The siRNA‐transfected HUVECs were plated inside a transwell chamber and incubated with serum for 18 h. The cells that invaded were observed after staining with crystal violet (CV) solution. Scale bars, 600 μm. The numbers of cells that invaded in each field of view were counted with the ImageJ software in (D). The data indicate the results of three independent experiments with ≥ 3 invasions per condition (mean ± SD). ***P < 0.001 (one‐way ANOVA with Tukey's post hoc test).

Tubulogenesis of control and CEP41‐knockdown cells for 18 h was compared via an in vitro angiogenesis assay. Scale bars, 600 μm. Quantification of tube node numbers in (F) and tube length in (G) from each field of view using the ImageJ angiogenesis analyzer at the indicated time points. The graph compares the relative length of control cells and CEP41‐knockdown cells. Data are shown as mean ± SD of five independent experiments with ≥ 5 tubulogenesis regions per condition. Statistical significance was assessed with the two‐way ANOVA followed by Tukey's post hoc test (**P < 0.01, ***P < 0.001).

Quantification of ISV lumen diameter in (B), the numbers of defective ISVs in (C), the numbers of embryos with aberrant DLAVs in (D), and the numbers of ruptured DLAVs in (E) from data observed in equivalent fields of view (within eight somites). The severity of blood vessel defects in cep41‐deficient zebrafish: B (narrowed ISVs) < C (shorten, fused, and missing ISVs) < D and E (ruptured DLAVs). Data are shown as mean ± SD of three independent experiments with ≥ 20 embryos per condition. Statistical significance was determined using the one‐way ANOVA followed by Dunnett's post hoc test (B) and the Kruskal–Wallis test by Dunn's post hoc test (C, E) (**P < 0.01, ***P < 0.001).

CV protrusions (arrowheads) and vascular loops (asterisks) were observed in zebrafish at 35 hpf. The images within the dotted rectangles are magnified in the right panels. A, anterior; P, posterior; CA, caudal artery; CV, caudal vein. Scale bars, 40 μm. Quantification of the length of the CV protrusions in (G) presented graphically by measuring them in equivalent fields of view (within five somites). Data are median of four independent experiments (n ≥ 5 embryos per experimental condition). Statistical significance was determined using the Kruskal–Wallis test followed by Dunn's post hoc test (***P < 0.001, ns: non‐significant).

Wild‐type and cep41‐deficient Tg(kdrl:eGFP) zebrafish embryos were fixed for immunostaining with Ac‐Tub‐ or GT335‐specific antibodies at 28 hpf in (C). The rectangles indicate the arterial cilia (AC) and venous cilia (VC) in zebrafish endothelial cells (EC). Magnified representative images are displayed in the right panels. Scale bars, 40 μm. Quantification of the labeled cilia observed in equivalent fields of view is presented graphically in (D). Data are shown as median of three independent experiments (n ≥ 20 embryos per condition). Statistical significance was determined using the Kruskal–Wallis test followed by Dunn's post hoc test (***P < 0.001, ns: non‐significant).

The indicated siRNA‐transfected HUVECs were subjected to in vitro angiogenesis assays. Tubulogenesis of HUVECs depleted for CEP41 and CCP5 was compared to that of control, CEP41 single‐knockdown, or CCP5 single‐knockdown cells (C). Scale bars, 600 μm. Quantification of tube node number in (D) and tube length in (E) from data examined within equivalent fields of view at the indicated time points using the ImageJ angiogenesis analyzer. Data are shown as mean ± SD of five independent experiments with ≥ 5 tubulogenesis regions per condition. Statistical significance was determined using the one‐way ANOVA followed by Tukey's post hoc test (***P < 0.001, ns: non‐significant).

The blood vessels in the trunks of cep41‐mutant/morphant zebrafish were compared to those of zebrafish co‐injected with ccp5 MOs (2 ng) at 40 hpf in (F). Asterisks and arrowheads indicate impaired ISVs and DLAVs, respectively. A, anterior; P, posterior. Scale bars, 100 μm. Quantification of the number of defective ISVs and DLAVs from data observed in equivalent fields of view (within eight somites). Data are median of three independent experiments with ≥ 20 embryos per condition. Statistical significance was determined using the Kruskal–Wallis test followed by Dunn's post hoc test (***P < 0.001).

HUVECs co‐transfected with CEP41 and CCP5 siRNAs were immunostained with ARL13b‐ and GT335‐specific antibodies. The staining results were compared to those of control, single CEP41, or CCP5 siRNA‐transfected cells. Representative images indicated with rectangles appear in magnified images in the lower panels. Scale bars, 20 μm. Quantification of relative intensity of GT335/ARL13b signals in the cilia (I) and ciliary length (J) in images (H) is the result of three independent experiments with ≥ 200 cells per condition (median). **P < 0.01, ***P < 0.001, ns: non‐significant (Kruskal–Wallis test with Dunn's post hoc test).

An in vitro angiogenesis assay was performed in static or shear stress‐exposed HUVECs transfected with the indicated siRNAs for 18 h in (C). Scale bars, 600 μm. Quantification of tube node number in (D) and tube length in (E) from data examined within equivalent fields of view at each time point using the ImageJ angiogenesis analyzer. Data are shown as mean ± SD of five independent experiments with ≥ 5 tubulogenesis regions per condition. Statistical significance was determined using the two‐way ANOVA followed by Tukey's post hoc test (**P < 0.01, ***P < 0.001, ns: non‐significant).

Control or CEP41‐deficient HUVECs cultivated under laminar shear stress and immunostained with ARL13b‐specific antibodies to measure deciliation. Representative images are indicated by rectangles with numbers and shown in magnified images in the right panels. The resulting data were compared to those of cells under static conditions in (F). Scale bars, 20 μm. Quantification of ARL13b‐positive ciliary cells (G) and of relative intensity of GT335/ARL13b signals in the cilia (H) in images (F) is the results of three independent experiments with ≥ 200 cells per condition (mean ± SD (G), median (H)). *P < 0.05, ***P < 0.001, ns: non‐significant (two‐way ANOVA with Tukey's post hoc test (G) and Kruskal–Wallis test with Dunn's post hoc test (H)).

Control, CEP41‐, or CCP5‐depleted hTERT‐RPE1 cells cultured under serum starvation for 48 h and fixed at 18 h after serum retrieval for immunostaining with both ARL13b‐ and GT335‐specific antibodies in (I). Images of the results 18 h after serum addition. Representative cells are indicated by rectangles. Scale bars, 20 μm. Quantification of ARL13b‐positive ciliated cells (J) and relative intensity of GT335/ARL13b signals in the cilia (K) under either serum starvation or serum retrieval (18 h) conditions are the results of three independent experiments with ≥ 200 cells per condition (mean ± SD (J), median (K)). **P < 0.01, ***P < 0.001, ns: non‐significant (one‐way ANOVA with Tukey's post hoc test (J) and Kruskal–Wallis test with Dunn's post hoc test (K)).

Immunoblot assays for phospho‐AURKA and AURKA were performed in HUVECs (C) and hTERT‐RPE1 cells (D) transfected with the indicated siRNAs. Protein levels were normalized against β‐actin in the same blots.

The mRNA levels of zebrafish vegfa (E) and vegfr2 (F) were quantified by qRT–PCR in eGFP‐positive ECs of control‐ or cep41‐MO‐injected Tg(kdrl:eGFP) zebrafish at 18 hpf (low shear stress) and 26 hpf (high shear stress). The expression of vegfa and vegfr2 was quantified in neuronal cells from control MO‐injected zebrafish for comparisons with that of ECs. The expression of zebrafish β‐actin was quantified for the normalization of those qRT–PCR results. Data are shown as mean ± SD of three independent experiments. Statistical significance was determined using the one‐way ANOVA followed by Tukey's post hoc test (***P < 0.001, ns: non‐significant).

Tg(kdrl:eGFP) zebrafish were injected with control‐ or cep41‐MOs and subjected to immunostaining with phospho‐AURKA‐specific antibodies (red) and DAPI at 26 hpf. The insets indicate the representative area from each immunostaining experiment, and the red dots indicate EC AURKA activation. Scale bars, 40 μm. Quantification of the phospho‐AURKA‐positive ECs (H) in equivalent fields of view for each MO‐injected zebrafish in images in (G) is the result of three independent experiments with ≥ 10 embryos per condition. The top and bottom whiskers represent the maximum and minimum values, respectively. **P < 0.01 (unpaired Student's t‐test with Welch's correction).

CEP41‐depleted hTERT‐RPE1 cells were transfected with vectors encoding nothing (MOCK), AURKA‐T288D, or AURKA‐K162R and then immunostained with ARL13b‐specific antibodies after serum starvation or serum retrieval. Images were taken from the results of staining done 18 h after serum retrieval. Rectangles indicate the representative cells that appear in the magnified images in the right panels. Scale bars, 20 μm. A quantification of ARL13b‐labeled ciliated cells under the indicated conditions in (J). These are the results of three independent experiments with ≥ 200 cells per condition (mean ± SD). *P < 0.05, ***P < 0.01, ns: non‐significant (two‐way ANOVA with Tukey's post hoc test).

Control and CEP41‐deficient HUVECs were transfected with expression vectors encoding nothing (MOCK), AURKA, AURKA‐T288D, or AURKA‐K162R and subjected to an in vitro angiogenesis assay for 18 h under static (K, L) or shear stress (M, N) states. Quantification of tube node number in (K, M) and tube length in (L, N) from data examined within equivalent fields of view at each time point using the ImageJ angiogenesis analyzer. Data are shown as mean ± SD of five independent experiments with ≥ 5 tubulogenesis regions per condition. Statistical significance was determined using the two‐way ANOVA followed by Tukey's post hoc test (*P < 0.05, **P < 0.01, ***P < 0.001, ns: non‐significant).

VEGFA (B) and VEGFR2 (C) mRNA levels were measured by qRT–PCR using cDNA from normoxic and hypoxic controls or CEP41‐depleted cells. The expression of GAPDH was quantified for the normalization of those qRT–PCR results. Data are shown as mean ± SD of three independent experiments. Statistical significance was determined with the one‐way ANOVA followed by Tukey's post hoc test (***P < 0.001, ns: non‐significant).

Tg(kdrl:eGFP) zebrafish were injected with control‐ or cep41‐MOs and then incubated under hypoxia at 28 hpf (a stage of high shear stress) for 2 h. They were subjected to immunostaining with phospho‐AURKA‐specific antibodies (red) and DAPI at 30 hpf. The insets indicate the representative areas from each immunostaining, and red dots indicate activated AURKA within ECs. Scale bars, 40 μm. Quantification of phospho‐AURKA‐positive ECs (E) in equivalent fields of view for each MO‐injected zebrafish in (D) is the result of three independent experiments with ≥ 20 embryos per condition. The top and bottom whiskers represent the maximum and minimum values, respectively. *P < 0.05 (unpaired Student's t‐test).

Zebrafish vegfa (F) and vegfr2 (G) mRNA levels were quantified by qRT–PCR in eGFP‐positive ECs of control‐ or cep41‐MO‐injected Tg(kdrl:eGFP) zebrafish subjected to either normoxia or hypoxia at 30 hpf. The expression of zebrafish β‐actin was quantified for the normalization of these qRT–PCR results. Data are shown as mean ± SD of three independent experiments. Statistical significance was determined using the one‐way ANOVA followed by Tukey's post hoc test (*P < 0.05, ***P < 0.001, ns: non‐significant).

Control and CEP41‐deficient HUVECs, transfected with expression vectors encoding nothing (MOCK), AURKA, AURKA‐T288D, HIF1α, or HIF1α with CEP41, were cultivated under hypoxia and subjected to an in vitro angiogenesis assay for 18 h. Scale bars, 400 μm. Quantification of tube node number in (I) and tube length in (J) from data examined within equivalent fields of view at each time point using the ImageJ angiogenesis analyzer. Data are shown as mean ± SD of five independent experiments with ≥ 5 tubulogenesis regions per condition. Statistical significance was determined using the two‐way ANOVA followed by Tukey's post hoc test (*P < 0.05, ***P < 0.001, ns: non‐significant).

Co‐immunoprecipitation assays using HIF1α‐ and CEP41‐specific antibodies were performed in HEK 293T cells expressing GFP‐CEP41 under normoxia or hypoxia. An IP using IgG‐specific antibodies was performed as a negative control. NOR, normoxia; HYP, hypoxia.

The CEP41‐depleted HUVECs were transfected with expression vectors encoding nothing (MOCK), CEP41, or HIF1α and cultivated under hypoxia. Whole‐cell lysates were used for immunoblot assays for phospho‐AURKA and AURKA, and the resulting data were compared to those of normoxic and hypoxic control cells. Protein levels were normalized against β‐actin in the same blots. NOR, normoxia; HYP, hypoxia.

The CEP41‐deficient HUVECs were transfected with expression vectors encoding nothing (MOCK), CEP41, HIF1α, or HIF1α with CEP41 and exposed to hypoxia. VEGFA (E) and VEGFR2 (F) mRNA levels were analyzed by qRT–PCR using cDNA from transfected cells. Data are shown as mean ± SD of three independent experiments. Statistical significance was determined using the one‐way ANOVA followed by Tukey's post hoc test (***P < 0.001, ns: non‐significant). NOR, normoxia; HYP, hypoxia.

Control or CEP41‐depleted HUVECs were transfected with expression vectors encoding nothing (MOCK) or HIF1α and treated with MG132 or not, and then, the cells were cultivated under hypoxia. Whole‐cell lysates were used for immunoblot assays for HIF1α, phospho‐AURKA, and AURKA, and the protein levels were normalized against β‐actin in the same blots. NOR, normoxia; HYP, hypoxia.

Quantification of tube node number in (H) and tube length in (I) from in vitro angiogenesis assays with HIF1α #1 or #2 siRNAs‐transfected HUVECs overexpressing nothing (MOCK) or CEP41 under hypoxia. Data are shown as mean ± SD of five independent experiments with ≥ 5 tubulogenesis regions per condition. Statistical significance was determined with the two‐way ANOVA followed by Tukey's post hoc test (***P < 0.001, ns: non‐significant).

VEGFA (J) and VEGFR2 (K) mRNA levels were analyzed by qRT–PCR using cDNA from hypoxic HIF1α‐depleted HUVECs transfected with control or CEP41 expression vectors. The results were compared to those of control and normoxic HIF1α‐depleted cells. Data are shown as mean ± SD of three independent experiments. Statistical significance was determined using the one‐way ANOVA followed by Tukey's post hoc test (***P < 0.001, ns: non‐significant).

A schematic diagram showing the molecular mechanism by which CEP41 drives hypoxia‐induced angiogenesis.

Wild‐type and cep41‐deficient Tg(kdrl:eGFP) zebrafish embryos were fixed for immunostaining with ARL13b‐specific antibodies at 28 hpf in (E). The rectangles indicate the arterial cilia (AC) and venous cilia (VC) in zebrafish endothelial cells. Magnified representative images are displayed in the right panels. CA, caudal artery; CV, caudal vein. Scale bars, 40 μm. Quantification of the labeled cilia observed in equivalent fields of view is presented graphically in (F). Data are shown as mean ± SD of three independent experiments (n ≥ 20 embryos per condition). Statistical significance was determined using the one‐way ANOVA followed by Tukey's post hoc test (ns: non‐significant).

Quantification of the numbers of embryos with defective ISVs in (B), the numbers of embryos with aberrant DLAVs in (C), and the numbers of ruptured DLAVs in (D) from data observed in equivalent fields of view (within eight somites). Data are shown as median of three independent experiments with ≥ 20 embryos per condition. Statistical significance was determined using the Brown–Forsythe ANOVA followed by Dunnett's T3 post hoc test (***P < 0.001).

Quantification of the numbers of ciliated cells (B), ciliary length (C, D), and relative intensity of GT335/ARL13b signals in the cilia (E) in images (A) is the result of three independent experiments with ≥ 200 cells per condition (mean ± SD or median). ***P < 0.001, ns: non‐significant (one‐way ANOVA with Dunnett's post hoc test (B–E)).

Quantification of the numbers of ciliated cells (B), ciliary length (C, D), and relative intensity of GT335/ARL13b signals in the cilia (E) in images (A) is the result of three independent experiments with ≥ 200 cells per condition (mean ± SD or median). **P < 0.001, ***P < 0.001, ns: non‐significant (one‐way ANOVA with Dunnett's post hoc test (B, D) and Brown–Forsythe ANOVA with Dunnett's T3 post hoc test (C) and Kruskal–Wallis test with Dunn's post hoc test (E)).

Quantification of tube node number in (B) and tube length in (C) from data examined within equivalent fields of view at each time point using the ImageJ angiogenesis analyzer. Data are shown as mean ± SD of five independent experiments with ≥ 5 tubulogenesis regions per condition. Statistical significance was determined using the two‐way ANOVA followed by Tukey's post hoc test (***P < 0.001, ns: non‐significant).