Genome-Wide Analysis of Gene Expression in Synovial Sarcomas Using a Cdna Microarray1

[CANCER RESEARCH 62, 5859–5866, October 15, 2002] Genome-wide Analysis of Gene Expression in Synovial Sarcomas Using a cDNA Microarray1

Satoshi Nagayama, Toyomasa Katagiri, Tatsuhiko Tsunoda, Taisuke Hosaka, Yasuaki Nakashima, Nobuhito Araki, Katsuyuki Kusuzaki, Tomitaka Nakayama, Tadao Tsuboyama, Takashi Nakamura, Masayuki Imamura, Yusuke Nakamura,2 and Junya Toguchida Laboratory of Molecular Medicine, Human Genome Center, Institute of Medical Science, The University of Tokyo, Tokyo 108-8639, Japan [S. N., T. K., Yu. N.]; Laboratory for Medical Informatics, SNP Research Center, RIKEN (Institute of Physical and Chemical Research), Tokyo 108-8639, Japan [Tat. T.]; Department of Tissue Regeneration, Institute for Frontier Medical Sciences [S. N., T. H., J. T.], and Departments of Orthopaedic Surgery [T. H., To. N., Ta. N.], Pathology [Ya. N], and Surgery and Surgical Basic Science [S. N., M. I.], Graduate School of Medicine, College of Medical Technology [Tad. T.], Kyoto University, Kyoto 606-8507, Japan; Department of Orthopedic Surgery, Osaka Medical Center for Cancer and Cardiovascular Diseases [N. A.], Osaka 537-8511, Japan; Department of Orthopedic Surgery, Kyoto Prefectural University of Medicine [K. K.], Kyoto 602-0841, Japan

ABSTRACT (10). These data support the hypothesis that SS may originate from cells that are widely distributed in a variety of tissues. Among a histologically heterogeneous group of soft tissue sarcomas, Among several histological findings, the most distinctive feature of synovial sarcoma (SS) is regarded as a “miscellaneous” entity of uncertain SS is epithelial differentiation. On the basis of the presence or absence origin. Although recent molecular analysis has disclosed involvement of a of an epithelial component, SS is classified into two major subtypes: specific chromosomal translocation in the pathogenesis of SS, its genetic features remain largely unclear. In the work reported here we examined biphasic, which is composed of distinct epithelial and spindle tumor genome-wide gene expression profiles of 13 SS cases and 34 other spindle- cells; and monophasic, which is composed of fibrosarcoma-like spin- cell sarcoma cases by cDNA microarray consisting of 23,040 genes. A dle tumor cells and no detectable epithelial components (1). However, hierarchical clustering analysis grouped SS and malignant peripheral because the proportion and features of the epithelial component vary nerve sheath tumor into the same category, and these two types of tumor significantly among biphasic tumors, transition from one to the other shared expression patterns of numerous genes relating to neural differ- subtype may be gradual rather than abrupt. entiation. Several genes were up-regulated in almost all SS cases, and the Although the histogenesis of SS remains unclear, molecular anal- presumed functions of known genes among them were related to migra- ysis of the mechanisms underlying tumorigenesis of SS progressed tion or differentiation of neural crest cells, suggesting the possibility of markedly with the discovery that a SYT-SSX fusion gene is a SS- neuroectodermal origin of SS. Moreover, we identified a set of genes that specific genetic alteration (11, 12). The SYT-SSX fusion product divided SS cases into two putative subclasses, a feature that may shed light contains both activator and repressor elements for transcription, and on novel biological aspects of SS in addition to those having to do with the net result seems to be transcriptional repression of certain genes in epithelial differentiation. These data have provided clues for understand- precursors of SS (13, 14). Target genes of the SYT-SSX protein seem ing the origin and tumorigenesis of SS. to be related not only to oncogenesis, but also to epithelial differentiation, because the SYT-SSX2 fusion protein has been found only in tumors with monophasic morphology, whereas biphasic tumors con- INTRODUCTION tain only the SYT-SSX1 subtype (15, 16). Therefore, identifying the STSs3 are difficult diseases to both diagnose and treat. STSs are target genes for the SYT-SSX protein will require determination of generally classified according to their histological resemblance to what the precursor cells are. mature, normal tissues (1). However, some sarcomas have no histo- In this study we analyzed the gene expression profiles of a panel of logical counterparts in normal tissues and therefore are grouped SS cases, using a genome-wide cDNA microarray containing 23,040 together as “miscellaneous soft tissue tumors” in the latest edition of genes. This approach has been useful for clarifying molecular mech- the WHO Soft Tissue Tumor Classification (2). A prototype of such anisms that underlie disease progression, for identifying novel cancer- tumors is SS, which predominantly affects the lower extremities of related genes, and for classifying human cancers at the molecular level adolescents and young adults 15–40 years of age (1). The clinico- (17–22). In addition to SS, we analyzed gene expression profiles pathological designation was originally given because SS occurs among four other types of STS: MFH, LMS, PLS and DLS, and primarily in the vicinity of large joints and histologically resembles MPNST. These tumors sometimes exhibit histological features that developing synovium (3). However, subsequent immunohistochemi- closely resemble those of SS. Although differential diagnosis among cal and ultrastructural studies (4, 5) have revealed significant differ- these other tumors and SS is now feasible through analysis for the ences between the SS tumor cells and synovial cells. In addition, SS SYT-SSX fusion gene (23), we judged that comparison of expression can arise where synovial structures are rare or absent, including the profiles among the various STSs should provide information about the lung (6), heart (7), kidney (8), digestive tract (9), and bone marrow histogenesis of SS. We report here that SS most closely resembles MPNST in terms of gene expression profiles, sharing differential expression of several Received 1/18/02; accepted 8/19/02. The costs of publication of this article were defrayed in part by the payment of page genes characteristic of neural crest-derived cells. This suggests a charges. This article must therefore be hereby marked advertisement in accordance with neuroectodermal origin of SS. Our results also indicate that SS can be 18 U.S.C. Section 1734 solely to indicate this fact. classified into two subgroups irrespective of the histological classifi- 1 This work was supported in part by Research for the Future Program Grant No. 00L01402 from the Japan Society for the Promotion of Science. cation. We also identified several genes expressed commonly in SS, 2 To whom requests for reprints should be addressed, at Laboratory of Molecular whose products should be suitable targets for development of novel Medicine, Human Genome Center, Institute of Medical Science, The University of Tokyo, 4-6-1 Shirokanedai, Minato-ku, Tokyo 108-8639, Japan. Phone: 81-3-5449-5372; Fax: therapeutic drugs. 81-3-5449-5433; E-mail: [email protected]. 3 The abbreviations used are: STS, soft tissue sarcoma; SS, synovial sarcoma; MFH, MATERIALS AND METHODS malignant fibrous histiocytoma; LMS, leiomyosarcoma; PLS, pleomorphic liposarcoma, DLS; dedifferentiated liposarcoma; MPNST, malignant peripheral nerve sheath tumor; MSC, mesenchymal stem cell; RT-PCR, reverse transcription-PCR; EST, expressed Tissue Samples. Primary or recurrent STS tissues were obtained from 47 sequence tag. patients who underwent surgical resection, including 13 with SS and 34 with 5859

spindle cell sarcomas (14 with MFH, 10 with LMS, 3 with PLS, 3 with DLS, list of selected genes, we then chose those showing slight or no expression and 4 with MPNST). Among the four MPNSTs, three had developed in (category D) in Ͼ80% of non-SS cases. patients with a clinical diagnosis of type 1 neurofibromatosis (patients Cluster Analysis of 13 SS Cases According to Gene Expression Profiles. MPNST248, MPNST397, and MPNST558). Tumor samples were snap-frozen To clarify the nature of the histological heterogeneity within the SS group, we in liquid nitrogen immediately after resection and stored at Ϫ80°C until focused on differences in expression patterns of 23,040 genes among the 13 preparation of RNA. Tissue specimens were obtained in the same manner from original SS cases. From the overall expression profiles of the SS group, we 15 additional SS patients to verify the expression patterns. All samples were chose 1405 genes for which data were present in 75% of the cases and that had approved for our analysis by the ethics committee of the Faculty of Medicine, expression ratios that varied by SDs Ͼ1.0. Clustering analysis was performed Kyoto University. Part of each tumor sample was fixed in 10% formalin and in the manner described above. routinely processed for H&E staining to establish a pathological diagnosis by Identification of Candidate Genes for Discriminating between SS Sub- two of us (Y. T. and J. T.). Histological subclassification of SS, either classes. We selected 7067 genes for which data were present in Ͼ10 of 13 SS monophasic or biphasic, was determined by the standard criteria, mainly based cases, and the mean (␮) and SD (␴) were calculated from the relative expres- on the presence of an epithelial component (1). Nine of the 13 cases were thus sion ratios of each gene in one of the two subclasses. A discrimination score ϭ ␮ Ϫ ␮ ␴ ϩ ␴ classified as monophasic, and the remaining 4 were classified as biphasic. At (DS) for each gene was defined as: DS 1 2/( 1 2), where the least 90% of the viable cells in each specimen were identified as tumor cells; subscripts refer to the same groups (26). A large DS indicates that a gene’s contamination with normal elements such as inflammatory cells was consid- expression varies greatly between the two groups but little within its own ered to be minimal. group. We invoked a permutation test to calculate the ability of individual RNA Preparation and T7-based RNA Amplification. Total RNAs were genes to distinguish between the two subclasses; samples were randomly extracted from each frozen specimen and from human MSCs purchased from permutated into each of the two groups 10,000 times. Because the DS dataset BioWhittaker, Inc. (Walkersville, MD) as a universal control, by use of TRIzol of each gene showed a normal distribution, a P for the user-defined grouping Ͻ reagent (Life Technologies, Rockville, MD) according to the manufacturer’s was calculated. If the P was 0.001, the gene was considered to have the instructions. After treatment with DNase I (Nippon Gene, Osaka, Japan), 10 power to distinguish the two groups. ␮ ␮g of total RNA from the tumors and MSC were amplified using an Amplis- Semiquantitative and Real-Time Quantitative RT-PCR. A3- g aliquot cribe T7 Transcription Kit (Epicentre Technologies, Madison, WI), and 5 ␮g of total RNA from each tissue sample was reverse-transcribed for single- of the amplified RNAs were labeled with Cy5-dCTP and Cy3-dCTP, respec- stranded cDNAs, using oligo(dT)12–18 primer and Superscript II (Invitrogen, tively, as described previously (24). Total RNA was also prepared from three Carlsbad, CA). Semiquantitative RT-PCR was carried out with the same SS cell lines (OUSS, HS-SY-II, and YaFuSS) to reinforce the microarray data. gene-specific primers as those prepared for constructing our cDNA microarray ␤ All samples used in this study were analyzed for SYT-SSX fusion transcripts or with a 2-microglobulin-specific primer as an internal control as described by RT-PCR using the following primers: for the SYT-SSX1 gene, 5Ј-CAA- previously (27). The primer sequences are listed in Table 1. PCR reactions were optimized for the number of cycles to ensure product intensity within the CAGCAAGATGCATACCA-3Ј and 5Ј-GGTGCAGTTGTTTCCCATCG-3Ј; linear phase of amplification. Real-time quantitative RT-PCR (TaqMan PCR; for the SYT-SSX2 gene, 5Ј-CAACAGCAAGATGCATACCA-3Ј and 5Ј- Applied Biosystems, Foster City, CA) was performed with the ABI Prism 7700 GGCACAGCTCTTTC CCATCA-3Ј. Sequence Detection system (Applied Biosystems) as described previously cDNA Microarray. We fabricated a “genome-wide” cDNA microarray (27). The primers and TaqMan probes are shown in Table 2. The expression with 23,040 cDNAs selected from the UniGene database (build no. 131) of the levels of each candidate gene were corrected by that of ␤ -microglobulin, and National Center for Biotechnology Information. Construction of the microar- 2 relative expression ratios (r) of each sample to MSC were calculated. ray, procedures for hybridization and washing, and photometric quantification Development of Predictive Formula for Discrimination of SS Sub- of signal intensities of each spot were performed as described previously (24), classes. From the quantitative results obtained by real-time RT-PCR, we except that all hybridization and washing procedures were carried out with an determined the discriminant coefficient (k ) of a predictor gene (j) and constant Automated Slide Processor (Amersham Bioscience, Buckinghamshire, United j value (C) by forward stepwise discriminant analysis. A predictive score (PS ) Kingdom). Each slide contained 52 housekeeping genes, and the Cy5/Cy3 ratio i of each sample (i) was calculated with the following formula: for each gene’s expression was adjusted so that the averaged Cy5/Cy3 ratio of the panel of housekeeping genes was 1.0. We assigned a cutoff value to each (A) PS ϭ ͸k ϫ log ͑r ) ϩ C microarray slide, using variance analysis. If both Cy3 and Cy5 signal inten- i j 2 ij sities were lower than the cutoff values, the expression of the corresponding j gene in that sample was assessed as low or absent. For other genes, we where r is the expression ratio (sample i/MSC) of gene j. Statistical analyses calculated Cy5/Cy3 as a relative expression ratio. ij were performed with statistical package SPSS (SPSS, Chicago, IL). Cluster Analysis of 47 STS Cases According to Gene Expression Pro- files. We applied a hierarchical clustering method to both genes and samples. To obtain reproducible clusters for classification of the 47 STSs, we selected RESULTS 1204 genes for which data were present in 90% of the experiments and that had expression ratios that varied by SDs Ͼ1.0. The analysis was performed using Cluster Analysis of Gene Expression Profiles of 47 STS Cases. web-available software (“Cluster” and “TreeView”) written by M. Eisen.4 We first examined all 47 tumors by RT-PCR for the presence of Before the clustering algorithm was applied, the fluorescence ratio for each SYT-SSX fusion transcripts. In all of the 13 tumors that had been spot was first log-transformed (log2), and then the data for each sample were diagnosed as SS, we identified fusion transcripts of either SYT-SSX1 median-centered to remove experimental biases. (11 cases) or SYT-SSX2 (2 cases), but found no evidence of SYT- Identification of Up-Regulated Genes Common to SS. The relative ex- SSX fusion transcripts in any of the 34 tumors diagnosed as other pression ratio of each gene (Cy5/Cy3 intensity ratio) was classified into one of histological types (data not shown). We then subjected the expression four categories: (A) up-regulated (expression ratio Ͼ2.0); (B) down-regulated profiles of all 47 STS cases to a hierarchical clustering analysis. (expression ratio Ͻ0.5); (C) unchanged (expression ratio between 0.5 and 2.0); and (D) not expressed (or slight expression but under the cutoff level for Reproducible clusters were obtained with 1204 genes (see “Materials detection). We used these categories to detect a set of genes for which changes and Methods”); their expression patterns across the 47 STS cases are in the expression ratios were common among samples as well as specific to a shown in Fig. 1. MFH, LMS, DLS, and PLS were scattered into certain subgroup in accordance with Schena et al. (25). To detect candidate several different clusters and failed to compose a disease-specific genes that were commonly up-regulated in SS, the overall expression patterns cluster. On the other hand, SS cases showed a distinct cluster along of 23,040 genes were first screened to select genes with expression ratios Ͼ3.0 with MPNST. Four tumors with biphasic features (SS190, SS334, that were present in Ͼ75% of the SS cases categorized as A, B, or C. From a SS487, and SS582) were clustered into one group, but nine tumors with monophasic features failed to make one cluster. Two SS cases 4 http://genome-www5.stanford.edu/MicroArray/SMD/restech.html. (SS213 and SS438) constituted one subcluster together with a case of 5860

Table 1 List of up-regulated genes common to the SS group LMMIDa Hs. Gene PCRb Primer Fc Primer R C0671d 31664 Frizzled homolog 10 260 5Ј-TCAGAAACCCTTCAGTGCTACAT-3Ј 5Ј-ATACACACGCAGAAACCACTCTT-3Ј A8647 105805 EST 255 5Ј-CCACTGTCTCATGAAGTGTCAAA-3Ј 5Ј-ACAGAATGGTAAGAAAGGAAGCC-3Ј A0102 1408 Endothelin 3 1004 5Ј-GACACAGATCATAGCTCTACAGGA-3Ј 5Ј-GAGTATTTGAGCAATTGATGGG-3Ј A5094N 53563 Collagen, type IX, ␣ 3 234 5Ј-GTGAGGAAGCAAGTGACAAGG-3Ј 5Ј-CACCCTACCTTCTCTCAAATGC-3Ј A0623 144879 Dual specificity phosphatase 9 308 5Ј-GAGAGCGCAATACCTCACG-3Ј 5Ј-GTGGAGAAACAGGGAGGTGA-3Ј B9059 25960 N-myc 808 5Ј-AAGACAGCAGCAGTTGCTAAAGA-3Ј 5Ј-GAAGAAACAGGCTAGGAAAAAGG-3Ј A6384 346950 Cellular retinoic acid-binding 511 5Ј-GGGGATCAGTTCTACATCAAGAC-3Ј 5Ј-CGTCTAACCAGTTTAATGACTTCG-3Ј protein 1 A0277 26988 Ephrin-B3 1001 5Ј-CATGAGAAGAAGTGTCCCGTTT-3Ј 5Ј-TAAAACTACTGAGGTGACGGCAT-3Ј A5044 31439 Serine protease inhibitor, Kunitz 239 5Ј-CCAACATCACTTCTGTGATGAGA-3Ј 5Ј-GATTTGAGTGATCATTAGGGCTG-3Ј type, 2 C0488 30743 Preferentially expressed antigen in 513 5Ј-CAACCTTAAGCTTCTACGGGATT-3Ј 5Ј-CCTCAAGTCAACATCTGCCTATC-3Ј melanoma A2246 73964 EphA4 1014 5Ј-GAAGGCGTGGTCACTAAATGTAA-3Ј 5Ј-CTTTAATTTCAGAGGGCGAAGAC-3Ј A0650 49585 Fibroblast growth factor 18 912 5Ј-ACTTGCCTGTGTTTACACTTCCT-3Ј 5Ј-GTGTTGGTTTCCTCATTCAAGTC-3Ј C1372 256311 Granin-like neuroendocrine peptide 365 5Ј-CTGTTGAGGTACTTGCTGGGAC-3Ј 5Ј-TCAGATCATGTTTATTGTGGGG-3Ј precursor E1451 198760 Neurofilament, heavy polypeptide 516 5Ј-CCAAAGAAACTCAGAAGAGTCC-3Ј 5Ј-GAAAGTGAACTCCAGTGGAAAG-3Ј (200 kDa) A2691N 2877 Cadherin 3, type 1, P-cadherin 806 5Ј-CTGAAGGCGGCTAACACAGAC-3Ј 5Ј-TACACGATTGTCCTCACCCTTC-3Ј (placental) B9201 284122 WNT inhibitory factor 1 860 5Ј-CACTGCAATAAAAGGTACGAAGC-3Ј 5Ј-TTCAGAAAACTAAAGCAGCACC-3Ј A2029 79404 Neuron-specific protein 1001 5Ј-CTCTGGCATCTTGGTAAGGAG-3Ј 5Ј-CCTCATGTTCTTTATTTGCACAGAG-3Ј C9473 92732 Homo sapiens X28 region near 286 5Ј-GTGAACTGAGGAAGGTGCTTAGA-3Ј 5Ј-CTTTATTCTTGAGATGCAGGGG-3Ј ALD locus A8857 11849 Hypothetical protein MGC15827 503 5Ј-CCCAGATGACCACATTTAATACC-3Ј 5Ј-AGAGAAGGGAATCACAACACAGA-3Ј C5852 55407 Homo sapiens cDNA 814 5Ј-CAAGGCTAGAAAGATGCTACGTT-3Ј 5Ј-CAGACACGCACTTGTGGTTTATT-3Ј DKFZp434K0621 A5313 BC009491 Clone MGC 16382 527 5Ј-AGAAGATGCCAATGTTTCATCC-3Ј 5Ј-GACTGTGTTGAGTAAGAGCCACA-3Ј D6309 129010 EST 351 5Ј-ATGCTGTCTCCAGACCCACT-3Ј 5Ј-AGTGACCCTGGCTCTGAAAG-3Ј D6252 128899 EST 227 5Ј-GGCTTATTCTTCAGGCACTAAGG-3Ј 5Ј-AGCAGTTGGAAATGTACTTGCAC-3Ј C9468 92679 Homo sapiens clone CDABP0014 206 5Ј-CTCCTTTCCAGACAGATGAGAGA-3Ј 5Ј-ATGCCTGTTTTTCCTACACTCAG-3Ј B8437 24583 Hypothetical protein 268 5Ј-TTACTGTTTTGTCTCTTGAGCCC-3Ј 5Ј-GTTACCCCTAGGTATGCTTCGTT-3Ј DKFZp434C0328 B7503 12714 EST 518 5Ј-AAAAGGATAGTTCCAGGCCATAG-3Ј 5Ј-GCCAGTAGACCCAAACAATAAGA-3Ј a LMMID, identification number of a gene used in our laboratory. Hs, UniGene accession number. b PCR product size (bp). c Primer F or R, forward or reverse primer sequence, respectively. d Italics indicate LMMID highly specific to SS.

MPNST (MPNST248), and one case of MPNST (MPNST558) fell confirmed the specific expression of these genes in SS or SS and into the major cluster of SS. These data suggested that SS and MPNST MPNST (Fig. 2). In addition, expression of all 26 genes was detected are closely related diseases in terms of gene expression, although both in three SS cell lines, indicating that this activity was intrinsic to SS are regarded as clearly distinct entities from a histological point of cells and not induced by the in vivo environment. In most of the view. known genes that were up-regulated commonly in SS and MPNST, Identification of Up-Regulated Genes Common to SS Cases. the proposed function and distribution of expression were related to We identified 26 genes, including four ESTs, that were commonly neural tissues, e.g., EphA4, ephrin-B3, and endothelin 3. Moreover, up-regulated in SS (Table 1). Among them, frizzled homologue 10 SS expressed additional markers of neural differentiation, such as (C0671) and one EST (A8647) were up-regulated specifically in SS; neurofilament, neuron-specific protein, and fibroblast growth factor the remaining 24 genes were also expressed in MPNST, at the same 18, which were expressed in MPNST at the same or slightly lower or lower levels. The results of semiquantitative RT-PCR experiments levels. These data suggested that the cellular precursor of SS was very

Table 2 List of primer sets and TaqMan probes LMMIDa Primer Fb Primer R TaqMan Probec A2673 5Ј-CGATGAGCTGGGAGTGAAGC-3Ј 5Ј-CATCGCTCTTGGATTCCCAC-3Ј 5Ј-FAM-CAGGCAAAAGTGAGCGCAGCTCCT-TAMRA-3Ј B2602 5Ј-TTTCGTTCTGTTTTCTCATGACAGA-3Ј 5Ј-GAGAGAGTCAGACTAATAAA 5Ј-FAM-CCCTTTCCCCACCCCTAAGTGCCTAA-TAMRA-3Ј CAGGCTGTT-3Ј A2041 5Ј-GGATTGCAGCTTCTGGGAAC-3Ј 5Ј-CAAGCAGTTTGGAGGCAGC-3Ј 5Ј-FAM-ATCTATGAGCTTCGAAATAAGGAACGCATCTCTG- TAMRA-3Ј D5183 5Ј-TTCGAGAAGCGCCACAAGA-3Ј 5Ј-GCTGAGAGGCCGGCACT-3Ј 5Ј-FAM-CACCTGTCCCCCTGCTTCAGGGA-TAMRA-3Ј C9540 5Ј-AGCCCTCGCGGCAAG-3Ј 5Ј-GCTGGCTCAACATGGAAGGA-3Ј 5Ј-FAM-CCCTCACTCTCTCGCCTGTTCTGTGTC-TAMRA-3Ј B9386 5Ј-TTGAAATGCTTTGATATTCT 5Ј-ATTCTTACGAACTTTAAAAAAATA 5Ј-FAM-AACAAGTTTTTTCCCTGCTCCCCAAATAGAAT- AATTGACA-3Ј GCAAAGT-3Ј TAMRA-3Ј A4266 5Ј-TGACGGACTTCGTGTGCAAA-3Ј 5Ј-ATTCACGCCGAAGAAGTTGG-3Ј 5Ј-FAM-CAGCAGAGTGAGCTGTTGACTCGATCG-TAMRA-3Ј A5183 5Ј-ACGCAGACAGAAGGTGGAGC-3Ј 5Ј-GAGACATGCAGCCGTTTCG-3Ј 5Ј-FAM-AGACGGCCAGCAGTCACAGACACAAAGT-TAMRA-3Ј A1181 5Ј-GACAAATGTTCGTCCTGTTA 5Ј-TTTTGCCACTGTGTATATCATCCA-3Ј 5Ј-FAM-TGTGGAGGTAGTTGGGTAGAAAAATTATTAGAACAT- ATTTATAGG-3Ј TCACTAMRA-3Ј A1848 5Ј-GAGGAGATCCGGCGCATAA-3Ј 5Ј-AGCACTCGCTGGAACATGAA-3Ј 5Ј-FAM-TACCAGCAGCAATATGGACGGAGCCTT-TAMRA-3Ј A8306 5Ј-CTCTTTGCACATGGGCATCA-3Ј 5Ј-TCTTTGTAAGGCTGTACCTTGCAT-3Ј 5Ј-FAM-TCTGTAGAATTTGACGGAACACAGCTATTTCCC- TAMRA-3Ј A1887N 5Ј-CTGGCCTGCCTTCGTTAACT-3Ј 5Ј-GGGCAGGATACCCAAACAAA-3Ј 5Ј-FAM-TGTCAATAAACAGCTTCATGCCTTTGTAAGTTATTTC- TTGTAMRA-3Ј a LMMID, identification number of a gene used in our laboratory. b Primer F or R, forward or reverse primer sequence, respectively. c FAM, 6-carboxyfluorescein; TAMRA, 6-carboxytetramethylrhodamine. 5861

similar to that of MPNST, possibly a cell derived from the neural crest. Subclassification of SS. On the basis of the expression patterns of the 1405 genes we selected (see “Materials and Methods”), the SS group was subdivided by a clustering analysis into two distinct subclasses (A and B; Fig. 3). As shown in the hierarchical clustering analysis for all tumors (Fig. 1), four biphasic tumors (SS190, SS334, SS487, and SS582) were again clustered closely, whereas monophasic tumors were divided into two groups. Therefore, the degree of epithelial differentiation may contribute to this subclassification. How- ever, because a tumor (SS646) that showed minimal epithelial differentiation and carried the SYT-SSX2 fusion gene (data not shown) was classified closely to one of the clusters of biphasic type, it is likely that factors other than epithelial differentiation also contribute to this subclassification. A permutation test identified a set of genes that may distinguish the two subclasses (Table 3). We applied forward stepwise discriminant analysis using 12 candidate genes and established a predictive formula with discriminant coefficients of 9 predictive genes that were further selected from the 12 genes and a constant value of 6.464 for subclassification of these learning cases (see “Materials and Methods” and Table 3). To further verify the power of a set of the nine genes to assign a tumor to one subgroup or the other, their expression levels were also determined by real-time RT-PCR analysis in 15 additional SS cases in which SYT-SSX fusion transcripts were identified, SYT- SSX1 gene in 12 cases and SYT-SSX2 in 3 cases. Histologically, 10 of the 15 additional cases were classified into the monophasic subtype, and 5 into the biphasic subtype. By the predictive formula, the 15 test cases also clearly fell into either subclass A (4 cases) or subclass B (11 cases; Fig. 4). All of the four cases in subclass A were monophasic, whereas all five biphasic cases were assigned to subclass B, confirm- ing that the subclassification reflected the degree of epithelial differentiation. However, a monophasic case with the SYT-SSX2 gene in the test cases, which may correspond to SS646 in the learning cases, fell into subclass B, again suggesting that factors other than epithelial differentiation also contribute to this subclassification. These results indicated a potential for this set of genes to classify SS cases in terms of their biological properties.

DISCUSSION Genome-wide analysis of gene expression patterns by use of a cDNA microarray revealed that SS and MPNST were closely related diseases in terms of the genes whose expression was changed in tumors compared with normal mesenchymal cells. As reported previously (28–30), MPNST is sometimes indistinguishable from SS, especially the monophasic subtype, by histological features alone. Moreover, a rare variant of MPNST exhibits extensive glandular differentiation (glandular MPNST) and closely resembles biphasic SS

Fig. 1. A, overall expression patterns of 1204 genes in 47 STS cases. Horizontal rows represent individual genes; vertical columns represent individual samples. Each cell in the matrix represents the expression level of a single transcript in a single sample, with red and green indicating transcript levels above and below the median for that gene across all samples, respectively. Black represents unchanged expression; gray indicates no or slight expression (intensities of both Cy3 and Cy5 under the cutoff value.) Color saturation is proportional to magnitude of the difference from the mean. The dendrogram at the top of the matrix indicates the degree of similarity between tumor samples. The dendrogram on the left side indicates the degree of similarity among the selected genes according to their expression patterns. B, enlarged view of the dendrogram, showing our biological classification of 47 STS cases; the shorter the branches, the greater the similarity. It is noteworthy that SS cases were separated from the others by virtue of distinct expression profiles and fell into the same category with MPNST cases. With respect to liposarcoma, cases 341D, 403D, and 602D were diagnosed as DLS and cases 391P, 406P, and 580P had pathological features of PLS. 5862

Figure 1. Continued.

(31), and both MPNST and SS contain a variant with extensive Our cluster analyses demonstrated that SS also had a feature of epithelial component (epithelioid MPNST and monophasic epithelial neuronal differentiation, irrespective of the presence of an epithelial SS, respectively; Refs. 32, 33). These morphological similarities component. Among the genes commonly up-regulated in SS, several might well reflect similarity in gene expression profiles. MPNST cells are known to be involved in migration or differentiation of neural are considered to originate in either Schwann cells or perineural cells, crest cells. For example, interactions between ephrin-B3 and EphA4 both of which arise from neural crest and migrate to peripheral regions are important for the proper migration of neural crest cells (34), and along with the peripheral nerve. endothelin 3 is essential for development of neural crest-derived cell

Fig. 2. Semiquantitative RT-PCR analyses of 10 representative up-regulated genes common to SS in 3 MFH (Lanes 1–3), 3 LMS (Lanes 4–6),1LS(Lane 7), 1 MSC (Lane 8), 13 SS (Lanes 9–21), and 4 MPNST (Lanes 22–25) cases and 3 SS cell lines (Lanes 26–28). Many of the genes were also expressed in MPNST at the same or lower levels compared with SS. FDZ10, frizzled homologue 10; COL9, collagen type 9; PRAME, preferentially expressed antigen in melanoma; CRABP1, cellular retinoic acid-binding protein-1; EDN3, ␤ ␤ endothelin 3; 2MG, 2-microglobulin (internal control). 5863

Fig. 3. A, overall expression patterns of 1405 genes in 13 SS cases. B, putative subclasses obtained by cluster analysis (see details in legend for Fig. 1). The SS group fell into two biolog- ically distinct subclasses (A and B); tumors with typical monophasic features (SS53 and 259) and with typical biphasic features (SS487 and 582) were each clustered in close relation- ship and separated from other categories in all cases.

5864

Table 3 List of candidate genes discriminating the SS subgroups Discriminant LMMIDa Hs. Gene Permutational P coefficient A2673 2730 Heterogeneous nuclear ribonucleoprotein L 5.43 ϫ 10Ϫ12 B2602 3542 Hypothetical protein FLJ11273 4.63 ϫ 10Ϫ11 Ϫ6.035 A2041 78596 Proteasome (prosome, macropain) subunit, ␤ type, 5 3.67 ϫ 10Ϫ8 Ϫ7.187 D5183 182740 Ribosomal protein S11 8.78 ϫ 10Ϫ8 Ϫ1.154 C9540 91142 KH-type splicing regulatory protein (FUSE binding protein 2) 1.99 ϫ 10Ϫ7 Ϫ2.591 B9386 27179 Homo sapiens cDNA FLJ12933 fis, clone NT2RP2004962 4.90 ϫ 10Ϫ7 A4266 82071 Cbp/p300-interacting transactivator, with Glu/Asp-rich COOH-terminal 5.24 ϫ 10Ϫ7 1.777 domain, 2 A5183 5011 RNA binding motif protein 9 5.37 ϫ 10Ϫ7 9.412 A1181 82173 TGFB inducible early growth response 3.39 ϫ 10Ϫ6 A1848 77840 Annexin A4 3.96 ϫ 10Ϫ6 Ϫ1.933 A8306 172572 Hypothetical protein FLJ20093 6.71 ϫ 10Ϫ6 7.464 A1887N 78465 v-jun avian sarcoma virus 17 oncogene homologue 1.24 ϫ 10Ϫ5 0.184 a LMMID, identification number of a gene used in our laboratory; Hs, UniGene accession number; TGF, transforming growth factor. lineages (35). In addition to being the major extracellular matrix of cells is likely to be induced by expression of a particular set of genes, cartilage, collagen IX also has a crucial function in migration of neural although no clear mechanism has been reported. One of the confound- crest cells (36); retinoic acid signaling is also indispensable for this ing facts is that there are variations in the degree of epithelial differ- migration, and cellular retinoic acid-binding protein 1 is thought to be entiation in different tumor samples, a feature that renders clear-cut a general marker of neural crest cells (37). Transduction of Wnt classification into one or the other of the two subclasses difficult. We signals is involved in the genesis of neural crest cells (38), and therefore attempted to classify SS cases according to their gene fibroblast growth factor 18 is expressed in such cells adjacent to expression patterns, irrespective of histological classification. As a primitive streaks (39). All of these findings suggest that SS cells are result, although all of the biphasic cases were clearly clustered to- derived from the neural crest. However, because the expression of gether, monophasic cases were divided into two subclasses. These some of these genes is not limited in neural tissues, our data are not data suggested the presence of other factors reflecting the differenti- conclusive for the origin of SS cells, and further investigation should ation stage of SS precursor cells or the concomitance of additional be required. genetic alterations during tumorigenesis. We believe that putative A previous report demonstrated that both spindle and epithelial subclassification, which could be predicted by a set of genes we components of SS contained the same type of fusion transcripts (40, selected, should be useful for understanding the histological hetero- 41), leading to the hypothesis that SS is monoclonal in origin. The geneity of SS. transition from uncommitted spindle cells to differentiated epithelial In contrast to the cluster formation of SS and MPNST, other histological types were not well clustered. Although MFH is the most common STS of elder adult life, there has been a long-standing debate whether MFH exists as an independent entity or represents an admix- ture of various mesenchymal and even nonmesenchymal tumors (1). Our failure to create a distinct cluster of MFH indicated that MFHs are not only morphologically, but also genetically heterogeneous tumors, supporting the idea that MFHs under the current histological definition may include tumors of various types (42). Liposarcomas in this study consisted of two rare subtypes, PLS and DLS. The definition of DLS is a neoplasm with well-differentiated liposarcoma juxtaposed to high-grade pleomorphic sarcoma, which usually resembles MFH (43). Distinction of PLS and MFH has been a difficult problem for pathol- ogists, especially for MFHs without a storiform pattern (44). Our cluster analyses demonstrated that DLS and PLS shared gene expression profiles with some cases of MFH, which may explain the his- topathological similarities. It is notable that three LMS cases (LMS181, LMS407, and LMS551), constituting a distinct group within a large cluster of MFHs, were diagnosed to be poorly differentiated types and clinically revealed very aggressive phenotypes with distant metastases (data not shown). Further analyses of the relation- ship between the expression profiles and clinicopathological features of tumors used in this study, including SS and MPNST, may provide insight for exploration of disease-specific genes and, hopefully, of potential genes for molecular target therapy.

ACKNOWLEDGMENTS

We thank Kie Naito, Hiroko Bando, Noriko Sudo, and Yukiko Tsuno for fabricating the cDNA microarray; Emi Ichihashi for analysis of data; Drs. Fig. 4. Scatter plot showing the predictive scores of SS subclasses (A and B)in13 Nobuyuki Hashimoto, Hideki Yoshikawa, and Hiroshi Sonobe for kindly learning cases (E) and 15 test cases (ࡗ). With the predictive formula determined by forward discriminant analysis in learning cases, 15 test cases were grouped into one of the providing cell lines of synovial sarcoma; Drs. Takeharu Nakamata, Tomoki putative subclasses. Horizontal bars indicate the median value for each group. Aoyama, Takashi Okamoto, Koichi Nishijo, and Norifumi Naka for prepara- 5865

Downloaded from cancerres.aacrjournals.org on September 27, 2021. © 2002 American Association for Cancer Research. GENOME-WIDE ANALYSIS OF GENE EXPRESSION IN SS tion of tumor and cDNA samples; and Dr. Suguru Hasegawa for helpful a downstream gene of the ␤-catenin/T-cell factor pathway and its involvement in discussions. colorectal carcinogenesis. Cancer Res., 61: 6345–6349, 2001. 22. Okabe, H., Satoh, S., Kato, T., Kitahara, O., Yanagawa, R., Yamaoka, Y., Tsunoda, T., Furukawa, Y., and Nakamura, Y. Genome-wide analysis of gene expression in REFERENCES human hepatocellular carcinomas using cDNA microarray: identification of genes involved in viral carcinogenesis and tumor progression. Cancer Res., 61: 2129–2137, 1. Weiss, S. W., and Goldblum, J. R. Enzinger and Weiss’s Soft Tissue Tumors, Ed. 4. 2001. St. Louis: Mosby, 2001. 23. van de Rijn, M., Barr, F. G., Collins, M. H., Xiong, Q. B., and Fisher, C. Absence of 2. Weiss, S. W., and Sobin, L. Histological typing of soft tissue tumours. In: World SYT-SSX fusion products in soft tissue tumors other than synovial sarcoma. Am. J. Health Organization International Histological Classification of Tumours, Ed. 2. Clin. Pathol., 112: 43–49, 1999. Berlin: Springer-Verlag, 1994. 24. Ono, K., Tanaka, T., Tsunoda, T., Kitahara, O., Kihara, C., Okamoto, A., Ochiai, K., 3. Smith, L. W. Synoviomata. Am. J. Pathol., 3: 355, 1927. Takagi, T., and Nakamura, Y. Identification by cDNA microarray of genes involved 4. Ghadially, F. N. Is synovial sarcoma a carcinosarcoma of connective tissue?. Ultra- in ovarian carcinogenesis. Cancer Res., 60: 5007–5011, 2000. struct. Pathol., 11: 147–151, 1987. 25. Schena, M. A., Shalon, D., Heller, R., Chai A., Brown, P. O., and Davis, R. W. 5. Smith, M. E., Fisher, C., Wilkinson, L. S., and Edwards, J. C. Synovial sarcoma lacks Parallel human genome analysis: Microarray-based expression monitoring of 1000 synovial differentiation. Histopathology, 26: 279–281, 1995. genes. Proc. Natl. Acad. Sci. USA, 93: 10614–10619, 1996. 6. Roberts, C. A., Seemayer, T. A., Neff, J. R., Alonso, A., Nelson, M., and Bridge, J. A. 26. Golub, T. R., Slonim, D. K., Tamayo, P., Huard, C., Gaasenbeek, M., Mesirov, J. P., Translocation (X;18) in primary synovial sarcoma of the lung. Cancer Genet. Cyto- Coller, H., Loh, M. L., Downing, J. R., Caligiuri, M. A., Bloomfield, C. D., and genet., 88: 49–52, 1996. Lander, E. S. Molecular classification of cancer: class discovery and class prediction 7. Iyengar, V., Lineberger, A. S., Kerman, S., and Burton, N. A. Synovial sarcoma of the by gene expression monitoring. Science (Wash. DC), 286: 531–537, 1999. heart. Correlation with cytogenetic findings. Arch. Pathol. Lab. Med., 119: 1080– 27. Yamanaka, Y., Tamari, M., Nakahata, T., and Nakamura, Y. Gene expression profiles 1082, 1995. of human small airway epithelial cells treated with low doses of 14- and 16-membered 8. Argani, P., Faria, P. A., Epstein, J. I., Reuter, V. E., Perlman, E. J., Beckwith, J. B., macrolides. Biochem. Biophys. Res. Commun., 287: 198–203, 2001. and Ladanyi, M. Primary renal synovial sarcoma: molecular and morphologic delin- 28. Folpe, A. L., Schmidt, R. A., Chapman, D., and Gown, A. M. Poorly differentiated eation of an entity previously included among embryonal sarcomas of the kidney. synovial sarcoma: immunohistochemical distinction from primitive neuroectodermal Am. J. Surg. Pathol., 24: 1087–1096, 2000. tumors and high-grade malignant peripheral nerve sheath tumors. Am. J. Surg. 9. Billings, S. D., Meisner, L. F., Cummings, O. W., and Tejada, E. Synovial sarcoma Pathol., 22: 673–682, 1998. of the upper digestive tract: a report of two cases with demonstration of the X;18 29. Smith, T. A., Machen, S. K., Fisher, C., and Goldblum, J. R. Usefulness of cytokeratin translocation by fluorescence in situ hybridization. Mod. Pathol., 13: 68–76, 2000. subsets for distinguishing monophasic synovial sarcoma from malignant peripheral 10. Hiraga, H., Nojima, T., Isu, K., Yamashiro, K., Yamawaki, S., and Nagashima, K. nerve sheath tumor. Am. J. Clin. Pathol., 112: 641–648, 1999. Histological and molecular evidence of synovial sarcoma of bone. A case report, 30. O’Sullivan, M. J., Kyriakos, M., Zhu, X., Wick, M. R., Swanson, P. E., Dehner, L. P., J. Bone Joint Surg. Am., 81: 558–863, 1999. Humphrey, P. A., and Pfeifer, J. D. Malignant peripheral nerve sheath tumors with 11. Clark, J., Rocques, P. J., Crew, A. J., Gill, S., Shipley, J., Chan, A. M., Gusterson, t(X;18). A pathologic and molecular genetic study. Mod. Pathol., 13: 1253–1263, B. A., and Cooper, C. S. Identification of novel genes, SYT and SSX, involved in the 2000. t(X;18)(p11.2;q11.2) translocation found in human synovial sarcoma. Nat. Genet., 7: 31. Christensen, W. N., Strong, E. W., Bains, M. S., and Woodruff, J. M. Neuroendocrine 502–508, 1994. differentiation in the glandular peripheral nerve sheath tumor. Pathologic distinction 12. Skytting, B., Nilsson, G., Brodin, B., Xie, Y., Lundeberg, J., Uhlen, M., and Larsson, from the biphasic synovial sarcoma with glands. Am. J. Surg. Pathol., 12: 417–426, O. A novel fusion gene. SYT-SSX4, in synovial sarcoma. J. Natl. Cancer Inst. 1988. (Bethesda), 91: 974–975, 1999. 32. Laskin, W. B., Weiss, S. W., and Bratthauer, G. L. Epithelioid variant of malignant 13. Nagai, M., Tanaka, S., Tsuda, M., Endo, S., Kato, H., Sonobe, H., Minami, A., peripheral nerve sheath tumor (malignant epithelioid schwannoma). Am. J. Surg. Hiraga, H., Nishihara, H., Sawa, H., and Nagashima, K. Analysis of transforming Pathol., 15: 1136–1145, 1991. activity of human synovial sarcoma-associated chimeric protein SYT-SSX1 bound to 33. Majeste, R. M., and Beckman, E. N. Synovial sarcoma with an overwhelming chromatin remodeling factor hBRM/hSNF2 ␣. Proc. Natl. Acad. Sci. USA, 98: epithelial component. Cancer (Phila.), 61: 2527–2531, 1988. 3843–3848, 2001. 34. Helbling, P. M., Tran, C. T., and Brandli, A. W. Requirement for EphA receptor 14. dos Santos, N. R., de Bruijn, D. R., and van Kessel, A. G. Molecular mechanisms signaling in the segregation of Xenopus third and fourth arch neural crest cells. Mech. underlying human synovial sarcoma development. Genes Chromosomes Cancer, 30: 1–14, 2001. Dev., 78: 63–79, 1998. 15. Kawai, A., Woodruff, J., Healey, J. H., Brennan, M. F., Antonescu, C. R., and 35. Dupin, E., Glavieux, C., Vaigot, P., and Le Douarin, N. M. Endothelin 3 induces the Ladanyi, M. SYT-SSX gene fusion as a determinant of morphology and prognosis in reversion of melanocytes to glia through a neural crest-derived glial-melanocytic synovial sarcoma. N. Engl. J. Med., 338: 153–160, 1998. progenitor. Proc. Natl. Acad. Sci. USA, 97: 7882–7887, 2000. 16. Antonescu, C. R., Kawai, A., Leung, D. H., Lonardo, F., Woodruff, J. M., Healey, 36. Ring, C., Hassell, J., and Halfter, W. Expression pattern of collagen IX and potential J. H., and Ladanyi, M. Strong association of SYT-SSX fusion type and morphologic role in the segmentation of the peripheral nervous system. Dev. Biol., 180: 41–53, epithelial differentiation in synovial sarcoma. Diagn. Mol. Pathol., 9: 1–8, 2000. 1996. 17. Alizadeh, A. A., Eisen, M. B., Davis, R. E., Ma, C., Lossos, I. S., Rosenwald, A., 37. Wendling, O., Dennefeld, C., Chambon, P., and Mark, M. Retinoid signaling is Boldrick, J. C., Sabet, H., Tran, T., Yu, X., Powell, J. I., Yang, L., Marti, G. E., essential for patterning the endoderm of the third and fourth pharyngeal arches. Moore, T., Hudson, J., Jr., Lu, L., Lewis, D. B., Tibshirani, R., Sherlock, G., Chan, Development, 127: 1553–1562, 2000. W. C., Greiner, T. C., Weisenburger, D. D., Armitage, J. O., Warnke, R., Staudt, 38. Echelard, Y., Vassileva, G., and McMahon, A. P. Cis-acting regulatory sequences L. M., et al. Distinct types of diffuse large B-cell lymphoma identified by gene governing Wnt-1 expression in the developing mouse CNS. Development, 120: expression profiling. Nature (Lond.), 403: 503–511, 2000. 2213–2224, 1994. 18. Perou, C. M., Sorlie, T., Eisen, M. B., van de Rijn, M., Jeffrey, S. S., Rees, C. A., 39. Ohuchi, H., Kimura, S., Watamoto, M., and Itoh, N. Involvement of fibroblast growth Pollack, J. R., Ross, D. T., Johnsen, H., Akslen, L. A., Fluge, O., Pergamenschikov, factor (FGF)18-FGF8 signaling in specification of left-right asymmetry and brain and A., Williams, C., Zhu, S. X., Lonning, P. E., Borresen-Dale, A. L., Brown, P. O., and limb development of the chick embryo. Mech. Dev., 95: 55–66, 2000. Botstein, D. Molecular portraits of human breast tumours. Nature (Lond.), 406: 40. Kasai, T., Shimajiri, S., and Hashimoto, H. Detection of SYT-SSX fusion transcripts 747–752, 2000. in both epithelial and spindle cell areas of biphasic synovial sarcoma using laser 19. Khan, J., Wei, J. S., Ringner, M., Saal, L. H., Ladanyi, M., Westermann, F., Berthold, capture microdissection. Mol. Pathol., 53: 107–110, 2000. F., Schwab, M., Antonescu, C. R., Peterson, C., and Meltzer, P. S. Classification and 41. Birdsall, S., Osin, P., Lu, Y. J., Fisher, C., and Shipley, J. Synovial sarcoma specific diagnostic prediction of cancers using gene expression profiling and artificial neural translocation associated with both epithelial and spindle cell components. Int. J. networks. Nat. Med., 7: 673–679, 2001. Cancer, 82: 605–608, 1999. 20. Kitahara, O., Furukawa, Y., Tanaka, T., Kihara, C., Ono, K., Yanagawa, R., Nita, 42. Hollowood, K., and Fletcher, C. D. Malignant fibrous histiocytoma: morphologic M. E., Takagi, T., Nakamura, Y., and Tsunoda, T. Alterations of gene expression pattern or pathologic entity? Semin. Diagn. Pathol., 12: 210–220, 1995. during colorectal carcinogenesis revealed by cDNA microarrays after laser-capture 43. Nascimento, A. G. Dedifferentiated liposarcoma. Semin. Diagn. Pathol., 18: 263– microdissection of tumor tissues and normal epithelia. Cancer Res., 61: 3544–3549, 266, 2001. 2001. 44. Downes, K. A., Goldblum, J. R., Montogomery, E. A., and Fisher, C. Pleomorphic 21. Lin, Y. M., Ono, K., Satoh, S., Ishiguro, H., Fujita, M., Miwa, N., Tanaka, T., liposarcoma: a clinicopathologic analysis of 19 cases. Mod. Pathol., 14: 179–184, Tsunoda, T., Yang, K. C., Nakamura, Y., and Furukawa, Y. Identification of AF17 as 2001.

5866

Satoshi Nagayama, Toyomasa Katagiri, Tatsuhiko Tsunoda, et al.

Cancer Res 2002;62:5859-5866.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/62/20/5859

Cited articles This article cites 42 articles, 11 of which you can access for free at: http://cancerres.aacrjournals.org/content/62/20/5859.full#ref-list-1

Citing articles This article has been cited by 11 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/62/20/5859.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/62/20/5859. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.