Plexin D1 negatively regulates zebrafish lymphatic development

The lymphatic system is a network of vessels that are essential for interstitial fluid homeostasis with secondary roles in immune cell trafficking and lipid absorption. Dysregulated lymphangiogenesis has been implicated in a number of health conditions that include, lymphoedema (Brouillard et al., 2021; Saito et al., 2013), tumour metastasis (Stacker et al., 2014) and graft rejection (Dashkevich et al., 2016; Dietrich et al., 2010; Ishii et al., 2010; Pedersen et al., 2020; Wong, 2020).

Lymphatic endothelial cells (LECs) are mostly specified in the venous endothelium, via the action of the transcription factor PROX1 (Srinivasan et al., 2007; Wigle et al., 2002; Wigle and Oliver, 1999; Yaniv et al., 2006). The migration and proliferation of LECs is stimulated through VEGFR3 signalling in LECs, where the binding of VEGFC to VEGFR3 stimulates downstream signalling through the AKT/PI3K and MEK/ERK pathways to promote migration, proliferation and survival (Baek et al., 2019; Coso et al., 2012; Deng et al., 2015; Grimm et al., 2019; Hägerling et al., 2013; Hogan et al., 2009b; Karkkainen et al., 2004; Küchler et al., 2006; Salameh et al., 2005; Shin et al., 2016). Although the pathways that drive lymphangiogenesis are relatively well studied, the mechanisms that regulate this process are less well understood.

The ability to image the entire larval lymphatic network, in vivo, make zebrafish an ideal model for discovering global regulators of lymphatic vessel growth (Jung et al., 2017; Okuda et al., 2012; Yaniv et al., 2006). In the zebrafish trunk, LECs migrate from the posterior cardinal vein (PCV) at 36 hpf to the horizontal myoseptum to form parachordal LECs (Bussmann et al., 2010; Hogan et al., 2009a; Yaniv et al., 2006). These parachordal LECs then migrate along the arterial intersegmental blood vessels (aISVs) to form the trunk lymphatic network, which by 6 dpf consists of intersegmental lymphatic vessels (ISLVs), the thoracic duct (TD) and the dorsal longitudinal lymphatic vessel (Bussmann et al., 2010; Okuda et al., 2012). In the zebrafish head, facial LECs are specified in the common cardinal vein, primary head sinus and from a non-venous progenitor called the ventral aortal lymphangioblast. By 6 dpf these progenitors have coalesced to form the otolithic lymphatic vessel (OLV), the medial facial lymphatic (MFL) and lateral facial lymphatic (LFL) (Eng et al., 2019; Okuda et al., 2012). Importantly, all of these vessels form in a reproducible pattern with little individual variation, suggesting that the migration and growth of both the trunk and facial lymphatic guidance is tightly regulated. It is known that specific populations of neurons, arterial mural cells and fibroblasts are required for growth of lymphatics in the zebrafish trunk, and that both mature Vegfc and the chemokine ligands Cxcl12a and Cxcl12b are involved in mediating the aISV/LEC interaction in the zebrafish trunk (Cha et al., 2012; Lim et al., 2011; Peng et al., 2022; Wang et al., 2020). However, to date, no global regulator of lymphatic vessel guidance has been identified.

We conducted a forward genetic screen to identify novel regulators of lymphatic development and uncovered a loss-of-function plexind1 mutant displaying a novel lymphatic misguidance and overgrowth phenotype. PLXND1 is part of a family of large transmembrane cell-signalling receptors that have roles in both axon and vascular guidance (Zhang et al., 2021). PLXND1 is expressed within endothelial cells and is activated primarily by binding to either secreted or membrane-tethered semaphorin (SEMA) ligands, which results in modulation of endothelial cell migration, growth and adhesion. Previous work has shown that PLXND1 has a conserved role in inhibiting the growth of blood vessels through antagonising VEGFR signalling and by regulating actin polymerisation within endothelial cells (Childs et al., 2002; Fukushima et al., 2011; Gitler et al., 2004; Gu et al., 2005; Kim et al., 2011; Moriya et al., 2010; Tata et al., 2014; Torres-Vázquez et al., 2004; Zhang et al., 2009; Zygmunt et al., 2011). Although the molecular mechanisms that underpin PLXND1-mediated anti-angiogenesis are still not clear, it is thought that upon ligand binding, the GTPase-activating protein (GAP) domain of PLXND1 is activated, leading to the deactivation of Rho-GTPases, resulting in the disruption of integrin-mediated cell adhesion, downregulation of both ERK and MAPK signalling, and modulation of the actin cytoskeleton (Gay et al., 2011; Zhang et al., 2021). In mice, SEMA3C, SEMA3D and SEMA4A have all been implicated in the negative regulation of blood vessel growth through PLXND1 (Gu et al., 2005; Toyofuku et al., 2007; Yang et al., 2015). In zebrafish, Sema3 and Plxnd1 signalling is essential for intersegmental blood vessel (ISV) guidance; Sema3aa and Sema3ab are expressed within the somites and this restricts ISV development to the intersomitic boundaries (Childs et al., 2002; Torres-Vázquez et al., 2004). In addition, the development of the zebrafish common cardinal vein, which requires collective endothelial cell migration, is guided by repulsive cues provided by Sema3d and Plxnd1 signalling (Hamm et al., 2016). Knockdown of mouse Plxnd1 demonstrated that it also has a role in lymphatic development. SEMA3G, which is expressed by dermal arteries, provides a repulsive cue to PLXND1-expressing LECs and knockout of either PLXND1 or SEMA3G results in the dermal lymphatics being aligned more closely with dermal blood vessels (Liu et al., 2016). PLXND1 mutants have additional lymphatic precursors in the intersomitic vessels and display extra branching in the developing cardiac lymphatics (Maruyama et al., 2021; Yang et al., 2012).

In this study, we show that loss of zebrafish Plxnd1 function results in misguided ISLVs, aberrant branching of the facial lymphatics and an increase in the number of facial LECs. Together, our data shows that Plxnd1 transduces repulsive Sema3 signals to guide lymphatic vessels and limit lymphangiogenesis, and is therefore a potential pathway that can be manipulated to control lymphatic vessel growth.

We conducted a forward genetic screen in the lymphatic-marking lyve1b:DsRed transgenic zebrafish line (Okuda et al., 2012) with the aim of identifying genes that regulate lymphatic vessel development. From this screen, we identified a recessive mutant (nz75) that displayed non-stereotypical growth of the lymphatics in the trunk and the head (Fig. 1) – a phenotype that had not been previously described in zebrafish lymphatic mutants. Mutants were viable and fertile as adults, and displayed no gross abnormalities (Fig. S1).

Using whole-genome sequencing coupled with SNP homozygosity linkage analysis (Leshchiner et al., 2012), we mapped the causative mutation to a 600 kb region at the telomeric region of chromosome 8. Within this region was plxnd1, which had been previously implicated in blood vessel guidance and was therefore a clear candidate (Fig. S1). Sequencing the coding regions of plxnd1 in the mutant revealed a single nucleotide substitution of thymine to guanine, resulting in an amino acid change from isoleucine to serine in residue 1030 within the third IPT domain of Plxnd1 (Fig. S1). This residue is highly conserved, and secondary structure prediction revealed that this mutation is likely to destabilize β-sheet formation, leading to a reduction of function (Fig. S1).

Plxnd1 has previously been shown to regulate blood vessel development, with the out of bounds (plxnd1^fov01b) mutant displaying aberrant intersegmental blood vessel (ISV) growth (Childs et al., 2002; Torres-Vázquez et al., 2004). Consistent with this, our plxnd1^nz75 mutants displayed a phenotype of abnormally patterned intersegmental blood vessels (ISVs). In wild-type animals, the primary ISVs emerged from the DA at the intersomitic boundaries and had no additional branches emerging from the main ISV (Fig. S2). In contrast, plxnd1^nz75 mutant fish displayed aberrant ISVs emerging at various locations across the DA and each ISV had sub-branches (Fig. S2). This phenotype was also observed in wild-type fish subjected to morpholino-mediated plxnd1-knockdown. The blood vessel phenotype was quantitated using two measures. Aberrant ISVs were defined as any ISV arising from the DA in a mid-somitic position. The second involved counting the number of branchpoints for each ISV at 36 hpf. plxnd1^nz75 mutant fish both displayed a 10-fold increase in the number of aberrant ISVs and 11-fold increase in ISV branchpoints (Fig. S2). Although a similar phenotype was observed in both plxnd1 morphants and plxnd1^fov01b mutants, the severity of the ISV branching phenotype was significantly higher in both the plxnd1^fov01b mutant and plxnd1 morphants (18-fold increase in morphants, 16-fold in fov01b mutants versus 11-fold in nz75 mutants), suggesting that the plxnd1^nz75 allele is a partial loss of function (Fig. S2). Importantly, plxnd1 mutant and morphant embryos displaying this blood vessel phenotype completely segregated with the previously observed lymphatic phenotype (Fig. 1) suggesting that plxnd1 is responsible for both blood and lymphatic vessel patterning.

We and others have demonstrated that there is an increase in primary angiogenesis (sprouting from the DA) in plxnd1^nz75 mutants and morphants (Fig. S2) (Torres-Vázquez et al., 2004), but the role of Plxnd1 signalling in secondary angiogenesis (sprouting from the PCV) was unknown. In order to investigate this, we imaged lyve1b:EGFP larvae at 39 hpf – a timepoint when secondary sprouts are emerging from the PCV – and observed an ∼1.5-fold increase in secondary angiogenic sprouts in both plxnd1^nz75 mutants and morphants (Fig. S3). Normally, half of these secondary sprouts contribute to the trunk lymphatic network by migrating to form parachordal LECs (PLs) at the horizontal myoseptum, while the other half anastomose with nearby ISVs to create veins (Geudens et al., 2019; Hogan et al., 2009b). We therefore examined both PL formation at 48 hpf as well as the number and ratio of arterial and venous ISVs in both plxnd1^nz75 mutants and morphants. Surprisingly, we found no increase in PL number in either plxnd1^nz75 mutants or morphants (Fig. S3). Instead, we found that there was an approximate twofold increase in the number of venous ISVs, which, because of the corresponding earlier increase in primary sprouting (Fig. S2) resulted in a twofold increase in arterial ISVs, and a 1:1 arterial:venous ratio in 72 hpf plxnd1 animals (Fig. S3). These data indicate that the additional secondary sprouting in plxnd1 mutants contributes towards the trunk blood vascular network.

We noted that there were mispatterned intersegmental lymphatic vessels (ISLVs) in plxnd1^nz75 mutants and morphants at 6 dpf (Fig. 1A-F). The ISLVs are typically guided during their development by growing along arterial ISVs (aISVs) and therefore the pattern of ISLV growth is normally congruent with the underlying pattern of aISVs (Bussmann et al., 2010). There are two explanations for the mispatterned ISLVs we observed in plxnd1^nz75 animals. The first is that the ISLVs are mispatterned indirectly, owing to the ISV defect induced by loss of plxnd1 (Fig. S2), in which case the lymphatics will still align with the ISVs. The second is that plxnd1 is directly required for ISLV guidance, in which case the lymphatics will develop independently of the ISVs. To distinguish between these two possibilities, we quantitated the percentage of ISLVs that did not align with an ISV in double transgenic lyve1b:DsRed; kdrl:EGFP larvae. Whereas wild-type and control morpholino-treated larvae almost never displayed ISLVs that were misaligned with ISVs (96% and 94% ISLV-ISV alignment, respectively), ISLVs were frequently observed in both plxnd1^nz75 mutants and in morphants developing separate from the ISVs (52% and 58% ISLV-ISV alignment, respectively), suggesting that plxnd1 is directly required for correct ISLV patterning (Fig. 1E). plxnd1^nz75/fov01b trans-heterozygotes also displayed an increase in misaligned ISLVs (59% ISLV-ISV alignment) (Fig. 1E), confirming that the ISLV misguidance phenotype was caused by loss of plxnd1.

We further examined the facial lymphatic phenotype at 6 dpf in plxnd1^nz75 mutants and in plxnd1 morphants. Any facial lymphatic branches that are not usually observed in wild-type animals were considered aberrant. We noted that plxnd1 animals frequently displayed aberrant branches from vessels within the facial lymphatic network (Fig. 1C,D). Quantitation revealed that plxnd1^nz75 mutants had a fivefold increase in the number of aberrant facial lymphatic branches (Fig. 1F) suggesting that plxnd1 normally prevents excessive facial lymphatic growth. plxnd1^nz75/fov01b trans-heterozygotes also displayed a threefold increase in aberrant facial lymphatic branches (Fig. 1F), confirming that this phenotype was caused by the loss of plxnd1.

To further confirm the ISLV misguidance phenotype, we examined ISLV patterning in soluble flt1 (sflt1) morphants. sFlt1 normally inhibits blood vessel growth by binding and sequestering Vegfa ligands; therefore, knockdown of this gene is known to cause aberrant growth of ISVs, but sFlt1 is not able to bind either Vegfd or Vegfc, and is not predicted to directly influence ISLV growth (Krueger et al., 2011; Vogrin et al., 2019). A previously validated sflt1 splice-blocking morpholino (Krueger et al., 2011; Wild et al., 2017) was injected into lyve1b:DsRed; kdrl:EGFP fish and the blood and lymphatic vessels in the trunks of these larvae were imaged at 6 dpf. Despite the expected 11-fold increase in ISV branching, there was no accompanying increase in misaligned ISLVs in sflt1 morphants (Fig. S4). This demonstrates that ectopic ISV growth is not sufficient to uncouple ISLV guidance from the ISVs, and provides further evidence for the direct guidance of ISLV growth by Plxnd1.

We examined the ISLV misguidance phenotype in detail by performing time-lapse imaging of both wild-type (Movie 1) and plxnd1^nz75 (Movie 2) animals from 60 to 70 hpf – when the parachordal LECs normally migrate from the horizontal myoseptum (HM) to form ISLVs alongside adjacent aISVs (Bussmann et al., 2010). In wild-type larvae, the developing ISLVs began migrating from the HM at the junction with each aISV but in plxnd1^nz75 larvae, the developing ISLVs would frequently migrate out of the HM at other locations (Fig. 2A-J). Furthermore, whereas wild-type ISLVs spent almost 100% of their migration path aligned with aISVs, in plxnd1^nz75 larvae they spent only 60% of their time in contact with an artery (Fig. 2K). Consequently, the developing ISLVs in plxnd1^nz75 larvae often deviate from the ventral-to-dorsal growth trajectory characteristic of wild types, instead growing in an anterior or posterior direction, as shown by the traces displaying the ISLV growth paths (Fig. 2I,J). Finally, whereas in wild-type larvae the developing ISLV generally had a single growing tip and therefore one direction of migration, in plxnd1^nz75 larvae the developing ISLVs were frequently observed to have multiple tips, with a fourfold increase in branching events per hour (Fig. 2L). The vessel tips would also frequently regress and change direction, taking a tortuous path to reach their final position, as seen by the 1.8-fold increase in ISLV meandering behaviour in plxnd1^nz75 larvae (Fig. 2M). Taken together, these data show that Plxnd1 normally functions to restrict lymphatic vessel migration to ensure the stereotypical pattern of the trunk and head lymphatic networks.

Our data indicate that Plxnd1 signalling regulates lymphatic vessel guidance. To test whether Plxnd1 performs this role in a cell-autonomous fashion, we first confirmed that plxnd1 is expressed in developing zebrafish lymphatic vessels by performing whole-mount in situ hybridisation. We observed plxnd1 expression in the trunk blood vasculature at 36 hpf, as shown previously (Fig. S5) (Torres-Vázquez et al., 2004). In lyve1b:EGFP embryos at 48 hpf – a timepoint when the developing facial and trunk lymphatics can be visualised using an anti-EGFP probe (Okuda et al., 2012), we observed plxnd1 expression in both the facial lymphatic sprout in the head and in the parachordal LECs in the trunk, confirming that plxnd1 is expressed in zebrafish LECs during lymphatic vessel development (Fig. 3A,B, Movies 3 and 4).

Next we investigated whether Plxnd1 acts cell autonomously within LECs. In order to test this hypothesis, transplantations were performed between plxnd1^nz75 donors and wild-type hosts at the blastula stage. If plxnd1 acts cell autonomously to guide lymphatic growth, then plxnd1^nz75 donor LECs would be expected to grow ectopically in wild-type hosts. To visualise the origin and fate of donor-derived endothelium, cells were transplanted from lyve1b:DsRed; kdrl:EGFP donor embryos into fli1a:EGFP host embryos (Fig. 3C). The expression of DsRed in the host embryos allowed the visualisation of donor-derived LECs at 6 dpf, while EGFP expression allowed visualisation of both host (fli1a:EGFP) and donor-derived (kdrl:EGFP) blood ECs at 6 dpf. In this way, it could be determined whether donor LECs are being guided normally by growing along blood vessels or whether they are misaligned with either host- or donor-derived blood ECs. The transplants were conducted using donor and host blastula embryos at 3 hpf. Chimeric embryos were raised and the facial and trunk vasculature imaged at 6 dpf.

As expected, wild-type LECs in wild-type hosts did not form aberrant branches in the facial lymphatics and did not result in misaligned ISLVs (Fig. 3D,F,H,I). In contrast, plxnd1^nz75 LECs in wild-type hosts displayed an 1.8-fold increase in the number of aberrant branches from plxnd1^nz75 (lyve1b:DsRed-expressing) facial vessels. In the trunk, only 66% of plxnd1^nz75 (lyve1b:DsRed-expressing) ISLVs were correctly aligned with aISVs, compared with 99% of wild-type (fli1a:EGFP-expressing) host ISLVs (Fig. 3E,G,H,I). Finally, there was a ninefold increase in aberrant ISV growth in the plxnd1^nz75 mosaic animals, which was consistent with the role of plxnd1 in regulating ISV migration (Childs et al., 2002; Torres-Vázquez et al., 2004) (Fig. S5). Overall, our results indicate that plxnd1 is acting cell-autonomously within LECs to guide lymphatic vessel growth.

We also noted an additional phenotype in the facial lymphatics of our plxnd1^nz75 mutant and morphants. At 6 dpf, the facial lymphatics appeared thickened, with enlarged LFLs, MFLs and OLVs when compared with controls – suggesting an increase in the number of facial LECs. To test this, we generated double transgenic lyve1b:DsRed; fli1a:nlsEGFP fish and counted the number of facial LECs (fLECS) present at 6 dpf. This revealed that the plxnd1^nz75 mutants and morphants had ∼16% more fLECs than controls (Fig. 4A-C). This phenotype was also observed in plxnd1^nz75/fov01b trans-heterozygotes (24% increase in fLECs over wild type), confirming that the loss of PLXND1 signalling results in facial lymphatic hyperplasia. To define when facial lymphatic hyperplasia occurs in the mutants, wild type and plxnd1^nz75 larvae were imaged at 3, 4, 5 and 6 dpf, with the plxnd1^nz75 larvae displaying significantly more facial LECs at 4 and 6 dpf (Fig. 4D). Next, we examined the numbers of prox1a-expressing lymphatic progenitors in the head and observed no difference between plxnd1^nz75 mutants and wild type (Fig. S6). Taken together, these data show that the facial lymphatic hyperplasia in plxnd1^nz75 animals occurs after lymphatic specification and sprouting. We also quantitated trunk LECs at 6 dpf, as well as prox1a-expressing lymphatic progenitor formation in the PCV at 36 hpf, and found no evidence of either trunk lymphatic hyperplasia or an increase in lymphatic progenitors (Fig. 4E, Fig. S6) – a result consistent with our earlier analysis showing normal numbers of parachordal LECs in plxnd1^nz75 animals (Fig. S3).

We next determined whether the facial lymphatic hyperplasia phenotype depends on Vegfr signalling. To test this, we used two small molecule inhibitors – sunitinib (a tyrosine kinase inhibitor) and tivozanib (a specific VEGFR inhibitor) – that block lymphatic vessel growth in zebrafish (Okuda et al., 2015). lyve1b:DsRed; fli1a:nlsEGFP larvae were placed into a solution of DMSO, 1 µM sunitinib or 2 nM tivozanib at 3 dpf – a timepoint before the onset of the hyperplasia phenotype and maintained in these compounds until 6 dpf when they were analysed. These inhibitor concentrations were chosen as they are the highest doses that still allow the complete development of both the facial and trunk lymphatic networks. Drug-treated wild-type larvae displayed a modest reduction in the average number of fLECs (DMSO=78, sunitinib 60, tivozanib=64) – a result consistent with inhibition of Vegfr signalling, whereas inhibitor treatment of plxnd1^nz75 larvae was able to rescue the facial lymphatic hyperplasia by normalising facial LEC numbers (plxnd1+DMSO=88, plxnd1+sunitinib=63, plxnd1+tivozanib=70). This rescue suggests that elevated Vegfr signalling drives the facial lymphatic hyperplasia phenotype seen in plxnd1^nz75 larvae (Fig. 5A). Of note, when the same level of Vegfr inhibition was applied from 2 to 6 dpf it had no significant effect on either the number of aberrant facial lymphatic branches or in the degree of misalignment of the ISLVs at 6 dpf, suggesting that dysregulated Vegfr signalling is not responsible for the misguidance of lymphatics in plxnd1^nz75 mutants (Fig. S7).

In order to investigate which signalling pathway downstream of Vegfr is overactivated in plxnd1 larvae, we examined the activity of the MEK/ERK pathway, as it is known to be inhibited by SEMA3/PLXND1, is downstream of VEGFR function and is also known to promote cell proliferation and survival in LECs (Baek et al., 2019; Carretero-Ortega et al., 2019; Grimm et al., 2019; Shin et al., 2016; Wortzel and Seger, 2011). Erk activity was measured by immunostaining against phospho-Erk (pErk) and anti-EGFP in 36 hpf lyve1b:EGFP larvae. The percentage of pErk-positive cells in the PCV of plxnd1 morphants was increased to 61% compared with only 39% in control morphants, indicating that the higher levels of secondary sprouting seen upon the loss of Plxnd1 may be a result of increased Erk signalling within the PCV (Fig. 5B). We then investigated Erk activity in the facial lymphatics at 3 dpf (Fig. 5C-E) and observed a significant increase in the proportion of pErk-positive fLECs in plxnd1 morphants (36% of fLECs were pErk-positive) over control morphants (20% of fLECs were pERK-positive) (Fig. 5C-E). To test the role of the Mek/Erk pathway in facial lymphatic hyperplasia phenotype, wild-type and plxnd1^nz75 larvae were treated with 5 µM of the MEK inhibitor SL327 from 3 to 6 dpf. This dose was chosen as it is the highest possible dose that still allows the complete development of the facial lymphatic network. Inhibition of Mek signalling was able to rescue the facial lymphatic hyperplasia phenotype in plxnd1^nz75 larvae, with fLEC number reduced to wild-type levels (WT+DMSO=75, plxnd1^nz75=86, plxnd1^nz75+SL327=71) (Fig. 5F), indicating that Plxnd1 signalling normally antagonises Vegfr signalling through the Mek/Erk pathway in order to limit fLEC numbers.

It was intriguing that lymphatic hyperplasia was observed only in the facial lymphatic network. One of the key differences between facial and trunk lymphatic development is that the facial lymphatics require Vegfd-induced Kdr signalling (Astin et al., 2014a; Bower et al., 2017; Vogrin et al., 2019). We hypothesised that Plxnd1 might normally antagonise Kdr-driven Erk signalling, which results in a specific facial lymphatic hyperplasia phenotype when plxnd1 is lost. As Vegfd can only bind to Kdr (Vogrin et al., 2019), we predicted that plxnd1 mutants should be resistant to vegfd knockdown. To test this, lyve1b:DsRed; fli1a:nlsEGFP wild-type or plxnd1^nz75 embryos were injected with either control morpholino or a previously validated vegfd morpholino (Astin et al., 2014a; Bower et al., 2017) and the fLECs were quantified. We found that, as expected, vegfd knockdown resulted in a 15% reduction in fLECs in wild-type larvae but, by contrast, there was no significant reduction in fLEC numbers in plxnd1^nz75 mutants (Fig. 5G). Taken together, these data suggest that Plxnd1 signalling normally antagonises Kdr-driven Mek/Erk signalling to limit fLEC numbers.

Plxnd1 is thought to negatively regulate ISV growth through the ligands Sema3aa and Sema3ab (Torres-Vázquez et al., 2004). These secreted Sema ligands are expressed from the somites where they could bind to and activate Plxnd1 signalling in the blood endothelial cells to prevent aberrant growth of the ISVs into the adjacent somites (Torres-Vázquez et al., 2004). Given the similarity in phenotype between the blood and lymphatic misguidance phenotypes in the trunk, we wondered whether these ligands may also be required for ISLV guidance. We first confirmed that both sema3aa and sema3ab are expressed in the somites by whole-mount in situ hybridisation at 24 hpf. Although we were unable to observe a reliable in situ signal for either gene at 30 or 48 hpf, RT-PCR confirmed that both of these genes are expressed in both the head and trunk at 48 hpf during lymphatic development (Fig. S8).

We analysed a sema3aa^sa10241 mutant generated by The Zebrafish Mutation Project (Kettleborough et al., 2013). This mutant contains a C-to-T substitution that converts Gln501 into a premature stop codon, truncating the protein by 359 amino acids within the semaphorin domain. We observed a significant increase in the frequency of misaligned ISLVs in sema3aa^sa10241 mutant embryos when compared with wild type (wild type=95% ISLV-ISV alignment, sema3aa=78% ISLV-ISV alignment) (Fig. 6A-G), demonstrating that Sema3aa ligands are involved in the Plxnd1-mediated guidance of ISLV growth. However, sema3aa^sa10241 mutants had no increase in aberrant facial lymphatic branches or in the number of fLECs (Fig. 6B,H,I), suggesting that the role of this ligand is restricted to the trunk lymphatic development.

Sema3c has been previously shown to be expressed in the head during facial lymphatic development (Callander et al., 2007; Yu and Moens, 2005) and was therefore a candidate Plxnd1 ligand to be involved in facial lymphatic growth and guidance. Whole-mount in situ hybridisation at 24 and 30 hpf showed that sema3c was also expressed in neuronal cells in the trunk and RT-PCR from both head and trunk RNA confirmed that expression was maintained in both the head and trunk at 48 hpf during lymphatic vessel development (Fig. S8). A sema3c^sa15161 mutant generated by The Zebrafish Mutation Project (Kettleborough et al., 2013) was obtained. This mutant contains a C-to-A substitution that converts Ser142 and Ser186 in the 713 and 757 amino acid Sema3c isoforms into a premature stop codon, thus truncating these proteins within the semaphorin domain. We observed a significant increase in misaligned ISLVs in the trunk of sema3c^sa15161 mutants (wild type=95% ISLV-ISV alignment, sema3c=70% ISLV-ISV alignment) (Fig. 6D-G), demonstrating that multiple Sema3 ligands are involved in the guidance of ISLV growth. However, sema3c mutants had no facial lymphatic defects (Fig. 6C,H,I), suggesting that sema3c is not required for facial lymphatic development.

We have shown that Plxnd1 acts cell autonomously in LECs to guide the pathfinding of the facial and trunk lymphatic networks (Fig. 7). In addition, we uncovered a role for plxnd1 as a negative regulator of facial lymphangiogenesis. We show that loss of plxnd1 results in facial lymphatic hyperplasia and that Plxnd1 normally antagonises Vegfr/Erk signalling to ensure the correct number of facial LECs (Fig. 7).

In the zebrafish trunk, the ISLVs are normally aligned with the aISVs (Bussmann et al., 2010). This interaction is, in part, mediated by chemokine signalling; Cxcl12b ligands, which are secreted from aISVs and associated mural cells, activate signalling in the Cxcr4a/b-expressing LECs that helps to anchor their migration along the aISV (Cha et al., 2012; Peng et al., 2022). We show that the loss of Plxnd1 function results in growth of ISLVs away from the blood vessels and across the somites. Live imaging revealed that growing ISLVs in plxnd1 larvae were more dynamic, spent significantly less time aligned with aISVs and, in some instances, even migrated directly across them. This suggests that the negative guidance provided by Plxnd1 signalling is the dominant guidance cue for the correct patterning of ISLVs.

We have also identified two Sema ligands involved in ISLV guidance; sema3aa mutants and sema3c mutants both displayed misguided ISLVs similar to those observed in plxnd1 mutants. We have confirmed that sema3aa is expressed in somites before lymphatic development (Torres-Vázquez et al., 2004) and that its expression is maintained in the trunk from 22 to 48 hpf. We have also shown that sema3c is expressed within the trunk at 48 hpf and observed expression in trunk neuronal structures by in situ hybridisation at 24-30 hpf. It is therefore likely that the expression of Sema ligands within the somitic region activates repulsive Plxnd1 signalling in LECs that grow towards the somite, restricting their growth to align with the aISVs within the intersomitic space (Fig. 7). The ISLV misguidance phenotype was less severe in both sema3 mutants than in plxnd1-deficient larvae, suggesting that they act redundantly, likely with other Sema ligands (such as Sema3ab), to regulate ISLV guidance. The role of Sema3/Plxnd1 signalling in the zebrafish trunk lymphatics is similar to its role in patterning the murine lymphatics. SEMA3G ligands released by arteries have been shown to repel PLXND1-expressing dermal lymphatics (Liu et al., 2016). In addition, loss of PLXND1 causes an increase in the branching of the cardiac lymphatic plexus around the truncus arteriosus (Maruyama et al., 2021). Taken together with our data, this demonstrates that PLXND1/SEMA signalling has a conserved role in lymphatic guidance.

The facial lymphatics, which are not guided by blood vessels, were also mispatterned in plxnd1 larvae, with an increase in aberrant branches that was similar to the phenotype observed in the trunk. Neither sema3aa or sema3c appear to be involved in facial lymphatic guidance despite both of these ligands being expressed in the head (Fig. S8) (Callander et al., 2007; Tanaka et al., 2007). The role and identity of Sema ligands involved in facial lymphatic guidance remains to be determined.

PLXND1 has a conserved role as a negative regulator of angiogenesis. It regulates blood vessel growth by antagonising VEGFR signalling and also by directly regulating cell migration through destabilising actin polymerisation and integrin-mediated cell adhesion in endothelial cells (Childs et al., 2002; Fukushima et al., 2011; Gitler et al., 2004; Gu et al., 2005; Kim et al., 2011; Maruyama et al., 2021; Moriya et al., 2010; Tata et al., 2014; Torres-Vázquez et al., 2004; Zhang et al., 2009; Zygmunt et al., 2011). We show that suppression of Vegfr signalling did not normalise the lymphatic misguidance phenotype in the trunk or the head. These data suggest that Plxnd1 inhibits pro-migration pathways that are somewhat independent of Vegfr signalling, such as actin assembly and the formation of integrin-mediated adhesions (Driessens et al., 2001; Tata et al., 2014). There are a number of intracellular modulators associated with PLXND1 signalling, such as Rho-GTPases, endocytic adaptors of the GIPC family, cytoskeletal proteins, integrins and kinases, that have been associated with PLXND1 regulation of cell growth that could be involved in lymphatic guidance (Burk et al., 2017; Carretero-Ortega et al., 2019; Choi et al., 2014; Sakurai et al., 2010, 2011; Tata et al., 2014; Uesugi et al., 2009; Wang et al., 2012). In vitro studies have shown that SEMA3E-PLXND1 signalling inhibits the formation of actin stress fibres and focal adhesions, which results in reduced motility and retraction of LECs (Liu et al., 2016; Maruyama et al., 2021). It is therefore possible that activation of Plxnd1 signalling causes localised disruption of the cell migration machinery, leading to withdrawal of cell protrusions and directed growth of the developing lymphatic vessel away from areas containing Sema3 ligands. Future experiments will focus on uncovering the downstream lymphatic migration pathways regulated by Plxnd1.

Plxnd1 signalling can also act as a negative regulator of lymphangiogenesis, which, unlike its guidance role, is context specific; loss of plxnd1 caused an increase in the number of facial LECs but had no effect on trunk LECs. Unlike the misguidance phenotype, the facial lymphatic hyperplasia was normalised by suppression of either Vegfr or Mapk/Erk signalling, indicating that Plxnd1 signalling normally antagonises this pathway. In support of this, we observed an increase in Erk signalling within the fLECs of plxnd1 morphant larvae. Plexins regulate MAP/ERK signalling in various contexts (Aurandt et al., 2006; Basile et al., 2007; Bribián et al., 2014; Ito et al., 2014). In addition, SEMA3E/PLXND1 signalling is known to inhibit VEGFR-mediated and VEGFR-independent ERK signalling in blood endothelial cells (Carretero-Ortega et al., 2019; Moriya et al., 2010) and has been shown to inhibit LEC cell proliferation (Maruyama et al., 2021; Yang et al., 2012). However the exact molecular mechanism by which PLXND1 antagonises VEGFR signalling has not yet been established.

One of the key differences between trunk and facial lymphatic development is the specific requirement for Vegfd/Kdr signalling in the development of the facial lymphatics (Astin et al., 2014a; Bower et al., 2017; Vogrin et al., 2019). We therefore hypothesised that Plxnd1 limits Kdr signalling, and thus inactivation of plxnd1 leads to facial lymphatic hyperplasia. In support of this, plxnd1^nz75 animals were resistant to morpholino knockdown of vegfd, a Kdr-specific ligand. Together, our data suggest that Plxnd1 antagonises Kdr/Vegfr signalling to limit lymphatic growth.

While Plxnd1 is known to regulate primary angiogenic sprouting (Childs et al., 2002; Torres-Vázquez et al., 2004), we show that it also regulates secondary sprouting from the posterior cardinal vein. Interestingly, despite an increase in secondary sprouting, there was no change in the number of prox1a-expressing lymphatic progenitors within the PCV or in parachordal LEC formation in our plxnd1^nz75 mutant fish. Instead, the extra secondary sprouts were biased towards a venous fate to match the earlier increase in primary sprouting from the DA. This resulted in a roughly twofold increase in the number of both aISVs and vISVs, and therefore the ratio of aISVs to vISVs was unchanged in plxnd1^nz75 mutants. This observation supports previous studies showing that the balance between arterial and venous ISVs is tightly regulated (Bussmann et al., 2010; Geudens et al., 2019, 2010).

It has been shown that the arteriovenous fate of ISVs is determined by the levels of Notch signalling in the primary ISVs and by an adaptive flow-mediated mechanism that together ensure a 1:1 ratio of arteries to veins (Geudens et al., 2019). Therefore, the fate of each secondary sprout is largely determined by the fate of the ISV that it engages with. While Prox1a signalling is required for lymphatic specification (Koltowska et al., 2015; Nicenboim et al., 2015), live imaging has revealed that lymphatic progenitors can arise from secondary sprouts that form no stable connection with an ISV (non-venous) but also from secondary sprouts that connect to an ISV and become venous (Geudens et al., 2019; Isogai et al., 2003). Together, these data support the idea that the mechanisms that promote the formation of secondary sprouts are distinct from the mechanisms involved in determining their fate. Our data suggest that while plxnd1 normally inhibits secondary sprouting, likely by antagonising the Vegfr3/Erk pathway, plxnd1 does not regulate arterial-venous fate decisions or the induction of Prox1a and lymphatic fate.

We have shown that Plxnd1 is an important regulator of developing lymphatics in zebrafish. The severity of the misguidance phenotype in plxnd1 mutants suggests that Sema3/Plxnd1 signalling plays a major role in shaping the anatomy of the lymphatic network. We have also identified a role for plxnd1 as an inhibitor of lymphangiogenesis by antagonising Kdr/Vegfr signalling within fLECs. These data demonstrate that not only is lymphatic migration regulated by plxnd1 but that, in certain contexts, it is also required to ensure the correct levels of lymphangiogenesis. Based on previously observed conservation of PLXND1 function, it is highly likely that it also acts as broad regulator of lymphatic growth in mammals (Liu et al., 2016; Maruyama et al., 2021; Yang et al., 2012). Dysregulated lymphatic development underpins a number of human health conditions; lymphatic overgrowth in tumours is associated with metastasis, while lack of lymphatic regeneration following lymph node removal promotes cancer-associated secondary lymphoedema (Gousopoulos et al., 2017; Saito et al., 2013; Yoon et al., 2003). Currently, the VEGFC/VEGFR3 pathway is the key target for regulating lymphatic development. Our data identify the PLXND1 pathway as a potential additional therapeutic target in the treatment of lymphatic pathologies. The use of antibodies, shRNA, injection of ligands or cell-permeable peptides that target the cytosolic domain of PLXND1 could be used to inhibit or stimulate PLXND1 signalling in order to therapeutically promote or suppress lymphangiogenesis (Dieterich and Detmar, 2015; Fukushima et al., 2011; Moriya et al., 2010; Stacker et al., 2014; Vivekanandhan et al., 2021; Zhou et al., 2018). Importantly, therapeutically targeting PLXND1 can likely synergise with existing strategies targeting the VEGFC/VEGFR3 pathway and allow additional therapeutic control of lymphatic growth.

All zebrafish strains were maintained under standard husbandry conditions. The lines used in this study were: wild type (AB), plxnd1^fov01b (Childs et al., 2002), sema3aa^sa10241 and sema3c^sa15161 (Kettleborough et al., 2013), Tg(lyve1b:EGFP)^nz150, Tg(lyve1b:DsRed)^nz101 (Okuda et al., 2012), Tg(fli1a:EGFP)^y1 (Lawson and Weinstein, 2002), Tg(fli1a:nlsEGFP)^y7 (Roman et al., 2002), Tg(kdrl:EGFP)^s843 (Jin et al., 2005), Tg(kdrl:nlsmCherry)^nz49 (Lam et al., 2010), Tg(kdrl:RFP)^la4 (Huang et al., 2005) and Tg(prox1a:KalTA4)^uq3bh; Tg(10xUAS: Venus) (known as prox1a:Venus; Koltowska et al., 2015). plxnd1^nz75 was generated in this study.

ENU mutagenesis was performed as previously described (Solnica-Krezel et al., 1994). Briefly, Tg(lyve1:DsRed)^nz101 males were mutagenised with weekly 1 h exposures to 3 mM ENU (Sigma) over 5 weeks and crossed with wild-type lyve1:DsRed females to produce F1 families. Subsequent incrossing of F2 progeny generated F3 embryos that were screened for misguided lymphatics.

Genomic DNA from pooled nz75 mutant larvae and wild-type siblings (48 larvae each) was isolated using DNeasy Blood+Tissue kits (Qiagen) and converted into libraries using the SureSelect^XT library preparation kit (Agilent Technologies) according to the manufacturer's instructions. Sequencing of libraries was performed on an Illumina NextSeq 500 with 150 bp paired-end reads. We obtained ∼65 million reads per library.

The nz75 mutation was then mapped using the raw fastq files with the SNPtrack algorithm (http://genetics.bwh.harvard.edu/snptrack/#), which maps the mutation to a genomic region by calculating the homozygosity score, expressed as a ratio of heterozygous SNP calls between wild-type and mutant pools multiplied by the number of informative homozygous SNP calls in the mutant pool (Leshchiner et al., 2012).

RNA was extracted from both plxnd1^nz75 mutant and wild-type siblings, and cDNA was generated using random hexamer primers. Eight overlapping PCR products (see Table S1 for primer sequences) within the coding region of plxnd1 were generated, sequenced and aligned to the reference to identify the nz75 mutation.

sema3aa^sa10241 and sema3^sa15161 mutants were genotyped through the Sanger sequencing of genomic PCR products from whole larvae after imaging at 6 dpf (see Table S1 for primer sequences).

In situ hybridisation for plxnd1, egfp, sema3ab and sema3c were performed on whole zebrafish larvae using the in situ HCR kit (Molecular Instruments) according to manufacturer's instructions (Choi et al., 2018). In situ hybridisation for sema3aa was performed as described previously (Jowett and Lettice, 1994) using a 717 bp antisense, digoxigenin-labelled (DIG-labelled) riboprobe (see Table S1 for primer sequences used).

Morpholino injections were conducted as previously described (Nasevicius and Ekker, 2000). Morpholino sequences and doses are given in Table S2. The plxnd1 (Torres-Vázquez et al., 2004) and vegfd MOs (Astin et al., 2014a; Bower et al., 2017) have been previously validated by comparison with a corresponding null mutant allele as per field guidelines (Stainier et al., 2017).

The embryos were raised to 2 dpf (in order to assess the effect on the misguidance phenotype) or 3 dpf (in order to assess the effect on the overgrowth phenotype), and then placed in a solution of 0.5% DMSO, 2 nM tivozanib (Aveo Pharmaceuticals), 1 µM sunitinib (Sigma-Aldrich) or 5 µM SL327 (Abcam), and the larvae were maintained in this solution until 6 dpf.

Cell transplantation was performed on 3 hpf embryos after manually dechorionating them on an agar bed. lyve1b:DsRed; kdrl:EGFP donor and fli1a:EGFP recipient embryos were placed into the troughs of an agar bed that was made using a transplantation mould (Adaptive Science Tools). Cells in the donor embryo were removed from ∼1/10th of the distance from the yolk sac to the animal pole using a borosilicate microinjection needle attached to an Eppendorf CellTram, and then 10-20 cells were transplanted into the same region of the blastula on the recipient embryo. Transplanted recipients were collected and maintained in E3 solution on an agar bed until 24 hpf, after which they were raised according to standard husbandry conditions. Successful grafts were identified by lyve1:DsRed expression and were imaged at 6 dpf.

Immunohistochemistry to detect pErk within the PCV and the facial lymphatics was performed on lyve1b:EGFP embryos using chicken anti-EGFP (Abcam, ab13970, 1:500) and rabbit anti-p44/42 MAPK (Cell Signaling Technology, 4370, 1:250) antibodies as described previously (Okuda et al., 2018).

Reverse transcription PCR (RT-PCR) was used to confirm the expression of semaphorin genes. Larvae at 22, 30 or 48 hpf had their head and trunk tissue removed by dissection with a scalpel and total RNA was extracted. cDNA synthesis was conducted using random hexamer primers on 2 μg of total RNA before PCR with gene-specific primers (see Table S1 for primer sequences used).

Embryos were imaged as described previously (Eng and Astin, 2018) with a Nikon D-Eclipse C1 confocal microscope or with an Olympus FV1000 confocal microscope for time-lapse microscopy. Still images were taken using z stacks 5 μm apart. For time-lapse microscopy, z stacks 5 μm apart were taken at 15 min intervals. Confocal images in this study are maximum projections of z series stacks. Images were processed using ImageJ (NIH), Photoshop CS5 (Adobe) and Volocity 5.4 image analysis software (Improvision/PerkinElmer Life and Analytical Sciences).

Thoracic duct formation was quantitated by live imaging lyve1b:DsRed larvae at 6 dpf and measuring the percentage of somites 7-17 containing a segment of TD (Astin et al., 2014a). Aberrant facial lymphatics were identified as any branches not typically seen in the OLV, MFL and LFL, as previously defined (Eng and Astin, 2018). Misaligned ISLVs were quantitated by live imaging lyve1b:DsRed; kdrl:EGFP or lyve1b:DsRed; fli1a:EGFP larvae at 6 dpf and counting any ISLVs that were not completely aligned with an ISV across somites 7-11. For grafted animals, quantitation was conducted in the regions containing donor cells, as identified by lyve1:DsRed-expressing cells. Facial LECs (Eng and Astin, 2018) and trunk LECs were quantitated as previously described (Okuda et al., 2018).

Aberrant ISVs were counted by live imaging kdrl:EGFP embryos at 48 hpf and identifying all the ISVs between somites 7 and 11 that could be seen just dorsal to the DA and then counting as aberrant those that did not arise from the DA at a position aligned with an intersomitic boundary. Artery:vein ratio was quantitated by live imaging kdrl:EGFP larvae at 3 dpf and identifying them as aISVs or vISVs based on their connection to the DA or PCV, respectively. Secondary sprouts were counted between somites 7-11 by live imaging lyve1b:EGFP embryos at 36 hpf. Parachordal LEC formation was quantitated by live imaging lyve1b:EGFP larvae at 48 hpf and measuring the percentage of somites 7-17 containing a LEC(s) (Astin et al., 2014b).

ISLV time-lapse analysis was conducted by time-lapse imaging lyve1b:DsRed; kdrl:EGFP larvae from 50-75 hpf and identifying an ISLV that grows from the horizontal myoseptum and subsequently displays 6 h of growth for analysis. The aISV contact percentage was calculated as the percentage of time that the ISLV was no more than 20 µm from an aISV. The meandering index was determined by dividing the total growth of the ISLV (including branches that later retracted) over the net migration distance and the number of branching events was counted for each ISLV to give branching events/hour.

Statistical analysis was performed using Prism 5.0 software (GraphPad). Normality of distribution of data was identified by Shapiro-Wilks test. Significance of two sets of data was determined by Mann–Whitney or t-tests depending on whether the data were normally distributed. Significance across three or more sets of data was determined by one-way ANOVA or Kruskal–Wallis tests, depending on whether the data were normally distributed. Welch's correction was performed if different variances were detected.

We thank Alhad Mahagaonkar for expert management of our zebrafish facility, and the Biomedical Imaging Research Unit (School of Medical Sciences, University of Auckland), Yang Deng, Michael Cammer and NYU School of Medicine's Microscopy Laboratory (NCRR S10 RR024708) for assistance in confocal microscopy.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200560.reviewer-comments.pdf.