Several of these critical molecular players in vasculogenesis and angiogenesis are members of the receptor tyrosine kinase superfamily, which mediates signal transduction by tyrosine phosphorylation of signaling intermediates and effector molecules. Transduction of signals by tyrosine phosphorylation is counter-regulated by a variety of transmembrane and cytoplasmic protein tyrosine phosphatases (PTPs), a few of which are known to be expressed in ECs. One of these PTPs, CD148/DEP-1/PTPη, causes vascular defects and embryonic lethality when mutated, although its expression is not restricted to ECs (3). The only known PTP whose expression appears to be restricted to vascular ECs in embryos is vascular endothelial PTP (VE-PTP) (4). VE-PTP, the mouse homologue of receptor-type human PTP-β (5), has been shown to associate with and dephosphorylate Tie2 (4) and VE-cadherin (6) but was shown to not interact with VEGF-R2 (4).

To study the expression pattern in embryonic and adult tissues and to elucidate the functional role of VE-PTP in vascular development, we generated mice in which the VE-PTP gene was replaced by LacZ, a high-resolution reporter gene. LacZ expression confirms that VE-PTP is expressed in both arterial and venous vascular endothelium in embryos, although more strongly in arterial vessels. We also show that, early in development, VE-PTP is highly expressed in the outflow tract and later in the valves of the heart. VE-PTP continues to be expressed in adult vascular endothelium and is also expressed in tumor vasculature.

VE-PTP-null mice die embryonically because of a variety of angiogenesis defects in the embryo and the yolk sac. Defects include failure of remodeling of the vascular plexus into large veins and branched vascular networks, which is clear throughout the embryo but is most prominent in the yolk sac. Our findings show that VE-PTP is essential for cardiovascular development, and its continued expression in the adult suggests it may also have a role in vascular homeostasis and may be involved in the pathological angiogenesis of tumors.

Although the genetic deletion in VE-PTP described here is distinct, the vascular phenotype is essentially the same as that described for another VE-PTP mutant published during the preparation of this manuscript (7). Baumer et al. (7) describe a mutation in VE-PTP that resulted in a soluble extracellular form of VE-PTP being secreted, and that may have some neomorphic function (7, 8). However, the symmetry of our phenotype with that of Baumer et al. (7) suggests the phenotypes resulting from both mutations are a consequence of loss of VE-PTP function.

VE-PTP Homozygous Mutant Mice Die Embryonically. VelociGene technology (9) was used to create a targeting vector in which exons 2 (except for the first 26 nucleotides, retaining the endogenous signal peptide) to 3 were replaced with a transmembrane-LacZ reporter gene and a loxP-flanked neomycin selection cassette (Fig. 1A). Correct gene targeting (frequency = 3.5%) in F1H4 (C57BL/6:129 hybrid) ES cell clones was scored by using the loss of native allele assay (data not shown, and see ref. 9). Male chimeric mice (complete transmitters of ES cell-derived sperm) from two independent ES cell clones were bred to C57BL/6 mice to generate F₁ offspring heterozygous for the modified VE-PTP allele (VE-PTP^Lz/+). When F₁ animals were intercrossed to generate F₂ offspring, no pups homozygous for the modified VE-PTP allele (VE-PTP^Lz/Lz) were born out of 25 litters (174 mice; 62% VE-PTP^Lz/+, 37% WT), leading to the suspicion that homozygous deletion of VE-PTP results in embryonic death. Timed matings of VE-PTP^Lz/+ mice were used to generate embryos 8.5–10.5 days postcoitum [embryonic day (E) 8.5–10.5]. Up to E9.5, VE-PTP^Lz/Lz, VE-PTP^Lz/+, and WT littermate embryos were obtained at approximate Mendelian ratios. However, by E9.5, the VE-PTP^Lz/Lz embryos exhibited obvious abnormalities (see below). The VE-PTP^Lz/+ and WT embryos are phenotypically indistinguishable and were both used as normal controls throughout this study.

Fig. 1. Gene targeting and embryo phenotype. (A) A schematic representation of the VE-PTP targeting strategy. An 11.8-kb VE-PTP genomic region along with flanking homology sequence (HBox) was used for bacterial homologous recombination with VelociGene technology (more ...)

To confirm that the loss of VE-PTP alleles results in the loss of VE-PTP mRNA expression and acquisition of LacZ expression, quantitative RT-PCR analysis was performed on E9.5 embryos (Fig. 1B). As expected, WT embryos express VE-PTP, but not LacZ, whereas VE-PTP^Lz/+ embryos express LacZ and have a reduction in VE-PTP mRNA levels. VE-PTP^Lz/Lz embryos express the LacZ reporter gene but lack VE-PTP transcripts.

At E8.5, the VE-PTP^Lz/Lz embryos are not visibly distinguishable from their WT littermates, either by size or appearance (data not shown). However, by E9.5, the VE-PTP^Lz/Lz embryos start dying, and no surviving embryos were recovered beyond E11. Compared with control littermates, VE-PTP^Lz/Lz embryos appear developmentally delayed starting at E9.0 and do not appear to progress beyond the E9.0 to early E9.5 stage, based on several objective criteria, including somite number and hallmarks of forelimb development. As such, they are significantly smaller than control littermates (compare Fig. 1 D with C). VE-PTP^Lz/Lz embryos have pericardial edema (Fig. 1D), which varies in severity. VE-PTP^Lz/Lz embryos collected at E10.0 and E10.5 are mostly dead, as assessed by the lack of heartbeat, or the appearance of necrotic tissue, and can have pronounced pericardial edema (data not shown).

Angiogenic Defects in Embryos. The time of embryonic death, developmental delay, and appearance of pericardial edema observed in the VE-PTP^Lz/Lz embryos are hallmarks of defective cardiovascular development. To evaluate the vascular system of VE-PTP mutant animals, CD-31 [platelet/EC adhesion molecule 1 (PECAM-1)] immunohistochemistry was used to visualize the vasculature in whole mounts and sections of VE-PTP^Lz/Lz and control littermate embryos and yolk sacs. Because we observed an overall developmental retardation in VE-PTP^Lz/Lz embryos, their vascular system was compared with stage-matched control embryos. At E8.5, when the vasculature consists primarily of a simple plexus, there is no obvious difference between VE-PTP^Lz/Lz embryos and their normal littermates (data not shown). However, by E9.5, when the vasculature of control littermates includes clearly defined large veins and networks of branching vessels (Fig. 2A), VE-PTP^Lz/Lz embryos have obvious vascular defects in areas where remodeling should take place (Fig. 2B). In control embryos, the vascular system has undergone extensive remodeling, as seen in the dorsal midbrain, where cephalic veins (Fig. 2C, blue arrowheads) and major branches of the mesencephalic artery (Fig. 2C, red arrowheads) can be readily distinguished. In contrast, the cephalic plexus in VE-PTP^Lz/Lz embryos fails to remodel into a hierarchical branched vascular network and instead remains as a plexus with only minor pruning (compare Fig. 2 D with C). In the lateral aspect of the developing control embryo, the cardinal veins are well defined from the head to the heart (Fig. 2E). In VE-PTP^Lz/Lz embryos (Fig. 2F), there is PECAM-1 staining in the correct location; however, the edges are lacy and not clearly delineated, especially in the more caudal region. Occasionally, the cardinal veins are absent in VE-PTP^Lz/Lz embryos (data not shown). Cross sections of PECAM-1-immunostained E9.0 embryos show that the dorsal aortas in VE-PTP^Lz/Lz embryos are much smaller and have a narrower lumen than in control littermates (compare Fig. 2 H with G), and the cardinal veins are revealed in cross section to be poorly defined and have diffuse PECAM-1 staining, indicating incomplete remodeling.

Fig. 2. Defects in vascular remodeling in VE-PTPLz/Lz embryos. (A–F) Whole-mount PECAM-1-stained E9.5 embryos. VE-PTPLz/+ (A, C, and E) compared with a VE-PTPLz/Lz littermate (B, D, and F). (C and D) A closeup of the head vasculature shows the normal (more ...)

Angiogenic Defects in Yolk Sacs. Beginning at E9.0, homozygous VE-PTP mutants can be visually identified during dissection, based on the unique appearance of their yolk sacs. Compared with control embryos, there are no visible blood-filled vessels, and the yolk sac surface appears dimpled in VE-PTP^Lz/Lz embryos (data not shown). At E9.5, there are still no visible vessels, and the VE-PTP^Lz/Lz yolk sac phenotype is even more obvious, because the surface is now dimpled and wrinkled (compare Fig. 3 B and A) because of a dramatic separation of the endoderm and mesoderm layers resulting from the disproportionate distention of the outer endodermal layer relative to the inner mesodermal layer (compare Fig. 3 F with E). The severity of this yolk sac phenotype is variable, ranging from the endoderm and mesoderm layers still being fairly well connected by multiple adhesion sites, to barely being held together by sparse elongated strands of endoderm. All VE-PTP^Lz/Lz embryos staged E9.0 or older have yolk sac defects, the severity of which was highly correlated with the severity of the vascular remodeling defects observed in the embryo proper.

Fig. 3. Vascular defects in yolk sacs of VE-PTPLz/Lz embryos. (A and B) Freshly dissected E9.5 VE-PTPLz/Lz yolk sacs (B) can be clearly distinguished from VE-PTPLz/+ littermates (A), because of a wrinkled appearance and a lack of visible vessels, which can be (more ...)

To study the morphology of the developing vasculature of the yolk sac in detail, whole-mount PECAM-1 immunostaining was performed. By E9.0, the preformed primitive vascular plexus begins to be remodeled in the yolk sacs of WT and VE-PTP^Lz/+ embryos and by E9.5, the yolk sacs have an organized vascular network of branching vessels (Fig. 3C), which are lined with PECAM-1-positive ECs. Although the initial primitive vascular plexus forms in VE-PTP^Lz/Lz yolk sacs, PECAM-1 immunostaining confirms the lack of any identifiable mature or remodeled blood vessels in VE-PTP^Lz/Lz yolk sacs (Fig. 3D), as observed during dissection. Instead, the vessels of the primitive plexus appear to merge together. These hyperfused vessels continue to expand, to the point of almost completely separating the endoderm and mesoderm layers, except for a few remaining adhesion sites (Fig. 3 D and F). In the VE-PTP^Lz/Lz yolk sacs with the most severe manifestation of these defects, the two layers are completely separated and held together only by sparse PECAM-1-positive strands (data not shown). However, even with this severe morphological defect, the EC layers appear intact, lining the entire surfaces of the mesoderm and endoderm layers of the yolk sac (Fig. 3 D and F), essentially making VE-PTP^Lz/Lz yolk sacs one large endoderm- and mesoderm-enveloped blood vessel.

VE-PTP Is Expressed in Embryonic Arterial and Venous Endothelium. To determine the sites of VE-PTP expression with high resolution, VE-PTP promoter activity was followed in embryos, yolk sacs, and adult tissues of VE-PTP^Lz/+ mice by using the β-gal reporter. Consistent with the observed cardiovascular defects, VE-PTP promoter activity is restricted to the developing cardiovascular system and in particular to the vascular endothelium. A careful analysis of VE-PTP promoter activity in embryos was conducted up to E15.5. At E9.5, the time when the VE-PTP^Lz/Lz cardiovascular defects become apparent, the highest sites of expression are in the aortic sac, outflow tract, umbilical artery, and dorsal aorta (data not shown), with limited or weak expression in smaller-caliber vessels. At E10.5, the strongest expression of the β-gal reporter is still in the endothelium of the heart, including the atrioventricular canal, aortic sac, and outflow tract (Fig. 4 C and D), dorsal aortas (Fig. 4 A and C–E), and umbilical arteries (Fig. 4A), but it is also strongly expressed in the endothelium of vessels of all sizes, including arteries, capillaries, and somewhat more lightly in veins (Fig. 4 A–C). Expression is also observed in the vitelline vessels of the yolk sac (data not shown). At E15.5, β-gal is highly expressed in the heart valves (Fig. 5 E and F) and continues to be expressed in the endothelium of the major arteries and veins and in the arterioles and venules of organs, with expression being higher in arteries (Fig. 4F). This expression in organs correlates with previous results from in situ hybridizations done on organs from E15.5 embryos (4). In the adult, expression is very high in the vasculature of lung, spleen, and kidney, as well as in the heart valves, and is also present in the endothelium of arterioles and venules (data not shown).

Fig. 4. VE-PTP LacZ reporter gene expression in embryos. (A–E) Whole-mount β-gal staining in E10.5 VE-PTPLz/+ embryos. (A) The lines depict the level of the cross sections in D and E. (B) Head vasculature showing reporter expression in arteries (more ...)

Fig. 5. Cardiac defects and reporter gene expression in VE-PTPLz/Lz embryos. (A and B) Freshly dissected E9.5 embryos. VE-PTPLz/Lz embryos have pericardial edema (blue arrow) (B). Although knockout embryos are smaller and developmentally delayed compared with (more ...)

VE-PTP Heart Defects. In addition to pericardial edema, VE-PTP^Lz/Lz embryos have hearts that are relatively enlarged compared with control littermates. At E9.5, although VE-PTP^Lz/Lz embryos are significantly smaller than their control littermates, their hearts are of similar size (compare Fig. 5 B with A). PECAM-1 immunostaining in VE-PTP^Lz/+ hearts shows the developing ventricles, atria, and connecting atrioventricular canal (Fig. 5C). In contrast, VE-PTP^Lz/Lz littermates have limited PECAM-1 staining in the relative position of the developing atria (Fig. 5D, dotted circle), which may be due in part to the VE-PTP^Lz/Lz embryos being developmentally delayed. As previously shown, the outflow tract and the atrioventricular canal is the site of highest expression of the β-gal reporter (Fig. 4C). Interestingly, this expression corresponds to the precursors of the developing cardiac valves, and VE-PTP reporter gene expression continues to remain high in all of the developing heart valves (Fig. 5 E and F) at E15.5. In addition, in adult heart, VE-PTP expression remains high in endothelium covering all of the cardiac valves (data not shown).

VE-PTP Is Expressed in Tumor Vasculature. Subcutaneous implantation of Lewis lung carcinoma cells in adult VE-PTP^Lz/+ mice causes tumor growth and induces angiogenesis. β-Gal staining shows that VE-PTP is expressed in the newly formed tumor blood vessels (Fig. 6 A–C). The tumor invades the overlying skin and coopts the resident vasculature, which leads to the dramatic induction of VE-PTP expression in the coopted vessels (Fig. 6 B, D, and E).

Fig. 6. VE-PTP reporter expression in Lewis lung carcinoma tumors implanted in VE-PTPLz/+ mice. Whole-mount β-gal staining of tumors is shown. (A) Thick cross section of the tumor. Boxes depict the areas taken at higher magnification in B and C. (B) Closeup (more ...)

Deletion of the VE-PTP gene causes severe cardiovascular defects and embryonic lethality by E10. Because the vascular disruption begins at E9.0, after the primary vasculature is established during vasculogenesis, it appears VE-PTP is not necessary for this process. However, it is required for angiogenesis in embryos, and its absence results in lethality due to defective remodeling. This finding is consistent with previous biochemical studies in which VE-PTP was shown to interact with and dephosphorylate Tie2 (which, like VE-PTP, is critical for angiogenesis but not vasculogenesis) but not VEGF-R2 (4), which has been clearly shown to be critical for vasculogenesis, and in fact, the formation of ECs themselves (10) (see below for further discussion).

During development, the vascular plexus, formed during vasculogenesis, undergoes remodeling and expansion by pruning, coalescence, sprouting angiogenesis (11), and intussusception (12). In VE-PTP mutant embryos, the vascular plexus is established, but it does not remodel correctly. Defects include failure to form branched vascular networks (e.g., in the venous plexus of the head), possibly because of improper pruning and intussusception. Also, the cardinal veins, which form by coalescence of smaller vessels, are indiscernible in VE-PTP mutant embryos, except for partial assembly at its most anterior segment. Instead, the portion of the vascular plexus destined to be cardinal veins remains as a collection of vessels that fail to reshape into a cohesive vessel. The dorsal aortas, which formed normally during vasculogenesis, become stenotic in VE-PTP^Lz/Lz embryos. Because the mutant embryos do not have a normal vascular framework, normal blood flow, which is a key component in arterio-venous differentiation (13), is prevented. The VE-PTP^Lz/Lz embryos are also developmentally delayed and have pericardial edema, which are indicators of cardiovascular insufficiency. These two characteristics are commonly seen with other gene deletions that cause midgestation lethality because of cardiovascular defects [e.g., VEGF-R3 (14), Dll4 (15–17), and CD148/DEP-1 (3)]. The hearts of VE-PTP^Lz/Lz embryos also do not develop properly, which may be due in part to inappropriate vascularization and underdevelopment.

The vascular defects in VE-PTP mutants are most striking in the yolk sac, which completely lacks a remodeled vascular network. Instead, the yolk sac vascular plexus essentially becomes one large vessel fully lined with PECAM-1-positive cells. There is a range in the amount of endodermal and mesodermal detachment in different VE-PTP^Lz/Lz yolk sacs. In those that are less detached, the yolk sac is formed of large interconnected channels, and in those more detached, only sparse PECAM-1-positive strands connect the endoderm and mesoderm layers. This variation shows the progression of this vascular defect as a case of unchecked coalescence, where the plexus vessels merge to the point of completely separating the two yolk sac layers, suggesting that VE-PTP plays a role in controlling that remodeling process. Other genes, when deleted, also cause a similar phenotype in yolk sacs (summarized in ref. 18), which has been termed vascular “hyperfusion” (19). Similar hyperfusion defects were described in quail embryos when excessive VEGF was administered, suggesting that this phenomenon can be VEGF driven (19). However, VE-PTP does not appear to play a direct role in the modulation of VEGF-R2, as suggested by prior biochemical data (4) and given the fact that vasculogenesis proceeds normally in VE-PTP mutant embryos. In addition, expression of VE-PTP was low during vasculogenesis and became most highly expressed during angiogenic stages (data not shown and Fig. 4). However, it remains possible that VE-PTP indirectly regulates production of VEGF itself, a possibility not explored in the current study. Thus, the mechanistic role that VE-PTP plays in the hyperfusion process remains unclear.

In embryos, VE-PTP is expressed in both arterial and venous endothelium with the highest apparent levels in arterial vessels and components of the heart. VE-PTP continues to be expressed in adult vascular endothelium, indicating a potential function in vessel maintenance. To explore whether VE-PTP is involved in adult angiogenesis, as occurs during tumor growth, Lewis lung carcinoma cells were xenografted into the flanks of VE-PTP^Lz/+ mice, and VE-PTP expression was examined in the collected tumors. VE-PTP promoter activity, as evaluated by LacZ expression, was clearly present in the new tumor vessels and was elevated in the vasculature of the skin overlying the tumors, as compared with the control skin on the contralateral flank. These results suggest VE-PTP may be important in tumor angiogenesis, in addition to embryonic angiogenesis, and may in fact represent a novel therapeutic target. Conditional deletion or specific inhibition of VE-PTP during tumor angiogenesis will be necessary to test this hypothesis.

The cardiovascular defects described here for VE-PTP are reminiscent of those reported for the CD148/DEP-1/PTPη mutant mouse, the only other known receptor-type phosphatase expressed (although not exclusively) in ECs (3). In addition, a recently published report (7) describes a distinct VE-PTP mutation, in which an insertion is made between exons 17 and 19, and which results in mice in which the majority of the extracellular region of VE-PTP remains intact and is in fact found to be secreted. It was speculated that this secreted form of VE-PTP ectodomain may represent a neomorphic allele with the potential to retain some function or possibly function in an antagonistic form (7, 8). However, despite the differences in knockout strategy, the described exon 17–19 insertion/deletion and our exon 2–3 deletion mutants appear to have essentially the same spectrum of vascular phenotypes, suggesting that both represent loss of function rather than neomorphic phenotypes.

Tie2 (4) and VE-cadherin (6) have been shown to associate with VE-PTP and as such should be considered, along with any other substrates, as having a role in the mechanism that causes the observed vascular defects. The VE-PTP mutants share many common characteristics with the Tie2 and Ang1 loss-of-function phenotype (20–22). However, deletion of a phosphatase that negatively regulates this signaling pathway would be presumed to have a gain-of-function phenotype rather than to mimic the loss of function of Tie2 or Ang1. Thus, it could be that perturbation of Tie2 signaling in either direction would lead to similar lethal consequences. In allantois explant studies done with ECs from various E8.5 homozygous mutants, it was found that Tie2^−/− had normal networks of endothelial cords compared with VE-PTP^−/− (7). Further biochemical analysis of the signaling of Tie2 and other VE-PTP targets in VE-PTP-null ECs, as well as the generation of a temporally controlled VE-PTP deletion later in development or in the adult vasculature, will be necessary to further elucidate many of these issues.

Targeting the VE-PTP Gene in Mice. Using VelociGene technology (9), an 11.8-kb segment of the VE-PTP gene extending from 27 nucleotides (in exon 2) through exon 3 was replaced with a β-gal reporter gene and a loxP-flanked neomycin selection cassette (pTM-Zen6-loxP). Briefly, a BAC containing the 11.8-kb VE-PTP region to be deleted and 195 kb of flanking sequences (clone 386g24 from a 129/SvJ BAC library from Incyte Genomics, Wilmington, DE) was used to generate a BAC-based targeting vector for replacement of the VE-PTP gene in F1H4 (C57BL/6:129 hybrid) mouse ES cells. Two independent correctly targeted ES cell clones were selected by using the loss of native allele (LONA) assay [as described (9)] and used to generate chimeric male mice. Germ-line transmitting chimeras were bred to C57BL/6 females to generate offspring heterozygous for the modified VE-PTP allele. Timed matings of VE-PTP^Lz/+ mice and VelociMouse technology (23) were used to generate embryos. The embryos were staged and checked for beating hearts when collected, and the yolk sacs and embryos were photographed during dissection. Yolk sacs and placentas were also collected for analysis. Embryos were genotyped by using the LONA assay. Animals were cared for under the terms and conditions of our Institutional Animal Care and Use Committee.

Quantitative RT-PCR Analysis of VE-PTP and lacZ mRNA Levels. Quantitative RT-PCR (TaqMan) was performed as described (24) on an Applied Biosystems (Foster City, CA) 7900HT real-time PCR system. The results are expressed as the ratio of the amount of the RNA of interest to the amount of control RNA (GAPDH) (25) by using specific primers and probes as follows: primers VE-PTP (Ex2–3)-278F, TCAGGATTGTTTCTCTGGATGGA and VE-PTP (Ex2–3)-385R GGGTGGTTGATGCTGTTTTCTC, and probe VE-PTP (Ex2–3)-326T CAGATCCGTTGCCTCC; primers LacZ-33F GGAGTGCGATCTTCCTGAGG and LacZ-102R CGCATCGTAACCGTGCATC and probe LacZ-54T CGATACTGTCGTCGTCCCCTCAAACTG. The analysis was done on RNA extracted from E9.5 embryos (three WT, three VE-PTP^Lz/+, and four VE-PTP^Lz/Lz).

Immunostaining and Reporter Detection. Whole-mount embryos, yolk sacs, and placentas, as well as sections of embryos and adult tissues, were immunostained with PECAM-1 antibody to detect vascular endothelium as described (26). Whole-mount embryos, yolk sacs, placentas, and adult tissues, as well as sections of embryos and adult tissues, were stained for lacZ as described (27). Paraffin sections of PECAM-1-stained yolk sacs were counterstained with hematoxylin.

Tumor Implantations. Lewis lung carcinoma cells (American Type Culture Collection, Manassas, VA) were implanted s.c. into the flank of VE-PTP^Lz/+ mice, and the tumors were collected after 13 days. To visualize reporter expression, whole-mount β-gal staining was done on the tumors and on skin adjacent to the tumor, as well as from the contra-lateral side as described (27).

We thank other Regeneron team members, including Stanley Wiegand, John Rudge, David Frendewey, Thomas M. DeChiara, Aris N. Economides, Yingzi Xue, Wojtek Auerbach, William Poueymirou, Joyce McClain, Sandra Coetzee, and Nicholas Papadopoulos, for assistance in generating the VelociGene ES cells and mice, technical assistance, and many helpful discussions. We also acknowledge the Regeneron executive management and in particular Leonard Schleifer for his ongoing support of this work.