Abstract
Cardiovascular lineages develop together with kidney, smooth muscle, and limb connective tissue progenitors from the lateral plate mesoderm (LPM). How the LPM initially emerges and how its downstream fates are molecularly interconnected remain unknown. Here, we isolated a pan-LPM enhancer in the zebrafish draculin (drl) gene that provides specific LPM reporter activity from early gastrulation. In toto live imaging and lineage tracing of drl-based reporters captured the dynamic LPM emergence as lineage-restricted mesendoderm field. The drl pan-LPM enhancer responds to the transcription factors EomesoderminA, FoxH1, and MixL1 that combined with Smad activity drive LPM emergence. We uncovered specific drl reporter activity in LPM-corresponding territories of several chordates including chicken, axolotl, lamprey, Ciona, and amphioxus, revealing a universal upstream LPM program. Altogether, our work provides a mechanistic framework for LPM emergence as defined progenitor field, possibly representing an ancient mesodermal cell state that predates the primordial vertebrate embryo.
Introduction
Key cell fates and organ systems in vertebrates emerge from multipotent progenitors within the embryonic mesoderm. Following gastrulation, the vertebrate mesoderm has been classically described to partition into axial, paraxial, and ventro-lateral domains (Gilbert, 2000; Gurdon, 1995). The latter, referred to as lateral plate mesoderm (LPM), is composed of highly motile cells and is mainly defined by its position adjacent to the somite-forming paraxial mesoderm (Hatada and Stern, 1994). Transplantation and lineage tracing experiments in several species have established that the LPM contains progenitor cells of the circulatory system, smooth muscle lineages, the kidneys (in amniotes often demarcated as intermediate mesoderm), and the limb connective tissue anlagen (Chal and Pourquié, 2017; Takasato and Little, 2015). Several transcription factors including Hand1/2, Tbx5, Osr1, FoxF1, Prrx1, Mesp1, and Etv2 are expressed in LPM territories and play overlapping roles in cell fate determination (Chal and Pourquié, 2017; Davidson and Zon, 2004; Takasato and Little, 2015), albeit not always with an evolutionarily conserved function (Yabe et al., 2016). During segmentation, the LPM principally segregates into the anterior LPM (ALPM) and posterior LPM (PLPM), which further divides into dorsal and ventral domains (somatopleure and splanchnopleure, respectively) (Gilbert, 2000). Several studies have uncovered mechanisms that control the lateral-to-medial or anterior-posterior specification of LPM-descendant organ precursors (Chal and Pourquié, 2017; Helker et al., 2015; Tonegawa et al., 1997). In contrast, it remains incompletely understood how the LPM arises from an initial mesendodermal population that goes on to form distinct endodermal and mesodermal progenitors. This is in part due to the lack of tools and markers to track LPM emergence genetically during development. Further, whether the LPM initially emerges as morphogenetic field in a molecularly coherent unit or as a loosely connected assembly of progenitor cells remains unclear.
Assessing the evolutionary context by which the LPM emerged as a developmental entity also remains challenging, in particular in extant jawless vertebrates such as lamprey or chordate models that do not form the full spectrum of LPM derivatives. Ancestral gene-regulatory repertoires that control higher-order structures in vertebrates previously have been indicated for somatic muscle in lamprey (Kusakabe and Kuratani, 2007) or for the putative equivalents of cardiac and hematopoietic progenitors in amphioxus (Pascual-Anaya et al., 2013). Anterior-to-posterior expression domains of key LPM transcription factors including Tbx1/10 and Hand are conserved in lampreys and amphioxus (Onimaru et al., 2011). Furthermore, the tunicate Ciona forms cardiac lineages that display genetic regulatory circuits homologous to the cardiac LPM progenitors found in vertebrates (Kaplan et al., 2015; Racioppi et al., 2019). These observations suggest the existence of an ancient, defined regulatory program that delineates prospective LPM progenitors in a common chordate ancestor, dating back to the Cambrian explosion 520-540 million years ago.
Several mammalian cis-regulatory elements with broad LPM activity have been reported; these include a small upstream enhancer of mouse and human HoxB6 (Becker et al., 1996), an upstream enhancer of mouse Gata4 (Rojas et al., 2005), and a downstream enhancer of mouse Bmp4 (Chandler et al., 2009). In line with a ventral LPM origin, the Gata4 LPM enhancer responds to Smads downstream of BMP signaling (Rojas et al., 2005). Nonetheless, the activities driven by these enhancer elements in mice confine to the PLPM and not pan-LPM readouts. In zebrafish, the ventrally emerging LPM forms during somitogenesis into a patchwork of bilateral gene expression domains, including the conserved LPM genes hand2, pax2.1, scl, lmo2, etv2, and tbx5 (Davidson and Zon, 2004). In contrast, transgenic reporters based on the 6.35 kb cis-regulatory region of the zebrafish-specific gene draculin (drl) selectively label the entire LPM from its emergence during gastrulation through initial differentiation (Mosimann et al., 2015). Cre/lox-mediated genetic lineage analysis has established that early drl reporter expression labels the LPM progenitors forming cardiovascular, blood, kidney, intestinal smooth muscles (iSMCs), and pectoral fin mesenchyme fates (Felker et al., 2016; Felker et al., 2018; Gays et al., 2017; Henninger et al., 2017; Mosimann et al., 2015). These observations suggest that the 6.35 kb drl region harbors cis-regulatory elements active throughout the prospective LPM starting from gastrulation, raising the possibility that these regulatory elements read out a hypothetical pan-LPM program.
Here, we dissected the 6.35 kb drl cis-regulatory elements and uncovered an intronic enhancer, +2.0drl, that is necessary and sufficient in zebrafish for driving LPM-specific expression in all presumptive LPM progenitors from early gastrulation to early somitogenesis. Panoramic SPIM and Cre/lox-mediated genetic lineage tracing of drl reporters demonstrated that the zebrafish LPM forms from a restricted mesendoderm territory during gastrulation. To uncover the upstream regulatory program read out by the +2.0drl pan-LPM enhancer, we combined ChIP data, Cas9-based crispant analysis, and reporter assays. This enabled us to identify the combination of mesendoderm transcription factors EomesA, FoxH1, and MixL1 as sufficient to drive pan-LPM activity. Our data suggest that the combined activity of these transcription factors demarcates a dedicated mesendoderm progenitor field during zebrafish gastrulation, restricting the emergence of LPM progenitors to their future lateral position. In cross-species assays, we observed specific activity of the zebrafish +2.0drl pan-LPM enhancer in LPM-corresponding territories in chicken, axolotl, lamprey, Ciona, and amphioxus embryos. These results demonstrate that the +2.0drl enhancer reads out a universal LPM progenitor program that is conserved across chordates, defining a core transcription factor code for LPM formation. Our data provide a developmental framework for charting the earliest emergence of LPM progenitors across chordates.
Results
The LPM emerges as a dedicated mesendoderm population
To resolve the dynamics of LPM emergence in toto, we performed time course experiments using single-plane illumination microscopy (SPIM) (Fig 1A-D) and panoramic projections (Schmid et al., 2013) (Fig 1E-H) of reporter-transgenic zebrafish embryos based on the full-length drl cis-regulatory region. drl:EGFP-expressing LPM precursors became detectable by early gastrula stages (50% epiboly) and continuously condensed along the embryo margin through the end of gastrulation (tailbud stage) (Fig 1A,B,F; Movie S1, S2). From tailbud stage onward, drl:EGFP-marked LPM formed a continuous band of cells with condensing anterior and posterior segments (Fig 1C,D,G,H). We confirmed that this EGFP-positive cell band encompasses the bilateral stripes of several established LPM sub-domain markers by comparing a series of overlapping expression domains from distinct reporter lines (Fig 1I-L). First, lmo2:dsRED2 labels embryonic hematopoietic and vascular tissues (Zhu et al., 2005), and its expression overlaps with medial drl:EGFP-expressing cells in the ALPM and PLPM (Fig 1I). scl:EGFP (Zhang and Rodaway, 2007) also co-expressed with drl:mCherry in the most medial PLPM domain and in a small ALPM population (Fig 1J). We find that while pax2.1:EGFP in early somitogenesis marks midbrain-hindbrain precursor cells of ectodermal origin, the likewise pax2.1:EGFP-expressing PLPM-derived pronephric epithelial precursors (Picker et al., 2002) also express drl:mCherry (Fig 1K). Moreover, hand2:EGFP expression, which demarcates the lateral-most PLPM domain plus parts of the ALPM-derived heart field and pectoral fin precursors (Perens et al., 2016; Yelon et al., 2000), was also fully situated within the pan-LPM expression domain of drl:mCherry (Fig 1L). Taken together, these data provide a continuous view of the emerging LPM stripes from gastrulation in zebrafish and document that the LPM emerges around the entire circumference of the zebrafish embryo (Fig 1M).
We next sought to capture how the drl-expressing LPM emerges relative to the endoderm. Panoramic SPIM of the sox17:EGFP-positive endoderm reporter together with drl:mCherry revealed a population of double-positive cells from the onset of reporter detection through late gastrulation (Fig 2A-D; Fig S1A-D). After gastrulation, we detected a continuous band of drl reporter-positive cells around the developing embryo that was separated from the more medial endodermal sox17 expression domain (Fig 2D; Movie S3, S4). To confirm whether endoderm progenitors are also marked by the drl reporter during gastrulation, we performed drl:creERT2-based genetic lineage tracing with the ubiquitous hsp70l:loxP-STOP-loxP-EGFP (hsp70l:Switch) and 4-OHT-based CreERT2 induction at discrete time points ranging from shield to 5-6 somite stages (ss) followed by analysis of labeling patterns at 72 hpf (Fig 2E-J). 4-OHT induction at shield stage marked LPM lineages including blood, endothelium, kidney, and iSMCs as the only mesodermal fates (Fig 2F-H) (Gays et al., 2017; Mosimann et al., 2015), while lineage labeling also marked broad territories within endodermal organs, including pancreas, liver, and pharynx/gut epithelium (Fig 2H; Fig S1A). 4-OHT induction at increasingly later time points gradually decreased endoderm labeling, with minimal to no endodermal lineage signals following 4-OHT induction at 5-6 ss (Fig 2I,J, Fig S1B, Fig S2B,C). In contrast, LPM structures remained robustly labeled as the exclusive mesoderm fate, consistent with previous work (Felker et al., 2018; Gays et al., 2017; Henninger et al., 2017; Mosimann et al., 2015). Additionally, we observed that the spatio-temporal contribution of drl reporter-expressing progenitors to endoderm differs along the anterior-posterior axis. We divided the embryo into four non-overlapping regions along the anterior-to-posterior axis (region I – IV) (Fig. S2A) and quantified the switching efficiency. The amount of lineage-labeled gut endothelium increased within individual embryos from the anteriorly located pharynx (region I) towards the caudal gut (region IV), independent of the stage of 4-OHT administration (Fig S2B,C). These results indicate that progenitors expressing the drl reporter become progressively restricted to an LPM fate from anterior to posterior with ongoing development, until by early somitogenesis drl reporter expression labels only early LPM.
In contrast, sox17:creERT2 exclusively marked endoderm lineages (n = 6; Fig S1C), supporting data that sox17 expression demarcates zebrafish endoderm progenitors downstream of the key endoderm regulator sox32 (Alexander and Stainier, 1999). Remarkably, embryos that fail to form endoderm upon sox32 perturbation still generated a type of drl-traced LPM that partitioned into its major recognizable descendants including heart, blood, endothelium, and pronephros (Fig S3). These results are not skewed by lineage-bias of hsp70l:Switch reporter sensitivity, as we observed well-distributed lineage labeling using the ubiquitous ubi:creERT2 (Fig S1D).
Altogether, these data establish that during gastrulation the drl-marked LPM gradually refines from a ventral mesendoderm territory to a bilateral LPM domain as the sole mesodermal fate along the entire anterior-posterior axis of the embryo. The rare lineage labeling of somitic muscle by drl:creERT2 (Fig 2F) (Gays et al., 2017; Mosimann et al., 2015) further underlines that, in zebrafish, the paraxial mesoderm and the LPM develop as distinct mesoderm lineages with only minimal overlap (Warga and Nüsslein-Volhard, 1999).
A pan-LPM enhancer in the zebrafish drl locus
To identify cis-regulatory element(s) in the drl locus responsible for pan-LPM progenitor expression, we divided the 6.35 kb drl regulatory region into smaller fragments and assayed their activity using Tol2-based EGFP reporters in F0 and stable transgenics (Fig 3A). We found that the promoter-proximal region surrounding exon 1 recapitulated drl reporter expression in ALPM and PLPM from 5-7 ss onwards (proximal drl), while the upstream promoter region alone remained active in the posterior endothelial and blood precursors (−1.02drl) (Fig 3A-C). In addition, we identified a small (968 bp) region in the first intron (+2.0drl) that initially recapitulated early drl reporter expression in zebrafish embryos before fading between 5-10 ss (approx. 13 hpf) (Fig 3A,D). Genetic lineage tracing with +2.0drl:creERT2 and hsp70l:Switch (Fig 3E) specifically labeled LPM-derived organs including heart, blood, endothelium, kidney, pectoral fin mesenchyme, and iSMCs, and additionally marked endoderm lineages when induced with 4-OHT at shield stage (Fig 3F-I). These results correlate well with our lineage tracing using full-length drl:creERT2 initiated at the onset of gastrulation (Fig 2E-J) (Felker et al., 2018; Gays et al., 2017; Mosimann et al., 2015). Further, deletion of elements within the +2.0drl enhancer defined a minimal enhancer region of 432 bp (+2.4drl) that functioned as a pan-LPM enhancer, albeit with higher variability in stable transgenics (Fig S4). These regulatory analyses indicate that the entire drl expression pattern derives from distinct cis-regulatory elements that control drl expression in separable early mesendoderm/pan-LPM and later ALPM versus PLPM domains. The latter pattern is analogous to the hematopoietic lineages that arise during somitogenesis and are commonly marked with drl mRNA ISH (Davidson and Zon, 2004; Herbomel et al., 1999). This data implies that the +2.0drl enhancer contains the key regulatory modules that respond and integrate to early LPM-defining inputs.
Combined EomesA, FoxH1, and MixL1 activity integrates ventral BMP activity into LPM formation
We next sought to investigate the upstream inputs that control the +2.0drl pan-LPM enhancer. BMP and Activin/Nodal ligands of the TGF-β superfamily trigger key pathways in early vertebrate axis determination and mesendoderm induction (Hill, 2018; Langdon and Mullins, 2011; Whitman, 2001). Both pathways interpret ligand gradients and signal through type I and type II receptors that result in cytosolic phosphorylation of Smad transcription factors (Fig 4A) (Massagué, 2012; Rogers and Müller, 2019). In zebrafish embryos during early gastrulation, BMP ligands are secreted from the ventral side, while Nodal ligands are expressed along the margin and the dorsal side (Hill, 2018; Langdon and Mullins, 2011). In line with the emergence of ventral LPM, endogenous drl expression was virtually absent in embryos maternally mutant for dominant-negative Smad5 (MZsbn), which lack BMP activity (Fig 4B,C) (Hild et al., 1999). Similarly, treatment with the BMP inhibitor Dorsomorphin resulted in pronounced decrease of endogenous drl expression (Fig 4E,F). We also found decreased drl expression in embryos with perturbed Nodal signaling: i) maternal-zygotic mutant embryos lacking the key Nodal co-receptor Crypto/Oep (MZoep) that cannot transmit Nodal signaling around the embryo margin (Gritsman et al., 1999; Schier et al., 1997), and ii) embryos treated with the Nodal signaling inhibitor SB-505124 (Hagos and Dougan, 2007) (Fig 4D,G,H). These results indicate that the ventral drl domain is sensitive to both BMP and Nodal input. Consistent with our lineage tracing results (Fig S3), embryos devoid of endoderm upon sox32 knockdown still expressed endogenous drl, albeit with overall thinned-out expression and a marked decrease of dorsal drl activity (Fig 4I).
We mined published whole-embryo ChIP-seq data from zebrafish gastrulation stages (Dubrulle et al., 2015; Nelson et al., 2014; Nelson et al., 2017) and identified candidates for transcription factors binding to the +2.0drl enhancer. These include the T-box transcription factor EomesoderminA (EomesA), its interaction partner FoxH1, and BMP/Nodal-mediating Smads. Published evidence uncovered that they participate in control of mesendoderm genes (Charney et al., 2017; Chen et al., 1996; Germain et al., 2000; Slagle et al., 2011) and affect drl expression during early somitogenesis (Slagle et al., 2011) (Fig 5A). We found that mRNA injection- or ubi promoter-driven expression of constitutive-active forms of EomesA or FoxH1 strongly augmented and prolonged +2.0drl reporter and endogenous drl expression in their native LPM domain compared to controls (Fig 5B-E, Fig S5A-D). Ubiquitous expression of wildtype eomesA or foxh1 mRNA was sufficient to increase endogenous drl expression (Fig 5F-H,). Addition of a constitutively-active Smad2 (Dick et al., 2000) to EomesA and FoxH1 resulted in dorsal widening of +2.0drl reporter expression pattern (Fig S5E,F).
EomesA and FoxH1 are maternally contributed and are thus ubiquitously distributed during gastrulation (Bruce et al., 2003; Slagle et al., 2011), during which the LPM emerges in the ventral BMP and Smad activity domain (Fig 1,4). We therefore hypothesized that at least one additional, ventrally expressed transcription factor might be required for LPM formation. In published ChIP-seq data (Nelson et al., 2017), we identified the homeodomain protein MixL1 as a third possible transcription factor that acts together with EomesA and FoxH1 in controlling +2.0drl enhancer activity (Fig 5A). MixL1 is a downstream target of BMP and Nodal signaling implicated in controlling endoderm and mesoderm fates (Kunwar et al., 2003; Stainier et al., 1996) and it retains ventral expression in MZoep mutants (Slagle et al., 2011). Furthermore, MixL1 can form a complex with EomesA (Bjornson et al., 2005). Reminiscent of eomesA or foxh1 mRNA injections (Fig 5G,H), microinjected mixl1 mRNA also resulted in increased +2.0drl reporter expression within in the native LPM domain (Fig 5I).
Combining the triplet of wildtype mRNAs or Tol2-based DNA constructs encoding full-length EomesA, FoxH1, and MixL1 (shortened as e/f/m) led to ubiquitous +2.0drl reporter activation in embryos (Fig 5J). In MZsbn-mutant embryos without BMP signaling, e/f/m misexpression induced drl expression dorsally (Fig S5H). These observations suggest that in e/f/m overexpression conditions there is still a requirement for additional Smad activity, which in MZsbn embryos is only available in the dorsal Nodal-positive domain. Conversely, loss of Nodal signaling in MZoep mutants led to a ventral upregulation of drl expression upon mRNA-based e/f/m overexpression (Fig S5I). Combining native e/f/m in wildtype and MZsbn embryos devoid of endoderm following sox32 knockdown also resulted in ubiquitous +2.0drl reporter activation (Fig 5K, Fig S5J). This suggests that most of the +2.0drl reporter-positive cells have an LPM identity. Mutating mixl1 by CRISPR-Cas9 resulted in mosaic loss of +2.0drl reporter activity in F0 crispants (Fig 5L-N), while mutating the redundant mixl1 paralog mezzo (Poulain and Lepage, 2002) alone or together with mixl1 did not influence +2.0drl reporter activity (Fig 5L,O,P). This indicates that MixL1 is the predominant Mix paralog acting on the +2.0drl enhancer in zebrafish. Furthermore, CRISPR-Cas9-mediated mutagenesis of the +2.0drl enhancer in the region of predicted FoxH1 and MixL1 sites in the context of the full-length drl:EGFP transgene resulted in specific perturbation of early LPM reporter expression, without affecting the later ALPM and PLPM patterns (Fig S6).
EomesA, FoxH1, and MixL1 misexpression also induced expression of tmem88a, which is highly enriched in the early native LPM (Mosimann et al., 2015) (Fig S7A,B). In contrast, the expression domains of other early expressed mesodermal genes either showed a slight broadening of their native domains or appeared generally unaffected (Fig S7). This is further illustrated by the lack of changes in hand2 expression, which normally initiates in the lateral-most LPM after gastrulation (Fig S7K-M). Together, these data suggest a regulatory model in zebrafish whereby the combination of EomesA, FoxH1, and MixL1 potentiates ventral Smad-relayed BMP signals to demarcate a mesendoderm territory that becomes prospective LPM.
The drl pan-LPM enhancer reads out an LPM program across chordates
We next explored whether the +2.0drl enhancer could read out a putative pan-LPM program in diverse chordate species. First, we tested several previously characterized enhancers with activity in the posterior LPM of mice: Gata4 (Rojas et al., 2005), Bmp4 (Chandler et al., 2009), and HoxB6 (Becker et al., 1996). Reporter transgenes based on mouse Gata4 and Bmp4 showed restricted activity in the outward migrating endothelial/blood progenitors and in the PLPM when electroporated into the primitive streak of ex ovo-cultured chicken embryos after the onset of gastrulation (HH3+/4) (Fig S8B,C). The mouse HoxB6 LPM enhancer showed no regulatory activity in this assay (Fig S8D). In contrast, when microinjected into zebrafish embryos, reporters based on these three mouse enhancers all resulted in expression mainly in the notochord without specific LPM activity (Fig S8E-H). This indicates that while some of the previously isolated LPM enhancers of mouse expressed faithfully in the PLPM of chick embryos, their activity did not recapitulate an LPM pattern in zebrafish. These results suggest that these mammalian LPM enhancers may have specialized during amniote evolution.
Electroporation of the zebrafish +2.0drl reporter into the primitive streak of HH3+/4 chicken embryos resulted in reporter activity specifically in the forming LPM: depending on the exact stage and region of electroporation, we observed specific reporter activity in several LPM territories. Most frequently observed expression patterns included medial and posterior LPM domains (Fig 6A-C, Fig S9), and we also observed ALPM reaching the head fold in individual embryos (Fig S9). These observations suggest that a basic upstream program underlying LPM formation, as read out by the +2.0drl reporter, continues to function in birds as representative amniotes.
We then tested whether +2.0drl is responding to LPM-inducing cues in other tetrapods. Axolotl embryos microinjected with the +2.0drl EGFP reporter marked putative endodermal and LPM territories beginning from early somite-stages (stage 21), additionally marking the pharyngeal regions at tailbud stages (st 27 and 32 shown in Fig 6D-G). Transversal sections of stage 32 embryos confirmed the presence of EGFP-positive cells in the endoderm (asterisk) and lateral mesendoderm (Fig S9G-J). Interestingly, EGFP fluorescence was present throughout axolotl development and could be readily detected in the gut, as well as LPM-derived tissues including the limb bud (Fig 6H-I), blood vessels (Fig 6J-K), and gut lining (Fig S9K-N). These results support the notion that the +2.0drl enhancer also interprets an LPM program active in amphibians.
Next, we asked if the +2.0drl enhancer reads out a pan-LPM program in more distantly related vertebrates. Lampreys are jawless vertebrates (cyclostomata) that can provide unique insights into vertebrate evolution due to the early divergence of their lineage from jawed vertebrates (Shimeld and Donoghue, 2012). Microinjection of the +2.0drl EGFP-based reporter into sea lamprey embryos (Petromyzon marinus) consistently resulted in robust EGFP expression in the lateral mesendoderm starting during neurulation (st 18-21) (n = 145/231) (Fig 6L,M), as well as in the developing pharynx at st 22-24 (Fig 6N). Transverse embryo sections revealed that this early expression domain includes the anterior-most, LPM-linked expression of lamprey pmHandA (Onimaru et al., 2011), with the later pharyngeal expression of EGFP being restricted to the endoderm and mesoderm (Fig S10A-K). We conclude that the +2.0drl enhancer is capable of integrating regulatory outputs from an upstream LPM program that remains conserved across vertebrates.
Next, we asked if the +2.0drl enhancer also responds to upstream activity in LPM-linked cell fates dating back to the chordate radiation (Fig 7A). We first electroporated the +2.0drl reporter into embryos of the tunicate Ciona robusta, a chordate species belonging to a sister clade of vertebrates (Fig 7A). While missing the full complement of LPM-derived organ systems found in vertebrates, the LPM is echoed in the cardiopharyngeal progenitors forming in Ciona embryos (Diogo et al., 2015; Kaplan et al., 2015). We detected +2.0drl:EGFP reporter activity in emerging cardiac and pharyngeal muscle lineages at Ciona larval stage (st 26; 18 hpf (Hotta et al., 2007)) (Fig 7B-D): we observed +2.0drl:EGFP reporter activity in the atrial siphon muscle precursors (ASMPs) and in both first and second heart precursors (FHPs and SHPs). This was confirmed by co-localization of Mesp:H2B-mCherry expression that labels the cardiopharyngeal cell lineage (Davidson et al., 2005) (n =15/92; Figure S10L-M). In agreement with the drl-based LPM lineage tracing in zebrafish, we found minimal to no overlap with paraxial mesoderm progenitors, and rarely observed EGFP expression in the anterior tail muscles (ATMs) of the Ciona larval tail (Fig 7B-D). These results indicate that the zebrafish +2.0drl enhancer responds to regulatory input in the emerging multipotent cardiopharyngeal progenitors in Ciona.
Lastly, we examined the cephalochordate amphioxus (Branchiostoma lanceolatum), which belongs to the most basally divergent lineage of chordates (Bertrand et al., 2011). In amphioxus, the LPM forms from a continuous sheet of cells that encompass the dorsally emerging somites, the LPM, and the ventral-most forming endoderm (Bailey and Miller, 1921; Holland, 2018; Holland et al., 2003; Kozmik et al., 2001; Onimaru et al., 2011). At mid-neurula stage (approximately equivalent to early somitogenesis in zebrafish), the ventral wall of the somites evaginates as nascent ventral mesoderm, which by late neurula stage fuses at the midline under the gut endoderm (Holland et al., 2003; Kozmik et al., 2001; Onimaru et al., 2011). Indeed, the amphioxus orthologs of LPM-expressed transcription factors including Hand, Csx, Vent1, and AmphiNk2-tin are expressed in the ventral half of the somite territory at mid-neurula stage (Holland et al., 2003; Kozmik et al., 2001; Pascual-Anaya et al., 2013). We observed that injection of +2.0drl-based reporters into amphioxus embryos showed specific reporter activity in the ventral half of somites and in the elongating somites at mid-neurula stage (6/7 ss) (n = 61/65) (Fig 7E-G, Fig S10N,O). At early larvae stage the activity of +2.0drl-based reporter is present in the pharyngeal region (Fig S10P) where LPM is located (Holland, 2018). Hence, also in amphioxus as cephalochordate, the +2.0drl enhancer reads out the positional input active in the LPM-corresponding territory during development.
Taken together, these observations establish that the intronic +2.0drl enhancer reads out a position-dependent LPM program that remains active in tunicates, cyclostomes, teleosts, amphibians, and amniotes, and thus across all tested chordates.
Discussion
The dynamic nature of the LPM has made it challenging to monitor precisely its emergence and morphogenesis and has hindered comparative studies of its properties during chordate evolution. Here, we show that an enhancer from the zebrafish drl locus (+2.0drl) reads out an LPM-demarcating transcriptional activity in six chordate species, ranging from cephalochordates to amniotes. This suggests that the LPM-underlying transcriptional program is of ancient evolutionary origin. Characterization of the properties of this enhancer in zebrafish revealed that the transcription factors Eomes, FoxH1, and MixL1 are sufficient to trigger this basic LPM program. These observations in zebrafish suggest a regulatory model whereby, among their roles in mesendoderm regulation, Eomes, FoxH1, and Mixl1 cooperate in inducing LPM together with position-dependent Smad activity. These factors have been individually implicated in mesendoderm development in several vertebrates (Arnold et al., 2008; Bjornson et al., 2005; Germain et al., 2000; Nelson et al., 2014; Slagle et al., 2011; Zhang et al., 2009), in diverse LPM-associated contexts such as in blood formation (Mead et al., 1996) and in reprogramming towards cardiac and renal fates (Costello et al., 2011; Pfeiffer et al., 2018; Takasato and Little, 2015; Takasato et al., 2014). Based on our series of comparative and mechanistic studies, we postulate that LPM-like origins in ancestral chordates already featured the basic molecular building blocks that enabled the increasing specialization of the LPM into its sophisticated descendant cell fates observed in vertebrates.
The LPM as early established, specialized mesoderm
In vertebrates, the LPM is readily detectable after gastrulation through its position lateral to the forming somites and by several progenitor markers for hemangioblast, renal, and smooth muscle fates (Jin et al., 2006; McDole et al., 2018; Picker et al., 2002; Yin et al., 2010; Zhu et al., 2005). Genetic tracking of some aspects of the LPM has been achieved previously by various means in different models. In mouse, transgenic strains based on HoxB6, Prrx1, Bmp4, Hand1/2, Gata2, and FoxF1 enable labeling of the LPM post-gastrulation (Becker et al., 1996; Chandler et al., 2009; Firulli et al., 1998; Martin and Olson, 2000; Ormestad et al., 2004; Osterwalder et al., 2014; Rojas et al., 2005). Recent work in chick used electroporation of ubiquitous reporters specifically at the position of emerging LPM progenitors to chart the forming limb and interlimb domains (Moreau et al., 2019). In zebrafish, we had previously found and applied the cis-regulatory region of the drl gene to genetically track LPM emergence during both gastrulation and early somitogenesis (Felker et al., 2018; Gays et al., 2017; Henninger et al., 2017; Mosimann et al., 2015). While drl encodes a putative zinc-finger protein of unknown function (Herbomel et al., 1999) without a clear ortholog outside of zebrafish, the early LPM-confined expression mediated by its +2.0drl enhancer provides a unique tool to investigate LPM origins across chordates. Using in toto live imaging of drl-based reporters together with lineage-restricted reporter transgenics, we charted LPM formation in zebrafish as a continuous process building around the entire circumference of the forming embryo (Fig 1, Movie S1). This mode of progenitor formation is distinct from paraxial and axial mesoderm, which both form by progressive extension over time (Alev et al., 2013; Gurdon, 1995).
Our reporter and lineage tracing data documents a close developmental relationship between endoderm and the LPM. Expression of the key vertebrate endoderm gene sox17 demarcates prospective endoderm progenitors also in zebrafish (Alexander and Stainier, 1999; Sakaguchi et al., 2006), as evident by the exclusive endoderm lineage-labeling with sox17:creERT2 (Fig S1C) (Hockman et al., 2017). We detected sox17/drl reporter double-positive cells throughout gastrulation that begin to separate from each other from tailbud stage (Fig 2A-D). Cre/lox-based lineage tracing confirmed that drl:creERT2-expressing cells at shield stage contribute to all endoderm-derived organs, while their mesodermal contribution was exclusively to the LPM (Fig 2F). The endoderm contribution gradually decreased from mid-gastrulation to early somitogenesis in an anterior-to-posterior fashion, while LPM derivatives remained labeled throughout (Fig 2E-J, Fig S1, S2). These findings indicate that drl labels a mesendoderm population that becomes progressively dedicated towards an LPM fate in a spatio-temporal manner.
Despite this close developmental relationship, drl-labeled LPM progenitors do not seem to require endoderm for their initial morphogenesis. While LPM midline migration is perturbed in sox32 (casanova) mutants or morphants devoid of endoderm progenitors, such embryos still form bilateral, contracting hearts (Alexander et al., 1999; Dickmeis et al., 2001; Kikuchi et al., 2001) and maintain kidney and iSMC progenitor markers (Chou et al., 2016; Reichenbach et al., 2008). Our LPM lineage tracing confirmed that the LPM stripes still form even without endoderm and it documented how they migrate in the trunk to the midline and develop into structures resembling pronephros, iSMC-like structures, endothelium, red blood cells, and pectoral fin mesenchyme (Fig S3). These data imply that, despite close or even joint origin, the sox17-positive endoderm progenitors have minimal influence on initial LPM fate determination and LPM morphogenesis. Our imaging and lineage tracing data further indicates minimal overlap between early paraxial mesoderm progenitors and LPM progenitors, as evident in the rare occurrence of somatic muscle labeling by drl-expressing precursors (Fig 2F). While there is considerable heterogeneity of cell fate domains among the post-gastrulation LPM (Fig 1), our findings collectively suggest that the LPM initially emerges as a field of cells endowed with common properties.
A transcription factor code for mesendoderm-to-LPM formation
Guiding the therapeutically relevant differentiation of cultured embryonic or induced pluripotent stem (iPS) cells towards cardiovascular, hematopoietic, or renal cell fates remains challenging (De Los Angeles and Daley, 2013; Lee et al., 2017; Slukvin II, 2013; Song et al., 2012; Takasato et al., 2014). Initial differentiation leads to broadly defined mesodermal progenitors that, depending on the protocol, show a preference to early versus late primitive streak regions, mimicking the anterior-to-posterior progression of vertebrate body axis formation (Mendjan et al., 2014). Other protocols combined expression of transcription factor combinations to drive direct differentiation into specific cell fates, such as achieved for cardiomyocytes or kidney cells (Ieda et al., 2010; Song et al., 2012; Takasato and Little, 2015; Takeuchi and Bruneau, 2009). Nonetheless, directed differentiation of uncommitted cells into correct LPM progenitor states, such as mesendoderm or mesoderm, would be highly desirable to achieve increased efficiency in cardiomyocyte, blood, or kidney reprogramming (Chia et al., 2019; Lee et al., 2017; Murry and Keller, 2008). In this regard, our functional analyses in zebrafish showed that Eomes, FoxH1, and MixL1 together with BMP-induced Smads are able to drive cells towards an LPM program.
Eomes and FoxH1 cooperate in controlling BMP/Nodal target genes together with Smads(Arnold et al., 2008; Bjornson et al., 2005; Miyazono et al., 2018; Slagle et al., 2011). Eomes, FoxH1, and Mixl1 have been implicated separately or in pairwise combinations in mesendoderm development (Chen et al., 1996; Du et al., 2012; Henry and Melton, 1998; Kikuchi et al., 2000; Slagle et al., 2011; Xu et al., 2014). In particular, Eomes and MixL1 have been found to form a tripartite complex with Gata5 in early endoderm determination (Bjornson et al., 2005). Our findings reveal that the combination of Eomes, FoxH1, and Mixl1 modulates mesendodermal target genes required for progression towards LPM formation. The requirement for the combined action of all three factors becomes apparent when testing each factor individually, as there was only a marginal increase in +2.0drl expression. In contrast, the combination of all three factors was sufficient to ubiquitously induce the +2.0drl pan-LPM reporter (Fig 5K) and tmem88a, which is an early LPM marker gene (Fig S7A,B) (Mosimann et al., 2015). The three factors do however not merely boost mesendoderm fate per se, as demonstrated in embryos devoid of endodermal progenitors following sox32 perturbation (Fig 5L). The dependency of early drl expression on BMP, an less so on Nodal based on comparing genetic mutants, is further in line with the classic definition of a ventral emergence of the LPM (Gurdon, 1995). From this data, we propose the following model: maternal Eomes and FoxH1 cooperate with BMP- and Nodal-triggered Smads to prime mesendodermal target genes. Ventral induction of MixL1 provides an instructive signal that cooperates with the previous permissive mesendoderm state to trigger an LPM fate in ventral BMP-receiving blastomeres (Fig 4E-I). These findings provide a developmental framework for the contribution of Eomes, FoxH1, and MixL1 in programming of naïve pluripotent stem cells into cardiovascular and renal lineages by generating the correct mesendodermal precursor lineage. In-depth analysis of genomic targets of the three transcription factors is warranted to i) establish how this program conveys key LPM properties to uncommitted progenitor cells for their differentiation, and ii) if or which orthologs of these T-box, Forkhead, and Homeobox factors drive LPM progenitor formation across chordates.
An ancient LPM program dating back to chordate ancestors
The evolutionary origin of the LPM has remained unaddressed. In part, the discussion of evolutionary origins of key features in the vertebrate body plan is tangled by the deduction of ancestral versus derived features without an existing common chordate ancestor (Technau and Scholz, 2003). Jawed vertebrate species share thousands of conserved non-coding regulatory regions (McEwen et al., 2009) and a greatly reduced number can be traced to jawless vertebrates like lamprey(Parker et al., 2011). Nonetheless, while some of the previously isolated LPM enhancers of mouse Gata4, Bmp4, and Hoxb6 (Becker et al., 1996; Chandler et al., 2009; Rojas et al., 2005) expressed faithfully in the PLPM of chick embryos, their activity did not recapitulate an LPM pattern in zebrafish (Fig S8). These observations suggest that these mammalian LPM enhancers have specialized during amniote evolution. In contrast, our cross-species regulatory analyses of the +2.0drl enhancer uncovered a remarkable degree of regulatory conservation. While cryptic reporter activity in cross-species assays can bias results, we established that the zebrafish +2.0drl enhancer drove specific fluorescent reporter expression in LPM or LPM-related structures in six analyzed chordates: chick, axolotl, zebrafish, lamprey, Ciona, and amphioxus (Fig 6,7). Of note, in Ciona, the +2.0drl:EGFP reporter also resulted in EGFP activity in cells of the developing mesenchyme; if this reporter activity is specific considering the lineage origin of this cell type (Beh et al., 2007) or ectopic activity of the used transgene plasmid as previously reported (Stolfi and Christiaen, 2012), remains to be determined. Nonetheless, the +2.0drl enhancer provides a unique tool to investigate the upstream regulatory networks and the emergence of LPM structures across chordate development. This remarkable conservation of upstream reporter inputs further suggests that +2.0drl enhancer activity uncovers a deeply rooted LPM-inducing program, dating back to the last shared chordate ancestor.
Our findings further provide a genetic approach for investigating an LPM-delineating program that sets this mesodermal lineage apart from axial and paraxial mesoderm progenitors. The LPM program responds to ancient upstream regulatory inputs that defines the LPM from its early developmental origins across chordates. The activity of the +2.0drl reporter in the prospective LPM of amphioxus is particularly striking (Fig 7), as cephalochordates form few and only rudimentary equivalents of the vertebrate LPM-derived organ systems. The LPM in amphioxus forms as a non-segmented mesodermal sheet that is continuous with the ventral prospective endoderm and the more dorsally folding somites (Fig 7) (Bailey and Miller, 1921; Holland, 2018; Holland et al., 2003; Kozmik et al., 2001). This configuration makes it tempting to speculate that the LPM evolved from mesenchymal mesendoderm that did not integrate into the definitive endoderm or into the paraxial somites, providing ample material for diversification over deep time.
Author Contributions
K.D.P., C.H., S.N., E.C.B., S.B., C.M. designed, performed, and interpreted zebrafish experiments; S.N., E.C., A.B. established and performed chicken experiments; K.D.P. performed lightsheet imaging with technical and equipment support by G.S., J.H.; K.W.R., P.M. provided and generated mutants and maternal-zygotic mutant zebrafish; H.J.P., M.B., R.K. designed, performed, and interpreted lamprey experiments; A.C., D.K., M.H.Y. designed, performed, and interpreted axolotl experiments; C.R., L.C. designed, performed, and interpreted Ciona experiments; I.K., Z.K. designed, performed, and interpreted amphioxus experiments; A.C., D.K., and M.H.J. designed and performed axolotl experiments; K.D.P., C.H., S.N., and C.M. assembled and wrote the manuscript with contributions from all co-authors.
Declaration of Interests
The authors declare no competing interests.
Materials and Methods
Animal experiments and husbandry
Zebrafish and chick experiments were carried out in accordance with the recommendations of the national authorities of Switzerland (Animal Protection Ordinance). The protocols and the experiments were approved by the cantonal veterinary office of the Canton Zurich (Kantonales Veterinäramt, permit no. 150). Zebrafish care and all experimental procedures were carried out in accordance with the European Communities Council Directive (86/609/EEC), according to which all embryo experiments performed before 120 hours post fertilization are not considered animal experimentation and do not require ethics approval. Adult zebrafish for breeding were kept and handled according to animal care regulation of the Kantonales Veterinäramt Zürich (TV4209). All zebrafish (Danio rerio) were raised, kept, and handled as described (Westerfield, 2007). White mutant (d/d) axolotls (Ambystoma mexicanum) were obtained from the axolotl facility at the TUD-CRTD Center for Regenerative Therapies Dresden, Germany.
Lamprey studies were conducted in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health, and protocols were approved by the Institutional Animal Care and Use Committees of the California Institute of Technology (Protocol # 1436-11).
Transgenic constructs and transgenic zebrafish lines
The upstream cis-regulatory elements of the zebrafish drl gene (ENSDARG00000078004; ZDB-GENE-991213-3) were amplified from zebrafish wildtype genomic DNA (Extended Data Table 1; regulatory elements) and TOPO-cloned into the pENTR™ 5’-TOPO® TA Cloning® plasmid (Invitrogen) according to the manufacturer’s instructions.
Subsequent cloning reactions were performed with the Multisite Gateway system with LR Clonase II Plus (Life Technologies) according to the manufacturer’s instructions. −1.0drl:EGFP (pDestTol2pA2_-1.0drl:EGFP) and [proximal]drl:EGFP (pDestTol2pA2_[proximal]drl:EGFP) were assembled from pENTR/5’_−1.0drl or pENTR/5’_[proximal]drl together with Tol2kit vectors #383 (pME-EGFP), #302 (p3E_SV40polyA), and #394 (pDestTol2A2) as backbone. +2.0drl:EGFP (pDestTol2pA2_+2.0drl:EGFP) and +2.4drl:EGFP (pDestTol2pA2_ +2.4drl:EGFP) were assembled from pENTR/5’_+2.0drl or pENTR/5’_+2.4drl together with pME-β-globin_minpromoter_EGFP (Tamplin et al., 2015), Tol2kit vectors #302 (p3E_SV40polyA), and #394 (pDestTol2A2) as backbone (Kwan et al., 2007). +2.0drl:creERT2 (pDestTol2CY_+2.0drl:creERT2,alpha-crystallin:YFP) and - 5.0sox17:creERT2 (pDestTol2CY_-5.0sox17:creERT2,alpha-crystallin:YFP) were assembled from pENTR/5’_+2.0drl or pENTR/5’_-5.0sox17 together with pCM293 (pENTR/D_creERT2) (Mosimann et al., 2011), Tol2kit vector #302 (p3E_SV40polyA), and pCM326 (pDestTol2CY) as backbone (Mosimann et al., 2015). Genomic coordinates for the +2.0drl enhancer used in the described constructs are chr5:61,649,227-61,650,194 (GRCz11/danRer11). Transcription factor binding sites were predicted using the JASPAR online interface (Khan et al., 2018).
The regulatory elements of mouse-specific LPM enhancers were PCR-amplified (Extended Data Table 1; regulatory elements), TOPO-cloned into the pENTR5’ plasmid, and assembled together with pKD001 (pME-β-globin_minpromoter_EGFP with improved Kozak sequence), Tol2kit vectors #302 (p3E_SV40polyA), and #394 (pDestTol2A2) as backbone.
Assembled reporter constructs were injected at a concentration of 25 ng/μl together with 25 ng/μl Tol2 mRNA for Tol2-mediated zebrafish transgenesis (Felker and Mosimann, 2016). Injected F0 founders were screened for specific EGFP or alpha-crystallin:YFP expression. Zebrafish were raised to adulthood and screened in F1 for germline transmission. Single-insertion transgenic strains were generated, and microscopy images were taken on a Leica M205FA with a Leica DFC450C digital camera or a Leica SP8 confocal microscope with Plan-Apochromat 20x/0.5 objective. Images were processed using Leica LAS and Fiji (Schindelin et al., 2012).
Established transgenic and mutant lines used in this study included drl:EGFP (Mosimann et al., 2015), drl:mCherryzh705 (Sánchez-Iranzo et al., 2018), drl:creERT2 (Mosimann et al., 2015), ubi:creERT2 (Mosimann et al., 2011), ubi:switch (Mosimann et al., 2011), hsp70l:Switchzh701(Felker et al., 2018), lmo2:dsRED2 (Zhu et al., 2005), scl/tal1:EGFP (Jin et al., 2006), pax2.1:EGFP (Picker et al., 2002), hand2:EGFP (Yin et al., 2010), actb2:h2afva-mCherry (Krens et al., 2011), maternal and zygotic EGF-CFC co-receptor oep gene mutants (MZoep) (Gritsman et al., 1999), and maternal-zygotic somitabundtc24 mutants (sbn, dominant-negative smad5) (Hild et al., 1999; Mullins et al., 1996).
Zebrafish CreERT2-based lineage tracing
Lineage tracing experiments were performed by crossing female hsp70l:Switch or ubi:Switch (Mosimann et al., 2011) reporter carriers with the male creERT2 drivers drl:creERT2 (Mosimann et al., 2015), sox17:creERT2, and +2.0drl:creERT2. Embryos were induced using 4-Hydroxytamoxifen (4-OHT) (H7904; Sigma H7904) from fresh or pre-heated (65°C for 10 min) stock solutions in DMSO at a final concentration of 10 μM in E3 embryo medium (Felker et al., 2016) at indicated time points. Embryos were washed in untreated E3 medium at 24 hpf. To induce EGFP transcription in hsp70l:Switch carrying embryos, the embryos were incubated at 37°C for 1 hour, 3 hours before fixation. Embryos were fixed with 4% paraformaldehyde (PFA) overnight at 4°C at 3 dpf and processed for confocal analysis.
Zebrafish transverse vibratome sections
Transverse sections were generated as previously described (Gays et al., 2017). Fixed embryos were washed in PBS, embedded in 6% low-melting agarose (Sigma-Aldrich) in PBS/0.1% Tween-20 (Sigma-Aldrich), and cut into 130 μm thick sections using a vibratome (Leica VT 1000S). Sections were mounted in DAPI-containing Vectashield (Cat#H-1200; Vector Laboratories). Sections were analyzed with a Zeiss LSM710 confocal microscope with a Plan-Apochromat 40x/1.3 oil DIC M27 objective. Images were cropped and adjusted for brightness using ImageJ/Fiji (Schindelin et al., 2012). Graphs were generated in Graphpad Prism 5.
Zebrafish morpholino and crispant experiments
The previously established sox32 morpholino (Sakaguchi et al., 2001) was synthesized by GeneTools (sequence: 5’-TGCTTTGTCATGCTGATGTAGTGGG-3’) and dissolved in nuclease-free water to a stock concentration of 10 µg/µl. The sox32 morpholino injection mix of 4 µg/µl was incubated for 10 min at 65°C, and 8 ng were microinjected into one-cell stage zebrafish embryos.
sgRNAs were obtained in an oligo-based approach as previously described (Burger et al., 2016). Briefly, primer extension was performed using Phusion polymerase (NEB) of forward primer 5’-GAAATTAATACGACTCACTATAGG-N20-GTTTTAGAGCTAGAAATAGC-3’ (including a 20 nt target site) and invariant reverse primer 5’-AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTA TTTCTAGCTCTAAAAC-3’ (PAGE-purified) (Bassett et al., 2013). The following target sites are shown in this manuscript: 5’-AGATTTGTTTAGTCAGTGTC-3’ (drl ccG), 5’-GTCTGGAAACAGTCTGAATC-3’ (drl ccH), 5’-GAGACTTCGCCCTTCGGTTC-3’ (mixl1 ccB), and 5’-GACAGAACAGGCCACGTTGA-3’ (mezzo ccA).
In vitro transcription of sgRNAs was performed as previously described (Burger et al., 2016) using MAXIscript T7 (Ambion). Afterwards, RNA was precipitated with ammonium acetate, washed with 75% ethanol, and dissolved in DEPC water. RNA quality was checked on a MOPS gel. RNPs of Cas9 protein (Cas9-mCherry, pMJ293 (Burger et al., 2016) (available from Addgene)) and sgRNA were in vitro-assembled for 10 min at 37°C in 300 mM KCl to ensure maximum cleavage efficiency and microinjected into the cell of one-cell stage embryos (Burger et al., 2016). The CRISPR target regions in the transgenic locus were amplified using specific target region amplification primers (Table S1; target region amplifications).
Sequencing analysis was performed with the R software package CrispRVariants for allele- and sequence-level variant counting and visualization (Lindsay et al., 2016).
Zebrafish overexpression experiments
For early developmental transcription factor genes, ORFs were PCR-amplified from mixed-stage zebrafish cDNA using ORF-specific primers (Table S1; coding sequence primers); full-length zebrafish eomesA was derived from I.M.A.G.E clone IRBOp991A0739D (Source BioScience LifeSciences, UK). All CDS were TOPO-cloned into the pENTR/D™ Directional TOPO® plasmid (Invitrogen) according to the manufacturer’s instructions. Constitutive-active Smad2, which encodes an N-terminal truncation of zebrafish Smad2 (Dick et al., 2000), was amplified from zebrafish cDNA and subcloned into pENTR1A (pCM269). Subsequent cloning reactions were performed with the Multisite Gateway system with LR Clonase II Plus (Invitrogen) according to the manufacturer’s instructions and as previously described (Felker and Mosimann, 2016). The CDS as pENTR/D vectors (pENTR1A for smad2(ca)) were assembled with pENTR5’_ubi (Mosimann et al., 2011), pENTR5’_T7-VP16, pENTR5’_T7-eng or pENTR5’_T7 together with Tol2kit vectors #302 (p3E_SV40polyA), and #394 (pDestTol2A2) as backbone (Kwan et al., 2007). Afterwards, plasmids were linearized (in case of T7-VP16, T7-eng, or T7) and in vitro transcribed (Roche). The EomesA-VP16 plasmid (containing 153aa-431aa of the zebrafish EomesA ORF) (Bruce et al., 2003) was kindly provided by Dr. Rebecca Burdine. The plasmid was linearized via NotI followed by SP6 in vitro transcription. The Ntl-VP16 plasmid was kindly provided by Dr. Ashley Bruce. The etv5a-VP16 and etv5a-eng plasmids were kindly provided by Dr. Michael Tsang. All mRNAs were precipitated with ammonium acetate, washed with 75% ethanol, and dissolved in DEPC water. mRNA quality was checked on MOPS gels as described (Burger et al., 2016).
Zebrafish chemical treatments
Chemicals for performed zebrafish treatments were dissolved in DMSO. Dorsomorphin (10-30 µM; Sigma-Aldrich) (Yu et al., 2008) and SB-505124 (30-60 µM; Sigma-Aldrich) (DaCosta Byfield et al., 2004; Rogers et al., 2017) were administered at 1-cell stage and embryos kept in the treated E3 until fixation.
Zebrafish whole-mount in situ hybridization
Total RNA was extracted from zebrafish embryos from various stages during development. This RNA was used as template for generation of first-strand complementary DNA (cDNA) by the Superscript III First-Strand Synthesis kit (Cat#18080051; Invitrogen). In situ hybridization (ISH) probes were designed with an oligonucleotide-based method (including T7 promoter added to the reverse primers) using zebrafish cDNA (Table S1; in situ hybridization probes). The following oligonucleotide pairs (including T7 promoter added to the reverse primers) were used to amplify the DNA template from zebrafish cDNA. The ISH probe for admp was obtained from a pCS2_ADMP plasmid, and sizzled from pCS2_Sizzled. Admp and sizzled were linearized by ClaI. The gata2a probe was obtained from the middle entry vector pCM238. For in vitro transcription, T7 RNA polymerase (Roche) and digoxigenin (DIG)-labeled NTPs (Roche) were used. Afterwards, RNA was precipitated with lithium chloride, washed with 75% ethanol, and dissolved in DEPC water. RNA quality was checked on a MOPS gel. ISH on whole-mount zebrafish embryos was executed as described before (Thisse and Thisse, 2008). After ISH, embryos were transferred to 80-95% glycerol (Sigma-Aldrich), and microscopy images were taken on a Leica M205FA with a Leica DFC450C digital camera. Images were cropped and adjusted for brightness using ImageJ/Fiji. The pathway schematic in Fig 4A was generated using BioRender.
Zebrafish selective plane illumination microscopy (SPIM)
At 30-50% epiboly, embryos in the chorion were embedded into 1% low-melting agarose with optional 0.016% Ethyl 3-aminobenzoate methanesulfonate salt (Tricaine, Cat#A5040; Sigma) in E3 embryo medium, and sucked into an FEP tube (inner diameter: 2.0 mm, wall thickness: 0.5 mm). 6-7 embryos were positioned on top of each other. The FEP tube was mounted in the microscope imaging chamber filled with E3 medium. Panoramic (3D) SPIM/lightsheet microscopy and subsequent image processing (Mercator projections) were performed as previously described (Schmid et al., 2013). A z-stack of 402 planes was obtained from every embryo with an interval of 2 min for a period of 14-17 hours. Images were processed using Leica LAS, ImageJ/Fiji, and Photoshop CS6.
Chicken embryo incubation and ex-ovo culturing
Ex ovo culturing was adapted from previously established protocols (New, 1955). Fertilized chicken eggs were obtained from a local hatchery and stored at 12°C up to maximum of 14 days. Prior to use, eggs were incubated horizontally for 17 hours until Hamburger-Hamilton (HH) 3+/4 in a 39°C incubator with 55-65% humidity. After incubation, the eggs were kept for at least 30 min at RT before opening. Eggs were opened in a petri dish and a layer of thick albumin together with the chalaziferous layer was removed using a plastic Pasteur pipette. A paper ring was placed around the embryo on the yolk and dissection scissors were used to cut the yolk membrane around the ring. The paper ring with the embryo was cleaned from remaining yolk and transferred and placed upside down on a semisolid albumin/agarose (43.5 ml thin albumin incubated for 2 hours at 55°C, 5 ml 2% agarose, 1.5 ml 10% glucose in 30 mm petri dishes) culturing plate. Embryos were recovered for at least 2 hours at RT before electroporation.
Chicken embryo injection and electroporation
For electroporations, a customized electroporation chamber was used containing an electrode with a positive pole on the bottom of the chamber and separate negative electrode on a holder (kindly provided by the lab of Dr. Jerome Gros). Both electrodes were mounted and connected to a square wave electroporator (BTX ECM 830). The electroporation chamber was filled with HBSS (Gibco Life Technologies), and the embryo-containing paper ring was placed in the chamber with the dorsal side up. The DNA mixture was injected by a mouth injector along the primitive streak beneath the pellucid membrane. The positive electrode holder was placed on top of the streak to allow electricity pulses flow through the embryo (3 pulses, 8V x 50 milliseconds, 500 milliseconds interval). The embryos were placed back on the albumin culturing plates with the ventral side up and placed back at 39°C until HH8-9. Microscopy images of the embryos were taken at HH8-9 on a Leica M205FA with a Leica DFC450C digital camera. Images were cropped and adjusted for brightness using ImageJ/Fiji. All injection mixtures for electroporations contained 0.1% fast green dye, 0.1% methyl-cellulose, 300 ng/µl control plasmid pCAGGs (pCMV:H2B-CAGG-RFP, abbreviated for chicken β-actin promoter CAGG-mCherry) and 1 µg/µl of the plasmid of interest.
Axolotl experiments
The generation of transgenic animals and determination of developmental stages were performed as described previously (Bordzilovsakya et al., 1989; Khattak et al., 2014). Animals at stage 43 were anaesthetized by bathing in 0.01% benzocaine (Khattak et al., 2014).
Live imaging was performed on an Olympus SZX16 fluorescence stereomicroscope. Time lapse movies were acquired using an Axio Zoom.V16 (Zeiss) stereomicroscope. Confocal images were acquired on a Zeiss LSM780-FCS inverted microscope.
For immunostaining, embryos were fixed in MEMFA at 4°C overnight, washed in PBS, embedded in 2% low melting temperature agarose and sectioned by vibratome into 200 µm thick sections. Fibronectin was detected using mouse anti-Fibronectin antibody (ab6328, Abcam) at 5µg/ml.
Lamprey experiments
The +2.0drl regulatory element was amplified from the zebrafish vector +2.0drl:EGFP by PCR using KOD Hot Start Master Mix (Novagen) (Table S1; regulatory elements). The amplified enhancers were cloned into the HLC vector for lamprey transgenesis (Parker et al., 2014b), containing the mouse c-Fos minimal promoter, by Gibson assembly using the Gibson Assembly Master Mix (NEB). Injections for I-SceI meganuclease-mediated lamprey transient transgenesis were performed using P. marinus embryos at the one-cell stage with injection mixtures containing 0.5 U/µl I-SceI enzyme and 20 ng/µl reporter construct as described previously (Parker et al., 2014a). Selected EGFP-expressing embryos were fixed in MEM-FA and dehydrated in methanol for in situ hybridization. EGFP-expressing embryos were imaged using a Zeiss SteREO Discovery V12 microscope with variable zoom and a Zeiss Axiocam MRm camera with AxioVision Rel 4.6 software. Images were cropped and adjusted for brightness using Adobe Photoshop CS5.1.
For mRNA ISH, total RNA was extracted from st 21-26 P. marinus embryos using the RNAqueous Total RNA Isolation Kit (Ambion). This was used as a template for 3’ rapid amplification of cDNA ends (RACE) with the GeneRacer Kit and SuperScript III RT (Invitrogen). A 339bp-long pmHandA in situ probe was designed based on a characterized cDNA sequence from the closely related Arctic lamprey (Lethenteron camtschaticum) (Kuraku et al., 2010), and this sequence was amplified by PCR from 3’ RACE cDNA using KOD Hot Start Master Mix (Novagen) with the following primers: 5’-GCGGAGGACATTGAGCATC-3’ (forward) and 5’-TGGAATTCGAGTGCCCACA-3’ (reverse). The cDNA fragment was cloned into the pCR4-TOPO vector (Invitrogen). The 709bp-long eGFP probe was described previously (Parker et al., 2014b).
DIG-labelled probes were generated and used in lamprey whole-mount ISH as described previously (Sauka-Spengler et al., 2007). Embryos were cleared in 75% glycerol and imaged using a Leica MZ APO microscope with variable zoom and Lumenera Infinity 3 camera with Lumenera Infinity Capture v6.5.3 software. Images were cropped and adjusted for brightness using Adobe Photoshop CS5.1.
After ISH, selected embryos were transferred into 30% sucrose in PBS, embedded in O.C.T. Compound (Tissue-Tek), and cut into 10 µm-thick cryosections using a CryoStar NX70 cryostat (Thermo Scientific). Images were taken using a Zeiss Axiovert 200 microscope with an AxioCam HRc camera and AxioVision Rel 4.8.2 software. Original data underlying the lamprey experiments in this manuscript are accessible from the Stowers Original Data Repository at http://odr.stowers.org/websimr/.
Ciona experiments
+2.0drl was amplified from the zebrafish vector +2.0drl:EGFP and sub-cloned upstream of unc76:GFP to generate a Ciona reporter construct including minimal promoter (pBuS24; see Table S1, regulatory elements for primer sequences). Gravid Ciona robusta adults were obtained from M-REP (San Diego CA, USA). To test the activity of the zebrafish enhancers in Ciona robusta, 80 μg of +2.0drl:EGFP was injected in a mixture with the reporter plasmid for Mesp (Davidson et al., 2005) to mark the B7.5 cardiopharyngeal lineage with H2B:mCherry (10 μg). For antibody staining, embryos were fixed in 4% MEM-PFA for 30 min, rinsed several times in PBT (PBS/0.1% Tween-20), and incubated with anti-GFP (1:500, mouse mAb, Roche) with 2% normal goat serum in PBT at 4°C overnight. Embryos were washed in PBT and then incubated with donkey anti-mouse secondary antibody (1:1000) coupled to Alexa Fluor 488 (Life Technologies) in PBT with 2% normal goat serum for 2 hours at RT, then washed in PBT (Racioppi et al., 2014).
Amphioxus experiments
The regulatory elements drl (entire 6.35 kb) and +2.0drl were amplified from the zebrafish reporter vector drl:EGFP (Mosimann et al., 2015) and +2.0drl:EGFP and subcloned upstream of a EGFP reporter in the pPB vector carrying PiggyBac transposon terminal repeats (Kozmikova and Kozmik, 2015). Adults of Branchiostoma lanceolatum were collected in Banyuls-sur-mer, France, prior to the summer breeding season and raised in the laboratory until spawning. The spawning of amphioxus male and females was induced by shifting of the temperature as described (Fuentes et al., 2007). For microinjection of amphioxus eggs, a mixture of pPB-drl:EGFP or pPB-+2.0drl:EGFP (200 ng/μl) with PiggyBac transposase mRNA (100 ng/μl) in 15% glycerol was used. Transgenic embryos were allowed to develop until neurula stage, fixed in 4% PFA overnight at 4°C, mounted with Vectashield with DAPI (Vector Laboratories), and analyzed using a Leica SP5 confocal microscope. The confocal images were adjusted for brightness and contrast with ImageJ/Fiji.
Acknowledgements
We thank Seraina Bötschi, Lukas Obernosterer, and Vesna Barros for technical and husbandry support; the lab of Dr. Stephan Neuhauss for zebrafish husbandry support; the labs of Dr. Esther Stöckli and Dr. Jerome Gros for chicken experimentation support; the ZMB at UZH for imaging support; Dr. Fiona Wardle for input and support on the ChIP-seq panel in Figure 5A; the lab of Dr. Magdalini Polymenidou for vibratome access; Karolína Ditrychová for cloning the pKD001 construct; Dr. Ashley Bruce, Dr. Rebecca Burdine, and Dr. Michael Tsang for sharing transcription factor constructs; Mark Miller for animal illustrations; the Stowers Institute histology facility for assistance with lamprey embryo sectioning; Dr. Hans-Henning Epperlein for discussions on salamander embryology; and all members of the Mosimann lab for constructive input. This work has been supported by a Swiss National Science Foundation (SNSF) professorship [PP00P3_139093] and SNSF R’Equip grant 150838 (Lightsheet Fluorescence Microscopy), a Marie Curie Career Integration Grant from the European Commission [CIG PCIG14-GA-2013-631984], the Canton of Zürich, the UZH Foundation for Research in Science and the Humanities, the Swiss Heart Foundation, and the ZUNIV FAN/UZH Alumni to C.M; a UZH CanDoc to C.H.; EuFishBioMed and Company of Biologists travel fellowships to K.D.P.; the Stowers Institute (grant #1001) to H.J.P. and R.K.; NIH/NHLBI R01 award HL108643, trans-Atlantic network of excellence award 15CVD01 from the Leducq Foundation to L.C.; a long-term fellowship ALTF 1608-2014 from EMBO to C.R; Alexander von Humboldt fellowship to A.C. and DFG Research Center (DFG FZ 111) and Cluster of Excellence (DFG EXC 168) funds to M.H.Y.; Czech Science Foundation 17-15374S to Z.K.
Footnotes
Expanded manuscript with more details on molecular and cross-species data. Added Axolotl data for drl reporter. Nomenclature changes for genetic elements to adhere to nomenclature guidelines.