Critical roles of CD146 in zebrafish vascular development

Authors

  • Barden Chan,

    1. Division of Nephrology, Center for Study of the Tumor Microenvironment, Beth Israel Deaconess Medical Center, and Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • Sonia Sinha,

    1. Division of Nephrology, Center for Study of the Tumor Microenvironment, Beth Israel Deaconess Medical Center, and Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • Dan Cho,

    1. Division of Nephrology, Center for Study of the Tumor Microenvironment, Beth Israel Deaconess Medical Center, and Harvard Medical School, Boston, Massachusetts
    2. Division Hematology-Oncology, Center for Study of the Tumor Microenvironment, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • Ramani Ramchandran,

    1. Division of Nephrology, Center for Study of the Tumor Microenvironment, Beth Israel Deaconess Medical Center, and Harvard Medical School, Boston, Massachusetts
    Current affiliation:
    1. National Cancer Institute, Key West Center, Bldg 14, 9610, Medical Center Drive, Rockville, MD 20850
    Search for more papers by this author
  • Vikas P. Sukhatme

    Corresponding author
    1. Division of Nephrology, Center for Study of the Tumor Microenvironment, Beth Israel Deaconess Medical Center, and Harvard Medical School, Boston, Massachusetts
    2. Division Hematology-Oncology, Center for Study of the Tumor Microenvironment, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, Massachusetts
    • 330 Brookline Avenue, RW 563, Beth Israel Deaconess Medical Center, Boston, MA 02215
    Search for more papers by this author

Abstract

In this report, we use zebrafish as a model system to understand the importance of CD146 in vascular development. Endothelial-specific expression of CD146 was verified by whole-mount in situ hybridization. Suppression of CD146 protein expression by antisense morpholino oligonucleotides (MO) resulted in poorly developed intersomitic vessels (ISVs). In CD146 morphants, we observed a lack of blood flow through the ISV region, despite that fluorescence microangiography showed that the ISVs were present. This finding suggests that the lumens of the developing ISVs may be too narrow for proper circulation. Additionally, remodeling of the caudal vein plexus into functional vascular tubes appeared to be affected. Suppression of CD146 protein expression resulted in a circulation shunt that caused the circulation to by-pass part of the caudal artery/vein system. The same vascular defects were recapitulated by using a second morpholino oligonucleotide. This morphant expressed a truncated CD146 protein with amino acids V32 to T57 at the N terminus deleted in an in-frame manner. This region, therefore, is likely to contain elements critical for CD146 function. This study provides the first in vivo functional assessment of CD146 in embryonic development by showing that knockdown of CD146 protein expression severely hinders vascular development in zebrafish. Developmental Dynamics 232:232–244, 2005. © 2004 Wiley-Liss, Inc.

INTRODUCTION

CD146, also known as MUC18, A32, Mel-CAM, and S-Endo1 (Lehmann et al., 1989; Shih, 1999), is a transmembrane glycoprotein that belongs to the immunoglobulin superfamily with a characteristic V-V-C2-C2-C2 domain structure (Williams and Barclay, 1988; Sers et al., 1993; Shih, 1999). Although it was first identified as a human melanoma-associated antigen (Holzmann et al., 1987; Lehmann et al., 1987), this protein is also highly expressed in the endothelium (Bardin et al., 1996a, b). In fact, it was determined that CD146 was the antigen recognized by a highly endothelial-specific monoclonal antibody called P1H12 (Solovey et al., 2001).

Although its cellular function is relatively poorly understood, CD146 is believed to play a role in cell–cell adhesion. It has been shown by immunohistochemical studies that CD146 is localized at the intercellular junctions of confluent endothelial cells in vitro (Bardin et al., 1996a, 2001), although CD146 does not appear to form part of the tight junctions (Lampugnani and Dejana, 1997) nor the adherens junctions (Bardin et al., 2001). In addition, the cell surface expression of CD146 was shown to increase by twofold as human umbilical vein endothelial cells (HUVECs) grew from subconfluency to complete confluency (Bardin et al., 2001). Furthermore, overexpression of CD146 in fibroblasts decreased paracellular permeability (Bardin et al., 2001), whereas treatment of confluent human microvascular endothelial cells (HMVECs) with monoclonal antibody P1H12 increased permeability to albumin in vitro (Solovey et al., 2001).

The nature of CD146 adhesive property has been addressed in vitro mainly through cell-aggregation and solid-phase binding assays. Results from these experiments suggest that CD146 confers cell–cell adhesion by means of interactions with a heterophilic ligand that has yet to be identified (Johnson et al., 1997; Shih et al., 1997). This interaction is divalent-cation independent and can be inhibited by a recombinant full-length CD146 protein and a polyclonal antiserum raised against CD146 (Johnson et al., 1997; Shih et al., 1997).

In addition to its potential role in cell–cell adhesion, CD146 has also been shown to be a signaling molecule that induces several intracellular molecular events upon cross-linking by a monoclonal antibody. For example, treatment of HMVEC with the monoclonal antibody P1H12 caused caveolin to redistribute from the cell periphery to the cytoplasm and nuclear entry of the transcription factor NF-κB (Solovey et al., 2001). In addition, antibody-mediated cross-linking of CD146 in vitro recruited Fyn to the cytoplasmic domain of CD146, triggered a phospholipase C γ–dependent store-operated calcium mobilization and induced phosphorylation of cellular proteins such as Pyk2, FAK, CAS, and paxillin (Anfosso et al., 1998, 2001; Solovey et al., 2001). However, CD146 has not been shown to interact with other cell-surface adhesive molecules such as VE-cadherin, PECAM-1, or β1 integrin (Bardin et al., 2001).

Recently, a fully humanized CD146 monoclonal antibody ABX-MA1 (Mills et al., 2002) and a mouse monoclonal antibody AA98 (Yan et al., 2003) were developed. When tested in mouse tumor models, both antibodies inhibited tumor growth and metastasis and reduced microvascular density in tumors (Mills et al., 2002; Yan et al., 2003). Of interest, the antibody AA98 was shown to bind only to stimulated but not serum-starved HUVECs in vitro (Yan et al., 2003). Most importantly, this antibody was shown to recognize an epitope in CD146 that was only present in tumor endothelium but not in endothelial cells present in normal tissues (Yan et al., 2003). AA98 monoclonal antibody was shown to inhibit proliferation and migration of HUVECs in vitro and significantly reduce angiogenesis in chicken chorioallantoic membrane assays (Yan et al., 2003). Collectively, these studies suggest that CD146 plays a significant role in blood vessel development as well as in solid tumor growth.

The ease of manipulation makes the zebrafish system ideal to identify endothelial specific genes. Specifically, the antisense morpholino-oligonucleotide technology (Heasman et al., 2000; Nasevicius and Ekker, 2000) provides a rapid way to suppress protein expression to address the functionality of a specific gene in a developmental context. One of our goals is to use this model to rapidly characterize endothelial-specific genes with unknown in vivo functions. We chose CD146 as our first target gene, because a CD146 genetic knockout mouse is not available to date. Here, we report the use of the zebrafish system to understand the potential role of CD146 in vascular development.

RESULTS

Cloning of the Zebrafish CD146 Gene

A cDNA clone (clone number, 3512; GenBank accession no. BG985668) reported in the zebrafish cDNA database (http://zf.nichd.nih.gov/pubzf/) shows a high level of sequence identity to mouse and chicken CD146 at the RNA and protein levels, respectively (Kudoh et al., 2001). Therefore, we performed 5′ and 3′ rapid amplification of cDNA ends (RACEs) based on this sequence information to clone the full-length zebrafish CD146 (Fig. 1).

Figure 1.

Full-length mRNA and protein sequences of zebrafish CD146. The 2.3-kb transcript harbors an open reading frame that encodes a 617-amino-acid protein. The amino acids are numbered, starting from the translational start codon. The putative polyadenylation site is underscored with a wavy line. The nucleotides deleted due to aberrant splicing in the presence of splice blocker 1 and splice blocker 2 are in bold and underlined, respectively. The premature stop codon present in the aberrantly spliced mRNA induced by splice blocker 1 is in bold and italic.

The transcript is approximately 2.3 kb long, harboring an open reading frame that encodes a 617-amino-acid protein. The 5′ untranslated region is relatively short (approximately 73 nucleotides), similar to that of the transcript of human origin (26 nucleotides; Sers et al., 1993). The 3′ untranslated region is approximately 350 nucleotides long and contains a canonical polyadenylation signal sequence AAUAAA just 5′ to the poly-A tail (Fig. 1). When this mRNA sequence was compared with the zebrafish genomic sequence, a 200-kb genomic fragment (ctg24502.1) was identified to contain the first 13 predicted exons of CD146 (not shown). This genomic fragment covers nucleotides 1 to 1685 of the mRNA. A similar search was performed at the ENSEMBL Web site (http://www.ensembl.org/Danio_rerio/blastview). The search result indicates that CD146 is located on chromosome 19 of the zebrafish genome.

The zebrafish CD146 protein shows approximately 30% overall sequence identity to that of the human, avian, mouse, and rat orthologs (Fig. 2A). Of interest, the transmembrane and the cytoplasmic domains are the most conserved regions. When the protein sequence was scanned for protein motifs (at http://scansite.mit.edu/; Yaffe et al., 2001), five immunoglobulin domains (Ig1, K35 to V123; Ig2, C168 to V232; Ig3, G273 to S324; Ig4, G359 to V408; Ig5, N445 to A507) were identified in the extracellular region of the protein, consistent with the characteristic Ig domain structure present in the subgroup of immunoglobulin superfamily that CD146 belongs to (Johnson et al., 1996). The two putative PKC phosphorylation sequences (S593GK and S614DK) in human (Alais et al., 2001) are also conserved in the zebrafish protein (S575GK and S596GK). The result of a phylogenetic analysis is shown in Figure 2B.

Figure 2.

A: Protein sequence of zebrafish CD146 aligned to the putative orthologs from chicken (accession no. Y08855; Alais et al., 2001), mouse (accession no. AAH26985), rat (accession no. BAB16048), and human (accession no. NP_006491; Lehmann et al., 1989). The alignment was performed at http://prodes.toulouse.inra.fr/multalin/multalin.html (Corpet, 1988). Highly conserved amino acid residues are highlighted. The leader sequence and the transmembrane domain of human CD146 are underlined by a solid line and a wavy line, respectively. Amino acids of the zebrafish protein are also numbered. B: Phylogenetic analysis based on the CD146 protein sequences. The analysis was conducted with the software MEGA3 (Kumar et al., 2001) using the Neighbor-Joining method. The long isoform of DIM-1 from Caenorhabditis elegans (accession no. AAM23316) is a member of the immunoglobulin superfamily (Rogalski et al., 2003) and was used as the outgroup in the analysis. Bootstrap support values based on 100 resamplings are presented at branch points.

Transient expression of zebrafish CD146 tagged with a fluorescent protein at the C terminus in 293 cells showed cell surface expression pattern similar to that of transiently expressed human CD146/YFP (data not shown). In addition, the first 11 amino acids of zebrafish CD146 directed the secretion of YFP into the culture medium when fused to the N terminus of the protein and transiently expressed in 293 cells, suggesting that this region contains a leader-peptide sequence and that Met1 is the authentic initiation methionine (data not shown). We also tested two commercially available antibodies against human CD146 (a polyclonal antibody from ZYMED and a monoclonal antibody from CHEMICON) in Western blot analysis and determined that the antibodies did not cross-react with the zebrafish protein overexpressed in 293 cells (data not shown).

Expression of CD146 in Zebrafish Embryos Is Highly Endothelial Specific

Next, CD146 expression pattern in zebrafish embryo was examined by whole-mount in situ hybridization by using a probe that was derived from the first 0.6 kb of the mRNA (Fig. 3A). In a 24-hr-old embryo, mRNA of CD146 was detected in the dorsal aorta (Fig. 3B, red arrow), caudal artery (Fig. 3B, red arrowhead), caudal vein (Fig. 3B, blue arrowhead), and the sprouting intersomitic vessels (Fig. 3B, green arrow). Expression in neuronal cells was also detected (Fig. 3B, black arrow). This expression pattern is essentially identical to that reported in the zebrafish cDNA database (http://zf.nichd.nih.gov/pubzf/), where the whole-mount in situ experiment was performed using a 0.5-kb probe derived from the 3′ end of the mRNA. Of interest, only a very low level of CD146 mRNA expression was detected in the posterior cardinal vein at this stage (Fig. 3B, blue arrow). The expression specificity was best illustrated by a cross section of the embryo (Fig. 3D). Intense staining of the message was detected in the aorta (Fig. 3D, red arrow); only a very low level of staining was detected in the posterior cardinal vein (Fig. 3D, blue arrow). In contrast, CD146 expression in the posterior cardinal vein was detected at day 2 (Fig. 3E, blue arrow) and day 3 (data not shown). This finding highlights the dynamics of the expression pattern of this endothelial marker during development. mRNA staining of CD146 in the head showed clear endothelial expression pattern (Fig. 3C). Certain cranial vessels (e.g., the middle cerebral vein) were stained, while others (e.g., the dorsal ciliary and median palatocerebral veins) did not appear to express CD146. Some other vascular structures stained included the aortic arches and the heart.

Figure 3.

Whole-mount in situ hybridization of CD146 in zebrafish. The digoxigenin-labeled antisense probe was derived from the first 0.6 kb of the mRNA. A: Overall staining pattern of CD146 in a 24-hour postfertilization (hpf) embryo. B: A magnification at the trunk-tail region of a 24-hpf embryo. Notice the staining of the intersomitic vessels (green arrow), dorsal aorta (red arrow), caudal artery (red arrowhead), and the caudal vein (blue arrowhead). Staining of the posterior cardinal vein was noticeably absent at this stage (blue arrow). Staining of neuronal cells was also seen at this stage (black arrow). C: Staining pattern in the head region of an embryo at 3 days postfertilization. D: Cross-section of an embryo at 24 hpf at the trunk region showing the expression of CD146 in the artery and a very low level of expression in the vein. E: Expression pattern of CD146 mRNA in a 2-day-old embryo showing endothelial specificity. Notice the expression of CD146 in the posterior cardinal vein. F: Cross-section of a 3-day-old embryo at the trunk region showing expression of CD146 in both the dorsal aorta and posterior cardinal vein.

Protein Expression Knockdown of CD146 Inhibits Proper Development of Intersomitic Vessels

To address the functional importance of CD146 in embryonic vascular development, we performed protein expression knockdown experiments by using antisense MOs. The first MO (splice blocker 1) was designed to target the seven nucleotides at the 3′ end of exon 10 and the first 18 nucleotides of intron 10 (Fig. 4A). Microinjection of splice blocker 1 into the embryos resulted in aberrant splicing. When total RNA harvested from control MO-injected embryos was used as the template for reverse transcriptase-polymerase chain reaction (RT-PCR), a specific band of ∼300 bp was amplified (Fig. 5A, lane 1). The identity of this PCR product was confirmed to be wild-type CD146 by DNA sequencing (Fig. 6A). When the same analysis was performed using total RNA harvested from splice blocker 1–injected embryos, a PCR product with the size <300 bp was obtained (Fig. 5A, lanes 3 and 5). DNA sequencing revealed that this product was derived from an aberrantly spliced CD146 mRNA with nucleotides 1260 to 1282 deleted (Fig. 6B). Apparently, with the predicted splice donor site of intron 10 blocked by the morpholino oligonucleotide (see Fig. 4A), G1260 of the mRNA was used as an alternate, cryptic splice donor site (Fig. 6B), generating an aberrantly spliced mRNA harboring a premature stop codon 186 amino acids N terminal to the predicted transmembrane domain (Fig. 1, deleted amino acids underlined and the premature stop codon in bold and italic).

Figure 4.

Design of splice blocking antisense morpholino oligonucleotides. Blast analysis revealed a 200-kb genomic fragment (ctg24502.1) in the zebrafish genomic database containing the first 13 presumed exons of CD146. A,B: Exons 10 (A) and 2 (B) were chosen as the targets. The regions complementary to the morpholino oligonucleotides used are underlined. A: Splice blocker 1 targets the 3′ end of exon 10 (bold) and the predicted GT splice donor site of intron 10. B: Splice blocker 2 targets the 3′ end of exon 2 (bold) and the predicted GT splice donor site of intron 2.

Figure 5.

Agarose gel analysis showing splice blocker 1 inhibited proper splicing of CD146 pre-mRNA. Fish embryos were injected with either control morpholino oligonucleotide (MO) or the splice blocker 1. Total RNA was extracted at day 2 of development. Reverse transcriptase-polymerase chain reaction (RT-PCR) was performed to determine whether splicing was blocked. A: PCR using a set of primers that amplified both the full-length and aberrantly spliced CD146 transcripts. The detection of a 300-bp DNA fragment indicated that proper splicing occurred when a control MO was injected (lane 1). A slightly shorter DNA fragment was amplified when the embryos were injected with splice blocker 1 (lanes 3 and 5). Mock treatments without RT (lanes 2, 4, and 6) were also included as controls. B: PCR using a set of primers that specifically amplified the full-length but not the truncated CD146 mRNA. Experiments were performed in triplicate. RT-PCR specifically amplified a 1.2-kb band using total RNA derived from control MO-injected embryos. No PCR product was detected when the embryos were injected with splice blocker 1, confirming that aberrant splicing took place. As controls, RT-PCR was performed to detect the expression of another endothelial specific gene called Fli 1. No-template PCR controls are indicated by a minus sign.

Figure 6.

Sequence of CD146 cDNA generated from control morpholino oligonucleotide (MO)-injected and splice blocker–injected embryos. Nucleotides are numbered starting from the A in the initiation codon ATG as 1. A: Sequence of CD146 coding region from nucleotide 1246 to 1323 in control MO-injected embryos reveals that the control MO had no effects in RNA splicing in this region. Exon 10 (black line) was joined to exon 11 (red line) as predicted. B: Sequence analysis reveals that splice blocker 1 inhibited normal splicing between exons 10 and 11. With the predicted splice donor site of intron 10 blocked by the morpholino oligonucleotide (see Fig. 4A), G1260 of the mRNA was used as an alternate, cryptic splice donor site. This aberrantly spliced mRNA results in a premature stop codon within the open reading frame of CD146. C: Sequence of CD146 coding region from nucleotide 88 to 183 in control MO-injected embryos reveals that the control MO had no effects in RNA splicing in this region. Exon 2 (black line) was joined to exon 3 (red line) as predicted. D: Sequence analysis reveals that splice blocker 2 inhibited normal splicing between exons 2 and 3. With the predicted splice donor site of intron 2 blocked by the morpholino oligonucleotide (see Fig. 4B), G94 of the mRNA was used as an alternate, cryptic splice donor site. This aberrant splicing event yielded a mutant CD146 with an in-frame deletion of 26 amino acids (V32 to T57) at the N terminus, which contains a conserved cysteine residue (C42) presumably important for the folding of the first immunoglobulin domain.

To confirm the effect of splice blocker 1 on CD146 mRNA splicing, another set of PCRs was performed in triplicate by using a pair of isoform-specific primers. This primer set was designed to detect the full-length, wild-type CD146 cDNA specifically. As shown in Figure 5B, the full-length CD146 was only detected in the control MO injected but not the splice blocker 1–injected embryos. As control, detection of another endothelial specific gene fli-1 was also shown.

To ascertain the role of CD146 in vessel development, blood circulation of embryos at 3 days postfertilization (dpf) was monitored (Fig. 7, boxed regions). Control injected fish showed normal circulation. Blood cells were seen traveling through the length of the embryo by means of the artery/vein system (Supplementary Movie 1A, which is available online at www.interscience.wiley.com/jpages/1058-8388/suppmat). Importantly, circulation was also detected throughout the intersomitic vessels, indicating that these vessels were functional at this stage of development. In contrast, circulation was abnormal in embryos injected with splice blocker 1. Although blood cells were seen traveling in the artery and vein, little or no circulation was detected in the intersomitic vessels (Supplementary Movie 1B). This observation suggests that suppression of CD146 protein expression blocked the development of the intersomitic vessels. This effect was tabulated, and the results are shown in Table 1. We decided to score the vessel development of the embryos by the percentage of functional intersomitic vessels present at 3 dpf. This titration experiment was done multiple times in small-scale pilot experiments, and a similar inhibition pattern was consistently observed. Next, we chose to inject the embryos with either 1.2 pmol control MO or 0.8 pmol splice blocker 1 in three additional independent experiments. The results are summarized in Figure 7C. Splice blocker 1 clearly had a deleterious effect on the development of intersomitic vessels. The deleted amino acids, therefore, are critical to CD146 functions. Because the transmembrane and cytoplasmic domains of CD146 were deleted in the morphants, intracellular signaling events triggered by CD146 were presumably abolished. A subpopulation of embryos injected with the MO showed different degrees of curvature at the trunk/tail region. An example is shown in Figure 7D. However, there did not appear to be a correlation between vascular defects and the body curvature.

Figure 7.

Effects of morpholino oligonucleotide (MO) in vascular development in zebrafish. A,B: Embryos were injected with 0.8 pmol of either the control MO (A) or splice blocker 1 (B) and examined at 3 dpf. Movie clips of blood circulation (boxed regions) observed under light microscopy were also recorded (see Supplementary Movies 1A and B). C: Histogram showing the distribution of embryos according to percent of functional intersomitic vessels present at 3 days postfertilization (dpf). Amounts of the control MO or splice blocker 1 injected were 1.2 or 0.8 pmol, respectively. Results were the averages of three different experiments. The percentages do not add up to 100% due to rounding errors. D: A 3-day-old embryo injected with splice blocker 1 showing a curvature in the trunk/tail region. Scale bars = 0.1 mm in A,B,D.

Table 1. Effects of Splice Blocker 1 Injection on Intersomitic Vessel Developmenta
Percent of functional ISVPercent of embryos
Control MOSplice blocker 1
1.2 pmol (n = 30)1.2 pmol (n = 25)0.8 pmol (n = 34)0.4 pmol (n = 33)
  • a

    MO, morpholine oligonucleotide; ISV, intersomitic vessels.

80–10010043100
50–800000
20–5008120
0–20088850

Reduced Protein Expression of CD146 Causes Disruption of Caudal Plexus Reorganization

Additional vascular defects were also detected in CD146 morphants. For example, the angiogenic remodeling process of the caudal plexus into functional vascular tubes was complete at day 3 in the control MO-injected fish (Movie 2A). Blood could be seen traveling through the entire length of the embryo by means of the dorsal aorta, the caudal artery, the caudal vein, and the posterior cardinal vein. However, this vascular reorganization of the plexus appeared to be defective in the splice blocker 1–injected embryos (Supplementary Movie 2B). The caudal plexus failed to completely remodel and organize into functional tubes. There was a circulation shunt that caused the circulation to by-pass much of the caudal artery/vein system. A stagnant pool of blood cells accumulated at the tail. Although no circulation was detected, these trapped cells appeared to exhibit slight pulsatile motion, suggesting that the caudal plexus might not be developed and organized properly for circulation but connected enough to allow hemodynamic forces to reach the tail region.

CD146 Appears To Be Important in Formation of Vessels With Proper Lumen Size

To investigate the integrity of the vasculature, we performed microangiography to detect lumenized vessels. Embryos at 3 dpf were injected with red fluorescent microbeads with an average diameter of 0.02 μm. The results are shown in Figure 8. In control MO-injected fish, vascular development as seen by microangiography appeared to be normal. In contrast, the overall vessel integrity of the splice blocker 1–injected fish appeared to be compromised. For example, distinct cranial vessels were seen in the control-injected fish (Fig. 8A). In contrast, the cranial vessels of the splice blocker 1-injected embryo appeared to form less-distinct tubular structures (Fig. 8B). In particular, a group of vessels called the cranial division of the internal carotid was less organized compared with the control (arrowheads in Fig. 8A,B). Second, the intersomitic vessels were narrower compared with the control, even though these vessels were detected by microangiography. A movie clip of the circulation again revealed that the development of intersomitic vessels was severely affected (Supplementary Movie 3). However, these ISVs were lumenized as seen by fluorescence microangiography (Fig. 8D). A similar defect was observed in the caudal plexus region. When comparing the microangiogram and the blood circulation of the splice blocker–injected embryos, we observed that blood did not flow through part of the lumenized caudal artery and vein vessels (corresponded to the region marked between the arrowhead and the arrow in Fig. 8D), resulting in a circulation shunt. This finding suggests that the lumen at the site where the circulation shunt occurred was too narrow for blood cells to pass through. Our observations suggest that CD146 may be important in formation of vessels with proper lumen size.

Figure 8.

Visualization of vasculature by microangiography. A–D: Embryos were injected with 0.8 pmol of either the control morpholino oligonucleotide (MO; A,C) or splice blocker 1 (B,D) and examined at 3 days postfertilization. The vessels in the head region of the splice blocker 1–injected embryo were less developed when compared with the control MO-injected embryo (see text). The arrowheads in A and B indicate a group of vessels called the cranial division of the internal carotid. Importantly, intersomitic vessels were detected in the angiogram of the morphants, while blood did not circulate through most of these vessels (see Supplementary Movie 3). The white arrow in D indicates the point at which blood exits the artery and enters the vein. No circulation was observed between the arrowhead and the arrow, even though this segment was lumenized. The boxed regions in A and B were magnified and are shown in C and D, respectively. Orientation of the embryo in C and D: left/caudal and right/rostral. Scale bars = 0.1 mm in A–D.

Elements Critical to CD146 Function Are Present at the N-Terminus of the Protein

To validate our data with splice blocker 1, we designed a second antisense morpholino oligonucleotide to target the 3′ end of exon 2 and the predicted GT splice donor site of intron 2 (Fig. 4). When this MO was delivered into the embryos by microinjection, it also caused aberrant mRNA splicing. RT-PCR analyses were performed with total RNA harvested from control MO- and splice blocker 2–injected embryos. The control MO did not affect splicing at the exon2/intron2 junction of CD146 pre-mRNA (Fig. 6C). However, sequencing data revealed that splice blocker 2 inhibited normal splicing between exons 2 and 3 (Fig. 6D). With the predicted splice donor site of intron 2 blocked, G94 of the mRNA was used as an alternate, cryptic splice donor site. This aberrant splicing event yielded a mutant CD146 with an in-frame deletion of 26 amino acids (V32 to T57) at the N terminus (Fig. 1, deleted region in bold).

Deletion of amino acids V32 to T57 of CD146 also affected the development of intersomitic vessels in zebrafish embryos. At day three, circulation was normal in embryos injected with the control MO (Supplementary Movie 4A). In splice blocker 2–injected fish, no circulation was seen in the intersomitic vessels (Supplementary Movie 4B and Table 2), a phenotype very similar to that generated by splice blocker 1–induced CD146 protein expression knockdown. Therefore, these two MOs appeared to have phenocopied each other. Furthermore, this result strongly suggests that the region between V32 and T57 is likely to contain elements critical for CD146 function.

Table 2. Effects of Splice Blocker 2 Injection on Intersomitic Vessel Developmenta
Percent of functional ISVPercent of embryos
Control MOSplice blocker 2
1.2 pmol (n = 53)1.2 pmol (n = 50)0.8 pmol (n = 31)0.4 pmol (n = 33)
  • a

    MO, morpholine oligonucleotide; ISV, intersomitic vessels.

80–1008722994
50–8062263
20–5026230
0–20690233

DISCUSSION

In this report, we cloned and characterized the CD146 ortholog in zebrafish. The overall amino acid sequence similarity between the zebrafish and human proteins was approximately 30%. Whole-mount in situ hybridization results indicate that the CD146 has a dynamic expression pattern in the vascular tree during development. At 1 dpf, CD146 transcript is detected in the dorsal aorta, caudal artery, and the intersomitic vessels. In the venous system, however, CD146 is detected only in the caudal vein but not in the posterior cardinal vein at this stage (Fig. 3B). However, CD146 began to express in the posterior cardinal vein at 2 dpf (Fig. 3E). The functional importance of the dynamic expression of CD146 in the venous system is yet to be understood. CD146 is also expressed in the heart of developing zebrafish embryos and may play a role in cardiac development. We occasionally observed pericardial edema in the morphants, but there did not seem to be a correlation between this phenotype and the vascular defects described here. Future studies will include the analysis of the relationship between heart development and CD146 functions. In this report, we focused on the vascular functions of CD146 primarily because of the apparent roles CD146 plays in tumor angiogenesis (Mills et al., 2002; Yan et al., 2003).

When the protein expression of CD146 was knocked down, vascular development in the zebrafish embryo was severely affected. One most prominent defect was the lack of blood circulating in the intersomitic vessels by 3 dpf (Supplementary Movies 1A and B). This lack of circulation was not due to the absence of intersomitic vessels per se, because these vessel structures were clearly detected by fluorescence microangiography, although many of them appeared to have uneven thickness and stenosis was evident even when observed under a fluorescence microscope (Fig. 8B,D). Although the lumen size of the intersomitic vessels in zebrafish has not been reported in literature, the diameters of artery and vein in zebrafish embryos have been measured to be 12.8 and 9.0 μm, respectively (Fritsche et al., 2000). The use of fluorescent microbeads with an average diameter of 0.02 μm allowed us to detect vessels with a diameter >400 times smaller than that of the vein. Based on our results, we speculate that, in the absence of CD146, the intersomitic vessels cannot develop into functional tubes with a lumen opening large enough for blood cells to pass through.

The development of the intersomitic vessels is a highly organized and choreographed process. It was reported recently that formation of the intersomitic vessels involves three different endothelial cell types sprouting from the dorsal aorta (Childs et al., 2002). The first two form a T-like structure at the dorsal side of the dorsal aorta and the ventral side of the dorsal longitudinal anastomotic vessel. The third one connects these two T structures. Because the intersomitic vessels were detected in the morphants with CD146 protein expression suppressed, this finding suggests that CD146 may play a role in the step(s) after the basic architecture of the intersomitic vessels has been laid out by these three types of endothelial cells.

Data from our morpholino experiments indicate that CD146 may play an important role in the remodeling of the caudal vein plexus into functional vascular tubes as well. In the CD146 morphants, circulation flowed from the dorsal aorta into the caudal artery and returned prematurely into cardinal vein, creating a circulation shunt in the caudal artery/vein system (Supplementary Movie 2B). This circulation shunt appears to be distinct from that observed in the zebrafish mutant mindbomb (mib) defective in Notch signaling (Lawson et al., 2001). In the mib mutant, the circulation shunt occurred in the trunk region where circulation flowed directly from the dorsal aorta into the posterior cardinal vein without entering the caudal artery/vein system. This circulation by-pass was speculated to arise from disorganization of the boundary between the dorsal aorta and the posterior cardinal vein based on histological data (Lawson et al., 2001). In the CD146 morphants described here, blood circulation entered the caudal artery but returned into the venous system prematurely into the caudal vein. However, fluorescence microangiography revealed that the caudal plexus was lumenized (Fig. 8D, see region between the arrow and the arrowhead). Therefore, the caudal shunt present in the CD146 morphants may be a result of restricted lumen diameter, a structural defect similar to that observed in the intersomitic vessels.

Our observation that CD146 is important in vascular development is in agreement with several studies reported in the literature. For example, it was shown the CD146 expression was up-regulated in human microvascular endothelial cells that had undergone tubular morphological changes on matrigel (Glienke et al., 2000). In additional, the ability of endothelial cells to form tubes on matrigel can be blocked by a CD146 neutralizing antibody (Mills et al., 2002). Functional suppression of CD146 by an antibody can also inhibit endothelial cell migration in vitro and neovascularization in chicken chorioallantoic membrane assays (Yan et al., 2003).

We have also identified V32 to T57 of CD146 to be an important sequence element in its protein function. We used an antisense morpholino oligonucleotide targeting the predicted exon 2/intron 2 junction of zebrafish CD146 pre-mRNA to create a mutant protein with those 26 amino acids deleted in an in-frame manner. This truncated protein appeared to be nonfunctional. This region may contain a recognition epitope for CD146 functions. Alternatively, the deletion may have affected proper protein folding, as the region deleted contains a conserved cysteine residue in the first immunoglobulin domain.

Vascular development requires extensive dynamic changes of endothelial morphology. These dramatic morphological changes were recorded recently by time-lapse microscopy in a developing zebrafish embryo in real-time (Lawson and Weinstein, 2002). Endothelial cells were seen to extend and retract long filopodial processes during path finding. Communication appears to occur between these activated endothelial cells through filopodial contacts, resulting in proper vascular remodeling and regression to ensure the development of functional lumens (Lawson and Weinstein, 2002). The exact molecular basis of these cellular behaviors is certainly complex, but highly orchestrated and dynamic changes of the cytoskeleton organization are probably involved (Davis et al., 2002). For example, Cdc42, Rac1, and integrins are some of the signaling molecules intimately linked to cytoskeleton functions that have been shown to be critical in endothelial tube and lumen formation in vitro (Davis and Camarillo, 1996; Bayless et al., 2000; Bayless and Davis, 2002; Davis and Bayless, 2003).

Although the molecular mechanism by which CD146 affects proper lumen formation observed in this report is at this point unclear, it is tempting to speculate that CD146 may achieve such function by regulating the cytoskeleton. It is known that antibody-mediated cross-linking of CD146 in vitro initiates a cascade of intracellular signaling events that includes phosphorylation of focal adhesion kinase, paxillin, and p130CAS (Anfosso et al., 1998, 2001). Additionally, CD146 has been shown to interact with the actin cytoskeleton in HUVEC (Bardin et al., 2001). Moreover, the CD146 ortholog in chicken has been shown to modulate expression of several integrins that are known to be expressed in endothelial cells (Alais et al., 2001; Hodivala-Dilke et al., 2003). All these cellular events appear to be related to the cytoskeleton. Perhaps, CD146 mediates the formation of proper vessels through regulating the dynamic changes in the organization of the cytoskeleton during development. Further studies that are designed to dissect the intracellular signaling events important in CD146 functions are currently under way.

EXPERIMENTAL PROCEDURES

Zebrafish Care

Zebrafish (Danio rerio) was raised and maintained as described (Westerfield, 2000).

5′ and 3′ RACES

Total RNA was collected from zebrafish embryos at 48 hours postfertilization (hpf) by using either the RNAaqueous-4PCR kit from Ambion (Austin, TX) or the RNeasy Mini kit from Qiagen (Valencia, CA). The 5′ and 3′ RACEs were carried out by using the GeneRacer Kit from Invitrogen (Carlsbad, CA) according to the manufacturer's protocol. The kit was chosen because the transcriptional start site of the mRNA can be obtained directly. Briefly, 5′ RACE was performed by using 5′-GGCACAGGGAGTGTCTTCTTGCCTTATA-3′ as the gene-specific primer for first-strand cDNA synthesis. The cDNA was amplified by nested PCR using the GC-Advantage DNA polymerase from Clontech (Palo Alto, CA). The primers for the first round of PCR were the GeneRacer 5′ Primer (provided in the GeneRacer Kit) and the gene-specific primer. For the second round of amplification, the GeneRacer 5′ Nested Primer (provided in the GeneRacer Kit) and a primer with the sequence 5′-GCAGCTTTCCCTTCTTGTAGAGGTAAT-3′ were used. The PCR products were cloned into the TOPO cloning kit from Invitrogen. Protein sequence alignment was performed at http://prodes.toulouse.inra.fr/multalin/multalin.html (Corpet, 1988). The mRNA sequence has been deposited in GenBank (accession no. AY497061).

In Situ Hybridization

Nucleotides spanning the 5′ untranslated region and the first 592 bases of the open reading frame of zebrafish CD146 were amplified and cloned into a vector derived from pCR-BluntII-TOPO (Invitrogen). To generate digoxigenin (DIG)-labeled antisense probes, the DNA plasmid was linearized with EcoRI, followed by transcription using SP6 polymerase and the 10× DIG RNA labeling mix from Roche (Indianapolis, IN). In situ hybridization was then performed according to a published procedure (Broadbent and Read, 1999; Read, 2000).

Microinjection and Fluorescence Microangiography

Two antisense (splice blocker 1. 5′- CAGTTGTAGTGACCCACCTCGCACG-3′; splice blocker 2. 5′-GTCTTTTTCTCACCGTAAACCAGAT-3′) and a control (5′-GACTCTATATTGCCTACCTTTACAG-3′) morpholino oligonucleotides were purchased from Gene-Tools (Philomath, OR). The morpholino oligonucleotides were injected into the yolk sac of zebrafish embryos between one- and four-cell stages. Embryos were raised in egg water in the presence of 0.003% (w/v) 1-phenyl-2-thiourea (Sigma, St. Louis, MO) to prevent pigmentation (Westerfield, 2000). To verify successful blockage of CD146 pre-mRNA splicing by the morpholino oligonucleotides, total RNA was collected from the zebrafish embryos at 2 dpf by using the RNeasy Mini kit from Qiagen. Reverse transcription was carried out by using either Thermoscript or Superscript III from Invitrogen with the gene-specific primer (5′-GGCACAGGGAGTGTCTTCTTGCCTTATA-3′). PCR was then performed by using GC-Advantage DNA polymerase (Clontech) and the following amplification primers: 5′-GATATCACCTTGACTTGCAACGCATTGTCT-3′ and 5′-GATTGGGTAGCCTTGTGCATAGCAGGTCAG-3′ for splice blocker 1 injection; 5′-ATGACCTACACCGCACTGCTGCTCGCGGGA-3′ and 5′-AGGATCGAGAGGGAAGCTGCCCTTGTACCA-3′ for splice blocker 2 injection. A pair of isoform specific primers that amplified only the wild-type CD146 cDNA (nucleotides 126 to 1287) has the following sequences: 5′-TACTTACACTCTGACGGAAGAAGCCAAGAT-3′ and 5′-CTTTCCTCGCACGTTTACTTGCACTGA-3′. Primers used to amplified Fli-1 were 5′-GAAAATGGACGGAACTATTAAGGAGGCGCT-3′ and 5′-TCTTCGAGTGCAGTTCAAGTTTTGGCATTTAAGGGT-3′.

Blood circulation in anesthetized fish embryos was recorded as time-lapse images using an inverted phase contrast microscope (Nikon Corporation, Tokyo, Japan), a Spot RT camera, and the Advanced SPOT software (Diagnostic Instruments, Inc., Sterling Heights, MI). The images were converted into QuickTime movies using a program called ImageJ.

Fluorescence microangiography was performed by using the red fluorescent carboxylate-modified FluoSpheres (0.02 μm) from Molecular Probes (Eugene, OR) as described (Weinstein et al., 1995). Fluorescence microangiograms were recorded with the Nikon fluorescence microscope using a filter set (41035:ex. 546; em. 605) from Chroma (Rockingham, VT).

Acknowledgements

This study was supported by seed funds from Beth Israel Deaconess Medical Center (Boston, MA).

Ancillary