VEGF-C was cloned as the ligand for VEGFR-3 (Joukov et al. 1996). Both receptor and ligand expression were suggestive for a role in lymphatic development (Kaipainen et al. 1995; Kukk et al. 1996). On the other hand, VEGF-C induced the proliferation and migration of blood vascular endothelial cells in vitro (Joukov et al. 1996; Lee et al. 1996). To identify its primary in-vivo target we created a transgenic mouse model. Expression of VEGF-C from the human full-length VEGF-C cDNA was driven by the keratin-14 promoter (Vassar et al. 1989) specifically in the basal keratinocytes of the skin epidermis.
The choice of the promoter was pragmatic: Expression is local, the skin is easily amenable to observation and manipulation, and the same promoter had been used by Detmar et al. (1998) to create a VEGF-overexpressing mouse.
Transgenic mice with high expression levels could be easily recognized by their smaller size, edematous facial features and disturbed hair-growth. Beneath the epidermis large dilated sinus-like vascular structures devoid of blood cells could be detected. Using immunohistochemistry, electron microscopy and mRNA in-situ hybridization we identified the origin of these structures as lymphatic.
Fluorescence microlymphangiography of the tail showed that the superficial lymphatic network was not altered in its geometry, but that the vessel diameter had enlarged approximately twofold, probably caused by proliferation of the endothelial cells. Remarkably, no sprouting of lymphatics was detected, just like in pathological processes such as lymphangioma or lymphangiectasia. The reason for the absence of sprouting is still unclear. Previous data suggests that the extracellular matrix is instructive in lymphatic vascular proliferation (Clark and Clark 1932) and the extracellular matrix in the superficial lymphatic network might be non-permissive for sprout formation. Apart from VEGFR-3, other receptors were detected on these vascular structures, notably VEGFR-2 and Tie-1.
The high degree of lymphatic specificity was unexpected, particularly as VEGF-C can also activate VEGFR-2. The lack of appropriate proteases might be the cause, since VEGF-C needs to be processed in order to show significant binding to VEGFR-2 (Joukov et al. 1997). Supportive for this explanation is the fact, that in most systems the dominant product of proteolytic processing is the 29/31-kDa form, which binds VEGFR-2 only marginally. Unfortunately, we were not able to demonstrate which form of the protein was predominant in the transgenic mouse skin. The transgene expressed the human protein and although human VEGF-C does bind mouse VEGFR-2, it might do so with a low affinity or it might be less potent due to different receptor internalization kinetics. Lee et al. were unable to immunoprecipitate VEGF-C using mouse VEGFR-2 (1996). There can be considerable interspecies differences of receptor binding affinities and kinetics: human VEGF-D binds VEGFR-2, while the mouse orthologue does not (Baldwin et al. 2001a).
No rapid in-vivo lymphangiogenesis assay has been devised to date. Using the newly discovered VEGF-C, this study characterizes the lymphatics of the CAM and demonstrates its suitability as an in-vivo model for lymphangiogenesis.
To produce the mature form of human VEGF-C, the cDNA coding for the VHD of VEGF-C (nucleotides 658-996, EMBL accession number X94216) was cloned into a baculoviral transfer vector in between the sequences coding for the honey bee melittin signal peptide (Tessier et al. 1991) and a hexahistidine tag. Protein was isolated from serum-free conditioned supernatant of HighFive cells infected with recombinant baculovirus using Ni2+ nitriloacetic acid affinity chromatography.
By evaporation of a sessile drop a radial protein concentration gradient (Deegan 2000) was created on a thermanox cover slip and applied to the differentiated CAM (day 13). Although VEGF-C induced a weak angiogenic response, its main effect was on the lymphatics, which were very abundant in the entire application area. In the region of the highest growth factor concentration a huge lymphatic sinus formed, that could be retrograde-injected with Mercox resin from the lymphatic trunks of the allantoic stalk.
Similar to the results in I, VEGF-C induced proliferation of lymphatic endothelial cells. The existing vessels increased in size, fused with neighboring vessels and formed plexuses. It appeared as if new vessels formed directly under the chorionic epithelium, although the mechanism of formation remained obscure. The mature form of VEGF-C does not require proteolytic processing to allow VEGFR-2 binding. Considering that this form was used, most surprising was again the near absence of any angiogenic effect. Human VEGF-C does bind avian VEGFR-2 (Eichmann et al. 1998) and subsequently the same results have been obtained with avian VEGF-C (own unpublished data). Cao et al. have shown that VEGF-C is angiogenic in the CAM when applied between days 6 to 10.5 (1998). Formation of blood islands commences at E3 in the allantoic bud (Papoutsi et al. 2001), and angiogenesis is considerable until the CAM becomes fully differentiated at day 12. Endothelial labeling indexes peak between days 8 and 10 (Ausprunk et al. 1974). VEGFR-3 is still expressed by the allantoic blood vessels at embryonic day 6 (Eichmann et al. 1993), which equals approximately mouse day 14 (Butler and Juurlink 1987). It is intriguing that avian blood vessels continue expressing VEGFR-3, while mouse blood vessels of a comparable developmental stage have already downregulated it (Kaipainen et al. 1995; Kukk et al. 1996).
It is noteworthy in this context that so far all attempts to achieve a lymphatic phenotype in transgenic mice overexpressing the mature form of VEGF-C have failed (Yulong He, personal communication). It is unlikely, that the lack of the C-terminal tail interferes with correct folding, since biologically active mature VEGF-C can be produced from a truncated cDNA in a variety of expression systems (II; Joukov et al. 1997). The role of the C-terminus of VEGF-C is thus still enigmatic, apart from its apparent function of masking the VEGFR-2 binding epitopes. It does not stimulate tyrosine phosphorylation of VEGFR-2 and VEGFR-3 (Joukov et al. 1997), but might be responsible for the interaction with other receptors, like neuropilin (Makinen et al. 1999; Karkkainen et al. 2001). Based on its homology to BR3P it was suggested that it might regulate the bioavailability of VEGF-C (Joukov et al. 1996).
PlGF was not angiogenic in the CAM in this study. It is difficult to interpret this finding, since PlGF is apparently angiogenic in the early CAM; at least until embryonic day 8 (Ziche et al. 1997; Maglione et al. 2000). Although the avian VEGFR-1 has been cloned (EMBL accession number AB065372), VEGFR-1 specific avian VEGF homologues have not been reported to date.
Shortly after the discovery of VEGF-C, two groups reported the cloning of a close VEGF-C paralogue, which was named VEGF-D or c-fos-induced growth factor (FIGF). Despite the significant homology of FIGF to VEGF family members, Orlandini et al. chose fibroblasts instead of endothelial cells to test the mitogenic activity of VEGF-D (1996). The high homology of VEGF-D to VEGF-C suggested that VEGF-D might have a similar receptor binding profile and should be tested for binding to endothelial cell-specific receptors.
The truncated cDNA corresponding to the VHD of VEGF-D (nucleotides 651-998, EMBL accession number AJ000185) was cloned into a baculoviral transfer vector as described in II. Full length VEGF-D was expressed in a similar fashion using the endogenous signal peptide of VEGF-D (nucleotides 411-1472, EMBL accession number AJ000185). Conditioned medium was used for receptor phosphorylation experiments, and purified protein was prepared as described in II and used for bioassays. The truncated forms of VEGF-D used in this study corresponded only approximately to the mature endogenous form of VEGF-D, as both predictions of the N-terminal propeptide cleavage site later appeared to be wrong (Stacker et al. 1999a).
Both the truncated and the full-length form of VEGF-D bound to and stimulated tyrosine-phosphorylation of VEGFR-3. However, only the truncated form was able to activate VEGFR-2. The amounts of protein expressed from the full-length cDNA might have been too low to induce detectable receptor phosphorylation. The 29/31-kDa form of human VEGF-D does bind to VEGFR-2 (Stacker et al. 1999a) and baculoviral expression of VEGF-D in insect cells from full-length cDNA results - similarly to VEGF-C - in the uncleaved and the 29/31-kDa forms. Small amounts of the mature 21-kDa form can only be detected after prolonged periods (>72 h) of infection (Hu et al. 1997; own unpublished data). In vivo, both the mature and the 29/31-kDa from were detected from embryonic lung tissue (Stacker et al. 1999a), but VEGF-D might have undergone additional proteolytic processing during lysis and sample preparation.
Surprisingly, it was shown that the receptor binding profile of VEGF-D is not conserved between humans and mice: human VEGF-D binds both VEGFR-2 and VEGFR-3 while mouse VEGF-D fails to bind mouse VEGFR- 2 (Baldwin et al. 2001a). This and the fact that mice do not possess the short splice isoform of VEGFR-3 (Galland et al. 1993; Pajusola et al. 1993), but do posses an additional splice-isoform of VEGF-D (Baldwin et al. 2001b), points to potential inter-species differences in lymphangiogenic signaling.
At least in humans, the biochemical properties of VEGF-C and VEGF-D appear interchangeable. Their long forms heterodimerize (own unpublished data). Their expression pattern is overlapping, but not identical, and the keratin-14 transgenic mice show a virtually identical phenotype (Veikkola et al. 2001). Both of them are clearly involved in lymphangiogenesis, but it is not well understood how exactly they divide the labor.
The aim of this study was to perform a screen for structural elements involved in VEGFR-3 interaction, and to comprehensively dissect the effects mediated by individual VEGF receptors. Non-random DNA family shuffling was invented to create a library of mosaic molecules from the two prototype members of the VEGF family - VEGF and VEGF-C. All 512 mosaic molecules of the library were screened for their receptor binding profiles and the results were correlated with their composition to identify structural elements responsible for receptor specificity.
The results indicate that all VEGFs bind their receptors in a very similar fashion. Within the VEGF family of growth factors, specificity is achieved by a limited subset of structural elements of the receptor-binding interface. The bottom groove of VEGF contains critical amino acids that prevent its interaction with VEGFR-3. The corresponding specificity-determining element in VEGF-C is the N-terminal fragment, which prevents interaction with VEGFR-1. These elements are unlikely to contribute much to the binding energy, since by combining the permissive structural elements from both receptors (the N-terminal helix of VEGF and the bottom groove of VEGF-C) mosaic VEGFs able to interact with all three VEGF receptors could be obtained. This concept is schematically explained in Figure 4D.
A panel of 10 VEGF mosaic molecules was selected for detailed analysis, including receptor phosphorylation, cell survival using VEGF/Epo receptor chimeras, and in-vivo (lymph)angiogenesis on the chick CAM. To exclude influences of the heparin binding domain of VEGF or the C-terminal domain of VEGF-C, all of the mosaic molecules were applied as minimal peptides of 109 amino acids. VEGFR-2 activation proved sufficient to induce angiogenesis and VEGFR-3 activation sufficient to induce lymphangiogenesis. VEGFR-1 specific mosaics and VEGF-B did not induce any obvious biological effects. The mosaic molecules that showed binding to both VEGFR-1 and VEGFR-3 induced lymphangiogenesis. Their potency though was lower when the affinity towards VEGFR-1 increased, indicating that VEGFR-1 acts also for these artificial molecules as a decoy receptor. None of the mosaic molecules was as potent as the control proteins in the CAM assay and the survival assay using VEGFR-2/EpoR and VEGFR-3/EpoR chimeras. To reach comparable effects in the CAM approximately 5-10 times more protein was applied, even for the most potent mosaics. In the cell survival assay the difference was even more pronounced with the exception of VEGFR-1-mediated cell survival, where several mosaic molecules showed the same potency as VEGF at similar concentrations. Some, but not all of the reduced angiogenic potency can be explained by the lack of the heparin binding domain: also VEGF121 is less potent compared to VEGF165 (Keyt et al. 1996a).
Usually several rounds of shuffling are performed to create molecules that surpass their parents in one feature (Chang et al. 1999). Although the results using one round of shuffling proved useful to identify structural elements of receptor interaction and specificity, additional optimization is clearly required to convert these molecules into useful research reagents.