THE JOURNAL OF BIOLOGICAL CHEMISTRY 0 1992 by The Amerlcan Society for Biochemistry and Molecular Blology. Inc

Vol. 267, No. 23, Issue of August 15, pp. 16317-16322, 1992 Printed in 1J. S. A

Vascular Endothelial Growth Factor REGULATION BY CELL DIFFERENTIATION AND ACTIVATED SECOND MESSENGER PATHWAYS*

(Received for publication, October 8, 1991)

Kevin P. Claffeyz, William 0. Wilkison$, and Bruce M. Spiegelmann From the Dana-Farber Cancer Institute and DeDartment of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, Boston, Massachusetts 021 15

Vascular endothelial growth factor (VEGF) is an an- giogenic polypeptide that has been isolated from a va- riety of tumorigenic and nontransformed cell lines. Because of the importance of blood vessel growth to cell and tissue development, we have examined VEGF gene expression in a variety of mouse tissues and ro- dent models of cellular differentiation. Using a cloned murine VEGF cDNA we show that VEGF mRNA is expressed at relatively low levels in many adult mouse tissues examined. However, this message is dramati- cally induced in two models of cell differentiation: 3T3-adipose conversion and CzClz myogenic differen- tiation. VEGF protein secretion is also induced in adi- pocyte differentiation. VEGF mRNA is markedly reg- ulated in a pheochromocytoma (PC12) cell model of transformation and differentiation. The transformed undifferentiated cells express moderate levels of VEGF mRNA and this expression is virtually extinguished when cells differentiate into non-malignant neuron- like cells. Experiments employing phorbol esters and CAMP analogues indicate that VEGF mRNA expression is stimulated in preadipocytes by both protein kinase C and protein kinase A-mediated pathways. These re- sults suggest that VEGF mRNA levels are closely linked to the process of cellular differentiation; they also clearly demonstrate that expression of this angi- ogenic factor is specifically regulated in a transformed cell line, possibly via aberrant activation of cellular second messenger pathways.

Proper cell and tissue development requires growth of new blood vessels that is tightly controlled in both space and time. The mechanisms involved in the regulation of vascularization during tissue development are not well understood. Develop- mentally regulated angiogenesis can be observed in many organs but it is particularly well illustrated in adipose tissue (1). Adipose tissue is the major site of energy storage in vertebrate animals, requiring extensive vascularization for proper function. The development of the vascular bed in

Institutes of Health (DK 42420) and the U. S. Army (DAMD 17-91- *This work was supported in part by grants from the National

2-1010). 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 the GenBankTM/EMBL Data Bunk with accession numberfs) The nucleotide seguence(s) reported in this paper h.as been submitted

M95200. 2 Supported by National Research Service Award Fellowship

DK08278. I Supported by National Research Service Award Fellowship

DK08277. 1 Established Investigator of the American Heart Association.

adipose tissue is tightly connected to both the number and size of differentiating adipocytes (I).

Immortalized cell lines, such as 3T3-F442A and 3T3-L1, can differentiate from fibroblast-like preadipocytes into adi- pocytes that demonstrate a rounded morphology and exten- sive triglyceride accumulation (2,3). Adipocytes differentiated in vitro express a large number of important genes that are identical to those present in fat tissue in vivo (4). These include lipogenic enzymes such as glycerophosphate dehydro- genase ( 5 ) , lipoprotein lipase (6), and acyl-CoA ligase ( 7 ) , as well as adipose-specific proteins like the serine protease adip- sin (8) and the lipid-binding protein adipocyte P2 (9, 10). 3T3-F442A preadipocytes also have the ability to form highly vascularized fat pads when injected into nude mice (1 1). Thus, these cell lines are useful models for cellular and tissue development.

The 3T3 adipocytes have been shown to secrete angiogen- esis-related activities, such as the stimulation of endothelial cell chemotaxis and mitosis (12, 13) in a differentiation- dependent manner. One such factor that stimulates angiogen- esis in chick chorioallantoic membrane assays has been iso- lated and identified as the novel lipid, 1-butyryl glycerol (13). Although this factor is angiogenic, it has no direct mitogenic activity toward endothelial cells. The mitogenic activity pro- duced by cultured adipocytes is specific for endothelial cells as opposed to vascular smooth muscle or fibroblast cell types (14). This feature distinguishes the adipocyte-derived activity from other peptide growth factors, such as acidic fibroblast growth factor (15), epidermal growth factor (16), and platelet- derived growth factor (17), which are less selective in their target cell types.

A number of groups have recently described a family of endothelial cell growth factors that possess mitogenic activity specific for endothelial cells, termed vascular endothelial growth factor (VEGF)’ or vascular permeability factor (VPF). These factors were isolated and purified from a variety of cell types including bovine folliculostellate cells (18), human ad- enocarcinoma (19), guinea pig tumor (20), mouse AtT-20 pituitary (21), and a rat glioma1 cell line (22). VEGF was identified and purified by its potent mitotic activity toward endothelial cells (21, 23) and the VPF by a skin permeability assay (19, 20). The VEGF/VPF are secreted proteins which have a dimeric structure and have sequence homology to the platelet-derived growth factor family of growth factors (23, 24).

At present very little is known about the regulation of VEGF expression during cellular processes such as differen- tiation and transformation. T o investigate the possible role

The abbreviations used are: VEGF, vascular endothelial growth factor; VPF, vascular permeability factor; DMEM, Dulbecco’s modi- fied Eagle’s medium; PMA, phorbol 12-myristate 13-acetate; SDS, sodium dodecyl sulfate.

16317

16318 VEGF Regulation in Differentiation

of VEGF in normal and pathological development, we have studied the expression of this mRNA in adipose differentia- tion and other models of cellular differentiation. We find a striking regulation of the mRNA encoding this endothelial growth factor in models of normal cellular differentiation and in a transformed cell model which can be converted into a non-malignant state by inducing differentiation. In addition, several hormone and second messenger pathways appear to be involved in modulating VEGF gene expression.

EXPERIMENTAL PROCEDURES

Cloning of the Murine Adipocyte VEGF-A cDNA library was synthesized from cellular RNA obtained from differentiated 3T3- F442A cells. The library was custom synthesized (Stratagene Inc.) using Lambda Zap I1 as a propagation vector. Library screening was performed on 250,000 plaques. Hybridization was carried out using the Hybond-N+ membrane according to the manufacturers instruc- tions for the 65 "C hybridization protocol (Amersham Corp.). A probe encoding the entire 600-base pair open reading frame of human VEGF cDNA (a gift from California Biotechnology Inc.) was used to screen the adipocyte library. Five positive clones obtained in the primary screening were plaque purified. Inserts were excised from the Lambda Zap I1 vector and confirmed positive by Southern blot analysis using the human VEGF cDNA as a probe. The largest clone obtained had an insert of 980 base pairs in length which was subsequently se- quenced using double stranded sequencing with the Sequenase kit (United States Biochemical Corp.).

VEGF Protein Amino Acid Sequences and Comparison-Amino acid sequences were generated by translation of the murine VEGF sequence. The GenBank accession numbers for the other VEGF cDNA sequences used are given: rat glioma-derived VEGF (M32167), human VEGF (M32977), bovine VEGF (M32976), and human VPF (M27281). Translated amino acid sequences were then aligned with their apparent homologous sequences.

Cell Lines and Differentiation Protocols-The preadipocyte cell line 3T3-F442A was grown and stimulated to differentiate as described previously (2). The myogenic CzClz and the pheochromocytoma PC12 cell lines were obtained and maintained as per instructions of the distributor (American Type Culture Collection). The C2CI2 cell line was stimulated to differentiate into multinucleated myotubes within 48 h by replacing the Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum with DMEM containing 4% calf serum. The differentiation of PC12 cells into neuron-like cells was performed as described (25) by addition of nerve growth factor (5 ng/ ml) to the media for a period of 2 days.

RNA Isolation and Northern Blot Analysis-Total cellular RNA was isolated from cultured cell lines as described previously (26). Mouse tissue samples were obtained by excising tissues and briefly washing in cold phosphate-buffered saline prior to RNA isolation according to the method of Chirgwin et al. (27). Northern blot analyses were performed using Biotrans nylon supported membranes (ICN) as described by the manufacturer. A mouse 0-actin cDNA was used as a control probe (28). All cDNA probes were radiolabeled with a random prime synthesis kit (Multi-Prime, Amersham, Inc.).

Cellular Labeling and Immunoprecipitation Analysis-Cultures of preadipocytes or adipocytes were washed once with DMEM and placed in methionine-free DMEM (Sigma) containing L-glutamine, penicillin/streptomycin, and [35S]methionine/cysteine Trans-label (ICN) at 1 mCi/ml. The cells were incubated for 6 h at 37 "C and the media collected. The media was centrifuged at 1000 X g for 5 min, the supernatant assayed for 35S label incorporation into protein by trichloroacetic acid precipitation. Supernatants were aliquoted to yield equal counts of incorporated protein and incubated with non- immune serum, anti-human VEGF peptide serum (kindly provided by California Biotechnology Inc.), or anti-human VEGF peptide serum preincubated with excess human VEGF peptide amino acids 4-24 of the mature protein. The immunoprecipitation samples were incubated at 4 "C for 16 h.

Protein-A Sepharose was washed 2 times in 10 mM Tris-HC1, pH 8.0, 150 mM NaCl (TBS), and resuspended in a slurry at a 1:l ratio. The slurry (150 pl) was added to the immunoprecipitation samples and incubated for 1 h at 4 "C. The mixtures werefhen centrifuged at 1000 X g and washed 3 times in TBS and once in 10 mM Tris-HC1, pH 7.4. The washed pellet was resuspended in 100 pl of reducing sodium dodecyl sulfate (SDS) sample buffer (10 mM Tris-HC1, pH 6.8, 100 mM dithiothreitol, 0.15% SDS, 10% glycerol, and 0.05%

bromophenol blue), boiled for 5 min, and centrifuged 2 min. The supernatant was electrophoresed in a 15% acrylamide-SDS gel, treated with EN3HANCE (Du Pont-New England Nuclear), dried, and exposed to Kodak X-Omat film.

Cell Incubations with Hormone and Second Messenger Actiuators- The 3T3-F442A preadipocytes or adipocytes were maintained in 10% calf serum in DMEM or 10% fetal bovine serum in DMEM with insulin at 5 Fg/ml, respectively. The cells were fed 24 h prior to addition of drugs, which were added directly to the media at the following final concentrations: phorbol 12-myristate 13-acetate (PMA) (1 pM), calcium ionophore A23187 (10 PM), dibutyryl-CAMP (50 pM), and 8-bromo-CAMP (50 FM). Incubation of hormones and second messenger activators with cell cultures was for 4 h.

RESULTS

Cloning and Sequence Analysis of the Murine VEGF- Previous work had demonstrated that adipose cell differentia- tion is accompanied by an increase in the secretion of an endothelial cell-specific mitotic activity (14). Since this bioac- tivity was suggestive of a VEGF-like molecule, we investigated the expression of VEGF in differentiating adipocyte cultures. The human VEGF cDNA encoding the entire open reading frame of 600 base pairs was used to probe preadipocyte and adipocyte RNA via Northern blot analysis. A single hybridi- zation signal was observed from adipocyte RNA suggesting the presence of a murine VEGF message (data not shown). This provided the impetus to clone the authentic murine VEGF from a cDNA library synthesized from 3T3-F442A adipocyte RNA. One cDNA clone out of five obtained from the initial primary screen contained a complete open reading frame of 190 amino acids and the nucleotide and deduced amino acid sequence of this clone is given in Fig. 1. The deduced sequence contained a 26-amino acid signal sequence which, when processed, would yield a mature secreted poly- peptide of 164 amino acids. The predicted sequence of 9 amino acids at the N-terminal end of the processed mature polypep- tide (Ala-Pro-Thr-Thr-Glu-Gly-Glu-Gln-Lys) is identical to the limited protein sequence obtained from the mouse AtT- 20 cell-derived VEGF (20). Like other VEGFs, the sequence revealed a putative N-linked glycosylation site (asparagine- isoleucine-threonine) at amino acids 74-76. The conserved signal peptide sequence and glycosylation sites are strikingly homologous to the bovine, human, and rat VEGF cDNA sequences previously described.

Alignment of the predicted amino acid sequences from the murine adipocyte VEGF with the rat glioma-derived (22), human promyelocytic leukemic (18), bovine folliculostellate (18), and human histocytic lymphoma (24) VEGF or VPF proteins is shown in Fig. 2. The murine VEGF amino acid sequence is most closely related to the rat glioma-derived VEGF sequence, revealing a 98% identity at the amino acid level with the observed changes being conservative substitu- tions. The rat and mouse VEGF forms both have some non- conservative amino acid differences from the bovine and human forms of VEGF and the 189-amino acid alternative splice form, VPF. These differences include a pair of threonine residues at positions 3 and 4 of the mature mouse and rat polypeptide and the presence of serine residues at positions 71,87, and 99. Also, the mouse and rat proteins have a cluster of 5 amino acids which differ from the human and bovine sequences at positions 110-112, 115, and 117. The functional implications of these sequence differences are unknown, but the serine and threonine sequences in the rat and mouse VEGF proteins may provide sites for additional 0-linked glycosylation along with the N-linked glycosylation site at amino acid 74.

Induction of VEGF mRNA and Protein during Adipocyte Differentiation-The temporal expression of VEGF mRNA

VEGF Regulation in Differentiation 16319

CAA G(1: ca3 AAG ACA c(x: COc Tcc Cn: CCA CTC CGA GGA

GAG AGG GAG ca3 GAC a OX OX a GCA OX CCT CU: AAACC

ATC AAC TTT Cn: TGc Cn; CAC TGc ACC Cn; CCT T T A Cn; M e t A s n P h e L e u Leu S e r T r p V a l His T r p T h r L e u A l a L e u L e u - 1 7 . a ~

a TAC CAC CAT Coc AACTCC Tcc CAC CCT CCA Cn: ACG ACA L e u T y r L e u His His Ala Lys Trp Ser G ln A la I Ala P r o T h r T h r + 4 a a

GAA GGA GAG CAG A A G Tcc CAT GAA Cn; ATC AAC TIC ATC CAC CK ClU G l y Glu G l n L y s S e r His Glu V a l I I c L y s P h c M e t A s p V a l + 1 9 a a

TAC CAC CGA ACC TAC Tcc CGT CCA A T T GAG ACC Cn; Cn; CAC ATC T y r G l n A r g S e r T y r C y r A r p P r o IIe Glu T h r L e u V a l A s p l l c +34aa

TM) CAC GAG TAC ClX GAC GAG A T A GAG TAC ATC TIC AAG Oa; Tcc P h e G l n G l u T y r P r o A s p Glu Ile Glu T y r I l e Phr Lys P ro Se r +49aa

TGT (;111 CCC Cn; ATG CGC TGT GCA Tcc TGT AAC GAT CAA Coc C y s V a l P r o L e u M e 1 A r g C y s A l a C I y Cy5 Cys Asn Asp Glu Ala +6Jaa

GAG Tcc CIC Cn: ACC TCA GAG ACC AAC ATC ACC ATG CAC ATC L e u G l u C y s V a l P r o T h r S e r Glu Ser b n I l c ‘Lhr Me1 Gln l l e +19aa

ATG CM: ATC AAA CCT CAC CAA AGC CAC CAC A T A GAG ACA ATG AGC M e t A r g I I e L y s P r o H i s Gln Ser Gln His I I e Glu Arg Me1 Se r +94aa

TTC CTA CAG CAC ACC AGA TCT CAA Tcc AGA CCA AAC AAA CAC ACA P h e L e u G l n His S e r A r g Cys Glu C y s A r g P r o L y s L y s A s p A r g +109aa

ACA,AAG CCA C A A A A T CAC TCT GAG KT TGT TCA GAG aX: ACA AAC T h r L y s P r o Glu A s n His Cys Glu Pro Cys Ser Clu A r g A r g L y s + 1 2 4 a a

CAT TTG TTT CTC CAA GAT Oa; CAG ACG TGT AAA TGT TtX TCC AAA His L e u P h e V a l G l n A s p P r o G l n T h r Cys Lys Cys Ser Cys Lys +139aa

AAC ACA CAC TCG CCX X€ AAC AGG CAC ClT GAG TTA AAC CAA A s n T h r A s p S e r A r g C y s L y s A l a A r e Gln L e u Glu L e u A s n Glu +154aa

CCT ACT Tcc AGA TGT CAC AAG CCA ACG CM: TGA CCC ACC Cn; CCA A r g T h r C y s A r g C y s A s p L y s P r o A r g A r g *** +164aa

C C A A C C A C C C T C C T C A C C C T C G G C A A C C A G A C C T C A A C

CATCTCACCACCATGCCATCATCGTCACCGTTGACAGAACAGTCCTTAATCCAGAAAG

CCT(;ATATGAACGAAGAGGAGATCClTCGAGGAGCAC~GGGTCCGGAGGGCGAGAC

TCCGCCAGACCCATM3CCCCCCAGGTGACCAAGCACGTGCCTCGTGGGACEGAlTCCC

CAmTCTTATATCTGCTCCTAAATCGCCAAGCCCGGAAGATTAGGG~G~CTGGGA

TTCCTGTACACCTCCX;

FIG. 1. Nucleotide and deduced amino acid sequence of the murine VEGF cDNA. The complete nucleotide sequence (980 nu- cleotides) of the cloned murine VEGF cDNA obtained was determined and translated into a single open reading frame of 190 amino acids. The amino acids are numbered starting with -26 for the first amino acid and number 1 for the N-terminal amino acid of the predicted mature cleaved polypeptide. Location of the cleavage site for the 26- amino acid signal sequence is indicated by a uerticul line (I). The conserved sequence for N-linked glycosylation is underlined. The translational stop codon TGA is indicated by asterisks (***I.

during murine adipocyte differentiation was evaluated. Fig. 3 demonstrates that there is a dramatic increase in the levels of VEGF mRNA during the transition from the preadipocyte (Fig. 3, lane 1 ) to the adipocyte phenotype (Fig. 3, lanes 2-7) with maximal expression obtained 9 days after confluence. The mouse adipocyte VEGF mRNA appears to be a single species of 4.2 kilobases, in contrast to the multiple forms of VEGF mRNA observed previously in bovine folliculostellate cells (18, 29). As shown previously, actin mRNA decreases during adipocyte differentiation (28). We have asked whether there is a similar regulation of VEGF protein, using an anti- peptide antibody prepared against a human VEGF-derived peptide. Immunoreactive polypeptides of 23 and 17 kDa were observed in the precipitates from adipocyte (Fig. 4A, lane 21, but not preadipocyte (Fig. 4A, lane 1 ) conditioned media. These were similar in size to the guinea pig vascular perme- ability factor polypeptides previously observed (20). To con- firm the identity of the putative adipocyte VEGF, immuno- precipitat,ions were performed with nonimmune sera (Fig. 4B, lane I ) , anti-VEGF antisera (lane 2) , or anti-VEGF antisera which was neutralized by preincubation with the VEGF pep- tide antigen (lane 3 ) . The immune sera that was preincubated with antigen peptide showed no ability to immunoprecipitate the 23- and 17-kDa species, confirming their identity as VEGF-related polypeptides.

- 2 6 MUVEGF MNFLLSWVHW SLALLLYLHH AKWSQAAPTT EGE-QKSHEV IKFMDVYQRS

+ I 23

RATGDVEGF MNFLLSWVHW SLALLLYLHH AKWSQAAPTT EGE-QKAHEV VKFMDVYQRS BOVEGF MNFLLSWVHW SLALLLYLHH AKWSQAAPMA EGG-QKPHEV VKFMDVYQRS HUVEGF MNFLLSWVHW SLALLLYLHH AKWSQAAPMA EGGGQNHHEV VKFMDVYQRS HUVPF MNFLLSWVHW SLALLLYLHH AKWSQAAPMA EGGGQNHHEV VKFMDVYQRS

MUVEGF YCRPIETLVDI FOEYPDEIEY IFKPSCVPLM RCAGCCNDEA LECVPTSESN 24 74

RATGDVEGF YCRPIETLVDI FQEYPDEIEY IFKPSCVPLM RCAGCCNDEA LECVPTSESN ~~ ~ ~~

BOVEGF FCRPIETLVDI FQEYPDEIEF IFKPSCVPLM RCGGCCNDES LECVPTEEFN HUVEGF YCHPIETLVDI FOEYPDEIEY IFKPSCVPLM RCGGCCNDEG LECVPTEESN HUVPF YCHPIETLVDI FQEYPDEIEY IFKPSCVPLM RCGGCCNDEG LECVPTEESN

75 MUVEGF ITMQIMRIKP HQSQHIERMS FLQHSRCECR PKKDRIKPEN HCEPCSERRK RATGDVEGF VTMOIMRIKP HOSOHIGEMS FLOHSRCECR PKKDRTKPEN HCEPCSERRK

124

BOVEGF ITMQIMRIKP HQSQHIGEMS FLQHNKCECR PKKDKARQEN PCGPCSERRK

HUVPF ITMQIMRIKP HQGQHIGEMS FLQHNKCECR PKKDRARQEK* HUVEGF ITMQIMRIKP HQGQHIGEMS FLQHNKCECR PKKDRARQEN PCGPCSERRK

(KSVRGKGKGQKRKRKKSRYKSWSV) PCGPCSERRK

MUVEGF HLFVQDPQTC KCSCKNTDSR CKARQLELNE RTCRCDKPRR I t 4 RATGDVEGF HLFVODPOTC KCSCKNTDSR CKAROLELNE RTCRCDKPRR I 6 4

I 25

BOVEGF HLFVQDPQTC KCSCKNTDSR CKARQLELNE RTCRCDKPRR 1 6 4 HUVEGF HLFVQDPQTC KCSCKNTDSR CKARQLELNE RTCRCDKPRR 165 HUVPF HLFVQDPQTC KCSCKNTDSR CKARQLELNE RTCRCDKPRR 1R9

FIG. 2. Alignment of the deduced murine VEGF amino acid sequence with other VEGFJVPF proteins. Amino acid sequence alignment was performed comparing the murine VEGF (MUVEGF) to the rat glioma-derived VEGF (RATGDVEGF), bovine VEGF (BOVEGF), human VEGF (HUVEGF), and the human vascular permeability factor (HUVPF). Sequences are aligned starting with amino acid -26 of the signal sequence conserved among all family members. The additional 24 amino acids in the alternative spliced human VPF are given below the line a t amino acid 115 and indicated by an asterisk (*) to maintain alignment.

1 2 3 4 5 6 7 ” . I, “ I

- VEGF 4.2 kb

FIG. 3. VEGF mRNA expression during the differentiation of 3T3-F442A adipocytes. Total cellular RNA (10 pg) obtained from 3T3-F442A preadipocytes a t confluence (lune I ) and differen- tiating adipocytes at 1, 2, 3, 5, 7, 9 days post-confluence (lanes 2-7, respectively) were electrophoresed, blotted, and hybridized to the murine VEGF cDNA probe and @-actin as described under “Experi- mental Procedures.” Exposure of film was for 16 h. kb, kilobase.

A 0

UKd#

I7 Kd-

I 2 1 2 3

FIG. 4. Secretion of VEGF-related polypeptides are depend- ent on adipocyte differentiation. Cells were metabolically labeled and 3sS-labeled proteins (2 X IO6 cpm) were immunoprecipitated as described under “Experimental Procedures.” Panel A , immunoprecip- itation of preadipocyte (lune 1 ) and adipocyte (lane 2) conditioned media with anti-human VEGF antisera. Panel B, immunoprecipita- tion of adipocyte conditioned media. Lune 1, nonimmune sera; lune 2, anti-VEGF antisera; lune 3, anti-VEGF antisera preincubated with excess VEGF peptide.

VEGF Expression Is Differentially Regulated in Tissues and Other Cellular Models of Differentiation and Malignant Trans- formation-Because the requirement for neovascularization during cell and tissue development may be general, the expres-

16320 VEGF Regulation in Differentiation

sion of VEGF mRNA was also examined in two other well established models of cell differentiation. We also examined the tissue specificity of VEGF mRNA expression in adult mice.

Similar to the adipocyte model of differentiation, an in- crease in VEGF mRNA was observed during the differentia- tion of muscle cells. Undifferentiated C2C12 myoblasts showed relatively little VEGF mRNA expression (Fig. 5A, lune I), while a large increase was observed during the transition to the myotube multinucleated phenotype (Fig. 5A, lunes 2-4).

Increased angiogenesis is not always correlated with cell differentiation; this is especially true in tumor cells where transformation is often accompanied by the ability to power- fully stimulate angiogenesis (30). In order to probe the rela- tionship between VEGF gene expression and a transformed uersus a differentiated state, we utilized the pheochromocy- toma cell line, PC12. These cells, derived from a rat tumor of neuroendocrine origin, lose the transformed state upon stim- ulation with nerve growth factor and differentiate into non- mitotic neuron-like cells (25). As shown in Fig. 5B, lune 1, the undifferentiated transformed PC12 cells express VEGF mRNA. Upon initial stimulation with nerve growth factor for 8 (lune 2) or 24 h (lune 3 ) the levels show no apparent regulation. However, coincident with morphological differen- tiation (appearance of axonal projections) a t 48 h, the level of VEGF mRNA is dramatically reduced (lune 4 ) . Thus, VEGF is positively regulated during the early differentiation and development of both adipose and muscle cells. In contrast to models of normal development, we observe dramatic down- regulation of VEGF gene expression in a cell model where the transformed state of cells is reversed by differentiation.

We have examined the levels of VEGF in the fully differ- entiated tissues of adult mice. VEGF mRNA was observed in adipose, heart, and skeletal muscle (Fig. 6, lunes 3, 5 , and 6, respectively), although at a lower level than was observed in the differentiating cultured adipocytes (Fig. 6, lune 2). Other tissues demonstrating detectable levels of VEGF mRNA were liver, brain, and kidney (Fig. 6, lunes 4, 7, and 8, respectively). However, there was no detectable VEGF mRNA in either thymus or spleen tissues (Fig. 6, lunes 9 and 10, respectively).

A Variety of Hormonal Activation Pathways Stimulate VEGF Expression-We have also examined VEGF regulation by various second messenger pathways. Since cellular trans- formation may involve activation of protein kinase C (31), we examined the effects of the protein kinase C activator, PMA. Treatment of preadipocytes with PMA for 4 h induced a dramatic increase (approximately 10-fold) in VEGF mRNA compared to untreated cells (Fig. 7, lunes 1 and 2). The PMA

A B c2c12 PC12

1 2 3 4 1 2 3 4 - ". .. . , " ""

I - VEGF - 6 4.2 kb

** - ACTIN - w+.e 2.2 kb

FIG. 5. VEGF mRNA expression in myogenic and neuro- genic differentiation. Northern blot analysis performed on total cellular RNA (10 pg) obtained from myogenic C2CI2 and neurogenic PC12 cell lines. Punel A , confluent C2C12 cells (lune I ) , and differen- tiating C2CI2 cells a t 1, 2, and 3 days (lunes 2-4, respectively). Punel B, control PC12 cells a t day 1 (lune 1 ); PC12 cells treated with nerve growth factor (5 ng/ml) for 8 h (lune 2); 1 day (lune 3); and 2 days (lune 4 ) . Murine VEGF and (3-actin signals are indicated for a 16-h exposure.

1 2 3 4 5 6 7 6 9 1 0 ""~"""-

ACTIN 2.2 kb - * ' 5 .%

FIG. 6. Northern blot analysis comparing VEGF expression in 3T3-F442A cells and mouse tissues. Total cellular RNA (15 pg) was obtained from 3T3-F442A cells and the tissues of 8-week-old mice and subjected to Northern blot analysis using the murine VEGF cDNA as probe. Samples analyzed were: lune I, 1-day post-confluent 3T3-F442A preadipocytes; lune 2, 8-day post-confluent 3T3-F442A adipocytes; lune 3, epididymal adipose tissue; lune 4, liver; lune 5, heart; lune 6, skeletal muscle; lane 7, brain; lane 8, kidney; lune 9, thymus; lune 10, spleen. The signal specific to VEGF and (3-actin are indicated. The blot was exposed for 16 h. kb, kilobase.

1 2 3 4 5 6 7

VEGF - 7 - 4.2 kb

FIG. 7. Regulation of VEGF mRNA expression in preadi- pocytes by second messenger effectors. Northern blot analysis was performed on total cellular RNA (10 pg) obtained from preadi- pocytes treated as described lune 1, untreated; lune 2, PMA (1 p ~ ) for 4 h; lune 3, PMA (1 p ~ ) for 16 h followed by PMA (1 p ~ ) for 4 h; lune 4, PMA (1 pM) for 16 h; lane 5, dibutyryl-CAMP (50 p M ) for 4 h; lune 6, 8-bromo-CAMP (50 p ~ ) for 4 h; and lune 7, calcium ionophore A23187 (10 PM) for 4 h. The signal specific to VEGF and (3-actin are indicated along with their respective sizes. kb, kilobase.

induction of steady state VEGF message levels was completely inhibited by a 16-h PMA preincubation (lune 3 ) which has been shown to down-regulate protein kinase C activity (32). The VEGF message was not affected by a 16-h PMA treat- ment alone (lune 4) , suggesting no role for protein kinase C on the steady state VEGF expression. VEGF mRNA could also be readily induced with the calcium ionophore A23187 (lune 7). The induction of the VEGF mRNA by calcium and PMA is suggestive of regulation of VEGF expression via a protein kinase C-mediated signal transduction pathway.

VEGF mRNA could also be induced with cell-permeable cAMP analogues in preadipocytes. Treatment of preadipo- cytes for 4 h with dibutyryl-CAMP or 8-bromo-CAMP results in an induction of the VEGF mRNA (Fig. 7, lunes 5 and 6, respectively). In additional experiments designed to evaluate the acute regulation of VEGF mRNA in differentiated adi- pocytes, we demonstrated that PMA, A23187, and cAMP analogues also induced VEGF mRNA in these cells in a similar manner (data not shown). These data strongly suggest that the VEGF mRNA is induced by second messenger path- ways mediated by the activation of both protein kinase C and CAMP-dependent protein kinase.

DISCUSSION

Angiogenesis is a critical step in the development of normal tissues and in the progression of malignant tumors. The differentiation of tissues typically requires the formation of new blood vessels to support their proper function. Although a large number of angiogenic factors have been isolated, the

VEGF Regulation in Differentiation 16321

identity of those involved in cell and tissue development have not been clearly established. Angiogenic factors include both polypeptides and nonpolypeptide compounds such as prosta- glandins (33) and 1-butyryl glycerol (13). Many angiogenic polypeptide growth factors, like epidermal growth factor, platelet-derived growth factor, and fibroblast growth factor, have mitotic activities toward a variety of cell types, including vascular endothelial cells. Recently, however, several polypep- tides with mitogenic activity toward endothelial cells but not other cell types have been identified. These include the related vascular endothelial growth factor (VEGF) and vascular permeability factor (VPF), as well as platelet-derived endo- thelial cell growth factor (34). In an attempt to understand the possible roles of these angiogenic polypeptides, we em- ployed systems of cellular and tissue development which regulate their angiogenic activities during differentiation. VEGF, one polypeptide which is a potent stimulator of endo- thelial cell growth, was found here to be dramatically regu- lated in fat and muscle models of cell differentiation and in one cell line in which cellular transformation and differentia- tion are mutually exclusive pathways.

Over the past several years, we have demonstrated that preadipocyte cell lines secrete angiogenic activities during their in vitro differentiation in a manner analogous to the activation of angiogenesis during embryological development of these cells (12). Specifically, 3T3-adipocytes secrete both endothelial cell-specific mitogenic activities and chemotactic substances that are biochemically distinct (13). An adipocyte- derived chemotactic agent was purified and its structure de- termined to be l-butyryl glycerol (monobutyrin); synthetic monobutyrin stimulates both endothelial cell chemotaxis and angiogenesis, but has no detectable mitogenic activity (13). Since endothelial cell proliferation is required to promote neovascularization, an endothelial cell mitogen coordinately secreted during adipocyte differentiation is of considerable interest. The specificity that this adipocyte-derived mitogenic activity shows for endothelial cells suggested that VEGF could be a candidate for this important angiogenic function. We demonstrate here that VEGF mRNA and protein are ex- pressed in a differentiation-dependent manner in 3T3-adipo- cytes, consistent with the endothelial cell mitogenic activity previously observed (14). Moreover, we did not detect the expression of an mRNA encoding another endothelial cell- specific mitogen, platelet-derived endothelial cell growth fac- tor, in the adipocyte cell model (data not shown). While it is likely that VEGF represents a substantial portion of the total endothelial cell mitogenic activity secreted by adipocytes, the low affinity of the antibody preparations used in this study did not permit immunoabsorption experiments which could quantitatively estimate the total VEGF contribution.

Examination of another model of cellular development was used to determine if the induction of VEGF expression during differentiation is general for cell types that are dependent on extensive vascularization. The C2CI2 myogenic cell line also expressed VEGF mRNA in a differentiation-dependent man- ner. Thus, VEGF expression appears to be linked to early differentiation and development in both fat and muscle cell lineages.

Malignant cells also require extensive neovascularization as part of the process of tumor progression (30). Thus there has been great interest in the specific molecules that stimulate tumor angiogenesis. A number of polypeptides with angi- ogenic activities have been isolated from tumors and tumor- derived cell lines, including acidic and basic fibroblast growth factor and VEGF. The use of the pheochromocytoma (PC12) cells in this study provided a convenient model for the com-

parison of the transformed and nontransformed phenotypes in a single cell line (25). In this regard, we demonstrate a strong correlation between the high levels of expression of VEGF mRNA and the tumorigenic state, whereas the mRNA is nearly absent in the differentiated nontumorigenic state. Thus, these data provide a clear linkage in the expression of a potent angiogenic factor and the transformed phenotype. The high level of expression of VEGF in the undifferentiated PC12 cells, as well as previous demonstrations of VEGF protein produced by a number of tumor-derived cells, suggest that activation of VEGF expression may be a common (al- though not necessarily universal) event in transformation.

The experiments performed on the differentiating fat, mus- cle, and PC12 cells suggest that there are highly regulated signal transduction pathways involved in activating VEGF expression. Clearly, the pathways that lead to VEGF expres- sion in normal and pathological cellular development will be important to explore. Based on the data shown here, it is apparent that VEGF mRNA can be induced by a variety of second messenger effectors. Activation of VEGF expression by phorbol ester, calcium ionophore, and CAMP analogues indicates the involvement of protein kinase C and CAMP- dependent protein kinase pathways. A very recent report has also observed the induction of VEGF expression by phorbol ester in human vascular smooth muscle cells (35). Activation of protein kinase C generally involves translocation of the enzyme to a membrane surface (36). Thus, the actual levels of membrane-associated protein kinase C observed in trans- formed cells may be the determining factor regulating VEGF expression. This suggests a general mechanism for linking tumorigenesis and angiogenesis.

Vascular endothelial growth factor is one factor among several that possesses the functional properties suggesting a role in the neovascularization process. Collectively, the data presented here on the regulation of VEGF expression suggest that it may play a very important role in the angiogenesis linked to normal and pathological tissue development.

Acknowledgments-We gratefully acknowledge the gift of anti- human VEGF antisera, VEGF peptide, and recombinant VEGF pro- tein from Drs. Judy Abraham and Stewart Thompson, California Biotechnology Inc. Also, thanks to Yuan-Di Halvorsen for a number of tissue samples.

REFERENCES 1. Wasserman, F. (1965) in Handbook OfPhysiology (Renold, A. E., and Cahill,

G. F., eds) pp. 87-100, American Physiological Society, Washington, D. C.

2. Green, H., and Kehinde, 0. (1976) Cell 7 , 105-113 3. Green, H., and Kehinde, 0. (1973) Cell 1 , 113-114 4. Green, H. (1978) in 10th Miami Symposium on Differentiation and Deuel-

opment (Ahmad, F., Schultz, J., Russell, T. R., and Werner, R., eds) pp. 13, Academic Press, New York

5. Wise, L. S.. and Green, H. (1979) J. Biol. Chem. 254. 273-275 6. Wise, L. S., and Green; H. (1978) Cell 13, 233-242 '

7. Coleman, R. A., Reed, B. C., Mackall, J. C., Student, A. K., Lane, D. M.,

8. Cook , K. S., Groves, D. L., Min, H-Y., Spiegelman, B. M. (1985) Proc. and Bell, R. M. (1978) J. Baol. Chem. 253,7256-7261

Natl. Acad. Sci. U. S. A. 82.6480-6484

10. Bernlohr, D. A,, Angus, C. W., Lane, M. D., Bolanowski, M. A,, and Kelly, 9. Spiegelman, B. M., and Green; H. (1983) J. Biol. Chem. 2 5 8 , 10083-10089

11. Green, H., and Kehinde, 0. (1979) J. Cell. Physiol. 1 0 1 , 169-172 12. Castellot, J. J., Jr., Kamhe, A. M., Dobson, D. E., and Spiegelman, B. M.

13. Dobson, D. E., Kamhe, A. M., Block, E., Dion, T., Lu, H., Castellot, J. J.,

14. Castellot, J. J., Jr., Karnovsky, M. J., and Spiegelman, B. M. (1980) Proc.

15. Gospodarowicz, D., Baird, A,, Cheng, J., Lui, G. M., and Esch, F. (1986)

16. Camenter. G.. and Cohen. S. (1975) J. Cell. Phvsiol. 88. 227-2.18

T. J., Jr. (1984) Proc. Natl. Acad. Sci. 81,5468-5472

(1986) J. Cell. Physiol. 127, 323-329

Jr., and Spiegelman, B. M. (1990) Cell 61,223-230

Natl. Acad. Sei. U. S. A . 77, 6007-6011

Endocrinology 118,82-90

17. Deiel, T. F . (1987) Annu.'Reu. Cell'Biol. 3 , 445-492 18. Leung, D. W., Cachianes, G., Kuang, W. J., Goeddel, D. U., and Ferrara,

19. Lobb, R. R., Key, M. E., Alderman, E. M., and Ferr, J. W. (1985) Int. J.

~~ I~~ ~~-

N. (1989) Science 2 4 6 , 1306-1309

20. Senger, D. R., Connolly, D. T., Van De Water, L., Feder, J., and Dvorak, Cancer 36,473-478

H. F. (1990) Cancer Res. 5 0 , 1774-1778

16322 VEGF Regulation 21. Plouet, J., Schilling, J., and Gospodarowicz, D. (1989) EMBO. J. 8, 3801-

3806 22. Conn, G., Bayne, M. L., Soderman, D. D., Kwok, P. W., Sullivan, K. A,,

Palisi, T. M., Hope, D. A,, and Thomas, K. A. (1990) Proc. Natl. Acad.

23. Conn, G., Soderman, D. D., Schaeffer, M-T., Wile, M., Hatcher, U. B., and

24. Keck, D. J., Hauser, S. D., Krivi, G., Sanzo, K., Warren, T., Feder, J., and Thomas, K. A. (1990) Proc. Natl. Acad. Sci. U. S. A . 8 7 , 1323-1327

25. Greene, L. A., and Tischler, A. S. (1976) Proc. Natl. Acad. Sci. 73, 2424- Connolly, D. T. (1989) Science 246 , 1309-1312

26. Chomczynski, P., and Sacchi, N. (1987) Anal. Biochem. 162, 156-159 2428

27. Chirgwin, J. J., Przbyla, A. E., MacDonald, R. J., and Rutter, W. J. (1979)

28. Spiegelman, B. M., Frank, M., and Green, H. (1983) J. Biol. Chem. 258,

Sci. U. S. A. 87,2628-2632

Biochemrstry 18,5294-5301

10083-10089

in Differentiation 29. Tischer, E., Gospodarowicz, D., Mitchell, R., Silva, M., Schilling, J., Lau,

Res. Comm. 165,1198-1206 K., Crisp, T., Fiddes, J. C., and Abraham, J. A. (1989) Biochem. Biophys.

30. Folkman, J., and Klagshurn, M. (1987) Science 235,442-447 31. Nishizuka, Y. (1984) Nature 308,693-698 32. Blackshear, P. J., Witters, L. A,, Girard, P. R., Kuo, J. F., and Quamo, S.

N. (1985) J. B i d . Chem. 260 , 13304-13315 33. Form, D., and Auerbach, R. (1983) Proc. Soc. Exp. Bid . Med. 172, 214-

218 34. Usuki, K., Heldin, N-E., Miyazono, K., Ishikawa, F., Takaku, F., Wester-

mark, B., and Heldin, C-H. (1989) Proc. Natl. Acad. Sei. U. S. A . 8 6 ,

35. Tischer, E., Mitchell, R., Hartman, T., Silva, M., Gospodarowicz, D., Fiddes, 7427-7431

36. Kikkawa, U., Takai, Y., Minakuchi, R., Inohara, S., and Nishizuka, Y. J. C., and Abraham, J. A. (1991) J. Biol. Chem. 266, 11947-11954

(1982) J. Bid. Chem. 257, 13341-13348