Arteriolar and venular patterning in retinas of mice selectively expressing VEGF isoforms

The murine VEGF gene is alternatively transcribed to yield the VEGF₁₂₀, VEGF₁₆₄, and VEGF₁₈₈ isoforms, which differ in their potential to bind to heparan sulfate and neuropilin-1 and to stimulate endothelial growth. Here, their role in retinal vascular development was studied in mice selectively expressing single isoforms. VEGF^164/164 mice were normal, healthy, and had normal retinal angiogenesis. In contrast, VEGF^120/120 mice exhibited severe defects in vascular outgrowth and patterning, whereas VEGF^188/188 mice displayed normal venular outgrowth but impaired arterial development. It is noteworthy that neuropilin-1, a receptor for VEGF₁₆₄, was predominantly expressed in retinal arterioles. These findings reveal distinct roles of the various VEGF isoforms in vascular patterning and arterial development in the retina.

The essential role of VEGF in blood vessel formation in the embryo and in the adult is well established (1, 2). A single VEGF gene gives rise, by alternative splicing, to multiple isoforms (VEGF₁₂₁, VEGF₁₆₅, and VEGF₁₈₉ in humans versus VEGF₁₂₀, VEGF₁₆₄, and VEGF₁₈₈ in the mouse) that differ in molecular mass, solubility, and receptor binding (3). VEGF₁₂₀ lacks exons 6 and 7, encoding extracellular matrix binding structures, and is therefore the most soluble. VEGF₁₈₈ contains all exons and avidly binds to the cell surface and extracellular matrix. VEGF₁₆₄ lacks only exon 6 and has intermediate properties. The VEGF isoforms bind to several receptors: VEGFR-1 (Flt-1), VEGFR-2 (Flk-1), and neuropilin-1 (NP-1) (3). NP-1 is a semaphorin receptor involved in neuron guidance. As a VEGF₁₆₄-specific coreceptor for VEGFR-2, NP-1 also affects angiogenesis (4) and enhances the angiogenic activity of VEGF₁₆₄ (5).

To define the differential role of the VEGF isoforms in vivo, mice expressing single VEGF isoforms were generated. Impaired myocardial angiogenesis has been demonstrated in VEGF^120/120 mice (expressing VEGF₁₂₀) (6). Here we report that loss of VEGF₁₆₄ (in VEGF^120/120 and VEGF^188/188 mice) impaired retinal arterial development, whereas loss of VEGF₁₆₄ and VEGF₁₈₈ (in VEGF^120/120 mice) led to dysregulated vessel outgrowth and patterning in the retina. These observations suggest possible mechanisms for the distinct roles of the VEGF isoforms in retinal vascular patterning and arterial endothelial cell specification.

Targeting of VEGF isoform-specific alleles. The VEGF gene is alternatively spliced to the VEGF₁₂₀ isoform (lacking exon 6 and 7), the VEGF₁₆₄ isoform (lacking exon 6), and the VEGF₁₈₈ isoform (encoded by all eight exons). Generation of mice selectively expressing single VEGF isoforms was achieved by Cre/lox P–mediated recombination in ES cells, using a “knock-in” strategy (see Methods; Figure 1). Quantitative RT-PCR analysis of brain RNA confirmed the selective expression of the respective VEGF isoform transcripts in each of these genotypes. Only the VEGF₁₆₄ isoform was detected in VEGF^164/164 mice (340 ± 46 VEGF₁₆₄ mRNA copies per 1,000 hprt copies, mean ± SD, n = 5) and the VEGF₁₈₈ isoform in VEGF^188/188 mice (770 ± 210 VEGF₁₈₈ transcript copies per 1,000 hprt copies, n = 5), while the other isoforms were undetectable (<0.3 VEGF₁₆₄ or VEGF₁₈₈copies per 1,000 hprt copies). Total VEGF transcript levels of the VEGF isoforms were comparable to those in VEGF^+/+ mice (total VEGF mRNA copies per 1,000 hprt copies: 250 ± 99 in VEGF^+/+, 230 ± 75 in VEGF^164/164, and 400 ± 120 in VEGF^188/188 mice; n = 5, P = not significant [NS]); the data in the VEGF^120/120 mice have been reported previously (6).

Inheritance and viability of VEGF-isoform mice. VEGF^+/120, VEGF^+/164, and VEGF^+/188 mice were normal, healthy, and had a normal life span. Offspring of VEGF^+/120 breeding pairs were born at a normal Mendelian frequency, but half of VEGF^120/120 neonates died shortly after birth because of cardiorespiratory distress (6). The other VEGF^120/120 mice died within 2 weeks after birth, in part due to impaired myocardial angiogenesis resulting in cardiac failure (6).

VEGF^164/164 mice were born at a normal Mendelian frequency. Of 366 live neonates born from VEGF^+/164 breeding pairs, 25% were VEGF^+/+, 51% were VEGF^+/164, and 24% were VEGF^164/164. They gained weight normally (32 ± 6 g in VEGF^+/+ mice, n = 8, versus 32 ± 4 g in VEGF^164/164 mice, n = 23, at 7 weeks; P = NS), were fertile (litters per breeding pair during 4 months: 4.8 ± 1.3 for VEGF^+/+ mice, n = 4, versus 4.8 ± 0.5 for VEGF^164/164 mice, n = 4; P = NS), and produced litters with normal size (11 ± 3 pups for VEGF^+/+ litters, n = 35, versus 10 ± 3 for VEGF^164/164 litters, n = 58; P = NS).

VEGF^188/188 mice were underrepresented at birth; out of 73 live-born pups, 24% were VEGF^+/+, 64% were VEGF^+/188, and only 12% were VEGF^188/188, indicating that a fraction of the VEGF^188/188 mice died in utero (P < 0.05 for observed vs. expected). Surviving VEGF^188/188 mice gained less weight than VEGF^+/+ mice at 7 weeks of age (32 ± 6 g in VEGF^+/+ mice, n = 8, versus 22 ± 3 g in VEGF^188/188 mice, n = 16; P < 0.05). VEGF^188/188 breeding pairs were less fertile (number of litters per breeding pair during 4 months: 4.8 ± 0.5 for VEGF^+/+ mice, n = 4, versus 3.7 ± 0.6 for VEGF^188/188 mice, n = 3; P < 0.05) and had smaller litter sizes (11 ± 3 pups for VEGF^+/+ litters, n = 35, versus 7 ± 2 pups for VEGF^188/188 litters, n = 53; P < 0.05).

Retinal vascular development in wild-type mice. Retinal vascularization was monitored by immunostaining of whole-mount preparations for CD31/PECAM or lectin for endothelial cells (ECs), smooth muscle alpha actin (SMA) for vascular smooth muscle cells (vSMCs) and pericytes (PCs), and GFAP for astrocytes (ACs). To study recruitment of periendothelial cells (PECs), VEGF-isoform mice were intercrossed with mice expressing LacZ in PECs (PEC^LacZ mice) (9). Arterial specification of ECs was examined by in situ hybridization for ephrin-B2 and by intercrossing VEGF isoform mice with mice harboring a LacZ gene under the control of the endogenous ephrin-B2 promoter (EB2^LacZ mice) (7).

Before birth, the murine retina is avascular and receives oxygen from the hyaloid and choroidal vessels. Postnatally, the increased retinal thickness and metabolism cause physiological hypoxia, which upregulates VEGF expression and initiates outgrowth of vessels from the optic disc (10). This results in the formation of a uniform capillary-like vascular labyrinth by postnatal day 2 (P2), which covers approximately 40% of the retina (not shown). Analysis of PEC^LacZ mice revealed that these primitive vessels were already covered by PECs, which, however, did not express SMA (not shown). Using whole-mount in situ hybridization, ephrin-B2 was detectable in 50% of the retinal vessels (e.g., in future arterioles) in an alternating pattern (not shown), indicating that arteriovenous specification had already occurred at this early stage of vascular development.

By P5, the vascular bed covered approximately 70% of the retina and had remodeled into a more mature network of branching vessels, which could be morphologically distinguished as arterioles, venules, and capillaries (Figure 2a). Arterioles had a small lumen and were surrounded by a capillary-free zone (a result of capillary pruning), whereas venules were wider and embedded in capillary-rich regions. Ephrin-B2 was expressed in retinal arterioles but not in venules (Figure 2, b and c). Analysis of PEC^LacZ mice revealed that PECs had spread along all vessels (Figure 2d), whereas expression of SMA was detectable in larger arterioles but not in venules or capillaries (Figure 2e). Transmission electron microscopy (EM) confirmed the presence of PCs along the retinal capillaries, which contracted during fixation (Figure 2f).

Figure 2

Retinal vascular development in VEGF^+/+ and VEGF^164/164 mice. (a) PECAM staining in VEGF^+/+ at P5. Capillaries reach two-thirds to three-quarters of the retinal radius. Arterioles (A) are thin and have a capillary-free zone. Venules (V) are broader and lack a capillary-free zone. La, arteriolar length; Lv, venular length; Lr, radius of retina; Lva, radius of vascular bed. (b and c) In situ hybridization for ephrin-B2 (EB2) (b) with or (c) without collagen type IV fluorescent immunostaining on VEGF^+/+ retina at P5 (EB2/CollIV). EB2 expression is confined to morphologically defined arterioles (see above) and extends into arterial capillaries (arrows). (d and e) Triple staining with LacZ (in PECs, blue dots), lectin (in ECs, red), and anti-SMA (in vSMCs, green) on a VEGF^+/+ × PEC^LacZ P5 retina (LacZ/Lec/SMA). (d) PECs have spread along all retinal vessels, whereas (e) only arterioles are SMA positive. (f) EM of VEGF^+/+ capillary at P5, which appears constricted. (g) PECAM staining on VEGF^+/+ P9 retina. (h) SMA staining in VEGF^+/+ at P14 is confined to arterioles and their side branches. (i) VEGF^+/+ × PEC^LacZ retina at 4 weeks. Hyaloid arteries (HA) have regressed (HA remnant, arrow). (j) Double staining for lectin (red) and GFAP (in ACs, green) (Lec/GFAP). AC network in front of vasculature. (k) PECAM staining of VEGF^164/164 retina at P5. Normal vascular development.

By P9, the vascular network had reached the periphery of the retina (Figure 2g). Arterioles and venules grew out over 88% ± 4% and 79% ± 6% of the retinal radius, respectively (Table 1). Venules were larger than arterioles, whereas capillaries had the smallest lumen (Table 1). On average, between five and six arterioles and venules were present per retina (Table 1). PECs covered arterioles, venules, and capillaries, resulting in a capillary EC-to-PC ratio (EC/PC ratio) of 5:1 (Table 1). However, only the primary and secondary side branches of arterioles expressed detectable SMA levels, while the venules and capillaries remained negative at P9 and beyond (Figure 2h).

Table 1

Characteristics of retinal vessels in VEGF-isoform mice

The hyaloid vasculature is present during development, but regresses once the developing retinal vasculature provides sufficient oxygen to the retina (15) (Figure 2i). Consistent with previous findings that the hyaloid vasculature contains only arteries (15), all hyaloid vessels in EB2^LacZ mice expressed ephrin-B2 (not shown). Hyaloid vessels were adjacent to the retina at P0 but did not make contact with retinal vessels and could easily be peeled away from the retina upon dissection of the retina from P0 up until their regression. They showed signs of regression by P9 and had disappeared by 4 weeks (Figure 2i).

Astrocyte development (ACs) have been implicated in guiding vessel outgrowth by producing a VEGF gradient (10). Whole-mount double immunofluorescent staining for GFAP and lectin revealed that spindle-shaped ACs were spreading across the avascular retina at P0.5 and had reached the periphery of the retina by P2–3 (not shown). ACs in these avascular regions expressed abundant VEGF (not shown). At later times they became organized into a network of tubes wrapped around outgrowing vessels (Figure 2j). Once the vasculature had reached the retinal periphery (P9),VEGF expression in the central vascularized regions was minimal.

Normal retinal vascular development in VEGF^164/164 mice. In VEGF^164/164 mice, the formation of a primitive labyrinth and subsequent remodeling was indistinguishable from that in VEGF^+/+ mice, resulting in similar numbers and sizes of arterioles, venules, and capillaries by P5 (Figure 2k, Table 1). Arterioles grew out over 89% ± 10%, capillaries over 99% ± 5%, and venules over 80% ± 12% of the retinal radius (Table 1). Ephrin-B2 and SMA expression (not shown), as well as PEC numbers and EC/PC ratios (Table 1), hyaloid regression, AC accumulation, and VEGF expression were comparable to VEGF^+/+ at all ages (not shown).

Impaired retinal arterial development in VEGF^188/188 mice. Retinal vessels. In mice expressing only VEGF₁₈₈, a normal capillary labyrinth containing PECs (not yet expressing SMA) developed by P3 (not shown). By P5, only half of the normal number of large vessels was present. These were morphologically identified as venules (Figure 3a), did not express ephrin-B2 (Figure 3, b and c), and failed to stain for SMA (Figure 3e), although PECs were present (Figure 3d), which contracted upon fixation (Figure 3f). At P6, ephrin-B2 was detectable in rudimentary retinal arterioles, suggesting that arterial specification was not lost (inset in Figure 3c). At P9, only half of the normal number of arterioles developed, grew out over only 17% ± 13% of the retina and were smaller in size (Table 1; Figure 3g). They were also covered by fewer LacZ-positive PECs (Table 1), which stained faintly for SMA at P9 and P14 (Figure 3h). In contrast to the arterial defects, venous and capillary development appeared largely normal. On average, there were five to six venules per retina, which grew out normally (Figure 3, a and g; Table 1), while capillaries grew out over 99% ± 2% of the retina at P9 (Table 1). The venules were somewhat enlarged and covered by PECs, which were positive for SMA (Figure 3h), possibly due to the arterial underdevelopment and resultant abnormal retinal perfusion. In addition, capillary pruning was decreased, resulting in a more dense capillary bed and a slight increase in the number of PECs per retinal area (Table 1; Figure 3, d and e), but in a normal EC/PC ratio (Table 1).

Figure 3

Retinal vascular development in VEGF^188/188 mice. (a) PECAM staining at P5. Capillary outgrowth comparable to VEGF^+/+ retinas, but no morphologically defined arterioles. (b and c) EB2/CollIV at P5. No EB2 expression at P5. Inset in c: VEGF^188/188 × EB2^LacZ retina at P6 shows small ephrin-B2–positive vessel (arrow). (d and e) LacZ/Lec/SMA on a VEGF^188/188 × PEC^LacZ retina at P5, revealing (b) high PC density, (c) but no SMA immunoreactivity. (f) EM at P5 shows normal capillary constriction. (g) PECAM staining at P9 with increased capillary density in the midperipheral retina (arrow) and partially removed hyaloid arteries. (h) SMA immunoreactivity is attenuated in arterioles, but has become positive in venules and capillaries (arrow) at P14. (i and j) VEGF^188/188 × PEC^LacZ retina at 4 weeks. (i) Persistence of hyaloid arteries and underdevelopment of arterioles. (j) Same retina after dissection of hyaloid arteries. Central part, but not peripheral part (arrows), of hyaloid arteries can be removed. (k) Lec/GFAP. ACs are organized into a tubular network well in front of the vascular bed.

Hyaloid vessels could not be easily dissected away from the retina and had established numerous connections with the retinal vascular network at P5. By P9, several hyaloid vessels had migrated into peripheral retina and established an arterial network (Figure 3g). Hyaloid arteries persisted up to 4 weeks of age (Figure 3i) and penetrated the inner limiting membrane distal to the site of maximal outgrowth of the rudimentary arterioles in the midretina (Figure 3, i and j). Possibly, the hyaloid vessels infiltrated the retina in response to the increased VEGF expression (not shown) that occurred as a result of insufficient arterial supply, thereby salvaging the hypoperfused periphery.

ACs migrated out normally to the retinal periphery by P2 (not shown) and had remodeled into a network that preceded the outgrowing vessels at P5 (Figure 3k). At P9, ACs accumulated in increased numbers at sites where hyaloid vessels made connections to the retinal vascular bed (not shown).

Impaired vascular outgrowth in VEGF^120/120 retinas. Severe retinal vascular defects were detected in VEGF^120/120 mice. Only a primitive vascular labyrinth of capillary-like vessels with uniform size developed by P5 (Figure 4a). Even though arteries and veins could not be morphologically distinguished, 50% of the retinal vessels expressed ephrin-B2 (Figure 4, b and c), suggesting that arteriovenous specification occurred before the remodeling stage. PEC recruitment was delayed; only scattered PECs were detected in and around the optic disc (Figure 4d), and retinas were still SMA negative at P5 (Figure 4e). EM illustrated that the sparsity of the PC resulted in a relaxed appearance of capillaries upon fixation (Figure 4f), as in diabetic retinopathy associated with PC dropout.

Figure 4

Retinal vascular development in VEGF^120/120 mice. (a) PECAM staining at P5. Dramatic impairment of vascular outgrowth. Arterioles and venules are morphologically indistinguishable. (b and c) EB2/CollIV at P5. Clusters of EB2-positive and -negative vessels alternate. (d and e) LacZ/Lec/SMA on VEGF^120/120 × PEC^LacZ retina at P5. (d) Sparse blue cells near optic disc. (e) SMA immunoreactivity is undetectable. (f) EM at P5 shows capillary relaxation. (g) PECAM staining at P9 with more pronounced impairment in arteriolar than venular outgrowth. (h) Arterial SMA staining is fainter, while venules and capillaries have become strongly SMA positive at P14. (i) VEGF^120/120 × Tie1^LacZ retina at P9, with persisting dilated and tortuous hyaloid arteries. (j) Lec/GFAP. AC organization into a network is present in the vascularized area, but rapidly alters to the primitive bipolar shape and parallel organization in the more peripheral retina (arrow). (k) Unstained whole-mount preparation at P9 with retinal hemorrhages (arrows).

In contrast to the other genotypes, the capillary bed in VEGF^120/120 mice covered the retina only over two-thirds in the small fraction of VEGF^120/120 mice that survived until P9 (Table 1). The vascular network had partially remodeled into morphologically recognizable arteries, capillaries, and veins (Figure 4g). To distinguish between a nonspecific general vascular defect and a specific arteriolar defect, outgrowth of arterioles was expressed as a percentage of the radius of the retina as well as the vascular bed. Arterial development was most significantly impaired; less than half of the number of arterioles grew out over only 35% of the retina and 53% of the vascular bed (Table 1; Figure 4g). Only half of the number of venules could be morphologically identified in VEGF^120/120 mice at P9, but venules had a normal lumen size and grew out over only 52% of the retinal radius, but over 80% of the vascular bed (Table 1; Figure 4g). There were more numerous and dilated capillaries in the central retina (Table 1). Fewer PECs surrounded the arterioles, venules, and capillaries, resulting in an increased EC/PC ratio (Table 1). The impairment was also reflected by a fragility of the capillaries, causing hemorrhages (Figure 4k). Possibly because of the increased hemodynamic stress, these PECs stained somewhat more strongly for SMA (Figure 4h). VEGF expression was markedly upregulated in the avascular retina, suggesting that the peripheral retina was highly ischemic (not shown).

No signs of hyaloid regression were observed in VEGF^120/120 mice at P9, and the hyaloid vessels appeared dilated and tortuous (Figure 4i). However, unlike in VEGF^188/188 mice, the hyaloid arteries in VEGF^120/120 mice neither penetrated the retina nor made connections with the retinal vasculature.

ACs initially migrated normally out to the periphery of the retina (not shown). By P5 and P9, ACs in the central vascularized retina associated with retinal vessels showed strong GFAP expression. In the avascular retinal periphery, however, GFAP immunoreactivity was weak (Figure 4j), possibly due to tissue damage secondary to the ischemia. VEGF expression was more abundant at the edge of the vascular bed, likely the result of local hypoxia (not shown).

Expression of VEGF receptors. Whole-mount in situ hybridization of retinas from VEGF^+/+ pups at P5 and P9 revealed that NP-1 was expressed in retinal arterioles and, at a much lower level, in retinal venules (Figure 5a). Higher magnification suggested that NP-1 is expressed in ECs (Figure 5b) and likely also in PECs (Figure 5, b and c), which were also labeled for PDGF receptor-β (Figure 5d). NP-1 mRNA expression has been reported previously in ECs and PECs (16). By triple immunofluorescent staining for NP-1, lectin, and SMA, NP-1 was detectable in retinal neurons, but not in ECs or PECs in the retina (Figure 5, e and f). Immunostaining for VEGFR-1 or VEGFR-2 indicated that VEGFR-1 was expressed in ECs and vSMCs of arteries and veins (Figure 5, g and h), whereas VEGF-R2 was expressed only in ECs of capillaries and venules in the retina (Figure 5, i and j), brain, heart, and kidney (not shown). Specificity of the VEGFR-1 and VEGFR-2 staining has been reported elsewhere (17). Expression of VEGFR-1 (18) and NP-1 (19) has been documented previously in PECs.

Figure 5

VEGF receptor expression in VEGF^+/+ retinas. (a) Whole-mount in situ hybridization for NP-1 at P5. Expression is prominent in arterioles and discrete in venules. (b) In situ hybridization on retinal sections at P7 is suggestive of NP-1 expression in ECs as well as PECs (arrows). (c and d) High magnification of whole-mount in situ hybridization for (c) NP-1 and (d) PDGF receptor-β (PDGFRβ) illustrates that PECs are positive for PDGFRβ and suggest NP-1 expression in PECs (arrows). (e and j) Fluorescent triple immunostaining using lectin, anti-SMA, and anti-VEGFR on sections from VEGF^+/+ eyes at P9. Lectin (blue), VEGFR (red), and SMA (green). (e and f) NP-1 is detectable in retinal neurons (arrowheads), but its expression is beyond the detection limit in vessels by immunostaining. (g and h) VEGFR-1 (R1) is expressed in vSMCs as well as ECs (arrow). (i and j) VEGFR-2 (R2) expression is only detectable in the ECs (arrows) in capillaries and venules.

Arterial development in other organs. A similar number of glomerular arterioles was present in anti-SMA–stained sections of VEGF^+/+, VEGF^164/164, and VEGF^188/188 kidneys (arterioles per kidney section: 45 ± 2, 49 ± 9, and 42 ± 7, respectively; n = 5; P = NS). Comparable densities of small, medium, and large coronaries were found in VEGF^+/+ (vessels per square millimeter: 8.5 ± 2.9, 4.3 ± 1.0, and 3.4 ± 1.6; n = 6), VEGF^164/164 (vessels per square millimeter: 8.1 ± 3.1, 4.2 ± 2.0, and 3.5 ± 1.8; n = 4, P = NS), and VEGF^188/188 hearts (vessels per square millimeter: 9.6 ± 2.8, 6.2 ± 2.6, and 2.5 ± 1.7; n = 6, P = NS). VEGF^120/120 mice had fewer renal and coronary arteries (6).

The present study investigates the distinct role of the different VEGF isoforms in retinal vascular development. Retinal vascular development was normal in VEGF^164/164 mice, indicating that this isoform contains all necessary information for normal outgrowth and remodeling of blood vessels. In contrast, VEGF^120/120 mice exhibited severe vascular defects, with impaired venous and severely defective arterial vascular development in the retina. VEGF^188/188 mice had normal venous development, but aborted arterial outgrowth.

Role of the VEGF isoforms in endothelial outgrowth. Astrocytes migrate in front of ECs and guide them to the periphery of the retina by releasing VEGF (20). Outgrowth of the vascular bed was normal in VEGF^164/164 and VEGF^188/188 mice (except for the arterioles in VEGF^188/188 retinas). In contrast, in VEGF^120/120 mice there was a defect in EC outgrowth, affecting all vessel types. This may be due to the fact that VEGF₁₂₀ is less potent in inducing EC mitosis (21) or has reduced affinity for VEGFR-1 (22), which could affect EC function or survival (6). Alternatively, VEGF₁₆₄ and VEGF₁₈₈, in binding the extracellular matrix, may provide matrix-associated guidance cues that facilitate EC migration through the retina, possibly through the establishment of a VEGF gradient or “trail” along which ECs migrate. Such cues would likely be absent in VEGF^120/120 mice in which diffusion of the soluble VEGF₁₂₀ isoform would induce random EC migration. These hypotheses are consistent with the finding that VEGF₁₂₀-expressing tumors are hypovascular, in contrast to VEGF₁₆₄- or VEGF₁₈₈-expressing tumors (23). This would implicate an important role for the different VEGF isoforms in correct patterning of nascent vessels. Vessel patterning is an essential characteristic of normal angiogenesis, which remains poorly understood at the molecular level.

Role of the VEGF isoforms in retinal arterial development. Genotype-specific defects were observed in the formation of retinal arterioles and venules. In VEGF^+/+ and VEGF^164/164 mice, arterioles reached out over approximately 90% of the entire vascular bed at P9. In VEGF^120/120 mice, there were fewer and smaller venules and arterioles but, whereas venules reached out over 80%, arterioles extended over only approximately 50% of the vascular bed. Arteriolar defects were even more pronounced in VEGF^188/188 mice, in which arterioles (but not venules) were smaller and extended over only 18% of the vascular bed. Thus, the VEGF₁₆₄ isoform is sufficient for venular and arteriolar development, the VEGF₁₈₈ isoform allows venular but not arteriolar development, and the VEGF₁₂₀ isoform is insufficient for venular, but especially for arteriolar, development.

Arterial EC specification. The current hypothesis suggests that “arterialization” of an undifferentiated vascular network requires remodeling and recruitment of vSMCs in response to hemodynamic forces (reviewed by ref. 24). More recent studies proposed that arterial and venous ECs are molecularly distinct from the earliest stages of vascular development, distinguishable by their expression of gridlock (25), D114 (26), Bmx (27), NP-1 (16), ephrin-B2 (on arterial ECs) and its receptor Eph-B4 (on venous ECs) (7), and activin receptor-like kinase-1 (28). Consistent with this revised model, ephrin-B2 was expressed in half of the vessels in VEGF^+/+ and VEGF^164/164 mice, even before arterioles could be morphologically distinguished. Since ephrin-B2 was also expressed in unremodeled retinal vessels in VEGF^120/120 mice, the impaired arteriolar development was not attributable to an aborted arteriolar EC specification. In contrast, ephrin-B2 expression in VEGF^188/188 mice was delayed until P6, when rudimentary arterioles began to grow out, indicating that these ECs were arterially differentiated but failed to grow out normally. Whether defective arterial outgrowth in VEGF^188/188 mice resulted from delayed or impaired arterial specification of EC precursors and/or outgrowth of arterial ECs, and why the VEGF₁₂₀ and VEGF₁₆₄ isoforms are more efficient than VEGF₁₈₈ in inducing arterial specification and/or outgrowth, remain to be determined. Intriguing possibilities are that the short isoforms induce expression of arterial markers (as demonstrated for Bmx) (27) or affect angioblasts even at an earlier stage or that the VEGF₁₈₈ isoform might be unable to provide spatial guidance or differentiation cues for retinal arteriolar ECs because it remains associated with the extracellular matrix.

PEC recruitment. PEC recruitment was normal in VEGF^164/164 mice. In contrast, VEGF^120/120 mice exhibited generally impaired PEC recruitment resulting in capillary dilatation and extravasation of red blood cells, as described in the hearts of these mice (6), as well as in other conditions of PEC dysfunction or dropout (29). ECs release signals (such as PDGF-BB) essential for PC recruitment (10). Since expression of PDGF-BB is reduced in VEGF^120/120 mice (6), it is possible that the impairment of PEC recruitment is attributable to the general endothelial defects in VEGF^120/120 mice. Interestingly, VEGFR-1 was expressed in retinal (as well as renal, brain, and cardiac) arteriolar and venular PECs (this study and ref. 18), indicating that VEGF may have direct effects on PC recruitment (10). Since the VEGF₁₂₀ isoform binds with a lower affinity to VEGFR-1 than the VEGF₁₆₄ isoform (22), this short isoform may be unable to recruit PECs as efficiently as the other isoforms. VEGF^188/188 mice, on the other hand, showed normal PEC recruitment to venules and capillaries, but reduced coverage of underdeveloped retinal arterioles. This may again suggest that the impaired PC recruitment is secondary to abnormal EC function in the underdeveloped arterioles. Alternatively, diffusion of the smaller VEGF isoforms may be essential for chemoattracting VEGFR-1–positive PECs. The specific defect of PC recruitment around arterioles may also suggest an essential interaction of the VEGF₁₆₄ isoform with NP-1, which is more abundantly expressed in retinal arterioles than in venules. Although not yet described, the VEGF₁₈₈ isoform likely binds to NP-1 since it contains exon 7, encoding the essential amino acid residues for such interaction (30). Perhaps this long isoform (which binds to the matrix) is unable to reach the NP-1 receptor on PECs or impairs PC migration because it immobilizes the NP-1–positive PCs at the site of VEGF₁₈₈ production. A role for NP-1 in cell adhesion has been suggested previously, although the nature of the ligand remains unknown (31). Another possibility is that blood flow in the underdeveloped arterioles is abnormal, which could impair their maturation. Finally, the retinal phenotype may have been influenced by the impaired angiogenesis in the heart, bone, and lungs of VEGF^120/120 mice (6, 32, 33), the growth retardation in the VEGF^188/188 mice, or the relative overexpression of a single isoform. Indeed, each of these knock-in mice expressed a single isoform at a level comparable to the total expression of the three isoforms in VEGF^+/+ mice (this study and ref. 6). However, it is unlikely that the retinal phenotype was only secondary to impaired vascular growth in other organs, because no such specific arterial defects were detected in other organs.

Arterial development in other organs. While it may be a surprise that this long isoform has a restricted role in establishing arterioles in the retina, emerging evidence indicates that the mechanisms of vessel formation (34), as well as the differential expression of the VEGF isoforms (32), are highly dependent on the tissue microenvironment. Noteworthy in this context is that retinal PECs are derived from neural crest cells, whereas PECs in other tissues are derived from other sources (35).

Role of VEGF isoforms in the regression of hyaloid vessels. In VEGF^188/188 mice, the retinal arteriolar underdevelopment prompted ingrowth by hyaloid arteries, which established connections with the retinal vascular network. In contrast, hyaloid vessels in VEGF^120/120 mice did not enter the retina, but were dilated and tortuous. The dissimilarity in hyaloid vessel phenotype may be attributable to the difference in solubility between the VEGF isoforms. VEGF₁₈₈ is matrix bound and presumably remains associated within the retina. Since hyaloid vessels lie closely apposed to the retina, the increased VEGF₁₈₈ levels might attract these hyaloid vessels into the ischemic peripheral retina (Figure 1g). Alternatively, VEGF₁₈₈ can be proteolytically cleaved into a shorter, soluble isoform, which would establish a VEGF gradient attracting hyaloid vessels (36). Soluble VEGF₁₂₀ would be expected to diffuse into the vitreous cavity (Figure 1f), explaining why hyaloid vessels persisted, became dilated and tortuous, and failed to enter the retina. Whether binding of VEGF₁₈₈ to NP-1 mediates adhesion of hyaloid vessels to the retina remains to be determined.

The persistence of hyaloid vessels in mice lacking the VEGF₁₆₄ isoform resembles the persistent hyperplastic primary vitreous disorder in humans, a common blinding congenital anomaly resulting from a failure of hyaloid vessel regression (37). Taken together, these findings suggest a novel role for the various VEGF isoforms in endothelial outgrowth and arterial development.

The authors thank A. Van Lommel (Pathology Department, Catholic University Leuven) and A. Bouché, K. Vandevelde, W. Man, L. Kieckens, A. Manderveld, F. De Wever, K. Maris, T. Vancoetsem, E. Gils, K. Bijnens, A. Vandenhoeck, P. Van Wesemael, L. Godde, S. Lucas, C. Van Huylebroeck, S. Wyns (The Center for Transgene Technology and Gene Therapy, Leuven) for assistance. They acknowledge K. Alitalo for providing the Tie1 promotor. I. Stalmans is a Research Assistant of the Fonds voor Wetenschappelijk Onderzoek (FWO), Belgium and P. D’Amore is a Jules and Doris Stein Research to Prevent Blindness Professor. This work is further supported by grants from the FWO (G012500), the Geconcerteerde Onderzoeksacties Belgium (2001/09), and the European Union (BMH4-CT98-3380) to P. Carmeliet; from the NIH (HL-6221 and CA-45548) to D.J. Anderson and P. D’Amore; and from Deutsche Forschungsgemeinschaft to H.-P. Hammes.

1. Carmeliet, P, et al. Abnormal blood vessel development and lethality in embryos lacking a single VEGF allele. Nature 1996. 380:435-439.
   View this article via: PubMed CrossRef Google Scholar
2. Carmeliet, P, Jain, RK. Angiogenesis in cancer and other diseases. Nature 2000. 407:249-257.
   View this article via: PubMed CrossRef Google Scholar
3. Neufeld, G, Cohen, T, Gengrinovitch, S, Poltorak, Z. Vascular endothelial growth factor (VEGF) and its receptors. FASEB J 1999. 13:9-22.
   View this article via: PubMed Google Scholar
4. Miao, HQ, Klagsbrun, M. Neuropilin is a mediator of angiogenesis. Cancer Metastasis Rev 2000. 19:29-37.
   View this article via: PubMed CrossRef Google Scholar
5. Soker, S, Takashima, S, Miao, HQ, Neufeld, G, Klagsbrun, M. Neuropilin-1 is expressed by endothelial and tumor cells as an isoform-specific receptor for vascular endothelial growth factor. Cell 1998. 92:735-745.
   View this article via: PubMed CrossRef Google Scholar
6. Carmeliet, P, et al. Impaired myocardial angiogenesis and ischemic cardiomyopathy in mice lacking the vascular endothelial growth factor isoforms VEGF164 and VEGF188. Nat Med 1999. 5:495-502.
   View this article via: PubMed CrossRef Google Scholar
7. Wang, HU, Chen, ZF, Anderson, DJ. Molecular distinction and angiogenic interaction between embryonic arteries and veins revealed by ephrin-B2 and its receptor Eph-B4. Cell 1998. 93:741-753.
   View this article via: PubMed CrossRef Google Scholar
8. Korhonen, J, et al. Endothelial-specific gene expression directed by the tie gene promoter in vivo. Blood 1995. 86:1828-1835.
   View this article via: PubMed Google Scholar
9. Tidhar, A, et al. A novel transgenic marker for migrating limb muscle precursors and for vascular smooth muscle cells. Dev Dyn 2001. 220:60-73.
   View this article via: PubMed CrossRef Google Scholar
10. Benjamin, LE, Hemo, I, Keshet, E. A plasticity window for blood vessel remodelling is defined by pericyte coverage of the preformed endothelial network and is regulated by PDGF- B and VEGF. Development 1998. 125:1591-1598.
    View this article via: PubMed Google Scholar
11. Kawakami, A, Kitsukawa, T, Takagi, S, Fujisawa, H. Developmentally regulated expression of a cell surface protein, neuropilin, in the mouse nervous system. J Neurobiol 1996. 29:1-17.
    View this article via: PubMed CrossRef Google Scholar
12. Bergemann, AD, Cheng, HJ, Brambilla, R, Klein, R, Flanagan, JG. ELF-2, a new member of the Eph ligand family, is segmentally expressed in mouse embryos in the region of the hindbrain and newly forming somites. Mol Cell Biol 1995. 15:4921-4929.
    View this article via: PubMed Google Scholar
13. Lindahl, P, Johansson, BR, Leveen, P, Betsholtz, C. Pericyte loss and microaneurysm formation in PDGF-B-deficient mice. Science 1997. 277:242-245.
    View this article via: PubMed CrossRef Google Scholar
14. Hammes, HP, et al. Acceleration of experimental diabetic retinopathy in the rat by omega-3 fatty acids. Diabetologia 1996. 39:251-255.
    View this article via: PubMed CrossRef Google Scholar
15. De Schaepdrijver, L, Simoens, P, Lauwers, H, De Geest, JP, Charlier, G. The hyaloid vascular system of the pig. A light and scanning electron microscopic study. Anat Embryol 1989. 180:549-554.
    View this article via: PubMed CrossRef Google Scholar
16. Moyon, D, Pardanaud, L, Yuan, L, Bréant, C, Eichmann, A. Plasticity of endothelial cells during arterial-venous differentiation in the avian embryo. Development 2001. 128:3359-3370.
    View this article via: PubMed Google Scholar
17. Carmeliet, P, et al. Synergism between vascular endothelial growth factor and placental growth factor contributes to angiogenesis and plasma extravasation in pathological conditions. Nat Med 2001. 7:575-583.
    View this article via: PubMed CrossRef Google Scholar
18. Suzuma, K, et al. Vascular endothelial growth factor induces expression of connective tissue growth factor via KDR, Flt1, and phosphatidylinositol 3-kinase-akt-dependent pathways in retinal vascular cells. J Biol Chem 2000. 275:40725-40731.
    View this article via: PubMed CrossRef Google Scholar
19. Ishida, A, et al. Expression of vascular endothelial growth factor receptors in smooth muscle cells. J Cell Physiol 2001. 188:359-368.
    View this article via: PubMed CrossRef Google Scholar
20. Stone, J, et al. Development of retinal vasculature is mediated by hypoxia-induced vascular endothelial growth factor (VEGF) expression by neuroglia. J Neurosci 1995. 15:4738-4747.
    View this article via: PubMed Google Scholar
21. Keyt, BA, et al. The carboxyl-terminal domain (111–165) of vascular endothelial growth factor is critical for its mitogenic potency. J Biol Chem 1996. 271:7788-7795.
    View this article via: PubMed CrossRef Google Scholar
22. Gitay-Goren, H, et al. Selective binding of VEGF121 to one of the three vascular endothelial growth factor receptors of vascular endothelial cells. J Biol Chem 1996. 271:5519-5523.
    View this article via: PubMed CrossRef Google Scholar
23. Grunstein, J, Masbad, JJ, Hickey, R, Giordano, F, Johnson, RS. Isoforms of vascular endothelial growth factor act in a coordinate fashion to recruit and expand tumor vasculature. Mol Cell Biol 2000. 20:7282-7291.
    View this article via: PubMed CrossRef Google Scholar
24. Yancopoulos, GD, Klagsbrun, M, Folkman, J. Vasculogenesis, angiogenesis, and growth factors: ephrins enter the fray at the border. Cell 1998. 93:661-664.
    View this article via: PubMed Google Scholar
25. Zhong, TP, Rosenberg, M, Mohideen, MA, Weinstein, B, Fishman, MC. Gridlock, an HLH gene required for assembly of the aorta in zebrafish. Science 2000. 287:1820-1824.
    View this article via: PubMed CrossRef Google Scholar
26. Shutter, JR, et al. Dll4, a novel Notch ligand expressed in arterial endothelium. Genes Dev 2000. 14:1313-1318.
    View this article via: PubMed Google Scholar
27. Rajantie, I, et al. Bmx tyrosine kinase has a redundant function downstream of angiopoietin and vascular endothelial growth factor receptors in arterial endothelium. Mol Cell Biol 2001. 21:4647-4655.
    View this article via: PubMed CrossRef Google Scholar
28. Urness, LD, Sorensen, LK, Li, DY. Arteriovenous malformations in mice lacking activin receptor-like kinase-1. Nat Genet 2000. 26:328-331.
    View this article via: PubMed CrossRef Google Scholar
29. Lindahl, P, Hellstrom, M, Kalen, M, Betsholtz, C. Endothelial-perivascular cell signaling in vascular development: lessons from knockout mice. Curr Opin Lipidol 1998. 9:407-411.
    View this article via: PubMed Google Scholar
30. Gitay-Goren, H, Soker, S, Vlodavsky, I, Neufeld, G. The binding of vascular endothelial growth factor to its receptors is dependent on cell surface-associated heparin-like molecules. J Biol Chem 1992. 267:6093-6098.
    View this article via: PubMed Google Scholar
31. Shimizu, M, Murakami, Y, Suto, F, Fujisawa, H. Determination of cell adhesion sites of neuropilin-1. J Cell Biol 2000. 148:1283-1293.
    View this article via: PubMed CrossRef Google Scholar
32. Ng, YS, Rohan, R, Sunday, ME, Demello, DE, D’Amore, PA. Differential expression of VEGF isoforms in mouse during development and in the adult. Dev Dyn 2001. 220:112-121.
    View this article via: PubMed CrossRef Google Scholar
33. Maes, C., et al. 2002. Impaired angiogenesis and enchondral bone formation in mice lacking the vascular endothelial growth factor isoforms VEGF164 and VEGF188. Dev. Dyn. In press.
34. Lecouter, J, et al. Identification of an angiogenic mitogen selective for endocrine gland endothelium. Nature 2001. 412:877-884.
    View this article via: PubMed CrossRef Google Scholar
35. Etchevers, HC, Vincent, C, Le Douarin, NM, Couly, GF. The cephalic neural crest provides pericytes and smooth muscle cells to all blood vessels of the face and forebrain. Development 2001. 128:1059-1068.
    View this article via: PubMed Google Scholar
36. Plouet, J, et al. Extracellular cleavage of the vascular endothelial growth factor 189-amino acid form by urokinase is required for its mitogenic effect. J Biol Chem 1997. 272:13390-13396.
    View this article via: PubMed CrossRef Google Scholar
37. Silbert, M, Gurwood, AS. Persistent hyperplastic primary vitreous. Clin Eye Vis Care 2000. 12:131-137.
    View this article via: PubMed CrossRef Google Scholar