Localised Collagen2a1 secretion supports lymphatic endothelial cell migration in the zebrafish embryo

The lymphatic vasculature drains tissue fluid and fatty acids, regulates immune cell trafficking and is important in a range of pathologies that include cancer and cardiovascular disease (Ma et al., 2018; Petrova and Koh, 2018). The development of new lymphatic vessels occurs primarily from pre-existing veins and is controlled by mechanisms that are conserved between zebrafish and mammals (Hogan and Schulte-Merker, 2017). In the zebrafish trunk, secondary sprouting occurs from the posterior cardinal vein (PCV) from 32 h post fertilisation (hpf) and produces the first lymphatic endothelial cell (LEC) progenitors (parachordal LECs; PLs) as well as the intersegmental veins (vISVs). In this dorsal angiogenic sprouting event, half the secondary sprouts will form vISVs by anastomosing with adjacent arteries, whereas the remaining half migrate dorsally to the horizontal myoseptum (HM) to form PLs (Bussmann et al., 2010). The PLs proliferate, leave the HM and migrate dorsally and ventrally to form the major lymphatic vessels of the trunk: the thoracic duct (TD), dorsal longitudinal lymphatic vessel (DLLV), and intersegmental lymphatic vessels (ISLVs) (Hogan et al., 2009; Yaniv et al., 2006). LECs migrate progressively between different locations in the embryo, involving serial contact with different cell types and substrates along the way. We are beginning to understand some of the cellular interactions and molecular guidance cues involved along the LEC migration route. For example, chemokine signalling from arteries and local secretion of Vegfc-processing machinery from fibroblasts and neurons are essential non-autonomously acting signals driving lymphangiogenesis (Cha et al., 2012; Lim et al., 2011; Wang et al., 2020). Yet, the identity of all of the key constituents of the LEC microenvironment and how they act in sequence to control LEC migration remains to be fully appreciated.

The development of LECs is intimately associated with and requires the extracellular matrix (ECM) (Frye et al., 2018; Planas-Paz et al., 2012). The ECM is composed of many structural proteins required for cell migration and adhesion, such as collagens, Fibronectin, integrins and Hyaluronan, and also acts as an environment for normal processing and presentation of growth factors such as Vegfc (Jeltsch et al., 2014; Joukov et al., 1997; Le Guen et al., 2014). In vitro studies have demonstrated that Collagen 1 facilitates LEC and blood endothelial cell (BEC) capillary formation and migration, involving collagen ligation to the collagen binding integrins α1β1 and α2β1 with α2β1(Davis and Camarillo, 1996; Podgrabinska et al., 2002; Seo et al., 2014; Whelan and Senger, 2003). In vivo, several ECM components and regulators can influence the development, maturation and function of lymphatics, including Reelin, Itga9, Emilin1, Hyaluronan and Polydom (Bazigou et al., 2009; Danussi et al., 2008; De Angelis et al., 2017; Karpanen et al., 2017; Lutter et al., 2012; Morooka et al., 2017). How different matrix factors pattern the early steps in lymphatic development and the in vivo roles of the many members of large families of structural proteins, such as the collagens, remains to be fully understood.

In this study, we used a forward genetic screen to identify new regulators of lymphatic vessel development. We found that sec23a and mbtps1 mutants exhibit defects in the migration of PLs from the HM to form a functional trunk lymphatic vasculature. Both mutants display a failure of Collagen2a1 (Col2a1) secretion by notochord sheath cells (NSCs). PLs migrate immediately along the NSCs following their departure from the PCV. Analysis of col2a1a mutants and cell autonomy studies suggest that Col2a1 is essential to regulate a discrete step in LEC migration and represents a key and spatially patterned component of the ECM regulating embryonic lymphangiogenesis.

In an ENU mutagenesis screen for genes involved in lymphangiogenesis (see Koltowska et al., 2015b), we identified two mutants, uq^27bh and uq^22bh, that displayed craniofacial defects, a short body axis and loss of lymphatic vessels at 5 days post fertilisation (dpf) (Fig. 1A-F, Figs S1A,B and S2A,B). Using a whole-genome sequencing approach to map the location of the causative genes (Koltowska et al., 2015b), we identified a region of homozygosity for the uq^27bh family on chromosome 18 and a C/T nonsense mutation encoding Y511* in mbtps1 (Fig. S1C-D′). In the uq^22bh family, we identified a region of homozygosity on chromosome 17 and a C/T nonsense mutation encoding Q549*, predicted to disrupt the α-helical domain of sec23a (Fig. S2C-D′). To confirm these mutations as causative, we used CRISPR/Cas9 to generate a 1 bp insertion at position 1157 of the coding sequence of mbtps1, causing a frameshift and subsequent stop codon (uq^28bh D405*). In sec23a, we generated a 3 bp deletion within the codons for amino acids 532 and 533, causing the introduction of a stop codon (uq^23bh, Y532*). Compound heterozygous embryos for uq^28bh;uq^27bh and for uq^23bh;uq^22bh displayed the same phenotypes as the ENU-generated uq^27bh and uq^22bh mutants (Figs S1E-I and S2E-I). This identifies the two ENU-generated mutants as mbtps1^uq27bh and sec23a^uq22bh.

Mbpts1 and Sec23a are both essential for the function of the coat-protein-II (COP-II) secretory pathway: the anterograde protein trafficking route which regulates the selection and packaging of protein cargo in the endoplasmic reticulum (ER) and its transport to the Golgi (Brandizzi and Barlowe, 2013; Szul and Sztul, 2011). The main function of Sec23a is to concentrate fully folded proteins into vesicles on the ER. Mbtps1 is a transcription factor serine protease, which cleaves and activates several transcription factors including Creb3l2 which is essential for the transcription of sec23a (Saito et al., 2009). Thus, both factors act in a pathway mediating ER-Golgi anterograde transport and are essential for secretion of protein cargos.

mbtps1^uq28bh and sec23a^uq22bh mutants were further examined on a transgenic background labelling vascular nuclei Tg(fli1a:nEGFP), veins and lymphatics Tg(-5.2lyve1b:dsRed). To understand when defects in lymphangiogenesis first arose, we quantified the number of PLs located in the HM at 54 hpf and 5 dpf. Both mbtps1^uq28bh and sec23a^uq22bh mutants showed a reduction in initial PL numbers in the HM (Fig. 1G,H), but at later stages PL numbers recovered (Fig. 1I), suggesting a delay in colonising the HM. Subsequently, PLs accumulated in the HM in mutants, suggesting eventual trapping of PLs (Fig. 1J,K). The formation of vISVs remained unaffected in both mutants at 54 hpf, suggesting specificity in this trunk lymphatic phenotype (Figs S1M and S2M). The craniofacial lymphatic network displayed reduced numbers of LECs in both mbtps1^uq28bh and sec23a^uq22bh mutants. However, both mutants also have craniofacial cartilage defects (Haffter et al., 1996; Schlombs et al., 2003) that could contribute to facial LEC phenotypes, and thus these vessels were not further studied (Figs S1J-L and S2J-L). To address the potential impact of general developmental delay, we examined development of the sub-intestinal vasculature, which expands from 2-5 dpf. At 3 dpf (when sec23a mutants already show defects in lymphangiogenesis) there was no difference in EC number in these vessels between siblings and mutants (Fig. S2N,O,R). At 5 dpf the mutants had a mildly reduced number of ECs in intestinal vessels (Fig. S2P,Q,S) but a complete loss of lymphatics, suggesting a mild general delay at later stages that is unlikely to account for the severe lymphatic defects.

We next investigated the timing of PL accumulation in the HM. We time-lapse imaged PLs in sec23a^uq22bh mutants and wild-type siblings from 65 hpf to 78 hpf (Movies 1 and 2). We found that although sibling PLs migrated dorsally and ventrally out of the HM during this period (Fig. 1L,N,O), mutant PLs failed to migrate out of the HM (Fig. 1M,O). These data suggest that PLs in sec23a^uq22bh mutants stall or exhibit a block in their normal migration route. As motoneurons and chemokine signalling can regulate PL migration out of the HM, we examined motoneuron development using immunofluorescence (IF) and expression of the chemokines cxcl12a and cxcl12b at the level of transcription, but found no differences in sec23a^uq22bh mutants or siblings (Fig. S3A-H). At earlier stages of development, we also found that Vegfc/Flt4 signalling was active in both mutants and siblings using IF analysis of downstream ERK-signalling (pErk1/2), and the downstream marker of LEC specification, Prox1, was also unchanged (Fig. S4A-H). Together, these data suggest a defect in LEC migration in the COP-II secretory pathway mutants.

Correct composition of the ECM and cell-ECM interactions are essential for vascular development (Aszodi et al., 2006; Hynes, 2007). sec23a (crusher) mutants, were previously shown to display reduced secretion of cartilage ECM proteins including type II Collagen (Lang et al., 2006). The zebrafish creb3l2 (feelgood) and mbtps1 (goz) mutants, also have decreased collagen deposition and abnormal aggregates of Col2 protein respectively (Melville et al., 2011; Schlombs et al., 2003). Thus, we performed IF analysis to examine the localisation of Col2a1 (which is produced by the NSCs) at the stages when PL migration fails in sec23a and mbtps1 mutants. We observed increased intracellular accumulation of Col2a1 within the NSCs of both sec23a^uq22bh and mbtps1^uq28bh mutant embryos at 54 hpf (Fig. 2A-H). In situ hybridisation revealed no obvious difference in col2a1a transcription, indicating a likely defect in protein secretion rather than increased transcript levels (Fig. S4I,J). In contrast, Fibronectin and Laminin (Lama1), secretion appeared to be unaffected in sec23a mutants (Fig. 2I-N). The NSCs that normally secrete Col2a1 were located immediately adjacent to the HM and the migrating PLs at 2 dpf (Fig. 2B,B′,G,H arrows). Together, these data suggest a selective defect in secretion of Col2a1 in the region of the embryo where we observe PL migration stalling (at the HM alongside the NSCs) in mutants defective in COP-II secretory pathway function.

We next asked whether the loss of Col2a1 contributes to the LEC migration phenotype in sec23a^uq22bh and mbtps1^uq28bh mutants. We generated a col2a1a^uq36bh mutant using CRISPR-mediated genome editing. The specific mutation introduced a 3 bp deletion and 4 bp insertion at position 637 of the coding sequence, predicted to cause a frameshift and subsequent stop codon, E235* (Fig. S4K,L). The col2a1a^uq36bh mutant displayed a shortened body axis (Fig. S4M,N), a lack of Col2a1 production (Fig. 3I,J) and abnormal lymphatic vessel development (Fig. 3C-H). Analysis of the number of PLs at the HM revealed that PLs could still develop in mutant embryos at 54 hpf (Fig. 3A,B,K). However, analysis of embryos from 3-5 dpf revealed that PLs failed to emigrate from, and remained trapped in, the HM (Fig. 3C-H,L). This resulted in reduced numbers of LECs in the TD and ISLVs (Fig. 3M).

To further understand the timing of Col2a1 function in LEC migration, we analysed the cellular defects in these mutants using high resolution, high-speed, time-lapse imaging of PL migration from 60-80 hpf (Movie 3). We found that, although PL migration occurred along the HM [anterior-posterior (a-p) axis], this was at a reduced velocity, indicating that PL motility was reduced (Fig. 4A,B). Quantification of the sphericity of PL nuclei (as a proxy for polarity and cell motility) showed no difference at 3 and 4 dpf, suggesting that cells polarise but have defects in migration (Fig. 4C). Analysis of migration on the embryonic dorso-ventral (d-v) axis revealed significantly less migration in mutants in comparison with siblings (Fig. 4D-H). This data indicates that, although mutant LECs maintain polarity in the HM, a lack of Col2a1 is sufficient to inhibit their ability to migrate dorsally and ventrally from the HM. Of note, the data also demonstrated significant variation in individual PL behaviours (Fig. 4G,H) indicating that some mutant cells migrate while others are severely impaired: overall a milder phenotype than in the COP-II mutants. This suggests contributions from other COP-II-dependent protein cargos, although Col2a1 clearly plays a key functional role.

Finally, to confirm that the cells contributing to the sec23a^uq22bh and col2a1a^uq36bh mutant phenotypes were the same, we injected a Tol2 DNA construct that drove expression of sec23a from the col2a1a promoter (col2a1a:Sec23a.T2A.eGFP). As previously reported, using this construct we were able to observe expression by the presence of EGFP specifically in the NSCs (Cox et al., 2018). Injected sec23a mutant embryos selected for the presence of EGFP-expressing NSCs had a partial rescue of lymphatic vessel development compared with siblings. Injected sibling and mutant embryos selected for the presence of EGFP-expressing NSCs also developed ectopic lyve1b-positive vessels in regions adjacent to EGFP expression around the notochord (Fig. 4I-N). The formation of ectopic vessel fragments contributed to a reduction in TD formation in siblings (Fig. 4I-N). These data indicate that secretion from the NSCs can partially restore lymphangiogenesis in sec23a mutants and, furthermore, that increased sec23a in NSCs can induce local ectopic vessel formation. Taken together with our findings above, this suggests that a defect in secretion of Col2a1 by NSCs along the LEC migration route contributes to the severe lymphatic phenotype in COP-II secretory pathway mutants (Fig. 4O,P).

Using a forward genetic screen, we find that, in addition to previously reported phenotypes, mbtps1 and sec23a mutants exhibit a loss of the lymphatic network in the trunk. Although these COP-II components have general roles in protein secretion, we find that at the developmental stages analysed, these mutants display a selective defect in Col2a1 secretion from NSCs. This work identifies that NSCs produce Col2a1 to support LEC migration in the embryo in a spatially regulated and restricted manner. This adds to other known molecular mechanisms involving cell-cell and cell-matrix interactions during PL migration, including chemokine signalling from arteries (Cha et al., 2012), guidance by motoneurons in the HM (Lim et al., 2011) and Ccbe1, Vegfc- and Vegfc-processing machinery produced by fibroblast-like cells and neurons (Wang et al., 2020). This work further indicates that LECs move progressively through the embryonic environment, interacting in a ‘relay’ with cells and matrix components as they go, to ultimately achieve normal developmental lymphangiogenesis. It seems likely that patterned deposition of matrix components along LEC migration routes could play broader roles in lymphangiogenesis in different vessel beds in both developmental and pathological settings.

All zebrafish transgenic strains Tg(fli1a:nEGFP)^y7 (Lawson and Weinstein, 2002), Tg(lyve1:DsRed)^nz101 (Okuda et al., 2012), Tg(kdrl:gfp)^s843 (Jin et al., 2005) were maintained in accordance with the animal ethics guidelines of the University of Queensland. The forward genetic screen and whole-genome sequence mapping were performed as previously described (Koltowska et al., 2015a).

CRISPR-genome editing of sec23a, mbtps1 and col2a1a was performed as previously described (Gagnon et al., 2014). The guide RNA (gRNA) sequences for sec23a and mbtps1 were designed using the Zhang laboratory software (http://crispr.mit.edu) and the CRISPR-scan software (http://www.crisprscan.org) was used for designing the gRNA for col2a1a. Both sec23a and mbtps1 gRNAs were transcribed using the Ambion Megascript SP6 promoter kit (AM13301502034), whereas col2a1a gRNA transcription was performed using the T7 Megascript™ high yield transcription kit (Invitrogen, AM1354). gRNA clean-up was performed using the RNA clean & concentrator™ (Zymoresearch, R1015) and the gRNA sequences are provided in Table S1.

We mixed 1 μl of sec23a, mbtps1 and col2a1a gRNA (300 ng/μl) with 1 nl of Cas9 mRNA (1 µg/µl) transcribed from pTST3-nCas9 (Jao et al., 2013) and 1 nl injected into the yolk of one-cell-stage embryos (Table S1). DNA extraction and identification of fish harbouring and transmitting lesions was achieved using high-resolution melt analysis (HRMA) as previously described (Dahlem et al., 2012). sec23a^uq22bh siblings and mutants were genotyped using a KASP assay (LGC Genomics). Primer blast was used to design HRMA and sequencing primers, which are listed in Table S1. The Tg(-1.7col2a1a:sec23a-v2a-EGFP) construct was kindly provided by the Knapik lab (Cox et al., 2018) and 100 pg was co-injected with Tol2 transposase RNA as previously described (Koltowska et al., 2015a).

Whole-mount in-situ hybridization (WISH) riboprobe templates for sec23a and col2a1a were amplified by PCR (Table S1.). Probes targeting cxcl12a and cxcl12b were used as previously described (Cha et al., 2012). The amplified product was gel purified (Zymoclean gel DNA recovery kit) and the DIG-incorporated RNA probe was transcribed by T3 RNA polymerase for 4 h at 37°C and purified with the RNA clean & concentrator™ kit (Zymoresearch, R1015). WISH was performed as previously described (Kartopawiro et al., 2014).

Dechorionated embryos were fixed in 4% paraformaldehyde (PFA) at various embryonic stages (32 hpf, 36 hpf, 2 dpf, 3 dpf, 4 dpf, 5 dpf) by placing the tubes of embryos on a rocker at 4°C overnight. Embryos were subsequently washed with 4% PFA and different protocols were followed for probing different proteins through IF staining.

Stage-fixed embryos were washed in 100% methanol, three times for 5 min each on ice. The following day, the embryos were incubated for 60 min in methanol on ice. Subsequently, the embryos were washed with 100% methanol on ice three times for 5 min each. The embryos were stored at −20°C in 100% methanol for a minimum of 2 days before performing IF. Embryos were equilibrated with 1× PBT (1× PBS+0.1% Tween 20) three times for 10 min each. Embryos were then incubated in PBT+30% sucrose (w/v) overnight at 4°C to cryoprotect the embryos. Embryos were equilibrated in 1×PBT three times for 10 min each. Subsequently, the embryos were incubated in 150 mM Tris-HCl (pH 9.0) diluted in milliQ water (MQ) for 5 min at room temperature (RT) followed by an incubation at 70°C for 15 min to facilitate antigen retrieval. The embryos were then washed in 1× PBT twice for 10 min each; followed by a wash in distilled water (dH₂O) twice, for 5 min each. The water was removed and ice-cold acetone was added to the embryos, which were incubated at −20°C for 20 min. After removing acetone, embryos were washed once with 1× PBT before adding Proteinase-K (10 mg/µl, #25530015, Invitrogen). For embryos at 32 hpf Proteinase-K treatment was not required, whereas embryos at 36 hpf were incubated in Proteinase-K (1:500 dilution in 1× PBT) for 30 min without placing the embryos on a rocker. After Proteinase-K treatment, embryos were fixed in 4% PFA for 20 min at RT, without placing the embryos on a rocker. After removing Proteinase-K, the embryos were washed five times in 1× PBT for 5 min each on a rocker. Embryos were then washed once in 1× cold wash buffer made up of TBS (pH 7.6), 0.1% Triton X-100. Embryos were then transferred into the 0.5 ml Eppendorf tubes using a Pasteur pipette. TBST was made using 10× TBS (pH 7.6), 10% Triton X-100 diluted in MQ. The blocking solution was made with TBS (pH 7.6), 0.1% Triton X-100, 1% bovine serum albumin (BSA) and 10% horse heat inactivated serum, and used to incubate embryos overnight on a rocker at 4°C. Subsequently, embryos were incubated with primary antibodies anti-GFP (1:250, #ab13970, Abcam), anti-Prox1 (1:250, # 11-002, AngioBio) and phospho-p44/42 MAPK (Erk1/2) antibody (1:250, # 4370, Cell Signaling Technology) diluted in the blocking solution on a rocker at 4°C overnight. Embryos were subsequently washed in TBST/0.1% Triton X-100 (wash buffer) five times for 30 min each, placed in a rocker at RT. The embryos were later washed in Maleic buffer [150 mM Maleic acid, 100 mM NaCl, 0.001% Tween20 (pH 7.4) saturated with 10N NaOH] once for 30 min at RT. Embryos in 2% blocking reagent (#11096176001, Roche Applied Sciences) diluted in Maleic buffer were placed on a rocker at RT for 3 h. Embryos were incubated on a rocker overnight in 4°C in the secondary fluorescence-conjugated antibodies Alexa 488 (1:400, #A11039, Invitrogen) and HRP-conjugated secondary antibody (1:1000, #W4011, Promega) diluted in Maleic buffer and 2% blocking reagent. The following day, embryos were washed five times in Maleic buffer for 30 min each at RT on a rocker. Embryos were washed once in 1× PBS and incubated in TSA Plus reaction solution (1:50 dilution of Cy3 in 1× amplification diluent; Perkin Elmer) at RT for 3 h. Embryos were washed in TBS (pH 7.6), 0.1% Triton X-100 wash buffer for at least 2 days before imaging.

Embryos were fixed overnight in 4% PFA at 4°C, washed five times in PBST (PBS with 0.2% Triton X-100) and incubated in DNAse I buffer [66 mM Tris (pH 7.5), 5 mM MgCl₂, 1 mM 2-Mercaptoethanol, 50 U/ml DNAse I in PBST] at 37°C for 75 min. Embryos were incubated in Proteinase-K (1:1000) for 20 min at RT followed by blocking for 60 min in blocking solution (1% Triton X-100, 10% horse heat inactivated serum, 1% DMSO, 100 mM maleic acid in PBS). Embryos were incubated with the primary antibodies anti-collagen type II (1:200, #II-II6B3, Developmental Studies Hybridoma Bank); anti-laminin-a (1:250, L9693, Sigma-Aldrich), anti-living colors DsRed (1:500, 623496-100UL, Fisher Scientific) and anti-GFP (1:250, #ab13970, Abcam), diluted in blocking solution overnight at 4°C. They were then washed six times for 30 min in wash solution (1% DMSO, 100 mM maleic acid in 0.2% PBST) and incubated with fluorescence-conjugated secondary antibodies Alexa 488 (1:400, #A11039, Invitrogen), Alexa 546 (1:500, #A11010, Invitrogen), Alexa 647 (1:300, #A21240, Invitrogen) and DAPI (1:1000, #D9542-10MG, Sigma-Aldrich) in blocking solution overnight at 4°C. Embryos were washed in PBST six times for 30 min each and imaged. IF assays on the embryos for fibronectin were also fixed in 4% PFA and washed five times in PBST (PBS with 0.3% Triton X-100). Embryos were incubated in Proteinase-K (1:500) for 20 min at RT followed by blocking overnight in blocking solution (4% BSA, 10% DMSO, 0.1% PBS-Triton X-100). Embryos were incubated with primary antibodies: anti-GFP (1:250, #ab13970, Abcam) and anti-fibronectin (1:300, #F3638, Sigma-Aldrich) diluted in blocking solution overnight at 4°C. They were then washed three times for 1 h each in 0.3% Triton X-100/PBS with 1% DMSO. After embryos were blocked for 1 h in blocking buffer, embryos were incubated with fluorescence-conjugated secondary antibodies Alexa 488 (1:400, #A11039, Invitrogen) and Alexa 647 (1:300, #A21240, Invitrogen) in blocking solution overnight at 4°C. Embryos were washed three times for 1 h each in 0.3% Triton X-100/PBS, fixed in 4% PFA at RT for a maximum of 2 h and washed several times in 0.3% Triton X-100/PBS before imaging.

Embryos were fixed in 4% PFA at 4°C overnight, and rinsed in a methanol series of 75%, 50% and 25% diluted in PBST (PBS with 0.3% Triton X-100) for 10 min each, followed by three washes in PBST for 5 min each. The embryos were incubated in the blocking solution (10% horse heat inactivated serum in PBST) for 1 h at RT on a rocker. The embryos were then incubated overnight on a rocker at 4°C with primary antibodies anti-GFP (1:250, #ab13970, Abcam) and anti-tubulin (1:200, #T6793, Sigma-Aldrich) diluted in the blocking solution. The next day, the embryos were washed three times in PBST for 1 h each followed by an incubation in the blocking buffer for 1 h at RT on a rocker. The embryos were then incubated in fluorescent-conjugated secondary antibodies Alexa 488 (1:400, #A11039, Invitrogen) and Alexa 647 (1:300, #A21240, Invitrogen) diluted in the blocking buffer overnight at 4°C on a rocker. The following day, embryos were washed in PBST three times for 1 h followed by re-fixation of the embryos in 4% PFA for 30 min. The embryos were then washed in PBST three times for 5 min before imaging.

After fixing embryos overnight in 4% PFA at 4°C the embryos were washed in 1× PBT (PBS with 0.1% Tween-20) three times for 5 min on a rocker. Embryos were incubated in Alcian Blue solution [0.1% Alcian Blue (w/v) in 80% ethanol and 20% glacial acetic acid, filtered through a 0.2 µm syringe filter before use] at RT for 6 h. Embryos were washed in an ethanol series of 75%, 50% and 25% diluted in in PBT (PBS with 0.1% Tween-20) for 5 min each. Embryos were then incubated in bleach solution (3% hydrogen peroxide and 1% potassium hydroxide diluted in PBT) at RT for 2 h in 2 ml Eppendorf tubes with caps open to facilitate removal of pigments from the embryos. The embryos were then washed in PBT for 5 min before imaging.

Embryos were mounted laterally on 0.7% low melting temperature agarose (Sigma-Aldrich, A9414-100G) and imaged on a Zeiss LSM 710 FCS confocal microscope or an Andor dragonfly spinning disc confocal microscope with the Zyla 4.2 sCMOS camera (exclusively for Movie 3). All images were processed using Fiji software (Schindelin et al., 2012) and time-lapse movies were processed in Imaris (Bitplane). Quantification of LEC number was performed across specific somites specified in the figure legends as previously described (Koltowska et al., 2015b) by quantifying the nuclei in z-stacks of confocal micrographs using the Tg(fli1a:nEGFP) Tg(-5.2lyve1b:DsRed) transgenic line. The sphericity of PL nuclei in Fig. 4 was generated using the Spot detection algorithm from Imaris (Bitplane).

Statistical analysis was performed using Prism software (GraphPad software) using the two-tailed unpaired Student's t-test for all statistical analysis. Error bars indicate s.e.m. P-values are represented as *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

The Institute for Molecular Bioscience (IMB) Sequencing Facility performed all sequencing. Imaging was performed in the Australian Cancer Research Foundation's Cancer Ultrastructure and Function Facility at IMB.