Oncogene (2006) 25, 7106–7116 & 2006 Nature Publishing Group All rights reserved 0950-9232/06 $30.00 www.nature.com/onc ONCOGENOMICS Genomic profiling of bone and soft tissue tumors with supernumerary ring using tiling resolution bacterial artificial microarrays

M Heidenblad1, KH Hallor1, J Staaf2,GJo¨ nsson2,A˚ Borg2,MHo¨ glund1, F Mertens1 and N Mandahl1

1Department of Clinical Genetics, Lund University Hospital, Lund, Sweden; and 2Department of Oncology, Lund University Hospital, Lund, Sweden

Ring chromosomes and/or giant marker chromosomes Keywords: sarcoma; array-based CGH; genome-wide; have been observed in a variety of human tumor types, but amplification; ring chromosome they are particularly common in a subgroup of mesen- chymal tumors of low-grade or borderline malignancy. These rings and markers have been shown to contain amplified material predominantly from 12q13–15, but also sequences from other chromosomes. Such amplified Introduction sequences were mapped in detail by genome-wide array comparative genomic hybridization in ring-containing Several bone and soft tissue tumor entities are characte- tumor samples from soft tissue (n ¼ 15) and bone rized cytogenetically by supernumerary ring chromosomes (n ¼ 6), using tiling resolution microarrays, encompassing or giant rod marker chromosomes in karyotypes with 32 433 bacterial artificial chromosome clones. The DNA few or no other chromosome aberrations (Mitelman copy number profiles revealed multiple amplification et al., 2005). The majority of these tumors are low-grade targets, in many cases highly discontinuous, leading to malignant neoplasms that include well-differentiated delineation of large numbers of very small amplicons. A liposarcoma/atypical lipomatous tumor (ALT), subgroups total number of 356 (median size: 0.64 Mb) amplicons of malignant fibrous histiocytoma (MFH), myxofibro- were seen in the soft tissue tumors and 90 (median size: sarcoma, parosteal osteosarcoma and dermatofibrosar- 1.19 Mb) in the bone tumors. Notably, more than 40% of coma protuberans (DFSP). The origin of the ring all amplicons in both soft tissue and bone tumors were chromosomes, which frequently vary in number and mapped to chromosome 12, and at least one of the size between cells from the same tumor, cannot be previously reported recurrent amplifications in 12q13.3– disclosed by chromosome banding techniques owing 14.1 and 12q15.1, including SAS and CDK4, and MDM2, to the diffuse banding pattern and morphological respectively, were present in 85% of the soft tissue tumors variability. Fluorescence in situ hybridization (FISH) and in all of the bone tumors. Although chromosome 12 analyses have revealed that these structures, in most was the only chromosome displaying recurrent amplifica- cases, contain material from chromosome 12, apart tion in the bone tumors, the soft tissue tumors frequently from in DFSP where chromosome 17 and 22 sequences, showed recurrent amplicons mapping to other chromo- including the PDGFB/COL1A1 fusion gene, are present somes, that is, 1p32, 1q23–24, 3p11–12, 6q24–25 and in rings (Dal Cin et al., 1993; Szymanska et al., 1996; 20q11–12. Of particular interest, amplicons containing Chibon et al., 2002; Sirvent et al., 2003). With few excep- involved in the c-jun NH2-terminal kinase/mitogen- tions, it is the central part of the long-arm of chromo- activated kinase pathway, that is, JUN in 1p32 some 12 (12q) that is present in multiple copies in the ONCOGENOMICS and MAP3K7IP2 (TAB2) in 6q24–25, were found to be ring chromosomes. The amplifications may be discon- independently amplified in eight of 11 cases with 12q tinuous and vary in size, but frequently include chromo- amplification, providing strong support for the notion that some bands 12q14 and 12q15 (Berner et al., 1996). The aberrant expression of this pathway is an important step MDM2 gene is always amplified, whereas several other in the dedifferentiation of liposarcomas. genes, such as SAS, GLI, CDK4 and HMGA2, are Oncogene (2006) 25, 7106–7116. doi:10.1038/sj.onc.1209693; frequently coamplified (Meltzer et al., 1991; Nilbert published online 29 May 2006 et al., 1994; Pedeutour et al., 1994; Gamberi et al., 2000; Gisselsson et al., 2002). The composition of the ring and giant marker Correspondence: Dr M Heidenblad, Department of Clinical Genetics, chromosomes is often quite complex and may include Lund University Hospital, Lund SE-221 85, Sweden. E-mail: [email protected] material from two or more chromosomes. Apart Received 3 January 2006; revised 3 March 2006; accepted 18 April 2006; from the inclusion of chromosome 12 material, direct published online 29 May 2006 evidence through FISH analysis has shown that a Genomic profiling of bone and soft tissue tumors M Heidenblad et al 7107 variety of chromosomes may be involved, in particular amplicons per sample). The number of chromosomes chromosome 1 (Pedeutour et al., 1994; Forus et al., involved in amplification in each case varied from 1 to 1995a; Meza-Zepeda et al., 2001; Nilsson et al., 2004). 14 (case 9); only chromosome 4 was never involved. Similarly, indirect evidence has been obtained from Analysis of primary tumor data from all 19 patients on chromosome-based comparative genomic hybridization average indicated a higher amplification frequency in the (CGH) (Pedeutour et al., 1999; Chibon et al., 2002; soft tissue tumors than in the bone tumors (23 vs 15 Micci et al., 2002; Coindre et al., 2004). The profiles of amplicons per case). Moreover, the analysis showed that genomic imbalances show many similarities between amplicon sizes were significantly smaller among the ALT, dedifferentiated liposarcoma (DDLS) and MFH, former (median size: 0.64 Mb) than among the latter which all share amplification of 12q14–15 in practically (median size: 1.19 Mb; Po0.01, Mann–Whitney U-test). all cases, and in which gain of 1q21–23 is common. Neither the number nor the distribution of amplicons However, also some distinct differences have been distinguished the four soft tissue tumors that meta- reported (Chibon et al., 2002). For example, gain of stasized from those that did not, or the high-grade from sequences from 1p32, 6q23 and 12q24 were absent or the low-grade osteosarcomas. very rare in ALT but fairly common among the other Regarding the genomic localization of the amplicons, soft tissue tumor types. Available data indicate a consi- chromosome 12 was by far most frequently affected derable biological overlap between low-grade MFH (17 of 19 cases), harboring more than 40% of the and lipomatous tumors (Chibon et al., 2002; Coindre amplicons in both tumor groups. In the soft tissue et al., 2003). Based on chromosome CGH data, the 12q tumors, also chromosomes 1 and 6 were commonly amplicon extends from 12q13 to 12q21 in most cases, affected, containing 25 and 6% of the amplicons in with a peak in 12q14 and 12q15, but may occasionally this group, respectively. Apart from chromosome 12, include both more distal and more proximal regions. the only chromosome amplified in more than one bone The 1q amplicon is usually located between 1q21 and tumor was chromosome 5. However, this chromo- 1q25 with a peak incidence in 1q23, whereas the 1p some did not contain any overlapping amplicons. A amplicon seems to be narrow and commonly includes complete description of all amplicons is available only 1p32. in Supplementary Table 1. In the present study, soft tissue and bone tumors To identify genomic regions likely to contain genes of with ring and/or giant marker chromosomes and few or importance for tumor development, a list of recurrent no other chromosome aberrations were investigated by amplifications was compiled (Table 2). In the soft tissue array CGH. For this purpose we used whole-genome tumors, segments amplified in 2–11 of 13 cases were tiling resolution bacterial artificial chromosome (BAC) identified in 1p, 1q, 3p, 6q, 12p, 12q and 20q. Because of microarrays, encompassing 32 433 BAC clones. To the high amplification occurrence and complex distri- determine whether any association between patterns of bution within chromosomes 1 and 12 in this group, DNA copy number alterations and tumor histotype amplification frequency plots were generated for these could be detected, tumors with several different diag- chromosomes (Figure 3a and b). The profile for noses were included in the investigation. chromosome 1 showed a distinct peak in 1p32, indicat- ing a recurrently amplified region of 1.18 Mb that contains four annotated genes, including JUN (Table 2). In 1q, three frequently amplified regions, all located Results within 1q23.2–1q24.3, were detected. The chromosome 12 profile revealed two narrow peaks of amplification The array CGH analysis revealed DNA copy number in 12q13.3–14.1 and 12q15, respectively, affected in alterations in all 13 soft tissue tumors and in all six bone more than 60% of the soft tissue tumors. The proximal tumors (Figure 1a, Table 1). Although several single- 0.42 Mb region contains 17 genes, including SAS and copy gains and losses of larger chromosomal segments CDK4, and the distal 0.75 Mb region comprises 16 were observed, the copy number changes were domi- genes, including MDM2. Other, less frequent, recurrent nated by genomic amplifications. As a lot of these amplifications in the soft tissue tumors were seen in demonstrated highly discontinuous patterns, large num- 3p11.1–12.1, 6q24.3–25.1 and 20q11.2–12.1 (Table 2). bers of very small amplicons were seen in the majority of The only chromosome showing overlapping amplifi- cases. In many instances, the amplification levels were cations in the bone tumors was chromosome 12. The higher than fivefold, including extreme cases in which amplification analysis revealed three small regions in more than 30-fold amplification were observed. Even 12p, all amplified in two of six cases (Table 2). The most though individual amplicons were dispersed throughout distal segment, located in 12p13.3, is 0.62 Mb in size and most of the genome, marked aggregations were observed contains seven genes, including CCND2, FGF6 and in specific chromosomes, mainly 1 and 12 (Figures 1 and FGF23. In 12q, the frequency plot clearly showed two 2). The tumor recurrences, samples 1b and 11b, showed major target regions, both amplified in all six cases DNA copy number profiles that were highly similar to (Figure 3c). The proximal, in 12q13.3–14.1, is 0.91 Mb the corresponding primary tumors (Figure 1, Table 1). in size and covers 17 genes, including SAS and CDK4. Altogether, 446 amplicons were identified among The distal 1.02 Mb region, located in 12q15, contains the 21 samples, 356 in the soft tissue tumors (2–54 eight genes, including MDM2 (Table 2). Notably, amplicons per sample) and 90in the bone tumors (4–28 the 12q13.3–14.1 commonly amplified region was highly

Oncogene Genomic profiling of bone and soft tissue tumors M Heidenblad et al 7108

Figure 1 DNA copy number profiles based on array CGH. (a) Genome-wide overview of copy number alterations in 21 soft tissue and bone tumor samples. Each row represents a separate sample, with case numbers indicated to the left; tumors derived from the same patient are indicated by red bars. Each column shows one of the 32 433 different BAC clones on the microarray, ordered from 1pter to Yqter, and within each chromosome according to the position in the UCSC May 2004 assembly (http://genome.ucsc.edu/). Test over reference fluorescence ratios (moving average, symmetric two nearest neighbors) based on a log2 pseudocolor scale (right) are shown, where red denotes fold amplification, green represents fold deletion, and black indicates normal DNA copy number. (b) Enlarged view of chromosomes 1 and 12. Below each chromosome, sequence positions according to the UCSC Genome browser (May 2004 freeze) are specified.

similar in the soft tissue and bone tumors, whereas (Figure 1a). In case 3, there was loss of one copy of 9p, the equivalent in 12q15 only overlapped partially. together with gain of one copy of 9q. In case 4, a Combining data from all 19 cases, additional recurrently heterozygous partial deletion of 9p, from 9p13.3 to amplified regions were observed in 5p14.1–15.2, 8q13.3, 9pter, was found. Finally, case 17 showed loss of one 14q11.2 and 16q22.2–22.3 (Supplementary Table 1). entire copy of chromosome 9. Genomic losses were few and mostly sporadic. The To validate the array CGH results and to deter- only recurrent aberration encompassed a large chromo- mine the chromosomal organization of a few selected some segment, namely the entire or the major part of 9p amplicons, a whole-chromosome-painting (wcp) Y probe

Oncogene Genomic profiling of bone and soft tissue tumors M Heidenblad et al 7109 Table 1 Clinical data on 19 patients with soft tissue or bone tumors with ring chromosomes Case Age Diagnosis Gradea Location Sizeb Metastasisc Status and Karyotype no. (years) / follow-upd sex

Soft tissue tumors 1ae 42/M SRPF 2 Shoulder 17 No NED 92 47–51,XY,+1–5r,+1–3mar/47–51,idem, add(9)(q34) 1b Recurrence 2 48–51,XY,+2–4r,+1–2mar/ 51,XY,add(9),+4r,+mar 2e 80/M Myxofibro 2 Groin 14 No DoC 128 46–51,X,ÀY,–22,+2–7r/45–50,idem,À18/46,XY 3e 75/F Dediff lipo 3 Thigh 18 5 DoD 24 46–54,XX,der(11)t(11;14)(p13;q11),À14,À17, 1À21,+1–4r, inc/47,XX,+X,add(1)(q44),(5;19) (q10;p10),ins(7;?) (p15;?),+8,+8,i(9)(q10), add(13)(q32),+14,der(14;21) (q10;q10)/ 80–96,idemx2 4 22/M MFH 2 Thigh 8 No NED 83 48–49,XY,+2–3r/96–98,idemx2/46,XY 5e 73/M Dediff lipo 2 Thigh 4 No NED 63 47,XY,+r/47–48,XY,add(1)(q21), t(12;18)(p13;q21), ins(18;?)(q21;?),+1–2r 6 58/M MFH 3 Thigh 20No DoC 12 47,XY,+r/45,XY, À14,+mar/46,XY 7 67/M MFH 2 Groin 11 122 AwD 173 47,XY,+mar/94,idemx2/48,XY,+2r/46,XY 8 55/M MFH 3 Groin 3 No NED 152 39–48,ÀX,–Y,del(6)(q15),À10,add(11)(p15), +1–2r 9 51/F Myxofibro 3 Retroperit ? ? LTF 46–50,XX,+1–5r,+mar/85–98,XXXX, +2–3r,+2–3mar/ 46,XX 1068/M MFH, infl 1 Finger 4 No NED 62 47,XY,+r/48,idem,+r/94,idemx2 11a 72/M Liposarcoma 2 Scrotum 17 33 DoD 42 49,XY,+3r 11b Metastasis — 12 71/M Desmoid 0Thigh 5 No DoC 152 47,XY,+r/46,XY 13 46/F Leiomyosarc, soft 2 Retroperit 8 36 DoD 6047–49,XX,+1–2r,+mar tissue metastasis

Bone tumors 14 30/F Parost OS 1–2 Fibula ? No NED 147 47,XX,+r/48,XX,+r,+mar 15 36/M Parost OS 1–2 Femur 3 No NED 71 48,XY,+2r 16 7/M Fibr OS 3 Femur 8 No NED 155f 47–48,XY,+1–2r 17 34/M Osteobl OS 2–3 Femur 5 26 NED 59g 45–50,X,Y,del(1)(q21),add(2)(q3?5),add(3)(q2?7), À9,À12,À13,À19,À21,À22,+4–5r,+2mar 18 41/M Parost OS/Conv OS 1 and 3 Femur 18 No NED 13f 43–49,XY,+r,+1–2mar/90–96,idemx2 19 25/F Parost OS 1–2 Femur 10No NED 48 47,XX,+r

Abbreviations: AwD, alive with disease; Conv OS, conventional, high-grade osteosarcoma; Dediff lipo, dedifferentiated liposarcoma; DoC, dead of other causes; DoD, dead of disease; F, female; Fibr OS, fibroblastic osteosarcoma; LGFMS, low-grade fibromyxoid sarcoma; LTF, lost to follow- up; M, male; MFH, malignant fibrous histiocytoma; MFH, infl, inflammatory MFH; Myxofibro, myxofibrosarcoma; NED, no evidence of disease; Parost OS, parosteal osteosarcoma; Osteobl OS, osteoblastic osteosarcoma; SRPF, sarcoma resembling proliferative fasciitis. aThree-grade scale. bLargest diameter in centimeter. cTime in months after diagnosis. dFollow-up time in months after primary surgery. eKaryotype previously published in O¨ rndal et al. (1992) or Mertens et al. (1998). fChemotherapy and amputation. gChemotherapy after lung metastases and amputation after local recurrence.

and BAC clones on chromosomes 3, 6 and 12 were used on both chromosomes 6 and 12, according to the array for metaphase FISH analyses of four primary tumors. CGH analysis (Figure 4c). Two of the amplicons on Array CGH analysis of an inflammatory MFH (case 10) these chromosomes, located in 6q24 and 12q15, respec- showed a copy number profile with amplification of 3p tively, were confirmed by FISH analysis (Figure 4c). as the sole aberration. This amplification was confirmed In the fourth tumor, an osteosarcoma (case 18), a by FISH, using three BAC clones within a 6 Mb region unique amplification of the Y chromosome was identi- in 3p11–12. The amplified sequences were localized in fied at array CGH analysis (Figure 1a). This case also giant marker chromosomes in some cells, and in rings showed amplification of several segments in both 12p in others (Figure 4a). The copy number profile of the and 12q. FISH analyses using BAC clones mapped to second tumor, a DDLS (case 3), indicated amplifications 12p11 and 12q15, and a wcp Y probe, corroborated on several chromosomes, including sequences over- the array CGH findings and demonstrated tumor lapping with the 3p amplicon in case 10and various heterogeneity with chromosome 12 sequences amplified segments on 12q (Figure 4b). FISH analysis confirmed a in various combinations of rings and/or giant markers recurrently amplified region in 3p11–12, and demon- in different cells (Figure 4d). The wcp probe confirmed strated high-level amplification of a sequence covering amplification of the Y chromosome in a smaller the MDM2 gene in 12q15 (Figure 4b). The third tumor, proportion of the cells, always localized at the ends of an MFH (case 7), showed several amplification peaks marker chromosomes.

Oncogene Genomic profiling of bone and soft tissue tumors M Heidenblad et al 7110

Figure 2 DNA copy number profiles of case 4. (a) Genome-wide copy number profile showing tumor/reference (T/R) log2 ratios across the genome. The BAC clones, which are displayed as a moving average are ordered from 1pter to Yqter. Individual chromosomes are separated by vertical bars. The profile indicates loss of 9p and amplifications in chromosomes 1, 6, 8, 12 and 16. (b) Enlarged view of chromosomes 1 and 12, showing several amplification subpeaks in both chromosomes. BAC clones within each chromosome are located according to their sequence position (indicated on the X-axes).

Table 2 Selected recurrently amplified regions Cytogenetic region Mb start position (BAC)a Mb end position (BAC)a Size (Mb) No. of cases No. of genesb Examples of genes

Soft tissue tumors 1p32.1–p32.2 58.57 (RP11-342F23) 59.76 (RP11-316F18) 1.18 5 4 JUN 1q23.2 156.83 (RP11-536C5) 157.03 (RP11-196A16) 0.21 4 8 1q23.3 157.82 (RP11-749M10) 158.36 (RP11-5K23) 0.53 4 25 USF1, APOA2 1q24.3 168.69 (RP11-776I22) 169.26 (RP11-806N18) 0.57 6 4 3p11.1–p12.1 85.31 (CTD-2120G8) 90.12 (RP11-371F2) 4.81 2 9 VGL-3, POU1F1 6q24.3–q25.1 147.81 (RP11-64E10) 150.07 (RP11-25G17) 2.26 3 9 MAP3K7IP2 12p12.3 17.19 (RP11-776I15) 17.56 (CTD-2220B12) 0.37 2 0 12q13.3–q14.1 56.27 (RP11-571M6) 56.69 (RP11-620J15) 0.42 9 17 SAS, CDK4 12q15 67.44 (RP11-797C20) 68.20 (RP11-663D20) 0.75 11 6 MDM2, YEATS4 20q11.2 35.31 (RP11-549C4) 36.94 (RP11-663D16) 1.62 2 21 SRC 20q12 38.23 (RP11-583C10) 41.06 (RP11-107D20) 2.83 2 8 TOP1, PTPRT

Bone tumors 12p13.3 3.85 (RP11-358F19) 4.47 (RP11-8E2) 0.62 2 6 CCND2, FGF6, FGF23 12p11.2 26.88 (RP11-35E22) 27.40(RP11-58P11) 0.52 2 6 STK38L 12p11.2 29.51 (RP11-794M10) 29.73 (RP11-478E11) 0.21 2 2 12q13.3–q14.1 56.27 (RP11-571M6) 57.18 (RP11-267H12) 0.91 6 17 SAS, CDK4 12q15 66.52 (RP11-554D4) 67.54 (RP11-450G15) 1.02 6 9 RAPB1, MDM2

Abbreviations: BAC, bacterial artificial chromosome; CDK, cyclin-dependent kinase. aThe start and end positions of each region are given together with the name of the respective size-delineating BAC clone. bThe number refer to Ref.Seq. genes in the UCSC genome browser (Human May 2004 assembly, hg 17).

Discussion microamplifications involving only a few BAC clones (Figures 1 and 2). In the present study, we used genome-wide tiling- Apart from heterozygous loss of material from the resolution BAC microarrays to characterize DNA copy short arm of chromosome 9 in two of 13 soft tissue number alterations in 21 bone and soft tissue tumor tumors and one of six bone tumors, no recurrent samples, containing supernumerary ring and/or giant deletions were identified. In two of these cases, the marker chromosomes. The investigated specimens findings were expected from the G-band karyotype. demonstrated aberrations, ranging from single-copy Case 17 showed monosomy 9 and case 3 harbored an imbalances of entire chromosomes to high-level i(9)(q10) in the more complex of the two hyperdiploid

Oncogene Genomic profiling of bone and soft tissue tumors M Heidenblad et al 7111

Figure 3 Amplification frequency plots of chromosomes 1 and 12. The BAC clones in each chromosome, illustrated as black dots, are located according to their sequence position in the UCSC genome browser May 2004 freeze (indicated on the X-axes). Amplification incidence along (a) chromosome 1 in soft tissue tumors (n ¼ 13), (b) chromosome 12 in soft tissue tumors (n ¼ 13) and (c) chromosome 12 in bone tumors (n ¼ 6). clones (Table 1). The resulting loss of 9p and gain of 9q the majority of gene copy number alterations consisted in the latter tumor was clearly demonstrated at array of genomic amplifications. In total, 446 amplicons CGH analysis. The loss of the major portion of 9p in were identified; 356 in the soft tissue tumors and 90in case 4 was not indicated by the karyotype and might be the bone tumors. The amplicons frequently spanned explained by the presence of a subclone that passed genomic segments less than 1 Mb in size, and occasion- undetected at cytogenetic analysis. Monosomy 9 or partial ally demonstrated more than 30-fold amplification, losses of 9p are common in many tumor types, including corresponding to over 60gene copies. The number of MFH of several subtypes (Mairal et al., 1999; Simons separate amplicons that could be detected in most et al., 2000; Sabah et al., 2005). The tumor suppressors tumors by the present tiling resolution BAC array by far CDKN2A and CDKN2B in 9p21, and possibly other loci, exceeded the number of amplicons identified in previous have been implicated in MFH tumorigenesis. By chromo- studies using chromosome CGH, and allowed a much some-based CGH studies, 9p losses in MFH were found more accurate delineation of the amplicons. What in 10–55% of the cases, included the major portion of may appear as a single amplified unit distinguished distal 9p, rarely only 9p21, and four of nine informative at the chromosome band or sub-band level by conven- cases showed simultaneous loss of 9p21 and over- tional CGH could thus be shown to consist of several representation of MDM2 (Simons et al.(2000)andseparate amplicons. references therein). Whether the large segmental losses Although the amplification patterns in soft tissue and of 9p seen in three of 19 cases in the present study play bone tumors overlapped with regard to the frequent any pathogenetic role remains unknown. involvement of 12q segments, several differences As expected from the present tumors with few were seen. The amplicons were smaller in the soft other changes than supernumerary ring chromosomes, tissue tumors, bone tumors generally contained fewer

Oncogene Genomic profiling of bone and soft tissue tumors M Heidenblad et al 7112

Figure 4 FISH analyses of DNA copy number alterations detected by array CGH. (a) Array CGH and FISH results from the 3p11.1– 12.1 amplicon in case 10(inflammatory MFH). In the DNA copy number profile (left), the position and designation of BAC probes used for FISH are indicated. The FISH images (center and right) reflect tumor heterogeneity, and show multiple signals for all three BACs in a giant marker chromosome and in a ring, respectively. (b) FISH confirmation of 3p11.1–12.1 (left) and 12q15 (center) amplifications in case 3 (dedifferentiated liposarcoma). The results, based on the chromosome 3 probe RP11-81P15, also used in the previous analysis, and the chromosome 12 probe, including the MDM2 gene, show a coamplification of 3p and 12q sequences in ring chromosomes (right). (c) DNA copy number profiles and FISH results of the 6q24.3 and 12q15 amplifications in case 7 (MFH), showing amplified 6q and 12q sequences dispersed along a giant marker chromosome. (d) Gene copy number and FISH results from case 18 (parosteal/conventional osteosarcoma), showing several amplifications along chromosome 12. BAC clones from two of these regions, 12p11.23 and 12q15 confirm amplification, and indicate tumor heterogeneity in the FISH analysis (center and right). FISH analysis of the Y chromosome amplification, indicated by array CGH, showed a normal signal in some cells (center), but clear amplification at the ends of a giant marker chromosome in other cells.

Oncogene Genomic profiling of bone and soft tissue tumors M Heidenblad et al 7113 amplicons, amplification of 12p sequences was more The complex and frequent involvement of chromo- common in bone tumors, and whereas amplicons were some 1 in soft tissue tumors is further illustrated by the commonly seen in 1p and 1q in the soft tissue tumors, finding of a recurrent amplicon in its short arm. Also these chromosome arms were affected in none and one gain of 1p sequences has been described before (Chibon of the bone tumors, respectively. Although the number et al., 2002), but our data could delineate a 1.18 Mb of cases was small, and many different histotypes were region in 1p32, containing only four known genes, of included, these differences suggest that the development which the JUN oncogene is the most likely target. of rings and markers is associated with different events Chibon et al. (2002) have previously shown by conven- or selection in bone and soft tissue tumors. tional CGH analysis that some 20% of MFH tumors Amplification of sequences from chromosome 12 was display high-level amplification of 12q14–15 sequences, observed over the entire tumor panel, and a less marked a feature suggesting a biological relationship with well- clustering of amplicons to chromosome 1 was seen in differentiated and dedifferentiated liposarcomas. Inter- soft tissue tumors (Figure 1). The two commonly estingly, out of 22 such cases, 15 showed concomitant amplified regions on 12q were strikingly similar in soft 1p32 (n ¼ 9) or 6q23 (n ¼ 6) amplification. Further tissue and bone tumors. A 0.42 Mb amplicon in bands studies of the 6q23 amplicon revealed consistent 12q13.3–q14.1, present in nine of the soft tissue tumors, amplification and overexpression of the MAP3K5 gene shared proximal borders (RP11-571M6) with a 0.91 Mb (Chibon et al., 2004), encoding a protein that is part of amplicon seen in all six bone tumors (Table 2). The the c-jun NH2-terminal kinase (JNK)—mitogen-acti- 0.42 Mb region that was in common for bone and soft vated protein kinase (MAPK) signaling pathway. Based tissue tumors contains several oncogenes, including SAS on these results, they speculated that aberrant activation and CDK4, previously shown to be amplified in bone of this pathway could affect tumor cell differentiation, a and soft tissue tumors with or without known ring hypothesis elegantly supported by evidence of restored chromosomes (Meltzer et al., 1991; Nilbert et al., 1994; adipocytic differentiation in an MFH cell line treated Pedeutour et al., 1994; Berner et al., 1996; Gamberi with inhibitors of MAP3K5. In this context, it is of et al., 2000; Meza-Zepeda et al., 2001; Gisselsson et al., particular interest to note that also we found frequent, 2002; Nilsson et al., 2004). The other amplicon on 12q and mutually exclusive, coamplification of 1p32 and 6q that was present in all six bone tumors, covering a sequences, seen in five and three soft tissue tumors, 1.02 Mb region in band 12q15, partially overlapped with respectively, with 12q14–15 amplification, and that the a 0.75 Mb amplicon present in 11 soft tissue tumors; candidate target gene in the 1p32 amplicon – JUN –isa these two amplicons share a 95 kb region containing a downstream effector in the JNK-MAPK pathway. The single candidate target gene – MDM2. Thus, the present 2.26 Mb amplicon we detected in 6q24–25 was, however, study provides strong support for the importance of distal to the one described by Chibon et al. (2004), and MDM2 amplification for the development of certain did thus not include the MAP3K5 locus. On the other subsets of both bone and soft tissue tumors. hand, it encompassed another gene in the JNK-MAPK Three separate recurrent amplicons were detected in pathway, MAP3K7IP2 (a.k.a. TAB2). Expression of the 1q in the soft tissue tumors. Gain or amplification of 1q MAP3K7IP2 protein induces MAPK8 (JNK1), which in sequences have previously been described in a variety of turn binds to, and phosphorylates, JUN. In line with the soft tissue tumors, and the existence of at least two results by Chibon et al. (2002) only one of eight soft separate amplicons in 1q21 and 1q23, respectively, has tissue tumors with concomitant 12q14–15 amplification been suggested (Forus et al., 1995a, b, 1998; Szymanska and 1p32 or 6q24–25 amplification was diagnosed as a et al., 1997; Kresse et al., 2005). In the former amplicon, liposarcoma, whereas the three tumors with the 12q three candidate target genes have been proposed, amplicon, but not the 1p or 6q amplicons, were classified COAS1–3 (Meza-Zepeda et al., 2002). However, their as liposarcoma (two cases) or desmoid tumor. Taken significance for sarcoma development remains unclear, together, the present results thus provide strong support and it could be noted that they were not part of any for the hypothesis presented by Chibon et al. (2004) that recurrent amplicon in the present series of tumors. A aberrant activation of the JNK-MAPK pathway is an 7 Mb region around the APOA2 locus in band 1q23 was important step in the dedifferentiation of liposarcomas recently studied more extensively in a series of 10soft with 12q amplification. tissue sarcomas with previously known rearrangements The finding of an isolated 6 Mb amplification in of 1q21–23 (Kresse et al., 2005). The results suggested chromosome bands 3p11–12 in case 10, an inflammatory the presence of an 0.8 Mb core amplicon containing 11 MFH of the finger, was an odd exception to the known genes, of which ATF6 and DUSP12 were found otherwise ubiquitous 1q and/or 12q amplification to show the highest level of amplification. This 0.8 Mb pattern seen in the remaining soft tissue tumors. Soft amplicon partly overlaps with one of the three tissue sarcomas are quite rare in the distal parts of the minimally amplified regions we detected in 1q23–24, extremities, but it could be noted that one case of an that is, the 0.53 Mb amplicon extending from RP11- entity showing a particular predilection for the hands 749M10to RP11-5K23. However, the overlap is and feet – myxoinflammatory fibroblastic sarcoma – restricted to the segment covered by BAC RP11-5K23, recently was shown to contain amplified material from which contains three known genes: FCGR2A, FCGR3A/ unspecified parts of chromosome 3, but not 12, in 2–4 B and HSPA6. At present, none of these genes has a supernumerary ring chromosomes (Mansoor et al., clear role in tumorigenesis. 2004). Thus, it is tempting to speculate that there may

Oncogene Genomic profiling of bone and soft tissue tumors M Heidenblad et al 7114 exist a distinct subset of inflammatory/myxoinflamma- genes whose overexpression would counteract tumor- tory soft tissue sarcomas with a predilection for the igenesis are positioned just proximal to the amplicon. distal parts of the extremities characterized by amplifi- This hypothesis is supported by the finding that none of cation of genes on chromosome 3 rather than on the 20tumors showed amplification of a 1.5 Mb segment chromosome 12. centromeric to the sharp border, and that this segment Previous studies of gene amplifications have clearly indeed contains potentially growth-suppressing genes, demonstrated that far from all detected gene amplifica- for example, INHBC and NAB2. Another recurrent tions in tumor cells are of pathogenetic significance, as, amplicon, in 1p32, partially includes the very large for instance, many amplified genes are not over- DAB1 gene, which was recently shown to map to the expressed (Platzer et al., 2002; Heidenblad et al., FRA1B common fragile site (Smith et al., 2005). Even 2005). Serendipitous gene amplifications may be parti- though none of the five tumors containing this amplicon cularly common in ring-containing tumors, as the showed identical breakpoints, they all displayed distal mitotically unstable ring chromosomes often transform amplification borders within DAB1, again exemplifying into rod-shaped chromosomes, typically capped in both how high-resolution array CGH may act as a tool, not ends by telomeres and subtelomeric sequences from only to allow precise mapping of very small imbalances other chromosomes. One good example of this phenom- but also to pinpoint genomic segments prone to DNA enon was seen in an osteosarcoma (case 18) with breakage. Future use of tiling resolution BAC micro- amplification of material from chromosome 12, as well arrays and even higher resolution oligonucleotide-based as the Y-chromosome. When analysed by metaphase arrays will undoubtedly provide new insights as to how FISH, however, it could be demonstrated that the and why these chromosomal rearrangements occur. Y-chromosome material was never present in the ring chromosomes, but always located at the ends of the clonally occurring giant marker chromosomes. Thus, Materials and methods there is reason to be cautious when interpreting the significance of amplified regions from distal parts of Tumor and reference samples chromosomes in ring-containing tumors. Based on the presence of ring and/or marker chromosomes at The mechanisms behind ring chromosome formation previous G-banding analysis, six bone and 15 deep-seated soft remain unknown. Typically, bone and soft tissue tissue tumor samples from 19 patients were selected for array tumors with supernumerary ring chromosomes harbor CGH analysis (Table 1). From two patients (cases 1 and 11), a two normal chromosomes 12, in spite of the fact that local recurrence and a distant metastasis obtained four and 2 years, respectively, after the primary tumors, were included. the rings, with few exceptions, contain chromosome 12 The patients consisted of 14 men and five women, 7–80years sequences. We have previously demonstrated that old at the time of diagnosis. As a control for normal gene copy heterodisomy is retained for those parts of chromosome number in the array CGH experiments, a DNA pool derived 12 that are not included in the rings, suggesting that the from multiple healthy male donors (Promega, Madison, WI, ring formation occurs owing to errors in the G2 phase of USA) was used. the cell cycle (Mertens et al., 2004). However, models previously proposed for the formation of gene amplifi- 32k BAC microarrays cation in the form of homogeneously staining regions or The 32k microarrays were constructed from purified degen- double minutes are difficult to apply in the case of ring erate oligonucleotide-primed polymerase chain reaction pro- chromosomes, which, as shown in the present study, ducts (Jo¨ nsson et al., 2005), produced from 32 433 individual are composed of large genomic regions, often from BAC clones, processed and purified at the BACPAC Re- several different chromosomes. A further problem when sources center (Oakland, CA, USA). A complete list of the studying ring chromosomes is that they are mitotically BAC clones, together with genomic position and other detailed information, is available at http://bacpac.chori.org, including unstable, presumably leading to a strong selection direct links to the UCSC browser. For the against sequences that do not add proliferative advan- mapping presented herein, the BACPAC May 2004 data tage. Thus, the ring chromosomes that are present at (hg17) were used. tumor diagnosis may be very different from those initially formed. Nevertheless, the consistent presence DNA isolation, labeling and microarray hybridization of material from 12q strongly suggests that the initiating Genomic DNA was extracted from archived freshly frozen DNA breakage occurs here. In this context it is of tumor biopsies. After brief thawing, soft tissue tumors were interest to note that the frequent 12q13–14 amplicon, disaggregated through scalpel mincing, and bone tumors using containing SAS and CDK4, had a very sharp proximal a dismembrator (Mikrodismembrator II, B.Braun, Melsungen, border, delineated by BAC RP11-571M6 in eight cases Germany). DNA was isolated using the DNeasy Tissue Kit and by BAC RP11-746D11 in five cases. This observa- (Qiagen, Valencia, CA, USA), including the optional RNaseH tion, and the fact that the DDIT3 (a.k.a. CHOP) gene, treatment. which is rearranged in the pathognomonic t(12;16) in Labeling of test and reference DNA was performed essentially as described previously (Jo¨ nsson et al., 2005). In myxoid liposarcomas, is located in RP11-746D11, brief, 1.0 mg of tumor and reference DNA was fluorescently indicate that sequences in this region are particularly labeled with Cy3-dCTP and Cy5-dCTP (Amersham Bios- susceptible to breakage, possibly due to recombination- ciences, Uppsala, Sweden), respectively, using the Array CGH prone sequences in the vicinity of DDIT3 (Kanoe et al., labeling kit (Invitrogen, Carlsbad, CA, USA), and purified 1999). However, the sharp border could also imply that using filter-based spin columns (Cyscribe GFX Purification

Oncogene Genomic profiling of bone and soft tissue tumors M Heidenblad et al 7115 kit, Amersham Biosciences). Differentially labeled DNA was cons were considered ended when separated by at least two pooled, mixed with 100 mg Human Cot-1 DNA (Invitrogen), BAC clones associated with log2 ratios below 0.5. For the and lyophilized before resuspension in 57 ml hybridization amplification frequency plots, a BASE implementation of solution (50% formamide, 10% dextran sulfate, 2 Â standard CGH-Plotter (Autio et al., 2003) was used to define each BAC sodium citrate (SSC), 2% sodium dodecyl sulfate (SDS), 10 mg/ clone as amplified (numerical value 1) or unamplified ml yeast tRNA). Probes were heated at 701C for 15 min and (numerical value 0), using an amplification log2 ratio threshold at 371C for 30min before hybridization to microarrays for of X0.99, and a moving mean sliding window of three clones. 48–72 h at 371C. Before hybridization, microarrays were Using the values given from CGH-Plotter, amplification UV-crosslinked at 500 mJ/cm2 and pretreated using the frequencies for BAC clones along frequently affected chromo- Universal Microarray Hybridization Kit (Corning, Acton, somes were calculated. When indicated, gene copy number MA, USA) according to the manufacturer’s instructions. The profiles are presented as a moving average (symmetrical two slides were washed in 2 Â SSC, 0.1% SDS for 15 min, followed nearest neighbors). Primary data are available upon publica- by 2 Â SSC, 50% formamide (pH 7.0) for 15 min at 451C, 2 Â tion at ArrayExpress (http://www.ebi.ac.uk/arrayexpress). SSC, 0.1% SDS for 30 min at 451C, and 0.2 Â SSC for 15 min at room temperature. Fluorescence was recorded using an FISH analysis Agilent G2565AA microarray scanner (Agilent Technologies, The probes used for FISH analysis consisted of a commercially Palo Alto, CA, USA). available wcp probe specific for the Y chromosome (Vysis, Downers Grove, IL, USA), and BAC clones spanning amplified Image and data analysis regions; RP11-80H24 (3p12.1), RP11-81P15 (3p12.1–3p11.2), Primary data were collected using the GenePix Pro 4.0 RP11-91A15 (3p11.1), RP11-307P5 (6q24.3), RP11-1137N1 software (Axon Instruments Inc., Foster City, CA, USA), (12q15), and RP11-564A17 (12p11.2) (BACPAC Resources and raw result files were deposited into the web-based database center). Isolated BAC DNA was labeled with biotin-, fluor- BioArray Software Environment (BASE) (Saal et al., 2002). escein isothiocyanate-, or Cy3-conjugated dUTP by random Spots were background-corrected using the median fore- hexamer priming (Megaprime DNA labeling system, Amer- ground minus the median background signal intensities for sham, Buckinghamshire, UK). Biotin-labeled probes were both dyes, and log2 ratios were calculated. Unreliable features, detected using DEAC-conjugated avidin. FISH analysis was marked in the feature extraction software, and spots not performed as previously described (Dahle´ n et al., 2003). showing signal-to-noise ratios X3, for both channels, were removed. Data normalization was performed per array subgrid using the Lowess curve fitting (Yang et al., 2002) with a Acknowledgements smoothing factor of 0.33, excluding BAC clones on chromo- somes X and Y during the estimation of the normalization This work was supported by the Swedish Cancer Society, the function. In the data interpretation, amplicons were defined as Swedish Children’s Cancer Foundation, the Crafoord Foun- regions with at least two consecutive BAC clones showing log2 dation, the Gunnar Nilsson’s Cancer Foundation, the Knut ratios X1.0, and amplicon sizes given as the longest distance and Alice Wallenberg foundation via the Swegene program, between the two outermost amplified BAC clones. In uncertain the Ingabritt and Arne Lundberg Foundation, and the Royal regions, showing highly discontinuous amplification, ampli- Physiographic Society in Lund.

References

Autio R, Hautaniemi S, Kauraniemi P, Yli-Harja O, Astola J, Forus A, Weghuis DO, Smeets D, Fodstad O, Myklebost O, Wolf M et al. (2003). Bioinformatics 19: 1714–1715. Geurts van Kessel A. (1995b). Genes Chromosomes Cancer Berner JM, Forus A, Elkahloun A, Meltzer PS, Fodstad O, 14: 15–21. Myklebost O. (1996). Genes Chromosomes Cancer 17: Gamberi G, Ragazzini P, Benassi MS, Ferrari C, Sollazzo 254–259. MR, Molendini L et al. (2000). Clin Orthop Relat Res 377: Chibon F, Mariani O, Derre J, Mairal A, Coindre JM, Guillou 195–204. L et al. (2004). Genes Chromosomes Cancer 40: 32–37. Gisselsson D, Pa˚ lsson E, Ho¨ glund M, Domanski H, Mertens Chibon F, Mariani O, Derre J, Malinge S, Coindre JM, F, Pandis N et al. (2002). Genes Chromosomes Cancer 33: Guillou L et al. (2002). Cancer Genet Cytogenet 139: 24–29. 133–140. Coindre JM, Hostein I, Maire G, Derre J, Guillou L, Leroux Heidenblad M, Lindgren D, Veltman JA, Jonson T, A et al. (2004). J Pathol 203: 822–830. Mahlama¨ ki EH, Gorunova L et al. (2005). Oncogene 24: Coindre JM, Mariani O, Chibon F, Mairal A, De Saint 1794–1801. Aubain Somerhausen N, Favre-Guillevin E et al. (2003). Jo¨ nsson G, Bendahl PO, Sandberg T, Kurbasic A, Mod Pathol 16: 256–262. Staaf J, Sunde L et al. (2005). J Natl Cancer Inst 97: Dahle´ n A, Debiec-Rychter M, Pedeutour F, Domanski 1377–1382. HA, Ho¨ glund M, Bauer HC et al. (2003). Int J Cancer Kanoe H, Nakayama T, Hosaka T, Murakami H, Yamamoto 103: 616–623. H, Nakashima Y et al. (1999). Oncogene 18: 721–729. Dal Cin P, Kools P, Sciot R, De Wever I, Van Damme B, Kresse SH, Berner JM, Meza-Zepeda LA, Gregory SG, Kuo Van de Ven W et al. (1993). Cancer Genet Cytogenet 68: WL, Gray JW et al. (2005). Mol Cancer 4: 39. 85–90. Mairal A, Terrier P, Chibon F, Sastre X, Lecesne A, Aurias A. Forus A, Berner JM, Meza-Zepeda LA, Saeter G, Mischke D, (1999). Cancer Genet Cytogenet 111: 134–138. Fodstad O et al. (1998). Br J Cancer 78: 495–503. Mansoor A, Fidda N, Himoe E, Payne M, Lawce H, Magenis Forus A, Weghuis DO, Smeets D, Fodstad O, Myklebost O, RE. (2004). Cancer Genet Cytogenet 152: 61–65. Geurts van Kessel A. (1995a). Genes Chromosomes Cancer Meltzer PS, Jankowski SA, Dal Cin P, Sandberg AA, Paz IB, 14: 8–14. Coccia MA. (1991). Cell Growth Differ 2: 495–501.

Oncogene Genomic profiling of bone and soft tissue tumors M Heidenblad et al 7116 Mertens F, Fletcher CDM, Dal Cin P, De Wever I, Mandahl Pedeutour F, Suijkerbuijk RF, Forus A, Van Gaal J, Van de N, Mitelman F et al. (1998). Genes Chromosomes Cancer 22: Klundert W, Coindre JM et al. (1994). Genes Chromosomes 16–25. Cancer 10: 85–94. Mertens F, Panagopoulos I, Jonson T, Gisselsson D, Isaksson Platzer P, Upender MB, Wilson K, Willis J, Lutterbaugh J, M, Domanski HA et al. (2004). Cytogenet Genome Res 106: Nosrati A et al. (2002). Cancer Res 62: 1134–1138. 33–38. Saal LH, Troein C, Vallon-Christersson J, Gruvberger S, Borg Meza-Zepeda LA, Berner JM, Henriksen J, South AP, A˚ , Peterson C. (2002). Genome Biol 3: SOFTWARE0003. Pedeutour F, Dahlberg AB et al. (2001). Genes Chromosomes Sabah M, Cummins R, Leader M, Kay E. (2005). Virchows Cancer 31: 264–273. Arch 447: 842–848. Meza-Zepeda LA, Forus A, Lygren B, Dahlberg AB, Godager Simons A, Schepens M, Jeuken J, Sprenger S, van de Zande G, LH, South AP et al. (2002). Oncogene 21: 2261–2269. Bjerkehagen B et al. (2000). Cancer Genet Cytogenet 118: Micci F, Teixeira MR, Bjerkehagen B, Heim S. (2002). 89–98. Cytogenet Genome Res 97: 13–19. Sirvent N, Maire G, Pedeutour F. (2003). Genes Chromosomes Mitelman F, Johansson B, Mertens F. (2005). Mitelman Cancer 37: 1–19. database of chromosome aberrations in cancer Available Smith DI, Zhu Y, McAvoy S, Kuhn R. (2005). Cancer Lett at:http://cgap.nci.nih.gov/Chromosomes/Mitelman. 232: 48–57. Nilbert M, Rydholm A, Wille´ n H, Mitelman F, Mandahl N. Szymanska J, Mandahl N, Mertens F, Tarkkanen M, (1994). Genes Chromosomes Cancer 9: 261–265. Karaharju E, Knuutila S. (1996). Genes Chromosomes Nilsson M, Meza-Zepeda LA, Mertens F, Forus A, Myklebost Cancer 16: 31–34. O, Mandahl N. (2004). Int J Cancer 109: 363–369. Szymanska J, Virolainen M, Tarkkanen M, Wiklund T, O¨ rndal C, Mandahl N, Rydholm A, Wille´ n H, Brosjo¨ O, Heim Asko-Seljavaara S, Tukiainen E et al. (1997). Cancer Genet S et al. (1992). Cancer Genet Cytogenet 60: 170–175. Cytogenet 99: 14–18. Pedeutour F, Forus A, Coindre JM, Berner JM, Nicolo G, Yang YH, Dudoit S, Luu P, Lin DM, Peng V, Ngai J et al. Michiels JF et al. (1999). Genes Chromosomes Cancer 24: 30–41. (2002). Nucleic Acids Res 30: e15.

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

Oncogene