Introduction

Embryonic heart development is directed by an evolutionarily conserved network of transcription factors. Notably, members of the GATA, NK-2 homeobox, TBX, HAND, and MEF2 transcription factor families play crucial, conserved, and well characterized roles in specifying cardiac cells from the mesoderm^1–4. Through a careful balance of positive and negative regulatory signals, the cardiac transcription regulatory network coordinates with other transcriptional networks to ensure that the appropriate number of cardiac progenitors are specified. In zebrafish, this activity partitions the anterior lateral plate mesoderm (LPM) into multiple fates including the cardiac mesoderm, which is bounded anteriorly by blood and endothelial progenitors and posteriorly by pectoral fin progenitors^5–8.

Chromatin remodeling and epigenetic modifications provide an additional level of transcriptional control at cardiac genes by regulating the binding of transcription factors and the binding and elongation rate of RNA Polymerase II (RNA Pol II). For example, the BAF (Brg1/Brm-associated factor) complex, an ATP-dependent chromatin remodeling complex that actively positions nucleosomes, is critical for normal vertebrate cardiogenesis^9,10. The BAF complex directly interacts with cardiac transcription factors including Tbx5, Nkx2-5, and Gata4¹¹ and controls temporal changes in chromatin accessibility during cardiac differentiation⁹. Intriguingly, transplanted cells overexpressing the BAF component Baf60c/Smarca3b and Gata5 preferentially contribute to the embryonic zebrafish heart¹². Similarly, forced expression of Baf60c along with Tbx5/Nkx2-5/Gata4 can induce cardiac differentiation from non-cardiac mesoderm¹³, suggesting that nucleosomal positioning and chromatin accessibility are key mechanisms regulating the transcription of cardiac genes.

Accumulating evidence suggests that cardiac gene expression is also highly sensitive to the activity of transcriptional pausing and elongation factors. Spt6 (Suppressor of Ty 6) is a histone chaperone that promotes RNA Polymerase II elongation and co-transcriptionally modulates chromatin structure^14,15. In zebrafish, loss of function of Spt6 disrupts cardiomyocyte differentiation and interacts genetically with the transcriptional pausing/elongation factor Spt5 (Suppressor of Ty 5) to support Nkx2.5+ cardiac progenitor specification¹⁶. While both Spt6 and Spt5 promote transcription elongation, they likely do so via distinct mechanisms. Spt5 directly interacts with Spt4 to form the DSIF (DRB Sensitivity Inducing Factor) complex, which has both inhibitory and stimulatory roles in transcription¹⁷. Unphosphorylated Spt5 fosters promoter-proximal pausing of RNA Polymerase II, but phosphorylation of Spt5’s C-terminal repeat region by Cdk9 (cyclin dependent kinase 9) produces a binding site for the Polymerase Associated Factor 1 complex (PAF1C) component Rtf1, promotes transcriptional pause release, and converts DSIF into an elongation-promoting factor^18–20. How the transcription regulatory activities of these elongation and pausing factors facilitate cardiac development is not yet understood.

We have previously shown that Rtf1 and other PAF1C members differentially regulate cardiogenesis²¹. The PAF1C consists of five core proteins, Paf1, Ctr9, Cdc73, Rtf1, and Leo1, and plays wide-ranging transcription regulatory roles including promoting co-transcriptional epigenetic modification of histones, supporting both pausing and elongation of RNA Polymerase II, influencing mRNA 3’ end formation, and serving as a bridge between specific transcription factors and RNA Polymerase II^22–24. In zebrafish, Leo1 is dispensable for the specification of cardiomyocytes, but is required for differentiation of the atrioventricular boundary and maturation of the cardiac chambers²⁵. Embryos lacking Ctr9, Cdc73, or Paf1, on the other hand, exhibit a deficiency of cardiomyocytes and only form small, dysmorphic hearts. Intriguingly, loss of the Rtf1 subunit of the PAF1C results in the most dramatic defects in cardiogenesis. Rtf1 deficient embryos have severe reduction in nkx2.5+ cardiac progenitors and fail to generate any cardiomyocytes²¹. Given the numerous roles played by the PAF1C in transcription regulation, it remains an open question how PAF1C members mechanistically regulate cardiogenesis.

In this study, we investigated the molecular mechanisms of Rtf1-dependent cardiac progenitor formation. Using CRISPR mutagenesis we showed that zygotic Rtf1 mutants, like rtf1 morphants, lack expression of most cardiac progenitor marker genes and produce no cardiomyocytes. In addition, knockout of Rtf1 in the cardiogenic mesoderm of mouse embryos and knockdown of Rtf1 in mouse ESCs reduced the formation of cardiac tissue, indicating that the role of Rtf1 in specification of the cardiac lineage is conserved among vertebrates. Using FACS and single cell sequencing-based approaches, we show that differentiation of the LPM into cardiac precursors requires Rtf1 activity. We investigated which of Rtf1’s functional domains participated in cardiac development and found that its Plus3 domain is critical for cardiac progenitor formation, while its HMD (Histone modification domain) is dispensable. Consistent with the Plus3 domain’s function as an interaction point between Rtf1 and the DSIF component Spt5, Rtf1 deficient embryos display a genome-wide decrease in promoter-proximal pausing of RNA Polymerase II. Excitingly, chemical and morpholino inhibition of pause release improved pausing at cardiac genes and restored cardiac progenitor formation to Rtf1-deficient embryos. Together, our findings suggest that during development, Rtf1’s primary function in cardiac gene regulation is to support promoter proximal pausing, and that normal levels of pausing facilitate expression of many genes, including key members of the cardiac transcription program.

Results

Rtf1 is essential for heart development in zebrafish

Morpholino knockdown of rtf1 abolishes cardiac progenitor formation in zebrafish, indicating an essential role for Rtf1 in cardiogenesis²¹. To genetically corroborate this finding, we used CRISPR/Cas9 to generate stable rtf1 mutant lines. We identified two independent lines (rtf1^LA2678 and rtf1^LA2679) with small deletions in exon 3 of the rtf1 gene that are predicted to result in frameshifts and early stop codons and which nearly eliminate the production of Rtf1 protein (Fig 1A-C). Like rtf1 morphants, both rtf1^LA2678 and rtf1^LA2679 mutants display no cardiac tissue at 24 hours post fertilization (hpf) and lack the nkx2.5+ cardiac progenitor population (Figure 2A,B,E,F). Rtf1 deficient embryos also fail to express normal levels of other cardiac transcription factors including mef2ca and tbx5a (Fig 2C,D,G,H), and exhibit reduced expression of tbx20, a key regulator of cardiac cell number that demarcates the anterior lateral plate mesoderm (ALPM) at the 8-somite stage (Fig 2I,J)²⁶. Knockdown and knockout of Rtf1 activity produced comparable changes in cardiac gene expression (Fig 2A-J), demonstrating that morpholino knockdown faithfully recapitulates the rtf1 mutant phenotype. Together, these data show that Rtf1 activity is required for the expression of multiple cardiac transcription factors and suggest that the absence of cardiomyocytes in Rtf1 deficient embryos is likely a result of this dramatic gene expression defect.

Figure 1.

Figure 2.

Conserved role of Rtf1 in mammalian cardiac development

The sequence of the Rtf1 protein is highly conserved among vertebrates (83% identity at the amino acid level between fish and humans) indicating that it is likely playing a fundamental role in biology. To examine whether Rtf1 activity is also needed for cardiogenesis in mammals, we expressed Cre-recombinase under the control of the Mesp1 promoter in Rtf1-flox mouse embryos (Fig 3A)²⁷. The Mesp1 promoter drives expression of Cre in the early cardiogenic mesoderm (E6.25)²⁸, allowing for knockout of Rtf1 activity prior to the onset of heart development. Rtf1 Mesp1-Cre knockout embryos die prior to birth, suggesting that Rtf1 function in the cardiogenic mesoderm is essential for survival. We next examined the developing hearts of Rtf1 Mesp1-Cre knockouts and their wild type siblings by in situ hybridization using Nkx2.5 and Tbx20 as cardiac markers and found that while Cre+;Rtf1flox/+ and Cre-embryos display a heart tube undergoing looping at E8.5, Rtf1 Mesp1-Cre knockouts have little or no expression of these markers (Fig 3B). Cardiac muscle, marked by expression of MLC2v at E9.5, was also dramatically reduced in Mesp1-Cre Rtf1 knockouts (Fig 3B). These data suggest a conservation of the critical role of Rtf1 in differentiation of the cardiac lineage from the mesoderm.

Figure 3.

To further investigate Rtf1’s role in cardiac differentiation, we knocked down Rtf1 activity in mouse embryonic stem cells (mESCs) using short hairpin RNA (Fig 4A,B). Unmanipulated embryoid bodies (EBs) derived from mESCs are capable of differentiating into lineages representing all three germ layers, as evidenced by temporally dynamic expression of Brachyury (mesoderm), Afp (endoderm), and Ncam1 (ectoderm) (Fig 4E-G). Similarly, we found that Rtf1 shRNA mESCs induce expression of Brachyury, Afp, and Ncam1 in a temporally appropriate manner, indicating that Rtf1 activity is dispensable for germ layer formation (Fig 4E-G). However, EBs derived from Rtf1-deficient mESCs produced significantly fewer beating patches (∼67% reduction) and also displayed a dramatic reduction in cardiac genes like Myh6, Nkx2.5, and ANF/Nppa (Figure 4C,D). Together, these data demonstrate that Rtf1 is essential for differentiation of the cardiac lineage from mesoderm in mammals, and reveal a conservation of Rtf1’s function in cardiogenesis between fish and mammals.

Figure 4.

Rtf1 is required for expression of genes associated with the progression of cardiac precursor formation

We previously found that loss of Rtf1 activity does not disrupt expression of the LPM marker hand2²¹. To better understand how Rtf1 directs cardiac differentiation from multipotent mesoderm cells, we took advantage of a hand2:GFP zebrafish transgenic reporter line to isolate LPM cells from wild type and Rtf1 deficient embryos for transcriptomic analysis (Fig 5A-C, Supplementary Figure 1). Compared to wild type control samples, the rtf1 morphant LPM displayed 1420 significantly downregulated and 1185 significantly upregulated genes (adjusted p-value < 0.05) at the 10-12 somite stage when cardiac progenitor formation is underway in the LPM (Supplementary Data File 1). We examined the most significantly differentially expressed genes and observed that many of the significantly downregulated genes, including nkx2.5, rbfox1l, atp2a2a, tnnt2a, isl2b, mef2ca, mef2cb and wnt2bb, are cardiac-lineage markers or have critical known roles in cardiac development (Fig 5D). Indeed, the most significantly downregulated genes were highly enriched for gene ontologies related to cardiac development, embryonic heart morphogenesis, and muscle cell development (Fig 5E). These data show that Rtf1 activity is required for expression of a broad cardiac gene program in the LPM that is likely under the control of multiple cardiac transcription factors.

Figure 5.

To further analyze the deployment of the cardiac gene program during cardiac progenitor formation from the ALPM, we performed Single Cell Multiome sequencing of control and rtf1 morphant embryos at the 11-12 somite stage. Clustering of control and rtf1 morphant cells based on gene expression and chromatin accessibility resolved 39 cell types corresponding to expected major cell lineages (Supplementary Data File 2, Supplementary Figures 2 & 3). Interestingly, while rtf1 morphant cells of each cell type were present, the frequencies of cell types differed significantly between controls and rtf1 morphants (Supplementary Figure 4). Notably, we also observed that while rtf1 morphant cells expressing the precardiac mesoderm marker mef2cb were present, they failed to express other cardiac marker genes including ryr2b and tnni1b (Supplementary Figure 3).

To examine developmental trajectories within the ALPM, we reclustered ALPM cells (sema3e+/gata5+ LPM) and ALPM derivatives (putative precardiac mesoderm, endothelial precursors, rostral blood precursors) (Figure 6A,B). We next calculated pseudotime trajectories based on the assumption that undifferentiated ALPM, intermediates, and more highly differentiated derivatives are present at this developmental timepoint (Figure 6C). In support of this notion, we observed a gradual increase in the expression levels of cardiac markers including ttn.2, mef2cb, tnnt2a, ryr2b, and myh7bb along the ALPM-cardiac developmental trajectory in control embryos (Fig 6D,E). Interestingly, Rtf1 deficient embryos properly downregulate expression of the ALPM marker sema3e, maintain the expression of gata5 in cardiac precursors, and initiate expression of the early cardiac lineage genes bmp6, ttn.2, and mef2cb (Fig 6D-G). However, loss of Rtf1 activity prevented the activation of cardiac lineage-restricted markers that are normally expressed later in the ALPM-cardiac trajectory including tnnt2a, ryr2b, and myh7bb (Fig 6D-G). Taken together, these data suggest that Rtf1 activity is required for gene expression changes associated with the progression of cardiac precursors into a more mature state, and that Rtf1-deficient cardiac precursors are arresting in an immature state rather than transfating to a different cell lineage.

Figure 6.

Rtf1’s Plus3 functional domain is necessary for cardiac progenitor formation

The Rtf1 protein is a multifunctional platform for transcription regulation: it consists of several defined functional domains that mediate its interaction with other transcription regulatory proteins and serve as additional points of contact for these proteins with the transcription apparatus (Figures 1A, 7A)^29,30. The N-terminal half of Rtf1 notably contains its Histone Modification Domain (HMD), which mediates its interaction with E2 ubiquitin-conjugating enzymes and supports the ubiquitylation of histone H2B K120^31–33. Rtf1’s C-terminus contains the Plus3 domain, which interacts with the C-terminal repeat region of the pausing and elongation factor Spt5^18,19,34, a Pol II-interaction domain that allosterically stimulates transcription elongation³⁰, and a PAF1C interaction domain.

We found that injection of rtf1 mRNA with 8 silent mutations that render it morpholino resistant (Rtf1 wt) into rtf1 morphant embryos was able to robustly rescue the knockdown phenotype and support normal formation of the embryonic heart tube (Figure 7C). This knockdown-rescue assay provides an in vivo platform to test whether Rtf1’s HMD and Plus3 domains are required for the formation of cardiac progenitors during development. We generated expression constructs for mutant versions of Rtf1 lacking these functional domains and tested whether they could support cardiogenesis in an rtf1 morphant background. Mutant FLAG-tagged Rtf1 proteins lacking the HMD or Plus3 domain were expressed at levels comparable to Rtf1 wt and properly localized to the nucleus demonstrating that the activity of these proteins was not compromised by instability or mislocalization (Fig 7B). Interestingly, we found that Rtf1’s HMD was dispensable for its role in cardiac progenitor formation (100% of Rtf1 wt injected rtf1 morphants with nkx2.5 signal, n=25 vs. 100% of Rtf1 ΔHMD rtf1 morphants with nkx2.5 signal, n=25) (Figure 7C). On the other hand, Rtf1’s Plus3 domain was essential for supporting the formation of cardiac progenitors (0% of Rtf1 ΔPlus3 rtf1 morphants with nkx2.5 signal, n=25) (Figure 7C). These data suggest that Rtf1’s domains function in a modular fashion for regulating cardiac gene expression and point to potential specific effects on transcriptional pausing and elongation in cardiogenesis.

Figure 7.

Rtf1 deficiency disrupts promoter-proximal transcriptional pausing

Given the critical role of Rtf1’s Plus3 domain in cardiac progenitor formation, we hypothesized that Rtf1 deficiency may impact the transcriptional activity of RNA Pol II at cardiac genes. We then examined the occupancy of RNA Pol II by performing ChIP-seq on embryos at the 10-12 somite stage, when cardiac progenitors are arising from the lateral plate mesoderm. We compared the occupancy of Pol II between uninjected control and rtf1 morphant embryos and found that Pol II was substantially diminished at the transcription start site (TSS) of many genes compared to the remainder of the gene body (Figure 8D-I and data not shown). Pausing of RNA Pol II at the TSS is a normal aspect of transcription in vertebrates and has been hypothesized to regulate both the timing and expression level of genes^35,36. To quantify transcriptional pausing at each gene, we calculated the Pause Release Ratio (PRR), a ratio of Pol II occupancy in the promoter proximal region of a gene (−300 to +300) to the occupancy in the downstream region (+301 to +1301). A higher PRR indicates increased pausing of Pol II near the transcription start site (TSS), while a lower PRR signifies that a gene is in a less paused state. We found that on a global scale, pausing is diminished upon loss of Rtf1 activity during embryonic development (Figure 8A,B). Out of 6,078 genes with substantial Pol II ChIP signal that were analyzed, 5,799 displayed a significantly decreased PRR (Benjamini–Hochberg false discovery rate of 10%) in Rtf1-deficient embryos compared to 37 genes that displayed a significantly increased PRR (Figure 8B). The median PRR in control embryos was 3.34, while 95% (5,793/6,078) of genes had a PRR of less than 3.34 in Rtf1-deficient embryos (Figure 8A), reflecting a widespread decrease in promoter proximally paused Pol II. These changes in PRR suggest that Rtf1 is critical for normal levels of pausing at transcriptionally active genes during the developmental window when cardiac progenitor cells arise from the lateral plate mesoderm.

Figure 8.

Attenuation of pause release permits cardiogenesis in an Rtf1-deficient background

The transition between the pausing and elongation phases of transcription is regulated by PTEF-b, a complex consisting of a cyclin (T1, T2a, or T2b) and the kinase Cdk9^37–39. Phosphorylation of the negative elongation factor (NELF) complex, the pausing/elongation factor Spt5, and the C-terminal tail of RNA Pol II by Cdk9 promotes changes in the factors associated with RNA Pol II and triggers a switch to processive elongation⁴⁰. Pharmacological inhibition of Cdk9 with the small molecule flavopiridol can prevent or attenuate pause release, leading to genome-wide increases in pausing at most genes⁴¹. We explored the effects of Cdk9 inhibition on Rtf1-dependent transcription by performing ChIP-seq for RNA Pol II in flavopiridol-treated rtf1 morphant embryos at the 10-12 somite stage. Flavopiridol treatment caused a small but highly significant increase in promoter-proximal pausing in an Rtf1 deficient background (Fig 8A), with 87.1% of genes with significantly altered PRR values displaying an increase in PRR in flavopiridol-treated rtf1 morphants compared to untreated rtf1 morphants (Figure 8C). Furthermore, we found that critical LPM and cardiac genes including hand2, gata5, aplnrb, bmp4, nkx2.7, and rbfox1l exhibited reduced promoter-proximal pausing in Rtf1 deficient embryos that was partially rescued by treatment with flavopiridol (Figure 8D-I).

If diminished promoter-proximal pausing at genes crucial for cardiogenesis is responsible for the absence of cardiac progenitor formation in Rtf1-deficient embryos, then we would predict that reducing pause release may restore cardiogenesis in an Rtf1-deficient background. We examined expression of the cardiomyocyte marker myl7 in wild type, rtf1 mutant, and flavopiridol-treated rtf1 mutant embryos and found that partial attenuation of Cdk9 activity (5 ug/ml dosage) potently restored cardiomyocyte formation in Rtf1-deficient embryos (Figure 8J,K). Similarly, we found that knockdown of Cdk9 by injection of antisense morpholino oligonucleotides^42,43 also restored the formation of myl7+ cardiomyocytes in an Rtf1-deficient background (Figure 8L,M), confirming that decreased activity of the Cdk9 kinase is responsible for rescuing the rtf1 cardiac phenotype. These dramatic results suggest that the primary role of Rtf1 in cardiac progenitor formation is regulating promoter-proximal pausing of critical members of the cardiac gene program. Altogether, our data support a model in which loss of Rtf1 causes excessive pause release, a dysregulation of the cardiac transcription program, and an arrest in the developmental trajectory of cardiac progenitors prior to the formation of mature cardiomyocytes.

Discussion

In this study we investigated the molecular mechanism of the PAF1C member Rtf1’s role in cardiac progenitor formation. We found that Rtf1 plays a critical and conserved role in cardiogenesis and cardiac gene expression in vertebrate and mammalian models. Zebrafish rtf1 mutants and mouse Rtf1 cardiac mesoderm Rtf1 conditional knockouts do not properly initiate expression of cardiac transcription factors in the LPM and fail to generate mature cardiomyocytes. Using single cell sequencing, we found that Rtf1 deficient embryos do generate immature cardiac progenitors, but that these progenitors fail to express genes associated with progression to a more mature state. Excitingly, we also showed that the Rtf1 protein can function in a modular fashion and that its Plus3 domain, a site of interaction with the pausing/elongation factor Spt5, is essential for supporting cardiac development. In line with this finding, we showed that loss of Rtf1 diminishes promoter-proximal pausing of RNA Polymerase II, and that partially restoring normal pausing by chemical inhibition of pause release can support cardiogenesis in an Rtf1-deficient background.

Despite having a nearly genome-wide decrease in promoter-proximal pausing of RNA Polymerase II, Rtf1-deficient embryos display many upregulated and downregulated genes. How exactly pausing regulates gene expression is not fully understood and likely involves both general and gene/locus-specific mechanisms^35,44. By limiting the release of RNA Polymerase II into the elongation phase of transcription, pausing can restrain gene expression. This inhibitory role of pausing on transcription allows for another layer of transcription regulation and has been shown to modulate the expression of heat shock and immune stimulus-induced genes^45–47.

Pausing-dependent restraint of gene expression is also important for the timing of gene expression in a developmental context. Supporting this notion, Spt5 and NELF-deficient zebrafish embryos exhibit aberrant upregulation of TGFβ signaling and a resulting inhibition of hematopoiesis⁴³. Intriguingly, the PAF1C has also been shown in mammalian cells to restrain transcription from a subset of promoters by inhibiting nearby enhancer activation⁴⁸.

On the other hand, pausing also positively regulates gene expression in some contexts. At genes where nucleosome formation is favored in promoter regions, pausing of RNA Polymerase II can promote a chromatin architecture that is permissive to transcription by competing with nucleosomes for promoter occupancy^49,50. This mechanism has been suggested to support expression of the IFN-gamma pathway during zebrafish development⁴³. Pausing may also positively regulate gene expression during neural crest development. Loss of Paf1 activity disrupts neural crest differentiation, but this defect can be partially rescued by inactivation of Cdk9. Interestingly, zygotic loss of Cdk9 alone is sufficient to expand the neural crest population in zebrafish embryos, suggesting that promoter-proximal pausing promotes the neural crest gene program⁵¹.

Recent work using a rapid protein degradation system in human cell culture has shed light on the mechanisms by which the PAF1C regulates transcriptional pausing and gene expression outputs⁵². In DLD-1 cells, acute depletion of PAF1 destabilizes the PAF1C and stimulates pause release. Loss of PAF1 also decreases chromatin recruitment of INTS11, a component of the Integrator-PP2A complex (INTAC) which interacts with PAF1C and counteracts the activity of P-TEFb. This disruption in the balance of INTAC and P-TEFb activities at transcribed genes results in hyperphosphorylation of SPT5 and likely explains the increase in pause release caused by PAF1 depletion. Similar to our finding that Rtf1-deficiency positively and negatively regulates gene expression, depletion of PAF1 resulted in both upregulation and downregulation of many genes. Intriguingly, the gene expression changes caused by PAF1 depletion correlated with the release frequency and processivity of RNA Polymerase II at a given gene, with genes exhibiting low levels of release frequency tending to display decreases in their expression level. These observations suggest that the combined regulatory functions of pausing and elongation determine the transcriptional output of PAF1C-regulated genes.

In conclusion, our study reveals that the cardiac gene program is highly sensitive to loss of promoter-proximal pausing. Future studies focused on the chromatin architecture of cardiac gene promoters and other regulatory regions, and the dynamics of pause release and elongation at cardiac genes are needed to provide a complete picture of the manner by which promoter-proximal pausing fosters their expression.

Materials and Methods

Zebrafish

Zebrafish colonies were cared for and bred under standard conditions⁵³. Developmental stages of zebrafish embryos were determined based on somite number or hours of development post fertilization at 28.5 °C ⁵³. Fish husbandry and experiments were performed according to the Institutional Approval for Appropriate Care and Use of Laboratory Animals by the UCLA Institutional Animal Care and Use Committee (Protocols ARC-2000-051-TR-001 and BUA-2018-195-002-CR).

CRISPR mutagenesis

Template DNA for rtf1 exon 3 guideRNA (rtf1-e3 gRNA) synthesis was amplified with KOD polymerase (MilliporeSigma) using the pXTW-gRNA plasmid and the primers rtf1-e3-F (5’-TAATACGACTCACTATAGGAAGAAGGGGAAACCGAGCAAGTTTTAGAGCTAGAAATAGC-3’) and gRNA-R (5’-AAAAAAAGCACCGACTCGGTGCCAC-3’). rtf1-e3 gRNA (target sequence: 5’-GGAAGAAGGGGAAACCGAGCAA-3’) was synthesized using the MEGAshortscript T7 Transcription Kit (ThermoFisher Scientific) and purified using the RNeasy Mini Kit (Qiagen). Zebrafish embryos were injected at the 1-cell stage with 300 pg of Cas9 mRNA and 200 pg of rtf1-e3 gRNA. Injected embryos were raised to adulthood and then outcrossed to wild type AB fish to establish lines carrying specific rtf1 mutations.

Mice

All mice were maintained according to the Guide for the Care and Use of Laboratory Animals published by the US National Institute of Health. Housing and experiments were performed according to the Institutional Approval for Appropriate Care and Use of Laboratory Animals by the UCLA Institutional Animal Care and Use Committee (Protocols ARC-2020-045-TR-001 and ARC-2020-043-TR-001). ES cells were obtained from KOMP and used to rederive the Rtf1^{tm1a(KOMP)Wtsi} mouse line. This line was then crossed to a ROSA26::FLPe strain⁵⁴ (The Jackson Laboratory, JAX stock #009086) to create an Rtf1 conditional knockout allele where exon 3 of Rtf1 is flanked by two loxP sites (Rtf1^flox). These mice were crossed into the Mesp1^tm2(cre)Ysa background²⁸ to produce a strain capable of generating Mesp-Cre⁺;Rtf1^flox/^flox offspring.

Whole mount in situ hybridization

Whole-mount in situ hybridization was performed as described previously⁵⁵. Embryos for in situ hybridization were fixed in 4% paraformaldehyde in 1xPBS. After a brief proteinase K digestion, embryos were incubated with digoxigenin-labeled antisense RNA probes overnight at 65-70°C. Probes were detected using an anti-digoxigenin antibody conjugated with alkaline phosphatase (Roche 11093274910). The zebrafish antisense RNA probes used in this study include myl7, nkx2.5, mef2ca, tbx20, and tbx5a. The mouse probes used include MLC2v, Tbx20, and Nkx2.5.

Genotypes of zebrafish embryos were confirmed by PCR with primers:

rtf1-e3-F: 5’-TACAGTGACATCCACCATGCTG-3’

rtf1-e3-R: 5’-GTAAATAGGGGAGAAACTAATTGAAG-3’.

Genotypes of mouse embryos were confirmed by PCR with primers:

Rtf1-loxP-F: 5’-GACTGAGAAGGGAAATCTTTGAAAC-3’

Rtf1-loxP-R: 5’-CTAGTCCTTGAACAAGTTCAGCT-3’

Cre-F: 5’-GTAGCTGATGATCCGAATAACTAC-3’

Cre-R: 5’-ATCCAGGTTACGGATATAGT-3’.

Constructs

To prevent complementarity between rtf1 morpholino and mRNA sequences, PCR was used to introduce 8 silent mutations to the morpholino target sequence using pCS2/3XFLAG-rtf1 as a template²¹. The morpholino-resistant wild type rtf1 sequence was cloned into pCS2/3XFLAG to produce an N-terminally FLAG-tagged morpholino-resistant rtf1 construct (pCS2/3XFLAG-rtf1-mr). Rtf1 deletion mutant sequences were amplified from pCS2/3XFLAG-rtf1-mr by SOE-PCR to produce Rtf1ΔHMD (missing a.a. 126-211 of Rtf1) and Rtf1ΔPlus3 (missing a.a. 319-450 of Rtf1), which were subsequently cloned into pCS2/3XFLAG to produce N-terminally tagged constructs.

Microinjections

0.5 ng of rtf1 morpholino (rtf1MO; 5’-CTTTCCGTTTCTTTACATTCACCAT-3’) was injected at the 1-cell stage^21,56 along with 1 ng of p53 morpholino (5’-GCGCCATTGCTTTGCAAGAATTG-3’) to prevent p53-dependent apoptosis⁵⁷. For cdk9 knockdown rescue experiments, 5 ng of cdk9 morpholino (cdk9MO; 5’-ACACACAAACATCAAATACTCACCC-3’)^42,43 was injected into an incross of rtf1^LA2679 heterozygotes at the 1-cell stage. 150-200 pg of rtf1 or rtf1 deletion mutant mRNAs were co-injected with rtf1 and p53 morpholinos at the 1-cell stage for mRNA rescue experiments.

Mouse ESC culture and differentiation

E14TG2a mESCs (ATCC) were cultured on 0.1% gelatin covered plates without feeder cells. The feeder-free mESCs were maintained in Dulbecco’s modified Eagle’s medium (Invitrogen) supplemented with 15% fetal bovine serum (FBS; Invitrogen, USA), 1% non-essential amino acids (Invitrogen), 0.1 mM β-mercaptoethanol (Sigma-Aldrich), 1% (v/v) penicillin-streptomycin-Glutamine (ThermoFisher Scientific), and 10 μg/μl recombinant mouse leukemia inhibitory factor (LIF; Sigma-Aldrich) in a 5% CO2 incubator at 37°C. The ESC culture medium was changed every two days. mESCs were subcultured every 3 days using 0.25% trypsin-EDTA (Invitrogen). mESCs were differentiated in hanging drops (600 cells/20 µL) without LIF under non-adherent conditions. Day 5 embryoid bodies (EBs) were plated onto 0.1% gelatin covered plates. Media was changed every two days during adherent culture.

Lentivirus transduction

To knockdown Rtf1 activity in mESCs, Rtf1 or NT (non-target) shRNA lentivirus harboring puromycin resistance were purchased from Sigma-Aldrich. mESCs were transduced with the indicated virus at a multiplicity of infection of 1 and selected in 1.2 μg/ml puromycin. Rtf1 shRNA knockdown efficacy was tested by Western blotting analysis that detected Rtf1 protein level. Antibodies used were anti-Rtf1 (Bethyl Labs A301-329A) and anti-β-actin (Sigma A1978). Relative protein levels were determined by densitometry measurements using Adobe Photoshop.

Quantitative RT-PCR

Total RNA from NT and shRNA transduced EBs was extracted using TRIzol RNA isolation reagents (ThermoFisher Scientific). cDNA was synthesized using iScript cDNA Synthesis Kit from Bio-Rad. Real-time quantitative PCR was carried out using the Roche LightCycler 480 Real-Time PCR System. Primers for qPCR are listed in Supplementary Table 1.

Antibody staining

Embryos injected with mRNA encoding FLAG-tagged zebrafish Rtf1 constructs were fixed in 4% PFA in PBS at 75% epiboly. The fixed embryos were incubated in primary antibody (1:50 mouse anti-FLAG M2, F1804, Sigma) in blocking solution (10% goat serum in PBT) for 2 hours at room temperature followed by detection with fluorescent secondary antibody (1:200 anti-mouse IgG-Alexa Fluor 555, A-31570). Nuclei were stained with DAPI and embryos were embedded in 1% low-melt agarose and imaged on a Zeiss LSM800 confocal microscope.

Western Blotting

Embryos were lysed in Rubinfeld’s Lysis Buffer at 1 day post fertilization as previously described²¹. Two embryo equivalents were loaded per lane of a 8% polyacrylamide gel. Following electrophoresis, proteins were transferred to a nitrocellulose membrane and detected with antibodies against Rtf1 (Bethyl Labs A301-329A, 1:4000) and beta-actin (Sigma A1978, 1:5000). ImageJ was used to perform densitometry comparing the levels of Rtf1 protein between samples using beta-actin as the control.

Chemical treatment

Flavopiridol (Cayman Chemicals) was dissolved in DMSO at a stock concentration of 10 mg/ml. Zebrafish embryos were treated with 1.375 – 5 ug/ml flavopiridol diluted in E3 buffer beginning at 60-70% epiboly. For rtf1 mutant rescue experiments, genotypes were confirmed using the primers rtf1-e3-F and rtf1-e3-R.

Facs

10-12 somite stage TgBAC(hand2:EGFP)^pd24 uninjected and rtf1 morpholino-injected zebrafish embryos⁵⁸ were nutated in 10 mg/ml protease (Protease from Bacillus licheniformis, Sigma P5380) in DMEM/F12 media for 30 minutes at 4°C and pipetted to dissociate cells. Debris was removed by filtering through a 40 um cell strainer and cells were then pelleted by centrifugation at 300 rcf for 5 minutes at 4°C. Cells were resuspended in cold FACS buffer (1xPBS, 2% FBS, 1mM EDTA) and refiltered immediately prior to sorting on a BD FACSAria II using a 100 um nozzle. Dead cells were excluded by staining with 0.1 µg/ml DAPI. Cells from batches of 100 embryos were sorted directly into a mixture of 350 ul Buffer RA1 and 3.5 ul 2-mercaptoethanol from the NucleoSpin RNA kit (Machery-Nagel) for RNA isolation.

mRNA-seq

RNA was purified from FACS-sorted cells with the NucleoSpin RNA kit (Machery Nagel). Approximately 5-10 ng of input RNA was used to produce libraries with the NEBNext Single Cell/Low Input RNA Library Prep Kit for Illumina (NEB) which were sequenced on an Illumina HiSeq 3000 to produce 50 bp single-end reads. Reads from triplicate samples were mapped to the danRer7 genome using Tophat v2.1.1⁵⁹ using default parameters. Mapped reads were counted with FeatureCounts v2.0.0⁶⁰ and differential expression analysis was carried out with DESeq2 v1.32.0⁶¹. Gene ontology enrichment was examined using the R package clusterProfiler⁶² in R v4.1.0. For gene ontology analysis, the most significantly downregulated genes in rtf1 morphant LPM were selected as those genes that were downregulated among the 300 most significantly differentially expressed genes based on adjusted p-value.

ChIP-seq

Embryos for ChIP were dissociated with 4 mg/ml Collagenase Type IV and 0.25% Trypsin at room temperature with intermittent pipetting. Dissociation was stopped with 10% FBS in DMEM and cells were washed once with PBS. Fixation and preparation of chromatin was carried out using the Covaris truChIP Chromatin Shearing Kit and 7 minutes of sonication in a Covaris E220 ultrasonicator using the recommended settings. Immunoprecipitations were carried out as previously described⁶³ on approximately 5 ug of chromatin (minimum 2.5 ug, maximum 8.8 ug) using 5 ul of total RNA Pol II antibody (Cell Signaling D8L4Y). Libraries were prepared using the NEBNext Ultra II DNA Library Kit for Illumina (NEB) and sequenced with an Illumina Novaseq 6000 system to produce 50 bp paired-end reads. Reads from triplicate samples were mapped to the danRer11 genome using bowtie2 v2.4.2⁶⁴ (options: --local -X 1000) and read densities of UCSC RefSeq gene regions were quantified using the package csaw v1.24.3^65,66 in R v4.0.3. For our pausing analyses, we defined the transcription start site (TSS) region as −300 bp to +300 bp surrounding the TSS. RNA Pol II ChIP-seq read densities for the TSS regions of genes displayed a bimodal distribution which was assumed to be two partially overlapping distributions of genes with high and low RNA Pol II TSS signal. We examined those genes with a substantial level of RNA Pol II signal in the TSS region of control samples, defined as those genes present in the upper mode of the distribution of log-transformed TSS read densities based on a cutoff for type-I error of 10% using the R cutoff package v0.1.0 (https://github.com/choisy/cutoff). The read density in the TSS region was divided by the read density in the region extending from +301 to +1301 bp with respect to the TSS to calculate the pause release ratio (PRR). ChIP-seq data was visualized in Python 3.10.0 using the plotting tool SparK v2.6.2⁶⁷.

Nuclei isolation for single cell sequencing

Nuclei were isolated from 200 11-12 somite stage zebrafish embryos for 10x Genomics Multiome sequencing using recommended procedures with slight modifications. Embryos were deyolked in Deyolking buffer (55 mM NaCl, 1.8 mM KCl, 1.25 mM NaHCO3) and then embryonic cells were partially dissociated by pipetting in DMEM/F12 media (Gibco 21041025). Cells were pelleted and then lysed on ice for 5 minutes in 100 µl of chilled 0.1X Lysis buffer (10 mM Tris pH 7.4, 10 mM NaCl, 3 mM MgCl₂, 0.01% Tween-20, 0.01% IGEPAL CA-630, 0.001% digitonin, 1% BSA, 1 mM DTT). 1 ml of Wash buffer (10 mM Tris pH 7.4, 10 mM NaCl, 3 mM MgCl₂, 0.1% Tween-20, 1% BSA, 1 mM DTT, 0.1 U/µl Sigma Protector RNase Inhibitor) was then added to the cell lysis and the nuclei were pelleted by centrifuging at 500xg for 5 minutes at 4°C. Two more washes with 1 ml of Wash buffer were performed. Nuclei were then resuspended in 50 µl of Nuclei Dilution Buffer (1X 10x Genomics Single Cell Multiome Nuclei Buffer, 1 mM DTT, 1 U/µl Sigma Protector RNase Inhibitor) and filtered with a 40 µm Flowmi cell strainer. Approximately 10,000 nuclei per sample were targeted for capture with the 10x Chromium Single Cell Multiome ATAC + Gene Expression system by the UCLA Technology Center for Genomics & Bioinformatics.

Single cell sequencing analysis

10x Chromium Single Cell Multiome ATAC + Gene Expression libraries were prepared by the UCLA Technology Center for Genomics & Bioinformatics following the manufacturer’s recommended protocols and sequenced with an Illumina Novaseq 6000 system. Reads from RNA-seq and ATAC-seq assays were aligned to the zebrafish GRCz11 genome with the Lawson laboratory’s improved transcriptome annotation⁶⁸ using 10x Genomics’ Cell Ranger ARC software v2.0.0 with default parameters. Mapped reads from single cell RNA-seq and ATAC-seq assays were analyzed in R v4.2.2 using Seurat v4.3.0⁶⁹. Good quality cells were retained by filtering using Seurat with the following thresholds: 15000 > RNA UMI > 2000, Genes > 1500, 20000 > ATAC UMI > 5000, TSS enrichment score > 3.5. Filtered control and rtf1 morphant cell datasets were integrated using Harmony v0.1.1⁷⁰, and preliminary dimensional reduction, cell clustering, and marker gene identification were carried out in Seurat. Clusters of cells that exhibited enriched expression of mitochondrial and ribosomal transcript expression levels but lacked enrichment for other genes were considered abnormal cells and were manually removed from the dataset. Following manual filtering, a final dataset of 14340 control and 13839 rtf1 morphant cells was again integrated with Harmony and multimodal dimensional reduction (weighted-nearest neighbor method with RNA and ATAC assays), cell clustering, and marker gene identification were carried out in Seurat. Cluster cell number proportions and gene expression levels were visualized with Seurat and with published code⁷¹. Based on marker gene expression, cells that were members of clusters corresponding to anterior lateral plate mesoderm (sema3e+), endothelium (flt4+), cardiac precursor (mef2cb+), and rostral blood (spi1b+) were extracted, independently integrated with Harmony, and dimensional reduction (RNA assay only), cell clustering, and marker gene identification were carried out in Seurat. Pseudotime analysis of Harmony/Seurat cell embeddings and plotting of gene expression over pseudotime was performed with Monocle v3.0⁷².

Statistical analysis

Statistical significance of gene expression changes in qPCR analyses was determined based on 2 (for Nkx2-5 and Afp expression) or 3 (for Myh6, Nppa, Brachyury, and Ncam1 expression) independent experiments using a two-tailed Student’s t-test with a p-value of less than 0.05 considered to be significant. Statistical significance of the difference between beating cardiomyocyte cluster formation in control and Rtf1 shRNA mESCs was determined based on 4 independent experiments using a two-tailed Student’s t-test with a p-value of less than 0.05 considered to be significant. To compare PRR means between uninjected, rtf1 morphant, and flavopiridol-treated rtf1 morphant embryos, mean PRR values were first log-transformed (log10) and then a Welch’s paired two-tailed t-test was carried out with a p-value of less than 0.05 considered to be significant. Significantly changed PRRs for individual genes were determined based on 3 independent replicates per condition using the rowttests function in the R package genefilter v1.72.1 to perform two-tailed t-tests. The p.adjust function in the R stats package v4.0.3 using a Benjamini and Hochberg (false discovery rate) correction for multiple comparisons was then applied with adjusted p-values of less than 0.1 considered to be significant.

Acknowledgements

We thank the UCLA Technology Center for Genomics & Bioinformatics for technical assistance with Next Generation Sequencing and the Broad Stem Cell Research Center Flow Cytometry Core Facility for technical assistance with cell sorting. Figures 3A, 4A, and 5C were created with BioRender.com. This work was supported by funding from the NIH/NHLBI to J.-N.C. (R01HL155905 and R01HL140472).

Author Contributions

A.D.L.: study design, data acquisition, analysis, and interpretation, manuscript writing. F.L.: study design, data acquisition, analysis, and interpretation. L.T.: data acquisition. Z.Y.S.H.: data acquisition. B.K.: data acquisition. Z.T.: data acquisition. A.E.: data acquisition. M.P.: data analysis. H.N.: study design and data interpretation. A.N.: study design and data interpretation. J.-N.C.: study design, data interpretation, manuscript writing. All authors approved the final manuscript.

Competing Interests

The authors declare no competing interests.

Data Availability Statement

The RNA-seq, ChIP-seq, and single cell-seq data presented in this study are available in the NCBI SRA (BioProject PRJNA1015262).