Endocardial primary cilia and blood flow regulate EndoMT during endocardial cushion development

Berg, Kathryn; Gorham, Joshua; Lundt, Faith; Seidman, Jonathan; Brueckner, Martina

doi:10.1038/s44161-025-00697-z

Download PDF

Article
Open access
Published: 26 August 2025

Endocardial primary cilia and blood flow regulate EndoMT during endocardial cushion development

Nature Cardiovascular Research volume 4, pages 1114–1134 (2025)Cite this article

2828 Accesses
1 Altmetric
Metrics details

Subjects

Abstract

Valvular heart disease affects 2.5% of the population and is frequently associated with congenital heart disease. Blood flow is critical for valve formation, but the cellular mechanosensors translating flow into the transcriptional regulation of valve development remain undiscovered. Here, we identify that primary cilia and blood flow in mouse embryos regulate early valve development in vivo by regionally controlling endothelial-to-mesenchymal transition (EndoMT) through the modulation of Krüppel-like factor 4 (Klf4) in the endocardial cushions. Endocardial ciliation decreases during cushion development in regions of high shear stress, correlating with KLF4 downregulation and EndoMT progression. Mouse embryos lacking cilia exhibit blood flow-dependent accumulation of KLF4 and impaired cushion cellularization. Single-nucleus RNA sequencing revealed that the cilia-knockout and contractility-knockout endocardium fails to progress through EndoMT pseudostages, retains endothelial markers, and has reduced EndoMT and mesenchymal genes that KLF4 antagonizes. These data indicate that endocardial primary cilia function as mechanosensors in cushion development through the regional regulation of KLF4.

A change of heart: new roles for cilia in cardiac development and disease

Article 03 December 2021

Cell diversity and plasticity during atrioventricular heart valve EMTs

Article Open access 09 September 2023

Generation and characterization of cardiac valve endothelial-like cells from human pluripotent stem cells

Article Open access 06 September 2021

Main

Valvular heart disease occurs in 2.5% of the population and is a component of many types of congenital heart disease (CHD), a leading cause of infant mortality¹. Valve development takes place in stages: small pockets of cardiac jelly, called endocardial cushions (ECCs), form and become cellularized through endothelial-to-mesenchymal transition (EndoMT), during which the endocardial luminal lining of the heart differentiates into mesenchyme and migrates into the cardiac jelly, populating the ECCs². This occurs in the outflow tract (OFT) and the atrioventricular canal (AVC). The cushion then goes through stages of remodeling, including mesenchymal proliferation and condensation, resulting in functional valve leaflets. Many CHDs exhibit valve or septal defects linked to improper ECC formation, remodeling or fusion during development³; an example is hypoplastic left heart syndrome, a CHD characterized by defective aortic and mitral valves and directly tied to abnormal EndoMT⁴.

ECCs form at the atrioventricular and ventriculoarterial junctions, which are areas of high wall shear stress (WSS) due to the flow of blood through a small luminal diameter. Blood flow and mechanical forces have been linked to cardiogenesis, including ECC development (reviewed in refs. ^2,5). Lack of contractility results in the absence of the critical endocardial ring before ECC formation in zebrafish and a complete loss of cushion cellularization in mice, indicating that EndoMT requires mechanical stimuli^6,7. Shear stress was demonstrated to be both necessary and instructive for valve development in zebrafish⁸. However, the cellular sensors that translate blood flow into the genetic regulation of cardiac cushion formation remain poorly understood, particularly in the four-chambered hearts of mammals.

Primary cilia are rod-like organelles found on most cell types, including the cells of the endocardium and vasculature^9,10,11. In vitro work has demonstrated that shear stress induces the loss of primary cilia (called ‘deciliation’), which ‘primes’ the EndoMT of cultured endothelial cells^12,13. During cardiac development, primary cilia are best known for their role in left–right asymmetry, where motile primary cilia generate extraembryonic fluid flow at the left–right organizer (LRO), which is then detected by immotile primary cilia to instruct the direction of cardiac looping¹⁴. Immotile primary cilia are also essential during later valve remodeling and zebrafish ventricular regeneration, and ciliary genes are expressed in the endocardium undergoing EndoMT^15,16,17. However, whether primary cilia act as endocardial mechanosensors during in vivo cushion formation, especially in four-chambered hearts, is unknown.

Here, we demonstrate a newly identified role for primary cilia as blood flow sensors in heart cushion formation in mice. Flow-dependent, dynamic ciliation correlates with the spatially selective regulation of the EndoMT antagonist Krüppel-like factor 4 (KLF4) and subsequent cushion EndoMT. Single-nucleus RNA sequencing (snRNA-seq) data show the transcriptional consequences of constitutive loss of cilia and cardiac contractility during early heart development, including the dysregulation of KLF4 targets and EndoMT pathways. EndoMT initiates but does not progress in embryos lacking primary cilia or blood flow, resulting in decellularized cushions. This highlights a requirement for primary cilia as flow sensors guiding EndoMT during heart formation independent of left–right patterning and upstream of valve remodeling.

Results

Spatiotemporally dynamic endocardial ciliation during cushion EndoMT is blood flow dependent

We first evaluated ECC endocardial cell ciliation (the percentage of cells with a cilium) before and during cushion formation. Cilia ranging from 0.5 to 1.5 µm in length extended into the lumen from the endocardial lining of both the OFT and the AVC from embryonic day 8.5 (e8.5) to e10 (Fig. 1a,b and Extended Data Fig. 1a,b). Mesenchymal cells inside the ECCs were predominately ciliated, suggesting endocardial cells that underwent EndoMT are able to reciliate as they migrate away from the lumen (Extended Data Fig. 1c,d). Between e8.5 and e9.0, we observed a decrease in total luminal endocardial ciliation in both the OFT and AVC ECCs, suggesting a deciliation event following the onset of laminar flow at e8.25 (Fig. 1d). Subsequently, total endocardial ciliation in the ECCs remained constant until the end of cushion EndoMT, despite increasing shear stress. The initial loss and subsequent plateauing of endocardial ciliation are evolutionarily conserved, with similar results observed in live zebrafish embryo AVCs during developmentally equivalent time points (Extended Data Fig. 1e,f).

Although total endocardial ciliation remained constant following the initial loss between e8.5 and e9.0, the spatial patterning of endocardial ciliation across the developing ECC was dynamic. Ciliation progressively disappeared at the AVC center and remained at the ventricular and, to a lesser degree, atrial junctions (Fig. 1c). A similar pattern emerged at e9.5 in the OFT cushions (Extended Data Fig. 1g). The areas of low ciliation correlated with areas previously shown to experience high WSS¹⁸.

To determine whether the regional cilia loss was due to WSS, we examined ECC ciliation in e9.5 Ncx1^−/− mutant mouse embryos that lacked cardiac contractility due to the loss of the Na⁺/Ca²⁺ exchanger protein¹⁹. Ncx1^−/− mutant embryos were viable until e9.75; at e9.5, they did not exhibit altered cell death or externally visible cardiac phenotypes beyond a lack of heartbeat compared to littermates (Extended Data Fig. 1h,i). Ncx1^−/− mutant embryos retained endocardial cilia, which remained ubiquitously distributed across the AVC and OFT (Fig. 1e–g and Extended Data Fig. 1j). These results demonstrate that blood flow is necessary for the dynamic regional endocardial distribution of cilia during cushion formation.

ECC EndoMT requires primary cilia independent of heart looping

Having established that endocardial ciliation is spatially dynamic downstream of blood flow, we investigated the necessity of primary cilia for normal ECC development. As ciliary proteins are continuously recycled and are consequently long-lived, previous studies using conditional cilia knockouts (KOs) have been unable to document cilia loss within a timeframe relevant for ECC formation¹⁷. Therefore, we used two independent constitutive cilia KO mouse models to ensure the loss of cilia during and before cushion development. These models, Kif3a^−/− (kinesin family member 3A) and Ift20^−/− (intraflagellar transport protein 20)^20,21, lead to a complete loss of cilia through distinct mechanisms (Fig. 2a,b and Extended Data Fig. 2a,b). As the cardiac phenotype and embryonic lethality are indistinguishable between these models, they will be shown aggregated and referred to as cilia KOs. Both Kif3a^−/− and Ift20^−/− mutant embryos are viable until e9.75 and exhibit randomized left–right asymmetry, occasional pericardial edema, thinned compact myocardium and loss of neural tube closure¹⁰. Importantly, at e9.5, cilia KO hearts beat at similar frequencies to wild-type (WT) littermate hearts (Supplementary Video 1) and show no cell death and cell cycle differences compared to WT e9.5 hearts (Extended Data Fig. 2c–e).

To determine the impact of constitutive cilia loss on EndoMT, we examined cushion cellularization in WT and cilia KO hearts and observed a complete lack of interstitial cells in the cilia KO cushions (Fig. 2c). The embryos were matched with respect to somite count. To confirm that interstitial cells were of endocardial origin, we measured the proportion of cells that had infiltrated the cushion while also retaining residual CD31 expression (Fig. 2d). Cilia KO mutants showed 0% cellularization of the AVC by e9.5 (identical to contractility-mutant Ncx1^−/− embryos), compared to an average of 17% cellularization in WT embryos. This indicates that cilia KO hearts lack cushion cellularization specifically due to an EndoMT defect.

Cilia KOs exhibit a random heart looping direction due to the requirement for cilia motility and sensing at the LRO before AVC development²². To determine whether the EndoMT defect in cilia KO hearts was due to hemodynamic and structural changes secondary to variable heart looping, rather than a direct effect of AVC intracardiac cilia sensing, we evaluated Dnah11^−/− (dynein axonemal heavy chain 11) mouse embryos. The constitutive KO of Dnah11 results in immotile LRO cilia and identical heart looping phenotypes as observed in cilia KO mice without affecting intracardiac primary cilia²³. Endocardial ciliation in Dnah11^−/− embryos was identical to that in WT siblings, both in amount and regionalization across the AVC (Fig. 2e,f). Dnah11^−/− embryos were viable and showed normal cellularization of the AVC ECC, even in A-loop hearts, suggesting that EndoMT and cushion formation are unaffected by variable heart looping (Fig. 2g). Together, these data demonstrate a previously unknown requirement for endocardial ciliation for proper cushion EndoMT independent of cardiac left–right asymmetry.

The mechanosensitive transcription factors KLF2 and KLF4 are involved in human valve development and are dynamically expressed in the mouse endocardium during cushion EndoMT

KLF2 and KLF4 are mechanosensitive zinc finger-containing transcription factors. We have previously identified dominant loss-of-function and damaging missense variants in both KLF2 and KLF4 in patients with aortic and mitral valve defects associated with hypoplastic left heart syndrome²⁴. KLF2 and KLF4 have overlapping functions in vascular development²⁵, but they are upregulated differently in response to mechanical forces²⁶. Both Klf2 and Klf4 mRNAs were expressed in the endocardial luminal lining of the AVC and OFT (Fig. 3a,b and Extended Data Fig. 3a,b). While Klf4 was most prominent in the ECCs, Klf2 was more widely expressed, including in the intersomitic vessels, cardinal vein, ventricle, dorsal aorta and atrium, consistent with the role of KLF2 in vascular identity. Klf2 and Klf4 mRNA expression increased following the onset of contraction (Fig. 3c,d). Analysis of published single-cell RNA-seq (scRNA-seq) data on mouse e7.75–e9.25 embryo hearts²⁷ found that, in contrast to Klf2, Klf4 (which is required earlier in development as a pluripotency factor) becomes predominantly expressed in the endocardium only between e8.25 and e9.25 (Fig. 3e). The expression levels of both Klf2 and Klf4 were lower in the OFT than in the AVC until e9.5 (Fig. 3d); KLF4 protein-positive cells in the OFT increased to the level observed in the AVC by e10 (Extended Data Fig. 3c), consistent with the known developmental delay between the OFT and AVC²⁸.

**Fig. 3: The mechanosensitive transcription factors KLF2 and KLF4 are dynamically expressed in the mouse endocardium during cushion EndoMT.**

Klf2 expression across the AVC endocardium resembled a bell curve mirroring known WSS levels¹⁸ (Fig. 3f), as has been observed previously at later time points of valve development^29,30. In contrast, Klf4 expression (both mRNA and protein) was higher at the ends of the AVC, especially on the ventricular side (Fig. 3f,g). This cluster of cells with high expression of KLF4 was first noticeable at e8.5 but comprised approximately 20% of the total AVC length by e9.5. A second peak of KLF4 protein-positive cells could also be seen at the atrial end by e10. A similar ventricular pattern could be observed in the OFT, which, like the ciliation pattern, was not as robust as in the AVC (Extended Data Fig. 3d,e). These data uncover two distinct populations of KLF4-expressing cells: KLF4-low cells in the center of the AVC and KLF4-high cells at the ends.

KLF4 expression negatively correlates with cushion EndoMT progression

KLF4 has been shown to inhibit or activate EndoMT both in vitro and in vivo, but its role in cushion EndoMT is unclear^13,31,32,33. To determine the effect of KLF4 expression on cushion EndoMT, we examined the expression patterns of vimentin (Vim, a mesenchymal marker) and the kinase insert domain receptor (Kdr, an endothelial marker) in KLF4-high and KLF4-low regions. Double-positive cells (considered to be undergoing EndoMT) were found in the center of the AVC at e9.5 (Extended Data Fig. 4a). Vim expression decreased in the KLF4-high region (Fig. 4a,b). Cushion cellularization was also first noticeable in the center of the AVC from e9.0 to e10 (Fig. 4c,d). The observed correlation between the KLF4-low population and the expression of EndoMT markers in the center of the AVC suggests that KLF4 inhibits cushion EndoMT.

To explore how endocardial Klf4 expression correlates with EndoMT progression at the transcriptional level, we defined pseudostages of EndoMT by measuring the ratio of endocardial and mesenchymal gene set enrichments on a previously published mouse heart scRNA-seq dataset²⁷ (Fig. 4e,f). These pseudostages were validated by running Slingshot³⁴ for pseudotime analysis. We then performed snRNA-seq on microdissected e9.5 WT mouse hearts and confirmed equivalent pseudostages to the scRNA-seq dataset²⁷. From this, we were able to identify gene networks associated with EndoMT progression by highlighting markers for each pseudostage (Supplementary Tables 1–4), such as valvular endocardium (Tmem100, Emcn and Gja4), early EndoMT (Wnt4, Fn1 and Fabp5), mid EndoMT (Trpm3, Bmper and Col23a1) and late EndoMT (Pdgfra, Vcan and Twist1/2). Immunofluorescence staining of FN, identified in our dataset as an early EndoMT marker, was found exclusively in the center of the AVC where cellularization first occurs (Fig. 4c), suggesting EndoMT progression in KLF4-low regions.

Total Klf4 mRNA peaked in mid EndoMT and was significantly decreased by late EndoMT (Fig. 4g). As KLF4 inhibits the EndoMT factors Snai1 and Snai2 (Slug)^35,36, we investigated nascent Snai1 and Snai2 expression in cells undergoing EndoMT in our snRNA-seq dataset. Analysis of the endocardium from this snRNA-seq dataset confirmed a negative correlation between Klf4 and Snai1, Snai2 and other essential EndoMT genes, such as Twist1, Twist2, Hand2, Tbx20, Msx1 and Sox9, which predominantly peaked in expression at late EndoMT (Extended Data Fig. 4b). The expression of genes encoding EndoMT antagonists, such as Vegfa, was positively correlated with Klf4. Together, these data suggest that the transition from mid to late EndoMT is linked to the downregulation of Klf4 expression.

Endocardial KLF4 expression correlates with ciliation and depends on blood flow

We observed that KLF4 expression does not spatially align with previously reported WSS patterns in which the highest WSS occurs at the center of the AVC¹⁸; instead, Klf4 mRNA (Fig. 3f) and KLF4 protein (Fig. 3g) peak at the distal parts of the AVC. Mechanical forces in the developing ECCs are complex and dynamic, and they include circumferential/radial stress and laminar/disturbed flow (reconstruction of previously published data^18,37,38 shown in Fig. 5a). KLF4 is not induced by circumferential/radial stress³⁹ or disturbed flow⁴⁰, so we propose that an endocardial mechanosensor capable of discriminating and responding to a range of laminar WSS specifies KLF4-low and KLF4-high regions. As cilia are shear mechanosensors present on the endocardium in a WSS-specific pattern, we explored the possibility of intracardiac cilia fulfilling this role. We found that KLF4-high regions were predominantly ciliated, whereas KLF4-low regions were predominantly nonciliated (Fig. 5b,c and Extended Data Fig. 5a), with an increasing correlation between ciliation and KLF4 regionalization during cushion development (Fig. 5d). KLF4 immunofluorescence intensity was quantitatively increased in ciliated compared to nonciliated endocardial cells (Fig. 5e). These results integrate ciliation and KLF4 expression and define two cell populations: KLF4-high ciliated cells and KLF4-low nonciliated cells.

As we observed that the spatial patterning of endocardial ciliation was dependent on WSS, we next examined whether regional KLF4 expression in the cushion similarly depends on blood flow. KLF4 expression was significantly reduced in the AVC and OFT ECCs of contractility-mutant Ncx1^−/− embryos (Fig. 5f–h and Extended Data Fig. 5b). The distribution of the limited KLF4-positive cells did not replicate the ventricular bias observed in WT but peaked instead in the center of the AVC (Fig. 5g), demonstrating that blood flow is necessary for regional KLF4 expression.

Failure of ciliogenesis results in abnormal KLF4 expression

We next investigated whether cilia are necessary for normal KLF4 expression in ECCs. KLF4 expression was indistinguishable between the constitutive cilia KO models (Ift20^−/− and Kif3a^−/−), so they will continue to be aggregated and referred to as cilia KOs (Extended Data Fig. 6a,b). At e8.5 (somite-matched to WT), cilia KO hearts had fewer KLF4-positive cells than WT littermate hearts (Fig. 6a and Extended Data Fig. 6c), consistent with in vitro observations showing decreased Klf4 mRNA in cilia-knockdown human umbilical vein endothelial cells¹³. Strikingly, by e9.0, cilia KO mutants showed significantly higher Klf4 mRNA expression and a greater number of KLF4-positive cells. The KLF4-low cell population in the center of the AVC observed in WT littermates was lost in cilia KO embryos, leaving only KLF4-high cells (Fig. 6b and Extended Data Fig. 6d). The KLF4-high cells were symmetrically distributed across the center of the AVC of e9.5 cilia KO hearts at both the protein and RNA levels, contrasting with the ventricular-biased distribution of KLF4-high AVC cells in WT littermates (Fig. 6c–e). The distribution of KLF4 expression in cilia KO OFTs exhibited a similar but subtler shift toward the OFT center (Extended Data Fig. 6e,f). Misexpression of KLF4 in cilia KOs occurs independent of heart looping, as Dnah11^−/− embryos with randomized left–right asymmetry showed no change in KLF4 expression compared to WT embryos (Fig. 6f).

To determine whether the overexpression of KLF4 in cilia KO cells depends on flow, we evaluated Ift20^−/−/Ncx1^−/− double-KO embryos. Ift20^−/−/Ncx1^−/− hearts were not significantly different from single-KO Ncx1^−/− hearts in terms of KLF4 expression (Fig. 6g,h and Extended Data Fig. 6g), indicating that blood flow acts upstream of cilia-dependent KLF4 regulation. Together, these data suggest that, while cilia contribute to upregulating KLF4 in early development when WSS is low, regional deciliation may be necessary for KLF4 downregulation. As cilia KO cells intrinsically do not possess cilia, they are unable to spatially regulate the downregulation of KLF4, resulting in KLF4 overexpression and loss of patterning.

Defects in both ciliogenesis and blood flow block EndoMT

To assess the EndoMT defect observed in cilia KO hearts, we first examined the early EndoMT marker FN. We found that FN levels in the cilia KO AVC endocardium were comparable to those in WT (Extended Data Fig. 6h,i). Cilia KO embryos then completely failed to cellularize their cushions, suggesting that they can initiate, but not complete, ECC EndoMT at e9.5.

To identify the EndoMT pseudostage that is blocked in cilia KOs, we performed snRNA-seq on microdissected e9.5 beating hearts from somite-matched Ift20^−/− and sibling (heterozygous and WT) embryos (Fig. 7a and Extended Data Fig. 7a,b). We obtained 30,823 single nuclei (29,313 after quality control) with 27,998 genes over two trials of five pooled hearts per genotype (nuclei counts: Ift20^+/+, 9,882; Ift20^+/−, 14,825; Ift20^−/−, 6,116). We used established marker genes to identify cluster identities and found that the proportions of major cell types were not significantly different between genotypes (Fig. 7b,c and Extended Data Fig. 7c,d,f–i). This includes the epicardium, a tissue layer that first appears at late e9.0, demonstrating that Ift20^−/− hearts develop similarly to littermate hearts. Subclustering of the endocardial population identified several cell types, including ECC-specific endocardial cells (Fig. 7c). Cells of the endocardium not in the ECC (referred to as ‘endocardial cells’) cluster with hematoendothelial cells and are marked by Irx5 (Iroquois homeobox transcription factor 5), which is not expressed in ECC regions⁴¹. ECC endocardial cells (referred to as ‘valvular endocardial cells’ (vECs)) cluster with EndoMT groups and are marked by Hey1 (Hes-related family bHLH transcription factor with YRPW motif 1), which is specifically expressed in the ECCs and is necessary for ECC EndoMT⁴².

**Fig. 7: Defects in both ciliogenesis and blood flow block EndoMT.**

Although the proportion of endocardial cell types did not significantly differ between genotypes, there were transcriptional differences. We analyzed differentially expressed genes (DEGs) between Ift20^−/− and WT endocardium (Fig. 7d and Supplementary Table 5). As we found increased KLF4 protein in the cilia KO endocardium, we analyzed whether these DEGs include KLF4 transcriptional targets. KLF4 can both activate and repress genes. In the Ift20^−/− endocardium (high KLF4), we observed increased levels of KLF4-activated genes, such as Lama3 (encoding a laminin subunit)⁴³, Malat1 (encoding metastasis-associated lung adenocarcinoma transcript 1)⁴⁴, and Vegfa and Vegfc (encoding vascular endothelial growth factors)⁴⁵, as well as decreased levels of KLF4-repressed genes, such as Col3a1 (encoding collagen)⁴⁶, Pdgfra (encoding a platelet-derived growth factor receptor)⁴⁷, Wnt4 (ref. ⁴⁸) and Cdkn1c (p57, encoding the cyclin-dependent kinase inhibitor)⁴⁹. Although nascent Klf4 was expressed at low levels and therefore not detected as a DEG by snRNA-seq, these differentially expressed KLF4 targets are consistent with the increased KLF4 expression in cilia KOs that we previously observed through hybridization chain reaction fluorescence in situ hybridization (HCR-FISH) and immunofluorescence analysis.

Using our WT EndoMT pseudostage markers and gene set enrichment (Fig. 7e,f), we identified the extent to which cilia KO embryos can progress through EndoMT. We validated these pseudostages with Slingshot and found that Ift20^−/− embryos did not progress similarly to control littermate embryos (Fig. 7g,h). In Ift20^−/− embryos, the proportion of cells in mid EndoMT was 2.13 times higher than in WT and heterozygous littermates, while the proportion of cells in late EndoMT was 6.1 times lower (Fig. 7i). This indicates that the cilia KO endocardium has difficulty progressing to the final pseudostage of EndoMT. This correlates with the timing of Klf4 downregulation in WT hearts and suggests that Klf4 downregulation is essential for the completion of EndoMT. We evaluated whether the decreased proportion of late EndoMT cells in cilia KOs was specific to EndoMT progression rather than caused by nonspecific changes in cell number, such as increased apoptosis or reduced proliferation within the vEC/EndoMT cluster. The expression of proapoptotic markers (Trp53, Casp3, Casp7) and the proportion of G2M cells were not significantly different in the Ift20^−/− vEC/EndoMT cluster (Extended Data Fig. 10a,b). This indicates that the loss of cilia inhibits EndoMT progression by affecting cell-fate trajectories.

As we previously determined that blood flow appears to act upstream of the ciliary regulation of KLF4, we next sought to investigate whether embryos lacking contractility experience similar difficulties in EndoMT progression at a transcriptional level. We performed snRNA-seq on microdissected e9.5 nonbeating hearts from somite-matched Ncx1^−/− and sibling (heterozygous and WT) embryos as described above for Ift20^−/− hearts. This dataset contained 36,296 single nuclei with 27,998 features (nuclei counts: Ncx1^+/+, 12,763; Ncx1^+/−, 12,030; Ncx1^−/−, 11,503) (Extended Data Fig. 8a,b). Similar to the Ift20 dataset, we were able to identify the main cell identities of the heart and found minimal differences in the major types (Extended Data Figs. 8c,d and 9a,b). However, there were differences in the endocardial subcluster, with fewer Ncx1^−/− nuclei in the EndoMT cluster compared to littermate controls (Extended Data Figs. 8e,f and 9c,d). Further subclustering of the vEC/EndoMT groups revealed a more severe EndoMT defect in Ncx1^−/− hearts than that observed in Ift20^−/− hearts (Extended Data Fig. 8g–i). While Ift20^−/− hearts can initiate but not complete EndoMT, Ncx1^−/− hearts stall at entry into early EndoMT (on average 15.37-fold higher than littermates at the vEC stage and 2.74-fold lower at early EndoMT). This finding was determined not to be due to differences in cell cycle or apoptotic marker gene expression between genotypes (Extended Data Fig. 8j,k). This suggests that primary cilia function as one, but not the only, flow sensor during ECC formation, consistent with the observation that KLF4 expression in Ift20^−/−/Ncx1^−/− hearts reflects that observed in Ncx1^−/− hearts.

EndoMT molecular pathways are dysregulated in cilia and ciliary calcium signaling KO hearts

To understand why the cilia KO endocardium fails to progress to late EndoMT, we identified DEGs in the vEC/EndoMT clusters of Ift20^−/− hearts compared to those in littermates. Upregulated DEGs (average log₂(fold change) of ≥0.4, adjusted P value of <0.05) in Ift20^−/− hearts were predominantly expressed in vECs and during mid EndoMT, including genes related to vasculature development, such as Tmem100 (encoding a transmembrane protein), Slc2a1 (encoding a glucose transporter) and Ldb2I (encoding a LIM domain-binding protein). In contrast, downregulated DEGs (average log₂(fold change) of −0.4 or lower, adjusted P value of <0.05) were expressed in late EndoMT and included migration-regulatory genes such as Robo2 (encoding a roundabout guidance receptor), Epha7 (encoding an ephrin receptor) and Erbb4 (encoding the Erb-B2 receptor tyrosine kinase) (Extended Data Fig. 10c). Additionally, Ift20^−/− hearts exhibited decreased expression of mesenchymal markers (Sox5, Vcan and Ncam1) and EndoMT markers (Snai1, Twist1 and Tbx20), along with increased expression of endothelial markers (Emcn, Vwf and Kdr) and EndoMT antagonist markers (Vegfa) (Fig. 8a and Supplementary Table 6). A subset of the DEGs was validated in vivo using HCR-FISH (Extended Data Fig. 10d,e). Gene set enrichment analysis of the top DEGs showed downregulation of Gene Ontology biological processes terms related to mesenchymal differentiation and cushion/valve development (30/166) in Ift20^−/− hearts, whereas terms related to endothelial/vascular identity/integrity (33/140) were upregulated (Fig. 8b). These same Gene Ontology terms were identified in Ncx1^−/− hearts, in addition to terms related to response to laminar fluid shear stress (as a positive control of the Ncx1^−/− phenotype) (Extended Data Fig. 9e). Further, the SMAD2/3/4 transforming growth factor-β (TGFβ) pathway required for EndoMT was perturbed in cilia KOs, as SMAD2/3/4 transcriptional targets were downregulated in the Ift20^−/− vEC/EndoMT cluster but not in other cell types such as epicardial cells (Extended Data Fig. 10f), demonstrating the inability of the Ift20^−/− endocardium to adopt a mesenchymal identity during EndoMT.

To determine the transcriptional changes driving the abnormal transition between mid and late EndoMT in Ift20^−/− hearts, we identified genes that Ift20^−/− hearts were unable upregulate or downregulate between each pseudostage of EndoMT compared to WT hearts (Supplementary Table 7). Gene set enrichment analysis of these gene networks showed that Ift20^−/− hearts exhibited upregulation of cell–cell/adherens junction genes (#1 Gene Ontology cellular components, adjusted P value: 2.589 × 10⁻⁸) and downregulation of collagen synthesis genes (#1 Gene Ontology cellular components, adjusted P value: 4.88 × 10⁻⁶) between mid and late EndoMT. This suggests that Ift20^−/− AVC endocardial cells have reduced ability to separate from adjacent cells and migrate into the cushion, providing a possible explanation for the lack of cellularization in cilia KO cushions.

As the KLF4 protein was found to be overactive in the Ift20^−/− endocardium and is involved in SMAD2/3/4 signaling³⁵, we evaluated whether the transcriptional changes in the vEC/EndoMT cluster of Ift20^−/− hearts were due to abnormally sustained KLF4 expression. We assessed the potential impact of KLF4 on the DEGs between the WT and Ift20^−/− endocardium undergoing EndoMT, using the TRRUST 2019 bioinformatics tool, which is currently the largest database of literature-curated transcription factor–target interactions⁵⁰. KLF4 was a significant hit (adjusted P value: 0.016) for DEGs that were higher in the Ift20^−/− vEC/EndoMT cluster, indicating that KLF4 is a key contributor to the transcriptional landscape in Ift20^−/− EndoMT. To solidify this conclusion, we assessed enrichment for the KLF4 binding motif (KRRRVWGGGTGKGGC) in the DEGs using motif enrichment analysis from MEME Suite. This analysis revealed that DEGs in the Ift20^−/− vEC/EndoMT cluster disproportionately possessed the KLF4 binding motif compared to a shuffled control sequence (enrichment ratio of 5.27 for higher DEGs (P = 2 × 10⁻⁹) and 1.82 for lower DEGs (P = 0.003)). This is consistent with a model in which the EndoMT defects observed in Ift20^−/− hearts are driven by the sustained expression and/or overexpression of KLF4 in the endocardium.

We next wanted to explore possible mechanisms linking primary cilia to KLF4 expression. It is well established that KLF2 expression can be dampened through the loss of TRP (transient receptor potential) channel calcium signaling, indicating a role for ciliary-localized calcium signals upstream of mechanosensitive transcription^{8,18,51,52,53}. In endothelial cells, mutations in either polycystin-1 (Pkd1) or Pkd2 (TRPP2), which form a calcium channel complex on the cilium, result in a loss of calcium influx to the same degree as that caused by a complete loss of primary cilia⁵⁴. Pkd2^−/− embryos are viable until e13.5 and exhibit ciliopathy-associated defects, such as atrioventricular septal defects and cystic kidneys, as well as decreased cushion mesenchyme¹⁴. Similar to Dnah11^−/− hearts, Pkd2^−/− embryos exhibit laterality defects due to the requirement for ciliary calcium signaling at the LRO²²; however, they display equivalent levels of endocardial ciliation compared to WT littermate embryos, indicating normal deciliation (Extended Data Fig. 10g). Although there was no decrease in KLF4-positive endocardial cells at e8.5 in Pkd2^−/− embryos, as we had previously observed in cilia KOs, KLF4-positive endocardial cells increased in Pkd2^−/− hearts by e9, even surpassing those in cilia KO embryos (Fig. 8c,d). By e9.5, Pkd2^−/− hearts had equivalent amounts of KLF4-positive endocardial cells compared to cilia KO embryos. Pkd2^−/− embryos had decreased AVC cellularization of endocardial origin compared to WT or Dnah11^−/− hearts, but not to the same degree as in cilia KO embryos (Fig. 8e). Together, these data indicate that KLF4 at the AVC is, at least in part, regulated through PKD2-mediated ciliary calcium.

Discussion

We used mice with defective cilia or cardiac contractility to show that intracardiac primary cilia transiently sense and integrate blood flow to spatially control EndoMT in early cardiac cushions through endocardial KLF4 regulation. We found endocardial primary cilia extending into the cardiac lumen, which become spatially regionalized to the ventricular–ECC junctions in a blood flow-dependent manner as development progresses. KLF4 expression spatially correlates with endocardial ciliation during cushion development. Mouse embryos that constitutively lack cilia overexpress and misexpress KLF4 and are unable to cellularize their ECCs, independent of the left–right defects associated with cilia loss. Comparison of snRNA-seq datasets from microdissected WT and cilia KO mouse hearts demonstrated that the cilia KO endocardium was unable to progress from mid to late EndoMT and had misregulated KLF4 targets.

Primary cilia are excellent candidate ECC mechanosensors due to their architecture, their role as signaling hubs and their ability to be endogenously removed. Blood flow-dependent WSS is the predominant force in the AVC and OFT^18,37. At the start of contraction, mouse embryonic hearts experience shear stress averaging <2–5 dyn cm⁻² (ref. ⁵⁵), with increments as low as approximately 1.2 dyn cm⁻² between the center and edge of the AVC as development progresses³⁸. This is smaller than the shear sensitivity of NOTCH1 (>10 dyn cm⁻²)⁵⁶ or ion channels/PIEZO1 (5–10 dyn cm⁻²)^57,58. In contrast, primary cilia can detect shear forces as low as 0.2 dyn cm⁻² (ref. ⁵⁹) and elicit a calcium signal in response to 0.5 dyn cm⁻² (ref. ²²). Intracardiac primary cilia are biased toward areas of lower shear stress, adding an additional layer of control in WSS-mediated genetic outcomes⁶⁰. KLF4 expression and ciliation in the OFT cushions (conal and truncal) resembled the pattern observed at the AVC but lagged behind, possibly due to differential levels of shear stress, geometry and the contribution of non-endocardial-derived cells⁶¹. Primary cilia disassemble at approximately 15 dyn cm⁻² (ref. ¹²), and the AVC center during late cushion formation in zebrafish and avian hearts can experience an average of approximately 30 dyn cm⁻² while the edges experience an average of approximately 10 dyn cm⁻² (refs. ^18,37,62). Consistent with this, we found that endocardial primary cilia were maintained on the ECC edges and reduced in the ECC centers, showing a correlation between shear stress and ciliation. This pattern was absent in mouse mutants lacking contractility, indicating that blood flow is necessary to establish endocardial ciliary regionalization.

For cilia to act as cardiac mechanosensors, they need not only to sense flow but also to translate it into gene regulation. The idea that primary cilia promote mechanosensitive factors is not new, as previous work has proposed that vECs can regulate Klf2 (ref. ⁶⁰). Additionally, primary cilia were previously found to sensitize endocardial cells for mechanosensitive gene expression in vitro⁶³. Our data show that, in vivo, constitutive cilia KOs cannot upregulate KLF4 at early time points, while at later time points (when WSS is higher), cilia KOs exhibit KLF4 overexpression. Further, the active loss of cilia in response to high WSS in WT hearts correlates with KLF4 downregulation. These observations suggest a mechanism by which cilia can interpret a gradient of mechanical forces and translate them into developmentally relevant signaling: first by sensing low shear stress and upregulating KLF4, and second by high-WSS-mediated deciliation and KLF4 downregulation. Our study validates and expands on previous work⁶⁰ by using newer technologies to establish a higher level of detail in the spatiotemporal dynamics between endocardial primary cilia and downstream mechanosensitive genes.

WSS-dependent cytoskeletal deformation is necessary for ERK5 phosphorylation upstream of Klf4 (ref. ⁶⁴). Previous studies have postulated that primary cilia amplify mechanical forces on the cytoskeleton in low-flow environments, allowing gene expression upregulation before the WSS levels are strong enough to deform the cytoskeleton alone (10 dyn cm⁻²)^11,63,65. In this model, primary cilia act as cytoskeletal levers to amplify low WSS and induce Klf4 during the early stages of cushion development, but they are unnecessary at later stages when WSS is higher. Further, TGFβ signaling is known to inhibit KLF4 in vivo and in vitro⁶⁶. TGFβ receptors 1 and 2 localize to cilia, and the constitutive loss of cilia results in decreased SMAD2/3/4 signaling⁶⁷, which we confirmed in the EndoMT cluster of cilia KO hearts. Therefore, a mechanism by which the constitutive loss of cilia leads to lower TGFβ-dependent KLF4 inhibition seems plausible. This could also be tightly regulated temporally, providing a possible explanation for the subsequent reestablishment of KLF4 expression after EndoMT that is not spatially controlled by ciliation^29,30.

The question remains: what distinguishes KLF4 expression after the endogenous loss of cilia due to shear stress, such as in WT hearts, from that in cilia KO hearts that intrinsically do not possess a cilium? Prior studies show that a mechanically driven loss of primary cilia accelerates epithelial-to-myofibroblast transition through TGFβ signaling⁶⁸ and that selective cilia loss in high-shear areas regulates bone morphogenic protein signaling during vascular regression⁶⁹. In both examples, the presence and subsequent loss of primary cilia are necessary for regional signaling. Our results support the function of primary cilia as mediators of shear forces that are transient by design, directing KLF4 patterning through selective regional deciliation and the resulting TGFβ. Mutants that intrinsically do not possess cilia are unable to activate TGFβ in the endocardium undergoing EndoMT, resulting in the overexpression and mislocalization of KLF4.

Other mechanisms may involve ciliary calcium signaling, as we observed a similar overexpression and misexpression of KLF4 in Pkd2^−/− hearts. KLF4 upregulation has been found to be affected by the loss of cytosolic calcium⁴⁹, possibly providing an additional method for KLF4 downregulation upon deciliation and the loss of ciliary-mediated calcium influx. Interestingly, instead of observing a decrease in KLF4 in Pkd2^−/− embryos, we noted an overabundance as WSS increased. This may be because decreases in cytosolic calcium due to Pkd2 mutations can enhance cAMP, causing KLF4 upregulation and stability⁷⁰.

Whether KLF4 is an EndoMT activator or inhibitor remains unclear^31,32. By investigating the spatial localization of genetic and morphological EndoMT markers, we found a positive correlation between areas of active EndoMT and KLF4-low regions, as well as a negative correlation between Klf4 expression and multiple essential EndoMT genes, suggesting that KLF4 has an inhibitory role in cushion EndoMT. This is consistent with a model in which KLF4 overexpression in cilia KOs leads to their observed EndoMT defect. We confirmed this KLF4 abundance transcriptionally in cilia KOs by observing upregulated KLF4 targets. Functionally, KLF4 not only inhibits multiple essential EndoMT genes, but it can also promote antagonistic EndoMT factors such as VEGF signaling⁴⁵. We observed increased expression of the VEGF pathway members Vegfc and Vegfa in cilia KO hearts, as well as increased enrichment of VEGF signaling-related Gene Ontology biological processes terms. Collectively, these results indicate that abnormal KLF4 overexpression stalls EndoMT progression in cilia KO hearts. Our data show a more severe EndoMT phenotype in the contractility-deficient mutants compared to the cilia-deficient mutants, indicating that primary cilia are not the only mechanosensors involved in EndoMT and are specific to a certain stage of the transition but not for initiation. There are many proposed mechanosensors and mechanosensitive signaling cascades that can trigger EndoMT, including NOTCH and ALK5 (ref. ⁷¹). However, the spatiotemporal patterning of KLF4 during early cushion formation is a unique pattern that requires cilia due to their ability to be regionally disassembled.

As cilia mechanosensation at the LRO is essential for establishing left–right asymmetry and the direction of the heart loop, it is unknown whether the wide range of cardiac defects observed in mice and humans with abnormal cilia are due to a primary requirement for intracardiac cilia or a secondary manifestation of abnormal blood flow due to defective heart looping. Dnah11^−/− mouse hearts, which have normal intracardiac cilia but exhibit randomized heart looping similar to the cilia KO hearts, were indistinguishable from WT with respect to the number or location of KLF4-positive cells, endocardial cilia or cushion cellularization. These data support a model in which the loss of KLF4 regulation and cushion cellularization observed in cilia KO hearts is due to a specific role of intracardiac primary cilia during cushion development, as opposed to upstream functions in left–right asymmetry.

Our investigation of KLF4 as a mechanosensitive transcription factor during valve development was prompted by the discovery of KLF4 variants in patients with a particularly severe form of CHD-associated valve disease, hypoplastic left heart syndrome, which is characterized by severely abnormal or absent aortic and mitral valves²⁴. Further support for a cilia–KLF4 axis in human valve development is provided by the association of another congenital valve abnormality, AVC defect, which is characterized by the failure to form separate mitral and tricuspid valves, with ciliopathies caused by abnormal ciliogenesis and function. It is interesting to speculate that cilia function as cardiac mechanosensors throughout the lifespan. Identifying patients with cilia defects underlying their valve disease may provide an approach to risk stratification and potential pharmacological mitigation.

Taken together, our data suggest that primary cilia integrate blood flow and the regulation of mechanosensitive factors to spatially control cushion growth in vivo. Cilia-dependent KLF4 expression maintains endocardial identity, as has been previously observed in lung fibrosis³², thus providing an endocardial pool for replenishing differentiated luminal cells. The ciliated endocardium becomes Klf4-positive following the onset of blood flow and then selectively deciliates in the high-shear-stress center of the ECCs, leading to reduced KLF4 (Fig. 8f). No longer inhibited by KLF4, cells at the ECC center can undergo EndoMT but must be replaced by neighboring luminal endocardial cells as they migrate into the cardiac jelly from the lumen. These neighboring cells will then be subjected to the higher shear stress in the center of the AVC, causing their own deciliation and restarting the process. We call this the ‘churning’ model of ECC formation. First, it provides a method by which a large quantity of endocardial cells can migrate to and rapidly cellularize the cushion. Second, it provides a mechanism by which the spatial identity of the ECC can be maintained through a positive mechanosensitive feedback loop during development: high shear stress leads to regional deciliation, downregulating KLF4 and promoting EndoMT, thereby increasing the thickness of the AVC, decreasing the diameter of the lumen and increasing shear stress, as the same amount of blood must now flow through a smaller orifice (Fig. 8f).

Methods

Experimental model details

Mice

Dnah11^−/− (lrd^{GFPΔneo/GFPΔneo})²³ embryos were obtained by crossing homozygous female and male mice. Ift20^−/− (ref. ⁷²), Kif3a^−/− (ref. ²¹), Ncx1^−/− (ref. ¹⁹) and Pkd2^−/− (ref. ⁷³) embryos were obtained by crossing heterozygous female and male mice. We generated Ift20^+/−/Ncx1^+/− mice that were viable and exhibited no phenotype and crossed a female and male to obtain Ift20^−/−/Ncx1^−/− embryos. All the lines used in this study were maintained on the C57BL/6 background, regularly outcrossed to fresh C57BL/6 WT and fed breeder diets. Timed matings were conducted for each strain, with noon on the day of vaginal plug detection considered e0.5. Plugged females were weighed before the day of dissection to confirm pregnancy. Embryonic staging was determined by somite count (6–8 somites for e8, 11–14 somites for e8.5, 15–19 somites for e9, 20–25 somites for e9.5 and 26–29 somites for e10). Both sexes were pooled for each embryonic experiment. For all mutant line analyses, equivalent numbers of randomized C57BL/6 somite-matched WTs and heterozygotes were taken from a littermate pool of each line as controls, unless otherwise stated in the figure legend (‘for comparison, the WT line is provided’). This research complies with all ethical regulations; mouse experiments were performed in a manner approved by the Yale University Institutional Animal Care and Use Committee (IACUC protocol no. 2022-08012). Mice were maintained on a 12-h light–dark cycle at 72 ± 2 °F and 50% ± 20% humidity.

Zebrafish

Tg(Sco:eGFP;Kdrl:mCherry;cmlc2:eGFP)^74,75 zebrafish were obtained from crossing homozygous transgenic females and males. Embryos were obtained through natural spawning. All zebrafish experiments were conducted according to the Yale Animal Resources Center and Institutional Animal Care and Use Committee guidelines (IACUC protocol no. 2021-10778).

Histology

Embryos were dissected in cold 1× PBS. A sample of the allantois was taken for genotyping. Embryonic somites were then counted, and embryos were fixed in 4% paraformaldehyde in 1× PBS overnight (16 h) at 4 °C. Embryos were then prepared for cryosectioning. Briefly, embryos were saturated with 30% sucrose, mounted in 10 × 10 × 5 mm cryomolds filled with OCT medium (Tissue-Tek) and frozen in chilled 100% ethanol. Following freezing, blocks were sliced into 12-µm sections using a Leica CM3050 S Research cryostat and mounted on Superfrost Plus slides (Thermo Fisher). Defrosted slides were washed with 1× PBS before being counterstained with Nuclear Fast Red solution (Ricca) for 5 min. They were then rinsed with water and dehydrated before being mounted with CytoSeal (Thermo Fisher) on a coverslip. The slides were then imaged on a Zeiss Axiovert microscope using an Axiocam 1077 driven by Axiovision software.

Whole-mount and cryosection in situ hybridization

Colorimetric

Digoxigenin (DIG)-labeled antisense probes were prepared using SP6 and T7 RNA polymerase (NEB) from linearized expression plasmids. pCX-Klf2 and pCX-Klf4 were gifts from B. Cohen (Addgene plasmid nos. 66655 and 66656, respectively). For linearization and transcription, EcoRV and SP6 were used for pCX-Klf2 and HindIII and T7 were used for pCX-Klf4. Embryos were dehydrated in a methanol series, rehydrated, permeabilized and incubated overnight in hybridization buffer with probe (~1 μg ml⁻¹) at 68 °C. The next day, the embryos underwent a series of washes in 1× maleic acid buffer with Tween (MABTw), blocked for 2 h in 2% blocking reagent (Roche) in MABTw with 20% normal goat serum, and incubated overnight in anti-DIG antibody (1:2,000, Roche) at 4 °C. The next day, embryos were washed in MABTw, followed by NTMT (pH 9.5), and then developed in BM Purple solution (Roche) containing 2 mM levamisole and 1% Tween. Once the reaction was complete, development was stopped with PBSTw. Whole embryos were mounted in 1× PBS and imaged using a Nikon SMZ 745T dissection scope equipped with an Excelis HDS HD camera and monitor system. Embryos were then prepared for cryosectioning and counterstaining with Nuclear Fast Red (Ricca) according to the standard product protocol. Slides were imaged on a Zeiss Axiovert microscope using an Axiocam 1077 driven by Axiovision software.

HCR-FISH

HCR oligonucleotide probes were synthesized for Klf2, Klf4, Vim, Kdr, Vcan and Emcn by Molecular Instruments (see Supplementary Table 14 for more information). The standard company protocol for whole-mount e9.5 mouse embryos was followed, with the following change: ProK digestion (step 12) was performed for 10 min. Following the last wash with 1× SSCTw, the embryos were incubated for 20 min with a 1:2,000 dilution of Hoechst 33342 before another 1× SSCTw wash. Embryos were stored protected from light in 1× SSCTw at 4 °C until they were imaged using a Zeiss Z.1 lightsheet microscope. Imaged embryos were collected again and frozen in OCT medium (Tissue-Tek) for later cryosectioning. Defrosted slides were rinsed with 1× PBS and mounted on coverslips with ProLong Gold antifade reagent (Invitrogen, P36935). Sections were examined using a Zeiss Axiovert microscope equipped with Apotome imaging.

Immunofluorescence

Whole mount

Embryos were dissected in cold 1× PBS. A sample of the allantois was taken for genotyping. Embryonic somites were then counted, and the embryos were fixed in 4% paraformaldehyde in 1× PBS overnight (16 h) at 4 °C. The next day, embryos were briefly washed with 1× PBS and then subjected to an ethanol dehydration and rehydration series (25%, 50%, 75%, 100% at −20 °C for 30 min, 75% back at room temperature, 50%, 25%, 0%). Embryos were then further permeabilized with two washes of 1× PBS containing 1% Triton-X (1× PBSTx), followed by blocking for >2 h (3% BSA in 1× PBSTx). Embryos were then incubated in blocking solution with the primary antibody overnight at 4 °C. The next day, embryos were washed four times for 15 min each with 1× PBSTx before incubating with a secondary antibody in the blocking solution for 2 h. Embryos were then rinsed three times for 15 min each in 1× PBS and incubated for 20 min with a 1:2,000 dilution of Hoechst 33342 (Thermo Fisher) before being rinsed again. Whole embryos were kept in 1× PBS at 4 °C while protected from light until they were imaged on a Zeiss Z.1 lightsheet microscope and a Zeiss 880 confocal microscope.

Cryosections

Embryos were embedded in OCT medium (Tissue-Tek) and sectioned using a cryostat microtome. Defrosted slides were then subjected to the same protocol as whole-mount immunofluorescence, except for the slides labeled with anti-γ-tubulin, which were postfixed in 100% methanol at −20 °C for 10 min. After the final rinse with 1× PBS, the slides were mounted on coverslips with ProLong Gold antifade reagent (Invitrogen, P36935). Sections were examined using a Zeiss Axiovert microscope equipped with Apotome imaging.

Cell death assay

Cell death was measured using the Click-IT Plus TUNEL Assay 594 kit (Invitrogen, cat. no. C10618). The following changes were made to the standard protocol provided for this kit: step 2.4 (incubation) was performed at 37 °C, step 2.6 (postfixation) was performed at room temperature, and 0.1% Triton-X was added to the PBS in step 2.7.

Antibodies

The following antibodies were used: mouse anti-ARL13B (1:200, NeuroMab), rabbit anti-ARL13B (1:200, ProteinTech), goat anti-KLF4 polyclonal (1:100, R&D Systems), rat anti-CD31 (1:200, BD Biosciences), rabbit anti-FN (1:200, Sigma Aldrich), Alexa 488 anti-mouse (1:500, Invitrogen), Alexa 488 anti-rabbit (1:500, Invitrogen), Alexa 594 anti-goat (1:500, Invitrogen), Alexa 647 anti-rat (1:500, Invitrogen), Hoechst 33342 (1:2,000, Thermo Fisher), sheep anti-DIG-AP (1:2,000, Roche), mouse anti-PCNA (1:250, ProteinTech) and rabbit anti-γ-tubulin (1:100, Sigma). See Supplementary Table 8 for the catalog and RRID numbers.

Lightsheet microscopy

Fixed

Embryos were embedded in Zeiss glass capillaries suited for their size (blue for e9.5–e10 mice, black for e8–e9 mice and zebrafish) in 1% purified low-melt agarose (Thermo Fisher). The chamber was filled with filtered 1× PBS before imaging on a Zeiss Z.1 lightsheet microscope.

Live

At 24 h after fertilization, 1× PTU (1-phenyl 2-thiourea, Sigma) was added to the zebrafish medium to block pigment formation. Zebrafish embryos were then screened for fluorescence signals using a Leica fluorescence stereomicroscope. Before embedding in glass capillaries, tricaine (Sigma) was added to the embryo medium at a concentration of 168 μg ml⁻¹ to immobilize the embryos. The chamber was filled with embryo medium containing tricaine and heated to 26 °C before imaging. The embryos were then embedded in glass capillaries and imaged as previously described⁷⁶. Briefly, 200 heartbeats were recorded for every 3 μm of a z-stack encompassing the entire heart at the maximum frame rate. Embryos were imaged every 12 h between 36 and 96 h after fertilization.

Imaging parameters and analysis

Microscope details

Images of cryosections were obtained using a Zeiss Axiovert microscope equipped with Apotome optical interference and a 20×, 40× or 63× objective. Images acquired at 20× and used as overviews were single plane, whereas 40× and 63× images were obtained as z-stacks of the entire 12-μm cryosection and represented with maximum intensity projections. The W Plan-Apochromat 20×/1.0 NA objective was used on the Zeiss Z.1 lightsheet microscope for whole-mount imaging. All whole-mount images were obtained through z-stack imaging of the entire region and were visualized either with maximum intensity projections or through three-dimensional (3D) rendering using FluoRender. The following software versions were used: Axiovert-ZEN (blue edition) v.3.3.89.0002 and Lightsheet-ZEN (black edition) v.2.0.

Imaging details

The entire AVC or OFT was acquired for z-stacks of cryosectioned or whole-mount embryos, starting at the dorsal aorta and ending at the gut. Roughly, we found an average of 116 AVC cells and 135 OFT cells at e8.5, 269 AVC cells and 262 OFT cells at e9.0, 435 AVC cells and 338 OFT cells at e9.5, and 756 AVC cells and 779 OFT cells at e10.

Analysis details

Cell, cilia, TUNEL and PCNA quantification was performed using CellProfiler (v.4.2.8)⁷⁷ (RRID:SCR_007358), Imaris Software (RRID:SCR_007370) and Fiji⁷⁸ (RRID:SCR_003070). Cilia were identified using the 3D Object Counter plug-in in Fiji, applying a relative threshold for the ARL13B signal and object size. This resulted in the expected one ciliary object per cell body. Similarly, KLF4-positive cells and endocardial nuclei were identified by creating a region of interest around CD31-positive areas and using the 3D Object Counter plug-in in Fiji with a relative threshold for the KLF4 or Hoechst signal and object size. The KLF4 signal was confirmed to overlap with Hoechst-defined objects. CellProfiler was used to confirm select slides by identifying primary objects for cilia, KLF4-positive nuclei and nuclei and relating the objects to find the number of daughter cilia per nucleus. For distance analysis, objects for cilia, KLF4 and nuclei were counted in discrete intervals spanning the AVC or OFT in whole-mount samples. The distance was then normalized between samples as a percentage of the total AVC or OFT length. Four-dimensional analysis of live zebrafish used Fiji to create hyperstacks for one cardiac cycle and FluoRender (v.2.26.3)⁷⁹ (RRID:SCR_014303) for rendering and figure and video creation. HCR quantification was done using Fiji’s ‘Measure’ and ‘Plot Profile’ functions. To compare replicates with various percentage lengths, we used MATLAB to automatically fill in the points between the experimentally provided points.

snRNA-seq

Single-nucleus collection, sequencing and alignment

e9.5 embryos were dissected in cold 1× PBS with diethyl pyrocarbonate. A sample of the allantois was taken for genotyping. Whole hearts were removed, allowed to self-perfuse and flash-frozen in liquid nitrogen. These were kept frozen in liquid nitrogen until processing. Pools of five embryonic hearts were used per genotype, with one replicate for Ncx1^−/−, Ncx1^+/− and Ncx1^+/+, two replicates for the Ift20^−/− and Ift20^+/− genotypes, and three replicates for Ift20^+/+. Briefly, RNA quality was assessed using the Agilent TapeStation with an RNA integrity number cutoff of ≥5. The sample was then homogenized in buffer (250 mM sucrose, 25 mM KCl, 5 mM MgCl₂, 10 mM Tris–HCl, 1 mM dithiothreitol, 1× protease inhibitor) using the TissueLyser II instrument (Qiagen) and filtered through a 40-μm strainer (Corning). The nuclei were then centrifuged at 1,200g for 5 min at 4 °C and resuspended in storage buffer (1× PBS, 4% BSA, 0.2 U μl⁻¹ Protector RNase inhibitor). Nuclei were sorted by fluorescence-activated cell sorting (Aria, BD Biosciences) using the NucBlue Live ReadyProbes reagent (Thermo Fisher) and assessed for morphology, size and concentration with a Countess automated cell counter (Thermo Fisher). Intact nuclei were fractionated using the 10X Genomics Chromium iX instrument according to the manufacturer’s protocol, with a targeted nuclei recovery of 10,000 per run. The 10X Genomics Chromium Single Cell Reagent Kit (v.3.1) was used to generate 3′ gene expression libraries. Libraries were sequenced on the NovaSeq 6000 platform (Illumina) with a minimum depth of 30,000–40,000 read pairs per nucleus. Reads were aligned to the mm10 mouse genome using Cell Ranger (v.3.0.1 for the Ift20 dataset and v.6.1.0 for the Ncx1 dataset) (RRID:SCR_017344).

Single-nucleus Seurat object creation

Doublet prediction was performed using Scrublet (v.0.2.3)⁸⁰ (RRID:SCR_018098) in Python (v.3.8.8)⁸¹ (RRID:SCR_008392) with the following parameter settings: expected_doublet_rate of 0.06, min_counts of 2, min_cells of 3, min_gene_variability_pctl of 85 and n_prin_comps of 30. Data files were read into R (v.4.1.2) with Seurat (v.4.3.0)⁸² (RRID:SCR_016341) and filtered based on the Scrublet score (Ift20 dataset (WT 1–3: 0.2, 0.2, 0.15; heterozygous 1–2: 0.3, 0.3; mutant 1–2: 0.3, 0.2), Ncx1 dataset (WT: 0.18; heterozygous: 0.25; mutant: 0.21)). Cutoff scores were determined by the bimodal distribution in the observed transcriptome graph created by Scrublet. Seurat objects were then merged by dataset, log-normalized (1 × 10⁴) and examined for variable features (‘vst’, 2,500 features). The resulting Seurat object was then filtered based on the number of counts (>300, <15,000), the number of features (>300, <5,000) and the percentage of mitochondrial reads (<0.02). Principal component analysis was performed using the Seurat (v.4.3.0) runPCA function, with the identified variable features excluding mitochondrial and ribosomal genes. The object then underwent Harmony (v0.1.1)⁸³ (RRID:SCR_022206) integration, and the ElbowPlot function in Seurat was used to establish the ndims (number of dimensions = 50). The RunUMAP (dims = 1:50, reduction = ‘harmony’), FindNeighbors (dims = 1:50, reduction = ‘harmony’) and FindClusters (resolution = 0.1 Ift20, 0.15 Ncx1) functions were then used for clustering. The FindClusters resolution was determined by clearly separated UMAP (Uniform Manifold Approximation and Projection) dimensions and at least 20 DEGs between each cluster. The resulting Ift20 object contained 29,313 single nuclei with 27,998 features (nuclei counts: Ift20^+/+, 9,882; Ift20^+/−, 14,825; Ift20^−/−, 6,116). The resulting Ncx1 object contained 36,296 single nuclei with 27,998 features (nuclei counts: Ncx1^+/+, 12,763; Ncx1^+/−, 12,030; Ncx1^−/−, 11,503).

Single-nucleus analysis

To find differential gene expression across all clusters, we used the FindAllMarkers function in Seurat (only.pos = TRUE, min.pct = 0.25). For differential gene expression between genotypes or specific clusters, we used the FindMarkers function in Seurat with default settings. For differential expression shown in dot plots, the presto v.1.0.0 function wilcoxauc was used and refined with the top_markers function (n = 10, auc_min = 0.5, pct_in_min = 20, pct_out_max = 20). Dot plots, violin plots and heat maps were produced using the Seurat functions DotPlot, VlnPlot and DoHeatmap to create ggplot2 v.3.5.0 objects with the default settings. Cell cycle analysis was conducted as previously described⁸⁴. To create the endocardial and vEC/EndoMT subclusters, we used the Seurat subset function, followed by FindVariableFeatures, ScaleData and RunPCA with default settings. The ElbowPlot function in Seurat was used to establish the ndims (20). RunUMAP (dims = 1:20), FindNeighbors (dims = 1:20) and FindClusters (resolution = 0.4 (Ift20, both subclusters), 0.5 (Ncx1, endocardial subcluster), 0.2 (Ncx1, vEC/EndoMT subcluster)) were then applied for clustering. To transfer anchors from the Ift20 dataset to the Ncx1 dataset for the endocardial subcluster, the FindTransferAnchors function in Seurat with dims = 1:20 was used. The AddModuleScore function in Seurat was used for gene set enrichment (endocardial genes: Nppa, Vwf, Plvap, Aplnr, Sox17, Hey1, Emcn, Ecscr, Cldn5, Gja5, Cdh5, Igfbp3, Flt1, Kdr, Tek, Tjp1, Notch1, Gja4, Npr3; EndoMT genes: Ncam1, Sox9, Sox5, Slit2, Snai1, Snai2, Twist1, Twist2, Hand2, Msx1, Tbx20, Gata4, Smad6, Ecm1, Aes, Mmp2, Tgfb2, Acta2, Vim, S100a4, Cdh2, C3, Col3a1, Bmper, Col1a2, Fn1, Ptgis, Timp3, Fbln1, Col1a1, Dpt, Bnc2, Pkhd1l1; mesenchymal genes: Mmp2, Sox9, Acta2, Col1a1, Col1a2, Vcam1, Icam1, Zeb2, Mmp9, Pdgfra, Vcan, Col3a1, Snai1, Snai2, Twist1, Twist2, Msx1). Gene Ontology analysis (biological processes and cellular compartments) and TRRUST 2019 enrichment⁵⁰ were done using Enrichr⁸⁵ (RRID:SCR_001575). Pseudotime was calculated using Slingshot (v.2.12.0)³⁴ (RRID:SCR_017012) with default parameters. Statistics for box plots were calculated with rstatix v.0.7.2.999 using a t test with Bonferroni correction. Statistics for violin plots were calculated with ggpubr v0.6.0 using Student’s t test. Motif enrichment analysis was performed with MEME Suite (v.5.5.5)⁸⁶ Simple Enrichment Analysis (RRID:SCR_001783). P values are represented as follows: no significance, P > 0.05; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001.

Single-cell analysis

For the single-cell analysis from de Soysa et al.²⁷, the expression matrix and metadata files were downloaded for the ‘early mouse cardiogenesis’ and ‘endocardial/endothelial’ subclusters from the UCSC (University of California, Santa Cruz) Cell Browser⁸⁷. The endocardial/endothelial subcluster was filtered to remove hematoendothelial cells and ages e7.75 and e8.25 (too young for EndoMT). The object was then processed (RunPCA, FindNeighbors (dims = 1:20), FindClusters (resolution = 0.5), RunUMAP (dims = 1:20)). The AddModuleScore function in Seurat was used for gene set enrichment. The vEC/EndoMT clusters were determined by gene set enrichment, subset and processed (ScaleData, RunPCA, FindNeighbors (dims = 1:20), FindClusters (resolution = 0.6), RunUMAP (dims = 1:20)). Pseudotime was calculated using Slingshot v.2.12.0 (RRID:SCR_017012) with default parameters. Statistics for violin plots were calculated with ggpubr v.0.6.0 using Student’s t test. P values are represented as follows: no significance, P > 0.05; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001.

Statistics and reproducibility

All statistical analyses outside of R were performed in GraphPad Prism v.9.2.0.332 for Windows (www.graphpad.com). Statistical details, such as the number of animals (n), can be found in the figure legends. Each experiment was repeated at least two times, with the exception of the Ncx1 snRNA-seq, in which five hearts were pooled per genotype with one run. Blinding was administered during data acquisition and data analysis of all samples, except for snRNA-seq due to the analysis parameters. Sample sizes for snRNA-seq experiments were chosen in agreement with guidelines for those analyses. Animal sample sizes were based on the data points needed for statistical rigor and reproducibility using a two-tailed t test. Graphs were created using the mean and s.e.m. for error bars. Simple linear regression was used to measure the R² values of xy graphs for the best fit. A two-tailed, one-sample t test was performed to measure reproducibility between biological replicates, and a two-tailed unpaired t test with Welch’s correction was conducted to assess variance between experimental groups. Statistics for snRNA-seq box plots were calculated with rstatix v.0.7.2.999 using a t test with Bonferroni correction. Statistics for snRNA-seq and scRNA-seq violin plots were calculated with ggpubr v.0.6.0 using Student’s t test. P values are represented as follows: no significance, P > 0.05; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All data supporting the findings of this study are available within the paper and its supplementary information files. The snRNA-seq data reported in this paper have been deposited in the Gene Expression Omnibus (GEO) under accession numbers GSE252341 (Ift20) and GSE293814 (Ncx1) and are publicly available. The scRNA-seq data from de Soysa et al.²⁷ are available in GEO under accession number GSE126128 and through the UCSC Cell Browser at https://mouse-cardiac.cells.ucsc.edu. The accession numbers for both datasets are listed in Supplementary Table 12. Source data are provided with this paper.

Code availability

This study did not generate original code. See ‘snRNA-seq’ in Methods and Supplementary Table 15 for published packages and analysis parameters used in this paper.

References

GBD 2017 Congenital Heart Disease Collaborators. Global, regional, and national burden of congenital heart disease, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet Child Adolesc. Health 4, 185–200 (2020).
Article Google Scholar
Ahuja, N., Ostwald, P., Bark, D. & Garrity, D. Biomechanical cues direct valvulogenesis. J. Cardiovasc. Dev. Dis. https://doi.org/10.3390/jcdd7020018 (2020).
Article PubMed PubMed Central Google Scholar
Umapathi, K. K. & Agasthi, P. Atrioventricular canal defects. StatPearls www.ncbi.nlm.nih.gov/books/NBK557511/ (2023).
Miao, Y. et al. Intrinsic endocardial defects contribute to hypoplastic left heart syndrome. Cell Stem Cell 27, 574–589 (2020).
Article CAS PubMed PubMed Central Google Scholar
Andrés-Delgado, L. & Mercader, N. Interplay between cardiac function and heart development. Biochim. Biophys. Acta 1863, 1707–1716 (2016).
Article PubMed PubMed Central Google Scholar
Bartman, T. et al. Early myocardial function affects endocardial cushion development in zebrafish. PLoS Biol. 2, E129 (2004).
Article PubMed PubMed Central Google Scholar
Huang, C. et al. Embryonic atrial function is essential for mouse embryogenesis, cardiac morphogenesis and angiogenesis. Development 130, 6111–6119 (2003).
Article CAS PubMed Google Scholar
Fukui, H. et al. Bioelectric signaling and the control of cardiac cell identity in response to mechanical forces. Science 374, 351–354 (2021).
Article CAS PubMed Google Scholar
Van der Heiden, K., Egorova, A. D., Poelmann, R. E., Wentzel, J. J. & Hierck, B. P. Role for primary cilia as flow detectors in the cardiovascular system. Int. Rev. Cell Mol. Biol. 290, 87–119 (2011).
Article PubMed Google Scholar
Slough, J., Cooney, L. & Brueckner, M. Monocilia in the embryonic mouse heart suggest a direct role for cilia in cardiac morphogenesis. Dev. Dyn. 237, 2304–2314 (2008).
Article PubMed PubMed Central Google Scholar
Goetz, J. G. et al. Endothelial cilia mediate low flow sensing during zebrafish vascular development. Cell Rep. 6, 799–808 (2014).
Article CAS PubMed Google Scholar
Iomini, C., Tejada, K., Mo, W., Vaananen, H. & Piperno, G. Primary cilia of human endothelial cells disassemble under laminar shear stress. J. Cell Biol. 164, 811–817 (2004).
Article CAS PubMed PubMed Central Google Scholar
Egorova, A. D. et al. Lack of primary cilia primes shear-induced endothelial-to-mesenchymal transition. Circ. Res. 108, 1093–1101 (2011).
Article CAS PubMed PubMed Central Google Scholar
Djenoune, L., Berg, K., Brueckner, M. & Yuan, S. A change of heart: new roles for cilia in cardiac development and disease. Nat. Rev. Cardiol. 19, 211–227 (2022).
Article PubMed Google Scholar
Li, X. et al. Primary cilia mediate Klf2-dependant Notch activation in regenerating heart. Protein Cell 11, 433–445 (2020).
Article CAS PubMed PubMed Central Google Scholar
Farhat, B. et al. Understanding the cell fate and behavior of progenitors at the origin of the mouse cardiac mitral valve. Dev. Cell 59, 339–350 (2024).
Article CAS PubMed Google Scholar
Toomer, K. A. et al. Primary cilia defects causing mitral valve prolapse. Sci. Transl. Med. https://doi.org/10.1126/scitranslmed.aax0290 (2019).
Article PubMed PubMed Central Google Scholar
Heckel, E. et al. Oscillatory flow modulates mechanosensitive klf2a expression through trpv4 and trpp2 during heart valve development. Curr. Biol. 25, 1354–1361 (2015).
Article CAS PubMed Google Scholar
Koushik, S. V. et al. Targeted inactivation of the sodium–calcium exchanger (Ncx1) results in the lack of a heartbeat and abnormal myofibrillar organization. FASEB J. 15, 1209–1211 (2001).
Article CAS PubMed Google Scholar
Follit, J. A., Tuft, R. A., Fogarty, K. E. & Pazour, G. J. The intraflagellar transport protein IFT20 is associated with the Golgi complex and is required for cilia assembly. Mol. Biol. Cell 17, 3781–3792 (2006).
Article CAS PubMed PubMed Central Google Scholar
Marszalek, J. R., Ruiz-Lozano, P., Roberts, E., Chien, K. R. & Goldstein, L. S. Situs inversus and embryonic ciliary morphogenesis defects in mouse mutants lacking the KIF3A subunit of kinesin-II. Proc. Natl Acad. Sci. USA 96, 5043–5048 (1999).
Article CAS PubMed PubMed Central Google Scholar
Djenoune, L. et al. Cilia function as calcium-mediated mechanosensors that instruct left–right asymmetry. Science 379, 71–78 (2023).
Article CAS PubMed PubMed Central Google Scholar
McGrath, J., Somlo, S., Makova, S., Tian, X. & Brueckner, M. Two populations of node monocilia initiate left–right asymmetry in the mouse. Cell 114, 61–73 (2003).
Article CAS PubMed Google Scholar
Sierant, M. C. et al. Genomic analysis of 11,555 probands identifies 60 dominant congenital heart disease genes. Proc. Natl Acad. Sci. USA 122, e2420343122 (2025).
Article CAS PubMed PubMed Central Google Scholar
Chiplunkar, A. R. et al. The Krüppel-like factor 2 and Krüppel-like factor 4 genes interact to maintain endothelial integrity in mouse embryonic vasculogenesis. BMC Dev. Biol. 13, 40 (2013).
Article PubMed PubMed Central Google Scholar
Zhang, Y. et al. Genetic reporter for live tracing fluid flow forces during cell fate segregation in mouse blastocyst development. Cell Stem Cell 30, 1110–1123 (2023).
Article CAS PubMed Google Scholar
de Soysa, T. Y. et al. Single-cell analysis of cardiogenesis reveals basis for organ-level developmental defects. Nature 572, 120–124 (2019).
Article PubMed PubMed Central Google Scholar
Todorovic, V. et al. Long form of latent TGF-β binding protein 1 (Ltbp1L) regulates cardiac valve development. Dev. Dyn. 240, 176–187 (2011).
Article CAS PubMed PubMed Central Google Scholar
Goddard, L. M. et al. Hemodynamic forces sculpt developing heart valves through a KLF2–WNT9B paracrine signaling axis. Dev. Cell 43, 274–289 (2017).
Article CAS PubMed PubMed Central Google Scholar
Lázár, E. et al. Spatial dynamics of the developing human heart. Preprint at bioRxiv https://doi.org/10.1101/2024.03.12.584577 (2024).
Cuttano, R. et al. KLF4 is a key determinant in the development and progression of cerebral cavernous malformations. EMBO Mol. Med. 8, 6–24 (2016).
Article CAS PubMed Google Scholar
Mastej, V., Axen, C., Wary, A., Minshall, R. D. & Wary, K. K. A requirement for Krüppel like factor-4 in the maintenance of endothelial cell quiescence. Front. Cell Dev. Biol. 10, 1003028 (2022).
Article PubMed PubMed Central Google Scholar
Wang, X. et al. Kruppel-like factor 4 (KLF-4) inhibits the epithelial-to-mesenchymal transition and proliferation of human endometrial carcinoma cells. Gynecol. Endocrinol. 32, 772–776 (2016).
Article CAS PubMed Google Scholar
Street, K. et al. Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics. BMC Genomics 19, 477 (2018).
Article PubMed PubMed Central Google Scholar
Ghaleb, A. M. & Yang, V. W. Kruppel-like factor 4 (KLF4): what we currently know. Gene 611, 27–37 (2017).
Article CAS PubMed PubMed Central Google Scholar
Subbalakshmi, A. R. et al. KLF4 induces mesenchymal–epithelial transition (MET) by suppressing multiple EMT-inducing transcription factors. Cancers (Basel) https://doi.org/10.3390/cancers13205135 (2021).
Article PubMed Google Scholar
Peterson, L. M. et al. 4D shear stress maps of the developing heart using Doppler optical coherence tomography. Biomed. Opt. Express 3, 3022–3032 (2012).
Article PubMed PubMed Central Google Scholar
Lee, J. et al. Spatial and temporal variations in hemodynamic forces initiate cardiac trabeculation. JCI Insight https://doi.org/10.1172/jci.insight.96672 (2018).
Article PubMed PubMed Central Google Scholar
Alejandre Alcazar, M. A. et al. Elafin treatment rescues EGFR–Klf4 signaling and lung cell survival in ventilated newborn mice. Am. J. Respir. Cell Mol. Biol. 59, 623–634 (2018).
Article PubMed PubMed Central Google Scholar
Andueza, A. et al. Endothelial reprogramming by disturbed flow revealed by single-cell RNA and chromatin accessibility study. Cell Rep. 33, 108491 (2020).
Article CAS PubMed PubMed Central Google Scholar
Kim, K.-H., Rosen, A., Bruneau, B. G., Hui, C.-C. & Backx, P. H. Iroquois homeodomain transcription factors in heart development and function. Circ. Res. 110, 1513–1524 (2012).
Article CAS PubMed Google Scholar
Fischer, A. et al. Combined loss of Hey1 and HeyL causes congenital heart defects because of impaired epithelial to mesenchymal transition. Circ. Res. 100, 856–863 (2007).
Article CAS PubMed Google Scholar
Wei, D., Kanai, M., Huang, S. & Xie, K. Emerging role of KLF4 in human gastrointestinal cancer. Carcinogenesis 27, 23–31 (2006).
Article CAS PubMed Google Scholar
Yang, H., Xi, X., Zhao, B., Su, Z. & Wang, Z. KLF4 protects brain microvascular endothelial cells from ischemic stroke induced apoptosis by transcriptionally activating MALAT1. Biochem. Biophys. Res. Commun. 495, 2376–2382 (2018).
Article CAS PubMed Google Scholar
Wang, Y. et al. KLF4 promotes angiogenesis by activating VEGF signaling in human retinal microvascular endothelial cells. PLoS ONE 10, e0130341 (2015).
Article PubMed PubMed Central Google Scholar
Li, H. et al. Cardiac fibroblast activation induced by oxygen–glucose deprivation depends on the HIF-1α/miR-212-5p/KLF4 pathway. J. Cardiovasc. Transl. Res. 16, 778–792 (2023).
Article PubMed Google Scholar
Aksoy, I. et al. Klf4 and Klf5 differentially inhibit mesoderm and endoderm differentiation in embryonic stem cells. Nat. Commun. 5, 3719 (2014).
Article PubMed Google Scholar
Malaab, M. et al. Antifibrotic factor KLF4 is repressed by the miR-10/TFAP2A/TBX5 axis in dermal fibroblasts: insights from twins discordant for systemic sclerosis. Ann. Rheum. Dis. 81, 268–277 (2022).
Article CAS PubMed Google Scholar
Choi, D. et al. ORAI1 activates proliferation of lymphatic endothelial cells in response to laminar flow through Krüppel-like factors 2 and 4. Circ. Res. 120, 1426–1439 (2017).
Article CAS PubMed PubMed Central Google Scholar
Han, H. et al. TRRUST v2: an expanded reference database of human and mouse transcriptional regulatory interactions. Nucleic Acids Res. 46, D380–D386 (2018).
Article CAS PubMed Google Scholar
Rothschild, S. C., Francescatto, L., Drummond, I. A. & Tombes, R. M. CaMK-II is a PKD2 target that promotes pronephric kidney development and stabilizes cilia. Development 138, 3387–3397 (2011).
Article CAS PubMed Google Scholar
Duchemin, A.-L., Vignes, H. & Vermot, J. Mechanically activated piezo channels modulate outflow tract valve development through the Yap1 and Klf2–Notch signaling axis. eLife https://doi.org/10.7554/eLife.44706 (2019).
Article PubMed PubMed Central Google Scholar
Zheng, Q. et al. Mechanosensitive channel PIEZO1 senses shear force to induce KLF2/4 expression via CaMKII/MEKK3/ERK5 axis in endothelial cells. Cells https://doi.org/10.3390/cells11142191 (2022).
Article PubMed PubMed Central Google Scholar
AbouAlaiwi, W. A. et al. Ciliary polycystin-2 is a mechanosensitive calcium channel involved in nitric oxide signaling cascades. Circ. Res. 104, 860–869 (2009).
Article CAS PubMed PubMed Central Google Scholar
Jones, E. A. V., Baron, M. H., Fraser, S. E. & Dickinson, M. E. Measuring hemodynamic changes during mammalian development. Am. J. Physiol. Heart Circ. Physiol. 287, H1561–H1569 (2004).
Article CAS PubMed Google Scholar
Mack, J. J. et al. NOTCH1 is a mechanosensor in adult arteries. Nat. Commun. 8, 1620 (2017).
Article PubMed PubMed Central Google Scholar
Kefaloyianni, E. & Coetzee, W. A. Transcriptional remodeling of ion channel subunits by flow adaptation in human coronary artery endothelial cells. J. Vasc. Res. 48, 357–367 (2011).
Article CAS PubMed PubMed Central Google Scholar
Lai, A. et al. Analyzing the shear-induced sensitization of mechanosensitive ion channel Piezo-1 in human aortic endothelial cells. J. Cell. Physiol. 236, 2976–2987 (2021).
Article CAS PubMed Google Scholar
Gopalakrishnan, J. et al. Emerging principles of primary cilia dynamics in controlling tissue organization and function. EMBO J. https://doi.org/10.15252/embj.2023113891 (2023).
Article PubMed PubMed Central Google Scholar
Van der Heiden, K. et al. Monocilia on chicken embryonic endocardium in low shear stress areas. Dev. Dyn. 235, 19–28 (2006).
Article PubMed Google Scholar
Jiang, X., Rowitch, D. H., Soriano, P., McMahon, A. P. & Sucov, H. M. Fate of the mammalian cardiac neural crest. Development 127, 1607–1616 (2000).
Article CAS PubMed Google Scholar
Buskohl, P. R., Jenkins, J. T. & Butcher, J. T. Computational simulation of hemodynamic-driven growth and remodeling of embryonic atrioventricular valves. Biomech. Model. Mechanobiol. 11, 1205–1217 (2012).
Article PubMed PubMed Central Google Scholar
Hierck, B. P. et al. Primary cilia sensitize endothelial cells for fluid shear stress. Dev. Dyn. 237, 725–735 (2008).
Article CAS PubMed Google Scholar
Li, P. et al. Cytoskeletal reorganization mediates fluid shear stress-induced ERK5 activation in osteoblastic cells. Cell Biol. Int. 36, 229–236 (2012).
Article CAS PubMed Google Scholar
Ishii, T., Warabi, E. & Mann, G. E. Mechanisms underlying unidirectional laminar shear stress-mediated Nrf2 activation in endothelial cells: amplification of low shear stress signaling by primary cilia. Redox Biol. 46, 102103 (2021).
Article CAS PubMed PubMed Central Google Scholar
Tiwari, A. et al. KLF4 regulates corneal epithelial cell cycle progression by suppressing canonical TGF-β signaling and upregulating CDK inhibitors P16 and P27. Invest. Ophthalmol. Vis. Sci. 60, 731–740 (2019).
Article CAS PubMed PubMed Central Google Scholar
Labour, M.-N., Riffault, M., Christensen, S. T. & Hoey, D. A. TGFβ1-induced recruitment of human bone mesenchymal stem cells is mediated by the primary cilium in a SMAD3-dependent manner. Sci. Rep. 6, 35542 (2016).
Article CAS PubMed PubMed Central Google Scholar
Rozycki, M. et al. The fate of the primary cilium during myofibroblast transition. Mol. Biol. Cell 25, 643–657 (2014).
Article PubMed PubMed Central Google Scholar
Vion, A.-C. et al. Primary cilia sensitize endothelial cells to BMP and prevent excessive vascular regression. J. Cell Biol. 217, 1651–1665 (2018).
Article CAS PubMed PubMed Central Google Scholar
Shan, F., Huang, Z., Xiong, R., Huang, Q.-Y. & Li, J. HIF1α-induced upregulation of KLF4 promotes migration of human vascular smooth muscle cells under hypoxia. J. Cell. Physiol. 235, 141–150 (2020).
Article CAS PubMed Google Scholar
Lin, Q.-Q., Zhao, J., Zheng, C.-G. & Chun, J. Roles of notch signaling pathway and endothelial–mesenchymal transition in vascular endothelial dysfunction and atherosclerosis. Eur. Rev. Med. Pharmacol. Sci. 22, 6485–6491 (2018).
PubMed Google Scholar
Jonassen, J. A., San Agustin, J., Follit, J. A. & Pazour, G. J. Deletion of IFT20 in the mouse kidney causes misorientation of the mitotic spindle and cystic kidney disease. J. Cell Biol. 183, 377–384 (2008).
Article CAS PubMed PubMed Central Google Scholar
Wu, G. et al. Cardiac defects and renal failure in mice with targeted mutations in Pkd2. Nat. Genet. 24, 75–78 (2000).
Article CAS PubMed Google Scholar
Yuan, S., Zhao, L., Brueckner, M. & Sun, Z. Intraciliary calcium oscillations initiate vertebrate left–right asymmetry. Curr. Biol. 25, 556–567 (2015).
Article CAS PubMed PubMed Central Google Scholar
Hogan, B. M. et al. ccbe1 is required for embryonic lymphangiogenesis and venous sprouting. Nat. Genet. 41, 396–398 (2009).
Article CAS PubMed Google Scholar
Schlaeppi, A., Graves, A., Weber, M. & Huisken, J. Light sheet microscopy of fast cardiac dynamics in zebrafish embryos. J. Vis. Exp. https://doi.org/10.3791/62741 (2021).
Article PubMed Google Scholar
Stirling, D. R. et al. CellProfiler 4: improvements in speed, utility and usability. BMC Bioinformatics 22, 433 (2021).
Article PubMed PubMed Central Google Scholar
Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. Nat. Methods 9, 676–682 (2012).
Article CAS PubMed Google Scholar
Wan, Y. et al. FluoRender: joint freehand segmentation and visualization for many-channel fluorescence data analysis. BMC Bioinformatics 18, 280 (2017).
Article PubMed PubMed Central Google Scholar
Wolock, S. L., Lopez, R. & Klein, A. M. Scrublet: computational identification of cell doublets in single-cell transcriptomic data. Cell Syst. 8, 281–291 (2019).
Article CAS PubMed PubMed Central Google Scholar
Van Rossum, G. & Drake, F. L. The Python Language Reference Manual: For Python Version 3.2 (Network Theory Ltd., 2011).
Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587 (2021).
Article CAS PubMed PubMed Central Google Scholar
Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat. Methods 16, 1289–1296 (2019).
Article CAS PubMed PubMed Central Google Scholar
Kowalczyk, M. S. et al. Single-cell RNA-seq reveals changes in cell cycle and differentiation programs upon aging of hematopoietic stem cells. Genome Res. 25, 1860–1872 (2015).
Article CAS PubMed PubMed Central Google Scholar
Chen, E. Y. et al. Enrichr: interactive and collaborative HTML5 gene list enrichment analysis tool. BMC Bioinformatics 14, 128 (2013).
Article PubMed PubMed Central Google Scholar
Bailey, T. L., Johnson, J., Grant, C. E. & Noble, W. S. The MEME Suite. Nucleic Acids Res. 43, W39–W49 (2015).
Article CAS PubMed PubMed Central Google Scholar
Speir, M. L. et al. UCSC Cell Browser: visualize your single-cell data. Bioinformatics 37, 4578–4580 (2021).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We thank the Yale Center for Genome Analysis for DNA sequencing, the Yale Imaging Facility on Science Hill for lightsheet microscopy, and the Yale Center for Cellular and Molecular Imaging for confocal microscopy and analysis workstations. We thank J. Drozd, T. Nuzzo and L. Baconguis for mouse support (Yale School of Medicine Animal Resources Center). We thank D. DeLaughter for snRNA-seq alignment (Harvard Medical School, Department of Genetics). We also thank Z. Sun (Yale School of Medicine, Department of Genetics) for sharing transgenic zebrafish with us, and D. Breslow, Z. Sun, S. Yuan and M. Pownall for their critical review of the manuscript. This work was supported by the National Heart, Lung and Blood Institute (NHLBI) Ruth L. Kirschstein National Research Service Award (NRSA-F31) to K.B. (F31HL158091) and R35HL145249 to M.B. NIH 5R01HL162356 to J.S.

Author information

Authors and Affiliations

Department of Genetics, Yale University School of Medicine, New Haven, CT, USA
Kathryn Berg & Martina Brueckner
Department of Genetics, Harvard Medical School, Boston, MA, USA
Joshua Gorham & Jonathan Seidman
Department of Molecular, Cellular and Developmental Biology, Yale College, New Haven, CT, USA
Faith Lundt
Department of Pediatrics, Yale University School of Medicine, New Haven, CT, USA
Martina Brueckner

Authors

Kathryn Berg
View author publications
Search author on:PubMed Google Scholar
Joshua Gorham
View author publications
Search author on:PubMed Google Scholar
Faith Lundt
View author publications
Search author on:PubMed Google Scholar
Jonathan Seidman
View author publications
Search author on:PubMed Google Scholar
Martina Brueckner
View author publications
Search author on:PubMed Google Scholar

Contributions

Conceptualization: K.B. and M.B.; formal analysis, data curation and visualization: K.B.; investigation: K.B., F.L., J.G., J.S. and M.B.; writing—original draft: K.B. and M.B.; supervision, project administration and funding acquisition: M.B.

Corresponding author

Correspondence to Martina Brueckner.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Cardiovascular Research thanks Dagmar Wachten and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Spatiotemporally dynamic endocardial ciliation of the ECCs during cushion EndoMT is blood-flow dependent and is conserved in zebrafish.

a) Immunofluorescence on mouse embryo sections for cilia (Arl13b, green) on endocardial cells (CD31, white) of the OFT over cushion development (e8.5, e9.5, e10). Nuclei are shown in blue (Hoechst). Close ups with and without CD31 are given; white arrows indicate cilia; luminal endocardial cells outlined with white and whole OFT (including cushion mesenchyme) outlined in red. b) Diagram of an e9.5 mouse heart for cryosection orientation (yellow: mesenchyme, red: myocardium, blue: endocardium, grey: lumen). c) Immunofluorescence on e10 mouse embryo sections for cilia (Arl13b, green) in the AVC. Nuclei are shown in blue (Hoechst). Luminal endocardial cells are outlined with dashed white and the whole AVC (including cushion mesenchyme) outlined in dotted white. d) Number of ciliated mesenchymal cushion cells as a percentage of total mesenchymal cushion cells in the AVC over time (e9.5 (n = 3), e10 (n = 3), e10.5 (n = 3)) (p = 0.0101). e) Still of live 54hpf tg(Arl13b:eGPF;Kdrl (Flk-1):mCherry;Cmlc2:eGFP) zebrafish heart showing cilia and myocardium in green and endocardium in red. Closeup of cilia provided in yellow box. f) Zebrafish endocardial ciliation over time in the atrioventricular canal (n = 3). (48 hpf vs 54 hpf p = 0.0213). g) Ciliated endocardial cells as percentage of all endocardial cells versus distance in the mouse OFT at e8.5 (n = 2) and e9.5 (n = 3). Distance runs left to right, Ventricle to Dorsal Aorta. h) Immunofluorescence on e9.5 wildtype and Ncx1^−/− mouse heart sections for TUNEL (red) in endocardial cells (CD31, white). Nuclei are shown in blue (Hoechst). Version without CD31 is provided; endocardium is outlined in white. i) Quantification for TUNEL staining in h) (WT n = 7 sections, Ncx1^−/− n = 11 sections). j) Endocardial ciliation in the OFT over time as percentage of all endocardial cells in wildtype mice (blue) and Ncx1^−/− mice (red) (e8.5 (WT n = 7, Ncx1^−/− n = 7), e9.0 (WT n = 4, Ncx1^−/− n = 4), e9.5 (WT n = 7, Ncx1^−/− n = 7)) (e8.5 p = 0.3438, e9 p < 0.0001, e9.5 p = 0.0397). Statistics: ns (p > 0.05), * (p ≤ 0.05), ** (p ≤ 0.01), *** (p ≤ 0.001), **** (p ≤ 0.0001). Statistical test used was two-sided t-test with Welch’s correction. Data are represented as mean ± SEM. Unless otherwise stated, n is number of embryos. Abbreviations: A-Atrium, AVC-Atrioventricular Canal, EC-Endocardial Cell, LV-Left Ventricle, V-Ventricle, OFT-Outflow Tract, RV-Right Ventricle, WT-Wildtype.

Source data

Extended Data Fig. 2 Endocardial cushion EndoMT requires primary cilia independent of heart looping.

Ciliated endocardial cells in the a) AVC and b) OFT over time as percentage of all endocardial cells in Ift20^−/− (dark green) and Kif3a^−/− (light green). Biological replicates for AVC (e8.5 (Ift20^−/− n = 6, Kif3a^−/− n = 4), e9.0 (Ift20^−/− n = 3, Kif3a^−/− n = 3), e9.5 (Ift20^−/− n = 3, Kif3a^−/− n = 4)); OFT (e8.5 (Ift20^−/− n = 4, Kif3a^−/− n = 4), e9.0 (Ift20^−/− n = 2, Kif3a^−/− n = 2), e9.5 (Ift20^−/− n = 4, Kif3a^−/− n = 3)). Wildtype (blue) provided as comparison (e9.5 n = 8, e9.0 n = 4, e9.5 n = 6) (WT vs Ift20^−/−: AVC e8.5 p < 0.0001, AVC e9 p = 0.0003, AVC e9.5 p < 0.0001, OFT e8.5 p < 0.0001, OFT e9.5 p = 0.0004/ WT vs Kif3a^−/−: AVC e8.5 p < 0.0001, AVC e9 p = 0.0002, AVC e9.5 p < 0.0001, OFT e8.5 p < 0.0001, OFT e9.5 p = 0.0004). Immunofluorescence on e9.5 wildtype and cilia KO (Ift20^−/−) mouse heart sections for c) TUNEL (red) or d) PCNA (green) in endocardial cells (CD31, white). Nuclei are shown in blue (Hoechst); the AVC is outlined in white. Closeup of PCNA signal from the white box in d) with nuclei outlined in white is given. White arrows indicate negative nuclei. e) Percentage of TUNEL or PCNA positive endocardium as a percentage of all e9.5 endocardium in the AVC and OFT (TUNEL WT n = 21 sections, Cilia KO n = 24 sections/PCNA WT n = 10 sections, Cilia KO n = 12 sections). Statistics: ns (p > 0.05), ** (p ≤ 0.01), *** (p ≤ 0.001), **** (p ≤ 0.0001). Statistical test used was two-sided t-test with Welch’s correction. Data are represented as mean ± SEM. Unless otherwise stated, n is number of embryos. Abbreviations: A-Atrium, AVC-Atrioventricular Canal, LV-Left Ventricle, OFT-Outflow Tract, WT-Wildtype.

Source data

Extended Data Fig. 3 The mechanosensitive transcription factors KLF2 and KLF4 are dynamically expressed in mouse endocardium during cushion EndoMT.

Whole mount ISH of e9.5 mouse embryos for a) Klf2 mRNA and b) Klf4 mRNA with closeups and cryosections of the OFT. c) KLF4 protein positive endocardial cells over time as percentage of all endocardial cells in the OFT (e8.5 (n = 11), e9 (n = 3), e9.5 (n = 6), e10 (n = 2)) and AVC (e8.5 (n = 14), e9 (n = 6), e9.5 (n = 8), e10 (n = 3)) (AVC: e8.5 vs e9 p = 0.0007, e9 vs e9.5 p = 0.0002, e9.5 vs e10 p = 0.0092). d) mRNA expression as measured by Grey_Value over distance for Klf4 (n = 3) in the e9.5 OFT. Shading indicates +/- SEM from mean (line). e) KLF4 protein positive endocardial cells versus distance in the e9.5 OFT (n = 2). For d) and e), distance runs left to right, Right Ventricle to Dorsal Aorta. Statistics: ns (p > 0.05), * (p ≤ 0.05), ** (p ≤ 0.01), **** (p ≤ 0.0001). Statistical test used was two-sided t-test with Welch’s correction. Data are represented as mean ± SEM. Unless otherwise stated, n is number of embryos. Abbreviations: DA-Dorsal Aorta, EC-Endocardial Cell, LV-Left Ventricle, OFT-Outflow Tract, RV-Right Ventricle, WT-Wildtype.

Source data

Extended Data Fig. 4 KLF4 expression negatively correlates with cushion EndoMT progression.

a) HCR-FISH on whole mount mouse embryos at e9.5 showing Kdr mRNA (red), Vim mRNA (green), and nuclei (Hoechst, blue). The AVC endocardium is outlined in white. b) Correlation matrix between Klf4 and EndoMT regulators in the endocardium; mRNA expression from SingleNuc RNAseq (4,291 nuclei). Abbreviations: OFT-Outflow Tract, WT-Wildtype.

Extended Data Fig. 5 Endocardial KLF4 expression correlates with ciliation and depends on blood flow.

a) Closeup from Fig. 5b” showing ciliated, KLF4-high cell (cilia: Arl13b, green, KLF4, red) and nonciliated, KLF4-low cell (white arrows). Endocardial lumen is outlined with dashed white line. b) KLF4 protein positive endocardial cells in the OFT over time as percentage of all endocardial cells in wildtype mice (blue) and Ncx1^−/− mice (red) (e8.5 (WT n = 10, Ncx1^−/− n = 10), e9.0 (WT n = 4, Ncx1^−/− n = 4), e9.5 (WT n = 5, Ncx1^−/− n = 5)) (e8.5 p = 0.0444, e9 p = 0.0135, e9.5 p = 0.0097). Statistics: * (p ≤ 0.05), ** (p ≤ 0.01). Statistical test used was two-sided t-test with Welch’s correction. Data are represented as mean ± SEM. Unless otherwise stated, n is number of embryos. Abbreviations: OFT-Outflow Tract, WT-Wildtype.

Source data

Extended Data Fig. 6 Failure of ciliogenesis results in abnormal KLF4 expression.

a) KLF4 protein positive endocardial cells in the AVC over time as percentage of all endocardial cells in Ift20^−/− (dark green) and Kif3a^−/− (light green) (e8.5 (Ift20^−/− n = 4, Kif3a^−/− n = 3), e9.0 (Ift20^−/− n = 2, Kif3a^−/− n = 2), e9.5 (Ift20^−/− n = 4, Kif3a^−/− n = 2)). b) KLF4 protein positive endocardial cells in the OFT over time as percentage of all endocardial cells in Ift20^−/− (dark green) and Kif3a^−/− (light green) (e8.5 (Ift20^−/− n = 3, Kif3a^−/− n = 2), e9.0 (Ift20^−/− n = 2, Kif3a^−/− n = 2), e9.5 (Ift20^−/− n = 4, Kif3a^−/− n = 2)). c) KLF4 protein positive endocardial cells in the OFT over time as percentage of all endocardial cells in wildtype (blue) and cilia KO mice (green) (e8.5 (WT n = 4, cilia KO n = 4), e9.0 (WT n = 3, cilia KO n = 3), e9.5 (WT n = 6, cilia KO n = 6)) (e8.5 p = 0.1402, e9 p = 0.0007, e9.5 p = 0.0544). d) Immunofluorescence on e9.5 wildtype and cilia KO (Ift20^−/−) whole mount hearts for KLF4 (red) in endocardial cells (CD31, white). Nuclei are shown in blue (Hoechst); the AVC is outlined in white. e) Number of KLF4 protein positive endocardial cells versus distance in the OFT of e9.5 cilia KO mice (green, n = 3); wildtype is given for comparison in blue (n = 2). f) Klf4 mRNA expression as measured by Grey_Value over distance in the OFT of e9.5 cilia KO mice (green, n = 3); wildtype is given for comparison in blue (n = 3). Shading indicates +/- SEM from mean (line). g) KLF4 protein positive endocardial cells in the e9.5 OFT as percentage of all endocardial cells in Ift20^+/−/ Ncx1^+/− (blue, n = 3) and Ift20^−/−/ Ncx1^−/− (brown, n = 4). Ift20^−/− (green, n = 4) and Ncx1^−/− (red, OFT n = 5) are given for comparison (Ift20^+/−/ Ncx1^+/− vs Ncx1^−/− p < 0.0001, Ift20^+/−/ Ncx1^+/− vs Ift20^−/−/ Ncx1^−/− p < 0.0001). h) Endocardial cells with a leading edge of fibronectin as percentage of all endocardial cells in e9.5 AVCs of wildtype (blue, n = 4) and cilia KO (green, n = 5) mice. i) Immunofluorescence on e9.5 wildtype and cilia KO AVC sections for Fibronectin (red). Nuclei are shown in blue (Hoechst); AVC is outlined in white. For e) and f), distance runs left to right, Right Ventricle to Dorsal Aorta. Statistics: ns (p > 0.05), * (p ≤ 0.05), ** (p ≤ 0.01), *** (p ≤ 0.001), **** (p ≤ 0.0001). Statistical test used was two-sided t-test with Welch’s correction. Data are represented as mean ± SEM. Unless otherwise stated, n is number of embryos. Abbreviations: A-Atrium, AVC-Atrioventricular Canal, DA-Dorsal Aorta, EC-Endocardial Cell, LV-Left Ventricle, OFT-Outflow Tract, RV-Right Ventricle, WT-Wildtype.

Source data

Extended Data Fig. 7 Defects in ciliogenesis block EndoMT.

a) Integration of biological replicates for snRNAseq of the whole embryonic e9.5 heart (29,313 nuclei). b) Number of features and counts across biological replicates (Cilia WT nuclei=12,763, Cilia Het nuclei=12,030, Cilia Mut nuclei=11,503). Graphs of the percentage of cell types in the c) whole heart (29,313 nuclei) and d) endocardium (Cilia WT nuclei=1,340, Cilia Het nuclei=2,029, Cilia Mut nuclei=922) for each genotype (Replicates: Cilia WT n = 3, Cilia Het n = 2, Cilia Mut n = 2) (Late EndoMT Cilia WT vs Cilia KO p = 0.034). Statistically significant numbers are given in red. Dotplots for top DEGs for each cell type in the e) vEC/EndoMT cluster (1,496 nuclei), f) whole heart (29,313 nuclei) and i) endocardium (4,291 nuclei) from snRNA-seq data. Established marker genes for clusters found in the g) whole heart (29,313 nuclei) and h) endocardium (4,291 nuclei). Statistics: ns (p > 0.05), ** (p ≤ 0.01). Statistical test used was two-tailed t-test with Bonferroni correction. Statistical Data are represented as mean ± SEM. Unless otherwise stated, n is number of embryos. Box-plot stats for c): EC Cilia WT (min: 27.34, Q1: 28.12, median: 28.91, Q3: 30.73, max: 32.56, lower whisker: 27.34, upper whisker: 32.56), EC Cilia Het (min: 12.05, Q1: 19.53, median: 27.01, Q3: 34.48, max: 41.96, lower whisker: 12.05, upper whisker: 41.96), EC Cilia Mut (min: 25.57, Q1: 27.82, median: 30.07, Q3: 32.33, max: 34.58, lower whisker: 25.57, upper whisker: 34.58), EndoMT Cilia WT (min: 19.77, Q1: 20.94, median: 22.10, Q3: 27.60, max: 33.09, lower whisker: 19.77, upper whisker: 33.09), EndoMT Cilia Het (min: 15.14, Q1: 17.19, median: 19.23, Q3: 21.28, max: 23.33, lower whisker: 15.14, upper whisker: 23.33), EndoMT Cilia Mut (min: 30.51, Q1: 32.02, median: 33.54, Q3: 35.05, max: 36.57, lower whisker: 30.51, upper whisker: 36.57), HE Cilia WT (min: 3.11, Q1: 4.43, median: 5.76, Q3: 7.53, max: 9.30, lower whisker: 3.11, upper whisker: 9.30), HE Cilia Het (min: 1.38, Q1: 3.93, median: 6.47, Q3: 9.02, max: 11.57, lower whisker: 1.38, upper whisker: 11.57), HE Cilia Mut (min: 2.59, Q1: 4.71, median: 6.84, Q3: 8.96, max: 11.09, lower whisker: 2.59, upper whisker: 11.09), Late EndoMT Cilia WT (min: 6.95, Q1: 6.96, median: 6.98, Q3: 9.04, max: 11.11, lower whisker: 6.95, upper whisker: 11.11), Late EndoMT Cilia Het (min: 6.06, Q1: 6.26, median: 6.46, Q3: 6.66, max: 6.86, lower whisker: 6.06, upper whisker: 6.86), Late EndoMT Cilia Mut (min: 0.82, Q1: 1.34, median: 1.86, Q3: 2.39, max: 2.91, lower whisker: 0.82, upper whisker: 2.91), PIM Cilia WT (min: 5.26, Q1: 6.34, median: 7.43, Q3: 8.37, max: 9.30, lower whisker: 5.26, upper whisker: 9.30), PIM Cilia Het (min: 3.42, Q1: 3.45, median: 3.47, Q3: 3.50, max: 3.53, lower whisker: 3.42, upper whisker: 3.53), PIM Cilia Mut (min: 2.45, Q1: 3.38, median: 4.30, Q3: 5.23, max: 6.15, lower whisker: 2.45, upper whisker: 6.15), Prolif. EC 1 Cilia WT (min: 14.39, Q1: 17.66, median: 20.93, Q3: 23.19, max: 25.45, lower whisker: 14.39, upper whisker: 25.45), Prolif. EC 1 Cilia Het (min: 8.24, Q1: 14.21, median: 20.19, Q3: 26.16, max: 32.13, lower whisker: 8.24, upper whisker: 32.13), Prolif. EC 1 Cilia Mut (min: 2.77, Q1: 6.53, median: 10.29, Q3: 14.04, max: 17.80, lower whisker: 2.77, upper whisker: 17.80), Prolif. EC 2 Cilia WT (min: 0.00, Q1: 0.00, median: 0.00, Q3: 0.48, max: 0.96, lower whisker: 0.00, upper whisker: 0.96), Prolif. EC 2 Cilia Het (min: 0.00, Q1: 6.91, median: 13.82, Q3: 20.74, max: 27.65, lower whisker: 0.00, upper whisker: 27.65), Prolif. EC 2 Cilia Mut (min: 0.00, Q1: 0.08, median: 0.16, Q3: 0.24, max: 0.32, lower whisker: 0.00, upper whisker: 0.32), Valvular EC Cilia WT (min: 1.16, Q1: 2.13, median: 3.11, Q3: 4.08, max: 5.04, lower whisker: 1.16, upper whisker: 5.04), Valvular EC Cilia Het (min: 2.17, Q1: 2.75, median: 3.34, Q3: 3.92, max: 4.51, lower whisker: 2.17, upper whisker: 4.51), Valvular EC Cilia Mut (min: 8.09, Q1: 10.51, median: 12.94, Q3: 15.36, max: 17.78, lower whisker: 8.09, upper whisker: 17.78). Box-plot stats for d): Endocardium Cilia WT (min: 7.86, Q1: 10.50, median: 13.13, Q3: 14.79, max: 16.45, lower whisker: 7.86, upper whisker: 16.45), Endocardium Cilia Het (min: 13.05, Q1: 13.46, median: 13.87, Q3: 14.28, max: 14.69, lower whisker: 13.05, upper whisker: 14.69), Endocardium Cilia Mut (min: 15.07, Q1: 16.02, median: 16.96, Q3: 17.91, max: 18.85, lower whisker: 15.07, upper whisker: 18.85), Epicardium Cilia WT (min: 1.41, Q1: 1.76, median: 2.11, Q3: 2.33, max: 2.56, lower whisker: 1.41, upper whisker: 2.56), Epicardium Cilia Het (min: 1.32, Q1: 1.64, median: 1.96, Q3: 2.29, max: 2.61, lower whisker: 1.32, upper whisker: 2.61), Epicardium Cilia Mut (min: 1.22, Q1: 1.65, median: 2.08, Q3: 2.50, max: 2.93, lower whisker: 1.22, upper whisker: 2.93), Erythroid Cilia WT (min: 1.63, Q1: 2.14, median: 2.65, Q3: 4.74, max: 6.83, lower whisker: 1.63, upper whisker: 6.83), Erythroid Cilia Het (min: 1.31, Q1: 1.42, median: 1.52, Q3: 1.63, max: 1.74, lower whisker: 1.31, upper whisker: 1.74), Erythroid Cilia Mut (min: 0.89, Q1: 2.24, median: 3.58, Q3: 4.93, max: 6.28, lower whisker: 0.89, upper whisker: 6.28), Fibroblasts Cilia WT (min: 0.83, Q1: 1.05, median: 1.28, Q3: 1.63, max: 1.98, lower whisker: 0.83, upper whisker: 1.98), Fibroblasts Cilia Het (min: 0.87, Q1: 1.14, median: 1.41, Q3: 1.68, max: 1.95, lower whisker: 0.87, upper whisker: 1.95), Fibroblasts Cilia Mut (min: 0.61, Q1: 1.14, median: 1.68, Q3: 2.21, max: 2.75, lower whisker: 0.61, upper whisker: 2.75), Immune Cilia WT (min: 0.25, Q1: 0.26, median: 0.28, Q3: 0.83, max: 1.37, lower whisker: 0.25, upper whisker: 1.37), Immune Cilia Het (min: 0.31, Q1: 0.32, median: 0.32, Q3: 0.33, max: 0.34, lower whisker: 0.31, upper whisker: 0.34), Immune Cilia Mut (min: 0.64, Q1: 0.69, median: 0.74, Q3: 0.80, max: 0.85, lower whisker: 0.64, upper whisker: 0.85), Multipotent Progenitor Cilia WT (min: 1.28, Q1: 1.41, median: 1.54, Q3: 1.59, max: 1.65, lower whisker: 1.28, upper whisker: 1.65), Multipotent Progenitor Cilia Het (min: 0.51, Q1: 0.91, median: 1.31, Q3: 1.72, max: 2.12, lower whisker: 0.51, upper whisker: 2.12), Multipotent Progenitor Cilia Mut (min: 0.92, Q1: 2.58, median: 4.25, Q3: 5.91, max: 7.57, lower whisker: 0.92, upper whisker: 7.57), Myocardium 1 Cilia WT (min: 41.61, Q1: 42.25, median: 42.90, Q3: 49.70, max: 56.49, lower whisker: 41.61, upper whisker: 56.49), Myocardium 1 Cilia Het (min: 39.26, Q1: 42.10, median: 44.95, Q3: 47.79, max: 50.63, lower whisker: 39.26, upper whisker: 50.63), Myocardium 1 Cilia Mut (min: 21.86, Q1: 27.48, median: 33.11, Q3: 38.73, max: 44.36, lower whisker: 21.86, upper whisker: 44.36), Myocardium 2 Cilia WT (min: 26.14, Q1: 29.34, median: 32.55, Q3: 33.89, max: 35.23, lower whisker: 26.14, upper whisker: 35.23), Myocardium 2 Cilia Het (min: 27.64, Q1: 31.14, median: 34.64, Q3: 38.14, max: 41.64, lower whisker: 27.64, upper whisker: 41.64), Myocardium 2 Cilia Mut (min: 26.91, Q1: 32.26, median: 37.60, Q3: 42.95, max: 48.29, lower whisker: 26.91, upper whisker: 48.29). Abbreviations: A-Atrium, DEG-Differentially Expressed Gene, EC-Endocardial Cell, HE-Hematoendothelium, LV-Left Ventricle, PIM-Pro-inflammatory, vEC-Valvular Endocardial Cell, WT-Wildtype.

Source data

Extended Data Fig. 8 Defects in blood flow block EndoMT.

a) Number of features and counts across biological replicates for the Ncx1^−/− snRNAseq dataset (36,296 nuclei). b) Integration of biological replicates for snRNAseq of the whole embryonic e9.5 heart (36,296 nuclei). c) UMAP plot of e9.5 whole hearts (36,296 nuclei). d) Graph of the percentage of cell types in the whole heart (Ncx1^+/+ 12,763 nuclei, Ncx1^+/− 12,030 nuclei, Ncx1^−/− 11,503 nuclei). e) UMAP plot of the endocardial subcluster (6,548 nuclei). f) Graph of the percentage of cell types in the endocardial subcluster (Ncx1^+/+ 2,589 nuclei, Ncx1^+/− 2,265 nuclei, Ncx1^−/− 1,694 nuclei). g) GeneSet Enrichment for EndoMT progression for vEC/EndoMT clusters (2,601 nuclei). h) UMAP plot of the vEC/EndoMT subcluster (2,601 nuclei). i) The percentage of nuclei in each EndoMT pseudostage per genotype (Ncx1^+/+ 1,284 nuclei, Ncx1^+/− 918 nuclei, Ncx1^−/− 399 nuclei). j) Percentage of cells in each cell cycle phase per genotype in the snRNA-seq vEC/EndoMT cluster (2,601 nuclei). k) Violin plots for cell death gene expression in each genotype in vEC/EndoMT cluster (2,601 nuclei). Abbreviations: EC-Endocardium, Endo-EndoMT, HE-Hematoendothelium, Mes-Mesenchyme, MP-Multipotent, PIM-Pro-inflammatory, vEC-Valvular Endocardial Cell, WT-Wildtype.

Extended Data Fig. 9 Defects in blood flow block EndoMT.

Dotplots for top DEGs for each cell type in the a) whole heart (36,296 nuclei) and d) endocardial subcluster (6,548 nuclei) from snRNA-seq data. Established marker genes for clusters found in the b) whole heart (36,296 nuclei) and c) endocardial subcluster (6,548 nuclei). e) Gene Ontology-Biological Processes lower and higher in Ncx1^−/− vEC/EndoMT nuclei compared to Ncx1^+/+ and Ncx1^+/− littermates. Cushion/contractility relevant terms are highlighted (p < 0.05). Abbreviations: A-Atrium, DEG-Differentially Expressed Gene, EC-Endocardium, HE-Hematoendothelium, MP-Multipotent, PIM-Pro-inflammatory, vEC-Valvular Endocardial Cell, WT-Wildtype.

Source data

Extended Data Fig. 10 EndoMT molecular pathways are dysregulated in cilia and ciliary-calcium signaling KO hearts.

a) Violin plots for cell death gene expression in each genotype in vEC/EndoMT cluster (1,496 nuclei). b) Percentage of cells in each cell cycle phase per genotype in the snRNA-seq vEC/EndoMT cluster (Replicates: Cilia WT n = 3, Cilia Het n = 2, Cilia Mut n = 2) (1,496 nuclei). c) The top 10 highest and lowest DEGs between cilia KO (Ift20^−/−) and wildtype hearts in the vEC/EndoMT cluster (1,496 nuclei) over EndoMT pseudostages, ordered by avg_log2FC. d) Whole mount HCR-FISH for Kdr (red), Vim (green), Emcn (green) and Vcan (red) in wildtype and cilia KO (Ift20^−/−) e9.5 AVCs. Nuclei are shown in blue (Hoechst); the AVC is outlined in white. e) Corrected total cell fluorescence (CTCF) analysis of HCR-FISH for Emcn, Kdr, Vcan, and Vim in the e9.5 AVC (WT n = 3, Cilia KO n = 3). f) Violin plots for GeneSet Enrichment of Smad2/3/4 transcriptional targets (GSEA-MSigDB:MM14876) in each genotype for the vEC/EndoMT cluster (1,496 nuclei) or the Epicardial Cluster (545 nuclei) (vEC/EndoMT Cluster: Cilia WT vs Cilia KO p = 2.9e-06, Cilia Het vs Cilia KO p = 0.00041). g) Endocardial ciliation in the AVC over time as percentage of all endocardial cells in wildtype mice (blue) and Pkd2^−/− mice (purple) (OFT: e8.5 (WT n = 5, Pkd2^−/− n = 5), e9.5 (WT n = 6, Pkd2^−/− n = 10); AVC: e8.5 (WT = 5, Pkd2^−/−=5), e9.5 (WT n = 8, Pkd2-/- n = 10)). Statistics: ns (p > 0.05), ** (p ≤ 0.01), **** (p ≤ 0.0001). Statistical test used for a) and f) was students two-tailed t-test. Statistical test used for b) was two-tailed t-test with Bonferroni correction. Statistical test used for g) was two-sided t-test with Welch’s correction. Data are represented as mean ± SEM. Unless otherwise stated, n is number of embryos. Box-plot stats for b): G1 Cilia WT (min: 33.33, Q1: 39.20, median: 45.07, Q3: 49.13, max: 53.19, lower whisker: 33.33, upper whisker: 53.19), G1 Cilia Het (min: 42.94, Q1: 45.38, median: 47.81, Q3: 50.24, max: 52.68, lower whisker: 42.94, upper whisker: 52.68), G1 Cilia Mut (min: 46.26, Q1: 46.41, median: 46.55, Q3: 46.70, max: 46.84, lower whisker: 46.26, upper whisker: 46.84), G2M Cilia WT (min: 14.89, Q1: 17.97, median: 21.05, Q3: 23.02, max: 25.00, lower whisker: 14.89, upper whisker: 25.00), G2M Cilia Het (min: 15.77, Q1: 17.34, median: 18.90, Q3: 20.46, max: 22.03, lower whisker: 15.77, upper whisker: 22.03), G2M Cilia Mut (min: 22.45, Q1: 22.73, median: 23.02, Q3: 23.30, max: 23.59, lower whisker: 22.45, upper whisker: 23.59), S Cilia WT (min: 31.91, Q1: 32.90, median: 33.88, Q3: 37.78, max: 41.67, lower whisker: 31.91, upper whisker: 41.67), S Cilia Het (min: 31.55, Q1: 32.42, median: 33.29, Q3: 34.16, max: 35.03, lower whisker: 31.55, upper whisker: 35.03), S Cilia Mut (min: 29.57, Q1: 30.00, median: 30.43, Q3: 30.86, max: 31.29, lower whisker: 29.57, upper whisker: 31.29). Abbreviations: A-Atria, AVC-Atrioventricular Canal, LV-Left Ventricle, OFT-Outflow Tract, vEC-Valvular Endocardium, WT-Wildtype.

Source data

Supplementary information

Reporting Summary

44161_2025_697_MOESM2_ESM.xlsx

Supplementary Tables Supplementary Table 1. Differentially expressed genes between each pseudophase of EndoMT. Quiescent endocardium. Generated from snRNA-seq. Avg_logFC is given for the named cluster compared to the other three. Cutoff adjusted P value of 0.05. Supplementary Table 2. Differentially expressed genes between each pseudophase of EndoMT. Early EndoMT. Generated from snRNA-seq. Avg_logFC is given for the named cluster compared to the other three. Cutoff adjusted P value of 0.05. Supplementary Table 3. Differentially expressed genes between each pseudophase of EndoMT. Mid EndoMT. Generated from snRNA-seq. Avg_logFC is given for the named cluster compared to the other three. Cutoff adjusted P value of 0.05. Supplementary Table 4. Differentially expressed genes between each pseudophase of EndoMT. Late EndoMT. Generated from snRNA-seq. Avg_logFC is given for the named cluster compared to the other three. Cutoff adjusted P value of 0.05. Supplementary Table 5. Differentially expressed genes between the Ift20^+/+ and Ift20^−/− endocardium. Positive values indicate a higher expression in Ift20^+/+. Supplementary Table 6. Differentially expressed genes between the Ift20^+/+, Ift20^+/− and Ift20^−/− endocardium undergoing EndoMT. A cutoff of 0.4 avg_logFC in Ift20^+/+ versus Ift20^−/− was used to guarantee differential expression. Positive values indicate a higher expression in Ift20^+/+ or Ift20^+/−. Supplementary Table 7. Disorganized gene expression in the Ift20^−/− endocardium between each pseudophase of EndoMT. Genes that are significantly either increased or decreased between each pseudophase in Ift20^+/+ and Ift20^+/− but not Ift20^−/−. Genes ‘increased’ are not turned on in Ift20^−/−, and genes ‘decreased’ are not turned off. Supplementary Table 8. Resources used: antibodies. Supplementary Table 9. Resources used: bacterial and viral strains. Supplementary Table 10. Resources used: chemicals. Supplementary Table 11. Resources used: critical commercial assays. Supplementary Table 12. Resources used: deposited data. Supplementary Table 13. Resources used: model organism strains. Supplementary Table 14. Resources used: oligonucleotides. Supplementary Table 15. Resources used: software.

Supplementary Video 1 Cilia KO (left) e9.5 hearts beat at the same frequency as wild-type (right) littermate hearts.

Source data

Source Data Fig. 1

Statistical source data.

Source Data Fig. 2

Statistical source data.

Source Data Fig. 3

Statistical source data.

Source Data Fig. 4

Statistical source data.

Source Data Fig. 5

Statistical source data.

Source Data Fig. 6

Statistical source data.

Source Data Fig. 7

Statistical source data.

Source Data Fig. 8

Statistical source data.

Source Data Extended Data Fig. 1

Statistical source data.

Source Data Extended Data Fig. 2

Statistical source data.

Source Data Extended Data Fig. 3

Statistical source data.

Source Data Extended Data Fig. 5

Statistical source data.

Source Data Extended Data Fig. 6

Statistical source data.

Source Data Extended Data Fig. 7

Statistical source data.

Source Data Extended Data Fig. 9

Statistical source data.

Source Data Extended Data Fig. 10

Statistical source data.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Berg, K., Gorham, J., Lundt, F. et al. Endocardial primary cilia and blood flow regulate EndoMT during endocardial cushion development. Nat Cardiovasc Res 4, 1114–1134 (2025). https://doi.org/10.1038/s44161-025-00697-z

Download citation

Received: 10 June 2024
Accepted: 17 July 2025
Published: 26 August 2025
Issue date: September 2025
DOI: https://doi.org/10.1038/s44161-025-00697-z

Subjects

Abstract

Similar content being viewed by others

Main

Results

Spatiotemporally dynamic endocardial ciliation during cushion EndoMT is blood flow dependent

ECC EndoMT requires primary cilia independent of heart looping

The mechanosensitive transcription factors KLF2 and KLF4 are involved in human valve development and are dynamically expressed in the mouse endocardium during cushion EndoMT

KLF4 expression negatively correlates with cushion EndoMT progression

Endocardial KLF4 expression correlates with ciliation and depends on blood flow

Failure of ciliogenesis results in abnormal KLF4 expression

Defects in both ciliogenesis and blood flow block EndoMT

EndoMT molecular pathways are dysregulated in cilia and ciliary calcium signaling KO hearts

Discussion

Methods

Experimental model details

Mice

Zebrafish

Histology

Whole-mount and cryosection in situ hybridization

Colorimetric

HCR-FISH

Immunofluorescence

Whole mount

Cryosections

Cell death assay

Antibodies

Lightsheet microscopy

Fixed

Live

Imaging parameters and analysis

Microscope details

Imaging details

Analysis details

snRNA-seq

Single-nucleus collection, sequencing and alignment

Single-nucleus Seurat object creation

Single-nucleus analysis

Single-cell analysis

Statistics and reproducibility

Reporting summary

Data availability

Code availability

References

Acknowledgements

Author information

Authors and Affiliations

Contributions

Corresponding author

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Extended data

Supplementary information

Source data

Rights and permissions

About this article

Cite this article

Share this article

Search

Quick links