Proinflammatory cytokines regulate expression of the lymphatic endothelial mitogen vascular endothelial growth factor-C.

Vascular endothelial growth factor (VEGF) is a prime regulator of normal and pathological angiogenesis. Three related endothelial cell growth factors, VEGF-B, VEGF-C, and VEGF-D were recently cloned. We have here studied the regulation of VEGF-C, a lymphatic endothelial growth factor, by angiogenic proinflammatory cytokines. Interleukin (IL)-1beta induced a concentration- and a time-dependent increase in VEGF-C, but not in VEGF-B, mRNA steady-state levels in human lung fibroblasts. The increase in VEGF-C mRNA levels was mainly due to increased transcription rather than elevated mRNA stability as detected by the nuclear run-on method and by following mRNA decay in the presence of an inhibitor of transcription, respectively. In contrast, angiopoietin-1 mRNA, encoding the ligand for the endothelial-specific Tek/Tie-2 receptor, was down-regulated by IL-1beta. Tumor necrosis factor-alpha and IL-1alpha also elevated VEGF-C mRNA steady-state levels, whereas the IL-1 receptor antagonist and dexamethasone inhibited the effect of IL-1beta. Experiments with cycloheximide indicated that the effect of IL-1beta was independent of protein synthesis. Hypoxia, which is an important inducer of VEGF expression, had no effect on VEGF-B or VEGF-C mRNA levels. IL-1beta and tumor necrosis factor-alpha also stimulated the production of VEGF-C protein by the fibroblasts. Cytokines and growth factors have previously been shown to down-regulate VEGF receptors in vascular endothelial cells. We found that the mRNA for the VEGF- and VEGF-C-binding VEGFR-2 (KDR/Flk-1) was stimulated by IL-1beta in human umbilical vein endothelial cells, whereas the mRNA levels of VEGFR-1 (Flt-1) and VEGFR-3 (Flt-4) were not altered. Our data suggest that in addition to VEGF, VEGF-C may also serve as an endothelial stimulus at sites of cytokine activation. In particular, these results raise the possibility that certain proinflammatory cytokines regulate the lymphatic vessels indirectly via VEGF-C.

grown endothelium differentiates into tubelike structures, which are stabilized by mesenchymal components, preventing leakage and fused to allow blood circulation. The significance of angiogenesis is now widely accepted in many physiological and pathological conditions (2). Angiogenesis is needed for embryonic development, for several female reproductive functions, and for wound healing and other repair processes, including collateral blood vessel formation in ischemic limb and heart diseases (1)(2)(3). Angiogenesis also occurs in nonmalignant diseases, such as diabetic retinopathy, atherosclerosis, psoriasis, and rheumatoid arthritis, and its importance in solid tumor growth has been demonstrated in multiple experimental models. Several growth factors and cytokines promote angiogenesis in animal models. Most of them, including tumor necrosis factor (TNF) 1 -␣ and interleukin (IL)-1, do not stimulate endothelial cell growth in culture and are thus called indirect angiogenic factors (4). Once tubes composed of endothelial cells are formed, angiogenesis is apparently completed by stabilization of these vessels through migration of pericytes and production of basement membrane components. Both of these processes have been suggested to be facilitated via endothelial cell specific Tek/Tie-2 receptor ligand angiopoietin-1 (Ang-1) (5)(6)(7)(8). In addition to angiogenic inducers, several endogenous inhibitors of angiogenesis have been recently characterized (2,9). The switch to an angiogenic phenotype in tumors may thus be triggered by overproduction of factors in favor of angiogenesis or reduction of inhibitor release, both of which can depend on the activation of oncogenes and inactivation of antioncogenes (1).
Vascular endothelial growth factor (VEGF), also known as vascular permeability factor or vasculotropin, is a glycoprotein with homology to platelet-derived growth factor (PDGF), and it is translated into multiple protein forms due to alternative exon splicing (reviewed in Refs. 10 -14). VEGF is a prototype of directly acting angiogenic factors, because it is secreted in a biologically active form, its receptors are found at sites of angiogenesis, and it is the most specific endothelial cell growth factor known thus far. VEGF also induces vascular permeability, regulates production of proteases and their inhibitors, and promotes endothelial cell differentiation, movement, and survival, which are related to its angiogenic properties.
Up-regulation of VEGF synthesis by hypoxia and by several indirectly acting angiogenic factors also supports its role in angiogenesis. VEGF transduces its signals through Flt-1 (VEGFR-1) and KDR/Flk-1 (VEGFR-2) tyrosine kinase receptors. However, mitogenesis of endothelial cells may be signaled exclusively via VEGFR-2. Genetic disruption of VEGF or its receptors has indicated that they are necessary for vasculogenesis (differentiation of endothelial cells from their precursors) and angiogenesis (15)(16)(17)(18). Further, injection of VEGF can induce blood vessel formation, whereas antibodies against VEGF, soluble extracellular domain of VEGFR-1 or overexpression of a dominant negative VEGFR-2 inhibit angiogenesis and subsequent tumor growth in animal models (13,14,19,20).
VEGF-B (also known as VEGF-related factor) (21)(22)(23)(24)(25), VEGF-C (also known as VEGF-related protein and VEGF-2) (26 -30), and c-fos-induced growth factor (also known as VEGF-D) (31,32) are recently discovered members of the VEGF family. Thus, this family is currently composed of five members, including also the placenta growth factor (33). VEGF-B is an endothelial cell mitogen, which is expressed in most tissues, but especially highly in the heart and skeletal muscle (23). VEGF-C induces endothelial cell migration and vascular permeability (26,34), but, unlike VEGF, it is a relatively weak mitogen for blood vascular endothelial cells (28,30). VEGF-C mRNA is expressed at low levels in many tissues, most prominently in the heart, placenta, skeletal muscle, ovary, and the small intestine, and it is also present in certain tumor cell lines (26,35). VEGF-C is the ligand for VEGFR-2 and for the third member of the VEGF receptor family, VEGFR-3 (also known as Flt-4), which is espressed mainly on venous endothelium in early embryos and lymphatic endothelium in adult tissues (27,36,37). Indeed, mice overexpressing VEGF-C in basal keratinocytes developed hyperplastic lymphatic vessels in the skin (38). Although VEGF induces growth of vascular and lymphatic endothelial cells in vitro, it seems to be specific for blood vessel growth in vivo (39).
The existence of a gene family consisting of several related growth factors suggests that these family members have overlapping but distinct functions and that they may be differentially regulated. The lymphatic endothelial growth factor VEGF-C (38,39) in particular could be important in inflammatory conditions by controlling the composition and pressure of interstitial fluid and by facilitating lymphocyte trafficking. We have therefore studied the regulation of the expression of VEGF-C in comparison with VEGF and VEGF-B by the angiogenic proinflammatory cytokines and anti-inflammatory agents.
Hypoxia Treatment-Confluent cultures of IMR-90 cells were grown on 10-cm diameter tissue culture plates containing 10 ml of DMEM and 0.5% FCS. These cultures were exposed to hypoxia by incubating the cells for 6 h in the Anaerocult A anaerobic culture jar supplied by Merck (Darmstadt, Germany).
Analysis of Transcription-For nuclear run-on analysis IMR-90 cells were first starved for 48 h and then incubated for 3 h with or without IL-1␤ (10 ng/ml) in DMEM supplemented with 0.5% FCS (10 15-cm dishes each). Total RNA was extracted from one control dish and one IL-1␤-treated dish and used for Northern blotting. The nuclei were isolated from nine control and nine IL-1␤-treated dishes as described by Greenberg and Ziff (42). Extraction of the RNA and nuclear transcription were performed as described by Celano et al. (43). [ 32 P]UTP (3000 Ci/mmol)-labeled nuclear RNA samples (1 ϫ 10 7 cpm/sample) were then hybridized to linearized plasmid DNA (pcDNA I) containing inserts of VEGF-C and GAPDH, or no insert. The probed blots were quantitated with the Bio-Imaging Analyser using the MacBas software, after which background was subtracted from VEGF-C and GAPDH signals.
Analysis of mRNA Half-life-Degradation of the mRNAs were studied in human IMR-90 lung fibroblasts. The cells were starved in DMEM supplemented with 0.5% FCS for 48 h and shifted to medium containing 5% FCS for 4 h prior to mRNA stability measurements. After two washes with DMEM, medium containing 0.5% FCS and actinomycin D (10 g/ml; Sigma) as an inhibitor of transcription, with or without IL-1␤ (10 ng/ml), was added for 1, 2, or 4 h.
Detection of VEGF-C Protein-Confluent IMR-90 cell cultures were first starved for 48 h in DMEM supplemented with 0.5% FCS and then preincubated with or without IL-1␤ (10 ng/ml) or TNF-␣ (50 ng/ml) for 4 h. After two washes with phosphate-buffered saline, metabolic labeling was carried out by addition of 100 Ci/ml of Pro-mix L-35 S in vitro cell labeling mix (Amersham, Buckinghamshire, UK) to methionineand cysteine-free Eagle's minimal essential medium, with or without the cytokines. After 6 h the media were collected and centrifuged to remove cellular debris. Aliquots of the culture media were precipitated with trichloroacetic acid and measured for incorporated radioactivity. Equal counts/min of the media were immunoprecipitated with anti-VEGF antiserum as described earlier (34). The immunoprecipitated proteins were separated in SDS-polyacrylamide gel electrophoresis and visualized and quantitated by the Bio-Imaging analyzer system.
Statistical Analysis-Statistical significance was calculated by Stu-dent's t test. All results are shown as means Ϯ S.E., and p Ͻ 0.05 was selected as the statistically significant value.  Fig. 1B, indicating that the maximal effect was obtained by 1 ng/ml. Proinflammatory Cytokines Induce VEGF-C mRNA Expression-Expression of VEGF-C transcripts was studied in serumstarved human IMR-90 fibroblasts. The cells were exposed to IL-1␤, IL-1␣, or TNF-␣ for 6 h, after which total RNA was isolated, electrophoresed, and subjected to Northern blot hybridization using a mixture of the VEGF, VEGF-B, and VEGF-C probes. The filters were also hybridized separately with probes for cyclooxygenase-2, serving as a known cytokineresponsive gene (41) and for GAPDH, which served as a loading control. As shown in Fig. 2A, IL-1␤, IL-1␣, and TNF-␣ elevated VEGF-C mRNA steady-state levels. Fig. 2B shows a densitometric evaluation of the VEGF-C mRNA levels induced by IL-1␤ from eight independent experiments (a 3.41 Ϯ 0.35-fold increase). IL-1␤ also stimulated VEGF mRNA expression (a 2.55 Ϯ 0.55-fold increase), but not that of VEGF-B. Interestingly, although the steady-state mRNA levels for Ang-1 were readily detectable under basal conditions, they were downregulated by IL-1␤ (10 ng/ml for 6 h) by 50 Ϯ 9% as calculated from seven separate experiments (the controls versus IL-1␤treated, p Ͻ 0.05).

Kinetics and Concentration Dependence of IL-1␤-induced VEGF-C mRNA Expression-To
Effect of Dexamethasone, Indomethacin, Inhibition of Protein Synthesis, and IL-1ra on IL-1␤-induced VEGF-C mRNA Expression- Fig. 3 shows the inhibitory effect of the anti-inflammatory and antiangiogenic glucocorticoid dexamethasone (4) on IL-1␤-induced VEGF and VEGF-C mRNA expression. Furthermore, while dexamethasone completely blocked the effect of IL-1␤ on VEGF mRNA induction, it inhibited only partially the elevation of VEGF-C mRNA steady-stale levels. In contrast, indomethacin, which is a nonsteroidal anti-inflammatory drug and an inhibitor of prostanoid production, did not inhibit the effects of IL-1␤ (not shown).
To investigate whether ongoing protein synthesis is necessary for IL-1␤-induced VEGF-C expression, cycloheximide, an inhibitor of protein synthesis, was administered to the cells. Fig. 3 shows that cycloheximide did not inhibit the effect of IL-1␤, indicating a protein synthesis-independent mechanism of action. IL-1ra was used to investigate the specificity of IL-1␤ effects in IMR-90 cells. This protein binds to the signal transducing IL-1 receptor type I and blocks the effects of IL-1␤ and IL-1␣ (44). Fig. 4 shows that IL-1ra inhibited IL-1␤-induced VEGF-C and VEGF mRNA expression, confirming the receptordependent action of IL-1␤.
Effect of Hypoxia on IL-1␤-induced VEGF-C mRNA Expression-Hypoxia is an important stimulus for angiogenesis and an inducer of VEGF expression (10 -14). In IMR-90 cells, hypoxia induced VEGF mRNA but not VEGF-B or VEGF-C mRNAs (Fig. 5). Further, when hypoxia was combined with IL-1␤ treatment, no additional VEGF stimulation was found in comparison to that of hypoxia alone. This is in contrast with the stimulation obtained with PDGF, which was potentiated by hypoxia (45) (Fig. 5). Moreover, cyclooxygenase-2 expression was recently reported to be induced by hypoxia in human vascular endothelial cells (46), but this did not seem to be the case in the IMR-90 cells (Fig. 5).
Effect of IL-1␤ on VEGF-C mRNA Transcription and Stability-To investigate a possible mechanism contributing to IL-1␤-induced VEGF-C mRNA expression, we performed nuclear run-on experiments, which indicated that transcription of VEGF-C is induced to similar extent as are the steady state levels of VEGF-C mRNA (Fig. 6A). We also measured the half-life of VEGF-C mRNA in the presence of an inhibitor of transcription, actinomycin D, with or without IL-1␤ for 1-4 h. As shown in Fig. 6B, the half-life of VEGF-C in the absence of IL-1␤ was approximately 3.5 h and it was somewhat prolonged in the presence of IL-1␤ (t1 ⁄2 Ͼ 4 h). VEGF mRNA was degraded considerably faster with a half-life of approximately 1 h, whereas the IL-1␤-stabilized mRNA had a half-life of approximately 2 h, which is consistent with the results obtained in rat aortic smooth muscle cells (47). The level of VEGF-B mRNA was not reduced during the 4 h experiment. These data suggest that the mechanism of the IL-1␤-induced VEGF-C mRNA expression operates mostly at the transcriptional level.
IL-1␤ and TNF-␣ Stimulate the Production of the VEGF-C Protein-IL-1␤ and TNF-␣ stimulated the production of metabolically labeled VEGF-C protein (6-and 9-fold, respectively), as detected by immunoprecipitation of the metabolically labeled proteins secreted to the culture media with the use of VEGF-C-specific polyclonal antibodies (Fig. 7). The major protein precipitated from the culture medium corresponds to the partially processed VEGF-C isoform, which is biologically active (34).

Effect of IL-1␤ on Expression of VEGFs and VEGFRs in Vascular Endothelial
Cells-Since very little is known about VEGF, VEGF-B, or VEGF-C expression in vascular endothelial cells, we investigated their mRNA levels in HUVECs. Both VEGF-B and VEGF-C, but not VEGF, were detected in HUVECs, and the VEGF-C mRNA level was increased by treatment of the cells with either IL-1␤ or TNF-␣ (Fig. 8A). Furthermore, IL-1␤ stimulated VEGFR-2 mRNA (Fig. 8B). In agreement with the earlier reports (48,49), PMA stimulated VEGFR-1 and VEGFR-2 mRNA levels, whereas TNF-␣ decreased VEGFR-1 mRNA. Tek/Tie-2 and the related endothelial-specific tyrosine kinase receptor Tie-1 (50) were not affected by treatments with the cytokines or PMA (Fig. 8B). DISCUSSION We have here studied the regulation of the expression of three members of the VEGF family in response to proinflammatory cytokines. All VEGFs studied were expressed in human lung fibroblasts. In serum-starved cells, the levels of VEGF-B and VEGF-C mRNAs were similar, but higher than those of VEGF. IL-1␣, IL-1␤, and TNF-␣ induced VEGF and VEGF-C For the nuclear run-on the nuclei were isolated and labeled nuclear RNA samples were hybridized to immobilized VEGF-C and GAPDH cDNAs. The signals were quantitated using the Bio-Imaging Analyser system. Values represent fold induction of the controls after subtraction of the background and normalization for the GAPDH signals. B, effect of IL-1␤ on degradation of VEGF, VEGF-B, and VEGF-C mRNAs. Serum-starved IMR-90 cells were first preincubated with 10% FCS for 4 h and then in 0.5% FCS in the presence of actinomycin D (10 g/ml) with or without IL-1␤ (10 ng/ml) for 1, 2, or 4 h. Total RNA was analyzed by Northern blotting, and the mRNA hybridization signals were quantified by densitometric scanning. Values in the graph indicate the percentage of VEGF and VEGF-C mRNA signals remaining under specified conditions after normalization for the GAPDH signals.

FIG. 7. Effect of IL-1␤ and TNF-␣ on production of the VEGF-C protein.
Metabolically labeled IMR-90 cells were incubated with or without IL-1␤ (10 ng/ml) or TNF-␣ (50 ng/ml) for 6 h, after which the proteins were immunoprecipitated with VEGF-C-specific polyclonal antibodies. Immunoprecipitated proteins were then separated in SDS-PAGE and visualized by the Bio-Imaging analyzer system. mRNAs, but VEGF-B expression was not affected by these cytokines. In contrast, IL-1␤ decreased Ang-1 mRNA levels. This is consistent with the hypothesis that the Tek/Tie-2 signaling pathway may be important in the destabilization of blood vessels during angiogenesis and play a role later during the stabilization of the newly formed vessels (5)(6)(7)(8). We also found that IL-1␤ and TNF-␣ stimulate the production of a partially processed form of the VEGF-C protein, which is an active ligand of the lymphatic endothelial cell receptor VEGFR-3 (34). Since the mature form of VEGF-C protein is released from its precursor as a result of proteolytic cleavages, production or activation of such proteases could be also modulated at sites of cytokine activation in vivo. It remains to be studied whether cytokines regulate these post-translational processes. VEGF expression has previously been shown to be stimulated by IL-1␣ in human synovial fibroblasts (51), by IL-1␤ in rat aortic smooth muscle cells (47), and by IL-6 in tumor cell lines (52). In addition to our fibroblast model, TNF-␣ was found to stimulate VEGF expression in keratinocytes (53) and in tumor cell lines (52), but not in vascular smooth muscle cells (47). Hypoxia induces VEGF expression by transcriptional activation via hypoxia-inducible factor-1 and by post-transcriptional stabilization of the mRNA (54 -59). The mechanism of action of IL-1␤ was also suggested to depend both on transcriptional activation and on post-transcriptional mRNA stabilization (47). In the case of VEGF, the rapid mRNA decay depends on AU-rich protein binding instability elements found in the 3Ј-untranslated region (55,58), which are not present in the VEGF-C mRNA (60). This is consistent with our data, which show that VEGF-C mRNA has a relatively long half-life and that only a small stabilizing effect can be obtained with IL-1␤. All this suggests that mRNA stabilization is not as important for VEGF-C regulation as it is for VEGF. Indeed, our nuclear run-on data suggest that IL-1␤ induces expression of VEGF-C at the transcriptional level. Li et al. (47) hypothesised that IL-1␤ may facilitate transcriptional activation of the VEGF gene via AP-1 sites found in its 5Ј-flanking region (61). Data published by Chilov et al. (60) indicate that the 5Ј-flanking region of the VEGF-C gene contains a putative binding site for NF-B, which is an important transcription factor for signal transduction of proinflammatory cytokines (62). Interestingly, NF-B sites have also been identified in the mouse, but not the human, VEGF gene (59,61).
Glucocorticoids are important anti-inflammatory agents that also inhibit angiogenesis and reduce brain tumor-associated vascular permeability (63,64). Heiss et al. (65) recently suggested that the inhibition of vascular permeability by dexamethasone is delivered via inhibition of VEGF production. Further, serum-and PDGF-induced, but not hypoxia-induced VEGF expression was inhibited by dexamethasone in glioma cell lines (65). We found that IL-1␤-induced expression of VEGF and VEGF-C was inhibited by dexamethasone in IMR-90 fibroblasts. Whether the effect of dexamethasone is due to direct inhibition of VEGF-C transcription or due to interference with the IL-1␤ signaling pathway remains to be studied. Proinflammatory cytokines stimulate the production of prostanoids and dexamethasone inhibits it via the inducible cyclooxygenase-2 pathway (41,66). Since prostanoids are considered to be indirect angiogenic agents acting by inducing synthesis of VEGF (51, 67), we investigated, whether the induction of VEGF-C by cytokines is prostanoid-dependent. Indomethacin, a cyclooxygenase-inhibiting nonsteroidal anti-inflammatory drug, did not inhibit the IL-1␤-induced VEGF-C mRNA expression confirming the mechanism to be independent of prostanoids.
Namiki et al. (68) recently reported that HUVECs and human dermal microvascular endothelial cells express VEGF mRNA, but only after induction with PMA or hypoxia. In HUVECs we detected VEGF-B and VEGF-C mRNAs but no VEGF mRNA. Furthermore, VEGF-C mRNA level was increased by treatment with IL-1␤ and TNF-␣. Since neither VEGF nor VEGF-C expression has been detected in endothelial cells in vivo, the significance of these findings remains to be determined. In fact, VEGF-C was detected in mesenchymal cells around the vessels expressing VEGFR-3, suggesting a paracrine mode of action (27). It is, however, tempting to speculate that, in defined conditions, such as during cytokine activation, VEGF-C could modulate endothelial cell responses in an autocrine manner.
VEGFRs are specifically expressed on endothelial cells in vivo (14). Their expression levels are high during embryogenesis, low in most adult tissues, and again elevated in many tumors. However, only a few examples of VEGFRs mRNA regulation exist. TNF-␣ was recently shown to down-regulate VEGFR-1 and VEGFR-2 mRNAs in HUVECs and in human aortic endothelial cells (48), whereas PMA induced all three VEGFR mRNAs in HUVECs (49). We found that IL-1␤ elevated the level of VEGFR-2 mRNA in HUVECs. These data support the hypothesis that cytokines could also induce responsiveness of endothelial cells to VEGF and VEGF-C by increasing the expression of VEGFR-2.
We have here shown that proinflammatory cytokines upregulate the expression of VEGF-C. Our data suggest that, in addition to VEGF, VEGF-C may serve as an endothelial stimulus at sites of cytokine activation. Since VEGF-C primarily targets endothelium of lymphatic vessels, these results raise the possibility that certain proinflammatory cytokines affect the lymphatic vessels indirectly via VEGF-C. This may be important in controlling the composition and pressure of interstitial fluid and in facilitating lymphocyte trafficking and thus have an important role in inflammatory processes.
Acknowledgments-We thank Dr. Ulf Eriksson for expert help with reagents, and Kaija Antila, Ritva Javanainen, and Tapio Tainola for excellent technical assistance.
FIG. 8. Effects of IL-1␤ and TNF-␣ in human vascular endothelial cells. HUVECs were incubated with or without IL-1␤ (10 ng/ml), TNF-␣ (50 ng/ml) or PMA (10 ng/ml) for 6 h. A, blots were hybridized with a mixture of the VEGF probes or the control probes, or B, alternatively with probes for VEGFRs or Tie-1 or Tek/Tie-2. Results are shown as in Fig. 1.