Loss-of-function mutations in KIF14 cause severe microcephaly and kidney development defects in humans and zebrafish

Abstract

Mutations in KIF14 have previously been associated with either severe, isolated or syndromic microcephaly with renal hypodysplasia (RHD). Syndromic microcephaly-RHD was strongly reminiscent of clinical ciliopathies, relating to defects of the primary cilium, a signalling organelle present on the surface of many quiescent cells. KIF14 encodes a mitotic kinesin, which plays a key role at the midbody during cytokinesis and has not previously been shown to be involved in cilia-related functions. Here, we analysed four families with fetuses presenting with the syndromic form and harbouring biallelic variants in KIF14. Our functional analyses showed that the identified variants severely impact the activity of KIF14 and likely correspond to loss-of-function mutations. Analysis in human fetal tissues further revealed the accumulation of KIF14-positive midbody remnants in the lumen of ureteric bud tips indicating a shared function of KIF14 during brain and kidney development. Subsequently, analysis of a kif14 mutant zebrafish line showed a conserved role for this mitotic kinesin. Interestingly, ciliopathy-associated phenotypes were also present in mutant embryos, supporting a potential direct or indirect role for KIF14 at cilia. However, our in vitro and in vivo analyses did not provide evidence of a direct role for KIF14 in ciliogenesis and suggested that loss of kif14 causes ciliopathy-like phenotypes through an accumulation of mitotic cells in ciliated tissues. Altogether, our results demonstrate that KIF14 mutations result in a severe syndrome associating microcephaly and RHD through its conserved function in cytokinesis during kidney and brain development.

Introduction

Congenital anomalies of the kidney and urinary tract (CAKUT) comprise a broad spectrum of renal and urinary tract malformations of varying severity. Kidney defects can range from unilateral or bilateral renal agenesis to renal hypodysplasia (RHD) or include multicystic kidney dysplasia (1). The latter can also be part of the phenotypic spectrum of fetal forms of ciliopathies [Meckel–Gruber syndrome (MKS)], linked to mutations in genes encoding key components of primary cilia (PC) (2). RHD results from defects during differentiation of the metanephros, initiated by a reciprocal induction between the ureteric bud (UB) and the metanephric mesenchyme. The UB invades the metanephric mesenchyme and undergoes several rounds of branching, expanding through the proliferation of UB tip epithelial cells to form the collecting ducts. Inductive signals from the UB drive the condensation of metanephric mesenchyme cells at the tips of the UB branches to form the cap mesenchyme. These cells undergo a mesenchyme-to-epithelial transition to form the renal vesicle, which will differentiate into several intermediate structures (comma body, then S-shaped body) to finally form the nephron, the functional unit of the kidney. The proximal region of the S-shaped body will differentiate to give rise to the blood-filtering glomerulus, while the distal region will connect to the collecting ducts. Nephrogenesis takes place in the cortex of the developing kidneys and ends before birth (35–36 weeks) in humans (3,4).

The mouse has been a widely used model to decipher the mechanisms controlling nephrogenesis, as many of the processes involved in metanephric kidney development are well conserved (3,4). However, zebrafish have recently become a powerful alternative model for both kidney development and disease modelling (5,6). The pronephros of zebrafish larvae is functional and consists of two nephrons with a fused glomerulus. Early development of the zebrafish pronephros requires the condensation of progenitors into two epithelial tubes, which retain a pattern of segmentation similar to that of mammalian nephrons. A functional kidney is formed by 48 h postfertilization (hpf) through the frontward migration of cells towards the glomerulus, along with spatially and temporally controlled waves of mitosis (5,6).

To date, mutations resulting in RHD in humans have been identified in more than 50 genes with a role in kidney development, revealing a great genetic heterogeneity (1,7). Most of these genes encode transcription factors (PAX2, SIX1, EYA1 and HNF1B) and actors in key signalling pathways (RET-GDNF, FGF and NOTCH). Few of these genes were associated with bilateral renal agenesis (8–10). Interestingly, biallelic loss-of-function (LOF) mutations in KIF14, encoding a mitotic kinesin required for cell abscission, were recently reported in fetuses presenting with bilateral renal agenesis or RHD, associated with complex brain malformations including very severe microcephaly (11,12).

Microcephaly is a condition in which fetuses or infants present with a very small head and brain and are part of a family of highly heterogeneous disorders grouped under the term of malformations of cortical development (MCD). The initial phase of brain development depends on an intense phase of progenitor proliferation within the ventricular zone. Symmetrical mitosis of these progenitor cells occurs at the apical face of the neuroepithelium, lining the lumen of the neural tube. Following this initial amplification phase, cells begin to divide asymmetrically and waves of committed, postmitotic cells undergo an active phase of migration (13). Defects in any of these complex events lead to MCD or microcephaly. Microcephaly, like CAKUT, is genetically heterogeneous, with mutations identified in genes regulating signalling pathways, microtubule dynamics and mitotic spindle orientation, including several kinesin family members (14).

Kinesins are molecular motors known to bind their respective cargos through their C-terminal domain and microtubules through their motor domain. Following these interactions, the kinesin then moves along microtubules, usually in a plus-end-directed manner, through the hydrolysis of ATP (15). KIF14 is a member of the kinesin-3 family, which is highly conserved during evolution and possesses a slow processive activity (16). It was identified as a key actor of cytokinesis (17,18), the last step of mitosis leading to separation of the two daughter cells (19), through its interaction with PRC1 and citron kinase (CIT) (18,20). KIF14 accumulates at the spindle midzone and the midbody, where it is recruited by CIT (18). KIF14 also acts reciprocally to ensure the correct distribution of CIT within this latter structure (20,21). Although initially considered an artefact, the formation of the midbody, or Flemming body, is now known to be crucial for the completion of mitosis (22). Proteins involved in cleavage furrow ingression accumulate between the two parts of the intercellular bridge and recruit components of the ESCRT (Endosomal Sorting Complex Required for Transport) machinery to ensure scission of the plasma membrane at one side or both sides of this central structure, which is then inherited by either one daughter cell or ‘secreted’ into the extracellular milieu (22). Unsurprisingly, therefore, depletion of KIF14 in HeLa cells leads to cytokinesis defects and binucleated cells (17,18). Besides its role during cytokinesis, KIF14 regulates migration in cancer cells via C-terminal interactions with RADIL (23).

In vivo, loss of KIF14 in Drosophila leads to cytokinesis and developmental defects (24–27) and, in mice, to microcephaly and growth retardation (28). In humans, in addition to the syndromic cases associating severe microcephaly and RHD (11,12), mutations in KIF14 were recently identified in isolated, mild to severe microcephaly (29,30), with a phenotypic spectrum similar to that observed in cases of CIT mutations (31–34). In these reports, microcephaly was not associated with major kidney malformations in any of the individuals harbouring mutations in either KIF14 or CIT. In addition, as the clinical features of syndromic cases showed a partial overlap with the phenotypic spectrum of MKS (intrauterine growth restriction, cystic kidneys and brain developmental defects, including cerebellar hypoplasia and vermis agenesis), KIF14 was subsequently termed MKS12 (OMIM 616258). All proteins encoded by MKS genes identified thus far are present at the PC and so KIF14 was proposed to have a ciliary function.

The main objective of this study was to further characterize the phenotypic spectrum associated with mutations in KIF14 and to understand better its role during kidney development. We describe novel biallelic KIF14 mutations in fetuses presenting with RHD and microcephaly. We analyse the functional effects of the mutations identified in four families and their consequences on cytokinesis and the potential role of KIF14 in ciliogenesis. We also use the zebrafish model to investigate the consequences of the loss of KIF14 in vivo, demonstrating a conserved function during both kidney and brain development. Finally, we provide direct evidence for similarities and distinctions between mitotic events occurring during proliferation in both brain and kidney development.

Results

Mutations in KIF14 cause a lethal, highly penetrant syndromic CAKUT with microcephaly

In addition to the initial family that we described several years ago (11), we further identified four families (families 1 to 4) with fetuses presenting with highly reminiscent phenotypes, microcephaly and cystic RHD or bilateral renal agenesis (Fig. 1A and B; Supplementary Material, Table S1), linked to biallelic mutations in KIF14. Mutations identified in families 1 and 2 have been reported recently (12), while those in families 3 and 4 have never been described.

Figure 1

Identification of mutations in KIF14 in a lethal, fetal syndrome associating severe microcephaly with bilateral renal agenesis, or RHD. (A) Pedigrees of the identified families (F1–F4) with homozygous or compound Htz mutations in KIF14. (B) Schematic representation of KIF14 cDNA (exons) and protein showing the position of the identified mutations. The PRC1 and citron kinase (CIT) binding domains are shown, along with the motor, FHA and the four coiled-coil (CC I–IV) domains. The compound Htz mutations of the previously identified family are also shown (11). Abbreviations: heterozygous (he), homozygous (ho). (C) Fetus 21 of family 1 (18 weeks gestation), pictures of brains and kidneys with representative histological sections from one kidney. (D) Fetus 22 (37 weeks gestation) and fetus 23 (18 weeks gestation) from family 2, pictures of brains and kidneys along with histology of F23 kidneys. (E) Fetus F23 of family 3 (18 weeks gestation), pictures of brain and kidneys along with kidney histology. Black arrows, kidneys; black arrowheads, glomeruli; asterisks, cystic glomeruli; white arrows, telencephalon; and white arrowheads, cerebellum.

Figure 2

C-terminally truncated variant KIF14 proteins are constitutively targeted to microtubules without impairing CIT binding. (A–F) Immunofluorescence of HeLa cells transfected with a GFP-fusion encoding WT or mutant KIF14 GFP fusions. Stable microtubules were stained using an antibody against AcTub (red), and nuclei using Hoechst (blue). Merge colour pictures (top panels) and black and white pictures corresponding to GFP fusions (KIF14; bottom panels) are shown. Scale bars: 5 μm. (G) Lysates from HEK293 cells cotransfected with plasmids encoding GFP-fusions of WT or mutant KIF14 and with an mCherry-fusion of the KIF14 binding domain of CIT (CCf) were immunoprecipitated using an anti-GFP antibody. Lysates and immunoprecipitated proteins were analysed through western blotting, using antibodies against GFP (KIF14) or mCherry (CIT-CCf). (H) Schematic representation of the constructs used in immunofluorescence and coimmunoprecipitation experiments.

Whole- or targeted-exome sequencing approaches enabled the identification of biallelic pathogenic variants in KIF14 in all these families (Fig. 1A and B; Supplementary Material, Table S1). The affected fetus in family 1 harboured a homozygous out-of-frame deletion of exons 23, 24 and 25 (c.[3567−?_4072+?del]) leading to a frameshift (p.Arg1189Argfs^*9). Compound heterozygous (Htz) variations associating a predicted damaging missense variation in the motor domain (c.1090C>T [p.Arg364Cys] or c.1367C>T [p.Thr456Met]) with a nonsense mutation predicted to lead to a loss of the C-terminal region of the protein (c.3910C>T [p.Gln1304^*] or c.4138C>T [p.Gln1380^*]) were identified in families 2 and 3, respectively. Finally, the affected fetus in family 4 harboured a homozygous nonsense mutation predicted to lead to an early truncation in the motor domain (c.1792C>T [p.Arg598^*]). All variations were absent from in-house and public databases (gnomAD), and missense variants were predicted damaging by PolyPhen and Sift (0.787/0.01 for Thr456Met) (12). Segregation of the identified mutations with the developmental defects was confirmed for all the families by Sanger sequencing. The parents were Htz, while affected fetuses, where DNA was available, harboured the corresponding biallelic variations in KIF14 (Supplementary Material, Table S1 and Fig. S1A–C) (12).

Despite the heterogeneity of the variations identified in these families, all the affected fetuses presented with strikingly similar severe brain and kidney phenotypes (Supplementary Material,Table S1). All affected fetuses presented with microcephaly with a flattened forehead (Fig. 1C–E; Supplementary Material,Fig. S1D). Autopsy revealed a strong delay in the development of the telencephalon (white arrows) leading to agenesis of the occipital lobes and corpus callosum. Additionally, all affected fetuses presented hypoplasia of the cerebellum (white arrowheads) with foliation delay (Fig. 1C–E; SupplementaryMaterial, Fig. S1 and Table S1). The kidney phenotype was equally severe, with either bilateral renal agenesis (3 of 11 cases), severe non-cystic (2 of 11) or cystic RHD (6 of 11; Fig. 1C and E, black arrows). Histochemistry of the cystic kidneys revealed, in addition to large cortical cysts, a lack of nephrogenic zone and absence of corticomedullary differentiation (Fig. 1C–E). Few glomeruli could be observed (black arrowheads), some were cystic (in families 1, 2 and 4; black asterisks), and tubules were rare (Fig. 1C and E; Supplementary Material, Figs S1 andS2) and surrounded by undifferentiated mesenchymal tissue. Immunohistochemistry confirmed the lack of nephrogenic zone (PAX2/SIX2 stainings) in fetuses from families 1 and 3 and the presence of few remaining glomeruli in fetus 1 from family 1 (Supplementary Material, Fig. S2A and B).

Altogether, these results clearly demonstrate that biallelic mutations in KIF14 lead to a syndrome characterized by severe developmental defects in both brain and kidney leading to arhinencephaly/atelencephaly and RHD.

Identified variations have a strong impact on KIF14 activity

Kinesins bind microtubules through their motor domain, which in many cases is negatively regulated through intramolecular interactions with the C-terminal cargo-binding domain (15). In the case of KIF14, this negative regulation of the motor domain is released upon interaction with CIT, which binds to the C-terminal region of KIF14 (amino acids 901–1233) through its N-terminal coiled-coil domain (CCf domain) (21). As the mutations identified affect the two key domains of KIF14, we predicted that they would have dramatic functional effects.

We first tested the consequences of the missense and truncation mutations on the steady state localization of KIF14. Depending on the activation state, KIF14 could either be found diffusely distributed in the cytoplasm when expressed in non-mitotic HeLa cells (17,18,21) or on microtubules upon activation (21). Accordingly, a GFP-fusion encoding wild-type (WT) KIF14 revealed a diffuse cytoplasmic localization (Fig. 2A), which was unaffected by the missense variations T456M and R364C (Fig. 2B and C). In contrast, the C-terminal truncations (Q1380^*, Q1304^* and R1189^*) all affected KIF14 localization and impacted upon cell morphology (Fig. 2D–F). The Q1380^* mutant strongly accumulated at the tips of transfected cells, which became highly elongated, and where it colocalized with acetylated α-tubulin (AcTub; Fig. 2D), a marker of stabilized microtubules. Both Q1304^* (Fig. 2E) and R1189^* (Fig. 2F) were present along stable microtubules, colocalizing with AcTub, and could occasionally be observed accumulated at the extremities of microtubules at the cell periphery. These observations suggest that the C-terminal truncations lead to constitutively active forms of KIF14, which are capable of binding microtubules. The procession of these active forms, particularly Q1380^*, towards the plus-ends of microtubules appears to cause a stabilization of microtubules, resulting in distension of the cell into protrusions, as was previously characterized for other kinesin family members (15).

As CIT is one of the main partners of KIF14 and regulates its activity (21), we next investigated the impact of the identified mutations on this interaction. We performed coimmunoprecipitation assays from lysates of HEK293 renal cells coexpressing WT or variant forms of KIF14 and the KIF14-binding domain of CIT (CCf; Fig. 2G). As expected, the CIT CCf domain was efficiently coimmunoprecipitated with WT KIF14 but not with the N-termi-nal PRC1-binding domain (1–356). This interaction was unaffected by the missense mutations in the motor domain (R364C and T456M) and by the two more C-terminal truncation mutations (Q1304^* and Q1380^*), which are localized after the previously mapped CIT binding site (amino acids 901–1233; Fig. 2H). Surprisingly, the most N-terminal truncation variation, R1189^*, also retained the interaction (Fig. 2G) despite being expected to result in the partial truncation of the CIT-binding domain (Fig. 2H). This result indicates that the CIT interaction domain of KIF14 in fact lies between amino acids 901 and 1189 (Fig. 2H).

Figure 3

Identified motor domain mutations affect KIF14 activity and impair microtubule binding. (A) Structure of mouse KIF14-MD is shown (PDB ID: 4OZQ) (16) with the mutated residues labelled (mouse residues in blue and the corresponding human residues in black). This image was produced in PyMol (The PyMOL Molecular Graphics System, Version 2.0; Schrödinger, LLC, New York, NY). (B) The microtubule-binding activity of human KIF14 motor construct (MBP-hKIF14) was assessed in an ultracentrifugation-based cosedimentation assay. A representative coomassie blue-stained gel, from four independent experiments, is shown. KIF14 motor constructs and tubulin are indicated. S, supernatant; P, pellet; MT, microtubules. (C) Basal and MT-stimulated ATPase activities of a KIF14 motor construct (MBP-hKIF14) were shown. ^*P ≤ 0.05; ^***P ≤ 0.001; from n=4 independent experiments, t-test. (D–F) Immunofluorescence of RPE1 cells cotransfected with plasmids encoding GFP-fusion of WT (D) or mutant KIF14 [T456M (E) or R364C (F)], along with an mCherry-fusion of the KIF14 binding domain of CIT (CCf, red) stained for α-tubulin (grey). Merge colour pictures (top panels) and black and white pictures corresponding to GFP fusions (KIF14; bottom panels) are shown. White arrows stress colocalization of KIF14 and α-tubulin. Nuclei were stained using Hoechst (blue). Scale bars: 5 μm.

As the nonsense mutation in the motor domain (family 4) was very similar to that previously described (11) and those of the mouse model (28), their negative impact was not investigated here. In contrast, the functional consequence of the missense variations identified in families 2 and 3 remained to be demonstrated. Site-directed mutagenesis has identified amino acids crucial to the activity of the motor domain, usually relating to ATP binding and hydrolysis. Arg364 (Fig. 3A), which is replaced by Cys in family 2, is analogous to Arg14 in conventional kinesin, which, when mutated, was shown to result in a severe reduction of MT-gliding velocity (35). The p.Arg364Cys variation is therefore expected to result in LOF of the KIF14 motor domain. Based on the resolved structure of the murine Kif14 motor domain (16), the Thr456 residue (Thr491 in mouse), replaced by Met in family 3, is part of the ATP-binding pocket of KIF14 where it faces Arg364 (Fig. 3A). This variation was therefore also expected to severely impact the function of the motor domain, which was investigated using a MT sedimentation assay. As previously demonstrated for the murine Kif14 motor domain (16), the human KIF14 motor domain efficiently copelleted with microtubules (Fig. 3B). The introduction of the p.Thr456Met variation (T456M) severely impaired MT binding, and the mutant motor domain remained within the supernatant (Fig. 3B). In addition, measurement of the ATPase activity of the motor domains revealed that ATP hydrolysis remained unchanged after the addition of microtubules to the T456M mutant (Fig. 3C). Similar results were obtained with the murine motor domain, where the equivalent human variant was introduced and compared to the WT (T491M; Supplementary Material, Fig. S3).

A previous work established that the CIT interaction activates KIF14 and allows its binding to microtubules (21). In contrast to WT KIF14 (Fig. 3D, arrows), R364C and T456M variants did not colocalize with α-tubulin when coexpressed with the CCf domain of CIT (Fig. 3E and F). As the missense variants retain the ability to bind CIT (Fig. 2), these data show that both variations severely impair the ability of the motor domain to bind microtubules, thus rendering these variant kinesins non-functional.

In conclusion, all identified variants could thus be considered as strongly damaging mutations.

Loss of KIF14 result in defects in cytokinesis in vitro and in vivo

A previous in vitro work has focused on the impact of a transient loss of KIF14 in HeLa cells (17,18). In order to better characterize the effect KIF14-loss in stable conditions in kidney cells, we performed a knockout (KO) of the gene in murine inner medullary collecting duct cells (mIMCD3), a widely used model of kidney epithelial cells. We used CRISPR/Cas9 (nickase) with guides targeting exon 5 encoding part of the motor domain (Supplementary Material, Fig. S4A). Clones were isolated and sequenced to identify homozygous mutational events. Two clones (KO1 and KO2), both with partial deletions of the 3′ region of exon 5 and 5′ region of introns 5–6 (Supplementary Material,Fig. S4B), were selected for further analysis. Sequencing of RT-PCR products revealed that the exon/intron deletions resulted in exon skipping in both clones, which could lead to an in-frame deletion in KO1 (exon 5 + 6) and/or to a frameshift and early stop codon in exon 6 in both KO1 and KO2 (Supplementary Material,Fig. S4C–E). These events likely result in non-functional proteins since they generate short truncated forms and/or proteins lacking a stretch of 28 amino acids within the motor domain. Two clones in which no mutational events in Kif14 could be identified (WT1 and WT2) and which expressed similar levels of Kif14 to the parental cell line (Supplementary Material, Fig. S4F) were maintained as controls.

While Kif14 (green) was detected at the midbody in parental IMCD3 cells (not shown) and control WT clones, it was absent from the intercellular bridges (AcTub staining, red) of KO clones (Fig. 4A), which were elongated compared to those of controls (Fig. 4B). In addition, CIT localization at the midbody was affected in KO clones, showing either a diffuse distribution along bridges or being present as two rings instead of a unique one (Fig. 4C and D), as previously described (21). Consequently, the loss of Kif14 in KO clones led to an increase in the proportion of binucleated cells (Fig. 4E and F). Interestingly, similar defects were observed in fibroblasts obtained from the affected fetus 21 from family 4 with longer intercellular bridges (Fig. 4G and H) and increased proportion of binucleated cells compared to control (Fig. 4I and J).

Figure 4

Loss of Kif14 leads to cytokinesis defects in mIMCD3 cells and in cells from the affected fetuses. (A, B) Cycling Kif14 WT (WT1 and WT2) and KO (KO1 and KO2) mIMCD3 clones were stained for KIF14 (green), AcTub (red) and DNA (Hoechst, blue). Insets show higher magnification of intercellular bridges stained with AcTub. Length of intercellular bridges in parental IMCD3 cells (IMCD3 WT) and WT and KO clones were quantified using ImageJ (B), as described in materials and methods. (C, D) WT and KO clones were immunostained for CIT (green) and AcTub (red; C) to quantify the localization of CIT to the midbody expressed as the percentage of cells with a ‘normal’ single ring CIT staining (D). (E, F) WT and KO clones were stained for F-actin using phalloidin (red), and nuclei using Hoechst (blue) in order to identify (asterisks; E) and quantify the percentage of binucleated cells (F). (G–J) Fibroblasts derived from control and fetus 24 from family 4 were stained for either AcTub (green) and KIF14 (red; G) or AcTub (green) only (I), as well as for DNA (Hoechst, blue). Length of mitotic bridges (H) and percentage of binucleated (J) were quantified based on these stainings. ns = not significant, ^**P < 0.001, ^***P < 0.0001 from n = 3 independent experiments [except in H (n = 2)], Dunn's Multiple Comparison Test following the Kruskal–Wallis ANOVA (Analysis of Variance) test for length of intercellular bridges (B, H) and Fischer's exact test for CIT localization and binucleated cells (D, F and J). Scale bars: 5 μm (A, C, E and G), 10 μm (I).

Human kidney development is severely impaired in cases where KIF14 function is lost, with affected fetuses presenting with either bilateral renal agenesis or cystic RHD (Fig. 1; Supplementary Material, Fig. S1). To explore the role of this mitotic kinesin in development, we analysed the protein localization in tissue sections from normal human fetal kidneys. Sections were stained for both KIF14 (green) and PAX2 (red), a marker of epithelial and mesenchymal components of the developing kidney (3,4). Strikingly, the lumen of some PAX2-positive epithelial structures were filled with KIF14-positive dots, which, when observed at a higher magnification, appeared as rings (Fig. 5A), indicative of the presence of KIF14-positive midbodies in UB tips lumen. In agreement with this observation, CIT (green) showed a similar ring-shaped pattern in the lumen of PAX2-positive structures surrounded by SIX2-positive cells (cyan; Fig. 5B), a marker of the cap mesenchyme. Finally, we observed a colocalization of both KIF14 (green) and CIT (cyan) in ring-shaped structures within clearly identifiable PAX2-positive UB branch tips (red; Fig. 5C).

Figure 5

KIF14 is required for cytokinesis in developing human kidney and brain. (A–C) Sections of control human fetal kidneys (23 weeks) were stained for (A) KIF14 (green), PAX2 (red) and Hoechst (blue); (B) CIT (green), SIX2 (red), PAX2 (cyan) and Hoechst (blue); and (C) KIF14 (green), PAX2 (red), CIT (cyan) and Hoechst (blue). Insets show higher magnification of regions of interest, denoted by the white outline, and arrows point to ring-like structures corresponding to midbodies. Similar results were obtained in three other control fetuses. (D–G) Haematoxylin–eosin staining of kidney sections from fetus 23 from family 3 (D–F) and from fetus 21 from family 1 (G) showed the presence of binucleated cells (black boxes), which were enlarged in the corresponding insets. (H–O) Haematoxylin–eosin staining of the brain cortex (H–K) and cerebellum (L–O) from a control fetus (37 weeks) and fetus 22 from family 2, as indicated. Binucleated pyramidal neurons and Purkinje cells are highlighted by black arrows. Scale bars: 10 μm (A–C), 50 μm (D–O).

These results were suggestive of a luminal accumulation of midbodies or midbody remnants following mitosis. Interestingly, mitosis of the epithelial cells of the UB tips was shown to occur in the lumen, with the daughter cell bodies subsequently reinserting into the epithelium (36). In this context, it was expected that the cytokinesis failure induced upon loss of KIF14 would generate binucleated cells, which should then be found in the lumen or epithelium layer. Careful analysis of kidney sections from affected fetuses revealed the presence of binucleated cells protruding from the surface of the epithelium along the lumen of large cysts (Fig. 5D and E) or inside the lumen of smaller cystic structures (Fig. 5F and G). As expected, numerous binucleated cells were also found in the cortex of the older affected fetus [51.5% (102/198) binucleated pyramidal cells versus 0.5% (1/200) in control; Fig. 5H–K]. Binucleated Purkinje cells could be found in the cerebellum of the affected fetus (Fig. 5L–O), they were, however, rare compared to cells in the cortex, reflecting the less pronounced defect in cerebellar development in the affected fetus (Supplementary Material, Table S1).

In conclusion, KIF14 plays a crucial function in cytokinesis during kidney and brain development.

Loss of kif14 in zebrafish causes microcephaly and ciliopathy-related phenotypes

A previous in vivo study in mice demonstrated that the loss of Kif14 resulted in microcephaly and growth restriction but failed to replicate the kidney phenotype observed in humans (28). We therefore chose to use the zebrafish as an alternative in vivo model in order to explore a potential conserved kidney-specific role for KIF14.

A mutant line was acquired from the Zebrafish Mutation Project (Sanger Institute; sa24165) (37), in which a single nucleotide substitution resulted in a nonsense mutation and an early truncation after the motor domain (c.1870C>T [p.Gln624^*]; Fig. 6A), a mutation similar to those identified in affected fetuses. The presence of the mutation was confirmed by Sanger sequencing (Fig. 6B). Macroscopically visible phenotypes (see below) were observed in one-fourth of the embryos from crosses of Htz individuals and Sanger sequencing confirmed that the affected larvae were homozygous for the mutation, hereafter kif14^−/−. These phenotypes were 100% penetrant in kif14^−/− homozygous embryos, while Htz embryos were indistinguishable from WT siblings. kif14^−/− embryos were not viable and died at day 5 postfertilization. By 48 hpf, they demonstrated microcephaly, eye defects, body curvature, which could be dorsal, ventral or lateral and cardiac oedema (Fig. 6C). Similar phenotypes were also observed in kif14 morphants (Supplementary Material, Fig. S5).

Figure 6

In vivo zebrafish model reveals loss of kif14 causes severe developmental defects of the brain and kidneys. (A) Schematic representation of zebrafish kif14 showing the position of the nonsense mutation. The motor and coiled-coil domains are also shown. (B) DNA trace from Sanger sequencing of exon 11 of kif14 in WT, Htz and homozygous embryos showing the nucleotide and subsequent amino acid, change. (C–E) Representative images of the phenotypes observed in WT/Htz and kif14^−/− embryos at 24 hpf (top row) and 48 hpf (bottom row). Scale bars: 100 μm. (D) Higher magnifications of the head region to better observe morphology of the brain and eyes. (E) Top view of the brain of WT/Htz and kif14^−/− embryos. F, forebrain; M, midbrain; H, hindbrain. (F) Immunofluorescence of WT/Htz and kif14^−/− embryos at 24 hpf immunostained for PH3 (green) and AcTub (grey) shown in dorsal views. Scale bars: 200 μm. (G) Representative images of the eyes of WT/Htz and kif14^−/− embryos at 24, 48 and 72 hpf. Asterisks denote observed coloboma. Scale bars: 100 μm. (H) Quantification of eye size (diameter). ns = not significant, ^***P < 0.0001 from n = 3 independent experiments, t-test. (I) Live imaging of Tg(wt1b:GFP) WT/Htz and kif14^−/− embryos at 48 hpf, dorsal view, with anterior to the left. Inset shows a higher magnification of proximal pronephros. Scale bars: 200 μm.

Data from zebrafish models of brain development have demonstrated that morphogenesis occurs first through the establishment of the forebrain, midbrain and hindbrain ventricles between 17 and 21 hpf, followed by expansion of the ventricles between 21 and 36 hpf (38). Microcephaly in mutant kif14^−/− embryos was evident around 22 hpf, with an absence of clear ventricular definition on both lateral and dorsal views (Fig. 6D and E). As shown in many microcephaly models (39), immunofluorescence staining for phospho-histone H3 (PH3), a marker of cells in G2/M, revealed a significant increase in the number of mitotic cells in kif14^−/− embryos, not only in the brain and along the spine, but also in peripheral tissues (Fig. 6F). Similarly, kif14^−/− mutant embryos also present microphthalmia (Fig. 6D, G and H), with a delay of optic fissure closure and occasional coloboma (Fig. 6G, blue arrows). In addition, the kif14^−/− embryos presented with cardiac oedema (Fig. 6C and D, lower row), often an indication of pronephric cysts (5,40). In order to examine the potential kidney defects, we used the Tg(wt1b:GFP) transgenic line expressing GFP under the control of the wt1b promoter (41). GFP expression in the proximal pronephros allowed us to confirm the presence of glomerular cysts in 73% of kif14^−/− mutant embryos at 48 hpf (n = 47, 4 separate clutches; Fig. 6I). Altogether, these data show that the loss of kif14 in zebrafish recapitulates both the brain and kidney developmental phenotypes observed in the affected fetuses.

Loss of kif14 causes ciliopathy-related phenotypes possibly due to an accumulation of mitotic cells in ciliated tissues

Pronephric cysts and body curvature are among the phenotypes widely seen in zebrafish models of known ciliopathy genes (5,40). Otolith defects are also commonly observed in ciliopathy models. Otoliths are structures present within the lumen of the otic vesicle and composed of biominerals. Their formation at the tips of tether cilia of sensory neurons has been shown to rely, at least partially, on the presence of motile cilia at the luminal surface of the neuroepithelium (42).

We observed a variety of otolith defects at 72 hpf in kif14^−/− embryos (Fig. 7A and B). Most otoliths were smaller, with many incorrectly positioned within the vesicle (misplaced). Occasionally, otoliths were fused, or present as a single structure, whilst extranumerary (extra) otoliths were rare. To directly explore whether these defects were related to cilia dysfunction within the otic vesicle, we used a transgenic (Tg) line in which the ciliary membrane marker Arl13b is fused to GFP and ubiquitously expressed [Tg(act2b:Arl13b-GFP)] (43). Confocal imaging of WT larvae (24 hpf) showed that the Arl13b fusion stained cilia present all cilia on neuroepithelial cells lining the lumen of the otic vesicle. Analysis in mutant embryos revealed that the number of cilia per otic vesicle was severely decreased compared to WT siblings (Fig. 7C and D). Taken together with the other phenotypes, these observations indicated that the loss of kif14 was impacting ciliogenesis.

Figure 7

In vivo zebrafish model presents ciliopathy-like phenotypes likely due to an accumulation of cells blocked/delayed in mitosis. (A, B) Representative images (A) and quantification (B; n = 61, 3 separate clutches), of otolith phenotypes observed at 72 hpf in WT/Htz and kif14^−/− embryos. (C, D) Live confocal imaging in brightfield (top row) and for GFP (cilia, bottom row) of Tg(act2b:Arl13b-GFP) of a representative otic vesicle of WT/Htz and kif14^−/− embryos at 24 hpf (C). Quantification of the number of cilia (Arl13b-GFP) in the otic vesicle (D). (E, F) Immunofluorescence of PH3 (white), and Hoechst (blue) stained Tg(cldnB:GFP) otic vesicles at 24 hpf (E) and quantification of PH3+ nuclei in the otic vesicle (F). ^*P < 0.01, ^***P < 0.0001 from n = 2 independent experiments, t-test (D, F). Scale bars: 10 μm.

To elucidate whether KIF14 plays a direct role in ciliogenesis, we initially tested two widely used in vitro models, IMCD3 and RPE1 cells, in which (PC) formation can be easily monitored and are therefore the two most widely used models to study ciliogenesis. Loss of KIF14 by either transient siRNA-mediated knockdown in RPE1 (Supplementary Material, Fig. S6A–D) or stable KO in IMCD3 (Supplementary Material, Fig. S6E–G) did not have a significant effect on the percentage of ciliated cells and cilia length compared to controls. We subsequently investigated the impact of the loss of kif14 on cilia in the cloaca region of zebrafish embryos, where cilia are affected in many ciliopathy models (44,45). Ciliogenesis in this distal part of the pronephros was not affected in kif14^−/− embryos (Supplementary Material,Fig. S6H–J) supporting the conclusion that KIF14 does not play a general role in ciliogenesis.

Cilia are present in quiescent cells and resorbed when cells enter in mitosis; mitotic cells are therefore not ciliated (Supplementary Material, Fig. S7A) (46). Interestingly, analysis of kif14^−/− zebrafish revealed a drastic increase in the number of mitotic PH3+ cells (Fig. 6F). In addition, AcTub stainings revealed the presence of numerous cells with mitotic spindle and condensed chromosomes around the distal end of the pronephros (Supplementary Material, Fig. S6H, arrows). These results then suggested that the ciliogenesis defects observed in the otic vesicle of kif14^−/− embryos could be linked to the accumulation of mitotic, non-ciliated cells. To test this hypothesis, we looked for the presence of mitotic cells (PH3 staining) in the Tg(cldnB:lynGFP) line in which neuroepithelial cells lining the otic vesicle express GFP (47). Confocal imaging of 24 hpf embryos revealed a drastic increase in the number of mitotic cells in the otic vesicle in kif14^−/− mutant embryos (Fig. 7E and F), mirroring ciliogenesis defects in this organ (Fig. 7C and D) and correlating with the accumulation of mitotic non-ciliated cells (Supplementary Material Fig. 7B). We then similarly analysed mitosis in the pronephros. As expected (48), very few PH3+ cells could be observed in proximal (wt1b:GFP; Supplementary Material, Fig. S8A and B) and distal (cldnB:GFP; Supplementary Material, Fig. S8C and D) tubules at 48 hpf, in WT/Htz embryos. However, the number of mitotic cells was drastically increased in kif14^−/− embryos in both regions.

Taken together, these data show that the loss of kif14 in zebrafish causes ciliopathy-like phenotypes, including in the otic vesicle and pronephros, which are likely linked to an accumulation of mitotic, and therefore non-ciliated cells in those tissues. In addition, this accumulation of mitotic cells indicates that kif14 may play an additional role during the early stages of mitosis in zebrafish.

Discussion

The present study, in accordance with our previous one, describes a novel syndrome linked to mutations in KIF14 and associated with severe developmental defects of both the brain (microcephaly/atelencephaly) and kidney (RHD). Here, we provide functional evidence that mutations in the syndromic form cause an LOF of KIF14. In addition, we used the zebrafish model to provide strong evidence that kif14 plays a crucial and conserved role during both brain and kidney development.

Two other groups recently identified mutations in KIF14 in 8 families with individuals presenting mild to severe microcephaly without major renal malformation (except 1 case out of 18; Supplementary Material, Fig. S9). Comparing the effect of these mutations provides some clues to explain the broad phenotypic spectrum associated with KIF14 mutations. Among the syndromic forms, two families harbour mutations leading to an early stop codon within the motor domain. These mutations are likely to correspond to a KO of KIF14 as even if alternative splicing were induced, any exon skipping in this region would lead to disruption of the motor domain. In addition, our functional studies clearly show that the mutations identified in the three other syndromic cases are also LOF mutations leading either to variant proteins unable to bind microtubules (motor domain), or to active non-functional forms, which lack the C-terminal domain, encoding at least a regulatory domain (this study) and the Radil binding site (23). Interestingly, our results also show that a truncation mutation of kif14 in the zebrafish, similar to those identified in humans, mimics the human syndromic form, further indicating that these truncation mutations result in non-functional proteins.

On the contrary, analysis of the mutations described in the eight families with isolated microcephaly (29,30) indicates that they are likely to be less severe. Three missense mutations and a single amino acid deletion are located in the forkhead-associated (FHA) domain, a domain that has regulatory functions in some kinesin family members (15) but whose function has not yet been characterized in KIF14. One missense mutation leads to a deletion of a smaller part of the C-terminal tail and missense or silent mutations affect splice sites, leading to the coexpression of C-terminally truncated proteins and missense variant or WT proteins, respectively. Interestingly, one mutation was also found in the motor domain (p.Gly459Arg); however, unlike both the Arg364 and Thr456 residues, the Gly459 residue is not present within the nucleotide binding pocket. It is likely, therefore, that the functional consequences of the p.Gly459Arg mutation would be far less severe than the p.Arg364Cys and p.Thr456Met mutations described here. Finally, the c.246delT/p.Asn83Ilefs^*3 homozygous mutation described in a family with individuals presenting with microcephaly without kidney defects would be expected to be even more severe than the ones in our syndromic families. Interestingly, it is in the same region as the c.263T>A/p.Leu88^* mutation that was shown to activate a cryptic splice site and to lead to an in-frame deletion within exon 2, resulting in a deletion of 124 amino acids (p.G58_L181del) within the PRC1-binding domain. The potential consequences on splicing of the c.246del/p.Asn83Ilefs^* mutation remain to be tested in order to conclude on the lack of genotype/phenotype correlation for this specific case.

Therefore, we can conclude that LOF mutations in KIF14 lead to syndromic forms associating microcephaly and brain malformations with RHD, whereas hypomorphic mutations cause mild to severe microcephaly with no developmental kidney defects. The observation of small echogenic kidneys in 1 affected individual (out of 18), however, indicates the potential for kidney manifestations in individuals harbouring hypomorphic KIF14 mutations.

Altogether, these results suggest that KIF14 likely has an important and similar function during the early steps of both kidney and brain development. There are very few reports of monogenic mutations identified in syndromes associating MCD or microcephaly with kidney phenotypes (Galloway-Mowat (49) and DREAM-PL (Dysmorphic facies, Renal agenesis, Ambiguous genitalia, Microcephaly, Polydactyly and Lissencephaly; 34) syndromes). However, the kidney phenotypes in these syndromes, including glomerulopathy or occasional unilateral agenesis, are much less severe than those observed in the case of KIF14 mutations. It is worth noting that associations of renal agenesis/RHD and MCD/microcephaly, reminiscent of those observed in individuals with KIF14 mutations, were described in cases with copy number variations of varying sizes and at various loci, suggesting the involvement of one or several genes, which were not characterized further (50). Finally, while mutations in genes involved in microtubule dynamics, mitosis and/or orientation of the mitotic spindle represent a major cause of microcephaly (14), it is not the case for renal agenesis and RHD with none of the mutations identified thus far affecting genes from these functional families [for a recent list of CAKUT genes, see (12,50)].

Ciliary defects were a plausible pathophysiological explanation for the two distinct phenotypes in fetuses harbouring LOF KIF14 mutations, as suggested earlier (11). Central nervous system manifestations are also observed in patients presenting with renal ciliopathies, such as cerebellar hypoplasia or agenesis of the corpus callosum, two defects that are present in all the affected fetuses. A plethora of brain developmental disorders is recognized as clinical signs in many ciliopathy phenotypes (2), but the direct role of mutations in ciliary genes, particularly for microcephaly, has not been frequently explored. Interestingly, dominant mutations in KIF2A, a kinesin involved in both mitotic spindle dynamics and ciliogenesis, were shown to cause MCD through their impact on both cilia and the cell cycle (51). In addition, bilateral agenesis and non-cystic RHD are very rarely reported among the renal phenotypic spectrum of ciliopathies (52). However, a recent report identified a homozygous LOF mutation in IFT27, encoding a component of the intraflagellar transport complex, causes bilateral renal agenesis (53), suggesting that defects in ciliary genes can lead to RHD.

We investigated a potential role for KIF14 in ciliogenesis. We found that, upon overexpression, KIF14 localizes to cilia (not shown) but failed to find any evidence for a direct role in ciliogenesis in vitro. Intriguingly, the zebrafish embryos harbouring LOF mutations in kif14 demonstrate ciliopathy phenotypes, but, once again, our analyses failed to identify a direct impact on ciliogenesis in vivo. However, we cannot formally exclude its potential role in the transport of cargo into the cilium, which remains to be explored. Based on these results, it is therefore unlikely that a disruption of ciliary function can account for a shared mechanism at the origin of renal and brain developmental defects.

The observation that KIF14-stained midbodies accumulate within the lumen of UB tips in human fetal kidneys provides a key clue to better understand the mechanism by which the loss of KIF14 affects both brain and kidney development in humans. It has previously been shown that midbody remnants are released or ‘secreted’ into the cerebrospinal fluid in mice. This accumulation, during the early stages of brain development, corresponds to the amplification of neural progenitors. During symmetrical division of these cells, cytokinesis occurs at the apical membrane and midbodies are formed within the lumen, where they are released following abscission (54). In UB tips, mitosis of proliferating UB epithelial cells occurs in the lumen of the expanding/branching tubular structures by a more complex process (36). Dividing cells protrude at the surface of the epithelium, within the lumen, and while one of the daughter cells remains attached to the basal membrane and retracts after mitosis, the other one intercalates between adjacent cells to integrate into the epithelium layer. Subsequently, midbodies are likely formed in the lumen of UB tips, where they could be released and accumulate, as occurs in the brain. To our knowledge, there is no other organ where a similar process resulting in the massive release of midbody remnants has been reported. The fact that both organs are dramatically affected by KIF14 LOF mutations strongly indicates that KIF14 plays a key role in vivo during these ‘symmetric’ divisions, whereby abscission occurs on both sides of the midbody and leads to their release into the extracellular milieu.

Intriguingly, LOF mutations in CIT lead to mild to severe microcephaly in humans (31–34), which is only occasionally (one case) found associated with severe kidney defects (32). These observations suggest that KIF14 may play a specific function during kidney development. Recent transcriptomic and single-cell transcriptomic analyses revealed expression of mitosis-related genes during the early development of the human kidney (55,56), including both CIT and KIF14, as well as other genes known to play crucial roles at the midbody (KIF23, KIF20A, ANLN etc.). These data suggest that other proteins required for cytokinesis could also have important functions during kidney development and may be able to compensate one another, except in the case of KIF14. Compensation mechanisms likely explain why the loss of Kif14 in the mouse does not lead to kidney defects (28). Of note, a similar lack of kidney phenotypes was previously observed in mouse models for a number of human kidney diseases (8,57,58).

The key question for future investigations is do ‘secreted’ midbodies have a function or not. Recent work has changed the view of midbodies and midbody remnants from passive structures eliminated after mitosis. Several groups identified possible postmitotic functions of midbodies when inherited or captured by daughter or neighbouring cells, including cell fate (stemness), polarity or ciliogenesis (22,59). In the brain, midbody remnants were purified from the ventricular fluid, indicating that most of them are not immediately recaptured by one of the daughter cells. Their function within the developing brain remains, however, to be determined. The images that we obtained clearly show the accumulation of midbodies in the lumen of UB tips. We cannot exclude that they may remain associated with the apical membrane of the UB tip cells where they could be involved in local signalling events and/or ciliogenesis. However, their localization in the central part of the lumen in most UB tip profiles suggests that they are released within the lumen, as in the developing brain. In this case, it is tempting to speculate that the connection of these midbody-filled UB tips to a newly formed and functional nephron would result in the expulsion of the midbodies by the incoming filtrate, leading, either, to signalling in the collecting duct or to elimination in the urine. All these exciting issues remain open and a source of future investigations.

Materials and Methods

Ethics statement

This study was conducted with the approval of the Comité de Protection des Personnes pour la Recherche Biomédicale Ile de France II. Approval was obtained under numbers 2007-02-09/DC-2008-229 and 2009-164/DC-2011-1449. For each fetus, written informed consent was obtained from the parents. For studies using animal data, housing and handling of fish were performed in accordance with the guidelines established by the French Council on animal care ‘Guide for the Care and Use of Laboratory Animals’: EEC86/609 Council Directive—Decree 2001-131. The project was approved by the departmental director of Services Vétérinaires de la Préfecture de Police de Paris and by the ethical committee of the Paris Descartes University (approval number: A75-15-34).

Genetics analyses, WES and Sanger sequencing

Family 3. Fetuses with severe kidney development defects, possibly associated with extrarenal defects, were collected through Centre de référence des Maladies Rénales Héréditaires de l'enfant et de l'adulte. DNA was extracted from frozen lung biopsies of the two affected fetuses of family 3 and whole-exome sequencing (WES) was performed using the 50 Mb Agilent SureSelect All Exon V3 and a HiSeq2500 (Illumina, San Diego, CA) sequencer. Sequence data were aligned to the human genome reference sequence (hg19 build) using BWA-MEM aligner (illumina). Downstream processing was carried out with the Genome Analysis Toolkit (GATK), SAMtools and Picard (http://www.broadinstitute.org/gatk/guide/topic?name=best-practices). The average coverage was 50, with 93.5% of the targeted regions covered >15×. Variants were annotated using a software system developed by the Paris Descartes University Bioinformatics Platform, based on the Ensembl (GRCh37/hg19), dbSNP, EVS, 1000 genome, ExAC and GnomAD databases. Variants were then prioritized according to their damaging effect (nonsense, frameshift, acceptor/donor splice site mutations, missense variants predicted to be damaging and absence or low frequency in GnomAD and in our in-house database (>10 000 exomes). For missense variants, prediction of damaging effects was based on PolyPhen2, Sift, Grantham and CADD (Combined Annotation Dependent Depletion) scores. According to a suspected autosomal recessive inheritance model, only one gene, KIF14, was identified as carrying two mutations common to both fetuses. Sanger sequencing in fetuses and parents confirmed compound heterozygosity.

In order to look for additional mutations in KIF14, probes covering the 30 exons of the gene were designed and added to a previously described SureSelect panel of 330 genes used for molecular diagnosis of CAKUT cases. KIF14 mutations identified in fetuses from families 1 and 2 have been briefly described in (12).

Family 4. Patients were included via the AGORA (Aetiologic research into genetic and occupational/environmental risk factors for anomalies in children) biobank project. The study protocol was approved by the Regional Committee on Research Involving Human Subjects (CMO Arnhem/Nijmegen 2006/048). DNA from fibroblasts and (umbilical cord) blood was obtained from fetuses 1, 2 and 4 and from blood from both parents. Written informed consent was obtained from both parents for whole-genome single-nucleotide polymorphism (SNP)-array analysis and WES of DNA from fetus 4 and analysis of skin fibroblasts from fetus 1. Because parents were consanguineous and autosomal recessive inheritance was suspected based on the phenotype, SNP-array analysis was performed to investigate genes located inside regions of homozygosity. Genomic DNA was extracted from fibroblasts obtained from a skin biopsy from fetus 4 according to standard protocols. Analysis of regions of homozygosity and copy number profiling were performed using the CytoSNP-850K BeadChip SNP-array (Illumina) according to standard procedures. SNP-array results were visualized and data analysis was performed using Nexus software version 7 (BioDiscovery, Los Angeles, CA) and the reference human genome build Feb. 2009 GRCh37/hg19. Results were classified with Cartagenia BENCH software (Cartagenia, Leuven, Belgium).

WES was performed on DNA from fetus 4. Protein coding genes and flanking (splice-site consensus) sequences were captured and enriched using the SureSelectXT Human All Exon V5 capture library (Agilent, Santa Clara, CA) and sequenced in rapid run mode on the HiSeq 2500 Sequencing system (Illumina) at a mean target depth of 100×. The target is defined as all coding exons of UCSC (60) and Ensembl (61) ±20 bp intron flanks. At this depth ~95% of the target is covered at least 15×. Reads were aligned to the reference human genome build Feb. 2009 GRCh37/hg19 using BWA (BWA-MEM v0.7.5a) and variants were called using the GATK haplotype caller (v2.7-2). Results were imported into Cartagenia BENCH Laboratory NGS module (Cartagenia, Leuven, Belgium) for variant annotation, filtering and prioritization. Filtering was performed for variants in genes associated with neurodevelopmental abnormalities (for a full list of genes, see Supplementary Methods S1). Identified variants were validated and segregation analysis was performed using standard Sanger sequencing.

Plasmids

Plasmids encoding WT KIF14 GFP-fusion and the CCf domain of CIT (amino acids 429–835) fused to mCherry were kind gifts from F. Barr and S. Narumiya, respectively (18,21). Variants from the affected fetuses were introduced in the GFP-KIF14 construct using Pfu turbo site-directed mutagenesis as previously described (44). The GST-mKif14 construct (pGex6P1) was described previously (16). Mutant mKIF14-MD-T491M DNA was generated by gene synthesis (GenScript, Piscataway, NJ) and cloned into pGex6P1.

Antibodies

The primary antibodies used were mouse anti-AcTub (1:10 000; Sigma-Aldrich, Saint-Quentin Fallavier, France; T6793), rabbit anti-KIF14 (1:300; Bethyl Laboratories, Inc., Montgomery, TX; A300-233A and A300-912A and 1:1000; Abcam, Cambridge, UK; ab3746), mouse anti-α-tubulin (1:5000; Sigma-Aldrich, T5168), mouse anti-CRIK (1:500; BD Biosciences, 611376), rabbit anti-SIX2 (1:200; Proteintech, Manchester, UK; 11562-1-AP), goat anti-PAX2 (1:200; R&D System, Lille, France; AF3364), rabbit anti-PH3 (1:200; Cell Signaling Technology, Leiden, the Netherlands; #3377S), rabbit anti-GFP (Thermo Fisher Scientific, Waltham, MA; A-11122), rabbit anti-dsRed (Clontech Laboratories, Inc., Mountain View, CA; 632496), mouse anti-GAPDH (Sigma-Aldrich, MAB374) and phalloidin-TRITC (1:300; Sigma-Aldrich, P1951). Cells were incubated with secondary antibodies (donkey) conjugated to Alexa Fluor® 488, 555 or 647 (Molecular Probes, Thermo Fisher Scientific). Sheep horseradish peroxidase-coupled antibodies were from GE Healthcare (anti-mouse: NA931V; anti-rabbit: NA934V).

Cell culture

Fetal control- and fetus-derived fibroblasts, human embryonic kidney 293 (HEK293) and HeLa cells (ATCC) were cultured in Dulbecco's Modified Eagle Medium (DMEM; Gibco®, Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS; Invitrogen, Thermo Fisher Scientific), glutamine and penicillin/streptomycin. IMCD3 and hTERT-retinal pigment epithelial (RPE1) cells were cultured in DMEM F:12, (Gibco®, Thermo Fisher Scientific) also supplemented with 10% FBS, glutamine and penicillin/streptomycin. Hela, HEK293 and IMCD3 cells were transfected using Lipofectamine 2000 (Thermo Fisher Scientific, 11668-019) and RPE1 using FuGENE® (Promega, Madison, WI; E2311) and maintained in culture for 24 h.

Biochemistry

Transfected cells were lysed in 0.5% triton, 150 mm NaCl and 50 mm pH 7.5 Tris–HCl. For immunoprecipitations, cleared lysates were incubated with mouse isotypic control antibodies and G-protein beads (Sigma-Aldrich, P7700) for 2 h at 4°C. Precleared proteins (1 mg) were incubated with mouse monoclonal anti-GFP antibodies (Sigma-Aldrich, 11814460001) coupled to G-protein beads for 3 h at 4°C. Beads were washed three times with increasing amounts of NaCl (150, 300 and 600 nm in 50 mm Tris–HCl pH 7.5), resuspended in 2× sample buffer (Sigma-Aldrich, S3401) and boiled at 95°C for 5 min. For immunoblotting, lysates and immunoprecipitates were separated by polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto polyvinylidene fluoride transfer membranes (GE Healthcare). Immunoblotting was performed using the indicated primary antibodies and revealed using the ECL+ Detection Kit (GE Healthcare).

Microtubule cosedimentation assay

Recombinant WT and mutant MBP-hKIF14-MD and GST-mKIF14-MD proteins were expressed and purified in the same manner as previously described (16). Microtubule preparation and cosedimentation assays were performed as previously described (62), with some modifications. Briefly, taxol-stabilized microtubules were prepared by polymerizing tubulin at 50 μm in 1× BRB80 (80 mm Pipes pH 6.8, 1 mm EGTA and 1 mm MgCl₂), 1 mm GTP and 1 mm DTT in the presence of 10% DMSO at 37°C for 25–30 min, then diluted to 10 μm working stocks in 1× BRB80 containing 20 μm taxol. Binding reactions were performed by mixing 500 nm of the indicated KIF14 constructs with 2 μm taxol-stabilized microtubules in 1× BRB80 supplemented with 25 mm KCl, 1 mm DTT, 0.01% Tween-20 and 20 μm taxol. After 20 min of incubation, mixtures were spun at 240 000×g for 5 min at 25°C. The supernatant and pellet fractions were recovered and resuspended in Laemmli buffer. Samples were resolved by SDS-PAGE. Gels were stained with Coomassie blue R-250 dye, destained and scanned with a digital scanner.

ATPase activity assay

KIF14 ATPase activity was monitored by a malachite green-based phosphate detection method, as previously described (16). Briefly, reactions were carried out in BRB80-based buffer (80 mm PIPES pH 6.8, 1 mm MgCl₂, 1 mm EGTA, 20 μm taxol, 25 mm KCl, 0.25 mg/ml BSA, 1 mm DTT and 0.02% Tween), supplemented with 1 mm ATP, with 2 μm taxol-stabilized microtubules and 50 nm KIF14 protein constructs. Basal activity of the Kif14 constructs was determined using the same reaction condition without microtubules. Reactions were allowed to proceed for 10 min, quenched with perchloric acid and malachite green reagent. The signal was quantified by measuring the absorbance at 620 nm in a Genios Plus plate reader (Tecan, Männedorf, Switzerland).

CRISPR

Guide couples were designed using CRISPOR and MIT targeting exon 5 of Kif14 and cloned into Cas9 (nickase) plasmids expressing either GFP or mCherry. The Cas9 nickase encoding plasmid pSpCas9n(BB)-2A-GFP was a gift from Feng Zhang (Addgene, Watertown, MA; plasmid #48140) (63). It was modified to generate an mCherry version. Briefly, the cleavable peptide 2A (P2A) sequence was cloned upstream of mCherry (Cold Fusion Cloning kit, System Biosciences, Palo Alto, CA) and a BbsI/BpiI restriction site was mutated (Quick Change kit, Stratagene, La Jolla, CA) as it was interfering with sgRNA subcloning in the final desired mCherry vectors. The recipient vector pSpCas9n(BB)-2A-GFP (#48140) was digested with EcoRI to excise the GFP, which was replaced by the P2A-mCherry insert. The sequences of the guides used were as follows: #4-GFP, AGTGTCCACTCGCCAGCGTGAGG; and #17-mCherry, ATTGACAGGCCTTCAACATACGG. IMCD3 cells were cotransfected with the GFP and mCherry vectors, sorted by flow cytometry for green and red fluorescence and clones were isolated. Nuclear DNA was extracted using QuickExtract™ DNA Extraction Solution (tebu-bio, Le-Perray-en-Yvelines France; QE09050) and Sanger sequenced using the following primers: forward ^5′gcacatctcgtgagaac^3′ and reverse ^5′gaacaaagaactaagagccc^3′.

Immunofluorescence

Cells were cultured on cover slips and fixed in 4% paraformaldehyde. Primary antibody incubation was performed in Dulbecco's phosphate buffered saline (PBS; Sigma-Aldrich), 0.1% Triton X-100 (Sigma-Aldrich) and 1 mg/ml or 3% bovine serum albumin (BSA; Sigma-Aldrich). After 3 washes with the incubation buffer, cells were incubated with secondary antibodies (1:300) in PBS–BSA (1 mg/ml) and, after 3 washes in PBS, nuclei stained using Hoechst (#33342, Sigma-Aldrich). Cover slips were mounted onto glass slides using Mowiol® 4-88 (Sigma-Aldrich).

Frozen tissues from control and affected kidneys were sectioned (8 μm) using a Leica microtome. Melted sections were postfixed in acetone (10 min, room temperature). The slides were washed twice in PBS and then blocked for 45 min at 4°C in 10% normal donkey serum diluted in PBT (PBS, 0.1% Tween) and then incubated with the primary antibodies diluted in PBT (overnight, 4°C). After three washes in PBT, sections were incubated with secondary antibodies diluted in PBT and, after two washes in PBS, nuclei stained using Hoechst (#33342, Sigma-Aldrich) and finally mounted onto glass slides using Mowiol® 4-88 (Sigma-Aldrich) or Fluoromount-G (Cell Laboratory, Beckman Coulter, Brea, CA).

Stained cells and tissues were imaged using an epi-illumination microscope (DMR, Leica , Wetzlar, Germany) with a cooled charge-coupled device camera (Leica DFC3000G). Images were acquired with LAS (Leica V4.6) and processed with ImageJ and Photoshop CS2 (Adobe Systems Inc., San Jose, CA). Confocal imaging was performed to acquire stainings of fetal fibroblasts using a Zeiss Confocal laser microscope LSM700, and images were processed with ZEN 2011 software.

Zebrafish experiments

The zebrafish kif14 mutant line sa24165 was acquired from the Zebrafish Mutation Project (Wellcome Sanger Institute, Hinxton, UK) and maintained at 28.5°C under standard conditions and according to European law. Kif14^+/− zebrafish were crossed with Tg(βactin:arl13bGFP), Tg(wt1b:GFP) and Tg(cldnB:lynGFP) to allow analysis of cilia (43), proximal (41) and distal pronephros (47), respectively. Genotyping was performed by placing embryos in 10 mm Tris pH 8.0, 1 mm EDTA and 1.2 mg/ml at 55°C overnight, and the reaction was stopped by incubating at 95°C for 5 min. DNA from exon 11 was Sanger sequenced using the following primers: forward ^5′gtgtgagattcgagtgttttc^3′ and reverse ^5′gttgcatatttaaacggaatg^3′.

Live embryos were analysed using a Leica M165FC stereoscope. For immunofluorescence, embryos were fixed in 4% PFA, washed in PBS and blocked in PBS, 0.3% Triton and 4% BSA. Both primary and secondary antibody incubations were performed overnight at 4°C. Fixed and stained embryos were stored in Mowiol and mounted in 1% low gelling temperature agarose (Sigma-Aldrich, A9414) and imaged using a Zeiss Axio Observer Z1 inverted microscope equipped with a Yokogawa CSU-X1 spinning disk. Images were acquired with a 40× water immersion objective (1.46) through a Hamamatsu Orca Flash 4.0 sCMOS camera. Images were processed using ImageJ and Photoshop CS2 (Adobe Systems Inc.).

Acknowledgements

We are grateful to the families for their participation. We greatly acknowledge N. Goudin and M. Garfa-Traoré (Necker cell imaging facility) for providing expert knowledge on confocal microscopy, members of the bioinformatic and genomic facilities of the Imagine Institute and C. Arrondelle and F. Legendre for their help on tissue sections and immunohistochemistry. We acknowledge E.D. Peters for immunofluorescence experiments. We would like to thank F. Barr, U. Gruneberg, S. Angers and S. Narumiya for their generous gifts of KIF14 and citron kinase encoding plasmids. The Imagine Institute is supported by a grant from Agence Nationale de la Recherche (ANR) (ANR-A0-IAHU-01). We acknowledge the Imagine Institute for the purchase of Leica SP8 STED and Zeiss Spinning Disk microscopes and the Fondation ARC (EML20110602384) for the purchase of the LEICA SP8 confocal microscope.

Conflict of Interest statement. None declared.

Funding

Fondation pour la Recherche Médicale (DEQ20130326532 to S.S.); European Union's Seventh Framework Programme FP7 (305608 to C.J.); GIS-Institut des Maladies Rares (AMA11025KSA to C.J. and S.S.); Canadian Institutes of Health Research (MOP-97928 to B.K.); Natural Sciences and Engineering Research Council of Canada (B.K.); Canadian Cancer Society Research Institute (703405 to B.K.); Fonds de Recherche du Québec-Santé (FRQS) (B.K.); Swiss National Science Foundation (320030-160200 to I.F.); Dutch Kidney Foundation (KOUNCIL).