Heg1 and Ccm1/2 proteins control endocardial mechanosensitivity during zebrafish valvulogenesis

Endothelial cells respond to different levels of fluid shear stress through adaptations of their mechanosensitivity. Currently, we lack a good understanding of how this contributes to sculpting of the cardiovascular system. Cerebral cavernous malformation (CCM) is an inherited vascular disease that occurs when a second somatic mutation causes a loss of CCM1/KRIT1, CCM2, or CCM3 proteins. Here, we demonstrate that zebrafish Krit1 regulates the formation of cardiac valves. Expression of heg1, which encodes a binding partner of Krit1, is positively regulated by blood-flow. In turn, Heg1 stabilizes levels of Krit1 protein, and both Heg1 and Krit1 dampen expression levels of klf2a, a major mechanosensitive gene. Conversely, loss of Krit1 results in increased expression of klf2a and notch1b throughout the endocardium and prevents cardiac valve leaflet formation. Hence, the correct balance of blood-flow-dependent induction and Krit1 protein-mediated repression of klf2a and notch1b ultimately shapes cardiac valve leaflet morphology.


Introduction
Biophysical forces including shear stress caused by blood-flow trigger activation of mechanotransduction pathways within endothelial cells (ECs), a process which causes cellular changes that contribute to the sculpting of vascular networks and of the heart (Baeyens and Schwartz, 2016;Haack and Abdelilah-Seyfried, 2016). Currently, our knowledge is incomplete regarding the modes by which the sensitivity of different vascular beds is attuned to the constantly changing stimulus of blood-flow. One well-studied example of a blood-flow-sensitive developmental process involves remodeling of endocardial cushions into cardiac valve leaflets in the zebrafish embryo Pestel et al., 2016;Scherz et al., 2008;Steed et al., 2016). Prior to the formation of cardiac cushions, the oscillatory pattern of blood-flow induces expression of the mechanotransduction pathway components zinc-finger transcription factors Krü ppel like factors 2a (klf2a) (Vermot et al., 2009) and klf2b (Renz et al., 2015). KLFs are important blood-flow-responsive genes within ECs that are strongly activated within regions of high shear stress in tissue culture and various in vivo models (Dekker et al., 2002(Dekker et al., , 2005Groenendijk et al., 2004Groenendijk et al., , 2005Lee et al., 2006;Novodvorsky and Chico, 2014;Parmar et al., 2005;Zhou et al., 2016). As zebrafish cardiac atrioventricular (AV) cushions are remodeled into AV valve leaflets, endocardial klf2a is highly expressed on the luminal side of the developing valve leaflet, which is exposed to blood-flow, whereas its expression is low on the abluminal side of the leaflet (Steed et al., 2016). Zebrafish klf2a mutants exhibit defective cardiac valve leaflets (Steed et al., 2016) and a loss of murine Klf2 also results in atrioventricular cushion defects (Chiplunkar et al., 2013). One of the targets of klf2a within cardiac cushions is notch1b (Vermot et al., 2009). The precise spatiotemporal regulation of Klf2 expression within cardiac valve leaflets indicates that fine-tuning the response to blood-flow has a conserved role in shaping functional leaflets during cardiac morphogenesis.
The control of Klf2 expression has been linked to the cerebral cavernous malformations protein complex (Maddaluno et al., 2013;Renz et al., 2015;Zhou et al., 2015Zhou et al., , 2016. Within this complex, KRIT1 (also known as CCM1) and the transmembrane protein HEG1 (also known as Heart of glass) physically interact with each other. This interaction results in stabilization of both proteins at EC junctions (Gingras et al., 2012;Kleaveland et al., 2009). The loss of HEG1, KRIT1, CCM2, or PDCD10 (also referred to as CCM3) causes severe cardiovascular and cardiac developmental defects (Kleaveland et al., 2009;Mably et al., 2003;Mably et al., 2006;Whitehead et al., 2009;Yoruk et al., 2012;Zheng et al., 2010). However, functional studies in zebrafish suggest that Ccm3 functions in a manner that is separate from Krit1 and Ccm2 (Yoruk et al., 2012). The name CCM refers to pathologies with familial inheritance that have been attributed to mutations in human KRIT1, CCM2, or PDCD10. Affected individuals exhibit morphological malformations of low-perfused venous endothelial beds of the neuro-vasculature that can result in dangerous cerebral bleeding [reviewed in Riant et al., 2010]. Similarly, the conditional endothelial-specific loss of Krit1, Ccm2, or Pdcd10 in postnatal mice causes CCM lesions (Boulday et al., 2011;Chan et al., 2011). Strikingly, postnatal loss of Heg1 in mice does not cause CCM lesions and familial forms of the CCM pathology have never been associated with mutations in HEG1 (Zheng et al., 2014). These findings suggest a developmental role for HEG1-CCM signaling which differs from postnatal functions of the other CCM proteins.
Functional studies in zebrafish ccm and murine Ccm mutants have linked cardiovascular developmental defects to elevated expression levels of Klf2 (Renz et al., 2015;Zhou et al., 2015;. Two major pathways have been implicated in regulation of KLF2 by the CCM proteins KRIT1 and CCM2. First, the core adaptor protein KRIT1 and its associated b1-Integrin binding protein 1 (ICAP1) suppress b1-Integrin signaling in human umbilical vein endothelial cells (HUVECs) (Faurobert et al., 2013). A loss either of CCM proteins or of ICAP1 enhances Integrin signaling and this causes strong upregulation of KLF2 expression in HUVECs, which is suppressed by the depletion of b1-Integrin (Renz et al., 2015). Similarly in zebrafish, ccm mutant cardiovascular defects are suppressed by a knockdown of b1-Integrin (Renz et al., 2015). Secondly, the binding of CCM2 to MEKK3 (also known as MAP3K3) (Cullere et al., 2015;Uhlik et al., 2003;Zhou et al., 2015) inhibits the MEKK3-MEK5-ERK5 mechanotransduction pathway, which controls expression levels of murine Klf2 (Zhou et al., 2015). These findings point to the importance of CCM proteins in regulating the physiological responsiveness of ECs to blood-flow. In support of such a physiological role, CCM proteins antagonize activation of b1-Integrin and thus interfere with the orientation of endothelial cells in response to shear stress (Macek Jilkova et al., 2014).
The developmental connection of Ccm proteins with the regulation of klf2a expression led us to investigate whether expression of zebrafish heg1 and krit1 within the vasculature is controlled in a mechanosensitive manner. We find that expression of heg1 is positively regulated in response to blood-flow and klf2a/b. We also explored the role of Heg1 and Krit1 in Klf2-dependent endothelial mechanotransduction during the formation of cardiac valve leaflets. Here, we show that Krit1 dampens levels of klf2a and notch1b expression. In tune with this finding, loss of Krit1 results in strong expression of klf2a and notch1b within the endocardium and prevents formation of an abluminal population of valve leaflet cells from endocardial cushions. Our findings reveal an important developmental role of Ccm proteins in fine-tuning the sensitivity of ECs in response to the strength of the biomechanical stimulus of blood-flow during valvulogenesis.

Results
Expression levels of heg1 mRNA are positively regulated by blood-flow and Klf2a/b To test whether the regulation of heg1 or krit1 mRNAs responds to changes in blood-flow, we measured their expression levels using RT-qPCR in troponin T type 2a (tnnt2a) morphants that have a non-contractile heart and thus lack blood-flow (Sehnert et al., 2002). Under this condition, mRNA levels of heg1 but not of krit1 were significantly lower than in wild-type (WT) embryos at 54 hr post fertilization (hpf) ( Figure 1A). Similarly, antisense oligonucleotide morpholino (MO)-mediated knockdown of Klf2a/b caused significant downregulation of heg1 mRNA. In comparison, the expression levels of krit1 mRNA were not significantly altered ( Figure 1A). To test whether heg1 is positively regulated by Klf2a, we generated transgenic lines of zebrafish for heat-shock-mediated induction of klf2a [Tg(hsp70l:klf2a_IRES_EGFP) pbb22 ]. We found that overexpression of klf2a upon heat-shock at 48-50 hpf led to a significant upregulation of heg1 mRNA expression by 54-55 hpf ( Figure 1B). However, the treatment did not significantly upregulate levels of krit1 mRNA.
It has previously been shown that heg1 is expressed within endocardium (Mably et al., 2003). To determine the regional expression of heg1 mRNA within the endocardium, we performed wholemount in situ hybridizations in WT and tnnt2a morphants. Whereas heg1 was initially expressed throughout the entire endocardium at 30 hpf, by 48 hpf, expression became more restricted to cells exposed to a high fluid shear stress including the atrioventricular canal (AVC) endocardium, and this was even more striking at 72 hpf [ Figure 1-figure supplement 1A-C;(Münch et al., 2017)]. In tune with flow-dependent regulation within the endocardium, expression of heg1 mRNA was reduced in tnnt2a morphants at 48 hpf ( Figure 1-figure supplement 1D).   As Heg1 and Krit1 proteins are known to interact in vitro (Kleaveland et al., 2009), we next tested whether Heg1 protein affects protein levels of its binding partner Krit1. To this end, we first generated a transgenic line of zebrafish for expression of EGFP-Krit1 [Tg(UAS:EGFP-krit1) pbb21 ]. Next, we analyzed the levels of EGFP-Krit1 protein in heg1 MO-injected embryos and found that they were significantly lower than in WT control embryos in the heart ( Figure 1C-E) and the caudal vasculature at 30 hpf ( Figure 1-figure supplement 2). Taken together, the levels of heg1 mRNA expression are positively regulated by blood-flow and Klf2a-dependent mechanotransduction. In turn, Heg1 has a stabilizing effect on Krit1 levels.
Overexpression of heg1 or krit1 mRNA dampens expression levels of klf2a mRNA A loss of Ccm proteins in zebrafish or mice results in higher levels of klf2 mRNA expression, suggesting that the physiological role of Ccm proteins is to modulate klf2 expression levels in response to blood-flow (Renz et al., 2015;Zhou et al., 2015Zhou et al., , 2016. Given that levels of heg1 mRNA expression are affected by blood-flow, we explored whether upregulation of Heg1 or Krit1 would have an impact on endothelial mechanotransduction pathways. First, we injected heg1 mRNA at the one-cell stage and assessed klf2a mRNA levels at 24 and 48 hpf. At both times, high levels of heg1 mRNA correlated with significantly lower levels of klf2a expression ( Figure 2A). Next, we tested whether overexpression of krit1 also downregulates levels of klf2a mRNA by using a transgenic line with heat-shock-inducible krit1 [Tg(hsp70l:krit1_IRES EGFP) md6 ]. Treatment with multiple heat-shocks at 14, 24, and 40 hpf resulted in significant downregulation of klf2a mRNA levels throughout the entire embryo at 48 hpf as determined by RT-qPCR (klf2a: fold change 0.63, p<0.05, n = 4 replicates; krit1: fold change 9.77, p<0.001, n = 4 replicates). Cardiac cushions formed normally after this treatment ( Figure 2-figure supplement 1) and blood-flow was not affected as assessed by visual inspection (n > 100 embryos analyzed showed normal blood circulation). Hence, the forced upregulation of heg1 or krit1 dampens expression levels of klf2a mRNA. This finding suggests that the strength of the biomechanical forces resulting from blood-flow impacts the expression levels of Ccm proteins; this modulates mechanosensitive signaling within endothelial cells upstream of klf2a.
Krit1 and Ccm2 impact expression of the mechanosensitive Notch activity reporter Tg(TP1:VenusPEST) To further assess the physiological effects of altered levels of CCM proteins for endocardial patterning, we analyzed expression levels of the transgenic line Tg(TP1:VenusPEST) s940 , which expresses destabilized Venus protein in regions of strong Notch activity (Ninov et al., 2012). In zebrafish, notch1b is a blood-flow-responsive gene that is highly expressed at cardiac cushions (Vermot et al., 2009). By 54 hpf, the expression of Tg(TP1:VenusPEST) s940 is restricted to the ventricular chamber and is particularly strong within the AVC region ( Figure 2B). To induce upregulation of krit1 or ccm2, Tg(hsp70l:krit1_IRES EGFP) md6 or Tg(hsp70l:ccm2_IRES EGFP) md12 embryos were treated with multiple heat-shocks at 14, 24, 40, and 48 hpf. On krit1 or ccm2 overexpression, and consistent with a dampening of blood-flow responses, expression from the Tg(TP1:VenusPEST) s940 reporter was generally weakened and was not restricted to the ventricle and AVC region at 54 hpf ( Figure

Krit1 is required for generation of abluminal cell fates during valvulogenesis
Given the strong involvement of Ccm proteins in mechanotransduction pathways, we next explored whether they have a developmental role during remodeling of cardiac cushions into valve leaflets, a morphogenetic process that is highly sensitive to the biomechanical stimulus of blood-flow Pestel et al., 2016;Scherz et al., 2008;Steed et al., 2016;Vermot et al., 2009). Previously, it was not possible to address such a potential developmental role as zebrafish ccm mutant endocardial cells fail to form cardiac cushions [ Figure 3B; (Renz et al., 2015). This is in contrast to WT endocardial cushion cells that acquire cuboidal shapes at 48 hpf and express activated leukocyte cell adhesion molecule (Alcam) ( Figure 3A; Beis et al., 2005). To elucidate whether CCM proteins play a role in this process, we performed a rescue experiment by injecting mRNA encoding EGFP-Krit1 into krit1 ty219c mutants, which rescued the cardiovascular defects associated with the loss of Krit1 at 48 hpf ( Figure 3C, Figure 3-figure supplement 1). Zebrafish ccm mutants have increased endocardial cell numbers by 48 hpf (Renz et al., 2015). To determine whether injection of mRNA encoding EGFP-Krit1 into krit1 ty219c mutants could rescue endocardial cell numbers, we compared ventricular endocardial cell numbers of WT and krit1 ty219c rescued embryos at 48 and 55 hpf ( Figure 3-figure supplement 2). Consistent with the strong rescue of the krit1 ty219c mutant phenotype at 48 hpf, the number of ventricular endocardial cells was normal at this stage. However, by 55 hpf, ventricular endocardial cell numbers had increased beyond those in WT, which suggests that at this stage the krit1 ty219c mutant phenotype becomes expressed again. To assay whether the re-appearance of a late krit1 ty219c mutant phenotype correlated with decreasing egfp-krit1 mRNA levels, we monitored krit1 mRNA levels over time using RT-qPCR. Indeed, WT embryos injected with egfp-krit1 mRNA showed a significant decrease of krit1 mRNA levels between 24 and 96 hpf (Figure 3-figure supplement 3). This result is consistent with recurrence of the krit1 ty219c mutant phenotype at 55 hpf. During the stage at which cardiac cushions form, we selected egfp-krit1 mRNA injected krit1 ty219c mutant embryos with normal blood-flow and found that they had developed normal endocardial   cushions by 48 hpf ( Figure 3C). Next, we analyzed whether these rescued krit1 ty219c mutant embryos would revert to a mutant cardiovascular phenotype by 96 hpf, a point at which endocardial cushions have been transformed via complex cellular rearrangements into double-layered valve leaflets Pestel et al., 2016;Scherz et al., 2008;Steed et al., 2016). In WT, those endocardial cells that face the luminal side of the valve leaflet and are exposed to the strongest shear stress, express high levels of klf2a (Steed et al., 2016) and can be identified by strong Alcam ( Figure 3D) and Cdh5 expression (Steed et al., 2016) In comparison, endocardial cells on the abluminal side of the valve leaflet have lower levels of klf2a, (Steed et al., 2016) and Alcam ( Figure 3D) and Cdh5 are only weakly expressed in abluminal cells . This is in tune with the more mesenchymal identity ascribed to this endocardial cell population (Pestel et al., 2016;Steed et al., 2016). We found that krit1 ty219c mutant embryos with a cardiac rescue that were exposed to normal blood-flow at 48 hpf, developed strongly dysmorphic valve leaflets by 96 hpf (n = 17/23 krit1 ty219c mutant embryos). At that stage, most mutants had valve leaflet endocardial cells on the abluminal sides that strongly

Krit1 dampens klf2a expression and Notch activation during cardiac valve morphogenesis
Ingression of abluminal cells from cardiac cushions has been associated with a bias in klf2a expression levels: whereas some endocardial cushion cells express high levels of klf2a and contribute to the luminal part of the valve leaflet, others express lower levels of klf2a which correlates with a mesenchymal-like morphology and their contribution to the abluminal portion of the valve leaflet (Steed et al., 2016). Our findings suggest that elevated levels of klf2a mRNA are the cause for defects in valve leaflet morphogenesis. To assess klf2a expression in rescued krit1 ty219c mutant hearts, we used whole-mount in situ hybridization. By 54 hpf, klf2a was strongly expressed within cardiac cushions at the AVC, whereas in other regions of the endocardium expression levels were lower or undetectable ( Figure 3G). In krit1 ty219c mutants, klf2a was strongly expressed throughout the entire endocardium and was not restricted to the AVC ( Figure 3H). At 54 hpf, in krit1 ty219c mutant embryos that were rescued by injection of mRNA encoding EGFP-Krit1, the expression of klf2a had a normal pattern with a localized strong expression at the AVC cushions (n = 6/6) ( Figure 3I). However, by 96 hpf, the expression of klf2a was not restricted to the AVC and outflow tract regions of the endocardium as in WT ( Figure 3J) but had reverted to wide and strong expressionpresent throughout the entire endocardium (n = 8/8) ( Figure 3L).
klf2a is a regulator of notch1b expression at cardiac cushions (Vermot et al., 2009). In addition, several studies have demonstrated that the reduction of CCM protein levels is associated with decreased Notch activity in human endothelial cells (Wüstehube et al., 2010;You et al., 2013). Hence, changes in Notch activity may contribute to defective cardiac valve leaflet morphogenesis in ccm mutants. To address this question, we first assessed notch1b expression by whole-mount in situ hybridization. At 48 hpf, notch1b was strongly expressed within high shear stress regions including the AVC of the WT heart ( Figure 4A). In contrast, notch1b was expressed throughout the entire endocardium in krit1 ty219c mutant embryos ( Figure 4B). Next, we characterized Notch signaling activity using the Tg(TP1:VenusPEST) s940 reporter. In WT, Notch activity was in a mosaic pattern within endocardial cushions and a few cushion cells neighboring the ventricle were consistently downregulating Notch activity ( Figure 4C'; n > 30 embryos analyzed). Those cells lacking Notch activity were in positions that have been shown to initiate endocardial sprouting behaviors and to contribute to formation of the abluminal portion of the cardiac valve leaflets (Steed et al., 2016). In striking contrast, in ccm2 m201 and krit1 ty219c mutants, Notch reporter expression was expanded throughout large regions of the endocardium ( Figure 4D 1B), and was especially active within all endocardial cells at the AVC region which demonstrated that a singling-out process and downregulation of Notch signaling among few cushions cells did not occur ( Figure 4D'; n > 30 embryos analyzed). In comparison, the expression of tbx2b, another AVC marker gene (Sedletcaia and Evans, 2011), was not expanded in ccm2 m201 mutants (Figure 4-figure supplement 1C-D). Hence, expansion of notch1b expression domain in ccm2 m201 or krit1 ty219c mutants is not a result of regional expansion of the AVC. Taken together, our findings suggest a critical role for Krit1 in modulating expression levels of klf2a and for Notch signaling among endocardial cushion cells which may control endocardial sprouting during the formation of cardiac valve leaflets.

Discussion
Our study establishes CCM proteins as having an important role in endothelial mechanosensitive signaling. We report that levels of heg1 mRNA expression are positively regulated by blood-flow in a Klf2-dependent manner and that Heg1 stabilizes Krit1 protein levels. We also show that overexpression of Heg1 and Krit1 dampens expression of klf2a mRNA. This establishes a negative feedback loop that desensitizes endothelial cells to blood-flow-induced mechanosensitive signaling by dampening expression levels of klf2a. The fine-tuning of blood-flow-sensitive responses is particularly important during developmental processes such as cardiac valvulogenesis, which involves establishment of regionalized differences in klf2a expression levels (Steed et al., 2016). While the endocardial cells facing the lumen of the heart experience high fluid shear forces, and consequently begin to express high levels of klf2a, those on the other side do not. In tune with this observation, the process of valvulogenesis is critically dependent on the biophysical stimulus by blood-flow Heckel et al., 2015;Pestel et al., 2016;Steed et al., 2016) and is affected by the loss of klf2a (Steed et al., 2016). Hence, mechanosensitive signaling pathways actively reshape cardiac valve leaflets in a manner that is highly sensitive to constantly changing blood-flow conditions. The present work suggests that Krit1 plays an important role in modulating expression levels of klf2a during this complex morphogenetic process. Our results suggest that downregulation of endocardial mechanosensitivity by CCM proteins generates a bias in expression levels of klf2a which is critical for correct valvulogenesis.
In accordance with our findings, expression of the KRIT1-CCM2-associated protein CCM2L has also been reported in a blood-flow-dependent manner (Cullere et al., 2015). In a complex with KRIT1 and CCM2, CCM2L directly binds to the MAP kinase MEKK3 and inhibits its activation. This reduces the ability of MEKK3 to phosphorylate MEK5 which is required to activate ERK5 (Cullere et al., 2015). The MEKK3-MEK5-ERK5 pathway is involved in activation of KLF2 expression (Zhou et al., 2015;Zhou et al., 2016), the control of endoMT (reviewed in Drew et al., 2012), and murine cardiovascular development (Yang et al., 2000;reviewed in Rose et al., 2010). Activation of klf2a expression and control of endoMT processes could be particularly important during formation of functional cardiac valve leaflets. Within abluminal endocardial cells of the zebrafish cardiac valve leaflet, the loss of Krit1 prevents downregulation of Cdh5 and Alcam proteins, a function that might ultimately be related to elevated levels of Klf2a. As yet, no blood-flow-independent mechanism has been identified that would trigger such an upregulation of Klf2a in the absence of Ccm proteins (see model, Figure 5). As Klf2a regulates the expression of endocardial notch1b within endocardial cushions (Vermot et al., 2009), the formation of abluminal valve leaflets may be regulated by a Notchdependent lateral inhibition process similar to angiogenic sprouting of tip cells. Whereas high klf2a expression corresponds to expression of the endothelial junctional proteins Cdh5 and Alcam within luminal endocardial valve cells, abluminal endocardial cells that are shielded from blood-flow and express lower levels of klf2a and notch1b downregulate the expression of Cdh5 (Steed et al., 2016) and Alcam and acquire a more mesenchymal-like state. Several publications have discussed the possibility that downregulation of Cdh5 and Alcam in abluminal endocardial cells and the behaviour of these endocardial cells during zebrafish valvulogenesis Lagendijk et al., 2011;Scherz et al., 2008;Steed et al., 2016) is highly reminiscent of the endoMT process that occurs during cardiac valve formation in the mouse (Chiplunkar et al., 2013). Our findings now implicate CCM proteins in providing a bias in klf2a and notch1b expression levels within cardiac cushions, which may be further elaborated by Notch-dependent lateral inhibition that could allocate cells to an abluminal fate (see model, Figure 5). In tune with such a model, Notch activity was in a mosaic pattern within WT endocardial cushions, while in ccm mutants, Notch signaling was widely active. Interestingly, pharmacological inhibition of Notch signalling has been shown to cause similar cardiac valve leaflet morphogenesis defects . Hence, it remains an important question for future research to elucidate whether Notch-dependent endocardial sprouting morphogenesis and/or endoMT are involved in the process of valve leaflet formation.
Taken together, our findings uncover crucial roles of Heg1 and Krit1 in controlling the sensitivity of endothelial cells to hemodynamic forces. This may promote developmental adaptations in response to changes in the strength of hemodynamic forces that occur during cardiovascular development. Within the early zebrafish heart tube, shear forces resulting from blood-flow increase rapidly within the first 4-5 days post fertilization (Hove et al., 2003). Similar changes in blood-flow dynamics also occur during the remodelling of the vasculature or during pathological changes of Blood flowindependent mechanism Figure 5. Model of the molecular pathways involved in endocardial cushion cell sprouting at 54 hpf. In wild-type, endocardial cells exposed to blood-flow have high levels of Klf2a, Notch, Alcam, and maintain cell adhesion. Ccm proteins provide a bias in Klf2a and Notch expression levels. Some cells that express lower levels of klf2a and have lower Notch activity establish protrusions and migrate into the cardiac jelly (endocardial sprouting). In ccm mutants, endocardial cells overexpress klf2a and have higher Notch activity because of a blood-flow-independent mechanism. As a consequence, high levels of Notch activity throughout the entire endocardium result in failure of endocardial sprouting. DOI: https://doi.org/10.7554/eLife.28939.015 blood-flow patterns. It will be important to assess the physiological roles of CCM proteins during these different adaptations of the vasculature.

Generation of transgenic lines of zebrafish
Transformation plasmids (25 pg/embryo) were co-injected together with mRNA encoding Tol2 transposase (50 pg/embryo) into one-cell-stage zebrafish embryos. Several independent transgenic lines were established for each construct. In functional tests and localization studies, these independent lines resulted in comparable phenotypes. One transgene for each construct was selected for further analyses:

mRNA injection experiments
Capped mRNA encoding EGFP-Krit1 or Heg1 was synthesized using SP6 polymerase (mMessage Machine kit, Ambion). For rescue experiments, 150 pg of egfp-krit1 mRNA was injected into onecell-stage zebrafish embryos. Embryos were selected for EGFP fluorescence at 6 hpf and the genotype was assessed by sequencing. For heg1 overexpression, 100 pg of heg1 mRNA was injected into one-cell-stage zebrafish embryos.

Statistical analysis of the efficiency of mRNA rescue experiments
Statistical analysis of the efficiency of egfp-krit1 mRNA rescue experiments (as shown in Figure 3figure supplement 1) was done using GraphPad Prism6 (Student's t-test, p=0.008 and p=0.005, respectively). Statistical analysis of the rescue efficiency of egfp-krit1 mRNA injection into krit1 ty219c mutant embryos was based on the presence of blood-flow at 48 hpf for all embryos, and at 96 hpf for those embryos that had blood-flow at 48hpf. The percentages of embryos with blood-flow was recorded for three individual experiments and compared with GraphPad Prism6, using a 2way ANOVA with Multiple comparisons without correction.

Quantifications of ventricular endocardial cell numbers
Endocardial cell numbers of the ventricles at 48 hpf and 55 hpf of WT and krit1 ty219c mutants were quantified as previously shown (Renz et al., 2015). Nuclei were visualized by Tg(kdrl:GFP) s843 expression and were counted within the ventricle (for endocardium). Cell numbers are shown as means with SEM. Prism 6 (GraphPad) was used to perform an unpaired t-test (Figure 3-

Ratiometric corrected total tissue fluorescence (CTTF) image analysis
The ratiometric measurements of Tg(TP1:VenusPEST) s940 fluorescence in atrium versus ventricle and AVC of the embryonic heart ( Figure 2E) was done as previously described (McCloy et al., 2014). As overexpression of krit1 also resulted in an upregulation of EGFP, the overlap of EGFP with Tg (TP1:VenusPEST) s940 fluorescence was safely separated into two different channels by recording the images using the online fingerprinting function of the Zen 8.1 Software (Zeiss). Images were acquired maintaining fixed recording settings. From a 3D confocal image, the z-portion corresponding to the entire heart was selected. Regions of single z-planes corresponding to atrium or to ventricle plus the AVC were selected and fluorescence was measured until entire hearts were covered. The corrected total tissue fluorescence [CTTF = integrated density -(area of selected tissue x mean fluorescence of background readings)] was calculated using Excel 2010 (Microsoft Office). Next, all CTTF values collected for atrium as well as for ventricle plus AVC of each heart were averaged. The mean of atrium CTTF values was divided by the mean of ventricle plus AVC CTTF values. A mean value of 1 corresponds to an equal expression of Tg(TP1:VenusPEST) s940 in both heart chambers. All measurements and tissue selection were performed using Fiji software (Schindelin et al., 2012). Statistical analysis was done with Excel 2010 (Microsoft Office) using an unpaired t-test (n: number of hearts analyzed, hs: heat-shock, *p<0.05, **p<0.01). To quantify the fluorescent levels of Tg(TP1:VenusPEST) s940 in the AVC of WT (heat-shock control) ( Figure 2F), hs:Krit1 and hs:Ccm2, a section of 20 mm corresponding to the entire AVC was selected from a maximum projection of confocal z-stacks of the AVC. The CTTF value of each AVC was measured and all the values were divided by the average CTTF of the controls for normalization. To quantify the fluorescent levels of EGFP-Krit1 in WT and heg1 morphants, maximum projections of confocal z-stacks of manually extracted 30 hpf hearts ( Figure 1C-E) or of the caudal plexus region (Figure 1-figure supplement 2) were used. The background intensity was measured in five different areas of each image. To measure the caudal plexus region, an area between two intersegmental vessels from the most dorsal to the most ventral side of the caudal plexus was selected. This was repeated in three different areas for each embryo. All measurements and tissue selection were performed using Fiji software (Schindelin et al., 2012). The CTTF was calculated using Excel 2010 (Microsoft Office). All the values were divided by the average intensity of the controls.

Quantifications of mRNA expression by RT-qPCR
For RT-qPCR experiments, 25 zebrafish embryos were pooled for each condition (three biological replicates). For heat-shock conditions, controls of each biological replicate were composed of heatshocked siblings from the same clutch lacking EGFP expression. Total RNA was extracted with Trizol (Sigma) and Phase Lock Gel Heavy tubes (1.5 mL, Prime 5) and the corresponding cDNA was synthesized from total RNA with the RevertAid H Minus First Strand cDNA Synthesis kit (ThermoFisher Scientific). RT-qPCR experiments were performed as described (Renz et al., 2015) using 18 ng cDNA per technical replicate and the KAPA Sybr Fast qPCR kit (Peqlab) on a PikoReal 96 Real-Time PCR System (ThermoFisher Scientific). Cycle threshold (Ct) values were determined by PikoReal software 2.2 (ThermoFisher Scientific). eif1b was used as a housekeeping gene. The following primers were used for qRT-PCR: Results were analyzed using the comparative threshold cycle method (2-DDCt) to compare gene expression levels between samples as previously described (Livak and Schmittgen, 2001). As an internal reference, we used zebrafish eif1b (Renz et al., 2015). Control sample values were normalized to 1. In the table below, the mean of the fold changes and the corresponding SEM of each biological replicate after normalization is shown. As each single biological replicate represents an independent experiment from an independent clutch of embryos, ratio paired t-tests were done with Prism 6 (GraphPad) (Stg: developmental stage of zebrafish embryos at RNA extraction, n: number of independent biological replicates analyzed, *p<0.05, **p<0.01, hs: heat-shock, mean >1 corresponds with target upregulation; mean <1 represents target downregulation).

Treatment
Stg ( To prove that differences in average values of krit1 mRNA levels between 24 hpf and 48 hpf, 48 hpf and 96 hpf, 24 hpf and 96 hpf were significant, a ratio paired t-test was performed, as in each single biological replicate, embryos of 24, 48, and 96 hpf were from the same clutches. The mean of the ratio indicated in the table below is always between 1 and 0, and this accounts for target downregulation. Hodge, and team members. The group has been generously supported by the excellence cluster REBIRTH, SFB958, and by Deutsche Forschungsgemeinschaft (DFG) projects SE2016/7-2 and SE2016/10-1 to SA-S. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.