Wnt5a Cloning, Expression, and Up-Regulation in Human Primary Breast Cancers’

Vol. 1, 215-222, February 1995 Clinical Cancer Research 215

Sue Lejeune, Emmanuel L. Huguet, brane or matrix (7). Writs produce morphological effects on Andrew Hamby, Richard Poulsom, some mouse mammary cancer cell lines by transfection in autocnine (8) and paracrine mechanisms (9). and Adrian L. Harris2 A survey of expression in normal mouse mammary gland Imperial Cancer Research Fund, Molecular Oncology Laboratory, development showed that some Wnts are expressed in virginal Institute of Molecular Medicine, John Radcliffe Hospital, Headington, Oxford, 0X3 9DU, United Kingdom breast, some in pregnancy, and others in lactation (10). How- ever, Wntl, which is involved in carcinogenesis is not expressed in normal mouse mammary tissue. This implicates Writ gene ABSTRACT family members in normal breast development and suggests Wnt genes are involved in mouse mammary cancer, but aberrant expression of other members can contribute to malig- their role in human cancer is unknown. Human Wnt5a was nancies (1 1). cloned from a placental cDNA library and used to assess Evaluation of the normal expression of Wnt genes in hu- expression by ribonuclease protection and in situ hybridiza- man breast epithelium and cancer would contribute to under- tion in human breast cell lines and in normal, benign, and standing the role of Wnt genes in human cancer. It has been malignant breast tissues. Human WntSa shows over 99% shown that some of the Wnt genes are expressed in human breast homology at amino acid level with mouse Wnt5a, and 90% tissue, and that quantitative differences exist in the Win expres- with Xenopus Wnt5a. It was expressed only at low levels in sion profile of normal and proliferative lesions (12). breast cell lines and normal breast tissue. Benign prolifera- We chose Wnt5a as a candidate human gene to clone and tions and invasive cancer respectively showed 10-fold and evaluate because it is expressed in normal mouse breast epithe- 4-fold higher Wnt5a than normal breast tissues. The greater hum to a low extent, and also in mouse breast cancer cell lines up-regulation in benign conditions suggests a robe in aber- (10). Furthermore it has some different properties from the rant differentiation. In situ hybridization localized the signal human Wnt genes previously cloned in its effects on cell gap to the epithelial component. Wnt5a is the first member of the junctions (13) and Xenopus development (14). Interactions be- Wnt family to demonstrate overexpression in human breast tween Wnt genes with different normal functions may contribute cancer. It was not associated with factors known to affect to malignant transformation (1 1). breast cancer prognosis such as lymph node status or epidermal growth factor receptor status. MATERIALS AND METHODS

INTRODUCTION Isolation of 384-Base Pairs Fragment of Human WntSa from Fetal Brain cDNA Library. Two hundred ng of a In mouse mammary tumor virus-induced breast cancer, human fetal brain library in plasmid pCDM8 (obtained from Dr. analysis of insertion sites has shown activation of endogenous D. Simmons and Dr. J. Fawcett, Institute of Molecular Medi- genes, mt genes (1). mt-i (now WntJ) was the first gene isolated and shows strong homology to a Drosophila developmental cine, Oxford, United Kingdom) was used as a template for PCR. gene, wingless (2), involved in pattern development. Two other Amplification of cDNA was carried out using 500 ng of each of members of this family, Wnt3 and Wnt2 (also called mt-related the degenerate forward (5’-GGGGAATTCCA’’/GGA”/GTGT/ protein, irp) are also involved in mouse mammary cancer (3, 4). AA”/TGT/CCAT3’) and reverse (5’-AAAATCTAGA’’/ Fifteen Wnt genes have been isolated in vertebrates. They GCA/GCACCA/GTG/GAA3) oligonucleotide primers pre- are expressed in many adult and embryonic tissues (5) and are viously described by Gavin et a!. (5). PCR products were involved in morphological development. The importance of separated on a 2% agarose gel. Products of the correct size their role is reflected by the severity of the phenotypic abnor- (predicted from known Writ sequences) were recovered, then malities that result from aberrant Wnt expression. Thus, Wnt further amplified using the above primers and ligated into the genes have important roles in development and in cancer. plasmid pBluescript KS+ (Stratagene). JM1O9 cells were trans- formed with the reaction products, and the nucleotide sequences The role of Wnts in mouse mammary carcinogenesis has been extensively studied and there is evidence that they are of clones containing inserts of correct size were determined by secreted proteins, processed via the Golgi apparatus (6), which dideoxy chain termination sequencing. Clones with significant Wnts remain tightly associated with the extracellular plasma mem- homology to known were identified using the FASTA program (GCG) and included one clone containing a 384-base pair fragment of human Wnt5a. Isolation of Human Wnt5a from Placental cDNA Li- brary. Replica colony lifts of approximately i0 recombinant Received 9/23/94; accepted 9/28/94. clones of a human placental library in the plasmid pCDM8 were , This work was funded by the Imperial Cancer Research Fund and Oxfordshine Health Authority. prepared using Hybond-N membranes (Amersham). The mem- 2 To whom requests for reprints should be addressed. branes were hybridized to a Wnt5a probe synthesized using the

384-base pair fragment isolated as described above. The probe Ribonuclease Protection Assays. Ribonuclease protec- was generated by incorporating [a-32PJdCTP (Amersham) in a tion assays were carried as described in Ref. 16. Briefly, a PCR amplification of the 384-base pair fragment using the 384-base pair fragment of human Wnt5a was cloned into pBlue- above degenerate primers. The resultant species were end filled script KS. The EcoRV linearized plasmid was used to generate and separated from unincorporated nucleotides by passage antisense [a-32P]CTP-labeled probes with T7 RNA polymerase through a Sephadex G-50 spin column (Boehninger Mannheim). (Gibco). Probes were hybridized to 10-jig samples of total RNA After hybridization of the probe to the membranes, positive extracted from tissues and cell lines by a single-step extraction clones were identified by autoradiognaphy, and subjected to method (17). Hybridization was carried out at 45#{176}Cfor 16 h. further rounds of screening. After four rounds of screening, Each hybridization also contained an antisense probe for positive clones were selected and sequenced. Clones with sig- GAPDH3 as a loading control. GAPDH probes were prepared nificant homology to known W,its were identified and included from a 120-base pain b fragment of GAPDH cloned into pBlue- one containing 798 base pairs of the 3’ end of the human Wnt5a script 5K (18). Unhybridized probe was digested with RNase A cDNA as well as some 3’ untranslated sequences. and Ti, and protected fragments were electrophoresed on poly- Further sequence 5’ to the 798-base pair 3’ terminus was acrylamide gels. Dried gels were autoradiographed. obtained using a nested PCR strategy. A primary PCR reaction Image Analysis. Autoradiographs of nibonuclease pro- was carried out using 200 ng of a placental cDNA library in tection assays were scanned using a Bio-image analyser (Milli- pCDM8 as template. The primers used were a mouse Wnt5a- gan Bioresearch) to determine RNA abundance. Wnt5a values specific forward primer FPI (5’-ATGAAGAAGCCCATTG- were normalized to GAPDH to allow for loading. MCF1O RNA GAATA-3’) corresponding to the 21 extreme 5’ nucleotides of was included in all assays as a positive control. This was mouse Wnt5a, and a human Wnt5a-specific reverse primer RPI assigned a unit level of expression and all other values were (5 ‘-GCACGCCCGGCTCATGGCGTF-3 ‘) corresponding to a standardized to this. known sequence in the 798-base pair partial clone. Fragments of Cell Lines. The following breast cell lines were obtained correct size (approximately 462 base pairs) were recovered and from ATCC (Bethesda, MD): T47D (ATCC HTB 133), used as template for nested PCR. In the nested PCR reaction the MDA231 (ATCC HTB26), MCFJO (ATCC CRL1O3I7), forward and reverse primers (FP2 and RP2) were 3’ and 5’ of MDA415 (ATCC HTB128), MDA453 (ATCC HTBI31), the primers FP1 and RP1, respectively. Primer FP2 (5’- MDA157 (ATCC HTB24), BT2O (ATCC HTB19), SKBR3 AANTCNTGGTGGTCNCTNGG-3’) was a fully degenerate (ATCC HTB3O). ZR9B1 1 (ZR-75-1), ZR4, and ZR1 1 were primer, and primer RP2 (5’-GTGTI’ATCCACAGTGCT-3’) obtained from Dr. E. Valvenius (Department of Pathology, Uni- was a human Wnt5a-specific primer corresponding to the ex- versity Hospital, Uppsala, Sweden). MCF7s were obtained from treme 5’ region of the 798-base pair partial clone. Fragments of Dr. B. Durkacz (Cancer Research Unit, University of Newcastle correct size (approximately 234 base pairs) were recovered, upon Tyne). Adniamycin-resistant MCF7s were obtained from subcboned into plasmid pBluescnipt 5K, and sequenced. Clones Dr. K. Cowen (National Institutes of Health). Lines 2-5-2a, with significant homology to Wnt5a were identified and in- 3-4-1, 5-3-1, 6-1-1, MTSV1-7, and MTSV4-1 were obtained cluded one containing 234 base pairs of human Wnt5a sequence from Dr. J. Taylor (Imperial Cancer Research Fund, London, 5’ to the 798-base pair partial clone. United Kingdom). A nested PCR strategy was also used in order to clone the Cell Culture. T47D, MDA23I, MCF1O, MDA453, extreme 5’ region of Wnt5a. In a primary PCR reaction, 200 ng SKBR3, MCF7, Adniamycin-resistant MCF7, ZR9B1 1, ZR4, of human placental cDNA library in pCDM8 was used as and ZR1 1 were maintained in DMEM with 10% FCS. BT2O was template. The primers used were a pCDM8-specific primer and maintained in Eagle’s MEM supplemented with 15% FCS and 2 the human Wnt5a-specific internal primer RP1. Products of a mM glutamine. MDA4I5 was maintained in DMEM, iS% FCS, large enough size to contain the 5’ end of Wnt5a were recovered 1 jiM hydrocortisone, 10 jig/mI insulin, and 10 jig/mb glutathi- and used as template in the nested PCR. In the nested PCR one. MDA1S7 was maintained in RPMI 1640 and 10% FCS. reaction, primers FP1 and RP2 were used. Fragments of correct MTSV1-7, MTSV4-l, 2-5-2a, 3-4-1, 5-3-1, and 6-1-1 were size (approximately 366 base pairs) were recovered, cloned into maintained in DMEM: Ham’s F12 (i:1), 10% FCS, 10 jig/mb TA cloning vector (Invitrogen), and sequenced. A clone con- insulin, and 5 jig/ml hydrocortisone. All cultures were grown on taming the 5’ end of Wnt5a was identified. Thus the sequences plastic dishes in 5% CO,-95% air in humidified incubators. All of the original 798-base pain partial clone and the two nested cultures were free of Mvcoplastna. PCR products combined to give the full-length sequence. Handling of Clinical Samples. Protocols for handling of Chromosomal Localization of Human Wnt5a. Twen- clinical samples and assays for hormone and growth factor ty-jig DNA samples from a panel of human-rodent hybrid cell receptors were followed as detailed in Leieune et a!. (19). lines, and control human, mouse, and hamster DNA (obtained Briefly, Tumors were considered to be ER positive if they from Dr. N. Spurn, Imperial Cancer Research Fund, London, contained at least 10 fmol of specific binding sites per mg of United Kingdom) were digested with EcoRI, fractionated on a cytosolic protein, and EGFR positive if they contained at least 0.7% agarose gel, and transferred to Hybond-N membranes according to Southern’s protocol (15). The membranes were then hybridized to a random primer generated [a-32P]dCTP- labeled probe, using a 1.4-kilobase XbaI fragment of the original 3 The abbreviations used are: GAPDH, glycenaldehyde-3-phosphate de- (placental) partial human Wnt5a clone as template. Signals were hydnogenase; ER, estrogen receptor; EGFR, epidenmal growth factor detected by autoradiography. receptor; ATCC, American Type Culture Collection.

FPI 20 fmol of specific binding sites per mg of membrane protein. ataaaaaaacccattaaaatattaagcccaggagt’ gctttggggatggctggaagtgca Human tissue samples were selected to represent normal breast H KKP IGI L S P 1 V /, L G M A (. S A tissue, benign breast disease, and breast cancer. atgtcttccaagttcttcctagtggctttggccatatttt tctccttcgcc-aggt tqta Hybridization. A single-stranded antisense RNA In Situ 61 , , +.------, 121) probe to Wnt5a was transcribed from EcoRI-linearized Wnt5a MSS KFFLVALA1FFSFSQ 2 V DNA using T3 DNA polymerase (Promega), and 35S-UTP FP2 attgaagccaattcttaataatcactaacitatgaataaccctgttcagatgtcagaaqt a (-800 Ci/mmol; Amersham International) as the sole source of 121 + , , IKANSWWSLGHNNPV Q MSF UTP. Histological sections of human tissues that had been fixed tatattataggagcacagcct ctctgcagccaactggcaggactttctcaaggacaqaag in neutral-buffered formalin and embedded in paraffin wax were 181 ‘ --‘ + ‘ + ‘ 245 Y I I S A Q 1’ L C S Q L A G L S 0 5 Q F treated in the manner described by Senior et a!. (20) with minor modifications. In summary, 1 X 10#{176}cpm of unhydrolyzed probe 241 - ‘ - i. in 10 ml buffer was hybridized overnight at 55#{176}Cto sections K L C H L V Q 1) H N QS : C ECAKV - permeabilized with proteinase K. Posthybnidization steps in- SF2 atcaaagaatgccagtatcadt tccgacatcgaaggtggaactgcacacLaLauatac cluded several large volume washes in a 50% formamide buffer 301 -- ,--- ------‘ I F F C 0 5 0 F H H P Is W N C S T .‘ II 1 at 55#{176}Cto remove unhybrized probe, RNase A treatment to (ctgtttttggcagggtqatg(aqataggcagccgcgagacggccttcacara I digest single-stranded and imperfectly hybridized domains, and 361 ------,------, -- -4V T 2 V F G S .-‘ H (I 5 5 F S A F 2 5 1 extensive washing to remove these cleaved fragments were performed. The final washes were in 0.SX SSC (1X SSC is gtuaqcqcagcaggggtggt9aacQCCat0aQCCUQUCatUCCgCgagggCq4gCt 1 0.15 M NaC1 and 0.015 M sodium citrate) at 65#{176}Cfor30 mm twice. 421 ------‘ --- : :- AAG ‘.‘ 2 1. 5 ‘ 5 15 A C H F C 51:. Slides were dehydrated and processed for autoradiography (Ilford

K2) at 4#{176}Cfor7 to 10 days. Latent images were developed with 481 . -. 54) S C V C S H A A H P K 1) 1. 1 5 1) W 1 H Kodak D-19, and sections were Giemsa counterstained. ggctgcggcgacaacatcqaCtatggctacCgCtttgCCaaqgagttCgtggaCgCc:3 The 3-actin mRNA was used as a positive control for the 541 + --‘-- ‘ ------‘ + 15_la G C G D N I 1) Y (1 5 5 FAKFFVDAF presence of mRNA species, as described previously by Wright gagcgggagcgcat ccacgccaacqqct cIt acgagagt gct cgcat cct cat qaa as et a!. (21). Labeled Wnt5a sense transcripts were used as con- 601 ------‘ _-_-- , trols for nonspecific niboprobe binding. EPEF I HAK S V F S A P 1 F H N caaacaa cgaggccggccgcaggaCggtgtaCaacctggctgatgtggccgcsa4t4C 661 ------‘-- -‘ ---‘ --- 7,))) RESULTS HNNKA C) P P TVVNLADVACKC catqgggt gt ccggctcatgt agcctgaagaCatgctggctgcagctqgCagacttCcq(

Isolation of WntSa cDNA. A partial length cDNA was 721 ---- ------ 7817 H(7SSSCSSKTCNLQ1ASFF initially isolated by using PCR primers to homologous domains in Wnt genes. This was used to isolate a cDNA from a human aaggtgggtgatgccctgaaggagaagtacgacagcqcggcggcCatgCggctcaacagc 781 ------‘------+- , 845 placental library, which, in combination with nested PCR prod- KVCDALKEK Y D S A A A M S L N 77 ucts from the same library provided the full human Wnt5a cggggcaagt tggtacaggtcaacagccgct tcaactcgcccaccacacaagacctggtc 841 -- #{149}- - -, ca: cDNA coding sequence shown in Fig. 1 with the corresponding SG KLV Q V N S S F N S P T T Q S 1. V amino acid sequence. The sequence comparison of human tacat cgaccccagcCCtgaCtaCtgCgtgcgCaatgagagCaccggct cgctq5gcacg 901 ------‘ ‘ ‘------. 1131) Wnt5a with mouse and Xenopus Wnt5a shows extensive con- V I [3 FSPDSCVPNFS I 77 77 1. 77 1 servation with amino acid level homology of 99% and 90%, cagggccqcctgtgCaaCaag8CgtCggagggCatggatggCtqC3dqCt catgtgctgc respectively. 961 ‘ ‘ -‘------‘ a:: Q GS L CN KTSKCMDG C F: :. 0 CC Chromosomal Localization of Wnt5a. The chromo- ggccgtggctaC9dCCagttC8agaCCgtgCag8CggagCgCtqCC.Ct 3caaqt tccac some location of the human Wnt5a was determined by screening 1021 -‘ ‘ G R G V 0 0 F K TVQT05CHCKFH hamster-human and mouse-human hybrid cell lines which had tggtgccgctaCgtCaagtgCaagaagtgCaCggagatCgtggaCCgL’JtgtgtgCdag known human chromosome karyotypes. Southern blotting 1051 + + . 141) w CY ) F C F showed a Wnt5a signal located on human chromosome 3 (Fig. C VKCKFCTEIVD 2), which was absent in a mouse-human hybrid cell line carrying tag 1141 ---1143 a p1 1 to p terminus deleted chromosome 3 (data not shown). This suggests that Wnt5a is on chromosome 3p. Fig. I Nucleotide sequence of human Wnt5a. The location of PCR Expression of Wnt5a in Human Breast Cell Lines. To primers FP1, FP2, RP1, and RP2 are underlined. Predicted amino acid assess expression in human breast cancer, a panel of human sequence shown below nucleotide sequence. Conserved cysteine resi- breast cell lines was initially analyzed, since they represent a dues are shown in bold. , stop codon. pure epithelial population. Cell lines from normal breast duct luminal epithelium, benign epithelial proliferation, and in situ or invasive components of breast epithelium were studied (Table 1 and Fig. 3). Levels of expression measured by nuclease protec- in all but one case (BT2O). Thus, in vitro cell lines rarely tion assays were low, requiring 7 days of development. One line expressed Wnt5a. from luminal epithelium had higher levels (MTSV1-7) than the Expression in Human Breast Tissue. Expression was others. then studied in normal breast tissue, tissue from benign breast Cell lines established from malignant pleural effusions and diseases, and primary breast cancers. The clinical details of the metastasis had lower levels than the luminal cell line MTSV1-7 patients (e.g., age, tumor receptor status) are given in Table 2.

ahcdefghijklmnopqrl23

..otr . I

Fig. 2 Southern blot of human-rodent hybrid cell lines with human Wnt5a probe. Lane 1, human genomic DNA; Lane 2, mouse genomic DNA; Lane 3, hamster genomic DNA. Lanes a-r, DNA from hybrid human-rodent cell lines, each beaning human chromosome DNA. Lanes: a, chromosome 1; b, chromosome 2; c, chromosome 3; d, chromosome 5; e, partial chromosome 6;f, chromosome 7; g, chromosome 8; h, chromosome 9; i, chromosome I 1 ; j, chromosome 12, k, chromosome 13; 1, chromosome 14; m, chromosome 15; n, chromosome 16; o, chromosome 17; p, chromosome 19; q, chromosome 20 and partial 4; r, chromosome 21. Arrows in Lanes I and c, signal for human genomic DNA (Lane 1) corresponds to that in Lane

F’, i.e. , chromosome 3. The absence of Wnt5a signal in a human-mouse hybrid cell line carrying a p terminus deleted chromosome 3 (data not shown) suggests that Wnt5a is on human chromosome 3p.

Table I Comparison of Wnt5a expression in breast cell lines derived phyllode tumors (n = 2). Tumors (n = 28) were selected to from ductal epithelium, in situ and invasive components of breast represent different subgroups according to known prognostic carcinoma, and breast cancer metastases factors (node, ER, and EGFR status). In contrast to the cancer Expression cell lines, Wnt5a was commonly expressed in primary breast standardized cancer (Fig. 4). Comparing normal breast tissue to tumor tissue Cell line Type to MCF1O MCFIO Immortal nonmalignant I showed levels that were 4-fold higher on average in the latter ZR9B1I ER+ breast cancer 0 (Table 2). Ten of 28 tumors had levels higher than the highest expression in the cell lines from normal breast tissue. Thus, the ZR1I ER+ breast cancer <1 cell lines reflected normal breast expression, but not tumor ZR4 ER + breast cancer 0 levels. Tumor levels were significantly greater than normal T47D ER+ breast cancer tissue levels (P 0.0004, Mann-Whitney U test). MCF7 ER+ breast cancer MDA231 ER- breast cancer 0 Benign breast tissue also showed much higher expression MDA415 ER- breast cancer < I than normal breast tissues, similar to and often greater than MDA453 ER- breast cancer 0 levels in many of the tumors. Two different types of benign MDAI57 ER- breast cancer BT2O ER- amplified EGFn ?8 breast disease were studied: fibroadenomas and fibrocystic dis- SKBR3 ER- amplified erbB-2 <1 ease. The former is a benign lesion involving epithelial and Adniamycin-resistant Drug-resistant MCF7 < 1 stromal elements. The latter is a nontumorous collection of cysts MCF7 and ducts with some epithelial hyperplasia. In both types of 2-5-2A Derived from benign 0 benign breast disease there was high expression of Wnt5a (Fig. component of breast 3-4-I Derived from in situ component 1 nuclease protection). Benign levels were significantly greater 5-3-I Derived from in situ component 1 than those of normals (P = 0.0012 Mann-Whitney U test) and 6-1-I Derived from in situ component <1 also than those of tumors, although to a smaller extent (P = MTSV1-7 Luminal normal cells, SV4O 5 0.03). immortalized < 1 In situ hybridization using the Wnt5a probe was carried out MTSV4-1 Luminal normal cells, SV4O to assess localization of expression. This showed that levels immortalized were low and difficult to detect in normal breast but were detectable in a few ducts and lobules. In some larger ducts the

Normal tissues (n = 15) were obtained either from reduction mRNA appeared to be expressed in luminal cells but not the

mammoplasties (n = 7), or from normal tissue adjacent to myoepithelial population. In fibroadenomas there was high ex-

tumors (n = 8). Benign breast disease samples consisted of pression in the epithelial component uniformly throughout the

fibroadenomas (n = 5), fibrocystic disease (n = 3), and benign tumor (Fig. 5, a and b). In fibrocystic disease it was again

ab c de f gh i j k 1 mnop q r s

wnt5a

-- - w--- --_ gapdh LT:____a.____ A . A A -A..

Fig. 3 RNase protection assay on cell lines. Wnt5a and GAPDH signals are shown. Lanes a-f, human mammary epithelial cell lines; Lanes g-s, human breast cancer cell lines. a, 2-5-2a; b, 3-4-1; c, 5-3-1; d, 6-1-1; e, MThV1-7;f, MRSV4-1; g, MCF1O; h, ZR1 1; i, ZR4;j, MDA41S; k, MDA4S3; I, MDA4S7; m, BT2O; n, SKBR3; o, Adniamycin-resistant; p, MCF7; q, AR9B11; r, T47D; s, MDA231.

Table 2 Wnt5a expression assayed by RNase protection in human high-expressing cases showed no evidence of this (data not breast cancer shown). Sample Type Features of carcinomas known to relate to tumor phenotype were compared with Wnt5a expression. There was no correba- Normal tion of either ER or EGFR with Wnt5a expression (P = 0.3 and Benign Tissue Primary 0.5, respectively, Spearman’s rank correlation coefficient; Table Total (a = 10) (a 15) (n - 28)

2), nor was there a correlation with node status (P = 0.1). There Age (yr) Mean 49 39 42 56 was no association with menopausal status, as indicated by age SE 2.13 3.1 4.8 2.3 >50 years or <50 years old (Table 2). ER and EGFR were Median 49 38 46 57 inversely related to each other, as previously reported (Ref. 22; Min,a max 18, 86 25, 52 18, 78 32, 86 P = 0.005, Spearman’s rank correlation coefficient). ER was Wnt5a (RNA units) related to age and EGFR inversely as has been described before Mean 7.1 18.7 1.5 5.9 (Ref. 22; P = 0.0007 and 0.036, respectively, Spearman’s rank SE 10.4 17.2 1.5 6.4 correlation coefficient). Median 3.0 11.5 1.0 4.0 In relation to patient age, there was no significant differ- Mm, max 0, 45 0, 45 0, 5 0.5, 31 ence between normal and benign samples but the primary cancer ER (fmol/mg) cytosol patients were significantly older than the normal (P = 0.02) and

protein benign (P = 0.0013) samples. However the differences in Mean 128 Wnt5a expression between the samples cannot be a function of SE 168 the different age distributions, as Spearman’s rank correlation of Median 50.5 Mm, max 3, 695 Wnt5a with age as a continuous variable shows no correlation EGFR (fmol/mg) (P = 0. 1 1). Furthermore the group of tumors which overlap the

membrane normal and benign samples in age (n = 9, age <52 years) have protein no different a level of Wnt5a than those which are older (n = 19, Mean 29 age >52 years, P = 0.86). Also, the differences in the level of SE 27 Median 25 Wnt5a between normal samples and cancers are maintained Mm, max 0, 127 whether the cancer patients considered are over or under 52 Node status years of age. Thus, the significant differences in Wnt5a expres- Positive 11 sion seen between the groups are not related to age distribution Negative 15 differences. a Mm, minimu m; max, maximum.

DISCUSSION The human Wnt5a gene shows marked homology to other species including mouse (23) and Xenopus (24). This is char- expressed in the epithelial component (Fig. 5, c and d). Simi- acteristic of all of the Wnt family members which are highly larly, in the malignant tumors (Fig. 5, e and f) Wnt5a was conserved (25). They show greater homology to the family expressed within the epithelial element of the tumor and in member in different species than to other family members clumps of invading cells throughout the stroma. There was no expressed in the same species. Wnt5a is located on the terminal increased localization at the invading edge. The sense Wnt5a region of chromosome 3 beyond the 3pl 1 band, whereas mouse probes used as controls for nonspecific hybridization showed a Wnt5a is on chromosome 14. However, human chromosome 3 is uniform and low background signal (data not shown). syntenic with mouse chromosome 9, suggesting chromosomal Thus, in many cases of benign breast proliferation and rearrangements at this locus during evolution. The region breast cancers. Wnt5a is overexpressed in the epithelium. This 3p2l-25 is known to be involved in loss of heterozygosity in was not due to gene amplification since Southern blots of the human cancer (26), including 30% of breast cancers (27). How-

NOFfl11II gap

.4 Fig. 4 RNase protection assay on human breast tissues. Wnt5a and GAPDH signals are shown for 15 normal breast tissue sam- Benign ples, 10 samples of benign disease, and 21 of the 28 tumors gap - -- examined (data for remaining 7 tumors not shown in this figure but included in Table 2).

5 a --- _. -. -- sIaIignant gap w

.- . 5.. ‘ %_ . a,

. -: _,_ . . -.

- . ,; ‘.. - a.. ___ Fig. 5 In situ hybridization of Wnt5a in human breast tissues, a, fibroadenoma, light field; b, fibroadenoma, dark field; c, fibrocystic disease, light field; d, fibnocystic disease, dank field; e, carcinoma, light field; f, carcinoma, dark field. e, epithelial component; s, stroma; c, carcinoma.

ever, the work of Clark et a!. (28) suggests that Wnt5a is outside as those having amplification of EGFR or erbB-2. With the this region and therefore not the gene involved. exception of BT2O, expression in the cell lines was very low During the course of this work, the isolation of overlapping with levels similar to or lower than those found in normal clones was reported giving the total sequence of the human tissues. The breast cancer cell line BT2O showed high wnt5a Wnt5a cDNA isolated from a human fetal fibroblast library. We expression, similar to the elevated levels found in benign pro- have similarly cloned overlapping clones from a human placen- liferative lesions. In the mouse, Wntl expression in mammary tab library. Our results independently confirm the sequence epithelium produces abnormal morphology in a hormone-inde- published by Clark et a!. (28), as well as the chromosomal pendent fashion (29). In relation to this, the level of expression localization of the gene. Our sequence differs from that of Clark of Wnt5a in breast cancer cell lines appears unrelated to hor- ci a!. (28) at a few nucleotides but these do not give rise to mone receptor status. amino acid differences. In tissues, Wnt5a expression was generally higher than that Expression was studied in a range of human breast cancer in cell lines. Benign and malignant proliferative lesions of the cell lines representing ER-positive and -negative types as well breast respectively showed levels of Wnt5a 10-fold and 4-fold

higher than those in normal tissue. In the mouse Wnt5a is To assess localization of Wnt5a, in situ hybridization was expressed in normal breast tissue and its regulation has been carried out on normal tissues, fibrocystic disease, fibroadeno- studied over a short period of reproductive history (10). Wnt5a mas, and carcinomas. In normal tissues, the level of expression is present in early pregnancy but is undetectable by day 17.5 of was generally too low to detect above background, but could be pregnancy. It is clearly difficult to reproduce such studies in seen in a few ducts. In fibrocystic disease, fibroadenomas, and humans, but our results suggest that human breast tissue gener- carcinomas Wnt5a was expressed in epithelium. In some cases ally resembles that of the mouse in its expression of Wnt5a, it was possible to clearly resolve the basal myoepithelial layer of although no major endocrine effects involving steroid hormones the breast from the luminal epithelium and Wnt5a was expressed were demonstrated in that pre- and postmenopausal breast levels in the luminal cell layer. were no different in any of the patient groups. This does not, In fibroadenomas stromal cells are present in a much however, exclude pregnancy-specific regulation events. Wnt5a greater proportion than in normal epithelium. Dilution of RNA was highest in the benign proliferative lesions, and analysis in from the epithelium by matrix and stromal cells shows how high human breast cancer showed up-regulation of Wnt5a in 10 of 28 the level of RNA must be in the fibroadenoma epithebium carcinomas above the highest level in normal breast tissues. The compared to normal tissue epithelium. It is possible that the up-regulation was not due to gene amplification in the breast stromal cells are involved in regulation of Wnt5a expression in cancers and may therefore be related to transcriptional activa- the epithelium or conversely that Wnt5a expression effects tion or RNA stabilization. The elevated level of Wnt5a is inde- stromal growth. Coculture of luminal cells with stromal cells to pendent of several other factors known to affect prognosis or measure Wnt5a regulation would be helpful to assess the role of biology of human breast cancer such as EGFR and estrogen such stromal epithelial interactions. receptor. In order to understand further the basis of up-regula- The up-regulation of Wnt5a in both benign and malignant tion of Wnt5a in human breast disease, human breast cell lines proliferative disease of the breast suggests an important role of transfected with a panel of oncogenes and cultured in different Wnt genes in breast pathology. Wnt5a is the first member of the extracellular matrices have been studied for regulation of family to demonstrate up-regulation in human breast cancer. Wnt5a. There is evidence that Wnt5a is up-regulated severalfold Cell lines established from benign or normal tissues may be in these circumstances.4 Thus, Wnt5a may be secondarily reg- suitable models for further evaluation of the role of Wnt5a to ulated in response to a range of other genetic changes. Since assess its regulation under variable growth and differentiation Wnts have recently been shown to modulate cell adhesion by conditions. regulation of cadherins, it is possible that Wnt5a may also contribute to this mechanism of regulation of growth and dif- REFERENCES ferentiation. 1. Nusse, R., vanOoyen, A., Cox, D., Fung, Y. K., and Varmus, H. In the mouse, Wnts have a role in the development of Mode of provinal activation of a putative mammary oncogene (mt-i) normal mammary tissue, and aberrant Wnt expression can con- oncogene promoter and its mechanism of activation by insertion of proviral DNA of the mouse mammary tumour virus. Mol. Cell. Biol., tribute to mammary hyperplasia and mammary carcinomas. 10: 4170-4179, 1984. Drawing a parallel between these observations and human 2. Rijsewijk, F., Schuermann, M., Wagenaar, E., Parren, P., Weigel, D., Wnt5a, the expression of the gene in normal breast suggests that and Nusse, R. The Drosophila homolog of the mouse mammary onco- it has a role in this tissue, and its aberrant expression is associ- gene mt-i is identical to the segment polarity gene wingless. Cell, 50: ated with proliferative lesions or aberrant differentiation as in 649-657, 1987. fibroadenomas and fibrocystic disease. Olson and Papkoff (30) 3. Roelink, H., Wagenaar, E., Lopes, d. S. S., and Nusse, R. Wnt-3, a gene activated by proviral insertion in mouse mammary tumors, is homologous have reported that the level of Wnt5a expression is inversely to int-1/Wnt-l and is normally expressed in mouse embryos and adult brain. related to the proliferative rate of the mouse mammary epithelial Proc. Natl. Acad. Sci. USA, 87: 4519-4523, 1990. cell line C57MG in vitro. Although this munine result cannot 4. Roelink, H., Wagenaar, E., and Nusse, R. Amplification and provinal necessarily be extrapolated to the human situation, it argues activation of several Wnt genes during progression and clonal variation against Wnt5a having a simple growth promoting role in the of mouse mammary tumors. Oncogene, 7: 487-492, 1992. benign and malignant lesions we analyzed. Thus, Wnt5a is 5. Gavin, B. J., McMahon, J. A., and McMahon, A. P. Expression of expressed at its highest level in benign disease, although it is multiple novel Wnt-1/int-I-related genes during fetal and adult mouse development. Genes Dev., 4: 2319-2332, 1990. also elevated in cancer. It is possible that high Wnt5a expression 6. Papkoff, J., Brown, A. M., and Vanmus, H. E. The mt-i protoonco- occurs at certain stage of normal differentiation, and that cells gene products are glycoproteins that appear to enter the secretory blocked in this phase show the observed regulation. Olson and pathway. Mob. Cell. Biol. 7: 3978-3984, 1987. Papkoff (30) have shown that in the mouse mammary epithelial 7. Smolich, B. D., McMahon, J. A., McMahon, A. P., and Papkoff, J. cell line C57MG, transfection of Wntl and Wnt2 causes a Wnt family proteins are secreted and associated with the cell surface. 30-fold reduction in the expression of endogenous Wnt4 and a Mol. Biol. Cell, 4: 1267-1275, 1993. smaller decrease in the expression of Wnt5a. Thus, it is possible 8. Blasband, A., Schryver, B., and Papkoff, J. The biochemical properties and transforming potential of human Wnt-2 are similar to Wnt-1. that Wnt5a itself may interfere with the function of other Wnts Oncogene, 7: 153-161, 1992. in human breast tissue where it is overexpressed. 9. Jue, S. F., Bradley, R. S., Rudnicki, J. A., Varmus, H. E., and Brown, A. M. The mouse Writ-i gene can act via a paracnine mechanism in transformation of mammary epithelial cells. Mol. Cell. Biol., 12: 321- 328, 1992. 10. Gavin, B. J., and McMahon, A. P. Differential regulation ofthe Wnt 4 T. Bui, E. Huguet, and A. L. Harris, unpublished data. gene family during pregnancy and lactation suggests a role in postnatal

development of the mammary gland. Mol. Cell. Biol., 12: 2418-2423, 21. Wright, N. A., Poubsom, R., and Stamp, G. W. Epidermal growth 1992. factor receptor induces expression of regulatory peptides in damaged I 1. Buhlen, T. A., Dale, T. C., Kieback, C., Humphreys, R. C., and human gastrointestinal tissue. J. Pathol., 162: 279-284, 1990. Rosen, J. M. Localization and quantification of Wnt-2 gene expression 22. Fox, S. B., Smith, K., Hollyer, J., Greenall, M., Hastnich, D., and in mouse mammary development. Dcv. Biol., 155: 87-96, 1993. Harris, A. L. The epidermal growth factor receptor as a prognostic 12. Huguet, E. L., McMahon, J. A., McMahon, A. P., Bicknell, R., and marker: results of 370 patients and review of 3009 patients. Breast Harris, A. L. Differential expression of human Wnt genes 2, 3, 4, and Th Cancer Res. Treat., 29: 41-49, 1994. in human breast cell lines and normal and diseased states of human 23. Gavin, B. J., McMahon, J. A., and McMahon, A. P. Expression of breast tissue. Cancer Research, 54: 2615-2621, 1994. multiple novel Wnt-1/int-I-related genes during fetal and adult mouse 13. Olson, D. J., and Moon, R. T. Distinct effects of ectopic expression development. Genes Dcv., 4: 2319-2332, 1990. of Wnt-1, activin B, and bFGF on gapjunctional permeability in 32-cell 24. Christian, J. L., Gavin, B. J., McMahon, A. P., and Moon, R. T. Xenopus embryos. Dev. Biol., 151: 204-212, 1992. Isolation of cDNAs partially encoding four Xenopus Wnt-1/int-1-related 14. Moon, R. T., Campbell, R. M., Christian, J. L., McGrew, L. L., proteins and characterization of their transient expression during em- Shih, J., and Fraser, S. Xwnt-SA: a maternal Wnt that affects morpho- bryonic development. Dev. Biol., 143: 230-234, 1991. genetic movements after ovenexpression in embryos of Xenopus !aevis. Development, 119: 97-1 1 1, 1993. 25. Sidow, A. Diversification of the Wnt gene family on the ancestral lineage of vertebrates. Proc. Nail. Acad. Sci. USA, 89: 5098-5102, 1992. 15. Southern, E. M. Detection of specific sequences among DNA fragments separated by gel electrophoresis. J. Mol. Biol., 98: 503-517, 26. Mastro, R., Gasparotto, D., Vukosavljevic, T., Barzan, L., 1975. S., and Boiocchi, M. Three discrete regions of deletion at 3p in head and 16. Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D., Seidman, neck cancers. Cancer Res., 53: 5775-5779, 1993. J. G., Smith, J. A., and Stnuhl, K. Current Protocols in Molecular Biology, 27. Callahan, R., Cropp, C., Menlo, G., Campbell, G., and Lidereau, R. Vol. 1, pp. 4.7.1-4.7.8. New York: Green Publishing Associates and Wiley Mutations in breast cancer. In: The Therapeutic Implications of the Interscience, 1991. Molecular Biology of Breast Cancer, pp. 57-67. Rome: John Libbey 17. Chomczinski, P., and Sacchi, N. Single step isolation of RNA by CIC, 1991. acid guanidium thiocyanate-phenol-chbonoform extraction. Ann. Bio- 28. Clark, C. C., Cohen, I., Eichstetter, I., Cannizzano, L. A., McPher- chem., 162: 321-328, 1987. son, J. D., Wasmuth, J. J., and lozzo, R. V. Molecular cloning of the 18. McCarthy, S., and Bicknell, R. Responses of pertussis toxin treated human proto-oncogene Wnt-SA and mapping of the gene (WNT5A) to microvasculan endotheliab cells to transforming growth factor 3. J. Biol. chromosome 3pl’4-p2l. Genomics, 18: 249-260, 1993. Chem., 267: 21617-21622, 1992. 29. Lin, T. P., Guzman, R. C., Osborn, R. C., Thordarson, G., and 19. Lejeune, S., Leek, R., Horak, E., Plowman, G., Greenall, M., and Nandi, S. Role of endocrine, autocnine, and paracnine interactions in the Harris, A. L. Amphiregulin, EGF receptor, and estrogen receptor in development of mammary hyperplasia in W,,t-1 tnansgenic mice. Cancer human breast cancer. Cancer Res. 53: 3597-3602, 1993. Res., 52: 4413-4419, 1992. 20. Senior, P. V., Cnitchley, D. R., Beck, F., Walker, F. A., and Varley, 30. Olson, D., and Papkoff, J. Regulated expression of Wnt family J. M. The localization of lamini mRNA and protein in the postimplan- members during proliferation of C57mg mammary cells. Cell Growth & tation embryo and placenta of the mouse: an in situ hybnidisation and immunocytochemical study. Development, 104: 431-446, 1988. Differ., 5: 197-206, 1994.

S Lejeune, E L Huguet, A Hamby, et al.

Clin Cancer Res 1995;1:215-222.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/1/2/215

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/1/2/215. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.