Cell adhesion mediated by VCAM-ITGa9 interaction enables lymphatic development

Abstract

Objective

Adhesive ligand-receptor interactions play key roles in blood vessel angiogenesis but remain poorly characterized during lymphatic vessel growth. In this study we use genetic approaches in both fish and mice to address the roles of cell surface integrin ligand VCAM and its two receptors, integrins a9 and a4, during lymphatic vascular development.

Approach and Results

Conditional deletion of the Vcam gene was used to test VCAM function in lymphatic growth in mid-gestation mice. Morpholino knockdown and cRNA rescue of the two zebrafish vcam alleles as well as integrins a9 and 4 were used to test the role of these ligands and receptors during lymphatic growth in the developing fish. We show that VCAM is essential for lymphatic development in the zebrafish embryo, and that integrin alpha9 (Itga9) rather than Itga4 is the required VCAM receptor in the developing fish. VCAM is expressed along lines of lymphatic migration in the mouse intestine, but its loss only retards lymphatic growth.

Conclusions

These studies reveal an unexpected role for cell-cell adhesion mediated by Itga9-VCAM interaction during lymphatic development in the fish but not in the mouse. We propose that the relative importance of cellular adhesive ligands is magnified under conditions of rapid tissue growth when cell number increases faster than cell matrix, such as in the early zebrafish embryo.

Introduction

Lymphatic vessels develop in all vertebrates where they function to drain fluid and cells that leak from the closed blood vasculature into the interstitium ¹. Lymphatic development has been studied in detail using genetic models in the mouse and fish ². These studies reveal that lymphatic endothelial cells arise from venous blood endothelial cells and then migrate away from blood vessels to form primary lymphatic structures that sprout to create a network of lymphatic vessels. In the mouse this primary structure is the lymph sac ³, while in the fish it is a line of parachordal lymphangioblasts ⁴. In both fish and mice lymphatic vessels arise in response to conserved molecular cues such as the secreted factors VEGF-C and CCBE1 ^5–10. However, the molecular mechanisms by which lymphatic endothelial cells (LECs) rapidly migrate to generate this network remain incompletely understood.

The development of the lymphatic vascular network requires endothelial cell proliferation and migration, e.g. in response to chemotactic factors that drive directional cell movement ¹¹. The physical movement of endothelial cells requires the engagement and release of adhesion receptors, especially integrins that are dynamically affinity modulated by extracellular signals such as those chemokines. Integrin-dependent adhesion and cell anchoring is also necessary for endothelial cell survival ¹². Blood vessels form in matrix-rich tissues and endothelial cell adheions and migration studies performed ex vivo utilize conditions that favor matrix production; thus the adhesive ligands that bind endothelial cell integrins in vivo are thought to be matrix proteins such as collagen and fibronectin ^{13, 14}. Consistent with this model of endothelial adhesion, genetic studies in mice have demonstrated endothelial cell requirements for integrin receptors during vascular development and post-natal angiogenesis (reviewed in ¹⁵). In contrast, the role of cellular integrin ligands such as VCAM remain restricted to leukocyte extravasation from the blood ¹⁶ and specific requirements for cell-cell adhesion during cardiac and placental development ^{17, 18}.

Less is known regarding the adhesion receptors and ligands that drive lymphatic vessel development and growth. Itga9 interaction with the EIIIA form of fibronectin is required for the formation of lymphatic valve leaflets ^19–21, and the broadly utilized fibronectin receptor Itga5 is required for lymphovenous valve development ²². Whether integrins play a more generally significant role during lymphatic endothelial cell migration and, if so, what ligands they utilize are not known. Blockade of integrin function is a promising anti-angiogenic strategy for treatment of cancer ¹⁵, thus identifying the integrin receptor-ligand interactions that participate in LEC migration might also prove valuable for anti-lymphangiogenic therapies.

To address the role of integrin receptors and ligands during lymphatic development we have examined integrins a4 and a9 that have been implicated in postnatal lymphatic growth ²³ and lymphatic valve development ²⁰ respectively, and their ligands fibronectin and VCAM1. We find that both VCAM and fibronectin are expressed along the path of migrating LEC in the developing mouse intestine but that loss of VCAM only slightly delays lymphatic network formation in that context. In contrast, loss of VCAM completely blocks lymphatic development in the zebrafish embryo. Endothelial Itga4 is not required for blood or lymphatic vessel development in the mouse ²⁴, and we find that it is also dispensable in the developing fish. Instead, loss of Itga9 blocks lymphatic development in a manner identical to that observed with loss of VCAM, and genetic studies reveal synergistic phenotypes with partial loss of both VCAM and Itga9 consistent with function as a critical receptor-ligand interaction. These studies reveal an unexpected role for integrin-mediated cell-cell adhesion during lymphatic development that is essential in the rapidly developing fish embryo but dispensable in the more slowly growing mouse embryo. There may also exist postnatal or other developmental contexts in which VCAM and cell-cell adhesion play key roles during lymphatic growth.

Materials and Methods

Materials and Methods are available in the online-only Data Supplement.

Results

VCAM and fibronectin are expressed along the path of lymphatic endothelial growth in the developing intestine

During lymphatic development in the mouse LECs reach the intestine from the mesentery at E15.5 (Fig. 1A–B, G–H) and subsequently migrate through the submucosa of the intestine wall to the villi to form a complete lymphatic vascular network ²⁵. To assess whether both cell-associated and matrix integrin ligands might participate in LEC migration during this process we first stained for fibronectin and VCAM1 and LECs in the developing mouse intestine. At E16.5, a timepoint at which LYVE1+ LECs have entered the submucosa but not yet migrated through it to reach the villi, VCAM expression was detected in both the submucosa and the villus (Fig. 1C, D). By E18.5, LYVE1+ LECs could be detected in the villi where they were surrounded by VCAM+ cells (Fig. 1E, F). Staining for fibronectin revealed abundant protein in the submucosa and villus at both E16.5 and E18.5 (Fig. 1I–L). Thus both cell surface and matrix integrin ligands are present along the path of LEC growth in the developing intestine.

Figure 1. VCAM and Fibronection are expressed along the path of LEC migration in the developing mouse intestine.

(A, B) VCAM+ cells are detected in the submucosa prior to the arriveal of LYVE1+ LEC in the mouse intestine at E15.5. (C, D) VCAM+ cells are detected in the submucosa and villi at E16.5, a timepoint at which LYVE1+ LEC are found in the submucosa. (E, F) LYVE1+ LECs are surrounded by VCAM+ cells in both the submucosa and villi of mouse intestine at E18.5. (G–L) Fibronectin is abundant throughout the submucosa and villi at E15.5 (G, H), E16.5 (I, J) and E18.5 (K, L). Boxed regions are shown at higher magnification in the below. Data shown one representative of 3 experiments performed.

Loss of VCAM retards lymphatic growth in the intestine but not skin of the developing mouse

Almost all VCAM-deficient animals die prior to E15 due to requirements for VCAM-Itga4 adhesion in the heart and placenta, but rare surviving VCAM knockout animals have been described that live to adulthood without overt phenotypes ^{17, 18, 26}. The lack of edema in rare surviving VCAM-deficient animals could indicate that VCAM is not essential for lymphatic growth, or merely reflect the ability of some animals to escape VCAM-deficient phenotypes due to a favorable genetic background. To more rigorously assess the requirement for VCAM during lymphatic development we generated Ub-CreERT2;Vcam1^fl/− animals and activated Cre activity in utero with maternal tamoxifen administration starting at E12.5 (Supp. Fig. 1A). Analysis of E16.5 intestine from Ub-CreERT2;Vcam1^fl/− animals revealed virtually complete loss of VCAM protein assayed both by immunostaining of the intestine wall (Supp. Fig. 1B, C) and immunoblot analysis of total intestine protein (Supp. Fig. 1D). At E16.5 Ub-CreERT2;Vcam1^fl/− animals exhibited 75% fewer LYVE1⁺ Prox1⁺ LECs in the submucosa of the intestine compared with Ub-CreERT2;Vcam1^fl/+ and Vcam1^fl/+ control littermates (Fig. 2A–E), and had not yet formed larger lymphatic vessels (Fig. 2A–D). At E18.5, the number of lymphatic vessels visualized by anti-LYVE1 wholemount staining of the intestine wall was reduced by approximately 50% in the proximal intestine and by >75% in the distal intestine in Ub-CreERT2;Vcam1^fl/− animals compared with littermate controls (Fig. 2F–K). By P0, however, the number of lymphatic vessels in both the proximal and distal intestine did not differ significantly between Ub-CreERT2;Vcam1^fl/− and control littermates (Fig. 2L–Q). The loss of VCAM therefore retards but does not completely block the migration of LECs into the intestine and the formation of the gut lymphatic network, an observation consistent with overlapping roles for cell-associated and matrix integrin ligands in this context. This possibility is supported by the observations that prenatal deletion of VCAM had no effect on lymphatic development in the matrix-rich dorsal skin (Supp. Fig. 2A–D), and postnatal deletion of VCAM did not affect lymphatic growth in the ear (Supp. Fig. 2E–H), a matrix-rich organ that develops after birth.

Figure 2. Inducible loss of VCAM in mid-gestation retards lymphatic growth into the mouse intestine.

(A–D) Staining for LYVE1+;PROX1+ LECs reveals fewer lymphatic endothelial cells and vessels in the E16.5 Ub-CreERT2;Vcam1^fl/− intestine. Arrows indicate individual PROX1+ nuclei in LECs. Boxes indicate regions in A and C that are shown at higher magnification in B and D. Images were obtained with a Nikon Eclipse 80i microscope using a 20x/0.50 or 40x/0.75 numeric aperature (NA) dry objective. (E) Quantitation of LEC number in Ub-CreERT2;Vcam1^fl/− and Vcam1^fl/+ control littermate intestines. N=5 for Vcam1^fl/+ and N=3 for Ub-CreERT2;Vcam1^fl/−. ** indicates P<0.01. (F–I) Wholemount staining for LYVE1+ lymphatic vessels in Ub-CreERT2;Vcam1^fl/− and control animals at E18.5 in the proximal (F, G) and distal (H, I) intestine. Images were acquired using a Leica TCS SP8 inverted microscope (Leica Microsystems, Wetzlar, Germany) with 10x dry objective and assembled and analyzed using ImageJ (NIH). (J, K) Quantitation of lymphatic vessel number in Ub-CreERT2;Vcam1^fl/− and Vcam1^fl/+ control littermates in the E18.5 proximal and distal intestine. N=4; ** indicates P<0.01. (L–O) Wholemount staining for LYVE1+ lymphatic vessels in Ub-CreERT2;Vcam1^fl/− and control animals at P0 in the proximal (L, M) and distal (N, O) intestine. Images were acquired using a Leica TCS SP8 inverted microscope (Leica Microsystems, Wetzlar, Germany) with 10x dry objective and assembled and analyzed using ImageJ (NIH). (P, Q) Quantitation of lymphatic vessel number in Ub-CreERT2;Vcam1^fl/− and Vcam1^fl/+ control littermates in the P0 proximal and distal intestine. N=5;

VCAM is required for lymphatic development in zebrafish embryos

The finding that VCAM contributes to lymphatic development in the intestine but not the ear in the mouse suggested that the role of cellular integrin ligands may be context-dependent and augmented when matrix ligands are less abundant. The lymphatic vasculature develops earlier in the zebrafish embryo than in the mouse embryo, suggesting that matrix may be less available and LECs may utilize cell-cell adhesion to a greater extent than in the mouse. Genome database analysis revealed two zebrafish genes that are syntenic with and exhibit similar levels of amino acid homology to mouse and chicken VCAM1 that were designated VCAM short (“VCAMs”) and VCAM long (“VCAMl”) (Supp. Figs. 3 & 4). VCAMs has fewer Ig domains and is similar to the single chicken VCAM orthologue, while VCAMl has a greater number of Ig domains and more closely resembles mammalian VCAM (Supp. Fig. 4). Morpholino knockdown of either VCAMl or VCAMs alone had no effect on lymphatic development, measured either at the parachordal lymphangioblast formation stage (52 hpf) or at the thoracic duct formation stage (5 dpf) using fli1a:eGFP;flk:mCherry double transgenic embryos in which blood endothelial cells express both GFP and mCherry while LECs express only GFP ²⁷ (Supp. Figs. 5 and 6 and Fig. 3).

Figure 3. VCAM is required for lymphatic development in zebrafish.

(A, B) Detection of parchordal lymphangioblasts (“PL”, indicated by arrows) in 52 hpf zebrafish embryos treated with control or VCAMl + VCAMs morpholinos (3ng+3ng). (C) Quantitation of parachordal lymphangioblast formation following injection of VCAMs and VCAMl morpholinos individually and together. ** indicates P<0.001. (D, E) Detection of the thoracic duct (“TD”) between the dorsal aorta (“DA”) and posterior cardinal vein (“PCV”) in 5 dpf embryos treated with control or VCAMl + VCAMs morpholinos (3ng+3ng). Images were acquired using a Leica TCS SP8 inverted microscope (Leica Microsystems, Wetzlar, Germany) with 20x dry objective and assembled and analyzed using ImageJ (NIH). (F) Quantitation of thoracic duct formation following injection of VCAMs and VCAMl morpholinos individually and together. ** indicates P<0.001. (G–J) Fish injected with VCAMl + VCAMs morpholinos (I, J) exhibit abdominal and peri-orbital edema compared with control morpholino injected fish (G, H). Images were acquired using an Olympus MVX10 microscope. (K) Parachordal lymphangioblast formation following injection with alternate morpholinos targeting VCAMl and VCAMs. (L) Thoracic duct following injection with alternate morpholinos targeting VCAMl and VCAMs. Numbers above each column indicate the number of embryos analyzed in each group.

Injection of low dose (3 ng each) of morpholinos targeting both VCAMs and VCAMl resulted in a nearly complete loss of parachordal lymphangioblast formation at 52 hpf (Fig. 3A–C) and thoracic duct formation at 5 dpf (Fig. 3D–F). Consistent with the loss of lymphatic function, VCAMs+VCAMl double morphant fish exhibited marked abdominal and peri-orbital edema at 7 dpf (Fig. 3G–J). To test the specificity of these findings we first assessed whether combinations of different morpholinos targeting VCAMs and VCAMl conferred similar defects in lymphatic vessel growth. Loss of parachordal lymphangioblast formation (Fig. 3K) and thoracic duct formation (Fig. 3L) was also observed with combined use of 2 distinct non-overlapping morpholinos directed against VCAMs and VCAMl, providing strong evidence for the specificity of the morpholino phenotypes.

Lymphatic defects conferred by vcam morpholinos are rescued by vcam cRNA injection but not p53 knockdown

Lymphatic phenotypes were conferred by low dose morpholinos designed to specifically knock down expression of the two zebrafish vcam alleles. The specificity of these findings is supported by the fact that phenotypes were observed only with simultaneous knockdown of both vcam alleles, but additional evidence that morpholino phenotypes are not due to off-target effects is required. A direct means of testing the specificity of MO-induced knockdown is to rescue the phenotype by injection of cRNA that encodes the target protein but is not blocked by the MO ²⁸. We co-injected VCAMl cRNA with morpholinos targeting VCAMl and VCAMs to test specific rescue of the lymphatic phenotype. Embryos injected with VCAMl cRNA alone did not exhibit developmental abnormalities or over-growth of parachordal lymphangioblasts (Fig. 4A–D). However, VCAMl cRNA significantly rescued parachordal lymphangioblast formation at 52 hpf (P<0.001, Fig. 4E–I), consistent with a specific requirement for VCAM during lymphatic development.

Figure 4. VCAM cRNA injection rescues lymphatic development in VCAM morphant fish embryos.

(A–D) Injection of 200pg vcaml cRNA does not interfere with parachordal lymphangioblast formation at 52 hpf. Arrows indicate GFP⁺ parachordal lymphangioblasts (“PL”). (E, F) Injection of 3ng of VCAMl MO1 and 3ng of VCAMs MO1 blocks parachordal lymphangioblast formation. (G, H) Co-injection of 200pg of vcaml cRNA rescues parachordal lymphangioblast formation following injection of VCAM morpholinos. (I) Quantitation of the rescue effect of vcaml cRNA in VCAM morphants on parachordal lymphangioblast formation at 52 hpf. Numbers indicate the number of embryos analyzed in each group. ** indicates P<0.001.

A common mechanism by which morpholinos confer off-target phenotypes is through p53 activation, and co-injection of morpholinos targeting p53 provides a second means of testing the specificity of morpholino phenotypes in zebrafish embryos ²⁹. To further address the specificity of the lymphatic phenotypes conferred by morpholino knockdown we co-injected p53 morpholinos. In contrast to vcam cRNA injection, knockdown of p53 failed to rescue lymphatic growth in vcam morphant embryos (Supp. Fig. 7), suggesting that loss of lymphatic growth is due to loss of vcam expression and is not an off-target morpholino effect.

The VCAM receptor Itga9 but not Itga4 is required for lymphatic development in zebrafish

The best characterized receptors for VCAM are integrin a4b1 that mediates leukocyte adhesion to endothelial cells ³⁰, and integrin a4b7 that functions primarily in B cells ³¹. VCAM-Itga4 interactions also play important roles during development, as Itga4-deficient mice exhibit cardiac and placental defects identical to those of VCAM-deficient mice ^{18, 26}. Importantly, a4b1 integrins have also been demonstrated to play important roles in postnatal lymphatic growth in mammals ²³. To identify the VCAM receptor required for lymphatic development in fish we therefore first tested loss of zebrafish itga4. Injection of 4ng of a splice morpholino directed against the splice donor of exon 6 of itga4 resulted in highly efficient knockdown of itga4 mRNA (Supp. Fig. 6C). However, itga4 morphant embryos exhibited normal lymphatic growth at both the parachordal lymphangioblast (Fig. 5A–B, I) and thoracic duct (Fig. 5C–D, J)stages, and were not edematous at 7 dpf (Fig. 5E–H). Embryos injected with a second splice morpholino targeting the splice donor of exon4 of itga4 also failed to exhibit a defect in lymphangiogenesis (data not shown). As in VCAM morphants, we observed no defects in cardiac development or pericardial edema in itga4 morphant embryos (e.g. Fig. 5F). These studies suggest that alpha 4 integrins are not the VCAM receptors required for lymphatic growth and development in fish.

Figure 5. itga4 is not required for lymphatic development in zebrafish.

(A, B) Parachordal lymphangioblast formation at 52 hpf proceeds normally in zebrafish embryos treated with 4ng of itga4 splice morpholino. (C, D) Thoracic duct formation in 5 dpf embryos treated with control or itga4 morpholino. (E–H) itga4 morphant embryos do not exhibit abdominal or peri-orbital edema. (I, J) Quantitation of parachordal lymphangioblast and thoracic duct formation in zebrafish embryos following injection of 6ng control and escalating doses of itga4 morpholinos. Numbers indicate the number of embryos analyzed in each group.

A second, less characterized receptor for VCAM is the integrin a9b1. a9b1 has been reported to mediate neutrophil adhesion to VCAM on activated endothelial cells ³², but affinity studies reveal a much lower affinity for VCAM compared with a4b1 ³³. In mice the in vivo role of Itga9 is tightly linked to the EIIIA fibronectin ligand as loss of either results in defective lymphatic valve formation, but otherwise normal lymphatic vascular growth and development ²⁰. Injection of two distinct morpholinos targeting the ATG and the exon 2 splice acceptor of zebrafish itga9 (Supp. Fig. 6D) resulted in a severe loss of parachordal lymphangioblasts (Fig. 6A–B, I and K)and thoracic duct (Fig. 6C–D, J and L)formation similar to that seen with loss of VCAM. Consistent with these findings, itga9 morphant fish also exhibited abdominal and peri-orbital edema at 7 dpf (Fig. 6E–H). These data reveal that itga9 is required for lymphatic development in zebrafish.

Figure 6. Loss of itga9 blocks lymphatic development in zebrafish.

(A, B) Parachordal lymphangioblast formation at 52 hpf is blocked in zebrafish embryos treated with 4ng of itga9 splice morpholino. (C, D) Thoracic duct formation in 5 dpf embryos treated with control or itga9 morpholino. (E–H) itga9 morphant embryos exhibit abdominal or peri-orbital edema. (I, J) Quantitation of parachordal lymphangioblast and thoracic duct formation in zebrafish embryos following injection of 4ng control and escalating doses of itga9 morpholino #1. (K, L) Quantitation of parachordal lymphangioblast and thoracic duct formation in zebrafish embryos following injection of 4ng control and escalating doses of ATG morpholino against itga9 (itga9 morpholino #2). Numbers indicate the number of embryos analyzed in each group. * indicates P<0.05, ** indicates P<0.001.

VCAM and Itga9 exhibit genetic interaction during zebrafish lymphatic development

The studies described above suggested that VCAM and Itga9 may function as a ligand-receptor pair during lymphatic network formation in the fish. To test this hypothesis we first examined vcam and itga9 gene expression in zebrafish embryos using in situ hybridization. At 52hpf, vcaml is expressed in the brain while itga9 is expressed in the heart and fin bud (Fig. 7A, C). Strikingly, both vcaml and itga9 are expressed in the midline of the embryo, where the parachordal lymphangioblast forms, and between the somites, where the intersegmental lymphatic vessels form (Fig. 7B, D). Thus the spatial pattern of VCAM and Itga9 expression parallels that of lymphatic development in the fish.

Figure 7. Synergistic lymphatic-deficient phenotypes with loss of both vcam and itga9.

(A–B) In situ hybridization of vcaml in zebrafish embryos at 52hpf. (C–D) In situ hybridization of itga9 in zebrafish embryos at 52hpf. Boxed regions are shown at higher magnification on the right. Shown is a single study that is representative of five separate experiments. (E–L) Parachordal lymphangiobasts and thoracic ducts were visualized at 52 hpf and 5 dpf respectively following injection of 5 ng control (E, F), 1.5 ng vcams+1.5 ng vcaml (G, H), 2 ng itga9 (I, J) or 1.5 ng vcams+1.5 ng vcaml + 2 ng itga9 (K, L) morpholinos. (M, N) Quantitation of parachordal lymphangioblast and thoracic duct formation in zebrafish embryos following injection of the indicated morpholinos. Numbers indicate the number of embryos analyzed in each group. ** indicates P<0.001

We next examined genetic interaction between the vcam and itga9 genes in developing fish embryos using compound knockdown experiments with low doses of morpholinos targeting VCAMs, VCAMl, and itga9. Low dose morpholino treatment against VCAMs and VCAMl genes or itga9 alone did not impair parachordal lymphangioblast or thoracic duct formation (Fig. 7E–J and M–N). Combined knockdown, however, resulted in significant loss of both lymphatic structures, consistent with a synergistic effect on lymphatic growth (Fig. 7K–N). These studies support genetic interaction between VCAM and Itga9 during lymphatic development and suggest that this is a functionally important receptor-ligand interaction during lymphatic development.

Discussion

The lymphatic vascular network forms rapidly as LECs proliferate and migrate from their site of venous origin to tissues throughout the body ³⁴. Such rapid growth and invasion of pre-existing tissues requires factors that support LEC proliferation (e.g. VEGFC ^{5, 6} and CCBE1^7–10), guide LEC migration (e.g. CXCL12 and CXCR4 ¹¹), and mediate the physical movement of LECs. The adhesive ligands and receptors that enable LEC movement during this process of rapid migration and vessel growth remain obscure. Our findings reveal a critical role for the cell surface ligand VCAM and the integrin alpha9 receptor during zebrafish lymphatic development. In contrast, in the developing mouse our studies and published work reveal only supportive roles for VCAM and ITGa9 in lymphatic growth of the developing intestine and the formation of lymphatic valve leaflets respective. These studies therefore reveal an unexpected mechanism of cell-cell adhesion that is essential during fish but not mammalian lymphatic development.

A recent study has revealed that many morphant fish phenotypes are not reproduced by genetic mutants ³⁵, raising the question of whether and to what extent our findings accurately reflect roles for VCAM and ITGA9 in fish lymphatic development. Several lines of evidence suggest that the differences in the requirement for VCAM and ITGa9 observed in the fish and mouse reflect real species differences rather than artificial phenotypes in the fish conferred by morpholinos that act non-specifically. First, vcam and itga9 are expressed in a coordinate manner along the path of lymphatic endothelial migration in the developing fish. In contrast, in the developing mouse intestine VCAM and ITGA4 exhibited minimal co-expression at sites of lymphatic growth (Supp. Fig. 8A–D), while ITGA9 expression was not spatially associated with lymphatics (Supp. Fig. 8E–F). Thus expression and function in the two species differ in coordinate ways. Second, we used multiple morpholinos to target the vcam and itga9 genes in fish and observe identical phenotypes with distinct morpholinos injected at very low dosages. Third, the lack of any phenotype in fish treated with only a single vcam morpholino speaks to specific functional redundancy of the two vcam genes rather than an off-target morpholino effect that is typically conferred in a dominant manner. Fourth, the lymphatic phenotypes conferred by the morpholinos used in this study are not rescued by reducing p53 levels (Supp. Fig. 7), indicating that p53 activation, a common mechanism of off-target morpholino effects ²⁹, does not play a role in the genesis of these phenotypes. Fifth, loss of vcam and loss of itga9 in the fish result in identical defects in lymphatic growth but not venous sprouting that are distinct from defects in both observed with loss of either ccbe1 or vegfc (Supp. Fig. 9). Finally, loss of lymphatic growth conferred by vcam morpholinos is rescued by co-injection of morpholino-resistant vcam-encoding cRNA. Thus while future studies using fish genetic models will be needed to fully explore this biology, our expression and functional studies support the conclusion that VCAM and ITGa9 play essential roles in early zebrafish lymphatic development, and non-essential roles in mouse lymphatic development.

Endothelial cell adhesion and migration during angiogenesis is thought to be mediated by cell-matrix interactions that are dynamically regulated by integrin receptors in response to external signals such as chemokines and VEGF ^{36, 37}. Studies of the transmembrane protein VCAM1 and related proteins such as MADCAM have demonstrated that they function as cell surface ligands for integrin receptors (e.g. a4b1 and a4b7) that enable hematopoietic cells to bind endothelial cells and exit the blood ^{30, 31}. The possibility that VCAM-integrin interaction might also play important adhesive roles during development and in non-hematopoietic cells was raised by studies of VCAM-deficient and ITGa4-deficient mice that exhibited similar lethal defects in myocardial-epicardial adhesion and placental growth prior to E12.5 ^{17, 18, 26}. Our studies of mouse embryos in which VCAM has been inducibly and efficiently deleted after E12.5 reveal a delay in lymphatic vascular development in the gut, a result consistent with the co-expression of fibronectin along the path of LEC migration in the developing intestine and the lack of an overt vascular phenotyope in embryos in which Itga4 has been deleted selectively in endothelial cells ²⁴. Thus in the mouse endothelial cell-cell adhesion appears redundant with cell-matrix adhesion and is not required for lymphatic development. In contrast, loss of VCAM function in zebrafish embryos resulted in a complete block in lymphatic development that, in contrast to the defects conferred by loss of vegfc or ccbe1, has no detectable effect on blood vessel development (Supp. Fig. 7 and ^{6, 8}). The reason for this striking species difference in the utilization of VCAM is not yet clear, but one possibility is that lymphatic outgrowth from the blood vascular system occurs very early in fish compared with mammals; thus there may be less time for matrix protein synthesis and deposition prior to lymphatic outgrowth and a correspondingly increased reliance upon the cellular adhesive ligand VCAM and its receptor ITGA9.

While LECs in the developing fish may require VCAM for cell adhesion, a second possible explanation for the VCAM-ITGA9 requirement is endothelial survival and proliferation. Integrin-matrix interactions have been shown to be required during active angiogenesis both in vivo ³⁸ and in vitro ³⁹. To address this possible mechanism we examined lymphatic endothelial proliferation and apoptosis in the E18.5 intestine of VCAM-deficient and control mouse embryos. These studies failed to reveal significant changes with loss of VCAM (Supp. Fig. 10), suggesting either that LEC adhesion and migration are more important than cell survival or that the VCAM-deficient lymphatic phenotype is too subtle in the mouse to detect important changes in cell survival. Whether for cell migration or cell survival, it seems most likely that the greater role of VCAM in the fish embryo is due to the rapid timecourse of its vascular development that provides less opportunity to synthesize and utilize matrix proteins.

Finally, if the dependence of lymphatic growth upon VCAM-ITGa9 interaction in the fish is contextual, e.g. due to a relative lack of matrix at that early timepoint in development, it is also possible that there exist mammalian contexts in which VCAM and/or ITGa9 play critical roles in lymphatic growth. Recent studies of lymphatic growth in adult mice during tumor growth or matrigel lymphangiogenesis have revealed a requirement for ITGa4, an integrin subunit that is expendable during lymphatic development ²³. Whether lymphatic growth in these more aggressive postnatal contexts may also require VCAM and/or ITGa9 remains to be investigated. Anti-angiogenic agents directed against integrin avb3 show promise in the blood vascular system; thus similar strategies directed against adhesive receptor-ligand interaction may also provide therapeutic approaches to blocking pathologic lymphangiogenesis.

Significance.

Lymphatic vessels develop after the formation of the blood vascular network through rapid expansion of a small number of lymphatic endothelial progenitors that arise in veins. Understanding how lymphatic vessels develop in the vertebrate embryo is expected to provide insight that may be applied toward therapeutic lymphangiogenesis, e.g. in individuals with chronic lymphedema. Using both mouse and zebrafish models we establish a role for cell-cell adhesion during lymphatic growth that is mediated by the integrin a9b1 and its cell-associated ligand VCAM1. These studies highlight the role of adhesive interactions during lymphatic growth and demonstrate important differences between mammals and fish in their use during this process.