Oncogene (2005) 24, 7281–7289 & 2005 Nature Publishing Group All rights reserved 0950-9232/05 $30.00 www.nature.com/onc

NRG1 rearrangements in clinical breast cancer: identification of an adjacent novel amplicon associated with poor prognosis

Leah M Prentice1,2, Ashleen Shadeo3, Valia S Lestou1,2,4, Melinda A Miller1,2, Ronald J deLeeuw3, Nikita Makretsov1,2, Dmitry Turbin1,2, Lindsay A Brown1,2, Nicol Macpherson5, Erika Yorida1,2, Maggie CU Cheang1,2, John Bentley1,2, Stephen Chia6, Torsten O Nielsen1,2, C Blake Gilks1,2, Wan Lam3 and David G Huntsman*,1,2

1Department of Pathology and Prostate Centre, Genetic Pathology Evaluation Centre, Vancouver General Hospital, Vancouver, BC, Canada V6H 3Z6; 2British Columbia Cancer Agency, Department of Pathology and Laboratory Medicine, University of British Columbia, Vancouver, BC, Canada V6H 3Z6; 3Department of Cancer Genetics and Developmental Biology, British Columbia Cancer Research Centre, Vancouver, BC, Canada V5Z 1L3; 4Department of Paediatric Laboratory Medicine, The Hospital for Sick Kids, Toronto, Canada M5G 1X8; 5Department of Medical Oncology, British Columbia Cancer Agency, Victoria, Canada V8R 6V5; 6Department of Medical Oncology, British Columbia Cancer Agency, Vancouver, BC, Canada V5Z 4E6

Rearrangements of the neuregulin (NRG1) gene have been Introduction implicated in breast carcinoma oncogenesis. To determine the frequency and clinical significance of NRG1 aberra- Chromosomal translocations are a common feature of tions in clinical breast tumors,a breast cancer tissue sarcoma, leukemia and lymphoma, but little is known microarray was screened for NRG1 aberrations by about their role in common epithelial malignancies. A fluorescent in situ hybridization (FISH) using a two-color translocation t(12;15) leading to ETV6-NTRK3 gene split-apart probe combination flanking the NRG1 gene. fusion is a specific event in secretory carcinoma, a rare Rearrangements of NRG1 were identified in 17/382 cases formof breast cancer (Tognon et al., 2002; Makretsov by FISH,and bacterial artificial array et al., 2004a). This was the first report of a recurrent and comparative genomic hybridization was applied to five of subtype-specific translocation in breast cancer and raises these cases to further map the chromosome 8p abnorm- the possibility that translocations could define other alities. In all five cases,there was a novel amplicon clinically relevant subtypes of breast cancer. The short centromeric to NRG1 with a minimum common region of arm of is a site of frequent chromosomal amplification encompassing two , SPFH2 and abnormalities in breast cancers and breast cancer cell FLJ14299. Subsequent FISH analysis for the novel lines, including gene amplification events and transloca- amplicon revealed that it was present in 63/262 cases. tions. In breast cancers, the NRG1 and FGFR1 genes Abnormalities of NRG1 did not correlate with patient have frequently been involved in cytogenetic abnormal- outcome,but the novel amplicon was associated with poor ities of 8p11–12 (Bernardino et al., 1998; Liu et al., 1999; prognosis in univariate analysis,and in multivariate Ugolini et al., 1999; Adelaide et al., 2000, 2003; analysis was of prognostic significance independent of Birnbaum et al., 2003; Ray et al., 2004). nodal status,tumor grade,estrogen receptor status,and NRG1 codes for at least 15 different isoforms through human epidermal growth factor receptor (HER)2 over- alternative splicing, including four heregulin (HRG) expression. Of the two genes in the novel amplicon, isoforms, which are ligands for members of the ERBB/ expression of SPFH2 correlated most significantly with human epidermal growth factor receptor (HER) family amplification. This amplicon may emerge as a result of (Adelaide et al., 2003; Falls, 2003). These HRG isoforms breakpoints and chromosomal rearrangements within the start transcription at exon 2, completely bypassing exon NRG1 locus. 1, which is located 955 kb away at the telomeric end of Oncogene (2005) 24, 7281–7289. doi:10.1038/sj.onc.1208892; the gene. Binding of HRG ligands to HER3 facilitates published online 18 July 2005 heterodimerization of HER3 with HER2 and stimulates a signaling cascade that effects proliferation, survival, Keywords: NRG1; breast cancer; amplicon; SPFH2;8p and differentiation (as reviewed by Yarden and Sliw- kowski, 2001). HER2 is overexpressed in 15–25% of ductal breast cancers and this overexpression correlates with poor prognosis (Yarden and Sliwkowski, 2001). HER1 and HER3 overexpression are also associated with poor prognosis in breast cancer (Wiseman et al., *Correspondence: DG Huntsman, British Columbia Cancer Agency, 2005). Furthermore, HRG has been shown to promote Department of Pathology and Laboratory Medicine, 600 West 10th tumor growth in vivo and induce metastasis (Atlas et al., Avenue, Vancouver, BC, Canada V5Z 4E6; E-mail: [email protected] 2003). The known oncogenic role of HRG has provided Received 18 April 2005; revised 1 June 2005; accepted 2 June 2005; the rationale for considering NRG1 as a potential target published online 18 July 2005 of 8p12 rearrangements in breast carcinoma. 8p abnormalities in clinical breast cancer LM Prentice et al 7282 In breast cancer cell lines, there have been seven were not assessable due to poor FISH hybridization, breakpoints described within or around NRG1,anda insufficient numbers of tumor cells, tissue loss, or tissue translocation with DOC4 located on chromosome 11 over/underdigestion. In all, 17 (4.7%) of the cases has been described in the MDA-MB-175 cell line (Liu showed abnormal NRG1 signals, excluding aneuploidy. et al., 1999; Wang et al., 1999; Adelaide et al., 2003). Three distinct types of abnormalities were seen: (A) low This translocation gives rise to a secreted fusion copy number of the 30 (centromeric)-end of the gene (Liu et al., 1999; Wang et al., 1999; Adelaide et al., 2003) with respect to the 50 (telomeric)-end of the gene (two that has been demonstrated to have an autocrine growth cases), (B) low copy number of the 50-end of the gene effect (Schaefer et al., 1997). with respect to the 30-end of the gene (12 cases), and (C) Huang et al. (2004) identified NRG1 breaks in 6% of amplification of both ends of the gene (three cases) breast cancers. The cases with NRG1 breaks tended to (Figure 1a). The two cases with type A NRG1 be HER2 negative and estrogen receptor (ER) positive. aberrations showed either loss of the 30-end of NRG1 They also reported a significant correlation with FGFR1 (case number 152) or gain of the 50-end of NRG1 overexpression. They hypothesized that these break- (case number 77) (Figure 1b). Type B aberrations were points result in translocation events that lead to breast the most frequently observed, with loss of the 50 signal cancer development through abnormal expression of the being present in 11 cases, two of which also had 30 30-end of NRG1. They have shown translocation events amplification events (cases 87 and 179). The remaining in four breast cancer cell lines (ZR-75-1, HCC1937, nine cases with type B abnormalities had increased copy UACC-812, and MDA-MB-175) and demonstrated gain number of the 30-end of NRG1 due to loss of the 50-end of the 8p11–12 region in three cell lines (SUM-52, ZR- of the gene without 30 amplification. A single case (case 75-1, and MDA-MB-134) (Adelaide et al., 2003). It is, 286) with type B abnormality showed amplification of however, possible that chromosomal rearrangements the 30-end of NRG1 only (Figure 1c). Three cases: 285, within intron 1 of the NRG1 gene do not represent gene- 303, and 382 showed type C abnormalities (Figure 1d), specific translocations and that these events may be part and this is most likely indicative of whole gene of some other cytogenetic process. Factors that make amplification. the gene-specific hypothesis less likely include: (1) the A second set of FISH probes was applied to the 17 lack of common fusion partners or a clearly oncogenic cases demonstrating NRG1 aberrations, one covering fusion transcript, (2) the lack of a consistent association exon 1 and the other covering exons 2–17 of the gene. between NRG1 breaks and NRG1 mRNA or protein Exon 2 is located 955 kb away fromexon 1 and levels, (3) the lack of any correlation between NRG1 specifically represents the start of the HRG coding region abnormalities and expression of the epidermal growth (Adelaide et al., 2003). This specific (internal) probe set factor receptors (HER1, HER2, HER3, and HER4), showed fewer abnormalities than were seen with the and (4) the lack of strong and specific clinical associa- flanking (external) probes. For example, both cases tions with NRG1 rearrangements, as has been seen for showing type A abnormalities (loss of the 30-end) no the t(12;15) translocation and secretory carcinomas. longer showed a difference in copy number between the To determine the frequency and further delineate the 50-andthe30-end of NRG1. The different locations of clinical significance of chromosomal rearrangements both specific and flanking probe sets are demonstrated in within and surrounding the NRG1 gene, we analysed Figure 2a and complete results are listed in Figure 1a. 438 clinically annotated breast cancers using fluorescent in situ hybridization (FISH) and immunohistochemistry (IHC). In five cases with NRG1 rearrangements, HRG correlates significantly with tumor grade, p53, bacterial artificial chromosome (BAC) array compara- and HER1 tive genomic hybridization (CGH) was performed. A second series of cases with matched frozen tissue There was no statistically significant correlation between samples was used to determine the association bet- NGR1 aberrations, either individually or taken as a ween gene copy number and expression levels of specific group, and the biomarkers, ER, p53, Ki67, HER1, genes identified as being amplified through BAC HER2, or HER3, nor with tumor grade, nodal status, or array CGH. histological subtype. NRG1 abnormalities were not associated with poor survival. There was no significant correlation between HRG IHC and NRG1 aberrations. HRG immunostaining was Results seen in 60% of the tumors, with 19% demonstrating strong immunostaining. HRG immunostaining intensity Three distinct types of NRG1 aberrations correlated positively with tumor grade (P ¼ 0.007), p53 A three-color FISH assay consisting of a centromere 8 (P ¼ 0.001), and Ki67 (P ¼ 0.01), and correlated nega- probe (aqua) together with probes flanking the 50-end tively with ER (P ¼ 0.016) and PR (P ¼ 0.017). A (green) and the 30-end (orange) of the NRG1 gene was positive and statistically significant correlation between performed on the tissue microarray (TMA) series of 438 HRG and HER1 (P ¼ 0.015) was observed but not with clinically annotated breast cancers. A total of 358 cases the other members of the epidermal growth factor produced FISH signals of sufficient intensity and clarity receptor family, HER2 (P ¼ 0.5), HER3 (P ¼ 0.4), and for analysis (81.7% of all cases). The remaining cases HER4 (P ¼ 0.2).

Oncogene 8p abnormalities in clinical breast cancer LM Prentice et al 7283

Figure 1 (a) Summary of NRG1 aberrations as determined by FISH. Two probe sets were used: one probe set flanks the 50- and 30-end of the NRG1 gene and the second probe set specifically covers the 50- and 30-end of the gene, and is internal to the first probe set. In all, 17 cases showed NRG1 abnormalities using the flanking probe set, while eight cases showed abnormalities with the internal, specific probes. (b–d) Microscopic images demonstrating examples of the three types of NRG1 aberrations as determined by FISH. Green spots represent the 50-end of the NRG1 gene, red spots represent the 30-end of the NRG1 gene, and blue spots are the chromosome 8 centromere. Increased spot count as compared to centromere signal indicates amplification; loss of signal as compared to centromere spot count indicates loss of genomic material. (b) Amplification of the 50-end of NRG1 (type A aberration) as shown by multiple green signals. (c) Amplification of the 30-end of NRG1 (type B aberration). (d) Amplification of both 50- and 30-end of the NRG1 gene as compared to the centromere (type C aberration)

BAC array CGH confirms NRG1 aberrations and five cases by array CGH (Figure 2b) and gain in four of identifies an amplicon centromeric to NRG1 five cases by FISH (Figure 2c). For all five cases, the minimal common region of amplification within the Five cases showing a variety of NRG1 aberrations by novel amplicon contains two Refseq annotated genes, FISH (cases 77, 179, 303, 285, and 286) were analysed SPFH2 and FLJ14299. SPFH2 was formally known as by BAC array CGH, focusing on chromosome 8p to C8orf2, and is the approved symbol as determined by gain more detail of the NRG1 region. In case 77, FISH the HUGO Committee (HGNC). with the flanking probe set showed 50 amplification, The minimal region of amplification was determined while the specific 50 probe did not show amplification using SMRT aCGH and SeeGH filtered raw data (type A); the CGH results confirmboth findings (Figure 3). Three of these five cases (285, 286, and (Figure 2b). In case 179, array CGH showed neither 50 303) also included PROSC as part of the amplicon; the deletion nor 30 amplification in contrast to FISH, which FISH probe encompassed all three genes due to their showed type B abnormalities. Case 303 shows slight proximity. The average amplification ratios for the amplification of both 50 and 30 NRG1 by array CGH; BACs representing the four different regions of 8p, as type C abnormalities were demonstrated by FISH. Case determined by FISH, are shown in Figure 2c. The 285 shows higher levels of amplification of both ends of amplification ratios relative to centromere copy number, NRG1 by array CGH, which is also consistent with the for the novel amplicon range from 3.8 to 5.9, compared type C FISH results. For case 286, array CGH results to FGFR1 amplification ratios that range from no also correlate with type B FISH and clearly demonstrate amplification (ratio 1.1) to a ratio of 4.0. high-level amplification of 30 NRG1 without gain of 0 the 5 -end. Discrepancies can be accounted for based on The novel amplicon is present in 24% of clinical breast the different BACs used for FISH and CGH. cancer cases and shows a positive correlation with poor All five cases show significant levels of amplification survival by array CGH and validated by FISH for an amplicon centromeric to the NRG1 gene and telomeric to FGFR1, To clarify the prevalence of this amplicon in breast while FGFR1 shows gain of copy number in only two of cancer, the same 438 case TMA was hybridized with a

Oncogene 8p abnormalities in clinical breast cancer LM Prentice et al 7284

Figure 2 (a) A schematic representation of the region of interest on the 8p arm, showing the location of BACs used in this study with corresponding BAC identification numbers. BACs located on the right side of the schematic are those used in BAC array CGH, while those on the left side are BACs used for FISH. (b) BAC array CGH results. For each individual case, there are three vertical lines representing the amplification scale. Vertical red left line represents a log 2 ratio between the sample and reference channels of À0.5, vertical purple middle line represents a log 2 ratio of 0, and the vertical green right line represents a log 2 ratio of þ 0.5. Each individual black spot represents one BAC clone and shows the amplification ratio as compared to normal DNA with ratios to left and right of the purple line representing losses and gains, respectively. There is consistent amplification of the novel amplicon in all five samples. (c)A graphical representation of amplification ratios obtained from FISH for all BACs located on the left side of the schematic, grouped according to color. Graph bars fromleft to right; 5 0 NRG1 (blue), 30 NRG1 (tan), novel amplicon (purple), and FGFR1 (burgundy). Only the novel amplicon consistently shows increased copy number in these five cases

three-color probe specific for the region of amplification or with the clinicopathological markers – tumor grade, (orange), FGFR1 (green), and centromere 8 (aqua). nodal status, or histological subtype, after a Bonferroni Results were obtained for 262 of 495 cases, with 63/262 adjustment for multiple comparisons. There is, however, (24%) cases showing a 1.5 or greater amplification ratio a highly significant correlation between the novel for the novel amplicon. Comparatively, only 39/262 amplicon and FGFR1 (Po0.001), and both the novel (15%) cases had an amplification ratio of 1.5 or greater amplicon and FGFR1 show significant correlation with for FGFR1. Of the 63 cases with the novel amplicon, NRG1 aberrations (Po0.001, P ¼ 0.002, respectively). 36 cases also demonstrate FGFR1 amplification. NRG1 aberrations do not correlate with amplification The novel amplicon shows a positive correlation with events in other chromosomal regions: C-MYC poor survival using the Kaplan–Meier method and a (P ¼ 0.241) and CCND1 (P ¼ 0.140) (Table 1). cutoff amplification ratio of 1.5 (P ¼ 0.0101) (Figure 4a and b). Significance of the novel amplicon was main- SPFH2 most significantly correlates overexpression with tained in multivariate analysis (P ¼ 0.0007) (Figure 4c). amplification as determined by real-time PCR In contrast, FGFR1 does not show a significant correlation with patient survival (P ¼ 0.0953). Amplifi- A second, independent breast cancer array was pre- cation of neither the novel amplicon nor FGFR1 showed pared from40 cases, each with paired RNA samples correlation with any of the biomarkers – ER, p53, Ki67, isolated fromsnap frozen malignant and surrounding C-MYC, Cyclin-D1, HRG, HER1, HER2, and HER3 – benign breast tissue. The array was hybridized with a

Oncogene 8p abnormalities in clinical breast cancer LM Prentice et al 7285

Figure 3 Minimum common region of amplification (MAR) at 8p12 using unsmoothed array CGH. Log 2 ratios of the data points obtained from the five cases are shown in multiple alignment, SeeGH Karyogram. Vertical line segments illustrate the individual BACs covering the corresponding chromosomal regions. Centerline is designated as 0 and movement towards the right line ( þ 0.5) indicates a gain in genomic material of the corresponding chromosomal region, while movement to the left (À0.5) indicates a loss. Horizontal lines specify the boundaries of the MAR. Case 77 determined the centromeric boundary and case 303 the telomeric boundary. FLJ14299 is entirely within the MAR, while SPFH2 is mostly within the boundaries. PROSC is excluded

Figure 4 (a) Kaplan–Meier survival curve demonstrating the prognostic significance of the novel amplicon. (b) Summary of univariate survival analysis for NRG1 aberrations, FGFR1, and the novel amplicon. Only the novel amplicon is prognostically significant. (c) Multivariate analysis results using Cox’s regression in a backwards stepwise elimination using the following variables: HER2, Grade, ER status, lymph node status, and the novel amplicon. Only the novel amplicon and lymph node status are of independent prognostic significance. Asterisk indicates significant P-value combination of three probes derived from the new 8p had the highest amplification ratios (8.4, 2.4, 2.1, 1.9). amplicon, FGFR1, and centromere 8. Using this array, Six cases showed overexpression of FLJ14299,witha 29 cases had interpretable FISH results and nine of these borderline significant correlation with gene amplifica- had the novel amplicon. Real-time PCR was used to tion (P ¼ 0.056). Of these six cases, two did not exhibit determine gene expression levels for FLJ14299, SPFH2, gene amplification. PROSC overexpression occurred in PROSC, and FGFR1 in order to correlate expression five cases and showed no significant correlation with with amplification. Using a twofold cutoff value as amplification (P ¼ 0.287). Two of these five cases did not indicative of overexpression, four cases were found to show gene amplification. FGFR1 was overexpressed in overexpress SPFH2 and overexpression correlated sig- nine cases, with a significant correlation with gene nificantly with gene amplification as determined by amplification (P ¼ 0.022), but more than half of these Fisher’s exact test (P ¼ 0.005) (Table 2). These four cases cases (5/9) did not have gene amplification.

Oncogene 8p abnormalities in clinical breast cancer LM Prentice et al 7286 Table 1 Correlations between 8p abnormalities and clinical/patholo- not part of a gene-specific oncogenic translocation gical variables event. Although NRG1 translocations in clinical breast Marker NRG1 aberrations FGFR1 Novel amplicon cancers could have a role in breast cancer development, a more likely scenario is that these breakpoints result in HER1 0.565 0.309 0.667 break–fusion–bridge (BFB) cycles that create amplified HER2 0.175 1.000 0.350 HER3 0.537 0.736 1.000 regions centromeric to the NRG1 breakpoint, leading to HRG 1.000 1.000 0.738 overexpression of novel oncogenes, a possibility sug- ER 0.598 0.330 0.688 gested by Birnbaum et al. (2003). Ki67 0.806 1.000 0.503 This hypothesis is supported by the discovery of the p53 0.792 0.194 0.021 novel amplified region centromeric to NRG1 and Tumor grade 0.446 0.613 0.593 Nodal status 0.807 0.398 0.171 its correlation with poor patient survival. This region Histological subtype 0.250 0.366 0.170 is amplified in 24% of breast cancer cases and is C-MYC 0.241 0.110 0.099 prognostically significant. FISH data fromthree breast Cyclin-D 0.140 0.304 0.520 cancer cell lines (T-47D, MDA-MB-361, ZR-75-1) NRG1 aberrations — 0.002* o0.001* demonstrated that translocations involving 8p12 are FGFR1 ——o0.001* accompanied not only by amplification of various genes *Indicates significance after a Bonferroni adjustment for multiple within the region but also by additional rearrangements comparisons (P ¼ 0.05/14 ¼ 0.00357) such as deletion, duplication, and inversion (Courtay- Cahen et al., 2000). These breakpoints and those found by Adelaide et al. (2003) could be the initiating events Table 2 Correlations between expression and amplification status of that lead to the BFB mechanism for gene amplification genes from chromosome 8p as has been suggested for Her2 and Topo2A (Jacobson Novel amplicon Amplified Not amplified et al., 2004). The BFB cycle is characterized by (N ¼ 9) (N ¼ 20) breakpoints surrounding the amplicon and loss of genes telomeric to the amplified region (Coquelle et al., 1997). SPFH2 Overexpressed (N ¼ 4) 4 0 0 P ¼ 0.005* Not overexpressed 520Consistent with the BFB model, the loss of the 5 -end of (N ¼ 25) the NRG1 gene, which is telomeric to the site of amplification, is seen much more frequently than loss FLJ14299 Overexpressed (N ¼ 6) 4 2 of the 30-end of NRG1 (11 cases versus one case). In P ¼ 0.056 Not overexpressed 518order for the BFB cycle to lead to successful amplifica- (N ¼ 23) tion, a second breakpoint region centromeric to the PROSC Overexpressed (N ¼ 5) 3 2 amplicon is required (Figure 5). There have been several P ¼ 0.287 Not overexpressed 618reports of entire 8p armloss in breast cancer cell lines (N ¼ 24)

FGFR1 amplicon Amplified Not amplified (N ¼ 5) (N ¼ 24)

FGFR1 Overexpressed (N ¼ 9) 4 5 P ¼ 0.022* Not overexpressed 119 (N ¼ 20)

*Indicates statistical significance

Discussion

We report the presence of NRG1 gene rearrangements in 4.7% of breast cancers. Our findings are similar to those of Huang et al. (2004), who found 6% of breast cancer cases with NRG1 gene rearrangements. These are the first two reports of NRG1 gene aberrations in clinical tumor samples, although these rearrangements have been previously characterized in breast cancer cell lines (Adelaide et al., 2003). Unlike the current study, Huang Figure 5 Schematic representation of the BFB mechanism. Initiation of the BFB cycle by a double-stranded break of sister et al. do not report any cases of amplification with their chromatids after S-phase replication. The breakpoint within the NRG1 probes, only loss. Neither Huang et al. nor our NRG1 gene is represented by a broken line. Broken ends become study shows correlation of NRG1 gene aberrations with fused and form a chromosomal bridge during mitosis with spindle the expression of HRG, the HRG receptors (HER1, attachment to the centromeres. The bridged chromosome breaks HER2, HER3, or HER4), or with any specific mole- during anaphase leaving duplicate copies of the novel amplicon and FGFR1 on the same chromosome. Cycle repeats with a second cularly or clinically defined subtype of breast cancer. breakpoint that creates specific amplification of the novel amplicon These findings suggest that NRG1 rearrangements are as compared to FGFR1 or NRG1

Oncogene 8p abnormalities in clinical breast cancer LM Prentice et al 7287 and this may reflect a second breakpoint region FGFR1 was also found to be coamplified with centromeric to the amplified region (Rummukainen CCND1 (11q13) in the breast cancer cell line MDA- et al., 2001). Two breakpoint regions centromeric to the MB-134 and 12 of 225 breast cancer specimens (Bautista novel amplicon, with one located telomeric to FGFR1, and Theillet, 1998). FGFR1 and CCND1 gene coampli- would explain the difference in amplification frequencies fication was associated with a worse prognosis in a between the novel amplicon (24%) and FGFR1 (15%). cohort of 640 breast cancer cases (Cuny et al., 2000). A In a region such as 8p11–12, the BFB-based mechanism link between FGFR and D cyclins has been found through which multiple amplicons can develop is likely within the G1 phase of the cell cycle. When FGFR very complex and requires multiple discrete breakpoints. activity is inhibited, cyclins D1 and D2 are down- The underlying genetic defects that permit the develop- regulated resulting in cell cycle arrest at G1 mediated by ment of multiple amplicons within breast cancer the RB phosphorylation pathway (Koziczak et al., genomes are not yet characterized. 2004). Chromosome 11q was assessed by BAC array The novel 8p12-derived amplicon described herein CGH in our five select cases to determine if coamplifica- contains two genes, FLJ14299 and SPFH2. When the tion events existed. Besides a minor amplification at expression levels of these genes, as assessed by quanti- 11q14.1 (results not shown), which did not include the tative real-time PCR, are compared with gene copy genes DOC4 and CCND1 or other regions previously number, SPFH2 expression most significantly correlates implicated in NRG1 translocations (Bautista and Theil- with gene amplification. Although the function of the let, 1998; Adelaide et al., 2000), our study did not find translated product of SPFH2 has yet to be elucidated, any coamplification events in the five select cases its gene product has been isolated fromcaveolae and analysed. Our study also showed no correlation between lipid raft-enriched fractions of the human endothelial FGFR1 and CCND1 amplifications as determined by membrane (Sprenger et al., 2004). By dissociating lipid FISH. While FGFR1 amplification may prove to be rafts and HER2 clusters, Nagy et al. (2002) demon- significant in breast cancer, it appears to be distinct from strated that HER2 association with HER3 decreased, as the novel amplicon. Compared to the novel amplicon, was EGF- and HRG-induced tyrosine phosphorylation FGFR1 is less commonly amplified as determined by of Shc. The location of SPFH2 within a clinically array CGH and FISH, and it is not a significant significant amplicon in breast cancers, the association prognostic indicator. between amplification and overexpression, and its Our study has demonstrated a novel amplification presence in lipid rafts all suggest an oncogenic role for event located within the 8p11–12 region that is SPFH2 in breast cancer. significantly associated with poor survival and maintains FLJ14299 amplification has been described in the cell significance as an independent prognostic indicator in lines SUM-44 and SUM-225 and is associated with multivariate analysis. This novel amplicon also corre- overexpression at the mRNA level (Ray et al., 2004). lates with NRG1 aberrations supporting a hypothesis FLJ14299 gene contains the C2H2 zinc-finger domain that breakpoints within the NRG1 gene lead to BFB present in several other tumor-related genes such as cycles that result in amplification of a novel oncogene. BCL6 (lymphoma), ZNF217 (breast carcinoma), and SPFH2 is a potential candidate for this oncogene. GLI (sarcoma), and could indicate a DNA-binding region (Ray et al., 2004). Our study shows that FLJ14299 is the most frequently overexpressed of the three genes and shows a borderline significant correla- Materials and methods tion with amplification. As PROSC was part of the Breast cancer case series and TMA construction amplicon in only three of five BAC array CGH cases and does not show correlation between amplification The initial, larger TMA was constructed using formalin-fixed, and overexpression in the smaller series, it is likely not paraffin-embedded tissue blocks received from the Department an oncogenically important element in the amplicon. of Pathology at Vancouver General Hospital during the period The adjacent FGFR1 gene has been shown to be 1974–1995. Case selection and TMA construction have been described previously (Parker et al., 2002; Makretsov et al., amplified in 9–15% of breast cancers, which is 2003, 2004b). Breast tumor samples used for the second, consistent with our findings (15%), and has been smaller TMA were obtained from women undergoing lum- implicated as the driver gene for the 8p11–12 amplicon pectomy or mastectomy for breast cancer in Victoria, British (Ugolini et al., 1999). Similar to our study, Huang et al. Columbia (Royal Jubilee Hospital and Victoria General (2004) found a significant association of FGFR1 Hospital), in collaboration with the Department of General expression with NRG1 aberrations. FGFR, like other Surgery and the Department of Anatomic Pathology from single transmembrane growth factor receptors, trans- June 1998 to June 2000. Briefly, representative areas of duces signals across the cell membrane after binding carcinoma were selected and marked on the hematoxylin and extracellular ligands (Powers et al., 2000). However, eosin slides and corresponding tissue blocks for TMA although high FGFR1 oncogene amplification was construction. The TMAs were assembled using a tissue- arraying instrument (Beecher Instruments, Silver Springs, demonstrated in three breast cancer cell lines SUM-44, MD, USA), with two 0.6 mm cores per case. Outcome data SUM-52, and SUM-225, FGFR inhibition failed to slow were available for all 495 patients on the larger TMA, with growth, suggesting that other genes may be responsible median follow-up of 15.4 years (range 6.3–26.6 years). For for this amplicon and/or breast cancer progression (Ray statistical analysis, only 438 cases with invasive breast et al., 2004). carcinoma and known tumor grade are analysed. Ethical

Oncogene 8p abnormalities in clinical breast cancer LM Prentice et al 7288 approval was obtained to performthis study fromthe Clinical BAC array CGH Research Ethics Board of the University of British Columbia. A subset of five cases representing all types of NRG1 gene rearrangements, that is, case 77 (type A), cases 179 and 286 RNA isolation (type B), and cases 303 and 285 (type C) were analysed by Total cellular RNA was extracted fromsamplesby the BAC array CGH to both verify the FISH results and provide acid–phenol guanidium method (Chomczynski and Sacchi, high-resolution mapping (B0.1 Mb) of the 8p11–12 region. 1987) with TRIzol as recommended by the manufacturer We performed array CGH on microarrays containing 32 433 (www.lifetech.com). After isolation, the RNA was treated with BAC-derived amplified fragment pools spotted in triplicate 1 U DNAse (RQ1, Promega, Wisconsin, USA) in the buffer (97 299 elements) over two 18 mm  54 mm arrays, as de- supplied to remove contaminating DNA. The quality of RNA scribed previously (Ishkanian et al., 2004). was assessed by confirming the presence of intact 18 and 26s ribosomal RNA bands by agarose gel electrophoresis. BAC array CGH imaging and analysis Fluorescent in situ hybridization Cyanine-3 and cyanine-5 images (16 bit) were acquired using a CCD camera system (Applied Precision, Issaquah, WA, USA). Sections (6 mmthick) of the TMA slides were pretreated for Images were then analysed using SoftWoRx Tracker Spot FISH as described elsewhere (Makretsov et al., 2004a). Locus- Analysis software (Applied Precision). Genomic imbalances specific FISH analysis for all genes, including both ends of and their associated breakpoints were identified using NRG1, FGFR1, and the novel amplicon, was performed using genetic local search algorithms within the software package the following BACs fromthe HumanBAC Library RPCI-11 aCGHsmooth developed by Jong et al. as described previously (BACPAC Resources Centre, Children’s Hospital Oakland (Jong et al., 2003; de Leeuw et al., 2004). Customsoftware Research Institute) (listed telomeric to centromeric): 566H8, (SeeGH) was used to visualize all data and is available upon 10L15, 97N12, 478B14, 15H14, 1002K11, 11N9, 692P18, request, as published elsewhere (Chi et al., 2004; Watson et al., 451O18 (for the NRG1 gene), 863K10 (novel amplicon), 2004). 350N15, and 675F6 (FGFR1). BACs 566H8, 10L15, 350N15, 675F6, 97N12, and 478B14 were directly labeled with SpectrumGreen and BACs 11N9, 692P18, 451O18, 863K10, Quantitative RT–PCR 1002K11, and 15H14 were directly labeled with Spectrum The probe sequences used for SPFH2 and FLJ14299 were Orange. The chromosomal locations of all BACs were CCAGAGGCAATCCG and AGACTAGCTTCAGCCTC, validated using normal metaphases (results not shown). The respectively. The forward and reverse primer sequences for well-characterized breast cancer cell line MDA-MB-175 was SPFH2 are CGGGTAACAAAGCCCAACAT, TCACTTTC also used to verify the flanking NRG1 probes and after CATCAACTCGTAGTTTCT, while the forward and reverse multicolor karyotyping (MFISH) treatment showed t(8;11). primer sequences for FLJ14299 are GCCATACGCGCTG The aqua-labeled a-satellite control DNA probe for chromo- TATGGA, GAGGAAGAGCTGTAGTTACTGGTATCC. some 8 (CEP8) was purchased from Vysis (Downers Grove, PROSC and FGFR1 probes with primers included were IL, USA). Probe labeling and FISH was performed using obtained as a 20x target assay fromApplied Biosystems Vysis reagents according to the manufacturer’s protocols (Foster, CA, USA). All probes were used at 100 nM and all (Vysis, IL, USA). Slides were counterstained with 4,6- primers at 900 nM concentrations. Cycling was set at 951C for diamidino 2-phenylindole for microscopy. For each of the 10 min, then 40 cycles of 951C (15 s)/601C (1 min). Results were TMA slides, FISH signals and patterns were identified on a assessed using ABI software and normalized to Stratagene’s Zeiss Axioplan epifluorescent microscope and were scored Universal Human Reference RNA mix consisting of 10 either manually (oil immersion  100) or using Metasystems different human cell lines (La Jolla, CA, USA). Overexpres- Metafer software (MetaSystems Group Inc., Belmont, MA, sion was defined as a 2.0-fold or greater level of mRNA as USA) and enumerated in approximately 40 morphologically compared to control. intact and nonoverlapping nuclei. MFISH of the cell line MDA-MB-175 was performed as published in detail elsewhere (Khoury et al., 2003). Amplification was defined as a copy Statistical analysis number to control copy number ratio of 1.5 or greater; loss was consider as a ratio of 0.6 or less. Statistical analysis was performed using SAS software version 8.02. The optimal amplification ratio cutoff value generated by X-tile (Camp et al., 2004) was 1.38. A cutoff value of 1.5 was Immunohistochemical analysis of HRG, HER3, and HER4 used for all analysis for more stringent results. The prognostic Immunohistochemical analysis was performed on 4-mm-thick significance of the novel amplicon, amplified FGFR1, and sections fromthe TMA after heat-induced epitope retrieval. NRG1 aberrations was assessed using Kaplan–Meier survival Primary antibodies were applied as follows: HRG (clone 7D5, estimates and the log-rank test. Fisher’s exact test was used to diluted 1 : 10), HER3 (rabbit polyclonal, diluted 1 : 400), and test for associations between the novel amplicon, amplified HER4 (clone HFR-1, diluted 1 : 160), all purchased from FGFR1, and NRG1 aberrations. After a Bonferonni adjust- NeoMarkers, Lab Vision (Freemont, CA, USA). HRG ment for multiple comparisons was made, associations were immunoreactivity was scored according to the percentage of deemed to be statistically significant if the adjusted P-value cells positive for the cytoplasmic reaction as follows: negative was less than 0.05/14 ¼ 0.00357. Multivariate analysis of the o10%, weak 10–40%, and strong >40% of cells stained. prognostic significance of the novel amplicon was performed HER3 and HER4 were scored as negative or positive using Cox’s proportional hazards model and a backwards (Wiseman et al., 2005). Linkage of the genetic data and stepwise method to remove variables from the model. Multi- HRG staining with tumor size, nodal status, the immunohis- variate analysis was performed with a model that included the tochemical biomarkers ER, PR, p53, Ki67, HER1, HER2, and novel amplicon, lymph node status, HER2, Grade, and ER as clinical outcome data was performed as described previously variables. Fisher’s exact test was used to test for correlations (Liu et al., 2002). between amplification and overexpression.

Oncogene 8p abnormalities in clinical breast cancer LM Prentice et al 7289 Abbreviations Acknowledgements TMA, tissue microarray; FISH, fluorescent in situ hybridiza- Funding for this project was provided by a Michael Smith tion; HRG, heregulin; HER, human epidermal growth factor Foundation for Health Research (MSFHR) Unit Grant and receptor; IHC, immunohistochemistry; BAC, bacterial artifi- an Aventis Canada Educational Grant. Torsten O Nielsen and cial chromosome; CGH, comparative genomic hybridization; David G Huntsman are MSFHR scholars. Leah M Prentice is MFISH, multicolor karyotyping; MAR, minimum common an MSFHR trainee. region of amplification; BFB, break–fusion–bridge.

References

Adelaide J, Chaffanet M, Mozziconacci MJ, Popovici C, Khoury H, Lestou VS, Gascoyne RD, Bruyere H, Li CH, Conte N, Fernandez F, Sobol H, Jacquemier J, Pebusque M, Nantel SH, Dalal BI, Naiman SC and Horsman DE. (2003). Ron D, Lafage-Pochitaloff M and BirnbaumD. (2000). Int. Mod. Pathol., 16, 716–724. J. Oncol., 16, 683–688. Koziczak M, Holbro T and Hynes NE. (2004). Oncogene, 23, Adelaide J, Huang HE, Murati A, Alsop AE, Orsetti B, 3501–3508. Mozziconacci MJ, Popovici C, Ginestier C, Letessier A, Liu CL, Prapong W, NatkunamY, Alizadeh A, Montgomery Basset C, Courtay-Cahen C, Jacquemier J, Theillet C, K, Gilks CB and van de Rijn M. (2002). Am. J. Pathol., 161, BirnbaumD, Edwards PA and Chaffanet M. (2003). Genes 1557–1565. Cancer, 37, 333–345. Liu X, Baker E, Eyre HJ, Sutherland GR and Zhou M. (1999). Atlas E, Cardillo M, Mehmi I, Zahedkargaran H, Tang C and Oncogene, 18, 7110–7114. Lupu R. (2003). Mol. Cancer Res., 1, 165–175. Makretsov N, Gilks CB, Coldman AJ, Hayes M and Bautista S and Theillet C. (1998). Genes Chromosomes Cancer, Huntsman D. (2003). Hum. Pathol., 34, 1001–1008. 22, 268–277. Makretsov N, He M, Hayes M, Chia S, Horsman DE, Bernardino J, Apiou F, Gerbault-Seureau M, Malfoy B Sorensen PH and Huntsman DG. (2004a). Genes Chromo- and Dutrillaux B. (1998). Genes Chromosomes Cancer, 23, somes Cancer, 40, 152–157. 100–108. Makretsov NA, Huntsman DG, Nielsen TO, Yorida E, BirnbaumD, Adelaide J, Popovici C, Charafe-Jauffret E, Peacock M, Cheang MC, Dunn SE, Hayes M, van de Rijn Mozziconacci MJ and Chaffanet M. (2003). Lancet Oncol., M, Bajdik C and Gilks CB. (2004b). Clin. Cancer Res., 10, 4, 639–642. 6143–6151. Camp RL, Dolled-Filhart M and Rimm DL. (2004). Clin. Nagy P, Vereb G, Sebestyen Z, Horvath G, Lockett SJ, Cancer Res., 21, 7252–7259. Damjanovich S, Park JW, Jovin TM and Szollosi J. (2002). Coquelle A, Pipiras E, Toledo F, Buttin G and Debatisse M. J. Cell Sci., 115, 4251–4262. (1997). Cell, 89, 215–225. Parker RL, Huntsman DG, Lesack DW, Cupples JB, Grant Chi B, DeLeeuw RJ, Coe BP, MacAulay C and LamWL. DR, Akbari M and Gilks CB. (2002). Am. J. Clin. Pathol., (2004). BMC Bioinform., 5, 13. 117, 723–728. Chomczynski P and Sacchi N. (1987). Anal. Biochem., 162, Powers CJ, McLeskey SW and Wellstein A. (2000). Endocr. 156–159. Relat. Cancer, 7, 165–197. Courtay-Cahen C, Morris JS and Edwards PA. (2000). Ray ME, Yang ZQ, Albertson D, Kleer CG, Washburn JG, Genomics, 66, 15–25. Macoska JA and Ethier SP. (2004). Cancer Res., 64, Cuny M, Kramar A, Courjal F, Johannsdottir V, Iacopetta B, 40–47. Fontaine H, Grenier J, Culine S and Theillet C. (2000). Rummukainen J, Kytola S, Karhu R, Farnebo F, Larsson C Cancer Res., 60, 1077–1083. and Isola JJ. (2001). Cancer Genet. Cytogenet., 126, 1–7. de Leeuw RJ, Davies JJ, Rosenwald A, Bebb G, Gascoyne Schaefer G, Fitzpatrick VD and Sliwkowski MX. (1997). RD, Dyer MJ, Staudt LM, Martinez-Climent JA and Lam Oncogene, 15, 1385–1394. WL. (2004). Hum. Mol. Genet., 13, 1827–1837. Sprenger RR, Speijer D, Back JW, De Koster CG, Pannekoek Falls DL. (2003). Exp. Cell Res., 284, 14–30. H and Horrevoets AJ. (2004). Electrophoresis, 25, 156–172. Huang HE, Chin SF, Ginestier C, Bardou VJ, Adelaide J, Iyer Tognon C, Knezevich SR, Huntsman D, Roskelley CD, NG, Garcia MJ, Pole JC, Callagy GM, Hewitt SM, Gullick Melnyk N, Mathers JA, Becker L, Carneiro F, MacPherson WJ, Jacquemier J, Caldas C, Chaffanet M, Birnbaum D and N, Horsman D, Poremba C and Sorensen PH. (2002). Edwards PA. (2004). Cancer Res., 64, 6840–6844. Cancer Cell, 2, 367–376. Ishkanian AS, Malloff CA, Watson SK, DeLeeuw RJ, Chi B, Ugolini F, Adelaide J, Charafe-Jauffret E, Nguyen C, Coe BP, Snijders A, Albertson DG, Pinkel D, Marra MA, Jacquemier J, Jordan B, Birnbaum D and Pebusque MJ. Ling V, MacAulay C and LamWL. (2004). Nat. Genet., 36, (1999). Oncogene, 18, 1903–1910. 299–303. Wang XZ, Jolicoeur EM, Conte N, Chaffanet M, Zhang Y, Jacobson KK, Morrison LE, Henderson BT, Blondin BA, Mozziconacci MJ, Feiner H, BirnbaumD, Pebusque MJ Wilber KA, Legator MS, O’Hare A, Van StedumSC, and Ron D. (1999). Oncogene, 18, 5718–5721. Proffitt JH, Seelig SA and Coon JS. (2004). Genes Watson SK, deLeeuw RJ, Ishkanian AS, Malloff CA and Lam Chromosomes Cancer, 40, 19–31. WL. (2004). BMC Genom., 5, 6. Jong K, Marchori E, van der Vaart A, Ylstra B, Weiss M and Wiseman SM, Makretsov N, Neilsen TO, Gilks B, Yorida E, Meijer G. (2003). Applications of Evolutionary Computing. Cheang M, Turbin D, Gelmon K and Huntsman DG. EvoBIO: Evolutionary Computation and Bioinformatics (2005). Cancer, 103, 1770–1777. Lecture Notes in Computer Science, Vol. 2611 Springer: Yarden Y and Sliwkowski MX. (2001). Nat. Rev. Mol. Cell. Berlin, pp. 54–65. Biol., 2, 127–137.

Oncogene