Regulation of VEGF and VEGF Receptor Expression inthe Rodent Mammary Gland During Pregnancy,Lactation, and InvolutionMICHAEL S. PEPPER,1* DANIELLE BAETENS,1 STEFANO J. MANDRIOTA,1 CORINNE DI SANZA,1

SARAH OIKEMUS,2 TIMOTHY F. LANE,3 JESUS V. SORIANO,1 ROBERTO MONTESANO,1 AND

M. LUISA IRUELA-ARISPE2

1Department of Morphology, University Medical Center, Geneva, Switzerland2Department of Molecular, Cell and Developmental Biology, Molecular Biology Institute, University of California,Los Angeles, California3Jonsson Comprehensive Cancer Center, University of California, Los Angeles, California

ABSTRACT Vascular endothelial growthfactors (VEGFs) are endothelial cell-specific mi-togens with potent angiogenic and vascular per-meability-inducing properties. VEGF, VEGF-C,and VEGFRs -1, -2, and -3 were found to be ex-pressed in post-pubertal (virgin) rodent mam-mary glands. VEGF was increased during preg-nancy (5-fold) and lactation (15–19-fold). VEGF-Cwas moderately increased during pregnancy andlactation (2- and 3-fold respectively). VEGF levelswere reduced by approximately 75% in clearedmouse mammary glands devoid of epithelial com-ponents, demonstrating that although the epi-thelial component is the major source of VEGF,approximately 25% is derived from stroma. Thiswas confirmed by the findings (a) that VEGFtranscripts were expressed predominantly inductal and alveolar epithelial cells, and (b) thatVEGF protein was localized to ductal epithelialcells as well as to the stromal compartment in-cluding vascular structures. VEGF was detectedin human milk. Finally, transcripts for VEGFRs-2 and -3 were increased 2–3-fold during preg-nancy, VEGFRs -1, -2 and -3 were increased 2-4-foldduring lactation, and VEGFRs -2 and -3 were de-creased by 20-50% during involution. These resultspoint to a causal role for the VEGF ligand-receptorpairs in pregnancy-associated angiogenesis in themammary gland, and suggest that they may alsoregulate vascular permeability during lactation.Dev Dyn 2000;218:507–524. © 2000 Wiley-Liss, Inc.

Key words: angiogenesis; vascular permeability;breast; growth factor

INTRODUCTION

Angiogenesis is required for the growth of normaland neoplastic tissues, and plays an important role inthe maintenance of the functional and structural integ-rity of the organism. Angiogenesis is particularly im-portant for normal reproductive function, including thecyclical growth of capillaries within the ovary (requiredfor ovulation and corpus luteum formation) and the

endometrium (required for regeneration followingmenstruation). Angiogenesis also occurs following im-plantation of the blastocyst, and is required for theformation of the placenta (Findlay, 1986).

The mammary gland is one of the few organs thatundergo cycles of growth, morphogenesis, differentia-tion, functional activity, and involution in post-natallife. At birth, the rodent mammary gland consists of15–20 branching ducts connected to the nipple by asingle primary duct. From about four weeks post-par-tum, the ducts grow and ramify under the influence ofovarian hormones, forming the mammary gland ductaltree. Following the onset of pregnancy, in response tosustained elevated levels of estrogens and progester-one, ductal elongation and branching resumes, andclusters of alveoli bud off from the growing ducts. Dur-ing the second half of gestation, alveolar morphogene-sis is followed by the structural and functional differ-entiation of alveolar epithelial cells in preparation formilk fat and protein secretion, which occurs duringlactation. After weaning, the mammary gland invo-lutes rapidly, ultimately leaving only a rudimentaryductal system and a few remaining alveoli (reviewed byDaniel and Silberstein, 1987).

Early whole-mount and light microscopic studieson the vascularization of the virgin mammary glanddescribed the presence of a periductal capillaryplexus, which becomes richly developed with advanc-ing pregnancy in association with epithelial growth.This is accompanied by concomitant growth of arte-rioles and venules. During lactation, there is a pro-

Grant Sponsor: Swiss National Science Foundation; Grant number:3100-043364.95; Grant sponsor: NIH-NCI; Grant number:RO1CA63356-01; Grant sponsor: DOD DAMD; Grant number: 17-945-J-4346; Grant sponsor: Fondation Sandoz pour l’avancement dessciences medico-biologiques.

Jesus V. Soriano’s present address is Laboratory of Cellular andMolecular Biology, National Cancer Institute, National Institutes ofHealth, Bethesda, Maryland.

*Correspondence to: Dr. Michael S. Pepper, Departement de Mor-phologie, Centre Medical Universitaire, 1, rue Michel-Servet, 1211Geneve 4, Switzerland. E-mail: michael.pepper@medecine.unige.ch

Received 10 January 2000; Accepted 7 April 2000

DEVELOPMENTAL DYNAMICS 218:507–524 (2000)

© 2000 WILEY-LISS, INC.

gressive increase in capillary size due to dilatation ofexisting capillaries in association with increasedmilk secretory function. During involution, the cap-illary bed progressively disappears, so that relativelythick-walled venules and arterioles appear dispro-portionately large compared with the capillary fieldwhich they supply. Gradually however, the latteralso begin to involute, and ultimately the vascular

bed resembles that of the virgin mammary gland(Wahl, 1915; Turner and Gomez, 1933; Soemarwotoand Bern, 1958). Despite these well described alter-ations in mammary gland vasculature, surprisinglylittle is known about mammary endothelial cell func-tion or the molecular mechanisms which regulateangiogenesis, permeability, and endothelial celldeath in the mammary gland.

Fig. 1. Vascular architecture of mouse mammary gland. Whole-mount staining with BSL at various stagesof the mammary gland cycle (stages indicated in each panel). Scale bars 5 150mm.

508 PEPPER ET AL.

Several polypeptide growth factors have been shownto promote angiogenesis in vivo, one of the most impor-tant of which is vascular endothelial growth factor(VEGF). VEGF is an endothelial cell-specific mitogenwith potent angiogenic and vascular permeability-in-ducing properties, both of which may be important formammary gland function. High affinity VEGF-cell sur-face interactions occur via two transmembrane ty-rosine kinase receptors, VEGFR-1 (Flt-1) andVEGFR-2 (KDR, Flk-1). VEGF-C is a recently de-scribed member of the VEGF family, which interactswith VEGFR-2 and VEGFR-3 (Flt-4). VEGF-C sharesthe major functions of VEGF including induction ofangiogenesis and vascular permeability, presumablythrough its interaction with VEGFR-2. However, byvirtue of its interaction with VEGFR-3, VEGF-C alsoinduces growth of the lymphatic vascular system, aprocess referred to as lymphangiogenesis (reviewed byDvorak et al., 1995; Ferrara, 1999; Stacker and Achen,1999; Veikkola and Alitalo, 1999).

The studies reported in this paper have focussed onthe rodent mammary gland. Our first objective was toquantitate vascularization during pregnancy, lactationand involution following weaning. We next assessedthe expression of VEGF and VEGF-C mRNA duringthis cycle, and attempted to localize VEGF mRNA andprotein. Experiments were performed to determine theimportance of lactation, the relative contributions ofepithelial and stromal components and the possiblecontribution of hormonal and polypeptide factors toVEGF synthesis. Levels of VEGF were also determinedin human milk. Finally, the expression of VEGFRs wasassessed at various stages of the mammary cycle.Our findings demonstrate that VEGF, VEGF-C, andVEGFR’s -1, -2, and 3 are expressed in the postpuber-tal mammary gland, and that alterations in expressionduring pregnancy, lactation and involution are compat-ible with a role for these ligand-receptor pairs duringpregnancy-associated vascular growth and permeabil-ity.

RESULTSAlterations in Vascular Density DuringPregnancy, Lactation, and Involution

To ascertain whether vascular density is altered dur-ing the mammary gland cycle, mouse mammary glandswere stained with Bandeiraea simplicifolia lectin(BSL) which recognizes mouse endothelium (Alroy etal., 1987) (Fig. 1 and Table 1). At 7–8 weeks the maturevirgin mammary gland shows a fully developed capil-lary system (Fig. 1A), which increases significantly indensity during early pregnancy (Fig. 1B). An increasein capillary loops can be seen as early as 3 days afterpregnancy. Vascular networks surround the expandingparenchyma and developing alveolar units (Fig. 1C).During lactation, expansion of terminal alveoli dis-tends the capillary networks; however this stage is notassociated with increased vascular growth (Fig. 1D); infact, the vascular density appears to decrease slightly.

This is probably artifactual and due to alveolar disten-sion during lactation. Regression of parenchyma uponweaning results in involution of capillary loops. Vesselsappear to be thinner and further apart; eventuallymany of these vessels regress. These events are alsoassociated with loss of pericytes and complete obliter-ation of the lumen (data not shown). A 7-day regressinggland is shown in Fig. 1E. Note dark epithelial unitsand thin vascular networks in between the involutingalveoli. Complete regression of the parenchyma is seen21 days after weaning (compare Fig. 1A, F). At thisstage, capillary density is comparable with that seen inthe early stages of pregnancy (Table 1).

Cloning of Partial Rat VEGF-C, VEGFR-3, andCD31 cDNAs

Partial cDNAs were cloned from adult rat lung asdescribed below. Nucleotide and amino acid sequencesfor VEGF-C, VEGFR-3, and CD31/PECAM-1 areshown in Figure 2. For rat VEGF-C (GenBank acces-sion no. AF010302), identity at the nucleotide level was97%, 88%, and 80% compared to mouse, human, andquail VEGF-C respectively. Identity at the amino acidlevel was 97%, 88%, and 80% for mouse, human, andquail. For rat VEGFR-3 (GenBank accession no.AF010131), identity at the nucleotide level was 93%and 82% compared to mouse and human VEGFR-3respectively. Identity at the amino acid level was90% and 74% for mouse and human. For rat CD31/PECAM-1 (GenBank accession no. U77697), identity atthe nucleotide level was 88%, 73%, 68%, and 73% com-pared to mouse, human, bovine, and porcine CD31/PECAM-1 respectively (data not shown). Identity atthe amino acid level was 84%, 66%, 59%, and 63% formouse, human, bovine, and porcine CD31/PECAM-1.

TABLE 1. Quantitation of Capillary Density in theMouse Mammary Gland During Pregnancy, Lactation,

and Involution

StageCapillary density (per mm2)

(mean 6 s.d.)Virgin

5 weeks 27 6 4.67 weeks 32 6 8.910 weeks 39 6 4.5

Pregnant5 days 67 6 6.514 days 110 6 3.419 days 198 6 7.6

Lactation2 days 119 6 5.74 days 105 6 7.57 days 109 6 6.5

Involution2 days 93 6 7.64 days 72 6 8.77 days 52 6 5.721 days 69 6 6.4

509VEGF AND RECEPTORS IN MAMMARY GLAND

Figure 2.

VEGF But Not VEGF-C Is Increased DuringPregnancy and Lactation

RNase protection analysis using a cRNA probe thatincludes the common and alternatively spliced regionsof the rat 164-amino acid VEGF isoform (VEGF164),revealed the presence of VEGF transcripts in the virginrat mammary gland (Fig. 3A). This was confirmed byRT-PCR, in which three bands were detected, whichbased on their size, represent mRNAs which code forthe rodent 120, 164, and 188 amino acid VEGF iso-forms (Breier et al., 1992) (data not shown).

To determine whether quantitative differences occurin expression of VEGF164, RNase protection analysiswas performed on total cellular mRNA from variousstages of the mammary cycle (Fig. 3A). To overcome theproblem of transcript dilution which occurs as a conse-quence of milk protein synthesis, an appropriatehousekeeping gene was sought for normalization ofmRNA levels. It has previously been reported thatexpression of the GAPDH gene is not uniform duringthe mammary gland cycle, but that it decreases pro-gressively as one proceeds through pregnancy and intolactation (Gavin and McMahon, 1992). A reciprocalincrease was seen in b-casein, reflecting the large in-crease in milk protein mRNAs during this period (Na-kahasi and Qasba, 1979; Fig. 4A). Levels of a secondhousekeeping gene, b-actin, paralled those of GAPDH(Gavin and McMahon, 1992; Fig. 3A). To correct for therelative decrease in non-milk protein mRNAs, VEGFwas therefore normalized relative to GAPDH. Figure3B shows that mRNA for VEGF164 is increased duringpregnancy (5.0-fold increase on day 12 with a subse-quent decrease on day 18) and during lactation (18.5-fold increase on day 7). Levels of VEGF appear to beminimally altered during involution. This experimenthas been repeated four times with similar results, andthe quantitative data in Figure 3B, which representthe means 6 s.d. of these four experiments, indicateminimal variation between experiments. The samplesanalyzed were pools of total cellular RNA from 3–7animals per stage. Similar results were obtained whenVEGF mRNA levels were determined by northern blotanalysis and expressed relative to GAPDH on the samefilter (data not shown). RT-PCR performed on the samesamples revealed that VEGF 120, 164, and 188 aminoacid isoforms were expressed during pregnancy, lacta-tion and involution, and their abundance relative toone another did not appear to change during the mam-mary cycle (data not shown).

Figure 2. (Continued.)

Fig. 2. Partial cDNA sequences of VEGF-C, VEGFR-3, and PECAM-1/CD31 from rat lung obtained by RT-PCR. (A) Partial cDNA sequence ofrat VEGF-C (GenBank accession no. AF010302), and comparison withamino acid residues of rat (r), mouse (m), human (h), and quail (q)homologues. (B) Partial cDNA sequence of rat VEGFR-3 (GenBankaccession no. AF010131), and comparison with amino acid residues ofrat (r), mouse (m), and human (h) homologues. (C) Partial cDNA se-quence of rat PECAM-1 (CD31) (GenBank accession no. U77697) andcomparison with amino acid residues of rat (r), mouse (m), human (h),bovine (b), and porcine (p) homologues.

Results obtained in rat mammary glands promptedus to extend our studies to the mouse. Northern blotanalysis of RNA isolated at various stages of the mam-mary cycle (Fig. 4A) demonstrated an increase inVEGF mRNA levels throughout pregnancy (maximum5.5-fold at 5 days) and a more marked increase duringlactation (maximal 9.7-fold at 7 days); levels of VEGFdecreased progressively during the phase of involution(Fig. 4B). The results shown in Fig. 4B were obtainedfrom 5 animals per time point. Samples from eachanimal were analyzed separately, and were normalizedto GAPDH to account for the dilution factor. These dataconfirm the increase seen in the rat during pregnancyand lactation.

RNAse protection analysis also revealed the pres-ence of VEGF-C transcripts in the virgin rat mammarygland (Fig. 3A). However, in contrast to VEGF, thelevels of VEGF-C mRNA, when expressed relative toGAPDH, were only modestly increased during preg-nancy (2.8-fold increase on day 4) and lactation (1.9-fold increase on day 2) (Fig. 3A, B).

To further establish the relationship between in-creased levels of VEGF mRNA and lactation, rat pupswere removed from lactating mothers, and after two

days were reintroduced for a further three days. In aparallel (control) group, pups were removed for a pe-riod of five days. Analysis of VEGF mRNA expression(relative to GAPDH) revealed a 4.1-fold increase in theexperimental group relative to controls (Fig. 5A andand data not shown), thereby establishing a clear rela-tionship between lactation and increased VEGF ex-pression. The increase in VEGF was in contrast to thedecrease in hepatocyte growth factor (HGF) and itsreceptor c-Met (71% and 86% respectively, when nor-malized to GAPDH) (Fig. 5A), previously observed un-der the same conditions (Pepper et al., 1995).

Mammary tissue is comprised of parenchymal (epi-thelial structures) and stromal (connective tissue,blood and lymphatic vessels, and nerves) compart-ments. Since fibroblasts, epithelium, endothelial, andsmooth muscle cells have been reported to synthesizeVEGF in vitro and in several organs in vivo, we eval-uated the relative contribution of parenchyma versusstroma by examining VEGF mRNA expression incleared murine mammary glands devoid of epithelialcomponents. For these experiments, the nipple andepithelial primordia were surgically removed from3-week-old female mice. After an additional 7 weeks,

Figure 3.

512 PEPPER ET AL.

RNA was isolated from operated and control glands(contra-lateral sham-operated glands). Northern blotswere hybridized with probes for VEGF and keratin 8,the latter of which was included to assess for the pres-ence and relative amount of epithelium in operated andsham-operated glands. A weak signal for VEGF mRNAwas detected in the operated glands (Fig. 5B). Quanti-tation revealed that 15–35% of VEGF mRNA is synthe-sized by stroma (operated glands), indicating thatmammary epithelium is the major source of VEGFmRNA.

To determine the localization of VEGF transcripts, insitu hybridization was performed on rat mammaryglands at various stages of the cycle. This revealedexpression of VEGF in the ductal epithelial cells of thevirgin mammary gland (data not shown), and in theductal and alveolar epithelial cells during pregnancy(Fig. 6B, D), confirming the data obtained in operatedmouse mammary glands described above. A positivesignal was also seen over alveolar and ductal cellsduring lactation (data not shown), although the preciselocalization was more difficult to determine due to the

gross morphological distortion which occurs during thisphase. Epithelial localization of VEGF mRNA was alsoobserved during involution (data not shown).

Localization of VEGF protein was achieved using anaffinity purified antibody against a synthetic peptidecorresponding to amino acids 1–20 of the amino termi-nus of human VEGF. In the rat, positive cytoplasmicimmunoreactivity was observed in ductal epithelialcells of the virgin mammary gland (Fig. 7A), in ductaland alveolar epithelial cells throughout the whole pe-

Fig. 3. (Continued.) Expression of VEGF and VEGF-C mRNA in therat mammary gland during pregnancy, lactation and involution—RNaseprotection analysis. (A) Total cellular RNA was analyzed for VEGF164 andVEGF-C mRNA expression (10 mg/lane). Lane 1: molecular size mark-ers; lane 2: cRNA probe; lane 3: cRNA probe 1 hybridization mixture;lane 4: yeast tRNA; lane 5: rat lung. Northern blots of the same sampleswere performed for GAPDH (5 mg/lane) and b-actin (10 mg/lane). 18Sribosomal RNA is shown as an indicator of RNA loading and integrity.Samples represent pools of total cellular RNA from 3–7 animals perstage. (B) Quantitation of the data shown in (A). All values were normal-ized to GAPDH, and expressed relative to 3-month virgins. Error bars inVEGF indicate the mean 6 s.d. from four independent experiments onthe same samples.

Fig. 4. Expression of VEGF mRNA in mouse mammary gland duringpregnancy, lactation and involution—Northern blot analysis. (A) 10 mg oftotal RNA were resolved on 1% denaturing gels and transferred to nylonmembranes. Blots were probed with cDNA fragments corresponding toVEGF, keratin 8 (K8) and GAPDH. Note the increased levels of K8 duringpregnancy, which relate to the relative increase of mammary epithelialcells during this period. Gels were stained with ethidium bromide (lowerpanel). (B) Values in the histogram correspond to relative levels of VEGFmRNA after normalization to GAPDH, and are expressed relative to5-week virgins. Scale bars represent the mean of 5 animals 6 s.d.

513VEGF AND RECEPTORS IN MAMMARY GLAND

riod of gestational evolution (Fig. 7B and data notshown) as well as during lactation (Fig. 7C), and inepithelial cells during involution (Fig. 7D). At allstages, DAB reaction product was also seen in micro-vascular profiles as well as in the wall of medium andlarge sized blood vessels. Immunostaining was com-pletely abolished after incubation of the sections withthe anti-VEGF antibody preabsorbed with its corre-sponding peptide at a concentration of $ 0.05 mg/ml(Fig. 7E, F). Incubation without primary antibody orwith non-immune serum as well as use of the secondantibodies or DAB alone failed to stain any structure

(not shown). To determine whether VEGF is alsopresent in human mammary tissue, immunohisto-chemistry was performed on non-pregnant humanmammary gland with the anti-VEGF antibody de-scribed above. Strong VEGF immunoreactivity was ob-served in the cytoplasm of ductal epithelial cells, aswell as in isolated cells and blood vessels in the stroma(Fig. 7G, H).

To assess whether hormonal or other factors mightincrease VEGF mRNA in cultured mammary epithelialcells, we exposed cells from a clonal mouse mammaryepithelial cell line (called TAC-2, Soriano et al., 1995)to a variety of mammotrophic factors for 18 hours inmedium containing 10% serum. Levels of VEGF mRNAwere determined by Northern blot hybridization andexpressed relative to GAPDH. With the exception ofHGF, which decreased levels of VEGF mRNA, none ofthe other factors analyzed significantly altered VEGFmRNA levels in these cells (Fig. 5C). The factors as-sessed included 17-b estradiol (10–100 nM), progester-one (1 mg/ml), prolactin (5 mg/ml), hydrocortisone(0.1–10 mg/ml), aldosterone (1 mg/ml), insulin (10 mg/ml), and heregulin-a (20 ng/ml), as well as the inactivesteroids tetrahydrocortisone (1 mg/ml) and 17-a hy-droxyprogesterone (1 mg/ml) (Fig. 5C and data notshown).

Finally, to determine whether VEGF might be se-creted into human milk, levels were measured byELISA in milk obtained from lactating mothers duringthe first week post-partum. VEGF levels ranged from38–220 ng/ml using a sandwich ELISA, or from 17–100ng/ml using a competitive ELISA (Fig. 8). The reasonsfor the approximately two-fold difference in detectionlevels using the two ELISA methods are not known.However, it is important to note that differences be-tween the different samples were reproducible with thetwo ELISA kits. Attempts at isolating milk from the ratmammary gland were repeatedly unsuccessful.

Fig. 5. (A) Induction of VEGF during lactation—RNase protection.Samples analyzed were as follows: 5 days post-lactation (5d post-lact.),2 animals; replacement of pups after 2 days of weaning for a further 3days (2 days p.l. 1 3d pups), 3 animals. All values were normalized toGAPDH, and expressed relative to the first 5-day post-lactation animal.Results are mean 6 s.d.. Values for HGF and c-Met are from datadescribed in Pepper et al. (1995), and are also expressed relative toGAPDH. (B) Contribution of epithelial versus stromal components toVEGF expression. Fifteen (15) mg of total cellular RNA were resolved on1% denaturing gels and transferred to nylon membranes. Blots wereprobed with cDNA fragments corresponding to VEGF and K8. To nor-malize for loading and transfer efficiency, blots were also probed withGAPDH. (C) regulation of VEGF expression in cultured murine mammaryepithelial cells. Confluent monolayers of TAC-2 cells were exposed toHGF (10 ng/ml) or heregulin-a (HRGa - 20 ng/ml) for 18 hr, and levels ofVEGF mRNA determined by Northern blot analysis. The results of threeindependent experiments are shown, in which VEGF and HRG-a werenormalized to GAPDH. Arrows in the left hand panel indicate meanvalues.

514 PEPPER ET AL.

VEGFR Expression During Pregnancy,Lactation, and Involution

VEGFR-1, -2, and -3 transcript levels were measuredin rat mammary glands by RNase protection analysis(Fig. 9A). Levels of VEGFR mRNA were normalized toCD31 mRNA as a means of accounting for alterationsin endothelial cell number. In these experiments, wehave assumed that CD31 mRNA expression is not al-tered in mammary endothelial cells under the condi-

tions we have analyzed, as has been assumed andpublished by others (Thomasset et al., 1998). A quan-titative analysis revealed an increase in VEGFR-2 (1.6-fold at 4 days) and VEGFR-3 (2.2-fold at 4 days) duringpregnancy. During lactation, VEGFR-1 (2.7-fold at 7days), VEGFR-2 (3.8-fold at 7 days), and VEGFR-3(1.5-fold at 21 days) were increased. Both VEGFR-2(45%, 50% and 34% on days 1, 2, and 3 respectively)and VEGFR-3 (33%, 21%, and 45% on days 1, 2, and 3

Fig. 6. Localization of VEGF transcripts in pregnant rat mammary gland by in situ hybridization. Paraform-aldehyde-fixed cryostat sections of 18-day pregnant rat mammary gland were hybridized with a [3H]-labelledrat VEGF cRNA probe. (B,D,F) are the dark field views of the corresponding bright field images in (A,C,E)stained with 0.2% methylene blue. (A–D) antisense probe. (E,F) sense probe. Scale bars 5 100mm.

515VEGF AND RECEPTORS IN MAMMARY GLAND

Fig. 7. Immunolocalization of VEGF in rat and human mammarygland. Bouin-fixed paraffin sections were incubated with an affinity puri-fied rabbit polyclonal antibody prepared against a synthetic peptide cor-responding to amino acids 1–20 of the amino terminus of human VEGF.(A–F) Rat mammary gland: (A) 3-month virgin rat; (B) 12-day pregnant

rat; (C and E) 7-day lactating rat; (D and F) 21 days post-lactation, rat. In(E) and (F) the antibody was preincubated with the corresponding antigenat the following concentrations: E 5 2 mg/ml; F 5 0.5 mg/ml. Scale bar 5100mm. (G,H) Non-pregnant human mammary gland: (G) Hematoxylinand eosin; (H) anti-VEGF antibody. Scale bar 5 150mm.

516 PEPPER ET AL.

respectively) were decreased in the early phases ofinvolution (Fig. 9B). To confirm the results obtainedwith VEGFR-2, the RNase protection analysis was re-peated using cRNA probes generated from non-over-lapping cDNAs for both VEGFR-2 (430 bp fragment)and CD31 (260 bp C9 fragment). The quantitative datashown in Figure 9B represent the means 6 s.d. for thetwo experiments. All samples analyzed were pools oftotal cellular RNA from 3–7 animals per stage.

To assess whether VEGFR-2 expression was modu-lated in the mouse, Northern blot hybridization wasperformed (Fig. 10A). VEGFR-2 mRNA levels weremoderately increased during pregnancy (maximum2.7-fold at 19 days), and increased further during lac-tation (maximum 3.7-fold increase on day 7) (Fig. 11B).The results shown in Fig. 10B were obtained from 7animals per time point. Samples from each animalwere analyzed separately, and as in the rat, VEGFR-2levels were normalized to CD31 in the same samples.These results confirm the observations in the rat mam-mary gland, namely that VEGFR-2 mRNA levels, whennormalized to CD31 (and presumably therefore for en-dothelial cell abundance), are increased in the preg-nant, and more significantly in the lactating mammarygland.

Attempts at localizing VEGFR-1 and -2 transcriptsin rat mammary glands by in situ hybridization were

repeatedly unsuccessful. Similarly, localization ofVEGFR-2 by immunohistochemistry using antibodiesfrom numerous sources did not produce acceptable re-sults. In contrast, antibodies to VEGFR-1 clearly de-tected immunoreactivity on endothelial cells, whichwere identified by staining with antibodies to RECA-1(a rat endothelial cell-specific antibody; Duijvestijn etal., 1992) and type IV collagen (basement membranes)(Fig. 11).

DISCUSSION

The existence of a precise sequence of tightly-sched-uled, hormonally-regulated changes in tissue functionand architecture makes the mammary gland an attrac-tive model for studying the mechanisms of angiogene-sis (i.e., during pregnancy), vascular permeability (i.e.,during lactation), and endothelial cell death (i.e., dur-ing post-weaning involution). Previous descriptivestudies have revealed that the vascular bed is a highlydynamic structure which undergoes profound quali-tatitive and quantitative alterations during the mam-mary cycle (Wahl, 1915; Soemarwoto and Bern, 1958;Yasugi et al., 1989; Matsumoto et al., 1992). We havedeveloped a novel whole-mount technique based on theaffinity of BSL for mouse endothelium (Alroy et al.,1987) in order to quantitate vascularization in themammary gland. We demonstrate that vascular den-sity increases progressively during pregnancy, reach-ing a maximum on day 19, and subsequently decreasesduring lactation, to levels observed during mid-preg-nancy. During involution, there is a further reductionin vessel density. Our findings in the mouse are con-sistent with a study in the rat using India ink injection(Yasugi et al., 1989). These authors found that vesseldensity increased progressively during pregnancy,reached a peak at day 10, and decreased progressivelythereafter through late pregnancy, lactation, and invo-lution.

In addition, we have found that VEGF (188, 164, and120 amino acid isoforms (Breier et al., 1992)), VEGF-Cand VEGFRs -1, -2, and -3 are expressed in the virginrodent mammary gland, with VEGF mRNA and pro-tein being localized predominantly in the cytoplasm ofmammary epithelial cells. Low levels of VEGF mRNAhave previously been observed in terminal duct epithe-lium of normal human breast tissue (Brown et al.,1995), and VEGF immunoreactivity has been observedin these cells in normal human and non-human pri-mate mammary gland (Nakamura et al., 1999, and ourown observations). Low levels of VEGF protein (asdetermined by ELISA) and VEGF mRNA (189, 165,and 121 amino acid isoforms) have been reported innon-neoplastic human breast tissue (Obermair et al.,1997; Greb et al., 1999), and weak VEGF-C immuno-reactivity has been reported in normal human mam-mary epithelium (Valtola et al., 1999). Using a dorsalskinfold chamber model in nude mice, Lichtenbeld etal. (1998) have reported that in contrast to tissue sam-ples from breast cancer, samples from normal healthy

Fig. 8. Detection of VEGF in human milk. Samples from 5 lactatingmothers in the first week post-partum (1–5) were analyzed by sandwichor competitive ELISA as described in Experimental Procedures.

517VEGF AND RECEPTORS IN MAMMARY GLAND

breasts had no angiogenic activity. The precise role ofthe VEGF ligand-receptor system in the normal mam-mary gland is not known.

We have observed that VEGF and VEGFR-2 (whichmediates VEGF signal transduction and induces endo-thelial cell mitogenesis) are increased during preg-nancy. These findings, which were observed in bothmouse and rat, point to a causal role for VEGF-VEGFR-2 interactions in the increase in vasculariza-tion which occurs during pregnancy-associated mam-mary growth. Whether the downregulation of VEGFRsobserved following weaning is causally related to invo-lution of the vasculature, is not known. Expression ofVEGF-C and VEGFR-3 was only modestly increasedduring pregnancy and lactation. VEGFR-3, which isknown to be specific for lymphatic vessels in most adulttissues, is also expressed in the blood capillary endo-thelial cells in the resting mammary gland (Valtola et

al., 1999). The function of VEGF-C in the mammarygland is still unknown.

The greatest increase in VEGF and VEGFRs -1 and-2 was seen during lactation. What is the relevance ofthese findings? The capillary is the primary site oftransport between the blood and alveolar epithelium,and constitutes the so-called blood-milk barrier. In thevirgin mammary gland, the capillary wall is usuallycontinuous with rare fenestrations of 30–55 nm(Stirling and Chandler, 1976; Matsumoto et al., 1992).During lactation, there is a marked increase in perme-ability as well as the number of cytoplasmic vesicles inendothelial cells surrounding alveoli (Matsumoto et al.,1992, 1994). These vesicles tend to fuse with one an-other to form clusters. These authors also describe astriking increase in the number and length of microvil-lous processes emanating from the surface of paren-chyma-associated capillary endothelial cells. Following

Figure 9.

518 PEPPER ET AL.

weaning, these features progressively disappear, withthe endothelium ultimately returning to its pre-preg-nancy state. VEGF, which was first identified as aresult of its capacity to increase vascular permeability(reviewed by Dvorak et al., 1995), induces morpholog-ical alterations in endothelial cells, including the ap-pearance of vesiculo-vacuolar organelles (reviewed byFeng et al., 1999), which are entirely consistent with

this function. We suggest that during lactation, themarked increase in VEGF induces functional alter-ations in endothelial cells which lead to the increase invascular permeability. The function, if any, of VEGF inhuman milk (Siafakis et al., 1999; this paper), is notknown.

It has previously been demonstrated that mammarygland fat pads devoid of parenchyma fail to develop thevascular pattern associated with pregnancy (Soemar-woto and Bern, 1958). The authors concluded that thevarious changes observed in the vascular bed occur inassociation with alterations in the epithelial structures(ducts and alveoli), and that these changes do not occurin the absence of parenchyma. Our in situ and immu-

Fig. 9. (Continued.) Expression of VEGFR-1, -2, and -3 mRNA in ratmammary gland during pregnancy, lactation and involution. (A) Totalcellular RNA was analyzed by RNase protection for VEGFR-1, VEGFR-2(350 bp fragment), VEGFR-3 and CD31 (450 bp 39 fragment) (10 mg/lane). Lane 1: molecular size markers; lane 2: cRNA probe; lane 3: cRNAprobe 1 hybridization mixture; lane 4: yeast tRNA; lane 5: rat lung forVEGFRs, rat kidney for CD31. 18S ribosomal RNA is shown as anindicator of RNA loading and integrity. Samples represent pools of totalcellular RNA from 3–7 animals per stage. (B) Quantitation of the datashown in (A). VEGFR values were normalized to CD31 and expressedrelative to 3-month virgins. Error bars in VEGFR-2 indicate the mean 6s.d. from two independent analyses with non-overlapping cRNA probesfor VEGFR-2 and CD31.

Fig. 10. Expression of VEGFR-2 mRNA in mouse mammary glandduring pregnancy, lactation and involution. Northern blot analysis. (A)Total cellular RNA was extracted and resolved as in Figure 4. Blots wereprobed with cDNA fragments corresponding to VEGFR-2 and CD31. Gelswere stained with ethidium bromide (lower panel). (B) Values in thehistogram correspond to relative levels of VEGFR-2 mRNA after normal-ization to CD31, and are expressed relative to 5-week virgins. Each barrepresent the mean of 7 animals 6 s.d.

519VEGF AND RECEPTORS IN MAMMARY GLAND

nohistochemistry data suggest that although VEGF isexpressed predominantly by epithelial cells, there isalso a significant contribution from the stroma. To fur-ther explore this issue, we have studied VEGF expres-sion in cleared murine mammary glands of 10-week-oldanimals, in which the epithelial component had beensurgically removed 7 weeks earlier. This allowed us tocompare epithelial- and non-epithelial-containing com-

partments. We found that, although VEGF is ex-pressed predominantly by epithelial cells, the stromalcomponent accounts for approximately 25% of VEGFexpression in operated glands. VEGF has been de-tected in early passage mammary fibroblast cultures,in which its expression is dramatically increased byhypoxia (Hlatky et al., 1994).

What are the factors that regulate VEGF expressionduring pregnancy and lactation? Using a clonal mousemammary epithelial cell line (Soriano et al., 1995), wehave been unable to identify hormonal factors whichmight be implicated in the regulation of VEGF expres-sion in vitro. However, VEGF was downregulated byHGF in these cells. Interestingly, the period of maxi-mal VEGF expression (i.e., lactation) is also the periodduring which HGF and c-Met are markedly downregu-lated (Pepper et al., 1995; Yang et al., 1995). It hasbeen demonstrated that HGF, which regulates manyepithelial cell functions (Matsumoto and Nakamura,1996) including the induction of mammary ductal mor-phogenesis (Soriano et al., 1995), is downregulatedduring lactation (Pepper et al., 1995; Yang et al., 1995),when the HGF-dependent tubulogenic process is essen-tially completed. It is not possible at this point to as-certain whether there is causalty in the inverse rela-tionship between VEGF and HGF levels.

Working with the mammary gland presents a num-ber of challenges. First, the changing cellular composi-tion during pregnancy, lactation, and involution re-quires separate consideration of the epithelial andstromal components. Indeed, we have found that bothepithelial and stromal components contribute to VEGFexpression. Second, we have used CD31 as an internalcontrol for endothelial cell number when assessing lev-els of VEGFR expression. This was prompted by aprevious report in which the authors expressed mam-mary gland vascularization in terms of CD31 expres-sion (Thomasset et al., 1998). However, like these au-thors, we have had to assume that endothelial CD31expression is not modulated during the various phaseswe have studied. The third problem concerns dilutionof the mRNAs of interest by milk protein mRNAs,whose production is massively increased during thelate stages of pregnancy and lactation (Nakahasi andQasba, 1979). To overcome this problem we have nor-malized all our samples to the housekeeping geneGAPDH, although very similar results were obtainedwhen b-actin was used in the same way. This approachhad previously been reported by others (Gavin andMcMahon, 1992).

Finally, the relationship between microvessel den-sity and breast cancer progression has previously beenreported (Weidner et al., 1991). The controlled increaseand subsequent decrease we have observed in VEGF/VEGFR expression in response to the physiological re-quirements of the mammary gland, contrast sharplywith the sustained elevation in VEGF and its receptorsreported in breast carcinoma (Brown et al., 1995;Gasparini et al., 1997; Obermair et al., 1997).

Fig. 11. Immunolocalization of VEGFR-1 in pregnant rat mammarygland. Cryostat sections of 18-day pregnant rat mammary gland wereimmunostained with (A) RECA-1 antibody, (B) anti-VEGFR-1 and (C)anti-type IV collagen. The section in (A) was counterstained with 0.06%Evans Blue for 20 seconds. Scale bar 5 50mm.

520 PEPPER ET AL.

In summary, the results obtained in this study pointto a causal role of VEGF and its receptors in the in-crease in vascularization that occurs during preg-nancy. They also support a role for VEGF in increasingvascular permeability during lactation, when increasedtransport of molecules from the blood is required forefficient milk protein synthesis.

EXPERIMENTAL PROCEDURESAnimals and Tissues

Mammary glands were obtained from adult femaleSprague-Dawley rats or Swiss Webster mice. Tissue sam-ples were frozen in liquid-nitrogen-cooled isopentane, andstored at 280°C until use. The staging of most animalswas confirmed on conventional hematoxylin and eosin-stained sections from Bouin-fixed paraffin-embedded tis-sues. In one series of experiments, surgical ablation of theepithelial components of the mammary gland was per-formed on 3-week-old mice, which were then allowed tomature to 10 weeks before isolation of mammary tissue.Verification of epithelial ablation was performed in par-allel experiments by staining with Carmine red and byevaluating levels of keratin 8 mRNA in the operatedglands. Contralateral sham-operated glands were used asa control. Human mammary tissue obtained from a 44year old woman, who died suddenly of a cerebral vascularaccident, was fixed in Bouin’s solution and embedded inparaffin. Details of fertility and parity were not available.

Whole-Mount Staining and Quantitation ofVessel Density

Dissected murine mammary glands were spread onhistological slides and fixed in 4% paraformaldehydefor 1 hr at 4°C. The tissues were then digested withtrypsin (50 mg/ml) at 37°C for 1 hr under agitation,incubated in 70% methanol containing 3% H2O2 for 30min to remove endogenous peroxidases, and subse-quently incubated overnight with 100 mg/ml of biotin-ylated Bandeiraea simplicifolia lectin B4 (Vector Lab-oratories, Burlingame, CA). Bound lectin was localizedwith an avidin-biotin-peroxidase amplification system(Vector), and mammary glands were mounted in Per-mount. Photographs were taken with a Zeiss Axiophotphotomicroscope. For quantitation of vascular density,five random images of whole-mount glands werescanned into a quantitation program (Image-Pro) usinga 3CCD Toshiba camera linked to a Dell computer. Thevessels were easily identified after staining by the im-aging program, and the number of pixels/mm2 wereconverted into vessels/mm2. The algorithms used forconversion were based on a V constant value estimatedafter scanning a single capillary vessel (10 mm in di-ameter by 300 mm in length).

Molecular Cloning of Rat VEGF-C, VEGFR-3,and CD31/PECAM-1 Partial cDNAs

Partially degenerate oligonucleotides were designed fromthe amino acid sequences TFFKPPC and NKELDED/E,SSQSSEE and QGRRRRP, or E/KAVYSVM and MSRPAA/

VP, which are conserved in the human (Joukov et al.,1996) and mouse (Kukk et al., 1996) VEGF-C cDNAs,the human (Pajusola et al., 1993; Lee et al., 1996) andmouse (Finnerty et al., 1993) VEGFR-3 cDNAs, or thehuman (Newman et al., 1990), mouse (Xie and Muller,1993) and bovine (Stewart et al., 1996) CD31 cDNAs,respectively. The primer sequences used were as fol-lows: VEGF-C, forward: 59-CGCGAATTCACCTTCTT-TAAACCTCC(A/G/C/T)TG(C/T) (containing an artifi-cial EcoRI site at the 59 end); reverse: 59-CGCGGA-TCC(A/G)TCTTCATCCA(A/G)CTC(C/T)TT(A/G)TT (con-taining an artificial BamHI site at the 59end); VEGFR-3,forward: 59-CGCGAATTCGGTACCAGCTCTCAGAG-CT(G/C)(A/C)GA(A/G)GA(A/G) (containing an artificialEcoRI site at the 59 end); reverse: 59-CGCGGATCCT-CTAGAGGGCCG(C/T)CGCCGCCT(A/G/C/T)CC(C/T)TG (containing an artificial BamHI site at the 59 end);CD31, forward: 59-CCGGAATTCGC(C/T)GT(G/C)TAC-TC(C/A)GT(G/C)ATG (containing an artificial EcoRIsite at the 59 end); reverse: 59-CGACGACGACGCGGC-CGC(A/G)C(G/T/A)GCTGGCCTGGACAT (containingan artificial NotI site at the 59 end). Total cellular RNA(2 mg) from adult rat lung was reverse transcribedusing random exanucleotides (Boehringer-MannheimBiochemicals, Indianapolis, IN) and Moloney murineleukemia virus reverse transcriptase (Promega, Madi-son, WI). One twentieth of the RT product was ampli-fied with degenerate VEGF-C, VEGFR-3, or CD31 oli-gonucleotides using PrymeZyme DNA Polymerase(Biometra). Polymerase chain reaction cycles were asfollows. For VEGF-C: 95°C, 2 min (13); 95°C, 1 min;52°C, 1 min; 72°C, 1 min (403); 72°C, 10 min (13); forVEGFR-3: 95°C, 2 min (13); 95°C, 1 min; 50°C, 1 min;72°C, 1 min (403); 72°C, 10 min (13); for CD31: 95°C,2 min (13); 95°C, 1 min; 52°C, 1 min; 72°C, 1.5 min(403); 72°C, 10 min (13). Approximately 430 bp, 400bp, and 1100 bp RT-PCR products were amplified withdegenerate oligonucleotides for VEGF-C, VEGFR-3, orCD31 respectively, digested at the extremities with theappropriate restriction enzymes, cloned into the corre-sponding sites of pBluescriptKS (Stratagene, La Jolla,CA) and sequenced on both strands.

Plasmid Construction

Rat VEGF: a partial 393-bp cDNA fragment contain-ing the common and alternatively spliced regions of therat 164aa isoform (nucleotides 242–632 in Conn et al.,1990), was provided by Dr. B. Berse (Beth Israel Hos-pital, Boston, MA).

Mouse VEGF: a cDNA containing the entire codingregion of murine VEGF (nucleotides -21 to 939 bp inClaffey et al., 1992) was provided by Dr. K. Claffey(Department of Physiology, University of Connecticut).

Rat VEGF-C: a partial 378-bp cDNA fragment wascloned as described above and in the Results section.

Rat VEGFR-1: a partial 600-bp cDNA (nucleotides800–1,400; Yamane et al., 1994), was provided by Dr.M. Shibuya (University of Tokyo, Japan).

521VEGF AND RECEPTORS IN MAMMARY GLAND

Rat VEGFR-2: non-overlapping partial 350-bp and430-bp cDNAs were subcloned from a 1.3-kb cDNAobtained from a Fisher rat placenta cDNA library (nu-cleotides 2962–3313 and 3837–4256 in Wen et al.,1998), and were provided by Dr. M. Shibuya (Univer-sity of Tokyo, Japan).

Mouse VEGFR-2: a partial 564-bp cDNA fragment(nucleotides 2345–2909 in Matthews et al., 1991) wasobtained by RT-PCR from mouse placenta.

Rat VEGFR-3: a partial 357-bp cDNA fragment wascloned as described above and in the Results section.

Rat CD31 (C9): a partial 260-bp KpnI/FspI cDNAfragment (nucleotides 322–583) was subcloned fromthe 1026 cDNA described above and in the Resultssection. Rat CD31 (39): the 1026 cDNA fragment de-scribed above and in the Results section was linearizedwith FspI; transcription from the T7 promoter pro-duced a 450-bp 39 cRNA probe.

Mouse CD31: a partial 1.6 kb cDNA fragment (nu-cleotides 639–2285 in Xie and Muller, 1993) cloned byRT-PCR from mouse placenta was provided by Dr. G.Wiedle (Basel Institute for Immunology, Switzerland).

Mouse keratin 8: a partial 670 bp cDNA fragment(nucleotides 12–682 in Baribault et al., 1993) was ob-tained by RT-PCR from mouse mammary gland.

RNA Extraction

Total cellular RNA was extracted from tissues andcells according to a modification of the acid guanidine-phenol-chloroform method of Chomczynsky and Sacchi(1987) as described by Chirgwin et al. (1979).

Ribonuclease Protection Assay

Prior to use in RNAse protection assays, RNA integ-rity was assessed by direct visualization of 18S and 28Sribosomal RNAs following electrophoresis in a 1.2%agarose gel, transfer to nylon membranes (Hybond,Amersham, Buckinghamshire, UK), and methyleneblue staining. RNAse protection assays were per-formed using an RPAII kit from Ambion Inc. (Austin,TX). Ten (10) mg total cellular RNA prepared from ratmammary glands as described above was hybridizedwith [32P]-labelled cRNA probes for 15 hr at 45°C.Hybridization, RNAse digestion, and detection of pro-tected fragments by polyacrylamide gel electrophoresiswere performed according to the manufacturer’s in-structions and previously published procedures (Pep-per et al., 1995). Dried gels were exposed to KodakXAR-5 films (Eastman Kodak, Rochester, NY) at280°C between intensifying screens. Autoradiogramswere quantitated using a densitometric scanner.

Northern Blotting and Hybridization

RNA was denatured with glyoxal, electrophoresed in1% agarose or formaldehyde-agarose gels, and trans-ferred to nylon membranes (Hybond, Amersham). Insome experiments, RNAs were crosslinked by exposureof filters to UV light (254 nm), and stained with meth-ylene blue to assess 18S and 28S ribosomal RNA integ-

rity. Filters were hybridized either with [32P]-labeledcRNAs as previously described (Pepper et al., 1990), orwith random primer [32P]-labeled cDNAs. Filters wereexposed to Kodak XAR-5 or MR film (Eastman Kodak)at room temperature or at 280°C between intensifyingscreens. Autoradiographic images were scanned andquantitated using a densitometric scanner or phospho-imager. To ensure linearity of signal, quantitationswere frequently performed at two different exposuretimes.

Immunohistochemistry

For VEGF, tissues were fixed in Bouin’s solution andembedded in paraffin. Six (6)-mm thick sections weredeparaffinized, transferred to 100% ethanol, andtreated for 10 minutes in methanol containing 0.5%H2O2. Sections were hydrated and exposed to 0.1%trypsin for 10 min at 37°C. Sections were preincubatedfor 30 min with 0.5% bovine serum albumin and thenincubated for 2 hr at room temperature with an affinitypurified rabbit polyclonal antibody prepared against asynthetic peptide corresponding to amino acids 1–20 ofthe amino terminus of human VEGF (cat no. sc-152,Santa Cruz Biotechnology, Santa Cruz, CA) at a 1/200dilution (0.5 mg/ml). A biotinylated anti-rabbit IgG wasthen applied for 30 min, followed by a streptavidin-peroxidase complex for a further 30 min. The presenceof VEGF immunoreactivity was revealed by diamino-benzidine tetrahydrochloride (DAB) (0.05 v/v) and0.01% H2O2 for 10 min. Sections were counterstainedwith hematoxylin, dehydrated in alcohol and mountedwith Eukitt. The following specificity controls wereused: omission of the primary antibody or incubationwith normal rabbit serum, incubation with biotinyl-ated anti-rabbit IgG or streptavidin-peroxidase com-plex alone followed by DAB treatment, and adsorptionof the specific anti-VEGF antibody with its homologousantigen at concentrations ranging from 0.05–1.50mg/ml for 12 hr prior to incubation.

For RECA-1, VEGFR-1, and type IV collagen, cryo-stat sections were fixed in cold acetone (RECA-1,VEGFR-1) for 4 min or 4% paraformaldehyde (collagenIV) for 10 min. RECA-1 (purchased from MEDACGmbH, Hamburg, Germany; Duijvestijn et al., 1992)was applied at a dilution of 1/30 followed by a goatanti-mouse fluorescent antibody at 1/600. ForVEGFR-1, an affinity-purified rabbit polyclonal anti-body raised against a synthetic peptide correspondingto amino acids 1312–1328 of the carboxy terminus ofhuman VEGFR-1 (cat. no. sc-316, Santa Cruz Biotech-nology, Santa Cruz, CA) was applied at a dilution of1/50 followed by a fluorescent goat anti-rabbit antibodyat 1/400. Collagen IV was detected by a sheep poly-clonal anti-human collagen IV antibody (provided byDrs. H. Kleinman and G. Martin, NIDR, Bethesda,MD) applied at a dilution of 1/400 followed by a rabbitanti-sheep fluorescent antibody at 1/400. All incuba-tions with primary antibodies were for 2 hr at roomtemperature; secondary antibodies were applied for 1

522 PEPPER ET AL.

hr. Sections were counterstained with 0.06% EvansBlue for 20 sec. The first antibody was omitted as acontrol of specificity.

In Situ Hybridization

Sections were prepared from rat mammary glandsfrozen as described above. Eight (8)-mm thick frozensections were cut using a Leica Jung CM 3000 cryo-tome and fixed in 4% paraformaldehyde for 5 min.Prehybridization, hybridization, and post-hybridiza-tion washes were as described (Sappino et al., 1989).Sections were hybridized using [3H]-labeled rat VEGFcRNAs prepared from the 393-bp VEGF cDNA frag-ment described above. Following hybridization and de-hydration in ethanol, sections were immersed in Kodakautoradiographic emulsion (Eastman Kodak) andstored at 4°C for 117 days. Slides were developed inKodak D-19 developper, fixed in Tetenal (Ott and WyssAG, Zofingen, Switzerland) and photographed underdark field illumination using a Zeiss Axiophot photomi-croscope.

Cell Culture

TAC-2 cells (Soriano et al., 1995), a clonally-derivedsubpopulation of the NMuMG mammary gland epithe-lial cell line (ATCC CRL 1636) (Owens et al., 1974),were cultured in collagen-coated tissue culture flasks(Falcon, Becton-Dickinson and Co., San Jose, CA) inhigh glucose DMEM (Gibco BRL, Life Technologies,Basel, Switzerland) supplemented with 10% fetal calfserum (Gibco). Total cellular RNA was prepared asdescribed above.

VEGF ELISA

Human milk was obtained from 5 lactating mothersduring the first week post-partum. VEGF levels weredetermined using ELISA kits from two differentsources. The first, a sandwich ELISA, was purchasedfrom R&D Systems Europe Ltd. (Abingdon, Oxford,UK). The second, a competitive ELISA, was purchasedfrom Cyt Immune Sciences Inc. (College Park, MD).ELISAs were performed according to instructions inthe enclosed manuals.

ACKNOWLEDGMENTS

We would like to thank Drs. Z. Werb, B. Vonderhaarand L. Orci for their advice, Dr. B. Berse for the ratVEGF cDNA, Dr. M. Shibuya for the rat VEGFR-1 andVEGFR-2 cDNAs, Dr. K. Claffey for the mouse VEGFcDNA and Dr. T. Nakamura for the recombinant hu-man HGF. We are grateful to M. Vallotton-Guisolanand M. Quayzin for technical assistance. Photographicwork was done by G. Negro and P.-A. Ruttiman. Thiswork was supported by grant no. 3100-043364.95 fromthe Swiss National Science Foundation (to RM andMSP), grants NIH-NCI-R29CA63356-01 and DODDAMD 17-945-J-4346 (to MLIA), and by a grant-in-aidfrom the “Fondation Sandoz pour l’avancement dessciences medico-biologiques” (to MSP).

REFERENCES

Alroy J, Goyal V, Skutelsky E. 1987. Lectin histochemistry of mam-malin endothelium. Histochem 86:603–607.

Baribault H, Price J, Miyai K, Oshima RG. 1993. Mid-gestationallethality in mice lacking keratin 8. Genes Dev 7:1191–1202.

Breier G, Albrecht U, Strerrer S, Risau W. 1992. Expression of vas-cular endothelial growth factor during embryonic angiogenesis andendothelial cell differentiation. Development 114:521–532.

Brown LF, Berse B, Jackman RW, Tognazzi K, Guidi AJ, Dvorak HF,Senger DR, Connolly JL, Schnitt SJ. 1995. Expression of vascularpermeability factor (vascular endothelial growth factor) and itsreceptors in breast cancer. Hum Pathol 26:86–91.

Chirgwin JM, Przybyla AE, MacDonald RJ, Rutter WJ. 1979. Isola-tion of biologically active ribonucleic acid from sources enriched inribonuclease. Biochemistry 18:5294–5299.

Chomczynski P, Sacchi N. 1987. Single-step method of RNA isolationby acid guanidinium thiocyanate-phenol-chloroform extraction.Anal Biochem 162:156–159.

Claffey KP, Wilkison WO, Spigelman BM. 1992. Vascular endothelialgrowth factor. Regulation by cell differentation and activated sec-ond messenger pathways. J Biol Chem 267:16317–16322.

Conn G, Bayne ML, Soderman DD, Kwok PW, Sullivan KA, PalisiTM, Hope DA, Thomas KA. 1990. Amino acid and cDNA sequencesof a vascular endothelial cell mitogen that is homologous to platelet-derived growth factor. Proc Natl Acad Sci USA 87:2628–2632.

Daniel CW, Silberstein GB. 1987. Postnatal development of the ro-dent mammary gland. In: Neville MC and Daniel CW, editors. Themammary gland. New York: Plenum Press. p 3–36.

Duijvestijn AM, van Goor H, Klatter F, Majoor GD, van Bussel E, vanBreda Vriesman PJC. 1992. Antibodies defining rat endothelialcells: RECA-1, a pan-endothelial cell-specific monoclonal antibody.Lab Invest 66:459–466.

Dvorak HF, Brown LF, Detmar M, Dvorak AM. 1995. Vascularpermeability factor/vascular endothelial growth factor, microvas-cular hyperpermeability, and angiogenesis. Am J Pathol 146:1029 –1039.

Feng D, Nagy JA, Pyne K, Hammel I, Dvorak HF, Dvorak AM. 1999.Pathways of macromolecular extravasation across microvascularendothelium in response to VPF/VEGF and other vasoactive medi-ators. Microcirculation 6:23–44.

Ferrara N. 1999. Role of vascular endothelial growth factor in theregulation of angiogenesis. Kidney Int 56:794–814.

Findlay JK. 1986. Angiogenesis in reproductive tissues. J Endocrinol111:357–366.

Finnerty H, Kelleher K, Morris GE, Bean K, Merberg DM, Kriz, R,Morris JC, Sookdeo H, Turner KJ, Wood CR. 1993. Molecular clon-ing of murine FLT and FLT4. Oncogene 8:2293–2298.

Gasparini G, Toi M, Gion M, Verderio P, Dittadi R, Hanatani M,Matsubara I, Vinante O, Bonoldi E, Boracchi P, Gatti C, Suzuki H,Tominaga T. 1997. Prognostic significance of vascular endothelialgrowth factor protein in node-negative breast carcinoma. J NatCancer Inst 89:139–147.

Gavin BJ, McMahon AP. 1992. Differential regulation of the Wnt genefamily during pregnancy and lactation suggests a role in postnataldevelopment of the mammary gland. Mol Cell Biol 12:2418–2423.

Greb RR, Maier I, Wallwiener D, Kiesel L. 1999. Vascular endothelialgrowth factor A (VEGF-A) mRNA expression levels decrease aftermenopause in normal breast tissue but not in breast cancer lesions.Br J Cancer 81:225–231.

Hlatky L, Tsionou C, Hahnfeldt P, Coleman CN. 1994. Mammaryfibroblasts may influence breast tumor angiogenesis via hypoxia-induced vascular endothelial growth factor up-regulation and pro-tein expression. Cancer Res 54:6083–6086.

Joukov V, Pajusola K, Kaipainen A, Chilov D, Lahtinen I, Kukk E,Saksela O, Kalkkinen N, Alitalo K. 1996. A novel vascular endo-thelial growth factor, VEGF-C, is a ligand for the Flt4 (VEGFR-3)and KDR (VEGFR-2) receptor tyrosine kinases. EMBO J 15:290–298.

Kukk E, Lymboussaki A, Taira S, Kaipainen A, Jeltsch M, Joukov V,Alitalo K. 1996. VEGF-C receptor binding and pattern of expression

523VEGF AND RECEPTORS IN MAMMARY GLAND

with VEGFR-3 suggests a role in lymphatic vascular development.Development 122:3829–3837.

Lee J, Gray A, Yuan J, Luoh SM, Avraham H, Wood WI. 1996.Vascular endothelial growth factor-related protein: a ligand andspecific activator of the tyrosine kinase receptor Flt4. Proc NatlAcad Sci USA 93:1988–1992.

Lichtenbeld HC, Barendsz-Janson AF, van Essen H, StruijkerBoudier H, Griffioen AW, Hillen HF. 1998. Angiogenic potential ofmalignant and non-malignant human breast tissues in an in vivoangiogenesis model. Int J Cancer 77:455–459.

Matsumoto K, Nakamura T. 1996. Emerging multipotent aspects ofhepatocyte growth factor. J Biochem 119:591–600.

Matsumoto M, Nishinakagawa H, Kurohmaru M, Hayashi Y, OtsukaJ. 1992. Pregnancy and lactation affect the microvasculature of themammary gland in mice. J Vet Med Sci 54:937–943.

Matsumoto M, Kurohmaru M, Hayashi Y, Nishinakagawa H, OtsukaJ. 1994. Permeability of mammary gland capillaries to ferritin inmice. J Vet Med Sci 56:65–70.

Matthews W, Jordan CT, Gavin, M, Jenkins NA, Copeland NG, Le-mischka IR 1991. A receptor tyrosine kinase cDNA isolated from apopulation of enriched primitive hematopoietic cells and exhibitingclose genetic linkage to c-kit. Proc Natl Acad Sci USA 88:9026–9030.

Nakahasi HL, Qasba PK. 1979. Quantitation of milk proteins andtheir mRNAs in rat mammary gland at various stages of gestationand lactation. J Biol Chem 254:6016–6025.

Nakamura J, Lu Q, Aberdeen G, Albrecht E, Brodie A. 1999. Theeffect of estrogen on aromatase and vascular endothelial growthfactor messenger ribonucleic acid in the normal nonhuman primatemammary gland. J Clin Endocrinol Metab 84:1432–1437.

Newman PJ, Berndt MC, Gorski J, White GC 2d, Lyman S, PaddockC, Muller WA. 1990. PECAM-1 (CD31) cloning and relation toadhesion molecules of the immunoglobulin gene superfamily. Sci-ence 247:1219–1222.

Obermair A, Kucera E, Mayerhofer K, Speiser P, Seifert M, Czer-wenka K, Kaider A, Leodolter S, Kainz C, Zeillinger R. 1997. Vas-cular endothelial growth factor (VEGF) in human breast cancer:correlation with disease-free survival. Int J Cancer 74:455–458.

Owens RB, Smith HS, Hackett AJ. 1974. Epithelial cell cultures fromnormal glandular tissue of mice. J Natl Cancer Inst 53:261–269.

Pajusola K, Aprelikova O, Armstrong E, Morris S, Alitalo K. 1993.Two human FLT4 receptor tyrosine kinase isoforms with distinctcarboxy terminal tails are produced by alternative processing ofprimary transcripts. Oncogene 8:2931–2937.

Pepper MS, Belin D, Montesano R, Orci L, Vassalli JD. 1990. Trans-forming growth factor-beta 1 modulates basic fibroblast growthfactor-induced proteolytic and angiogenic properties of endothelialcells in vitro. J Cell Biol 111:743–755.

Pepper MS, Soriano JV, Menoud PA, Sappino AP, Orci L, MontesanoR. 1995. Modulation of hepatocyte growth factor and c-met in therat mammary gland during pregnancy, lactation, and involution.Exp Cell Res 219:204–210.

Sappino AP, Huarte J, Belin D, Vassalli JD. 1989. Plasminogen acti-vators in tissue remodeling and invasion: mRNA localization inmouse ovaries and implanting embryos. J Cell Biol 109:2471–2479.

Siafakas CG, Anatolitou F, Fusunyan RD, Walker WA, Sanderson IR.1999. Vascular endothelial growth factor (VEGF) is present in hu-man breast milk and its receptor is present on intestinal epithelialcells. Pediatr Res 45:652–657.

Soemarwoto IN, Bern HA. 1958. The effect of hormones on the vas-cular pattern of the mouse mammary gland. Am J Anat 103:403–435.

Soriano JV, Pepper MS, Nakamura T, Orci L, Montesano R. 1995.Hepatocyte growth factor stimulates extensive development ofbranching duct-like structures by cloned mammary gland epithelialcells. J Cell Sci 108:413–430.

Stacker SA, Achen MG. 1999. The vascular endothelial growth factorfamily: signalling for vascular development. Growth Factors 17:1–11.

Stewart RJ, Kashxur TS, Marsden PA. 1996. Vascular endothelialplatelet endothelial adhesion molecule-1 (PECAM-1) expression isdecreased by TNF-alpha and IFN-gamma. Evidence for cytokine-induced destabilization of messenger ribonucleic acid transcripts inbovine endothelial cells. J Immunol 156:1221–1228.

Stirling JW, Chandler JA. 1976. The fine structure of the normal,resting terminal ductal-lobular unit of the female breast. VirchowsArch A Pathol Anat Histol 372:205–226.

Thomasset N, Lochter A, Sympson CJ, Lund LR, Williams DR, Be-hrendtsen O, Werb Z, Bissell, MJ 1998. Expression of autoactivatedstromelysin-1 in mammary glands of transgenic mice leads to areactive stroma during early development. Am J Pathol 153:457–467.

Turner CW, Gomez ET. 1933. The normal development of the mam-mary gland in the male and female albino mouse. Res Bull Mo AgricExper Sta 182.

Valtola R, Salven P, Heikkila P, Taipale J, Joensuu H, Rehn M,Pihlajaniemi T, Weich H, deWaal R, Alitalo K. 1999. VEGFR-3 andits ligand VEGF-C are associated with angiogenesis in breast can-cer. Am J Pathol 154:1381–1390.

Veikkola T, Alitalo K. 1999. VEGFs, receptors and angiogenesis. SemCancer Biol 9:211–220.

Wahl HM. 1915. Development of the blood vessels of the mammarygland in the rabbit. Am J Anat 18:515–524.

Weidner N, Semple JP, Welch WR, Folkman J. 1991. Tumor angio-genesis and metastasis—correlation in invasive breast carcinoma.N Engl J Med 324:1–8.

Wen Y, Edelman JL, Kang T, Zeng N, Sachs G. 1998. Two functionalforms of vascular endothelial growth factor receptor-2/Flk-1 mRNAare expressed in normal rat retina. J Biol Chem 273:2090–2097.

Xie Y, Muller WA. 1993. Molecular cloning and adhesive properties ofmurine platelet/endothelial cell adhesion molecule 1. Proc NatlAcad Sci USA 90:5569–5573.

Yamane A, Seetharam L, Yamaguchi S, Gotoh N, Takahashi T,Neufeld G, Shibuya M. 1994. A new communication system betweenhepatocytes and sinusoidal endothelial cells in liver through vascu-lar endothelial growth factor and Flt tyrosine kinase receptor fam-ily (Flt-1 and KDR/Flk-1). Oncogene 9:2683–2690.

Yang Y, Spitzer E, Meyer D, Sachs M, Niemann C, Hartmann G,Weidner KM, Birchmeier C, Birchmeier W. 1995. Sequential re-quirement of hepatocyte growth factor and neuregulin in the mor-phogenesis and differentiation of the mammary gland. J Cell Biol131:215–226.

Yasugi T, Kaido T, Uehara Y. 1989. Changes in density and architec-ture of microvessels of the rat mammary gland during pregnancyand lactation. Arch Histol Cytol 52:115–122.

524 PEPPER ET AL.