Published OnlineFirst April 15, 2015; DOI: 10.1158/1078-0432.CCR-14-2910
Biology of Human Tumors Clinical Cancer Research Gene Expression Profiling of Desmoid Tumors by cDNA Microarrays and Correlation with Progression-Free Survival Sebastien Salas1,2, Celine Brulard3, Philippe Terrier4, Dominique Ranchere-Vince5, Agnes Neuville3, Louis Guillou6, Marick Lae7, Agnes Leroux8, Olivier Verola9, Kurtz Jean-Emmanuel10, Sylvie Bonvalot11, Jean-Yves Blay12, Axel Le Cesne13, Alain Aurias3, Jean-Michel Coindre3,14, and Frederic Chibon3,15
Abstract
Purpose: Because desmoid tumors exhibit an unpredictable itive predictive value (PPV) and negative predictive value clinical course, translational research is crucial to identify the (NPV). predictive factors of progression in addition to the clinical Results: Multivariate analysis showed that our molecular sig- parameters. The main issue is to detect patients who are at a nature had a significant impact on PFS while no clinical factor had higher risk of progression. The aim of this work was to identify any prognostic value. Among the 1,000 random signatures gen- molecular markers that can predict progression-free survival erated, 56.7% were significant and none was more significant than (PFS). our 36-gene molecular signature. PPV and NPV were high Experimental Design: Gene-expression screening was con- (75.58% and 81.82%, respectively). Finally, the top two genes ducted on 115 available independent untreated primary downregulated in no-recurrence were FECH and STOML2 and the desmoid tumors using cDNA microarray. We established a top gene upregulated in no-recurrence was TRIP6. prognostic gene-expression signature composed of 36 genes. Conclusions: By analyzing expression profiles, we have iden- To test robustness, we randomly generated 1,000 36-gene tified a gene-expression signature that is able to predict PFS. This signatures and compared their outcome association to our tool may be useful for prospective clinical studies. Clin Cancer Res; define 36-genes molecular signature and we calculated pos- 21(18); 4194–200. 2015 AACR.
Introduction (1, 2), but because desmoid tumors exhibit an unpredictable clinical course and an indistinguishable morphology, translation- Desmoid tumors are mesenchymal fibroblastic/myofibroblas- al research is crucial to identify the predictive factors of progres- tic proliferations. The major obstacle in the management of sion in addition to the clinical parameters. A significant improve- desmoid tumors is their high propensity for local recurrence ment would be to be able to detect patients who are at a higher risk even after complete surgical removal. Currently, an initial of progression and those with no risk of progression. "wait-and-see" policy is explored as a possible standard of care Biologically, alterations of the APC (mutation or loss of the entire locus) and CTNNB1 mutation might constitute an initial mutually exclusive alteration (3, 4). Moreover, Salas and collea- 1 2 Aix Marseille Univ, CRO2, INSERM U911, Marseille, France. APHM, gues described three recurrent and relevant alterations of chromo- Timone Hospital, Department of Medicine, Division of Adult Oncology, Marseille, France. 3Department of Pathology, INSERM U916, Bergonie somes 8, 20, and 6 by array comparative genomic hybridization Institute, Bordeaux, France. 4Department of Pathology, Gustave (CGH) array. These alterations could be involved in the same Roussy Institute, Villejuif, France. 5Department of Pathology, Leon 6 pathway or could confer a selective advantage. Patients harboring Berard Center, Lyon, France. University Institute of Pathology, Lau- CTNNB1 sanne, Switzerland. 7Department of Pathology, Institut Curie, Paris, mutations, in particular CTNNB1 (45F) mutations, are France. 8Department of Pathology, AlexisVautrin Center, Nancy, at risk of recurrence and the wild-type appears to be a good France. 9Department of Pathology, Saint-Louis Hospital, Paris, France. prognostic marker (5, 6). There have been only a few reports 10Department of Oncology and Hematology, University Hospital, Strasbourg, France. 11Department of Surgery, Gustave Roussy Insti- concerning gene-expression analysis in desmoid tumors. One of tute, Villejuif, France. 12Department of Medicine, Leon Berard Center, them demonstrated that a gene-expression signature could dis- Lyon, France. 13Department of Medicine, Gustave Roussy Institute, tinguish desmoid tumors from nodular fasciitis and suggested 14 Villejuif, France. Victor Segalen University Bordeaux, Bordeaux, that selected tyrosine kinases, transcription factors, and members France. 15Translational Research, Bergonie Institute, Bordeaux, France. of the Wnt, TGFb, IFN, and TNF signaling pathways could dis- Note: The data used in this article were provided by the French Sarcoma tinguish these two entities (7). Another study has shown that it Group database as part of the ConticaBase (www.conticabase.org). was identified genes, in particular ADAM12, WISP-1, and SOX 1, Corresponding Author: Sebastien Salas, Aix Marseille University, 13005 which were uniquely overexpressed in 12 cases of aggressive Marseille, France. Phone: 33-4-91-38-57-08; Fax: 33-4-91-38-76-58; E-mail: fibromatosis compared with expression in normal skeletal tissues [email protected] and a variety of normal tissues. The authors concluded that gene- doi: 10.1158/1078-0432.CCR-14-2910 expression patterns may be useful in the classification of subtypes 2015 American Association for Cancer Research. of aggressive fibromatosis (8). Finally, Colombo and colleagues
4194 Clin Cancer Res; 21(18) September 15, 2015
Downloaded from clincancerres.aacrjournals.org on September 24, 2021. © 2015 American Association for Cancer Research. Published OnlineFirst April 15, 2015; DOI: 10.1158/1078-0432.CCR-14-2910
Expression Data from 115 Desmoid Tumors
dose of 50 Gray. Forty-three patients received radiotherapy. Sur- Translational Relevance gery was followed by radiotherapy in 16 patients. Clinical and The aim of this study was to identify molecular markers that histologic data were entered into a centralized computerized can predict progression-free-survival (PFS) and thus to distin- database (www.conticabase.org). All samples were obtained after guish those desmoid tumors for which the use of aggressive informed consent from patients. treatment is justified rather than a "wait-and-see" strategy. This study clearly demonstrates that there is prognostic molecular Pathology review signature of desmoid tumors that could benefit from different Histologic slides of all patients entered in this study were therapeutic strategies. The main question raised is how reviewed by the pathology subcommittee of the French Sarcoma patients should be managed. Should patients with poor prog- Group (GSF). This subcommittee included 20 pathologists and a nostic molecular signature be operated straightaway in a monthly slide review session was performed. For each tumor, one curative intent and/or benefit from early medical treatment? to eight slides were reviewed collegially. Histologic typing was This study is the starting point for prospective studies, the only based on the World Health Organization (WHO) histologic way to answer these questions and optimize the management typing of soft tissue tumors. Histopathologic diagnosis was con- of desmoid tumors. Prospective validation of the molecular firmed by the search for CTNNB mutations. All training samples signature is under way in France through a clinical trial had the CTNNB1 mutation. evaluating the "wait-and-see" strategy (ClinicalTrials.gov iden- tifier NCT01801176). RNA extraction and cDNA array Total RNAs were extracted from frozen tumor samples with TRizol reagent (Life Technologies, Inc.) and purified using the RNeasy Min Elute TM Cleanup Kit (Qiagen) according to the manufacturer's procedures. RNA quality was checked on an Agi- (9) compared the gene-expression profiles of 14 sporadic lent 2100 bioanalyzer (Agilent Technologies). Samples were then desmoid tumors to those of five normal tissues acquired from analyzed on Human Genome U133 Plus 2.0 array (Affymetrix), corresponding desmoid tumor patients and six solitary fibrous according to the manufacturer's procedures. All microarray data tumor specimens. The protein products of three of the upre- were simultaneously normalized using the GCRMA algorithm gulated desmoid tumor genes, ADAM12, MMP2,andmidkine, (Wu J and Gentry (2014) RIwcfJMJ. GRCMA: Background Adjust- werefoundtobecommonlyexpressedinalargecohortof ment Using Sequence Information. R package version 2.40.0). human desmoid tumor samples assembled on a tissue micro- Minimum information about a microarray experiment–compliant array. Overexpression of midkine was significantly correlated data have been deposited at Gene Expression Omnibus under with decreased time to primary recurrence. Moreover, midkine accession number GSE58697. was found to enhance the migration and invasion of primary desmoid tumor cell cultures. These studies suggest the utility of Gene-expression analysis midkine as a clinical desmoid tumor molecular prognosticator Differential expression was established using the limma and a potential therapeutic target. Midkine could be a novel R package and P values were adjusted using the Benjamini– b-catenin transcriptional target (9). However, no relationship Hochberg procedure (10). Analysis of variance was performed was noted between gene-expression profiling and the clinical by using GeneSpring GX software (Agilent Technologies) one-way course of the disease. ANOVA test and P values were adjusted by using the Benjamini– In this study, molecular markers predictive for progression-free survival (PFS) were investigated using gene-expression screening Table 1. Patient and disease characteristics at baseline that was conducted on 115 available independent untreated Characteristics Cohort A Cohort B Cohorts A þ B primary desmoid tumors using cDNA microarray. (n ¼ 66) (n ¼ 49) (n ¼ 115) Median follow-up 2.36 (0.03–12.04) 1.43 (0.33–11.07) 1.82 (0.08–11.97) Materials and Methods (years) (IC 95) Age at diagnosis (%) Patients and samples 37 years 27 (41) 25 (50) 52 (45) From February 1, 1987 to March 6, 2008, 115 consecutive >37 years 37 (56) 24 (50) 61 (53) patients with sporadic aggressive fibromatosis were diagnosed for Nd 2 (3) 2 (2) their first tumoral event in 16 participating cancer centers. Among Male sex (%) 23 (35) 20 (41) 43 (37) Location (%) tumors in the 115 patients, 66 formed the training samples. The Intra-abdominal 4 (6) 8 (16) 12 (10) diagnosis of desmoid tumors was confirmed in each case by Abdominal wall 13 (20) 9 (18) 22 (18) collegial histologic analysis (mesenchymal fibroblastic/myofibro- Extra-abdominal 49 (74) 32 (66) 71 (62) blastic proliferations). The following clinical data were collected: Size (%) gender, age at diagnosis, location (intra-abdominal, abdominal 7 cm 35 (53) 20 (41) 55 (48) < wall, and extra-abdominal), size of tumor, and follow-up 7 cm 17 (26) 22 (45) 39 (34) Nd 14 (21) 7 (14) 21 (18) (Table 1). All patients had an initial surgical resection. Histologic Progression number (%) evaluation of surgical margins was available in 91 (79%) cases. 0 48 (72) 14 (29) 62 (54) Forty-seven patients (41%) had R0 resection, 32 (28%) had R1 1 0 30 (61) 30 (26) resection, and 12 (10%) had R2 resection (macroscopic incom- 2 13 (20) 1 (2) 14 (12) plete as R2 resection; microscopic incomplete resection as R1 3 5 (8) 1 (2) 6 (5) > resection; microscopic complete resection as R0 resection). Radio- 3 0 3(6) 3(3) Mutations CTNNB1 (%) 66 (100) 29 (60) 95(82) therapy generally included photons or electrons with a median
www.aacrjournals.org Clin Cancer Res; 21(18) September 15, 2015 4195
Downloaded from clincancerres.aacrjournals.org on September 24, 2021. © 2015 American Association for Cancer Research. Published OnlineFirst April 15, 2015; DOI: 10.1158/1078-0432.CCR-14-2910
Salas et al.
Hochberg procedure. To assign prognosis, we applied the nearest Results centroid method. Centroids represent a centered mean of expres- Establishing molecular signature sion for the signature genes for each patient outcome (recurrence In unsupervised analysis, no molecular signature could be and no-recurrence) of the training sets. Each sample of the identified. To identify molecular markers that can predict PFS of validation sets was allocated to the prognostic class (centroid) desmoid tumors, we compared expression profiles in supervised with the highest Spearman correlation. Gene Ontology (GO) analyses of 48 patients without recurrence versus 18 patients analysis was performed to establish statistical enrichment in GO with two or three local recurrences. We identified 18 genes terms by using Gorilla software (11). upregulated in recurrence samples and 18 genes downregulated Statistical analysis in recurrence samples (Table 2). The supervised analysis compar- PFS is defined as time from the date of initial diagnosis to the ing patients without recurrence with those having one or more date of progression or recurrence or last follow-up. Local recurrences did not identify any prognostic molecular signature. recurrence-free survival (LRFS) is defined as time from the date Classification based on biologic process categories from GO of initial diagnosis to the date of recurrence or last follow-up. (www.geneontology.org) of the 36 genes demonstrated four We chose to study PFS because the validation cohort consisted enriched GO process terms: cellular component organization or of patients who underwent surgery that was not only R0 and R1 biogenesis (GO:0071840), cellular component organization but also R2. Survival curves were obtained by the Kaplan–Meier (GO:0016043), single-organism organelle organization method and compared with the log-rank test. All survival (GO:1902589), organelle organization (GO:0006996), which analyses were performed by using R software, version 2.14.1 include 16, 16, 10, and 12 genes, respectively (Table 3). The (R development Core Team, Vienna Austria, 2009) and survival interaction of the lists of genes from the ANOVA analysis com- package (Therneau T (2013). A Package for Survival Analysis in paring 0-versus-2, 0-versus-3, and 2-versus-3 recurrences identi- S. R package version 2.37-4). The Cox proportional hazards fied three genes as shown in the Venn diagram (Fig. 1): FECH, model was used to calculate adjusted hazard ratios (HR) and STOML2, and TRIP6. TRIP6 was overexpressed in the good out- their 95% confidence intervals (95% CI). Variables with a P come group unlike STOML2 and FECH, which were overexpressed valuelessthan0.05inunivariateanalysesweretestedinthe in the poor outcome group. multivariate analysis. Multivariate analysis was performed by using Cox regression with Firth's correction (R by Meinhard Prognostic factors for PFS and molecular signature validation Ploner and Fortran by Georg Heinze, 2012; coxphf: Cox regres- To test whether clinical factors can predict PFS, we performed sion with Firth's penalized likelihood. R package, version 1.09). univariate analysis for age at diagnosis (cutoff, 37 years), tumor The positive predictive value (PPV) is the ratio of the number of site (intra-abdominal versus abdominal wall versus extra- patients having neither a recurrence nor progression at the end abdominal) and tumor size (cutoff, 7 cm) in the whole cohort. of follow-up and classified in the group with a good prognostic Tumor site and tumor size were not significant but age at molecular signature over the total number of patients classified diagnosis (P ¼ 4.1e10 02)hadasignificant impact on PFS. as having a good prognostic signature. The negative predictive Survival analysis was also established for a 36-gene molecular value (NPV) is the ratio of the number of patients whose signature with centroid defined on cohort A (n ¼ 66) and disease recurred or progressed and were classified as having a classification and validation defined on the whole cohort poor prognostic signature over the total number of patients (n ¼ 115). Univariate analysis showed that the molecular classified as having a poor prognostic signature. Sensitivity, signature predicted PFS (P ¼ 1.7e10 07;Fig.2).Themolecular specificity, and accuracy were also calculated for the molecular signature also has a prognostic value in LRFS (only the signature. recurrence is considered as an event) in the validation cohort
Table 2. Final gene set containing 36 genes: 18 genes were upregulated in no-recurrence samples and 18 genes were downregulated in no-recurrence samples Downregulated genes in recurrence Upregulated genes in recurrence Probe set id Corrected P value Gene symbol Probe set id Corrected P value Gene symbol 209129_at 8.93e 05 TRIP6 203116_s_at 3.23e 07 FECH 226728_at 1.17e 04 SLC27A1 215416_s_at 1.09e 05 STOML2 204986_s_at 1.77e 04 TAOK2 225439_at 2.49e 05 NUDCD1 229377_at 2.55e 04 GRTP1 200014_s_at 1.34e 05 HNRNPC 206846_s_at 4.27e 04 HDAC6 226312_at 2.26 e05 RICTOR 220128_s_at 5.22e 04 NPAL2 230465_at 2.73e 05 HS2ST1 231767_at 6.35e 04 HOXB4 202854_at 3.58e 04 HPRT1 236229_at 8.01e 04 Hs.661286 211727_s_at 3.66e 04 COX11 203204_s_at 8.22e 04 JMJD2A 227211_at 3.70e 04 PHF19 214251_s_at 9.82e 04 NUMA1 229253_at 7.60e 04 THEM4 215692_s_at 9.93e 04 MPPED2 200006_at 8.07e 04 PARK7 236123_at 1.01e 03 ST7L 226776_at 8.22e 04 ENY2 228929_at 1.05e 03 DNASE1 203312_x_at 8.27e 04 ARF6 244614_at 1.48e 03 TFG 203606_at 1.40e 03 NDUFS6 232463_at 1.62e 03 CXYorf10 200749_at 1.47e 03 RAN 237317_at 1.68e 03 Hs.127312.0 217880_at 2.69e 03 CDC27 232331_at 2.26e 03 Hs.25717.0 202121_s_at 3.17e 03 CHMP2A 234106_s_at 2.43e 03 FLYWCH1 226596_x_at 3.32e 03 LOC729852
4196 Clin Cancer Res; 21(18) September 15, 2015 Clinical Cancer Research
Downloaded from clincancerres.aacrjournals.org on September 24, 2021. © 2015 American Association for Cancer Research. Published OnlineFirst April 15, 2015; DOI: 10.1158/1078-0432.CCR-14-2910
Expression Data from 115 Desmoid Tumors