ETV2 gene is deleted in aves

ETV2 is essential for hemangioblast development in mice and important for vascular and myeloid development in zebrafish (Sumanas & Lin, 2006; Sumanas et al, 2008; Kataoka et al, 2011; Liu et al, 2015). In the chicken genome, we were unable to find the ETV2 ortholog (Weng et al, 2020). To clarify when this gene may have been lost during avian evolution, we performed phylogenomic analysis of the ETV2 gene in birds and non-avian reptiles. ETV2 belongs to the ETS family of transcription factors, the first member of which was chicken ETS1, a cellular counterpart of v-ets oncogene in avian leukemia retrovirus E26 (Leprince et al, 1983; Sizemore et al, 2017). Among 12 subfamilies of ETS-domain proteins (Laudet et al, 1999), ETV2 is a member of the ER71 subfamily (ETV2 is also called ER71 or ETSRP), closely related to another subfamily (the ETS subfamily) that includes ETS1 and ETS2 genes. Our analysis revealed that all three genes, ETV2, ETS1, and ETS2, were paralogous to each other, likely formed as a result of genome duplications in ancestral vertebrates (Lautenberger et al, 1992; Garrett-Sinha, 2013; Yu et al, 2023), and that each had a neighboring ETS family gene belonging to the ERG subfamily, forming three tandemly duplicated ETS gene pairs (ETS1–FLI1; ETS2–ERG; ETV2–FLI1B) (Fig 1B). Both the ETS1–FLI1 and ETS2–ERG pairs are conserved, whereas the ETV2–FLI1B pair and its neighboring genes are deleted in the chicken genome. After extensive search using newly generated avian genome data from the B10K initiative (363 species representing 218 bird families) (Feng et al, 2020) (see the Materials and Methods section), we found that loss of ETV2–FLI1B gene pair was conserved in all avian species analyzed (Fig 1B–D), strongly suggesting a deletion event predating the appearance of modern birds. Syntenic analysis revealed that neighboring genes COX6B1, RBM42, and HAUS5 also exhibited various degrees of lineage-specific loss in bird genomes (Fig 1D).

The squamates (a group of non-avian reptiles including modern-day lizards and snakes) retained both the ETV2 and FLI1B genes and the neighboring COX6B1, RBM42, and HAUS5 genes (Fig 1C and D). Mammalian genomes lost the FLI1B gene when retaining the three neighboring genes (Fig 1D). These data suggested that ancestral amniotes had both ETV2 and FLI1B genes in a genomic locus linked to COX6B1, RBM42, and HAUS5 genes, with subsequent loss of ETV2 and FLI1B in birds (Fig 1D), of FLIB in mammals (Fig 1D), and of ETV2 in turtles (UCSC genome browser gateway; data not shown). Linkage of ETV2 gene to RBM42 and HAUS5 genes was seen in amphibia (Tibetan frog, X. levis, and X. tropicalis) and to RBM42 gene in teleost fish (zebrafish, medaka, and tetraodon) (UCSC genome browser gateway), suggesting that syntenic organization of ETV2 and its neighboring genes in ancestral amniotes was the result of rearrangements of homologous syntenic blocks during tetrapod evolution (Sacerdot et al, 2018; Damas et al, 2021, 2022).

Avian NPAS4L marks mesoderm progenitors that will give rise to the blood and endothelium, but not the smooth muscle

Lack of ETV2 in avian genomes suggested that, unlike in mammals, another transcription factor may function as a hemangioblast master regulator in birds. RNA in situ hybridization analysis showed that endogenous chicken NPAS4L was expressed transitorily in hemangioblasts and that ectopic activation of either chicken or zebrafish NPAS4L was able to induce both SCL/TAL1 and LMO2, two conserved hemangioblast markers (Weng et al, 2020). At the single cell level, RNAseq analysis (HH4–HH11) (see the Materials and Methods section) (Williams et al, 2022; Rito et al, 2023 Preprint) revealed that NPAS4L+ cells were co-positive for SCL/TAL1 (Fig 2A, I, and J) and LMO2 (Fig 2B, I, and J) and were mutually exclusive with HAND2+ smooth muscle progenitor cells (Fig 2C, D, I, and J), in agreement with our whole-mount expression and functional analysis data (Shin et al, 2009). Because NPAS4L marks the early phase of hemangioblast development, at later stages, SCL/TAL1 and LMO2-positive cells became negative for NPAS4L, and only a very small percentage of NPAS4L-positive cells co-expressed terminal differentiation markers for endothelium (CDH5) (Fig 2E, I, and J) and blood (HBZ, hemoglobin pi) (Fig 2F, I, and J). SNAI2 (Fig 2G, I, and J) and ZEB2 (Fig 2H, I, and J), two EMT (epithelial mesenchymal transition) transcription factors expressed during ventral mesoderm differentiation, exhibited different and dynamic co-expression patterns with NPAS4L, suggesting that nascent ventral mesoderm cells are under complex morphological and migratory regulation in addition to cell fate specification (see later part of this work).

Exogenously expressed mouse ETV2 up-regulates avian LMO2, but not SCL/TAL1 expression

Mouse ETV2 was shown to regulate both erythropoietic and endothelial gene regulatory networks, and its knockout affected the development of both lineages (Kataoka et al, 2011; Sumanas & Choi, 2016; Koyano-Nakagawa & Garry, 2017; Sinha et al, 2022). Coding and noncoding sequences regulating a given developmental trait, however, may follow separate evolutionary trajectories (e.g., in genes and regulatory sequences controlling mammalian hairlessness) (Kowalczyk et al, 2022). If deletion of avian ETV2 gene was a recent event, its target genes may still retain cis-regulatory elements reflecting its ancestral roles in erythropoietic and endothelial lineage specification. Indeed, optimal binding sites for both ETV2 (Shrestha et al, 2022) and NPAS4L (Marass et al, 2019) were detected in chicken SCL/TAL1 and LMO2 loci (Fig 2K).

To investigate whether these binding sites reflected bona fide direct transcriptional regulation as was reported for ETV2 during mouse hemangioblast differentiation (Wareing et al, 2012), we cloned full-length mouse ETV2 cDNA (NM_007959) (see the Materials and Methods section) into the pCAGGS-2A-eGFP expression vector (Weng & Sheng, 2014) and tested its ability to induce blood (SCL/TAL1+) and/or endothelial (LMO2+) lineage differentiation in chicken mesoderm. The construct contained an eGFP-encoding gene separated from ETV2 by a 2A-peptide sequence (Weng & Sheng, 2014), and electroporated cells were identified by their GFP-positive signals. Early primitive streak stage (HH2-3) chicken embryos were electroporated with either control (GFP-only) or ETV2-expressing construct and analyzed for SCL/TAL1 and LMO2 expression at HH5-6, when blood and endothelial lineages started to be specified (Fig 3A), and lineage-specific segregation of SCL/TAL1-positive (in blood) and LMO2-positive (in endothelium) cells was initiated. Control construct did not induce either SCL/TAL1 or LMO2 expression as we had previously reported (Weng & Sheng, 2014; Weng et al, 2020) (in this experiment, 0/5 for SCL/TAL1; 0/6 for LMO2) (Fig 3F). Mouse ETV2 was able to induce LMO2 expression strongly (Fig 3C) (10/12), but not SCL/TAL1 (Fig 3B) (0/9) or the hemangioblast marker NPAS4L (Fig 3E) (0/9), suggesting that exogenously expressed ETV2 could still induce endothelial lineage specification despite its deletion in all bird genomes.

Avian ETS1 does not compensate the putative hemangioblast-inducing function of ETV2

As normal chicken hemangioblast development does not require ETV2, we asked whether its function might have been replaced by another ETS-domain protein. Of 22 ETS-domain-containing genes found in the chicken genome (ETS1, ETS2, ELF1, ELF2, ELF3, ELF5, ETV1, ETV3, ETV3L, ETV4, ETV5, ETV6, ETV7, EHF, ELK3, ELK4, FLI1, ERG, FEV, SPDEF, SPI1, and SPIC) (GRCg6a; ensembl.org), 12 (ETS1, ETS2, ELF1, ELF2, ELF5, ETV1, ETV3, ETV5, ETV6, ELK3, ELK4, and FLI1) were shown to be expressed at stage HH3-8 (stages spanning hemangioblast formation and early differentiation) based on chicken developmental promoterome data we had published previously (Lizio et al, 2017). None of these genes matched temporal expression profiles of known hemangioblast markers (SCL/TAL1, LMO2, or NPAS4L) (Nakazawa et al, 2006; Shin et al, 2009; Weng et al, 2020). Two of them (ETS1 and FLI1) exhibited temporal patterns suggestive of their potential involvement in endothelial, but not blood, development (Nakazawa et al, 2006) (http://geisha.arizona.edu/).

In mice, loss of function of ETS1 or FLI1 gene did not affect hemangioblast development but rather resulted in vascular integrity and late-stage hematopoietic lineage differentiation phenotypes (Bories et al, 1995; Spyropoulos et al, 2000; Gao et al, 2010; Li et al, 2015). Although sharing low sequence homology outside the ETS domain, ETS1 is paralogous to ETV2. We asked whether ETS1 could be functioning like ETV2 in chicken hemangioblast development. Exogenous ETV2 did not induce endogenous ETS1 expression (Fig 3D) (0/3). We cloned both versions of chicken ETS1 (ETS1-p68 and ETS1-p54) (Crepieux et al, 1993) into pCAGGS-2A-eGFP vector and analyzed their ability to induce SCL/TAL1 or LMO2 gene. Chicken ETS1 was unable to induce SCL/TAL1 (0/5 for ETS1-p68 and 0/5 for ETS1-p54) (Fig 3G and I) and only induced LMO2 sporadically (1/6 for ETS1-p68 and 1/6 for ETS1-p54) (Fig 3H and J). Similar results were observed when we used an alternative method (CRISPRon) (Weng et al, 2020) to turn on endogenous ETS1 transcription (0/4 for SCL/TAL1, 1/13 for LMO2), suggesting that chicken EST1 could not perform the putative role of ETV2 as a hemangioblast regulator. These data showed that mouse ETV2 was able to induce endothelial, but not hematopoietic, differentiation in chicken mesoderm and that neither ETV2 nor its paralog ETS1 met the criterion of a master regulator of hemangioblast specification.

Squamate genome contains both ETV2 and NPAS4L, and both genes are expressed in putative hemangioblast cells during early development in Madagascar ground gecko

These observations suggested that, in stem amniote, ETV2 may have functioned primarily as an endothelial inducer instead of hemangioblast inducer. However, it may also be possible that ETV2 gene has evolved in its DNA and cofactor binding specificities so that mouse ETV2 could not regulate hematopoietic gene expression in birds. To test this, we decided to clone ETV2 gene from a non-avian reptile species. Our comparative phylogenomic analysis (Fig 1) revealed that both NPAS4L and ETV2 genes are retained in green anole, Madagascar ground gecko, and python, suggesting that squamates (lizards and snakes) resemble anamniotic vertebrates (e.g., amphibians and teleosts) in having both the NPAS4L and ETV2 genes in their genomes. Madagascar ground gecko (Paroedora picta) was recently shown to be a tractable animal model for evolutionary developmental studies (Hara et al, 2015, 2018; Yoshida et al, 2016; Kajikawa et al, 2020), and its developmental staging system has been well documented (Noro et al, 2009; Yoshida et al, 2016). We decided to investigate ETV2 and NPAS4L expression in Madagascar ground gecko embryos.

Pre-ovipositional (before egg-laying) development of P. picta takes ∼10 d, and freshly laid eggs are at a developmental stage equivalent to HH11-12 chicken embryos, too late for hemangioblast specification study. We therefore retrieved pre-ovipositional embryos from gecko oviducts, and embryos equivalent to HH5-7 (gecko Stage 7–9; [Yoshida et al, 2016; Kajikawa et al, 2020]) were used for RNA in situ hybridization analysis, using the same protocol as reported for chicken embryos (Alev et al, 2013). P. picta NPAS4L and ETV2 genes (see the Materials and Methods section) were amplified from the cDNA prepared from pre-ovipositional stage P. picta embryos. RNA whole-mount in situ hybridization study showed that both gecko NPAS4L (PpNPS4L) (Fig 3M and N) and ETV2 (PpETV2) (Fig 3Q and R) genes were expressed in the lateral plate/extraembryonic territories. Paraffin-sections of stained embryos revealed hemangioblast-specific staining of both genes (Fig 3O and P for PpNPAS4L and Fig 3S and T for PpETV2). Sequence comparison revealed that ppETV2 was most closely related to amphibian (X. tropicalis) ETV2 (84.7% amino acid sequence identity in the ETS domain; 34.2% amino acid sequence identity in the full length), followed by teleost (D. rerio) ETV2 (76.2% and 32.3%, respectively) and mammalian (M. musculus) ETV2 (63.9% and 28.8%, respectively). These data suggested that in squamates, both NPAS4L and ETV2 genes are expressed at stages and in cell lineages supportive of their potential involvement in hemangioblast specification in stem amniotes.

Gecko ETV2 induces LMO2, but not SCL/LMO2 expression in chicken mesoderm

Gecko ETV2 likely resembles ETV2 in ancestral sauropsids in both its protein sequence and molecular functions. We then asked whether gecko ETV2 was capable of regulating both hematopoietic and endothelial development, as its expression pattern would suggest. Manipulation of post-ovipositional gecko embryos has been reported (Nomura et al, 2015). However, molecular perturbation before egg-laying is still impractical in any reptilian embryo. We therefore cloned gecko ETV2 in pCAGGS-2A-eGFP expression construct (see the Materials and Methods section) and tested its ability to induce SCL/TAL1 and LMO2 in chicken mesoderm. Similar to mouse ETV2, gecko ETV2 induced ectopic LMO2 expression strongly (Fig 3L) (5/6), but failed to induce SCL/TAL1 (Fig 3K) (0/5) or NPAS4L (0/6).

EMT transcription factor SNAI2 does not interfere with hemangioblast lineage specification

These data suggested that in ancestral amniotes, as in extant teleost fish, both NPAS4L and ETV2 were expressed in ventral mesoderm cells and were involved in blood and endothelial differentiation, with NPAS4L functioning as the major inducer for both lineages and ETV2 as a potent inducer for the endothelial lineage. However, from the perspective of phylogenetic conservation, neither can be considered the hemangioblast master regulator in amniotic vertebrates. The concept of “master regulator” was coined to describe those genes that sit at the top of a lineage specification hierarchy. Pinpointing an exact step for the origin of hemangioblasts is difficult as vertebrate development is a continuous process. In mammals, Flk1+ (VEGFR2+) mesoderm cells give rise to ETV2+ hemangioblasts (Kataoka et al, 2011; Zhao & Choi, 2017). In birds, we had previously described the origin of hemangioblasts as part of ventral mesoderm precursors located in the posterior two-thirds of the primitive streak (Nakazawa et al, 2006; Alev et al, 2010). But the identity of hemangioblasts appears after gastrulation EMT (epithelial–mesenchymal transition) (Nakazawa et al, 2006; Nakaya et al, 2008; Shin et al, 2009), after smooth muscle lineage segregation from the same pool of post-EMT mesoderm cells (Shin et al, 2009) (Fig 1A), during early migration of post-EMT mesoderm cells (Weng et al, 2007; Weng & Sheng, 2014; Nagai et al, 2018) and before aggregation of individually specified hemangioblasts as blood islands (Weng et al, 2007; Sheng, 2010). To test whether hemangioblast specification is affected by gastrulation EMT and mesoderm migration, we overexpressed SNAI2 in chicken mesoderm. SNAI2 is an EMT regulator and is normally expressed in the primitive streak and a subset of nascent mesoderm cells (Fig 4A and B) that do not overlap with NPAS4L positive cells (Fig 2G), whereas another EMT transcriptional factor, ZEB2, is mainly expressed in neural ectoderm territory, not in the streak or nascent mesoderm, and weakly in later-stage blood island cells (Fig 4C and D). Overexpression of SNAI2 did not affect normal expression of SCL/TAL1 (0/10) (Fig 4E), LMO2 (0/4) or HAND2 (0/4) (Fig 4F), suggesting that cellular differentiation is regulated separately from cellular morphogenesis. This was in agreement with data from our previous explant experiments (Nakazawa et al, 2006), in which posterior primitive streak tissues could differentiate into hematopoietic cells without undergoing proper EMT or migration.

Receptor tyrosine kinase signaling functions upstream of NAPS4L-mediated hemangioblast specification in birds, similar to ETV2-mediated hemangioblast specification in mammals

To test whether receptor tyrosine kinase-mediated signaling could regulate avian hemangioblast specification, as demonstrated in mammals, we treated early gastrulation chick embryos (stage HH2/3) with SU5402 (a potent inhibitor of receptor tyrosine kinases, including VEGRs and FGFRs). We had previously shown SU5402 treatment after HH6/7 (when hemangioblast specification is complete) could promote blood lineage differentiation among the hemangioblasts (Nakazawa et al, 2006), a role mediated through FGFR signaling. When chick embryos were treated at HH1 (before gastrulation), both streak formation and gastrulation were severely affected (data not shown). However, when embryos were treated at early mid-gastrulation stages (HH2/3), effect on mesoderm formation was mild. Interestingly, SU5402 treatment at HH2/3 completely abolished hemangioblast marker expression, including NPAS4L (Fig 4G and H), SCL/TAL1 (Fig 4I and J), and LMO2 (Fig 4M–P), whereas smooth muscle cell marker (HAND2) expression was strongly increased (Fig 4K and L), suggesting that VEGFR-mediated signaling plays a role in hemangioblast specification in birds, in agreement with similar observations in mammals. In the presence of SU5402, however, ETV2 (6/7) retained its ability to induce endogenous chicken LMO2 expression (Fig 4M and N), but was still unable to induce endogenous SCL/TAL1 (0/5) or endogenous NPAS4L (0/5). This was partly in agreement with the observation that in the mammalian model, exogenous ETV2 could rescue FLK1-mutant phenotypes (Rasmussen et al, 2013). With SU5402, CRISPRon-mediated ectopic expression of endogenous NPAS4L (as we had previously reported [Weng et al, 2020]) retained its ability to induce SCL/TAL1 and LMO2 (data not shown). Either ETV2 or NPAS4L alone could induce endogenous LMO2 expression, and only NPAS4L could induce SCL/TAL1 expression (Fig 3) (Weng et al, 2020). Combined expression of exogenous ETV2 and CRISPRon-mediated endogenous NPAS4L did not have synergistic effect on either SCL/TAL1 or LMO2 induction (Fig 4O–R) (i.e., SCL/TAL1 induction by NPAS4L; LMO2 induction by both ETV2 and NPAS4L), which was not affected in the presence of SU5402 (compare Fig 4O with Fig 4P and Fig 4Q with Fig 4R).

Taken together, in this work, we showed that ancestral amniotes had both ETV2 and NPAS4L genes in their genomes and that the ETV2 gene was lost in the reptilian lineage leading to modern birds. A separate event led to the loss of NPAS4L gene in ancestral mammals. In reptilian species retaining both genes in their genome, ETV2 promotes endothelial lineage specification, similar to the scenario in anamniotes. Receptor tyrosine kinase-mediated hemangioblast specification may act by promoting chromatin accessibility of phylogenetically conserved downstream transcriptional regulators (e.g., SCL/TAL1 and LMO2) through NPAS4L in birds and ETV2 in mammals. A recent epigenetic study showed that ETV2 binding to its target sequence in mouse SCL/TAL1 enhancer could be distinguished from ETV2-mediated activation of SCL/TAL1 transcription (Steimle et al, 2023), suggesting that ETV2 may function as a pioneer factor in hematopoietic and vascular development (Gong et al, 2022; Steimle et al, 2023). Although NPAS4L has not been shown to bind to closed chromatin or function as a pioneer factor (Marass et al, 2019), its paralogue, NPAS4, can regulate neuronal-specific gene expression in an activity-dependent manner and through differential cofactor recruitment by its heterodimeric partner ARNT2 (Sharma et al, 2019). These data indicate that both NPAS4L and ETV2 may regulate lineage specification via uncoupling of target DNA binding-site occupancy and target gene transactivation, permitting additional steps of molecular regulation for fine-tuned control of hemangioblast differentiation. It is also worth noting that our current study focuses on primitive hematopoietic development and its associated endothelial lineage specification, taking place before the establishment of circulation. Involvement of NPAS4L in definitive hematopoiesis, especially with regard to the generation of hematopoietic stem cells from the dorsal aorta endothelium, awaits further study.

Phylogenomics analysis and single cell RNAseq data analysis

To determine whether the Etv2 and Fli1b genes were lost in extant birds, we annotated these two genes and three neighboring genes (Cox6b1, Rbm42, Haus5) on 363 bird genomes from the B10K project ([Feng et al, 2020] for list of bird species). As reference sets, we collected protein sequences of these genes from the following five species: Homo sapiens, Lacerta agilis, Sphaerodactylus townsendi, Gallus gallus, and Taeniopygia guttata. Two homologous ETS gene pairs (Ets1-Fli1 and Ets2-Erg) of Etv2-Fli1b were also included in the reference set to avoid annotation errors caused by similarity. The reference protein sequences were aligned to the avian genomes by Exonerate (v2.4.1) for gene annotation. To filter out low-quality annotation outcomes, all predicted gene models were translated into protein sequences and then aligned with the reference protein sequence with Muscle (v3.8.1551). Gene models with fewer than 30 amino acids and less than 40% identity to the reference protein were removed. For the loci with multiple annotated gene models, we only kept the one with the highest identity to the reference protein. Etv2 and Fli1b genes were not found in any of the B10K genomes, but neighboring genes were annotated in some avian genomes. To minimize false loss caused by poor assembly quality, we further checked the NCBI gene set database and noted no Etv2 and Fli1b genes in any birds, but we did locate neighboring genes in some birds. We chose birds with high-quality genomes from different clades to demonstrate the presence or absence of Etv2 and Fli1b and three neighboring genes.

Raw reads for chicken early embryogenesis scRNAseq were obtained from GEO NCBI (GSE181577 [Williams et al, 2022] and GSE223189 [Rito et al, 2023 Preprint]) and processed with cellranger-7.0.1 using galGal6 (GCF_000002315.6) as a reference genome assembly. Next expression matrices were filtered (min cell size 500) and normalized by Pagoda2 (v1.0.11; number of top overrepresented genes 2,000, number of PCs 50). Integration was conducted with Conos (v1.5.0) (Barkas et al, 2019) using k.self = 10, ncomps = 30, and n.odgenes = 2000. Human gastrulation data were used for annotation inference (Tyser et al, 2021). Cell clusters were identified with the Leiden algorithm.

Embryology, RNA in situ hybridization, and expression constructs

Fertilized hens’ eggs were obtained from a local farm in Aso. Madagascar ground gecko colony was maintained in RIKEN BDR. Embryological analysis of chicken embryos followed standard protocols as previously described (Alev et al, 2013; Weng et al, 2020). Intra-uterine gecko embryo collection was described previously (Kajikawa et al, 2020), and RNA in situ analysis with gecko embryos followed the chicken protocol (Alev et al, 2013).

Madagascar ground gecko (Paroedora picta) NPAS4L (comp58593_c0_seq1) was identified by tblastx on Reptiliomix (https://transcriptome.riken.jp/reptiliomix/) with green anole (Anolis carolinensis) NPAS4 (XM_008104927) (this gene is the NPAS4L ortholog but is incorrectly annotated as NPAS4) and confirmed by NCBI blast. A 1,068 bp fragment corresponding to nucleotides 798–1,865 of comp58593_c0_seq1 was amplified and used for RNA in situ hybridization. Sequence (comp295360_c0_seq3) containing Madagascar ground gecko ETV2 was identified by tblastx on Reptiliomix with green anole ETV2 (XM_008120938) and verified by NCBI blast. A 567 bp fragment corresponding to nucleotides 526–1,092 of comp295360_c0_seq3 was selected for RNA in situ hybridization.

Application of CRISPRa (also known as CRISPRon) technology in avian embryos was described previously (Lizio et al, 2017; Weng et al, 2020). Four sgRNA sequences (GCCCATGTCTGAGGGAGAGA, TCTCTCCCTCAGACATGGGC, GGAAACCCAGAGGTGCCCAG, CTGGGCACCTCTGGGTTTCC) within the 400 bp region before the transcription start site of chicken p68ETS1 (https://fantom.gsc.riken.jp/zenbu; galGal5::chr24:857052..961686+) were cloned into pAC154 dual-dCas9VP160-sgExpression vector (#48240; Addgene) for ETS1 CRISPRa. Full-length coding regions of gecko ETV2 (comp295360_c0_seq3), mouse ETV2 (NM_007959), chicken p68ETS1 (XM_015297968), and chicken p54ETS1 (XM_040652038) were inserted into pCAGGS-2AGFP (Weng & Sheng, 2014) expression vector for overexpression in chick embryos. Expression constructs for p68ETS1 and p54ETS1 were generated by cloning full-length sequences (154–1,608 of XM_015297969.2 and 176–1,548 of X13027.1, respectively) into pCAGGS-2A-eGFP vector described previously (Weng & Sheng, 2014). cDNA sequence for chicken SNAI2 (XM_040664699.1; coding region 3,450–4,256) was cloned into pCAGGS-2A-eGFP for making SNAI2 expression construct. Plasmid constructs were diluted before electroporation to a final concentration of 1 mg/ml per construct for electroporation in Pannett-Compton solution with glycerol (final concentration 10%) and Fast Green (#061-00031; Wako; final concentration 0.1%).