Oncogene (2007) 26, 298–307 & 2007 Nature Publishing Group All rights reserved 0950-9232/07 $30.00 www.nature.com/onc ONCOGENOMICS Correlated break at PARK2/FRA6E and loss of AF-6/Afadin expression are associated with poor outcome in breast cancer

A Letessier1,7, S Garrido-Urbani2,7, C Ginestier1, G Fournier2, B Esterni3,4, F Monville1, J Ade´ laı¨ de1, J Geneix1, L Xerri5,6, P Dubreuil2, P Viens4,6, E Charafe-Jauffret1,5,6, J Jacquemier1,5, D Birnbaum1, M Lopez2 and M Chaffanet1

1Centre de Recherche en Cance´rologie de Marseille, De´partement d’Oncologie Mole´culaire, UMR599 Inserm et Institut Paoli-Calmettes, Marseille, France; 2He´matopoie`se Fonctionnelle et Mole´culaire, UMR599 Inserm et Institut Paoli-Calmettes, Marseille, France; 3De´partement de Biostatistiques, Institut Paoli-Calmettes, Marseille, France; 4De´partement d’Oncologie Me´dicale, UMR599 Inserm et Institut Paoli-Calmettes, Marseille, France; 5De´partement de BioPathologie, Institut Paoli-Calmettes, Marseille, France and 6Faculte´ de Me´decine, Universite´ de la Me´diterrane´e, Marseille, France

Common fragile sites (CFSs) are regions of chromosomal 2001). Breaks at common fragile sites (CFS) may break that may play a role in oncogenesis. The most play a role in tumor initiation and/or progression frequent alteration occurs at FRA3B, within the FHIT (Huebner and Croce, 2001; Dhillon et al., , at chromosomal region 3p14. We studied a series of 2003; Popescu, 2003). Among them, the most frequently breast carcinomas for break of a CFS at 6q26, FRA6E, altered CFSs, FRA3B at 3p14.2 and FRA16D at and its associated gene PARK2, using fluorescence in situ 16q23.3, are encompassed by FHIT (Fragile Histidine hybridization on tissue microarrays (TMA). We found Triad) and WWOX (WW domain-containing oxydo- break of PARK2 in 6% of cases. We studied the PARK2- reductase) loci, respectively (Huebner et al., 1998; encoded protein by using immunohistochemistry Bednarek et al., 2000). FHIT and WWOX are on the same TMA. Loss of Parkin was found in 13% of both considered as tumor suppressor (TSG) samples but was not correlated with PARK2 break. (Bednarek et al., 2001; Paige et al., 2001; Huebner and PARK2 break but not Parkin expression was correlated Croce, 2003; Fabbri et al., 2005). Their alterations are with prognosis. Alteration of PARK2/FRA6E may cause found in various types of cancer and lead to the loss or haplo-insufficiency of one or several telomeric potential inactivation of their respective (Paige et al., tumor suppressor genes (TSG). The AF-6/MLLT4 gene, 2001; Huebner and Croce, 2003). We have previously telomeric of PARK2, encodes the Afadin scaffold protein, reported that the loss of FHIT expression is a marker of which is essential for epithelial integrity. Loss of Afadin adverse evolution in good prognosis localized breast was found in 14.5% of cases, and 36% of these cases cancer (Ginestier et al., 2003). FHIT and WWOX are showed PARK2 break. Loss of Afadin had prognostic coordinately inactivated in a subset of invasive breast impact, suggesting that AF-6 may be a TSG. Loss of cancers (Guler et al., 2005). Afadin was correlated with loss of FHIT expression, The third most frequent CFS, FRA6E, is located in suggesting fragility of FRA6E and FRA3B in a certain region 6q26 within approximately 3.6 Mb. proportion of breast tumors. The PARK2 gene encompasses the distal half of the Oncogene (2007) 26, 298–307. doi:10.1038/sj.onc.1209772; FRA6E locus. The most unstable region of FRA6E is published online 3 July 2006 localized between exons 2 and 8 of PARK2. The PARK2 gene encodes Parkin, a cytoplasmic E3 ubiquitin protein Keywords: Afadin; AF-6 gene; breast cancer; FRA6E; ligase, whose mutations cause autosomal-recessive Parkin; PARK2 gene; tissue microarrays juvenile Parkinsonism (Kitada et al., 1998; Marin et al., 2004). Frequent loss of heterozygosity (LOH) in ONCOGENOMICS introns 2 and 6 and downregulation of Parkin are found in ovarian tumors (Cesari et al., 2003; Denison et al., Introduction 2003b; Picchio et al., 2004; Wang et al., 2004), which suggests that PARK2 may be involved in oncogenesis as The human contain several regions called a TSG. fragile sites that are particularly susceptible to break The long arm of contains potential in response to environmental carcinogens (Richards, TSG involved in various types of cancer such as melanoma (Millikin et al., 1991), ovarian carcinoma Correspondence: Dr D Birnbaum, UMR599 Inserm, 27 Bd. Leı¨ Roure, (Saito et al., 1992; Tibiletti et al., 1996), breast 13009 Marseille, France. carcinoma (Orphanos et al., 1995; Noviello et al., E-mail: [email protected] 7These authors contributed equally to this work. 1996), non-Hodgkin’s lymphoma (Menasce et al., Received 21 December 2005; revised 27 April 2006; accepted 22 May 1994a) and acute leukemia (Hayashi et al., 1990; 2006; published online 3 July 2006 Menasce et al., 1994b). Tumorigenicity of breast cell PARK2/FRA6E and Afadin in breast cancer A Letessier et al 299 lines can be suppressed by microcell-mediated transfer locus and a chromosomal derivative detected by red of regions 6q21–q23 and/or 6q26–q27 suggesting that signals without green signals. This indicated that the these regions may indeed contain TSG (Negrini et al., 50 region of PARK2 (green signal) was lost (Figure 1Bc). 1994). LOH and comparative genomic hybridization The second type of alteration, observed in only one case experiments have shown that the 6q26-qter region is (a ductal carcinoma), displayed a wild-type locus and a frequently altered in breast cancer (Noviello et al., 1996; chromosomal derivative detected by green signals with- Kerangueven et al., 1997; Rodriguez et al., 2000; out red signal, indicating that the 30 region of PARK2 Teixeira et al., 2002). was lost (Figure 1Bd). The 6q26-qter region contains several genes including PARK2, FOP, TTLL2, AF-6/MLLT4, KIF25, THBS2 Correlation of PARK2 break with histoclinical factors and TBP, potentially involved in cancer (Prasad et al., and clinical outcome 1993; Taki et al., 1996; Whitcomb et al., 2003; Denison We next examined the relation between PARK2 break et al., 2003a; Guasch et al., 2004; Agirre et al., 2005). and histoclinical factors. We did not take into account Among these genes, AF-6/MLLT4, located telomeric of the monosomy cases but only the cases with break. PARK2, encodes the Afadin protein, which is crucial for PARK2 break was not associated with any histoclinical epithelial physiology and development (Ikeda et al., factor, including age, histological type, pathological 1999; Zhadanov et al., 1999). Afadin is a scaffold tumor size, SBR grade, peritumoral vascular invasion, protein located at adherens and tight junctions (Mandai axillary lymph node status, and ER, PR, ERBB2, P53 et al., 1997). It participates to the establishment and the and Ki67 expression (Table 1). maintenance of epithelial polarity, a process disrupted We examined if the PARK2 break had an impact on during oncogenesis. prognosis. When we considered the whole population of We report here an analysis of PARK2/FRA6E, Parkin patients, the 5-year metastasis-free survival (MFS) was and Afadin in breast cancers. We first established the 61.4% (range 37.7–99.9) for patients with a tumor frequency of PARK2 break at FRA6E by using showing a break at PARK2, and 79.9% (73.9–86.5) for fluorescence in situ hybridization (FISH) on tissue patients with a wild-type PARK2 tumor (P ¼ 0.0216) microarrays (TMA). The break of PARK2, but not the (Figure 2). PARK2 break was associated with decreased loss of expression of Parkin, had an impact on disease 5-year MFS in patients with breast cancer. outcome. We then showed indirectly that the PARK2 break could affect the telomeric gene AF-6, and the expression of its encoded protein Afadin. Loss of Parkin expression in breast carcinomas expression of Afadin measured by immunohistochem- To determine if the break of the PARK2 gene affected istry (IHC) was associated to poor outcome and seemed Parkin expression, we studied expression of the protein to be a good marker of metastasis appearance in the by IHC on the same TMA. Parkin was strongly population of patients without axillary lymph node expressed in the cytoplasm of the epithelial cells of invasion. Finally, we showed that loss of Afadin normal breast tissue (Figure 3Aa). Of the 547 tumors, expression after RNA silencing experiments affects 473 immunostained cases (86%) were available for cell–cell contacts. Our data suggest that AF-6 may be quick score analysis (Supplementary Table 1). In 412 a TSG in breast cancer whose loss is a marker of adverse tumors (87%), Parkin was expressed with a level of prognosis. expression from low to fully positive (Figure 3Ab). In 61 tumors (13%), no Parkin expression was found (Q ¼ 0) (Figure 3Ac). None of the tumors with break of PARK2 Results showed loss of Parkin expression. Other mechanisms, such as mutations, intragenic deletions and epigenetic Characterization of break in the PARK2 gene in modifications, may explain loss of expression in the breast carcinoma tumors without PARK2 break. We searched for alteration of the PARK2 locus at We examined the impact of the loss of Parkin FRA6E in 547 breast tumors by using FISH on TMA expression on the 5-year MFS. The 5-year MFS for sections. Biotinylated and digoxigenin-labeled sequences patients with a Parkin-negative tumor was 75.7% (65.4– used as probes in the 50 and 30 regions of PARK2 87.7), and 79.1% (75.1–83.3) for patients with a Parkin- (Figure 1a) were seen as green or red fluorescent signals, positive tumor (P ¼ 0.994). Thus, in contrast to the respectively. Of the 547 cases, only 190 (35%) gave PARK2 break, the loss of Parkin expression had no reliable results (Supplementary Table 1). Loss of data impact on prognosis. resulted either from the loss of the sample during stringent FISH pretreatment or from high background. Afadin expression in breast carcinomas Of these 190 samples, 177 (94%) showed integrity of the As there was no correlation between PARK2 gene status PARK2 locus revealed by clustered green/red signals and Parkin expression, we studied the consequences of a (Figure 1Ba), two showed monosomy for this locus PARK2 break on the expression of genes located in the (Figure 1Bb), and 11 (6%) showed break of the locus 6q26-qter region telomeric of PARK2. Among the genes (Figure 1Bc and Bd). Two types of PARK2 break were located in this region (FOP, TTLL2, AF-6, KIF25, found. The first type, observed in 10 cases (nine ductal THBS2, TBP), AF-6/MLLT4 is located in a region of and one lobular carcinomas), displayed a wild-type frequent LOH and displays frequent deletions in breast

Oncogene PARK2/FRA6E and Afadin in breast cancer A Letessier et al 300

Figure 1 Break of the PARK2 locus in breast cancer. (A) Map of the locus. (a) Ideogram of chromosome 6 showing the PARK2 locus at 6q26 (chr6:161 740 081–163 119 211) and the AF-6/MLLT4 locus at 6q27 (chr6:168 046 404–168 183 532). (b) The two pools of BAC clones used as probes in FISH experiments. To detect a break of PARK2, tumors present on a TMA were hybridized using biotinylated and dig-labeled sequences in the 50 (green) and 30 (red) regions of PARK2 revealed after detection in green and red, respectively. (c) Enlarged PARK2 locus showing exon–intron organization and the two overlapping BAC clones used as probes. (B) Examples of PARK2 status in breast cancer determined by FISH on TMA. (a) Tumor with intact PARK2 gene: two wild-type copies of PARK2 are seen as clusters of green and red signals (arrows). (b) Tumor with monosomy of PARK2 gene: only one wild-type copy is seen and revealed by clustered green and red signals (arrow). (c) Tumor with break of PARK2 gene seen as a split of the clustered signals and disappearance of one green signal (arrow), suggesting the loss of the 50 region of the locus. (d) Tumor with break of PARK2 gene seen as a split of the clustered signals and disappearance of one red signal (arrow), suggesting the loss of the 30 region of the locus.

cancers. Moreover, AF-6 encoded product, Afadin, is with high pathological size (P ¼ 0.05389). Interestingly, expressed in normal breast epithelium and is involved in 36% of Afadin-negative tumors (four cases) had break epithelial cell architecture. We thus focused our atten- of PARK2, whereas only 5% of Afadin-positive tumors tion on Afadin. We first validated the two different anti- (seven cases) had break of PARK2 (Po0.001). Thus, Afadin antibodies (Table 2) on mammary cell lines by loss of Afadin was associated with a PARK2 break. Western blot and IHC (data not shown). Other pathological mechanisms, such as mutations, We investigated the expression of Afadin in breast intragenic deletions and epigenetic modifications, may tumors by IHC on TMA sections. Afadin was strongly explain loss of expression in the 64% of Afadin-negative expressed in the cytoplasm and at the cellular membrane tumors without break of PARK2. of the epithelial cells of normal breast (Figure 3Ba). As These results led us to examine the impact of Afadin measured by the quick score (QS), two distinct levels expression on clinical outcome. When we considered of Afadin expression were observed in the 352 tumors the whole population of patients analysed for Afadin with reliable results (Supplementary Table 1). In 301 expression (N ¼ 352), the 5-year MFS was 67.8% tumors (85.5%), the level of Afadin was similar to that (55.3–82.9) for patients with an Afadin-negative tumor, of normal breast tissue (Figure 3Bb). In 51 tumors and 81.8% (77.4–86.5) for patients with an Afadin- (14.5%), there was no Afadin expression (QS ¼ 0) positive tumor (P ¼ 0.046) (Figure 4a). When we (Figure 3Bc). considered the lymph node-negative population of patients (N ¼ 181), the 5-year MFS was 72.8% (56.1– Loss of Afadin expression correlates with PARK2 break 94.5) for patients with an Afadin-negative tumor, and and prognosis 89.5% (84.5–94.7) for patients with an Afadin-positive Loss of Afadin did not correlate with any of the studied tumor (P ¼ 0.0496) (Figure 4b). Loss of Afadin was thus histoclinical factors (Table 3). It tended to be associated associated with poor outcome in this group of patients.

Oncogene PARK2/FRA6E and Afadin in breast cancer A Letessier et al 301 PARK2 Table 1 Correlation between the break of determined by 1.0 FISH and histoclinical factors

Characteristics No. of patients (%) P-value No PARK2 break 0.8 Cases with Cases without break of PARK2 break of PARK2 (N ¼ 11) 6% (N ¼ 177) 94% 0.6

Age (years) PARK2 break o50 6 (55) 49 (28) NS 0.4 X50 5 (45) 128 (72) free survival (MFS)

Histological type Proportion of Metastasis 0.2 Ductal 10 (91) 138 (78) p= 0.0216 Lobular 1 (9) 17(10) NS Other 0 (0) 22 (12) 0.0

Pathological tumor size 020 40 60 80 100 120 pT1 5 (45) 70 (40) Months after surgery pT2 6 (55) 85 (48) NS Figure 2 PARK2 status and associated survival in breast cancer. pT3 0 (0) 22 (12) Impact of the break of PARK2 on MFS of the whole population analysed for PARK2 break (N ¼ 188). Kaplan–Meier curves SBR grade illustrate MFS according to the status of PARK2. I 2 (18) 51 (29) II 6 (55) 78 (44) NS III 3 (27) 48 (27) Loss of Afadin expression correlates with loss of FHIT Peritumoral vascular invasion Absent 8 (73) 108 (61) NS expression Present 3 (27) 69 (39) We next looked for variation of expression of FHIT on the same TMA (data not shown). We found a loss of Axillary lymph node status expression of both FHIT and Afadin in 26% of cases Negative 5 (45) 95 (54) NS (Table 3). The loss of these two proteins was correlated Positive 6 (55) 81 (46) (P ¼ 0.01). Patients with an Afadin-negative/FHIT- Estrogen receptor status negative tumor showed a poor prognosis as compared Negative 0 (0) 40 (23) NS to patients with an Afadin-positive/FHIT-positive Positive 10 (100) 136 (77) tumor (P ¼ 0.0148) (data not shown). This result Progesterone receptor status suggests that the determination of Afadin expression Negative 2 (18) 61 (35) NS in patients with a loss of FHIT protein expression, Positive 9 (82) 112 (65) which is a marker of adverse evolution in good prognosis localized breast cancer (Ginestier et al., ERBB2 status 2003), may help distinguish a subpopulation with a 0–1 8 (89) 146 (86) NS 2–3 1 (11) 24 (14) poorer outcome.

P53 status Loss of Afadin expression affects cell–cell junctions Negative 9 (82) 126 (74) NS To document the potential role of Afadin in breast Positive 2 (18) 44 (26) cancer we analysed the effect of the loss of Afadin Ki67 status expression on cell behavior. We abolished Afadin o20 9 (100) 121 (76) NS expression in MDCK II epithelial cells using artificial X20 0 (0) 38 (24) microRNA (miRNA). We designed these miRNA to have 100% identity with a sequence of Afadin conserved Abbreviations: FISH, fluorescence in situ hybridization; NS, not sufficient. between human, simian, canine and rodent. Cocistronic expression of emerald green fluorescence protein (EmGFP) and the miRNA in a pcDNA-based vector (pCDmiR-Afadin) allowed the correlation of GFP level We did a Cox multivariate analysis of MFS in and Afadin expression. EmGFP miR-Afadin was used which the values for Afadin, tumor size, age, grade, in Cos cells and showed a marked reduction of peritumoral vascular invasion, ER, PR and Ki67 endogenous Afadin expression (Figure 5a, top). This were considered as categorical variables. Afadin expres- knockdown was specific. Irrelevant controls (EmGFP- sion remained significant as well as Ki67 status miR-neg and EmGFP-miR-LacZ) did not change and grade according to the Akaike Information Afadin level and no miRNA affect the expression of criterium when dichotomized negative vs positive, unrelated protein p85 (Figure 5a, bottom). Analysis of o20 vs X20 and I vs III, respectively (Table 4). confluent MDCKII cells stably transfected with the The relative risk of recurrence was 2.96 for Afadin- pCDmiRAF6-3481 vector showed a marked Afadin negative disease compared to Afadin-positive disease knockdown only in cells with high EmGFP levels (P ¼ 0.028). (Figure 5b). The absence of Afadin expression was

Oncogene PARK2/FRA6E and Afadin in breast cancer A Letessier et al 302

Figure 3 Parkin and Afadin expression in breast cancer. (A) Parkin expression in normal breast and breast tumors determined by IHC on TMA. (a) Normal breast tissue with Parkin expression; (b) Tumor with preserved Parkin expression (arrow) and (c) Tumor showing loss of Parkin expression. (B) Afadin expression in normal breast and breast tumors determined by IHC on TMA: (a) Normal breast tissue with Afadin expression (arrow); (b) Tumor with Afadin expression (arrow) and (c) Tumor showing loss of Afadin expression in tumoral component (arrow) and Afadin expression in normal tissue.

Table 2 List of proteins tested by and characteristics of the corresponding antibodies Protein Antibody1 Origin Clone Dilution

Estrogen receptor mmab Novocastra Laboratories 6F11 1/60 Progesterone receptor mmab DakoCytomation PgR 636 1/80 ERBB2 rpab DakoCytomation HercepTest 1/400 P53 mmab Immunotech DO-1 1/4 MIB1 mmab DakoCytomation Ki67 1/100 Parkin rpab Neomarkers RB9215 1/30 Afadina mmab Transduction Laboratories 35 1/50 Afadin rmab HyCult Biotechnology b.v 3 1/50 FHIT rpab Zymed Laboratories ZR44 1/300

Abbreviations: mmab, mouse monoclonal antibody; rpab, rabbit polyclonal antibody; rmab, rat monoclonal antibody; IHC, immunohisto- chemistry; TMA, tissue microarrays. aThis Afadin antibody was used in IHC on TMA.

associated with a profound disorganization of epithelial CFSs are highly unstable genomic regions. They cell–cell contacts confirming the fundamental role of could predispose chromosomes to break, generate Afadin in epithelial physiology previously suggested by chromosomal rearrangements in cancer cells and the studies of knockout mouse models (Ikeda et al., play a role in tumor initiation and/or progression 1999; Zhadanov et al., 1999). (Huebner et al., 1998; Smith et al., 1998; Sutherland et al., 1998). The cloning and the characterization of FRA6E at 6q26 identified eight genes associated Discussion with this fragile site. Among them, PARK2 contains the most unstable region of FRA6E between its exons Breast cancer is an heterogeneous cancer with 2–8 (Denison et al., 2003a). PARK2 is a large gene that multiple forms and distinct entities (Sorlie et al., is mutated in patients with autosomal recessive juvenile 2003; Charafe-Jauffret et al., 2005). A better under- Parkinsonism and spans the telomeric half of FRA6E. standing of the molecular basis of this heterogeneity PARK2 shows similarities with the two most active should allow a better management of the disease. CFS-associated genes, FHIT and WWOX. FHIT at A possible molecular substratum for tumor develop- FRA3B (3p14) and WWOX at FRA16D (16q23) are ment and heterogeneity is genetic instability. The both large genes. They both suppress tumor cell growth identification of a distinct subclass of breast cancer with in vitro and in vivo and have been classified as TSG genomic fragility may help understand disease hetero- (Bednarek et al., 2001; Paige et al., 2001; Roz et al., geneity. This subclass could be recognized by alterations 2002; Huebner and Croce, 2003; Ishii et al., 2003; Fabbri at CFS. et al., 2005).

Oncogene PARK2/FRA6E and Afadin in breast cancer A Letessier et al 303 Table 3 Correlation between Afadin expression determined by IHC and histoclinical factors Characteristics No. of patients (%) P-value

Afadin negative Afadin positive (N ¼ 51) 14.5% (N ¼ 301) 85.5%

Age, years o50 17 (33) 76 (25) NS X50 34 (67) 225 (75)

Histological type Ductal 40 (78) 218 (72) Lobular 5 (10) 32 (11) NS Other 6 (12) 51 (17)

Pathological tumor size pT1 15 (29) 137 (46) pT2 25 (49) 127 (42) 0.05389 pT3 11 (22) 37 (12) NS

SBR grade I 13 (26) 99 (33) II 20 (39) 118 (39) NS III 18 (35) 84 (28)

Peritumoral vascular invasion Absent 31 (62) 190 (63) NS Present 19 (38) 111 (37)

Axillary lymph node status Negative 25 (50) 156 (52) NS Positive 25 (50) 144 (48)

Estrogen receptor status Negative 14 (28) 66 (22) NS Figure 4 Afadin status and associated survival in breast cancer. Positive 36 (72) 232 (78) (a) Impact of the loss of Afadin expression on MFS of the whole population analysed for Afadin expression (N ¼ 352). (b) Impact of Progesterone receptor status the loss of Afadin expression on MFS of patients with absence of Negative 19 (37) 97 (33) NS axillary lymph node invasion (N ¼ 181). Kaplan–Meier curves Positive 32 (63) 201 (67) illustrate MFS according to Afadin expression. ERBB2 status 0–1 41 (85) 245 (87) NS 2–3 7 (15) 36 (13) Table 4 Cox multivariate analysis of metastasis-free survival for P53 status patients without axillary lymph node invasion Negative 38 (76) 224 (74) NS Variable Coefficient value Hazard ratio 95% CI P-value Positive 12 (24) 77 (26) Grade Ki67 status I 1 0.012 o20 40 (85) 223 (78) NS III 1.173 3.233 (1.291–8.096) X20 7 (15) 64 (22) Ki67 PARK2 break status o20 1 0.018 Yes 4 (36) 7 (5) 0.001 X20 1.127 3.087 (1.218–7.826) No 7 (64) 130 (95) Afadin FHIT status Negative 1 0.028 Negative 12 (26) 31 (11) 0.01 Positive À1.085 0.3378 (0.1283–0.8891) Positive 34 (74) 246 (89)

expression, and the presence of aberrant transcripts PARK2 encodes the Parkin E3 ubiquitin ligase and is and occurrence of genomic deletions have been observed altered in various tumors. It is also considered a in malignant samples (Denison et al., 2003a). The same potential TSG (Cesari et al., 2003; Denison et al., observations have been made in hepatocellular carcino- 2003a; Wang et al., 2004). PARK2 alterations were mas (Wang et al., 2004) and non-small-cell lung cancer initially reported in breast and ovarian cell lines and (Picchio et al., 2004). tumors (Cesari et al., 2003; Denison et al., 2003a, b). The We first determined the status of FRA6E breaks in absence of normal PARK2 transcript and Parkin breast tumors by using FISH on TMA. We measured

Oncogene PARK2/FRA6E and Afadin in breast cancer A Letessier et al 304 a subsequent effect on a 6q gene, telomeric of PARK2. fadin Genes located telomeric of FRA6E may be potential -neg -A -LacZ iR iR iR breast cancer genes. Frequent LOH are observed in this -m -m FP FP-m FP region in breast cancer (Orphanos et al., 1995; Noviello G G G ot transfectedm m m N E E E et al., 1996; Cesari et al., 2003). The tumorigenicity of breast cell lines can be suppressed by microcell-mediated

Afadin transfer of a part of human chromosome 6 (Negrini 175 - et al., 1994). The 6q26–6q27 region contains several genes including FOP, TTLL2, AF-6/MLLT4, KIF25, THBS2, and TBP. FOP encodes a centrosomal protein 83 - that is fused to FGFR1 kinase in myeloproliferative p85 disorders (Popovici et al., 1999 ; Delaval et al., 2005). It 62 - could be interesting to examine if this gene could play a role in oncogenesis out of its fusion with FGFR1. b TTLL2 (Tubulin Tyrosine Ligase-like 2) encodes a member of the TTL homology domain protein family, which catalyses the ligation of glutamic acid to tubulin. In neuronal systems, tubulin polyglutamination could regulate the organization of microtubule network (Bonnet et al., 2001) thus controlling centriole stability and mitosis (Bobinnec et al., 1998a, b). KIF25 encodes a protein of the kinesin superfamily, KIF25, which is involved in molecular transport away from the centro- some (Miki et al., 2005). No role in neoplasia has been reported yet for KIF25. The promoter of THBS2 (thrombospondin 2) gene is methylated (62.5%) in primary endometrial carcinoma (Whitcomb et al., 2003). High THBS2 expression may be associated with an angiogenic phenotype in endometrial cancer and Figure 5 Afadin knockdown in MDCKII cells leads to the THBS2 expression is a marker of poor prognosis in this destabilization of cell–cell junctions. (a) Different miRNA expres- disease (Seki et al., 2001). The TATA-binding protein sion vectors were tested in Cos cells. Two different miRNA expression plasmids (EmGFP-miR-neg and EmGFP-miR-LacZ) encoded by TBP is associated with transcriptional were used as irrelevant controls to assess the specificity of the cellular systems. Modulation of TBP concentration has EmGFP-miR-Afadin plasmid. Cos lysates 10 mg were separated on an impact on that can mediate potential SDS–PAGE then immunoblotted with Afadin mAb (top) and p85 cell transformation (Johnson et al., 2003a, b). to control loading (bottom). (b) MDCKII cells stably expressing AF-6/MLLT4 encodes Afadin, which is involved in the EmGFP-miR-Afadin plasmid. Afadin is localized at cell–cell junctions in confluent MDCKII cells as previously described epithelial physiology. It is ubiquitously expressed in (arrowheads). High EmGFP expression correlates with a marked normal epithelial cells, where it localizes at adherens and knockdown of Afadin expression, the destabilization of cell–cell tight junctions (Mandai et al., 1997). Afadin is a scaffold junctions and the extinction of Afadin signal at cell–cell junctions protein that links adhesion proteins, cellular receptors (arrows). and signaling effectors to the actin cytoskeleton (Mandai et al., 1997; Buchert et al., 1999). Mice lacking the frequency of PARK2 break and its potential impact the Af-6 gene die at 10 days post coitum of placenta on clinical outcome. Break of one allele of PARK2 was failure (Ikeda et al., 1999; Zhadanov et al., 1999). AF-6 observed in 6% of tumors. It correlated with decreased could play a key role in the development of carcinomas. 5-year MFS. Loss of Parkin expression was observed in AF-6 is fused to the MLL gene in the t(6;11)(q27;q23) around 13% of cases. Similarly, Parkin expression is chromosomal translocation, which is the most common decreased in a large proportion of ovarian tumors translocation found in acute lymphoid leukemia (Prasad (Cesari et al., 2003; Denison et al., 2003a). However, we et al., 1993). We thus chose to study Afadin because of found that loss of Parkin was not correlated with a its role in epithelial physiology and potential involve- break of PARK2. This was in sharp contrast to the good ment in cancer, but also for technical reasons. The two correlation observed between alterations at FRA3E and anti-Afadin antibodies work well in IHC on paraffin FRA16D and FHIT and WWOX protein levels, embedded tissues whereas no appropriate antibody is respectively (Ginestier et al., 2003; Park et al., 2004). available for the other proteins encoded by the FRA6E- The absence of correlation between FRA6E break and telomeric genes. Parkin expression suggests that other mechanisms are The Afadin status in breast tumors was addressed by responsible for abnormal expression of Parkin. Abnor- using IHC on TMA. Complete loss of Afadin was mal methylation may be one of these mechanisms observed in 14.5% of tumors. Loss was correlated with (Agirre et al., 2005). the break of PARK2 and with a bad outcome for We hypothesized that the consequences of the PARK2 patients without lymph node invasion. We propose that break on the clinical outcome may be due to a loss of Afadin expression can be due to a break of

Oncogene PARK2/FRA6E and Afadin in breast cancer A Letessier et al 305 FRA6E/PARK2. The multivariate analysis showed that by IHC with the 0–3 þ score as illustrated by the HercepTest Afadin could be a good prognosis marker. In associa- kit scoring guidelines (DakoCytomation, Coppenhagen, tion with grade and Ki67 status, Afadin status may help Denmark), and Ki67 status as evaluated by IHC with a in the detection of patients with poor prognosis in the positive cutoff value at 20%. lymph node-negative population. Moreover, in the FHIT-negative patient population, Afadin is a marker TMA construction of poor prognosis. The combined analysis of FHIT and TMA were prepared as described previously (Ginestier et al., Afadin could be useful to discriminate patients with 2002). For each tumor, three representative tumor areas were carefully selected from a hematoxylin-eosin-safran stained adverse outcome in the whole population. The coordi- section of a donor block. Core cylinders with a diameter of nated loss of FHIT and WWOX expression has been 0.6 mm each were punched from each of these areas and found in breast cancers (Guler et al., 2005). We have deposited into three separate recipient paraffin blocks using a similarly shown here that loss of break of PARK2 and specific arraying device (Beecher Instruments, Silver Spring, loss of FHIT expression are concomitantly found in MD, USA). In addition to tumor tissues, the recipient block some samples. These cases are associated with a poor also received normal breast tissues and cell lines pellets. prognosis. We have not tested WWOX expression on Sections 5-mm of the resulting microarrays block were our series. made and used for FISH and IHC analysis after transfer onto Finally, we showed that loss of Afadin expression glass slides. affects adherence of cells in culture. This is in perfect agreement with in vivo data; cell–cell adherens Fluorescence in situ hybridization analysis and tight junctions are improperly organized in the FISH on TMA was carried out according to a published ectoderm of Af-6 (À/À) mice and embryoid bodies protocol (Chin et al., 2003; Huang et al., 2004). Based on the (Ikeda et al., 1999; Zhadanov et al., 1999). A thorough split-signal FISH approach (van der Burg et al., 2004), we used a combination of two differently labeled pools of BAC clones study has recently described the role of Afadin in the overlapping the PARK2 locus as probes (Figure 1a): from recruitment of E-cadherin and tight junction compo- telomere to centromere, biotinylated RP11-157B17 (chr6: nents at cell–cell junctions (Sato et al., 2006). Like 163,481,350-163,680,324), RP11-117I16 (AC058815; chr6: E-cadherin, Afadin expression may be lost in a 163,177,577- 163,342,875), RP11-153I8 (chr6: 162,898,135- subgroup of breast cancers. 163,041,573) (revealed in green, FITC) and digoxigenin-labeled In conclusion, our data suggest that: (i) frequent RP11-431E19 (chr6: 161,659,375- 161,837,993), RP11-479C23 breaks in tumors at a CFS should not automatically (chr6: 161,442,754- 161,620,713), RP11-158E9 (chr6: point to an intrasite gene as involved in cancer; breaks 161,336,365-161,484,919) (revealed in red, TRITC). PARK2 is may induce loss of nearby TSG; (ii) Afadin may be a located on chromosome arm 6q, in the 161,740,081-163,119,211 new marker of adverse evolution in patients with genomic interval. RP11-153I8 and RP11-431E19 were the two overlapping BAC clones of PARK2 used in this combination. apparent good prognosis at diagnosis; (iii) breast tumors They overlap PARK2 on 221kb and 98kb, respectively with concomitant loss of FRA6E break, Afadin loss and (Figure 1a). Genomic information was taken from the UCSC FHIT loss may constitute a subclass with increased Genome Browser on Human (http://genome.ucsc.edu – May genomic fragility; (iv) Afadin may have a role in 2004 Assembly), which is based on NCBI Build 35 (National mammary oncogenesis. It acts as a tumor suppressor Center for Biotechnology Information, National Library of whose loss of expression disrupts epithelial integrity and Medicine, Bethesda, USA). may favor metastasis. DNA from BAC clones were purified, labeled and indivi- dually verified for their specificity for chromosome 6. All BAC clones were obtained from the BACPAC resource (Children’s Patients and methods Hospital Oakland – BACPAC Resources, Oakland, CA, USA). After counterstaining with Vectashield containing 4,6- Patients and histological samples diamidino-2-phenylindole (DAPI) (Vector, Burlingame, CA, A consecutive series of 547 unilateral localized invasive breast USA), images were analysed with a microscope (DMRXA, carcinomas from women treated at the Institut Paoli- Leica Microsystems, Marseille, France), captured with a CCD Calmettes between October 1987 and December 1999 was camera, filtered and processed with ISIS software (In Situ studied. According to the WHO classification, this series Imaging Systems, Metasystems Hard- und Software GmbH, comprised 386 ductal, 72 lobular, 37 tubular, 8 medullary Altlussheim, Germany) (described in www.metasystems.de). carcinomas and 44 other histological types. The average age at Fluorescence was scored on a minimum of 50 nuclei per tumor. diagnosis was 59 years (range 25–94 years). A total of 254 The 50 nuclei of cancer cells were representative of the overall tumors were associated with lymph node invasion and 403 cell heterogeneity of the tumor. Two observers (AL and CG) were positive for estrogen receptor. Of the 547 cases, 190 cases read the TMA independently. were available for FISH analysis, and 473 and 352 were available for Parkin and Afadin immunostaining, respectively Immunohistochemical analysis (Supplementary Table 1). The characteristics of the antibodies used are listed The various histoclinical factors collected for this series in Table 1. IHC was carried out on five-mm sections of tissue included: patient age, invasive histological type, pathological fixed in alcohol formalin for 24 h and included in paraffin. tumor size, Scarff–Bloom–Richardson (SBR) grade (I–III), Sections were deparaffinized in Histolemon (Carlo Erba peritumoral vascular invasion, axillary lymph node status, Reagenti, Rodano, Italy) and rehydrated in graded alcohol estrogen receptor expression (ER), progesterone receptor solution. Antigen enhancement was carried out by incubating expression (PR), P53, Parkin, Afadin, as evaluated by IHC the sections in target retrieval solution (DakoCytomation, with a positivity cutoff value of 1%, ERBB2 status, evaluated Coppenhagen, Denmark) as recommended. The reactions were

Oncogene PARK2/FRA6E and Afadin in breast cancer A Letessier et al 306 carried out using an automatic stainer (Dako Autostainer, semidry transferred to polyvinylidene difluoride membranes Copenhagen, Denmark). Staining was carried out at room (Immobilon-P, Millipore, Boston, MA, USA), probed with the temperature as follows: after washes in phosphate buffer, indicated antibody and visualized with ECL (Amersham followed by quenching of endogenous peroxidase activity by Pharmacia Biotech, Uppsala, Sweden). treatment with 0.1% H2O2, slides were first incubated with blocking serum (DakoCytomation) for 30 min and then with the affinity-purified antibody for 1 h. After washes, slides were Immunofluorescence studies incubated with biotinylated antibody against rabbit immuno- MDCK II cells were cultured on glass coverslips globulin for 20 min followed by streptadivin-conjugated (VWR, West Chester, PA, USA) in Dulbecco’s modified peroxidase (DakoCytomation LSABR2 kit). Diaminobenzidine Eagle’s medium supplemented with 10% fetal calf serum or 3-amino-9-ethylcarbazole was used as the chromogen. (FCS) until confluence. Cells were then fixed in 4% parafor- Slides were counter-stained with hematoxylin, and cover- maldehyde in phosphate-buffered saline (PBS) for 30 min, slipped using Aquatex (Merck, Darmstadt, Germany) mount- permeabilized in PBS 0.1% Triton X-100 for 2 Â 5 min and ing solution. Results were evaluated under a light microscope blocked with PBS 1% bovine serum albumin (Euromedex, by two pathologists (EC-J, JJ) and scored by the Souffelweyersheim, France) for 30 min. Then the cells were quick score (QS) as previously performed (Ginestier et al., stained with anti-Afadin mAb. Between each incubation, 2002), except for Ki67 status which was expressed in terms of coverslips were washed with PBS containing 0.1 mM CaCl2 percentage of positive cells, and ERBB2 status, which was and 1 mM MgCl2. Coverslips were mounted on slides with evaluated with the Dako scale (HercepTest kit scoring Prolong Gold (Invitrogen). Images were recorded with an Axio guidelines, DakoCytomation, Coppenhagen, Denmark). For Zeiss LSM 510 Meta confocal microscope. each tumor, the mean of the score of a minimum of two core biopsies was calculated. Statistical methods Descriptive data were summarized by frequency and percen- RNA silencing and cell transfection TM tage for categorical variables and by means, median and range RNA silencing was performed using the BLOCK-iT Pol II for continuous variables. Associations between molecular RNAi expression vector kit as recommended by the manu- markers and other categorical variables were examined using facturer (Invitrogen, Carlsbad, CA, USA). Artificial Afadin w2 analysis, or Fisher’s exact test for small sample sizes. The microRNA (miRNA) was cloned in the pcDNA 6.2-GW/ metastasis-free interval was calculated from the date of EmGFP-miR leading to a cocistronic expression of Emerald diagnosis. The nonmetastatic patients are rightcensored at GFP (EmGFP) with the miRNA of Afadin. Different the last follow-up visit. MFS curves were estimated by the sequences of Afadin miRNA were designed using an algorithm Kaplan–Meier method using the first metastatic recurrence as developed by Invitrogen. Sense and antisense DNA sequences first event definition, and the curves were compared by the log were: AF6_3481S: TGCT GAGG ACTA GGAG GCTG rank test. All the tests where two-sided and a P-value of less ATTT GCGT TTTG GCCA CTGA CTGA CGCA AATC than 0.05 was considered statistically significant. For graphical ACTC CTAG TCCT, and AF6_3481AS: CCTG AGGA representation, follow-up was truncated at 120 months. CTAG GAGT GATT TGCG TCAG TCAG TGGC CAAA Multivariate analyses for response were performed using Cox’s ACGC AAAT CAGC CTCC TAGT CCTC, respectively. This proportional hazards regression model using a backward sequence is located at residue 3481 in AF-6 cDNA and is stepwise selection procedure to evaluate the effect of inter- present in the different Afadin isoforms. MDCK II cells (50% action between the different variables. confluent) were transfected with 10 mg of the indicated vectors and selection of transfected cells were selected by 5 mg/ml blasticidin (Invitrogen). Acknowledgements

Western blot analysis We thank F Birg and D Maraninchi for encouragements, Cos cells were lyzed in ice-cold lysis buffer containing 50 mM C Chabannon for biobank sample management and JM Durey Hepes, pH 7.5, 150 mM NaCl, 1.5 mM MgCl2,1mM EGTA, for his help with iconography. This work was supported by 1% Triton X-100 and 10% glycerol. A protease inhibitor Institut Paoli-Calmettes, INSERM, and grants from Ligue mixture was added as recommended by the manufacturer Nationale Contre le Cancer (LNCC) (Label 2003-2006), (Roche Diagnostics, Meylan, France). the Association pour la Recherche contre le Cancer Ten mg of cell lysate was heated in sodium dodecyl sulfate (ARC-3128) and Ministries of Health and Research (Cance´ ro- (SDS) sample buffer (60 mM Tris-HCl, pH 6.7, 3% SDS, poˆ le PACA). SGU, CG and FM are supported by a fellowship 2% (v/v) 2-mercaptoethanol, and 5% glycerol) separated from Ministry of Research and AL by a fellowship from by 7.5% SDS–polyacrylamide gel electrophoresis (PAGE), LNCC.

References

Agirre X, Roman-Gomez J, Vazquez I, Jimenez-Velasco A, Bobinnec Y, Moudjou M, Fouquet JP, Desbruyeres E, Garate L, Montiel-Duarte C et al. (2005). Int J Cancer 118: Edde B, Bornens M. (1998b). Cell Motil Cytoskeleton 39: 1945–1953. 223–232. Bednarek AK, Keck-Waggoner CL, Daniel RL, Laflin KJ, Bonnet C, Boucher D, Lazereg S, Pedrotti B, Islam K, Bergsagel PL, Kiguchi K et al. (2001). Cancer Res 61: Denoulet P et al. (2001). J Biol Chem 276: 12839–12848. 8068–8073. Buchert M, Schneider S, Meskenaite V, Adams MT, Canaani Bednarek AK, Laflin KJ, Daniel RL, Liao Q, Hawkins KA, E, Baechi T et al. (1999). J Cell Biol 144: 361–371. Aldaz CM. (2000). Cancer Res 60: 2140–2145. Cesari R, Martin ES, Calin GA, Pentimalli F, Bichi R, Bobinnec Y, Khodjakov A, Mir LM, Rieder CL, Edde B, McAdams H et al. (2003). Proc Natl Acad Sci USA 100: Bornens M. (1998a). J Cell Biol 143: 1575–1589. 5956–5961.

Oncogene PARK2/FRA6E and Afadin in breast cancer A Letessier et al 307 Charafe-Jauffret E, Ginestier C, Monville F, Fekairi S, Miki H, Okada Y, Hirokawa N. (2005). Trends Cell Biol 15: Jacquemier J, Birnbaum D et al. (2005). Int J Oncol 27: 467–476. 1307–1313. Millikin D, Meese E, Vogelstein B, Witkowski C, Trent J. Chin SF, Daigo Y, Huang HE, Iyer NG, Callagy G, Kranjac T (1991). Cancer Res 51: 5449–5453. et al. (2003). Mol Pathol 56: 275–279. Negrini M, Sabbioni S, Possati L, Rattan S, Corallini A, Delaval B, Letard S, Lelievre H, Chevrier V, Daviet L, Barbanti-Brodano G et al. (1994). Cancer Res 54: Dubreuil P et al. (2005). Cancer Res 65: 7231–7240. 1331–1336. Denison SR, Callahan G, Becker NA, Phillips LA, Smith DI. Noviello C, Courjal F, Theillet C. (1996). Clin Cancer Res 2: (2003a). Genes Chromosomes Cancer 38: 40–52. 1601–1606. Denison SR, Wang F, Becker NA, Schule B, Kock N, Phillips Orphanos V, McGown G, Hey Y, Boyle JM, Santibanez- LA et al. (2003b). Oncogene 22: 8370–8378. Koref M. (1995). Br J Cancer 71: 290–293. Dhillon VS, Husain SA, Ray GN. (2003). Teratog Carcinog Paige AJ, Taylor KJ, Taylor C, Hillier SG, Farrington S, Scott Mutagen (Suppl 1): 35–45. D et al. (2001). Proc Natl Acad Sci USA 98: 11417–11422. Fabbri M, Iliopoulos D, Trapasso F, Aqeilan RI, Cimmino A, Park SW, Ludes-Meyers J, Zimonjic DB, Durkin ME, Zanesi N et al. (2005). Proc Natl Acad Sci USA 102: Popescu NC, Aldaz CM. (2004). Br J Cancer 91: 753–759. 15611–15616. Picchio MC, Martin ES, Cesari R, Calin GA, Yendamuri S, Ginestier C, Bardou VJ, Popovici C, Charafe-Jauffret E, Kuroki T et al. (2004). Clin Cancer Res 10: 2720–2724. Bertucci F, Geneix J et al. (2003). Int J Cancer 107: 854–862. Popescu NC. (2003). Cancer Lett 192: 1–17. Ginestier C, Charafe-Jauffret E, Bertucci F, Eisinger F, Geneix Popovici C, Zhang B, Gregoire MJ, Jonveaux P, Lafage- J, Bechlian D et al. (2002). Am J Pathol 161: 1223–1233. Pochitaloff M, Birnbaum D et al. (1999). Blood 93: Guasch G, Delaval B, Arnoulet C, Xie MJ, Xerri L, Sainty D 1381–1389. et al. (2004). Blood 103: 309–312. Prasad R, Gu Y, Alder H, Nakamura T, Canaani O, Saito H Guler G, Uner A, Guler N, Han SY, Iliopoulos D, McCue P et al. (1993). Cancer Res 53: 5624–5628. et al. (2005). Pathol Int 55: 471–478. Richards RI. (2001). Trends Genet 17: 339–345. Hayashi Y, Raimondi SC, Look AT, Behm FG, Kitchingman Rodriguez C, Causse A, Ursule E, Theillet C. (2000). Genes GR, Pui CH et al. (1990). Blood 76: 1626–1630. Chromosomes Cancer 27: 76–84. Huang TJ, Huang BJ, Liang QW, Huang CW, Fang Y. (2004). Roz L, Gramegna M, Ishii H, Croce CM, Sozzi G. (2002). Hepatobiliary Pancreat Dis Int 3: 62–68. Proc Natl Acad Sci USA 99: 3615–3620. Huebner K, Croce CM. (2001). Nat Rev Cancer 1: 214–221. Saito S, Saito H, Koi S, Sagae S, Kudo R, Saito J et al. (1992). Huebner K, Croce CM. (2003). Br J Cancer 88: 1501–1506. Cancer Res 52: 5815–5817. Huebner K, Druck T, Siprashvili Z, Croce CM, Kovatich A, Sato T, Fujita N, Yamada A, Ooshio T, Okamoto R, Irie K McCue PA. (1998). Recent Results Cancer Res 154: 200–215. et al. (2006). J Biol Chem 281: 5288–5299. Ikeda W, Nakanishi H, Miyoshi J, Mandai K, Ishizaki H, Seki N, Kodama J, Hashimoto I, Hongo A, Yoshinouchi M, Tanaka M et al. (1999). J Cell Biol 146: 1117–1132. Kudo T. (2001). Int J Oncol 19: 305–310. Ishii H, Zanesi N, Vecchione A, Trapasso F, Yendamuri S, Smith DI, Huang H, Wang L. (1998). Int J Oncol 12: Sarti M et al. (2003). FASEB J 17: 1768–1770. 187–196. Johnson SA, Dubeau L, Kawalek M, Dervan A, Schonthal Sorlie T, Tibshirani R, Parker J, Hastie T, Marron JS, Nobel AH, Dang CV et al. (2003a). Mol Cell Biol 23: 3043–3051. A et al. (2003). Proc Natl Acad Sci USA 100: 8418–8423. Johnson SA, Dubeau L, White RJ, Johnson DL. (2003b). Cell Sutherland GR, Baker E, Richards RI. (1998). Trends Genet Cycle 2: 442–444. 14: 501–506. Kerangueven F, Noguchi T, Coulier F, Allione F, Wargniez V, Taki T, Hayashi Y, Taniwaki M, Seto M, Ueda R, Hanada R Simony-Lafontaine J et al. (1997). Cancer Res 57: et al. (1996). Oncogene 13: 2121–2130. 5469–5474. Teixeira MR, Pandis N, Heim S. (2002). Genes Chromosomes Kitada T, Asakawa S, Hattori N, Matsumine H, Yamamura Cancer 33: 1–16. Y, Minoshima S et al. (1998). Nature 392: 605–608. Tibiletti MG, Bernasconi B, Furlan D, Riva C, Trubia M, Mandai K, Nakanishi H, Satoh A, Obaishi H, Wada M, Buraggi G et al. (1996). Cancer Res 56: 4493–4498. Nishioka H et al. (1997). J Cell Biol 139: 517–528. van der Burg M, Poulsen TS, Hunger SP, Beverloo HB, Smit Marin I, Lucas JI, Gradilla AC, Ferrus A. (2004). Physiol EM, Vang-Nielsen K et al. (2004). Leukemia 18: 895–908. Genom 17: 253–263. Wang F, Denison S, Lai JP, Philips LA, Montoya D, Kock N Menasce LP, Orphanos V, Santibanez-Koref M, Boyle JM, et al. (2004). Genes Chromosomes Cancer 40: 85–96. Harrison CJ. (1994a). Genes Chromosomes Cancer 10: Whitcomb BP, Mutch DG, Herzog TJ, Rader JS, 286–288. Gibb RK, Goodfellow PJ. (2003). Clin Cancer Res 9: Menasce LP, Orphanos V, Santibanez-Koref M, Boyle JM, 2277–2287. Harrison CJ. (1994b). Genes Chromosomes Cancer 10: Zhadanov AB, Provance Jr DW, Speer CA, Coffin JD, Goss 26–29. D, Blixt JA et al. (1999). Curr Biol 9: 880–888.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc).

Oncogene