Sox18 and Sox7 play redundant roles in vascular development

The acquisition of arterial and venous identity is an early event in endothelial cell differentiation, preceding the onset of blood circulation. Pioneering work in the zebrafish has revealed that the acquisition of arterial identity is governed by a genetic network including Shh, VEGF, and Notch signaling.^1,–3  Many aspects of this regulatory cascade are also conserved in mammals.⁴  More recently, the acquisition of venous identity was shown not to occur by default but rather to be genetically controlled. The orphan nuclear receptor COUP-TFII induces venous endothelial cell differentiation by suppressing Notch signaling.⁵  Although chick-quail grafting experiments point to an initial plasticity (ie, nascent endothelial cells are able to change their arteriovenous identity depending on the surrounding context), it is becoming clear that arterial and venous angioblasts segregate from the beginning of vasculogenesis.⁶  The ephrin/Eph system is involved in the establishment of arteriovenous (AV) cell identity, and is important for the segregation between arteries and veins. The transmembrane ligand ephrinB2, and its cognate tyrosine kinase receptor EphB4, are exclusively expressed by arterial and venous endothelial cells, respectively, both in mouse and in zebrafish, and this molecular distinction is essential to establish a functional hierarchical vascular network.⁶ 

Arteriovenous malformations (AVMs) are frequently induced by a loss of arterial or venous cell identity.⁷  In humans, hereditary hemorrhagic telangiectasia (HHT) is a vascular dysplasia characterized by anomalous fusion of arterioles and venules, leading to telangiectases in the skin and mucocutaneous tissues and intrahepatic and pulmonary AVMs. HHT type 1 and type 2 are both caused by autosomal dominant mutations in the transforming growth factor beta (TGF-β) receptors ENG and ALK1, respectively. Alk1^−/− homozygous mouse embryos exhibit large AV shunts between major arteries and veins, linked to a loss of ephrinB2 expression in the vasculature.⁸ 

Hypotrichosis-lymphedema-telangiectasia (HLT) is a human syndrome characterized by the association of the 3 pathologic traits.⁹  Dominant and recessive forms of this syndrome are linked to mutations in the SOX18 gene. SOX (SRY-related HMG-box) proteins are a family of transcription factors characterized by an HMG-box DNA binding domain; they are present throughout the animal kingdom and play key roles in embryonic development.^10,11  Twenty SOX genes are present in mammals, subdivided in several groups on the basis of the similarities among HMG-boxes.¹² SOX18 belongs to SOX subgroup F, together with the closely related SOX7 and SOX17. The spontaneous ragged mutant mice, first described in the 1950s,^13,14  represent the murine counterpart of the HLT syndrome and are characterized by defects in vascular, hair follicle, and lymphatic development. Closely spaced mutations within the mouse Sox18 gene underlie the 4 allelic variants of the ragged phenotype.^15,16  These mutations alter the transactivation and C-terminal domains of the Sox18 protein, without affecting its DNA binding domain, and they act genetically in a semidominant fashion. The most severe allele, ragged-opossum (RaOp), causes embryonic lethality before embryonic day (E) 11.0 in homozygous mutant mice.^15,17  On the other hand, Sox18-null mice are viable and present only a mild coat defect.¹⁸  It has been suggested that related SOX proteins can compensate for the absence of SOX18 in null mice, while the mutations in ragged mice give rise to truncated SOX18 proteins that act in a dominant-negative fashion and interfere with functionally redundant SOX transcription factors.^18,19 Sox7 and Sox17 have been found to be expressed together with Sox18 during vascular development in mouse,²⁰  and may provide the suggested redundancy.

We decided to specifically address the functional redundancy of Sox-F genes in vascular development by making use of the zebrafish model system. We first identified the zebrafish sox18 and sox7 genes and showed that they have largely overlapping expression in the developing vasculature. Simultaneous partial knockdown of both genes, but not of either one, using moderate amounts of specific morpholinos, caused a block of trunk/tail blood circulation. The expression of several endothelial markers was altered in the double morphants from the very first stages of vasculogenesis. Venous endothelial markers were more affected than arterial ones. The double morphants presented several shunts between the dorsal aorta and the cardinal vein.

Our data indicate that sox18 and sox7 are functionally redundant, but collectively essential in the differentiation of vascular endothelial cells and in their acquisition of a full AV identity.

Zebrafish were raised and maintained according to established techniques.²¹  The following strains were used: AB (obtained from the Wilson lab, University College London, London, United Kingdom), tg(fli1:EGFP)^y1 (Ref. 22) (from the Lawson lab, University of Massachusetts Medical School, Boston), tg(flk1:EGFP) (Ref. 23) (from the Stainier lab, University of California at San Francisco), and tg(gata1:dsRed) (Ref. 24) (from the Zon lab, Children's Hospital, Boston, MA). Fixed clo^s5 embryos were kindly provided by M. Gering (Nottingham, United Kingdom).

Antisense morpholinos (MOs; Gene Tools, Philomath, OR) were designed against the AUG translation start site region and the 5′ untranslated region (UTR): sox7-MO1, 5′-ACGCACTTATCAGAGCCGCCATGTG-3′; sox18-MO1, 5′-TATTCATTCCAGCAAGACCAACACG-3′; sox7-MO4, 5′-CTCAAACTTTGTTTCTCGGGCGCAG-3′; and sox18-MO4, 5′-CCAGCAAGACCAACACGATTAAAGC-3′. Splice MOs targeting the 5′ splice site of the intron within the HMG-box coding region were designed as follows: sox7-MO2, 5′-GTTAAAATCTTACCAAGCATCTTGC-3′; and sox18-MO2, 5′- GTGAGTGTCTTACCCAGCATTTTAC-3′. MOs, diluted in Danieau buffer,²⁵  were injected at 1- to 2-cell stage. Escalating doses of each MO were tested for phenotypic effects; as control for unspecific effects, each experiment was performed in parallel with a std-MO (standard control oligo) with no target in zebrafish embryos. For double knockdown experiments, we usually injected 250 fmol/embryo of each MO1, 750 fmol/embryo of each MO2, and 500 fmol/embryo of each MO4. scl spliceMO,²⁶  kindly provided by the Patient lab (Oxford, United Kingdom), was injected at 1 pmol/embryo.

RNA samples were extracted with TOTALLY RNA isolation kit (Ambion, Austin, TX), treated with RQ1 RNase-Free DNase (Promega, Madison, WI), and retrotranscribed with random hexamers (High Capacity cDNA Archive kit; Applied Biosystems, Foster City, CA), according to the manufacturers' instructions. Real-time polymerase chain reaction (PCR) was performed using the TaqMan methodology (Applied Biosystems) and an ABI thermocycler (Applied Biosystems); details on primers and probes are available on request. Additional details can be found in the legend for Figure S1 (available on the Blood website; see the Supplemental Materials link at the top of the online article).

Whole-mount in situ hybridizations (ISHs) were carried out essentially as described.^26,27  For double ISHs, fluorescein probes were purified on Spin Post-Reaction Clean-Up Columns (Sigma-Aldrich, St Louis, MO) prior to use, and red staining was obtained with an INT Red (Sigma-Aldrich)/BCIP (Roche, Mannheim, Germany) mix. The following probes were synthesized as described in corresponding papers: flk1,²⁸ fli1,²⁹ lmo2 and flt4,³⁰ gata1 and gata2,³¹ notch3,¹ deltaC,³² scl,³³ vegf165,³⁴ dab2,³⁵ etsrp and admr,³⁶ ang1,³⁷ radar,³⁸ cdh5,³⁹ vsg1,⁴⁰ grl,²  and kdrb.⁴¹ flt1 was kindly provided by S. Schulte-Merker (Ultrecht, The Netherlands); efnB2a and ephB4 were kindly provided by R. Patient.

For the sox18 probe, a BglII-KpnI fragment from IMAGE clone 6790334 was subcloned into pBSKS⁺; the resulting plasmid was digested with SmaI and transcribed with T7 RNA Polymerase (Roche). For the sox7 probe, the IMAGE clone 7044541 was digested with BamHI and transcribed with T7 RNA Polymerase (Roche).

Images were taken with a Leica MZFLIII epifluorescence stereomicroscope equipped with a DFC 480 R2 digital camera and IM50 Leica imaging software (Leica, Wetzlar, Germany). Confocal microscopy was performed on a Leica TCS SP2 AOBS microscope, equipped with an argon laser. Images were processed using the Adobe Photoshop software (Adobe, San Jose, CA). Movies were processed using the QuickTime Player software (Apple, Cupertino, CA).

Alkaline phosphatase staining was performed as described.⁴² 

2,3-butanedione monoxime (2,3-BDM; Sigma) was added at 8 mM to 24-hpf (hours postfertilization) dechorionated embryos to block heartbeat, as previously described.⁴³ 

Tg(fli1:EGFP)^y1 embryos at 30 hpf were fixed at room temperature in 4% paraformaldehyde/phosphate-buffered saline (PBS). Cryosections (24 μm) were obtained as described⁴²  and analyzed by confocal microscopy.

Embryos for histologic analysis were fixed at room temperature in 1.5% glutaraldehyde/4% paraformaldehyde/cacodylate buffer (pH7.2), postfixed in 1% osmium tetroxide, embedded in an Epon-Araldite mixture, and sectioned with an ultramicrotome. Semithin plastic sections (1-1.5 μm) were stained with violet gentian and basic fuchsin. Images were taken with an Olympus BH2 microscope (Tokyo, Japan), equipped with a Leica DFC320 digital camera.

By using the mouse sequences as queries in combined searches of genomic and expressed sequence tag (EST) databases (Ensembl, National Center for Biotechnology Information [NCBI]), we identified cDNA clones corresponding to putative Sox7 and Sox18 orthologs in zebrafish. These clones were obtained from the I.M.A.G.E. consortium (Lawrence Livermore National Laboratory, Livermore, CA) through MRC geneservice (Cambridge, United Kingdom). The complete sequences (Sox7, 390 aa; Sox18, 431 aa) were highly related to previously identified homologs in mammals, chick, Xenopus, and Fugu rubripes (Figure S1). Both proteins contain the β-catenin binding motif⁴⁴  that is present in all Sox-F group proteins with the notable exception of zebrafish Sox17. Zebrafish Sox7, Sox17, and Sox18 share very high amino acid identity within the HMG-box domain (> 85% in pairwise comparisons), but are less related to each other than to their mammalian homologs.

About 20% of the zebrafish genome is duplicated due to an event before the teleost radiation,⁴⁵  but we failed to find duplicates for sox7 and sox18, which map on chromosomes 20 and 23, respectively (Zv6 assembly; http://www.sanger.ac.uk/Projects/D_rerio). Both genes have a single intron within the sequence coding for the HMG-box, in a conserved position relative to their mammalian orthologs.¹¹ 

We analyzed the expression of the 3 zebrafish Sox-F group genes during embryonic and early larval development by real-time reverse transcription (RT)–PCR (Figure S2). As expected from its role in endoderm development,⁴⁶  the sox17 gene is expressed at maximum levels during gastrulation while its expression drops during somitogenesis. In contrast, sox7 and sox18 expression levels are low during gastrulation; they start increasing in somitogenesis and peak at around 1.5 dpf (days postfertilization). At this stage, blood circulation has already started, and the vascular tree is actively being remodeled.

Sox18 is expressed transiently in the endothelial component of the developing blood vessels in mouse and chicken (reviewed by Downes and Koopman¹⁹ ). Mouse Sox7 and Sox18 expressions are parallel in diverse locations, including the developing embryonic vasculature.^20,47 Sox7 expression in the embryonic vasculature was recently reported also in Xenopus laevis.⁴⁸ 

Our whole-mount ISH analysis (Figure 1) reveals that during somitogenesis zebrafish sox7 and sox18 are both expressed in bilateral stripes corresponding to the posterior lateral plate mesoderm (PLM) and, at a lower level, in the anterior lateral plate mesoderm (ALM). ALM and PLM are the regions where hemangioblasts, the presumptive common precursors of blood and endothelial cells, are thought to reside.^29,33 sox7 transcripts are detectable at 4-somite stage, and sox18 at 6- to 8-somite stage (Figure 1A-D). sox7 and sox18 are expressed in the innermost PLM fli1-positive cells, corresponding to endothelial cell precursors,^29,49  as revealed by double ISH (Figure 1E,F). The sox7 expression domain in the PLM appears slightly broader than that of sox18 (Figure 1E,F; data not shown).

During vasculogenesis in zebrafish, 2 waves of angioblast migration from lateral plate mesoderm to the midline give rise to 2 axial vascular cords (the arterial one preceding the venous) that subsequently lumenize. Secondary vessel formation by angiogenesis follows after the coalescence of arterial and venous angioblasts at the midline. We could see sox7 and sox18 expressing cells migrating from the PLM to the midline (data not shown). Both sox7 and sox18 marked the developing axial and intersomitic vessels (ISVs), and the intermediate cell mass (ICM), where endothelial and blood cell precursors reside. sox7 and sox18 were also expressed in the developing head vasculature.

sox7 and sox18 are differentially expressed in other regions of the embryo: sox7 staining was clearly associated with the developing otic vesicles (Figure 1I), whereas more diffuse sox18 staining was visible in the rostral region of the embryo from early somitogenesis (Figure 1C,D,F). At around 24 hpf, sox18 was expressed in rostral parts of the central nervous system, notably in the eye region (Figure 1L).

Based on their expression domains, it seemed likely that sox7 and sox18 could play a role in zebrafish vascular development. To place both genes in the hierarchy of genes controlling endothelial cell differentiation and blood vessel formation, we examined their expression in the cloche mutant (clo) and in wild-type embryos injected with a morpholino directed against scl (scl^mo). The zebrafish clo mutation affects a very early step in differentiation of blood and endothelial cells,⁵⁰  at the level of the putative hemangioblast. The scl gene acts downstream of clo⁵¹  and is required for hematopoietic and endothelial development.^26,52  At 6 to 8 somites, the expression of sox7 and sox18 in the ALM and PLM was lost in approximately one-quarter of the embryos (ie, in the expected clo homozygous mutants), along with a strong reduction in gata1 signal (Figure 2A,B,E,F; data not shown). At 24 hpf, expression of sox7 and sox18 was detected in putative clo embryos only in nonvascular domains (Figure 2D,H). When embryos of the same clutch were hybridized with cdh5 (coding for the endothelial-specific adhesion protein VE-cadherin), no expression was detectable in approximately one-quarter of the embryos except for a few cells in the cardinal vein plexus region, where also some residual sox18 staining was noticeable (Figure 2H; data not shown). A few remaining endothelial cells have indeed been observed in this region in clo mutants.^30,36,53  The vascular expression of sox7 and sox18 is thus controlled by cloche.

In contrast, the onset of sox7 and sox18 expression did not appear to be controlled by scl. In scl^mo embryos, the expression pattern of sox7 and sox18 at 6 to 8 somites in the ALM and PLM was not altered (Figure 2; compare panels J,N with I,M), whereas gata1 expression was absent or severely reduced (13 of 19; data not shown). The situation is different at 22 hpf: in scl morphants, the expression of sox7 is limited to the developing otic vesicles, the region of the first aortic arch, and the ICM (Figure 2L), but sox18, similarly to cdh5 (data not shown), is still expressed in the region of the posterior cardinal vein (PCV) and the caudal vein plexus (Figure 2P). As expected, sox18 staining in the dorsal aorta (DA) and in the ISVs is absent, because of the crucial role of scl in the development of the DA.²⁶  Scl thus appears to control the maintenance of the endothelial expression of sox7 but not of sox18.

Overall, the expression pattern of sox7 and sox18 genes points to a potential role in zebrafish vascular development.

To test the hypothesis that sox18 and sox7 play redundant roles in vascular development, as suggested for mammalian homologs, we designed morpholinos against regions surrounding the AUG translation start codon of sox18 or sox7 (sox18-MO1 and sox7-MO1).

Low doses of morpholinos (250 fmol/embryo) caused no morphologic alterations when injected separately (Figure S3). When low doses of sox18-MO1 and sox7-MO1 were coinjected in the same embryos, gross morphologic abnormalities were absent at 1 dpf (Figure 3). However, functional defects started to appear after the onset of blood circulation and became more evident by 2 dpf (Table S1). More than 90% of the double morphants showed no sign of blood circulation in the trunk and tail regions; blood flow was present only in the rostral part of the embryos, from heart to yolk through the anterior cardinal vein, on both sides or just on one side (Videos S1,S2). A premature venous return around the anus was observed in a few double morphants with residual limited blood flow in the trunk, possibly indicative of an AV shunt (Videos S3,S4).

By 3 dpf, a large number of double morphants showed extensive heart edema (Figure 3F), which we interpret as a secondary effect due to the block in circulation. Many double morphants showed blood poolings in various positions, including eyes, trunk, and pericardiac region (Figure 3H-J,L).

Another set of morpholinos (sox18-MO4 and sox7-MO4), targeting different sequences in the 5′ UTR of sox18 or sox7 transcripts, gave qualitatively similar results (Table S1). This confirms that the massive circulatory phenotype is specifically linked to knockdown of both genes, but not of either gene at a time. When combined, the 2 sets of morpholinos could cause a massive circulatory phenotype even at doses that were basically ineffective on their own, confirming their targeting specificity (Table S1). Splice morpholinos (sox18-MO2 and sox7-MO2) interfering with the maturation of sox18 and sox7 transcripts also reproduced the circulatory phenotype when combined, though with a lower penetrance (Figure S4; Table S2).

To gain insight into the molecular events following the double knockdown of sox7 and sox18, we analyzed the expression of several endothelial and nonendothelial markers. A subtle reduction and/or partial disorganization in the expression of fli1, etsrp, and lmo2 genes in the PLM was detectable by ISH in the double morphants at early somitogenesis (Figure 4A,A′-C,C′), whereas the expression of scl and of the early hematopoietic markers gata2 and gata1 appeared unaffected (Figure S5; data not shown). A reduced expression of cdh5, but not of other endothelial markers, in the primordium of the PCV was noticed at 22 somites (Figure 4D,D′, H,H′; data not shown).

At 22 to 26 hpf, around the normal onset of blood circulation in control embryos, the expression of several endothelial markers (flk1, the vessel-specific gene 1 [vsg1], cdh5, the adrenomedullin receptor [admr], and dab2) appeared reduced in the double morphants, particularly in the trunk/tail region (Figure 4E,E′-G,G′,I,I′,J,J′). These defects were particularly evident in the PCV and in the caudal vein plexus.

Endothelial expression of some markers was specifically delayed in the double morphants. Arterial expression of flt4, for example, persisted in the double morphants at 26 hpf, whereas it was already restricted to venous endothelial cells in control embryos (Figure 4K,K′). The vein-specific expression of flt4 was indeed only detectable at 29 hpf in the double morphants (Figure 4L,L′). The expression of ephrinB2 (efnb2a) in the DA was markedly reduced in the double morphants at 22 hpf (Figure 4M,M′), but became comparable with control embryos by 26 hpf (Figure 4N,N′). Very little or no expression of the venous marker ephb4 could be detected in the caudal vein plexus region of the double morphants in the same interval and even at a later stage (29 hpf; Figure 4O,O′,P,P′).

In contrast, double knockdown of sox7 and sox18 does not alter the expression of DA markers such as gridlock (grl), notch3, deltaC, and kdrb genes (Figure S5; data not shown). A number of genes (vegf, angiopoietin 1 [ang1], the growth differentiation factor 6a [gdf6a]/radar) expressed in nonendothelial cells are known to influence vascular development, but their expression was not affected in the double morphants (Figure S5; data not shown).

The ISH analysis of key markers on double morphants injected with the second set of morpholinos (DMO4) gave qualitatively similar, though quantitatively less pronounced, results (Figure S6). The same markers were unaffected in the single morphants.

These data support the notion that sox7 and sox18 have a role in the establishment of the venous identity.

To better characterize vascular abnormalities, we made use of the tg(fli1:EGFP)^y1 transgenic line,²²  where the enhanced green fluorescent protein (EGFP) expression is driven by the promoter region of the early endothelial gene fli1. The overall fluorescence of this transgenic line is not substantially affected by the double knockdown of sox7 and sox18 (compare Figure 5A′ with 5B′), thus enabling us to use confocal microscopy to better define vascular defects. Starting from 2 dpf, the subintestinal vessels (SIVs) basket showed severe abnormalities in the double morphants, as also revealed by endogenous alkaline phosphatase (AP) staining at 3 dpf (Figure 5J,K and G,H). The SIV basket appeared less regularly organized and severely enlarged, with thinner vessels and very elongated spikes protruding toward the yolk. No enlargement of the SIV basket was associated with 2,3-BDM treatment, which weakens myocardial force through a block of myofibrillar ATPase and therefore eliminates blood flow⁴³  (Figure 5L).

Vascular defects were also detectable in the caudal region. The caudal vein plexus was severely reduced. Moreover, the double morphants showed enlarged, not patent ISVs with increased vessel branching and aberrant anastomoses, particularly in the dorsal somites (Figure 5N,O). The simple block of blood flow induced with 2,3-BDM did not induce similar abnormalities, apart from some thickening of the caudal ISVs (Figure 5P).

Coinjection of sox18-MO1 and sox7-MO1 in the tg(flk1:EGFP) line led to a severe reduction in the expression of the transgene which is driven by the flk1/vegfr2/kdra promoter. Residual fluorescence in the trunk axial vessels is limited to the DA (Figure 5D′). The strong difference in EGFP fluorescence in the 2 transgenic lines, together with the fact that fli1 is expressed earlier than flk1 in endothelial development, lead us to conclude that endothelial cells are specified but do not properly follow their differentiation program when sox7 and sox18 are simultaneously knocked down.

The ISH analysis suggested that the alterations in the expression of endothelial markers in the double morphants were more pronounced in the trunk/tail region and more profound in the axial vein than in the DA.

Histologic analysis was carried out on double morphants starting from 29 hpf. In the anterior trunk (ie, in the region of DA/PCV bifurcations), a single axial vessel was detectable in 1 or 2 points due to shunts between the DA and the PCV (Figure 6B′). The fusion of the DA and PCV endothelia in the trunk was also clearly visible in cryosections of tg(fli1:EGFP)^y1 double morphant embryos at 30 hpf (Figure 6A′). Around this stage, arteriovenous shunts leading to premature blood flow return in the arterior trunk were also visible in vivo in double morphants (Videos S5,S6). The anterior trunk fusions were clearly detectable also at 2.5 dpf and 3 dpf (Figure 6C′; data not shown).

By 3 dpf, the axial vein was significantly dilated in the trunk/tail region whereas the separation between the major axial vessels was reduced (Figure 6E′). These anomalies were also present in control embryos treated with 2,3-BDM (Figure S7) and could therefore be due to the lack of blood flow. In the double morphants, multiple AV shunts between the axial vessels were detectable in the posterior trunk/tail region starting from 2.5 dpf and became more extended with development (Figure 6D′,F′,G). On the contrary, no AV fusions were present in 2,3-BDM–treated embryos (Figure S7).

We conclude that the extensive AV shunts in the tail of the double morphants are late events specifically due to the sox7/sox18 double knockdown, whereas axial vascular tube formation per se is not affected. Our histologic analysis reveals putative occluding thrombi followed by huge blood poolings at or just following some of the AV shunts in the trunk (data not shown; Figure 6A′,C′).

We used zebrafish as a model system to study the role of Sox18 in endothelial cell differentiation and vascular development and to directly test the hypothesis of functional redundancy among Sox-F proteins.

We identified single-copy genes as bona fide sox18 and sox7. Our real-time RT-PCR analysis revealed that sox18 and sox7 have similar temporal expression patterns and suggested that they both act from early somitogenesis. In contrast, sox17 is mainly expressed during gastrulation and its expression drops afterward. The residual sox17 expression reported in somitogenesis with a transgenic line is limited to endodermal cells.⁵⁴  In contrast, all 3 Sox-F members are coexpressed in vascular endothelial cells, and Sox17 and Sox18 act redundantly in postnatal angiogenesis in mice.⁵⁵ 

Both sox18 and sox7 are expressed at very early steps of endothelial differentiation as: in the anterior and posterior lateral plate mesoderm, where endothelial cell precursors reside; in the angioblasts migrating toward the midline and coalescing to form the main axial artery and vein; in the endothelial cells of the nascent intersomitic vessels; and in the forming head vasculature. Each gene also shows unique expression domains (eg, in the central nervous system or in the otic vesicles), suggesting that they act redundantly in vascular development while acting nonredundantly in other developmental processes.

Endothelial expression of sox18 and sox7 is differently regulated: scl controls the endothelial expression of sox7, but not of sox18, at 22 hpf. scl knockdown does not affect their initial expression, consistent with the model in which scl acts downstream of cloche, playing a role in vascular development but not in the very initial phases of endothelial cell development, where instead etsrp would be a major player.^56,57 

Knocking down each single gene caused no or limited effect on blood circulation. In contrast, the double knockdown of sox18 and sox7 led to a massive and specific block in trunk/tail blood circulation. The effects of the combined knockdown of the 2 genes are more than additive, strongly implying that sox18 and sox7 act redundantly in vascular development. In contrast, our preliminary knockdown experiments with morpholinos against sox18 and sox17 did not reveal functional redundancy in zebrafish vascular development. It is noteworthy that SOX7 and SOX18 proteins were shown to be functionally redundant in cardiogenesis in Xenopus.⁵⁸ 

The emerging picture is that the double knockdown of sox18 and sox7 disturbs the normal acquisition of specific markers during endothelial cell differentiation. More specifically, the acquisition of the correct AV identity appears to be impaired. In axial veins, we found a moderate to strong reduction in the expression of panendothelial or venous markers. The expression of both ephrinB2 and EphB4 was affected, though differently since venous expression of EphB4 was stably reduced, while arterial expression of ephrinB2 appeared only transiently affected. In contrast, at the developmental stages we analyzed we could not detect alterations in several genes controlling arterial differentiation.

A common outcome of the lack of proper AV identity is the formation of AV shunts. Mutations or functional knockdown of genes of the Notch pathway, controlling arterial differentiation, and endothelial-specific overexpression of COUP-TFII, which controls vein identity, all lead to fusion of arteries and veins.^1,5,59,60  A similar picture emerges in sox18/sox7 double morphants in which we describe the presence of AV shunts at developmental stages close to the first onset of blood circulation.

Already at 29 hpf, AV shunts can be clearly detected in the trunk region of the double morphants, the DA being fused to the PCV. The fusions occur in recurrent sites along the axial vessels, and histologic analysis reveals that the extent of the AV shunts increases as development proceeds. At 3 dpf, multiple AV shunts lead to a single dilated axial vessel in the double morphants' tails, though distinct DA and CV (caudal vein) initially form. In the vast majority of the double morphants, no blood flow is detectable in the trunk/tail region, although the formation of the AV shunts is not apparently linked to the consequent change in hemodynamic forces. In fact, a chemical treatment blocking blood flow (2,3-BDM) does not result in AV fusions, and no such AVM have been reported for silent heart/tnnt2 (sih) MO-injected embryos, which also lack circulation.⁶¹ 

Anomalous fusions of arterioles and venules can lead to telangiectases, as reported in patients with HHT.⁷  It is tempting to speculate that our studies in zebrafish are unraveling a molecular basis for telangiectases in patients with HLT carrying mutations in SOX18.⁹ 

Endothelial cells from blood vessels of different caliber or anatomical sites are distinct differentiated cell types with distinct intrinsic gene expression programs,⁶²  exposed to different extrinsic influences.⁶  Consistently, in the sox18/sox7 double morphants, the alteration in the expression of the affected genes is, in general, more pronounced in the trunk/tail region than in the head. Axial vessels, and in particular the PCV and the caudal vein plexus, appear more affected than the ISVs in our ISH analyses. Nevertheless, the most caudal ISVs are specifically altered in the double morphants, as they appear thickened, not lumenized, and show anomalous branching in their dorsalmost aspect. The SIV basket, which is highly enlarged in the double morphants, presents instead thinner vessels than in control larvae. These vessels are more irregularly shaped and present abnormal spikes protruding toward the yolk. Overall, these observations suggest that the role of sox18 and sox7 is specific for different regions of the vascular tree.

In conclusion, we introduce here novel concepts on the role of sox18 and sox7 in endothelial cell differentiation and vascular development. We report that these genes act in concert at very early stages of angioblast differentiation and, most important, control the acquisition of AV identity. The described vascular phenotype, characterized by AV shunts, suggests links to the human HLT syndrome associated with mutations in the SOX18 gene. This supports the use of zebrafish to help deciphering the molecular basis of hereditary vascular diseases.

Lymphedema is the most prominent feature of the human HLT syndrome. It is worth noting that lymphatic endothelial cells derive from venous endothelial cells both in mammals⁶³  and, as recently reported, in zebrafish.^64,65  The double knockdown of zebrafish sox7 and sox18 affects mainly venous endothelial cell differentiation; it will be therefore interesting to address whether they also have a role in lymphatic endothelial cell differentiation.

We would like to thank the following colleagues for probes and reagents: M. Gering, R. Patient, N. Lawson, S. Schulte-Merker, B. Weinstein, S. Sumanas, J. Larson, D. Stainier, C. Wilson, S. Wilson, Z. Wen, K. Crosier, and F. Argenton. We appreciate the help of A. Conti and F. Orsenigo for the analysis of quantitative PCR results, and of S. Rodighiero for confocal analysis. We would like to thank C. Lora-Lamia for teaching us histologic techniques, G. Cappellano for his contribution to the initial phase of this project, G. Brunetti for his help in fish husbandry, S. Nicoli and A. Rissone for helpful suggestions, and M. E. Bianchi and M. Presta for critical reading of the manuscript.

We acknowledge financial support from the Telethon Foundation (GGP04255 to M.B.), Italian Ministry of Education, University, and Research (MIUR-Cofin 2004 to M.B. and Cofin 2006 to E.D.), Fondazione Cariplo (Cariplo 2006 to M.B. and N.O.B.E.L. to F.C.), Leducq Foundation, and Istituto Superiore di Sanità (project on Rare diseases to E.D.). S.M. is the recipient of a PhD fellowship from the Fondazione Fratelli Confalonieri.

Contribution: S. Cermenati and S.M. performed research and analyzed data; S. Cimbro, P.C., L.D.G., and R.A. participated in performing research; P.K. discussed the results and provided helpful suggestions; E.D. and F.C. analyzed data and provided helpful suggestions; and M.B. designed and supervised the research project, analyzed data, and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Monica Beltrame, Dipartimento di Scienze Biomolecolari e Biotecnologie, Università degli Studi di Milano, Via Celoria 26, 20133 Milano, Italy; e-mail: monica.beltrame@unimi.it.