Endothelial Neuropilin Disruption in Mice Causes DiGeorge Syndrome-Like Malformations via Mechanisms Distinct to Those Caused by Loss of Tbx1

The spectrum of human congenital malformations known as DiGeorge syndrome (DGS) is replicated in mice by mutation of Tbx1. Vegfa has been proposed as a modifier of DGS, based in part on the occurrence of comparable phenotypes in Tbx1 and Vegfa mutant mice. Many additional genes have been shown to cause DGS-like phenotypes in mice when mutated; these generally intersect in some manner with Tbx1, and therefore impact the same developmental processes in which Tbx1 itself is involved. In this study, using Tie2Cre, we show that endothelial-specific mutation of the gene encoding the VEGFA coreceptor neuropilin-1 (Nrp1) also replicates the most prominent terminal phenotypes that typify DGS. However, the developmental etiologies of these defects are fundamentally different from those caused by absence of TBX1. In Tie2Cre/Nrp1 mutants, initial pharyngeal organization is normal but subsequent pharyngeal organ growth is impaired, second heart field differentiation is normal but cardiac outflow tract cushion organization is distorted, neural crest cell migration is normal, and palatal mesenchyme proliferation is impaired with no change in apoptosis. Our results demonstrate that impairment of VEGF-dependent endothelial pathways leads to a spectrum of DiGeorge syndrome-type malformations, through processes that are distinguishable from those controlled by Tbx1.


Introduction
DiGeorge syndrome (DGS) is a human genetic disorder characterized by a range of cardiac outflow tract (OFT) and great vessel malformations, craniofacial malformations, hypoplasia of the thymus and parathyroid pharyngeal organs, and cognitive impairment [1]. Most cases of DGS are associated with hemizygous microdeletions of chromosome 22q11.2. The gene encoding TBX1, a member of the T-box containing family of transcription factors, is located within this region and haploinsufficiency of this gene appears to be responsible for most of the manifestations of DGS in humans. In mice, heterozygosity of Tbx1 can result in impaired 4 th pharyngeal arch artery development and thereby in type B interrupted aortic arch [2][3][4], one of the prominent cardiovascular phenotypes in human DGS. Complete absence of TBX1 in mice results in a high severity and penetrance of almost all of the phenotypes seen in human DGS, including cardiac outflow and arch artery defects, pharyngeal organ hypoplasia, and craniofacial malformations [3,5].
A common theme in the development of the organs that are impacted in DGS is their dependence on neural crest cells. However, Tbx1 is not expressed in and does not function in neural crest cells [6][7]. Rather, Tbx1 functions in pharyngeal ectoderm, mesoderm, or endoderm (depending on the organ system investigated), and acts through a variety of tissue interactions and downstream effectors in different developmental processes.
Tbx1 is expressed in endothelial cells, although mutation of Tbx1 in vascular endothelium had no consequence in cardiac, pharyngeal, or craniofacial morphogenesis [8][9]. Tbx1 function in endothelium is required for proper lymphatic vessel development [10], although lymphatic defects are not generally thought to contribute to or be part of the clinical spectrum associated with DiGeorge syndrome.
A number of genes cooperate with Tbx1 and therefore result in some or all DGS phenotypes when mutated in mice. One such gene is Crkl [11][12], which is located near Tbx1 and is therefore also deleted in most cases of DGS. Many other genes that are not physically linked to Tbx1 have been shown to interact functionally or genetically with Tbx1 in one or more of the developmental processes in which Tbx1 participates [2,[12][13][14][15][16][17][18][19][20][21][22]. Finally, other genes are known to cause some DGS-like phenotypes when mutated in mice, but have not been analyzed to determine if or how they intersect with Tbx1.
Mouse embryos lacking VEGF164, the major isoform of the Vegfa gene (which is not chromosomally linked to Tbx1), exhibit craniofacial, pharyngeal organ, pharyngeal arch artery, and cardiac outflow tract defects and therefore resemble DGS [23]. Among other sites, Vegfa is expressed in the myocardium of the heart outflow tract, in pharyngeal endoderm adjacent to the pharyngeal arch arteries, and in the developing thymus and palate [23][24][25], which is consistent with these phenotypes. Based on a number of arguments (see Discussion), Vegfa has been proposed to interact genetically with Tbx1 and thereby function as a modifier of the DiGeorge syndrome [23]. Vegfa deficiency is also associated with generally impaired vascular development, specifically in the formation of small and medium caliber vessels [23,[26][27], representing a known function of VEGF in promoting angiogenic vessel branching [27][28].
Neuropilin-1 (NRP1) is a dual ligand coreceptor, mediating the effects of VEGF164 when coupled to the VEGF receptor FLK1 (VEGFR2), and the effects of members of the Sema3 family when coupled to several plexins [29]. NRP1 can also mediate VEGF signaling in the absence of VEGFR2 in some contexts [30]. Global knockout of Nrp1 in mice is embryonic lethal around E12.5; cardiovascular defects in these embryos included a general deficiency in small and medium caliber blood vessel formation, altered pharyngeal arch artery formation, and a failure to septate the outflow tract into the ascending aorta and pulmonary trunk [31], all similar to phenotypes seen in Vegfa mutants [23]. Conditional Nrp1 disruption in endothelial cells using Tie2Cre [32] allowed survival to full term, and replicated several of the cardiovascular defects seen in the global Nrp1 knockout, including general peripheral vascular branching deficiency and failure to septate the outflow tract [33].
In this study, we have investigated the developmental role of Nrp1 function, specifically in endothelial cell lineages, with particular attention to the developmental processes that are impacted in DGS. In addition to the cardiac outflow tract defect (lack of outflow tract septation) that has already been reported [33], we find that Tie2Cre/Nrp1 mutants have pharyngeal arch artery, pharyngeal organ, and craniofacial defects that collectively resemble those seen in Vegfa mutants and in human DGS. However, we find that the developmental basis of each defect in Tie2Cre/Nrp1 mutants is dissimilar to what occurs in Tbx1 null embryos. We suggest that defective vascular organization and endothelial cell dysfunction underlie the phenotypes of Vegfa and Nrp1 mutants. More importantly, our results demonstrate a requirement for Nrp1 function in endothelium that when compromised leads to DiGeorge syndrome-like defects through mechanisms distinct to those associated with Tbx1 deficiency.

Results
Endothelial-specific deletion of Nrp1 causes DiGeorge syndrome-like defects in mice Embryos with conditional knockout of Nrp1 driven by Tie2Cre survived to full term in normal numbers, but died in the immediate postnatal period. The overall size of E18.5 Tie2Cre/ Nrp1 mutants was not obviously different, and limb development was normal at all stages (Fig. S1A), implying no general embryowide growth deficiency or developmental delay. In more than 20 E18.5 mutant embryos, we confirmed that all Tie2Cre/Nrp1 mutants had a single cardiac outflow vessel (a phenotype called persistent truncus arteriosus or common arterial trunk; Fig. 1A-B), as previously reported [33]. All Tie2Cre/Nrp1 mutants also had a ventricular septal defect, which is a hemodynamic necessity of a common arterial trunk. In addition, all Tie2Cre/Nrp1 mutant embryos examined at late stages had some malformation of the great vessels ( Fig. 1A-B): a type B interrupted aortic arch was most common, additional great vessel malformations included retroesophageal right subclavian artery and a right sided dorsal aorta. In 10 out of 10 Tie2Cre/Nrp1 mutants examined at E18.5, the thymus was hypoplastic (Fig. 1C); morphometric quantitation ( Fig. 1D) indicated an average 50% decrease in thymic size in mutants relative to littermate controls. Although not quantitated, the thyroid was also hypoplastic in all Tie2Cre/Nrp1 mutants  Fig. S1D-E), although in most Tie2Cre/Nrp1 mutants, the mandible was not obviously dysmorphic. In summation, Tie2Cre/Nrp1 mutants have a high or complete penetrance of cardiac outflow tract, great vessel, pharyngeal organ, and craniofacial defects that collectively typify DiGeorge syndrome in humans and that are also seen with high or complete penetrance in Tbx1 and Vegfa mutant mice. Tie2Cre/Nrp1 mutation also resulted in two additional phenotypes seen in late gestation embryos. There was a disruption in the continuity of the atrial myocardium in all of five Tie2Cre/Nrp1 mutants examined for this phenotype (Fig. S1F-G), and in one of seven examined mutants, there were mild defects in the structure of the vertebral bodies (Fig. S1H). Comparable atrial and more severe vertebral phenotypes were observed in Tie2Cre/plexinD1 mutants [34], but have not been reported in Vegfa mutants and are not associated with DiGeorge syndrome. These defects therefore are likely to represent defective semaphorin, not VEGF, signaling through NRP1/plexinD1 complexes in endothelial cells.
Global heterozygosity of Nrp1, and endothelial-specific conditional Nrp1 heterozygosity driven by Tie2Cre, are both viable and apparently normal (i.e., no malformations were reported for these genotypes) [31,33]. In our analysis of 14 Tie2Cre conditional heterozygous Nrp1 mutants at E18.5, we did not see any malformations in the cardiac outflow tract, the great vessels, pharyngeal organs, or craniofacial structures.
Altered endothelial cell organization and outflow tract cushions in Tie2Cre/Nrp1 mutants We next explored the morphogenesis of the cardiac outflow tract, the pharyngeal arch arteries and pharyngeal organs, and the palate to determine if the comparable DGS-like phenotypes seen in Tie2Cre/Nrp1 and Tbx1 mutants at late embryonic stages share a common underlying developmental and mechanistic basis.
Tbx1 normally functions in pharyngeal and splanchnic mesoderm [8,35] to maintain the second heart field (SHF), a progenitor population that is added progressively to the distal end of the cardiac outflow tract during the E8-11 period of development [36]. This process allows the OFT to lengthen and thereby achieve a characteristic looped shape by E10.5, and is necessary for septation to occur during the E10.5-11.5 period. In Tbx1 null embryos at E10.5, as also previously reported [37], the OFT was shortened and poorly looped ( Fig. 2A-B). In contrast, the morphology (length and degree of looping) of the OFT was normal in all E10.5 Tie2Cre/Nrp1 mutants ( Fig. 2C-D). ISL1, a marker of the SHF, was normally expressed in Tie2Cre/Nrp1 mutants ( Fig. 3A-B), whereas the domain of ISL1 expression is reduced in Tbx1 mutants [38]. MF20, a marker of differentiated myocardium, was properly expressed in Tie2Cre/Nrp1 mutants up to the distal end of the OFT (Fig. 3C-D), whereas MF20 expression extends into the splanchnic mesoderm in Tbx1 mutants because of premature myocardial differentiation [38][39]. These observations imply that the early aspects of SHF differentiation that are controlled by Tbx1 are normal in Tie2Cre/Nrp1 mutants.
Neural crest cells are required for OFT septation [40]. NC cells that reach the distal end of the OFT begin to differentiate and express smooth muscle actin (SMA). In Tbx1 mutants, NC cell migration to the OFT is disrupted [2,41]. In contrast, we observed a normal number and distribution of SMA-positive cells (of presumptive NC origin) at and around the distal OFT of Tie2Cre/ Nrp1 mutants (Fig. 3E-F). We also crossed the conditional lacZ reporter allele R26R into the Tie2Cre conditional mutant background and visualized endocardium and endocardiumderived cells in the OFT by Xgal staining. Unstained mesenchymal cells in the OFT are of presumptive NC origin, as the only known sources of OFT mesenchyme are NC and transformed endocardium [32,42]. The amount and distribution of unstained and stained mesenchyme in the OFT was normal in control and Tie2Cre/Nrp1 mutants ( Fig. 3G-J). Thus, there does not appear to be any apparent defect in NC migration into the OFT that might account for septation failure. Proliferation of mesenchymal neural crest cells and of endothelial cells in the distal OFT was normal Outflow septation involves the growth and fusion of two opposing mesenchymal cushions across the lumen of the OFT [43]. These cushions form by the migration of NC cells between the endocardium and myocardium, and close to the right ventricle, the outflow tract cushions also contain endocardium-derived mesenchyme [42]. In Tie2Cre/Nrp1 mutants, OFT cushion number and morphology were disorganized. This was seen most clearly in sections orthogonal to the outflow tract ( Fig. 3G-J): in normal embryos, the endocardium and myocardium were juxtaposed at two opposite locations (between the two mesenchymal cushions), whereas in mutants, there were multiple sites of association between the endocardium and myocardium (i.e., multiple cushions), and with abnormal attachment of the endocardium to the myocardium.
Valve leaflets form from the distal (truncal) segment of the outflow tract cushions. In normal embryos following septation, both vessels have three valve leaflets. In Tie2Cre/Nrp1 mutants, we observed a variable number of leaflets (2-4 or more), with these having a range of sizes and morphologies (Fig. S2). Abnormal OFT cushion number and organization therefore might be the initial anatomical defect that leads to improper septation and then also to valve leaflet disorganization. Defective arch artery formation in Tie2Cre/Nrp1 mutants The pharyngeal arch arteries form sequentially as the connections between the outflow tract and systemic circulation. At E10.5, the 1 st and 2 nd arch arteries have regressed and the 3 rd , 4 th , and 6 th arch arteries are organized in a bilaterally symmetric pattern, connecting to the right and left dorsal aortae. From E11, these arch arteries and segments of the dorsal aortae are reorganized in a highly asymmetric manner to form the mature great vessels [44]. We performed ink injection into the hearts of E10.5 embryos in order to visualize the pattern of arch arteries in control vs. Tie2Cre/Nrp1 mutants (Fig. 4). The most common defect, seen in 8 of 10 mutant embryos, was a bilateral absence or hypoplasia of both 4 th arch arteries (Fig. 4B). In two of these, both 6 th arch arteries were also missing, and in one embryo, only the right 6 th arch artery was also missing. One mutant had a missing right 6 th arch artery with normal 4 th arch arteries (Fig. 4C), and one mutant had what appeared to be a normal arch artery organization. These arch artery defects at E10.5 are consistent with the great vessel malformations seen in E18.5 mutant embryos as described above. In Tie2Cre/Nrp1 mutants, smooth muscle differentiation around arch arteries that were properly formed appeared normal (Fig. 3E-F).
Heterozygosity of Tbx1 in mice results in a moderate frequency of 4 th arch artery defects [2][3][4]. In contrast, no arch artery defects were observed in 13 Tie2Cre endothelial-specific conditional Nrp1 heterozygotes at E10.5 (littermates of those described above), and as noted earlier, no defects were observed in the great vessels or elsewhere in E18.5 conditional Nrp1 heterozygous embryos. Arch artery defects were also not observed among 24 littermate embryos at E10.5 that lacked the Tie2Cre gene.

Impaired thymic vascular development in Tie2Cre/Nrp1 mutants
The thymus and parathyroids are derived from the 3 rd pharyngeal pouch. As also previously observed [2,35,45], Tbx1 mutants lacked proper segmentation of the pharyngeal arches and pouches ( Fig. 5A-B), which compromises the formation of the thymus and parathyroid organs. In contrast, in every Tie2Cre/Nrp1 mutant embryo, we observed normal pharyngeal pouch organization, even as pharyngeal arch artery development was obviously compromised (Fig. 5C-D). Thus, the pharyngeal organ defects of E18.5 Tie2Cre/Nrp1 mutants do not share a common developmental etiology with the similar malformations seen in Tbx1 Upper row in all pairs of panels is a control, lower row is a littermate Tie2Cre/Nrp1 mutant. A,B, Immunohistochemical detection of ISL1 at E10.5 (sagittal section); the brackets indicate the ISL1 + second heart field. C,D, Immunohistochemical detection of MF20 at E10.5 (sagittal section); the arrows point to the distal limit of myocardium in the outflow tract. E,F, Immunohistochemical detection of smooth muscle actin at E11.25 in the outflow tract (transverse section); the asterisks show the location of immunoreactive mesenchyme of presumptive neural crest derivation at the distal end of the outflow tract. In the control, OFT septation is underway although not yet completed, whereas in the mutant, this process fails, resulting in a common arterial trunk (CAT). Positive staining is also apparent in scattered mesenchyme cells lower in the truncal cushions of the outflow tract, and in differentiating smooth muscle surrounding the 6 th arch arteries (6). At this stage, SMA also labels myocardium. G-J, Orthogonal sections of Xgal-stained Tie2Cre/R26R distal outflow tracts from two different control (G,I) and two different Tie2Cre/R26R/Nrp1 (H,J) E10.5 embryos; the endocardium is Xgal-positive whereas the mesenchyme is unlabeled and therefore of presumptive neural crest derivation. Arrows point to the juxtapositions between the endocardium and the myocardium that separate OFT cushions; there are two such points in controls, and several in mutants indicating abnormal cushion number and organization. Note also the normal extent of Xgal + mesenchyme derived by endocardial transformation in the atrioventricular canal (AVC; I,J). K, OFT cell proliferation at E10.5; embryos were BrdU labeled and sections through the distal OFT were immunstained to detect proliferating cells. Mesenchymal cells are of presumptive neural crest derivation, endothelial cells (EC) were identified based on morphology. doi:10.1371/journal.pone.0032429.g003 mutants. Furthermore, these observations also show that the arch artery and great vessel defects in Tie2Cre/Nrp1 mutants described above do not occur as the indirect result of pharyngeal disorganization.
Thymic morphology and size at E12.5 was normal in Tie2Cre/ Nrp1 mutants, but hypoplasia became apparent at E13.5 and persisted proportionally as development proceeded. TUNEL staining showed no difference in apoptosis at any stage (Fig. S3). We used PECAM1 immunostaining to visualize vascular organization in the developing thymus. In normal development, blood vessels initiate assembly around the periphery of the thymus at E12.5 [46], and this occurred properly in mutants (Fig. 6A-B). At E13.5, in control embryos, peripheral vessels were still present and in addition several intrathymic vessels had also formed, whereas in Tie2Cre/Nrp1 mutants, peripheral vessels were present but with few if any internal vessels (Fig. 6C-D). In control embryos at E14.5, extensive intrathymic vessels were present and now with few peripheral vessels, whereas in mutants, the majority of vessels were still at the thymic perimeter with fewer and less complex intrathymic structures (Fig. 6E-F). Thus, thymic vascular formation is deficient or delayed in Tie2Cre/Nrp1 mutants.

Reduced palatal mesenchyme cell proliferation in
Tie2Cre/Nrp1 mutants The palate forms from opposed shelves in the oral cavity that first elevate, grow towards each other, and then fuse. In normal development, the palatal shelves contact each other at E14.5 and by subsequent breakdown of the medial epithelium become fused at E15.5 [47]. Cleft palate was present in approximately half of E18.5 Tie2Cre/Nrp1 mutants (described above). We examined younger embryos to determine the basis of this phenotype. At E14.5, we observed a spectrum of palatal shelf hypoplasia, ranging from failure to elevate to various degrees of incomplete growth (Fig. 7A-D). We also observed fusing palatal shelves in approximately half of examined mutant embryos at E14.5 (Fig. 7E), which is consistent with the range of phenotypes scored at E18.5.
In Tbx1 nulls, cleft palate is a highly penetrant phenotype, and analysis at E14.5 showed a diminishment of palate shelf size [48], similar to what we observed in approximately half of the Tie2Cre/ Nrp1 mutants. In Tbx1 nulls, there is an increase in palatal mesenchyme cell proliferation that is more than counterbalanced by an increased level of apoptosis, resulting in insufficient net growth and thereby in clefting [48]. In contrast, a similar analysis of Tie2Cre/Nrp1 mutants revealed a significant decrease in cell proliferation ( Fig. 7F-H) and no change in apoptosis (Fig. 7I-J). Thus, the cellular basis of cleft palate in the two models is distinct.
The mesenchyme of the palatal shelves is derived almost completely from neural crest, so a decrease in palatal mesenchyme cell proliferation must be an indirect consequence of Nrp1 disruption in endothelium. We used PECAM1 immunostaining to visualize the developing vasculature in the palatal shelves. At E12.5, we consistently observed fewer blood vessels in mutants compared to controls (Fig. 7K-L). This may be the initial defect that leads to decreased mesenchymal cell proliferation and thereby to cleft palate. At E13.5 and E14.5, vascular distribution was more variable in mutants and often resembled controls; this may explain why palate fusion was achieved in approximately half of mutant embryos by E14.5.

Lack of genetic interaction between Nrp1 and Tbx1
In mice, the primary phenotype in Tbx1 heterozygotes is a variable penetrance of 4 th arch artery hypoplasia (resulting in type B interrupted aortic arch in later embryos), although the penetrance of this phenotype is strain-dependent. All other DGS defects occur when Tbx1 is homozygous, or when Tbx1 heterozygosity is combined with mutation of another interacting gene. We addressed whether heterozygous Nrp1 deficiency in the Tie2Cre domain synergized with global heterozygous Tbx1 deficiency. In the outbred  genetic strain background of our mice, there was no incidence among 9 Tbx1 heterozygous embryos analyzed of interrupted aortic arch nor of any other malformation. All of 7 Tie2Cre,Nrp1 fl/+ ,Tbx1 2/+ embryos, analyzed at E18, were also normal in all regards (Fig. S4), implying no detectable enhancement of phenotypic penetrance. The power of this analysis is limited by the absence of observed phenotypes, and detection of a possible epistatic interaction might require backcrossing onto a more susceptible genetic background. However, the apparent lack of genetic interaction is fully consistent with the observations described above showing a different cellular and developmental basis for each DGS phenotype in Tie2Cre/Nrp1 and Tbx1 mutants. Collectively, our results imply that Tbx1 and Nrp1 are parts of distinct pathways that lead to DGS malformations.

Discussion
In this study, we report that endothelium-specific disruption of Nrp1 causes DiGeorge syndrome-type malformations in mice. An unseptated outflow tract (common arterial trunk) was noted in an earlier description of Tie2Cre/Nrp1 mutants [33], and global Nrp1 disruption was previously shown to result in early pharyngeal arch artery defects [31] that would be manifest as great vessel malformations if such embryos were able to survive to late gestation. However, the constellation of cardiovascular outflow tract, great vessel, pharyngeal organ, and craniofacial defects that collectively typify DiGeorge syndrome had not been previously recognized in Tie2Cre/Nrp1 mutants, and could not have been seen (because of early lethality) in global Nrp1 mutants.
Because NRP1 functions as a coreceptor with VEGFR2 for VEGF signals, and with plexins for Sema3 signals, Nrp1 mutant phenotypes might reflect either or both signaling pathways. Cleft palate or pharyngeal organ malformations are not known to result from plexin or semaphorin gene mutations, but do occur in Vegfa mutants, implying that these phenotypes in Tie2Cre/Nrp1 mutants represent Vegfa signaling pathways. Cardiac outflow tract defects are observed in Sema3c and in plexinD1 mutants as well as in Vegfa mutants; however, a previous detailed genetic analysis showed that NRP1 is required for VEGF but not for Sema3 signaling in this tissue [33,49]. We conclude therefore that NRP1 is required as a receptor for VEGF signals that are important in cardiac outflow tract, arch artery, pharyngeal organ, and craniofacial development (i.e., the organs impacted in DiGeorge syndrome). In contrast, the atrial myocardium and vertebral defects seen in Tie2Cre/Nrp1 mutants are not known to occur in Vegfa mutants but are manifest in Tie2Cre/plexinD1 mutants, implying that NRP1 serves as a plexinD1 coreceptor for Sema3 signals that support the integrity of the atrial myocardium and that mediate vertebral skeletal organization.
One of the best known roles of VEGF is to support blood vessel vasculogenesis and angiogenesis, and previous descriptions of Vegfa [23,[26][27] and Nrp1 [31,33] mutants reported a deficiency of blood vessel density, specifically of medium and small caliber vessels, in tissues throughout the embryo. We observed the same type of deficiency in the palate and thymus of Tie2Cre/Nrp1 mutants, which may underlie the diminished growth of these organs during mid-to late gestation. The great vessel defects seen in this mutant background result from a failure of some of the initial arch arteries to form. Vegfa is expressed in pharyngeal endoderm [23][24], and is therefore in close proximity to influence the behavior of NRP1-expressing endothelial cells in the pharyngeal arches. We do not yet know if the primary defect is in the initial organization of endothelial cells into nascent vessels, or in their subsequent interaction with neural crest cells that is required for vessel stabilization [50].
The outflow tract phenotype of Tie2Cre/Nrp1 mutants appears to be qualitatively distinct from the vascular defects seen elsewhere in these embryos. Vegfa is expressed in outflow tract myocardium [23][24][25], and our results suggest that it signals through NRP1 in the endocardium to facilitate cushion organization, which is required for subsequent septation and valve formation. Distortion of OFT cushion organization is likely to be the underlying cause of the single persisting outflow vessel in Tie2Cre/Nrp1 mutants, as the ultimate fusion of these malformed and mispositioned cushions would result in the obstruction of one of the two outflow channels [51]. DiGeorge syndrome in humans is most commonly associated with mutation of the Tbx1 gene, and in mice, the terminal DGS phenotypes of Tie2Cre/Nrp1 and Tbx1 mutants are similar. Nonetheless, our analysis revealed fundamental differences in the developmental process that lead to these phenotypes in the two mutant backgrounds. Thus, in Tbx1 mutants, a second heart field deficiency leads to a shortened and misaligned outflow tract [8,35,37] (Fig. S2A-B), defective neural crest cell migration and/ or smooth muscle differentiation appear to be responsible for arch artery defects [2,7,41,52], disorganization and lack of segmentation of the pharyngeal endoderm results in pharyngeal organ defects [2,35,45] (Fig. 5A-B), and elevated mesenchymal apoptosis leads to cleft palate [48]. None of these events occur in Tie2Cre/ Nrp1 mutants. Instead, we have described alternate developmental processes that explain the various phenotypes. If Nrp1 and Tbx1 converge in a functionally relevant manner, we would predict not only similar terminal mutant phenotypes but also similar intermediate developmental manifestations, which we did not see. Our results lead to the conclusion that a broad range of DGS phenotypes can occur as the result of disruption of a single gene (Nrp1) and through mechanisms that are distinct from those associated with Tbx1.
We believe that our conclusion of a Tbx1-independent pathway leading to the spectrum of defects that typify DGS is novel. Many mouse gene mutations result in a similar range of DGS-type malformations, but these genes have either been shown to interact genetically or mechanistically with Tbx1 and therefore impair the same developmental processes as does Tbx1, or simply have not been examined in detail. Many other gene mutations cause individual phenotypes but not the larger constellation of DGS malformations, and these clearly need not intersect with Tbx1. It is also noteworthy that no previous genetic manipulation that results in the broad spectrum of DGS-type defects has been shown to involve the endothelium, and in the case of Tbx1, an endothelial role has been disproven [8][9][10]. One possible example similar to Nrp1 involves the endothelin signaling system, where mutation of the ligand gene Edn1 [53][54], of the receptor gene Ednra [55], or of the converting enzyme gene Ece1 [56][57] all lead to DGS-type malformations. Analysis of Tbx1-Ece1 trans heterozygotes failed to reveal genetic interaction [58], as we also observed between Tbx1 and Nrp1. However, if endothelin is part of the VEGFA-NRP1 axis rather than part of the TBX1 pathway, the relationship must be indirect: Ednra function in neural crest cells is required for the normal morphogenesis of the cardiac outflow tract, arch arteries, pharyngeal organs, and craniofacial structures [59], although endothelial-specific mutation of Edn1 caused no developmental malformations [60].
Vegfa has been proposed as a genetic modifier of the chromosome 22q11 deletion syndrome [23], based on several criteria: the occurrence of similar phenotypes in mouse Vegfa mutants as in Tbx1 mutants, the observation of genetic synergy between vegfa and tbx1 in zebrafish embryos specifically for arch artery defects, a Vegfa haplotype in humans that was associated with increased incidence of cardiovascular phenotypes in chromosome 22q11 patients, and finally by the decreased expression of Tbx1 in the pharyngeal region of E10.5 Vegfa mutant mouse embryos. Tie2Cre/Nrp1 and Vegfa mouse mutants share most DGSrelated phenotypes, yet our results imply that Tbx1 is not involved in the etiology of any of the Tie2Cre/Nrp1 defects. Thus, it may also not be necessary to invoke Tbx1 to explain most of the Vegfa deficient phenotypes. A prior study also questioned whether non-Tbx1-dependent mechanisms explain the effects of VEGF on pharyngeal arch artery development [9]. At the same time, in some tissues that are outside of the Tie2Cre recombination domain, VEGF may also regulate or intersect with Tbx1. This may for example explain the reported decrease in Tbx1 expression in the pharyngeal arch region of Vegfa mutants, and the higher incidence and severity of mandible hypoplasia in Vegfa mutants compared to Tie2Cre/Nrp1 mutants.
With the exception of the outflow tract phenotype, the other VEGF-dependent defects observed in Tie2Cre/Nrp1 mutants appear to reflect deficient vascular formation (great vessel defects) or angiogenesis (for the thymus and palate). Deficient vascular organization has previously been associated with teratogeninduced and spontaneous cleft palate [61][62], and a ''vascular hypothesis'' has been proposed to more broadly explain DiGeorge syndrome phenotypes as the consequences of vascular insufficiency [63]. Clearly in the case of chromosome 22q11 deletion syndrome and for Tbx1 mutation, a disruption of earlier processes is a more suitable explanation for these defects. However, our results suggest that VEGF-and NRP1-dependent vascular defects can result in many DGS phenotypes in a manner that is independent of and later in developmental time than when the Tbx1 gene functions, at least as Tbx1 function is currently understood.
There is an alternative interpretation of our observations and of those from past studies that we cannot formally exclude. TBX1 clearly controls many early developmental processes in an NRP1independent manner that when compromised result in DGS. However, Tbx1 might in addition also control VEGF signaling to endothelial NRP1 in a way that accounts for the later alternative defects that we observed in Tie2Cre/Nrp1 mutant embryos. The lack of genetic interaction between Tbx1 and Nrp1 argues against this model, although the negative result does not disprove its possibility. If this model is true, Tbx1 controls two independent sets of processes (earlier NRP1-independent and later VEGF/NRP1dependent) that both lead to the same spectrum of DGS phenotypes. This would be a novel aspect of Tbx1 function that has been obscured by all of the early (NRP1-independent) defects that alone are sufficient to explain DGS phenotypes in Tbx1-null embryos.
In Tie2Cre/Nrp1 mutants, deficient vascular organization is not confined to only the cardiac outflow tract, pharyngeal arch arteries, pharyngeal organs, and palate; similar vascular deficiencies occur throughout the embryo [33], including many organs that are not grossly impacted by Nrp1 mutation and that are not part of the DGS spectrum. This is also true for Vegfa mutation [23,[26][27]. VEGF signaling through endothelial NRP1 is therefore responsible for local vasculogenesis and angiogenesis, although most tissues are not apparently affected when this process is compromised. The unexpected observation is that the tissues that are most commonly impacted in DiGeorge syndrome appear to be particularly dependent on VEGF-NRP1 signaling. It has long been appreciated that neural crest cells contribute to the tissues that are compromised in DGS. Thus, NC cells are required for OFT septation [40], support the formation of the pharyngeal arch arteries [50], form the pericytes and smooth muscle of the thymic vasculature [64][65][66], and contribute virtually all of the mesenchyme that allows growth of the palatal shelves [47]. Moreover, ablation of the NC can phenocopy DGS [67]. We speculate that a VEGF-dependent interaction between endothelial and neural crest cells may account for the restricted (DGS spectrum) distribution of phenotypic consequences in Tie2Cre/ Nrp1 and Vegfa mutants. For arch artery formation and cardiac outflow tract septation, this might involve a direct physical or signaling interaction between VEGF-activated endothelial cells and neural crest cells. Alternatively, neural crest cells are thought to be particularly sensitive to hypoxia [68][69], and organs such as the thymus or palate that depend on NC cells for their growth might be indirectly compromised as the consequence of vascular deficiency.
Most cases of DGS in humans are associated with Tbx1 deficiency; however, a number of DGS patients lack chromosome 22 deletions and do not appear to have mutations in the Tbx1 gene [70]. A second DGS locus has been postulated to reside on human chromosome 10 [71][72], in the vicinity of the Nrp1 gene. Nrp1 may therefore be a candidate gene to explain DGS phenotypes in patients without Tbx1 mutations.

Immunohistochemistry
Embryos were fixed in 4% PFA overnight at 4uC, then dehydrated in an ethanol series and embedded in paraffin, and sectioned at 10 m thickness. Following dewaxing and rehydration, primary antibodies used were against MF20 (Developmental Studies Hybridoma Bank; 1:20 dilution), ISL1 (Developmental Studies Hybridoma Bank; 1:20), and SMA (Sigma, 1:1000). For BrdU detection, pregnant females were injected with 100 mg/g body weight BrdU 2 hr before sacrifice, followed by fixation, paraffin embedding, and sectioning as above; primary antibody was from the BrdU Staining kit (Invitrogen). For detection of CD31 (BD Pharmingen MEC13.3; 1:400) and phospho-histone H3 (Cell Signaling; 1:100), embryos were fixed as above then transferred to 10% and then 30% sucrose, then embedded in OCT on dry ice, and 10 m frozen sections were cut. Biotinconjugated secondary antibodies (Santa Cruz, 1:300) were used according to the primary antibody species. Streptavidin-HRP was applied to slides after secondary antibody incubation; signal was detected with the DAB Plus kit (Invitrogen). In general, sections were counterstained with hematoxylin prior to mounting.

In situ hybridization
Embryos were isolated in RNase free PBS on ice. The isolated embryos were fixed in 4% PFA overnight at 4uC, then transferred to 10% and then 30% sucrose, then embedded in OCT on dry ice. 10 m sections were cut. Digoxygenin-labeled PTH probe was made by in vitro transcription from a template obtained from N. Manley. Signals were detected with anti-digoxygenin alkaline phosphatase-coupled antibody (Roche) and BM Purple substrate kit (Roche).

Ink injection
Embryos were isolated and kept in PBS on ice. The body wall was opened and the right ventricle exposed; India ink (Martin's Bombay Black, diluted in PBS as needed) was injected directly into the right ventricle by mouth pipetting with a drawn out glass micropipet. Embryos were then fixed in 4% PFA overnight.

X-gal staining
Embryos were isolated on ice and fixed with 0.2% glutaraldehyde at 4uC for 10 min, then whole mount stained in Xgal at room temperature overnight. For sections, stained embryos were processed to 10% and 30% sucrose, then embedded and frozen in OCT. 10 m sections were rinsed in PBS and counterstained with nuclear fast red.

TUNEL assay
Paraffin sections were cut at 10 m; the remaining procedures were done according to the In Situ Death Detection kit protocol (Roche). The positive controls were prepared from sections of the same embryos by preincubation with DNaseI to induce DNA strand breaks, according to the kit protocol.
Bone and cartilage staining E18.5 embryos were isolated and fixed in 95% ethanol overnight. Following a brief rinse in deionized water, cartilage was stained with alcian blue (0.05% alcian blue 8GX/5% acetic acid in distilled water) for 24 hours. Stained embryos were washed in 70% ethanol for 6-8 hr, then transferred to 1% potassium hydroxide overnight. Bone was stained with alizarin red (0.005% (w/v) in 1% KOH) for 2-4 hr. The tissue was then cleared in 1% KOH/20% glycerol for 2 or more days.

OFT valve staining
The OFT valve region of E18.5 embryos was dissected manually, then fixed in 4% PFA overnight. The tissue was whole mount stained with alcian blue (0.05% alcian blue 8GX/5% acetic acid in distilled water) for 24 hr, then rinsed in water and photographed using darkfield optics. Figure S1 Embryogenesis in Tie2Cre/Nrp1 mutants. A, Comparison of E14.5 forelimbs from a control (top) and Tie2Cre/ Nrp1 mutant (below) littermates to show normal limb development. B,C, In situ hybridization detection of parathyroid hormone to visualize the parathyroids in a control (A) and mutant (B) embryo at E18. D,E, Images of E14.5 heads of control (C) and mutant (D) embryos; the arrowhead in E points to the one obvious example of mandibular hypoplasia seen in this mutant background in this study. The shiny appearance of the embryo in panel E is a lighting artifact. F,G, Atrial myocardium at E14.5 in a control (F) and mutant (G) embryo; in the mutant, the myocardium is detached from the epicardium and is split into multiple pieces (two shown by arrows). H, Skeletal preparation of one mutant embryo at E18.5 showing vertebral defects, specifically an incompletely formed lumbar L2 vertebral body, and a partial fusion of the L7 and L8 vertebrae.