Possible participation of autocrine and paracrine vascular endothelial growth factors in hypoxia-induced proliferation of endothelial cells and pericytes.

Hypoxia is the principal factor that causes angiogenesis. These experiments were conducted to explore how it induces the proliferation of vascular cells, a key step in angiogenesis. Human umbilical vein endothelial cells and bovine retinal pericytes were grown in controlled atmosphere culture chambers containing various concentrations of oxygen. The numbers of both endothelial cells and pericytes increased significantly under hypoxic conditions; the O2 concentrations that achieved maximal growth promotion were 10% for endothelial cells and 2.5% for pericytes. Quantitative reverse transcription-polymerase chain reaction analysis revealed that mRNAs coding for the secretory forms of vascular endothelial growth factor (VEGF), a mitogen specific to endothelial cells, were present in both endothelial cells and pericytes and that their levels increased significantly in the two cell types as the atmospheric O2 concentration decreased. The two genes for VEGF receptors, kinase insert domain-containing receptor (kdr) and fms-like tyrosine kinase 1 (flt1), were found to be constitutively expressed in endothelial cells, and their relative mRNA levels were ranked in that order. On the other hand, only flt1 mRNA was detected in pericytes under hypoxic conditions. Furthermore, most antisense oligodeoxyribonucleotides complementary to VEGF mRNAs efficiently inhibited DNA synthesis in endothelial cells cultured under hypoxic conditions. These results indicate that autocrine and paracrine VEGFs may take part in the hypoxia-induced proliferation of endothelial cells.

Hypoxia is the principal factor that causes angiogenesis. These experiments were conducted to explore how it induces the proliferation of vascular cells, a key step in angiogenesis. Human umbilical vein endothelial cells and bovine retinal pericytes were grown in controlled atmosphere culture chambers containing various concentrations of oxygen. The numbers of both endothelial cells and pericytes increased significantly under hypoxic conditions; the O 2 concentrations that achieved maximal growth promotion were 10% for endothelial cells and 2.5% for pericytes. Quantitative reverse transcription-polymerase chain reaction analysis revealed that mRNAs coding for the secretory forms of vascular endothelial growth factor (VEGF), a mitogen specific to endothelial cells, were present in both endothelial cells and pericytes and that their levels increased significantly in the two cell types as the atmospheric O 2 concentration decreased. The two genes for VEGF receptors, kinase insert domain-containing receptor (kdr) and fms-like tyrosine kinase 1 (flt1), were found to be constitutively expressed in endothelial cells, and their relative mRNA levels were ranked in that order. On the other hand, only flt1 mRNA was detected in pericytes under hypoxic conditions. Furthermore, most antisense oligodeoxyribonucleotides complementary to VEGF mRNAs efficiently inhibited DNA synthesis in endothelial cells cultured under hypoxic conditions. These results indicate that autocrine and paracrine VEGFs may take part in the hypoxia-induced proliferation of endothelial cells.
Angiogenesis is a process by which new vascular networks are formed from pre-existing capillaries (1). Physiologically, it is essential for embryogenesis, development, ovulation, corpus luteum formation, and wound repair. In addition, it occurs during the progression of various pathological conditions such as cancers, diabetic retinopathy, rheumatoid arthritis, and occlusive vascular diseases, e.g. neovascularization is needed by solid tumors to access sufficient nutrients and oxygen for growth (1,2). To determine how angiogenesis is induced under these circumstances is therefore important for clarifying the pathogenesis, prevention, and treatment of such diseases as well as for understanding the basis of the physiological processes involved.
A decrease in tissue oxygen concentrations has been considered as the leading cause of angiogenesis (3). However, the mechanisms underlying the induction of angiogenesis by hypoxia are still poorly understood. Using primary cultured vascular cells, we have been investigating the biochemical basis underlying various vascular functions and disturbances (4 -7). In this study, we employed a hypoxic culture system and examined how low oxygen tensions affect the proliferation of endothelial cells and pericytes, a key step in angiogenesis; the latter cell type is the microvascular constituent encircling the endothelium, which we showed plays important roles in the growth, function, and damage of endothelial cells (4,5). The first part of this paper describes the accelerated growth of both endothelial cells and pericytes caused by hypoxia.
The following parts of this paper deal with the molecular mechanism underlying the hypoxia-induced proliferation of the vascular cells. Recently, there has been an explosive growth in knowledge regarding angiogenic growth factors (8 -10), including vascular endothelial growth factor (VEGF), 1 an endothelial cell-specific mitogen. This factor was initially identified in the conditioned medium of bovine pituitary follicular stellate cells (11,12), and its expression was subsequently shown to increase in human gliomas under hypoxic conditions (13,14). The presence of VEGF in vascular cells, including endothelial cells (15), smooth muscle cells (16), and mesangial cells (17), has also been noted. Here we show that endothelial cells and pericytes per se can produce secretory forms of VEGF in response to hypoxia. We also determined the VEGF receptor subtypes expressed in these two cell types. Furthermore, a functional relationship between the vascular VEGF system and hypoxiadriven endothelial cell growth was tested by manipulating VEGF gene expression with antisense DNA. Cells-Vascular endothelial cells were primary-cultured from human umbilical cord veins as described previously (4) and maintained in a gelatin-coated Falcon 3110 tissue culture flask in RPMI 1640/Medium 199 (1:1) supplemented with 15% FBS, 100 g/ml endothelial cell growth supplement, and 25 g/ml heparin at 37°C under 5% CO 2 , 95% air. Cells at 15-20 passages were used for experiments; Ͼ95% of the cells were identified as endothelial cells by their uptake of acetylated low density lipoprotein (4). Pericytes were isolated from bovine retina as described (4) and maintained in a Falcon 3110 culture flask in Dulbecco's modified Eagle's medium supplemented with 20% FBS under 5% CO 2 , 95% air. Cells at 5-10 passages were used for experiments; Ͼ95% of the cells were identified as pericytes by their characteristic size and contour, lack of "hill and valley" growth pattern, and positive stain for the ␣-isoform of smooth muscle actin. U251 cells, a human glioma cell line, were donated from RIKEN Cell Bank (Tsukuba, Japan) and maintained in Dulbecco's modified Eagle's medium supplemented with 10% FBS under 5% CO 2 , 95% air.
Antisense and Sense Oligodeoxyribonucleotides-A 22-mer antisense oligodeoxyribonucleotide complement of the 5Ј-region of human VEGF mRNA containing the initiator AUG codon and the corresponding sense oligodeoxyribonucleotide were synthesized with the Model 392 DNA synthesizer by the phosphorothioate approach using tetraethylthiuram disulfide (25). Sequences of antisense and sense oligodeoxyribonucleotides were 5Ј-CCCAAGACAGCAGAAAGTTCAT-3Ј (complement of nt 57-78) and 5Ј-ATGAACTTTCTGCTGTCTTGGG-3Ј (nt 57-78) (11), respectively. Additional antisense oligodeoxyribonucleotides and the respective sense controls were raised against nt Ϫ698 to Ϫ678, nt Ϫ624 to Ϫ603, and nt Ϫ424 to Ϫ411 of the 5Ј-untranslated region; nt 155-174, Culture under Low Oxygen Tensions-After seeding in a cluster dish or flask, cells were grown at 37°C under 5% CO 2 , 95% air for 24 h to ensure cell attachment. The resultant monolayer cultures were then placed in a controlled atmosphere culture chamber (Bellco, Vineland, NJ), a humidified airtight incubation apparatus with inflow and outflow valves, into which a gas mixture containing 5% CO 2 and 0, 2.5, 5, 10, or 20% O 2 balanced with N 2 was flushed for 5 min at a flow rate of 10 liter/min. The chamber was sealed to maintain a constant gas composition and kept at 37°C. The gas phase was renewed every 24 h, and the medium every 48 h. The O 2 and CO 2 levels in the atmosphere culture chambers were monitored with FYRITE gas analyzers (Bacharach, Inc., Pittsburgh, PA) prior to and after gas and medium renewals; they were immediately equilibrated to the set values by the 5-min flushing and kept constant for at least 24 h to an accuracy of Ϯ5%.
Measurement of Cell Growth-The number of viable cells was estimated by the dye exclusion method (27). [ 3 H]Thymidine incorporation was determined essentially as described (28). Briefly, endothelial cells were seeded at a density of 2 ϫ 10 4 cells/well of a Costar 24-well cluster dish and placed at 37°C overnight. After cell attachment, antisense or sense oligodeoxyribonucleotides were added to the medium, and cells were incubated for 24 h. Then, [ 3 H]thymidine was added to a final concentration of 1 Ci/ml, and cells were further incubated for 4 h. After incubation, cells were fixed with ice-cold 5% (w/v) trichloroacetic acid for 20 min. The resulting precipitates were washed three times with ice-cold 5% trichloroacetic acid and solubilized by mixing with 200 l of 1 N NaOH at room temperature for 20 min, followed by neutralization with the same volume of 1 N HCl. 3 H radioactivity was measured by liquid scintillation counting.
Poly(A) ϩ RNA Isolation-Poly(A) ϩ RNAs were isolated as described (29) from cells that had been grown to confluency in a Falcon 3112 culture flask under standard atmospheric conditions and then incubated for 24 h under various oxygen tensions. Isolated poly(A) ϩ RNAs were stored at Ϫ80°C until assayed.
Quantitative RT-PCR-RT-PCR was performed as described (30). A 10-l aliquot of each RT-PCR reaction mixture was electrophoresed on a 2% agarose gel, transferred to a Hybond-N ϩ nylon membrane, and hybridized with specific oligodeoxyribonucleotide probes that had been 32 P-end-labeled with [␥-32 P]ATP and polynucleotide kinase. Autoradiographic exposures of washed membranes were done at Ϫ80°C for various time periods, and the radioactivities of the hybridization bands were measured with a Fujix BA100 BioImage analyzer (Fuji Photo Film Co. Ltd., Hamamatsu, Japan).
Metabolic Labeling and Immunoprecipitation of VEGF-The experiments were carried out as described (31,32). Briefly, subconfluent cultures of endothelial cells were incubated in the presense or absence of antisense or sense oligodeoxyribonucleotides against VEGF mRNA for 24 h and further incubated for 18 h at 37°C under 10% O 2 in methionine-free RPMI 1640 medium/complete RPMI 1640 medium (9:1) containing 0.2 mCi/ml [ 35 S]methionine with or without the antisense or sense oligodeoxyribonucleotides. The cells were lysed, and cell lysates were immunoprecipitated with monoclonal antibody A4.6.1. Pericytes and U251 cells grown under 2.5% O 2 were also processed in the same way. The samples were analyzed on a 15% SDS-polyacrylamide gel under reducing conditions. The gel was dried and autoradiographed. The radioactivities of the bands were measured with the Fujix BA100 BioImage analyzer.

Hypoxia Stimulates the Growth of Both Endothelial Cells
and Pericytes-Human umbilical vein endothelial cells and bovine retinal pericytes were cultured in controlled atmosphere culture chambers containing 2.5, 5, 10, and 20% O 2 . Fig. 1A shows the growth curves of endothelial cells. Under the standard oxygen tension (20%), the viable cell number began to gradually increase on day 3 and doubled on day 7. When exposed to 10% O 2 , endothelial cells started to grow within 24 h without noticeable lag and increased in an almost linear fashion, reaching on day 7 a level ϳ3.5-fold higher than at the beginning of the hypoxic culture; at every time point examined, the viable cell number under 10% O 2 was significantly larger than that under 20% O 2 . Five percent O 2 also supershifted the curve, but to a lesser extent than did 10% O 2 . When the O 2 tension was further reduced to 2.5%, the viable cell number instead decreased, and almost all the endothelial cells were dead within 3 days. The growth of pericytes was also stimulated under hypoxic conditions (Fig. 1B). In the case of pericytes, however, the maximal growth promotion was achieved by 2.5% O 2 . The rank order of O 2 tensions in the promotion of pericyte growth was 2.5% Ͼ 5% Ͼ 10% Ͼ 20% on day 6.
Both Endothelial Cells and Pericytes Express mRNA for Secretory Forms of VEGF in Response to Hypoxia-Poly(A) ϩ RNAs were isolated from cells that had been exposed to low oxygen tensions and analyzed by an RT-PCR technique to determine the expression of the VEGF gene. It has been known that four alternatively spliced products can be generated from the single VEGF gene, yielding different protein products composed of 121, 165, 189, and 206 amino acids, designated as VEGF 121 , VEGF 165 , VEGF 189 , and VEGF 206 , respectively. Among them, only VEGF 121 and VEGF 165 are secreted and induce mitogenesis of endothelial cells; VEGF 189 and VEGF 206 are membrane-anchored and act as vascular permeability factors (31,33). Since Northern blot analysis could not clearly discriminate the four mRNA species, we employed a more sensitive RT-PCR technique to determine which VEGF forms are expressed in the vascular cells. For this, we designed primers and an internal probe against the regions common to the four alternatively spliced products ( Fig. 2A); with them, 486-, 618-, 690-and 741-bp-long cDNA fragments would be amplified from respective VEGF 121 , VEGF 165 , VEGF 189 , and VEGF 206 mRNA templates (19). Fig. 2 (B and C) shows titration curves of RT-PCR products for determining the quantitative range in which the reactions proceeded exponentially. Poly(A) ϩ RNA from U251 cells, a human glioma cell line, was used as a standard template; this cell line has been known to produce high amounts of VEGF (13,14). Signal intensities of the products obtained with U251 poly(A) ϩ RNA were plotted as functions of template amount and cycle number. The products increased linearly up to 50 ng (Fig. 2B) and up to 30 cycles (Fig. 2C); hence, we chose 30-ng templates and 25 cycles as the conditions for VEGF mRNA analysis.
As shown in Fig. 2D, poly(A) ϩ RNAs from endothelial cells and pericytes gave signals at 486 and 618 bp, which corresponded to mRNAs for VEGF 121 and VEGF 165 , respectively, as did U251 poly(A) ϩ RNA. The levels of the two mRNA species increased significantly in these cells as the atmospheric O 2 concentration decreased from 20 to 0%. The sum of the 486-and 618-bp band intensities was strongest in anoxic cultures; in endothelial cells and pericytes, it was 8-and 9-fold higher than respective normoxic cultures when standardized with the signal intensities of ␤-actin mRNA as an internal control. On the other hand, signals for VEGF 189 and VEGF 206 mRNAs were not detected in either the vascular cells or the glioma cells throughout these experiments.
Determination of VEGF Receptor Subtypes Expressed in Endothelial Cells and Pericytes-To exert its action, locally produced VEGF requires locally expressed receptors. We then examined VEGF receptor gene expression in endothelial cells and pericytes. So far, two kinds of VEGF receptors, flt1 (20,34) and kdr (22,23), and their homologue, flt4 (21,35), have been identified. Although flt4-encoded protein has recently been shown not to act as a receptor for VEGF (36), the three have a quite similar structure, each containing seven extracellular immunoglobulin-like domains, one transmembrane domain, and two cytoplasmic tyrosine kinase domains; their mRNAs exhibit high homologies in both nucleotide and deduced amino acid sequences (35). Accordingly, the RT-PCR method rather than Northern blotting seemed suitable for differential mRNA detection as in the case with VEGF mRNAs. Primers and probes were designed so that they would correspond to the regions where sequence homologies among the three are relatively low and would generate products of differing lengths (Fig. 3A). Data in Fig. 3B show that each set of primers and a probe worked well in specifying their corresponding mRNA; 1098-bp DNA fragments amplified with flt1 primers hybridized to the flt1 probe, but not the kdr probe, and 555-bp RT-PCR products with kdr primers hybridized only to the kdr probe. Based on the titration of the products (Fig. 3, C and D), RT-PCR was performed with 30 ng of template poly(A) ϩ RNA for 25 cycles for the quantitative detection of each receptor mRNA.
As shown in Fig. 3E, endothelial cells were found to contain mRNAs for flt1 and kdr. Between the two, the intensities of the hybridization signals were much stronger for kdr mRNA than for flt1 mRNA. In contrast with VEGF mRNAs, the levels of kdr and flt1 mRNAs were essentially unchanged when atmospheric O 2 tensions were lowered.
On the other hand, in pericytes, neither of the two mRNA species was detected under normoxic conditions. However, as shown in Fig. 3F, the 1098-bp band corresponding to flt1 mRNA was visible in pericytes grown at 2.5% O 2 and became clearly marked at 0%. kdr mRNA remained undetected at 2.5 and 0% O 2 even when the template amount and the cycle number were raised to 100 ng and 35 cycles, respectively (data not shown).
Antisense Oligodeoxyribonucleotides against VEGF mRNA Inhibit DNA Synthesis in Hypoxic Endothelial Cell Culture-We next tested whether vascular VEGF could be functionally related to the hypoxia-induced proliferation of endothelial cells. For this, a 22-mer antisense oligodeoxyribonucleotide complement of the 5Ј-region of human VEGF mRNAs and the corresponding sense oligodeoxyribonucleotide were synthesized and administered to the culture medium in which endothelial cells were grown. After attachment to the dish, cells were incubated with various concentrations of oligodeoxyribonucleotides under 10% O 2 for 24 h, and then [ 3 H]thymidine incorporation was measured. As shown in Fig.  4A, the antisense oligodeoxyribonucleotide was found to inhibit [ 3 H]thymidine incorporation into the endothelial cells in a dose-dependent manner; at 1, 2, 5, and 10 M, 6.9, 55.1, 71.6, and 86.5% inhibitions were achieved, respectively. Control sense oligodeoxyribonucleotides showed no significant change.  (Fig. 4C).
Evidence That the Antisense Oligodeoxyribonucleotides Block VEGF Expression-To confirm whether the antisense oligodeoxyribonucleotides did block the expression of VEGF mRNA, we examined by immunoprecipitation de novo VEGF synthesis in endothelial cells that had been treated with or without the antisense or sense oligodeoxyribonucleotides. As shown in Fig.  5, the anti-VEGF monoclonal antibody specifically recognized 35 S-labeled proteins that migrated to the positions of 22 and 18 kDa, and the amounts of the 22-and 18-kDa proteins were found to be consistently lowered by the antisense oligodeoxyribonucleotides in a dose-dependent manner. At 2 and 10 M, 52 and 78% inhibitions were achieved, respectively. On the other hand, the sense oligodeoxyribonucleotides did not inhibit VEGF synthesis in endothelial cells.
Antisense Oligodeoxyribonucleotides against Different Re- , pericytes, and U251 cells, which had been incubated under various oxygen tensions, underwent quantitative RT-PCR analysis. The products were electrophoresed on 2% agarose gel, transferred onto nylon membranes, and hybridized with 32 P-end-labeled probes specific to VEGF (upper panels) and ␤-actin (lower panels) mRNAs. PCR amplification for the latter was performed for 15 cycles. Bars indicate size markers in base pairs. Endothelial cell and pericyte blots were exposed for 12 h, and the U251 blot for 6 h. RT(Ϫ), the reaction without reverse transcriptase. , ␤-actin. E, RT-PCR analysis of VEGF receptor mRNAs in endothelial cells. Thirty nanograms of poly(A) ϩ RNA from endothelial cells incubated under the indicated O 2 concentrations was amplified by RT-PCR and hybridized with 32 P-end-labeled probes specific to flt1 and kdr (upper panels) and ␤-actin (lower panels) mRNAs. PCR amplification for the latter was performed for 15 cycles. The flt1 blot was exposed for 32 h, and the kdr blot for 1 h. RT(Ϫ), the reaction without reverse transcriptase. F, RT-PCR analysis of VEGF receptor mRNAs in pericytes. Thirty nanograms of poly(A) ϩ RNA from pericytes incubated under the indicated O 2 concentrations was amplified by RT-PCR and hybridized with 32 P-end-labeled probes specific to flt1 (upper panel) and ␤-actin (lower panel) mRNAs. PCR amplification for the latter was performed for 15 cycles. The film was exposed for 7 days.
gions of VEGF Also Inhibit Endothelial Cell Synthesis of DNA-As additional controls, we prepared 10 independent antisense oligodeoxyribonucleotide species against different regions of VEGF mRNA, including the 5Ј-and 3Ј-untranslated regions and the VEGF open reading frame. As shown in Fig. 6, six antisense species out of the 10 could significantly inhibit endothelial cell synthesis of DNA under 10% O 2 .

DISCUSSION
The process of angiogenesis is thought to consist of the following four steps: 1) proteolytic degradation of the basement membrane, 2) migration of endothelial cells, 3) proliferation of endothelial cells, and 4) tube formation. The final step is completed when new capillaries are covered with pericytes (1,2). In this study, we have focused on how the proliferation of endothelial cells and pericytes is affected by hypoxia, the leading cause of angiogenesis.
This study has confirmed that low atmospheric O 2 tensions can result in the stimulation of the proliferation of both human umbilical vein endothelial cells and bovine retinal pericytes. The O 2 concentrations that induced maximal growth promotion were 10% for endothelial cells and 2.5% for pericytes. This is comparable with earlier reports that human umbilical vein endothelial cells grew at the greatest rate under 7.5% O 2 (37) and bovine brain microvascular pericytes under 3% O 2 (38).
This study has also demonstrated for the first time that mRNAs coding for VEGF, a potent endothelial cell mitogen, are present not only in endothelial cells per se, but also in pericytes, and that their level is significantly elevated in both cell types as atmospheric O 2 concentrations decrease. Moreover, it is VEGF 121 and VEGF 165 , the secretory forms of VEGF, that are coded for by the mRNAs expressed in the vascular cells. VEGF is a growth factor known to be present in a variety of tissues, including ovary (39,40) and malignant tumors (41)(42)(43)(44), where angiogenesis takes place. Its expression has been shown to be localized around necrotic foci in brain tumors (45) and to be enhanced by O 2 depletion in glioblastoma cell cultures (13,14). Iizuka et al. (46), using an RT-PCR analysis similar to that conducted in this study, recently showed that VEGF mRNA levels are elevated during hypoxia in human osteosarcoma  (Student's t test). q and E, cumulative DNA synthesis without oligodeoxyribonucleotides; and , with antisense oligodeoxyribonucleotides; f and Ⅺ, with sense oligodeoxyribonucleotides.

FIG. 5. Effect of antisense oligodeoxyribonucleotides on VEGF synthesis in endothelial cells.
Endothelial cells were incubated with [ 35 S]methionine in the presence or absence of the indicated concentrations of antisense or sense oligodeoxyribonucleotides, lysed, and immunoprecipitated as described under "Experimental Procedures." A, fluorogram of total labeled proteins. Proteins were electrophoresed on a 12% SDS-polyacrylamide gel under reducing conditions. Ten-microliter aliquots of cell lysate were loaded per lane. Bars on the left indicate molecular mass markers in kilodaltons. B, immunoprecipitates. Immunoreacted materials that had been prepared from 8.0 ϫ 10 6 dpm of lysate each except for pericytes (4.0 ϫ 10 6 dpm) were electrophoresed under the same conditions as described for A. Specific immunoprecipitates were marked at 22 and 18 kDa. Note that pericytes synthesized VEGF, as did U251 cells. cells. From these observations, VEGF has been regarded as the principal angiogenic factor under ischemic and hypoxic conditions. Our finding that vascular cells per se can express the VEGF gene in response to hypoxia would seem, therefore, to have significant implications in angiogenesis. Vascular VEGF may participate in the process of angiogenesis in autocrine and paracrine manners.
In general, the actions of growth factors are mediated by receptors on their target cells (47,48). In this study, we have also identified VEGF receptor subtypes expressed in endothelial cells and pericytes. Endothelial cells were found to constitutively express the two genes for flt1 and kdr, regardless of the atmospheric O 2 tensions. The rank order of mRNA abundance was kdr Ͼ Ͼ flt1, suggesting that Kdr is the major VEGF receptor species expressed in human umbilical vein endothelial cells. On the other hand, only flt1 mRNA was detected in pericytes under hypoxic conditions. The difference in the mode of flt1 gene expression between the two cell types is of interest in understanding the regulation of VEGF receptor gene expression. Transcripts of the kdr gene were not detected in pericytes; given the relatively low level of the transcript as determined by Northern blot analysis with human kdr cDNA (data not shown) and the lack of bovine sequence, it was unfortunate that RT-PCR could not be used to evaluate the level of the kdr transcript.
The results obtained indicate that vascular endothelial cells per se possess a system for triggering their own growth. In endothelial cells, VEGF expression was inducible by low O 2 concentrations, whereas receptor expressions were constitutive. This may be an indication that ligand expression is a rate-limiting step in the putative autocrine actions of VEGF, as in the case of the keratinocyte growth factor system in dermal wound healing (49). We therefore manipulated VEGF expression with antisense oligodeoxyribonucleotides to test the functional role of the VEGF system. Antisense oligodeoxyribonucleotides complementary to the 5Ј-region of VEGF mRNA, encompassing the initiator codon, were found to cause a doseand time-dependent inhibition of DNA synthesis in endothelial cells cultured under hypoxic conditions (Fig. 4), probably through a blockage of the translation of VEGF mRNA (Fig. 5). We also conducted control experiments with additional antisense oligodeoxyribonucleotide species against different regions of VEGF mRNA, including the 5Ј-and 3Ј-untranslated regions and the VEGF open reading frame, the majority of which (6 of 10) could also efficiently inhibit endothelial cell synthesis of DNA (Fig. 6); arrests of the ribosome transition (50) and an RNase H-like activity-driven degradation of target mRNA (51) would account for the antisense DNA action. The results could be regarded as evidence that vascular VEGF is causally related to the hypoxia-induced proliferation of endothelial cells. 2 In light of these findings, we propose a model for the mechanism of the hypoxia-induced proliferation of vascular cells (Fig. 7). When the local oxygen concentration is lowered, VEGF gene expression would be induced in endothelial cells and in pericytes to produce secretory forms of VEGF. VEGF in turn may act on Kdr and Flt1 receptors on endothelial cells in autocrine and paracrine manners, thereby causing the proliferation of endothelial cells, which may lead to angiogenesis. Basal amounts of vascular VEGF synthesized in normoxic states may promote the maintenance of microvascular homeostasis, as suggested by our observation that the antisense VEGF oligodeoxyribonucleotide could modify endothelial cell DNA synthesis under 20% O 2 (Fig. 4C). According to this model, we can also suggest a possible approach for the prevention of angiogenesis. Interruptions of the series of biochemical events at certain steps might halt the process of angiogenesis, e.g. antisense DNA/RNA against VEGF mRNA may have therapeutic potential in the treatment of proliferative angiopathies or tumors. In support of the above-mentioned model, Aiello   al. (53) recently reported that ocular VEGF levels were abnormally high in a large population of patients with actively proliferative diabetic retinopathy, but dropped following successful treatment.
The model proposed was based on in vitro observations with cultivated endothelial cells and pericytes. Although it can partially explain how hypoxia causes endothelial proliferation, the involvement of other growth factors must also be taken into account when considering the processes that would occur in vivo. In addition to VEGF, acidic fibroblast growth factor, basic fibroblast growth factor (bFGF), epidermal growth factor, transforming growth factor-␣, transforming growth factor-␤ (TGF-␤), and platelet-derived growth factor B (PDGF-B) have been implicated in angiogenesis (54 -57). Among them, bFGF, TGF-␤, and PDGF-B seem particularly important because they can be produced by endothelial cells themselves, thus possessing potential autocrine/paracrine activities. bFGF is a mitogen for a wide variety of cell types, playing diverse roles in vascular and nervous systems as well as in connective tissues (58). Concerning vascular functions, bFGF has been reported to stimulate not only endothelial cell mitosis and chemotaxis, but also tube formation. Furthermore, it can induce collagenases and plasminogen activator, which would promote degradation of the basement membrane of the parental vessels (59). These observations indicate that bFGF may be related to all the steps of angiogenesis. However, the expression of this growth factor has been shown not to be influenced by hypoxia (60). TGF-␤ inhibits the proliferation of endothelial cells in culture (61), but is known to induce new capillary formation in vivo (56). This apparently paradoxical effect of TGF-␤ may be explained by the fact that this factor is chemotactic for macrophages and causes them to release angiogenic factors (56). However, TGF-␤ expression has also been shown not to be affected by hypoxia (60). PDGF-B is a major serum mitogen for mesenchymally derived cells. Since PDGF-B is not only released by platelets, but also secreted by cells involved in inflammatory responses, it has been suggested to play a role in wound healing (10). Although PDGF-B was previously thought to be devoid of mitogenic activity on endothelial cells, Funa et al. (62) have recently demonstrated that functional PDGF-B receptors are expressed on hyperplastic capillary endothelial cells in malignant glioma, suggesting that autocrine PDGF-B has a role in the proliferation of endothelial cells. In contrast to bFGF and TGF-␤, hypoxia-induced up-regulation of the PDGF-B gene has also been reported (60). Available evidence thus suggests that the major autocrine/paracrine growth factors involved in the control of endothelial cell growth under normoxic conditions would be bFGF, VEGF, and PDGF-B. Under hypoxic conditions, induced VEGF and PDGF-B would mainly account for the endothelial proliferation.
We have shown previously that pericytes can restrict the replication of co-cultured endothelial cells and suggested TGF-␤ and heparan sulfate as the candidate regulatory molecules (4). We also have shown that endothelium-dependent stimulation of pericyte growth is mediated mainly by endothelin-1 (5), a hypoxia-inducible mitogen (63). In the present study, pericytes were demonstrated to express the VEGF gene in response to hypoxia, as do endothelial cells. Therefore, pericytes may be regarded as cells that can exert both negative and positive growth effects on their microvascular counterpart. Although the role of pericytes during angiogenesis is poorly understood, it is likely that hypoxia would turn them predominantly mitogenic, thereby promoting the growth of endothelial cells together with that of pericytes through synergistic actions of VEGF, PDGF-B, and endothelin-1. Whether VEGF in fact stimulates pericyte replication, however, remains to be established because of the following observations. Midy and Plouët (64) reported that VEGF stimulated the migration, but not mitosis, of bovine osteoblasts, a cell type of the same mesenchymal origin as pericytes. Vascular smooth muscle cells, the equivalent of pericytes in larger vessels, have also been reported not to respond to VEGF (16). Seetharam et al. (65) showed that transfection of flt1 cDNA failed to confer responsiveness to VEGF on NIH3T3 cells.