairs valve-forming cell specification Rasa1 is a Ras GTPase-activating protein (RasGAP) that accelerates conversion of Ras protein from an active GTP-bound state back to an inactive GDP-bound state, hence acting as a negative regulator of Erk signaling. 46It has been shown that loss-of-function Rasa1 mutations in mouse embryos likewise result in impaired valve development, and administration of ERK inhibitor AZD6244 partially rescues LVV development in the absence of RASA1 in mice. 45,47However, it is yet uncertain if Erk signaling has to be suppressed by Rasa1 during the process of valve formation.
Since valve-forming cells in the FCLV-LV are largely derived from the FCLV LECs, it is clear that FCLV's developmental status is critical for valve-forming cell formation in FCLV-LV.So, at 3 dpf when FCLV had formed but valve-forming LEC specification was about to begin, we applied selumetinib and found enhanced ectopic gata2a:EGFP expression in the FCLV surrounding FCLV-LV (Fig. 5k).This observation is consistent with ectopic gata2a:EGFP expression in erk1 −/− ;erk2 −/− mutant embryos (Supplementary Fig. 7g).Inhibition of Erk activity by selumetinib could not induce high or ectopic gata2a:EGFP expression in trunk blood or lymphatic vessels (Supplementary Fig. 8a, b).Thus, in erk1 −/− ;erk2 −/− double mutants or embryos treated with selumetinib from 2 dpf onwards, the bimodal impact on FCLV-LV development was likely caused by dynamic alterations in Erk signaling during lymphatic development, resulting in some embryos with deficient FLVs having valve hypoplasia and others with normal FLV formation having valve hyperplasia.
Efnb2-Ephb4 signaling is essential for lymphatic valve formation Since mutations in the human EFNB2-EPHB4 cassette are linked to both vascular and lymphatic diseases, and Rasa1 probably interacts with Ephb4 to inactivate Ras in endothelial cells to regulate vascular development, 38,69 we wondered if Ephb4 and its ligand Efnb2 could inhibit Erk signaling to promote lymphatic valve specification upstream of Rasa1.
Then, we used MEK inhibitors to rescue valve defects in these two mutants.We applied five different MEK inhibitors, selumetinib, cobimetinib, trametinib, U0126-EtOH, and PD0325901, to ephb4b −/ − mutants from 2 to 4 dpf and used pericardial edema as the phenotypic readout (Fig. 7).MEK inhibition reduced the ratio of mutant embryos with pericardial edema (Fig. 7 and Supplementary  11c, d).In contrast, mTOR inhibitors rapamycin and BEZ235 did not rescue ephb4b −/− mutant edema (Supplementary Fig. 11d).We checked the formation of LVs and LVVs in ephb4b −/− mutants after treatment with 10 μM selumetinib and found that the gata2a:EGFP-labeled valve structure could be restored to a certain extent around 77 hpf (Supplementary Fig. 11e, f).
Rescue analyses in efnb2a −/− ;efnb2b −/− mutants showed different results.At 4 dpf, selumetinib treatment failed to reduce mutant edema.However, we discovered that more gata2a:EGFPlabeled valve-forming cells were present in efnb2a −/− ;efnb2b −/− mutant embryos given 100 μM selumetinib from 3 dpf, not only at the putative valve sites but also ectopically in the aLFL at 4 dpf (Fig. 7), even though these valve-forming cells were unable to form mature valve structures probably due to their poor organization.Thus, we propose that Efnb2, which is likewise critical for valve cell organization, 45,72 enhances valve-forming cell specification together with Ephb4 by inhibiting Erk signaling.
Moreover, we found that Efnb2 expression is dynamic in FCLV LECs during lymphatic valve formation.At 60 hpf, FCLV LECs express a small amount of Efnb2 protein on their plasma membrane.Over time, its expression on the plasma membrane of FCLV LECs increases and becomes more constricted and elevated to a subset of FCLV LECs with Prox1a hi expression, which indicates valve-forming LEC specification, at 77 hpf.Efnb2 remains on the plasma membrane of the valve cells at 5 dpf (Fig. 7).The dynamic expression of Efnb2 may be connected to its dual roles in regulating the specification of valve-forming cells and the organization of valve cells.

DISCUSSION
Previous genetic studies suggest that polarized expression of shear stress-responsive genes, including gata2, prox1, and foxc2, is required for the specification of valve-forming LECs during lymphatic valve development. 10,60An intriguing question is how gata2 and prox1 expression are particularly up-regulated in LECs that generate valve cells.Recent studies have shown that oscillatory shear stress can activate the canonical Wnt/β-catenin pathway to regulate valve specification. 10,73Unknown is whether another signaling mechanism is involved.Here, we discovered that, although Erk signaling must be engaged during the LEC formation, Erk inhibition is required for the specification of valveforming cells.Firstly, prior to the initiation of the valve-forming cells, it is necessary to suppress Erk activity in the FCLV.Secondly, in addition to regulating the maturation and survival of valves, Efnb2-Ephb4-Rasa1 signaling is also required to suppress Erk activity in valve-forming cells to promote their specification in zebrafish (Fig. 8).
Activation of MAPK/Erk signaling by Vegfc-Vegfr3 is critical for the induction of LEC fate and thus promotes lymphatic vessel formation, whereas excessive Erk activation invariably leads to the lymphatic abnormalities observed in related lymphatic diseases. 24,26,74,75Loss of both Spred1 and Spred2 in mice, two genes that act to inhibit Vegfr3 signaling-activated Erk1/2, causes not only LEC overgrowth, but also embryonic edema and blood filling of lymphatic vessels. 76It has been reported that the abundance of epsin1 and epsin2 in collecting lymphatic vessels, but not in valve regions, mediates the internalization and degradation of Vegfr3, resulting in the termination of Vegfr3 signaling.Mice with LECspecific deficiency of epsin1 and epsin2 had dilated lymphatic capillaries, immature lymphatic valves, and defective lymph drainage. 77Here, by using ERK-KTR sensor system, we discovered that Erk signaling is dynamically regulated during lymphatic valve development.In contrast to its beneficial function in lymphatic vessel development, suppression of Erk activity is essential for the initiation of valve-forming program, which might also be controlled by spred1/2 or epsin1/2 at this stage.This needs further investigation in zebrafish.However, because FCLV-LV cells are mostly derived from FCLV LECs, the developmental status of FCLV is also closely related to the formation of valve structures that follow.Therefore, the spatiotemporal regulation of Erk activity is critical for the generation of putative valve-forming cells.
Vegfc-Vegfr3 also activates PI3K-Akt signaling in LECs. 78PI3K-Akt signaling has been shown to stimulate de novo LV growth in mice, possibly by inactivating Foxo1, a key negative regulator, to inhibit the expression of valve-forming genes, such as Gata2, Foxc2 and Prox1. 12Taking into account that PI3K-Akt inhibits Erk signaling via Akt1-dependent phosphorylation of Raf1 on Ser259 in LECs, 24,79 Akt1-Raf1 crosstalk may be implicated in valveforming LEC fate specification by controlling Erk1/2 activation.Combined with our findings, we conclude that the suppression of Erk activity is vital for LV development and may be achieved through the synergistic interaction of several signaling pathways.However, the molecular mechanism underlying the regulation of valve-forming genes by suppressed Erk signaling remains elusive and needs further investigation.One potential pathway implicated in this process is Notch signaling.Notch signaling has been shown to suppress Prox1 expression, ensuring the appropriate specification of LECs from venous ECs. 80Notch targets have been found to exhibit oscillatory patterns during somitogenesis and muscle stem cell differentiation, [81][82][83] thereby regulating Erk activity oscillation. 63Additionally, studies in cancer have revealed a positive regulation of Notch signaling by the MAPK/Erk pathway. 84There might be a relationship between Notch pathway with Erk signaling in regulating valve formation.
We noticed that LVs and LVVs almost fail to form in efnb2a −/− ;efnb2b −/− double mutants, but partially form in ephb4b −/− mutant larvae, suggesting that valve abnormalities are not the same in these mutants.Consistently, when using the MEK inhibitor to restore the valve formation in efn-b2a −/− ;efnb2b −/− and ephb4b −/− mutants, respectively, we found that the rescue effect in efnb2a −/− ;efnb2b −/− is inferior to that of ephb4b −/− mutant.This phenomenon may be attributable to the fact that Efnb2 ligand activates other Eph receptors, [85][86][87][88] including Epha4, Ephb1, and Ephb3, which may be expressed in FCLV LECs and hence influence valve development.Contrary to the observations made by Greysson-Wong et al., 89 we didn't observe similar vascular defects in our rasa1 −/− mutants.We suspect there may be a leaky Rasa1b expression due to translation reinitiation after the premature termination codon, helping embryos avoid blood vascular defects. 90These ideas, however, need further investigation.
2][93] In addition to malformed blood vessels, lymphatic abnormalities can be observed in some CM-AVM and VOGM patients, and EPHB4 mutations are found in lymphatic disorder CCLA patients. 35,37ese findings indicate that EFNB2-EPHB4-RASA1 signaling regulates blood and lymphatic vessels in a complicated manner.Our study not only elucidates a possible mechanism underlying the lymphatic abnormalities in such diseases, but also establishes several zebrafish genetic disease models, including ephb4b −/− , efnb2a −/− ;efnb2b −/− and rasa1a −/− ;rasa1b −/− , to better understand the pathogenesis of human lymphedema and evaluate potential therapeutic agents.Importantly, we found that treatment with MEK inhibitors significantly improves the lymphedema phenotype and restores valve formation in ephb4b −/− or rasa1a(-3) −/− ;rasa1b −/− mutant larvae, providing a potential treatment strategy for valve-deficient disorders that currently lack specific molecular treatments.
Whole mount in situ hybridization (WISH) Embryos were fixed for 1-2 days at 4 °C with 4% paraformaldehyde (Sigma, P6148) in PBS, and then transferred to 100% methanol and stored at −20 °C for at least 1 day.Whole mount in situ hybridization was carried out following the standard protocol.A DIG-labeled antisense RNA probe for erk1 was synthesized using Roche DIG RNA Labeling Mix (11277073910).Refer to Supplementary Table 3 for the primers.
Imaging and image processing Embryos or larvae were anesthetized with 0.2 mg/ml Tricaine (Sigma, A5040) and embedded in 1% or 0.8% low melting agarose/Holtfreter's water in 35-mm glass bottom culture dishes.Imaging was carried out on a Perkin Elmer Spinning Disk confocal or Dragonfly Spinning Disk confocal microscope (Andor) using a 20x objective.For time-lapse imaging of Erk reporter fish, embryos were embedded in 0.4% low melting agarose and imaged at 40 min intervals using a 40x oil objective lens on a 30.5 °C heated stage.Time-lapse imaging of embryos with Tg(ef1α-ERKKTR) from shield stages were carried on a 20× objective lens of an FV3000 with a 10-min interval on a 30.5 °C heated stage.For immunostaining, a 40× oil objective was used.Blood and lymph autofluorescence can be detected in the λex 561 nm channel.Embryos with unknown genotypes were imaged first, and then genotyped (see Table S1).As for adult fish caudal heart, fish was first fixed in 10 ml 4% PFA for 1 h at room temperature.Then washed with PBS for several times.The skin and muscle near the caudal heart were removed using forceps.Imaging was performed on the remaining tails.For photoconversion, the Tg(lyve1b:Kaede) larva were mounted in 1% low melting point agarose and observed under an FV3000 inverted confocal microscope (Olympus) with 20× objective.The focused LECs were irradiated by the 405 nm laser for 3 s.Imaris 9.0.1 was used for image processing.Single plane or 3D images were generated by Snapshot and processed by Adobe Photoshop 2020.Movies were generated using Imaris 9.0.1, imported into Adobe Premiere Pro 2020 for labeling, and finally exported as .movfiles.

Lymphangiography and angiography
In tail lymphatic vessel lymphangiography, 5 mg/ml 2000 kDa Dextran-Rhodamine (D7139, Invitrogen) was injected directly into the tail lymphatics using an oil injector under fluorescence stereomicroscopy.The video was taken immediately after injection.For lymphangiography in the trunk, 1-2 nl 10 mg/ml 2000 kDa Dextran-Rhodamine was directly injected into the trunk.After 10-30 min, the larvae were embedded for confocal imaging.Adult retro-orbital injection was performed as described before 98 with modifications.Injections could be made into the tissue behind the eyes as well as the retro-orbital venous sinus.The surrounding lymphatic vessels could take up the Dextran-fluorescein and finally entered the blood circulation.A 10 μl hamilton syringe was used, and the tip was hand honed to be shorter and smoother.Adult fish was anesthetized with 0.2 mg/ml tricaine and positioned dorsal up, head right.The needle was inserted 1-2 mm deep at a 45-degree angle to the fish at seven o'clock position.3 μl 5 mg/ml 2000 kDa Dextran-fluorescein (D7137, Invitrogen) was injected.For the angiography at 4 dpf, 1-2 nl 5 mg/ml 2000 kDa Dextran-Rhodamine was injected into the CCV.Dextran-Rhodamine would be observed circulating in the bloodstream immediately after injection.The successful injection ratio was around 50%.Embryonic Western-blot 4 dpf embryos were collected and anesthetization with Tricaine after being treated with DMSO or 100 μM selumetinib.Yolk was removed by blowing and sucking using a 200 μl pipette.TNE (0.15 M NaCl, 5 mM EDTA, 10 mM Tris-HCl and 1% Triton X-100) plus protease inhibitor (Roche, 04693132001) and phosphatase inhibitor NaF and Na 3 VO 4 was used to lyse the embryos, which were then homogenized using a TGrinder (TIANGEN, OSE-Y50).After boiling in SDS loading buffer, the samples were centrifuged at 12,000 × g for 5 min.The supernatant was used for a typical western blot with a 4-20 % precast gel (EASYBIO, BE6931).Mouse Erk 1/2 antibody (1/2000; Santa Cruz, 514302), rabbit p-Erk1/2 antibody (1/5000; CST, 4370), and rabbit Actb antibody (1/100000; Abclonal, AC026) were used for primary incubation.

Data processing and statistical analysis
To quantify Erk activity in FLV, we used two methods.The first method is to measure the cytoplasm-to-nuclei total intensity ratio.The 3D images were rotated and processed into maximum intensity projection in Imaris 9.0.1.Only cells with obvious EGFP and NLS-mCherry expression were chosen for analysis in Fiji.The nucleus was selected by NLS-mCherry fluorescence with thresholding.Image Calculate-Subtract function was used to generate the ERKKTR-EGFP fluorescence in cytosol.Single cell was marked manually as region of interest (ROI).Total fluorescence intensities of the whole cell and cytosol were calculated with the same thresholding using one ROI.Finally, the cytoplasm-to-nuclei ratio was calculated to represent Erk activity.Same method for calculating Erk activity in whole FCLV and aLFL, as well as the C/N ratio after selumetinib treatment.
The second method is to measure the mean intensity ratio of cytoplasm (define a ring from nuclei) to nuclei.The procedure was carried out as previously described. 99,100The images were first deconvoluted by Huygens and removed unnecessary parts before converted to 2-D by maximum intensity projections in Imaris.Embryos with low KTR expression in FCLV were not suited for calculation because they produced falsely high Erk activity.The nuclei were identified using the Bernsen Auto Local Threshold method, and defined as cells after being dilated twice (two pixels).Fluorescence intensity was measured with Huang auto threshold values.
To quantify Erk activity at the margin region in Tg(ef1a:ERKKTR) embryos, the z-stacked images were first processed into maximum intensity projections using Imaris 9.0.1.The margin region was cropped for analysis in Fiji.First, images were subjected to Subtract Background for signal intensity analysis.In DMSO-treated group, the images were first smoothened and Gaussian blurred with sigma = 1, and then nuclei were found using the Triangle threshold method with low value (from 2 to 7 or 10) because the nuclei were dimmer.The nuclei were bright in selumetinib-treated group, and were identified using an auto Triangle threshold method (argument, ignore black).The nucleus masks were dilated twice (2 pixels) to define cell masks.Nucleus and cell regions and mean intensities were measured and the C/N ratio of mean intensity was calculated in Excel.
When analyzing the ERKKTR-EGFP fluorescence in the nucleus in time-lapse imaging from 72 to 78 hpf, we compared the ratio of nuclear total intensity changes in the nuclei. 101Nucleus equivalent surfaces were created based on the expression of nls-mCherry using Imaris 9.8.0 with a smooth factor 0.7 and thresholding around 112. Surfaces objects with low ERKKTR expression were deleted.Unsuitable Surfaces with multiple nuclei were manually removed or separated.Nuclei tracking was corrected thereafter.These Surfaces were then used to calculate the ERKKTR-EGFP fluorescence intensity in the nucleus.To depict the Erk activity, intensities were normalized by the first value and then scaled by log10.
Cells with Prox1a hi expression and gata2a:EGFP expression were considered as valve cells.When gata2a:EGFP was not detected,

Fig. 1
Fig. 1 New valves in the zebrafish lymphatic system.a Structure of the remaining facial lymphatic sprout (RFLS, arrowheads) during zebrafish larvae development as labeled by Tg(lyve1b:TopazYFP).The white and yellow arrows indicate the sprouts or connections of the RFLS with the otolithic lymphatic vessel (OLV) and trunk superficial lateral lymphatic vessels (LL), lymphatics around the kidney (KL), respectively.Bottom, enlarged images of the boxed regions in the upper panels.SL, standard length.Scale bars, 100 μm.b Development of the RFLS-CCV LVV (yellow arrows) labeled with gata2a:EGFP at different developmental stages.Uncropped images are presented in Supplementary Fig. 1b.Scale bars, 50 μm.c Prox1a immunostaining for the RFLS-CCV LVV in Tg(gata2a:EGFP;lyve1b:DsRed2) transgenic larvae at 6 dpf.Uncropped images are presented in Supplementary Fig. 1c.d Transmission electron microscopy shows the valve (magenta) structure at the RFLS-CCV conjunction site.The boxed region is enlarged at the right panel.CCV, common cardinal vein.Scale bars, 5 μm.e The existence of red blood cells (arrow) with epb41b:mCherry expression in the caudal heart (CH).Scale bar, 50 μm.f The CH structure is labeled by fli1a:EGFP, and a single-layer picture in the bottom panel shows the connection between the CH and CV.CH, caudal heart; CV, caudal vein; CA, caudal artery.Scale bars, 100 μm.g The CH structure is labeled by lyve1b:TopazYFP but not flk:mCherry.gata1:DsRed is used to identify certain red blood cells in the CH.Note that the tail lymphatics are blind-ended.h the CH-LVs and CH-CV LVV labeled by gata2a:EGFP in the CH structure at 26 dpf.The CH (the region with dotted lines) is labeled by lyve1b:DsRed2.Bottom, enlarged images of the boxed region in the upper panel.The white arrow indicates tail lymphatic vessels, yellow arrows indicate lymphatic valves in the caudal heart (CH-LVs), and the white arrowhead indicates the lymphovenous valve that connects the caudal heart and caudal vein (CH-CV LVV).Scale bars, 100 μm.i Transmission electron microscopy shows the lymphovenous valve (magenta) at the CH-CV connection site.Two ultra-thin sections 4 μm apart were chosen to show the two leaflets (the blue and yellow arrows) of this valve structure.The box regions are enlarged at the bottom panel.Scale bars, 5 μm.j Left panel, schematic diagram of valves in the facial lymphatic vasculature (purple) at 5 dpf and juvenile stage.The JLV connects with both thoracic (TD) and kidney lymphatics.Blue, vein; red, artery; green, valves.Right panel, schematic diagram of valves in the tail lymphatic vasculature.Purple, caudal heart; blue, caudal vein (CV); red, caudal artery (CA); green, valves; cyan, lymphatic vasculature.All the images are anterior to the left, dorsal upward

Fig. 2
Fig. 2 Repressed Erk activity in FCLVs during valve formation.a Development of the lymphatic valve (arrowheads) and FCLV-PHS LVV (arrows) labeled with gata2a:EGFP in the anterior part of facial lymphatic vessels (dashed box) at different developmental stages.Middle panel shows gata2a:EGFP expression presented in Fire LUT (Fiji).The arrowheads indicate the putative valve-forming sites.Scale bars, 50 μm.b Detection of valve-forming LECs labeled by high expression of Prox1a at different developmental stages.Middle panel shows Prox1a expression presented in Fire LUT (Fiji).The numbers of embryos with the Prox1a expression pattern (cyan) are indicated.Arrowheads point out valve-forming regions.Scale bars, 50 μm.c Lineage tracing of the valve cells from the surrounding lymphatic vessels with photoconverted Kaede.The LECs at FCLV, LFL near the FCLV, or LFL away from the FCLV were photoconverted at 3 dpf and the valve formation was observed at 5 dpf.The middle two panels are the enlarged and section views of the boxed regions on the left, and valve structures are labeled by dotted lines.The numbers of embryos with the indicated phenotype are shown.Photoconversion results are summarized as schematic drawings on the right.Scale bars, 50 μm.d Upper panel depicts the formation of anterior facial lymphatic vessels.Two populations of LEC precursors from PHS, PHS-LP A and PHS-LP P , contribute to the FCLV, and LEC precursors from ventral aorta, VA-L, contribute to the anterior part of LFL (aLFL) along with PHS-LP P .Lower panels, Erk activity detected by ERKKTR biosensor in Tg(gata2aECE:ERKKTR;gata2aECE:nls-mCherry) fish during lymphatic development, from 52 hpf to 77 hpf.The FCLV region (white solid lines) and anterior part of the LFL (aLFL, yellow solid line) are shown.White arrows indicate cells with high Erk activity, and yellow arrows indicate cells with relative low Erk activity.The individual cells used for ERKKTR-EGFP analysis are listed in Supplementary Fig. 2c, d.Scale bars, 50 μm.e Quantification of the cytoplasmic/nuclear (C/N) intensity ratios used as readout for Erk activity in individual cells.Unpaired two-tailed t test, n = 2 independent experiments, one of them is shown here.All images are anterior to the left, dorsal upward

Fig. 8
Fig. 8 Inhibition of Erk activity in FCLV contributes to valve-forming LEC specification.a In wild-type embryos, FCLV-LVs and FCLV-PHS LVVs can be formed appropriately to ensure the unidirectional flow of lymphatic fluid and prevent the backflow of blood into the lymphatic circulation.b In the embryos with Efnb2-Ephb4-Rasa1 cassette mutations, FCLV-LVs and FCLV-PHS LVVs are defective, therefore leading to a kind of lymphatic blood-filling phenotype.c Vegfc-Vegfr3-mediated Erk activation in luminal LECs of the aLFL.d Efnb2-Ephb4-Rasa1 works in Erk inhibition to induce extremely high Prox1 expression in valve-forming cell specification.PHS primary head sinus, FCLV facial collecting lymphatic vessel, aLFL anterior part of lateral facial lymphatic vessel, pLFL posterior part of lateral facial lymphatic vessel, LV lymphatic valve, LVV lymphovenous valve.Venous endothelial cells (blue), the aLFL LECs with activated Erk signaling (orange), the FCLV LECs with repressed Erk signaling (green), Efnb2-expressing LECs (cyan), lymphatic valve to be formed (cyan with dashed lines), and FCLV-PHS LVV (purple) are indicated.The figure was created using Adobe Illustrator 2021