expression of vascular endothelial growth factor (vegf) and its two receptors (vegf-r1-flt1 and...

J. Pathol. 188: 369–377 (1999)

EXPRESSION OF VASCULAR ENDOTHELIAL GROWTHFACTOR (VEGF) AND ITS TWO RECEPTORS(VEGF-R1-Flt1 AND VEGF-R2-Flk1/KDR) IN

NON-SMALL CELL LUNG CARCINOMAS (NSCLCs):CORRELATION WITH ANGIOGENESIS AND SURVIVAL

1,2, 1,2, 1,2, 2,3, 1,2, 2,3 1,2*

1Laboratoire de Pathologie Cellulaire, Hôpital Albert Michallon, BP217, 38043 Grenoble Cedex 9, France2Lung Cancer Research Group (CJF INSERM 97-01), Institut Albert Bonniot, CHU, 38043 Grenoble Cedex 9, France

3Service de Pneumologie, Hôpital Albert Michallon, BP 217, 38043 Grenoble Cedex 9, France

SUMMARY

The formation of new vessels (angiogenesis) is essential for primary tumour growth and metastasis and is induced by severalangiogenic factors, including vascular endothelial growth factor (VEGF). The microvascular density (MVD) in tumours was assessedand the expression of VEGF and its receptors VEGF-R1-Flt1 and VEGF-R2-KDR/Flk1 was investigated in the different cellularcompartments in vivo, in order to establish their interrelationship and their prognostic influence. Immunohistochemical study of 69 stageI–II non-small cell lung carcinomas (NSCLCs) was performed on paraffin sections with CD34 antibody to estimate MVD, using aChalkley eye-piece graticule and VEGF, VEGF-R1, and VEGF-R2 antibodies. There was strong expression of VEGF and its receptorsin tumour cells, endothelial cells, and stromal fibroblasts. In tumour cells, the level of VEGF was correlated with that of VEGF-R1(p=0·018) but not that of VEGF-R2. In fibroblasts, high expression of VEGF was correlated with that of VEGF-R1 (p=0·0001) andVEGF-R2 (p=0·0001). In endothelial cells, expression of VEGF was correlated with that of VEGF-R1 (p<0·0001) and VEGF-R2(p=0·04). The level of VEGF in fibroblasts was correlated with that of VEGF-R1 (p=0·0028) and VEGF-R2 (p=0·01) in endothelialcells. There was no correlation between the level of MVD and that of VEGF or VEGF-R1 or VEGF-R2. Neither the level of MVD, northe level of expression of VEGF and VEGF receptors in any compartment influenced the patient’s survival. In conclusion, althoughangiogenesis is essential for tumour growth, this study failed to demonstrate that MVD, VEGF, VEGF-R1, and VEGF-R2 areprognostic markers for stage I–II NSCLC. VEGF, however, might act as a direct autocrine growth factor for tumour cells via VEGF-R1and angiogenesis could be promoted in a paracrine loop, where VEGF is produced by fibroblasts and tumour cells and then binds toendothelial cells via induced VEGF receptors. VEGF and its receptors thus appear as relevant therapeutic targets in NSCLC. Copyright? 1999 John Wiley & Sons, Ltd.

KEY WORDS—angiogenesis; vascular endothelial growth factor; vascular endothelial growth factor receptors; Flt1; Flk1; lung carcinoma;microvessel count

*Correspondence to: Professor Elisabeth Brambilla, Laboratoire dePathologie Cellulaire, BP 217, 38043 Grenoble Cedex 9, France.E-mail: [email protected]

Contract/grant sponsor: ARC (Association pour la Recherche sur leCancer).

INTRODUCTION

Lung carcinoma is the leading cause of cancer deathamongst men and has become second amongst womenin industrialized countries. Despite therapeutic efforts,fewer than 10 per cent of patients can be cured andenjoy long-term survival.1 Furthermore, TNM stagingremains the only reliable determinant of prognosis innon-small cell lung carcinomas (NSCLCs) and the mainfactor in the choice of curative treatment.2 The biologi-cal factors that determine a different individual outcome(recurrence, new primary malignancies, survival) withina similar stage of disease remain obscure, illustrating theneed to identify new prognostic factors that may helpclinicians optimize therapeutic efforts.

CCC 0022–3417/99/090369–09$17.50Copyright ? 1999 John Wiley & Sons, Ltd.

Angiogenesis is the formation of new blood vesselsfrom the pre-existing vascular bed. It is characteristicof several pathological conditions, including woundhealing, chronic inflammatory diseases, and tumourgrowth.3 After solid tumours reach a volume of1–2 mm3, their further outgrowth is strictly dependenton angiogenesis.4 An association between microvasculardensity (MVD), quantified by microvessel counting, andtumour metastasis or prognosis has been reported inseveral human tumours, including breast cancer,5cutaneous melanoma,6 and stomach cancer.7 Morerecently, lung cancers have been studied8–19 (Table I)with divergent results.

Angiogenesis is a complex multistep process involvingextracellular matrix remodelling, endothelial cellmigration and proliferation, capillary differentiation,and anastomosis.3 These processes are controlled byangiogenic factors which regulate one or more of thesekey events. Vascular endothelial growth factor (VEGF),also known as vascular permeability factor, is a multi-functional cytokine and a major factor implicated in

Received 19 October 1998Revised 5 March 1999

Accepted 22 March 1999

370 M. DECAUSSIN ET AL.

angiogenesis.20 VEGF is an endothelial cell-specificmitogen and an angiogenesis inducer in vivo.21 Tumourgrowth is dependent on the growth of local blood vesselsand inhibition of VEGF-induced angiogenesis sup-presses tumour growth in vivo.22 High expression ofVEGF has been correlated with relapse and stage pro-gression in superficial bladder cancer,23 and with MVDin cervical intraepithelial neoplasia24 and human coloniccancer.25 In lung cancers, a correlation between MVDand VEGF has been reported,26,27 but few reports coulddemonstrate a relationship with survival and this waslimited to squamous cell carcinoma28 or stage I lungadenocarcinoma.29 VEGF has two identified receptors,VEGF-R1-Flt1 and VEGF-R2-KDR/Flk1,21 expressedprincipally on vascular endothelium. Their expressionin vivo in NSCLC and their correlation with VEGF hasnot been studied.

The aim of the present study was to quantify MVD inorder to estimate its correlation with survival inNSCLC, and to investigate by immunohistochemistrythe expression of VEGF and its two receptors in thesetumours, in order to evaluate their function in thedifferent cellular compartments in vivo and theircorrelation with MVD and survival.

MATERIALS AND METHODS

Patient characteristics

The population studied consisted of 69 consecutivepatients with operable NSCLC (stages I and II) whounderwent curative surgery from 1989 to 1994 atGrenoble Hospital. Representative samples of freshtumour tissue were fixed in formalin for histological

Copyright ? 1999 John Wiley & Sons, Ltd.

study. Haematoxylin and eosin-stained sections wereused for the diagnosis of the histological type. Part ofthe fresh material was directly frozen in isopentane andsorted at "80)C. The tumours consisted of 69 NSCLCstages I and II, including 27 squamous carcinomas,36 adenocarcinomas, three large cell undifferentiatedcarcinomas, according to the 1981 World HealthOrganization classification, and three basaloidcarcinomas, according to Brambilla et al.30 (Table II).Eight lung specimens with minimal histological lesions,obtained from resection of subpleural emphysema, werealso examined and considered a control group.

Table I—Results of studies correlating microvascular density and survival in non-small cell lung carcinoma

Authors YearNo. ofpatients Stage

Immunohisto-chemistry Analysis p value

Pastorino et al.8 1997 515 I CD 31 U 0·77, NSChalkley

Macchiarini et al.9 1992 87 I FVIII M <0·0001Harpole et al.10 1996 275 I FVIII U 0·006

M 0·02Duarte et al.11 1998 106 I FVIII U 0·025

CD31 U and M NSApolinario et al.12 1997 116 I–IIIA CD31 U 0·37

25 II CD 31 U 0·02Giatromanolaki et al.13 1996 107 I–II CD31 U 0·0004

Chalkley M NSGiatromanolaki et al.14 1997 134 I–II FVIII U 0·01

CD31 U 0·004Fontanini et al.15 1996 253 I–III FVIII U 0·001

M 0·003Fontanini et al.16 1997 407 I–III CD 34 U <0·00001

M <0·00001Yuan et al.17 1995 55 I–III FVIII and CD31 U <0·001Angeletti et al.18 1996 96 IIIA FVIII U 0·0007

M 0·02Yamazaki et al.19 1994 42 I–IV FVIII U 0·027

M=multivariate analysis; U=univariate analysis.

Immunohistochemistry

Immunohistochemistry was performed on a represen-tative section of each tumour, avoiding areas withextensive necrosis or haemorrhage. The antibodies usedon paraffin sections in the study were as follows: CD34(1/1000, monoclonal, clone (QBEN10; Immunotech,Marseille, France), VEGF (1/400, polyclonal; SantaCruz Biotechnology, Santa Cruz, Ca, U.S.A.), VEGF-R1/Flt1 (1/100, polyclonal; Santa Cruz Biotechnology,Santa Cruz, CA, U.S.A.), VEGF-R2/Flk1 (1/100, mono-clonal; Santa Cruz Biotechnology, Santa Cruz, CA,U.S.A.). Immunohistochemistry was performed onserial sections of paraffin-embedded samples fixed withformalin. Endogenous peroxidases were suppressedusing 0·3 per cent of hydrogen peroxide in distilled waterfor 5 min at room temperature. In order to retrieve theVEGF-R1 and VEGF-R2 antigens, the deparaffinizedslides were placed in 10 m citrate buffer (pH 6) heatedin a microwave oven for 2#5 min, and then washedtwice with phosphate-buffered saline (PBS) for 5 min,

J. Pathol. 188: 369–377 (1999)

371EXPRESSION OF VEGF AND ITS TWO RECEPTORS IN NSCLC

prior to incubation with the primary antibody. Non-specific staining was inhibited with normal donkeyserum, 2 per cent, for 30 min before the primary anti-body and 10 min after it. After overnight incubation at4)C with the primary antibody, slides were washed inPBS, pH 8·6, and then exposed to the secondary anti-body, a biotinylated donkey F(ab*)2 anti-mouse anti-body (1/500; Jackson Laboratories, West Grove, PA,U.S.A.) or a biotinylated donkey F(ab*)2 anti-rabbitantibody (1/1000, Jackson Laboratories, West Grove,PA, U.S.A.) for 1 h at room temperature. The slideswere then washed in PBS, pH 8·6 and incubated with thestreptavidin–biotin–peroxidase complex (1/200; Dako,Glostrup, Denmark) for 1 h at room temperature. Thechromogenic substrate of peroxidase was a solution of0·05 per cent 3,3*-diaminobenzidine tetrahydrochloride/0·03 per cent H2O2/10 mmol/l imidazole in 0·05 mol/lTris buffer, pH 7·6. The slides were counterstained withHarris’ haematoxylin. Normal mouse IgG for mono-clonal antibody or normal rabbit IgG for polyclonalantibody, at the same concentration as the primaryantibody, served as negative controls.

In order to control the specificity of the polyclonalantibodies against VEGF and VEGF-R1, specific block-ing peptides were used. For neutralization, antibodiesagainst VEGF and VEGF-R1 were incubated for 2 h atroom temperature with a 10 times excess weight of thecorresponding peptide antigen (Santa Cruz Bio-technology, Santa Cruz, CA, U.S.A.), in PBS, pH 8·6.The slides were then incubated with this solution insteadof the primary antibody.

The results were recorded by two experiencedpathologists (MD and EB) independently assessing thepercentage of positive cells (1–100 per cent) and theintensity of staining (1+ =mild labelling; 2+ =moderatrelabelling; 3+ =intense labelling) in one section of eachpatient. Immunostaining scores were calculated by mul-tiplying the percentage of labelled cells by the intensityof staining.

Microvessel counting

The microvessel density (MVD), defined as capillariesand small venules, was evaluated using a 25-pointChalkley eye-piece graticule, as previously described.31

Briefly, MVD was measured on CD34 immunostainedparaffin sections. MVD was determined in the threeareas of maximal vascularization, where the highestnumber of microvessels was stained. These ‘hot spots’were chosen at low magnification (#100). MVD was


then estimated using the Chalkley graticule at #250magnification. The graticule was rotated in order tooverlay the maximum number of graticule dots withimmunohistochemically identified vessels. These dotswere considered positive and the Chalkley score wasthe sum of the positive dots in the three fields. Eachinvestigation was performed independently by twopathologists (MD and EB).

Statistical analysis

Correlations between the expression patterns ofgrowth factors, growth factor receptors, and MVD wereinvestigated with Fisher’s exact test. Actuarial survivalrates were calculated from the time of treatment until1 June 1998, using the Kaplan–Meier method.Comparison of survival was performed using the log-rank test. A p value less than 0·05 was consideredsignificant.

RESULTS

Microvessel counting

Taking a 10 per cent difference between counts as thethreshold value, the concordance between two readingson the same sections was 95 per cent. The mediannumber of vessels was 18 (range 11–39). MVD could beevaluated in 66 of the 69 cases (Fig. 1a), three tumourshaving diffuse and non-interpretable immunostaining.The level of MVD among the different histologicalsubtypes is shown in Table III. There was no significantdifference in expression between the different histologi-cal subtypes of NSCLC. The median was chosen as thecut-off point to separate two groups of patients with lowMVD (¦18) or high MVD (>18).

Table II—Distribution of histological subtypes and stages of the NSCLCs studied

Histology

Squamouscell

carcinomaAdeno-

carcinomaBasaloid

carcinomaLarge cellcarcinoma Total

Stage I 16 33 3 3 55Stage II 11 3 0 0 14Total 27 36 3 3 69

VEGF, VEGF-R1, and VEGF-R2 expression

In normal lung, VEGF, VEGF-R1, and VEGF-R2expression was limited to bronchial mucosa (score50–100), smooth muscle cells (score 200–300), hyper-plastic type II pneumonocytes (200–300), and alveolarmacrophages (200–300) (Fig. 1b). In contrast, neitherendothelial cells nor fibroblasts in interstitial lung tissuesexpressed these proteins with a score higher than 60. Thehigh and reproducible expression (score 300) of VEGFin smooth muscle cells was used as an internal positivecontrol.

J. Pathol. 188: 369–377 (1999)


Fig. 1—Angiogenesis in NSCLC. (a) Microvessel density evaluated by CD 34 immunostaining. (b) VEGF expression innormal lung. Immunostaining of bronchial mucosa and smooth muscle cells. Immunohistochemical staining of anadenocarcinoma (c, e, g) and a squamous cell carcinoma (d, f, h) with antibodies to VEGF (c, d) VEGF-R1 (e, f), andVEGF-R2 (g, h). Tumour cells, endothelial cells, and stromal fibroblasts are immunostained with the three antibodies

Copyright ? 1999 John Wiley & Sons, Ltd. J. Pathol. 188: 369–377 (1999)


In lung cancers, the three proteins were expressed inthe cytoplasm of tumour cells, endothelial cells, andstromal fibroblasts (Fig. 1c–h), with sometimes mem-branous accentuation of the immunostaining for recep-tors. VEGF-R1 immunostaining could be evaluated in67 out of 69 patients, two tumours being negative,including internal positive controls. The tumours wereconsidered positive with a score §10. VEGF wasexpressed in all tumours, and VEGF-R1 and VEGF-R2in 97 and 96 per cent, respectively. There was nosignificant difference in expression of VEGF, VEGF-R1or VEGF-R2 between the different histological subtypesof NSCLC (see Table IV). When the tumour sectionsincluded both the periphery of the tumour adjacent to


normal lung structures and the more central parts, theexpression of VEGF was obviously enhanced, as well asthat of its receptors in peripherally situated, as opposedto centrally located tumour cells. The same increasingintensity of VEGF and VEGF receptor expression wasobserved in the peripheral rim of large areas of necrosisand in small clusters of tumour cells disseminating in thetumour stroma.

When the antibodies against VEGF and VEGF-R1was neutralized by an excess of peptide antigen prior toincubation, no staining was observed.

Table III—Distribution of microvascular density (MVD)ranges according to histological type

MVD ¦18* MVD >18

10–14 15–18 19–22 23–39

Adenocarcinoma 2 13 7 5Squamous cell carcinoma 5 8 6 15Basaloid carcinoma 0 3 0 0Large cell carcinoma 1 1 0 0Total 8 25 13 20

*Median vascular density (MVD)=18 as assessed by Chalkleygraticule microvessel counting.

Table IV—Number of cases showing immunostaining for VEGF and its receptors in tumor cells (TC),endothelial cells (EC), and fibroblasts (F)

Histology

No. of cases immunostained

VEGF VEGF-R1 VEGF-R2

T F EC T F EC T F EC

Adenocarcinoma 36 34 33 35 25 30 34 25 31Squamous cell carcinoma 27 26 25 24 23 24 26 23 26Basaloid carcinoma 3 3 3 3 2 3 3 2 3Large cell carcinoma 3 3 3 3 2 2 3 2 3Total 69/69 66/69 64/69 65/67 52/67 59/67 66/69 52/69 63/69

100% 96% 93% 97% 77% 88% 96% 75% 91%

Table V—Overexpression of VEGF, VEGF-R1, and VEGF-R2 in NSCLC

Histology

No. of cases with more than the mean score (m)

VEGF>m VEGF-R1>m VEGF-R2>m

T F EC T F EC T F EC

Adenocarcinoma 21 10 17 22 8 10 21 8 12Squamous cell carcinoma 15 15 14 11 8 11 15 5 13Basaloid carcinoma 3 3 3 2 1 2 2 1 2Large cell carcinoma 1 1 1 1 1 1 1 1 1Total 40/67 29/67 35/67 36/67 18/67 24/67 39/67 15/67 28/67

60% 43% 52% 54% 27% 36% 58% 22% 42%

T=tumour cells; F=fibroblasts; EC=endothelial cells.

Correlation between VEGF and its receptors

The mean scores (m) of staining were used as acut-off (m=150 for VEGF and VEGF-R1; m=120 forVEGF-R2) to delineate two groups of patients. Wethus defined a group with high level of expression anda group with low level, for each protein. The resultsare shown in Table V. The majority of tumoursshowed high levels of expression of VEGF (60 percent) and of both VEGF-R1 (54 per cent) andVEGF-R2 (58 per cent) receptors. High levels ofexpression of VEGF, VEGF-R1, and VEGF-R2 wereobserved respectively in 43, 27, and 22 per cent ofcases in stromal fibroblasts. Endothelial cells expressedhigh levels of VEGF in 52 per cent of cases and highlevels of VEGF-R1 and VEGF-R2 in 36 and 42 percent tumours, respectively.

J. Pathol. 188: 369–377 (1999)


Table VI—Correlation between the levels of expression of VEGF and its receptors in each cellcompartment

VEGF-R1 T VEGF-R2 TLow High p value Low High p value

VEGF T Low 18 9 Low 13 14High 13 27 0·018 High 15 25 0·45

VEGF-R1 EC VEGF-R2 ECLow High p value Low High p value

VEGF EC Low 28 3 Low 22 9High 15 21 <0·0001 High 16 19 0·04

VEGF-R1 F VEGF-R2 FLow High p value Low High p value

VEGF F Low 38 2 Low 38 2High 11 16 0·0001 High 13 14 0·0001

T=tumour cells; EC=endothelial cells; F=fibroblasts.Low and high: number of cases with low (less than mean score) or high (more than mean score) levels of expression

of the respective marker in each cell type.p value: probability for a direct correlation between VEGF and VEGF-receptor level according to Fisher’s exact

test.

Table VII—Correlation between the levels of expression of VEGF and its receptors in different cellularcompartments

VEGF-R1 F VEGF-R2 FLow High p value Low High p value





VEGF F Low 33 12 Low 28 10High 7 15 0·0028 High 11 17 0·01

T=tumour cells; F=fibroblasts; EC=endothelial cells.Low and high: number of cases with low or high levels of expression of (less or more than the mean score, respectively)

for the indicated marker in each cellular compartment.p value: probability for a direct correlation according to Fisher’s exact test.

A high level of expression of VEGF on tumour cellswas correlated with that of VEGF-R1 (p=0·018), butnot of VEGF-R2. A high level of expression of VEGFin endothelial cells was correlated with a high levelof VEGF-R1 (p<0·0001) and VEGF-R2 (p=0·04). Ahigh level of expression of VEGF in fibroblasts was


correlated with a high level of VEGF-R1 (p=0·0001)and VEGF-R2 (p=0·0001) (see Table VI).

A high level of expression of VEGF in fibroblastswas also correlated with a high level of expression ofVEGF-R1 (p=0·0028) and VEGF-R2 (p=0·01) inendothelial cells (Table VII). However, the high level

J. Pathol. 188: 369–377 (1999)


of expression of VEGF in tumour cells was notcorrelated with a high level of expression of thereceptors in endothelial cells or in fibroblasts(Table VII).

Correlation of VEGF, VEGF-R1, and VEGF-R2expression with MVD

Levels of VEGF, VEGF-R1 or VEGF-R2 in thedifferent cellular compartments were not correlated witha high level of MVD, using Fisher’s exact test, with themean values as cut-off.

Correlation of MVD and VEGF expression with stagesand survival

Using the mean as a cut-off, we failed to demonstratea correlation between VEGF expression in any cellularcompartment, or MVD and tumour size (T1 versus T2),or metastatic nodal involvement (N0 versus N1). Noneof the angiogenic factors tested influenced patient sur-vival, as shown by log-rank analysis of Kaplan–Meiercurves in patients with high or low MVD and highor low levels of VEGF, VEGF-R1 or VEGF-R2expression.

DISCUSSION

Angiogenesis is essential for tumour growth and cellsurvival. The microvascular density (MVD) was used toevaluate the level of angiogenesis and was correlatedwith survival in a variety of tumours. In lung cancer, thisrelationship is not fully established (Table I). Fontaniniet al.16 studied 407 cases of stage I–III NSCLC andfound a high correlation between MVD and survival(p<0·0001), whereas Pastorino et al.,8 in a series of 515cases of stage I NSCLC, failed to establish any corre-lation between MVD and survival, in agreement withour study of 69 cases of stages I and II, with a majorityof stage I. MVD was correlated with survival when thestudies included different tumour stages, showing that itis more strongly correlated with TNM and stage thanpredictive of survival at any given stage. In contrast, ina series of 275 stage I NSCLC, MVD correlated withsurvival, but Factor VIII was used as the endothelial cellmarker.10 Furthermore, Factor VIII, although highlyspecific for the vasculature, is absent from part of thecapillary endothelium in tumour tissue, and an inter-national concensus on the methodology to quantifyangiogenesis has recommended CD31 or CD34 asspecific endothelial markers.32 CD34 was only discussedfor its lymphatic immunostaining, but neoplasms are notassumed to elicit the formation of a new lymphaticdrainage system, so the stroma of tumours contains nolymphatic vessels. The use of a Chalkley eyepiece grati-cule to count vessels in the present study could notexplain this lack of correlation, since it is an easy andobjective counting technique.32 We confirmed the repro-ducibility of this method, since we found the samemedian for MVD as Pastorino et al. in a larger series of515 cases (m=18). Furthermore, some studies reported a


correlation between MVD and survival with this grati-cule in stages I–II NSCLC14 and stage II had a signifi-cantly higher MVD than stage I.14 However, in this lastseries, MVD was not an independent factor of survivalwhen stage was included in the multivariate analysis.Thus, the independent prognostic significance of MVDat any given stage in lung cancer remains unclear.

Angiogenesis is dependent on many cytokines andgrowth factors. Vascular endothelial growth factor(VEGF) is a key factor implicated in angiogenesis.20,21 Afew studies have found a correlation between VEGFexpression in tumour cells and angiogenesis in a varietyof tumours (cervical intraepithelial neoplasia,24 humancolonic cancer25). In lung cancers, a correlation betweenVEGF and MVD was reported in 91 human epidermoidlung carcinomas with a limited power of significance(p=0·05)26 and in 44 stage I lung adenocarcinomas(p<0·05).29 A correlation between VEGF and over-expression of p53 protein on highly vascularizedtumours was also observed in 73 stage I–III NSCLCs(p=0·02),27 but not between VEGF and MVD. How-ever, our data did not demonstrate a correlationbetween high levels of expression of VEGF and highMVD in stage I and II NSCLC. Likewise, no correlationbetween VEGF and the hot-spot of MVD could beestablished in a study of 45 invasive breast cancers.33

Despite VEGF expression in tumour cells being pro-posed as a prognostic marker in a series of 44 stage Ilung adenocarcinomas29 and in a series of 91 squamouscell lung carcinomas,28 our study failed to demonstrate acorrelation between VEGF and survival, probably dueto the fact that we studied early stages of NSCLC, witha majority of stage I, where VEGF did not correlate withMVD. Other larger studies are necessary to confirm this.

Although previous studies in lung carcinoma havereported VEGF expression in tumour cells, its presencein endothelial cells and stromal fibroblasts, and theexpression of the VEGF receptors were not mentioned.Our data showed the expression of VEGF and its tworeceptors, not only in tumour cell but also in fibroblastsand endothelial cells. VEFG-R2-KDR/Flk1 is essentialfor the development of endothelial cells, whereas VEGF-R1-Flt1 is necessary for organization of the embryonicvasculature.34,35 In tumour cells, the levels of expressionof VEGF and VEGF-R1 was correlated, suggesting anautocrine function of VEGF on tumour cells via theVEGF-R1 receptor. VEGF may act as a growth factorfor tumour cells. A possible autocrine loop in tumourcells via VEGF-R1 and VEGF-R2 was also proposed ininvasive breast cancer, since high co-expression of thethree proteins was also demonstrated.36 The clearenhancement of VEGF and VEGFR expression at theperiphery of tumours and in disseminating tumour cellsis consistent with VEGF functioning in a growth factorautocrine loop contributing to tumour expansion.

The fact that VEGF did not correlate with the level ofangiogenesis and was not a significant prognostic factorin our study does not rule out a crucial role for VEGF asa key factor in tumour angiogenesis. Moreover, thestrong correlation that we observed between exogenousVEGF and its receptors on endothelial cells strengthensa potential role for VEGF and its receptors in

J. Pathol. 188: 369–377 (1999)


endothelial cell growth. In endothelial cells in vitro, thespecific localization of VEGF-R1 and VEGF-R2 hasbeen demonstrated20 and our study showed that highexpression of VEGF was correlated with that of the tworeceptors, suggesting a paracrine loop. Indeed, VEGFmRNA was not detected in blood vessels,20 indicatingan exogenous origin for VEGF. The expression ofVEGF receptors VEGF-R1 and VEGF-R2 was consid-erably increased in endothelial cells and stromal fibro-blasts of lung cancer, compared with their low level ofexpression in these cells in normal lung. Although acuteand chronic hypoxia increases gene expression ofVEGF-R1 and VEGF-R2 receptors and VEGF bind-ing to VEGF-R1 triggers receptor activation byautophosphorylation,20 it has not been demonstratedthat VEGF itself could trigger VEGF receptor geneexpression. In this context, chronic and acute hypoxia intumours may increase both VEGF and receptor geneexpression. Our finding that VEGF and VEGF receptorexpression increased at the periphery of necrosis isconsistent with this hypothesis. Alternatively, the hugeamount of VEGF present in the vicinity of stromalvessels may induce VEGF receptor expression.

Although we did not find a correlation betweenVEGF expression in tumour cells and VEGF-R1 andVEGF-R2 expression in endothelial cells, we found astrong correlation between the high level of expressionof VEGF in fibroblasts and the high level of expressionof its receptors in endothelial cells, suggesting a moreefficient paracrine loop between fibroblasts and endo-thelial cells than between tumour cells and endothelialcells. The neo-angiogenesis may therefore take place in aparacrine fashion, where VEGF may be produced in thestromal fibroblasts and then be bound to the endothelialcells to allow their growth via induced VEGF receptors.Indeed, no cells other than endothelial cells are thoughtto be able to proliferate in response for VEGF.37 Alter-natively, the simultaneous presence of ligand/receptorcombinations does not necessarily point to auto- orparacrine loops, since there may be no effect whenreceptor and ligand are not produced at the samelocation. A possible explanation for the lack of corre-lation between VEGF accumulation in tumour cells andthe expression of the receptors on endothelial cells couldbe that VEGF is secreted and secondarily deposited andstored in the extracellular matrix before its use byendothelial cells. However, we could not discern specificimmunostaining in the extracellular matrix around ves-sels, suggesting that the main source of VEGF forendothelial cell growth may be stromal fibroblasts ratherthan tumour cells. Although the presence of a protein isnot absolute proof of its functional role in tumourgrowth, in situ study by immunocytochemistry can helpto focus on those possible autocrine and paracrine loopsthat merit validation by further functional studies.

In conclusion, our study could not confirm thatmicrovascular density, VEGF or its receptors areprognostic markers of stage I–II NSCLC. VEGF may bea growth factor for several types of cells, includingendothelial cells and tumour cells. Our findings suggestan autocrine loop on tumour cells mediated by VEGF asgrowth factor, via VEGF-R1 receptor. Angiogenesis


could be induced in a paracrine fashion, in a scenariowhere VEGF is produced by stromal fibroblasts and bytumour cells and then binds to the endothelial cells,through induced VEGF receptors. These findingsemphasize the potential benefit to be expected fromcellular or genetic therapies targeting VEGF and itsreceptors,22–38 on both tumour cells and stromal cells,which may be more accessible by the vascular bed.

REFERENCES

1. Boring CC, Squires TS, Tong T. Cancer statistics. Chem. Abstr. 1994; 44:7–26.

2. Mountain CF. New international staging system for lung cancer. Chest1986; 89: 225s–233s.

3. Folkman J. Angiogenesis in cancer, vascular, rheumatoid and otherdiseases. Nature Med 1995; 1: 27–31.

4. Folkman J, Watson K, Ingber D, Hanahan D. Induction of angiogenesisduring the transition from hyperplasia to neoplasia. Nature 1989; 339:58–61.

5. Weidner N. Tumoural vascularity as a prognostic factor in cancer patients:the evidence continues to grow. J Pathol 1998; 184: 119–122.

6. Fallowfield ME, Cook MG. The vascularity of primary cutaneousmelanoma. J Pathol 1991; 164: 241–244.

7. Maeda K, Chung YS, Takatsuka S, et al. Tumor angiogenesis and tumorcell proliferation as prognostic indicators in gastric carcinoma. Br J Cancer1995; 72: 319–323.

8. Pastorino U, Andreola S, Tagliabue E, et al. Immunocytochemical markersin stage I lung cancer: relevance to prognosis. J Clin Oncol 1997; 15:2858–2865.

9. Macchiarini P, Fontanini G, Hardin MJ, Squartini F, Angeletti CA.Relation of neovascularisation to metastasis of non-small-cell lung cancer.Lancet 1992; 340: 145–146.

10. Harpole DH, Richards WG, Herndon JE II, Sugarbaker DJ. Angiogenesisand molecular biologic substaging in patients with stage I non-small celllung cancer. Ann Thorac Surg 1996; 61: 1470–1476.

11. Duarte IG, Bufkin BL, Pennington MF, et al. Angiogenesis as a predictor ofsurvival after surgical resection for stage I non-small-cell lung cancer.J Thorac Cardiovasc Surg 1998; 115: 652–658.

12. Apolinario RM, Van der Valk P, De Jong JS, et al. Prognostic value of theexpression of p53, bcl-2, and bax oncoproteins, and neovascularization inpatients with radically resected non-small-cell lung cancer. J Clin Oncol1997; 15: 2456–2466.

13. Giatromanolaki A, Koukourakis M, O’Byrne K, et al. Prognostic value ofangiogenesis in operable non-small cell lung cancer. J Pathol 1996; 179:80–88.

14. Giatromanolaki A, Koukourakis MI, Theodossiou D, et al. Comparativeevaluation of angiogenesis assessment with anti-Factor VIII and CD31immunostaining in non-small cell lung cancer. Clin Cancer Res 1997; 3:2485–2492.

15. Fontanini G, Vignati S, Bigini D, et al. Neoangiogenesis: a putative markerof malignancy in non-small-cell lung cancer (NSCLC) development. Int JCancer 1996; 67: 615–619.

16. Fontanini G, Lucchi M, Vignati S, et al. Angiogenesis as a prognsoticindicator of survival in non-small cell lung carcinoma: a prospective study.J Natl Cancer Inst 1997; 89: 881–886.

17. Yuan A, Yang PC, Yu CJ, et al. Tumor angiogenesis correlates withhistologic type and metastasis in non-small-cell lung cancer. Am J RespirCrit Care Med 1995; 152: 2157–2162.

18. Angeletti CA, Lucchi M, Fontanini G, et al. Prognostic significance oftumoral angiogenesis in completely resected late stage lung carcinoma (stageIIIA-N2). Cancer 1996; 78: 409–415.

19. Yamazaki K, Shosaku A, Takekawa H, et al. Tumor angiogenesis in humanlung adenocarcinoma. Cancer 1994; 74: 2245–2250.

20. Dvorak HF, Brown LF, Detmar M, Dvorak AM. Vascular permeabilityfactor/vascular endothelial growth, microvascular hyperpermeability, andangiogenesis: review. Am Med J 1995; 146: 1029–1039.

21. Ferrara N. Vascular endothelial growth factor. Trends Cardiovasc Med1993; 3: 244–250.

22. Kim KJ, Li B, Winer J, et al. Inhibition of vascular endothelial growthfactor-induced angiogenesis suppresses tumour growth in vivo. Nature 1993;362: 841–844.

23. Crew JP, O’Brien T, Bradburn M, et al. Vascular endothelial growth factoris a predictor of relapse and stage progression in superficial bladder cancer.Cancer Res 1997; 57: 5281–5285.

24. Dobbs SP, Hewett PW, Johnson IR, Carmichael J, Murray JC. Angio-genesis is associated with vascular endothelial growth factor expression inexpression in cervical intraepithelial neoplasia. Br J Cancer 1997; 76:1410–1415.

J. Pathol. 188: 369–377 (1999)


25. Takahashi Y, Kitadai Y, Bucana CD, Cleary KR, Lee ME. Expression ofvascular endothelial growth factor and its receptor, KDR, correlates withvascularity, metastasis, and proliferation of human colon cancer. CancerRes 1995; 55: 3964–3968.

26. Mattern J, Koomagi R, Volm M. Association of vascular endothelialgrowth factor expression with intratumoral microvessel density and tumourcell proliferation in human epidermoid lung carcinoma. Br J Cancer 1996;73: 931–934.

27. Fontanini G, Vignati S, Bigini D, et al. Neoangiogenesis and P53 proteinin lung cancer: their prognostic role and their relation with vascularendothelial growth factor (VEGF) expression. Br J Cancer 1997; 75:1295–1301.

28. Imoto H, Osaki T, Taga S, Ohgami A, Ichiyoshi Y, Yasumoto K. Vascularendothelial growth factor expression in non-small-cell lung carcinoma:prognostic significance in squamous cell carcinoma. J Thorac CardiovascSurg 1998; 115: 1007–1014.

29. Shibusa T, Shijubo N, Abe S. Tumor angiogenesis and vascular endothelialgrowth factor expression in stage I lung carcinoma. Clin Cancer Res 1998; 4:1483–1487.

30. Brambilla E, Moro D, Veale D, et al. Basaloid carcinomas of lung: a newphenotypic entity with prognostic significance. Hum Pathol 1992; 23:993–1003.

31. Fox SB, Leek RD, Weekes MP, et al. Quantification and prognostic valueof breast cancer angiogenesis: comparison of microvessel density, Chalkleycount, and computer image analysis. J Pathol 1995; 177: 275–283.


32. Vermeulen PB, Gasparini G, Fox SB, et al. Quantification of angiogenesisin solid human tumours: an international consensus on the methodologyand criteria of evaluation. Eur J Cancer 1996; 32A: 2474–2484.

33. De Jong JS, Van Diest PJ, Van der Valk P, Baak J. Expression of growthfactors, growth inhibiting factors, and their receptors in invasive breastcancer. II: Correlation with proliferation and angiogenesis. J Pathol 1998;184: 53–57.

34. Fong GH, Rossant J, Gertsenstein M, Beritman ML. Role of the Flt-1receptor tyrosine kinase in regulating the assembly of vascular endothelium.Nature 1995; 376: 66–70.

35. Shalaby F, Rossant J, Yamaguchi RP, et al. Failure of blood-islandformation and vasculogenesis in Flk-1 deficient mice. Nature 1995; 376:62–66.

36. De Jong JS, Van Diest PJ, Van der Valk P, Baak J. Expression of growthfactors, growth inhibiting factors, and their receptors in invasive breastcancer. I: An inventory in search of autocrine and paracrine loops. J Pathol1998; 184: 44–52.

37. Seetharam L, Gotoh N, Maru Y, Neufeld G, Yamaguchi S, Shibuya M. Aunique signal transduction from Flt tyrosine kinase, a receptor for vascularendothelial growth factor VEGF. Oncogene 1995; 10: 135–147.

38. Goldman CK, Kendall RL, Cabrere G, et al. Paracrine expression of anative soluble vascular endothelial growth factor receptor inhibits tumorgrowth, metastasis, and mortality rate. Proc Natl Acad Sci U S A 1998; 95:8795–8800.

J. Pathol. 188: 369–377 (1999)

expression of vascular endothelial growth factor (vegf) and its two receptors (vegf-r1-flt1 and...

Documents