Cerebral cavernous malformations (CCMs) are a prevalent class of vascular anomalies characterized by thin-walled clusters of malformed blood vessels in the brain. Heritable forms are caused by mutations in CCM1 , CCM2 and CCM3 , but despite the importance of these factors in vascular biology, an understanding of their molecular and cellular functions remains elusive. Here we describe the characterization of a zebrafish embryonic model of CCM. Loss of ccm1 in zebrafish embryos leads to severe and progressive dilation of major vessels, despite normal endothelial cell fate and number. Vascular dilation in ccm1 mutants is accompanied by progressive spreading of endothelial cells and thinning of vessel walls despite ultrastructurally normal cell–cell contacts. Zebrafish ccm2 mutants display comparable vascular defects. Finally, we show that ccm1 function is cell autonomous, suggesting that it is endothelial cellular morphogenesis that is regulated by CCM proteins during development and pathogenesis.
Cerebral cavernous malformation (CCM) is a prevalent disease characterized by enlarged thin-walled capillary clusters in the brain. Familial CCM is inherited as an autosomal dominant trait and in some populations can account for up to 50% of clinical presentation. Loss of function mutations in three genes, CCM1 , CCM2 or CCM3 , have been demonstrated to be responsible for familial CCM ( 1 ).
In vitro studies have shown that the three CCM proteins associate with each other in the same complex, which is associated with the cytoskeleton as well as with components of signal transduction pathways and the cell junctions ( 2–9 ). Although in vivo analysis has been limited, Ccm1 -deficient mice display vascular dilation phenotypes that strongly resemble the human disease. The dilations in Ccm1 -deficient mice are associated with the loss of arterial gene expression and increased endothelial mitosis, suggesting that changes in cell fate or proliferation may contribute to CCM phenotypes and pathogenesis ( 10 ). The zebrafish-dilated heart mutants santa and valentine have recently been shown to correspond to ccm1 and ccm2 , respectively ( 11 ), but a thorough analysis of the mutant vasculature remains to be reported and these mutants have yet to yield genuine mechanistic insights into CCM function.
Here we describe in detail the vascular phenotypes associated with the loss of Ccm1 in zebrafish. Zebrafish ccm1 mutants display conserved progressive dilations of embryonic vessels, resembling both the murine and human CCM phenotypes. In contrast to previous findings in the mouse, these conserved vascular dilation phenotypes can occur independent of changes in cell number or cell fate. Vascular dilations are associated with specific changes in cell shape although endothelial cell–cell contacts appear normal. Furthermore, ccm1, ccm2 and ccm1/2 double mutants displayed indistinguishable vascular phenotypes, implicating a conserved functional relationship. Finally, we demonstrate that the function of ccm1 is autonomous to endothelial cells. This study identifies zebrafish ccm1 as an endothelial specific regulator of cellular morphogenesis, implicating specific endothelial cellular morphogenesis defects in the dilation of CCM deficient vessels during development and pathogenesis.
The vascular morphogenesis mutant t26458 encodes ccm1/krit1
In a large-scale forward genetic screen ( 12 ), we identified a mutant which displayed normal vascular patterning but dramatic and completely penetrant vascular morphogenesis defects. This mutant failed to initiate circulation and starting from 32 hours post fertilization (h), dilation was observed in the primordial midbrain channel (PMBC) (Fig. 1 A and B) and the heart, with the caudal vein (CV) also showing severe morphogenesis defects and apparent dilation (not shown). At 50 h, severe prominent dilation of the sub-intestinal vessels (SIVs), dilation of the posterior cardinal vein (PCV) in the region dorsal to the sub intestinal vessels, as well as aberrant morphogenesis and dilation of the CV were observed in mutants (Fig. 1 C–H). Severe heart and vascular dilation was also observable using standard histological approaches ( Supplementary Material, Fig. S1 ). Genetic mapping localized the mutation to a region containing 9 genes ( Supplementary Material, Fig. S1 ). One gene in this region, krit1/ccm1 , had been previously associated with vascular dilation phenotypes ( 1 , 10 ). Sequencing of ccm1 identified a cytosine to adenosine transversion mutating a tyrosine into a premature stop codon at a position corresponding to amino acid 237 of the protein (Fig. 1 I). In contrast to two previously reported alleles, which led to a small in frame deletion and a C-terminal premature stop codon retaining most functional domains ( 11 ), this mutant protein would lack almost all known functional domains (Fig. 1 I). The injection of a morpholino (MO) targeted to the ccm1 start codon recapitulated the phenotypes observed, whereas the injection of a silent heart targeting MO ( 13 ), which inhibited circulation, did not lead to vascular dilations, demonstrating that the vascular phenotypes are caused by the loss of ccm1 and not loss of circulation ( Supplementary Material, Fig. S1 ).
ccm2 mutants display identical phenotypes to ccm1 mutants
We next examined a previously identified retroviral insertional mutant for ccm2 with the insertion located 162 base pairs into the first intron of the ccm2 gene ( 14 ). By analysis in the fli1:GFP background, we found that the vascular and heart dilation phenotype was variable in this line (Fig. 2 A, data not shown). The phenotypic variability in heart dilation associated with this allele was different from previously described alleles ( ccm2m201 ) which did not show an intermediate phenotype, indicating that our allele is a hypomorphic allele of ccm2 ( 11 ). ccm2hi296 mutants displayed mild or severe vascular phenotypes, with severe vascular dilation and morphogenesis defects from 48 h identical to those observed in ccm1 mutants (Fig. 2 B–E). Although early defects of the CV (comparable to ccm1 mutant phenotypes) were apparent, dilation of the PMBC was variable, apparently reduced and not clearly quantifiable in these hypomorphic mutants (data not shown). In addition, we produced double mutants for ccm1 and ccm2hi296 and examined their vasculature. We found that the dilation of embryonic vessels in double mutants was indistinguishable from ccm1 or severe ccm2hi296 mutants (data not shown). This phenotypic similarity and double mutant epistasis correlate with the biochemical analysis placing CCM proteins within the same complex and demonstrates non-redundant and essential functions for both ccm1 and ccm2 .
Arterial venous fates and cell number are normal in ccm1 mutants
Analysis of Ccm1 -deficient mice has suggested that CCM vascular dilations are associated with altered arterial fate and increased endothelial mitosis ( 10 ). Zebrafish santa mutants have been recently identified in a separate screen for vascular defects and a cursory analysis of the phenotype identified increased tie2 expression in mutant vessels leading to the suggestion that endothelial cell number is also increased ( 15 ). Incongruously, the number of endocardial cells, which are endothelial in origin, is reportedly unchanged in santa mutants ( 11 ).
To determine if altered cell fate or altered proliferation were responsible for CCM phenotypes, we carefully examined both in mutants. We found no difference between ccm1 mutant and wild-type embryos for the expression of notch3, dll4, hey2, ephrinB2, flt4 and dab2 at both 24 and 32 h (preceding and concurrent with the first phenotypes) (Fig. 3 A–D, Supplementary Material, Fig. S2 ). We also analysed large clutches of embryos produced from double heterozygotes for ccm1 and ccm2 for the expression of ephrinB2 and flt4 and observed no change in expression (at 32 h, n = 118/118 embryos displayed normal expression of ephrinB2 and n = 67/67 embryos displayed normal expression of flt4 ). Furthermore, we observed normal development of the dorsal aorta and intersomitic vessels, indicating that arterial fates were normal in ccm1, ccm2 and ccm1/2 double mutants. To investigate cell number in mutant vessels, we focused on the SIVs due to their prominent mutant phenotype and their relative isolation over the yolk, away from other embryonic tissues which might impede on accurate cell counting. We directly counted the number of endothelial cells in the SIV in mutant and sibling vessels using an anti-GFP antibody counter-stained for individual endothelial cell nuclei (DAPI) at 72 h. We observed no significant difference in the number of cells between mutant and sibling embryos (Fig. 3 E and F). We also saw no difference in cell number using the fli1:nucGFP transgenic line to count cells in the SIV [SIV at 58 h: wt = 33 ± 3; ccm1 = 35 ± 5 (mean±SD; P = 0.402 by Student’s t -test; n = 5)]. Determination of cell number in the PCV and CV is hampered by the transient presence of variable numbers of haemopoietic stem cells which seed and differentiate in both the PCV and CV ( 16 ) and express the marker gene fli . However, cell counts in equivalent regions of these vessels revealed no indication of an increase in the number of fli1: nucGFP nuclei in 2, 3 and 4 dpf mutants compared with wild-type (2 dpf PCV: wt = 52 ± 9.5; ccm1 = 43.8 ± 7.5, 2 dpf CV: wt = 95 ± 10.2; ccm1 = 83 ± 12, 3 dpf PCV: wt = 67.8 ± 4.1; ccm1 = 56.8 ± 4, 3 dpf CV: wt = 87.8 ± 7; ccm1 = 85 ± 10.1, 4 dpf PCV: wt = 68.6 ± 4.6; ccm1 = 63.2 ± 3.4, n = 5 for all time-points and genotypes). Furthermore, we observed no change in PCNA ( in situ hybridization at 24 h), phosphorylated histone (immunohistochemistry at 24 and 32 h) or BrdU incorporation (24 and 32 h) in wholemount stainings (data not shown). Collectively, these data indicate that increased endothelial cell number and proliferation do not contribute to the vascular dilation phenotype observed.
Vascular dilation is associated with progressive dilation and cell spreading but normal cell–cell contact
To better understand the defects observed, we examined dilation and morphogenesis of the SIVs using live confocal timelapse imaging in the nuclear localized fli1:nucGFP transgenic line ( 17 ). We found that between 34 and 58 h, the dilation of mutant SIVs was associated with an apparent progressive and dynamic spreading of endothelial cells over the embryonic yolk (Fig. 4 A and B, Supplementary Material, Movies S1 and S2 ). This defect was not a secondary effect of loss of circulation, as silent heart MO injected embryos ( 13 ), with no observable circulation, develop constricted SIVs ( Supplementary Material, Fig. S3 ). We next compared vascular dilation in the trunk of ccm1 mutants with wild-type siblings at 32, 60 and 80 h. While the diameter of the PCV was not increased in mutants compared with siblings at 32 h, mutant PCV dilation was prominent at 60 h, and at 80 h mutant PCVs were vastly dilated with a dramatic thinning of vessel walls (Fig. 4 C–H). Importantly, these observations suggest that ccm1 mutant vascular dilations likely occur due to progressive thinning and spreading of endothelial cells during vascular morphogenesis. Intuitive explanations for such cellular morphogenesis defects could be altered cell junctions and cell–cell adhesion leading to spreading and dilation or, alternatively, the morphogenesis defects may be inherent to the mutant endothelial cells themselves and independent of their contacts with neighbouring cells. Using electron microscopy to examine endothelial cells within the PCV in ccm1 and ccm2 (severe phenotype) mutant vessels at 48 h, we observed no obvious ultrastructural remodelling of cell junctions despite marked thinning of the vessel wall in mutants (Fig. 4 I–M). These data suggest that cell-shape changes occur independent of any obvious changes in endothelial cell–cell contact.
Given that CCM in humans almost exclusively presents in the vasculature of the brain, we examined the morphology and ultrastructure of the brain vasculature at 72 h, a time-point prior to general necrosis and death but when later cranial phenotypes were apparent. We examined the secondary vessels that form via angiogenesis and permeate the brain as they most resemble those present throughout the mammalian CNS and we found that they were extremely abnormal and failed to lumenise in ccm1 mutants. These defects were most likely due to the failure to initiate circulation as silent heart MO injected embryos displayed apparently identical defects ( Supplementary Material, Fig. S3 ). This phenotype precludes a thorough analysis of lumenised mutant secondary vessels ( Supplementary Material, Fig. S4 ). However, an analysis of the ultrastructure of these highly abnormal ccm1 mutant brain vessels did identify endothelial junctions ( Supplementary Material, Fig. S4 ).
Quantitation of progressive venous dilations in ccm1 mutants
To quantitate endothelial cell spreading and dilation, we next analysed ccm1 mutants using the fli1:nucGFP transgenic line at 2, 3 and 4 dpf. We focused on the PCV due to its simple, predictable morphology in wild-type and mutant embryos. We used three-dimensional GFP and transmitted light z-stack images to measure vessel diameter as well as the distance between nearest neighbouring nuclei within each PCV. Quantitation of approximate vessel diameter (based on the dorsal–ventral extent of the lateral surface of the PCV) indicated that dilation progressed over time in mutants with no quantifiable difference observed at 48 h, but a significant increase of approximately 8.5 µm ( P = 0.00085) at 72 h, and of 26.7 µm ( P = 2 × 10 −4 ) at 4 dpf (Fig. 5 ). Quantitation of cell spreading was based on multiple measurements of the distance between nearest neighbouring endothelial nuclei located in the same plane of the vessel and at equivalent locations between somites immediately anterior to the cloaca. Measurements were taken between wild-type and mutant vessels and indicated that the spreading of endothelial cells was progressive over developmental time with no quantifiable difference observed in this region of the PCV at 48 h, but a significant increase of approximately 3 µm ( P = 1.19 × 10 −5 ) at 72 h, and of 9.3 µm ( P = 4.6 × 10 −14 ) at 4 dpf (Fig. 5 ). Interestingly, the ratio of vascular diameter to distance between nearest neighbours remained virtually unchanged throughout the analysis, providing an independent measurement indicating that cell-number changes are unlikely to contribute to dilation. These data indicate that progressive, quantifiable, cellular spreading and vascular dilation underpin the ccm1 phenotype.
ccm1 is required cell autonomously for the regulation of endothelial cell shape
All three Ccm genes are expressed broadly during development, making it unclear whether CCMs function in endothelial or surrounding tissues ( 18–20 ) ( Supplementary Material, Fig. S5 ). In order to formally determine whether ccm1 function is cell autonomous to the endothelium, we aimed to perform transplantation assays. We transplanted wild-type vegfR4:GFP or fli1:GFP cells into non-transgenic ccm1 mutant recipient embryos and found that wild-type cells were capable of forming a wild-type elongated morphology in multiple vessels in mutant host embryos (Fig. 6 ). To better understand morphology in mutant vessels, we focused on mutant cells in the region of the CV and found that they appeared distinctly and uniformly spread in appearance, whereas wild-type cells were elongated forming the CV plexus (Fig. 6 A and B, Supplementary Material, Movies S3 and S4 ). This cell spreading in the CV was specific as endothelial cells are elongated in silent heart MO injected CVs (Fig. 6 C and D). We counted the relative number of elongated and spread cells in the lateral wall of the CV in silent heart MO injected and ccm1 mutant embryos and found that mutant vessels never displayed elongated cells at 72 h ( Supplementary Material, Fig. S6 ). We also injected DNA encoding vegfR4 promoter driven tandem Tomato into ccm1 mutant and sibling embryos and counted the number of isolated elongated cells versus isolated spread cells in the lateral wall of the CV in mosaic animals (Fig. 6 E–G). We found no elongated cells in mutant embryos although we did observe spread cells in wild-type CVs ( Supplementary Material, Fig. S6 ). Importantly, isolated transplanted wild-type endothelial cells in an otherwise mutant CV were capable of marked elongation and could even constrict dilated mutant vessels (Fig. 6 J, Supplementary Material, Fig. S6G ). Reciprocal transplantations of mutant cells into wild-type embryos would not be interpretable in this assay given the occasional presence of spread cell morphologies in wild-type CVs. This ability of wild-type cells in a mutant environment to independently elongate and to influence the mutant phenotype demonstrates that ccm1 functions autonomously within endothelial cells.
Understanding CCM phenotypes using embryonic zebrafish vessels
Human mutations in CCM1, CCM2 and CCM3 cause cavernous malformations, but despite the importance of these genes in the aetiology of this disease, very little is known about their molecular and cellular functions. A genetically tractable in vivo model allowing genetic, molecular and cellular dissection of the effects of loss of CCMs will greatly aid our understanding in this field. The zebrafish vascular phenotypes associated with the loss of ccm1 and ccm2 described here are strikingly similar to those associated with the loss of the CCM genes in mice and humans. In particular, the hypomorphic allele of ccm2 will be of use for performing MO-based reverse genetic or pharmacological modifier screens which can be used to identify phenotypically relevant genetic or chemical interactors in the CCM pathway.
Here, we show that CCM phenotypes can occur in vivo , despite normal endothelial cell fate, normal endothelial proliferation and in the presence of ultrastructurally normal cell–cell contacts. The fact that zebrafish mutant embryos show no alterations in arterial-venous specification or endothelial proliferation, challenges findings in mice ( 10 ) and warrants further analysis of the murine phenotype. Given that circulation through forming vessels has been shown to be necessary for their normal expression of arterial marker genes in the chick ( 21 ), it seems plausible that the loss of arterial gene expression in Ccm1 knockout mice may be secondary to a loss of circulation. We also show here that the progressive vascular dilation phenotype is associated with specific thinning of vessel walls concurrent with the broadening and spreading of endothelial cells. This requirement for ccm1 in defining endothelial cell morphology is specific and intrinsic to endothelial cells themselves and not to surrounding or supporting cells because transplanted wild-type endothelial cells in an otherwise mutant vessel establish their wild-type elongated morphology and influence the structure of the surrounding mutant vessel.
Our findings complement recent in vitro findings, which demonstrate that the CCM complex associates with cytoskeletal elements, signal transduction components and junctions ( 2–9 ). The presence of the CCM complex on the cytoskeleton and at the cell junctions, combined with the in vivo analysis described here, point towards a working model to explain CCM function during development and pathogenesis. We suggest that CCMs, acting within endothelial cells, at the interface between the cytoskeleton and the cell junctions, primarily regulate endothelial cellular morphogenesis and thereby vascular tubular morphogenesis and that this occurs downstream of normal differentiation events. Further studies focused on the role of CCM family protein function at the sub-cellular level are now required to fully elucidate the mechanism by which they regulate the morphogenesis of endothelial cells and vascular tubes.
MATERIALS AND METHODS
All zebrafish strains were maintained in the Hubrecht Institute using standard husbandry conditions. Animal experiments were approved by the Animal Experimentation Committee of the Royal Netherlands Academy of Arts and Sciences (DEC). Transgenic lines used ( 17 , 22 , 23 ) were TG(fli1a:gfp)y1 , TG(fli1:nucGFP)y7 and TG(vegfR4: gfp)s843 (also known as flk1, kdra or kdr-like ) ( 24 ). ccm2 insertional mutants have been previously identified ( 14 ) and were a generous gift from Adam Amsterdam and Nancy Hopkins.
Genetic mapping and genotyping
Bioinformatics construction of the genomic region and synteny analysis was performed using the Ensembl database ( http://www.ensembl.org ), release 44, April 2007. Meiotic mapping of the ccm1 mutation was performed using standard simple sequence length polymorphisms. Subsequent genotyping of ccm1 mutants was performed on individual embryos using the PCR primers: ccm1exon5F ; 5′-CCACAAGCGTAACGTAAATG-3′ and ccm1exon5R ; 5′-ATCTATGGACGCAATGCAG-3′. DNA sequencing was performed with the ccm1exon5F primer.
The ccm1 start codon targeting MO ( ccm1 MO 5′-CTCCTCTAGCTCTTGGTTTCCCATC-3′) was injected at a concentration of 1.25 ng/embryo. The silent heart ( 13 ) MO (5′-CATGTTTGCTCTGATCTGACACGCA-3′) was injected at a concentration of 1 ng/embryo.
In situ hybridization and immunohistochemistry
In situ hybridization and immunohistochemistry using anti-GFP antibody (Torrey Pines Biolabs, http://www.chemokine.com/ ) were performed as previously described ( 24 ). Cell counts in immuno-stained embryos were normalized between wt and mutant embryos by focusing on the region encompassed from the first SIV endothelial cell immediately posterior to the pectoral fin tip to the region where the yolk extension meets the yolk ball. Previously described probes used for in situ hybridization were: notch3 ( 25 ), dll4 ( 26 , 27 ), hey2 ( 28 ), ephrinB2 ( 25 ), flt4 ( 29 ) and dab2 ( 30 ). ccm1 probe was synthesized by first cloning a PCR fragment amplified from cDNA (primer: Forward; 5′-GCGCGAATTCACCATGGGAAACCAAGAGCTAGAGGAGG-3′ and Reverse; 5′-GCGCCTCGAGTTACCCATACGCATATTTATCAGAC-3′) into the EcoR I and Xho I sites of the pCS2+ vector ( 31 ) followed by in vitro transcription of RNA with T7 RNA polymerase (Promega) from EcoR I digested plasmid DNA.
Vibratome sections were cut to a thickness of 250 µm using a HM650V vibratome (Microm). Filamentous actin was visualized with rhodamine phalloidin counter staining (Fluka) by incubation overnight at a 1/1000 dilution in phosphate buffered saline. Samples were mounted and imaged in Aquamount (BDH laboratory supplies).
Embryos were mounted in 0.5% agarose in a six-well culture plate with a cover slip replacing the bottom of each well. Imaging was performed with a Leica TCS SP confocal microscope (Leica Microsystems, http://www.leica-microsystems.com/ ) using a 10× or 40× objective with digital zoom. Scale is given as image width in the absence of distinct morphological features. The timelapse analysis was compiled using ImageJ software ( http://rsb.info.nih.gov/ij/ ). Time points were recorded every 20 min for 24 h. A heated stage maintained the embryos at approximately 28.5°C. Movies of 3D reconstructions were made using the Volocity software package (Improvision). Analysis of fli1:nucGFP vessels including cell counting, distance to nearest neighbour and dorsal–ventral extent calculation was aided by the Volocity software package (Improvision) for object recognition, 3D reconstruction and distance calculations. Diameter calculations were made based on combined GFP z-stack and light (DIC) images to best observe the venous lumen. Cell counts in the PCV and CV using the fli1:nucGFP strain were normalized by imaging equivalent regions of each vessel and counting cells in a region of each vessel spanning three somite widths. Embryos for electron microscopy were fixed in 2.5% glutaraldehyde in 0.2 M cacodylate buffer.
Transplantation and microinjections
Transplantation was performed essentially as previously described ( 32 ). Briefly, wild-type donor embryos of the genotype TG(fli1:GFP) or TG(vegfR4:GFP) were injected with 70 kDa Tetramethyl Rhodamine (TAMRA) (Molecular Probes) at the one cell stage and utilized as donors at pre-dome stages. Approximately 10–20 cells were transferred from donor to recipient embryos with recipient embryos utilized between sphere and 30% epiboly stages. Genotypes were inferred from phenotype at 48 h. Analysis of cell shape in mosaic mutant and wild-type vessels was performed using DNA injections of a construct consisting of approximately 6.5 kb of the previously described vegfR4 promoter ( 23 ) driving the expression of tandem Tomato ( 33 ) and flanked by I- Sce I meganuclease recognition sites.
Supplementary Material is available at HMG Online.
B.M.H. was funded in part by an EMBO long-term fellowship and in part by an NH&MRC (Australia) CJ Martin fellowship. J.B. was funded by a PhD stipend from Boehringer Ingelheim Fonds and a fellowship from the Stichting Vrienden van het Hubrecht Institute.
The t26458 allele was discovered thanks to the Tubingen 2000 Screen Consortium (see Supplementary material ). We would like to thank Nancy Hopkins and Adam Amsterdam for supplying the ccm2 mutant line, Gabi Frommer-Kästle for expert assistance with electron microscopy, Merlijn Witte and Josi Peterson-Maduro for technical assistance, Terhi Kärpänen for the vegfr4:tandem Tomato construct and Kelly Smith, T.Kärpänen and Andy Oates for critical reading of the manuscript.
Conflict of Interest statement . The authors declare no conflicts of interest.