The short arm of chromosome 8, 8p, is often rearranged in carcinomas, typically showing distal loss by unbalanced translocation. We analysed 8p rearrangements in 48 breast, pancreatic and colon cancer cell lines by fluorescence in situ hybridization (FISH) and array comparative genomic hybridization, with a tiling path of 0.2 Mb resolution over 8p12 and 1 Mb resolution over chromosome 8. Selected breast lines (MDA-MB-134, MDA-MB-175, MDA-MB-361, T-47D and ZR-75-1) were analysed further. Most cell lines showed loss of 8p distal to a break that was between 31 Mb (5′ to NRG1) and the centromere, but the translocations were accompanied by variable amplifications, deletions and inversions proximal to this break. The 8p12 translocation in T-47D was flanked by an inversion of 4 Mb, with a 100 kb deletion at the proximal end. The dicentric t(8;11) in ZR-75-1 carries multiple rearrangements including interstitial deletions, a triplicated translocation junction between NRG1 and a fragment of 11q (unconnected to CCND1), and two separate amplifications, of FGFR1 and CCND1 . We conclude that if there is a tumour suppressor gene on 8p it may be near 31 Mb, for example WRN; but the complexity of 8p rearrangements suggests that they target various genes proximal to 31 Mb including NRG1 and the amplicon centred around ZNF703/FLJ14299.
Rearrangement of 8p, with loss of distal 8p, is one of the most frequent genomic events in a range of common epithelial cancers, but, in spite of much effort, their role in carcinogenesis is unclear (reviewed by Birnbaum et al., 2003). Classical cytogenetic analysis of several cancers, including breast, shows unbalanced translocations breaking around 8p12, with loss of distal 8p (Dutrillaux, 1995; also Teixeira et al., 2002). Comparative genomic hybridization (CGH) also shows frequent losses of distal 8p in breast, colorectal, prostate, bladder and other cancers (e.g. Davidson et al., 2000; Abdel-Rahman et al., 2001; Karan et al., 2003; Hurst et al., 2004; Hwang et al., 2004), and more recently array CGH has begun to provide more detail of these losses (e.g. Paris et al., 2003; Douglas et al., 2004; Hurst et al., 2004; Nakao et al., 2004; Garcia et al., 2005). Distal 8p also frequently shows loss of heterozygosity (LOH), which would often reflect loss through unbalanced translocation (for references see Adams et al., 2005).
There is also, often, amplification on 8p, proximal to the loss. In breast, for example, classical cytogenetic banding shows hsrs (homogeneously staining regions, i.e. large-scale amplifications) attached to 8p12 (Bernardino et al., 1998) and fluorescence in situ hybridization (FISH), CGH and Southern blotting show amplification generally in the region of 8p12 (Dib et al., 1995; Bautista and Theillet, 1998; Davidson et al., 2000), which often, but not invariably, includes FGFR1, and may occasionally extend as far distally as NRG1 (Ugolini et al., 1999). Recent array CGH studies that examine the 8p12 region in breast cancer have concentrated on the definition of the 8p12 amplicon (Ray et al., 2004; Garcia et al., 2005; Prentice et al., 2005). The amplicon has been defined to extend from 36.9 to 37.9 Mb (Garcia et al., 2005). The ZNF703/FLJ14299, SPFH2, BRF2 and RAB11FIP1 genes map to this region and their amplification correlates with overexpression. This amplicon was present in 24% of primary breast tumours and found to correlate with poor survival (Garcia et al., 2005; Prentice et al., 2005). The long-suspected amplification target, FGFR1, was excluded from the current consensus amplicon and detected as amplified in fewer, 10–15%, of cases (Theillet et al., 1993; Ugolini et al., 1999; Prentice et al., 2005). Amplification of proximal 8p is also seen in other carcinomas (Veltman et al., 2003; Nakao et al., 2004).
A further level of complexity is suggested by detailed analysis of the translocations that result in loss of distal 8p in cell lines. Firstly, in five breast and two pancreatic cell lines the breakpoints were clustered at the NRG1 gene (Adelaide et al., 2003, Huang et al., 2004). NRG1 encodes the heregulins/neuregulin 1 family of ligands, which bind to ErbB2, 3 and 4. Our working hypothesis is that the translocations result in aberrant expression of normal or modified heregulins, which act in an autocrine loop on the cancer cells, as in at least one of the cell lines, MDA-MB-175, the translocation results in fusion of NRG1 and secretion of a fusion protein with soluble heregulin-like activity (Schaefer et al., 1997; Liu et al., 1999; Wang et al., 1999). We have subsequently shown that breakpoints in NRG1 are present in about 6% of uncultured breast tumours (Huang et al., 2004 also supported by Prentice et al., 2005).
We had previously found another possible cluster of breaks about 2.5 Mb proximal to NRG1: three breast cancer cell lines, MDA-MB-361, T-47D, and ZR-75-1, had a break at about 34–35 Mb (Courtay-Cahen et al., 2000). This breakpoint cluster region is clearly distinct from the NRG1 cluster, as in two of the lines NRG1 is lost altogether. These translocations were accompanied by additional deletions, duplications and low-level amplifications (most of these amplifications are distinct from the consensus amplicon described above), suggesting that 8p rearrangements in general might be complex.
We describe here a high-resolution analysis of rearrangements of 8p in carcinoma cell lines by FISH and array CGH. Our analysis shows a clear pattern of 8p loss extending, at least, to 30 Mb suggesting that the rearrangements could have multiple genetic targets, and shows that the rearrangements are often highly complex.
Rearrangements of 8p were identified in a set of breast, colon and pancreatic cancer cell lines, for most of which we had previously obtained SKY (spectral karyotyping) karyotypes (Davidson et al., 2000; Abdel-Rahman et al., 2001; Sirivatanauksorn et al., 2001; available at http://www.path.cam.ac.uk/~pawefish).
Rearrangements were determined at the cytogenetic level by metaphase FISH (Figure 1). Of 35 lines examined that had rearrangements of chromosome 8, 27 had breaks on 8p or at the centromere, as determined by 8q arm paint (ideograms of these chromosome 8 rearrangements are available at http://www.path.cam.ac.uk/~pawefish/). These were further analysed using 8p11.2 and 8p22 band-specific paints, and BACs at the NRG1 and FGFR1 loci. Chromosomes with breaks between NRG1 and 8p22 were then tested with a BAC containing WRN. This gave a low-resolution cytogenetic classification of breakpoints on 8p (Figure 1).
Array comparative genomic hybridization
To map rearrangements in more detail in a number of lines, and to discover additional examples of cell lines with a breakpoint in 8p12, we used array CGH (Figure 2). Array CGH can detect breakpoints of unbalanced chromosome rearrangements because a copy number change occurs at the breakpoints. We used a custom array that included near-tiling-path coverage of the 8p12 region, approximately from WRN to FGFR1, together with roughly 1 Mb coverage of the rest of chromosome 8, and analysed 24 of the 27 cell lines identified in the FISH screen that had 8p rearrangements, and a further 24 breast cancer cell lines that have been karyotyped by SKY or M-FISH (Supplementary Table 1): a total of 48 cell lines, of which 39 show some type of rearrangement on 8p.
The ability of the array CGH to detect copy number changes and identify breakpoints was remarkable. In most cell lines clear step changes in copy number were visible and the breakpoint could generally be placed to the nearest BAC. As shown in Figure 3, in several lines we were able to compare the array CGH to the detailed FISH mapping of the chromosomes that is described below. For example, in T-47D, FISH had shown that there was a breakpoint within BAC RP11-419I6 and a deletion of most of BAC RP11-90P5: both are clearly visible in the array CGH. In MDA-MB-361 and ZR-75-1, where FISH showed multiple copy number changes in this region, the copy changes corresponded clearly to the array CGH (Figure 3).
The pattern of rearrangement that results in loss of distal 8p
Taken together, at low resolution, Figures 1 and 2a show that the majority of lines show the typical pattern of loss of distal 8p, to form unbalanced translocations, terminal deletions or isoderivatives. In Figure 2a, the cell lines have been approximately ranked according to the main breakpoint, to illustrate the distribution of breaks. This is necessarily crude, as in several lines the transition from gain to loss occurs in more than one step (e.g. SW403 has a break at the centromere and in proximal 8p12), and some lines show steps both up and down in copy number.
However, Figure 2a shows clearly that most lines have lost distal 8p and the break(s) between gain and loss are mostly in the interval NRG1 to FGFR1 or at the centromere. The WRN to FGFR1 region is covered at higher resolution by our array (average 1 BAC per 0.16 Mb) and Figure 2b shows 31 cell lines that show copy number changes in this region as a heat map. The lines have been ordered slightly differently from Figure 2a, roughly according to the most distal fall in copy number.
Several lines show a single step in copy number (Figure 2b), with loss of distal sequences, usually as a result of unbalanced translocation (e.g. PaTuI, and, at this resolution, T-47D), or sometimes as an isoderivative (e.g. MDA-MB-453). However, several lines have more than one copy number change in the region, showing that the low-resolution picture in Figure 2a, and the cytogenetic picture, oversimplifies the pattern of breaks. Several (Cama-1, HCC1500, MDA-MB-134, MDA-MB-175, MDA-MB-361, SUIT-2, SUM44, SUM52 and ZR-75-1) show at least low-level amplification of a fragment immediately proximal to the main break. All of these, except Cama-1, have been reported before (Lemieux et al., 1996; Courtay-Cahen et al., 2000; Adelaide et al., 2003; Ray et al., 2004; Garcia et al., 2005) and all, except those in Cama-1 and the SUM lines, have been verified by us using FISH with BACs and chromosome 8 paint. HCC38 and UACC812 have a small amplicon substantially proximal to the main break, and two BACs spanning the FGFR1 locus are amplified in HCC1599. Several lines show a small region of relative loss within the retained region proximal to the break (C70, MDA-MB-361, HCC1569, ZR-75-1 (more than one), HCC1806, SW837, UACC812), and one line, HCC70, has a small region of additional loss distal to the main break at the centromere. These regions of loss could be interstitial deletions, but they may also be gaps between regions that are present in extra copies, as shown below for MDA-MB-361 and ZR-75-1, or deletions at inversion junctions, as shown below for T-47D. MDA-MB-175 is an extreme example, with a small fragment around NRG1 present in several copies inserted into chromosome 11 material (Figure 3), and the remainder of 8p lost.
Several of the breast lines also showed substantial amplifications between 37 and 40 Mb, as we have described previously for breast tumour tissues and some cell lines (Garcia et al., 2005). Several cell lines not in our previous study that showed amplification or gain in this region included Cama-1, HCC1599, HCC38, UACC812, ZR-75-1, and the colon line SW837 (Figure 2b).
In summary, the overall pattern is very variable but there is loss of distal 8p; the region immediately proximal to the loss is often amplified (or at least duplicated), and in several cases a region proximal to this may actually show loss. In some cases there is a complex combination of losses and duplications/amplifications, as analysed in more detail below in specific cases.
Retention of distal 8p
Two lines showed an additional rearrangement: retention of a fragment of 8p extending from 8pter to a break around 8p22. Metaphase FISH detected this in PMC42 (Figure 1). Array CGH detected a further example in HCC1806, and mapped both breaks to between 12 and 14 Mb. (Although in Figure 2a Colo357 and UACC812 appear similar, this is due to small copy number fluctuations about the threshold and is not convincing). In MCF7 and SKBR3, FISH showed that fragments corresponding to the whole of 8p were present (Figure 1).
Detailed fluorescence in situ hybridization analysis of individual cell lines
The rearrangements of chromosome 8 in certain cell lines were mapped in more detail by FISH. Breaks were mapped by FISH to the nearest BAC in a near-contiguous set that covers the region, as summarized in Figure 4, and detailed in Figure 3 and Supplementary Table 2.
In the breast cancer cell line T-47D we found that the 8p translocation was accompanied by cryptic inversion and deletion. This line has a der(8)t(8;14)(p12;q12), which is present in two copies with two apparently normal chromosome 8 s (Morris et al., 1997; Courtay-Cahen et al., 2000, Figure 3).
We first mapped the expected breakpoints on chromosomes 8 and 14, but found they were not joined. Conventional FISH and array CGH both positioned the break on 8p within BAC RP11-419I6 (Figure 4). Fibre-FISH showed loss of about one-third of this BAC, placing the break at approximately 34.5 Mb (not shown). On chromosome 14, following FISH with a series of BACs on metaphase chromosomes (Supplementary Table 2), fibre-FISH placed the break in BAC RP11-893D14 at 24.4 Mb (Figure 5a). In expectation that the BAC RP11-419I6 from chromosome 8 and RP11-893D14 from chromosome 14 would be joined, we hybridized the two BACs together on T-47D fibres. We were unable to find any joined copies.
To explain this, we hypothesized that sequences on 8p adjacent to the break were inverted (Figure 5b), as inversions could explain the complex 8p rearrangements in ZR-75-1 and MDA-MB-361. The site of the proximal inversion break was suggested by the partial deletion of BAC RP11-90P5: this deletion had been detected as a weak RP11-90P5 signal on the translocated 8;14 chromosomes, and in the array CGH (Figures 3 and 4).
To test the hypothesis, we first used interphase FISH to show that the relative distances between BACs at each end of the inversion changed as predicted (Figure 5c). We then confirmed the predicted junctions by fibre-FISH: the proposed junction of the chromosome 14 break with the proximal end of the inverted fragment of chromosome 8 was detected as contiguous signals from BACs that span the breaks, respectively, RP11-893D14 and CTD-2277D22 (Figure 5e). The junction created by the inversion, between chromosome 8 sequences, was similarly confirmed using BACs RP11-90P5 and RP11-419I6 (Figure 5f).
The extent of the deletion of chromosome 8 sequence at the proximal end of the inversion, in RP11-90P5, was estimated to be between 80 and 150 kb using fosmid clones (Figure 5g) and BAC metaphase and fibre-FISH (Figure 5e and f). The distal end was between 38.07–38.08 Mb and the proximal end was between 38.170–38.215 Mb.
In conclusion, the segment from about 34.50 (RP11-419I6) to about 38.15 Mb (RP11-90P5) has been inverted, with loss of sequence contained within RP11-90P5, and the inverted piece is joined to chromosome 14 (Figure 5). Genes that might be affected include ASH2L to BAG4, in the deletion, and genes that flank the new junctions including EIF4EBP1, BAG4 or DDHD2, UNC5D, and, on chromosome 14, STXBP6 (Map, Figure 4).
The breast cancer cell line ZR-75-1 has an extraordinarily complex rearrangement of 8p. At cytogenetic resolution, it has a dicentric unbalanced translocation of chromosomes 8 and 11, dic(8;11)(p12;q12), and three apparently normal chromosome 8 s. One of the ‘normal’ 8 chromosomes proves to have some of the changes present on the dic(8;11) (Figures 3 and 4) (Courtay-Cahen et al., 2000).
A reconstruction of the dicentric chromosome is shown in Figure 6a based on the work described below, and our earlier less detailed work (Courtay-Cahen et al., 2000; Adelaide et al., 2003). The main features are as follows: a translocation of the NRG1 gene to chromosome 11 (Adelaide et al., 2003; Huang et al., 2004); triplication of a complex fragment that includes this 8;11 translocation junction and other neighbouring segments of chromosome 8 and 11 (previously reported as a duplication Courtay-Cahen et al., 2000); an interstitial deletion proximal to NRG1; a separate amplification of the FGFR1 region, which is also present on one of the untranslocated copies of chromosome 8; a second interstitial deletion proximal to FGFR1, which is associated with both copies of the FGFR1 amplification (Courtay-Cahen et al., 2000); and involvement of at least two fragments of chromosome 11q, including a separate amplification of the CCND1 region. This may not be complete as, for example, we would not have detected inversions.
Chromosome 11- NRG1 junction
We identified a fragment of 11q12 that was present in the dicentric and joined to the breakpoint on 8p in NRG1 (we had not previously investigated what the break in NRG1 was joined to (Adelaide et al., 2003)). Fluorescence in situ hybridization on extended metaphase chromosomes, followed by fibre-FISH (Figure 6f, Supplementary Table 2), showed that an approximately 300 kb fragment of 11q12, extending from approximately 62.8 (the midpoint of BAC RP11-996J9) to 63.1 Mb, was present in the dicentric and triplicated in the same locations as NRG1 (Figures 6b–d and Supplementary Table 2). The proximal end of this fragment was shown by interphase measurements then fibre-FISH to be fused to the break in NRG1 (Figure 6g).
Other 11q sequences present on the dic(8;11)
The only other 11q sequences we detected were a fragment of 4–7 Mb spanning CCND1 (maximally 64.7 to 71.9 Mb). Bacterial artificial chromosomes (BACs) RP11-126P21, RP11-300I6 (CCND1) showed amplification and RP11-512I24 was present in a single copy (Figure 6h and i; Supplementary Table 2). RP11-300I6 also hybridized in a separate location between the amplified and single copy of FGFR1 (Figure 6g). Other regions of 11q were absent by BAC FISH (Supplementary Table 2), and by hybridization of amplified, flow-sorted copies of the dicentric to the 1 Mb genomic array of Fiegler et al., 2003 (KAB, unpublished observations).
Details of chromosome 8 fragments
Different parts of the region between NRG1 and FGFR1 were present in anything from zero to three copies, as shown by FISH with BACs from 8p12, confirmed by array CGH (Figures 3 and 4; Supplementary Table 2). Three separate fragments were triplicated together with the triplicated fragment of 11q. Additionally there were two regions present in two copies, two regions present in single copies, and a deletion proximal to NRG1. We propose that this is due to the duplication of some parts and then reduplication of part of the already duplicated region. It appears that the second and third copies are successively placed further from the 8 centromere, because successively fewer of the chromosome 8 fragments are present. As the regions duplicated and triplicated were not originally contiguous, deletions and/or inversions must have occurred before or during the duplications. The proximal boundary of the triplication is within BAC RP11-90P5 (38.1 Mb), just distal to the FGFR1 BAC RP11-350N15.
Amplifications of FGFR1 and CCND1
The FGFR1 locus (detected in this study by BAC RP11-350N15), although not in the triplication, was present as a small amplification, in at least three copies clustered together, on both the translocated and one untranslocated chromosome (Figure 6h and i) as previously described (Courtay-Cahen et al., 2000). The hybridization signals were tightly clustered and separate from the triplication just described, hence it seems to be an independent event. Some sequences, hybridizing to BACs CTD-2343B20 and RP11-90P5, just distal to RP11-350N15, are also amplified on the untranslocated chromosome 8, although not on the dic(8;11). On the dicentric chromosome this region (CTD-2343B20 and RP11-90P5) is included in the triplication.
We confirmed that, in the dic(8;11) chromosome, the FGFR1 amplification was also well separated from the cluster of CCND1 signals (Figures 6h and i).
Centromeric to FGFR1 there is a second interstitial deletion, present on both the dicentric and the ‘normal 8’ that has amplified FGFR1. This was previously detected by a YAC (Courtay-Cahen et al., 2000). The array CGH suggests that the deletion extends roughly from 38.6 to 40.4 Mb. The array CGH also suggests the region proximal to this deletion and before the centromere may also be amplified, but this was not investigated further. This region may include the ANK1 and PLAT genes which can be amplified separately (Ugolini et al., 1999).
MDA-MB-361 has two complex 8p translocations that appear both to have evolved from the same 8;17 translocation (Figure 3) (Morris et al., 1997; Courtay-Cahen et al., 2000). Both have duplicated or triplicated fragments of 8p from the region proximal to the breakpoint, and the extra copies are inserted into the adjacent chromosome 17 material (Courtay-Cahen et al., 2000). We mapped the distal breakpoints of the translocation and extra copies to the nearest BAC by FISH (Figures 3 and 4; Supplementary Table 2). Array CGH placed the proximal limit of the duplicated segment at approximately 39 Mb.
MDA-MB-134 is known to have a high copy number amplification of sequences from 8p12, including FGFR1, forming an hsr with sequences from 8q and chromosome 11 (Figure 3) (Lemieux et al., 1996; Bautista and Theillet, 1998). The distal end of the 8p amplification was mapped to within BAC RP11-155H15, at 34.7 Mb by FISH and array CGH (Figure 3; Supplementary Table 2). The array CGH shows the amplification continuing proximally, beyond FGFR1, to about 42 Mb. The amplified region shows varying levels of gain, suggesting that the amplification may be the result of repeated joining and re-breaking with potentially multiple boundaries.
MDA-MB-175 has a translocation of 8p12 at NRG1 that creates a fusion with the DOC4 gene on chromosome 11 (Wang et al., 1999; Liu et al., 1999; Adelaide et al., 2003). The fragment containing the junction is amplified (Figure 3). We mapped the boundaries of the amplified fragment of 8p using array CGH and FISH, and it extends approximately from 32.4 to 33.2 Mb (Figures 3, 4 and Supplementary Table 2). The copies are present in a small cluster inserted into chromosome 11 material, remote from other parts of chromosome 8 (Figure 3).
Clusters of breakpoints at NRG1 and UNC5D
We have previously reported that seven cell lines had breaks within or immediately 5′ to NRG1, (Adelaide et al., 2003). The array CGH detected two more examples, in HCC1569 and HCC1806, which were confirmed by FISH: the breaks were respectively at about 32.1–31.5 Mb (between BACs RP11-22F19 and RP11-275E10 and between BACs RP11-212N14 and RP11-566H8). The known NRG1 breaks in MDA-MB-175, PaTuI, SUM52, SUIT-2, UACC812 and ZR-75-1 were also detected. The break in HCC1937 is reciprocal and so was not detected.
Interestingly, in two cell lines, HCC38 and MDA-MB-157, there is a break just proximal to NRG1 thus removing this gene in the translocated chromosomes (Figure 2b).
Several cell lines have breaks 5′ to UNC5D at 35.2–35.8 Mb. We previously found three breaks near 34.5 Mb which appeared to be a cluster, in the breast cell lines MDA-MB-361, T-47D and ZR-75-1 (Courtay-Cahen et al., 2000). These were mapped precisely by FISH and array CGH, to 34.2, 34.5 and 34.9 Mb, respectively (Figure 4). At least five more cell lines had breaks 5′ to UNC5D by array CGH: CF-PAC, C70 and HCC1569 at 34.1–34.4 Mb, MDA-MB-134 at 34.7 Mb, and MDA-MB-453 at 35.5 Mb. VP229 and BT20 may also have breaks nearby, but they were not precisely located.
Our main observations and interpretation are summarized in Figure 7. At cytogenetic resolution, the overall pattern of 8p rearrangement is simple (Figure 1, and 2a): the great majority of cases show loss of distal 8p, with breakpoints scattered between WRN (31 Mb) and the centromere (44–47 Mb), mostly between NRG1 and FGFR1 (32 and 38.4 Mb). This would be compatible with the ‘traditional’ view that there is a tumour suppressor gene on 8p of general importance in carcinomas (reviewed by Birnbaum et al., 2003). However, at higher resolution, there seems to be a great deal more complexity than this simple idea would predict.
Is there a tumour suppressor?
If there is a tumour suppressor gene on 8p, we would expect the breaks to be scattered between the centromere and the gene: Figure 2 shows breaks scattered between the centromere and the distal end of NRG1 (32 Mb), making genes immediately distal to NRG1, including WRN (31 Mb), obvious candidates. Most of the candidates proposed previously – none of which are particularly convincing – are in the region 3–24 Mb (e.g. DLC1, DBC2, FEZ1/LZTS1 and TUSC3; for references see Birnbaum et al., 2003; Toomes et al., 2003; Asatiani et al., 2005) and our array CGH (Figure 2) gives no obvious support to this region as it is substantially distal to the breaks we have detected. However, if breaks are concentrated in regions of relative fragility, this argument would not hold.
Alternative theories of 8p breakage
Among alternative explanations for 8p breakage is that distal loss occurs simply in order for amplification of the region around FGFR1 to occur by breakage-fusion-bridge cycles, and breaks may be clustered if there are sites of preferential breakage (Birnbaum et al., 2003). This idea does not fit the data particularly well: some lines, including HCC70, MCF7, and SKBR3, have no 8p material retained; others have 8p breaks but no detectable amplification – HCC1143, MDA-MB-453, BT20, T-47D, C70, for example; and there are cases where loss occurs proximal to amplification – MDA-MB-175 and ZR-75-1, for example. While amplification often occurs immediately proximal to a break, we favour the interpretation that oncogenic fusions occur at the break that are subsequently amplified by a mechanism independent of the original break. A possible variation on this idea would be that the breaks permit inversion.
Clusters of breakpoints at NRG1 and UNC5D
A completely different interpretation is that some of the 8p translocations break at specific genes and activate or inactivate them. Although 8p breakpoints are spread, some are clustered in the NRG1 gene and there may be a second cluster at UNC5D.
We recently reported that seven breast or pancreatic cancer cell lines have breaks at or within the 5′ end of NRG1 that could activate or fuse NRG1, and this study adds two more lines. Similar breaks occur in fresh tumours (Adelaide et al., 2003; Huang et al., 2004). These translocations seem to specifically target NRG1 and to be oncogenic: one of the NRG1 translocations, in HCC1937 is reciprocal; in at least one case, MDA-MB-175, the result is secretion of an Nrg1 fusion protein with biological activity (Schaefer et al., 1997); in several cases the translocation junction is amplified; and there is a strong correlation between translocation and expression of NRG1 transcripts (Y-L Chua and PAWE, unpublished). Nrg1 proteins are ligands for the ERBB receptor family and overactivity of NRG1 promotes mammary tumour development in transgenic mice (Krane and Leder, 1996). At present, these translocations appear to be distinct events from others affecting 8p – certainly in the two lines examined in detail by FISH, MDA-MB-175 and ZR-75-1, the NRG1 junction behaves as a separate rearrangement to other events on 8p12. On the other hand, most translocations of 8p result in complete loss of a copy of NRG1, so translocation of NRG1 is not a general explanation for 8p rearrangement.
UNC5D, at 35.2–35.8 Mb, is also a candidate target gene, its 5′-end lying proximal to a number of breaks between 34.1 and 35.5 Mb, in at least eight cell lines (HCC1569, MDA-MB-134, MDA-MB-361, MDA-MB-453, T-47D, ZR-75-1, CF-PAC and C70). We originally reported three of these which appeared to be a cluster (Courtay-Cahen et al., 2000). We had previously found four breast tumours with breaks at about 35.5 Mb (Garcia et al., 2005).
UNC5D could well be a critical gene, as it is one of the four human homologues of Caenorhabditis elegans unc5, which are receptors for Netrin. The Netrin system, including the Netrin receptors UNC5A-D, DCC and neogenin, were originally implicated in axon guidance, but are also active in survival signalling, and are involved in morphogenesis of the mammary gland (Hinck, 2004). The UNC5s have been proposed as tumour suppressors (Arakawa, 2004), but could equally be activated, which would be the more likely consequence of the translocations reported here.
At present, however, there is not much evidence that UNC5D is targeted. The breaks are mostly quite a long way 5′ of the gene, and with more breaks mapped in this region (Figure 2b), it is not clear that they form a distinct cluster. Furthermore, the complexity of the rearrangements associated with these breaks, particularly our finding that the adjacent segment is inverted in T-47D, may well mean that these breaks are not in themselves significant.
The overall pattern of amplifications is complex, but we suggest they can be divided into at least three separate categories.
First, there is the consensus amplicon around the genes ZNF703/FLJ14299 to RAB11FIP1at about 36.9–37.9 Mb. FGFR1, at 38.4 Mb, although originally proposed as the principle target for amplification in this region, is not included (Ray et al., 2004; Garcia et al., 2005; Prentice et al., 2005). This amplicon we defined using the same array on breast tumour tissues (Garcia et al., 2005), and it was independently identified by Prentice et al. (2005). In the present study Cama-1 provided another example of high-level amplification of the ZNF703/FLJ14299 region with only slight gain of FGFR1.
Secondly, our data make it clear that FGFR1 is sometimes a distinct target of amplification, though less common. FGFR1 is amplified or gained without the ZNF703/FLJ14299-RAB11FIP1 region in UACC812, and probably in HCC1599 and SW837 (colon), though these cases may include the proximal part of the consensus amplicon. Most clearly, in ZR-75-1, FGFR1 amplification occurs in a physically separate location from triplication of the region that includes ZNF703/FLJ14299- RAB11FIP1, and so cannot simply be a passenger.
A third class of low-level amplification, sometimes only duplication or triplication, was seen immediately proximal to breaks in many cases, suggesting either that the mechanism of translocation often results in their formation or that the translocation junction is oncogenic and so tends to be amplified during subsequent evolution. For example, several of the breaks at NRG1are amplified – in MDA-MB-175, SUM52, ZR-75-1 and Suit2 – as are the breaks in MDA-MB-134 (34.7 Mb), and MDA-MB-361 (34.2 Mb). Other examples extend into the ZNF703/FLJ14299 and FGFR1 regions.
We suggest that inversions may be frequent in these tumours. In T-47D we demonstrated an inversion accompanying the 8p translocation (Figure 5), and inversions, before or during translocation, would explain the alternating regions of deletion and duplication seen in ZR-75-1. Inversions are difficult to detect but are an attractive explanation for the amplification of apparently distinct regions of one chromosome (e.g. Orsetti et al., 2004): inversion, creating an oncogenic junction, would be followed by amplification of one inversion junction, manifest as two apparent regions of amplification by array CGH, in the same way as amplifications on distinct chromosomes can be coamplifications (Guan et al., 1994; Bautista and Theillet, 1998; Volik et al., 2003).
Rearrangements are complex
In all cases where we investigated rearrangement in detail, the translocations are accompanied by complex additional events, including inversion, deletion proximal to the main translocation junction, and multiple independent duplications or amplifications. In ZR-75-1, for example (Figure 6), on the single dic(8;11) there is translocation of NRG1, triplication of the translocation together with the region around the ZNF703/FLJ14299 amplicon, and physically separate amplification of CCND1 and FGFR1. This complexity cannot be dismissed as an in vitro artefact, as each of the events is common in uncultured breast tumours. In T-47D, an apparently simple 8;14 translocation also has an inversion and deletion of sequences at the inversion junction.
There are now many examples, where translocations are accompanied by additional rearrangements on a scale of hundreds of kilobases or larger. For example, the translocation-plus-inversion structure that we found in T-47D probably occurs in some examples of the recurrent translocations ETV6-ABL1 (Van Limbergen et al., 2001) and MLL-AF10 (Klaus et al., 2003 and references therein). In both cases the inversion is necessary for the two participating genes to be in the same orientation. Other translocations may be accompanied by deletions (e.g. Rousseau-Merck et al., 1999; Sinclair et al., 2000). Similarly, constitutional translocations are often accompanied by complex cryptic rearrangements (Gribble et al., 2005), and as these are likely to be single events, they suggest that the mechanism of translocation often produces complex rearrangements.
8p12 rearrangements in real tumours
Many of the rearrangements that we have reported in cell lines affecting 8p12 have also been detected in real tumours. The recurrent breakpoint effecting NRG1 (Huang et al., 2004), the ZNF703/FLJ14299-RAB11FIP1 amplicon and a recurrent breakpoint at UNC5D in breast cancers (Garcia et al., 2005) have been previously reported. In our previous publication we compare array CGH results from breast cancer cell lines and breast tumours and the overall pattern of rearrangement is very similar, showing loss of distal 8p, gain of 8q and the most frequent site of copy number change occurring at 8p12, often with amplification (Garcia et al., 2005). This comparison suggests that the events we describe here in cell lines are relevant to real carcinomas.
In conclusion, at present there is no grand unifying theory of 8p rearrangement. If it is primarily driven by loss of a distal tumour suppressor gene, we have evidence that the gene is close to NRG1. However, the tumour suppressor hypothesis does not account well for the details and complexity of rearrangements upstream, such as NRG1 translocations and amplification of more proximal sequences, unless these are merely consequences of the breakage event, or initial translocation or deletion that makes the chromosome unstable.
Materials and methods
Cell culture, metaphase preparation and fluorescence in situ hybridization
Cell lines are listed in Supplementary Table 1. Metaphase spreads and FISH were carried out using standard methods as detailed in Alsop et al. (2006). For metaphase FISH, chromosome paints kindly provided by Professor MA Ferguson-Smith were labelled using nick translation either directly with spectrum orange-dUTP (Vysis/Abbot Laboratories, IL, USA) or indirectly with biotin-dUTPs (Roche Diagnostics, Basel, Switzerland). Bacterial artificial chromosomes were labelled indirectly, using nick translation with digoxigenin-, estradiol- and biotin-labelled dUTP (Roche Diagnostics). Band-specific paints for 8p11.2, and 8p22 were from Research Genetics/Invitrogen (Paisley, UK). Chromosome 8q arm paint pre-labelled with digoxigenin was from Cambio (Cambridge, UK). Detection of labelled, incorporated dUTPs was performed using the appropriate antibody or avidin conjugates (sheep anti-digoxigenin-FITC (Roche Diagnostics), avidin-Cy3, or avidin-Cy5 (GE Healthcare, Buckinghamshire, UK) and biotinylated anti-avidin (Vector Laboratories Inc., Burlingame, CA, USA). Images were captured on a Nikon Eclipse 800 microscope using Cytovision software (Applied Imaging, Carlsbad, CA, USA).
Metaphase spreads with more extended chromosomes were made from ZR-75-1 by culturing the cells with 40 μg/ml bromodeoxyuridine for 16 h, addition of 5 μg/ml ethidium bromide, and harvesting an hour later without colcemid treatment.
The BAC and PAC clones used in FISH experiments are detailed in Supplementary Table 2. Clone positions are given according to Human genome sequence NCBI Build 35. Bacterial artificial chromosome clones were from the RPCI-11 (RP11) library, the GS1 library and Caltech whole genome libraries (CTA, CTC and CTD). They were identified from the emerging genome assembly or BAC end sequence data displayed on the UCSC Genome Browser website (http://genome.ucsc.edu), or the Washington University Fingerprint maps. All BACs had been checked for chimerism and location on 8p using conventional FISH on normal metaphases. They were obtained from the Wellcome Trust Sanger Institute (Hinxton, UK), Research Genetics/Invitrogen, or in some cases from the laboratory that had sequenced them.
Fosmids were selected from the UCSC genome browser (http://genome.ucsc.edu) and obtained from the Sanger Institute. They were also checked by hybridization to normal metaphase chromosomes. Isolated fosmid DNA was amplified using Templiphi (GE Healthcare), and labelled by nick translation.
Fibre-FISH was based on the method described by Mann et al. (1997). Briefly, 10 μl of fixed metaphase preparation was smeared across the width of a polylysine-coated slide (Polyprep, Sigma, Dorset, UK) approximately 1 cm from the top of the usable slide surface and allowed to dry for a minute before being submerged vertically in lysis buffer (0.5% (w/v) SDS, 50 mM EDTA, 0.2 M Tris-HCl pH7.4) in a coplin jar and left to stand with the cell preparation just below the buffer surface. After 5 min 94% ethanol was slowly run on to the buffer surface to form an upper layer until the entire slide was covered. After incubating for 10 min, the slide was pulled out of the solutions slowly and steadily at a 30° angle, submerged in 70% ethanol, incubated for 30 min, and dehydrated in an ethanol series. Slides were hybridized as for metaphase spreads, using digoxigenin-dUTP and biotin-UTP (both from Roche Diagnostics) haptens and mouse anti-digoxin (Sigma) followed by alexafluor-594 anti-mouse (Molecular Probes, Invitrogen) for the digoxin-labelled DNA and streptavidin-alexafluor-488 conjugate (Molecular Probes, Invitrogen), biotinylated anti-streptavidin (Vector laboratories) and streptavidin-alexafluor-488 conjugate in three layers for biotin-labelled DNA. Estimates of breakpoint position from fibre-FISH were made by comparing the length of a whole BAC of known length to the length of a BAC fragment, averaged over 20 captured images.
Distances between sequences were estimated as described by (van den Engh et al., 1992), by hybridizing BACs to interphase nuclei. The average separation of FISH signals in interphase nuclei is proportional to the square root of the genomic distance between the hybridization targets, up to 1 Mb separation, so it easily distinguishes sequences that are adjacent from sequences that are a few hundred kilobases or more apart (van den Engh et al., 1992). Distances between BAC signals were measured in arbitrary units, using Cytovision software (Applied Imaging), except one unit was subtracted from values to give a zero result for zero distance, and averaged over at least 15 examples.
Array CGH was performed essentially by the methods of Fiegler et al. (2003), on an in-house array (Huang et al., 2004; Garcia et al., 2005), that comprised BACs approximately 10 Mb apart across the whole genome; 1–1.5 Mb apart over chromosome 8; and at near-tiling path density over 8p12, with 58 BAC clones covering 9.6 Mb of 8p12 from 31.0 (RP11-473A17) to 40.6 Mb (RP11-51K12). All but one gap between successive BAC ends were 0.3 Mb or less, the one larger gap being 0.6 Mb near the proximal end. Bacterial artificial chromosome clones from outside the 8p12 region were from the 1 Mb cloneset of Fiegler et al. (2003). Labelling of genomic DNA, hybridization and post-hybridization washes were carried out as previously described (Garcia et al., 2005). Arrays were scanned using an Axon Genepix 4100 and the data acquired using Genepix4.1 software (Molecular Devices, Union City, CA, USA). Data analysis was carried out using Microsoft Excel, followed by CGHAnalyzer (http://acgh.afcri.upenn.edu; Margolin et al., 2005) to compare array CGH profiles and generate Figures 2a and b. Results were calculated relative to the median Cy3:Cy5 ratio for the whole array excluding sex chromosomes, and expressed on a log2 scale. The array CGH data will be available on our website http://www.path.cam.ac.uk/~pawefish.
Abdel-Rahman WM, Katsura K, Rens W, Gorman PA, Sheer D, Bicknell D et al. (2001). Proc Natl Acad Sci USA 98: 2538–2543.
Adams J, Williams SV, Aveyard JS, Knowles MA . (2005). Cancer Res 65: 66–75.
Adelaide J, Huang HE, Murati A, Alsop AE, Orsetti B, Mozziconacci MJ et al. (2003). Genes Chromosomes Cancer 37: 333–345.
Alsop AE, Teschendorff A, Edwards PA . (2006). Cancer Genet Cytogenet 164: 97–109.
Arakawa H . (2004). Nat Rev Cancer 4: 978–987.
Asatiani E, Huang WX, Wang A, Rodriguez Ortner E, Cavalli LR, Haddad BR et al. (2005). Cancer Res 65: 1164–1173.
Bautista S, Theillet C . (1998). Genes Chromosomes Cancer 22: 268–277.
Bernardino J, Apiou F, Gerbault-Seureau M, Malfoy B, Dutrillaux B . (1998). Genes Chromosomes Cancer 23: 100–108.
Birnbaum D, Adelaide J, Popovici C, Charafe-Jauffret E, Mozziconacci MJ, Chaffanet M . (2003). Lancet Oncol 4: 639–642.
Courtay-Cahen C, Morris JS, Edwards PA . (2000). Genomics 66: 15–25.
Davidson JM, Gorringe KL, Chin SF, Orsetti B, Besret C, Courtay-Cahen C et al. (2000). Br J Cancer 83: 1309–1317.
Dib A, Adelaide J, Chaffanet M, Imbert A, Le Paslier D, Jacquemier J et al. (1995). Oncogene 10: 995–1001.
Douglas EJ, Fiegler H, Rowan A, Halford S, Bicknell DC, Bodmer WF et al. (2004). Cancer Res 64: 4817–4825.
Dutrillaux B . (1995). Adv Cancer Res 67: 59–82.
Fiegler H, Carr P, Douglas EJ, Burford DC, Hunt S, Scott CE et al. (2003). Genes Chromosomes Cancer 36: 361–374.
Garcia MJ, Pole JC, Chin SF, Teschendorff A, Naderi A, Ozdag H et al. (2005). Oncogene 24: 5235–5245.
Gribble SM, Prigmore E, Burford DC, Porter KM, Ng BL, Douglas EJ et al. (2005). J Med Genet 42: 8–16.
Guan XY, Meltzer PS, Dalton WS, Trent JM . (1994). Nat Genet 8: 155–161.
Hinck L . (2004). Dev Cell 7: 783–793.
Huang HE, Chin SF, Ginestier C, Bardou VJ, Adelaide J, Iyer NG et al. (2004). Cancer Res 64: 6840–6844.
Hurst CD, Fiegler H, Carr P, Williams SV, Carter NP, Knowles MA . (2004). Oncogene 23: 2250–2263.
Hwang ES, DeVries S, Chew KL, Moore II DH, Kerlikowske K, Thor A et al. (2004). Clin Cancer Res 10: 5160–5167.
Karan D, Lin MF, Johansson SL, Batra SK . (2003). Int J Cancer 103: 285–293.
Klaus M, Schnittger S, Haferlach T, Dreyling M, Hiddemann W, Schoch C . (2003). Cancer Genet Cytogenet 144: 36–43.
Krane IM, Leder P . (1996). Oncogene 12: 1781–1788.
Lemieux N, Apiou F, Vogt N, Malfoy B, Dutrillaux B . (1996). Cancer Genet Cytogenet 90: 75–79.
Liu X, Baker E, Eyre HJ, Sutherland GR, Zhou M . (1999). Oncogene 18: 7110–7114.
Mann SM, Burkin DJ, Grin DK, Ferguson-Smith MA . (1997). Chromosome Res 5: 145–147.
Margolin AA, Greshock J, Naylor TL, Mosse Y, Maris JM, Bignell G et al. (2005). Bioinformatics 21: 3308–3311.
Morris JS, Carter NP, Ferguson-Smith MA, Edwards PA . (1997). Genes Chromosomes Cancer 20: 120–139.
Nakao K, Mehta KR, Fridlyand J, Moore DH, Jain AN, Lafuente A et al. (2004). Carcinogenesis 25: 1345–1357.
Orsetti B, Nugoli M, Cervera N, Lasorsa L, Chuchana P, Ursule L et al. (2004). Cancer Res 64: 6453–6460.
Paris PL, Albertson DG, Alers JC, Andaya A, Carroll P, Fridlyand J et al. (2003). Am J Pathol 162: 763–770.
Prentice LM, Shadeo A, Lestou VS, Miller MA, Deleeuw RJ, Makretsov N et al. (2005). Oncogene 24: 7281–7289.
Ray ME, Yang ZQ, Albertson DG, Kleer CG, Washburn JG, Macoska JA et al. (2004). Cancer Res 64: 40–47.
Rousseau-Merck MF, Versteege I, Legrand I, Couturier J, Mairal A, Delattre O et al. (1999). Cancer Res 59: 3152–3156.
Schaefer G, Fitzpatrick VD, Sliwkowski MX . (1997). Oncogene 15: 1385–1394.
Sinclair PB, Nacheva EP, Leversha M, Telford N, Chang J, Reid A et al. (2000). Blood 95: 738–743.
Sirivatanauksorn V, Sirivatanauksorn Y, Gorman PA, Davidson JM, Sheer D, Moore PS et al. (2001). Int J Cancer 91: 350–358.
Teixeira MR, Pandis N, Heim S . (2002). Genes Chromosomes Cancer 33: 1–16.
Theillet C, Adelaide J, Louason G, Bonnet-Dorion F, Jacquemier J, Adnane J et al. (1993). Genes Chromosomes Cancer 7: 219–226.
Toomes C, Jackson A, Maguire K, Wood J, Gollin S, Ishwad C et al. (2003). Genes Chromosomes Cancer 37: 132–140.
Ugolini F, Adelaide J, Charafe-Jauffret E, Nguyen C, Jacquemier J, Jordan B et al. (1999). Oncogene 18: 1903–1910.
van den Engh G, Sachs R, Trask BJ . (1992). Science 257: 1410–1412.
Van Limbergen H, Beverloo HB, van Drunen E, Janssens A, Hahlen K, Poppe B et al. (2001). Genes Chromosomes Cancer 30: 274–282.
Veltman JA, Fridlyand J, Pejavar S, Olshen AB, Korkola JE, DeVries S et al. (2003). Cancer Res 63: 2872–2880.
Volik S, Zhao S, Chin K, Brebner JH, Herndon DR, Tao Q et al. (2003). Proc Natl Acad Sci USA 100: 7696–7701.
Wang XZ, Jolicoeur EM, Conte N, Chaffanet M, Zhang Y, Mozziconacci MJ et al. (1999). Oncogene 18: 5718–5721.
We thank Huai-En Huang, Rachel Lyman, Nicholas J Roberts, Joanne M Staines, Melody Tabiner, and Katherine Williams for contributing to the FISH data and cell culture. Supported by Cancer Research UK, Biotechnology and Biological Sciences Research Council (BBSRC), Breast Cancer Campaign, Wellbeing, Medical Research Council (studentship for SLC), Cambridge Commonwealth Trust (studentships for AEA, KAB) and Sackler Fund (KAB).
About this article
Cite this article
Pole, J., Courtay-Cahen, C., Garcia, M. et al. High-resolution analysis of chromosome rearrangements on 8p in breast, colon and pancreatic cancer reveals a complex pattern of loss, gain and translocation. Oncogene 25, 5693–5706 (2006). https://doi.org/10.1038/sj.onc.1209570
- breast cancer
- array CGH
- chromosome translocation
- chromosome inversion
Breast Cancer Research (2021)
Applied Network Science (2019)
Cellular and Molecular Life Sciences (2018)
BMC Cancer (2012)
Structural analysis of the genome of breast cancer cell line ZR-75-30 identifies twelve expressed fusion genes
BMC Genomics (2012)