A dual genetic tracing system identifies diverse and dynamic origins of cardiac valve mesenchyme

Precise genetic manipulation of cells in vivo advances our understanding of organ development and tissue regeneration. The widely used DNA site-specific recombinase (SSR), Cre recombinase, allows predictable and efficient genomic modifications in Cre-expressing cells and their descendants, providing a powerful tool for genetic lineage tracing and gene manipulation in vivo (Buckingham and Meilhac, 2011; Kretzschmar and Watt, 2012). Nevertheless, Cre is usually driven by a single gene promoter in order to target one cell population during development, diseases or regeneration (Sauer and Henderson, 1988). By Cre-loxP recombination, the conventional single-colored loxP reporter system, such as Rosa26-LacZ, Rosa26-tdTomato or Rosa26-GFP, can be used to label Cre-expressing cells and their descendants (Soriano, 1999; Srinivas et al., 2001; Madisen et al., 2010). Although multicolor reporter systems have also been used for single cell clonal analysis, such as Brainbow or Confetti (Livet et al., 2007; Snippert et al., 2010) and the double markers (MADM) system (Devine et al., 2014), such a lineage-tracing strategy still falls within the Cre-expressing cell population defined by a specific gene promoter. To achieve genetic targeting of two or more distinct cell populations in vivo, other SSR systems that are orthogonal to the conventional Cre-loxP-based system should be explored for genetic engineering. Indeed, several SSR systems other than Cre-loxP have been identified, such as Flpe-Frt or Dre-rox (Rodriguez et al., 2000; Anastassiadis et al., 2009). Moreover, different SSR systems have been simultaneously used for more precise genomic engineering, such as intersectional genetics, sequential genetics or exclusive genetics (Jensen et al., 2008; Fenno et al., 2014; Schonhuber et al., 2014; He et al., 2017), allowing better understanding of cellular and molecular mechanisms in vivo.

Nigri-nox is a newly discovered SSR system, in which Nigri recombinase targets its own recombination DNA site nox (Karimova et al., 2016). The sequence identity (ID) and similarity (SIM) of Nigri with respect to Cre are 17% and 25%, respectively, while the ID and SIM of Dre with respect to Cre are 39% and 49%. The lower ID and SIM may indicate the lower crosstalk between those two systems. (Karimova et al., 2016). For recombination sites, nox is about 35% similar to loxP by DNA sequences. Moreover, in vitro cell experiments show that Nigri specifically and efficiently recombines its target site nox but not loxP (Karimova et al., 2016). Moreover, Nigri-nox recombination has high efficiency and specificity without cross-recombining with other known SSR systems by in vitro experiments. The new reported Nigri-nox is orthogonal to the present Cre-loxP, Flp-frt and Dre-rox. These data indicate that the Nigri-nox system could be used with other SSR systems for more precise in vivo genomic engineering and genetic manipulation of specific cell population.

To explore whether the Nigri-nox system could be used in vivo, we generated a Cdh5-Nigri mouse allele and tested this in vivo to address whether Nigri-nox recombination efficiently and specifically targets endothelial cells and their descendants. We generated a new reporter that allows either Cre-loxP or Nigri-nox recombination on the Rosa26 locus. Using this reporter line coupled with tissue-specific Cre and Nigri recombinases, we simultaneously and genetically targeted two distinct progenitor cells in development and followed their cell fates with distinct lineage-tracing reporters. Taking cardiac valve mesenchyme formation as an example, we traced different developmental origins for mesenchymal cells in cardiac valves of the developing heart. Dual lineage tracing of endocardial cells and epicardial cells showed their dynamic cell replacement in part of tricuspid and mitral valves from early to late embryonic development. Furthermore, we also used Nigri-nox and Cre-loxP systems to simultaneously trace the cell fate of endocardial cells and neural crest cells in the same mouse heart. We provide detailed dynamics of mesenchymal components from distinct developmental origins by using these two orthogonal systems. Our study therefore shows that the Nigri-nox system works efficiently in vivo and could be used together with other SSR systems for precisely delineating cell fates in vivo.

We first generated a new reporter mouse line that responded to Nigri recombinase. By incorporation of the interleaved reporter strategy previously described (He et al., 2017), we arranged one pair of nox sites together with one pair of loxP sites in Rosa26 allele. These two pairs of recombination sites were interleaved in position nox-loxP-nox-loxP such that the recombination of one SSR system removes the recombination site of another SSR system. We knocked the CAG-nox-loxP-Stop (transcriptional stop sequence)-nox-ZsGreen-polyA-loxP-tdTomato sequence into the Rosa26 gene locus and named this reporter line Rosa26-nox-loxP-reporter, R26-NLR (Fig. 1A). For R26-NLR, Cre-mediated loxP recombination led to expression of tdTomato reporter, whereas Nigri-mediated nox recombination resulted in expression of ZsGreen reporter (Fig. 1B). The Cre and Nigri recombinases were driven respectively by different cell-specific promoters. For example, cell type A expresses gene A, which drives Nigri; cell type B expresses gene B, which drives Cre. In the A-Nigri;B-Cre;R26-NLR triple-positive mouse, we could simultaneously and genetically target two distinct cell populations to understand their cell fates in vivo (Fig. 1C).

We next generated Cdh5-Nigri by placing Nigri cDNA in-frame with the start codon ATG of the endogenous Cdh5 gene via homologous recombination (Fig. 1D). Correct targeting was verified by long PCR spanning the homology arms. Cdh5-Nigri was then crossed with R26-NLR to generate Cdh5-Nigri;R26-NLR mouse. We collected both embryonic and postnatal mouse tissues for analysis. Whole-mount epi-fluorescence showed ZsGreen⁺ signals in a vascular pattern in embryonic day (E)14.5 embryos and postnatal (P)7 organs (Fig. 1E-F, Fig. S1B). We did not detect tdTomato⁺ signal in the tissues, indicating that Nigri recombined nox but not loxP sites. Immunostaining for ZsGreen, tdTomato and endothelial cell maker VE-cadherin (VE-cad) on heart sections showed that almost all endothelial cells (>99%) were ZsGreen⁺ (Fig. 1G). We did not detect tdTomato expression in endothelial cells of the heart. Immunostaining for another endothelial cell maker, PECAM, confirmed that almost all PECAM⁺ cells were ZsGreen⁺ (>99%, Fig. 1H). We also performed immunostaining for ZsGreen and tdTomato on serial sections of E15.5 Cdh5-Nigri;R26-NLR heart, but did not find tdTomato⁺ cells (Fig. S3E). Efficient ZsGreen labeling was also confirmed in other organs or tissues of the Cdh5-Nigri;R26-NLR mouse (Fig. S1B,C). To further verify if there was no crosstalk between Nigri-nox and Cre-loxP systems, we performed the following experiments. We isolated endothelial cells of E15.5 Cdh5-Nigri;R26-NLR tissues and then performed FACS analysis. We detected that 95.53%±1.22% CD31⁺ endothelial cells were ZsGreen⁺, but did not detect any tdTomato⁺ cell by flow cytometry (Fig. S3A). We also performed flow cytometric analysis of E15.5 Tie2-Cre;R26-GFP tissues and found that the recombination efficiency of Cre-loxP was 96.40%±2.01% (Fig. S3B). There was no significant difference between these two systems (Fig. S3C). Moreover, western blot analysis detected protein expression of ZsGreen but not tdTomato (Fig. S3D). We also crossed Cdh5-Nigri with Rosa26-loxP-Stop-loxP-tdTomato (R26-tdTomato) line) (Madisen et al., 2010) to generate the Cdh5-Nigri;R26-tdTomato mouse. We performed immunostaining for tdTomato on heart sections of the Cdh5-Nigri;R26-tdTomato mouse but did not detect any tdTomato⁺ cells (Fig. S1A). Co-transfection assays were also performed to examine the crosstalk in vitro. We co-transfected 293T cells with recombinase expression plasmids and recombination reporter plasmids. The pcDNA3.1(−) plasmids were used as controls. We co-transfected pcDNA3.1(−) plasmids with CAG-loxP-STOP-loxP-ZsGreen reporter plasmids or CAG-nox-STOP-nox-ZsGreen reporter plasmids. We then detected very rare leakiness in 4× fields using an Olympus microscope. To test whether crosstalk between Nigri-nox and Cre-loxP systems existed, we performed co-transfections of pcDNA3.1-Nigri with CAG-loxP-STOP-loxP-ZsGreen reporter plasmids, as well as co-transfections of pcDNA3.1-Cre with CAG-nox-STOP-nox-ZsGreen reporter plasmids. In these experiments, we detected very few ZsGreen⁺ cell 36 h post transfection, indicating that the crosstalk between Nigri-nox and Cre-loxP systems was negligible in vitro. (Fig. S3F,G). Taken together, these genetic data demonstrated that Nigri specifically and efficiently recombined nox but not loxP sites in vivo.

To further characterize the newly generated Cdh5-Nigri mouse line, we collected P7 Cdh5-Nigri;R26-NLR mouse heart and examined the cell fate of Cdh5-derived descendants. We stained heart sections with ZsGreen and other cell lineage markers, including the fibroblast marker PDGFRa, the cardiomyocyte marker troponin I (TNNI3), the smooth muscle cell marker aSMA and the pericyte marker PDGFRb. Consistent with previous reports (de Lange et al., 2004; Lincoln et al., 2004), endothelial cells labeled by Cdh5-Nigri contributed to the majority of PDGFRa⁺ mesenchymal cells/fibroblasts in the cardiac valve (Fig. 2A). We did not observe any contribution of endothelial cells to cardiomyocytes (Fig. 2B). Interestingly, a subset of Cdh5-derived cells also expressed aSMA or PDGFRb, indicating that endothelial cells might differentiate into smooth muscle cells or pericytes in the myocardium of developing heart (Fig. 2C,D). This finding is consistent with previous report demonstrating that endothelial cells are progenitors of coronary smooth muscle cells and pericytes (Chen et al., 2016).

We next tested whether Cre recombinase recombines nox sites. We used constitutive Cre driven by an ACTB promoter that is broadly active in tissues (ACTB-Cre) and crossed it with R26-NLR mouse line to generate ACTB-Cre;R26-NLR (Fig. 3A). Whole-mount epifluorescence of E14.5 ACTB-Cre;R26-NLR mouse showed tdTomato⁺ but not ZsGreen⁺ signal, whereas no tdTomato⁺ cells were detected in the absence of ACTB-Cre (Fig. 3B). Immunostaining for tdTomato and ZsGreen on the entire embryonic section and heart section showed tdTomato⁺ZsGreen^– cells (Fig. 3C,D). The same results were obtained in E12.5 and P7 ACTB-Cre;R26-NLR tissues (Fig. S2A-C). Western blot analysis of ACTB-Cre;R26-NLR tissue also detected protein expression of tdTomato but not ZsGreen (Fig. S3D). To further characterize Cre-loxP recombination in R26-NLR, we crossed it with an inducible CreER line that was driven by the ubiquitous gene promoter UBC-CreER (Fig. 3E). After tamoxifen induction, both whole-mount epifluorescence and sectional immunostaining showed that most cells were tdTomato⁺ZsGreen^– throughout embryos or heart tissues of the UBC-CreER;R26-NLR embryos (Fig. 3F-H, Fig. S2D,F). We did not detect tdTomato⁺ cells that had not received tamoxifen (Fig. 3G,H). These data demonstrated that both constitutive and inducible Cre recombinases efficiently mediated recombination of loxP site but not nox site. Furthermore, to verify if nox sites could been recognized by other available recombinases such as Dre, we crossed R26-NLR with Tie2-Dre to generate the Tie2-Dre;R26-NLR line. Whole-mount epifluorescence views and sectional staining of E14.5 Tie2-Dre;R26-NLR embryos showed no ZsGreen or tdTomato expression, demonstrating that Dre cannot recombine nox sites (Fig. S4). By Nigri-nox and Cre-loxP recombination systems, the new dual reporter R26-NLR line could be used to genetically trace the cell fates of two distinct cell populations simultaneously in vivo.

As our new dual reporter system combines two pairs of recombinase-recognition sites for simultaneous tracking of two distinct progenitor cells, we can study the dynamic contribution of different cell lineages during organ development (e.g. their component ratio and relative location from early embryonic to postnatal stages) within the same mouse. We then reassessed the dynamic cellular contribution during cardiac valve development. The mammalian heart has four sets of cardiac valves, including the atrioventricular valves (inlet valves such as mitral and tricuspid valves) and the outflow tract valves (outlet valves such as pulmonic and aortic valves) (de Vlaming et al., 2012). During valve formation and morphogenesis, the mesenchymal cells of valves have multiple developmental origins (Snarr et al., 2008; de Vlaming et al., 2012; von Gise and Pu, 2012; MacGrogan et al., 2014). Previous studies demonstrated that the mesenchymal cells of inlet valves are primarily derived from endocardium (de Lange et al., 2004; Timmerman et al., 2004; Lincoln et al., 2004; Rivera-Feliciano et al., 2006) and epicardium (Gittenberger-de Groot et al., 1998; Zhou et al., 2010; Wessels et al., 2012), whereas neural crest and endocardium mainly give rise to outlet valves (Jiang et al., 2000; de Lange et al., 2004; Nakamura et al., 2006). The relative localization and dynamic change of mesenchymal components and ratios of different cell origins remain incompletely understood. We first examined the inlet valve morphogenesis using Nigri-nox and Cre-loxP systems. To trace the fate of epicardial-derived cells and endothelial-derived cells simultaneously, we crossed the R26-NLR mouse line with Tbx18-Cre (Cai et al., 2008) and Cdh5-Nigri mouse line to generate the Tbx18-Cre;Cdh5-Nigri;R26-NLR triple-positive mouse (Fig. 4A). Tbx18-Cre marks the epicardium and is also expressed in a few cardiomyocytes (Christoffels et al., 2009); this study focused on the contribution of epicardial cells to mesenchymal cells in the valve. Cdh5-Nigri labeled endocardial cells as ZsGreen and Tbx18-Cre labeled epicardial cells as tdTomato (Fig. 4A, Fig. S5A). In order to monitor the dynamic changes of descendants of these two progenitor populations in inlet valve development, we collected hearts for analysis from E12.5 to postnatal day (P) 7 (Fig. 4B). Whole-heart sectional fluorescence of E12.5 Tbx18-Cre;Cdh5-Nigri;R26-NLR embryos showed expression of both tdTomato and ZsGreen in the same heart (Fig. 4C). tdTomato⁺ cells were mainly detected in epicardial cells, whereas ZsGreen⁺ cells were restricted in endocardial cushion cells as well as endothelial cells at E12.5 (Fig. 4C). At E12.5 and afterwards, epicardial cells form epicardial-derived mesenchymal cells (EPDCs) and differentiate into mesenchymal cells via epithelial-to-mesenchymal transition, EpiMT (Wessels and Pérez-Pomares, 2004; Kolditz et al., 2008; Zhou et al., 2008; Combs and Yutzey, 2009). Tbx18-Cre labeled epicardial cells gradually appeared in the parietal, but not septal, leaflet of the tricuspid valve at E14.5 that constituted the majority of mesenchyme at E17.5 and replaced endocardium-derived mesenchyme at P7 (Fig. 4C,D), indicating that EPDCs migrated into the valve and contributed significantly to valve mesenchyme. We also found that endocardium-derived cells constituted almost all valve mesenchyme in the mural leaflet of mitral valve at E12.5, with less contribution from EDPCs (Fig. 4E). At E14.5, EDPCs started to migrate into the proximal mural leaflet and more EDPCs could be found at E17.5 that gradually constituted the majority of mesenchymal cells in the mural leaflet but not aortic leaflet of the mitral valves by P7 (Fig. 4E,F). The majority of mesenchymal cells in the septal leaflet of tricuspid valve and aortic leaflet of mitral valve were still derived from endocardium (Fig. 4C-F). Quantification of the percentage of ZsGreen⁺ or tdTomato⁺ mesenchymal cells in labeled valve mesenchyme confirmed the gradual replacement of endocardium-derived cells by EPDCs in the parietal leaflet of the tricuspid valve and the mural leaflet of the mitral valve (Fig. 4G).

We next used Nigri-nox and Cre-loxP systems to simultaneously monitor contribution of two distinct progenitors in outflow track cushion morphogenesis, remodeling and subsequent semilunar valve formation. Neural crest cells (NCCs) have been reported to contribute primarily to semilunar valves (Jiang et al., 2000; Lincoln et al., 2004; Nakamura et al., 2006). To trace the fate of NCCs and endocardium-derived cells simultaneously, we crossed the R26-NLR mouse line with Wnt1-Cre and Cdh5-Nigri to generate the Wnt1-Cre;Cdh5-Nigri;R26-NLR triple-positive mouse (Fig. 5A). Cdh5-Nigri labeled endocardial cells as ZsGreen and Wnt1-Cre labeled neural crest cells as tdTomato (Fig. 5A, Fig. S5B). We analyzed the origins of semilunar valves in embryonic, neonatal and adult hearts of Wnt1-Cre;Cdh5-Nigri;R26-NLR mice (Fig. 5B). At E13.5, NCCs labeled by Wnt1-Cre (tdTomato⁺) contributed to the majority of mesenchymal cells in the right and left posterior cusp of pulmonary valves (PV) adjacent to the aorticopulmonary septum, whereas endocardial cells labeled by Cdh5-Nigri (ZsGreen⁺) contributed to a substantial number of the distal anterior leaflet of PV (Fig. 5C). For the aortic valves (AoV), tdTomato⁺ NCCs contributed to the majority of mesenchymal cells whereas ZsGreen⁺ endocardial cells contributed to fewer (Fig. 5D). Unlike the atrioventricular valves, endocardial cells constituted the minority of mesenchymal cells in the outflow track valves but they increased gradually from 21.78%±4.40% in the embryonic stage to 35.87%±1.99% in the adult stage of aortic valves, and from 20.64%±1.85% in the embryonic stage to 27.05%±5.15% in the adult stage of pulmonary valves during maturation (Fig. 5E,F). There was a significant difference in the percentage of ZsGreen⁺ mesenchymal cells between E13.5 and 6-week-old adult valves (Fig. 5E). Taken together, we provide examples to illustrate how Nigri-nox and Cre-loxP systems could be used simultaneously to lineage trace distinct cell populations during development in vivo. The dual recombination systems could also be used for studying cell fates of two distinct cell populations during tissue homeostasis and diseases.

In this study, we developed a new recombination system (Nigri-nox) for in vivo genetic lineage tracing. Nigri-nox recombination is efficient and specific for genomic engineering and does not operate through Cre-loxP (Karimova et al., 2016). A combination of dual recombinases using Nigri and Cre with new dual reporter lines allows genetic lineage tracing of two distinct cell populations. Our work provides proof-of-principle experiments showing that the Nigri-nox system can be used to trace cell fate in vivo. Coupled with other recombinases, our results show genetic targeting of multiple cell lineages in the same mouse.

To prove that this new genetic tool enables genetic tracing of two distinct progenitor populations simultaneously in vivo, we used both Nigri-nox and Cre-loxP to reassess the cellular contribution to mesenchyme of cardiac valve during heart development. The inlet atrioventricular valves are primarily derived from endothelium (de Lange et al., 2004; Lincoln et al., 2004) and epicardium (Gittenberger-de Groot et al., 1998; Wessels et al., 2012), whereas the outlet semilunar valves are mainly derived from neural crest and endothelium (Jiang et al., 2000; Nakamura et al., 2006). We generated the triple-positive Tbx18-Cre;Cdh5-Nigri;R26-NLR mouse line to trace epicardial and endothelial cells simultaneously. We found that these two populations contributed to almost all mesenchymal cells in the inlet valves during valve remodeling. In the early embryonic stage, endocardial cells contributed to the majority of mesenchymal cells in both atrioventricular cushions, whereas from mid-embryonic stage to neonatal stage, the resident endocardium-derived cells of lateral AV cushions were gradually replaced by infiltrating Tbx18-Cre-labeled EPDCs from proximal to distal. However, endocardium-derived cells still constituted the majority of mesenchyme in the septal leaflet of the tricuspid valves and in the mural leaflet of the mitral valves. We reasoned that this differential component of two leaflets in tricuspid and mitral valves could be due to their distinct anatomic locations in the developing heart. The parietal leaflet of the tricuspid valve is attached to the right ventricular wall and the mural leaflet of the mitral valve is attached to the left ventricular wall, where EPDCs were mostly distributed. Moreover, for analyzing outlet semilunar valves, we also generated a triple-positive Wnt1-Cre;Cdh5-Nigri;R26-NLR line to trace neural crest cells and endothelial cells simultaneously. During valve remodeling from the embryonic to adult stage, the contribution of endothelial cells increased gradually in the aortic and pulmonary valves (Fig. 5F). This result might indicate that endocardium-derived mesenchymal cells probably expanded more than neural crest cells during development. Alternatively, endocardium-derived mesenchymal cells could be less sensitive to signals driving apoptosis or valve remodeling.

The dual reporter system R26-NLR has advantages in revealing two distinct cell populations. However, this strategy cannot clearly distinguish two cell populations simultaneously if their two promoters are commonly active in an intersection part. If a lineage-traced cell exhibits activity of both driving promoters at the same time in development, we could label two cell populations by other previously reported intersectional reporters such as R26::FLAP (Awatramani et al., 2003), RC::Fela (Jensen et al., 2008), R26NZG (Yamamoto et al., 2009), RC::FrePe (Engleka et al., 2012) and RC::RLTG (Plummer et al., 2015). In these alleles, expression of the first reporter located near the promoter is controlled by one type of recombinase-mediated recombination, whereas expression of the second reporter located far from the promoter requires both types of recombination. For two different progenitor populations that are driven by two distinct gene promoters, reporter of the first recombination cannot be maintained but is replaced by the second reporter after the second round of recombination. Although these intersectional dual-recombinase-based reporters allow us to label one group of progenitor cells and its subpopulations, they could not label two separate cell populations defined by two exclusively distinct genes. Here, the property of our new interleaved dual-recombinase-based reporter bypassed this pitfall, allowing us to target two distinct progenitor populations simultaneously during development and to trace for their cell fates with distinct lineage-tracing reporters (Fig. 6). Together, our novel system provides additional resolutions in cellular dynamics during tissue development.

In summary, our work showed that Nigri-nox was efficient in mammalian genomic engineering and could be used with other recombinase systems for more precise genetic targeting in vivo. Specifically, Nigri-nox and Cre-loxP systems could target two distinct progenitor cell populations for delineating the contributions of endocardium, neural crest and epicardium cells during cardiac valve development and remodeling. This dual recombination system could also be broadly used in studying various cell lineages in development, diseases and tissue regeneration.

All mouse experiments were conducted in accordance with the guidelines from the Institutional Animal Care and Use Committee (IACUC) of the Institute of Biochemistry and Cell Biology, and Institute for Nutritional Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences. ACTB-Cre, UBC-CreER. (Liu et al., 2016; Ruzankina et al., 2007) and Tie2-Cre mice have been described previously (Kisanuki et al., 2001). Tie2-Dre mouse was reported previously by our lab (Pu et al., 2018). The R26-NLR mouse line was generated by knocking CAG-nox-loxP-Stop-nox-ZsGreen-polyA-loxP-wpre-Frt-Neo-Frt into intron 1 of Rose26 locus through homologous recombination. The Cdh5-Nigri knock-in mouse line was generated by homologous recombination using CRISPR/Cas9 methods. For the Cdh5-Nigri line, a cDNA encoding the Nigri cassette was inserted into the second exon of Cdh5. The Cdh5-Nigri line results in the heterozygosity of Cdh5. Cdh5-Nigri and R26-NLR knock-in lines were generated by Shanghai Biomodel Organism Science and Technology Development, Shanghai, China. All mice used in this study were maintained on a C57BL6/ICR background. Tamoxifen (Sigma, T5648-5G) was dissolved in corn oil and administrated to mice by oral gavage at the indicated time (0.2 mg/g).

Genomic DNA was extracted from mouse tail. Tissues were lysed in lysis buffer (100 μg/ml proteinase K) overnight at 55°C and the mixture was centrifuged at the maximum speed of 20,000 g for 8 min to obtain supernatant with genomic DNA the next day. DNA was precipitated using isopropanol, and then washed in 70% ethanol by centrifugation at 20,000 g for 3 min. Finally, DNA was dissolved in deionized water. All the mice or embryos were genotyped using genomic PCR to distinguish knock-in allele from wild-type allele. For the R26-NLR line, primers 5′-TCCCGACAAAACCGAAAATCTGTGG-3′ and 5′-TGCATCGCATTGTCTGAGTAGG-3′ were used to detect the R26-NLR positive allele, and 5′-TCCCGACAAAACCGAAAATCTGTGG-3′ and 5′-GGGGCGTGCTGAGCCAGACCTCCAT-3′ were used to detect the wild-type allele. For the Cdh5-Nigri line, primers 5′-GGAGCCAGGCGCATAGTGAGCA-3′ and 5′-GGTAGGGGCGGTGGTTCTGGTG-3′ were used to detect the Nigri positive allele, and 5′-TTTCCCCAGATCAGCTCCTCCAC-3′ and 5′-CCCCCGCACCACCAATGATA-3′ were used to detect the wild-type allele.

Immunostaining in this study was performed as previously described (Zhang et al., 2016). Briefly, mouse embryos or hearts were collected in PBS on ice and then washed gently to remove blood. They were then fixed in 4% PFA for 15-60 min at 4°C depending on tissue size. After washing several times in PBS, we used fluorescence microscopy (Zeiss AXIO Zoom V16) to take photographs of embryos or hearts with indicated fluorescence reporters. After that, the embryos or hearts were dehydrated in 30% sucrose/PBS for several hours or overnight at 4°C, and then embedded in OCT (Sakura) the next day. Frozen sections of 7-10 µm were collected on slides. For immunostaining, tissue sections were first washed in PBS to remove OCT, then blocked in 5% PBSST (5% donkey serum/0.1% Triton X-100/PBS) for 30 min at room temperature and then incubated with primary antibodies overnight at 4°C. Primary antibodies against the following proteins and dilution were used: tdTomato (Rockland, 600-401-379; 1:1000 dilution), ZsGreen (Clontech, 632474; 1:1000), VE-cad (R&D, AF1002; 1:100 dilution), PECAM (BD, 553370; 1:500), PDGFRa (eBioscience, 14-1401-81; 1:500), PDGFRa (R&D, AF1062; 1:500), PDGFRb (eBioscience, 14-1402; 1:500), aSMA (Sigma, F3777; 1:500) and TNNI3 (Abcam, ab56357; 1:200). The next day, sections were washed with PBS several times and then incubated with Alexa fluorescence secondary antibodies and counterstained with DAPI (Vector Laboratories) at room temperature for 40 min in dark. After washing several times in PBS, all slides were mounted with mounting medium. For weak signals, HRP-conjugated secondary antibodies were used to amplify the signals. Images were acquired using an Olympus confocal microscopy system (FV1200) and Leica confocal (TCS SP5). To provide z-stack images, we scanned 5-10 consecutive xy images on the z-axis using a confocal microscope. The images were analyzed using ImageJ (NIH) software.

Tissues of R26-NLR mice, ACTB-Cre;R26-NLR mice and Cdh5-Nigri;R26-NLR mice were collected and dissected. Protein supernatant was obtained after homogenized in Tris-SDS lysis buffer [1% SDS, 50 mM Tris-HCl (pH 8.0)], incubated at 4°C for 30 min and centrifuged at maximum speed for 5 min. After mixing 4:1 with 5× loading buffer [10% SDS, 0.3 M Tris-HCl, 1.5 M dithiothreitol (pH 6.8)], protein samples were boiled for 10 min, separated by 10% SDS-PAGE and then transferred by a Mini Trans Blot system (Bio-Rad) onto a polyvinylidene fluoride (PVDF) membrane (Immobilon, Millipore). The membrane was blocked with 5% BSA dissolved in TBS-T (150 mM NaCl, 0.5% vol/vol Tween-20, 10 mM Tris-HCl) at room temperature for 1 h and incubated with primary antibody at 4°C overnight. After washing three times with TBS-T, the membrane was incubated with HRP-conjugated secondary antibody at room temperature for 1 h. The membrane was then washed with TBS-T three times for 30 min. Proteins bands were observed by enhanced chemiluminescence (Pierce). Antibodies used in this study were as follows: ZsGreen (rabbit, Clontech, 632474; 1:1000), tdTomato (rabbit, Rockland, 600-401-379; 1:1000 dilution), anti-β-actin (mouse, Cell Signaling, 3700; 1:1000), HRP-anti-rabbit IgG (Jackson ImmunoResearch, 711-035-152; 1:4000) and HRP-anti-mouse IgG (Jackson ImmunoResearch, 115-035-174; 1:4000).

Endothelial cells were isolated from Cdh5-Nigri;R26-NLR and Tie2-Cre;R26-GFP mice. Briefly, tissues from Cdh5-Nigri;R26-NLR and Tie2-Cre;R26-GFP mice were dissected and cut into small pieces. They were then digested by lysis buffer that contained DNase I and collagenase type II at 37°C for 40 min. After digestion, this buffer was passed through 70 mm cell strainers. The cells were centrifuged at 1000 g for 5 min at 4°C. Blood cells were lysed by Red Blood Cell lysis buffer (eBioscience, 00-4333-57) and the cells were re-suspended and washed in PBS. After that, 500 μl isolation buffer (0.5% BSA and 2 mM EDTA in PBS) was used to wash the cells, then centrifuged for 2 min at 4600 g. The cells were stained with antibody containing CD45 PE-Cy7 (eBioscience, 25-0451, 1:400), Ter-119 PE-Cy7 (eBioscience, 25-5921, 1:400) or CD31 APC (eBioscience, 17-0311-82, 1:40) for 30 min at 4°C. Finally, the cells were washed in isolation buffer and re-suspended in PBS. FACS Aria II Flow Cytometer (BD Bioscience) was used to analyze the stained cells. Flowjo (TreeStar) was used to process the raw data with quantification analysis.

We co-transfected 293T cells with recombinase expression plasmids and recombination reporter plasmids to test whether crosstalk exists between Nigri-nox and Cre-loxP systems. Two recombinase expression plasmids (pcDNA3.1-Cre, pcDNA3.1-Nigri) and two recombination reporter plasmids (CAG-loxP-STOP-loxP-ZsGreen, CAG-nox-STOP-nox-ZsGreen) were used in this assay. The pcDNA3.1(-) plasmids were used as controls. The assay was performed using Lipofectamine 3000 Transfection Reagent (Invitrogen). Briefly, 293T cells were plated at the density of 5×10⁵ cells per well in 12-well dishes and the cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) containing 10% fetal bovine serum (Invitrogen) and 1% penicillin-streptomycin-glutamine (PSG, Invitrogen). Then we co-transfected 293T cells with plasmids. We designed six groups of experiments and three repeats in each group. Group one and group two were negative control groups and the cells were co-transfected with pcDNA3.1(−) plasmids containing CAG-loxP-STOP-loxP-ZsGreen reporter plasmids or CAG-nox-STOP-nox-ZsGreen reporter plasmids, respectively. Group three and group four were positive control groups. One group was co-transfected with pcDNA3.1-Nigri and CAG-nox-STOP-nox-ZsGreen reporter plasmids and another group was co-transfected with pcDNA3.1-Cre and CAG-loxP-STOP-loxP-ZsGreen reporter plasmids. Group five and group six were used to test whether crosstalk exists between Nigri-nox and Cre-loxP systems. One group was co-transfected with pcDNA3.1-Nigri and CAG-loxP-STOP-loxP-ZsGreen reporter plasmids and cells from another group were co-transfected with pcDNA3.1-Cre and CAG-nox-STOP-nox-ZsGreen reporter plasmids. The cells were then culture in 5% CO₂ at 37°C. The expression of ZsGreen was analyzed 36 h after transfection using an Olmypus BX53 microscope.

All data were obtained from four to six independent experiments, as indicated in each figure legend. Immunostaining of ZsGreen and tdTomato was used to indicate dual lineage tracing of progenitor-derived cells. ZsGreen indicates endocardial-derived cells (Cdh5-Nigri). tdTomato indicates epicardial-derived cells (Tbx18-Cre) or neural crest cell-derived cells (Wnt1-Cre). The proportion of fluorescent area between ZsGreen and tdTomato in heart valves indicates the ratio of two groups of labeled cells. The signals of ZsGreen removed PECAM was used for statistics because PECAM⁺ valve endothelial cells were also labeled by ZsGreen. Fluorescence signals were analyzed by ImageJ software. All data were presented as mean values±s.e.m. Two-tailed unpaired Student's t-test was used to compare the difference between two groups; ANOVA test was used to compare differences among three or more groups. P<0.05 was accepted as statistically significant.

We thank Dr Qiang She for providing the Tbx18-Cre mouse line and Shanghai Biomodel Organism for mouse generation. We also thank Baojin Wu, Guoyuan Chen, Zhonghui Weng and Aimin Huang for the animal husbandry; Wei Bian for technical help; and the National Center for Protein Science Shanghai for helping in confocal imaging.