Formal proof of the requirement of MESP1 and MESP2 in mesoderm specification and their transcriptional control via specific enhancers in mice

The mesodermal cells are derived from the primitive streak (PS) during gastrulation in mouse development and continue to be produced in the tail bud from the anterior to posterior. These mesodermal cells give rise to the heart, head mesenchyme, blood, blood vessels, somites, kidney and limbs. In order to differentiate into different tissues, specific regulators for each tissue should be expressed in a spatiotemporal-specific manner in mesodermal cells. MESP1, a bHLH transcriptional factor, is the earliest marker for mesodermal cells. The Mesp1 lineage, which is derived from the anterior mesoderm, gives rise to the heart and head mesenchyme, hematopoietic cells and blood vessels (Saga et al., 1996, 2000, 1999; Yoshida et al., 2008). Mesp1-null mutant mice exhibit heart morphology defects (Saga et al., 1999), suggesting that MESP1 is essential for cardiac formation. Although mice null for the Mesp1 paralogue Mesp2 did not exhibit anterior mesoderm formation defects (Saga et al., 1997), Mesp1 and Mesp2 double-null mice displayed the accumulation of undifferentiated mesoderm cells at the PS (Kitajima et al., 2000), suggesting that MESP1 and MESP2 have redundant functions in anterior mesoderm formation. Forced expression of MESP1 in mouse embryonic stem (ES) cells induced cardiac differentiation (Bondue et al., 2008; Chan et al., 2013; David et al., 2008; Lindsley et al., 2008), and introduction of the combination of MESP1 and other heart-specific transcriptional factors was proposed to reprogram cells into the cardiac lineage (Fu et al., 2013; Islas et al., 2012; Wada et al., 2013). This suggests that MESP1 and MESP2 are key regulators of anterior mesoderm differentiation. However, spatiotemporal regulation of Mesp1 and Mesp2 expression during early mesoderm formation has not been fully addressed. As Mesp1 expression starts immediately after mesoderm cells are formed and is downregulated during gastrulation (Saga et al., 1996, 1999), it is possible that the regulation of Mesp1 expression is coupled with PS formation and mesoderm progenitor cell specification. BMP from extra-embryonic tissue and epiblast Nodal have together been reported to induce the expression of Wnt3 (Ben-Haim et al., 2006), then a WNT3 and NODAL positive-feedback loop determines where the PS is induced (Ben-Haim et al., 2006; Brennan et al., 2001; Rivera-Perez and Magnuson, 2005; Tortelote et al., 2013; Yoon et al., 2015). Wnt3 is soon downregulated at the PS (Liu et al., 1999). However, Wnt3a expression is induced at the PS (Takada et al., 1994), and canonical Wnt signaling is continuously activated at the PS and functions in the maintenance of mesodermal progenitors (Dunty et al., 2008; Garriock et al., 2015; Nowotschin et al., 2012; Takemoto et al., 2011). Activation of canonical Wnt signaling can induce mesodermal cells from ES cells (Chal et al., 2015; Gouti et al., 2014; Zhao et al., 2014). On the other hand, Nodal signaling activity is downregulated in the posterior part of the PS and the Nodal signal gradient in the PS has been proposed to specify a different lineage, as the anterior PS gives rise to the definitive endoderm and the posterior PS gives rise to the mesoderm (Arnold and Robertson, 2009; Morgani and Hadjantonakis, 2020; Robertson, 2014; Tam and Behringer, 1997; Tam and Loebel, 2007). Lefty2 is a downstream target of the Nodal pathway and functions as an inhibitor of Nodal (Meno et al., 1999; Saijoh et al., 2000; Sakuma et al., 2002). It also likely plays a role in the specification of mesodermal cells in the posterior PS. Wnt3 conditional knockout (KO) in the epiblast, as well as Lefty2-null mice, exhibit defects in mesoderm migration from the PS and accumulation of mesodermal cells at the PS (Meno et al., 1999; Tortelote et al., 2013), which resembles the Mesp1 and Mesp2 double-null phenotype. It is therefore reasonable to hypothesize that tissue-specific expression of Mesp1 and Mesp2 is regulated by the continuous activation of canonical Wnt and downregulation of Nodal signaling in the posterior PS.

In zebrafish, Mesp overexpression causes cells to differentiate into the cardiac lineage (Deshwar et al., 2016), suggesting conserved Mesp family functions in early mesoderm formation. However, quadruple Mesp mutant zebrafish had only somite polarity defects, not early mesoderm formation defects (Yabe et al., 2016). Furthermore, the targeting strategy for Mesp2 affected the severity of Mesp2-null phenotypes. In particular, the targeting strategy using the PGK-neo cassette affected the expression of the neighbor gene Mesp1 (Takahashi et al., 2007). In this study, we re-established Mesp1 and Mesp2 single or Mesp1/Mesp2 double-null mutant mice without inserting the PGK-neo cassette using CRISPR/Cas9-mediated genome editing, and evaluated the spatiotemporal regulation of Mesp1 expression during anterior mesoderm formation.

The previously reported Mesp1/Mesp2 double knockout (dKO) mice lines were generated by integrating the PGK-neo cassette following the KO strategy (Kitajima et al., 2000). In order to establish Mesp1/Mesp2 dKO mice without using the PGK-neo cassette, we used the CRISPR/Cas9 genome-editing technique. We designed two sets of guide RNAs (sgRNAs), (1) Mesp1-1 and Mesp2-2, and (2) Mesp1-5 and Mesp2-5, to target both Mesp1 and Mesp2 genes simultaneously. We designed the sgRNAs upstream of the basic helix-loop-helix (bHLH) domain to create functionally null MESP1 and MESP2 proteins if a frame-shift occurred as the result of an indel caused by non-homologous end-joining (NHEJ) (Fig. S1A). Initially, we conducted F0 analyses. The bicistronic expression vectors expressing hCas9 mRNA and sgRNA for Mesp1 and Mesp2 (Table 1) were injected into fertilized eggs, and embryos were recovered at embryonic day (E) 9.5. The embryos were categorized into three groups by their phenotypes: type I exhibited normal morphology (Fig. 1A: n=99); type II displayed an ambiguous somite boundary, similar to Mesp2 KO embryos (Fig. 1B: n=15); and type III embryos were completely disorganized, with traces of a head fold (HF)-like structure and allantois, and no posterior structure, similar to Mesp1/Mesp2 dKO mutants (Fig. 1C: n=12). We observed no embryos with the double-heart phenotype expected for the Mesp1-single null mutant (Saga et al., 1999). The genotypes of embryos were confirmed by sequencing (Table S1). Type I contained Mesp1-null mutants caused by a frame-shift in both alleles. This suggests that Mesp1-null embryos developed normally, which is inconsistent with the previous report. Type I also contained homozygously modified Mesp2 mutants. However, in most cases, the mutation contained only a small indel without a frame-shift. Type II contained Mesp2 mutants with frame-shifts in both alleles, with at least one functional Mesp1 allele. Type III embryos contained homozygous frame-shift mutations for both Mesp1 and Mesp2 (Table S1), confirming that the most severe phenotype was caused by the dKO mutation of Mesp1 and Mesp2. The phenotype of Type II embryos was further analyzed by the expression of Uncx4.1, a caudal somite marker, and T, a presomitic mesoderm marker (Fig. S2). Most Type II embryos displayed a caudalized somite phenotype similar to previously reported Mesp2 KO embryos with variable severity. This variable severity may have been caused by the mosaicism of F0 mutants (Table S1). Although we used two sets of sgRNAs for Mesp1 and Mesp2, the mutation frequency in the Mesp2 gene was low in both cases. In order to address the possibility that the Mesp1 locus is more favorably mutated by the CRISPR/Cas9 technique than by the Mesp2 locus, the bicistronic expression vectors expressing hCas9 mRNA and sgRNA for Mesp1 or Mesp2 were separately injected into fertilized eggs, and embryos were recovered at embryonic day (E) 9.5 (Table S2). Among the 20 embryos obtained by pX330/Mesp2-2 injection, all contained the indel mutation and 19 were homozygously mutated. This high frequency was similar to that seen in response to pX330/Mesp1-1 injection, in which we obtained 10 embryos, including eight with homozygous mutations. On the other hand, of 26 embryos obtained by pX330/Mesp2-5 injection, 19 were mutants, but only four had homozygous mutations (Table S2). The target sites of Mesp2-2 and Mesp2-5 sgRNAs are only 36 bp apart (Fig. S1A), suggesting that the efficiency of mutation by these sgRNAs is not regulated by epigenetic modification of the locus, but simply reflects the double-strand break efficiency. We injected pX330/Mesp1-1 and pX330/Mesp2-2 together again, and the efficiency at each locus was similar to that when injected separately (Table S2).

To analyze the phenotypes of the mutants in more detail, Mesp1 and Mesp2 KO lines were established using the CRISPR/Cas9 technique (Tables 2 and 3). Mesp1^Δ2bp/Δ2bp, Mesp1^{Δ17bp/Δ17bp} and Mesp1^{Δ573bp/Δ573bp} mice developed normally and were fertile, confirming that Mesp1 is dispensable as long as the Mesp2 allele remains intact. We obtained several types of Mesp2 KO, including Mesp2^{Δ53bp/Δ53bp}, Mesp2^+1bp/+1bp, Mesp2^Δ1bp/Δ1bp and Mesp2^Δ5bp/Δ5bp. The CRISPR/Cas9-mediated Mesp2 KO embryos displayed phenotypes similar to those previously reported, such as upregulated Mesp1 expression compared with controls at the newly forming somite boundary (Fig. 2A,D,G), caudalized somites (Fig. 2E,H, Fig. S3B,C), and fused pedicles and laminae of the neural arches (Fig. 2F,I). Embryos with a few nucleotide deletions without frame-shifts in Mesp2 (Mesp1^{Δ14bp/Δ14bp};Mesp2^{Δ27bp/Δ27bp} and Mesp2^{Δ21bp/Δ21bp}) demonstrated normal development similar to controls (Fig. 2A-C,J-L, Fig. S3A,D), which was likely due to intact bHLH domains in these mutant embryos. We also established Mesp1/Mesp2 dKO mouse lines (Mesp1^+1bp/+1bp;Mesp2^{Δ10bp/Δ10bp} and Mesp1^{Δ573bp/Δ573bp}; Mesp2^Δ4bp/Δ4bp). Previously reported Mesp1/Mesp2-null phenotypes at the gastrulation stage (Kitajima et al., 2000), such as the accumulation of mesodermal cells around the PS (Fig. 3A-D) and loss of Lefty2 expression (Fig. 3E,F), were also recapitulated in the dKO embryos. Furthermore, Eomes, Fgf8 and Wnt3, which are expressed in the mesoderm at the gastrulation stage, were upregulated in the mesoderm of Mesp1/Mesp2 dKO embryos (Fig. 3G-L). However, Wnt3a, which is expressed in the mesoderm later than Wnt3, was not expressed in the mesoderm of Mesp1/Mesp2 dKO embryos (Fig. 3M,N). The expression of the epithelial-mesenchymal transition (EMT) marker Snai1 was downregulated and restricted to the PS in Mesp1/Mesp2 dKO embryos (Fig. 3O,P). This suggested that the mesodermal tissue in Mesp1/Mesp2-null embryos accumulated at an early mesodermal state and was prevented from further differentiation. Therefore, Mesp1 and Mesp2 are required for proper mesoderm formation in mouse but not in zebrafish.

We found that Mesp1-null mice generated by genome editing had no abnormalities, suggesting that the phenotypes of Mesp1 KO mice observed previously were caused by the PGK-neo cassette used for targeting. To address this possibility, we removed the PGK-neo cassette from the Mesp1^Neo allele using CRISPR/Cas9-mediated genome editing with Mesp1-5′ and Mesp1-3′-2 sgRNAs (Fig. S1, Fig. 4A), and designated this as the Mesp1^ΔNeo allele. Mesp1^Neo/Neo embryos displayed bifurcation of the cardiac primordia, as previously reported (Fig. 4B) (Saga et al., 1996). However, Mesp1^ΔNeo/ΔNeo embryos had correctly looped single-tube hearts at a similar stage (Fig. 4C) to controls (Fig. 4A). Furthermore, Mesp1^ΔNeo/ΔNeo mice developed normally and were fertile. We also made a new Mesp1-Cre line with a floxed PGK-neo cassette. This line was generated by almost the same strategy as the Mesp1^Cre line previously produced (Saga et al., 1999), but the PGK-neo cassette was floxed and removed to produce Mesp1^CreΔNeo (Fig. S4A). The homozygous Mesp1-Cre mice exhibited severe heart developmental defects similar to or even more severe than those in the Mesp1-null embryo, in which Mesp1 was replaced with the PGK-neo cassette (Lescroart et al., 2018; Saga et al., 1999). Similar to the Mesp1^ΔNeo/ΔNeo case, Mesp1^{CreΔNeo/CreΔNeo} mice also developed normally and were fertile. As MESP1 and MESP2 have redundant functions (Saga, 1998), Mesp2 expression was examined in Mesp1 KO embryos. In wild-type embryos, Mesp2 expression was observed from the stage when gastrulation starts until the early-bud (EB) stage (Fig. 4D) and was downregulated around the late-bud (LB) stage (Fig. 4G). Mesp2 expression became restricted to the lateral plate mesoderm around the early HF (EHF) stage (Fig. 4J) and was then further restricted at the boundary of newly forming somites (Fig. 4M). However, Mesp2 expression was higher at the EB stage (Fig. 4F), and retained at the LB stage in Mesp1^ΔNeo/ΔNeo (Fig. 4I) and Mesp1^{CreΔNeo/CreΔNeo} embryos (Fig. S4C). This upregulation was not observed in Mesp1^Neo/Neo embryos (Fig. 4E,H), but Mesp2 expression was retained at the LB stage, as previously reported (Kitajima et al., 2000), along with slight morphological defects, which may reflect a delay of mesodermal cell differentiation. All mutants exhibited the downregulation of Mesp2 by the EHF stage (Fig. 4J-L, Fig. S4C), but its expression re-started again at the somitogenesis stage (Fig. 4M-O). In order to examine whether the upregulation of Mesp2 expression in Mesp1^ΔNeo/ΔNeo is sufficient to rescue Mesp1-null phenotypes, we conducted RNA-seq analysis using controls (Mesp1^ΔNeo/+) and Mesp1^ΔNeo/ΔNeo embryos at the EB and LB stages. Based on the pairwise scatterplot, regardless of the genotype and stage, the expression levels were highly correlated as the Pearson correlation coefficients were r>0.985 for all pairs (Fig. S5A). In the principal component analysis (PCA), the samples with the same stage or genotype were grouped together. However, the distance between biological replicates was similar to that between the same stage controls and Mesp1^ΔNeo/ΔNeo samples. This suggested that the top 500 most variable genes were similar between the controls and Mesp1^ΔNeo/ΔNeo (Fig. S5B). Indeed, only six genes, including Mesp2, at the EB stage and four genes and the LB stage were differentially expressed genes (DEGs) according to the comparison between control and Mesp1^ΔNeo/ΔNeo samples (Fig. S5C, Table S3). The DEGs at the EB stage contained Tdgf1 and Eomes, which are indispensable for PS formation (Ding et al., 1998; Russ et al., 2000; Xu et al., 1999). In order to determine whether the upregulation of these genes plays a role in rescuing Mesp1-null phenotypes, we examined their expression in control and Mesp1^ΔNeo/ΔNeo embryos. However, their expression was unchanged between the control and Mesp1^ΔNeo/ΔNeo embryos (Fig. S5D,E). Mesp2 was the only gene among the DEGs whose expression was higher in both biological replicates of Mesp1^ΔNeo/ΔNeo than in controls at both stages. It is likely that the upregulation of Mesp2 expression during early mesoderm formation rescued the phenotypes caused by deletion of Mesp1, but the PGK-neo cassette insertion prevented this rescue mechanism.

Although we confirmed that proper expression of Mesp1 and Mesp2 is essential for early mesoderm formation in mice, the regulation of Mesp1 and Mesp2 expression during the gastrulation stage is not well understood. As mesoderm specification requires continuous Wnt signaling activation and the suppression of Nodal signaling in the PS, Mesp1 and Mesp2 expression may be regulated by these signaling pathways. To assess this possibility, we established CRISPR/Cas9-mediated Wnt3 KO and Lefty2 KO mice, and examined Mesp1 and Mesp2 expression. The phenotype of the previously reported Wnt3 null, which lacked the PS structure and expression of markers for the PS and early mesoderm (Liu et al., 1999), was recapitulated in the Wnt3 KO generated by Cas9 (Fig. S6B,E,H). In the Wnt3 KO, neither Mesp1 nor Mesp2 expression was observed (Fig. 5A-D). As reported previously, the Lefty2 KO exhibited the accumulation of mesodermal cells around the PS (Meno et al., 1999). In the Lefty2 KO embryos, Wnt3 expression, as well as expression of the PS and early mesoderm markers, were upregulated in the thickened mesoderm (Fig. 5E,F and Fig. S6C,F,I), similar to in Mesp1/Mesp2 dKO. However, Mesp1 expression was unaffected (Fig. 5G,H) and Mesp2 expression was retained in the Lefty2 KO (Fig. 5I,J). Taken together with Lefty2 expression being lost in Mesp1/Mesp2 dKO (Fig. 3F), it is possible that Lefty2 expression is regulated by MESP1 and MESP2 in the early mesoderm, and that it is essential for further mesoderm differentiation. To test this hypothesis, we established mice with Lefty2 knocked into the Mesp1 locus (Mesp1^Lefty2; Fig. S7). Mesp1^{Lefty2/Lefty2} mice developed normally and were fertile. We then further established Mesp1^{Lefty2/Lefty2};Mesp2-null mice. Although the Mesp1^{Lefty2/Lefty2};Mesp2-null mice expressed Lefty2 in the mesodermal cells at the gastrulation stage (Fig. 6A,B), the phenotypes were indistinguishable from those of Mesp1/2 dKO mice (Fig. 6C-N). This suggested that Lefty2 expression cannot rescue MESP1/2 loss in mesoderm differentiation.

We have previously reported that the enhancer to regulate Mesp1 expression in the early mesoderm (EME) is located between 3.1 kb and 4.4 kb upstream of the Mesp1 translational start site (Haraguchi et al., 2001). Further analysis of the EME using the VISTA tool revealed that there is an ∼340 bp highly conserved region (HCR1) among mammals, including mice, humans, dogs, cows and rats (Fig. S8A,B), which also contained the Mesp1 presomitic mesoderm-specific enhancer (P1-PSME) with T- and E-boxes (Fig. S8C) (Oginuma et al., 2008). As the Wnt and Nodal signaling pathways are essential for early mesoderm formation, Mesp1 may be a target. We found a TCF/LEF-binding site (T/LBS2) in the HCR1, but no Smad-binding site (Fig. S8C). In the EME, there were three more TCF/LEF-binding sites, making a total of four TCF/LEF-binding sites (Fig. 7A). To evaluate the possibility that Mesp1 is a direct target of the Wnt pathway, we examined enhancer activity using a transient transgenic assay. We introduced mutations into the four TCF/LEF-binding sites of the P1L-NB vector, which is a lacZ reporter containing the EME with the promoter region, designated as P1L-NB mutT/LBS. The P1L-NB reporter activity was not observed at streak stages (0 lacZ-positive embryos out of 15 transgenic embryos: 0/15), but was observed from bud stages in the mesodermal cells that emerged from the posterior PS (Fig. 7B: 8/14). The activity was maintained in the PSM from the headfold stages to early somitogenesis stages (Fig. 7C, 11/19). The reporter activity of P1L-NB mutT/LBS was markedly reduced compared with that of P1L-NB (Fig. 7D,E). As a previous report suggested that the sequence further upstream of the EME is also important for Mesp1 expression (Oginuma et al., 2008), the HindIII-NcoI fragment was inserted to P1L-NB mutT/LBS, designated as P1L-HB mutT/LBS. However, its reporter activity was almost the same as that of P1L-NB mutT/LBS (Fig. 7F), indicating that the TCF/LEF sites are responsible for the enhancer activity. In order to further examine the role of canonical Wnt signaling in the enhancer activity, the enhancer and promoter fragments of P1L-NB and P1L-NB mutT/LBS were ligated into the Luciferase reporter plasmid, designated as P1Luc-NB and P1Luc-NB mutT/LBS, respectively. These reporters were transfected into ES cells, and their activities were examined with and without Wnt activation. P1Luc-NB reporter activity was upregulated upon canonical Wnt signaling activation by treatment with a GSK3 inhibitor, CHIR99021, whereas P1Luc-NB mutT/LBS reporter activity was one-third that of P1Luc-NB (Fig. 7G). We further conducted a ChIP assay to examine whether the direct binding of LEF1 to the putative binding sites is responsible for the activation. ES cells expressing LEF1 were examined with and without Wnt activation. As a result, LEF1 was able to interact with these TCF/LEF binding sites (Fig. 7H) regardless of Wnt activation. This suggested that Mesp1 transcription is regulated through the TCF/LEF-binding sites in the EME upon canonical Wnt signaling activation.

In order to evaluate the EME activity for Mesp1 expression during early mesoderm formation in vivo, we established EME-deleted mice (Figs S1B and S8A,B). Mesp1 expression in the homozygously deleted embryos (Mesp1^EMEΔ1/Δ1) was not different from that in the controls at the EB stage (Fig. 8A,E). In the control embryos, Mesp1 expression decreased around the EHF stage (Fig. 8C) and was restricted to newly forming somites in the somitogenesis stage (Fig. 8D). The Mesp1 expression in Mesp1^EMEΔ1/Δ1 embryos decreased from the LB to EHF stage, earlier than in wild type (Fig. 8B,C,F,G), and almost no expression was observed at the somitogenesis stage (Fig. 8H), confirming that the EME also contained a somite-specific enhancer (PSME) for Mesp1. Mesp1^EMEΔ1/Δ1 mice were fertile and displayed no obvious phenotypes (Table S4). As Mesp2 expression is upregulated in Mesp1 null, we examined whether Mesp1 downregulation in EME-deleted embryos causes the upregulation of Mesp2 expression. However, Mesp2 expression was unaltered compared with controls during early mesoderm formation (Fig. 8I-N), which suggested that the EME does not affect Mesp2 expression, consistent with the previous report (Oginuma et al., 2008). Therefore, the maintenance, but not the onset, of Mesp1 transcription, might be regulated through the EME. Further analysis is necessary to reveal which transcriptional factors regulate the onset of Mesp1 expression in early mesoderm formation.

As BAC-Tg experiments suggested that not only the EME, but also regions further upstream, including highly conserved region 2 (HCR2), is important for Mesp1 expression in the early mesoderm (Oginuma et al., 2008), we further established mouse lines with deletion of upstream of the EME (Mesp1^EMEΔ2) and a larger deletion including a region upstream of the EME (Mesp1^EMEΔ1Δ2) (Fig. S1B). The Mesp1 expression in Mesp1^EMEΔ2/Δ2 embryos did not differ from that in controls (Fig. 9A,B), and Mesp1^{EMEΔ1Δ2/Δ1Δ2} embryos displayed similarly downregulated Mesp1 expression to Mesp1^EMEΔ1/Δ1 embryos (Figs 8B,F and 9A,C). As this region is between Mesp1 and Mesp2 genes, we examined the possibility that Mesp2 expression is regulated via this region. The Mesp2 expression in Mesp1^EMEΔ2/Δ2 embryos did not differ from that in controls (Fig. 9D,E), but it was downregulated in Mesp1^{EMEΔ1Δ2/Δ1Δ2} embryos (Fig. 9D,F). Mesp1^EMEΔ2/Δ2 mice were fertile and displayed no obvious phenotypes; however, Mesp1^{EMEΔ1Δ2/Δ1Δ2} was embryonic lethal (Table S4). In order to examine the cause of lethality, we obtained embryos from Mesp1^EMEΔ1Δ2/+ intercrossing at E8.5 and E9.5 stages. Mesp1^{EMEΔ1Δ2/Δ1Δ2} embryos at E8.5 (Fig. 9J,Q) exhibited developmental delay of ∼1 day with a less developed headfold and thickened mesodermal layer at the PS compared with similarly sized controls dissected at E7.5 (Fig. 9G,H). Mesp1^{EMEΔ1Δ2/Δ1Δ 2} embryos at E9.5 had a smaller head and turning defect (Fig. 9L,P) compared with controls (Fig. 9I,M). In addition, Mesp1^{EMEΔ1Δ2/Δ1Δ 2} embryos displayed bifurcation of the cardiac primordia (Fig. 9K), similar to Mesp1^Neo/Neo embryos (Fig. 4B), which was likely the cause of lethality. Of note, the Mesp1^{EMEΔ1Δ2/Δ1Δ 2} embryos at E9.5 had more than 10 somites, albeit smaller in size, which exhibited rostral-caudal polarity (Fig. 9M,P). Indeed, Mesp2 was expressed in Mesp1^{EMEΔ1Δ2/Δ1Δ 2} embryos at the lateral plate mesoderm, similar to controls (Fig. 9N,O,Q). This suggested that Mesp2 expression is regulated through the combination of EME and HCR2 regions during early mesoderm formation, and the regulation is switched to another enhancer, likely a previously reported enhancer, during somitogenesis (Haraguchi et al., 2001; Yasuhiko et al., 2006, 2008) (Fig. S9).

MESP1 was previously proposed as a master regulator of cardiac progenitors. However, our CRISPR/Cas9-mediated Mesp1 KO and Mesp1-cre mice without the PGK-neo cassette did not exhibit any phenotypes, including cardiac morphogenesis, although this does not mean that MESP1 is not important for cardiac formation. The CRISPR/Cas9-mediated Mesp1/Mesp2 dKO confirmed the previous report (Figs 1 and 3) (Kitajima et al., 2000) that both MESP1 and MESP2 play important roles in early mesoderm formation. Although there is a report suggesting the presence of a few MESP1-specific target genes that can regulate the migration of cardiovascular progenitors (Chiapparo et al., 2016), MESP2 can fully rescue MESP1 functions directly or indirectly when Mesp1 is knocked out. As MEPS1 and MESP2 are both important for early mesoderm formation, compensatory mechanisms exist and are activated when each gene is knocked out. Indeed, Mesp1 and Mesp2 single mutant mice demonstrated expression changes for the other gene (Saga et al., 1999; Takahashi et al., 2005, 2007). The existence of a heterologous promoter like the PGK promoter was previously suggested to affect expression of neighboring genes (Fiering et al., 1995; Olson et al., 1996; Seidl et al., 1999; West et al., 2016). In the Mesp1/Mesp2 locus, insertion of the PGK-neo cassette into one gene may have affected the regulation of a neighboring gene, but it more likely prevented the rescue expression of the other, thereby making the phenotype more severe than that of the KO created without using the PGK-neo cassette. The mechanism of how Mesp1/Mesp2 genes rescue each other when one is knocked out remains unknown. One possibility is that MESP1 and MESP2 act as transcription factors to suppress each other. To assess this possibility, we looked for putative E-boxes in the HCRs between Mesp1 and Mesp2 genes, where MESP1 and MESP2 may interact. We found one E-box in HCR1 (Fig. S8C) and two more in the EME. This is consistent with the report that MESP1 can interact around the EME in differentiated mouse ES cells (Bondue et al., 2008). However, deletion of the EME did not affect Mesp2 expression (Fig. 8I-N, Fig. S1), suggesting that even if MESP1 interacts with the E-boxes in the EME and can suppress Mesp2 transcription, its activity is not as strong as the rescue activity of the Mesp2 gene in Mesp1 KO mice. The other possibility is that there are common enhancers between Mesp1 and Mesp2 that compete for transcription factors. The T-box transcriptional factors EOMES and T/BRACHYURY have been reported to be the strongest candidate activators of Mesp1 and Mesp2 expression during early mesoderm formation. RNA-seq analyses revealed that Mesp1 and Mesp2 expression was repressed in the differentiating Eomes and T double KO ES cells (Tosic et al., 2019), and that EOMES and T/BRACHYURY directly interacted with the T-box in the Mesp1 and Mesp2 promoter and EME of Mesp1 to regulate Mesp1 and Mesp2 expression (Costello et al., 2011; David et al., 2011; Tosic et al., 2019; van den Ameele et al., 2012). We looked for putative T-boxes in the HCRs between Mesp1 and Mesp2 genes. In addition to the previously reported T-boxes, we found two additional T-boxes in HCR2 and HCR3 (Figs S8D,E and S9). Their positions were consistent with the ChIP-seq data using antibodies against EOMES and T/BRACHYURY in differentiating ES cells (Tosic et al., 2019). This suggested that Mesp1 and Mesp2 genes share EOMES and T/BRACHYURY for their activation at the early mesoderm formation stage through several shared T-boxes. A recent study revealed that once the promoter of a gene is removed or mutated, the enhancer for the gene can be used to activate the neighboring genes (Oh et al., 2021). Similar to this case, in the Mesp1 and Mesp2 loci, once one gene is knocked out, EOMES and T/BRACHYURY are no longer used for activation, and the other gene can use them via shared enhancers to promote its expression. This is consistent with results from our study. Deletion of the EME downregulated only Mesp1 expression (Fig. 8A-H, Fig. S1), whereas simultaneous deletion of the EME and HCR2 also downregulated Mesp2 expression during the early mesoderm formation stage (Fig. 9D-F, Fig. S1). This suggested that the EME is shared between Mesp1 and Mesp2 genes, and the T-box in the HCR2 is also important for Mesp2 gene expression. Importantly, removing both enhancers resulted in similar cardiac morphological abnormality to that in Mesp1^Neo/Neo embryos (Figs 4B and 9K). The total amount of MESP1 and MESP2 decreased in Mesp1^{EMEΔ1Δ2/Δ1Δ2} and Mesp1^Neo/Neo embryos, which caused the delay in mesodermal cell differentiation and migration out from the PS, affecting the development of head mesenchymal cells and the cardiac primordia. It is still unclear why the deletion of the EME affected Mesp1 expression, but deletion of the EME or HCR2 region alone did not alter Mesp2 expression (Figs 8A-N, 9D,E, Fig. S1). One possibility is that there are specific mechanisms for activating Mesp2 expression through cooperative activation of these two enhancer regions. The other possibility is the distance between the promoter and enhancer, which has been proposed to affect transcriptional activity (Fukaya et al., 2016). As the EME and HCR2 are closer to the Mesp1 promoter (Figs S1, S8 and S9), this may be sufficient to alter the expression level. However, the Mesp2 promoter is further away from the EME and HCR2, and removal of the two enhancers, especially T-box binding sites, is needed to observe the alteration. Further studies, such as deletion of HCR3, may provide more insight into the mechanisms of compensation between Mesp1 and Mesp2 genes.

The mesodermal cells of Mesp1/Mesp2 dKO were prevented from further differentiation and accumulated at the PS. Canonical Wnt and Nodal signaling is important to maintain the mesoderm progenitor cells, and must be downregulated for further differentiation of mesodermal cells. As the expression of T-box transcriptional factors Eomes and T in the PS is regulated by both Nodal and Wnt signaling pathways (Arnold et al., 2000; Barrow et al., 2007; Brennan et al., 2001; Simon et al., 2017; Yamaguchi et al., 1999), these signaling pathways may regulate Mesp1 and Mesp2 expression through the regulation of EOMES and BRACHYURY. Lefty2 is a direct target of Nodal signaling and suppresses the signaling pathway, and Lefty2 KO exhibited accumulation of the mesodermal tissue, similar to Mesp1/Mesp2 dKO (Fig. 5C-F; Meno et al., 1999). Lefty2 was also proposed as a downstream target of MESP1 and MESP2 (Fig. 3E,F; Costello et al., 2011; Kitajima et al., 2000; Lindsley et al., 2008). We therefore hypothesized that LEFTY2 is one of the factors responsible for early mesoderm specification. However, knock-in of Lefty2 cDNA into the Mesp1 locus did not rescue the Mesp1/Mesp2 dKO phenotypes. Although Lefty2 mRNA in Mesp1^{Lefty2/Lefty2};Mesp2 KO embryos was detected at comparable levels with those in controls (Fig. 6A,B), the LEFTY2 protein level may have been insufficient to be functional. On the other hand, it is more likely that MESP1 and MESP2 transcription factors induce other essential factor(s) for further differentiation, and this role cannot be substituted for by LEFTY2. LEFTY2 may regulate early mesoderm formation in parallel with MESP1 and MESP2. Detailed comparison between Mesp1/Mesp2 dKO and Lefty2 KO may reveal the mechanisms that specify early mesoderm differentiation.

In this study, we proposed that Mesp1 can be regulated by the canonical Wnt signaling pathway via the EME. The transient transgenic assay using the Mesp1 EME reporter revealed that mutations in TCF/LEF binding sites in the EME markedly reduced the reporter activity, but residual expression at the anterior part of the migrating mesoderm was observed (Fig. 7D). Deletion of the EME downregulated Mesp1 expression only after the LB stage (Fig. 8A-H). This suggests that the canonical Wnt signaling pathway has a role not in the induction, but rather in the maintenance of Mesp1 expression in the early mesoderm. As mentioned above, the EME also promotes Mesp2 expression and it is possible that its expression is also regulated by the canonical Wnt signaling pathway via TCF/LEF-binding sites in the EME. MESP1 has been proposed to control the EMT by regulating the expression of Snai1 (Lescroart et al., 2014; Lindsley et al., 2008), and we confirmed that Snai1 was downregulated in Mesp1/2 dKO embryos (Fig. 3O,P). Once the early mesoderm cells go through the EMT, the cells migrate out and away from the PS, which is the source of canonical WNTs. The downregulation of Mesp1 after mesodermal cells migrate out of the PS may be caused by the reduced amount of WNT proteins received by cells.

As mentioned above, deletion of the two independent enhancers, EME and HCR2, caused the downregulation of Mesp1 and Mesp2 expression, and early mesoderm formation defects. However, this mutant formed segmented somites, suggesting that the transcriptional regulation of Mesp1 and Mesp2 is switched from the early mesoderm formation stage to the somitogenesis stage. Activity of the Mesp2 enhancer in somitogenesis was identified around the transcriptional start site, and demonstrated to be regulated by a T-box transcriptional factor, TBX6, and the intracellular domain of Notch (NICD) (Haraguchi et al., 2001; Yasuhiko et al., 2006, 2008). The T-box in the EME was essential for Mesp1 expression during somitogenesis (Oginuma et al., 2008), which is also activated by TBX6 (Sadahiro et al., 2018). Thus, the switch of transcriptional regulation of Mesp1 and Mesp2 from the early mesoderm formation stage to the somitogenesis stage can be due to the use of T-box binding sites by different T-box transcriptional factors. Further investigation is needed to understand how these two genes are activated and cooperatively function in early mesoderm formation, specification and somitogenesis.

Mice carrying the Mesp1^{Cre floxed Neo} allele were established as described previously (Saga et al., 1999) using the floxed pgk-neo cassette instead of the pgk-neo cassette. Mesp1^{Cre floxed Neo} heterozygous mice were crossed with CAG-cre mice to remove the pgk-neo cassette. The resultant allele was designated as Mesp1^CreΔNeo. C57BL6, B6C3F1 and MCH strain mice were purchased from CLEA Japan. All mouse experiments were approved by the Animal Experimentation Committee at the National Institute of Genetics (Permit Number 27-13).

The bicistronic expression vector expressing sgRNA and hCas9 mRNA (pX330) (Cong et al., 2013) was purchased from Addgene (plasmid #42230) pX330 was linearized with BbsI, gel purified, and then ligated with an annealed pair of oligos for Mesp1-1 (5′-caccGACTCGGAGCGCGGCTCGCAC-3′ and 5′-aaacGTGCGAGCCGCGCTCCGAGTC-3′), Mesp1-5 (5′-caccGCTACAGCGGACCCAATGGTC-3′ and 5′-aaacGACCATTGGGTCCGCTGTAGC-3′), Mesp1 exon1-3′ (5′-caccGCCGAGTCGCCGCAGAATCGT-3′ and 5′-aaacACGATTCTGCGGCGACTCGGC-3′), Mesp1EME5′ (5′-caccGCAGTTCGGTCTGGAGTAAGC-3′ and 5′-aaacGCTTACTCCAGACCGAACTGC-3′), Mesp1EME3′ (5′-caccGGAGGGGCTGCCTATCTTAC-3′ and 5′-aaacGTAAGATAGGCAGCCCCTCC-3′), Mesp1EMEΔ2 (5′-caccGTGGAGATCTCTGCCCGCGTC-3′ and 5′-aaacGACGCGGGCAGAGATCTCCAC-3′), Mesp2-2 (5′-caccGGCGCTGGGCGCGTTCGTCG-3′ and 5′-aaacCGACGAACGCGCCCAGCGCC-3′), Mesp2-5 (5′-caccGTAGTGCGCGTGCTACGGGC-3′ and 5′-aaacGCCCGTAGCACGCGCACTAC-3′), Lefty2 exon3-1 (5′-caccGCGTCCCCTGCGCCGCGAAA-3′ and 5′-aaacTTTCGCGGCGCAGGGGACGC-3′), Lefty2 exon3-3 (5′-caccGAGGGGACGCCGGACGGCAAG-3′ and 5′-aaacCTTGCCGTCCGGCGTCCCCTC-3′), Wnt3-182 (5′-caccGATGCACGAAGGCCGATTCAC-3′ and 5′-aaacGTGAATCGGCCTTCGTGCATC-3′), Wnt3-1229 (5′-GTCCTTGAGGAAGTCGCCAA-3′ and 5′-aaacTTGGCGACTTCCTCAAGGAC-3′) and Wnt3-1569 (5′-caccGCGTAGATGCGAATACACTCT-3′ and 5′-aaacAGAGTGTATTCGCATCTACGC-3′), and the resulting vectors were designated as pX330/Mesp1-1, pX330/Mesp1-5, pX330/Mesp1 exon1-3′, pX330/Mesp1EME5′, pX330/MespEME3′, pX330/Mesp1EMEΔ2, pX330/Mesp2-2, pX330/Mesp2-5, pX330/Lefty2 exon3-1, pX330/Lefty2 exon3-3, pX330/Wnt3-182, pX330/Wnt3-1229 and pX330/Wnt3-1569, respectively. pX330/Mesp1-1, pX330/Mesp1 exon1-3′, pX330/Mesp1EME5′-3, pX330/MespEME3′-1, pX330/Mesp1EMEΔ2-1 pX330/Lefty2 exon3-1, pX330/Lefty2 exon3-3, pX330/Wnt3-182, pX330/Wnt3-1229 and pX330/Wnt3-1569 were then used as a template to amplify the PCR product using the T7 promoter-attached forward primer for Mesp1-1 (5′-TTAATACGACTCACTATAGGGACTCGGAGCGCGGCTCGCAC-3′), Mesp1 exon1-3′ (5′-TTAATACGACTCACTATAGGGCCGAGTCGCCGCAGAATCGT-3′), Mesp1EME5′ (5′-TTAATACGACTCACTATAGGGGCAGTTCGGTCTGGAGTAAGC-3′), Mesp1EME3′ (5′-TTAATACGACTCACTATAGGGGGAGGGGCTGCCTATCTTAC-3′), Mesp1EMEΔ2 (5′-TTAATACGACTCACTATAGGGGTGGAGATCTCTGCCCGCGTC-3′), Lefty2 exon3-1 (5′-TTAATACGACTCACTATAGGGCGTCCCCTGCGCCGCGAAA-3′), Lefty2 exon3-3 (5′-TTAATACGACTCACTATAGGGAGGGGACGCCGGACGGCAAG-3′), Wnt3-182 (5′-TTAATACGACTCACTATAGGGATGCACGAAGGCCGATTCAC-3′), Wnt3-1229 (5′-TTAATACGACTCACTATAGGGTCCTTGAGGAAGTCGCCAA-3) and Wnt3-1569 (5′-TTAATACGACTCACTATAGGGCGTAGATGCGAATACACTCT-3′), respectively, and common reverse primer (5′-AAAAAAGCACCGACTCGGTG-3′). sgRNAs and hCas9 mRNA were synthesized as described previously (Ajima et al., 2017). The following Alt-R CRISPR-Cas9 crRNAs and tracrRNA were purchased from Integrated DNA Technologies: Mesp1-5′ crRNA (5′-AltR1-CUACAGCGGACCCAAUGGUCGUUUUAGAGCUAUGCU-AltR2-3′), Mesp1-3′-2 crRNA (5′-AltR1-CCAGGAUCUACUCCCUCCGAGUUUUAGAGCUAUGCU-AltR2-3′), Mesp2-2 crRNA (5′-AltR1- GGCGCUGGGCGCGUUCGUCGGUUUUAGAGCUAUGCU-AltR2-3′), Mesp2-5 crRNA (5′-AltR1-GUAGUGCGCGUGCUACGGGCGUUUUAGAGCUAUGCU-AltR2-3′) and Mesp2-6 crRNA (5′-AltR1-ACAAUAAGGAGGUGUCACCCGUUUUAGAGCUAUGCU-AltR2-3′). GeneArt Platinum Cas9 Nuclease was purchased from Thermo Fisher Scientific.

To establish the Mesp1^Lefty2 targeting vector (TV), a 500 bp fragment upstream of the first ATG of the Mesp1 gene as a 5′ long arm (LA), a 500 bp fragment downstream of the Mesp1 exon1-3′ gRNA recognition site as a 3′ long arm and a Lefty2-coding sequence fragment were amplified with KOD Fx Neo (TOYOBO) using C57BL6J genomic DNA as a template for the 5′ LA and 3′ LA, and pBS/Lefty2 plasmid as a template for the Lefty2-coding sequence. These three fragments with a XhoI-EcoRI-digested pBluescript SK+ fragment were joined using NEBuilder HiFi DNA assembly (New England Biolabs). Single-strand DNA of the Mesp1^Lefty2 targeting vector was made using the Guide It long ssDNA Production System (TaKaRa) with phosphorylated T7 and T3 primers.

To establish the transgenes, mutations of the four putative TCF/Lef-binding sites (T/LBS) were introduced into the P1L-NB construct (Haraguchi et al., 2001) by PCR using the following primer sets with mutations, as underlined: mT/LBS-1 Fw (5′-GCATAGCACATTGGCGTAGCAAGGC-3′)+mT/LBS-1 Rv (5′-CGCCAATGTGCTATGCACAGTCCCT-3′); mT/LBS-2 Fw (5′-GGAGCGCCAATGGGCAGAGGAAGCAG-3′)+mT/LBS-2 Rv (5′-GCCCATTGGCGCTCCTTATCTGGCCAGT-3′); mT/LBS-3 Fw (5′-TGCCTAAATTGGCGGTTAATAAAATACAAC-3′)+mT/LBS-3 Rv (5′-CCGCCAATTTAGGCAGACATGTTT-3′); and mT/LBS-4 Fw (5′-AACCGCCAATTTCTATTAACAAGATGC-3′)+mT/LBS-4 Rv (5′-AATAGAAATTGGCGGTTGTATTTTATTAAC-3′). The construct with all four T/LBS mutations was designated as P1L-NB mutT/LBS. P1L-HB mutT/LBS was made by inserting the HindIII-NcoI fragment into the P1L-NB mutT/LBS construct. The HindIII-NotI fragment of P1L-NB, P1L-NB mutT/LBS or P1L-HB mutT/LBS was gel extracted using the QIAQuick Gel Extraction kit and used for embryo injection. To establish Luciferase reporters, NcoI fragments of P1L-NB and P1L-NB mutT/LBS were blunted and subcloned into the EcoRV site of the pGL4.10[luc2] vector (Promega).

To establish the 3xTy1-mLef expression plasmid, first 3xTy1 oligo DNAs (5′-CTAGCACCATGGAGGTGCACACCAACCAAGATCCGCTTGATGCTGAAGTTCATACAAACCAGGATCCCCTCGATGCCGAAGTCCATACTAATCAGGACCCACTGGACGCAA-3′ and 5′-AGCTTTGCGTCCAGTGGGTCCTGATTAGTATGGACTTCGGCATCGAGGGGATCCTGGTTTGTATGAACTTCAGCATCAAGCGGATCTTGGTTGGTGTGCACCTCCATGGTG-3′) were annealed and cloned into NheI and HindIII-digested pcDNA3.1 (Invitrogen). The resulting plasmid was designated as pcDNA3.1-3xTy1. The Lef1 cDNA fragment was amplified by PCR with HindIII-3xTy1-mLef1 Fw (5′-ACTGGACGCAAAGCTTCCCCAACTTTCCGGAGGAGG-3′) and 3xTy1-Lef1-NotI Rv (5′-TAGACTCGAGCGGCCGCTCAGATGTAGGCAGCTGTCATTCTGG-3′) primers with PrimeStar GXL DNA polymerase (TaKaRa) using cDNA of 3 µM CHIR99021-treated ES cells as a template. The fragment was subcloned into HindIII and NotI-digested pcDNA3.1-3xTy1 using In-fusion HD Mutagenesis (Clontech) and designated as pcDNA3.1-3xTy1-Lef1. The 3xTy1-Lef1 fragment was amplified by PCR with pcDNA3.1-3xTy1-Lef1 as a template, and subcloned into the XhoI-NotI-digested pPBhCMV_1HApA vector using In-fusion HD Mutagenesis (Clontech).

B6C3F1 (C57BL/6N×C3H/HeN) female mice were super-ovulated and mated with B6C3F1 males, and fertilized embryos were collected from oviducts. To establish Mesp1^ΔNeo lines, B6C3F1 (C57BL/6N×C3H/HeN) female mice were super-ovulated, and oocytes were collected and used for in vitro fertilization with sperm from a Mesp1^Neo/+ male.

To make gene-targeted mice, mixtures of different concentrations of pX330 vectors, sgRNA, hCas9 mRNA, GeneArt Platinum Cas9 Nuclease and a circular or single-strand targeting vector were dissolved in injection buffer [10 mM Tris-HCl and 0.1 mM EDTA (pH 7.5)], and injected into the pronucleus or cytoplasm of fertilized eggs in M2 medium (Sigma). For the transient transgenic assay, gel-purified transgene fragments were dissolved in injection buffer and injected into the pronucleus of fertilized eggs in M2 medium (Sigma). The injected zygotes were cultured in KSOM (Millipore) at 37°C under 5% CO₂ until the two-cell stage after 1.5 days. Thereafter, 20-32 two-cell stage embryos were transferred into the uterus of pseudo-pregnant MCH females at 0.5 dpc. Embryos were dissected on the indicated day or born naturally. Mouse lines established by CRISPR/Cas9-mediated genome editing were backcrossed several times with C57BL6 or B6C3F1 mice to eliminate any off-target mutations.

For direct sequencing of CRISPR/Cas9-mediated Mesp1/Mesp2 KO mice, ∼100 ng of the PCR products of the Mesp1 allele with Mesp1 GT Fw (5′-AGTCCTGGCCATAGGTGCCTGACTTACT-3′)+Mesp1 GT Rv (5′-CGAGTGTGCGCATACGTAGCTTCTCC-3′) and Mesp2 allele with Mesp2-L5 (5′-GGGGACCACCTCACACCTTTAGTCCAG-3′)+Mesp2-R4 (5′-AGGCTAGGACCAAGCATCTGAGCCTGTG-3′) were treated with ExoSAP-IT (Affymetrix), reacted with the Big Dye Terminator v3.1 Cycle sequencing kit (Applied Biosystems) with Mesp1 GT Fw and Mesp2-L4 (5′-CCGGTCCAGCTTCCCAGAGTCACACCAG-3′), respectively, according to the manufacturer's instructions, and then analyzed with the ABI 3130xl genetic analyzer (Applied Biosystems). The peaks were read manually or using TIDE web tool (https://tide.nki.nl). When the direct sequencing results of F0 analysis were unclear, the PCR products were subcloned into the pT7Blue-2 T-vector (Novagen) and sequenced with the T7 primer. For screening of the Mesp1^Lefty2 targeted mice, the following PCR primer sets were used: Lefty2KI 5′Fw (5′-ACCAAAGCAGCAGAGCTCCAGTC-3′)+Lefty2KI 5′Rv (-CTCAGGACCTGTTCCTCGGTCATC-3′) and Lefty2KI 3′Fw (5′-GTGTTGCCTCAGAGATGACCTCCTT-3′)+Lefty2KI 3′Rv (5′-AGGGTTGGAATGGTACAGTCTGGAT-3′). Genotyping primers for established mouse lines are listed in Table S5.

For embryonic day 9.5 (E9.5) and younger embryos, whole-mount in situ hybridization and probe synthesis were performed as previously described (Biris and Yamaguchi, 2014). For E10.5 embryos, the InsituPro system (M&S Instruments) was used for whole-mount in situ hybridization according to the manufacturer's instructions. For cloning of the Wnt3, Wnt3a and Eomes probes, total RNA from TT2 ES cells was isolated with the RNeasy mini kit (Qiagen), reverse transcription was conducted with SuperScript III (Invitrogen), cDNA was amplified with the primers listed below, and resulting PCR products were subcloned into SalI- and EcoRI-digested pBluescript SK+ using the In-Fusion HD cloning kit (Clontech). Primers used were: for the Wnt3 probe, Fw (5′-CCCCCTCGAGGTCGACTTCTAATGGAGCCCCACCTGCTC-3′)+Rv (5′-CGGGCTGCAGGAATTCTACTTGCAGGTGTGCACATCGTAGA-3′); for the Wnt3a probe, Fw (5′-CCCCCTCGAGGTCGACCGGCGATGGCTCCTCTCG-3′)+Rv (5′-CGGGCTGCAGGAATTCTCAGTGCGCGCGTTATGCC-3′); for the Eomes probe, Fw (5′-CCCCCTCGAGGTCGACTAAAGCATGCAGTTGGGAGAGCAG-3′)+ Rv (5′-CGGGCTGCAGGAATTCTAGGGACTTGTGTAAAAAGCATAATAAGC-3′). The other probes used in this study were prepared as described previously.

E18.5 mouse embryos were fixed with 100% ethanol. For genotyping, PCR was performed using a piece of embryonic skin digested with proteinase K (Roche). Alcian Blue and Alizarin Red staining were performed as described previously (Saga et al., 1997; Takahashi et al., 2000).

E7.5 mouse embryos were dissected in ice-cold PBS and photographed, and a small piece of extra-embryonic tissue was trimmed for genotyping. The remaining tissue was frozen in liquid nitrogen and stored at −80°C until preparing total RNA. Two stage-matched embryos were lysed together with 500 µl of TRIzol Reagent (Thermo Fisher Scientific), heated at 55°C for 5 min, mixed with 100 µl of chloroform, centrifuged at 12,000 g at 4°C for 15 min and the supernatant collected. Next, 300 µl of ethanol was added to the supernatant, which was applied to an RNeasy MinElute Spin Column and total RNA was harvested according to the manufacturer's instructions (RNeasy Micro Kit, Qiagen). Total RNA was used to make the RNA-seq library using the KAPA Stranded mRNA-Seq Kit Illumina platform (KAPA) and barcoded with DNA adapters using the KAPA single-indexed adaptor kit (KAPA). Quality assessment was performed using the Agilent DNA1000 Kit (Agilent). The libraries were 100 bp pair-end and were sequenced on an Illumina Hiseq2500 (Illumina).

For all libraries, low-quality sequences and adapters were trimmed or removed using Cutadapt (version 3.4) (Martin, 2011) with the following options: ‘-j 8 -m 36 -q 20 -a GATCGGAAGAGCACACGTCTGAACTCCAGTCAC’. The raw reads and processed reads were checked using FastQC (version 0.11.9, http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). For preparation for mapping reads to the mouse reference genome, the ensembl mouse reference genome (release-102, Mus_musculus.GRCm38.dna.primary_assembly.fa.gz) and the gene annotation file in General Transfer Format (GTF) (Mus_musculus.GRCm38.102.gtf.gz) were downloaded from the ensembl ftp site (http://ftp.ensembl.org/), and non-chromosomal sequences (scaffolds) and the related annotation data were removed in the reference genome and the gene annotation GTF file, respectively. To increase the mapping accuracy of splicing reads, splice site and exon information were extracted from the gene annotation GTF file using the Python scripts hisat2_extract_splice_sites.py and hisat2_extract_exons.py, respectively, from the HISAT2 (version 2.2.1) package (Kim et al., 2015). The HISAT2 index files of the reference genome were built including the extracted genomic information using ‘hisat2-build’ command with options: ‘--ss’ and ‘--exon’. Clean reads were then mapped to the HISAT2 index files using HISAT2 with default options. The obtained Sequence Alignment Map (SAM) files were sorted by genomic coordinates and converted to Binary Alignment Map (BAM) files using the SAMtools (version 1.13) (Li et al., 2009) ‘sort’ command with option: ‘-O BAM’. Raw read counts per gene were calculated using featureCounts (version 2.0.3) (Liao et al., 2014) with options: ‘-s 2 -T 8 -t exon -g gene_id -a Mus_musculus.GRCm38.102.gtf’. Only protein-coding genes were used for the downstream analysis of this study. Normalized counts and statistical values for differential gene expression analysis were calculated using the Bioconductor DESeq2 packages (version 1.32.0) (Love et al., 2014) in R (version 4.1.0) (https://cran.r-project.org/). To assess gene expression correlation between samples, the pair-wise scatterplot was produced using log2 (the normalized counts+1), the ‘cor’ function with the parameters “method=‘spearman’ and use=‘pairwise.complete.obs’”, and ggplot2 packages (version 3.3.5) (https://cran.r-project.org/web/packages/ggplot2/citation.html) in R. For PCA, the variance-stabilizing transformed (vst) normalized counts were calculated using the vst function of the DESeq2 with the default settings and PCA was performed with the top 500 most variable genes using the DESeq2 plotPCA function. To determine DEGs, MAplots for each comparison of sample groups (controls vs Mesp1^Δneo/Δneo in EB or LB) were produced using the results of the DESeq2 analysis with ggplot2. DEGs were detected using the DESeq2 with the cut-off criteria of an adjusted P-value <0.01.

X-gal staining of whole embryos to detect lacZ expression has been described previously (Saga et al., 1992). The embryos were subsequently examined for the presence of the transgene by PCR analysis with the following primer sets: P1LB-NB GT Fw (5′-GGTCCAGGTGGAGCAGACTGGACTA-3′)+LacZ-R5 (5′-GCCAGGGTTTTCCCAGTCACGACG-3′).

P1L-NB –luc or P1L-NB mutT/LBS-luc reporters with pRL-tk (Promega) internal control vector were transfected into feeder-depleted ES (TT2) cells with Lipofectamine 2000 (Invitrogen), and cultured with or without 3 µM CHIR99021 in DMEM supplemented with 10% FBS, 2 mM L-Glutamine, 1×MEM NEAA, 1×penicillin-streptomycin, 55 µM 2-mercaptoethanol, 3 mM adenosine, cytidine, guanosine, and uridine, and 1 mM thymidine. The samples were harvested 48 h after transfection and the Luciferase assay was conducted using the Dual Luciferase Reporter Assay System (Promega) according to the manufacturer's instructions.

To establish the DOX-inducible 3xTy1-Lef1-expressing ES cell lines, ES (TT2) cells on Neo^r feeder cells were transfected with pPBhCMV-3xTy1-Lef1, pPBCAGrtTA-IN and pBase plasmids using Lipofectamine 2000 (Invitrogen), and then selected with 150 µg/ml of geneticin for 6 days and picked up. The feeder-depleted DOX-inducible ES cells were cultured on gelatin-coated plates with or without 3 µM CHIR99021 in DMEM supplemented with 10% FBS, 2 mM L-Glutamine, 1×MEM NEAA, 1×penicillin-streptomycin, 55 µM 2-mercaptoethanol, 3 mM adenosine, cytidine, guanosine and uridine, and 1 mM thymidine for 1 day, cultured with or without 1 µg/ml of doxycycline for 8 h. Chromatin was fixed with 1% formaldehyde for 10 min, and cells were then collected and washed with ice-cold PBS. The cell pellets were suspended with SDS lysis buffer [1% SDS, 10 mM EDTA and 50 mM Tris (pH 8.1)], sonicated with Vibra-cell VCX-130 (80% amplitude, 15 s×20 times; Sonic) and centrifuged. The supernatants were aliquoted as the input, and the remaining supernatants were diluted ten times with dilution buffer [0.01% SDS, 1.1% TritonX-100, 1.2 mM EDTA, 16.6 mM Reis-HCl (pH 8.1) and 167 mM NaCl] with a Complete mini proteinase inhibitor tablet (Roche). The lysates were divided into two tubes and incubated with anti-Ty1 antibody (diagenode) or normal mouse IgG (Santa Cruz) overnight at 4°C with rotation, and with Dynabeads-ProteinG (Invitrogen)/sonicated salmon sperm DNA (BioDynamics laboratory) mixture for a further 4 h. The beads were washed with low-salt wash buffer [0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl (pH 8.1) and 150 mM NaCl], high-salt wash buffer [0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl (pH 8.1) and 500 mM NaCl], LiCl wash buffer [0.25 M LiCl, 1% IGEPAL-CA630, 1% deoxycholic acid, 1 mM EDTA and 10 mM Tris (pH 8.1)] and TE buffer [10 mM Tris-HCl and 1 mM EDTA (pH 8.0)] twice each. The precipitated complexes were eluted from the beads by incubation with elution buffer (1% SDS and 0.1 M NaHCO₃). To reverse histone-DNA crosslinks, a 1/25 volume of 5 M NaCl was added to the supernatants and inputs, and then incubated overnight at 65°C. These samples were treated with proteinase K and DNA was recovered by phenol/chloroform extraction followed by ethanol precipitation, and then used for real-time PCR (Thermal Cycler Dice, TaKaRa).

We are particularly thankful to Dr Kurumi Fukuda and Makoto Kiso (National Institute of Genetics, Japan) for technical support. We are grateful to Dr Hitoshi Niwa for providing pPBhCMV_1cHApA, pPBCAGrtTA-IN and pBase plasmids. We are grateful to Dr Atushi Toyoda (Advanced Genomics Center, National Institute of Genetics, Japan) for supporting RNA-seq analysis. We are grateful to Danelle Wright for editing the manuscript.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.194613.