Abnormal lymphatic vessel development in neuropilin 2 mutant mice

The vertebrate vascular system is composed of arteries, veins and capillaries, and develops early in embryonic development in order to accommodate the output of the first heart beat. The lymphatic vascular system serves to transport tissue fluid, extravasated plasma proteins and cells back to the circulation (for a review, see Karkkainen et al., 2002). Lymphatic endothelial cells (EC) are thought to develop by sprouting from veins in a process referred to as lymphangiogenesis (Sabin, 1909). Mesoderm-derived lymphangioblasts may also contribute to lymphangiogenesis (Wilting et al., 2000). Recent reports have focused on the role of members of the vascular endothelial growth factor (VEGF) family in the development of arteries, veins and lymphatic vessels. VEGF is a key regulator of EC functions and is required for the formation of the primitive vascular plexus (Carmeliet et al., 1996; Ferrara et al., 1996). The VEGF family also comprises placenta growth factor (PlGF), VEGFB, VEGFC and VEGFD. The biological effects of these growth factors are mediated by binding to three EC tyrosine kinase receptors, VEGFR1-VEGFR3 (for reviews, see Neufeld et al., 1999; Veikkola et al., 2000). VEGF binds to VEGFR1 and VEGFR2, while VEGFC and VEGFD bind to VEGFR2 and are the only known ligands for VEGFR3. The expression of VEGFR3 becomes restricted to lymphatic EC from mid-gestation stages onward both in murine and in avian embryos (Kaipainen et al., 1995; Wilting et al., 1997).

Several studies have shown that VEGF signals angiogenesis, while VEGFC and VEGFD induce lymphangiogenesis. Overexpression of VEGF in the skin of transgenic mice or on the chick chorioallantoic membrane stimulates growth of blood-vascular capillaries without affecting the lymphatic vessels (Detmar et al., 1998; Wilting et al., 1996). Conversely, overexpression of VEGFC in the same mouse or chick models stimulates lymphangiogenesis with only a weak effect on blood-vascular EC (Jeltsch et al., 1997; Oh et al., 1997). Lymphangiogenesis could be selectively inhibited by overexpression of a soluble VEGFR3 (Makinen et al., 2001). VEGFC and VEGFD have also been shown to stimulate metastasis via the lymphatic vessels (Karpanen et al., 2001; Mandriota et al., 2001; Skobe et al., 2001; Stacker et al., 2001). A role for VEGFR3 as the signaling receptor mediating VEGFC or VEGFD actions was suggested by overexpression of a VEGFC form that does not bind VEGFR2 (VEGF-C156S) in the skin of mice, which also stimulates lymphangiogenesis (Veikkola et al., 2001). Moreover, mutations in the tyrosine kinase domain of VEGFR3 have been linked to human hereditary primary lymphoedema, and mice harboring an ENU-induced mutation in the kinase domain of VEGFR3 also develop lymphoedema (Karkkainen et al., 2000; Karkkainen et al., 2001).

Several members of the VEGF family have been recently shown to bind to the non-kinase neuropilin (NRP) receptors. This small family of type I transmembrane proteins includes NRP1 (NRP – Mouse Genome Informatics) and NRP2. Neuropilins were initially discovered in the nervous system, where they function as receptors for the class III family of semaphorins (Sema) to mediate chemorepulsive guidance of developing axons (Chen et al., 1997; He and Tessier-Lavigne, 1997; Kolodkin et al., 1997). NRP1 and NRP2 show overlapping sema binding specificities in vitro: NRP1 binds with high affinity to Sema3a, Sema3c and Sema3f, while NRP2 is a high affinity receptor for Sema3c and Sema3f (Chen et al., 1997; Feiner et al., 1997). However, NRP1 function appears only necessary for Sema3a-mediated repulsive guidance events, both in vitro and in vivo, and genetic ablation of Nrp1 phenocopies the neuronal defects observed in Sema3a mutant mice (Kitsukawa et al., 1997). Gene inactivation of Nrp2 results in a distinct neuronal phenotype compared with the Nrp1 mutation (Chen et al., 2000; Giger et al., 2000). Homozygous Nrp2 mutants specifically lose their response to repulsive guidance events mediated by Sema3f. Fasciculation and guidance of distinct subsets of cranial nerves are perturbed in Nrp1 and Nrp2 mutants: Nrp1^–/– showed deficiencies in the cranial nerves VII, IX and X, which were not affected in Nrp2^–/– mice. Conversely, cranial nerves III and IV, which were normal in Nrp1 mutants, showed abnormal projections in Nrp2^–/– animals (Chen et al., 2000; Giger et al., 2000; Kitsukawa et al., 1997). Altogether, these results suggested specific, non-redundant functions for NRP1 and NRP2 in the nervous system.

Nrp1 mutant mice also showed severe defects in cardiovascular development, resulting in death of homozygous embryos by embryonic day (E) 14 (Kawasaki et al., 1999). Defects in vessel formation included failure of capillary ingrowth into the brain and abnormal formation of aortic arches and yolk-sac vasculature. These defects may be due to a requirement of NRP1 as a receptor for several members of the VEGF family, including the heparin-binding VEGF isoforms VEGF165 and VEGF145 as well as VEGFB, VEGFE and PlGF (for a review, see Neufeld et al., 2002). NRP1 enhanced VEGF165 binding to VEGFR2 and VEGFR2-mediated chemotaxis response of EC (Miao et al., 1999; Soker et al., 1998). In the developing avian vascular system, Nrp1 showed specific expression in arterial EC (Moyon et al., 2001; Herzog et al., 2001).

NRP2 has also been shown to bind several VEGF family members, including VEGF165, VEGF145, PlGF and VEGF-C (Karkkainen et al., 2001) (reviewed by Neufeld et al., 2002). A specific role for NRP2 in the vascular system has not been described previously. Indeed, many Nrp2 mutants are viable until adulthood and show a grossly normal cardiovascular system. We report that the formation of small lymphatic vessels and capillaries is abnormal in Nrp2 homozygous mice. In all tissues examined, including the heart, lung, diaphragm, gut and skin, these vessels are absent or severely reduced until postnatal stages. These observations show a selective requirement of NRP2 in lymphatic vessel development.

Nrp2 mutant mice in CD1 and C57B6 background were kindly provided by M. Tessier-Lavigne and maintained in a conventional animal facility. As previously described, Nrp2^+/– mice appear unaffected by the mutation, while Nrp2^–/– mice show deficiencies in the guidance and fasciculation of cranial nerves III and IV, defective development of the dorsal funiculus, absence or severe reduction of the anterior commissure and the habenular tract, and hippocampal mossy fiber projection defects (Chen et al., 2000). Adult Nrp2^+/– mice of each genetic background were intercrossed to maintain the colony and to obtain Nrp2^–/– animals. For RT-PCR, the procedure described by Chen et al. (Chen et al., 2000) was modified as follows: 2 bp were added on the 3′ end of primer Rc, specific primers (Rc and Rv) were used for the RT reaction and PCR was performed using ready-to-go PCR beads (Amersham Pharmacia Biotech).

In situ hybridization has been described previously (Moyon et al., 2001,) except that 6 μg/ml proteinaseK was used. Probes were a 1.2 kb mouse Nrp2 fragment (Chen et al., 1997), a 2.2 kb rat Nrp1 fragment (position 104-2388) (kindly provided by A. Chédotal) and a mouse Vegfr3 fragment (position 2411-4154).

Tissue biopsies were fixed in 4% paraformaldehyde, dehydrated and embedded in paraffin. Immunohistochemistry was performed using the Tyramide Signal Amplification system (TSA™, NEN™ Life Science Products, Boston, MA). Peroxidase activity was developed with 3-amino-9-ethyl carbazole (Sigma). Antibodies were a monoclonal anti-VEGFR3 (Kubo et al., 2000), biotinylated polyclonal anti-mouse VEGFR3, anti-NRP2 and anti-VEGFC (R&D Systems, Oxon, UK), PECAM (Pharmingen, San Diego, CA) and FITC-conjugated anti-smooth muscle actin (Sigma). X-gal staining was carried out as described (Puri et al., 1995).

We prepared serial 7.5 μm transverse sections between the neck and the tail of E17 embryos (n=5 for Nrp2^–/–, n=4 for Nrp2^+/–, n=2 for Nrp2^+/+), which were ordered such that each slide contained sections from embryos of each genotype at the same anteroposterior level. These slides were stained with anti-VEGFR3 antibody (see Fig. 5C,E,G) and examined under the microscope (×25 objective, final magnification ×110). For body skin, the area counted spanned the dermis of the dorsal half of the embryo. In initial counts, different anteroposterior levels of the body axis were counted separately, i.e. between the forelimbs, around the waist and between the legs; no significant differences in these regions were observed and the data were pooled. For counts of hindlimb skin, the area considered was the dermis covering the entire hindlimb. We counted manually on every fourth section the number of VEGFR3-positive lymphatic profiles, a profile was counted as one regardless of its size (i.e. Fig. 5C Nrp2^+/–; 11 profiles counted). The length of the dermis was measured, its thickness was similar in all embryos analyzed (see Fig. 5C-H) and was therefore not taken into account. No significant differences in embryos of the same genotype were observed, and data for embryos of the same genotype were pooled. The count thus represents the mean±s.d. of the sum of all vessel profiles per cm of dorsal half of the dermis. For counts of subserosal VEFGFR3-positive lymphatic profiles, we proceeded in the same way.

Pregnant females were injected intraperitoneally with 2 ml/100 g BrdU (Roche Diagnostics). After 6 hours, mice were sacrificed and embryos recovered for immunohistochemistry as described above. We used a biotinylated mouse anti-BrdU antibody (Pharmingen, San Diego, CA) and diaminobenzidine (Sigma). Slides were then incubated with biotinylated anti-VEGFR3 and developed after tyramide amplification with CY3.

FITC-dextran (2000 kDa, 8 mg/ml, Sigma) was injected intradermally into the ear and staining of the lymphatic network was followed by fluorescence stereomicroscopy.

We examined NRP2 expression during embryonic and postnatal stages (E10-P30) by in situ hybridization using a murine Nrp2 antisense riboprobe (Chen et al., 1997). In addition, we performed immunostaining with an anti-NRP2 antibody and X-gal staining of heterozygous Nrp2 mutant mice (Chen et al., 2000). These mice carry a gene trap insertion of a β-galactosidase reporter cassette fused to the CD4 transmembrane domain into an intron of the Nrp2 locus, producing a non-functional fusion protein retained in the endoplasmic reticulum. During embryonic development, we observed essentially the same Nrp2 mRNA expression pattern as previously described (Chen et al., 1997), including expression at E10 in the floor plate and neurons of the ventral region of the spinal cord, dorsal root and sympathetic ganglia, somitic tissue, and dorsal aspect of the forelimb (Fig. 1A). In addition, we observed expression of Nrp2 in the vascular system. E10 mouse embryos showed a signal in EC of the cardinal vein, but not the dorsal aorta (Fig. 1A,D). The Nrp1 receptor mRNA showed a complementary expression pattern in the dorsal aorta EC and in underlying mesenchymal cells, but was absent from the cardinal vein (Fig. 1B,E). Comparison with Vegfr3 expression (Kaipainen et al., 1995) showed that mRNA for this receptor had already become downregulated in arterial and venous EC at E10, but remained high in small vessels that appeared to branch from the cardinal vein (Fig. 1C,F).

During subsequent developmental stages, Nrp2 mRNA and protein expression in the cardinal vein decreased to very low levels (Fig. 2A,D). From E13 onwards, the only vessels strongly expressing NRP2 appeared to be lymphatics, as judged initially by their anatomical position: in the neck region, the major lymphatic vessels are formed in close proximity to the cardinal vein (Fig. 2). These vessels and their branches strongly expressed NRP2 (Fig. 2A,D). To precisely determine if the NRP2-positive vessels were lymphatics, we performed staining for VEGFR3, which is largely restricted to lymphatic EC (Kaipainen et al., 1995) and PECAM, which labels all EC but shows lower expression levels on mature lymphatic EC (Erhard et al., 1996). VEGFR3 and NRP2 both strongly labeled the jugular lymphatic vessels (Fig. 2A,B). Arteries did not express VEGFR3 or NRP2, but were strongly PECAM positive (Fig. 2A-C). Veins were PECAM positive, VEGFR3 negative and weakly labeled with NRP2 (Fig. 2C-E). This expression of NRP2 was maintained throughout development and persisted in the adult. Lymphatic EC of all tissues and organs examined (gut, mesenteries, skin, heart, lung, diaphragm) co-expressed NRP2 and the lymphatic EC markers VEGFR3 and podoplanin (Karkkainen and Alitalo, 2002) (Fig. 3). However, in all tissues and organs examined, NRP2 remained expressed at low levels in veins (Fig. 3A,D), which were not labeled with VEGFR3 or podoplanin antibodies (Fig. 3B,C,E-G). The expression of NRP2 therefore suggested possible functions in the development of veins and lymphatic vessels.

To study the function of NRP2 in vascular development, we used Nrp2 mutant mice previously described (Chen et al., 2000) (see Materials and Methods). The Nrp2 mutation was bred into two genetic backgrounds, C57B6 and CD1, both were analyzed in this study and results were indistinguishable. Genotyping was performed using RT-PCR essentially as described (Chen et al., 2000). Homozygous Nrp2 mutants were found to be viable, but were obtained at a reduced mendelian frequency. Of 328 mice analyzed between E13 and P30, the expected ratio of homozygous mutants was observed until P0 (175 animals: n=44 for Nrp2^–/–, n=86 for Nrp2^+/–, n=45 for Nrp2^+/+). Mice older than P1 were obtained at a reduced ratio (153 animals, n=29 for Nrp2^–/–, n=77 for Nrp2^+/–, n=47 for Nrp2^+/+). The reduced ratio of homozygotes was most probably due to the death of some animals between P0 and P1: 11/29 homozygous mice tested at P0 and P1 were dying. Death of these animals appeared unrelated to their vascular phenotype, which was similar in the dying and surviving mutants (see below). Additional causes for lethality may therefore exist, perhaps related to the neuronal phenotype of these mice (Chen et al., 2000; Giger et al., 2000). The surviving homozygous mutants could live until at least P30; however, they were smaller than their littermates (about 60% of body weight compared with Nrp2^+/– or Nrp2^+/+ littermates). A difference in size was apparent from about P8 onwards. This phenotype had not been reported previously.

Different tissues of Nrp2 mutants were examined on sections and by whole-mount X-gal staining. Serial transverse sections of E13 and E15 embryos (n=5 for Nrp2^–/–, n=10 for Nrp2^+/–, n=3 for Nrp2^+/+) were stained with PECAM antibodies, which revealed no differences in blood vessels and capillaries between Nrp2^–/– and Nrp2^+/– or Nrp2^+/+ animals (Fig. 4A-C). Staining with NRP2 and VEGFR-3 antibodies showed that jugular lymphatic vessels (Fig. 4D-I) as well as lymphatics of the thoracic duct were present and appeared normal. However, striking differences in the formation of lymphatic capillaries were noted between Nrp2^–/– and Nrp2^+/– or Nrp2^+/+ embryos from E13 until P3. In E13 Nrp2^+/+ and Nrp2^+/– embryos, numerous VEGFR3- and NRP2-positive capillaries were present in the dermis (Fig. 4D,E,G,H). These capillaries were absent or strongly reduced in number in the Nrp2^–/– littermates (Fig. 4F,I). NRP2 staining of sections of Nrp2^–/– embryos confirmed their genotype: intracellular punctuate staining was observed instead of membrane staining, reflecting the retention of mutant protein in the endoplasmic reticulum (Fig. 4I).

To determine if the NRP2 mutation affected lymphatic capillary formation in the skin, we performed whole-mount X-gal staining of the skin of E15 embryos (n=8 for Nrp2^–/–, n=13 for Nrp2^+/–). An almost complete absence of X-gal-positive vessels was observed in the dorsal skin overlying the neural tube in Nrp2^–/– animals, while the corresponding skin of Nrp2^+/– mice showed a readily visible capillary plexus (Fig. 5A,B). We next prepared serial transverse sections between the neck and the tail of E17 embryos (n=8 for Nrp2^–/–, n=8 for Nrp2^+/–, n=5 for Nrp2^+/+), which were ordered such that each slide contained sections from Nrp2^+/+, Nrp2^+/– and Nrp2^–/– embryos. Sections of the skin of newborn mice were also prepared (n=12 for Nrp2^–/–, n=16 for Nrp2^+/, n=7 for Nrp2^+/+). These slides were stained with antibodies recognizing VEGFR3, PECAM and smooth-muscle actin (SMA). In all cases except one (P0), a reduction of the number of VEGFR3-positive vessels was observed in Nrp2^–/– mutants as compared to Nrp2^+/– or Nrp2^+/+ skins (Fig. 5C,E,G). By contrast, PECAM staining did not reveal a reduction in the number of dermal blood-vascular capillaries, arteries or veins (Fig. 5D,F,H). In all mice analyzed, PECAM staining of the VEGFR3-positive lymphatic vessels in the skin was less intense than staining of arteries or veins (Fig. 5E,F). In contrast to arteries and veins, VEGFR3-positive lymphatic vessels were not surrounded by SMA-positive cells (Fig. 5I,J). The VEGFR3-positive lymphatic vessels that formed in the Nrp2^–/– mice sometimes appeared enlarged and were present at the border of the dermis and the subcutaneous tissue, when compared with Nrp2^+/+ and Nrp2^+/– littermates, where most of them were present roughly in the middle layer of the dermis (Fig. 5C,E,G).

The number of VEGFR3-positive lymphatic vessels present in the skin was counted on transverse serial sections of E17 embryos (n=5 for Nrp2^–/–, n=4 for Nrp2^+/–, n=2 for Nrp2^+/+) (see Materials and Methods). No significant difference in lymphatic vessel number was observed in Nrp2^+/+ compared with Nrp2^+/– skins (vessel number per cm sectioned body skin 20.6±1.1 in Nrp2^+/+, 19.4±1.1 in Nrp2^+/–; hindlimb skin 31.7±5.3 in Nrp2^+/+, 32.4±2.0 in Nrp2^+/–). In Nrp2^–/– embryos, the lymphatic vessel number was reduced three times in the body skin (vessel number per cm=6.1±1.7) and 1.8 times in the skin of the limb (18.0±2.3). To monitor the proliferation of lymphatic EC, we measured BrdU incorporation into the lymphatic vessels of the body skin of E17 embryos (n=5 for Nrp2^–/–, n=3 for Nrp2^+/–, n=1 for Nrp2^+/+). In the Nrp2^+/+ and Nrp2^+/– mice, 23.9±3.7% of VEGFR3-positive cells had incorporated BrdU (Fig. 6A,C), while only 13.6±3.4% of VEGFR3-positive cells were also BrdU positive in the Nrp2^–/– mutants (Fig. 6B,D). Thus, a reduction of about 1.8-fold was observed in the proliferation of skin lymphatic vessels in E17 Nrp2 mutants. By contrast, no obvious decrease in the number of BrdU-positive cells in the epidermis (Fig. 6A,B) or other tissues was detected in the Nrp2^–/– animals compared with Nrp2^+/– or Nrp2^+/+.

Whole-mount X-gal staining also revealed the absence or strong reduction of the number of lymphatic capillaries at the surface of the heart in the Nrp2^–/– compared with Nrp2^+/– littermates (Fig. 7A,B). Staining of sections prepared from hearts with VEGFR3 antibody confirmed that the X-gal stained vessels were lymphatics (Fig. 7C). The onset of the phenotype correlated with the onset of the development of these vessels: X-gal-positive lymphatic capillaries progressively covered the heart surface from E15 onwards in Nrp2^+/– mutants, while in Nrp2^–/– embryos, only collecting lymphatic trunks surrounding the great vessels were present at all developmental stages examined (Fig. 7A,B).

In the periphery of the diaphragm, a striking reduction of the number of X-gal-positive capillaries was observed from E15 onwards (Fig. 7F,G). Sectioning of X-gal-stained diaphragms and labeling with VEGFR3 antibodies confirmed that these capillaries were lymphatics (Fig. 7H,I). The onset of this phenotype coincided with the onset of the development of these capillaries between E15 and E17. Larger collecting lymphatics in the central portion of the diaphragm were present in the Nrp2^–/– animals and appeared normal (Fig. 7F,G). Some weaker X-gal-stained vessels, which may correspond to veins or to larger collecting lymphatic vessels, were present in all Nrp2^–/– mutants (Fig. 7G).

Altogether, of 36 Nrp2^–/– hearts and diaphragms examined between E15 and P3, 32 hearts showed an absence of lymphatic capillaries, the remaining four resembled Nrp2^+/– mice. All diaphragms showed absence of lymphatic vessels; so all the 36 Nrp2^–/– mice could be clearly distinguished from their Nrp2^+/– littermates by X-gal staining of these two tissues. Hearts and diaphragms of Nrp2^+/– mice (n=106) always showed a similar staining pattern (Fig. 7A,F). Staining with an antibody recognizing VEGF-C revealed labeling of the superficial epicardium in both Nrp2^–/– and Nrp2^+/– mice (Fig. 7D,E), suggesting that the observed decrease in lymphatic vessel density was not due to decreased levels of this growth factor.

To analyze a possible phenotype in the lung, we performed X-gal staining, VEGFR3 whole-mount immunohistochemistry of lung slices (not shown) and staining of transverse lung sections (n=3 for Nrp2^–/–, n=3 for Nrp2^+/– sectioned at E15; n=1 for Nrp2^–/–, n=1 for Nrp2^+/–, n=1 for Nrp2^+/+ at E17; n=4 for Nrp2^–/–, n=4 for Nrp2^+/–, n=2 for Nrp2^+/+ at P0) with antibodies directed against NRP2, VEGFR-3 and PECAM. Lymphatics of the thoracic duct were present in the Nrp2^–/– mice at all stages examined (Fig. 8A,B). However, lymphatic capillaries, numerous in Nrp2^+/– or Nrp2^+/+ lungs, were strongly reduced in Nrp2^–/– lungs (Fig. 8A,B,E-G). This phenotype was observed from E15 onwards and correlated with the appearance of VEFGR3-positive capillaries in the lungs at E15. PECAM staining did not reveal any differences in the blood-vascular EC of the different animals (Fig. 8C,D). At birth, lymphatic vessels surrounding the main bronchi had formed in the Nrp2^–/– lungs and were stained with anti-NRP2 and VEGFR3 antibodies (Fig. 8E-G). However, VEGFR3-positive capillaries were still absent in the Nrp2^–/– mutants (Fig. 8E-G). In the subserosal plexus of the gut, a reduction in the number of lymphatic vessels was observed in E15 and E17 embryos (Fig. 8H,I). Counting of the number of VEGFR3-positive vessels in the subserosal plexus confirmed this reduction: 5.2±0.4 VEGFR3-positive vessels per cm sectioned guts were observed in Nrp2^+/– and Nrp2^+/+ animals (n=4 sectioned at E17), while 2.3±0.7 vessels were counted per cm sectioned guts in Nrp2^–/– littermates (n=5).

Lack of lymphatic skin vessels is associated with hereditary or acquired lymphoedema (Karkkainen et al., 2000; Karkkainen et al., 2001). No signs of dermal thickening or oedema in the extremities of newborn, juvenile or adult Nrp2 mutant mice were, however, apparent. To determine the evolution of the lymphatic vessels in the skin of older Nrp2 mutants, we carried out a functional test based on the uptake of high molecular weight FITC-dextran into the lymphatics after intradermal injection into the ear or the tail of P7 (n=1 for Nrp2^–/–, n=1 for Nrp2^+/+) and 4-week-old mice (n=5 for Nrp2^–/–, n=4 for Nrp2^+/–, n=3 for Nrp2^+/+). No difference in dye uptake between different animals could be observed (Fig. 9A,B). Sectioning of ears showed the presence of VEGFR3 and NRP2 stained lymphatic vessels (not shown). Thus, growth of lymphatic vessels and capillaries occurred during postnatal life in the mutant mice.

X-gal or VEGFR-3 staining of other organs isolated from juvenile or adult animals confirmed this observation. We analyzed hearts and diaphragms from seven Nrp2^–/–, three Nrp2^+/– and four Nrp2^+/+ animals sacrificed between P7 and P12, as well as six Nrp2^–/–, five Nrp2^+/– and four Nrp2^+/+ animals sacrificed between P20 and P35. In the heart, four out of seven Nrp2^–/– mice between P7 and P12 showed very few lymphatic vessels, the remaining three resembled Nrp2^+/–. Older Nrp2^–/– mutants always showed lymphatic vessels, although mostly with an abnormal pattern (Fig. 9C,D). In the diaphragm, all juvenile and adult Nrp2^–/– mice showed lymphatic vessel growth (not shown). In the gut, the subserosal lymphatic plexus had formed in the Nrp2^–/– mice at P0, as judged by whole-mount immunohistochemistry with VEGFR3 antibody (not shown). Transport of lipids into the mesenteries occurred normally in the Nrp2^–/– mutants from birth onwards, as judged by visual inspection of milk uptake into the mesenteries after feeding (Fig. 9E,F). However, subserosal lymphatic vessels of adult mice were often enlarged (Fig. 9G,H). Lymph nodes were present in all mutant animals (not shown).

The data presented show a selective requirement for NRP2 in the formation of lymphatic vessels. Loss of NRP2 function resulted in the absence or severe reduction of the number of small lymphatic vessels and capillaries in all tissues examined, including the skin, gut, heart, diaphragm and lung. Lymphatic vessel number in all of these tissues was decreased as soon as these vessels developed in the embryo, and remained reduced or absent until after birth. BrdU incorporation studies revealed a reduction in the number of proliferating lymphatic EC in the skin, suggesting a requirement of NRP2 in the proliferation of these vessels. The lymphatic vessels present in the skin of mutant animals were also positioned abnormally and sometimes enlarged, suggesting that guidance defects may also occur in the absence of NRP2 function. Veins, which express lower levels of NRP2, developed normally in the Nrp2^–/– animals and PECAM staining revealed no difference in capillary formation between homo- and heterozygous mice, suggesting that the NRP2 mutation selectively affected the lymphatic compartment.

The mechanisms responsible for the development of lymphatic vessels are currently not well understood. The first detectable lymphatic vessels in the embryo are the thoracic duct and jugular lymph sacs, which are thought to develop by sprouting from veins (Sabin et al., 1909). Indeed, labeling with lymphatic-specific markers such as Prox1 (Wigle and Oliver, 1999) or VEGFR3 (Kaipainen et al., 1995) are suggestive of a sprouting process. NRP2 is likely to be downstream of these initial events in the formation of lymphatic EC, as its expression becomes downregulated in veins after that of VEGFR3. Wigle et al. (Wigle et al., 2002) have recently shown that in Prox1 mutant mice, the initial sprout formation from veins is not affected, but that sprout progression is impaired and that the lymphatic EC instead adopt a blood-vascular phenotype. How the remaining lymphatic vessels in the body develop is currently unknown. Quail-chick chimera experiments have suggested the participation of mesoderm-derived lymphangioblasts (Wilting et al., 2000), but the extent of their contribution to the developing lymphatic vasculature is currently unknown. Our observations of NRP2 staining suggest that there is a phase of lymphatic EC development from pre-existing lymphatic EC, analogous to the classically defined angiogenesis process. NRP2 staining carried out at different stages of development showed a progressive coverage of, for example, the heart and the diaphragm with lymphatic EC. These vessels did not appear to develop by sprouting from veins; they may be formed by mesoderm-derived lymphangioblasts, but the most likely explanation is that they are formed by sprouting from pre-existing lymphatic vessels. In homozygous NRP2 mutants, this sprouting process appears affected, while the formation of the lymphatic thoracic duct and other collecting lymphatic vessels occurs normally. It may therefore be speculated that NRP2 function is not required for sprouting of lymphatic vessels from veins, but for the sprouting of lymphatic EC from pre-existing lymphatic EC. The Nrp2 mutants analyzed here were previously reported to maintain very low levels of wild-type transcripts (Chen et al., 2000). Their neural phenotype was indistinguishable from a complete loss-of-function mutant generated by Giger et al. (Giger et al., 2000). The vascular phenotype of this second Nrp2 mutant has not been reported yet.

In the developing avian vascular system, expression of Nrp1 and Nrp2 has recently been shown to localize to arteries and veins, respectively (Moyon et al., 2001; Herzog et al., 2001). Preferential expression of NRP1 in developing retinal arteries has also been recently reported in mice (Stalmans et al., 2002). The observations reported here confirm and extend these findings: at early developmental stages, Nrp1 and Nrp2 are expressed in respectively arterial and venous EC. At later developmental stages, NRP2 mRNA and protein expression in veins decreases, but remains high in lymphatic EC, which can be also identified by expression of VEGFR3. The different expression patterns suggest non-redundant functions of NRP1 and NRP2 in the vascular system. Interestingly, inactivation of both Nrp1 and Nrp2 genes leads to embryonic lethality at E8.5 (Takashima et al., 2002). Embryos homozygous for one NRP mutation and heterozygous for the other NRP gene died at E10-E10.5 because of deficiencies in arterial and venous branching (Takashima et al., 2002). Taken together with the specific expression of Nrp1 and Nrp2 in respectively arterial and venous EC, these results suggest a requirement of both NRP genes for correct arterial and venous patterning during early remodeling of the primary vascular plexus. At later developmental stages, the Nrp1 mutation mainly affects the arterial compartment (Kawasaki et al., 1999), while we here report a specific requirement for NRP2 in lymphatic EC.

NRP2 is co-expressed with VEGFR-3 in lymphatic EC, although in mature vessels of adult skin its levels were decreased, and it has been shown to bind VEGFC (Karkkainen et al., 2001). It is therefore tempting to speculate that NRP2 may act as a co-receptor for VEGFR3 to mediate VEGFC-dependent lymphangiogenesis, analogous to NRP1, which enhances VEGF165 activity via VEGFR2. However, biochemical analysis of the signal transduction initiated by NRP2 has yet to be performed. Mutations in the tyrosine kinase domain of VEGFR3 lead to cutaneous aplasia of lymphatic vessels, causing the formation of edema in both mouse and humans (Karkkainen et al., 2000; Karkkainen et al., 2001). No signs of edema were apparent in NRP2 mutant mice, in spite of the reduction of lymphatic vessels in the skin and the internal organs. Larger collecting lymphatic vessels were, however, present in the mutant animals, which may be sufficient for tissue fluid drainage. Furthermore, from P7 onwards we observed growth of lymphatic vessels in the different organs of Nrp2^–/– animals. Regrowth of lymphatic vessels during postnatal development has also been observed in transgenic mice overexpressing a soluble VEGFR3 (Makinen et al., 2001). The precise mechanisms regulating embryonic and adult physiological and pathological lymphangiogenesis remain to be fully explored.

We thank M. Tessier-Lavigne for mice, A. Chédotal for probes, S. Gallier for animal care, P. Pourcelot for help with histology, members of the InsermU36 for discussion, and M. Tessier-Lavigne, A. Chédotal, J. L. Thomas, P. Corvol and J. Favier for reading the manuscript. This work was supported by grants from Association pour la Recherche contre le Cancer (ARC), Fondation pour la Recherche Médicale, Ministère de la Recherche (ACI) and Inserm (Avenir) to A. E. D. M. is supported by an ARC fellowship.