Norepinephrine Induces Vascular Endothelial Growth Factor Gene Expression in Brown Adipocytes through a b -Adrenoreceptor/cAMP/Protein Kinase A Pathway Involving Src but Independently of Erk1/2*

To identify the signaling pathway that mediates the adrenergic stimulation of the expression of the gene for vascular endothelial growth factor (VEGF) during physiologically induced angiogenesis, we examined mouse brown adipocytes in primary culture. The endogenous adrenergic neurotransmitter norepinephrine (NE) induced VEGF expression 3-fold, in a dose- and time-de-pendent manner (EC 50 ’ 90 n M ). Also, the hypoxia-mim- icking agent cobalt, as well as serum and phorbol ester, induced VEGF expression, but the effect of NE was additive to each of these factors, implying that a separate signaling mechanism for the NE-mediated induction was activated. The NE effect was abolished by propranolol and mimicked by isoprenaline or BRL-37344 and was thus mediated via b -adrenoreceptors. The NE-induced VEGF expression was fully cAMP mediated, an effect which was inhibited by H-89 and thus was dependent on protein kinase A activity. Involvement of other adrenergic signaling pathways ( a 1 -adrenoreceptors, Ca 2 1 , protein kinase C, a 2 -adrenoreceptors, and pertussis toxin- sensitive G i -proteins) was excluded. The specific inhibitor of Src tyrosine kinases, PP2, markedly reduced the stimulation by NE, which demonstrates that a cAMP-dependent Src-mediated pathway is positively connected to VEGF expression. However, inhibition of Erk1/2 MAP an Ultraspec solution (Biotecx), and the manufacturer’s procedure for RNA isolation was followed. The RNA concentration was measured on a Beckman DU 50 spectrophotometer with readings at 260 and 280 nm; the 260/280 nm ratio was routinely 1.7–1.8. 10 m g of total RNA was separated by electrophoresis in an ethidium bromide-containing agarose-formaldehyde gel, as described by Bronnikov et al. (20). The intensity of the 18 S and 28 S rRNA bands under UV light was routinely checked to verify that all samples were equally loaded and that no RNA degradation had occurred. blotted from the gel to a Hybond-N mem- brane, cross-linked, and hybridized as described by Bronnikov et al. (20). Hybond-N then washed 0.1 to a PhosphorImager 1 day. The screen was a Molecular and VEGF was quantified with the ImageQuant pro- When the same Hybond-N

For the study of angiogenesis, brown adipose tissue provides a physiologically interesting model system. In this tissue, norepinephrine (NE), 1 the adrenergic mediator in sympathetic nerves, is the acute inducer of thermogenesis (1). During the thermogenic process, the rate of lipid oxidation in this tissue, and thus the demand for oxygen delivery to the tissue, is immense. Heat production in brown adipose tissue accounts for nearly half of the total energy expenditure in a cold-acclimated animal in the cold, even though the tissue only constitutes 1-2% of the total body weight (2,3). The blood supply needed to support the oxygen requirement under such conditions is necessarily exceedingly high (2,3), demanding a rich vasculature, which is indeed increased during brown adipose tissue recruitment (2,4).
The regulation of this angiogenesis in brown adipose tissue has so far not been extensively investigated. Expression of the angiogenic factor basic fibroblast growth factor, is induced in the tissue by cold exposure (5) and by NE in cultured brown adipocytes (6). The VEGF-related factor (vascular endothelial growth factor-B) is expressed already during development (7), and expression of VEGF (vascular endothelial growth factor (8)) itself is induced by NE administration and cold exposure in brown adipose tissue (9), and VEGF expression and secretion is stimulated by NE in cultured brown adipocytes (10,11). Thus, NE appears to mediate cold-induced angiogenesis in brown adipose tissue by stimulating the expression of angiogenic factors, such as VEGF. In this respect, the effect of NE may be connected to the general recruitment effect of NE in this tissue, i.e. increased DNA synthesis (6,(12)(13)(14) and increased expression of the brown adipocyte-specific mitochondrial uncoupling protein UCP1 (15)(16)(17)(18).
The aim of the present investigation was to clarify through which intracellular signaling mechanisms NE stimulates expression of the VEGF gene. We conclude that NE-induced VEGF expression was exclusively dependent on a ␤-adrenoreceptor-mediated increase in cAMP levels and PKA activity. The mediation involved Src tyrosine kinases, unexpectedly and completely independently of downstream mediation via Erk1/2 MAP kinases. We thus demonstrate a positive connection between VEGF expression and a signaling pathway that branches off from the NE-activated Src-Erk1/2 cascade.

EXPERIMENTAL PROCEDURES
Cell Isolation-Brown adipocyte precursors were isolated from 3-4week-old mice of the NMRI strain, obtained from a local supplier (Eklunds), principally as described by Néchad et al. (19) and modified by Bronnikov et al. (14). Briefly, the cervical, interscapular, and axillary brown adipose tissue depots were dissected out from each mouse. The depots were pooled and incubated in the Hepes-buffered Ringer solution detailed by Néchad et al. (19), containing 0.2% (w/v) crude collagenase type II (Sigma). The tissue was digested for 30 min at 37°C, with vortexing every 5 min, after which the digest was filtered through a 250-m nylon screen. To allow lipid droplets and mature adipocytes to float, the digest was put on ice for 20 min, and the infranatant was then collected and filtered through a 25-m nylon screen. The cells in the * This work was supported by a grant from the Swedish Cancer Foundation and by the Swedish Natural Science Research Council. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. filtrate were pelleted by centrifugation, suspended in Dulbecco's modified Eagle's medium and repelleted, after which they were suspended in culture medium and used for cell culture.
Cell Culture-The cells were cultured either in Costar 6-well plates (growth area 9.4 cm 2 per well, with cells obtained from 2.5 animals inoculated per plate in 2 ml of culture medium per well) or in Costar 12-well plates (growth area 3.83 cm 2 per well, with cells obtained from 2 animals inoculated per plate in 1 ml of culture medium per well). The culture medium consisted of Dulbecco's modified Eagle's medium (routinely Life Technologies, Inc., without L-glutamine with 4 nM L-glutamine (Life Technologies, Inc.) added), supplemented with 10% newborn calf serum (PAN Systems), 4 nM insulin (Actrapid Human, Novo), 10 mM Hepes (Life Technologies, Inc.), and with 50 IU penicillin, 50 g of streptomycin (Life Technologies, Inc.), and 25 g of sodium ascorbate (Kebo) per ml, at 37°C in a water-saturated atmosphere of 8% CO 2 in air in a Heraeus CO 2 auto-zero B5061 incubator. On day 1, the cultures were washed with prewarmed Dulbecco's modified Eagle's medium, and fresh prewarmed medium was then added. Medium was then changed every other day until the day of experimentation, which was routinely day 6.
Cultures under serum-free conditions were cultured for 5-6 days in the above medium, after which they were washed and the same culture medium as described above, without newborn calf serum, but with 0.5% (w/v) albumin (fraction V, fatty acid free; Roche Molecular Biochemicals) added. The experiments were performed after approximately 20 h in this medium.
RNA Isolation and Determination of mRNA Levels for VEGF and UCP1-On the indicated day of culture (day 6 if not otherwise indicated) and after stimulation, the culture medium was discarded and the cells were dissolved in 0.8 ml of an Ultraspec solution (Biotecx), and the manufacturer's procedure for RNA isolation was followed. The RNA concentration was measured on a Beckman DU 50 spectrophotometer with readings at 260 and 280 nm; the 260/280 nm ratio was routinely 1.7-1.8. 10 g of total RNA was separated by electrophoresis in an ethidium bromide-containing agarose-formaldehyde gel, as described by Bronnikov et al. (20). The intensity of the 18 S and 28 S rRNA bands under UV light was routinely checked to verify that all samples were equally loaded and that no RNA degradation had occurred.
The RNA was blotted overnight from the gel to a Hybond-N membrane, cross-linked, and hybridized as described by Bronnikov et al. (20). The Hybond-N membrane was then washed three times in 0.1 ϫ SSC, 0.2% SDS at 50°C for 20 min. The membrane was sealed in a plastic envelope and exposed to a PhosphorImager screen for 1 day. The screen was analyzed on a Molecular Dynamics PhosphorImager and the overall VEGF mRNA signal was quantified with the ImageQuant program. When the same Hybond-N membrane was analyzed for both VEGF and UCP1 mRNA, the previous probe was removed by putting the membrane in 0.2% SDS solution at 100°C and allowing this solution to cool down to room temperature.
cDNA Probes-The VEGF cDNA was a gift from Dr. Georg Breier, Max-Planck Institute, Bad Nauheim, Germany. It contained the 580base pair mouse VEGF 164 cDNA, which encodes the VEGF 164 isoform (21). This cDNA had been cloned into the EcoRI-BamHI cloning sites of a pBluescript KS vector. The vector was amplified by transformation of a TG-1 Escherichia coli strain and the cDNA was isolated. The UCP1 cDNA was that earlier used (18). The probes were labeled with [␣-32 P]dCTP with a Random Primed DNA Labeling Kit (Roche Molecular Biochemicals), according to the manufacturer's instructions.
cAMP Analysis-On day 6 in culture and after stimulation, the culture medium was aspirated after the indicated time and the cells were treated principally as described by Bronnikov et al. (20). Briefly, 0.8 ml of 75% ethanol with 1 mM EDTA was added to each well, and after 10 min, cells were harvested by scraping. Ethanol was removed by SpeedVac centrifugation and pellets were suspended in 0.5 ml of 4 ϫ TE buffer and sonicated 5 s. After centrifugation, a 25-l supernatant aliquot from each sample was used for determination of cAMP levels with the Cyclic AMP Assay System (Amersham Pharmacia Biotech), according to the manufacturer's instructions.
Erk1/2 Phosphorylation Analysis-On day 6 in culture and after stimulation, the culture medium was discarded and the cells were harvested and analyzed for Erk1/2 phosphorylation as described earlier (22). Briefly, proteins were separated on 12% polyacrylamide gels. Electrotransfer to Hybond-C Extra nitrocellulose membranes (pore size 0.45 m; Amersham Pharmacia Biotech) was carried out in a semidry electroblotter. After transfer, the membranes were allowed to soak in Trisbuffered saline for 5 min followed by quenching (5% non-fat dry milk, 0.1% Tween 20 in Tris-buffered saline) of nonspecific binding for 1 h at room temperature. The membranes were incubated with primary anti-bodies, an Erk1/2 phosphospecific antibody (human Thr(P) 202/Tyr(P) 204; New England Biolabs) or an Erk1/2 antibody (New England Biolabs) overnight at 4°C. The primary antibody was detected with a secondary horseradish peroxidase-conjugated goat anti-rabbit antibody (New England Biolabs) and enhanced chemiluminescence (ECL; Amersham Pharmacia Biotech). The blots were exposed to Kodak X-Omat RP films and quantified on a Molecular Dynamics densitometer. Erk1/2 phosphorylation was routinely expressed as the ratio of phosphorylated Erk1/2 to total Erk1/2.

Mouse Brown Adipocytes Constitutively Express the VEGF Gene
To examine the regulation of VEGF expression in mouse brown adipocytes, we first investigated the ability of these cells to express the VEGF gene constitutively. Total RNA isolated from differentiated mouse brown adipocytes (on day 6 in culture) was analyzed by Northern blot analysis. The 580-base pair mouse VEGF 164 cDNA probe (21) was used. On the Northern blot (Fig. 1A), a major band at approximately 4 kilobases was observed (top arrow), and 4 further bands at lower molecular weight were distinguishable on densitometric analysis. This Northern blot pattern was similar to that observed in other cells, tissues, and species (9,21,23,24), including rat brown adipose tissue (9). During the different treatments used in this study, the intensity of these bands varied in parallel, and the total signal was therefore that used for analysis.
To investigate whether the expression of VEGF is a differentiation characteristic, we followed VEGF expression during cell differentiation. Brown adipocytes in primary culture proliferate until confluence is reached, which occurs at approximately day 5-6 after inoculation (14,19,20,25,26). At this time, differentiation spontaneously occurs. This is verified in Fig. 1B where the ability of NE to induce expression of the brown adipocyte-specific differentiation marker, mitochondrial uncoupling protein-1 (UCP1), is analyzed. As seen, rather high levels of UCP1 mRNA were induced by NE from day 5 (Fig. 1B, black bars), indicating that brown adipocyte differentiation had occurred. As seen in Fig. 1C (gray bars), VEGF was constitutively expressed already in undifferentiated cells, and basal VEGF mRNA levels were similar in all stages of differentiation. This is in contrast to VEGF expression observed in 3T3-F442A adipocytes, which was fully differentiation-dependent (27). Thus, the state of differentiation is not crucial for basal VEGF expression in brown adipocytes.

Norepinephrine Induces VEGF Gene Expression
To investigate the effect of NE on VEGF expression, we treated the cultures with NE for 1 h. VEGF mRNA levels increased approximately 3-fold after NE stimulation (Fig. 1A, two rightmost lanes), which is in accordance with observations in cultured rat brown adipocytes (11). The stimulation by NE was observed at all stages of brown adipocyte differentiation, from proliferating to fully differentiated cells, and was of similar relative magnitude (Fig. 1C, black bars).
When brown adipocytes were stimulated with increasing concentrations of NE, VEGF mRNA levels increased monopha-sically, following simple Michaelis-Menten kinetics, to 4-fold control levels, obtained with 10 M NE; the EC 50 was 70 Ϯ 22 nM ( Fig. 2A). Analysis of mean points from four similar experiments yielded a 3-fold increase, with similar kinetics; EC 50 93 Ϯ 27 nM (not shown).
In the maintained presence of NE, the VEGF mRNA levels were elevated only during the first few hours and had returned to control levels after 6 h (Fig. 2B). An independent experiment with 30 min NE stimulation of brown adipocytes resulted in only a 30% elevation of VEGF mRNA levels (not shown); thus maximum levels occurred at about 1 h of stimulation. The rapid disappearance of the mRNA signal in these experiments im-plies a short half-life of VEGF mRNA (see also below). This transient increase in expression is in principal agreement with the time course of VEGF expression induced in rat brown adipose tissue in situ by cold exposure: the expression peaked at 4 h, and had decreased down to control levels within 24 h (9). Based on these experiments, further experiments were performed with 10 M NE and analyzed after 1 h of stimulation.

Interaction between Norepinephrine and Other Inducers of VEGF Gene Expression
As demonstrated above, NE is a potent inducer of VEGF expression in mouse brown adipocytes. In various other cell types, other factors such as hypoxia, cobalt (as a hypoxiamimicking agent), serum, growth factors, and phorbol esters have been demonstrated to induce VEGF expression (27)(28)(29)(30)(31). We have therefore investigated whether NE is an exclusive inducer of VEGF expression in brown adipocytes, or whether the classical factors mentioned above could induce expression also in these cells. These experiments may also reveal initial information on the intracellular pathway mediating the NE signal, as non-additivity between the NE-induced VEGF expression and that induced by other factors would imply distinct signaling pathways.
Hypoxia (Cobalt)-The hypoxia-mimicking agent cobalt was not able to elevate VEGF expression after 1 h of treatment (not shown), but after 6 h, a 3-fold increase could be seen (Fig. 3A). The elevated VEGF mRNA levels persisted for at least 24 h (not shown). Thus, hypoxia is probably an inducer of VEGF expression also in brown adipocytes. The effect of NE was additive to that of cobalt (Fig. 3A). Thus, NE and hypoxia (cobalt treatment) utilize separate pathways for the induction of VEGF expression.
Serum-To investigate the effect of serum factors, serum was removed from the brown adipocyte cultures. When readded the next day, serum induced VEGF expression 4-fold (Fig. 3B). The effect of serum lasted longer than that of NE (Ͼ6 h; not shown). NE was able to induce VEGF expression, also in the absence of serum factors (Fig. 3B). The effect of NE was additive to that of serum (Fig. 3B), and thus the signaling pathway mediating the NE effect would also seem to be different from pathways utilized by serum factors.
Phorbol Esters-The phorbol ester TPA was also a potent inducer of VEGF expression (4-fold) (Fig. 3C), implying that protein kinase C (PKC)-activated pathways can stimulate VEGF expression in brown adipocytes. NE gave a marked increase of VEGF mRNA in addition to that induced by TPA alone (Fig. 3C). Thus, the pathway mediating the NE effect on VEGF expression appears not to involve PKC activation (see also below). Thus, also the classical inducers of VEGF expression, hypoxia (cobalt), serum, and phorbol ester, induced VEGF expression in brown adipocytes, but NE utilized a distinct signaling pathway.

Mediation of Norepinephrine-induced VEGF Gene Expression
To investigate which adrenergic receptors mediate the NE effect on VEGF expression in brown adipocytes, adrenoreceptor agonists and antagonists were used (Fig. 4A). The ␣ 1 -adrenoreceptor antagonist prazosin had no significant effect on NEinduced VEGF expression, but the ␤-adrenoreceptor antagonist propranolol completely abolished it. In accordance with this, the ␣ 1 -adrenoreceptor agonist cirazoline had no inducing effect (Fig. 4A), nor had the ␣ 2 -adrenoreceptor agonists clonidine (10 M) or UK14304 (10 M) (not shown), whereas the ␤-adrenoreceptor agonist isoprenaline was an equally potent inducer as was NE (Fig. 4C). Thus, only ␤-adrenoreceptors mediated the NE-induced VEGF expression.
Proliferating brown preadipocytes only express the ␤ 1 -adrenoreceptor subtype whereas fully differentiated brown adipocytes also express the ␤ 3 -adrenoreceptor subtype which, at this stage of differentiation, is the only ␤-adrenoreceptor subtype coupled to adenylyl cyclase (20). Thus, the NE-induced VEGF expression observed in proliferating undifferentiated brown preadipocytes (days 3-4, Fig. 1B) is most likely mediated by ␤ 1 -adrenoreceptors. In differentiated brown adipocytes, the ␤ 3adrenoreceptor agonist BRL-37344 induced VEGF expression to the same extent as did NE (not shown). The ␤ 3 -adrenoreceptor agonist CGP-12177 (32) (which is a ␤ 1 -and ␤ 2 -adrenoreceptor antagonist) also induced VEGF expression in these cells (not shown). Thus, both ␤ 1 -and ␤ 3 -adrenoreceptors were able to mediate NE-induced VEGF expression in these cells, the coupling depending on the differentiation stage. To identify which second messengers had the ability to induce VEGF expression, we examined the intracellular second messenger systems involved in adrenoreceptor-mediated signaling: Ca 2ϩ and PKC (from ␣ 1 -adrenoreceptors), the proposed ␤ 3 -adrenorecep-
The Ca 2ϩ ionophore A23187, which increases intracellular Ca 2ϩ levels and through this, e.g. induces c-fos expression (34), had no effect on VEGF mRNA levels (Fig. 4A). Concerning PKC activation, it was shown above that the phorbol ester TPA induced a marked elevation of VEGF expression (Fig. 3C), but it was concluded that the NE effect appeared to be independent of PKC, because of the additivity of the effects. To substantiate this, PKC was inactivated prior to NE stimulation, by prolonged exposure to TPA, a treatment that inactivates TPAsensitive PKC isoforms (35,36). After TPA pretreatment, no remaining elevation of VEGF mRNA levels was observed (Fig.  4B), and the acute effect of TPA on VEGF mRNA levels was abolished (Fig. 4B). Thus, TPA pretreatment effectively inactivated PKC. However, the NE-induced VEGF expression was not altered by TPA pretreatment (Fig. 4B) and was thus not mediated through a TPA-sensitive PKC pathway.
To investigate the possible involvement of the suggested cAMP-independent ␤ 3 -adrenoreceptor signaling through pertussis toxin (PTX)-sensitive G-proteins, i.e. G i -proteins (33,37,38), we pretreated brown adipocytes with PTX. PTX augmented ␤ 3 -adrenoreceptor-stimulated cAMP production, 2 confirming that G i proteins were active. However, NE-or isoprenaline-induced VEGF expression was not affected by PTX, and PTX had no effect in itself (Fig. 4C). Thus, no adrenergic stimulation of VEGF expression was mediated through PTXsensitive G i proteins in a cAMP-independent process.
To investigate the ability of cAMP to induce VEGF expression, brown adipocytes were treated with forskolin to activate adenylyl cyclases. Forskolin induced VEGF expression to the same extent as did NE (Fig. 4A). This effect was not altered by increased levels of intracellular Ca 2ϩ (Fig. 4A). The forskolin analogue 1,9-dideoxy-forskolin, which does not activate adenylyl cyclases, had no effect (not shown). The cAMP analogue 8-bromo-cAMP (1 mM for 1 h) also induced VEGF expression 3-fold (not shown). Thus, increased levels of cAMP are potent stimulators of VEGF expression in brown adipocytes and may mediate the NE effect.

cAMP Is the Mediator of the Norepinephrine-induced VEGF Expression
To establish that cAMP not only is able to induce VEGF expression, but also is the actual mediator of the NE effect, we first investigated the correlation between cAMP levels and VEGF expression in brown adipocytes treated with different doses of forskolin. The elevation of cAMP levels was dose-dependent (Fig. 5A), as was the elevation of VEGF mRNA levels (Fig. 5B). However, nearly maximal induction of VEGF expression occurred already at approximately 50 pmol of cAMP/well (Fig. 5C). Thus, the ability of cAMP to induce VEGF expression was saturated at this cAMP level.
To compare the ability of NE-and forskolin-derived cAMP to induce VEGF expression, we measured the cAMP levels induced by NE (Fig. 5D). The NE-induced cAMP levels reached the saturating levels for VEGF expression (approximately 50 pmol of cAMP/well), and the dose-response kinetics were very similar to those for NE-induced VEGF expression ( Fig. 2A). Indeed, when NE-derived results were plotted into the forskolin curve (Fig. 5E), they did not deviate from the forskolin results. These results suggest that NE-induced VEGF expression is exclusively dependent on cAMP. To confirm this, we treated brown adipocytes with NE at forskolin concentrations sufficient to saturate cAMP-induced VEGF expression. Under these conditions, NE did not affect forskolin-induced cAMP levels (not shown), nor did it affect forskolin-induced VEGF expression (Fig. 5F). This lack of additivity implies that cAMP is the common mediator of NE-and forskolin-induced VEGF expression (contrast the additivities in Fig. 3).

The Signaling Pathway Downstream of cAMP Is Mediated via PKA and Src but Not via Erk1/2
To identify the further signaling pathway involved in mediating the cAMP-dependent stimulation of VEGF expression, we used inhibitors of intracellular signaling factors.
The classical immediate downstream effector of cAMP, the cAMP-dependent protein kinase (PKA), could be expected to mediate the cAMP-dependent signaling. Therefore, we pretreated brown adipocytes with the PKA inhibitor H-89, which ␤-Adrenergic Regulation of VEGF Expression through Src had no effect in itself, but which abolished both NE-and forskolin-induced VEGF expression (Fig. 6). That also the forskolin effect was inhibited demonstrated that H-89 acted by inhibiting PKA, and not by competing with NE for ␤ 3 -adrenoreceptor binding (such an effect has been demonstrated for ␤ 1 -and ␤ 2 -adrenoreceptors (39), but does not seem to apply to ␤ 3adrenoreceptors). 3 Thus, the cAMP-dependent stimulation was mediated through a PKA-dependent signaling pathway.
In brown adipocytes, adrenergic activation may proceed via Src tyrosine kinases, which are activated via ␤ 3 -adrenorecep-tors in a PKA-dependent manner in these cells. 4 Therefore, to investigate whether Src mediates NE-induced VEGF expression, we pretreated brown adipocytes with the specific inhibitor of Src, PP2. PP2 was fully effective as a Src inhibitor, since NE-induced Erk1/2 phosphorylation was abolished (Fig. 7A). To ensure that PP2 had no unspecific effect on adrenergic pathways and gene expression, we first examined whether PP2 inhibited NE-induced UCP1 expression. Clearly PP2 did not have a general inhibitory effect; rather, surprisingly, PP2 markedly potentiated the effect of NE (Fig. 7B), an effect for which we have no explanation. Concerning the NE-induced VEGF expression, PP2 had a marked inhibitory effect (Fig. 7C). Thus, Src was involved in the mediation of the response, although a Src-independent pathway also existed.
The MAP kinases Erk1/2 could be expected to mediate the Src-dependent stimulation of VEGF expression. This cascade is activated by NE in brown adipocytes in a cAMP-dependent manner (22,40). In other cell types, VEGF expression (induced by other factors) has been implied to be induced via this signaling cascade (24,30,(41)(42)(43). Therefore, we pretreated brown adipocytes with the MAP kinase/Erk1/2 kinase (MEK1) inhibitor PD98059. In agreement with earlier observations (22), PD98059 totally abolished NE-induced Erk1/2 phosphorylation (Fig. 8A). However, PD98059 did not alter the NE-induced VEGF expression (Fig. 8B). Thus, despite the involvement of Src demonstrated above, Erk1/2 did not mediate NE-induced VEGF expression in brown adipocytes. In an attempt to characterize factors acting downstream of Src tyrosine kinases on the stimulation of VEGF expression, we used the general tyrosine kinase inhibitor, genistein (100 M, 1 h pretreatment), but it decreased NE-induced VEGF expression to a lesser extent than did PP2 (not shown).
Phosphatidylinositol 3-kinase is a substrate for Src (44) and is a factor implicated in the activation of VEGF expression (45). However, wortmannin (100 nM, 1 h pretreatment), an inhibitor of phosphatidylinositol 3-kinase, had no effect on NE-induced VEGF expression (not shown). Thus, phosphatidylinositol 3kinase-dependent signaling pathways did not appear to be involved.

Norepinephrine Does Not Require Induced Protein Synthesis to Induce VEGF Gene Expression
To investigate whether synthesis of new transcription factors, or other protein factors, were required for NE-induced

␤-Adrenergic Regulation of VEGF Expression through Src
VEGF expression, we examined the effect of the translational inhibitor cycloheximide. Cycloheximide in itself led to an induction of VEGF mRNA (Fig. 9). A similar effect has been described in other systems (46,47). These experiments imply either that a short-lived protein factor may constitutively inhibit VEGF transcription or, perhaps more likely, that such a factor may promote VEGF mRNA degradation (in other systems, cycloheximide has been demonstrated to stabilize VEGF mRNA (48)).
The effect of NE was additive to that of cycloheximide ( Fig. 9) and thus not dependent on protein synthesis. Thus, although NE may induce the gene expression of certain transcription factors (such as hypoxia-inducible factor-1␣), 5 such induction cannot be involved in the mediation of the NE effect. The lack of need for protein synthesis is also in accordance with observations in other systems, with other inducers of VEGF expression (46,49). Thus, all factors required for regulation of VEGF expression appeared to be constitutively produced and present in brown adipocytes.

The Effect of Norepinephrine on VEGF Gene Expression Is Transcriptional
The augmenting effect of NE on VEGF mRNA levels could be due to an increase in the rate of transcription of the VEGF gene or to an increase of VEGF mRNA half-life. Indeed, in several cell types, the half-life of VEGF mRNA is positively affected by other agents, such as hypoxia (50,51), PKC activation (48), and nitric oxide (52). To investigate whether NE functions by prolonging VEGF mRNA half-life, we treated brown adipocytes with the transcriptional inhibitor actinomycin D. The VEGF mRNA half-life, as measured after actinomycin D treatment, was approximately 1.5 h (Fig. 10) (compare also the rate of spontaneous decrease seen in Fig. 2B), which is similar to earlier observations (51,53). NE did not affect the half-life of VEGF mRNA (Fig. 10). Thus, the effect of NE was probably due to a direct increase in the rate of transcription. DISCUSSION In the present investigation, we have demonstrated that the adrenergic agent NE induces the expression of the gene for the potent angiogenic factor VEGF in primary cultures of mouse brown adipocytes. The stimulatory effect by NE occurred independently of that induced by other factors known to induce VEGF expression in other cell systems. Signaling was via a ␤-adrenoreceptor-induced increase in cAMP levels, further mediated by PKA. Downstream of cAMP/PKA, the signal was mediated by Src tyrosine kinases and also through a Src-independent pathway. However, the Src pathway did not involve Erk1/2 MAP kinases (Fig. 11).
Induced synthesis of protein factors was not required for mediation of the adrenergic stimulation. The mechanism of action was not prolongation of VEGF mRNA half-life, and was thus apparently an increase in the rate of transcription.
Norepinephrine Stimulation of VEGF Gene Expression-That adrenergic stimulation of cultured rat (11) and mouse brown adipocytes stimulated VEGF expression, indicates parenchymal localization of the NE-induced expression observed in intact animals by Asano et al. (9). Although NE is not generally recognized as an inducer of VEGF expression, the effect is not exclusive for brown adipocytes, since also in rat aortic smooth muscle cells, adrenergic stimulation induces VEGF expression (54).
Adrenergic stimulation of VEGF expression in brown adipose tissue is physiologically meaningful in that also the oxygen-demanding thermogenic processes are adrenergically stim-ulated. However, the transient induction of VEGF expression, demonstrated here in cultured brown adipocytes, where elevated VEGF mRNA levels were observed only for the first few hours during chronic stimulation, may be thought to imply that adrenergic stimulation of VEGF expression may not be of major significance in regulating physiological angiogenesis. VEGF VEGF expression was analyzed in brown adipocytes principally as described in the legend to Fig. 1A. The cells were treated with 5 g/ml actinomycin D (Act) (5 min before NE), with or without 10 M NE, for the indicated times; C, control levels. The values are means from one experiment with duplicate wells for each time and treatment, and mean start control value was set to 100%. Curves were drawn according to a simple exponential decay (R t ϭ R 0 ⅐ exp (Ϫk ⅐ t). For actinomycin-treated cells, Ϫk was Ϫ0. 43  expression induced by the hypoxia-mimicking agent cobalt was maintained for a longer period of time, and thus could be considered more physiologically relevant, as hypoxic conditions may be generated in brown adipose tissue in the cold-exposed animal, due to the extensive oxygen consumption by the adrenergic thermogenic processes (2,3). However, in brown adipose tissue in cold-exposed rats (9) and mice, 6 cold-induced VEGF expression is also transient and thus, in this respect, very similar to our observations in NE-stimulated cultured brown adipocytes. Thus, it is likely that NE is the physiological inducer of VEGF expression and thereby of cold-induced angiogenesis. Apparently, an initial short elevation of VEGF expression is sufficient to support angiogenesis during brown adipose tissue recruitment.
␤-Adrenoreceptors, cAMP, PKA, and Src but Not Erk1/2 Mediate the Norepinephrine-induced VEGF Gene Expression-The classical ␤-adrenoreceptor pathway, cAMP and PKA, was fully responsible for the mediation of the NE effect. In several cell types, forskolin and/or cAMP-analogues have earlier been demonstrated to induce VEGF expression (27, 46, 54 -62). In most of these cell types, as yet only an ability of cAMP to induce VEGF expression has been demonstrated, but it has been implied that the relevant physiological activator also in these cells induces VEGF expression via cAMP. Further mediation has, however, not been clarified.
It is in this context especially noteworthy that in brown adipocytes, Src tyrosine kinases are activated via a ␤ 3 -adrenoreceptor/cAMP/PKA pathway. 4 This is of relevance because Src earlier has been demonstrated to mediate hypoxia-induced VEGF expression (41) and has been suggested to be involved in tumor-induced angiogenesis by mediating enhanced VEGF expression (63,64). Indeed, in brown adipocytes we found that specific inhibition of Src tyrosine kinases, by PP2 treatment, markedly reduced VEGF expression in brown adipocytes stimulated by NE; in agreement with observations in hypoxiatreated cell lines (41), this inhibition was only partial. Thus, in brown adipocytes, Src tyrosine kinases appear to be involved in the mediation of NE-induced VEGF expression.
In glioblastoma U87 and fibrosarcoma HT1080 cells, Src-dependent VEGF expression is further mediated via Erk1/2 MAP kinases (41,43), and the Src-Erk1/2 signaling cascade appears to be important for the enhanced VEGF expression in other tumor cell types (42). However, the Erk1/2 MAP kinases were not involved in mediating the NE-induced VEGF expression in brown adipocytes. The unexpected lack of involvement of Erk1/2 demonstrates the existence of a cAMP-induced Srcmediated signaling pathway, positively connected to VEGF expression, which is not dependent on downstream mediation via Erk1/2.
Src has numerous substrates involved in various cellular processes (44), but the factors mediating the adrenergic stimulation of VEGF expression downstream of Src are presently unknown. As the effect of NE does not involve prolongation of VEGF mRNA half-life, an increase in the rate of transcription of the VEGF gene is to be expected. Furthermore, as the NE induction of VEGF expression is not dependent of protein synthesis, the transcription factor involved must be constitutively expressed. The Src-dependent NE-induced VEGF expression may involve the transcription factor Sp1, which can mediate Src-dependent gene expression (65) and PDGF-induced VEGF expression (66). Also NFB may mediate Src-dependent gene expression (44) and thus may be involved. For these transcription factors, putative binding sites (consensus sequences) exist in the VEGF promoter (67,68).
In conclusion, we have here demonstrated that adrenergic stimulation of VEGF expression in brown adipocytes is exclusively mediated by a ␤-adrenoreceptor/cAMP/PKA signaling pathway. This pathway can utilize a Src-dependent pathway that branches off from the Src-Erk1/2 signaling cascade to stimulate expression of the VEGF gene.