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

Abstract

Although cardiac abnormalities have been observed in a growing class of human disorders caused by defective primary cilia, the function of cilia in the heart remains an underexplored area. The primary function of cilia in the heart was long thought to be restricted to left–right axis patterning during embryogenesis. However, new findings have revealed broad roles for cilia in congenital heart disease, valvulogenesis, myocardial fibrosis and regeneration, and mechanosensation. In this Review, we describe advances in our understanding of the mechanisms by which cilia function contributes to cardiac left–right axis development and discuss the latest findings that highlight a broader role for cilia in cardiac development. Specifically, we examine the growing line of evidence connecting cilia function to the pathogenesis of congenital heart disease. Furthermore, we also highlight research from the past 10 years demonstrating the role of cilia function in common cardiac valve disorders, including mitral valve prolapse and aortic valve disease, and describe findings that implicate cardiac cilia in mechanosensation potentially linking haemodynamic and contractile forces with genetic regulation of cardiac development and function. Finally, given the presence of cilia on cardiac fibroblasts, we also explore the potential role of cilia in fibrotic growth and summarize the evidence implicating cardiac cilia in heart regeneration.

Cilia are hair-like, microtubule-based cellular organelles of several micrometres in length that protrude from the surface of most animal cells, from protozoa to humans¹. First described >300 years ago by van Leeuwenhoek² and documented by Zimmerman³, cilia were initially believed to function solely for cellular propulsion. However, research from the past two decades has highlighted the central role of cilia in orchestrating numerous cellular sensory and signalling events as well as in the development of numerous diseases^4–9. Of note, reports of developmental defects in the heart associated with ciliopathies have emerged in the past 10 years^10–18, but a consensus on the role of cilia in cardiogenesis has not yet been reached. In this Review, we summarize the mechanisms by which cilia contribute to cardiac left–right axis development, identify a role for cilia in congenital heart disease (CHD) and common cardiac valve diseases, explore the relationship between cardiac cilia and mechanosensation, which potentially links haemodynamic and contractile forces with genetic regulation of cardiac development and function, discuss the evidence linking cilia and fibrotic growth, and explore the potential implications of cardiac cilia in heart regeneration. Together, these findings have advanced our understanding of the role of cilia in cardiac development and function, can help delineate the regulatory mechanisms guiding heart development, and provide important insights into disease pathogenesis, prognostic indicators and novel therapeutic targets to treat CHD.

Cilia.

Membrane-bound, microtubule-based, antenna-like sensory organelles that project from the surface of most animal cells to coordinate numerous signalling pathways during development and in tissue homeostasis.

Congenital heart disease.

(CHD). A disease characterized by a structural abnormality of the heart that is not an acquired condition.

Cilia organization and signalling

Cilia consist of a membrane-bound axoneme containing nine circumferentially arranged doublet microtubules that are composed of A and B subfibres¹ (FIG. 1). These subfibres are held together by nexin links, which extend upwards from the basal body and out into the extracellular space. All types of cilia rely on intraflagellar transport for assembly, maintenance and signalling. Intraflagellar transport involves a specialized bidirectional trafficking system that moves axonemal precursors and cilia membrane proteins into and out of the cilium^19,20. Cilia are traditionally classified into motile, nodal or primary types that follow a 9 + 0 or 9 + 2 ultra-structural arrangement of axonemal microtubules (FIG. 1), which are observable with the use of transmission electron microscopy. Motile cilia typically beat in a planar motion and have two additional central microtubules connected to each of the surrounding doublets by radial spokes that are absent in 9 + 0 axonemes²¹. In addition, 9 + 2 motile cilia contain accessory structures used for motility such as outer and inner dynein arms and central pair projections. Cilia that have a typical 9 + 0 axonemal arrangement usually do not have the outer and inner dynein arms and other accessory structures required for motility and are thus often immotile. These 9 + 0 cilia are frequently referred to as primary cilia and are widely distributed among tissues in vertebrates. In most cases, there is only one primary cilium per cell and it arises from the mother centriole. Although 9 + 2 structures are commonly associated with motile cilia in the setting of multiciliated cells and 9 + 0 structures are associated with immotile cilia, many exceptions to this observation exist^22–27. Notably, motile cilia in the left–right organizer (LRO; also known as the ventral node in mice, Hensen’s node in chickens, Kupffer’s vesicle in zebrafish and the gastrocoel roof plate in Xenopus^28–30), an embryonic structure that determines left–right body axis asymmetry, are mostly 9 + 0 in ultrastructure and have a circular beating motion rather than a planar motion^31–33. Electron tomography studies have further challenged this canonical definition of ciliary subtype by demonstrating that primary cilia have the established 9 + 0 axonemal microtubule arrangement only at their proximal base, with fewer microtubule doublets near the distal tip^34–37. These reports underscore the fact that the ultrastructure of cilia is far more complex than previously thought.

Fig. 1 |. Diversity of ciliary structure.

A cilium is composed of multiple pairs of microtubules (axoneme) running from the basal body. The axoneme is the structural core of a cilium. The microtubule doublets, containing an A-tubule (A) and a B-tubule (B), are continuous with the microtubules at the transition zone. The doublets are connected by nexin links and are held in place by radial spokes that extend into the axonemal centre. Depending on the type of cilium, the cilium backbone can have 9 + 0 or 9 + 2 axoneme configurations. In motile cilia (9 + 2 configuration except at the left–right organizer), inner and outer dynein arms are connected to the A-tubules and mediate cilia bending by reversibly binding to the neighbouring B-tubule. Most 9 + 0 cilia (primary cilia) do not possess inner and outer dynein arms, radial spokes or central pairs. Some 9 + 0 cilia lack the central microtubules only and are motile (nodal cilia). The basal body of each cilium is a specialized centriole that anchors the axoneme to the cell. The basal body is composed of nine triplets of microtubules (9 × 3 microtubular structure), containing an A-tubule, a B-tubule and a C-tubule (C), rooted in pericentriolar material with a proximal amorphous disc, a cartwheel structure, a middle portion that lacks appendages and transition fibres at the distal end. Proximal transition fibres connect each microtubule doublet (without dynein arms) to the membrane. The transition zone functions as a regulated gate that controls the entry and exit of ciliary proteins into the cilium. The transition zone converts the 9 × 3 microtubular structure of the basal body into the axonemal doublet structure. The distal part contains stellate fibre arrays and an amorphous disc structure and gives rise to the central microtubules in 9 + 2 axonemes.

Left–right organizer.

(LRO). Evolutionarily conserved embryonic ciliated organ of laterality in which breaking of left–right asymmetry occurs.

Primary cilia are involved in the maintenance of tissue homeostasis by regulating signalling pathways and sensing biophysical and biochemical changes in the extracellular environment. For example, primary cilia can act as flow sensors and chemosensors that respond to a plethora of extracellular signals^38,39. During development and in the adult organism, primary cilia regulate multiple signalling pathways^39,40, including Hedgehog^41,42, WNT^43,44, transforming growth factor-β (TGFβ)⁴⁵ and Notch^46,47 (FIG. 2a). Critically, many components of these signalling cascades localize to or near the cilium, including core Hedgehog factors such as Patched, Smoothened and glioma-associated oncogene homologue (GLI)^41,42,48,49. Furthermore, primary cilia are specialized calcium signalling compartments, given that intraciliary calcium signalling is distinct from extraciliary calcium signalling and that numerous calcium channels and calcium-binding proteins localize to the cilium^50–55. Of note, signals from distinct ciliary channels can be spatially distinguishable from one another within the cilium because some of these channels and binding partners occupy unique microdomains in the organelle such as the inversin compartment^56,57 and linear arrays formed by polycystin 2 (PKD2; a mechanosensitive transient receptor potential channel)⁵⁸. Indeed, the cilium’s compartmentalization, strategic positioning on the cell surface and unique geometry with a large surface-area-to-volume ratio are all believed to have facilitated its adoption by vertebrate cells as a master signal transduction centre. The structural integrity and performance of the cilium is critical for its function given that defects in cilia contribute to a wide array of multiorgan conditions such as airway dysfunction, anosmia, cancer, cognitive disorders, laterality defects, polycystic kidney disease, retinal degeneration, obesity and sterility as well as hepatic, pancreatic, renal and skeletal defects^{4–9,50,59,60} (FIG. 2b).

Fig. 2 |. Signalling pathways involved in heart development and organs involved in ciliopathies.

a | Diagram of ciliary signalling pathways implicated in cardiac development. The cilium is a master signal transduction centre that orchestrates numerous signalling pathways involved in heart morphogenesis. Members of these pathways identified from screens in animal models and in patients with congenital heart disease are highlighted: WNT signalling pathway proteins are shown in yellow, calcium signalling proteins are shown in green, Hedgehog signalling proteins are shown in red and the proteins in the transforming growth factor-β (TGFβ)–BMP–Nodal pathways are shown in blue. b | Diagram of organ system involvement in ciliopathies. Because motile and immotile cilia are found throughout most cells in the human body, cilia defects result in pleiotropic disorders that affect numerous organ systems such as the heart, lungs, kidney, brain, eyes, ears, craniofacial system, limbs and reproductive system. Motile ciliopathies are shown in pink, sensory ciliopathies in purple and phenotypes attributed to both are shown in blue. GLI, glioma-associated oncogene homologue; PKD, polycystin. Part a adapted from REF.²³⁷, CC BY 4.0 https://creativecommons.org/licenses/by/4.0/. Parts c, e, f, g, h, i adapted with permission from REF.⁵³, Elsevier.

Left–right patterning of the heart

Initiation of cardiac development.

During vertebrate embryogenesis, the early symmetrical heart tube undergoes a dynamic looping stage that results in left–right asymmetry of the developing heart. The first stages of cardiac left–right development are initiated in a transient, evolutionarily conserved, ciliated epithelium known as the LRO. Within the LRO, which is formed after development of the dorsoventral and anteroposterior axes^61,62, both motile and primary cilia have a critical role in establishing the left–right asymmetry of the body axis and proper placement and patterning of the internal organs, including heart looping^32,63–65. Impaired LRO ciliary signalling results in major left–right patterning defects^31,66–68. Motile cilia in the LRO are oriented in a polarized fashion and rotate synchronously in a clockwise manner to produce a leftward fluid flow in the extracellular space (FIG. 3a–c). This leftward flow is essential for defining left–right asymmetry^31,66–70, as shown by the demonstration that artificially applied flow to mouse embryos in vitro can determine left–right patterning⁶⁸. Although approximately 300 cilia are present in the mouse LRO as few as two motile LRO cilia are enough to break the bilateral symmetry in the mouse embryo⁶². Accordingly, optical mapping of flow in the LRO of mouse⁷¹, zebrafish⁷² and Xenopus⁷³ embryos revealed that local regional areas of flow in the LRO, rather than across the entire structure, are critical for left–right development. Indeed, LRO fluid flow is sensed by primary cilia on the cells located solely along the peripheral edges of the LRO, known as crown cells⁷⁴.

Fig. 3 |. Model for intraciliary calcium signalling in left–right patterning.

a,b | Scanning electron microscopy images of left–right organizer (LRO) cilia at the embryonic mouse LRO at 8 days post-coitum. c | Model for intraciliary calcium signalling during left–right development in the zebrafish embryo LRO. At the bud stage (~10 hours post-fertilization (hpf)), cilia (blue) in the LRO are primarily immotile and bilateral symmetry is maintained. At the 1 to 4 somite stage (~11 hpf), cilia motility initiates and drives left-biased fluid flow (orange), triggering Pkd2-dependent (green) intraciliary calcium oscillations (ICOs; red) along the left side of the LRO (1). At the 6+ somite stage (12 hpf), ciliary-to-cytosolic Ca²⁺ waves (2) spread to neighbouring cells to establish leftward-biased stable elevations in calcium around the mesendoderm (3), which in turn direct asymmetrical gene expression (such as dand5, shown in purple) to establish left–right patterning (4). d–i | Calcium imaging of the zebrafish LRO reveals ICOs that are mainly located on the left side of the LRO. e | Representative live images of the LRO zebrafish embryo at the 1-somite stage (~10 hpf) expressing ratiometric genetically encoded calcium indicators (arl13b–GCaMP6s/arl13b–mCherry). The ratiometric signal is shown and false-coloured with a rainbow intensometric scale. White arrowheads indicate LRO cilia at resting stage (left panel), red arrowhead highlights an ICO inducing a cytosolic calcium wave (right panel). f | Rose diagram depicting the spatial distribution and mean percentage of cells per embryo displaying ICOs in the total LRO of embryos (n = 10) spanning the entire course of LRO development (bud to 16-somite stage: ~10–17 hpf). Left-sided waves are shown in blue, right-sided waves are shown in yellow. g | Representative montage of ratiometric fluorescent images of a single LRO cell from the left side or right side of the LRO over 55 seconds of total elapsed time. White arrowheads indicate cilia at resting state and red arrowheads indicate the onset of ICO. h | Two-colour fluorescent intensity (F/F₀) over time (seconds) plot from a single representative LRO cilium (left side) co-expressing arl13b–GCaMP6s and arl13b–mCherry at the 1-somite stage. Multiple ICOs are visible in the Arl13b–GCaMP6s channel (blue trace), whereas the Arl13b–mCherry signal (pink trace) is relatively quiescent. i | Ratiometric intensity (Arl13b–GCaMP6s/Arl13b–mCherry; gold trace) over time (seconds) plot. The arrowhead and black bars indicate the start of multiple ICOs. The ratiometric signal was thresholded (indicated by the grey horizontal line) to facilitate the analysis of calcium oscillation dynamics. A, anterior; L, left; P, posterior; R, right. Parts a and b reprinted from REF.²³⁸, Springer Nature Limited. Parts c and e–i modified with permission from REF.⁵³.

The precise mechanism by which LRO flow is detected is uncertain as is the function of cilia in this process and how flow regulates Nodal signalling to establish left–right asymmetry. Several models have been proposed^63,64. The ‘morphogen flow’ hypothesis suggests that LRO flow creates an asymmetrical distribution of short-lived molecules across the LRO, resulting in the accumulation of these molecules in embryonic fluid flow at the left side of the LRO^31,66. A variation of this morphogen flow model proposes that these soluble molecules are encapsulated in membrane-bound vesicles (also known as nodal vesicular parcels (NVPs)) that are delivered to the left side of the LRO^61,75. Although numerous studies have reported on the presence of ciliary vesicles reminiscent of NVPs in Caenorhabditis elegans, green algae and cultured cells from mice, whether these ciliary vesicles are present in the LRO and are functionally equivalent to NVPs remain uncertain^76–80. An alternative model, which has become known as the ‘two-cilia’ model^32,65, proposes that LRO leftward flow driven by motile cilia initiates a calcium influx in cells on the left side of the LRO via activation of the polycystin ion channels in primary cilia that function as shear flow mechanosensors^{32,53,54,74,81} (FIG. 3c). Indeed, calcium signalling has a critical role in left–right determination, given that one of the earliest molecular events in establishing asymmetry is an increase in cytoplasmic calcium in the left-sided cells of the LRO^32,53,81 and that the cation channel PKD2 is critical for proper left–right patterning of the embryo^74,82. The contribution of calcium signalling to left–right determination will be addressed in more detail below. Given the importance of proper left–right patterning for embryonic development, several mechanisms are likely to occur in concert to guarantee the integrity of this process.

After the detection of LRO flow by primary cilia, left-sided cytoplasmic mesodermal calcium transients are transmitted to surrounding tissues to induce the asymmetrical expression of the left-side determinant Nodal and its feedback inhibitor Lefty. Nodal and Lefty are highly conserved among vertebrates and some non-vertebrates and constitute a self-enhancement and lateral-inhibition system that provides robustness to the left–right determination process^83–85. Asymmetrical Nodal signalling induces left-sided expression of other genes, such as Pitx2 (REF.⁸⁶), which in turn induce asymmetrical morphogenesis of most visceral organs, including the heart, lungs and kidneys⁸⁷.

Calcium signalling and cardiac situs.

A growing line of evidence has highlighted the critical role of calcium signalling in left–right patterning. For more than a decade, the earliest observable molecular event in the process of left–right asymmetry was an increase in cytoplasmic calcium on the left-sided mesoderm of the LRO³². However, several studies have demonstrated that before these left-sided cytoplasmic calcium events take place, intraciliary calcium oscillations occur in the left side of the LRO and initiate vertebrate left–right asymmetry^53,54 (FIG. 3d–i). Intraciliary calcium oscillations were initially only observed in the zebrafish LRO but not in the mouse LRO^53,88; however, a subsequent study detected intraciliary calcium oscillations in mouse embryos that were strikingly similar to those observed in zebrafish embryos⁵⁴. These left-sided intraciliary calcium oscillations resist treatment by thapsigargin (a blocker of the sarco/endoplasmic reticulum calcium ATPase channel that depletes cytosolic calcium stores), thereby demonstrating that intraciliary calcium oscillations are generated independently of cytosolic calcium. These intraciliary calcium oscillations also depend on the presence of LRO flow and PKD2 in both mice⁵⁴ and zebrafish⁵³, occurring in both species at the time point and location at which bilateral symmetry is first broken. Of note, these intraciliary calcium oscillations are the earliest known left–right asymmetrical molecular signal.

Differences in the study protocols could account for the discrepancies between the findings on intraciliary calcium oscillations reported in published studies, in particular the use of different culture media. Indeed, in the study by Mizuno et al., mouse embryos were cultured in Dulbecco’s Modified Eagle Medium (DMEM) containing 75% rat serum⁵⁴, whereas Delling et al. analysed mouse embryos that were cultured in DMEM plus F-12 medium supplemented with 10% fetal bovine serum⁸⁸. Data demonstrating that mouse embryos require culture in medium containing 50–75% rat serum for proper establishment and maintenance of left–right asymmetry^54,89 might explain the failure to detect intraciliary calcium oscillations in embryos cultured without rat serum⁸⁸. Another discrepancy in study protocol that might influence the capacity to detect intraciliary calcium oscillations relates to differences between the genetically encoded calcium indicators (GECIs; particularly in the dissociation constant (K_d) for calcium) used in various studies^{38,53,88,90–92}. Our group⁵³ and Mizuno et al.⁵⁴ used the GCaMP6s and GCaMP6 sensors (~144 nM (REF.⁹⁰) and ~158 nM (REF.⁹¹), respectively) in their studies, whereas Delling et al. used G-GECO1.2 (450–1,100 nM (REF.⁹²))⁸⁸. If LRO cilia contain a low concentration of calcium, the choice of GECIs and the K_d might be instrumental in dictating its capacity for detecting intraciliary calcium oscillations. Another potential factor that can account for discrepant results between studies is variation in the design of GECIs^{37,38,88,93,94}. Our group co-expressed arl13b–GCaMP6s and arl13b–mCherry⁵³, while Mizuno et al. conjugated GCaMP6 to mCherry⁵⁴ using a p2A peptide to allow post-translational cleavage⁹³; both reporters were therefore expressed as a ratiometric system consisting of two separate proteins (with mCherry used as an internal control to correct for motion arte-fact). However, Delling et al. used G-GECO1.2 and mCherry expressed as a direct fusion protein⁸⁸, which might hamper the GECI via Förster resonance energy transfer⁹⁴ and steric inhibition between the fluorophores.

Although further research is required to understand the downstream molecular targets of intraciliary calcium oscillations, the likeliest ultimate target is Dand5 mRNA (also called Cerl2 in mouse, charon in zebrafish and Coco in Xenopus)⁹⁵. Dand5 encodes a Nodal antagonist and is the earliest known asymmetrically expressed gene involved in left–right development^96,97. Intraciliary calcium oscillations are likely to lead to the entry of calcium in the cell bodies on the left side of the LRO, triggering inositol triphosphate-receptor-dependent release of calcium from the endoplasmic reticulum, which induces the degradation of Dand5 mRNA along the left side of the LRO^54,74,81 (FIG. 3c). Given that decay of Dand5 mRNA is post-transcriptionally regulated via its 3′ untranslated region⁹⁸, microRNAs or microRNA-related proteins might be involved in left–right patterning⁹⁹. Once Nodal signalling inhibition is lifted on the left side of the lateral plate mesoderm, left-sided expression of downstream target genes can be induced, which ultimately leads to asymmetrical morphogenesis of most visceral organs¹⁰⁰.

Overall, this model strongly suggests that LRO cells respond to calcium signals. This concept is reinforced by the observation that many calcium signalling proteins are expressed in LRO cells and are required for proper left–right patterning such as calmodulin-dependent kinase II, which senses differential frequency of calcium spikes¹⁰¹, or inversin, a ciliary protein that has calmodulin-binding domains¹⁰². Nonetheless, the mechanism linking intraciliary calcium oscillations and subsequent cytoplasmic transients to Dand5 mRNA degradation remains to be determined.

Impaired LRO ciliary signalling in CHD.

As described above, cilia are essential for the development of vertebrate left–right asymmetry. Failure to establish normal left–right patterning results in severe CHD^13,14,18. Numerous studies have implicated impaired signalling in cilia as one of the mechanisms underlying CHD pathogenesis^10–18,103. Defects in cilia are also known to cause heterotaxy, a syndrome characterized by abnormal left–right asymmetrical body patterning that commonly affects the heart and is tightly correlated with a broad spectrum of CHDs^104–106 (FIG. 4), including most known CHD subtypes, but especially in atrioventricular septal defects (AVSDs) and transposition of the great arteries^104–106 (FIG. 4b). The incidence of heterotaxy is greatly increased in patients with cilia motility defects compared with the general population (6.3% versus 0.004%)^104,106. Mutations affecting cilia structure and function have been identified in patients with heterotaxy and other congenital diseases^107,108 (FIGS 4,5; TABLE 1). Indeed, the incidence of CHD is higher in patients with heterotaxy than in the general population (57% versus 1%), with AVSDs and transposition of the great arteries being the most prevalent CHD subtypes¹⁰⁴ (FIG. 4b).

Fig. 4 |. Cilia defects and left–right patterning impairment leading to heterotaxy and congenital heart disease.

a | Anatomical spectrum of organ and heart laterality associated with heterotaxy and congenital heart disease (CHD). b | Atrioventricular septal defects and transposition of the great arteries are CHDs associated with cilia and laterality defects. Oxygenated blood is shown in red, deoxygenated blood is shown in blue, and a mixture of oxygenated and deoxygenated blood is shown in purple.

Fig. 5 |. Cilia-related genes linked to human congenital heart disease.

Schematic of a cilium and related components that have been associated with the development of congenital heart disease (CHD). Heterotaxy and a range of other CHDs have been associated with alterations of most ciliary components. Dynein and dynein assembly genes are required for ciliary motility, and mutations in these genes are more frequently associated with situs inversus. Mutations in genes required for Hedgehog signalling are more frequently associated with atrioventricular canal defects. See TABLE 1 for a full list of cilia-related genes linked to CHD, the ciliary function or compartment they encode, and the associated heart diseases. AVSD, atrioventricular septal defect.

Table 1 |.

Cilia-related genes linked to human congenital heart disease

Cilia component	Gene	Molecular function	Cardiac phenotype	Refs
Cilia motility	DNAI1	Dynein intermediate chain	SI, HIX, CHD	104,193–195
	DNAAF3	Dynein assembly	SI	196
	DNAH5	Dynein outer arm	SI, HIX, CHD	104,197
	TXNDC3	Dynein outer arm	SI, HIX	198
	DNAH11	Dynein outer arm	SI, HIX, CHD	121,199,200
	DNAAF2	Dynein assembly factor	SI	201
	CCDC39	Dynein assembly	SI, HIX	202
	CCDC40	Dynein assembly	SI	203
	DNAL1	Dynein light chain	SI	204
	CCDC103	Dynein assembly	SI, HIX	205
	HEATR2	Dynein transport and assembly	SI, HIX, CHD	206
	LRRC6	Dynein transcription	SI	207
	CCDC114	Dynein docking	SI, HIX, CHD	208
	ZMYND10	Dynein transcription	SI	209,210
	ARMC4	Dynein assembly	SI	211
	DYX1C1	Dynein assembly	SI, HIX, CHD	212
	C21ORF59	Dynein assembly	SI, HIX	213
	SPAG1	Dynein transport and assembly	SI	214
	CCDC151	Dynein assembly	SI, HIX, CHD	215
	PIH1D3	Dynein assembly	SI	216
	DNAAF1	Dynein assembly	SI	217–219
	CCDC11	CentrioCe	SI, HIX	220
	DNAH6	Dynein inner arm	HIX	221
	TTC25	Outer arm docking	SI	222
PoCycystin complex	PKD2	Ciliary membrane cation channel	SI	223
	PKD1L1	Ciliary membrane cation channel	HIX	224
Transition zone or inversin compartment	NPHP2	Inversin compartment	SI, IGA	130,225,226
	NPHP3	Inversin compartment	VSD, HIX	227,228
	ANKS6	Inversin compartment	SI, CHD	229
	NPHP4	Transition zone	CHD, HIX	129
	NEK8	Inversin compartment	CHD, HIX	195,230
Centriole	MKKS	Centriole	CHD	231
	NEK2	Centriole	HIX	11
BBSome	BBS6	Cilia trafficking	HIX, AVC	124
	BBS17	Cilia trafficking	HIX, AVC	123
LRO formation	GDF1	LRO formation	RAI, CHD	15,232
	GALNT11	LRO formation	HIX	11,233
Others	EVC	Hedgehog signalling	AVC	234
	EVC2	Hedgehog signalling	AVC	234
	MKS1	Hedgehog signalling	AVC	116
	DCHS1	Not reported, possibly near ciliary base	Mitral prolapse	235
	NUP188	Nuclear or ciliary pore	CHD	11,236
	RNF20	Transcriptional regulation	CHD, HIX	108

AVC, atrioventricular canal; CHD, congenital heart disease; HTX, heterotaxy; LRO, left–right organizer; SI, situs inversus; RAI, right atrial isomerism; TGA, transposition of the great arteries; VSD, ventricular septal defect.

Heterotaxy.

(HTX). Any arrangement of the organs across the left–right axis differing from situs solitus and situs inversus.

Atrioventricular septal defects.

(AVsDs). Conditions characterized by improper atrial and ventricular septa and adjoining valve development.

Transposition of the great arteries.

(TgA). A congenital heart defect characterized by a switch in the position of the pulmonary artery and the aorta.

Cilia and cardiac morphogenesis.

The mechanisms involved in the translation of global embryonic asymmetry to asymmetrical cardiac morphogenesis remain unclear. Cells at the LRO do not directly contribute to the heart; cilia at the LRO break bilateral symmetry and provide the initial asymmetrical molecular signal, which is propagated through the lateral plate mesoderm to cardiac primordia^84–86. The signal is initiated by left-sided degradation of Dand5, which then leads to amplification of Nodal signalling on the left side of the embryo, culminating in the expression of transcription factor Pitx2 in the anterior lateral plate mesoderm and the primitive left atrium⁸⁵. The most pronounced cardiac asymmetry is the direction of the heart loop but striking asymmetrical development can also be observed in the venous and arterial vessels, atrial primordia and cardiac valves. The shape and direction of the heart loop have been suggested to result from a combination of ‘buckling’ of the heart tube as it elongates while its endpoints are fixed and amplification of small left–right asymmetries programmed into the poles of the heart by cilia-generated signals arising from the LRO¹⁰⁹. In chick embryos, cardiac cells have inherent chirality that can be oriented by the NODAL protein, providing a potential mechanism linking LRO signalling to cardiac cells¹¹⁰.

Laterality phenotypes can be classified as situs solitus (normal orientation), situs inversus (mirror image orientation of all organs) and heterotaxy (any orientation of organs across the left–right axis diverging from situs solitus or situs inversus) (FIG. 4a). The most severe laterality phenotypes are right and left isomerism resulting from complete failure to break bilateral symmetry¹¹¹. How cilia and LRO asymmetry can affect the final cardiac phenotype is not known but certain indicators can help predict the resulting CHD. When the LRO signal is ‘random’ owing to absent ciliary motility but preserved ciliary sensing, cardiac loop direction and global organ laterality are concordant and random, resulting in the most common phenotype being either situs solitus or situs inversus and normal intracardiac anatomy⁵⁹. Lack of control over loop direction is hypothesized to result in some loops that are abnormally shaped and interfere with normal intracardiac interactions, leading to structural heart disease, which has been seen in mice and patients harbouring mutations that affect cilia motility^104,112. Cardiac morphogenesis is most notably affected by mutations that alter ciliogenesis and signalling. These mutations can lead to bilateral right-sided molecular signatures, such as that observed in the absence of PKD2-mediated cilia sensing¹¹³, or bilateral left-sided molecular signatures resulting from loss of TMEM107 (transmembrane protein 107)¹¹⁴, which is a component of the cilia transition zone that is required for regulation of ciliary protein composition. Of note, cilia are found in the developing heart, and ciliary sensing has a role in cardiac development, including in valve development (discussed below). The presence of defective intracardiac cilia is predictive of more complex CHD owing to defective heart loop formation that results from abnormal left–right asymmetry in addition to aberrations in subsequent heart development^115,116. Although CHD that is linked to abnormal laterality is complex (FIG. 4b), left isomerism is thought to be more frequently associated with obstruction to systemic outflow and right isomerism with obstruction to pulmonary outflow and pulmonary venous return¹¹⁷. Furthermore, Hedgehog signalling (which requires cilia for normal function) at the dorsal mesenchymal protrusion is necessary for cardiac septation^118,119. Even pulmonary venous return might be directed by cilia function, because mice with a conditional mutation in Ift88 (encoding an intraflagellar transport component) had total anomalous pulmonary venous drainage, a form of CHD¹⁰³. Given that cilia function is required for directing heart asymmetry, looping, cardiac septation and valve development, complex forms of AVSD are a hallmark of ciliopathy and laterality-associated CHD¹²⁰.

Together, these studies show that mutations affecting the structure and function of LRO cilia can lead to left–right asymmetry defects associated with CHD. Given that these mutations might also have an effect on the assembly and function of primary cilia formed within the developing heart, establishing whether a heart defect results from impaired LRO cilia signalling, defective cardiac primary cilia or both remains a challenge.

Variants affecting cilia gene function in CHD

Defective cilia have been linked to a broad range of CHDs with and without abnormal laterality in both mice and humans^15,18 (FIGS 4b,5). Targeted sequencing of cilia-related genes in patients with CHD in association with other features of ciliopathies has led to the identification of variants affecting genes required for cilia biogenesis, signalling and motility, all of which contribute to CHD. The link between defective cilia motility and CHD has been well established. In this setting, the resulting CHD is most often laterality associated but also includes many other CHD subtypes such as tetralogy of Fallot, AVSDs and single ventricle defects¹⁵. The variable presentation of the congenital heart defect resulting from mutation of a single cilia dynein gene was first observed in mice homozygous for a mutation in Dnah11 (REF.¹¹²). Subsequently, targeted sequencing of 37 genes linked to primary cilia dyskinesia in 42 patients with CHD associated with heterotaxy syndrome resulted in the identification of possible disease-causing heterozygous variants in 45% of patients, with the DNAH11 gene accounting for 14 of 84 alleles¹²¹. Abnormal thoracoabdominal situs and cardiac looping with or without intracardiac defects are associated with mutations that affect most of the components and processes required for cilia motility, including dynein inner and outer arms, dynein assembly, dynein docking and regulation of dynein transcription¹⁰⁴ (FIGS 1,5). Interestingly, defects in the radial spoke proteins connecting the central pair microtubules to the peripheral A microtubules are associated with respiratory cilia dysmotility but not CHD, which suggests that radial spokes and the central pair microtubules do not have a fundamental role in cilia motility at the LRO¹²².

In addition to cilia motility, ciliary sensing is also required for normal heart development, and mutations affecting primary immotile cilia can likewise contribute to CHD^115,116. The BBSome is an octameric complex required for the trafficking of cilia membrane proteins^123–125. Variants in genes encoding BBSome components are responsible for Bardet–Biedl syndrome and are also associated with CHD, especially AVSDs^124,125. The transition zone and inversin compartment are the proximal components of the cilia axoneme involved in the regulation of cilia signalling, including Hedgehog signalling¹²⁶. Variants in the genes encoding these proximal components result in ciliopathies such as nephronophthisis, Joubert syndrome and Meckel–Gruber syndrome. Patients with these disorders present most commonly with polycystic kidney disease and cerebellar and neural tube defects but several of these gene variants have also been linked to abnormal heart development¹²⁶. Mice with mutations in Jbts17 have pulmonary valve defects¹²⁷, whereas mice with mutations in Mks1 have AVSDs¹¹⁶. Furthermore, variants in TMEM237 (REF.¹²⁸), NPHP4 (REF.¹²⁹) and NPHP2 (REF.¹³⁰) have also been identified in patients with CHD. In addition to evidence linking gene variants that affect cilia function to complex CHD, either isolated or in the setting of ciliopathies or abnormal laterality, these genetic variants have been implicated in heart valve development and disease (discussed extensively below).

The majority of cilia-related genes implicated in cardiac development cause disease in a recessive manner and encode proteins that make up the structure of the cilia or the ciliary signalling machinery^15,131,132. However, monoallelic variants affecting the RNF20 complex (which works in a dosage-dependent manner as a transcriptional regulator of motile cilia function) and transmitted heterozygous variants in genes related to signalling downstream of cilia (such as ROCK2 and SMAD2) have also been linked to CHD^108,133.

The findings from numerous targeted-sequencing studies implicating genes that encode various ciliary components in CHD suggest that cilia have a major role in the pathogenesis of CHD. Given that an increasing number of genes and proteins have been linked to cilia function, one of the major challenges in improving our understanding of the genetics of ciliopathies is the large size of the ‘ciliome’ (the genes and proteins required for cilia function). Estimates of the size and composition of the ciliome in the literature are wide-ranging. The SYSCILIA consortium developed a manually curated list of 428 cilia-associated proteins, of which 187 have been linked to ciliopathies¹³⁴. In 2019, Bayesian integration of 9 independent cilium datasets predicted a total of 956 cilia-related genes¹³⁵. Genome-wide analysis of the contribution of cilia to CHD has now been performed in both mice and humans. A forward recessive mutagenesis screen for severe CHD in mice resulted in the identification of 61 CHD-related genes, 34 of which have been linked to cilia¹⁸. Subsequently, whole-exome sequencing conducted to analyse 2,781 patients with CHD demonstrated that cilia-related gene variants were disproportionately found in patients with laterality disturbances but contribute broadly to CHD, including non-syndromic CHD¹⁵. Further analysis of the same patient cohort revealed that cilia-related genes are enriched for damaging rare, recessive genotypes¹³¹, similar to the observations made in mice. The substantial contribution of cilia-related genetic variation to CHD suggests that some CHDs can be classified as a ciliopathy. However, as some cilia-related proteins might have secondary or additional functions beyond the cilium, ciliopathic defects in the heart might also be caused by cilia-independent mechanisms.

Cilia in valve development and disease

Heart valve defects in ciliopathies.

Patients with ciliopathies often have heart valve defects¹⁷. In particular, patients with nephronophthisis¹³⁶, Ellis–van Creveld (EvC) syndrome¹³⁷, Joubert syndrome¹³⁸, Bardet–Biedl syndrome¹³⁹ or autosomal dominant polycystic kidney disease¹⁴⁰ have an increased risk of mitral valve prolapse, bicuspid aortic valve disease, and tricuspid or pulmonary valve defects^141,142. Variants in ciliary genes associated with Meckel–Gruber syndrome, Bardet–Biedl syndrome and EvC syndrome can cause cardiac valve defects irrespective of left–right patterning defects^104,116. A genome-wide association study revealed a strong link between bicuspid aortic valve disease and genetic variation in human primary cilia, whereby single nucleotide polymorphisms were identified in or near genes related to the regulation of ciliogenesis through the exocyst¹⁴², a protein complex that chaperones cilia cargo to the membrane. Similarly, numerous cilia proteins, such as Smoothened and EvC Ciliary Complex Subunits 1 and 2, are heavily expressed in the endocardial cushion; approximately 60% of patients with a ciliopathy and variants in these genes have endocardial cushion defects^143,144. Improper tissue formation from endocardial cushions as well as other structures, such as the dorsal mesenchymal protrusion, can contribute to the development of AVSDs¹⁴⁵. Interestingly, the first gene implicated in AVSDs in patient studies was CRELD1, which encodes a component of cilia¹⁴⁶. Subsequently, more than 25 known mouse ciliary genes and 7 human ciliary genes have been found to cause AVSDs¹¹⁶. A study that examined how cilia can influence AVSD prevalence revealed that mutations in genes related to motile cilia only contributed to AVSDs if heterotaxy was also present, whereas mutations in genes involved in ciliary signalling contributed to AVSDs irrespective of left–right patterning defects¹¹⁶. These findings suggest that ciliary signalling, not cilia motility, is a driving force behind AVSDs and imply a role for primary cilia in cushion formation beyond establishing left–right asymmetry^16,116,120.

Endocardial cushion.

(ECC). A small pocket of cardiac jelly between the endocardial lining and the myocardium in the atrioventricular canal and outflow tract.

Cilia localization in the developing valve.

The first step in valve formation is the secretion and accumulation of cardiac jelly between the endocardial lining and the myocardium in the atrioventricular canal and outflow tract¹⁴⁷ (FIG. 6a). This process creates small pockets known as endocardial cushions that must be cellularized through endothelial-to-mesenchymal transition (EndoMT), wherein the endocardial lining differentiates and migrates into the cardiac jelly, populating the cushions with mesenchyme¹⁴⁷. EndoMT spans from E9.0 to E10.5 in mice, Hamburger–Hamilton (HH; stages of chick development) 16 to HH22 in chicks and 72 hours post-fertilization (hpf) to 96 hpf in zebrafish. The cushion mesenchyme will then proliferate, increasing cushion size until cushion remodelling begins. Remodelling is a coordinated effort involving differentiation, migration and condensation of the mesenchymal cells, and spans from E12.5 to E16.5 in mice, HH22 to HH28 in chicks and 96 hpf to 166 hpf in zebrafish¹⁴⁷. Endocardial primary cilia are found both upstream and downstream of the developing atrioventricular and outflow tract cushions in chick hearts at HH17 (REF.¹⁴⁸). These cilia become increasingly scarce over time at the narrowest section of the atrioventricular canal and outflow tract where EndoMT is occurring owing to higher shear stress. Given that the atrioventricular canal is subjected to higher shear stress between two contractile regions, the outflow tract retains its cilia for a longer period of time. Mesenchymal primary cilia, which are not subject to shear stress, are present throughout valve development. In mice, cilia are similarly found on both the endocardium and mesenchyme of the atrioventricular and outflow tract cushions from E8.5 to E12.5, which encompasses the major stages of cushion development after the onset of contraction^{18,115,141,149}. In zebrafish, endocardial primary cilia can be found protruding into the lumen throughout the heart from 24 hpf to 96 hpf (REFS^150,151). Just as in chicks, mice and zebrafish show a regionalized decrease in endocardial cilia in areas of higher shear stress where EndoMT occurs^115,141,150 (FIG. 6a).

Fig. 6 |. Spatiotemporal distribution of primary cilia during valve development.

The initial critical steps of cushion formation and remodelling in three vertebrate model organisms are presented (part a). The cellular processes are conserved across mice, chicks and zebrafish, and the figure indicates the spatial context of the forming valves within the heart at relevant times during development in each model. Haematoxylin and eosin (H&E)-stained sections of mouse hearts at E12.5 (part b). Left panels show hearts from wild-type (WT) mice, middle panels from mice with mutant Lrd (encoding left–right dynein), which have left–right patterning defects, and right panels from mice with deletion of Kif3a (encoding a motor subunit of the heterotrimeric kinesin complex). The arrows pinpoint the lack of cushion cellularization in cilia mutants irrespective of left–right patterning defects. Histopathological images of mouse hearts with deletion of Cc2d2a (which encodes a basal body component) at E15.5, which have normal outflow tract cushions but atrioventricular (AV) cushions with dextrocardia (part c) and AV septal defects (AVSDs) (part d). Mutant AV cushions are improperly formed (part e). Mice with conditional deletion of Ift88 (Nfatc1^Cre;Ift88^f/f), which encodes an intraflagellar transport component, have cushion remodelling phenotypes that result in bicuspid aortic valve (BAV) defects compared with mice with WT Ift88 (Nfatc1^Cre−;Ift88^f/f) (parts f, g). This phenotype was highly penetrant, being observed in 19 of 28 mice with the genetic deletion. The left panel shows H&E-stained sections and the right panel shows 3D reconstruction images depicting the penetrance of this phenotype. A, atrium; Ao, aorta; Atr, atrium; AVC, atrioventricular canal; ECC, endocardial cushion; EndoMT, endothelial-to-mesenchymal transition; HH, Hamburger–Hamilton; hpf, hours post-fertilization; LA, left atrium; LC, left coronary; LV, left ventricle; mLV, morphological left ventricle; mRV, morphological right ventricle; NC, non-coronary; OFT, outflow tract; PA, pulmonary artery; RA, right atrium; RC, right coronary; RNC, right non-coronary; RV, right ventricle; V, ventricle. Part b reprinted with permission from REF.¹¹⁵, Wiley. Parts c–e reprinted from REF.¹⁸, Springer Nature Limited. Parts f and g reprinted with permission from REF.¹⁴¹, Wiley.

Deciliation and reduction in cilia length in response to strong fluid flow have been previously reported in human umbilical vein endothelial cells¹⁵². Interestingly, modulation of cilia length, as well as the presence of cilia, can alter genetic signalling in the tissues that these cilia occupy^153,154. Markers of mechanical stress, such as KLF2 (encoding Krüppel-like factor 2 (KLF2)), are expressed in regions with deciliation and reduced cilia length^148,155, suggesting that regional deciliation is involved in the modulation of mechanosensitive signalling in the endocardium.

Atrioventricular canal.

(AVC). The channel connecting the developing atrium and ventricles in embryonic development.

Outflow tract.

(OFT). The channel connecting the developing ventricles and dorsal aorta in embryonic development.

Endothelial-to-mesenchymal transition.

(EndoMT). The differentiation of endothelial cells to mesenchyme during cardiac cushion development.

Primary cilia in heart cushion formation.

Loss of primary cilia has been found to prime the endothelium for EndoMT, a critical process for cardiac cushion development¹⁵⁶. Mutant mouse endothelium that lacks cilia undergoes EndoMT in response to shear stress. However, when wild-type endothelium is exposed to sufficient shear stress, resulting in deciliation, the endothelium also undergoes EndoMT¹⁵⁶. Interestingly, while the mechanosensitive Klf2 and Klf4 genes are increased in response to shear stress in wild-type endothelium in mice, removal of cilia inhibits the induction of Klf2 expression and significantly downregulates KLF4 at the mRNA and protein levels in response to shear stress^150,157. This observation contradicts earlier findings of Van der Heiden et al., who reported that only regions of endocardium with few or no primary cilia expressed KLF2 in chick¹⁴⁸. Whether this discrepancy is related to differences between endothelial populations (kidney versus heart) or distinct regulatory mechanisms is not known. This finding suggests that primary cilia are required in regions of higher shear stress (such as the atrioventricular canal) to induce Klf2 expression before being lost, allowing Klf2 expression to persist while Klf4 (which encodes an inhibitor of EndoMT) is downregulated. This theory can help explain how modulation of gene expression and the resulting EndoMT and cushion formation can be controlled by gradual, regional deciliation.

Mutant mice lacking Kif3a or Kif3b (which encode motor subunits of the heterotrimeric kinesin complex) are unable to assemble cilia and are non-viable by E10.5 owing to heart failure⁶⁹. These mouse mutants have a complete loss of cushion cellularity not seen in other mutants with defective left–right patterning and irrespective of heart looping, indicating a severe EndoMT defect independent of its left–right randomization^69,115,158 (FIG. 6b). A less severe defect in cushion cellularization can be seen in mice lacking Pkd2 (which encodes a mechanosensitive transient receptor potential cation channel), suggesting that abnormal cilia calcium signalling can contribute to inhibition of cushion cellularity^115,159,160. Similarly, targeted insertional Ift88^−/− mice, which lack a subunit of the intraflagellar transport complex B, have residual, but few, cilia present with reduced cushion cellularity and are non-viable by E11.5 (REFS^158,161). Therefore, primary cilia might act as mechanosensors in cushion formation just as they do in the endothelium. Overall, these findings support a role for cilia, deciliation and shear stress in initiating EndoMT during cushion formation in the heart.

Primary cilia in valve leaflet remodelling.

In addition to their role in early valve formation, primary cilia also contribute to later stages of valve development during leaflet remodelling^{141,142,162,163}. In mice at E12.5, selective loss of mesenchymal primary cilia in the atrioventricular cushions through loss of the basal body component CC2D2A resulted in remodelling defects and AVSDs, whereas the outflow tract cushion remained unaffected¹⁸ (FIG. 6c–e). To discriminate between ciliary signalling in valvulogenesis and left–right patterning, study investigators assessed mutant mice with a hypomorphic allele of Ift88 (Ift88^cbbs) established from chemical mutagenesis that leads to reduced cilia numbers but normal left–right development¹⁴⁹. These mutants live longer than Ift88^−/− mice and also have impaired cushion remodelling and AVSDs. Other mutant mouse models that have reduced cilia numbers through conditionally deleted intraflagellar transport components also evade left–right patterning defects. Conditional knockout of Ift88 in the endocardium using a Nfatc1–Cre driver results in a phenotype that is reminiscent of myxomatous bicuspid aortic valve disease¹⁴¹ (FIG. 6f,g). The increased cusp size was secondary to increased extracellular matrix (ECM) production, suggesting that primary cilia can repress collagen deposition and maturation to control leaflet size. Subsequently, RNA sequencing confirmed that the conditional mutation leads to the upregulation of ECM-related gene pathways in the anterior mitral leaflets¹⁶². Interestingly, conditional knockout of Klf2 in the endocardium using the same Nfatc1–Cre driver resulted in a similar cushion remodelling phenotype, further suggesting a role for cilia in modulating the expression of Klf2 in vivo¹⁶⁴. Similar to the conditional Ift88 knockout, Ift88^orpk mutant mice (also referred to as polaris), which is a hypomorphic allele established by insertional mutagenesis, also have increased ECM deposition in the heart at E11.5 (REF.¹⁵⁶). Furthermore, Ift88^orpk mice have renal cysts, polydactyly and hydrocephalus but survive past birth and do not have left–right patterning defects¹⁶⁵. Mutations affecting the exocyst, a shuttling complex necessary for ciliogenesis and ciliary signalling, result in aortic valve defects in both mice and zebrafish¹⁴². Endocardial expression of Desert Hedgehog is also necessary for cilia-induced activation of the TIAM1–RAC1 axis, which in turn stimulates ECM remodelling necessary for proper valve remodelling¹⁶³. Perturbations in this paracrine system lead to adult myxomatous mitral valve disease, which is related to the mitral valve prolapse observed in many patients with ciliopathies¹⁶³. Although disease pathogenesis in patients with mitral valve prolapse was believed to be through gradual valve degeneration, the latest evidence points to two major disease phases, wherein secondary effects such as inflammation can aggravate a pre-existing congenital valve defect^162,163. These findings demonstrate how defects in cilia-dependent leaflet remodelling during development can prime conditions such as mitral valve prolapse.

In contrast to the findings by Slough et al., who reported that mutant mice lacking cilia had no cushion cellularization¹¹⁵, mice with NFatc1–Cre conditional Ift88 mutations have normal cushion cellularization, suggesting that endocardial primary cilia are not necessary for EndoMT¹⁴¹. However, given that some cilia components are long-lived, cilia priming of EndoMT should already be under way after induction of Nfatc1 expression¹⁶⁶ (which in mice occurs at E9.5 at the earliest¹⁶⁷). Removing Exoc5 (encoding a subunit of the exocyst complex) with the use of the same Nfatc1–Cre driver mediates cilia loss by E13.5 (REF.¹⁴²), well past the stage at which EndoMT occurs (E9–E10.5). The product of Nfatc1 has also been identified as an inhibitor of EndoMT; therefore, endocardial cells expressing Nfatc1 will not undergo EndoMT regardless of the presence of primary cilia¹⁶⁸. This finding might explain the discrepancy observed between the conditional Ift88 mutant mice and the Kif3a-knockout mice used by Slough et al.¹¹⁵, which have severe EndoMT defects irrespective of left–right randomization but still indicate a role for primary cilia in remodelling.

Cilia as mechanosensors in valve development

Primary cilia as endothelial mechanosensors.

Primary cilia are known to act as mechanosensors in endothelial tissue of the vasculature^169,170 and kidney¹⁷¹. Mechanical stimulation of cilia on Madin–Darby Canine Kidney (MDCK) cells resulted in an influx of cytosolic calcium¹⁷². The same response was elicited by fluid flow over the endothelium and removal of cilia by treatment with chloral hydrate or by genetic knockdown approaches ablated this mechanosensitive response^173,174. Mouse endothelial cells with knockdown or knockout of Pkd1 or Pkd2 have the same loss of calcium influx as endothelium lacking cilia, suggesting that cilia-localized PKD1 and PKD2 are important for the transduction of mechanical signals¹⁷⁵. This finding also mirrors the observation that cilia localization of PKD2 in the mouse LRO is required for sensing LRO flow and normal left–right development⁷⁴.

Increased fluid flow in zebrafish vasculature leads to increased cilia deflection and calcium influx in a dose-dependent manner in vivo³⁵. Furthermore, a reduction in blood viscosity or fluid shear stress as well as a loss of contractility, cilia or pkd2 expression can all individually result in loss of endothelial calcium influx in zebrafish³⁵. Similar to findings in the heart, endothelial cilia gradually deplete as flow increases. Intriguingly, vascular endothelial cilia have incomplete microtubules towards the end of the axoneme, which potentially allows for greater flexibility and sensitivity. This observation lends support to the hypothesis that vascular endothelial cilia are necessary for mechanosensation in low-flow settings and are subsequently lost when no longer required.

Primary cilia as endocardial mechanosensors.

The endocardium of chicken arterial vasculature has a higher sensitivity to mechanical stimuli than endothelium in vivo^155,176. Furthermore, shear stress-induced gene expression is higher in chicken ciliated endocardium than in non-ciliated endocardium¹⁷⁷. The removal of endocardial cilia in chicken by the administration of chloral hydrate diminished shear stress-induced gene expression to the level of expression in the arterial endothelium. These findings suggest that the developing heart is more sensitive to mechanical cues than the vasculature, wherein smaller stress increments can affect resulting genetic programmes. Furthermore, these observations indicate that specialized mechanosensors are present in the developing heart that are sensitive to these smaller stress increments.

The role of primary cilia as endocardial mechanosensors in both the setting of the cardiovascular system and cilia biology has been a topic of debate¹⁷⁸. The main argument against primary cilia as mediators of mechanical stress during valve development was formed on the basis of the observation that cilia are absent from the endocardium at seemingly critical stages of valve development owing to deciliation. Even before deciliation, not all cells are ciliated at one time, leading to the question of how mechanosensitive gene expression could be initiated ubiquitously without the mechanosensor present on every cell¹⁷⁹. However, a two-phase hypothesis might address this issue. First, primary cilia can influence the endocardial response to flow via cilia signalling. Second, the cilia can amplify mechanical signals on the cytoskeleton to allow instantaneous and lasting gene expression. In this model, cilia are transient by design to enable spatiotemporal regulation of mechanosensation^{35,148,156,177,180}. Cytoplasmic calcium can also be subsequently induced through cell–cell junctions or communication, removing the need for complete ciliation¹⁷⁴. However, given that deciliation has been found to influence gene expression, prior resorption of cilia on the endocardium might be a critical priming step in valve development¹⁵⁶.

Numerous lines of evidence have implicated primary cilia as endocardial mechanosensors through the same pathways used in vascular endothelium (FIG. 7). In zebrafish, mutations in pkd2 result in both the loss of klf2a expression and cushion cellularization to the same degree as loss of contractility with lidocaine treatment, thereby demonstrating the importance of cilia calcium influx for endocardial mechanosensation and EndoMT^160,181. Zebrafish with an ift88 knockdown have reduced endocardial notch1b expression in a flow-dependent manner but notch1b loss only occurred within a narrow timeframe when cilia were pharmacologically removed, indicating that primary cilia are needed for EndoMT-dependent Notch1b activation at the onset of flow¹⁵⁰. As in vascular endothelium, primary cilia in the endocardium are hypothesized to detect flow through activation of calcium transient receptor potential channels localized in cilia, such as Pkd2, that are stretched in response to low laminar flow^150,182. In zebrafish, endocardial Notch signalling is also involved in the regeneration of the heart, a process that requires blood flow and subsequent klf2a expression¹⁸³. Knockdown of ift88 prevented the upregulation of the expression of both klf2a and notch1b in the atrioventricular canal during ventricular regeneration^151,182. Whether these mechanisms are dependent on cilia calcium influx or whether primary cilia similarly act as endocardial valve mechanosensors outside of ventricular regeneration remains unclear. Together, these data implicate primary cilia as critical mediators between laminar flow and genetic regulation of pathways responsible for complex cellular processes in the endocardium.

Fig. 7 |. Primary cilia as endocardial mechanosensors.

Schematic depicting the hypothesis that primary cilia serve as endocardial mechanosensors through cilia-dependent calcium influx during low blood flow in a zebrafish embryo at 72 hours post-fertilization (hpf). Polycystin 2 (Pkd2) must be localized to primary cilia to initiate cytoplasmic calcium influx in the endothelium. Perturbations at every level of the mechanosensory pathway would result in the same loss of target genes downstream of this pathway: cessation of blood flow or contraction (induced by tnnt2a morpholinos (MO; nucleotide analogues that recognize and bind short nucleotide sequences to modify or block mRNA transcription) or the silent heart mutation), reduction in blood viscosity and resulting shear stress during blood flow (induced by gata1 MO or gata2 MO), blocking of cilia calcium influx (in pkd2 knockout (KO)), removal of cilia (with ift88 MO) or removal of the downstream mechanosensitive transcription factor Krüppel-like factor 2a (in klf2a KO). A, atrium; AVC, atrioventricular canal; OFT, outflow tract; V, ventricle.

Cilia in cardiac fibrosis and regeneration

Studies within the past 5 years have implicated primary cilia and polycystin signalling in the process of heart regeneration. Although primary cilia have been located in the embryonic and adult heart of several species¹⁸⁴, including humans^185,186, the specific cell location and function of cilia remain unclear. A 2019 study reported that, in mouse, rat and human hearts, primary cilia were specifically found in cardiac fibroblasts not cardiomyocytes¹⁸⁷. Of note, the study investigators found that, after heart injury, ciliated fibroblasts accumulated at the sites of injured myocardium. In these cells, cilia and PKD1 participate in the fibrogenic response observed after injury by inducing TGFβ1–SMAD3 activation, the production of ECM proteins and cell contractility, processes that are all compromised in fibroblasts depleted of primary cilia. These results suggest that signalling in primary cilia after cardiac injury in mammals leads to a maladaptive fibrogenic response that impairs cardiac regeneration. This notion is further reinforced by findings that, after myocardial infarction, the induction of cilia disassembly in epicardial cells via short hairpin RNA-mediated silencing of Ift88 between postnatal days 7 and 28 in mice stimulates cardiac regeneration by promoting epithelial-to-mesenchymal transition and reducing scar formation¹⁸⁸. Modulating cilia response after cardiac injury in mammals might therefore help to promote regenerative responses, and uncovering the mechanisms involving cilia in cardiac regeneration in adult animals might provide valuable insights. For example, in zebrafish, cardiac injury induces cardiomyocyte proliferation¹⁸⁹, with Notch signalling activation in the endocardium being essential for regeneration of the myocardium^190–192. This activation was mediated by haemodynamic changes that occur after injury¹⁸³. Interestingly, mechanical haemodynamic activation of Notch signalling after ventricular ablation was mediated by endocardial primary cilia via upregulation of klf2 expression¹⁵¹. How primary cilia transduce mechanical shear stress to regulate Notch signalling to ultimately promote heart regeneration remains to be determined.

Conclusions

Although cilia were once meekly described as ‘eyelashes’ on cells by early microscopists, this tiny but mighty organelle’s essential function in heart development and disease is now firmly in the limelight. Advances in research into intraciliary calcium signalling have addressed major long-standing gaps in our understanding of cardiac left–right development. Genomic and patient studies have critically linked cilia function and biogenesis to several cardiovascular disorders, namely CHD and mitral valve prolapse. Pioneering work in animal models has unravelled surprising roles for intracardiac cilia in haemodynamic mechanosensation, valvulogenesis and myocardial regeneration. Despite this immense progress, more research will be necessary to uncover how primary cilia integrate the different signalling pathways they are associated with during heart formation and how these affect different stages of heart development. To do so, discriminating between ciliary signalling in the LRO and embryonic and adult cardiac tissues will be crucial. Generating conditional knockout models of primary cilia from different heart tissues at distinct time points could help to establish the contribution of cardiac primary cilia to the various steps of cardiac morphogenesis and maturation independently of their involvement in the LRO during early left–right specification. The strong association between ciliopathies and congenital defects in the heart irrespective of heterotaxy indicates that primary cilia act as important genetic regulators of cardiac morphology and function throughout development. Ultimately, a deeper understanding of the mechanisms and function of cilia in the heart might lead to the development of novel diagnostic and therapeutic strategies to manage the increasing number of ciliopathic cardiovascular disorders.

Key points.

• Cilia are antenna-like organelles that extend from most eukaryotic cells to obtain and interpret information from the extracellular environment; impaired ciliary signalling has been linked with congenital heart disease.

• Intraciliary calcium signalling in the embryonic left–right organizer initiates vertebrate left–right patterning; abnormal left–right asymmetry is associated with major congenital heart disease, especially in heterotaxy syndrome.

• Impaired cilia function and signalling are associated with heterotaxy, congenital heart disease, mitral valve prolapse and numerous other cilia-related disorders with cardiac abnormalities.

• Primary cilia have a role in heart development beyond establishing left–right asymmetry.

• Endocardial primary cilia might translate mechanical signals to gene expression during heart valve development.

• Fibroblast cilia have been implicated in the regulation of cardiac regeneration.