Hypoxia in combination with FGF-2 induces tube formation by human microvascular endothelial cells in a fibrin matrix: involvement of at least two signal transduction pathways

The growth of new blood vessels or angiogenesis takes place in response to angiogenic factors by a series of discrete, but overlapping, phases, including vascular dilation and increased vascular permeability, endothelial cell activation, degradation of the endothelial cell basal lamina, endothelial cell migration and proliferation, lumen formation and capillary stabilisation (Pepper, 1997). Analysis of the array of angiogenesis stimulators strongly suggests that there must be several different mechanisms for stimulating capillary growth. Not only angiogenic growth factors and cytokines, but also environmental factors such as the matrix composition and oxygen tension, influence the angiogenic response.

In a wound area, fibrin exudates are formed as a temporary matrix and stimulate angiogenesis (Dvorak et al., 1987). Previously we have shown that human microvascular endothelial cells (hMVEC) form capillary-like tubular structures in a fibrin matrix after stimulation with an angiogenic growth factor (fibroblast growth factor (FGF)-2 or vascular endothelial growth factor (VEGF)) in combination with the cytokine tumour necrosis factor (TNF)α (Koolwijk et al., 1996). These angiogenic factors induce, among other effects, the enhancement of components of the plasminogen activator system, namely urokinase-type plasminogen activator (u-PA) and its cell-bound receptor u-PA receptor (u-PAR), thereby increasing proteolytic capacity and allowing hMVEC to degrade the extracellular matrix (Koolwijk et al., 1996).

Hypoxia is a common feature of many of the pathological conditions in which neovascular growth is observed. Different cell types are able to sense the drop in oxygen level and react in various ways. Macrophages secrete growth factors like VEGF (Xiong et al., 1998), FGF-2 (Ishibashi et al., 1995) and platelet derived growth factor (Kuwabara et al., 1995) in response to hypoxia. Endothelial cells enhance the expression of specific integrins (Suzuma et al., 1998) and enhance their proteolytic activity (Graham et al., 1998; Kroon et al., 2000). The mechanism by which endothelial cells sense the oxygen level is not yet known, but it has been shown that hypoxia influences the activation of several components of signal transduction pathways. The transcription factor nuclear factor κB (NF-κB), as well as the members of the mitogen-activated protein kinase (MAPK) superfamily, consisting of extracellular-related kinase (ERK)1/2, c-jun NH₂-terminal kinase (JNK) 1/2 and p38, have been shown to be activated by hypoxia (Laderoute et al., 1999; Conrad et al., 1999a; Conrad et al., 1999b; Koong et al., 1994b). Phosphorylation and thus activation of these MAPK after stimulation with growth factors such as FGF-2 and TNFα has been correlated with the process of angiogenesis; NF-κB and NF-κB-like sites are located in the promoter region of target genes in endothelial cells, whose expression is critical to the initiation of capillary formation by stimulated endothelial cells (Read et al., 1994; Gerritsen and Bloor, 1993). ERK activation has been implicated in FGF-2-mediated angiogenesis in the chorioallantoic membrane (Eliceiri et al., 1998). The function of the stress-activated kinase cascades in endothelial cell activation is not well understood. A role of p38 in the migration of human umbilical vein endothelial cells through actin reorganisation is, however, shown by Rousseau et al. (Rousseau et al., 1997).

Hypoxia can induce angiogenesis both in vitro and in vivo situations, and the angiogenic response is even more pronounced when a growth factor is present (Takeshita et al., 1995; Tsurumi et al., 1997). After delivering of FGF-2 (Baffour et al., 1992; Lazarous et al., 1995; Ueno et al., 1997) or VEGF (Takeshita et al., 1995; Banai et al., 1994) to ischaemic legs or myocardia, improved blood flow in these regions has been reported. The mechanism by which the endothelial cells are activated in such situations is not fully understood. In order to understand this mechanism, capillary-like tube formation in fibrin matrices by human microvascular endothelial cells, stimulated with hypoxia in combination with FGF-2, was studied. We focussed on the proteolytic capacity of the cells and the induced signal transduction pathways in hypoxic culture conditions.

Cell culture materials were purchased as described (Kroon et al., 1999). Human fibrinogen was obtained from Chromogenix AB (Mölndal, Sweden); Factor XIII was generously provided by Drs H. Boeder and P. Kappus (Centeon Pharma GmbH, Marburg, Germany) and thrombin came from Organon Technika (Boxtel, The Netherlands). FGF-2 was purchased from Pepro Tech EC (London, England), and VEGF₁₆₅ and soluble-VEGFR-1 were a kind gift of Dr H. Weich (GFB, Braunschweig, Germany). Human recombinant TNFα was a gift from Dr J. Travernier (Biogent, Gent, Belgium) and contained 2.45×10⁷ U/mg protein and less than 40 ng lipopolysaccharide per mg protein. Aprotinin was purchased from Pentapharm Ltd (Basel, Switzerland). Rabbit polyclonal anti-u-PA antibodies were prepared in our laboratory (Wijngaards et al., 1982) as well as the rabbit polyclonal anti-t-PA antibodies. The monoclonal u-PAR-blocking antibody H-2 was a kind gift from Dr U. Weidle (Boehringer Mannheim, Penzberg, Germany) (Bohuslav et al., 1995). The polyclonal antibodies recognising p65 and IκBα came from Santa Cruz (Santa Cruz, CA, USA). Polyclonal antibodies recognising phospho-p44/p42 (Erk1/2) MAP kinase and phospho-p38 came from New England Biolabs (Beverly, MA, USA), polyclonal anti-activeNAPCO® JNK antibodies from Promega (Madison, WI, USA). Horseradish-conjugated goat-antirabbit antibodies were bought at Roche Diagnostics Nederland BV (Almere, The Netherlands) and FITC-conjugated swine-antirabbit antibodies came from Dako (Glostrup, Denmark). PD98059 and SB203580 were purchased from Alexis Biochemicals (Läufelfingen, Switzerland).

The following cDNA fragments were used as probes in the hybridisation experiments: a 1023 bp fragment of the human u-PA cDNA (a kind gift from Dr W.-D. Schleuning, Schering AG, Berlin, Germany), a 585 bp BamHI fragment of the human u-PAR cDNA (a kind gift from Dr F. Blasi, Milan, Italy) and a 1200 bp PstI fragment of hamster actin cDNA.

Human foreskin microvascular endothelial cells (hMVEC) were isolated, cultured and characterised as previously described (Van Hinsbergh et al., 1987; Defilippi et al., 1991). HMVEC were cultured on gelatine-coated dishes in M199 supplemented with 20 mM Hepes (pH 7.3), 10% human serum (HS), 10% heat-inactivated newborn calf serum (NBCS), 150 μg/ml crude endothelial cell growth factor, 2 mM L-glutamine, 5 U/ml heparin, 100 i.u./ml penicillin and 100 μg/ml streptomycin at 37°C under a 5% CO₂/95% air atmosphere, unless mentioned otherwise. Experiments were performed with confluent cells (0.7×10⁵ cells/cm²), which had been cultured without growth factor for at least 24 hours.

For culturing in hypoxic conditions, hMVEC were placed in a NAPCO® incubator (serial number 7101-C1, Precision Scientific Inc, Chicago, USA), which controls the oxygen concentration by flushing with N₂. Oxygen levels in the incubator were monitored by an internal oxygen sensor as well as by external calibration using Dräger Tubes 6728081 (Drägerwerk Ag, Lübeck, Germany). The hypoxic condition is defined as culturing at 37°C under 1% O₂ /5% CO₂ atmosphere.

Human fibrin matrices were prepared by the addition of 0.1 U/ml thrombin to a mixture of 2.5 U/ml factor XIII (final concentrations), 2 mg/ml fibrinogen, 2 mg/ml sodium citrate, 0.8 mg/ml NaCl and 3 μg/ml plasminogen in M199 medium (mixture pH 7.4). 300 μl portions of this mixture were added to the wells of 48-well plates. After clotting at room temperature, the fibrin matrices were soaked with M199 supplemented with 10% (v/v) HS and 10% (v/v) NBCS for 2 hours at 37°C to inactivate the thrombin. Confluent endothelial cells (0.7×10⁵ cells/cm²) were detached and seeded in a 1.25:1 split ratio on the fibrin matrices to form a highly confluent monolayer. After culturing for 24 hours in M199 medium supplemented with 10% HS, 10% NBCS and penicillin/streptomycin, the endothelial cells were stimulated with the mediators for the time indicated. At the end of the culture period the media were collected and the formation of tubular structures of endothelial cells in the three-dimensional fibrin matrix was analysed by phase-contrast microscopy and the total length of tube-like structures of six randomly chosen microscopic fields (7.3 mm²/field) was measured using an Olympus CK2 microscope equipped with a monochrome CCD camera (MX5) connected to a computer with Optimas image analysis software, and expressed as mm/cm².

For histochemistry, matrices were fixed in 2% (w/v) p-formaldehyde in phosphate-buffered saline (PBS; 0.15 M NaCl, 10 mM Na₂PO₄, 1.5 mM KH₂PO₄, pH 7.4) embedded in Technovit 8100 (Heraeus Kulzer, Wehrheim, Germany), sectioned at 4 μm and stained with phloxin.

u-PA, t-PA, PAI-1 and VEGF antigen, as well as fibrin degradation product (FDP) determinations, were performed by commercially available immunoassay kits: uPA EIA HS Taurus (Leiden, The Netherlands), Thrombonostika t-PA (Organon-Teknika, Turnhout, Belgium), IMULYSE® PAI-1(Biopool, Umea, Sweden), VEGF ELISA (R&D system, Minneapolis, USA) and Fibrinostika® FDP ELISA (Organon-Teknika, Turnhout, Belgium). MCP-1 ELISA was previously described (Peri et al., 1994).

Determination of specific binding of diisopropylfluorophosphate-treated u-PA (Ukidan®) (DIP-u-PA) to hMVEC was described previously (Kroon et al., 1999).

Total RNA was isolated as described (Chomczynski and Sacchi, 1987) and electrophoresed in a 1.2% (w/v) agarose gel under denaturing conditions using 1 M formaldehyde. Northern blotting was performed as described (Lansink et al., 1998).

HMVEC were cultured to confluency on glass coverslips (14 mm diameter). Cells were treated with different stimulators and incubated for the time indicated. After the incubation, cells were fixed for 5 minutes in ice-cold methanol and air-dried. After three washes with PBS containing 0.1% bovine serum albumin (BSA), the coverslips were incubated for 20 minutes in a block-buffer (PBS + 10% NBCS). The cells were washed three times and incubated with the first antibody (anti-NF-κB p65, 1:100 in PBS + 0.1% BSA). After a 1 hour incubation followed by three washes, the cells were incubated with the second antibody (FITC-conjugated swine-antirabbit, 1:50 in PBS + 0.1% BSA). Finally, the cells were dehydrated in graded steps of ethanol and attached to a slide with a drop of Vectashield (Vector Laboratories, Burlingame, CA, USA).

Equal numbers of hMVEC, cultured on gelatin-coated wells (5 cm² wells/condition), were washed with ice-cold PBS and lysed in a lysis buffer containing 10% glycerol (v/v), 3% SDS (w/v), 0.125 M Tris-PO₄ pH 6.7 and 10% β-mercaptoethanol (v/v). Protein was separated on a 10% SDS-polyacrylamide gel and electrophoretically transferred onto polyvinylidene difluoride membrane (Amersham, Uppsala, Sweden) in a buffer of 192 mM glycine, 25 mM Tris (pH 8.3) and 20% (v/v) methanol. The filters were blocked with 5% (w/v) non-fat milk in 137 mM NaCl, 20 mM Tris (pH 7.6) and 0.25% Tween 20 (TBST) for 1 hour, followed by overnight incubation at 4°C with the primary polyclonal antibodies (anti-IκBα 1:10000, antiphospho-ERK1/2 1:1000, anti-active JNK1/2 1:2000, antiphospho-p38 1:1000) in TBST + 5% non-fat milk + 0.5 mM sodium orthovanadate. Subsequently, the blots were washed three times with TBST and incubated for 1 hour at room temperature with horseradish-conjugated goat-antirabbit antibodies (1:5000) in TBST + 5% nonfat milk + 0.5 mM sodium orthovanadate, as a conjugate. The bands were visualised with enhanced chemiluminescence (Sigma, St Louis, MI, USA).

Data are expressed as means ± s.d. The unpaired t-test was used for comparison of groups with equal variance and normal distribution. P<0.05 was considered statistically significant.

Under 20% oxygen atmosphere (normoxic conditions), hMVEC required exposure to an angiogenic growth factor (FGF-2) and the cytokine tumour necrosis factor (TNF)α to form capillary-like tubular structures in the fibrin matrix (Fig. 1A) (Koolwijk et al., 1996). When the experiments were performed under hypoxic conditions (1% oxygen atmosphere), an increase in the extent of tube formation was observed at all FGF-2 concentrations used (5-50 ng/ml) (Figs 1B, 2). Normoxic hMVEC, grown on a fibrin matrix for 5 days and stimulated with FGF-2 only, neither formed tubes nor sustained an intact monolayer (Fig. 1C), probably due to excessive fibrin degradation induced by t-PA (Collen et al., 1998). In cross sections of these three-dimensional fibrin matrices, gaps were visible in the monolayer (arrow, Fig. 1E). In the presence of anti-t-PA IgG, gap formation in the monolayers was prevented, but tubes could still not be induced (data not shown). Interestingly, in the hypoxic condition the sole addition of ≥30 ng/ml FGF-2 induced tubular structures (arrow, Figs 1D,F, 2). This difference was not due to an altered fibrin degradation, since no significant difference in fibrin degradation products was measured (15.1±1.3 μg/ml versus 14.2±1.4 μg/ml, respectively, n=3) in the conditioned media of normoxia/FGF-2- and hypoxia/FGF-2-stimulated cells.

To investigate whether VEGF₁₆₅ induction contributed to the hypoxia/FGF-2-dependent tube formation, soluble-VEGFR-1 was added (29 nM; a 50-fold excess over the VEGF₁₆₅ concentration used in the control experiments) to the medium in combination with FGF-2. Soluble-VEGFR-1 did not affect the tube formation of hypoxia/FGF-2-stimulated hMVEC (Fig. 3A). In the same group of experiments soluble-VEGFR-1 inhibited the VEGF₁₆₅+TNFα-induced tube formation by 74±3%, indicating the ability of soluble-VEGFR-1 to neutralise VEGF₁₆₅ in the medium.

The capillary-like structures, formed by hypoxia/FGF-2-stimulated hMVEC, were inhibited by antibodies against u-PA (93±4%, n=3, P=0.02), u-PAR (72±2%, n=3, P=0.01) and the plasminogen inhibitor aprotinin (94±5%, n=3, P=0.01), but not significantly by antibodies against t-PA (26±5%, n=3, P=0.06) (Fig. 3B), in agreement with our previous findings (Collen et al., 1998), indicating that t-PA does not play a significant role in tube formation in this in vitro system.

Because cell-bound u-PA activity is required to induce capillary-like structures under our experimental conditions,

we evaluated whether hypoxia affected the production of fibrinolytic proteins in non- and FGF-2-stimulated cells.

In hypoxia- and hypoxia/FGF-2-stimulated hMVEC, u-PAR mRNA was enhanced as compared to their normoxic counterparts (Fig. 4A). In addition, u-PAR antigen was enhanced in hypoxic hMVEC as measured by specific ¹²⁵I-DIP-u-PA binding. This increase was significant for hypoxia/FGF-2-stimulated hMVEC (138±11%, P=0.03, n=3, performed in duplicate wells) (Fig. 4B). In addition, u-PAR antigen was measured by a u-PAR ELISA. An increase of 146±2% (n=3) in u-PAR antigen was found in FGF-2/hypoxia-stimulated cells as compared to their normoxic counterparts.

In hypoxic conditions less u-PA antigen accumulated in the conditioned media (Fig. 4C). However, when the cellular uptake of the u-PA:PAI-1 complex was prevented by blocking the interaction of u-PA with its receptor u-PAR by the blocking antibody H-2, the decrease in u-PA accumulation in hypoxic conditions was completely abolished and comparable amounts of u-PA were produced (Fig. 4C). In addition, no difference in u-PA production was detected between control and FGF-2-stimulated cells. As only the u-PA:PAI-1 complex is internalised, this indicates that u-PA had been activated. A significant increase in the amount of relatively internalised u-PA was observed in hypoxic/FGF-2-stimulated cells as compared to their normoxic counterparts (6±0.3% increase; n=3; P=0.03). The amounts of t-PA produced by FGF-2-stimulated hMVEC were comparable in hypoxic and normoxic culture conditions (1.5±0.1 versus 1.4±0.3 ng t-PA/10⁵ cells, respectively, n=3). This was also the case for the PAI-1 levels in the conditioned media (162±16 versus 164±13 ng PAI-1/10⁵ cells, respectively, n=4).

To evaluate whether hypoxia induces an activation of endothelial cells similar to that mediated by TNFα, the activation of NF-κB was investigated. To this end, translocation of NF-κB to the nucleus was determined via immunofluorescence staining of the p65 subunit of NF-κB. In control hMVEC a diffuse immunofluorescence staining was observed in both normoxic and hypoxic cells (Fig. 5A,B). Translocation of p65 was observed in several cells, probably due to endothelial cell activation caused by culturing on glass coverslips. After stimulation with TNFα in normoxic or hypoxic conditions for 8 hours, all p65 was translocated to the nucleus (Fig. 5C,D). In normoxic conditions, stimulation with FGF-2 for 8 hours did not result in the translocation of p65 (Fig. 5E). However, in hypoxia/FGF-2-stimulated cells a considerable part of p65 was translocated to the nucleus (Fig. 5F). After 16 hours of incubation this translocation was no longer observed in hypoxia/FGF-2-stimulated cells, in contrast to TNFα-stimulated hMVEC (data not shown).

Because nuclear translocation of NF-κB was found after 8 hours of hypoxia, we investigated whether the cellular content of IκBα was altered by hypoxia. Fig. 6 shows that the amount of IκBα-antigen did not change at various times of exposure to hypoxia (1-16 hours), compared to normoxic exposure.

A sensitive reflection of endothelial activation is the production of monocytic chemoattractant protein (MCP)-1. Hypoxia did not significantly alter MCP-1 production in non-stimulated cells. However, hypoxia/FGF-2-stimulated hMVEC produced significantly more MCP-1 after an incubation period of 8 hours than their normoxic counterparts (2.0±0.9-fold, P=0.002, n=3 performed in duplicate wells) (Fig. 7). Hypoxia/TNFα-stimulated cells also produced more MCP-1 than normoxia/TNFα-stimulated cells, but this did not reach statistical significance (1.5±0.5-fold, P=0.066, n=3 performed in duplicate wells).

Subsequently the effect of hypoxia/FGF-2-stimulation on the activity of members of the MAPK family (JNK 1/2, p38 and ERK 1/2) was evaluated. In non-stimulated or hMVEC stimulated for 4 hours with FGF-2, hardly any phosphorylated JNK1/2 and p38 were detected in normoxic and hypoxic conditions, whereas TNFα induced phosphorylation of JNK1, JNK2 and p38 both in normoxic and hypoxic conditions (Fig. 8).

These data were confirmed by the observation that hypoxia/FGF-2-induced tube formation was not influenced by SB203580, a specific inhibitor of p38 MAPK (Fig. 9). FGF-2+TNFα-induced tube formation was partly blocked by SB203580 (58±8%, two independent experiments, data not shown). On the contrary, PD98059, a specific inhibitor of ERK1/2 phosphorylation, was able to inhibit hypoxia/FGF-2-induced tube formation completely (Fig. 9).

In normoxic conditions, treatment of hMVEC with FGF-2 or TNFα enhanced phosphorylation of ERK1 and ERK2 (Fig. 10A), FGF-2 being a stronger stimulator than TNFα. After 4 hours of hypoxia, phosphorylation of ERK1 as well as ERK 2 was increased as compared to normoxia, both under basal (176±21%, n=3) and FGF-2-stimulated (141±14%, n=3) or TNFα-stimulated conditions (119±7%, n=3)(Fig. 10A). No difference in phosporylation between ERK1 and 2 was detected. The same results were found with cells cultured on a gelatin or fibrin coating (data not shown). In addition, PD98059 was able to almost completely inhibit hypoxia/FGF-2-induced ERK 1/2 phosphorylation, while SB 203580 had no effect on ERK1/2 phosphorylation (Fig. 10B).

In normoxic culture conditions both a growth factor and the cytokine TNFα are required for the formation of capillary-like structures by human microvascular endothelial cells (hMVEC) in a fibrin matrix (Koolwijk et al., 1996; Kroon et al., 1999). In the current report we have shown that in hypoxic conditions, stimulation solely with the growth factor FGF-2 is sufficient to induce in vitro angiogenesis in a fibrin matrix. Hypoxia activates at least two signal transduction pathways in FGF-2-stimulated cells; the NF-κB pathway and the ERK1/2 pathway. Both pathways may be involved in the stimulating effect of hypoxia on tube formation.

Several studies have shown that the administration of angiogenic growth factors, such as VEGF and FGF-2, can result in an improvement of blood flow in ischaemic limbs and myocardia (Takeshita et al., 1995; Tsurumi et al., 1997; Baffour et al., 1992; Lazarous et al., 1995; Ueno et al., 1997). Our observation that FGF-2 induces capillary tubes in hypoxic conditions, but not in normoxic conditions, agrees with these findings. In contrast, FGF-2 induced the formation of capillary-like structures in bovine microvascular endothelial cells under normoxic conditions (Pepper et al., 1990). This induction was mediated by the induction of autocrine VEGF expression (Seghezzi et al., 1998). In human MVEC, FGF-2 only induced the formation of capillary-like structures if an additional factor such as TNFα or hypoxia was present (Koolwijk et al., 1996; this study). In this study, we demonstrated that under our experimental conditions the FGF-2/hypoxia-induced tube formation did not depend on the endogenous production of VEGF. This may be related to the fact that hMVEC do not produce VEGF (as quantified by RT-PCR; our own observations). Several mechanisms may contribute to the hypoxia/FGF-2-enhanced formation of capillary-like structures. FGF-2 and hypoxia may affect matrix-degrading proteases (Pepper et al., 1990; Wojta et al., 1988; Mignatti et al., 1991). Proteolytic enzymes may stimulate the invasion by stimulating controlled pericellular proteolysis, or prevent invasion of endothelial cells by excessive lysis of the fibrin matrix (Montesano et al., 1987). When hMVEC were stimulated with FGF-2 in the absence of TNFα, the intact monolayer of hMVEC could not be sustained on a fibrin matrix, due to excessive fibrin degradation by t-PA-dependent plasmin activation (Collen et al., 1998). However, although titration of anti-t-PA antibodies reduced the lysis in a concentration dependent way and restored monolayer integrity, no capillary-like tubular structures were formed (Collen et al., 1998). Therefore, it is unlikely that the fibrin lysis per se is the only factor that prevents the ingrowth of endothelial tubes. Induction of cellular signalling pathways and additional factors by hypoxia apparently underlies the hypoxia/FGF-2-stimulated tube formation.

Hypoxia/FGF-2-stimulated hMVEC had comparable overall production of u-PA, but expressed higher levels of u-PAR than their normoxic counterparts. Similar to previous observations in normoxic FGF-2/TNFα-stimulated conditions, the hypoxia/FGF-2-induced tube formation requires cell-bound u-PA and plasmin activities (Koolwijk et al., 1996; Kroon et al., 1999). Although FGF-2 affects the production of t-PA and PAI-1 (Yamamoto et al., 1994; Pepper et al., 1990), hypoxia did not alter the overall production of u-PA, t-PA and PAI-1 in FGF-2-stimulated cells, but significantly increased the relative uptake of u-PA by hypoxia/FGF-2-stimulated cells. It is unlikely that an accompaning moderate increase in u-PAR-dependent pericellular fibrinolysis in FGF-2-stimulated hMVEC is the major contributor to the induction of tubes by hypoxia. Alternatively, u-PAR may enforce cell attachment, in particular in relation to the vitronectin-binding integrins (Wei et al., 1996) and thus contribute to the maintenance of the cell monolayer on fibrin. In this context it is of interest to note that α_v-integrin is upregulated in endothelial cells by hypoxia (Suzuma et al., 1998; Kroon et al., 2000).

Hypoxia has multiple effects on endothelial cells and multiple signal transduction pathways important for the process of angiogenesis can be triggered. Our data imply that hypoxia in combination with FGF-2 induces the activation of the subunit p65 of NF-κB and the activation of members of the MAPK superfamily. The transcription factor NF-κB regulates a number of genes involved in angiogenesis (such as TNFα, IL-2, IL-6, VCAM-1 and u-PA) (Read et al., 1994; Gerritsen and Bloor, 1993; Guerrini et al., 1996), which indicates that activation of NF-κB may be of importance in tube formation by hypoxia/FGF-2-stimulated hMVEC. Although NF-κB is involved in the regulation of the u-PA gene, u-PA expression in hMVEC was not enhanced by hypoxia. It has been described that u-PA expression is regulated by the Rel-A subunit of NF-κB (Reuning et al., 1995), but also by C-Rel transcription factor (Reuning et al., 1999; Hansen et al., 1992). The complexity of the regulation of the u-PA gene might explain why hypoxia per se is not sufficient to increase in u-PA expression in hMVEC.

Koong et al. reported activation of NF-κB upon exposure to severe hypoxia (0.02% oxygen) (Koong et al., 1994a). This activation was accompanied by phosphorylation, dissociation and degradation of IκB (Koong et al., 1994a; Zhang et al., 1998). Under our experimental conditions we did not detect a decrease in the cellular IκBα content, but this does not exclude a transiently reduced IκB activity at 1% oxygen tension. The low oxygen tension probably lowers the nitric oxide (NO) production in endothelial cells (Liao et al., 1995). Because NO reduces NF-κB activation (De Caterina et al., 1995; Zeiher et al., 1995) by stabilising IκB (Peng et al., 1995), a reduced NO production will facilitate NF-κB activation and its subsequent nuclear translocation.

Other signal transduction pathways are also affected by hypoxia (Faller, 1999); the members of the MAPK superfamily, p38, JNK1/2 and ERK1/2, have all been reported to be influenced by hypoxia (Conrad et al., 1999a; Conrad et al., 1999b; Laderoute et al., 1999). In our experimental setting, p38 and JNK1/2 phosphorylation was only induced by TNFα-stimulation and was not changed by hypoxic conditions. In addition, SB203580, a specific inhibitor of p38, did not inhibit hypoxia/FGF-2-induced tube formation. These data indicate that p38 and JNK1/2 activity are not essential for hypoxia-induced tube formation by human MVEC in a fibrin matrix.

In contrast to p38 and JNK1/2, ERK1/2 phosphorylation in hMVEC was increased by hypoxia, as shown by other groups (Minet et al., 2000; Müller et al., 1997). There are several possible ways in which ERK1 activation can lead to increased angiogenesis in hypoxic conditions. Tanaka et al. showed that PD98059, a specific inhibitor of ERK1/2 phosphorylation, can inhibit induction of ETS-1 mRNA, a prototype of ets family transcription factors (Tanaka et al., 1999). ETS-1 converts endothelial cells to an invasive phenotype by inducing the expression of proteases including u-PA, MMP-1, -3 and -9 as well as β-3 integrin subunit as target genes in endothelial cells (Oda et al., 1999). Recently, it has been shown by several groups that ERK activation by hypoxia is involved in the activation of the hypoxia-inducible factor (HIF)-1α (Richard et al., 1999; Minet et al., 2000) as well as the activation of HIF-2α (Conrad et al., 1999a). HIF-1α regulates many genes, some of which are involved in the organisation of vascular networks (Ryan et al., 1998). Active ERK1/2 induced a rapid phosphorylation of HIF-1α, whereas p38 and JNK were not capable of phosphorylating HIF-1α (Richard et al., 1999).

The activation of NF-κB and ERK1/2 is observed very soon after hypoxic incubation. A rapid and short-lasting activation of NF-kB by hypoxia has previously been observed (Koong et al., 1994a; Koong et al., 1994b) and MAPK activation can be found as early as 5-15 minutes after hypoxic stimulation (Müller et al., 1997). Despite the quick and often short activation of these factors, their actions are described as important in such lengthy processes as angiogenesis (Stoltz et al., 1996; Tanaka et al., 1999). Obviously, the activation of transcription factors such as NF-κB and ERK1/2 is important for the initiation of the angiogenic response, for example by mediating the expression of factors that can propagate the angiogenic response. These factors could be essential for the angiogenic potential of the endothelial cells themselves or could indirectly influence the angiogenic response in hypoxic conditions. For example, NF-κB regulates the expression of MCP-1 (Marumo et al., 1999). It has been suggested that MCP-1 is a chemokine that recruits leukocytes to sites of inflammation, neovascularization and vascular injury. In hMVEC, MCP-1 production was stimulated by TNFα, but also by exposure to hypoxia in combination with FGF-2. Hypoxic induction of MCP-1 expression could be of importance because of the attraction of inflammatory cells, which are capable of producing angiogenic factors in hypoxic conditions, and in this manner indirectly contribute to hypoxia-induced angiogenesis.

In conclusion, this study shows that in hMVEC, hypoxia in combination with a growth factor induces at least two pathways that might be responsible for the induction of tube-like formation, via direct activation of the endothelial cells or via indirect mechanisms. Understanding the effect of hypoxia in combination with growth factors on endothelial cells may be useful for improving the treatment of patients with chronic ischaemia by administering growth factors.

We wish to thank Erna Peters and Mario Vermeer for their excellent technical assistance and Pascal Nègre for advice on western blotting. This work is supported by the Dutch Heart Foundation grant 95.193 and The Dutch Cancer Society grant TNOP 97-1511.