Tie2Cre-mediated inactivation of plexinD1 results in congenital heart, vascular and skeletal defects

Abstract

PlexinD1 is a membrane-bound receptor that mediates signals derived from class 3 secreted semaphorins. Although semaphorin signaling in axon guidance in the nervous system has been extensively studied, functions outside the nervous system including important roles in vascular patterning have also been demonstrated. Inactivation of plexinD1 leads to neo-natal lethality, structural defects of the cardiac outflow tract, peripheral vascular abnormalities, and axial skeletal morphogenesis defects. PlexinD1 is expressed by vascular endothelial cells, but additional domains of expression have also been demonstrated including in lymphocytes, osteoblasts, neural crest and the central nervous system. Hence, the cell-type specific functions of plexinD1 have remained unclear. Here, we describe the results of tissue-specific gene inactivation of plexinD1 in Tie2 expressing precursors, which recapitulates the null phenotype with respect to congenital heart, vascular, and skeletal abnormalities resulting in neonatal lethality. Interestingly, these mutants also have myocardial defects not previously reported. In addition, we demonstrate functions for plexinD1 in post-natal retinal vasculogenesis and adult angiogenesis through the use of inducible cre-mediated deletion. These results demonstrate an important role for PlexinD1 in embryonic and adult vasculature.

Introduction

In the central nervous system semaphorins promote axon guidance by mediating both attractive and repulsive cues (Dickson, 2002; Raper, 2000; Tessier-Lavigne, 2002; Tessier-Lavigne and Goodman, 1996; Tran et al., 2007). There are five vertebrate classes of semaphorins (classes 3–7) which can be either membrane bound or secreted (Fujisawa and Kitsukawa, 1998), and a wide variety of mammalian family members, domains of expression and receptor binding specificities contribute to the complexity of neuronal patterning and development. Most semaphorins bind to plexin receptors that transduce signals, and some class 3 secreted semaphorins bind to receptor complexes composed of plexin and neuropilin subunits (Tamagnone and Comoglio, 2000).

In recent years, functions for semaphorins and their receptors have been identified in the developing vasculature that are in many ways analogous to those described in the central nervous system (Autiero et al., 2005; Carmeliet and Tessier-Lavigne, 2005; Eichmann et al., 2005; Suchting et al., 2006). For example, neuropilins are expressed by vascular endothelial cells, where they heterodimerize with tyrosine kinase receptor subunits to compose receptors for heparin-binding isoforms of vascular endothelial growth factor (VEGF) (Neufeld et al., 2002a; Neufeld et al., 2002b). Semaphorins can modulate the migratory and growth characteristics of cultured endothelial cells, in part by competing with VEGF for neuropilin-containing receptors, but also by binding to receptors that do not recognize VEGF (Banu et al., 2006; Basile et al., 2005; Catalano et al., 2004; Gu et al., 2005; Guttmann-Raviv et al., 2007; Toyofuku et al., 2007). For example, Sema3E can function as a regulator of angiogenesis and vascular pathfinding in vivo by binding directly to plexinD1 on endothelial cells (Gu et al., 2005). PlexinD1 can mediate repulsive guidance cues in the vasculature, thus contributing to the formation of the complex but largely predictable vascular network that accompanies embryonic development (Carmeliet and Tessier-Lavigne, 2005; Eichmann et al., 2005). Other families of axon guidance molecules have also been implicated in vascular patterning, including the Slit/Robo, Ephrin/Eph and Netrin/Unc5 networks, suggesting significant parallels between the neuronal and vascular systems (Autiero et al., 2005; Carmeliet and Tessier-Lavigne, 2005; Eichmann et al., 2005; Suchting et al., 2006).

The discovery that semaphorin, plexin and neuropilin signaling affect vascular growth has important implications for the development of novel pro- or anti-angiogenic therapies. For example, Sema3F displays anti-angiogenic activities in animal models of cancer (Kessler et al., 2004), although the receptor complex that mediates this effect is unclear. PlexinD1, which can bind to class 3 semaphorins, is up-regulated in the tumor vasculature of some mouse models of cancer and can be targeted with intravenous injection of monoclonal antibodies directed against PlexinD1 (Roodink et al., 2005). Semaphorins also have direct effects on the migration and proliferation of some tumor cells (Guttmann-Raviv et al., 2007; Neufeld et al., 2005). Therefore, tissue-specific gene inactivation of semaphorins and their receptors will be useful to further delineate therapeutic targets in various cancer models.

We (Gitler et al., 2004b; Torres-Vazquez et al., 2004) and others (Gu et al., 2005; van der Zwaag et al., 2002) have previously shown that plexinD1 is expressed in the developing vasculature. In zebrafish, expression is similar to that of fli1 in the endothelial compartment, and mutations in zplexinD1 results in the out of bounds phenotype in which intersomitic blood vessels migrate inappropriately into domains where class 3 semaphorins are expressed (Torres-Vazquez et al., 2004). In mice, plexinD1 expression is prominent in developing endothelial cells throughout the vasculature, but expression is also apparent in the central nervous system (Gu et al., 2005; van der Zwaag et al., 2002), in the salivary gland (Chung et al., 2007), neural crest (Toyofuku et al., 2008) and in bone (Kanda et al., 2007). Analysis of the GEO database (Gene Expression Omnibus, http://www.ncbi.nlm.nih.gov/geo/) suggests additional sites of expression including lymphocytes.

Homozygous deficiency of plexinD1 in mice leads to neo-natal lethality with a high penetrance of a severe form of congenital heart disease (Gitler et al., 2004b) in addition to more subtle peripheral vascular patterning defects (Gitler et al., 2004b; Gu et al., 2005). The congenital heart defects involve the outflow tract of the heart, which normally arises as a single tube, the truncus arteriosus, and later septates into two major vessels, the aorta and the pulmonary artery. In plexinD1 mutants, this septation process fails to occur and mice are born with a persistent truncus arteriosus and other patterning defects of the aortic arch arteries (Gitler et al., 2004b). Septation of the cardiac outflow tract is known to be critically dependent upon cardiac neural crest cells, which migrate from the dorsal neural tube and contribute to the definitive aortico-pulmonary septum and the tunica media of the great vessels (Engleka et al., 2005; Epstein et al., 2000; Gitler et al., 2002; Kirby et al., 1983). Interestingly, inactivation of Sema3C, a class 3 semaphorin that is able to bind to PlexinD1, also results in cardiac outflow tract defects (Brown et al., 2001; Feiner et al., 2001). PlexinD1 expression in neural crest has been reported (Toyofuku et al., 2008) and a functional role for plexinD1 in neural crest precursors in addition to a potential role in endothelium could have explained the congenital cardiac defects present in the null mice. PlexinD1 deficient mice also display axial skeletal abnormalities involving the vertebral bodies, ribs and cartilage, and plexinD1 is expressed by osteoblasts (Kanda et al., 2007).

In order to determine the tissue-specific functions of plexinD1 that account for the phenotypes that have been described in homozygous null mice, we have created a new allele of plexinD1 in which critical exons are flanked by loxP sites allowing for cre-mediated tissue- and temporal-specific inactivation. We show that loss of function of plexinD1 in neural crest is well tolerated. In contrast, loss of plexinD1 function in cells derived from Tie2-expressing precursors recapitulates the congenital heart, vascular and skeletal defects observed after ubiquitous loss of function. Furthermore, inducible deletion after birth demonstrates functions for plexinD1 in postnatal retinal vasculature.

Results

PlexinD1 is expressed in the vasculature and other tissues

We examined the expression domain of plexinD1 RNA by performing in situ hybridization analysis on E9.5, E12.5, E14.5, E16.5 and E18.5 wild type embryos. At early embryonic stages from E9.5 to E12.5, plexinD1 is expressed in the endothelium and central nervous system (CNS) as has been previously published (Gitler et al., 2004b; Gu et al., 2005; van der Zwaag et al., 2002). From E14.5 to E18.5, plexinD1 expression continues to be robust in the endocardium and vascular endothelium (Figure 1A), and also becomes evident in the brain (forebrain, trigeminal ganglion and choroids plexus, Figure 1C), dorsal root ganglion, adrenal gland, kidney, lung mesenchyme, ossification center of the bones and small intestine (Figure 1D,G,J, Figure 2F). To verify the tissue types where the expression was observed, we stained adjacent sections for expression of von Willebrand factor (vWF), which is specific for vascular endothelial cells, the neuronal tissue marker neurofilament (2H3), the adrenergic neuron marker tyrosine hydroxylase (TH) and the in lung mesenchyme marker wnt2. The previously unappreciated broad expression pattern during late gestation of plexinD1 prompted us to re-consider the cause of the perinatal death of plexinD1 null mice and to rigorously examine the cell-autonomous role for plexinD1 in endothelium.

Figure 1. PlexinD1 is broadly expressed in late embryonic development and is required for Sema3A-mediated endothelial cell migration in vitro.

PlexinD1 mRNA was detected by in situ hybridization (red signal) in the heart and vascular endothelium (A, arrow), forebrain (C), dorsal root ganglia and adrenal gland (D, arrow), lung mesenchyme (G), and the ossification centers of vertebral bodies (J, arrow). Adjacent sections were stained for vWF (B, H), neurofilament (2H3) (E) and tyrosine hydroxylase (F) by immunohistochemistry and wnt2 by in situ hybridization (I). fb, fore brain; gV, trigeminal ganglion; cp, choroid plexus; ad, adrenal gland; drg, dorsal root ganglion; oc, ossification center. Scale bars equal to 500 μm in panel C and 100 μm in other panels. (K) Boyden chamber assays reveal the motility of primary endothelial cells (PEC) from plexinD1^−/− mouse is impaired as determined by calcein AM emission (see methods). PECs from plexinD1^+/− (het, dotted line) and plexinD1^−/− (ko, solid line) pups were assessed at basal condition (Ctrl), and when soluble Sema3A was added to the upper chamber (3A). (L) Aortic ring assays reveal reduced responsiveness of plexinD1^−/− aortic explants to Sema3A. The explants were incubated in complete medium (green) or complete medium plus Sema3A (red). Results are expressed as absolute outgrowth in mm: 1.26±0.19 mm (in control medium), and 1.02±0.18 mm (in SEMA3A conditional medium) from wild type and plexinD1^+/− explants (n=10). From plexinD1^−/− explants (n=7), the absolute outgrowth is 1.17±0.13 mm (in control medium) and 1.23±0.17 mm (in SEMA3A conditional medium). Representative images of aortic ring explants are shown in Supplemental Figure 1. n.s., not significant.

Figure 2. Targeted disruption of plexinD1.

(A) A schematic illustration of the plexinD1 locus, targeting vector, and the targeted locus after Flp-mediated recombination. (B) The 5′ external probe as shown in (A) was used for Southern blot analysis to detect heterozygous recombinants using BglII-digested genomic DNA. (C) PCR of tail DNA from plexinD1^flox/+ mice confirms germline transmission. The 580bp band represents the wild-type allele, while the 620bp band indicates the targeted allele. (D–I) Absence of plexinD1 mRNA in the vascular endothelium of E12.5 D1^ECKO embryos (G,H,I) is demonstrated by in situ hybridization, while plexinD1 transcripts are detected in other tissues including trigeminal ganglion (arrow) and the smooth muscle of the small intestine (double arrow). The D1^ECKO embryo has persistent truncus arteriosus (D). A, aorta. P, pulmonary artery. TA, truncus arteriosus.

PlexinD1 modulates primary endothelial cell migration in vitro

We performed Boyden chamber migration assays to study the motility of plexinD1-deficient primary endothelial cells. Endothelial cells were harvested from plexinD1^+/− and plexinD1^−/− P0 neonates using CD31-antibody selection and seeded in the upper chamber.

After various times of incubation, the number of the cells that migrated through the membrane was determined in the presence or absence of Sema3A. PlexinD1^−/− cells migrate less robustly than control cells either in the presence of growth factors, and they failed to respond to Sema3A (Figure 1K).

We also performed aortic ring assays to investigate endothelial cell migration using control and plexinD1^−/− aortic explants. In this assay, endothelial cells migrate from the aortic explants into the gel matrix and form vascular branches (Masson et al., 2002) (Figure S1). We quantified the absolute outgrowth distances from control and plexinD1^−/− explants in the absence or presence of Sema3A. Outgrowth distance was significantly inhibited by Sema3A when control explants were used, but Sema3A did not affect plexinD1^−/− explants (Figure 1L). Thus, endothelial cells lacking PlexinD1 display abnormal migratory characteristics consistent with a cell-autonomous role for PlexinD1 in endothelial cells.

Inactivation of plexinD1 in Tie2-derivatives results in cardiovascular defects

To delete plexinD1 in a tissue-specific manner, we generated a floxed allele of plexinD1 in the mouse. The first exon of plexinD1, containing the translational start codon and part of the SEMA domain, was flanked by loxP sites. Correctly targeted ES cells were identified by Southern blot analysis (Figure 2A,B) and used to create chimeric mice and germline heterozygotes after removal of the frt-flanked PGK-neo cassette. Germline transmission and genotyping was confirmed by PCR (Figure 2C).

We crossed plexinD1^flox/+ mice with Tie2-cre mice in order to inactivate plexinD1 in the endothelial compartment. (Hereafter, we refer to Tie2-cre; plexinD1^flox/− and Tie2-cre; plexinD1^flox/flox mice as D1^ECKO mice). Tie2-cre mice have been extensively characterized and efficiently mediate endothelial recombination in appropriate reporter mice beginning at E9.5 (Constien et al., 2001; de Lange et al., 2004; Kisanuki et al., 2001). In situ hybridization with a plexinD1-specific riboprobe using E12.5 embryo sections showed complete deletion of plexinD1 in the vascular endothelium in D1^ECKO mice compared to controls, while plexinD1 mRNA expression remained evident non-endothelial cells (Figure 2D–I).

Genotype data from Tie2-cre, plexinD1^+/− X plexinD1^flox/+ matings are shown in Table 1. Notably, we were unable to identify D1^ECKO mice after the first day of life, while other genotypes were found at close to expected ratios. We performed more matings than shown in Table 1 in order to produce embryos for all the studies shown and for each experiment, N is greater to or equal to 4. At birth (P0) D1^ECKO pups were identified, but nearly uniformly appeared cyanotic and succumbed by P1. Mutant pups had shorter stature and tail length (Figure 3A). Subcutaneous hemorrhage, especially in the head, was routinely apparent and allowed for prompt identification of mutant pups (Figure 3B). D1^ECKO mice displayed severe cardiac defects. At birth, wild type hearts are characterized by a fully septated outflow tract with a distinct aorta and pulmonary artery (Figure 3C). D1^ECKO mutant hearts showed persistent truncus arteriosus in the majority of cases (Figure 3D). In some cases, partial septation was evident, although the outflow tract was malrotated and a right-sided aortic arch was sometimes present (Figure 3E, F). The coronary arteries, which normally originate from the root of the aorta, were instead noted to originate more distally from the ascending portion of the truncus arteriosus, or from the ascending aorta (when partial or compete septation was present) in all mutant hearts examined (N=5) (Figure 3D–F), and multiple small hemorrhages were apparent on the ventricular walls. The atria were markedly dilated and abnormal with a “raspberry-like” appearance (Figure 3D,E). Histological sections of mutant hearts revealed ventricular septal defects and abnormally condensed and discontinuous atrial myocardial architecture when compared to controls (Figure 3G,H).

Table 1.

Genotypes resulting from Tie2-cre, plexinD1^+/− X plexinD1^flox/+ matings

	E11.5–E18.5			P0–P1			P2–P4
	Obsa	% Obsb	%Expc	Obs	%Obs	%Exp	Obs	%Obs	%Exp
plexinD1+/+	8	16	12.5	17	15	12.5	9	12	12.5
plexinD1flox/+	6	12		13	11		14	18
plexinD1+/−	8	16		10	9		13	17
plexinD1flox/−	3	6		15	13		13	17
Tie2cre, plexinD1+/+	9	18		14	12		4	5
Tie2cre, plexinD1flox/+	7	14		19	16		15	19
Tie2cre, plexinD1+/−	4	8		14	12		10	13
Tie2cre, plexinD1flox/−	6	12		13	11		0	0
Total	51			115			78

^aThe number of mice observed (Obs) for each genotype is recorded.

^bThe percent of mice observed (%Obs) for each genotype is recorded.

^cThe percent of mice expected (%Exp) if Mendelian ratios were obtained is indicated.

Figure 3.

D1^ECKO mice were born alive but died within 2 days and displayed cardiac and vasculature defects.

(A) P0 D1^ECKO pups have shorter statues and tails than wild type litter mates (arrows). (B) Hemorrhage is present in the head region of D1^ECKO at P0 (arrows). (C–F) Hearts from wild type (C) and D1^ECKO (D–F) P0 pups reveal persistent truncus arteriosus (TA, D) in some animals, while others show septation with abnormal patterning of the aorta (Ao) and pulmonary artery (P) (E,F). All mutants pups had abnormal coronary origins (arrows, D,E) and hemorrhages on the ventricular walls (arrow heads, D, E). (G,H) H&E staining of P0 heart sections shows condensed and discontinuous atrial myocardium which was separated from the epicardium in mutants (arrow, H). Some mutant hearts also displayed ventricular septal defects (arrowhead, H). (I, J) Whole-mount PECAM staining of E11.5 wild type and D1^ECKO embryos. LV, left ventricle; RV, right ventricle; A, atrium.

These results suggest that PlexinD1 function in endothelial cells accounts for cardiac defects in plexinD1 null mice. However, cardiac outlflow tract defects, including truncus arteriosus, are commonly associated with neural crest abnormalities (Stoller and Epstein, 2005), and plexinD1 expression in cardiac neural crest has been reported (Toyofuku et al., 2008). Nevertheless, we found that neural crest-specific deletion of plexinD1 using Wnt1-cre resulted in normal appearing and viable mice without congenital cardiac defects such as those seen in plexinD1 null pups. We obtained 241 P0–P3 pups from Wnt1-cre, plexinD1^flox/+ X plexinD1^+/− matings, of which 37 (15%) were Wnt1-cre, plexinD1^flox/− (compared to 12.5% expected) and outflow tract septation was normal (Figure S2).

In addition to structural heart, coronary and atrial defects, D1^ECKO embryos also displayed peripheral vascular defects. Whole-mount PECAM antibody staining of E11.5 embryos revealed immature intersomitic vessels with poorly remodeled vascular plexus (Figure 3I,J). At this time point, aortic outflow and atrial defects were already evident by routine histology (Figure 4A,B). The atrial myocardium of the D1^ECKO hearts was condensed, discontinuous and separated from the epicardium (Figure 4C,D). The ventricular myocardial wall was abnormal and unusual epicardial blood islands were present (Figure 4E,F). We used Foxp4 and MF20 as epicardial and myocardial markers, respectively, at this embryonic stage. Foxp4 has been shown to express in the mouse epicardium from E9.5 to E12.5 (Li et al., 2004) and MF20 recognizes myosin. Coronary vessel defects were evident by E12.5 as detected by whole-mount PECAM staining (Figure 4G–J). Interestingly, the mutant conotruncus was surrounded by PECAM positive cells (arrow, Figure 4H), which are normally not present in this location at this developmental stage, as shown in the wild type control (Figure 4G). It is interesting to note that Sema3C is expressed by outflow tract myocardium at this stage (Brown et al., 2001). Abnormal clusters of PECAM-positive cells were apparent on the ventricular surface of the mutant hearts (Figure 4J), consistent with abnormal coronary vessel formation.

Figure 4. Cardiovascular defects in D1^ECKO embryos are evident at mid-gestation.

(A, B) H&E staining of cross sections from control and D1^ECKO E11.5 hearts shows condensed and discontinuous atrial myocardium in the mutant (B, arrow), which also displays truncus arteriosus (TA), while the aorta (Ao) and pulmonary artery (P) are separated in the Tie2-cre control heart (A). (C, D) Sections adjacent to those shown in (A,B) were stained for Foxp4 (green) and MF20 (red), markers of epicardium and myocardium, respectively. (E, F) H&E stained sections of ventricular regions show abnormal blood islands in D1^ECKO embryos (arrows) and abnormal thickened regions of compact myocardium (brackets). (G, H) Whole-mount PECAM staining of E12.5 wild type (WT) and D1^ECKO hearts shows coronary vessel plexus defects in mutants (H). The base of the outflow trunk in the D1^ECKO embryo is covered with PECAM-positive cells (arrow, H), while the corresponding region of the wild type is not (arrow, G). (I,J) Higher magnification of the boxes in (G) and (H), showing large clusters of PECAM positive cells are apparent in the ventricular surface of the D1^ECKO heart (arrowheads, J). Scale bars equal to 50 μm in A, B, E, F and 100 μm in C, D.

Tie2-cre inactivation of plexinD1 results in skeletal defects

Previous research has shown that plexinD1^−/− mice have skeletal defects (Kanda et al., 2007). To determine if the skeletal abnormalities are due to a primary bone defect or are secondary to vascular defects, we analyzed bone formation in the D1^ECKO mice. Interestingly, D1^ECKO mice display skeletal defects indistinguishable from those seen in plexinD1^−/− mice at birth, as revealed by Alizarin Red S and Alcian Blue staining. In both plexinD1^−/− and D1^ECKO P0 pups, the vertebral bodies were poorly formed and frequently fused or split in both thoracic (Figure 5A–C) and lumbar regions (Figure 5D–F). Staining for vWF to identify endothelial cells revealed abnormal microvasculature within bones derived from plexinD1^−/− and D1^ECKO mice when compared to controls, though large vessels attached to the long bones appeared intact (Figure 5G–I). Surprisingly, one D1^ECKO mouse survived to adulthood, perhaps because of relatively inefficient cre-mediated recombination in this case. We performed μ-CT imaging to examine skeletal structures, which revealed multiple vertebral body fusions along the entire anterior-posterior axis (Figure 5J–K).

Figure 5. Tie2-cre inactivation of plexinD1 results in skeletal defects.

(A–F) Alizarin Red S and Alcian Blue Staining of P0 pup thoracic (A–C) and lumbar (D–F) skeletons. Both plexinD1^−/− and D1^ECKO pups revealed axial skeletal malformations, including fusions and splitting (circles and arrows, B,C,E,F) of the vertebral bodies. (G–I) vWF (green) staining of wild type, plexinD1^−/− and D1^ECKO P0 sections of bone reveal fewer vWF-positive microvascular cells in both plexinD1^−/− (H) and D1^ECKO (I) samples compared to control (arrows, G). Insets show intact vWF staining from the large epichondrial vessels invading adjacent bone in all three animals. (J, K) Control (J) and D1^ECKO (K) 3-dimensional reconstruction from μ-CT imaging in which structures are viewed from the left, posterior aspect reveal fusion of multiple vertebrae in the D1^ECKO animal (green arrows, K) that are not seen in the control (J).

We sought to determine if skeletal defects were related to activity of Tie2-cre in bone lineages as opposed to vascular endothelial cells. We harvested primary osteoblasts from plexinD1^flox/− and D1^ECKO P0 pups and examined plexinD1 mRNA levels by quantitative RT-PCR. No significant loss of plexinD1 mRNA was detected in the osteoblast culture from D1^ECKO samples (data not shown). Lineage tracing of Tie2-cre derivatives in Tie2-cre, Z/EG embryos shows that Tie2-cre derivatives are not detectable in bony structures at E14.5, prior to vascularization (Figure S3A,B). These results are consistent with prior data which indicate that the Tie2 promoter is not active in osteoblasts (Winslow et al., 2006). However, Tie2-cre is active in osteoclast progenitors (Wan et al., 2007) which first invade developing bone at E15.5. At this time point, we already detected significant skeletal abnormalities, including rib fusions (Figure S3C,D). Therefore, we believe that skeletal defects in plexinD1 null mice are most likely secondary to vascular abnormalities and are unlikely to be related to loss of plexinD1 in bone lineages, although a role for plexinD1 in osteoclasts cannot be completely ruled out.

Post-natal inducible deletion of plexinD1 produces post-natal defects of angiogenesis

In order to bypass neo-natal lethality and to study the function of PlexinD1 in post-natal mice, we crossed plexinD1^flox/+ mice with tamoxifen-inducible CMV-creER mice (Hayashi and McMahon, 2002) to produce CMV-creER, plexinD1^flox/flox newborns. (We henceforth refer to CMV-creER, plexinD1^flox/flox and CMV-creER, plexinD1^flox/− mice as D1^CMV-ERKO). Nuclear translocation and subsequent cre recombinase activity was induced by intraperitoneal injection of tamoxifen on each of 2 successive days, and 92% loss of plexinD1 mRNA expression was confirmed by quantitative RT-PCR (Figure 6A). Tamoxifen was injected on P3 and P4 and retinal blood vessel development was assessed by FITC-dextran injection and confocal microscopy at P10 (Figure 6B–I). Loss of PlexinD1 resulted in severe retinal vasculature defects, including lack of retinal vessel development in the deeper layers of the retina (Figure 6C), and vessel leakage and hemorrhage (Figure 6D–I). We have not identified other postnatal angiogenic defects in PlexinD1 knockout mice to date.

Figure 6. PlexinD1 functions in postnatal angiogenesis.

(A) Quantitative PCR of retina shows effective deletion of plexinD1 after tamoxifen treatment in D1^CMV-ERKO mice compared to control (CTRL). The relative expression of PlexinD1 is 0.08±0.04 fold of the control. (n=3) (B) Confocal images pseudocolored to indicate depth of vascular structures in retinal flatmounts from a control P10 pup reveals 2 layers of retinal vessels (green = top layer, 24 micron depth; red = bottom layer, 28.5 micron depth). (C) Confocal image of a D1^CMV-ERKO P10 pup shows only one layer of vessels on the surface of retina with fewer deep vessels. (D–F) FITC-dextran perfusion of control (D) and mutant (E,F) retina samples reveals vascular leakage in mutants both from large vessels (E) and from smaller vascular plexi at the periphery of the retina (F). scale bar equals to 50 μm. (G–I) Confocal images of control (G) and mutant (H,I) retina flat mounts show hemorrhage on the retina surface (H), and in the deep layer (I) of mutants.

In a separate set of experiments, tamoxifen or vehicle was administered at 6 weeks of age and matrigel plugs were implanted subcutaneously. After 7 days, the animals were sacrificed and vascular infiltration of the matrigel plugs was assessed by Masson’s trichrome staining (Figure S4A). Deletion of plexinD1 resulted in a marked decrease in the number and extent of endothelial cell migration into the plugs. Under basal conditions when the growth-factor reduced matrigel alone was implanted, slightly fewer numbers of cells invaded the plug in the D1^CMV-ERKO mice compared to the control. When endothelial cell growth factors (ECGF) were added to the matrigel plug, there was a significant increase in cell infiltration in the control animals, but in the D1^CMV-ERKO animals, the cell numbers remained at basal level, suggesting a significant defect in post-natal angiogenesis (Figure S4B).

Discussion

Although plexinD1 expression is not restricted to endothelium, inactivation of plexinD1 in Tie2-expressing cells is sufficient to recapitulate the phenotypes previously observed in plexinD1 null pups including perinatal lethality, congenital heart, coronary vessel, peripheral vasculature and skeletal defects. Additionally, inducible deletion of plexinD1 in post-natal mice produces angiogenic defects in the retina and in matrigel plug assays.

The most likely explanation for the phenotypes that we describe in D1^ECKO mice is that plexinD1 function in endothelial cells is required for cardiovascular development and neonatal viability, and that this activity is removed by Tie2-cre mediated inactivation. However, because Tie2-cre also mediates recombination in hematopoetic precursors (Gitler et al., 2004a; Schlaeger et al., 2005), we cannot formally exclude a critical function for plexinD1 in blood. Alternative endothelial cre mice exist, such as VE-cadherin-cre (Alva et al., 2006; Monvoisin et al., 2006), but in our experience efficiency has been far less robust than with Tie2-cre during development. Furthermore, we have not observed plexinD1 expression in blood lineages in the embryo. Hence, we interpret our results to strongly implicate loss of plexinD1 function in endothelium as the cause of neo-natal lethality and pathology in D1^ECKO mutants.

Interestingly, D1^ECKO mouse embryos also have myocardial defects not previously described. Myocardial development is known to depend upon signals from endocardium. For example, neuregulin-1 is expressed exclusively in the endocardium and provides paracrine signals via its receptors ErbB2 and ErbB4, which are expressed by myocardial cells, to orchestrate proper ventricle trabeculation (Meyer et al., 1997). Similarly, depleting myocardial angiopoietin-1 resulted in defects both in the endocardial monolayer and adjacent myocardial trabeculation (Suri et al., 1996). Our data suggest that loss of plexinD1 in endocardium affects myocardial development, perhaps by affecting cell-cell contacts or paracrine substances produced by endothelial cells that modulate myocardial development. Future studies will address the molecular nature of these signals.

The observation of skeletal defects in D1^ECKO mice was surprising, given the prior description of plexinD1 expression in osteoblasts and the suggestion that this might account for the skeletal abnormalities (Kanda et al., 2007). We confirmed that Tie2-cre was not active in osteoblast precursors and that osteoblast expression of plexinD1 was retained in D1^ECKO pups. Tie2-cre is active in precursors of osteoclasts, which function in bone resorption (Wan et al., 2007). Osteoclasts are multinucleated cells that enter the bone through blood vessels, first appearing in skeletal structures at E15.5 (Kahn and Simmons, 1975; Manolagas and Jilka, 1995). Osteoclast defects result in osteoporosis (when there is too much osteoclast activity), and osteopetrosis (when osteoclast activity is deficient) (Helfrich, 2003; Tondravi et al., 1997). We are unaware of any data that would link osteoclast abnormalities to vertebral body morphology defects such as those in D1^ECKO mice. Furthermore, we have detected bone defects prior to osteoclast mediated bone resorption (Figure S3). Also, prior studies have indicated that bone mineral density of plexinD1 knockout newborns is not altered (Kanda et al., 2007), suggesting that osteoclast activity is unaffected. Thus, it is likely that endothelial-related vascular defects account for the skeletal abnormalities that we observe, and this is consistent with alterations in vWF expression and microvascular patterning in bone of mutant mice. These axial patterning defects, characterized by hemivertebrae and vertebral fusions, are reminiscent of spondylothoracic and spondylocostal dysostosis (SCDO1) in human patients which can be caused by mutations in the Notch ligand encoded by DLL3 (Bulman et al., 2000). The Notch-Delta pathway plays an important role in the “segmentation clock” that patterns the presomitic mesoderm and forming somites which give rise to skeletal derivatives (Dunwoodie et al., 2002). Therefore, we examined expression of a variety of Notch signaling components and early somite and sclerotome patterning, but did not observe differences in D1^ECKO embryos compared to controls (Figure S3E,F and data not shown). We did note abnormalities of intersomitic vascular patterning as early as E11.5 (Figure 3I,J) and subtle defects of somite polarity, morphology or gene expression associated with these vascular patterning defects that were below our level of detection cannot be ruled out, and could be significant contributors to subsequent skeletal abnormalities.

Vascular invasion of the developing skeleton occurs during mid-gestation. At E14.5, the cartilage primordia of the endochondral skeleton are surrounded by blood vessels in wild type embryos. By E15.5, blood vessels have fully penetrated into the primary ossification centers, together with chondroclast/osteoclast recruitment. VEGF is robustly expressed in the perichondrium and surrounding tissues at E13.5, which probably stimulates vascular in-growth (Zelzer et al., 2002). At E17, Sema3A and Neuropilin-1 are expressed by the outer perichondrium and the hypertrophic chondrocytes within the primary ossification centers of the diaphysis and metaphysis (Gomez et al., 2005). Thus, plexinD1 is likely to play an important role in proper blood vessel invasion into the bone; loss of plexinD1 in the vasculature leads to axial patterning defects. The skeletal requirement for normal vascularization is well established (Towler, 2007). For example, mice deficient in matrix metalloproteinase-1 display skeletal defects due to deficient microvasculature (Zhou et al., 2000), and bone-derived hypoxia-inducible factor 1 couples angiogenesis to osteogenesis (Wang et al., 2007).

We show that plexinD1 is also necessary in postnatal angiogenesis, which may have important implications for the future analysis of plexinD1 as a therapeutic target for pro-or anti-angiogenitc therapies. It remains unclear if PlexinD1 on endothelial cells functions primarily to mediate attractive or repulsive signals, or both. Indeed, Sema3E/PlexinD1 signaling in neurons can result in either attraction or repulsion, depending on the presence or absence of Neuropilin (Nrp)-1 (Chauvet et al., 2007). PlexinD1 has been detected in tumor vasculature, and in a number of brain tumors (Roodink et al., 2005). Putative PlexinD1 ligands, Sema3A and Sema3C, are able to modulate the invasive and adhesive properties of prostate cancer cells (Herman and Meadows, 2007). Nrp-1 and -2 are highly expressed by a variety of tumors (Bielenberg et al., 2006). It is attractive to speculate that modulation of Sema/Nrp/PlexinD1 signaling may affect tumor angiogenesis and disease progression in a manner analogous to that developed for modulation of VEGF signaling (Holash et al., 2002; Pan et al., 2007). Our demonstration of a role for PlexinD1 in postnatal retinal vascularity suggests that PlexinD1 may be a viable target for the modulation of diabetic retinopathy or other retinal vascular diseases.

Materials and Methods

Generation of plexinD1-flox mice

Mouse genomic clones were derived from a 129/Sv genomic library (Stratagene). A targeting vector was constructed using 1.8 kb XhoI-SacII region containing 5′ untranslated sequences for the 5′ arm, 2.4 kb SacII-EcoRI region including exon1 flanked by two loxP sites, and 8.5kb EcoRI-SalI (derived from phage DNA) region for the 3′ arm. A frt-neo-frt-loxP cassette was inserted in a SacII site before exon1. Another loxP was inserted in an EcoRI site after exon1. A linearized targeting construct was electroporated into 129Sv/Ev derived ES cells. Cells were selected with G418 and screened by Southern blot analysis using a 0.8 kb Xba I-Xho I fragment, generating a 12 kb wild-type and a 4 kb mutant band. Another probe outside the 3′ arm was also used to confirm 3′ homologous recombination. The cells carrying the correct mutation were injected into C57BL/6J blastocysts. Chimeric offspring were mated with C57BL/6J mice. Germ-line transmission of the mutant allele was determined by Southern blot analysis of genomic DNA from tails of mice. Heterozygous F1 animals (plexinD1^flox/+) were intercrossed with a FLPe transgenic mouse (Farley FW, 2000) to delete the neo cassette.

Genotyping

Mice with plexinD1^flox, plexinD1^ko, Tie2-cre and CMVcreER alleles were identified by PCR strategy using the following primers: plexinD1^flox forward primer: 5′-ACA GGT GTG TGC TCA AGG CCA CCT C-3′; plexinD1^flox reverse primer: 5′-CAG CCC TAT AGT TCT CCA CCA AAG A-3′). plexinD1^ko forward primer1: 5′-GCC AAG TTC TAA TTC CAT CAG AAG CTG AC-3′; plexinD1^ko forward primer2: 5′-TCA CAC CCC CTA TGT TCT CCA AGC CTC-3′ plexinD1^ko reverse primer: 5′-GAT CAC CTG GGT TTC TAT GAC CTT AGG-3′; Tie2-cre forward primer: 5′-CGC ATA ACC AGT GAA ACA GCA TTG C-3′ Tie2-cre reverse primer: 5′-CCC TGT GCT CAG ACA GAA ATG AGA-3′; CMVCreER forward primer (MG1084) 5′-GCG GTC TGG CAG TAA AAA CTA TC-3′ CMVCreER reverse primer (MG1085) 5′-GTG AAA CAG CAT TGC TGT CAC TT-3′. All studies involving animals were approved by the University of Pennsylvania Institutional Animal Care and Use Committee.

Primary culture of endothelial cells

Primary endothelial cells were harvested from P0 neonates and selected twice with Dynabeads (Invitrogen, #110-35) conjugated with endothelial cell surface marker CD31 antibody (BD Pharmingen, MEC13.3). This protocol was modified from previously published work (Chittenden et al., 2006; Li et al., 2002). Dynabeads were washed with PBS/0.1% BSA three times and incubated with CD31 antibody at 4°C overnight (antibody:beads = 1:50 vol/vol). P0 pups were euthanized with CO_2.. Heart and descending aorta from each pup were dissected and minced, and incubated in 5 ml sterilized type I collagenase (Worthington biochemical corporation, cat#M6B109) solution (1 mg/ml in PBS) and rotated at 37°C for 45 minutes. The tissue was disassociated, filtered to remove particulates, resuspended in 2 ml PBS/0.1%BSA, and incubated with 40μl CD31 antibody-conjugated Dynabeads at 37°C for 10 minutes. The complexes were selected by a magnet separator, washed with DMEM/20%FBS medium four times and then plated onto 1% gelatin pre-treated 6-well tissue culture plates with 2 ml complete endothelial cell growth medium (EBM+20%FBS+EGM-2 (Clonetics, CC-4176)).

Boyden chamber assay

Primary endothelial cells (passage 3–6) were starved overnight (EBM2+0.5%FBS) and labeled with 5 μg/ml Calcein AM (Molecular Probes #C3100MP) for 2 hours at 37°C. Labeled cells were resuspended in cell dissociation solution (GIBCO #13150-016) for 5–10 minutes and spin down, resuspended in complete endothelial cell growth medium at a concentration of 4×10⁵ cells/ml. 250μl cells were loaded into uncoated BD Falcon Fluoroblok inserts (BD #351151) (10⁵ cell/well). Calcein AM emission was measured every hour in a bottom plate reader to record cell migration.

Mouse aortic ring assay

Mouse aortic ring assay was modified from previously described methods (Masson et al., 2002). The thoracic aortas were harvested from P0 mice and kept in ice-cold sterilized DMEM (GIBCO BRL). The peri-aortic fibro-adipose tissue was removed, and the aortas were sliced to 1 mm rings and washed with DMEM. 300μl matrigel (BD, CB40234B) was seeded into 12-well plates on ice and allowed to solidify at 37°C for 30 minute. Four to 6 rings were placed in each well on the solid matrigel and another 300μl of liquid matrigel was added on top of the explants. The matrigel “sandwich” was then incubated at 37°C overnight and 200μl endothelial cell complete growth medium (EBM+20%FBS +EGM-2) was added into each well the next morning. Explants were photographed using a Nikon Eclipse TE2000-U inverted microscope and an attached digital camera. The length of the outgrowth was measured using ImageJ software.

In situ hybridization and immunohistochemistry

Embryonic tissue was fixed in 4%PFA for 24 to 48 hours followed by dehydration in ethanol/PBS solutions, xylene treatment and paraffin embedding. Paraffin sections (6 μm) were then used for in situ hybridization and immunohistochemistry. The plexinD1 antisense riboprobe were derived from EST #AA138454 corresponding to bp 5901–6955 of mouse plexinD1 mRNA (Accession # NM_026376). In situ probe for wnt2 was designed to target exons 3 and 4 of the mouse wnt2 gene. PAX1 in situ probe was designed to target nt 1781 to 2440 of mouse PAX1 mRNA (GenBank Accession #AK134066). ³⁵S-labeled antisense riboprobes were synthesized by in vitro run-off transcription of linearized plasmids, using SP6, T7 or T3 RNA polymerase, and ³⁵S-UTP.

For immunohistochemistry on paraffin sections we used MF-20 antibody (HybridomaBank), von Willebrand Factor (vWF) antibody (Sigma, F3520), Neurofilament antibody 2H3 (HybridomaBank), PECAM antibody (PharMingGen, MEC 13.3), tyrosine hydroxylase (TH) antibody (Chemicom AB152) and smooth muscle actin (SMA) antibody (Sigma A2547). Runx2 antibody (Sanrz Cruz #M-79) was used for immunohistochemical staining at a concentration of 1:20.

For whole-mount PECAM staining we used the purified anti-mouse CD31 (BD Pharmingen, MEC13.3) antibody, HRP-goat-to-rat antibody (Abcam, ab6120-250) and the DAB peroxidase substrate kit (Vector#SK-4100). E11.5 embryos or embryonic hearts were fixed in 4% PFA overnight at 4°C. Fixed embryos were rinsed in PBT (PBS/0.1% TWEEN) and dehydrated in methanol/PBT, bleached in 5% H₂O₂/methanol, and rehydrated in methanol/PBT. The tissues were then blocked in 2% milk/PBT for 30 minutes and incubated with CD31 antibody (1:500) in 2% milk/PBT at 4°C overnight, followed secondary antibody HRP-Goat-anti-Rat IgG staining, and were developed using DAB substrate kit (VECTOR, SK-4100). Animal and section preparations were described in the methods and materials.

RT-PCR

For RT-PCR analysis, mouse tissues were placed in liquid nitrogen immediately after dissection and stored at −80°C until use. Total RNA was isolated using Trizol reagent (Invitrogen). First-strand cDNA was generated by random heximer priming utilizing the SuperScript III first strand cDNA synthesis system for RT-PCR (Invitrogen). To prevent trace amount of DNA contamination, RNA samples were treated with amplification grade DNase I (Invitrogen) before reverse transcription. Of the resulting cDNA, 1.0 μl was used in a subsequent 20 μl quantitative PCR reaction. All qPCRs were conducted in 384 plates using the 7900HT Sequence Detection System and the results were analyzed by the SDS2.0 software (Applied Biosystem). Primers for plexinD1, (D1_F: 5′-TGA ACA CAC TGG CCC ACT ACA AGA-3′), (D1_R: 5′-TGT CCA AGT CTT TCA CTC TGC CCA-3′).

Skeletal staining

Mouse P0 pups were euthanized and the skin and fat tissue were removed and then fixed in 90% ethanol. The liver, kidney and gut were removed prior to staining. Tissue was then placed in 20 ml freshly prepared Alcian Blue solution (16 ml 100% ethanol, 2 mg Alcian Blue and 4 ml glacial acetic acid dissolved in 0.8 ml ddH₂O) for 3 days at room temperature, then washed in 70%, 40% and 15% ethanol solution each for 2–3 hours and then in water. The preparation was immersed in freshly prepared 1% KOH until it became transparent and then Alizarin Red S solution (1% KOH with Alizarin Red S) until the bones were stained.

Primary osteoblast cultures

P0 pups were sacrificed in CO₂ followed by decapitation. Calvaria were surgically removed, excess soft tissue trimmed, and placed in a sterile 1.5 ml eppendorf tube with 1 ml of digestion solution [αMEM, 0.1% collagenase (Sigma–type 1A#C9891), 0.2% dispase (Boeringer Mannheim #165859)] and incubated at 37°C @ 180RMP plate rotator for 10 minutes. The calvaria would settle to the bottom of the tube and the supernatant was discarded. Another 1ml of digestion solution was added and rotated at 37°C for 20–30 minutes. Supernatant was centrifuged at 1500 RPM for 5 minutes at 4°C and the pelleted cells were plated in 6-well dishes in culture media (αMEM+10%FBS).

Micro-CT bone imaging and sample preparation

Euthanized animals were perfused via the left ventricle first with 2 ml of heparinized saline, then 2 ml of 2% paraformaldehyde in PBS, and finally with 2–4 ml of Microfil (Flow Tech, Inc.) using a 21-gauge butterfly needle. The intravascular Microfil was allowed to solidify for 4 hours and the specimens were then fixed and stored in 10% formalin. Micro-CT imaging was performed on an eXplore Locus SP specimen scanner (GE Healthcare, London, Ontario, Canada). Scan protocols used 80 kVp, 80 μA, 250 μm Al filter, short-scan (Parker) method with 400 views at 0.5° steps, 1.7 s exposure time, 2×2 bin mode, four averages, and Feldkamp conebeam filtered back projection reconstruction at either 29×29×29 μm³ or 58×58×58 μm³ voxels. Three-dimensional volume rendering and multi-planar reformatting were performed using the OsiriX dicom viewer (www.osirix-viewer.com). Imaging was performed at the Small Animal Imaging Facility in the Department of Radiology at the University of Pennsylvania School of Medicine.

Induction with Tamoxifen

Tamoxifen (TM) (Sigma, #T5648) was dissolved in corn oil at a concentration of 10 mg/ml and administered as described (Hayashi and McMahon, 2002). For neonatal mice, 50 μl of TM was injected intraperitoneally at P3 and P4. For adult mice, 4 mg of TM/40g of body weight was injected intraperitoneally daily for 5 consecutive days.

Matrigel plug assay

Matrigel plugs were prepared as described (Malinda et al., 1997). Matrigel (Beckton Dickinson #356231) with PBS or endothelial cell growth factor (ECGF) was implanted subcutaneously. We use EGM-2 mixture (clonetics, CC-4176) as the supply of ECGF, which contains Heparin, VEGF, R3-IGF-1 and hEGF. After 7 days the Matrigel plugs were excised and fixed in 4%PFA for 24 hours at room temperature, embedded in paraffin and stained with Masson’s Trichrome. ImageJ was used to calculate cellularity.