Elevated expression of eIF4E and FGF-2 isoforms during ... - Nature

138 downloads 0 Views 808KB Size Report
Cherie-Ann Nathan1,2 , Peggy Carter 1 , Li Liu1, Benjamin DL Li3, Fleurette Abreo4, Ann Tudor5,. Stephen G Zimmer5 and Arrigo De Benedetti1. 1Departments ...
Oncogene (1997) 15, 1087 ± 1094  1997 Stockton Press All rights reserved 0950 ± 9232/97 $12.00

Elevated expression of eIF4E and FGF-2 isoforms during vascularization of breast carcinomas Cherie-Ann Nathan1,2 , Peggy Carter 1 , Li Liu1, Benjamin DL Li3, Fleurette Abreo4, Ann Tudor5, Stephen G Zimmer5 and Arrigo De Benedetti1 1 5

Departments of Biochemistry, 2Otolaryngology, 3Surgery and 4Pathology at LSU Medical Center, Shreveport, Louisanna; and Markey Cancer Center, University of Kentucky, Lexington, Kentucky, USA

The translation initiation factor eIF4E is a novel protooncogene found over expressed in most breast carcinomas (Kerekatte et al., 1995), but the pathology where this elevation is initially manifested and its possible role in cancer progression are unknown. We report that eIF4E is markedly increased in vascularized malignant ductules of invasive carcinomas, whereas necrotic and avascular ductal carcinomas in situ display signi®cantly lower levels. eIF4E facilitates the synthesis of FGF-2, a powerful tumor angiogenic factor. Conversely, reducing eIF4E with antisense RNA in MDA-435 cells suppresses their tumorigenic and angiogenic properties, consistent with loss of FGF-2 synthesis. These ®ndings suggest a causal role for eIF4E in tumor vascularization. Keywords: eIF4E oncogene; FGF-2 isoforms; tumor vascularization; translation initiation; breast cancer progression; immunohistology

Introduction The vascular system is a complex organ that has evolved to provide essential transport functions in all sizable metazoans. It provides for the oxygenation and nutrition of all tissues; the circulation of cells that mediate host defenses (immune system); and the transport of proteins and cells involved in wound repair, in¯ammatory responses, homeostasis, and thrombolysis. The vascular system is plastic: it responds throughout life to the need for neovascularization in response to a milieu of cell types and cytokines to prevent hemorrhage and restore patency of vessels (Edginton, 1995). The process of neovascularization is complex and essentially involves: (i) degradation of the vascular basement membrane and the ®brin interstitial matrix; (ii) migration of endothelial cells; (iii) rapid proliferation of endothelial cells; and (iv) formation of new capillary tubules and deposit of a new basement membrane (Folkman, 1986). The entire process must be remarkably fast to minimize blood and ¯uid loss and yet, stop as soon as the area in need is adequately vascularized. Rapid reponses to injury, in¯ammation, or tumor angiogenesis would be impaired were they to rely entirely on de novo gene expression. Instead, a mechanism that is based largely on altered translational eciency may have evolved,

Correspondence: A De Benedetti C-AN and PC contributed equally to this work Received 18 February 1997; accepted 13 May 1997

originally in endothelial/mesenchymal lineages, to respond quickly to the need for vascularization. Basic Fibroblast Growth Factor (FGF-2) is a powerful modulator of neovascularization. FGF-2 immunoreactivity is commonly found in invasive breast carcinomas (reviewed in Wellstein and Lippman, 1991), but ampli®cation at the gene or mRNA level is only found sporadically (Theillet et al., 1989; Champeme et al., 1994; Andappa et al., 1994) suggesting that the mechanism of upregulation is post-transcriptional. The main form of FGF-2 does not contain export signal sequences and is secreted poorly (Schwiegerer et al., 1987), although there are isoforms that are secreted (Souttou et al., 1996; Halaban et al., 1988; Morrison et al., 1990; Kevil et al., 1995). These isoforms are all generated from a complex pattern of alternate translation start sites in the mRNA 5'-leader (Prats et al., 1989). The FGF-2 mRNA of mammals contains a *500 nt long, G+Crich 5'-leader, that is strongly inhibitory for translation. Additionally, translation can start from several CUG codons upstream of the main, AUG-initiated ORF. The resulting amino-terminal extension(s) provide a mechanism that can rapidly alter the intracellular distribution and release of FGF-2, but the factors that govern the relative synthesis of the di€erent isoforms have not been elucidated. The translation factor eIF4E was recently identi®ed as a powerful oncogene/growth activator when overexpressed in model cell lines (Lazaris-Karatzas et al., 1990; De Benedetti and Rhoads, 1990). Furthermore, eIF4E is elevated 3 ± 30-fold in breast carcinomas, but not in ®broadenomas, indicating that its overexpression may mark a critical transition in the genesis of breast cancer (Kerakatte et al., 1995). We have proposed that eIF4E causes some of its e€ects by speci®cally increasing the translation of certain growth factors, since conditioned medium from cells transformed with eIF4E is strongly mitogenic (Kevil et al., 1995; Koromilas et al., 1992). This was con®rmed with the identi®cation of a large translational increase of two powerful cytokines and angiogenic inducers: FGF-2 and Vascular Permeability Factor (Kevil et al., 1995, 1996).

Results In vitro translation of FGF-2 mRNA in rabbit reticulocyte lysate (RRL) A direct e€ect on FGF-2 synthesis by eIF4E (capbinding protein) or by eIF4F (an ATP-dependent helicase composed of the initiation factors 4E, 4G,

Expression of eIF4E and FGF-2 in DCIS C-A Nathan et al

and 4A) can be demonstrated in vitro by translation of a capped rat (rFGF) transcript (Figure 1a and b). Both eIF4E and eIF4F strongly stimulated rFGF translation. This indicates that RRL contain suboptimal amounts of these factors for translation of FGF, but not for a standard mRNA like globin, that does not contain a long and structured 5'-leader. The level of eIF4E and eIF4F in RRL is comparable to normal ®broblasts, and we estimate that the highest concentration of added factors was tenfold the endogenous level. One di€erence was that eIF4F stimulated synthesis of

a

-mRNA

rFGF

CUG1 CUG2/3 AUG

1

2

3

4

5

elF4F

b

both the CUG1 isoform and the AUG-initiated product, which is believed to occur by internal initiation (Kevil et al., 1995; Vagner et al., 1995). In contrast, eIF4E stimulated only synthesis of the CUG1 isoform. Therefore, excess eIF4E alone results in a speci®c increase in `cap-dependent initiation' at CUG1, whereas addition of the pre-assembled helicase (eIF4F) also results in increased translation from the internally initiated (AUG) site. Elevated eIF4E and FGF-2 isoforms in ductal carcinomas Neovascularization is a critical event in tumor progression, based on long-standing observations that tumor growth beyond 1 mm requires adequate vascularization, or the tumor becomes necrotic (Hanahan and Folkman, 1996). Ductal Carcinoma In Situ (DCIS) is considered to be a transitional phase between pre-malignant lesions and Invasive Ductal Carcinomas (IDC): 30 ± 50% of women diagnosed with DCIS develop breast cancer if left untreated after the initial biopsy (Frykberg and Bland, 1994). This also presents a complex therapeutic problem for breast conservation, since it is dicult to estimate the malignant character of di€erent DCIS lesions (in Lippman, 1996), although the degree of vascularization may give an indication. We have analysed 127 breast lesions (most stereotaxic core biopsies, which represent a greater spectrum

rFGF a

DCIS CNT DCIS CNTCNT DCIS FAD FAD FAD IDC IDC IDC IDC

eIF4E

CUG1 CUG2/3 AUG

1

7

8

9

10

2

3

CNT7 IDC1

b

4

5

6

7

IDC2 DCIS2 IDC3

8

IDC8

9

10

11

12 13

FAD12 IDC4 DCIS4

11 elF4E

elF4E FGF-2

1088

CUG1 CUG2/3 AUG

c 1

Figure 1 In vitro transcription/translation of rFGF-2 in RRL was carried out as described in (Kevil, 1995). The translation products were separated on 15% SDS polyacrylamide gels and ¯uorographed. Transcripts were capped by inclusion of m7GpppG during the polymerization and used at a concentration of &25 mg/ ml. Reactions (30 ml) were supplemented with 0, 16, 32, and 48 ng of eIF4F (a, lanes 2 ± 5); or 0, 3, 10, 30, and 100 ng eIF4E (b, lanes 7 ± 11). A diagram of the construct for in vitro transcription is shown in (c)

2

3

4

5

6

7

8

9

Figure 2 Elevated eIF4E and FGF-2 isoforms in Invasive Ductal Carcinomas. Homogenates of breast biopsies were prepared as described in (Kerekatte, 1995). Proteins were separated on duplicate 15% polyacrylamide-SDS gels, and transferred to PVDF membranes. After addition of HRP-conjugated secondary antibodies, the blots were developed with DAB. (CNT=Control normal breast; FAD=Fibroadenoma; DCIS=Ductal Carcinoma In Situ; IDC=Invasive Ductal Carcinomas). (a) Representative benign (CNT, FAD) and malignant (DCIS, IDC) biopsies were analysed with rabbit serum against human eIF4E. (b) was probed anti-eIF4E and anti-FGF-2. The bands corresponding to eIF4E and the various FGF-2 isoforms are indicated. The sample labeled DCIS2 (lane 5) corresponds to (a, lane 6); and DCIS4 (lane 9) corresponds to (a, lane 1)

Expression of eIF4E and FGF-2 in DCIS C-A Nathan et al

of pre-malignant lesions) by Western blots and by immunohistology with antiserum to eIF4E. Thirteen samples were diagnosed as DCIS or mixed DICS/IDC, 62 were more advanced stages, and the remainder were non-malignant lesions. Several representative samples are shown in Figure 2: normal breast controls (CNT) and ®broadenomas (FAD) constitute standard controls for the basal level of eIF4E. The level of eIF4E was determined densitometricaly and typically normalized to histone H2B (not shown). All cancer-containing biopsies had 3 ± 30-fold elevated eIF4E, extending our previous analysis (Kereakatte et al., 1995). Notably, the average eIF4E elevation in IDC was 10.5+5.3, in contrast to 2.4+1.1 in DCIS, although we recognize that this analysis is limited to a relatively small number of DCIS samples. In parallel with the eIF4E elevation, the level of FGF-2 (particularly the CUG isoforms) was highly increased in IDC samples (Figure 2b). This is consistent with results obtained for FGF-2 translation in RRL. Thus, the elevation of eIF4E corresponds with that of FGF-2, and this is reportedly concomitant with proliferation of endothelial cells and formation of new blood vessels in a€ected lobules (Visscher et al., 1995). It is interesting to note that DCIS2 (Figure 2a, lane 6; Figure 2b, lane 4) was a mixed DCIS/IDC biopsy and showed a signi®cant elevation in eIF4E and FGF-2. In

contrast, FGF-2 was not increased in DCIS4, which showed no evidence of invasive cancer at pathologic examination. eIF4E immunohistology in DCIS and IDC In the absence of a large number of DCIS, we used immunostaining techniques in situ to study a panel of biopsies containing DCIS or mixed DCIS/IDC. A ®broadenoma (Figure 3, FAD) does not stain for eIF4E any more than normal tissue; note the distorted and elongated shape of the ducts characteristic of FAD. The double-layer of epithelial cells lining the ducts are morphologically distinct from stromal ®broblasts; these cells are cuboidal (well-differentiated) and tightly packed. This gives the impression that the eIF4E staining is stronger in these cells, but it is actually comparable to that seen in stromal ®broblasts. In contrast, the level of eIF4E increases visibly in DCIS (Figure 3, purple/blue round cells); note that the entire duct is enlarged and ®lled with several layers of large tumor cells. In DCIS, the cancer is con®ned within the duct, highly necrotic in the center, and avascular. In the adjacent IDC, the ductal stroma is eroded and the cancer cells stain much darker, indicating a further eIF4E elevation, and they are more spindly and disorganized. Here, the

IDC FAD

DCIS

IDC8

Figure 3 Immunohistology of eIF4E expression in breast biopsies. Three representative samples are shown. The DCIS and IDC were on the same histological section (patient DCIS2). The slides were reacted with eIF4E antiserum, then with HRP-conjugated secondary antibodies and developed with DAB/CoC12. Cancer cells stain blue/purple; erythrocytes (which do contain eIF4E) stain dark cyan in this system because of the constrast with hemoglobin, and blood vessels can be visualized via the patches of erythrocytes. A large blood vessel in IDC is indicated with an arrow. In DCIS8, the cancer has fully invaded the surrounding stromas and replaced the normal ductal structures. All frames were photographed at 206magni®cation and reproduced at approximately the same magni®cation

1089

Expression of eIF4E and FGF-2 in DCIS C-A Nathan et al

Reduction of eIF4E in MDA-435 cells with antisense RNA We have previously described an episomal vector engineered to express a transcript complementary to the translation start region of eIF4E mRNA (De Benedetti et al., 1991). The vector confers uniform expression in uncloned Geneticin-selected cells, and we found that a 50 ± 60% reduction in eIF4E is sucient to inhibit the tumorigenic properties of Ha-Rastransformed cells (Rinker-Schae€er et al., 1993; Gra€ et al., 1995). However, we did not know if this technology was applicable to a natural human carcinoma, since Ras and eIF4E lie on the same signal transduction pathway (Rinker-Schae€er et al., 1993). MDA-435 cells stably transformed with the antisense construct were obtained (435AS); a control line was also made with the same vector without antisense oligonucleotide (435BK). The expression of eIF4E in 435AS cells was reduced 3 ± 5-fold (Figure 4a; lane 5 vs. 3 or 4). This level was only slightly higher (1.5 ± 2-fold) than in normal breast cells (lanes 1 and 2). This eIF4E reduction had little obvious e€ect on tissue culture properties (Table 1) or on the general protein synthesis pattern (not shown). The 435AS cells grew at the same rate as 435 and 435BK, and their capacity to form colonies in soft agar was not signi®cantly altered (15% vs 20% eciency). This indicates that AS cells still display anchorage independence and no apparent inhibition of growth in tissue culture in serum-rich medium. However, their capacity to synthesize FGF-2 was suppressed (Figure 4b). In extracts from normal breast cells, FGF-2 was barely detectable and then only the 18 kDa, AUG-initiated product. In contrast, 435 and 435BK expressed signi®cant quantities of FGF-2 (Wellstein and Lippman, 1991), and particularly the CUG1 isoform. The 435AS cells displayed a substantial reduction of all FGF-2 isoforms and a nearly complete loss of the CUG1 isoform (Figure 4b, lane 5). The signi®cance of this loss is emphasized in the lower blot (Figure 4b, Medium), since this isoform is the main immunoreactive species that can be recovered from the culture medium after concentration and anity puri®cation on Heparin-Sepharose. Little, if any, FGF-2 isoforms could be recovered from the medium of 435AS cells or from normal breast cells (lanes 5, 1, 2). Furthermore, addition of conditioned medium (1 : 10 vol/vol) from 435AS cells was marginally mitogenic to a culture of serum-depleted HUVE cells, which are strongly dependent on FGF-2 for proliferation. This is in contrast to the powerful stimulation obtained with medium from 435 cells

435AS

435BK

435

JA

MCF10A

a

actin

eIF4E

1

2

3

4

5

b CELL EXTRACT

tumor is highly vascularized and not necrotic, suggesting that the elevation in eIF4E precedes the angiogenic phase. To establish the validity of this hypothesis we used the mammary carcinoma line MDA-435, which is tumorigenic in nude mice, to investigate the role of eIF4E in breast cancer progression. We recently reported that this, and several other human breast cancer lines, have a 5 ± 12-fold elevation in eIF4E expression compared to normal breast cells (Anthony et al., 1996), similar to the level found in malignant biopsies.

CUG1

MEDIUM

1090

CUG1 CUG2 CUG3 AUG

CUG2 CUG3 AUG

1

2

3

4

5

c

28S 4.2 kb

2.2 kb 18S

1

2

3

4

5

Figure 4 Synthesis of FGF-2 isoforms in breast cell lines correlates with eIF4E expression. (a) Western blot of eIF4E in normal cells (JA and MCF10A) and in MDA435 cells, MDA435 transfected with the control vector (435BK), or with the antisense RNA vector (435AS). (b) Western blot of FGF-2 in the corresponding cell lines. The top blot shows FGF isoforms in total cell extract. The lower blot shows FGF isolated from 20 ml of conditioned medium, after anity puri®cation on HeparinSepharose (Kevil et al., 1995). The FGF-2 immunoreactive bands are indicated with their presumed initiation codon. (c) Northern blot of FGF-2. The two major transcripts, di€ering in the length of 3'UTR, are indicated as 4.2 and 2.2 kb

(Table 1). The loss of FGF-2 expression in 435AS is clearly post-transcriptional, since FGF-2 transcripts were detected in each cell line (Figure 4c). Inhibition of tumorigenic and angiogenic capacity of 435AS cells The most prominent di€erence found with the 435AS cells was their reduced tumorigenic capacity (Table 1). The 435 and 435BK cells formed sizable tumors in all the injected animals in 26 days. In contrast, 435AS

Expression of eIF4E and FGF-2 in DCIS C-A Nathan et al

Table 1

1091

Growth in soft agar and tumorigenic properties of the di€erent cell lines

Cell line

MDA-435

435BK

435AS

1

30+5 9.8+0.3 2200+500 12

30+5 11.4+0.2 2400+500 9.4

30+5 8.3+0.1 1600+350 3.2

8 in 8 animals 26 ± 52 days 7 days

14 in 14 animals 25 ± 34 days 7 days

2 in 14 animals 49 ± 80 days 10 days

Doubling time in culture (h) Saturation densities (6104/cm2) 3 Soft agar (colonies per 104 cells) 4 Growth stimulation of HUVEC (fold-DNA synthesis over D-MEM) 5 Tumor incidence: 106 cells subcut 6 Tumor latency (&0.5 cm/dia) 7 Average tumor doubling 2

1

Logarithmically growing cells were plated at a density of 26103 cells per cm2 in 6-well plates (in duplicates). The medium (D-MEN with 15% FCS) was changed every 3 days, and each day sample wells were trypsinized and the cells were counted with a hemocytometer during a 10-days time course. 2 Saturation densities were determined by seeding 56104 cells in 60 mm dishes. Cells were harvested when they appeared stationary, and the reported numbers are the average of triplicate samples and range of cell counts. 3 Colonies composed of 550 cells were scored (RinkerSchae€er et al., 1993). 4 Primary HUVEC (second passage) were seeded in 12-well plates and grown EGM (Clonetics, San Diego, CA). The EGM was then replaced with D-MEN with 0.5% FCS. After overnight incubation, the medium was replaced with either control (100% D-MEN with 5% FCS), or conditioned medium (1/10 vol/vol). After 3 h, DNA synthesis was measured by incorporation of [[6 ± 3H]thymidine (15 Ci/ mmol; New England Nuclear, Boston, MA) added directly to the medium at 5 mCi/ml for. The cells were washed twice with PBS and lysed with 1 ml of 2% SDS. 10%-TCA-precipitable material was collected on GF/A ®lters (Millipore, Bedford, MA) and counted. 5 The animals were considered negative for tumor-formation after a 3.5 months incubation (end-point). 6 Tumor latency was determined as the time lapsed for the tumors to grow to 0.5 cm (largest diameter). Statistical di€erences could not be established, since only 2 animals injected with 435AS developed tumors. 7 Tumor volume doubling time (explained in Methods)

cells developed tumors in only two out of 14 animals, and only after a long latency period. We thus, sought to establish whether this di€erence was due to poor angiogenic induction by AS cells, consequent to the reduction in FGF-2 expression. 435BK and 435AS cells were seeded on polyurethane sponges and allowed to colonize the supports for several days in culture. Visual inspection with a stereoscope revealed that the cells grew well in the sponges and to similar densities. The sponges were then implanted in the mammary adipose layer of female nude mice to monitor the extent of vascularization induced by these cells. Each animal (nine per cell line) also received a control, unseeded sponge on the opposite ¯ank. The mice were sacri®ced sequentially, and the extent of vascular in®ltration was determined at necropsy. The sponges seeded with 435BK became deeply embedded in the chest wall and were visibly ®lled with capillaries in 7 ± 9 days (Figure 5a). The unseeded sponge on the opposite ¯ank showed little evidence of capillary in®ltration and was not ingrown at necropsy. The sponges seeded with 435AS remained essentially avascular (Figure 5b); even after 2 weeks the sponges were `white' despite clear evidence of being engulfed in ®brous tissue at that point (not shown). The sponges were subsequently embedded in paran and thin-sectioned, to study the network of capillaries by staining endothelial cells with factor VIII antiserum. By this method, capillaries could be visualized also in sponges seeded with 435AS (Figure 5d), but their density was similar to that of unseeded sponges (not shown). In contrast, 435BK showed a much greater network of capillaries, relative to unseeded sponges (Figure 5c). Particularly obvious at higher magnifications was also the large number of extravascular erythrocytes, indicative of extensive angiogenic remodeling. This was very reproducible in all animals implanted with 435BK sponges. The number of cancer cells (stained blue and more concentrated onto the folds of sponge) was not very di€erent for 435BK as compared to 435AS (Figure 5c and d), although several better-developed foci can be seen in (Figure 5c). This con®rmed that the large di€erence in vascularization was not due to di€erences

in seeding of the sponges before implantation. The histologic examination established a profound difference in the density of capillaries induced by 435BK vs 435AS, and con®rmed that 435AS are impaired in their capacity to elicit tumor angiogenesis. Discussion In this work we report on the role of eIF4E as an important marker for breast cancer. Elevated levels of eIF4E are present in most IDC, and this may be essential for cancer progression. We propose that upregulating the level of eIF4E is a pre-requisite for escaping the partially anoxic environment of con®ned DCIS and for vascularization and metastasis of the primary tumor. Cancer cells, expressing elevated eIF4E, can accomplish this through increased synthesis of the larger isoform(s) of FGF-2 that are exported and strongly induce proliferation and migration of capillaries. In this respect, it is also important to point out that reduction of eIF4E with antisense RNA inhibits the expression of the 92 kDa collagenase (MMP9) in Ras-transformed CREF cells (Gra€ et al., 1995). This metallo-protease is reportedly crucial for remodeling of the basement membranes during vascularization and metastasis of primary tumors (Su et al., 1993). The experiments with 435AS cells underscore the critical need for elevated eIF4E in the angiogenic phase and progression of breast carcinomas, since reducing eIF4E to near normal level was sucient to cripple their angiogenic and tumorigenic properties. No tissue culture property of 435AS cells was grossly altered, but their capacity to translate FGF-2 was impaired, and presumably that of a few other mRNAs with complex 5'-leaders. These mRNAs, perhaps 1% of all cellular mRNAs, require excess eIF4E/F for ecient translation and are subject to tight translational control (Koromilas et al., 1992; Kevil et al., 1995; Sonenberg, 1996). Cancer cells, which constitutively overexpress eIF4E, may exploit a normal pathway of auto/ paracrine stimulation that originally evolved in vascular endothelium in response to pathologic

Expression of eIF4E and FGF-2 in DCIS C-A Nathan et al

1092

emergencies (e.g. wound healing, anoxia/ischemia). In such situations, reversible phosphorylation events may lead to changes in eIF4F composition and release from a set of inhibitory proteins that can also act as tumor suppressors: 4E-BPs (reviewed by Sonenberg, 1996) or DAP5/NAT1 (Levy-Strumpf et al., 1997; Yamanaka et al., 1997). Using a sensitive ELISA method, the expression of FGF-2 was found to increase very rapidly in endothelia following angiogenic induction

or oxidative stress (Gabra et al., 1994). In addition, a recent investigation of a panel of cell lines showed that only transformed or stressed cells express the CUGinitiated forms of FGF-2 (Vagner et al., 1996). Since eIF4E is elevated in many transformed cells (Miyagi et al., 1995; Anthony et al., 1996) we speculate that eIF4E directly causes synthesis of the CUG-initiated FGF-2 isoforms in carcinomas. As such, eIF4E should be regarded as a general and important factor in tumor

RT-FLANK

LF-FLANK

Figure 5 Sponge vascularization assays. Sponges were implanted in female nude mice as described in text. Sponges seeded with 435BK cells (a) or with 435AS (b) were implanted in right ¯ank of the animals; control unseeded sponges were implanted in the left ¯ank. In this experiment, the mice were sacri®ced after 9 days, when the vascularization is fully complete for sponges seeded with 435BK cells. The sponges were excised and placed in a multiwell plate with bu€ered formalin. The sponge seeded with 435BK was so imbedded that it had to be excised in two sections: note the red color from the network of blood vessels. (c and d) Paran sections and staining of capillary vessels. Thin-sections from the sponges in above animals (right ¯ank) were stained with antiserum to factor VIII, followed by HRP-antirabbit IgG and DAB development. The capillaries are stained magenta color. The slides were lightly counterstained with hematoxilin (blue colour) to visualize the cancer cells (435BK and 435AS). Microscopy was at 56. A diagram conceptualized from the visible structures in paran sections is shown on the left

Expression of eIF4E and FGF-2 in DCIS C-A Nathan et al

vascularization, beyond the presumed requirement for genetic alterations at loci involving angiogenic factors and their receptors. Histochemistry of breast biopsies with eIF4E antiserum o€ers several advantages: (1) The level of eIF4E gives an indiction of the aggressiveness of the tumor and of its angiogenic potential. This could be helpful in the classi®cation of di€erent types of DCIS: the patient labeled as DCIS2 has recently recurred with bilateral breast cancer (Li et al., 1997); (2) Erythrocytes (and indirectly blood vessels) can be visualized as well as cancer cells in a€ected ducts, giving an independent parameter of the vascularization of the tumor; (3) eIF4E staining is superior to staining directly for FGF2, since the cancer (producing) cells are stained, whereas FGF-2 is deposited in stromal layers containing normal epithelia and tumor cells. FGF-2 is also unstable and distributed intra- and extracelluarly, often leading to con¯icting interpretations (Yiangou et al., 1997). Finally, while eIF4E is an important marker in breast cancer biopsies, it may not be a suitable circulating indicator for persistent disease. However, our research immediately suggests that the CUGisoforms of FGF-2 should be. Indeed, we found a report that large isoforms of FGF-2 accumulate in sera of breast cancer patients (Takei et al., 1993), although this awaits independent con®rmation. Materials and methods Cell cultures transformation and selection were carried out essentially as described in (Rinker-Schae€er et al., 1993). The BK virus-based episomal vector constructed to express ASRNA against eIF4E is described in (De Benedetti et al., 1991). Growth in soft agar 16104 cells were mixed with 2 ml of melted Noble Agar (428C; 0.4% ®nal conc.) in complete medium and poured onto 60 mm gridded dishes layered with 0.4% Agar. The Agar was overlaid with 0.3 ml of medium, which was changed every 3 days. The colonies were counted after 3 weeks, after staining with 1 ml of 3-[4,5-dimethylthiazol2yl]-2,5-diphenyltetrazolium bromide (MTT; Sigma). Tumorigenicity assays Tumorigenicity assays were performed by injecting 106 cells in 0.1 ml PBS subcutaneously into nu/nu mice (Charles

River). Tumors were scored as positive if they grew to 50.5 cm/dia. Tumor doubling times were determined by measuring daily the smallest (d) and the largest (D) diameter, and the volume was estimated using the formula: V=d26D60.4. Sponge vascularization assay This was adapted from Ascher et al. (1979). Cores of commercial polyurethane foam were cut-out with a corkborer (260.5 cm). The sponges were washed repeatedly with boiling DD-H2O, then with Acetone/Ethanol (50 : 50), and ®nally stored in sterile PBS. The sponges were placed individually in 6-well (non-stick) plates with 3 ml of DMEM+15% FCS. Much of the trapped air was squeezed out with a rubber policeman, before addition of 16105 cells. The sponges were typically ready to be implanted 1 week later. The mice were anesthetized and scrubbed with antiseptic solution. An incision (&0.5 cm) was made near the hind leg and a subcutaneous tunnel was created in the ¯ank with Metzenbaum scissors. The sponge was squeezed through the tunnel with a hemostat, and the incision was sealed with a staple. The sponges were retrieved post-mortem, taking utmost precautions to avoid bleeding (we used a cauterizing knife) and to preserve the vascular architecture inside the sponge. Northern and Western blots were carried out as described in Kevil et al., 1995). Immunohistochemistry of histological sections After dewaxing the slides were slowly rehydrated with progressively wetter alcohol, and ®nally dipped for 3 min in PBS with 0.5% Tween, before application of primary antiserum for 30 min. The polyclonal antiserum to human eIF4E (Kereakatte et al., 1995) was used at 1 : 200 dil. and the factor VIII antiserum (vWF; BioGenex) was used as speci®ed. After the addition of HRP-conjugated antirabbit antiserum (1 : 500 dil., Sigma) the slides were developed with a solution of DAB (1 mg/ml) in 50 mM Tris (pH 7.6); 0.3% H2O2 [120 mM CoCl2 was added for DAB/metal staining (Harlow and Lane, 1988)].

Acknowledgements We are indebted to R McLellan for providing us with stereotaxic breast core biopsies, to MJ Bruce for help with the paran sections, and to C Kevil and G McLoney for advice and assistance. This work was supported by NSF grant MCB9513756 and NIH grant CA69148-01A1 to ADB, and by the KY Medical Center Research Fund and the Markey Cancer Ctr./McDowel Foundation to SGZ.

References Anandappa SY, Winstanley JHR, Leinster S, Green B, Rudland PS and Barraclough R. (1994). Br. J. Cancer, 69, 772 ± 776. Anthony B, Carter P and De Benedetti A. (1996). Int. J. Cancer, 65, 858 ± 863. Ascher N, Ferguson R, Ho€man R and Simmons R. (1979). Transplantation, 27, 254 ± 260. Champeme MH, Bieche I, Hacene K and Lidereau R. (1994). Modern Pathology, 7, 900 ± 905. De Benedetti A and Rhoads RE. (1990). Proc. Natl. Acad. Sci. USA, 87, 8212 ± 8216. De Benedetti A, Joshi-Barve S, Rinker-Schae€er CW and Rhoads RE. (1991). Mol. Cell. Biol., 11, 5434 ± 5445. Edginton T. (1995). FASEB J., 9, 841 ± 842. Folkman J. (1986). Cancer Res., 46, 467 ± 473.

Frykberg ER and Bland KI. (1994). World J. Surgery, 18, 45 ± 57. Gabra N, Khayat A, Calabresi P, Khiat A. (1994). BBRC, 205, 1423 ± 1430. Gra€ JR, Boghaert ER, De Benedetti A, Tudor DM and Zimmer SG. (1995). Int. J. Cancer, 60, 255 ± 263. Halaban R, Kwon BS, Ghosh S, Delli Bovi P, Baird A. (1988). Oncogene Res., 3, 177 ± 186. Hanahan D and Folkman J. (1996). Cell, 86, 353 ± 364. Harlow E and Lane D. (1988). Antibodies: A laboratory manual, pp 403. CSH press. Kereakatte V, Smiley K, Hu B, Smith A, Gelder F and De Benedetti A. (1995). Int. J. Cancer, 64, 27 ± 31. Kevil C, Carter P, Hu B and De Benedetti A. (1995). Oncogene, 11, 2339 ± 2348.

1093

Expression of eIF4E and FGF-2 in DCIS C-A Nathan et al

1094

Kevil C, De Benedetti A, Payne KD, Coe LL, Laroux S and Alexander S. (1996). Int. J. Cancer, 65, 785 ± 790. Koromilas AE, Lazaris-Karatzas A and Sonenberg N. (1992). EMBO J., 11, 4153 ± 4158. Lazaris-Karatzas A, Montine KS and Sonenberg N. (1990). Nature, 345, 544 ± 547. Levy-Strumpf N, Deiss LP, Berissi H and Kimchi A. (1997). Mol. Cell. Biol., 17, 1615 ± 1625. Li BDL, Liu L, Dawson M and De Benedetti A. (1997). Cancer, (in press). Lippman M. (1996). 19th Annual San Antonio Breast Cancer Symposium Breast Cancer Research and Treatment, 41, 191 ± 217. Miyagi Y, Sugiyama A, Asai A, Okzaki T, Kuchino Y and Kerr S. (1995). Cancer Lett., 91, 247 ± 252. Morrison RS, Gross JL and Herblin WF. (1990). Cancer Res., 50, 2524 ± 2529. Prats H, Kaghad M, Prats AP, Klagsbrun M, Lelias JM, Liazun P, Chalon P, Tauber JP, Amalric F, Smith JA, and Caput D. (1989). Proc. Natl. Acad. Sci. USA, 86, 1836 ± 1840. Rinker-Schae€er CW, Gra€ JR, De Benedetti A, Zimmer SG and Rhoads RE. (1993). Int. J. Cancer, 55, 841 ± 847. Schweigerer L, Neufeld G, Friedman J, Abraham JA, Fiddes JC and Gospodarowicz D. (1987). Nature, 325, 259 ± 261.

Sonenberg N. (1996). In Translational Control (CSH press), pp 245 ± 269. Souttou B, Gamby C, Crepin M and Hamelin R. (1996). Int. J. Cancer, 68, 675 ± 681. Su Z, Austin V, Zimmer S and Fisher P. (1993). Oncogene, 8, 1211 ± 1219. Takei Y, Kurobe M, Uchida A and Hayashi K. (1993). J. Clin. Biochem. Nutr., 15, 57 ± 64. Theillet C, Le Roy X, De Lapeyriere O. (1989). Oncogene, 4, 915. Vagner S, Gensac MC, Maret A, Bayard F, Amalric F, Prats H and Prats AC. (1995). Mol. Cell. Biol., 15, 35 ± 44. Vagner S, Touriol C, Galy B, Audigier S, Gensac MC, Amalric F, Bayard F, Prats H and Prats AC. (1996). J. Cell Biology, 135, 1391 ± 1402. Visscher DW, Demattia F, Ottosen S, Sarkar FH and Crissman JD. (1995). Modern Path., 8, 665 ± 670. Wellstein A and Lippman ME. (1991). In Molecular Foundations of Oncology S. Broder (ed), pp. 403 ± 417. Yamamaka S, Poksay KS, Arnold KS and Innerartity TL. (1997). Genes & Dev., 11, 321 ± 333. Yiangou C, Gomm JJ, Coope RC, Law M, Luqmani Y, Shousha S, Coomnes RC, Johnston CL. (1997). Br. J. Cancer, 75, 28 ± 33.