The aE-catenin gene (CTNNA1) acts as an invasion ... - Nature

Oncogene (1999) 18, 905 ± 915 ã 1999 Stockton Press All rights reserved 0950 ± 9232/99 $12.00 http://www.stockton-press.co.uk/onc

The aE-catenin gene (CTNNA1) acts as an invasion-suppressor gene in human colon cancer cells Stefan J Vermeulen1, Friedel Nollet2, Erik Teugels3, Krist'l M Vennekens1, Fransiska Malfait1, Jan PhilippeÂ4, Frank Speleman5, Marc E Bracke1, Frans M van Roy2 and Marc M Mareel1 1

Laboratory of Experimental Cancerology, Department of Radiotherapy and Nuclear Medicine, University Hospital Gent, De Pintelaan 185, B-9000 Gent; 2Department of Molecular Biology, Laboratory of Molecular Cell Biology, University of Gent and VIB, Ledeganckstraat 35, B-9000 Gent; 3Laboratory of Medical Oncology, University Hospital Brussels, Laarbeeklaan 101, B-1010 Jette; 4Department of Clinical Chemistry, Microbiology and Immunology, University Hospital Gent, De Pintelaan 185, B-9000 Gent; 5Department of Medical Genetics, University Hospital Gent, De Pintelaan 185, B-9000 Gent, Belgium

The acquisition of invasiveness is a crucial step in the malignant progression of cancer. In cancers of the colon and of other organs the E-cadherin/catenin complex, which is implicated in homotypic cell-cell adhesion as well as in signal transduction, serves as a powerful inhibitor of invasion. We show here that one allele of the aE-catenin (CTNNA1) gene is mutated in the human colon cancer cell family HCT-8, which is identical to HCT-15, DLD-1 and HRT-18. Genetic instability, due to mutations in the HMSH6 (also called GTBP) mismatch repair gene, results in the spontaneous occurrence of invasive variants, all carrying either a mutation or exon skipping in the second aE-catenin allele. The aE-catenin gene is therefore, an invasionsuppressor gene in accordance with the two-hit model of Knudsen for tumour-suppressor genes. Keywords: colon cancer; invasion; aE-catenin; genetic instability

Introduction Colon cancer progression is characterized by histopathological changes starting from the normal mucosa over hyperplasia and sometimes polyp formation, towards carcinoma in situ and ®nally invasive cancer. This histopathological evolution is associated with the accumulation of a number of distinct genetic alterations (Kinzler and Vogelstein, 1996). Activation of several oncogenes and inactivation of several tumoursuppressor genes has been associated with growth disturbance preceeding invasive lesions. Fewer genetic alterations could be associated so far with the acquisition of the invasive and potentially metastatic phenotype, which is crucial for cancer malignancy and responsible for therapeutic failure (Mareel et al., 1991, 1996). Experimental as well as clinical observations have demonstrated that the E-cadherin/catenin complex is a powerful inhibitor of invasion (Behrens et al., 1989; Vleminckx et al., 1991; Takeichi, 1991; Bracke et al., 1996; Mareel et al., 1997). Disturbance of one of the elements of this complex, as evidenced by loss or

Correspondence: MM Mareel Received 16 March 1998; revised 5 August 1998; accepted 6 August 1998

truncation of the protein, has been associated with the transition from the noninvasive towards the invasive phenotype and with a worse prognosis in many human cancers. We have only started to understand how activation or inactivation of cadherin and catenin genes are implicated in tumour progression. The E-cadherin gene acts as an invasion-suppressor gene in lobular breast cancer since mutation of one allele in combination with loss of the other allele leads to invasive tumours (Berx et al., 1995). b-Catenin acts as an oncogene in colon cancer and in melanoma since heterozygous mutation results in protein stabilization and correlates with tumour progression (Morin et al., 1997; Rubinfeld et al., 1997). For aE-catenin the eect on tumour progression of heterozygous as compared to homozygous mutations has not been examined. There are indications from immunohistochemical studies that aE-catenin is reduced in 80% of invasive colon cancers (Shiozaki et al., 1994, 1995). In some colon cancer cell lines, de®ciencies at the mRNA and protein level were described for E-cadherin and aE-catenin but not for b-catenin (Vermeulen et al., 1996). Cultivating the human colon cancer cell line HCT-8, we found that invasive variants regularly emerged from noninvasive clones (Vermeulen et al., 1995a). Such invasive variants were recognized through their round (R) morphotype against a background of epithelioid (E) cells. Similar R variants were observed in the cell lines HCT-15, HRT-18 and DLD-1 which all have the same genetic background as HCT-8 (Vermeulen et al., 1998, in press). R variants were de®cient in aE-catenin protein and this explained their invasiveness in vitro. We have used this tumour model to address the question whether or not the aE-catenin gene (CTNNA1) meets the criteria of an invasionsuppressor gene in accordance with the two-hit hypothesis of Knudson (1985) for tumour-suppressor genes.

Results Regular transition from the epithelioid (E) to the round (R) morphotype HCT-8 subclones that were either positive or negative for aE-catenin protein could be selected by their epithelioid (E) or round (R) morphotype respectively (Vermeulen et al., 1995a). Using phase contrast microscopy we estimated that R variants reproducibly

CTNNA1: an invasion-suppressor gene SJ Vermeulen et al

906

emerged from E clones at a frequency of 1 R cell per 26105 E cells. Twenty-two E strains were cloned from the HCT-8 cell line (Figure 1). R variants were harvested from the conditioned medium (see Materials and methods) of 20 out of 22 (91%) of these E clones and from ®ve out of ®ve non cloned E cell populations (HCT-8, E11TD1 to E11TD4; see Figure 1). Conditioned media used for harvesting contained E cells as well as R cells. Out of 29 such heterogeneous cell cultures we obtained 19 morphotypically homogeneous R strains by 2 ± 4 rounds of harvesting. R variants were also obtained through cloning after seeding of E cells at low density (E11R1 to R4; E21R to E25R; see Figure 1). Acquisition of phenotypes typical of invasive cells In vitro assays showed that all E cells formed domes in con¯uent epithelioid monolayers on solid substrate and compact aggregates in collagen type I gels; most of them did not invade into embryonic chick heart in

organ culture (Table 1, Figure 2). By contrast, none of the R variants formed domes and they all grew in clusters that were easily released from the solid substrate. Furthermore, they produced loose clusters in collagen and they invaded into chick heart organ cultures. Suspension cultures of E cells produced compact multicellular spheroids that were resistant to pipetting, whereas R cells formed loosely organized multicellular clusters that were dissociated by pipetting (Figure 2). All these assays suggest that R cells have a de®cient E-cadherin/catenin complex. The E to R transition is not spontaneously reversible The E11R1 variant remained morphotypically stable for more than 120 passages. In E11R1 cultures that were maintained for more than 10 weeks by refreshing the culture medium without subcultivation, E colonies were never observed. Neither could the E morphotype be induced by treatment of E11R1 cultures with azacytidine nor with other dierentiating agents such as dimethylformamide, dimethylsulphoxide, insulin-like growth factor I, all-trans-retinoic acid, sodium butyrate or conditioned medium from E11 cells. Aberration of aE-catenin results in the R morphotype

Figure 1 Pedigree of HCT-8 showing the derivation of E (epithelioid) and R (round) clones and variants obtained by cloning (full lines) or harvesting from conditioned medium (dashed lines); stippled lines indicate derivation from a nude mouse tumour (TD); single box, E clones producing R variants; encircled, stable R variants or clones; double box, stable E clones

Coimmunoprecipitation of radioactively labelled proteins from E cell strains with antibodies against bcatenin and E-cadherin showed bands at 120, 102, 95 and 84 kDa representing E-cadherin, aE-, b- and gcatenin, respectively. In contrast, all R variants lacked the 102 kDa band corresponding to aE-catenin (Figure 3). Western blotting showed that the absence of the 102 kDa aE-catenin band in coimmunoprecipitates from R variants was due to the lack of expression of normal aE-catenin and not to a failure in the association of normal aE-catenin with the complex. Indeed, R variants showed either no detectable 102 kDa band (15 out of 20 variants) or bands with lower molecular weights ranging from 45 ± 93 kDa (5 out of 20 variants). By contrast, nine out of nine E cell strains showed the normal aE-catenin band (Figure 4). As a control, we stained Western blots of the E clones and R variants with an antibody against E-cadherin. A band of 120 kDa corresponding to E-cadherin was found in all E cell strains and R variants (Figure 4), con®rming immunocytochemical observations (data not shown). Northern blotting revealed that all E cell clones tested expressed aE-catenin mRNA as a 4.0 kb doublet but each of the R variants analysed showed deviations in the aE-catenin mRNA (Figure 5). Alterations observed in the R variants were: a doublet around 2.8 kb (E12R2), a singlet at 4.0 kb (E15R1), a smaller amount of aE-catenin mRNA (e.g. E11R7) or no detectable aE-catenin mRNA at all (e.g. E12R1). The aE-catenin-speci®c mRNA was not detected in the human PC-3 prostate cancer cells, was of smaller size (3 kb) in the human PC-9 lung cancer cells and was present as a doublet around 4.0 kb in human MCF-7/6 breast cancer cells. RT ± PCR reactions generated two overlapping fragments. These fragments were ampli®ed by primer sets s1-a3 and s2-a1 (see Materials and methods) that


Table 1

Invasion in vitro and expression of aE-catenin in epithelioid (E) and round (R) HCT-8 derived cells

Acronyma

Cell line Invasion in vitro n/nb

Proteinc

Northern blotting

0/25 n.d.

+ 85 kDa

n.d. 7

E11 E11R1

4/19 11/11

+ 7

+ 7

E11R5 E11R6 E11R7

8/12 6/10 5/5

7 7 7

E21 to E25 E24R E11TD1 E11TD1R E41 E11TD2 to E11TD4 E11TD2R E11TD3R E12 E12R1 E12R2 E13 E13R1 E13R2

n.d. n.d. n.d. 0/16 n.d. n.d. n.d. 0/11 3/10 8/12 0/11 7/10 4/10

E14 E14R1 E15 E15R1

0/12 7/7 0/6 6/6

HCT-8 E1R1

aE-catenin mRNAd RT ± PCR PTT

Sequencing

n.d. D 290 bp D 500 bp + +

n.d. n.d. n.d. + D COOH (2)

7 + +

+ D 1300 bp D 130 bp D 240 bp

D COOH n.d. n.d. n.d.

Cys767STOP n.d. D exons 2a ± 3 Cys767STOP Gln718STOP; Cys767STOP Cys767STOP D exons 2a-9 n.d. D exon 2b

87 kDa + 7 + + 7 45 kDa + 7 61 kDa + 7 93 kDa

n.d. n.d. n.d. + n.d. n.d. n.d. + + + (2.8 kb) + 7 +

+ 7 + 7

+ + + +

n.d. n.d. n.d. n.d. n.d. n.d. n.d. + + D 1075 bp + + D 170 bp D 310 bp D 430 bp D 520 bp + + + +

n.d. n.d. n.d. n.d. n.d. n.d. n.d. + COOH n.d. n.d COOH n.d. n.d. n.d. n.d. n.d. COOH n.d. COOH

n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. Cys767STOP D exons 2a ± 6 n.d. Cys767STOP n.d. D exon 5 D exons 5 ± 6 n.d. n.d. Cys767STOP n.d. Cys767STOP

D D

D D

a

E, epithelioid; R, round; TD, tumour derived (see Figure 1). bNumber of cultures showing invasion into PHF over total number examined; n.d., not done. cDetection of 102 kDa aE-catenin (+), truncated aE-catenin (relative molecular weights in kilodalton) or no detectable protein (7) by Western blotting as described in Materials and methods. dDetection (+), or no detection (7) of aE-catenin mRNA as demonstrated by Northern blotting, reverse transcriptase polymerase chain reaction (RT ± PCR) and in vitro transcription translation (protein truncation test, PTT). RT ± PCR: +, no deletion detected; D, deletion basepairs (bp) as estimated by agarose gel electrophoresis relative to a 100 bp ladder and RT ± PCR products from E cells. PTT: +, no deletion detected. D COOH, deletion in the carboxyterminal region; (2), 2 truncated proteins detected. Sequencing: deleted (D) exon(s) are indicated

cover the whole open reading frame of aE-catenin and, therefore, have a theoretical size of 2109 and 1667 bp respectively. In 5 out of 11 R variants, the generated fragments had unexpected sizes (Table 1 and Figure 6). Alterations found in these R variants were: single large deletions or multiple shortened bands. All deletions, except the 1300 bp one, in s1-a3 products (Figure 6b) were found also in s1-a2 products (Figure 6d), indicating their 5' terminal location. Sequencing of RT ± PCR products showed skipping of one to more exons (Figure 6g; Table 1). In six R variants, the RT ± PCR fragments showed an apparently normal size. In order to detect possible aE-catenin mutations that will not alter the size of RT ± PCR fragments (e.g. nonsense mutations or small insertions/deletions creating frameshifts) the protein truncation test (PTT) was applied. The two primer sets, designed to analyse the entire coding region of aEcatenin, yielded partly overlapping in vitro translated proteins (Figure 7) with a molecular weight of, respectively, 87 kDa (aminoterminal part) and 61 kDa (carboxyterminal part). In E clones, protein truncations were not demonstrated. However, for the carboxyterminal PTT product of the six R variants analysed, a prominent 49 kDa truncated protein was observed. Additionally, E11R1 cells showed a second band around 45 kDa (Table 1; Figure 7a. Sequence analysis

revealed the presence in E11R1 cells of two nonsense mutations, one in exon 14 and one in exon 16 (Figure 7). aE-catenin mutations in parental HCT-8 cells The two nonsense mutations observed in exons 14 and 16 of E11R1 cDNA were con®rmed on the genomic level as a heterozygous C to T transition in codon 718 (presumably the 45 kDa truncated PTT product) and a C to A transversion in codon 767 (presumably the 49 kDa truncated PTT product; Figure 7c and d). These two mutations were found on dierent mRNA transcripts by PCR cloning of E11R1 cDNA, indicating the inactivation of the two CTNNAI alleles (data not shown). The codon 767 mutation was found also in the parental E clone (HCT-8/E11) and in the other R variants, strongly indicating a preexisting mutation in one aE-catenin allele of HCT-8 cells (Table 1; Figure 7). These mutations were not detected in a genomic fragment cloned into the PAC cloning vector and containing part of the CTNNAI gene (Figure 7). E cell strains not producing R variants From the tumour-derived E11TD1 cell strain we obtained an E clone (E41; see Figure 1), that failed

907


908

of propidium iodide stained cells and this was also found for the E42 cell strain (data not shown). All other variants tested were DNA near diploid. HCT-8 is a genetically unstable cell line

Figure 2 Formation of multicellular spheroids and invasion in vitro by E and R variants of HCT-8. Photomicrographs of 4-day old multicellular spheroids from E11 cells (a and b) and from E11R1 cells (c and d) before (a and c) and after (b and d) pipetting. Photomicrographs of paran sections from embryonic chick heart fragments confronted in organ culture with noninvasive E15 cells (e and g) or with invasive E15R1 cells (f and h) and ®xed after 4 days; staining was with hematoxylin-eosin (e and f) and immunostaining with an antiserum against chick heart (g and h). Arrows indicate a peripheral chick ®broblast layer interrupted and underlayed by invading E15R1 cells (arrowhead). H, heart tissue; Scale bars=100 mm

to produce R variants during more than 12 weeks of harvesting the conditioned medium. This nonproducer E41 cell strain was subcultured for more then 28 passages and R variants were never observed. The nonproducer cell strain E41 was compared with its R producing parental cell strain E11. All in vitro assays suggested a functional E-cadherin/catenin complex in both cell strains. Immunocytochemistry, co-immunoprecipitation, Western blotting and Northern blotting did not show dierences between the nonproducer E41 and the producer E11 cell strain. However, karyotyping indicated that the E41 cell strain had a near tetraploid karyotype as compared to the near diploid karyotype of E11 or E11R1 (Figure 8). This karyotype was stable for at least 24 passages. Loss of the second sex chromosome and trisomy 13 were the only detectable chromosomal abnormalities observed in E11 and E11R1. In E41 cells, besides tetraploidization of the progenitor E11 cells, two extra copies of chromosome 7 were consistently found in addition to a telomeric association between two chromosomes 3[tas(3;3)(q29;p26)]. Tetraploidization of the nonproducer E41 cell strain was con®rmed by ¯ow cytometry

Mutation rates at HPRT and CTNNAI (aE-catenin gene) loci in E11 as compared to E41 cells were inferred from, respectively, the number of 6-thioguanine-resistant colonies and from the number of colonies with an R morphotype. The mutation rates for HPRT in E11, E41 and MRC-5 (control) were, respectively, 1.361075, 5.061077 and 52.661077. The mutation rates for CTNNAI in E11 and E41 were, respectively, 4.561076 and 59.461078. The results show significantly higher (P50.005, Student's t-test) mutational stability for both loci (30- and 50-fold, respectively) in E41 cells as compared to E11 cells. RT ± PCR reactions generating fragments that cover part of the open reading frames of the DNA repair enzymes HMSH6 and DNA polymerase d were performed. None of the generated fragments had unexpected sizes. Sequencing of the HMSH6 and DNA polymerase d speci®c RT ± PCR products, respectively from E13R2 and E12R2, indicated the presence of inactivating mutations in both genes. A 5 bp deletion/substitution at codon 1171 and a nonsense mutation at codon 290 were found in the HMSH6 mismatch repair gene while the DNA polymerase d was inactivated by a G to A transition at codon 506. Discussion R (round) cell variants regularly emerge by CTNNA1 mutation from E (epithelioid) clones of pseudodiploid human colon cancer cells that have one mutated CTNNA1 allele as well as defects in their DNA repair genes. This E to R transition marks the acquisition of the invasive phenotype. It is unlikely that the R variants represent a minor pre-existing subpopulation. R variants were harvested from various E cultures that underwent one or two rounds of previous cloning. We could not ®nd any arguments to accept that the growth of such putative pre-existing minority would not keep pace with the growth of the E cells. When cultured separately, R and E variants have about the same growth rate. When both types of cells were seeded together, R variants became obvious in the mixed culture like in the originally homogeneous E cultures (data not shown). This indicates that E cell-derived paracrine negative growth factors are not responsible for the repression of R variants. The presence of dierent types of aberration in R variants argues against a single prexisting subpopulation. Finally, calculation of mutation rates yielded consistent values of about 561076. Results of Northern blotting, RT ± PCR and PTT from all R variants investigated were suggestive for splice variations and truncating mutations in aEcatenin and this was con®rmed by sequencing genomic and RT ± PCR products. Despite extensive PCR-SSCP analysis of the aE-catenin gene in several round cell variants, we were unable to ®nd a gene mutation in one


of the consensus splice sites that could be responsible for the aberrant mRNA splicing. It is however clear that aE-catenin downregulation in the HCT-8 variants is due to defects at the genomic and/or mRNA level. To our knowledge, there are no other examples of regular mutational downmodulation of aE-catenin in short-term experimental systems. De®ciencies in aEcatenin, associated with loss of function of the Ecadherin/catenin complex, were found in other human cancer cell lines from prostate (PC-3; PPC-1; ALVA31) (Bussemakers et al., 1996), lung (PC-9), and colon (Clone A; MIP 101; CCL222) (Vermeulen et al., 1996). Their derivation from aE-catenin-positive parental cells was, however, not investigated.

The E to R transition was found with ®ve (HCT-8, DLD-1, HCT-15, HRT-18 and LoVo) out of a series of human colon cancer cell lines tested (Vermeulen et al., 1995a) (our unpublished results). In LoVo cells, the E to R transition was due to loss of E-cadherin protein with conservation of wild type aE-catenin (data not shown). However, ®ngerprinting showed that the cell lines (HCT-8, DLD-1, HCT-15 and HRT-18) in which the transition was aE-catenin-regulated, have the same genetic background (Vermeulen et al., 1998 in press). This strongly indicates that the four cell lines were derived from the same patient (Shibata et al., 1994; Chen et al., 1995). This unexpected result indicated a cell line- or tumour-speci®c predisposition for loss of

Figure 3 Coimmunoprecipitation of the E-cadherin/catenin complex from HCT-8 derived E and R cell strains. Fluorographs of SDS ± PAGE gels from lysates obtained by metabolic labelling with Tran35S followed by immunoprecipitation with antibody HECD-1 against E-cadherin (upper panel) and with antibody 1522 against b-catenin (lower panel). Molecular weight markers are indicated at the right in kilodalton (kDa). E-CAD, E-cadherin; a-CTN, aE-catenin; b-CTN, b-catenin; g-CTN, g-catenin (identical to plakoglobin); open arrowheads indicate bands found also in controls processed as for coimmunoprecipitation but without antibody (data not shown)

Figure 4 R variants do not express the normal aE-catenin protein. Western blotting was performed for total lysates from HCT-8derived E and R cell strains. Immunostaining was with anity puri®ed antibody against aE-catenin (no 1597; upper panel) or with antibody against E-cadherin (HECD-1; lower panel); E-CAD, E-cadherin; a-CTN, aE-catenin. Molecular weight markers are indicated in kilodalton (kDa); asterisks indicate bands with aberrant molecular weight

909


910

Figure 5 Abberrant aE-catenin transcription patterns in HCT-8 R variants. Analysis of aE-catenin (a-CTN)-speci®c mRNA by Northern blotting of 15 mg RNA from HCT-8-derived E and R cell strains. Data for PC-9 (truncated band at 3 kb), PC-3 (no bands) and MCF-7/6 (4 kb) are shown for reference to, respectively, aberrant and normal transcription patterns. Hybridization was done with a-CTN-speci®c probes (upper panel) and glyceraldehyde 3-phosphate dehydrogenase-speci®c probes (GAPDH; lower panel). Horizontal bars (right) indicate RNA molecular weight markers in kilobases (kb). Numerical values (bottom line) indicate relative amounts of a-catenin mRNA calculated as described in Materials and methods

aE-catenin. Sequence analysis showed that the two alleles of CTNNA1 were mutated in the E11R1 variant. The nonsense mutation in codon 767 was present not only in all R variants tested but also in HCT-8 derivatives of the E type and in the HCT-8 parental cell population. This is in line with heterozygosity for CTNNA1 mutation in this cell type. Whereas the founder CTNNA1 mutation may explain the preferential loss of aE-catenin from the E-cadherin/catenin complex during malignant progression of HCT-8 cells, the regularity and relatively high frequency of the E to R transition points to genetic instability of the HCT-8 cell line. Indeed, in the present HCT-8-derived E clones mutation was not restricted to CTNNA1, as evidenced by high mutation rates in the monoallelic HPRT gene in line with observations in HNPCC (hereditary nonpolyposis colorectal cancer) (Lynch and Smyrk, 1996). Remarkably, HNPCC tumours and cell lines derived therefrom are pseudodiploid in stead of aneuploid (Shibata et al., 1994; Kouri et al., 1990; Frei, 1992). They lack one or more DNA repair enzymes (Rhyu, 1996), explaining the multiplicity of mutations, as evidenced by microsatellite instability and by sequencing of genes such as APC, HPRT, TGFb1-RII, IGF,-IIR and p53 (Bhattacharyya et al., 1994, 1995; Lazar et al., 1994; Markowitz et al., 1995; Parsons et al., 1995; Souza et al., 1996). These mutations are considered to be relatively early events and they tend to persist during tumour development (Kinzler and Vogelstein, 1996; Rhyu, 1996). Data on the cancer of origin are not available to us but the following arguments indicate that the genetic instability of HCT-8 cells (with the same genetic origin as DLD-1, HCT-15 and HRT-8) is due to a DNA repair de®ciency. Microsatellite instability and a high mutation rate at the HPRT locus was reported for DLD-1 cells (Bhattacharyya et al., 1994, 1995) and ascribed to frameshift mutations in the HMSH6 mismatch repair gene on chromosome 2 (Papadopoulos et al., 1995) and to a heterozygous

mutation in the proofreading domain of the DNA polymerase d gene on chromosome 19 (da Costa et al., 1995). We found similar high mutation rates for HPRT as well as for CTNNA1 in HCT-8 cells and the same mutations in the DNA polymerase d and HMSH6. HCT-8, DLD-1 and HCT-15 cells have a near diploid karyotype like other cell lines with DNA repair de®ciency (Chen et al., 1995). The sequence, CATTATG, was de®ned as a hotspot for mutation in the HPRT gene (Eshleman et al., 1996) of DNA repair de®cient cells and was found also in exon 4 which was deleted in aE-catenin mRNA of some R variants. The interest of our observations for understanding the molecular mechanisms of colon cancer progression is that loss of aE-catenin is directly related to the acquisition of the invasive phenotype by noninvasive cancer cells. Thus the CTNNA1 gene is an invasionsuppressor gene in accordance with the two-hit model (Knudson, 1985), with the ®rst hit being a nonsense mutation in codon 767 without apparent phenotype, and with the second hit, (e.g. another nonsense mutation in codon 718 of E11R1) resulting in the acquisition of the invasive phenotype. The E-cadherin/ catenin complex is an invasion-suppressor and it has been demonstrated repeatedly that aE-catenin is an essential element of this complex (Bracke et al., 1996; Kemler, 1993). Moreover, other cell lines that lacked aE-catenin were found to be also invasive (Vermeulen et al., 1995a; Breen et al., 1993). Finally, invasiveness could be abolished by introduction of aE-catenin cDNA (Breen et al., 1993) or of chromosome 5 (Ewing et al., 1995), carrying the CTNNA1 gene (Nollet et al., 1995), into aE-catenin-negative human cancer cell lines. We isolated, by serendipity, from a tumour-derived cell line an E clone that did not produce R variants. So far, we have been unable to reproduce such isolation of an aE-catenin stable variant; ten other clones from the same E11TD1 cell line and three other tumour-derived cell lines (see Figure 1) all yielded R variants. Near


tetraploidy suggests that these clones have emerged by cell fusion or by endomitosis. Doubling of the number of functional alleles, one for the X-linked HPRT and one for CTNNA1 might, indeed, explain why the E41

cell strain has a lower mutation rate for both CTNNA1 and HPRT. It is less probable that doubling chromosome number would correct the DNA repair de®ciency through a gene dosage eect or complemen-

Figure 6 Analysis by RT ± PCR and sequencing indicates aberrant aE-catenin (a-CTN)-speci®c mRNA splicing from R cell strains. mRNA was transcribed by reverse transcriptase (RT). RT ± PCR products were obtained with primer sets speci®c for glyceraldehyde 3-phosphate dehydrogenase (GAPDH; a) and for aE-catenin (s1-a3, b; s2-a1, c; s1-a2, d; s2-a3, e). No RT ± PCR products were obtained with the GAPDH primer set on native RNA samples. Arrowheads indicate PCR products with normal size. Horizontal bars (right) indicate migration position of DNA molecular weight markers in basepairs (bp). Asterisks indicate bands with aberrant molecular weights. (f) shows the aE-catenin mRNA with the position of the exons. One boundary between exons 2a and 2b is also shown (Nollet et al., in preparation) in addition to the previously described boundaries (Oda et al., 1993). The relative position of sense (s1, s2 and s3), antisense (a1, a2, and a3) primers (closed arrowheads) and expression (s1exp and s2exp) primers (strings indicate arti®cial expression cassette added to the primer for use in PTT) are indicated as arrows. UTR, untranslated region. Comparison of some of the sequences of the truncated RT ± PCR products with known aE-catenin cDNA sequences showed preserved exons (numbered under the rectangles) and lost exons (dashed line) as depicted in (g)

911


912

tation, since both alleles of the DNA repair gene are expected to be mutated in the parental HCT-8 cells. Fusion of DNA repair de®cient DLD-1 cells with cells showing another DNA repair de®ciency led to the correction of both de®ciencies presumably by complementation (Casares et al., 1995). Mismatch repair was also improved by microcell transfer of chromosome 3 into the human HNPCC type HCT-116 cells (Koi et al., 1994). Microcell transfer of chromosome 2

(HMSH6) into the HCT-8 cells resulted in loss of the E to R transition in ®ve out of 10 E clones whereas removal of this chromosome by an `in out' selection system again induced the E to R transition (our unpublished results in collaboration with Boukamp, Jiricny and Cuthbert). We are currently investigating whether this loss of E to R transition by chromosome 2 is indeed by correction of DNA repair de®ciency in these HCT-8 cells. The correction of the E to R

Figure 7 The protein truncation test (PTT) and sequencing localizes nonsense mutations. Autoradiographs of 35S-methioninelabelled proteins from in vitro transcribed and translated aE-catenin RT ± PCR products were prepared from E11 and E11R1 cells (a). a-CTN-NH2, amino terminal part of aE-catenin (RT ± PCR products prepared with s1exp-a3 primer set); a-CTN-COOH, carboxy terminal part of aE-catenin (PCR products prepared with s2exp-a1 primer set). Molecular weight markers (right) are indicated in kilodalton (kDa) and in vitro translated proteins (left) are indicated by arrowheads. Asterisk indicates major band with aberrant molecular weight; Control, PTT without RT ± PCR product. cDNA (b) and genomic sequences (c) indicate two heterozygous nonsense mutations (arrows) in the E11R1 clone of which one mutation was already present in the E11 clone


Figure 8 Representative G-banded karyotypes of cell strains E11 (upper panel), E11R1 (middle panel) and E41 (lower panel)

transition may reveal the importance of DNA-repair in the progression towards invasive colon cancer. Materials and methods Cells Cell lines were derived through cloning or harvesting from the human HCT-8 colon cancer cell line and selected for their epithelioid (E, previously described as S, forming Smooth edged epithelioid colonies) (Vermeulen et al., 1995a) or round (R) morphotype, as evident under phase contrast microscopy (Table 1). The HCT-8 cell line was originally derived from a male patient with an adenocarcinoma of the ileocecal region (Tompkins et al., 1974). Cultures were maintained in RPMI-1640 (Gibco ± BRL, Gent, Belgium), supplemented with 10% heat-inactivated foetal bovine serum, 1 mM sodium pyruvate, 100 mg/ml streptomycin, 250 IU/ml penicillin and 2.5 mg/ml fungizone (hereafter called culture medium). LoVo was cultured as described previously (Vermeulen et al., 1995a). All cell lines used and generated in the present study were mycoplasma-free (DAPI-staining, Sigma, St Louis, MO, USA). Cloning with glass cylinders was done from low density cultures (Vermeulen et al., 1995a). The clones E21 to E25 resulted from two consecutive rounds of cloning. Brie¯y, 500 E11 cells in 15 ml culture medium were seeded on 75 cm2 tissue culture substrate. After 3 weeks, colonies were selected for the E morphotype and isolated with cloning cylinders. Isolated cells were again seeded at clonal density and a second round of cloning was done with selection for the E morphotype.

Tumour-derived (TD) cell strains were isolated from subcutaneous nude mouse tumours in accordance with standard procedures. 56106 E11 cells were injected s.c. in 6-week old nu/nu mice (Ia Credo, Brussels, Belgium). After 8 weeks, tumours were dissected out, trimmed, fragmented and washed at 48C in Ringer's followed by three washes for 15 min with culture medium containing 1 mg per ml streptomycin and 2500 IU/ml penicillin. Fragments were then minced, incubated in 0.25% trypsin for 2 h at 378C, resuspended in culture medium and seeded on 25 cm2 tissue culture substrate. E11TD1, E11TD2, E11TD3 and E11TD4 cell lines were each obtained from a dierent mouse. To harvest R strains from conditioned medium, E cells were maintained as con¯uent cultures on 25 cm2 tissue culture substrate with biweekly culture medium refreshment but without passaging. Conditioned media (2.5 ml), containing shed cells, were harvested weekly, admixed with 2.5 ml fresh culture medium and seeded onto a new 25 cm2 tissue culture substrate. Such ®rst harvests yielded morphotypically heterogeneous cultures, that were used for one to three more rounds of harvesting to obtain homogeneous R variants. Human lung PC-9 (Shimoyama et al., 1992; Hirano et al., 1992) and prostate PC-3 cancer cells (Morton et al., 1993) were included in the present experiments because they carry a homozygous deletion in the aE-catenin gene, resulting in truncated mRNA or no mRNA, and absence of aE-catenin protein. Human mammary carcinoma cells MCF-7/6 (Bracke et al., 1991) were used as a positive control for aE-catenin expressing cells. In mutation rate analyses, human ®broblasts MRC-5 served as controls with low mutation rate (Bhattacharyya et al., 1994). The human origin of the HCT-8 cell strains was veri®ed by lactate dehydrogenase (LDH) zymography (Baxter diagnostics Inc., Irvine, CA, USA). E11, E11R1 and E41 produced all ®ve human LDH isoenzymes (data not shown). VNTR (variable number of tandem repeats) ®ngerprinting with the four microsatellites, CD4, THO1, D21S11 and SE33 demonstrated that E11, E11R1 and E41 were parentally related (our unpublished results). Functional evaluation of the E-cadherin/catenin complex The function of the E-cadherin/catenin complex was investigated in dierent assays used previously to characterize R and E subclones (Vermeulen et al., 1995a): (i) Calcium-dependent and E-cadherin-speci®c fast (30 min) aggregation, multicellular spheroid formation, cell dissociation and colony formation in collagen type I, to evaluate the cell-cell adhesion and compaction function; (ii) Morphotype and dome or cyst formation, respectively on solid substrate and in confronting organ culture, to evaluate the role in epithelial organization; (iii) Confronting organ cultures with embryonic chick heart fragments and seeding on collagen type I, for the evaluation of the invasion-suppressor function. For the dissociation assay, cells were detached, suspended in culture medium and kept under gyrotory shaking (Gyrotory shaker; New Brunswick Scienti®c, New Brunswick, NJ, USA) at 80 r.p.m. for 3 days. Multicellular spheroids were evaluated microscopically before and after they were passed 30 times through Pasteur pipettes with an inner diameter of 1.5 mm. RNA isolation and Northern blotting Total RNA was isolated from cell cultures with RNAzolTMB (Wak Chemie Medical GMBH, Bad Hamburg, Germany). Fifteen mg total RNA was glyoxylated, size fractionated by electrophoresis through 1% agarose gels, followed by transfer to a HybondTM-N+ membrane (Amersham, Buckinghamshire, UK) and consecutive hybridization with probe 2 covering bp 716 to 2857 of aE-catenin mRNA (Nollet et al., 1995) and with a probe

913


914

against GAPDH. Probes were labelled with a-32P-dCTP (Amersham) using the RadPrime DNA system (Gibco BRL). Hybridization signals were revealed by ¯uorography and the integrated optical densities (iod) were measured using a PhosphorimagerTM (Molecular Dynamics, Sunnyvale, CA, USA). Relative amounts of aE-catenin (a-CTN) in R and E cultures were normalized on the basis of the GAPDH hybridization signals and calculated using the following formula: e.g: relative amount E11R1= [iod (a-CTN)E11R1 6 iod(GAPDH)E11]/[iod(a - CTN)E11 6 iod (GAPDH)E11R1]. RT ± PCR The GeneAmp1 RNA PCR kit from Perkin Elmer (Branchburg, NJ, USA) was used to generate cDNA and PCR products. To avoid false positive signals from the aEcatenin pseudogene (Nollet et al., 1995), RNA samples were DNAse treated (Promega, Madison, WI, USA). To evaluate DNA contamination in RNA, GAPDH primers known to amplify both GAPDH cDNA and GAPDH pseudogene DNA (Clontech, Palo Alto, CA, USA) were used as controls under conditions described by the manufacturer. Sense (s) and antisense (a) primers for ampli®cation of aE-catenin cDNA (s1, s2, s3, a1, a2 and a3) were previously published by Oda et al., 1993 and purchased from Gibco. Previously described primers for the ampli®cation of DNA polymerase d (da Costa et al., 1995) and HMSH6 (Papadopoulos et al., 1995) were used. DNA-free RNA was reverse transcribed and ampli®ed with AmpliTaq GoldTM (Perkin Elmer) in a Programmable Thermal Controller-100 (MJ Research, Watertown, MA, USA). Brie¯y, PCR conditions consisted of primer sets s1a3, s1-a2, s2-a1 and s2-a3, activation of AmpliTaq GoldTM at 958C for 12 min followed by 35 cycles of annealing, extension and denaturation for 45 s at 628C, 2 min 20 s at 748C and 45 s at 958C, respectively. cDNA fragments were cloned using the AdvantageTM PCR cloning kit (Clontech). The dierent mutations were indicated by their restriction polymorphisms. The C718T mutation results in a restriction polymorphism using MwoI (Biolabs, Beverly, MA, USA) and the C767A mutation results in a restriction polymorphism using BslI (Biolabs). The restriction polymorphisms were con®rmed as mutations by sequencing several cDNA clones at Eurogentec (Seraing, Belgium). All PCR products were analysed on a 1.5% 16EB agarose gel. Protein truncation test (PTT) To detect translation-terminating mutations, s1 and s2 primers were modi®ed 5' terminally with an expression cassette containing a T7-promotor sequence followed by an eukaryotic translation initiation sequence (exp=5'-GCTAAT ACG ACT CAC TAT AGGAACAGACCACCATGG3', startcodon is underlined (Roest et al., 1993). The following expression primers were purchased from Eurogentec: slexp, 5'-exp-CTTCGGGCCTCTGGAATTTA-3'; s2exp, 5'-exp-AGACCAGGGACTTGCGTAGACA-3'. s1exp-a3 and s2exp-a1 primer sets were used in a PCR reaction as described above. In vitro transcriptiontranslation was done on total PCR products with the TnT1 coupled reticulocyte lysate system (Promega). Proteins were in vitro translated in the presence of [35S]methionine (Amersham), then separated by 12% SDS ± PAGE and bands were revealed by autoradiography. cDNA and genomic DNA sequencing RT ± PCR products were excised from a 1.5% agarose gel and puri®ed using the Sephaglass Bandprep kit (Pharmacia, Upsala, Sweden). DNA sequence was determined by the dideoxy chain termination method using ¯uorescent

dye terminators in a 373A automated DNA sequencer (Applied Biosystems, Foster City, CA, USA). Sequencing primers for a-catenin were the same as for the generation of the PCR-products. Primers for DNA polymerase d and HMSH6 were generated on the basis of the published sequences (da Costa et al., 1995; Papadopoulos et al., 1995). Sequences were analysed by the Lasergene software (DNASTAR, Madison, WI, USA). Immunoprecipitation, Western blotting and immunocytochemistry Immunoprecipitation, Western blotting and immunocytochemistry were exactly as described previously (Vermeulen et al., 1995a). We used the following antibodies: HECD-1 (Takara Shuzo, Kyoto, Japan) and MB2 (Bracke et al., 1993) mouse monoclonal antibodies against human Ecadherin; anity-puri®ed rabbit polyclonal antibody 1522, raised against the b-catenin-speci®c synthetic peptide CPGDSNQLAWFDTDL (Vermeulen et al., 1995b); anitypuri®ed rabbit polyclonal antibody 1597 raised against the aE-catenin-speci®c synthetic peptide C-HVDPVQALSEFK (Vermeulen et al., 1995a). Karyotyping and DNA ploidy Cytogenetic analysis was done on G-banded slides according to standard procedures (Speleman et al., 1992). For ¯ow cytometric analysis of DNA ploidy, cells were stained with the Coulter1 DNA-prep reagent system (Coulter, Hialeah, FL, USA). Fluorescence intensity was measured on single cells with a FACSort ¯ow cytometer (Becton Dickinson, Mountain View, CA, USA) with peripheral blood lymphocytes as a standard. Drugs and conditioned media The following drugs were dissolved and used at the concentrations indicated: dimethylphormamide (Sigma), and dimethylsulphoxide (Sigma) at 1% and 0.1%; alltrans-retinoic acid (Sigma) in ethanol at 1 mM and used at concentrations of 1076 and 1077 M; recombinant IGF-I (Boehringer Mannheim) at 1 mg/ml; 12-O-tetradecanoylphorbol-13-acetate (Sigma) in dimethylsulphoxide at 250 mg/ml and used at 250 and 25 ng/ml; sodium butyrate (Sigma) at 2 and 0.2 mM; azacytidine (Sigma) freshly dissolved at 5 mM in 10 mM Tris, pH 6.3 for use at 1 to 2 mM. Conditioned medium was harvested from E11 and E11R1 cell cultures, centrifuged during 5 minutes at 200 g, ®ltered through a 0.2 mm ®lter (Schleicher and Schuell, Dassel, Germany) and admixed with equal amounts of fresh culture medium. Mutation rate analysis E11, E41 and MRC-5 cells were seeded at respectively 200 (E11, E41) and 1000 (MRC-5) cells per 25 cm2 and grown to a ®nal density of 107 to 106 cells. 106 cells per 25 cm2 were then grown in the presence of 15 mM 6-thioguanine and resistant colonies (mutated in HPRT) were counted after 2 weeks in accordance with Bhattacharyya et al. (1994). We calculated mutation rates in accordance with the Luria and DelbruÈck (1943) formula using the table published by Capizi and Jameson (1973).

Acknowledgements We thank L Baeke, R Colman, A Verspeelt, K Staes and R Coquyt for technical assistance and J Roels for preparation of the illustrations. This work was supported by the Fund for Scienti®c Research-Flanders, The Sportvereniging tegen Kanker, The Belgian Cancer Association, The ASLK/


VIVA-verzekeringen and the GOA from the Vlaamse Gemeenschap, Brussels, Belgium. FN is a Research

Assistant and FVR a Research Director with the Fund for Scienti®c Research-Flanders.

References Behrens J, Mareel MM, Van Roy FM and Birchmeier W. (1989). J. Cell Biol., 108, 2435 ± 2447. Berx G, Cleton-Jansen A-M, Nollet F, de Leeuw WJF and van de Vijver MJ, Co. (1995). EMBO J., 14, 6107 ± 6115. Bhattacharyya NP, Skandalis A, Ganesh A, Groden J and Meuth M. (1994). Proc. Natl. Acad. Sci. USA, 91, 6319 ± 6323. Bhattacharyya NP, Ganesh A, Phear G, Richards B, Skandalis A and Meuth M. (1995). Hum. Mol. Genet., 4, 2057 ± 2064. Bracke ME, Van Larebeke NA, Vyncke BM and Mareel MM. (1991). Br. J. Cancer, 63, 867 ± 872. Bracke ME, Vyncke BM, Bruyneel EA, Vermeulen SJ, De Bruyne GK, Van Larebeke NA, Vleminckx K, Van Roy FM and Mareel MM. (1993). Br. J. Cancer, 68, 282 ± 289. Bracke ME, Van Roy FM and Mareel MM. (1996). In: GuÈnthert U and Birchmeier W (eds). Attempts to Understand Metastasis Formation I. Springer: Berlin, pp. 123 ± 161. Breen E, Clarke A, Steele Jr G and Mercurio AM. (1993). Cell Adhesion Commun., 1, 239 ± 250. Bussemakers MJG, van Bokhoven A, Dikho X, Jansen CFJ and Schalken JA. (1996). Clin. Exp. Metastasis, 14 (Suppl. 1): 15. Capizzi RL and Jameson JW. (1973). Mutat. Res., 17, 147 ± 148. Casares S, Ionov Y, Ge H-Y, Stanbridge S and Perucho M. (1995). Oncogene, 11, 2303 ± 2310. Chen TR, Dorotinsky CS, McGuire LJ, Macy ML and Hay RJ. (1995). Cancer Genet. Cytogenet., 81, 103 ± 108. da Costa LT, Liu B, El-Deiry WS, Hamilton SR, Kinzler KW, Vogelstein B, Markowitz S, Willson JKV, de la Chapelle A, Downey KM and So AG. (1995). Nat. Genet., 9, 10 ± 11. Eshleman JR, Markowitz SD, Donover PS, Lang EZ, Lutterbaugh JD, Li G-M, Longley M, Modrich P, Veigl ML and Sedwick WD. (1996). Oncogene, 12, 1425 ± 1432. Ewing CM, Ru N, Morton RA, Robinson JC, Wheelock MJ, Johnson KR, Barrett JC and Isaacs WB. (1995). Cancer Res., 55, 4813 ± 4817. Frei JV. (1992). Cancer, 69, 1108 ± 1111. Hirano S, Kimoto N, Shimoyama Y, Hirohashi S and Takeichi M. (1992). Cell, 70, 293 ± 301. Kemler R. (1993). TIG, 9, 317 ± 321. Kinzler KW and Vogelstein B. (1996). Cell, 87, 159 ± 170. Knudson AG Jr. (1985). Cancer Res., 45, 1437 ± 1443. Koi M, Umar A, Chauhan DP, Cherian SP, Carethers JM, Kunkel TA and Boland CR. (1994). Cancer Res., 54, 4308 ± 4312. Kouri M, Laasonen A, Mecklin J-P, JaÈrvinen H, Franssila K and PyrhoÈnen S. (1990). Cancer, 65, 1825 ± 1829. Lazar V, Grandjouan S, Bognel C, Couturier D, Rougier P, Bellet D and Bressac-de Paillerets B. (1994). Hum. Mol. Genet., 3, 2257 ± 2260. Luria SE and DelbruÈck M. (1943). Genetics, 28, 491 ± 511. Lynch HT and Smyrk T. (1996). Cancer, 78, 1149 ± 1167. Mareel MM, De Baetselier P and Van Roy FM. (1991). Mechanisms of Invasion and Metastasis. CRC Press, Boca Raton, Ann Arbor, Boston. ISBN 0-8493-6254-7. Mareel M, Berx G, Van Roy F and Bracke M. (1996). J. Cell. Biochem., 61, 524 ± 530. Mareel M, Leroy A and Bracke M. (1997). Handbook of Clinical Neurology, Neuro-Oncology, Part III. Vinken PJ and Bruyn GW (eds). Elsevier, Vol. 25 (69), pp. 105 ± 122.

Markowitz S, Wang J, Myero L, Parsons R, Sun L-Z, Lutterbaugh J, Fan RS, Zborowska E, Kinzler KW, Vogelstein B, Brattain M and Wilson JKV. (1995). Science, 268, 1336 ± 1338. Morin PJ, Sparks AB, Korinek V, Barker N, Clevers H, Vogelstein B and Kinzler KW. (1997). Science, 275, 1787 ± 1790. Morton RA, Ewing CM, Nagafuchi A, Tsukita S and Isaacs WB. (1993). Cancer Res., 53, 3585 ± 3590. Nollet F, Van Hengel J, Berx G, Molemans F and Van Roy F. (1995). Genomics, 26, 410 ± 413. Oda T, Kanai Y, Shimoyama Y, Nagafuchi A, Tsukita S and Hirohashi S. (1993). Biochem. Biophys. Res. Commun., 193, 897 ± 904. Papadopoulos N, Nicolaides NC, Liu B, Parsons R, Lengauer C, Palombo F, D'Arrigo A, Markowitz S, Willson JKV, Kinzler KW, Jiricny J and Vogelstein B. (1995). Science, 268, 1915 ± 1917. Parsons R, Myero LL, Liu B, Willson JKV, Markowitz SD, Kinzler KW and Vogelstein B. (1995). Cancer Res., 55, 5548 ± 5550. Rhyu MS. (1996). J. Natl. Cancer Inst., 88, 240 ± 251. Roest PAM, Roberts RG, Sugino S, van Ommen G-JB and den Dunnen JT. (1993). Hum. Mol. Genet., 2, 1719 ± 1721. Rubinfeld B, Robbins P, El-Gamil M, Albert I, Por®ri E and Polakis P. (1997). Science, 275, 1790 ± 1792. Shibata D, Peinado MA, Ionov Y, Malkhosyan S and Perucho M. (1994). Nat. Genet., 6, 273 ± 281. Shimoyama Y, Nagafuchi A, Fujita S, Gotoh M, Takeichi M, Tsukita S and Hirohashi S. (1992). Cancer Res., 52, 5770 ± 5774. Shiozaki H, Iihara K, Oka H, Kadowaki T, Matsui S, Gofuku J, Inoue M, Nagafuchi A, Tsukita S and Mori T. (1994). Am. J. Pathol., 144, 667 ± 674. Shiozaki H, Kadowaki T, Doki Y, Inoue M, Tamura S, Oka H, Iwazawa T, Matsui S, Shimaya K, Takeichi M and Mori T. (1995). Br. J. Cancer, 71, 250 ± 258. Souza RF, Appel R, Yin J, Wang S, Smolinski KN, Abraham JM, Zou T-T, Shi Y-Q, Lei J, Cottrell J, Cymes K, Biden K, Simms L, Leggert B, Lynch PM, Frazier M, Powell SM, Harpaz N, Sugimura H, Young J and Meltzler SJ. (1996). Nature Genetics, 14, 255 ± 257. Speleman F, Colpaert C, Goovaerts G, Leroy JG and Van Marck E. (1992). Cancer Genet. Cytogenet., 58, 176 ± 179. Takeichi M. (1991). Science, 251, 1451 ± 1455. Tompkins WAF, Watrach AM, Schmale JD, Schultz RM and Harris JA. (1974). J. Natl. Cancer Inst., 52, 1101 ± 1110. Vermeulen SJ, Bruyneel EA, Bracke ME, De Bruyne GK, Vennekens KM, Vleminckx KL, Berx GJ, Van Roy FM and Mareel MM. (1995a). Cancer Res., 55, 4722 ± 4728. Vermeulen SJ, Bruyneel EA, Van Roy FM, Mareel MM and Bracke ME. (1995b). Br. J. Cancer, 72, 1447 ± 1453. Vermeulen SJ, Chen TR, Speleman F, Nollet F, Van Roy FM and Mareel MM. (1998). Cancer Genet. Cytogenet., (in press). Vermeulen S, Van Marck V, Van Hoorde L, Van Roy F, Bracke M and Mareel M. (1996). Pathol. Res. Pract., 192, 694 ± 707. Vleminckx K, Vakaet Jr L, Mareel M, Fiers W and Van Roy F. (1991). Cell, 66, 107 ± 119.

915