Translational control of furina by an RNA regulon is important for left-right patterning, heart morphogenesis and cardiac valve function

Cardiovascular disease is a leading cause of death and accounts for a quarter of all deaths in the UK (Cheema et al., 2022). Even though cardiovascular development has been extensively studied, the molecular mechanisms underlying congenital heart disease are not fully known. The importance of RNA-binding proteins (RBPs) and RNA motifs in regulating organ development and diseases has been recently uncovered and understanding their interactions is fundamental to developing future diagnostic and therapeutic strategies (Brinegar and Cooper, 2016; Kalsotra et al., 2008; Sephton et al., 2010).

Zebrafish is a popular research model for studying heart development because the embryos are transparent, allowing in-depth observation of formation of many internal organs using live microscopy until larval stages. Additionally, although the heart in zebrafish embryos develops by 24 h post-fertilisation (hpf), embryos can survive the first week of life without a circulatory system. Zebrafish heart development is a dynamic process: at 24 hpf, a primitive beating organ is already formed, and by 48 hpf looping of the chambers and atrioventricular (AV) canal formation is complete (Liu and Stainier, 2012). Although heart formation occurs rapidly, many complex processes must be completed for the functional organ to develop. One key process is the establishment of left-right (LR) asymmetry, which is required for the correct placement of internal organs, including the heart (Bakkers, 2011).

In mammals, positioning of the heart on the left side and liver on the right is referred to as situs solitus. Disruptions to this process may cause left or right isomerism disorders that affect ∼1:10,000 individuals (Blum and Ott, 2018). Following heart looping, morphogenesis of the valves is initiated in the AV canal by formation of the endocardial cushions, which gradually differentiate and form functional heart valves (Gunawan et al., 2019). At similar stages, zebrafish embryos are small enough to obtain the required oxygen to live by passive diffusion. Therefore, unlike other model organisms, many mutations in the zebrafish heart do not cause embryonic lethality within the first week of zebrafish life (Liu and Stainier, 2012). This is especially useful for studying key processes in cardiac differentiation, and cardiac valve and septal development, which if disrupted in mammals result in foetal death.

The mechanisms that control LR asymmetry are well-conserved in vertebrates (Hamada et al., 2002). For instance, asymmetric left-sided expression of the secreted growth factor Nodal during somitogenesis in the lateral plate mesoderm (LPM) has been reported in many organisms including frogs, fish, chicks and mice (Levin et al., 1995; Long et al., 2003, 1996; Sampath et al., 1997, 1998; Shiratori and Hamada, 2006; Varlet and Robertson, 1997). Motile cilia generate calcium transients in the LR organiser, and drive leftward Nodal gene expression (Francescatto et al., 2010; Shinohara and Hamada, 2017). Expression of the Nodal inhibitor Lefty1 in the embryonic midline spatially restricts Nodal to the left LPM and results in asymmetric expression of the Nodal target Pitx2 (Chen and Shen, 2004; Lin et al., 1999; Lowe et al., 1996; Schottenfeld et al., 2007; Shiratori and Hamada, 2006; Thisse and Thisse, 1999).

The pro-protein convertase FurinA cleaves Nodal and is essential for heart development in many vertebrates (Beck et al., 2002; Constam and Robertson, 2000). Studies in zebrafish showed that FurinA is essential for maturation of the Nodal homologue Southpaw (Spaw) and establishes its signalling range in the LPM. Mutations disrupting zebrafish FurinA result in heart looping and trabeculation defects. The furina mutant embryos also display abnormal jaw development and 80% of maternal zygotic mutants die owing to lack of inflation of the swim bladder (Tessadori et al., 2015; Walker et al., 2006; Zhou et al., 2021). In the mouse, deletions in Furin cause embryonic lethality, and the mutants display a range of cardiac morphogenesis defects, including ventral closure, abnormalities of outflow tract and severe heart looping defects. Furin mutant mice also exhibit impairment of large vessel formation and defects in yolk sac vasculature (Dupays et al., 2019; Roebroek et al., 1998). Conditional deletion of Furin in endothelial cells in mouse did not reduce the severity of phenotype, and heart valve malformations in this allele also resulted in embryonic lethality (Kim et al., 2012), underscoring the importance of Furin during cardiac development.

Although the molecular mechanisms underlying transcriptional control of the Nodal pathway have been extensively studied, very few studies have explored regulation of the LR axis by non-coding regions and post-transcriptional regulation (Minegishi et al., 2021).We found that many Nodal pathway components in zebrafish harbour a short, conserved 3′ untranslated region (3′UTR) motif characterised by a AGCAC sequence motif, followed by a predicted stem and loop structure motif. The 3′UTR element was first identified in squint (sqt) [(nodal-related 1 (ndr1)] transcripts. A fluorescent sqt RNA reporter localises to dorsal progenitors in early zebrafish embryos (Gilligan et al., 2011; Gore et al., 2005; Kumari et al., 2013). We demonstrated that transcripts of the Nodal inhibitors lefty1 and lefty2 also contain the 3′UTR motif, and lefty reporter RNAs also localise in early embryos, indicating similar activities of the 3′UTR motifs (Zaucker et al., 2017). The conserved RBP Ybx1, binds the 3′UTR element in sqt, lefty1 and lefty2 and translationally represses these RNAs during early development (Kumari et al., 2013; Zaucker et al., 2017). Overexpression of full-length lefty1/2-gfp versus lefty1/2 lacking the 3′UTR motif showed that deletion of the element leads to a strong loss-of-nodal phenotype through inhibition of Nodal (Zaucker et al., 2017). Mutations in mammalian Ybx1 also led to de-repression of Nodal and Lefty in cultured cells, suggesting broader conservation of this mechanism.

Given that Ybx1 was found to translationally repress a range of different mRNAs through the 3′UTR motif, the motif was named ‘Ybx1 binding element’ (YBE) (Zaucker et al., 2019). By computational analysis of the zebrafish transcriptome, we now find that the YBE motif is present in a variant 3′UTR transcript (variant X1) of the Nodal maturation enzyme FurinA, but not in spaw (also known as ndr3). We show that the furina variant X1 harbours the YBE motif in its 3′UTR and is expressed prior to organ positioning in zebrafish. Disruptions of the furina 3′UTR element lead to LR patterning defects. Conditional ybx1 mutant embryos exhibit LR axis defects and abnormal visceral organ positioning. In addition, these ybx1 mutant embryos have an enlarged AV canal and show retrograde blood flow. This is similar to mitral valve regurgitation disorder, a condition in humans characterised by a leaky heart valve and back-flow of blood into the heart chambers. Our work shows that translational control by Ybx1 functions in LR axis formation, heart morphogenesis and valve development, and identifies a new zebrafish model for mitral valve regurgitation disorders.

Previous studies have demonstrated that FurinA proprotein convertase cleaves Nodal and facilitates establishment of LR asymmetry (Constam and Robertson, 2000; Tessadori et al., 2015). Through analysis of zebrafish genomic databases, we found that, in addition to the previously described furina transcript, a long furina variant X1 is also expressed during zebrafish embryogenesis. Variant X1 consists of an exceptionally large (∼4 kb) 3′UTR sequence that has been only computationally predicted. Analysis of available RNA-sequencing datasets shows that during somitogenesis, the furina variant X1 is expressed at levels higher than the previously identified short furina isoform (White et al., 2017; Fig. 1A,B). To determine whether variant X1 can be detected in embryos, we collected cDNA libraries from adult fish ovaries and 50% epiboly, 10-somite (10-som) and 18-somite (18-som) developmental stages and performed semi-quantitative PCR. In agreement with the RNA-sequencing data, we observed that the predicted variant X1 was not efficiently amplified in the ovary and at 50% epiboly. The highest expression levels of the variant were detected during somitogenesis (Fig. 1C). To quantify the relative expression levels of the transcript isoforms, we performed quantitative real-time PCR (qRT-PCR) with distinct primers to detect the coding sequences (CDSs) or variant X1 3′UTR, and normalised to 18S rRNA control (Fig. 1D,E). We did not detect a significant difference between the expression levels of the two transcripts in the ovary, at the 1 K cell stage or at 50% epiboly. Transcript X1 levels were high at 10-som and the isoforms were detected at similar levels by 18-som (Fig. 1E). Next, we examined whether there are differences in the expression pattern of the variant isoform in zebrafish embryos. Consistent with previous work (Tessadori et al., 2015), a furina coding region probe detected ubiquitous expression at all stages examined: 4-cell stage, bud, 10-som and 18-som. The expression pattern of variant X1 was similar to the Refseq-annotated transcript and no obvious differences were observed in lateral or dorsal views, although whole-mount in situ hybridisation (WISH) to detect the CDS gave a slightly stronger signal during early somitogenesis (Fig. 1F). At 10-som and 18-som, we noticed that there were areas where both transcripts were slightly enriched. Overall, we did not find any obvious difference in the spatial distribution of the two isoforms (Fig. 1F). Thus, both furina transcripts are ubiquitously expressed, throughout embryonic development, with a slight enrichment in the eyes, hindbrain, and somite structures.

We previously found that many Nodal pathway components harbour a 3′UTR element containing an AGCAC sequence motif, followed by a predicted stem and a variable length loop region (hairpin) (Zaucker et al., 2017). Computational analysis identified the furina variant X1 as a potential mRNA with a YBE element in the 3′UTR (Fig. 2A). The YBE-harbouring sqt RNA localises to one or two cells at the 4-cell stage (Gore et al., 2005; Gilligian et al., 2011) and the YBE in lefty1 and lefty2 mRNAs can localise exogenous reporter RNAs similar to sqt (Zaucker et al., 2017). If the furina YBE motif behaves similarly to the 3′UTR motif in these RNAs, they should colocalise in early zebrafish embryos. To test this hypothesis, furina variant X1 and lefty1 fluorescent mRNA reporters were co-injected into 1-cell-stage embryos and analysed for colocalisation at the 4-cell stage. The furina variant X1 fluorescent mRNA reporter colocalised with lefty1 in all injected embryos (n=16; Fig. 2B). To determine whether mutations affecting the 3′UTR YBE motif disrupt reporter mRNA localisation at the 4-cell stage, we generated deletion constructs that are predicted to change the YBE structure. The fluorescent reporter furina ΔYBE had the entire YBE motif deleted within the 3′UTR. furina Stem-break (SB) had a mutation disrupting base pairing within the stem of the YBE, and furina Stem-restore (SR) had a swapped sequence that restores base pairing within the stem of the hairpin. Injected fluorescent lacZ-βglobin mRNA served as a negative control and sqt RNA as a positive control. Embryos were injected with a fluorescent reporter and then scored for reporter RNA localisation at the 4-cell stage. lacZ-βglobin fluorescent mRNA reporter distribution was diffuse everywhere in the blastoderm, but, in contrast, sqt RNA localised asymmetrically in one or two cells in 95% of embryos (n=19) (Fig. 2C). furina variant X1 fluorescent reporter localised asymmetrically in 75% of embryos (n=17), whereas deletion of the YBE motif resulted in mislocalisation of the mRNA in 90% of embryos (n=18). Breaking of the stem caused 15% of SB reporter-injected embryos to show complete mislocalisation and 45% showed diffuse asymmetric distribution of the reporter. Injection with SR restored the localisation by 25% compared with SB (Fig. 2C). Thus, the furina X1 variant harbours a functional YBE motif.

The furina X1 UTR was also investigated by in vitro RNA structure probing (Fig. 2D,E). In the wild-type furina (furina WT), we observed a region of increased probe reactivity, flanked by regions of relatively low probe reactivity. Reactivity with the probe indicates high probability of single-strandedness, and this pattern approximately corresponded to the predicted stem-loop region in Fig. 2A. The region (blue dashed boxes) in the footprinting gel shows that there is a change in the pattern of bands from the wild-type pattern in the region encompassing the YBE region in the furina SB mutant, and this pattern is partially restored to the wild-type pattern in the furina SR mutant. The pattern noticeably flattened in furina SB, but was partially restored in furina SR and appeared more similar to the wild-type pattern (Fig. 2D). Quantification of the gel footprints (Fig. 2E) showed that the RNA structure outside of the YBE was largely unchanged in all RNAs tested, suggesting that structural mutations within the YBE only cause local changes in structure when probed in vitro. Comparing the quantified pattern of RNA structure by Pearson correlation (0.62 for WT versus SB and 0.84 for WT versus SR) indicated partial loss of the overall RNA structure in SB and partial restoration in SR. This partial loss and restoration of structure aligns with the in vivo reporter assays in Fig. 2C. Therefore, the furina X1 variant harbours a functional YBE motif with a structured RNA motif.

The RBP Ybx1 binds to the YBE motif in the 3′UTR of Nodal and Lefty mRNAs and forms a YBE/Ybx1 translational repression and RNA localisation module (Kumari et al., 2013; Zaucker et al., 2017). We hypothesised that the YBE we identified in the furina X1 variant RNA is likely to bind to Ybx1. To determine whether Ybx1 interacts with furina X1 mRNA in vivo, we performed RNA immunoprecipitation with zebrafish embryonic lysates from various stages (50% epiboly, 10-som and 18-som) using anti-Ybx1 antibodies, followed by qRT-PCR (Fig. 3). In western blots, we detected Ybx1 protein at 50% epiboly, 10-som and, to a lesser extent, at 18-som. Pull-downs with mouse IgG served as a negative control (Fig. 3B). Consistent with our published work, lefty1 was enriched at 50% epiboly and the 10-som stage. Using primers to distinguish between the furina coding sequence versus 3′UTR sequences, we found that at 50% epiboly and 10-som, there is enrichment of the furina long isoform over mouse IgG controls in pull-downs in both set of samples, whereas no significant difference in furina isoforms was detected in pull-downs from 18-som embryo lysates (Fig. 3C). Ribosomal 5S amplification served as a negative control in all cases, and lefty1 served as positive control. These results indicate that furina mRNA is in a complex with Ybx1 protein at 50% epiboly and at 10-som, but not at 18-som.

The YBE in nodal and lefty1/2 functions in translational control of these RNAs during early development. To test whether the furina variant X1 is similarly regulated by Ybx1, we performed our established translation assay that utilises ybx1^sa42 temperature-sensitive mutant fish to measure the expression of microinjected reporter constructs in ybx1 mutants by circumventing the early embryonic death phenotype of ybx1 null mutants (Kumari et al., 2013). The Ybx1^sa42 mutant protein is expressed in embryos at the permissive temperature, but Ybx1^sa42 protein levels are reduced in embryos at the restrictive temperature (Kumari, 2013). Cold-shock treatment during cleavage and blastula stages leads to developmental arrest during gastrulation and the embryonic phenotypes at 22°C are similar to ybx1 null mutants (Kumari et al., 2013). Therefore, to examine exogenous Furin reporter protein expression, we injected ybx1 mutant embryos at the 1-cell stage with furina-sfGFP:furina3′UTR RNA. Ybx1 protein function was disabled between 16-cell and 512-cell stages, as shown in the schematic, and embryos were analysed for protein expression between 1 K and epiboly stages (Fig. 4A,B). Control wild-type embryos that underwent the same cold-shock treatment or maternal ybx1 (Mybx1) mutant embryos kept at the permissive temperature of 28°C were used for comparison with mutant embryos. Analysis of the furina-GFP reporter showed a strong band in the Mybx1 lane, but not in control embryo or uninjected control lanes. We quantified the expression levels based on band intensity and normalised to Actin levels. Significantly higher FurinA GFP levels were detected in lysates from ybx1 mutant embryos (Fig. 4C,D).

As FurinA cleaves Spaw/Nodal to control the Nodal signalling range in the lateral plate mesoderm (Tessadori et al., 2015), we posited that overexpression of FurinA in ybx1 mutant embryos would lead to increased processing of Spaw and aberrant Nodal signalling in ybx1 mutant embryos. To test this, we injected spaw-GFP fusion mRNA into ybx1 mutants at the 1-cell stage, followed by temperature shift at the 16-cell stage, and collected embryonic lysates at the 500-cell stage (Fig. 4A,B). We detected two Spaw protein products, consistent with the existence of a Spaw precursor (75 kDa) and a mature Spaw (41 kDa) (Fig. 4C). Interestingly, in mutant lysates, we found elevated levels of Spaw mature peptide compared with controls. Quantification of band intensity of Spaw mature domain (n=3 experiments) and normalisation to Actin loading control (Fig. 4D) showed that, in the absence of Ybx1 function, FurinA protein levels are upregulated. Therefore, Ybx1 translationally represses furina variant X1 mRNA.

Mature Nodal/Southpaw is secreted into the extracellular space. We therefore examined Spaw maturation in ybx1 mutant embryos over time (Fig. 4E,F). We injected spaw GFP reporter at the 1-cell stage and analysed expression at the 1 K, sphere and 30% epiboly. At the 1 K, sphere and 30% epiboly stages, both intra- and extracellular GFP expression was detected by imaging the blastoderm of wild-type embryos (Fig. 4E,F). By contrast, in ybx1 mutant embryos, there was an obvious increase in accumulation of Spaw-GFP in the extracellular space (Fig. 4F). We measured total Spaw-GFP fluorescence levels across the blastoderm in wild-type and ybx1 mutant embryos and found that overall fluorescence levels were similar (Fig. S1). Therefore, even though overall Spaw-GFP levels were apparently similar in wild-type and mutant embryonic blastoderms, Spaw maturation by Furin is increased and faster in ybx1 mutants, resulting in excess and premature secreted mature Nodal protein in mutant embryos.

Increased processing of Spaw/Nodal and altered Nodal levels are associated with LR asymmetry defects (Long et al., 2003; Tessadori et al., 2015). Therefore, we next investigated whether ybx1 mutant zebrafish embryos show defects in LR asymmetry. WISH at 18-som showed a 15% increase in bilateral or right-sided spaw expression in ybx1 mutants (Fig. 5A,B, n=44). Abnormal spaw expression is associated with organ laterality defects, such as defective looping of the heart and mispositioning of the liver, pancreas and gut (Long et al., 2003). In ybx1 mutant embryos, we detected a 20% increase in an aberrant looping of the heart compared with both wild-type embryos at 22°C, and 28°C ybx1 controls (Fig. 5C,D). A proportion of the temperature-shifted ybx1 mutant embryos also showed no looping, or complete inversion of the heart chambers (Fig. 5C,D). Analysis of heart looping in ybx1 mutant embryos showed altered heart morphology in mutant embryos compared with wild-type controls. Further examination of bmp4 and notch1b expression in the heart showed that, in addition to the looping defects, ybx1 mutants also have a larger heart (n=8/24) and AV canal (n=4/29), respectively (Fig. S2).

To assess the positioning of visceral organs, such as liver, gut and pancreas, we examined expression of foxa3. In ybx1 mutants, we observed an increase in either duplication, mispositioning or inversion of the liver and pancreas (Fig. 5E,F). By contrast, the majority of control embryos showed correct organ positioning (Fig. 5E,F). Similarly, fabp10a expression revealed a similar trend: the majority of control embryos had a liver on the left-hand side and 20% of the ybx1 mutant embryos had either duplicated livers or inverted liver position (Fig. 5G,H, black arrowheads). Therefore, LR organ positioning is affected in ybx1 mutant embryos.

To determine whether heart morphogenesis is altered in the ybx1 mutant embryos, we performed immunofluorescence with the ZN8 antibody, which labels the AV canal, and quantified the AV canal width. At 55 hpf, we observed an overall increase in width of the AV canal compared with wild-type embryos. Additionally, measurements in wild-type controls were more consistent, whereas ybx1 mutant embryos presented a broad range in AV canal width (Fig. 6A-C). This raised the possibility that ybx1 mutant embryos harbour defects in heart valves and cardiac blood flow.

To determine the consequences of the larger AV canal in the mutants, we analysed the heartbeat of wild-type and MZybx1 mutant embryos by live imaging at 5 days post-fertilisation (dpf) (Fig. 6D,E). The movement of red blood cells was tracked, and the direction of blood flow during the contraction of the heart muscle was assessed. Live imaging at different stages of the heartbeat cycle and tracking of the red blood cells at 5 dpf revealed that in seven out of 13 MZybx1 mutant embryos red blood cells travelled back to the atrium, in contrast to wild-type controls, in which blood flowed from the atrium to the ventricle (Fig. 6E,F, Movies 1, 2). Quantification showed that over 50% of MZybx1 mutants exhibited retrograde blood flow (Fig. 6F), which is consistent with abnormal heart morphogenesis in these embryos.

Analysis of the ybx1 heartbeat revealed that during the ventricular systole, instead of blood cells being pushed out of the ventricle, a large number of red blood cells leak through the heart valves and move back into the atrium. To examine blood flow patterns in these embryos in depth, we analysed and quantified the movements of the red blood cells around the AV canal (Fig. 7A). We focused on two timepoints during ventricular contraction in ybx1 embryos with reverse blood flow (RF) versus wild-type controls. At the beginning of the ventricular systole, red blood cells moved in the same direction in wild-type and mutant embryos (Fig. 7B,D). However, at the end of the ventricular systole, in mutant embryos we observed that red blood cells change direction and start moving back, whereas in wild-type embryos the movement trajectory remained the same (Fig. 7C,E). Analysis of the velocity of the red blood cells during contraction of the ventricle confirmed that in ybx1 RF embryos there is increase in cell movement back into the atrium, whereas in control embryos little movement was detected in this direction (Fig. 7F). Furthermore, in ybx1 embryos, initially the cells moved out of the ventricle, but over time their velocity gradually decreased. In control embryos, however, movement of cells out of the ventricle remained steady and consistent (Fig. 7G). The altered movement of blood cells is consistent with our observations that ybx1 mutants have a larger AV canal (Fig. 6B,C), and heart valves appeared misaligned, resulting in reverse blood flow, showing a heart defect at 5 dpf. This shows the important role of translational control by Ybx1 protein, not only in LR asymmetry establishment during early embryogenesis in zebrafish, but also in correct heart morphogenesis and heart valve function.

To test directly the role of the YBE region in the furina X1 3′UTR, we performed CRISPR-Cas9 genome editing using guide RNAs targeting the YBE region of furina. Despite several attempts, we were unable to recover stable germline mutants in which the furina X1 3′UTR YBE region had been disrupted, whereas stable germline furina ATG region mutants similar to those described by Tessadori et al. (2015) were recovered (Nagorska, 2022). Therefore, we generated somatic furina X1 mutants (i.e. ‘crispants’; Burger et al., 2016) using Cas9 protein and a specific guide RNA targeting the furina ATG or a pair of gRNAs flanking the furina 3′UTR YBE region.

The morphology of these embryos was found indistinguishable from controls at prim5 (Fig. 8). WISH to detect spaw in the lateral plate mesoderm showed that, in contrast to control embryos that typically have left-sided expression of spaw (83%, n=41), a significant proportion of furina Δ3′UTR YBE embryos showed disruptions in LR asymmetry with bilateral or right-sided spaw (∼43%, n= 44; Fig. 8). By contrast, control embryos injected with Cas9 protein alone or Cas9 together with a control gRNA showed LR patterns similar to those of wild-type embryos. These results support a role for the furina 3′UTR and YBE in regulation of asymmetric spaw expression and LR patterning.

In this work, we describe a novel furina variant X1 mRNA with a YBE element in the 3′UTR. Similar to other nodal mRNAs, we show that furina YBE motif is recognised by the RBP Ybx1 and has an important role in translational repression of FurinA. Previous work by other groups showed that FurinA protein cleaves Spaw and establishes its signalling range (Tessadori et al., 2015). We observed that, in the absence of Ybx1, FurinA is premature and overexpressed and, as a result, Spaw processing is increased. This causes excess Spaw protein levels and ectopic Spaw expression outside the left lateral plate mesoderm. This leads to LR asymmetry defects, with abnormal heart looping and abnormal positioning of liver and pancreas. Our work thus identifies a previously unappreciated mechanism of regulation of LR asymmetry and cardiac morphogenesis through the RBP Ybx1. Although many transcription factors have been found to control LR patterning across animal species, very few RBPs and/or corresponding RNA elements have been identified in this process (Rago et al., 2019; Vejnar et al., 2019); therefore, our work could lead to potentially new clinical diagnostic tests for LR asymmetry defects that are currently unassigned.

RNA elements have a crucial regulatory role in mRNA biogenesis, stability, localisation and translation. The Hu RBP family in human neurons is a great example. The HuR, HuB, HuC and HuD (ELAVL1-4) proteins preferentially bind to ‘AU’ and ‘U’ rich sequences in 5′ and 3′UTRs of the RNAs. These interactions are key in regulating RNA stability and translation in changing cellular conditions (Lopez de Silanes et al., 2004). In early zebrafish embryos, 3′UTR elements and translational control play crucial roles in regulation and stability of maternal RNAs (Mishima and Tomari, 2016; Winata et al., 2021). Mammalian studies have also identified general motifs via CLIP-seq (cross-linking immunoprecipitation sequencing) that some RBPs bind to, for instance: UGAU, AAUAAA. Interestingly, the IGF2BP3 family of RBPs recognises CWWCATCA and TGCACTAT, to facilitate mRNA transport and enhance stability (Li et al., 2017). In this study, we experimentally validated the YBE motif in furina mRNA by in vitro RNA structure probing. Although we do not see a perfect recapitulation of the hypothesised 3 bp stem loop structure, we see some support for RNA structure in that region and note that our in vitro observations might be confounded by the lack of stabilising trans-acting factors that would normally be present in vivo. One possibility is that perhaps there is RNA structure heterogeneity in the region, whereby all these RNAs can generate a structure similar enough to that observed by structure probing, but with furina WT having the highest propensity and furina SB the lowest propensity for the predicted structure.

We show that Ybx1 interacts with furina mRNA at the embryonic stages prior to organ morphogenesis. The phenotype observed in ybx1 mutant embryos is likely due to a lack of translational control of furina mRNA and is consistent with overexpression of FurinA protein. When there is an excess of FurinA protein, cleavage of Spaw is upregulated and the midline of an embryo is saturated with Spaw, which causes bilateral expression and consequently LR asymmetry defects (Tessadori et al., 2015). To date, there are very few studies showing involvement of RBPs in LR asymmetry. One study described translational repression of Bicc1 on dand5 5′UTR mRNA and showed a role for this interaction in the establishment of laterality (Maerker et al., 2021). Dand5 is a negative inhibitor of Nodal signalling and probably acts at a different level of the Nodal pathway, downstream of furina and Ybx1 regulation. Interestingly, we observed the LR asymmetry defects in approximately 20% of the mutant embryos. Ybx1 is a global translational repressor and it also interacts with the Nodal inhibitors Lefty1 and Lefty2 (Sun et al., 2018; Zaucker et al., 2017). Although in the ybx1 temperature-sensitive allele the mutation was triggered in late epiboly stage, there is still upregulation of Lefty1 in ybx1 mutants. During somitogenesis, Lefty1 is expressed at the midline of the embryo preventing Spaw from crossing the midline barrier and ensuring its leftward expression (Lenhart et al., 2011). However, in ybx1 mutants, upregulation of Lefty1 activity likely represses elevated Spaw. This, as a result, might reduce the severity of the LR asymmetry phenotype in ybx1^sa42 embryos.

In addition to the LR organ-positioning defects, we also found that the AV canal in the ybx1 mutants is enlarged and, consequently, the embryos have retrograde blood flow. This phenotype could also potentially be independent of or in addition to the interaction of FurinA with Spaw. FurinA is a ubiquitously expressed protein that cleaves a range of targets, which are required for correct heart and AV canal development, such as Notch1b in zebrafish (Zhou et al., 2021). Mouse Furin is expressed in a range of cell types and loss of Furin activity in the extra-embryonic and cardiogenic mesoderm leads to yolk sac morphogenesis and heart looping defects (Constam and Robertson, 2000). Moreover, conditional loss of Furin activity in endothelial cells is associated with cardiac valve malformations and ventricular septal defects (Kim et al., 2012). These phenotypes seem similar to the cardiac morphogenesis defects in our ybx1 mutants with elevated Furin activity, and lend support to the possibility that Furin activity levels are tightly controlled in embryos and any misregulation might lead to defects in heart development. In Xenopus and mouse, Furin has been shown to interact with the bone morphogenic pathway (BMP) family proteins, which are also important for heart development (Constam and Robertson, 2000; Cui et al., 1998). The loss of translational control of furina in ybx1 mutants might also disrupt interactions with one of these targets and lead to additional heart defects. It is also possible that Ybx1 might regulate heart morphogenesis through other Nodal-independent mechanisms (Noël et al., 2013).

When we made deletions within the furina 3′UTR by CRISPR/Cas mutagenesis using a pair of gRNAs flanking the YBE, we could not recover stable germline mutants. Nonetheless, in somatic mutant embryos in which we disrupted the furina X1 3′UTR, we find increased LR patterning defects compared with wild-type or control-injected embryos. Together with the LR patterning, heart morphogenesis and valve defects in ybx1 mutants, these results support a role for the 3′UTR/Ybx1 regulon in translational control of furina and heart development.

There are reports of regulation of the heart development by non-coding molecules, such as microRNAs (miRNAs). Key cardiac development genes such as tbx2, notch1 and cspg2 are regulated by miRNAs in zebrafish (Kalayinia et al., 2021). Studies in mice have shown that increased levels of miR-25 lead to depressed cardiac function, but inhibition of miR-25 reversed the phenotype and increased survival in a mouse heart failure model (Wahlquist et al., 2014). Studies in humans have shown that reduced amounts of miR-26a, miR-30b and miR-195 in stenotic valves led to increased chance of developing valve leaflet fusion. These miRNAs are thought to be involved in modulation of calcification genes; when their levels are lower, calcium begins to accumulate at the aortic valve leaflet and changes the morphology (Nigam et al., 2010).

We find that ybx1 mutant embryos have retrograde blood flow at 5 dpf. Retrograde blood flow can be observed normally up to day four in larvae, and is important for modelling of the AV canal and valve development, and mutations in the heart valve precursor gene klf2a cause aberrant valve development (Vermot et al., 2009). In ybx1 mutants, we observe that red blood cells squeeze through the valves back to the atrium. This is similar to a human condition called mitral valve regurgitation, which is characterised by a leaky valve leading to backflow of blood. The extent of the symptoms varies, with some people being asymptomatic and others experiencing dizziness, breathlessness and irregular heartbeat (Adams et al., 2010). The expansion of the atrioventricular canal likely prevents heart valves from shutting properly, and during ventricular contraction the blood cells squeeze between the valves and go back to the atrium of mutant hearts. Therefore, our ybx1 mutant could serve as a zebrafish model for this disorder and potentially lead to better understanding and future treatments of heart valve regurgitation disorders.

Zebrafish were maintained in accordance with the University of Warwick institutional animal welfare regulations and the UK Home Office guidelines. All fish in this work were maintained in the Tübingen (TU) background. Zebrafish ybx1^sa42 fish lines and controls were set up as single pair crosses separated with a divider the previous day; on the next morning, the divider was removed to allow mating. Maternal ybx1 embryos (Mybx1; i.e. progeny of homozygous ybx1 females mated with wild-type males) as well as maternal zygotic ybx1 mutants (MZybx1; progeny of homozygous ybx1 male and female matings) arrest during early gastrulation regardless of the genotype of the male, whereas paternal ybx1 (Pybx1; progeny of homozygous ybx1 males mated with wild-type females) develop normally (Kumari et al., 2013). We used both Mybx1 and MZybx1 embryos in the experiments and they are generally referred to as ybx1 mutants.

Embryos were kept at 28°C in 0.3× Danieau's solution [17 mM NaCl, 2 mM KCl, 0.12 mM MgSO₄, 1.8 mM, Ca(NO₃)₂, 1.5 mM HEPES, pH 7.6] until 50% epiboly, and afterwards they were shifted to 22°C until the 20-somite stage. Embryos were then shifted back to 28°C until 50 hpf (long-pec stage) for live imaging or fixed in 4% paraformaldehyde, dehydrated and stored in methanol at −20°C. For heart imaging, MZybx1 and wild-type control embryos were collected at the same time and grown at 28.5°C until the 75% epiboly stage. They were subjected to temperature shifts to 22°C until the 21-som stage and shifted back to 28°C until 5 dpf.

Zebrafish embryos from stages 1 K, 50% epiboly, 10-som, 18-som and adult ovaries were collected and 50 eggs/embryos were used per stage. RNA was initially homogenised in TRIzol reagent (Thermo Fisher Scientific) and then extracted using an RNA miniprep kit (NEB). The integrity of purified RNAs was checked on agarose gel and concentration was measured on the a NanoDrop spectrophotometer. Next, RNAs were reverse transcribed using SuperScript III (Thermo Fisher Scientific). For qRT-PCRs, each cDNA was amplified using gapdh, furina and furina X1 primers in a standard 20 μl PCR reaction [5× GoTaq Green buffer (Promega), dNTPs 10 mM (Promega) of 2 µm each forward and reverse primer, home-made Taq polymerase and sterile water]. For qPCR, we used the 2× Luna Universal qPCR Master Mix (NEB), following the manufacturer's protocol. All primers were tested for efficiency to ensure they were comparable. Normalisation to 18S rRNA expression was carried out for qRT-PCR. The primer list is available in Table S1.

Embryos were fixed at the required developmental stages overnight at 4°C in 4% paraformaldehyde in PBS and WISH was performed according to established protocols (Thisse and Thisse, 2008).

Plasmids were linearised using NotI restriction enzyme and mRNA was transcribed from purified template DNA using SP6 RNA polymerase. Transcription master mix contained 500 ng of linear purified template, 5× transcription buffer, DTT, ribonuclear capping mix (rATP, rGTP, rCTP; 10 mM each) rUTP (100 mM), ⁷mG(5′)pppG cap (7.5 mM) and Alexa 488-5-UTP (1 mM) or Alexa 546-UTP (1 mM) (Thermo Fisher Scientific) fluorophore. Following transcription samples were treated for 30 min with a Turbo DNase I enzyme and unincorporated nucleotides were removed using mini-spin columns (Bio-Rad). Next, phenol-chloroform purification was performed on the fluorescent RNAs, followed by overnight precipitation in isopropanol and elution in nuclease-free water. RNA integrity was checked on an agarose gel and the concentration and incorporation rate of the fluorophore was checked using a NanoDrop spectrophotometer. Embryos were injected at the 1-cell stage with 10-25 pg aliquots of fluorescent mRNAs, imaged and/or scored at 4-cell stage.

Zebrafish embryos were collected at following stages: 50% epiboly, 10-som and 18-som, in aliquots of 250 embryos. Embryos were cross-linked in 1% formaldehyde solution for 20 min, rinsed three times with PBS-T (PBS with 0.1% Tween 20) and lysed using RIPA buffer in DEPC water (50 mM Tris-Cl pH 7.5, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS, 1 mM EDTA, 1 M NaCl, 1 M urea, protease inhibitor). Immunoprecipitation was performed according to Kumari et al. (2013) and Zaucker et al.(2017), with an anti-Ybx1 antibody (Sigma-Aldrich, 4F12), followed by qRT-PCR. 5S rRNA expression served as a control for RT-PCR. A list of primers used in RT-PCRs and qPCRs is available in the Table S1.

Antibody staining was performed as described by Santos et al. (2018). MF20, S46 and ZN-8 antibodies were purchased from Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242, and used according to the manufacturer's recommendations.

Full-length spaw-GFP and furina sfGFP were cloned into pCS2 vector by Gibson assembly and transcribed using the SP6 mMachine kit (Thermo Fisher Scientific). Capped in vitro-synthesised RNAs were treated with turbo DNase, phenol-chloroform purified, precipitated overnight in isopropanol and eluted in nuclease-free water. Next, 50 pg of RNA was injected into 1-cell-stage maternal ybx1^sa42/sa42 mutant embryos, and 50% of the injected embryos were temperature shifted to 22°C at the 16-cell stage, and 50% of injected embryos were kept at 28°C as controls. At the 512-cell stage, all embryos were collected and lysed in RIPA buffer (50 mM Tris-Cl pH 7.5, 1% NP-40, 0.5% sodium deoxycholate, 0.05% SDS, 1 mM EDTA, 150 mM NaCl) with proteinase inhibitor (Roche). The sample concentration was normalised using a Bradford assay kit (NEB), following which all samples were boiled for 5 min in Laemmli buffer, loaded on a 10% acrylamide gel and western blotting was carried out. Membranes were incubated with anti-GFP HRP conjugated antibody (1:2000; Santa Cruz, sc-9996 HRP) and HRP-conjugated Actin loading control (1:4000; Santa Cruz Biotechnology, sc-47778).

Imaging of immunofluorescence and for reporter translation was carried out using an Andor Revolution Spinning Disk system, with a Yokogawa CSU-X1 spinning disc unit, fitted with a 488 nm laser and captured with an iXon Ultra 888 EMCCD camera. Control, mutant and injected embryos were dechorionated and mounted in a mould (Zaucker et al., 2021) and imaged with identical settings and exposure times. As Ybx1 acts as a general translational repressor in mammalian cells and zebrafish embryos (Evdokimova et al., 2001; Minich et al., 1993; Sun et al., 2018), normalisation of Spaw and Furin translation reporter-injected embryos was carried out with rhodamine dextran injection control as described previously (Zaucker et al., 2017). Images for z-stacks were acquired using Andor software and final processing and analysis performed in ImageJ.

MZybx1 and wild-type control embryos were collected at the same time and grown at 28.5°C until the 75% epiboly stage. They were subjected to temperature shifts to 22°C until the 21-som stage and shifted back to 28°C until 5 dpf. At 24 hpf, embryos were treated with the phenylthiourea at 0.003% concentration in Danieau's solution to prevent pigment formation. To distinguish between general LR and heart morphogenesis defects, embryos were sorted for heart looping and only embryos with normal looping were assessed for cardiac valve function. Prior to imaging on day 5, embryos were immobilised by adding 25× tricaine solution to Danieau's solution and mounted in 0.6% low melt agarose on a 3-cm glass cover slip-bottom Petri dish for imaging at 13 frames per second using Nikon ECLIPSE Ni microscope with a HAMAMATSU digital camera C11440, ORCA-Flash4.OLT. Time-lapse movies of approximately 30 s to 1 min duration were acquired.

To analyse the direction and velocity of the blood flow in ybx1 and control embryos, the particle image velocimetry (PIV) application was used in MATLAB (Thielicke and Sonntag, 2021). Each movie was converted in ImageJ using the function ‘find edges’ and uploaded to PIV, the region of interest was set out to be a junction in the AV canal and frames were analysed. Vectors were saved and analysed for direction with a custom MATLAB code.

Several gRNAs targeting the 3′UTR of the furina transcript were designed using CHOPCHOP v3 software (Thyme et al., 2016) and CRISPRon (Xiang et al., 2021; Anthon et al., 2022). Off-target analysis was carried out using CRISPRoff to ensure selection of gRNAs with minimal off-target potential, i.e. all gRNAs were selected with high predicted specificity (Alkan et al., 2018; Anthon et al., 2022). The selected gRNAs (Table S1) were synthesised by in vitro transcription of PCR product DNA templates with the gRNA sequence fused to a T7 promoter, following an oligo-based strategy (Varshney et al., 2015). Several gRNAs were tested for efficient cutting after co-injection with Cas9 mRNA using a T7 endonuclease (T7E1) assay and two efficient gRNAs spanning an ∼1.3 kb region that contains the YBE and located 60 nt after the stop codon were selected (Table S1). We designed a primer pair that amplifies a 188 bp PCR product from the predicted Δ3′UTR deletion mutant allele, primer F and primer R2. The predicted 188 bp deletion mutant sequence was successfully amplified from gDNA extracted from embryos that had been injected with the two gRNAs with Cas9 mRNA. The PCR product was purified and sequenced to confirm the sequence resulting by fusion of sequences adjacent to the gRNA target sites, with an ∼1.3 kb deletion in the intervening region. Sequences of the gRNAs and genotyping primers are provided in the primer list (Table S1).

For generation of furin Δ3′UTR crispants we injected 1-cell-stage embryos with 2 nl aliquots containing 1 μM each of the gRNAs and 1.33 μM Cas9 protein (NEB, M0646T). Injected embryos were kept at 28.5°C until 60% epiboly, incubated at 22°C overnight, and fixed at the 19- to 21-somite stage in 4% paraformaldehyde in PBS. In one experiment, 22 embryos were genotyped using primers F and R2 together with primer R1, which anneals to a sequence that is deleted in Δ3′UTR crispants. Using the three primers together in a single PCR reaction, we can detect wild-type and deletion mutant products in parallel. To rule out non-specific phenotypes arising from toxicity of the injected materials, we used multiple controls: non-injected wild-type embryos, 1 μM Cas9 protein only-injected embryos, and embryos injected with 1 μM Cas9 together with gRNA2 and the control non-cutting gRNAc (instead of gRNA1).

An unpaired, two-tailed Student's t-test was used for statistical analysis of heart looping, qPCR datasets and western blot band and fluorescence intensity measurements. Single-factor ANOVA was used to analyse the width of the atrioventricular canal at 3 dpf. Fisher's exact test was used to analyse significance of spaw expression in ybx1^sa42 mutant embryos and furina 3′ UTR crispants and foxa3 expression. For all statistical analyses, P-values are represented as follows: *P<0.05, **P<0.01, ***P<0.001; P>0.05 was considered non-significant.

We thank members of the K.S. lab for feedback and comments on the manuscript, the University of Warwick Biological Services Unit aquatic facility staff for fish husbandry and care, and the Warwick fish community and the Balasubramanian laboratory for support and advice.