Involvement of Flt-1 Tyrosine Kinase (Vascular Endothelial Growth Factor Receptor-1) in Pathological Angiogenesis

INTRODUCTION

Angiogenesis is an essential process in normal embryogenic development and in a variety of pathological conditions, such as diabetic retinopathy and tumor growth in vivo (1) . VEGF 3 and its receptor (VEGFR) family, including Flt-1 (VEGFR1), KDR/Flk-1 (VEGFR2), and Flt-4 (VEGFR3), are well known to be a crucial regulatory system for normal and pathological angiogenesis (2, 3, 4, 5, 6, 7, 8) . Flt-1 as well as KDR and Flt-4 is structurally related to the Fms/Kit/platelet-derived growth factor receptor family and contains an extracellular domain carrying seven immunoglobulin-like sequences and a cytoplasmic TK domain with a long kinase insert.

Flt-1 is expressed as a full-length tyrosine kinase receptor and in some cases as a soluble form, which carries only the extracellular domain (9, 10, 11, 12, 13) . Biochemically, Flt-1 shows an affinity to VEGF that is at least 10-fold higher than that of KDR/Flk-1, but its TK activity is ∼10-fold weaker than that of KDR/Flk-1 (2, 3, 4 , 14, 15, 16, 17, 18, 19) , suggesting a function different from KDR/Flk-1. Flt-1 and KDR/Flk-1 are specifically expressed on vascular endothelial cells (20, 21, 22, 23, 24) , and as an exception, the flt-1 mRNA has been shown to be expressed in monocyte/macrophages (25 , 26) .

Recently, gene targeting studies were carried out on flt-1 and KDR/flk-1 in mice (7 , 8) . The KDR/flk-1 (−/−) homozygous mice died at embryonic day 8.5 (E8.5) because of a severe deficiency in vascular formation associated with a strong hematopoietic impairment (7) . The flt-1 (−/−) homozygous mice also showed embryonic lethality at almost the same stage (E8.5–9.0); however, the phenotype was quite different. The blood vessels in flt-1 (−/−) mice were disorganized, and the endothelial-like abnormal cells were overgrowing within the vascular lumens (8) . These results suggest that Flt-1 has a negative role in the early angiogenesis in embryo. More recently, we have shown that the Flt-1 TK domain-deficient[ flt-1 TK (−/−)] mice developed basically normal blood vessels and survived with an impairment of macrophage migration toward VEGF (27) . Thus, Flt-1 is considered to play a negative function in embryogenesis by trapping the endogenous VEGF and adjusting the levels of VEGF to an appropriate range.

Although one important function of Flt-1 was found to be a VEGF-absorbing activity, it is still not clear whether the Flt-1 has a positive role in pathological angiogenesis, such as tumor angiogenesis. To answer this question, we introduced an Flt-1-specific ligand into a tumor cell line and examined the tumor growth and tumor angiogenesis in both the wild-type flt-1 TK (+/+) and flt-1 TK (−/−) mice. Our results indicate clearly that the TK of Flt-1, although it does not have strong enzymatic activity, is important for angiogenesis under certain pathological conditions.

MATERIALS AND METHODS

Animals.

F₁ mice of flt-1 TK (+/−) heterozygotes were intercrossed, and among offspring the F₂ littermates of flt-1 TK (+/+) and flt-1 TK (−/−) mice were selected for analysis. Seven to 10 flt-1 TK (+/+) and flt-1 TK (−/−) mice derived from several litters were used for each assay.

Cells Overexpressing Flt-1 Ligand.

Human PlGF2 or human VEGF₁₆₅ cDNA was inserted into murine pSRαMSVtk-neo retroviral vector, and the high titer retroviruses were used to infect murine endothelial cell-derived F₂ (kindly provided by Dr. K. Toda, Department of Dermatology, Kyoto University, Kyoto, Japan; Ref. 28 ) or LLC cell line by the Polybrene method. G418-resistant cells, designated as F2-PlGF, LLC-PlGF, or LLC-VEGF, were harvested as a mixed population, and the production of human PlGF2 or VEGF₁₆₅ from these cells into culture medium was examined by Western blotting with antibody against PlGF or VEGF.

Injection of Ligand-containing Matrigels and Ligand-overexpressed LLC Cells into Mice.

F2-PlGF (5 × 10⁶ cells) or 200 ng/ml of PlGF purified from F2-PlGF culture medium using HiTrap Heparin Affinity Column (1 ml; Pharmacia, Uppsala, Sweden) as described previously (27) were mixed with 1 ml of Matrigel (Becton Dickinson Laboratory) and s.c. injected into the back of anesthetized mice. Mouse VEGF₁₆₄ and antimouse VEGF₁₆₄ neutralizing antibody were purchased from R&D Systems (Minneapolis, MN). Parental LLC, LLC-PlGF, or LLC-VEGF cells were collected by centrifugation, resuspended in DMEM at 5 × 10⁶ cells/ml, and injected into the backs of mice.

Preparation of Proteins and Western and Northern Blot Analyses.

LLCs were lysed in an ice-cold buffer containing 150 mm NaCl, 50 mm HEPES (pH 7.4), 10 mm EDTA, 1% Triton X-100, 10% glycerol, 2% aprotinin, and 1 mm phenylmethylsulfonyl fluoride and incubated with heparin beads for 2 h. The protein-bound beads were collected by centrifugation, washed in PBS, and analyzed by SDS-PAGE and Western blotting. For Northern blot analysis, 0.2 kb of mouse PlGF cDNA and 0.5 kb of mouse VEGF-B cDNA were used for hybridization probes.

Immunohistochemistry.

Tumor tissue sections were immunohistochemically stained with an antiserum against human vWF (Dako, Carpinteria, CA) as an endothelial cell-specific marker or antiserum against mouse F4/80 antibody (BMA) as a macrophage marker. Tissue sections (8-μm thick) were fixed with acetone at −20°C for 5 min and rehydrated in PBS. For inhibition of endogenous peroxidase, sections were incubated in 0.1% H₂O₂ methanol solution for 10 min at room temperature and washed three times with PBS.

Nonspecific binding of antibody was blocked by incubation with 10% normal goat serum-containing PBS for 30 min. Subsequently, sections were incubated with primary antibody for 2 h, washed three times with PBS, incubated with biotinylated goat antirabbit IgG (Vector, Burlingame, CA) for vWF, washed again with PBS, and then incubated with ABC kit (Vector). Horseradish peroxidase-conjugated sheep antirabbit IgG was used for F4/80. After washing with PBS, vWF and F4/80 were detected by 3-amino-9-ethyl carbazole (Vector). The smooth muscle cells in the thoracic aorta and tumor tissues were stained with a FITC-conjugated anti-α-smooth muscle actin antibody (Sigma) at 4°C overnight.

RESULTS

Murine F2 Cells Overexpressing PlGF-2 Induce an Angiogenic Response in flt-1 TK (+/+) Mice but not in flt-1 TK (−/−) Mice

To examine whether the TK domain of Flt-1 is important for any pathological angiogenesis in vivo, we took advantage of the pair of Flt-1 TK domain (+/+) and (−/−) mice. We have shown already that the flt-1 TK (−/−) mice develop essentially normal physiological angiogenesis except for an impaired response of macrophages to VEGF (27) .

To continuously stimulate the Flt-1 TK, at the first step, we used F2-PlGF cells, a murine endothelial-derived cell line (28) that expresses relatively low levels of endogenous VEGF but high levels of exogenously introduced Flt-1-specific ligand, PlGF-2 (Refs. 13 , 15, and 29, 30, 31, 32 ; see “Materials and Methods”). F2-PlGF cells were mixed with Matrigel and s.c. injected into flt-1 TK (+/+) or flt-1 TK (−/−) mice (Fig. 1) ⇓ .

After a 4-day incubation, we found that the s.c. tissues attached to the Matrigel/F2-PlGF mixture and the surface of the Matrigel had become reddish with dilated fine blood vessels in flt-1 TK (+/+) mice but not in flt-1 TK (−/−) mice (Fig. 1, C and D) ⇓ . Microscopically, the Matrigel/F2-PlGF mixture in flt-1 TK (+/+) mice showed a clear angiogenic response with several dilated vessels but not in flt-1 TK (−/−) mice. The original F2 cells without the exogenous PlGF expression showed no detectable angiogenic response either in flt-1 TK (+/+) or in TK (−/−) mice (Fig. 1, A and B) ⇓ .

During this incubation period, most of the F2 or F2-PlGF cells died within the Matrigel because of its weak transforming potential (Fig. 1) ⇓ . However, these results strongly suggest that Flt-1 TK is directly involved in angiogenesis under certain conditions, i.e., when the Flt-1-specific ligand is highly expressed.

Purified PlGF-2 Induces an Angiogenic Response in flt-1 TK (+/+) Mice

To avoid any effects produced by F2 cells on the angiogenesis assay, we next used the PlGF-2 purified from the culture medium of cells overexpressing PlGF-2 (see “Materials and Methods”) and mixed it with Matrigel for in vivo analysis. The same amounts (200 ng) of PlGF-2 or VEGF₁₆₄ mixed with Matrigel were inoculated s.c. for 4 days. The Matrigel containing PlGF-2 showed a mild angiogenic response with dilated blood vessels in flt-1 TK (+/+) mice but not in flt-1 TK (−/−) mice (Fig. 2, A and B) ⇓ . On the other hand, the Matrigel containing VEGF instead of PlGF induced small new blood vessels to form both in flt-1 TK (+/+) and TK (−/−) mice (Fig. 2, C and D) ⇓ .

In some areas of the PlGF-Matrigel in flt-1 TK (+/+) mice, we observed an infiltration of macrophage-like, F4/80-positive cells. Because macrophages were reported to secrete VEGF to some extent, we tested whether macrophage-mediated VEGF participates in this angiogenic response by using antimouse-VEGF neutralizing antibody. The Matrigel containing both PlGF and antimouse-VEGF neutralizing antibody still induced an angiogenic response in flt-1 TK (+/+) mice (Fig. 2E) ⇓ , indicating that the major angiogenic factor in this system is PlGF itself. In total, however, the responses to purified ligands were mild, even in the wild-type mice, possibly because of a rapid inactivation and degradation of ligands in Matrigels.

Comparison of Tumor Angiogenesis in flt-1 TK (+/+) Mice and in flt-1 TK (−/−) Mice

Expression of PlGF-2 or VEGF₁₆₅ in LLC Cells Does Not Modify the in Vitro Cell Growth.

Because F2 or F2-PlGF cells could not make tumors in vivo, we used the murine LLC cell line for an in vivo tumorigenicity assay. Before introducing human PlGF or VEGF cDNA to LLCs, we examined the endogenous VEGF-related ligands in LLCs. We found that after overnight culture, LLCs secreted ∼40 ng/ml of VEGF₁₆₄ (Fig. 3A) ⇓ , less than the ascites-generating sarcoma cell lines (170 to 850 ng/ml; Ref. 33 ). LLCs also expressed VEGF-B mRNA (34, 35, 36) at levels similar to a melanoma but higher than normal placenta and lung (Fig. 3C) ⇓ . LLCs did not express detectable levels of PlGF mRNA (Fig. 3B) ⇓ .

After infecting LLCs with retrovirus expression vector for PlGF or VEGF cDNA, we harvested the infected cells as a mixed population to avoid selection of specific clones. LLC-PlGF thus obtained secreted ∼200 ng/ml of PlGF, and LLC-VEGF secreted ∼40 ng/ml of exogenous VEGF₁₆₅ (total, 80 ng/ml) per 1 × 10⁷ cells in overnight culture (data not shown). The growth rate of these LLCs overexpressing either PlGF or VEGF₁₆₅ was essentially the same as that of the parental LLCs in in vitro culture (Fig. 4) ⇓ .

LLC-PlGF-induced Tumors in flt-1 TK (+/+) Mice Grow Faster Than Those in flt-1 TK (−/−) Mice, with Higher Angiogenesis.

Cells (5 × 10⁶) of LLC-PlGF or LLC-VEGF were injected into the back of flt-1 TK (+/+) or flt-1 TK (−/−) mice, and the growth rate of tumors was monitored for 1 month (Fig. 5) ⇓ . PlGF protein secreted from LLC-PlGF could be detected in tumor tissues 3 weeks after inoculation (data not shown).

Eight days after inoculation, the tumor growth of LLC-PlGF was faster in flt-1 TK (+/+) mice than flt-1 TK (−/−) mice. After 4 weeks, the tumor mass in flt-1 TK (+/+) mice was ∼300% that in flt-1 TK (−/−) mice (Fig. 5A) ⇓ .

On the other hand, LLC-VEGF cells induced tumors of almost the same size in the two genotypes of mice (Fig. 5C) ⇓ . This is to be expected because the overexpressed VEGF could stimulate KDR/Flk-1 in these genotypes, which carries potent TK activity and mediates a strong angiogenic signal.

The tumor growth rate of the original LLCs was basically low but slightly better in flt-1 TK (+/+) mice when compared with flt-1 TK (−/−) mice (Fig. 5B) ⇓ . This minor difference could be attributable to an endogenous VEGF-B expression in LLCs. The LLC-PlGF-induced tumors showed a clear increase in tumor vessel density, particularly small- to middle-sized blood vessels, as indicated in Table 1 ⇓ and Fig. 6 ⇓ .

Morphological Differences in the Blood Vessels between LLC-PlGF- and LLC-VEGF-induced Tumor Tissues.

The blood vessels induced by LLC-PlGF were morphologically different from those induced by LLC-VEGF. In LLC-PlGF, relatively large vessels >100 μm in diameter were often observed (Fig. 6, E and G) ⇓ . However, LLC-VEGF cells induced many small vessels not >50 μm in diameter (Fig. 6, I and J ⇓ ; Table 1 ⇓ ). The large vessels induced by LLC-PlGF, as well as small vessels induced by LLC-VEGF, carried a single layer of vWF-positive endothelial cells (Fig. 7 ⇓ , upper), but smooth muscle cells positive for α-smooth muscle actin staining were almost undetectable (Fig. 7 ⇓ , lower). These results suggest that the processes of angiogenesis stimulated by PlGF and by VEGF are not identical.

Number of Macrophages within the LLC-PlGF-induced Tumor Tissues.

Macrophages are known to be partly involved in the process of angiogenesis by secreting MMPs and angiogenic factors. To examine whether macrophage-like cells participate in the LLC-PlGF-induced tumor angiogenesis, we counted the number of cells positively stained with anti-F4/80 antibody, which was specific for macrophages. The numbers of the F4/80-positive macrophages at day 8 were almost the same between the tumors in flt-1 TK (+/+) and in flt-1 TK (−/−) mice, probably because of a nonspecific infiltration. At day 16, the number of these macrophages in the tumors were decreasing in both cases but 3-fold larger in flt-1 TK (+/+) mice than that in flt-1 TK (−/−) mice (Table 2) ⇓ . Because LLC-PlGF tumors are growing faster even at day 8 in flt-1 TK (+/+) mice when compared with those in flt-1 TK (−/−) mice, we do not think that the macrophages infiltrated into tumors play a major role in LLC-PlGF-induced tumor angiogenesis.

DISCUSSION

Mechanisms for the Tumor Angiogenesis Induced by the Flt-1 TK Receptor.

The TK activity of Flt-1 has been shown to be one order of magnitude weaker than that of KDR/Flk-1. We reported previously that PlGF-2, which binds both Flt-1 and neuropilin-1 (37) , a subtype-specific receptor for VEGF, showed a weak stimulatory activity for proliferation of endothelial cells in vitro (2, 3, 4 , 14, 15, 16, 17, 18, 19) . PlGF-1, which binds Flt-1 but not neuropilin-1, was found to induce angiogenesis using a highly sensitive system in vivo (38) . However, generally the biological activity of PlGF is not strong; thus, it is still not clear whether the Flt-1 tyrosine kinase induces pathological angiogenesis, such as tumor angiogenesis in vivo, because of the unavailability of a sensitive enough assay.

In this study, using LLC-PlGF and both wild-type flt-1 TK (+/+) and flt-1 TK (−/−) mice, we have shown that the tumor growth was ∼3-fold higher in the wild-type compared with flt-1 TK (−/−) mice. The increase in the tumor growth in the wild-type mice correlated well with the increase in tumor angiogenesis. These results strongly suggest that the TK domain of Flt-1 is required for a certain pathological angiogenesis, such as Flt-1-specific, ligand-induced tumor angiogenesis in vivo.

Because Flt-1 bears weak stimulatory activity for endothelial cell proliferation and because angiogenesis is a multistep process involving matrix proteolysis, migration, proliferation, and tube formation, several steps other than direct cell proliferation are also considered to be stimulated by Flt-1.

Recently, our preliminary results suggested that MMP-2 was activated in blood vessel-enriched regions in tumor tissues in wild-type but not in the flt-1 TK (−/−) mice. 4 MMP is known to be important for tumor growth because MMP-2 null mutation in mice resulted in the reduction of tumor growth via a decrease in angiogenesis (39, 40, 41) . In addition, batimastat (BB94), an inhibitor for MMPs, disturbed the growth of a transplanted LLC-induced tumor by 25% (42) .

Precursor forms of MMP-2 and MMP-1, as well as tissue inhibitor of MMP-1 and tissue inhibitor of MMP-2 are known to be produced in endothelial cells (39) . Therefore, signaling via the TK domain of Flt-1 appears to activate MMP-2, leading to a proteolytic digestion of matrix in tumor tissue to facilitate migration and tube formation of endothelial cells.

It is also possible that some angiogenic factors, such as VEGF secreted from the macrophages that migrated in tumor tissues, induce angiogenesis. However, this seems unlikely, because approximately the same number of macrophages have migrated into tumor tissues at day 8, when a difference in tumor angiogenesis between the two strains of mice was already observed (Tables 1 ⇓ and 2) ⇓ . Moreover, a partially purified PlGF-induced angiogenesis in wild-type mice, even in the presence of anti-VEGF neutralizing antibody, suggested a minor contribution of VEGF secreted from infiltrating inflammatory cells (Fig. 2E) ⇓ .

Taken together, a simple explanation for the rapid growth of LLC-PlGF in wild-type mice might be direct stimulation of both the activation of proteases and the proliferation of endothelial cells via a positive signal from the Flt-1 tyrosine kinase.

Possible Mechanisms for Morphological Differences in Blood Vessels Induced by PlGF and VEGF.

We found a difference in the formation of blood vessels between LLC-PlGF- and LLC-VEGF-induced angiogenesis. The vessels induced by LLC-PlGF were often larger in diameter than those induced by LLC-VEGF. In our preliminary experiments, the proteolysis surrounding the newly formed blood vessels was increased in LLC-PlGF-induced tumor tissues compared with LLC-VEGF-induced tumor. 4 Dvorak’s group reported previously that, in addition to the sprouting mechanism, angiogenesis occurred via an increase in the diameter of blood vessels because of a proliferation of endothelial cells, followed by segregation of an enlarged vessel to several small ones (43) . Therefore, the morphological difference induced by PlGF and by VEGF might be attributable to a difference in the remodeling process of angiogenesis. PlGF may induce less sprouting-type vessels but rather large-diameter-type vessels.

Another difference between PlGF- and VEGF-induced angiogenesis may be in the cells surrounding endothelial cells, such as smooth muscle cells and pericytes. However, this seems unlikely because no vessels induced by either PlGF or VEGF were stained with anti-α-smooth muscle actin antibody, except for preexisting arteries, indicating that in both cases the association of smooth muscle cells with vessels was equally low.

Differential Roles of Flt-1 during Embryonic and Pathological Angiogenesis.

We have shown previously that the extracellular domain of Flt-1 plays a negative role in embryogenesis. Although null mutation of the flt-1 gene results in embryonic lethality because of an overgrowth of endothelial-like cells, the flt-1 TK (−/−) mice developed an almost normal vascular system and could be mated as wild-type mice. Because the loss of a single allele of the VEGF gene in mice strongly disturbed angiogenesis and was lethal between the stage of E11–E12, a tight regulation of VEGF levels in vivo is thought to be necessary for normal angiogenesis in embryogenesis. Thus, the extracellular domain of Flt-1 may function as a ligand-trapping molecule and negatively regulate the levels of VEGF, decreasing the signals from KDR/Flk-1.

Recently, elevations of human PlGF mRNA in renal cell carcinoma as well as some brain tumors (44 , 45) and an increase in human VEGF-B mRNA in melanoma, lung carcinoma, and colon cancer have been reported (46) . These observations, as well as the results shown here, suggest that not only KDR/Flk-1 but also Flt-1 is involved in tumor angiogenesis in a positive manner. Therefore, in addition to the blocking of VEGF and KDR/Flk-1, suppression of the Flt-1-PlGF/VEGF-B pathway may be required to decrease certain types of pathological angiogenesis in vivo.

On the basis of these findings, Flt-1 is considered to have dual functions in endothelial cells, either negative or positive, depending on the biological conditions. Because Flt-1 can bind VEGF, PlGF, and VEGF-B with different affinity and because the gene regulation of each of these three ligands is different, further characterization is required to elucidate the network of the VEGF ligand family and Flt-1 in each pathological condition.