Gene amplification is a common phenomenon in malignant neoplasms of all types. One mechanism behind increased gene copy number is the formation of ring chromosomes. Such structures are mitotically unstable and during tumor progression they accumulate material from many different parts of the genome. Hence, their content varies considerably between and within tumors. Partly due to this extensive variation, the genetic content of many ring-containing tumors remains poorly characterized. Ring chromosomes are particularly prevalent in specific subtypes of sarcoma. Here, we have combined fluorescence in situ hybridization (FISH), global genomic copy number and gene expression data on ring-containing soft tissue sarcomas and show that they harbor two fundamentally different types of ring chromosome: MDM2-positive and MDM2-negative rings. While the former are often found in an otherwise normal chromosome complement, the latter seem to arise in the context of general chromosomal instability. In line with this, sarcomas with MDM2-negative rings commonly show complete loss of either CDKN2A or RB1 —both known to be important for genome integrity. Sarcomas with MDM2-positive rings instead show co-amplification of a variety of potential driver oncogenes. More than 100 different genes were found to be involved, many of which are known to induce cell growth, promote proliferation or inhibit apoptosis. Several of the amplified and overexpressed genes constitute potential drug targets.
Genomic copy number aberrations in malignant tumors vary from low-level gains and losses to high-level amplifications and homozygous deletions. High-level amplifications of restricted chromosomal regions have been shown to contain tumor-promoting oncogenes (1–3). The distribution of amplified regions varies considerably among tumor types, making amplification patterns useful as diagnostic markers (4). The amplified material can be scattered throughout the genome or organized in discrete structures. A subset of such structures usually lack telomeric sequences and present as circular units of variable size, ranging from submicroscopic episomes and double minutes to large ring chromosomes. Episomes are believed to arise by excision of genomic segments from the original chromosome and then gradually enlarge until they become double minutes (5,6). One single cell may contain hundreds of double minutes and their content may integrate into chromosomes as repeated units at a single locus, known as homogeneously staining regions (5). A survey of 56 806 published tumor karyotypes revealed cytogenetic signs of gene amplification— i.e. double minutes, homogeneously staining regions, and/or ring chromosomes—in 4.4% of the cases, with ring chromosomes accounting for two-thirds of them (7).
The mechanism behind ring chromosome formation in neoplasia is unknown. At the time of detection, these structures are much larger than double minutes, usually have centromeres and frequently contain amplified sequences from several different chromosomes (8,9). This material is almost always present also in the normal homologs, wherefore the episome model proposed for double minute formation does not apply to the origin of ring chromosomes (6,10). In the course of tumor progression, the ring chromosomes are frequently broken. They may then reseal or transform into rod-shaped giant marker chromosomes, which typically capture telomeres and subtelomeric sequences from other chromosomes (11,12). Due to this inherent mitotic instability, the size and content of ring and marker chromosomes vary considerably between cells from the same lesion. Also the number of rings per cell frequently varies within a tumor. This provides a basis for selection of the most beneficial genes and a strong selection against sequences associated with a proliferative disadvantage. However, passive bystander genes that neither promote nor inhibit tumor progression may be co-amplified simply due to their chromosomal location adjacent to driver oncogenes.
The frequency of ring chromosomes varies among different tumor classes, ranging from <2% in hematologic neoplasms to 12% in soft tissue tumors (7). Among sarcomas, which are malignant tumors of mesenchymal origin, ring chromosomes are particularly common in borderline and high-grade malignant lesions. The most extreme example is well-differentiated liposarcoma, for which 85% of published karyotypes contain one or more ring chromosomes. Also other common sarcomas like myxofibrosarcoma and undifferentiated pleomorphic sarcoma display ring chromosomes in 15–25% of the cases (7). Although the genetic content of ring chromosomes in well-differentiated liposarcoma is fairly well characterized (13–15), the composition in other sarcomas is less well known. Here, we have investigated various histological subtypes of predominantly high-grade sarcomas with ring chromosomes. We show that they harbor two essentially different types of ring chromosome, associated with distinct patterns of gene amplification and deletion.
Several recurrently amplified and lost chromosome segments are identified
In total, 47 sarcomas with ring chromosomes were analyzed by SNP arrays and a subset of the tumors were also investigated by global gene expression arrays (Supplementary Material, Table S1). Genomic alterations were identified in all cases investigated by SNP array analysis (Supplementary Material, Table S2). Forty-two of these showed amplified regions and were included in subsequent JISTIC analyses (Fig. 1A). This algorithm detected 82 recurrently amplified regions and 58 of these were so-called peak amplified subregions, which covered whole or parts of large aberrant regions (Fig. 1B). Frequently amplified chromosomal segments included chromosome arms 1q, 5p, 6q and 12q. Although less common, recurrently amplified regions were also found in chromosome arms 1p, 3p, 4q, 7p, 7q, 8q, 11q, 12p, 13q, 18p, 18q and 19q. Regions on chromosome arms 8p, 9p, 10p, 13q, 16q, 17p and 17q were recurrently lost and homozygous deletions affecting either of the genes CDKN2A, RB1 or TP53 were each found in four or more tumors (Figs 1B and 2).
Amplified regions map to ring and marker chromosomes
By using probes mapping to amplified regions, fluorescence in situ hybridization (FISH) analyses were successfully performed on metaphase chromosome preparations from 19 tumors and on interphase nuclei from two tumors. The latter showed amplification signals at FISH, but the chromosomal location of the signals could not be discerned. All cases combined, the position of 55 of 58 peak amplicons could be determined by FISH (Fig. 1C; Supplementary Material, Table S3); three amplicons in 1q, 3p and 12q could not be mapped to specific chromosomes. Of the 55 confirmed amplicons, all but 8 were located in ring chromosomes (Supplementary Material, Table S3).
Two fundamentally different types of ring chromosome are recognized by their MDM2 status
Fourteen tumors showed amplification of MDM2 and these tumors generally displayed complex amplification patterns with multiple amplicons primarily affecting chromosomes 1, 6 and 12 (Fig. 2). The amplicons in 1p31–32 (57.8–61.7 Mb) and 6q24-25 (148.9–149.9 Mb), harboring JUN and TAB2 (MAP3K7IP2), were mutually exclusive events. Despite their complex amplification pattern, MDM2-positive tumors generally showed simple karyotypes. Nine of 14 tumors displayed no or only one cytogenetic aberration in addition to supernumerary ring or giant marker chromosomes (Supplementary Material, Table S1). By FISH analyses, the positions of a total of 149 amplicons were investigated in this group of tumors. Of these, 80% (119 amplicons) mapped to ring chromosomes (Supplementary Material, Table S4). One case - a spindle cell sarcoma of the heart – was unique by showing both amplification of MDM2 and homozygous deletion of CDKN2A.
The remaining 33 tumors were MDM2 amplification-negative. These tumors displayed complex karyotypes but the number of amplicons was smaller than in MDM2-positive cases and few amplicons were recurrent (Supplementary Material, Table S1; Fig. 2). Further subclassification of MDM2-negative tumors could be done based on their status of CDKN2A and RB1. A group of 16 tumors showed homozygous deletion of CDKN2A. In the group of CDKN2A depleted tumors, the most recurrent amplicons mapped to chromosome arm 5p. A second group of nine MDM2-negative tumors displayed homozygous loss of RB1. Neither this group nor the remaining group of nine cases displayed any unique, recurrent amplification pattern. The latter nine cases were denoted ‘Other’ and gene expression data were available for three of these tumors. None of them showed an aberrant expression of MDM2, CDKN2A or RB1 (Fig. 3). All the 33 MDM2-negative tumors showed complex karyotypes with multiple aberrations in addition to the ring chromosomes. Mapping of amplicons by FISH could be performed in ten of the cases. These analyses identified 16/41 (39%) amplicons in ring chromosomes, while more than 60% of the amplified signals were seen on marker chromosomes of varying size (Supplementary Material, Table S4).
Taken together, these findings suggest that the formation of MDM2-positive ring chromosomes is an early genetic event, and that the content of these rings may affect the gene expression in a way that is sufficient for tumor development. On the contrary, MDM2-negative ring chromosomes may form as a secondary genetic event due to frequent chromosomal rearrangements. In these tumors, also other genetic aberrations, such as loss of tumor suppressor genes, are necessary for tumor formation.
MDM2-positive tumors display a multitude of amplified and overexpressed genes involved in pathways that promote proliferation or inhibit apoptosis
By comparing gene expression levels in tumors with and without the amplified regions, 123 amplified genes were found to be overexpressed [Supplementary Material, Table S5; Mann–Whitney U test, false discovery rate (FDR) corrected P < 0.05]. All of them were located in chromosomes 1, 6, 11 or 12 (Supplementary Material, Table S5), and were primarily found in tumors with MDM2 amplification (Fig. 2). Using searches in the PubMed database (January 2013), 25 of the overexpressed genes were found to positively regulate proliferation and/or prevent apoptosis: DDR2, FRS2, TAB2, MDM2, SPRR2A, UHMK1, CDK4, YAP1, PRRX1, YEATS4, CNOT2, YBX3, HMGA2, JUN, RECQL, TNFSF4, RAB32, CRYZ, BIRC2, BIRC3, DUSP23, GRIP1, PDE3A, TPR and NUP107 (Fig. 4). Amplification of these statistically confirmed overexpressed genes was only sporadically found in MDM2-negative tumors.
CDK4 and HMGA2 are often, but not always, co-amplified with MDM2
All but three of the cases (cases 7, 9 and 12) with MDM2 amplification showed concurrent amplification of CDK4 (Supplementary Material, Tables S2 and S6). Gene expression data were available for cases 7 and 9, both of which showed high expression of MDM2 and low expression of CDK4 (Supplementary Material, Fig. S1). They also displayed low expression values for CDKN2A; only cases with homozygous deletion of this gene displayed lower values (Fig. 3 and Supplementary Material, Fig. S1). Case 7 showed amplification and high expression of JUN and both cases displayed amplification and high expression of YAP1 and BIRC2 (Supplementary Material, Fig. S1). CDKN2B, CDK6, CCND1 and TAB2 were not differentially expressed in these two cases compared with the rest of the tumors (Supplementary Material, Fig. S1). Nine of the 14 cases with MDM2 amplification showed co-amplification and high expression levels of HMGA2 (cases 2, 4, 5, 8 and 10–14, Supplementary Material, Fig. S1and Table S6). However, high gene expression levels of HMGA2 were also detected in cases 7 and 9, with normal genomic HMGA2 status as well as normal genomic status and low expression of CDK4 (Supplementary Material, Fig. S1and Table S6). Four of the cases with MDM2 amplification (cases 1, 3, 4 and 7) displayed amplification and overexpression of the receptor tyrosine kinase-encoding DDR2 gene in 1q23 (Figs 2 and 3).
Candidate amplified target genes in MDM2-negative sarcomas
Amplicons were less often recurrent in MDM2-negative sarcomas which, in combination with restricted gene expression data, made the identification of target genes more difficult. For example, three MDM2-negative cases displayed amplification of YES1 in 18p11, but gene expression data were not available from any of them (cases 21, 42 and 43, Fig. 2). YES1 encodes a tyrosine kinase that phosphorylates the protein product of YAP1 in 11q22, and this gene was amplified and overexpressed in another four cases (cases 7, 9, 13 and 47) representing both MDM2-positive and -negative tumors (Figs 2 and 4; Supplementary Material, Tables S2, S5 and S7). It could also be noted that amplicons in both 5p and 18q contained cadherin genes (CDH7, CDH9, CDH10, CDH12 and CDH18). However, the genomic status did not correlate with gene expression levels in the few cases from which information on both genomic copy number and RNA levels was available (Supplementary Material, Tables S2 and S7). The amplicon in 4q (58.139–60.869 Mb) did not contain any known genes. However, amplification or single copy gain of two nearby genes in 4q—PDGFRA (54.790–54.859 Mb) and KIT (55.218–55.301 Mb)—was detected in six cases with variable MDM2 status (cases 10, 12, 14, 16, 30 and 45, Supplementary Material, Table S2). A seventh case (case 40) showed amplification of PDGFRA without concomitant amplification of KIT. Gene expression data were available in three of the cases and all of them showed high levels of PDGFRA, while the expression of KIT was not different from control tumors (cases 14, 30 and 45, Supplementary Material, Fig. S2). In cases 16 and 33, an amplicon was found in 3p12. Parts of ROBO1 mapped to this amplicon, and this gene showed a high expression in case 33 from which gene expression data were available. However, this case also showed amplification and high expression level of VGLL3 (Supplementary Material, Fig. S2and Tables S2 and S7), which is also located in 3p12. The amplicon in 19q12 detected in three sarcomas (cases 13, 30 and 38, 32.554–33.080 Mb) does not contain any known protein-coding gene. Case 30 displayed amplification solely of this region (Supplementary Material, Table S2). In the remaining two cases, and in two additional cases, also other parts of 19q12 were amplified (Supplementary Material, Table S2). Gene expression data were available for three of them and all displayed high expression levels of CCNE1 (cases 2, 38 and 45; Supplementary Material, Fig. S2 and Table S7). This gene, located at 34.994–35.007 Mb, could constitute a potential target. No putative target gene could be identified in or near the recurrent amplicons in chromosomes 7, 8 and 13.
Also genes that are not recurrently amplified may still be of importance for the development of ring-containing tumors. Therefore, a few examples of less frequently affected potential target genes are given below. MYOCD is involved in smooth-muscle differentiation and has previously been shown to be amplified in leiomyosarcoma. Here, this gene was highly expressed and part of a large gain or amplification in cases 14, 31 and 32, two of which were diagnosed as leiomyosarcomas (Supplementary Material, Fig. S2and Tables S2 and S7). EGFR, encoding a tyrosine kinase receptor, was amplified and highly expressed in case 26. Another gene that encodes a tyrosine kinase receptor—FLT3—was highly expressed but not amplified in case 25. FLT3 acts upstream of the RAS signaling transduction pathway and two other members of this pathway—KRAS and NRAS—were highly expressed in cases 14 and 26, respectively. In case 14, the high gene expression level was associated with genomic amplification.
Homozygous deletions in MDM2-negative sarcomas affect CDKN2A, RB1, TP53, NF1 and PTEN
Recurrent homozygous deletions were not found in MDM2-positive tumors (Fig. 2). In MDM2-negative tumors, such aberrations were common and affected CDKN2A (cases 14–29), RB1 (cases 30–38) and TP53 (cases 19, 29, 30 and 44). Thus, complete loss of TP53 was concurrently found with homozygous loss of CDKN2A in two cases and with RB1 in one case, as well as in one case without CDKN2A or RB1 loss (Fig. 2). These sarcomas, and two additional cases (cases 35 and 38), showed a low expression of TP53 (Supplementary Material, Fig. S2). NF1 encodes a tumor suppressor that negatively regulates the RAS signaling transduction pathway (Fig. 4). A homozygous deletion and an accompanying low expression level of NF1 were detected in case 29. This gene showed an equally low expression level in combination with a hemizygous loss in case 45. PTEN encodes an inhibitor of the PIK3CA-AKT1 signaling transduction pathway (Fig. 4), and this gene was homozygously deleted and consequently displayed a low expression in case 39.
Disease outcome correlates with tumor location and size but not with karyotypic complexity or MDM2 status
Of the 44 patients with follow-up data, 23 (52%) developed metastasis and/or died from their disease (Supplementary Material, Table S1). Disease outcome was correlated with tumor location; all eight patients with non-extremity-based tumors died from their disease. Also the tumor size was associated with outcome; none of the six patients with tumors ≤5 cm in size developed metastasis and/or died from their disease. Patient age did not correlate with outcome.
Karyotypic complexity did not correlate with clinical outcome. Simple karyotypes were found in nine tumors and seven of these patients developed metastases and/or died from their disease (Supplementary Material, Table S1). However, all cases with simple karyotype displayed amplification of MDM2 and complex genomic rearrangements despite few cytogenetic aberrations. The number of tumors of the present study precluded any comprehensive analysis regarding the impact of MDM2 amplification on the disease outcome. However, there was no obvious correlation between MDM2 status and outcome. Of 14 patients with MDM2-positive tumors, 8 developed metastases and 8 died from their disease. This was not significantly different from patients with MDM2-negative tumors; 13 of 33 patients developed metastases and 12 died from their disease (Fisher's exact test, P > 0.05).
In its simplest form, a ring chromosome is created by fusing the arms of a linear chromosome. Although such simple ring chromosomes occur also in neoplasia, tumor-associated ring chromosomes are often more complex and contain amplified genetic material. In sarcoma, three main types of ring chromosome have been characterized so far: (1) rings that carry fusion genes, e.g., in dermatofibrosarcoma protuberans, (2) rings with 3p12 amplification, seen in myxoinflammatory fibroblastic sarcoma and (3) rings with MDM2 amplification, characteristic for well-differentiated and dedifferentiated liposarcomas, as well as parosteal osteosarcoma (12–18).
Ring chromosomes are commonly found also in other subtypes of sarcoma (Mitelman et al., 2013), but their contents remain poorly characterized. By combining information on karyotype and diagnosis, we here selected all relevant sarcomas and excluded entities characterized by the known types of ring chromosome and fusion genes. The aggressive nature of the selected subset of sarcoma is reflected by the poor outcome of the patients, half of whom developed metastases and/or died of their disease. To identify gene pathways of importance for the development of these tumors, it is important to remember that not all amplified genes contribute to oncogenesis (19). It is thus necessary to separate driver oncogenes from co-amplified bystanders (20). This was attempted by extracting genes that were both recurrently amplified and overexpressed. By this approach, we first identified two fundamentally different types of amplification pattern. One group of tumors showed high-level amplification of MDM2 together with several co-amplified regions from chromosomes 1, 6 and 12. The remaining tumors did not amplify MDM2 and their amplification pattern was much more heterogeneous. These MDM2-negative tumors were instead frequently affected by complete loss of the tumor suppressor genes CDKN2A, RB1 or TP53. The inverse relationship between homozygous deletions targeting CDKN2A or RB1 and MDM2 amplification has, to our knowledge, not been reported for sarcomas or other tumors (15,21,22).
The MDM2-positive and MDM2-negative ring chromosomes are likely to have different roles in tumor development. We postulate that the MDM2-positive ring chromosomes are early, tumorigenic genetic aberrations. This assumption is based on the finding that the vast majority of such sarcomas showed relatively simple karyotypes with no or only one cytogenetic aberration in addition to the supernumerary rings/giant markers. Furthermore, 80% of the amplicons in these sarcomas were located in ring chromosomes. In sharp contrast, ring chromosomes detected in MDM2-negative sarcomas were part of highly complex karyotypes, and a substantial proportion (60%) of the amplicons could not be traced to ring chromosomes. In these sarcomas, the ring chromosome formation is probably not the primary pathogenetic event, but a secondary phenomenon caused by extensive chromosomal instability. This instability could partly be explained by the loss of CDKN2A and RB1, which both have been shown to guard genome integrity (23,24). MDM2-negative tumors displayed recurrent amplicons originating from chromosomes 1 and 5 and known cancer genes such as YES1, PDGFRA, VGLL3, CCNE1, MYOCD, EGFR, NF1 and PTEN were affected in one or more of these tumors. In contrast, MDM2-positive rings were found to amplify and overexpress more than 100 potential driver oncogenes, at least 25 of which are known to promote proliferation and/or inhibit apoptosis. Compared with previous reports on adipocytic sarcomas with MDM2 amplification, these data suggest that the number of genes that are selected for in MDM2-positive ring chromosomes is considerably larger than previously appreciated. This discrepancy may partly be explained by the fact that many driver genes identified here were located in amplicons that originated from the same, rather small chromosomal region. Previously, it has been sometimes assumed that distinct amplicons contain only one driver gene. For several reasons, we believe that this assumption has led to an underestimation of the number of target genes. First, we show that several closely located genes that are implicated in promoting proliferation and cell growth and/or inhibiting apoptosis are both co-amplified and overexpressed. Second, the high resolution of the present genomic analyses allowed for the detection of different copy number levels also for closely located genes. This indicates that what has previously been considered a single amplicon may indeed be closely located amplicons with different driver genes. Finally, it has been recently shown in lymphoma and dedifferentiated liposarcoma that focal genomic amplifications may contain multiple independent drivers (25,26).
MDM2-positive ring chromosomes are hallmarks of well-differentiated and dedifferentiated liposarcomas. The latter sarcoma type arises through transition from an atypical lipomatous tumor/well-differentiated liposarcoma to a non-adipocytic, poorly differentiated sarcoma of variable histological grade (27). When atypical lipomatous tumors develop in the mediastinum, retroperitoneum or intra-abdominally, there is a high risk of local recurrence and dedifferentiation. In fact, up to 80% of the patients eventually die from their disease. Atypical lipomatous tumors at such sites are therefore preferably labeled well-differentiated liposarcoma, to emphasize their malignant potential (27). Irrespective of location, atypical lipomatous tumors invariably harbor MDM2-positive supernumerary ring chromosomes and/or giant marker chromosomes, with few or no additional aberrations detectable by chromosome banding analysis. The amplification pattern described for atypical lipomatous tumors and dedifferentiated liposarcomas is almost identical to the present MDM2-positive tumors (14,28). It could therefore be argued that not only case 13 but also all other 13 sarcomas showing MDM2 amplification in the present study in fact represent dedifferentiated liposarcomas. For instance, mutually exclusive amplifications and high expression levels of JUN and TAB2 were here found in both leiomyosarcoma and undifferentiated pleomorphic sarcoma. One hypothesis is that amplification and activation of JUN, and possibly its upstream kinase MAP3K5, block adipocytic differentiation (29). Both MAP3K5 and TAB2 can activate MAPK8, which in turn phosphorylates and stabilizes JUN (Fig. 4). According to this hypothesis, tumors with amplification of JUN and TAB2 would represent dedifferentiated liposarcomas. However, there is also evidence suggesting that JUN affects cell proliferation rather than adipocytic differentiation and that the high expression of genes in this pathway simply reflects a more aggressive phenotype (30). For several reasons, we believe that it would be too simplistic to view all of the present sarcomas showing MDM2 amplification as dedifferentiated liposarcomas. First, the majority was located in the extremities, and atypical lipomatous tumors at such locations rarely dedifferentiate (27). Second, the same genetic aberration may well be present in morphologically and biologically distinct lesions, and have different phenotypic consequences depending on in which cell type it occurs. For instance, atypical lipomatous tumors display exactly the same cytogenetic and molecular features as parosteal osteosarcoma, although there is no morphologic resemblance between the entities (12,31,32). Third, the amplicons in 12q detected here differ slightly from those previously described in atypical lipomatous tumors and dedifferentiated liposarcomas (15,33–35). While HMGA2 has been reported to be consistently co-amplified with MDM2 in such tumors, we found HMGA2 amplification in only 8 of 13 cases. Hence, HMGA2 amplification seems to be much less common among extremity-based sarcomas with ring chromosomes and MDM2 amplification than among dedifferentiated liposarcomas. Combining all the arguments above, we suggest that sarcomas with MDM2 amplification arising in the extremities do not necessarily represent dedifferentiated liposarcomas, but rather constitute de novo high-grade malignant lesions.
In conclusion, integrative genome analyses reveal two distinct types of ring chromosome in high-grade soft tissue sarcomas. The MDM2-positive rings are early genetic aberrations and contain a large number of potential driver genes. Many of these amplified and overexpressed genes are known to induce cell growth, promote proliferation and inhibit apoptosis, and several of them are potential drug targets, such as genes encoding receptor tyrosine kinases. The majority of the present cases did not, however, display MDM2-positive rings, but rather a novel type of MDM2-negative ring chromosome. These rings seem to arise in the context of a general chromosomal instability. On the molecular level, this is supported by our finding that tumors with MDM2-negative rings display homozygous deletion of tumor suppressor genes CDKN2A or RB1—both known to be important for the maintenance of genome integrity (23,24). Although the present study employed several high-resolution methodologies, additional genetic events associated with the two types of ring chromosomes could probably be detected by deep sequencing approaches. Indeed, Taylor et al. detected several point mutations, fusion transcripts and epigenetic alterations by performing genome, exome and transcriptome analyses, as well as methylation studies, on two dedifferentiated liposarcomas with MDM2 amplification (14). As patients with ring chromosome-carrying sarcomas seem to have a dismal prognosis, such in-depth studies of extended series of tumors are warranted to identify potential novel treatment targets.
MATERIALS AND METHODS
Patients and tumor material
All malignant soft tissue tumors with ring chromosomes analyzed at the Department of Clinical Genetics in Lund between 1985 and 2009 were retrieved. Sarcomas diagnosed as dedifferentiated liposarcoma, well-differentiated liposarcoma and myxoinflammatory fibroblastic sarcoma were subsequently excluded because their ring chromosomes have been extensively studied before. Also sarcomas known to harbor pathogenetically important gene fusions were excluded. All the remaining tumors for which suitable material was available were analyzed by SNP arrays and a subset of the tumors was also investigated by global gene expression arrays (Supplementary Material, Table S1). In total, malignant soft tissue tumors from 47 patients were included in the study and the various subtypes were: undifferentiated pleomorphic sarcoma (17), leiomyosarcoma (13), myxofibrosarcoma (5), malignant peripheral nerve sheath tumor (3), fibroblastic/myofibroblastic sarcoma (2), atypical myofibroblastic tumor (2), low-grade myofibroblastic sarcoma (1), giant cell tumor of soft tissue (1), spindle cell sarcoma of the heart (1) and pleomorphic liposarcoma (1). Also, a dedifferentiated liposarcoma was included as control. All diagnoses were made by experienced pathologists at the sarcoma centers in Lund and Stockholm, Sweden, and followed the criteria outlined in the WHO classification of soft tissue tumors (36). The tumors ranged in size from 1 to 26 cm and all patients were adults (age range: 35–89 years; 22 women, 25 men). Of 44 patients with follow-up data, 23 had died of their disease (n = 20) and/or developed metastases (n = 21). The remaining 21 patients had been followed for 21–153 months (median 54 months). All samples were obtained after informed consent, and the study was approved by the Regional Ethics Committee of Lund University.
Cytogenetic analyses were performed according to the standard procedures and karyotypes were written following the recommendations of the ISCN (37). The complexity of the cytogenetic changes ranged from supernumerary ring and marker chromosomes as the sole anomalies to highly aberrant karyotypes with multiple gains, losses and structural rearrangements in addition to ring chromosomes (Supplementary Material, Table S1).
DNA and RNA extractions
Tumor biopsies were frozen in liquid nitrogen and homogenized using a Mikro-Dismembrator S (Sartorius AG, Goettingen, Germany). To ensure that genomic copy number and gene expression data were from the same cell population, DNA and RNA were extracted from the same homogenized material. Extractions were done using the DNeasy tissue kit including the optional RNase A treatment and the RNeasy lipid tissue kit, according to the manufacturer's instructions (Qiagen, Hilden, Germany). Quality and concentration of the extracted material were measured with a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and NanoDrop ND-1000 (Thermo Fisher Scientific, Waltham, MA).
Genomic copy number analyses
SNP array analysis was used to identify regions with aberrant DNA copy numbers. Briefly, tumor DNA was hybridized onto Illumina Human CNV370-Quad v3.0 BeadChip, containing 370 000 markers, following protocols supplied by the manufacturer (Illumina, San Diego, CA, USA). SNP positions were based on the UCSC hg18/NCBI Build 36 sequence assembly. Data were extracted from the GenomeStudio software (Illumina), and subsequently normalized and segmented using thresholded quantile normalization and B allele frequency segmentation, respectively (38,39). The segmented SNP array data were analyzed using the statistical algorithm JISTIC in order to identify regions that were gained or lost more often than would be expected by chance, with a greater weight given to high amplitude events (40). Amplified regions were defined as regions with a logR ratio of ≥0.3. The algorithm also detects multiple significant subregions with minimal q-value (peak amplified subregions) within large aberrant regions. Constitutional copy number variations were excluded through comparison with the Database of Genomic Variants (http://projects.tcag.ca/variation/) (41).
The chromosomal locations of recurrently amplified regions could be investigated in 21 cases using FISH analyses. Bacterial artificial chromosome clones mapping to amplified regions were labeled and hybridized to metaphase chromosome preparations as previously described (42). The metaphases were subsequently analyzed to estimate the level of amplification and to identify positive or negative hybridization to ring and marker chromosomes, and/or double minutes.
Global gene expression profiling
Global gene expression analyses were performed using Affymetrix Human Gene 1.0 ST arrays according to the manufacturer's instructions (Affymetrix, Santa Clara, CA, USA). RNA of good quality was available for 26 tumors, all of which had shown an aberrant SNP array profile. The material investigated thus represented tumor tissue. Gene expression data were normalized, background-corrected, and summarized using the Robust Multichip Analysis algorithm implemented in the Expression Console version 1.1 software (Affymetrix). For each amplified region, gene expression data were compared between tumors with and without the amplicon. Group comparisons were done using the Mann–Whitney U test, adjusting the P-values for multiple testing by Benjamini–Hochberg FDR correction. A corrected P–value <0.05 was considered significant.
This work was supported by the Swedish Cancer Society, the Swedish Research Council and the Gunnar Nilsson's Cancer Foundation.
We thank Srinivas Veerla for technical assistance and acknowledge the help with the microarray analyses from the Swegene Center for Integrative Biology at Lund University. This work was supported by the Swedish Cancer Society, the Swedish Research Council and the Gunnar Nilsson's Cancer Foundation.
Conflict of Interest statement. None declared.