clinically relevant molecular subtypes in leiomyosarcoma · biology of human tumors clinically...

Biology of Human Tumors

Clinically Relevant Molecular Subtypes inLeiomyosarcomaXiangqian Guo1, Vickie Y. Jo2, Anne M. Mills3, Shirley X. Zhu1, Cheng-Han Lee4,Inigo Espinosa5, Marisa R. Nucci2, Sushama Varma1, Erna Forg�o1, Trevor Hastie6,Sharon Anderson1, Kristen Ganjoo7, Andrew H. Beck8, Robert B.West1,Christopher D. Fletcher2, and Matt van de Rijn1

Abstract

Purpose: Leiomyosarcoma is a malignant neoplasm withsmooth muscle differentiation. Little is known about its molec-ular heterogeneity and no targeted therapy currently exists forleiomyosarcoma. Recognition of different molecular subtypesis necessary to evaluate novel therapeutic options. In a previousstudy on 51 leiomyosarcomas, we identified three molecularsubtypes in leiomyosarcoma. The current study was performedto determine whether the existence of these subtypes could beconfirmed in independent cohorts.

Experimental Design: Ninety-nine cases of leiomyosarcomawere expression profiled with 30end RNA-Sequencing (3SEQ).Consensus clustering was conducted to determine the optimalnumber of subtypes.

Results: We identified 3 leiomyosarcoma molecular sub-types and confirmed this finding by analyzing publicallyavailable data on 82 leiomyosarcoma from The CancerGenome Atlas (TCGA). We identified two new formalin-fixed,paraffin-embedded tissue-compatible diagnostic immuno-

histochemical markers; LMOD1 for subtype I leiomyosar-coma and ARL4C for subtype II leiomyosarcoma. A leiomyo-sarcoma tissue microarray with known clinical outcomewas used to show that subtype I leiomyosarcoma is associ-ated with good outcome in extrauterine leiomyosarcomawhile subtype II leiomyosarcoma is associated with poorprognosis in both uterine and extrauterine leiomyosarcoma.The leiomyosarcoma subtypes showed significant differ-ences in expression levels for genes for which novel targetedtherapies are being developed, suggesting that leiomyosar-coma subtypes may respond differentially to these targetedtherapies.

Conclusions: We confirm the existence of 3 molecular sub-types in leiomyosarcoma using two independent datasets andshow that the different molecular subtypes are associatedwith distinct clinical outcomes. The findings offer an opportu-nity for treating leiomyosarcoma in a subtype-specific targetedapproach. Clin Cancer Res; 21(15); 3501–11. �2015 AACR.

IntroductionLeiomyosarcoma is reported to represent as many as 24% of

all sarcomas (1, 2). Currently, no targeted therapy exists forleiomyosarcoma and the tumor responds poorly to conven-tional chemotherapy or radiotherapy. A significant proportionof leiomyosarcoma originates in the uterus, and the remainder

of leiomyosarcoma originates in various soft tissue siteswhere it is thought to often arise from smooth muscle cellsin vessel walls.

The successful stratification of several tumors (includingbreast cancer, lung cancer, and colon cancer) into molecularsubtypes in the past decades has significantly improved ourknowledge of these malignancies and has led to changes in thetherapeutic approach to these cancers (3–12). In a previousstudy, our group proposed the existence of three molecularsubtypes of leiomyosarcoma by using microarray-based geneexpression profiling on 51 fresh frozen samples of leiomyo-sarcoma (13). To validate this finding, we analyzed 99 leio-myosarcoma cases collected at different institutions from theyears 1991 to 2012 as an independent cohort, and performedexpression profiling with 30 end RNA sequencing (3SEQ;refs. 14–20). In addition, we analyzed the publically availableexpression profiling dataset from TCGA on 82 cases. Our ana-lysis confirms the existence of three molecular subtypes inleiomyosarcoma. Using two novel markers, we could distin-guish the three subtypes by immunohistochemistry and corre-late the subtypes with clinical outcome. The recognition ofthese molecular subtypes in leiomyosarcoma may have clinicalsignificance as the subtypes vary in their expression of a numberof genes for which novel targeted therapies either already existor are being developed.

1Department of Pathology, Stanford University School of Medicine,Stanford, California. 2Department of Pathology, Brigham andWomen's Hospital and Harvard Medical School, Boston, Massachu-setts. 3DepartmentofPathology,UniversityofVirginia,Charlottesville,Virginia. 4Division ofAnatomical Pathology, RoyalAlexandraHospital,Edmonton, Alberta,Canada. 5Department of Pathology, Hospital de laSanta Creu i Sant Pau, Autonomous University of Barcelona, Barce-lona, Spain. 6Department of Statistics, Stanford University, Stanford,California. 7Stanford Comprehensive Cancer Center, Stanford Univer-sity, Stanford, California. 8Department of Pathology, Beth IsraelDeaconess Medical Center and Harvard Medical School, Boston,Massachusetts.

Note: Supplementary data for this article are available at Clinical CancerResearch Online (http://clincancerres.aacrjournals.org/).

Corresponding Author: Matt van de Rijn, Department of Pathology, StanfordUniversity School of Medicine, 300 Pasteur Drive, Stanford, CA 94305. Phone:650-498-7154; Fax: 650-723-1950; E-mail: [email protected]

doi: 10.1158/1078-0432.CCR-14-3141

�2015 American Association for Cancer Research.

ClinicalCancerResearch

www.aacrjournals.org 3501

on March 27, 2021. © 2015 American Association for Cancer Research. clincancerres.aacrjournals.org Downloaded from

Published OnlineFirst April 20, 2015; DOI: 10.1158/1078-0432.CCR-14-3141

http://clincancerres.aacrjournals.org/

Materials and Methods3SEQ library construction and bioinformatics analysis

Paraffin blocks of 99 leiomyosarcomas, 4 myometrium, 3leiomyoma, and 6 undifferentiated pleomorphic sarcomas from1991 to 2012 from 9 hospitals were obtained with Institutionalreview board approval and a waiver of consent due to the archivalnature of the specimens. Multiple 2 mm-diameter cores were re-embedded into paraffin blocks longitudinally and sectionedagain to ensure the purity of tumor cells. After RNA isolation,3SEQ libraries for next-generation sequencing–based expressionprofilingwere sequenced and analysiswas performed as describedpreviously (14–20). All gene expression profiling data used forthis study have been deposited in the Gene Expression Omnibus(GEO) and are publicly accessible throughGSE45510, GSE53844and GSE54511.

After filtering data to genes with a SD greater than 100, trans-forming thedata by log2 andgene-based centering, theConsensusClustering (R package ConsensusClusteringPlus; ref. 21)was usedto determine the number of subtypes and to assign the subtype foreach leiomyosarcoma case. Thiswas ranover 1,000 iterationswithsetting "Distance – (1-Pearson correlation), a 80% sample resam-pling, 80% gene resampling, a maximum evaluated k of 12, andagglomerative hierarchical clustering algorithm." Silhouettewidth (R package cluster; ref. 22) was then calculated on the basisof the assignment from ConsensusClusteringPlus to measurethe accuracy of assignments from ConsensusClusteringPlus. Sep-arately, RNASeq data of 82 leiomyosarcoma cases were down-loaded from the TCGA database. To compare with 3SEQ data,the TCGA data were normalized into TPM and analyzed withConsensusClusteringPlus and Silhouette analysis as performedfor 3SEQ data. Subclass mapping (23) was performed to deter-mine the common leiomyosarcoma subtypes identified in 3SEQand TCGA RNASeq datasets.

To get subtype specific genes, SAMSeq (24) was performed onboth datasets between each subtype and all other subtypes with afalse discovery rate (FDR) of 0.05. Kaplan–Meier plots were usedto compute the survival curves, and Log-rank (Mantel–Cox) testwas used to determine the statistical significance of survivalbetween groups by using GraphPad Prism5 software. Significance

of contingency analysis between one subtype and the othersubtypes was measured by two-tailed Fisher exact test and c2 testwith GraphPad Prism5 software. Univariate and multivariateanalysis by the Cox proportional hazard method was performedusing the survival package in R. Analysis of gene ontology (GO)was done using DAVID Bioinformatics Resources version 6.7(25, 26). CSF1-signature (27) positive, and CINSARC signature(28) positive cases were identified as those cases that coordi-nately highly expressed these signature genes with >0.3 correla-tion with the centroid as performed previously (19, 27, 29, 30).Principal component analysis (PCA) was performed on squareroot transformed TPM with R package pcaMethods, ellipse con-tour of principal components (PC) was computed and drawnwith R package car.

TMA construction, immunohistochemistry staining, andscoring

The tissue microarrays used in this study (TA-201 andTA-381) were constructed with Tissue Arrayer (Beecher Instru-ments). TA-201 included the leiomyosarcoma cases with clinicoutcome data, while TA-381 was comprised of leiomyosarcomacases analyzed by 3SEQ. For immunohistochemical staining,sections underwent antigen retrieval and were stained withanti-ARL4C antibody (1:120, Sigma, CAT#HPA028927) andanti-LMOD1 antibody (1:20, Sigma, CAT#HPA028325) withthe EnVisionþsystem (Dako). The staining results were scoredas follows: 3 (þþþ), strong staining (>30% positivity); 2 (þþ),weak staining (10%–30%); 1 (þ), equivocal or uninterpretable(<10%); 0 (�), absence of any staining. Hierarchical clusteringof immunohistochemical (IHC) results was performed withsoftware "Cluster3" with "uncentered Correlation" and "Cen-troid Linkage" and a clustered heatmap was visualized withJava TreeView (31, 32).

ResultsConsensus clustering identifies three molecular subtypes ofleiomyosarcoma based on gene expression data

We analyzed 99 cases of leiomyosarcoma by 3SEQ, a next-generation sequencing–based approach that quantifies thenumber of 30 ends for each polyadenylated mRNA or lncRNAand that performs well on RNA isolated from FFPE material(14–20). Clinical information of the cases is shown in Table 1and Supplementary Table S1. The median age for this cohort is56 years, with 49 cases originating in the uterus and 49 casesin extrauterine sites. For one case the origin was unknown. Afemale:male ratio of 3.3:1 was noted in this cohort. Half ofthe extrauterine leiomyosarcomas (24 cases) were from theextremities, while the rest were from thoracic, abdominal, orretroperitoneal regions.

3SEQ yielded an average of 18 million total reads afterquality filtering and an average of 2 million uniquely mappingreads against human transcriptome (refMrna) per sample. Allreads that uniquely mapped to an individual gene (UCSCgenome browser, hg19) were combined to obtain a quantifi-cation of the expression level for that gene. The genes with thehighest degree of variation in expression levels across allsamples were identified by filtering the dataset for genes witha SD �100, yielding 1,300 genes. These were used to performConsensus Clustering (21), a method that uses data resamplingfollowed by iterative reclustering to estimate cluster stability

Translational Relevance

Leiomyosarcoma is a malignant neoplasm with varyingdegrees of smoothmuscle differentiation and complex genet-ic abnormalities. It can occur in a wide range of sites but isusually managed as a single disease in conventional therapyand in clinical trials. The recognition of molecular subtypescan lead to the identification of novel therapies that mayaffect one of these subtypes preferentially. We confirmedthe results of an earlier study and determined the existenceof 3 molecular subtypes in leiomyosarcoma using twoindependent cohorts. We identified two new formalin-fixed,paraffin-embedded (FFPE) tissue-compatible diagnosticimmunohistochemical markers for two of these subtypes:LMOD1 for subtype I and ARL4C for subtype II, and showedthat these molecular subtypes are associated with distinctclinical outcomes. Our study offers an opportunity for aleiomyosarcoma subtype–specific targeted treatment.

Guo et al.

Clin Cancer Res; 21(15) August 1, 2015 Clinical Cancer Research3502




and to help determine the optimal number of clusters in adataset. As shown in Fig. 1A and B, the greatest increase in thearea under the CDF curve (empirical cumulative distributionfunction) is seen when 3 molecular subtypes are assumed.Assuming 4 molecular subtypes showed a lesser degree ofincrease in the area under the CDF curve and showed nosignificant increase in the ability to classify the cases (Sup-plementary Fig. S1). To keep the number of subtypes to amanageable number, we focused on the existence of threemolecular subtypes of leiomyosarcoma for this study. A con-sensus matrix of the probability that 2 cases of leiomyosar-coma will belong to the same subtype is shown in Fig. 1C.Silhouette analysis (22) was performed to measure the confi-dence with which an individual leiomyosarcoma case couldbe assigned to its subtype (Fig. 1D). This analysis indicated thatall subtype II cases had a positive silhouette value while asubset of subtype I and III leiomyosarcoma had negative values.Similar to studies performed by others, we used only the 70leiomyosarcoma cases with positive Silhouette values as themost reliable examples ("core" leiomyosarcoma cases) forsubsequent analyses (3,8). In the group of 70 core leiomyo-sarcoma cases, 35 belonged to subtype I, 22 to subtype II, while13 were subtype III.

Molecular subtypes of leiomyosarcoma can be reproduced inan independent dataset

To further explore the reproducibility of the three leiomyo-sarcoma subtypes, we downloaded gene expression data for 82leiomyosarcoma cases from the TCGA database. Of these sam-ples, 80 cases were primary leiomyosarcoma while 2 cases wererecurrences. The origin of the tumors differed from the 99 casesanalyzed by 3SEQ with 19 leiomyosarcoma originating fromthe uterus, 13 cases from the extremities and 32 cases fromother extrauterine sites. The remaining 18 cases had an

unknown primary site. In addition to the differences in thesite of origin for the tumors examined there was a significantdifference in the percentage of samples derived from primarylesions versus recurrences between the two cohorts. In the3SEQ cohort 37% of the cases were recurrences or metastaseswhile only 2% of TCGA leiomyosarcoma cases were recurrentand none were metastatic (Supplementary Table S2). Thetwo datasets differed in two additional aspects. First, the TCGAdataset used fresh frozen material whereas our 3SEQ analysiswas performed on FFPE material. Second, whole transcriptomeRNASeq was used for the TCGA dataset while 3SEQ deter-mines the gene expression level based on quantification of 30

end reads. Despite these differences, analysis of the TCGAdataset produced results very similar to those found in the3SEQ cohort. The TCGA dataset yielded 1,174 genes whenfiltered for a SD �100 (1,300 genes in the 3SEQ set). ByConsensus Clustering and Silhouette analysis, the TCGA data-set also showed the greatest increase in area under the CDFcurve when three subtypes were used (Fig. 2A). Comparedwith the cases studied by 3SEQ in our cohort the homogeneityof the subtypes in the TCGA cohort was greater as determinedby consensus clustering (Fig. 2B), possibly due to the fact thatfrozen material and not FFPE material was used for the casesin the TCGA cohort. To measure the reproducibility betweenthe 3SEQ and TCGA cohorts the cases with positive Silhouettevalues from both sets were analyzed by Subclass mapping(23). The 3SEQ-subtype I and -subtype II types were signifi-cantly reproduced as subtype-C3 and -C1 in the TCGA cohortwith FDR corrected P values of 0.0090 for both. The 3SEQ-subtype III did not reach statistical significance when it wascompared with the remaining TCGA-subtype C2 (Fig. 2C;P ¼ 0.0959). A substantial number of genes identified bySAMSeq in the 3SEQ and TCGA datasets showed overlapbetween 3SEQ subtypes I and II and the corresponding TCGA

Table 1. Patient characteristics (N ¼ 99)

Patients, n (%) Subtype I Subtype II Subtype III Other LMSa

Age (year)Median 56 55 55 60 57Range 2–94 2–84 41–67 43–80 38–94

SexFemale 76 (77%) 22 18 12 24Male 23 (23%) 13 4 1 5

LocationUterine 49 (49%) 9 13 12 15TARb 25 (25%) 12 8 1 4Extremities 24 (24%) 13 1 0 10Unknown 1 (1%) 1 0 0 0

GradeLow 18 (18%) 10 3 0 2Intermediate 19 (19%) 10 4 3 5High 62 (63%) 15 15 10 22

RelapsePrimary 62 (63%) 22 9 12 19Local recur 27 (27%) 9 12 0 6Metastasis 10 (10%) 4 1 1 4

Therapyc

Yes 11 (11%) 4 4 1 2No 86 (87%) 29 18 12 27Unknown 2 (2%) 2 0 0 0

Total 99 35 22 13 29aOther leiomyosarcoma, leiomyosarcoma with negative silhouette value.bTAR, thoracic/abdominal/retroperitoneal sites.cAny prior chemotherapy and radiotherapy.

Molecular Subtypes in Leiomyosarcoma

www.aacrjournals.org Clin Cancer Res; 21(15) August 1, 2015 3503




subtypes (Supplementary Table S3). Importantly, five genes forwhich we previously used immunohistochemical markers(13, 33) to identify subtype I leiomyosarcoma (CASQ2,ACTG2, MYLK, SLMAP, and CFL2) were shown to be highlyexpressed on the mRNA level in TCGA subtype C3, as was thenew subtype I marker LMOD1. In addition, the new markerARL4C, which identifies 3SEQ subtype II cases, was highlyexpressed in the corresponding TCGA subtype C1 (see belowand Supplementary Table S4). Together, these data indicate

that the existence of 3 molecular subtypes in leiomyosarcomacan be reproduced on the mRNA level across these differentdatasets.

Clinical features of leiomyosarcoma molecular subtypes inthree datasets

In the 3SEQ dataset, approximately 26%, 59%, and 92% ofsubtype I, subtype II, and subtype III leiomyosarcoma werederived from the uterus with uterine leiomyosarcoma being

Figure 1.Identification of three molecular subtypes in leiomyosarcoma (LMS). A, empirical cumulative distribution plots as a function of the number of molecular subtypesin 3SEQ cohort. B, relative increase in the area under the CDF curve with increasing number of molecular subtypes. C, consensus clustering matrix ofleiomyosarcoma samples using 3 molecular subtypes. D, silhouette analysis of leiomyosarcoma samples based on the assignments from Consensus Clustering.

Guo et al.





significantly associated with subtype III (c2 test; P ¼ 0.0005).However, when evaluating only those 34 cases arising in theuterus, we found that uterine leiomyosarcoma had a roughlyequal chance of belonging to each of themolecular subtypes with9, 13, and 12 cases belonging to subtype I, II, and III, respectively.In contrast, the extrauterine leiomyosarcoma were overrepresent-ed in subtype I and II, with 25, 9, and 1 cases belonging to subtype

I, II, and III, respectively. These results indicate that a subset ofuterine leiomyosarcomas behave as independent molecular sub-type (subtype III) while another subset of uterine leiomyosar-coma together with extrauterine leiomyosarcoma belong to sub-type I and subtype II leiomyosarcoma. In the TCGA dataset, weobserved similar demographic characteristics as in the 3SEQcohort, with TCGA C2–like 3SEQ subtype III being significantly

Figure 2.Identification of molecular subtypes of leiomyosarcoma (LMS) in the TCGA data. A, empirical cumulative distribution plots as a function of the number ofmolecular subtypes in TCGA cohort. B, consensus clustering matrix of leiomyosarcoma samples using 3 molecular subtypes. C, subclass association(SA) matrix between 3SEQ data (subtype I, subtype II, subtype III) and TCGA data (subtype C1, subtype C2, subtype C3). P value is corrected with FDR.D, principal component analysis (PCA) of 70 core 3SEQ leiomyosarcoma cases with other diagnoses. Subtype I leiomyosarcoma, black circle; subtype IIleiomyosarcoma, red triangle; subtype III leiomyosarcoma, green triangle; leiomyoma, cyan solid circle; myometrium, orange solid circle; undifferentiatedpleomorphic sarcomas, blue solid pyramid. The elliptical contours represent the normal probability of 0.75.






associated with uterine leiomyosarcoma (P < 0.0001), whileTCGA C3 (subtype I) and TCGA C1 (subtype II) were overrep-resented in extrauterine sites (Supplementary Table S2). Theseresults correlate well with our findings on 51 leiomyosarcomacases from our previous dataset analyzed by gene microarrays,where the ratio of uterine leiomyosarcoma (13) alsowas higher insubtype III (10/24) than in subtype I (1/13) and subtype II (3/12),but did not reach significance (c2 test; P ¼ 0.1178).

Within the 3SEQ cohort, 18% (18) of tumors are low grade,19% (19) are intermediate grade, and 63% (62) are high gradeby histologic analysis. Low-grade lesions were more frequent insubtype I leiomyosarcoma, with 10 of 35 cases showing low-grade histology while 3 of 22 of subtype II and 0 of 13 ofsubtype III leiomyosarcoma cases are low grade, respectively;however, there was no statistically significant correlationbetween grade and molecular subtype (Table 1, c2 test; P ¼0.0989). The median age for subtype I, subtype II, and subtypeIII was 55, 55, and 60 years, respectively. One-way ANOVAanalysis showed that there was no significant difference in agebetween the three subtypes (P ¼ 0.3769).

Functional annotation of leiomyosarcoma molecular subtypesbased on subtype-specific genes in the 3SEQ dataset

SAMSeq (24) was used to identify genes that were significantlyover/underexpressed in each subtype of leiomyosarcoma ana-lyzed by 3SEQ, again using only those cases with a positiveSilhouette value. In subtype I leiomyosarcoma, 5,900 genes arerelatively overexpressed while 1,900 genes are underexpressedcomparedwith the other two subtypes (Supplementary Table S5).The overexpressed genes were enriched in Gene Ontology (GO)Biology processes that included muscle contraction, muscle sys-tem processes, and cytoskeleton organization. In addition, anumber of genes (LMOD1, MYOCD, CALD1, DES, CNN1,ACTG2, SLMAP, and CASQ2) known to be involved in smoothmuscle functionwere identified in this list, indicating that subtypeI was most associated with normal smooth muscle function(Supplementary Table S6). Consistent with this finding, 4 sam-ples of normal myometrium and 3 cases of benign uterineleiomyoma were grouped with subtype I leiomyosarcoma inprincipal component analysis (Fig. 2D) and clustered with sub-type I when analyzed by unsupervised hierarchical clustering withthe 70 leiomyosarcoma core samples (Supplementary Fig. S2A).As stated above, the overexpression of genes associated withmuscle functionwas also noted on themRNA level in correspond-ing subtype C3 from the TCGA dataset, while the subtype Iidentified previously through gene microarray analysis had highlevels of expression for these genes at both themRNA and proteinlevel (13).

In 3SEQ data on subtype II leiomyosarcoma, 5,100 genes weresignificantly overexpressed and 2,087 genes were expressed atlower levels as compared with the other two subtypes of leio-myosarcoma. These subtype II leiomyosarcoma-specific genes areenriched in GO biologic processes, including translation, trans-lational elongation, and protein localization (SupplementaryTable S6). Subtype II showed significantly less muscle-specificgenes than subtype I. Undifferentiated pleomorphic sarcoma is atumor type that fails to express differentiation markers and thatcan be difficult to distinguish from leiomyosarcoma. Six cases ofundifferentiated pleomorphic sarcomas grouped with subtype IIleiomyosarcoma in principal component analysis (Fig. 2D) andbyunsupervised hierarchical clustering (Supplementary Fig. S2B).

A search for the CSF1 signature, which we previously reported tobe associated with poor outcome for both gynecologic and non-gynecologic leiomyosarcoma (27), in the 3SEQ data showed thatthe cases with a centroid correlation > 0.3 for the CSF1 signaturewere overwhelmingly subtype II leiomyosarcoma (c2test; P <0.0001).

A total of 1,320 genes are significantly overexpressed and7,544 genes are underexpressed in subtype III leiomyosarcomawhen compared with the other two subtypes. Subtype IIIoverexpressed genes are enriched in GO biologic processes,including metabolic process, ion transport, and regulation oftranscription (Supplementary Table S6). When consideringonly the genes with a mean level of expression 50 reads (TPM)or more in any subtype in the Stanford 3SEQ cohort thenumber of overexpressed genes identified by SAMseq willdecrease from 11,125 to 3,350. The same approach willdecrease the number of genes in the TCGA cohort from11,068 to 3,845.

Identification of novel immunohistochemical markers forleiomyosarcoma subtype I and II

Through gene microarray analysis, we previously identified 5IHC markers (ACTG2, SLMAP, MYLK, CFL2, and CASQ2) thatcould identify subtype I leiomyosarcoma in FFPE material(13, 33). The quantitative nature of gene expression profiling by3SEQ allowed for a more detailed search for additional immu-nohistochemical markers for leiomyosarcoma subtypes. LMOD1mRNA was highly expressed in subtype I leiomyosarcoma. Atissue microarray (TA-381) was generated that contained 58 ofthe 70 leiomyosarcoma cases with positive Silhouette values.Immunohistochemistry showed strong staining of a subsetleiomyosarcoma (Fig. 3A). LMOD1 stained positive in 31 leio-myosarcoma cases, 19ofwhichwere subtype I leiomyosarcoma asassigned by 3SEQ analysis, in contrast, 21 of 27 LMOD1-negativeleiomyosarcomas were subtype II or subtype III leiomyosarcoma.LMOD1 protein expression therefore showed a significant asso-ciation with subtype I leiomyosarcoma (c2 test, P ¼ 0.0085; r ¼0.8964, P < 0.0001, Spearman correlation; Supplementary TableS7). Comparison of the LMOD1 staining results with the fivepreviously identified subtype I biomarkers (13, 33) showed thatthe highest correlation of IHC staining (0.65) was obtainedbetween genes ACTG2, SLMAP, and LMOD1. CASQ2 had thelowest correlation with the 5 other genes (r ¼ �0.22) and werefined our panel of subtype I biomarkers to include ACTG2,SLMAP, LMOD1, CFL2, andMYLK, while omitting CASQ2. Usingthis panel, the correlation for the new panel of markers was 0.47(Supplementary Fig. S3). Finally, staining of TA-381 showed asignificant association (P < 0.0001) between LMOD1 immunos-taining and mRNA levels as determined by 3SEQ LMOD1mRNAlevel (Fig. 3B).

Analysis of 3SEQ data showed high levels of mRNA expres-sion for ARL4C in subtype II leiomyosarcoma. Using a FFPEcompatible antibody, strong staining was seen in a subset ofleiomyosarcoma cases (Fig. 3C) with positive staining in 30leiomyosarcoma cases, 17 of which were subtype II leiomyo-sarcoma. Negative staining was seen in 28 leiomyosarcomacases, 23 of which were subtype I or subtype III leiomyosar-coma. This IHC result validates ARL4C to be a subtype IIleiomyosarcoma biomarker not only at the mRNA level (asdetermined by SAMSeq) but also at the protein level (c2 test,P ¼ 0.0033; r ¼ 0.5277, P < 0.0001, Spearman correlation;

Guo et al.





Supplementary Table S7). Comparison between 3SEQ mRNAlevels and ARL4C immunohistochemistry showed a good asso-ciation (P ¼ 0.0046; Fig. 3D). Comparison between LMOD1and ARL4C protein expression and mRNA expression levelsvaried as expected across the leiomyosarcoma subtypes asdefined by 3SEQ (Supplementary Fig. S4).

While in individual cases these markers may not definitelyidentify a case as belonging to a specific subtype, when com-bined these markers do allow us to distinguish large numbers ofcases represented on a TMA in distinct molecular groups. Nooutcome data were available for the leiomyosarcoma samplesused for 3SEQ analysis. To correlate the assignment of leiomyo-sarcoma subtypes with outcome, we used immunostaining dataon TA-201 that contains 127 cases of leiomyosarcoma withknown clinical outcome (27, 34). In this analysis, 48 of 127(38%) leiomyosarcoma cases were defined as subtype I leio-myosarcoma by their coordinate expression of all 5 subtype Ireactive antibodies. These cases demonstrated a better disease-specific survival (DSS) when uterine and extrauterine leiomyo-sarcoma were analyzed together and compared to the remaining

leiomyosarcoma cases (Log-rank test; Fig. 4A; P ¼ 0.0101).However, the difference in clinical outcome was mostly drivenby the extrauterine leiomyosarcoma (Fig. 4B; P ¼ 0.0208) andno difference in outcome was seen for subtype I leiomyosar-coma from uterus (Fig. 4C; P ¼ 0.2113). The association withgood outcome lost its significance in multivariate analysis whentwo prognostic signatures that we previously identified, ROR2(34) and the CSF1 response protein signature (27), were includ-ed (Supplementary Tables S8 and S9).

Subtype II leiomyosarcoma cases were defined as thosereacting for ARL4C on the outcome TMA. Staining for ARL4Cwas seen in 25 of 111 available leiomyosarcoma cases (23%)and these cases were associated with a worse DSS whenall leiomyosarcoma were combined (log-rank test, Fig. 4D;P ¼ 0.0019). This finding was seen in both extrauterine (P ¼0.0352) and uterine (P ¼ 0.0461) leiomyosarcoma. Whilesubtype II was associated with poor outcome in a statisticallysignificant manner in univariate analysis (P ¼ 0.0030, Waldtest), this significance was also lost in the multivariate analysiswhen prognosticators ROR2 and the CSF1 response protein

Figure 3.Immunohistochemical markers for subtype I andsubtype II leiomyosarcoma (LMS). A,representative staining for LMOD1 (score þþþ)and CASQ2 (negative) for a subtype Ileiomyosarcoma case (case # 10129; scale bar,100 mm, 10�). B, comparison of 3SEQ mRNAquantification and immunostaining results onTA-381 that contains 58 cases of leiomyosarcomaanalyzed by 3SEQ; leiomyosarcoma cases positivefor anti-LMOD1 staining have significantly highermRNA expression levels (TPM) thanleiomyosarcoma cases negative for anti-LMOD1(P < 0.0001). The P value was derived fromunpaired t test with Welch correction. C, positiveand negative staining by ARL4C antibody forrepresentative subtype II leiomyosarcoma (top,case # 10019) and subtype I leiomyosarcoma(bottom case # 9994; scale bar, 100 mm, 10�).D, leiomyosarcoma cases positive for anti-ARL4Cstaining have significantly higher mRNAexpression levels (TPM) than leiomyosarcomacases negative for anti-ARL4C (P ¼ 0.0046).






signature were included (refs. 27, 34; Supplementary Tables S8and S9). For correlation with outcome in the IHC analysis,subtype III was defined as those cases that failed to classify aseither subtype I or II. When thus defined, subtype III had anintermediate outcome (Supplementary Fig. S5) but failed toreach prognostic significance in univariate analysis (Supple-mentary Table S8).

Potential clinical applications of leiomyosarcoma subtypingTo explore the potential for clinical implications in the recog-

nition of leiomyosarcoma subtypes, we compared the genesthat are specifically overexpressed in each leiomyosarcomasubtype with genes from the TARGET V2 database known tobe activated by mutations or amplifications (35). This databasecontains genes for which targeted therapy is either currentlyavailable or under development. A large number of these geneswere found to be expressed at different levels between the 3molecular subtypes (Table 2). This suggests that leiomyosar-coma subtypes may respond differentially to those targetedtherapies, an observation that may play a role in determiningthe success rate of novel therapies in clinical trials. For several ofthese targets there may be an incomplete correlation betweenmRNA levels for the mutated or amplified target gene andresponse to drug, but this analysis nevertheless forms a firststep towards personalization of leiomyosarcoma patient care.

In addition to the genes identified from the TARGET V2,we studied the expression level of ROR2, a receptor tyrosinekinase that plays a role in tumor progression and that has a low

Figure 4.IHC markers for subtype I and subtype II leiomyosarcoma (LMS) predict different clinical outcomes. TA-201 with 127 cases of leiomyosarcoma with knownclinical outcome was stained for the panel of 5 subtype I leiomyosarcoma markers and the ARL4C marker for subtype II leiomyosarcoma. The P valuewas from log-rank (Mantel–Cox) test. Coordinate expression of subtype I biomarkers was associated with good outcome when analysed on all cases (A) butthis was mainly due to an effect on extrauterine (EU) leiomyosarcoma (B) but not uterine (U) leiomyosarcoma (C). Subtype II biomarker ARL4Cpredicted worse outcome in leiomyosarcoma (D), with equal effect on extrauterine and uterine leiomyosarcoma (not shown).

Table 2. Targets unique to leiomyosarcoma subtypes

Gene Therapeutic agents

Genes overexpressed in subtype IARAF Sorafenib, vemurafenib, dabrafenib,

RAF inhibitorsCCNE1 CDK2 InhibitorKDR KDR InhibitorsNOTCH1 Notch InhibitorsFGFR2 FGFR InhibitorsFGFR1 FGFR Inhibitors

Genes overexpressed in subtype IIMCL1 TubulinsCDK4 CDK4/6 InhibitorsCTNNB1 WNT InhibitorsAURKA AURKA InhibitorsRHEB mTOR InhibitorsCCND2 CDK4/6 InhibitorsCCND1 Hormone therapy, CDK4/6 inhibitorsMTOR Everolimus, temsirolimus, mTOR inhibitorsMAPK3 Erlotinib, fefitinib, EGFR inhibitorsMAPK1 Erlotinib, gefitinib, EGFR inhibitorsCCND3 CDK4/6 InhibitorsNOTCH2 Notch InhibitorsMAP2K2 MEK Inhibitors

Genes overexpressed in subtype IIIMDM4 MDM4 InhibitorsERBB3 PertuzumabEPHA3 Dasatinib, ephrin inhibitorsESR1 Hormonal therapy

Genes overexpressed in subtype II and IIIEGFR Erlotinib, gefitinib, EGFR inhibitors

Guo et al.





level of expression in normal human tissues. This molecule hasbeen proposed as a novel therapeutic target in leiomyosarcoma(34) and other tumors (36–38) and was found to be morefrequently expressed in subtype II leiomyosarcoma than inthe other subtypes (FDR < 0.05). Immunostaining for theROR2 protein on the TMA containing cases analyzed by 3SEQsimilarly showed a correlation with subtype II leiomyosarcoma(c2 test, P ¼ 0.0347,).

DiscussionLeiomyosarcoma is a malignant soft tissue tumor with com-

plex genetic abnormalities. In clinical practice, the tumors aretreated differently depending on whether they originate fromthe uterus or from extrauterine sites. However, this distinctionis not based on a molecular rationale and appears based, atleast in part, on clinical subspecialization. Recently Italiano andcolleagues demonstrated a molecular difference between leio-myosarcoma occurring in the retroperitoneum and extremities(39), but this article does not include uterine lesions. Leiomyo-sarcomas are tumors that exhibit varying degrees of smoothmuscle differentiation and often appear to be derived fromsmooth muscle cells in the myometrium. They can also arisefrom the walls of blood vessels or connective tissues through-out the body. Conventional chemotherapy and radiotherapyonly have limited effects and surgical resection remains thebest option for treatment. Characterization of molecular sub-types in malignancies can lead to better individualization oftreatment. In a prior study on 51 cases of leiomyosarcoma forwhich frozen tissue was available, we used 44,000 spottedcDNA arrays for gene expression profiling and obtained datathat indicated that at least 3 molecular subtypes exist inleiomyosarcoma (13). This study relied on the availability offresh frozen tissues and as a result only small numbers ofleiomyosarcoma cases could be studied. The 3SEQ approachcan use FFPE materials and renders a digital rather than ananalog readout for gene expression levels with a higher signal tonoise ratio and allows for accurate measurement of genesexpressed at a wide range of levels, including those with lowlevels of expression. 3SEQ analysis on 99 samples of leiomyo-sarcoma and data on a third cohort of 82 leiomyosarcoma casesfrom the TCGA study were obtained. Despite the differences inmethodology and sample selection between these threecohorts, remarkable similarities were apparent and indicatedthe existence of three molecular subtypes in leiomyosarcoma.

Subtype I leiomyosarcoma expresses most genes associatedwith muscle function by Gene Ontology annotation. The sim-ilarity of subtype I leiomyosarcoma to smooth muscle differ-entiation was also supported by the fact that subtype I casesclustered with samples of leiomyoma and myometrium. Final-ly, we observed high expression levels for LMOD1 in subtype Ileiomyosarcoma, a smooth muscle cell-restricted gene that ispreferentially expressed in differentiated smooth muscle cells(40). The expression of smooth muscle markers in subtype Isuggest a tumor subtype that is closer to the smooth musclecells than subtype II and III. However this similarity is notcaptured by the objective measurement of histologic grade, asno correlation between grade and molecular subtype wasfound. Others have noticed the existence of a subset of leio-myosarcoma that express muscle function related genes athigher levels than other leiomyosarcoma cases. Francis and

colleagues (41) identified a 200 gene signature for leiomyo-sarcoma by expression profiling different subtypes of 177 softtissue sarcomas, this signature was characterized by overexpres-sion of muscle-specific genes and was shown to be highlyexpressed in 12 of 40 leiomyosarcoma cases by supervisedclustering. Villacis and colleagues (42) performed gene expres-sion profiling and supervised clustering with 587 genesexpressed differentially between undifferentiated pleomorphicsarcomas and leiomyosarcoma, and found that a subset ofleiomyosarcoma that highly expressed muscle functionassociated genes clustered separately from undifferentiatedpleomorphic sarcomas and other leiomyosarcoma cases.

In contrast to subtype I leiomyosarcoma, subtype IIleiomyosarcoma showed no significant smooth muscle differ-entiation by GO analysis and therefore represents a less differ-entiated form of leiomyosarcoma. Undifferentiated pleomor-phic sarcoma is a type of sarcoma that lacks any indicationof differentiation by histologic examination and by immuno-histochemical analysis. Hierarchical clustering of the coreleiomyosarcoma cases with 6 cases of undifferentiated pleo-morphic sarcomas showed coclustering of undifferentiatedpleomorphic sarcomas with subtype II leiomyosarcoma. Othershave studied the relation between leiomyosarcoma and undif-ferentiated pleomorphic sarcomas by molecular studies but didnot take into account the existence of leiomyosarcoma subtypes(43). Villacis and colleagues performed gene expression pro-filing of 22 leiomyosarcoma and 22 undifferentiated pleomor-phic sarcomas, and did not identify a clear distinction ofexpression profiles of undifferentiated pleomorphic sarcomasand leiomyosarcoma by either unsupervised clustering or evensupervised clustering using a gene list derived from SAM anal-ysis. However, when performing supervised clustering with 587genes expressed differentially between undifferentiatedpleomorphic sarcomas and leiomyosarcoma, a subset of leio-myosarcoma that consisted predominantly of retroperitonealleiomyosarcoma clustered separately from undifferentiatedpleomorphic sarcomas and the remaining leiomyosarcoma, inpart based on higher expression of muscle function associatedgenes in the retroperitoneal leiomyosarcoma (42). These find-ings partially overlap with our results; in our study 6 retroper-itoneal leiomyosarcoma were analyzed, 4 of which grouped insubtype I with 2 cases grouping in subtype II, the subtype mostassociated with undifferentiated pleomorphic sarcomas.

Subtype III leiomyosarcoma is the only subtype that shows apreference for a specific anatomic site and was more likely to befrom uterus (Fisher exact test; P¼ 0.005). Only one of 13 subtypeIII leiomyosarcoma did not originate from the uterus and pre-sented in the scrotum. However, uterine leiomyosarcoma as agroup was evenly distributed over the 3 subtypes.

Neither uterine nor extrauterine leiomyosarcoma haveclinically accepted prognostic molecular markers, and histo-logic grading to predict clinical outcome is controversial inuterine leiomyosarcoma. Several molecular prognosticators(28, 39, 41, 44, 45) have previously been reported in leiomyo-sarcoma. By genomic and expression profiling of 183 sarcomas,Chibon and colleagues established a prognostic signature call-ed Complexity Index in Sarcoma (CINSARC), that includes 67genes related to mitosis and chromosome management. Thissignature was shown to be associated with metastatic outcomein sarcomas, including leiomyosarcoma, and even in lympho-mas and breast carcinoma. The CINSARC signature is a






powerful predictor of tumor metastasis and could improve thepatient selection for chemotherapy (28). To investigate thesignificance of leiomyosarcoma subtypes on clinical outcome,we stained a TMA containing cores from 127 leiomyosarcomacases with clinical outcome with the IHC markers for subtype Iand II leiomyosarcoma. Subtype I leiomyosarcoma was asso-ciated with good outcome in extrauterine leiomyosarcoma,while subtype II was associated with poor outcome in univar-iate analysis. In this context, it is interesting to note that theassociation between muscle gene expression and better clinicaloutcome is a finding that was also reported in various waysby other immunohistochemical studies (18, 45). While sub-typing of leiomyosarcoma by immunohistochemistry canhelp predict clinical outcome, in multivariate analysis thesubtype I and II biomarkers were outperformed by the previ-ously described markers ROR2 and the "CSF1 protein responsesignature" (27, 34). The association between immunohisto-chemically defined subtype II leiomyosarcoma with poor out-come was consistent with previously published gene expressionprofiling data. The CINSARC signature was significantly asso-ciated with subtype II leiomyosarcoma but not with the othersubtypes (P ¼ 0.0279 by Fisher exact test).

At this time, no targeted therapies exist for leiomyosarcoma.The recognition of molecular subtypes in malignancy has ledto the identification of targets that are unique to a subset of thecases in a variety of tumors that include breast cancer, coloncancer, bladder cancer and glioblastoma, and others (3–12). Inprevious studies, we have identified ROR2, a receptor tyrosinekinase (34) and the presence of intratumoral macrophages(27) as prognostic markers in both uterine and extrauterineleiomyosarcoma. On the basis of our finding, subtype IIleiomyosarcoma, which has high levels of ROR2 expression,would benefit most from drugs targeting this molecule eitheras a single drug or in combination with CD47-based therapy(34, 46, 47).

Here we report that the molecular subtypes in leiomyosar-coma show differential expression of a large number of genesfor which therapeutics either currently exist or are under devel-opment. These subtype specific targets may not show a signif-icant response when leiomyosarcoma cases are considered as asingle homogenous entity in a clinical trial. Recognition ofmolecular subtypes in leiomyosarcoma may highlight prefer-

ential response in that subtype for which the target gene isexpressed at higher levels.

In summary, we used gene expression profiling and immu-nohistochemical assays to demonstrate the existence of 3 dis-tinct molecular subtypes in leiomyosarcoma in two indepen-dent sets of cases and showed that they are associated withdistinct clinical outcomes. These findings indicate distinct bio-logic subclasses in leiomyosarcoma that may respond different-ly to novel therapeutic approaches.

Disclosure of Potential Conflicts of InterestC-H. Lee is a member of the AstraZeneca Ovarian Cancer Advisory Board.

No potential conflicts of interest were disclosed by the other authors.

Authors' ContributionsConception and design: X. Guo, R.B. West, M. van de RijnDevelopment of methodology: X. Guo, S. Zhu, R.B. West, M. van de RijnAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): X. Guo, V.Y. Jo, A.M. Mills, S.X. Zhu, C.-H. Lee,I. Espinosa, M.R. Nucci, S. Anderson, C.D. Fletcher, M. van de RijnAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): X. Guo, T. Hastie, K. Ganjoo, A.H. Beck, R.B. West,C.D. Fletcher, M. van de RijnWriting, review, and/or revision of the manuscript: X. Guo, V.Y. Jo, C.-H. Lee,I. Espinosa, E. Forg�o, K. Ganjoo, A.H. Beck, R.B. West, C.D. Fletcher, M. van deRijnAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): X. Guo, V.Y. Jo, S.X. Zhu, S. Varma, M. van deRijnStudy supervision: X. Guo, M. van de Rijn

AcknowledgmentsThe authors thank the Stanford Center for Genomics and Personalized

Medicine with help on Illumina sequencing and Alayne Brunner, PatriciaMarian Robinson, Joanna Przybyl, and other labmates for sample collection,helpful discussions and experimental design.

Grant SupportThis study was supported by the NIH (grant no. CA 112270; to M. van de

Rijn) and the Fund for Health Research (FIS PI12/01645; to I. Espinosa).The costs of publication of this article were defrayed in part by the payment

of page charges. This article must therefore be hereby marked advertisementin accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received December 17, 2014; revised April 7, 2015; accepted April 11, 2015;published OnlineFirst April 20, 2015.

References1. Gustafson P, Willen H, Baldetorp B, Ferno M, Akerman M, Rydholm A.

Soft tissue leiomyosarcoma. A population-based epidemiologic andprognostic study of 48 patients, including cellular DNA content. Cancer1992;70:114–9.

2. Toro JR, Travis LB,WuHJ, ZhuK, FletcherCD,Devesa SS. Incidence patternsof soft tissue sarcomas, regardless of primary site, in the surveillance,epidemiology and end results program, 1978–2001: An analysis of26,758 cases. Int J Cancer 2006;119:2922–30.

3. Witt H, Mack SC, RyzhovaM, Bender S, Sill M, Isserlin R, et al. Delineationof two clinically and molecularly distinct subgroups of posterior fossaependymoma. Cancer Cell 2011;20:143–57.

4. Marisa L, de Reynies A, Duval A, Selves J, Gaub MP, Vescovo L, et al.Gene expression classification of colon cancer into molecular subtypes:characterization, validation, and prognostic value. PLoS Med 2013;10:e1001453.

5. Lapointe J, LiC,Higgins JP, vandeRijnM,Bair E,MontgomeryK, et al.Geneexpression profiling identifies clinically relevant subtypes of prostatecancer. Proc Natl Acad Sci U S A 2004;101:811–6.

6. Shai R, Shi T, Kremen TJ, Horvath S, Liau LM, Cloughesy TF, et al. Geneexpression profiling identifies molecular subtypes of gliomas. Oncogene2003;22:4918–23.

7. Choi W, Porten S, Kim S, Willis D, Plimack ER, Hoffman-Censits J, et al.Identification of distinct Basal and luminal subtypes of muscle-invasivebladder cancer with different sensitivities to frontline chemotherapy.Cancer Cell 2014;25:152–65.

8. Verhaak RG, Hoadley KA, Purdom E, Wang V, Qi Y, Wilkerson MD, et al.Integrated genomic analysis identifies clinically relevant subtypes of glio-blastoma characterized by abnormalities in PDGFRA, IDH1, EGFR, andNF1. Cancer Cell 2010;17:98–110.

9. Damrauer JS, Hoadley KA, Chism DD, Fan C, Tiganelli CJ, Wobker SE,et al. Intrinsic subtypes of high-grade bladder cancer reflect the hall-marks of breast cancer biology. Proc Natl Acad Sci U S A 2014;111:3110–5.

10. Sorlie T, Tibshirani R, Parker J, Hastie T,Marron JS, Nobel A, et al. Repeatedobservation of breast tumor subtypes in independent gene expression datasets. Proc Natl Acad Sci U S A 2003;100:8418–23.

Guo et al.





11. Sorlie T, Perou CM, Tibshirani R, Aas T, Geisler S, Johnsen H, et al.Gene expression patterns of breast carcinomas distinguish tumorsubclasses with clinical implications. Proc Natl Acad Sci U S A 2001;98:10869–74.

12. The Cancer Genome Atlas Research Network. Integrated genomic analysesof ovarian carcinoma. Nature 2011;474:609–15.

13. Beck AH, Lee CH, Witten DM, Gleason BC, Edris B, Espinosa I, et al.Discovery of molecular subtypes in leiomyosarcoma through integrativemolecular profiling. Oncogene 2010;29:845–54.

14. Beck AH, Weng Z, Witten DM, Zhu S, Foley JW, Lacroute P, et al. 30-endsequencing for expression quantification (3SEQ) from archival tumorsamples. PLoS ONE 2010;5:e8768.

15. Brunner AL, Beck AH, Edris B, Sweeney RT, Zhu SX, Li R, et al.Transcriptional profiling of lncRNAs and novel transcribed regionsacross a diverse panel of archived human cancers. Genome Biol 2012;13:R75.

16. LeeCH,OuWB,Marino-EnriquezA, ZhuM,MayedaM,WangY, et al. 14-3-3 fusion oncogenes in high-grade endometrial stromal sarcoma. Proc NatlAcad Sci U S A 2012;109:929–34.

17. Brunner AL, Li J, Guo X, Sweeney RT, Varma S, Zhu SX, et al. A sharedtranscriptional program in early breast neoplasias despite genetic andclinical distinctions. Genome Biol 2014;15:R71.

18. Demicco EG, BolandGM, Brewer SavannahKJ, Lusby K, Young ED, IngramD, et al. Progressive loss of myogenic differentiation in leiomyosarcomahas prognostic value. Histopathology 2015;66:627–38.

19. Guo X, Zhu SX, Brunner AL, van de Rijn M, West RB. Next generationsequencing-based expression profiling identifies signatures from benignstromal proliferations that define stromal components of breast cancer.Breast Cancer Res 2013;15:R117.

20. Lee CH, Ali RH, Rouzbahman M, Marino-Enriquez A, Zhu M, Guo X,et al. Cyclin D1 as a diagnostic immunomarker for endometrial stromalsarcoma with YWHAE-FAM22 rearrangement. Am J Surg Pathol 2012;36:1562–70.

21. Wilkerson MD, Hayes DN. ConsensusClusterPlus: a class discovery toolwith confidence assessments and item tracking. Bioinformatics 2010;26:1572–3.

22. Peter RJ. Silhouettes: A graphical aid to the interpretation and validation ofcluster analysis. J Comput Appl Math 1987;20:5365.

23. Hoshida Y, Brunet JP, Tamayo P, Golub TR, Mesirov JP. Subclassmapping:identifying common subtypes in independent disease data sets. PLoSONE2007;2:e1195.

24. Li J, Tibshirani R. Finding consistent patterns: A nonparametric approachfor identifying differential expression in RNA-Seq data. Stat Methods MedRes 2013;22:519–36.

25. Huang DW, Sherman BT, Lempicki RA. Bioinformatics enrichment tools:paths toward the comprehensive functional analysis of large gene lists.Nucleic Acids Res 2009;37:1–13.

26. Huang DW, Sherman BT, Lempicki RA. Systematic and integrative analysisof large gene lists using DAVID bioinformatics resources. Nat Protoc2009;4:44–57.

27. Espinosa I, Beck AH, Lee CH, Zhu S, Montgomery KD, Marinelli RJ, et al.Coordinate expression of colony-stimulating factor-1 and colony-stimu-lating factor-1-related proteins is associated with poor prognosis in gyne-cological and nongynecological leiomyosarcoma. Am J Pathol 2009;174:2347–56.

28. Chibon F, Lagarde P, Salas S, Perot G, Brouste V, Tirode F, et al. Validatedprediction of clinical outcome in sarcomas andmultiple types of cancer onthe basis of a gene expression signature related to genome complexity. NatMed 2010;16:781–7.

29. Chen JL, Espinosa I, Lin AY, Liao OY, van de Rijn M, West RB. Stromalresponses among common carcinomas correlated with clinicopathologicfeatures. Clin Cancer Res 2013;19:5127–35.

30. Beck AH, Espinosa I, Edris B, Li R, Montgomery K, Zhu S, et al. Themacrophage colony-stimulating factor 1 response signature in breastcarcinoma. Clin Cancer Res 2009;15:778–87.

31. de HoonMJ, Imoto S, Nolan J, Miyano S. Open source clustering software.Bioinformatics 2004;20:1453–4.

32. Liu CL, Prapong W, Natkunam Y, Alizadeh A, Montgomery K, Gilks CB,et al. Software tools for high-throughput analysis and archiving of immu-nohistochemistry staining data obtained with tissue microarrays. Am JPathol 2002;161:1557–65.

33. Mills AM, Beck AH, Montgomery KD, Zhu SX, Espinosa I, Lee CH, et al.Expression of subtype-specific group 1 leiomyosarcoma markers in a widevariety of sarcomas by gene expression analysis and immunohistochem-istry. Am J Surg Pathol 2011;35:583–9.

34. Edris B, Espinosa I,Muhlenberg T,Mikels A, LeeCH, Steigen SE, et al. ROR2is a novel prognostic biomarker and a potential therapeutic target inleiomyosarcoma and gastrointestinal stromal tumour. J Pathol 2012;227:223–33.

35. Van Allen EM,Wagle N, Stojanov P, Perrin DL, Cibulskis K, Marlow S, et al.Whole-exome sequencing and clinical interpretation of formalin-fixed,paraffin-embedded tumor samples to guide precision cancer medicine.Nat Med 2014;20:682–8.

36. Rebagay G, Yan S, Liu C, Cheung NK. ROR1 and ROR2 in humanmalignancies: potentials for targeted therapy. Front Oncol 2012;2:34.

37. Morioka K, Tanikawa C, Ochi K, Daigo Y, Katagiri T, Kawano H, et al.Orphan receptor tyrosine kinase ROR2 as a potential therapeutic target forosteosarcoma. Cancer Sci 2009;100:1227–33.

38. Wright TM, Brannon AR, Gordan JD, Mikels AJ, Mitchell C, Chen S, et al.Ror2, a developmentally regulated kinase, promotes tumor growth poten-tial in renal cell carcinoma. Oncogene 2009;28:2513–23.

39. Italiano A, Lagarde P, Brulard C, Terrier P, Lae M, Marques B, et al. Geneticprofiling identifies two classes of soft-tissue leiomyosarcomaswith distinctclinical characteristics. Clin Cancer Res 2013;19:1190–6.

40. Nanda V,Miano JM. Leiomodin 1, a new serum response factor-dependenttarget gene expressed preferentially in differentiated smooth muscle cells.J Biol Chem 2012;287:2459–67.

41. Francis P, Namlos HM, Muller C, Eden P, Fernebro J, Berner JM, et al.Diagnostic and prognostic gene expression signatures in 177 soft tissuesarcomas: hypoxia-induced transcription profile signifiesmetastatic poten-tial. BMC Genomics 2007;8:73.

42. Villacis RA, Silveira SM, Barros-Filho MC, Marchi FA, Domingues MA,Scapulatempo-NetoC, et al. Gene expressionprofiling in leiomyosarcomasand undifferentiated pleomorphic sarcomas: SRC as a new diagnosticmarker. PLoS ONE 2014;9:e102281.

43. Carneiro A, Francis P, Bendahl PO, Fernebro J, AkermanM, Fletcher C, et al.Indistinguishable genomic profiles and shared prognostic markers inundifferentiated pleomorphic sarcoma and leiomyosarcoma: differentsides of a single coin? Lab Invest 2009;89:668–75.

44. Lee YF, John M, Falconer A, Edwards S, Clark J, Flohr P, et al. A geneexpression signature associated with metastatic outcome in human leio-myosarcomas. Cancer Res 2004;64:7201–4.

45. Perot G, Mendiboure J, Brouste V, Velasco V, Terrier P, Bonvalot S, et al.Smooth muscle differentiation identifies two classes of poorly differenti-ated pleomorphic sarcomas with distinct outcome. Mod Pathol 2014;27:840–50.

46. Edris B, Weiskopf K, Volkmer AK, Volkmer JP, Willingham SB, Contreras-Trujillo H, et al. Antibody therapy targeting the CD47 protein is effective inamodel of aggressivemetastatic leiomyosarcoma. Proc Natl Acad Sci U S A2012;109:6656–61.

47. Willingham SB, Volkmer JP, Gentles AJ, SahooD,Dalerba P,Mitra SS, et al.The CD47-signal regulatory protein alpha (SIRPa) interaction is a thera-peutic target for human solid tumors. Proc Natl Acad Sci U S A 2012;109:6662–7.






2015;21:3501-3511. Published OnlineFirst April 20, 2015.Clin Cancer Res Xiangqian Guo, Vickie Y. Jo, Anne M. Mills, et al. Clinically Relevant Molecular Subtypes in Leiomyosarcoma

Updated version

10.1158/1078-0432.CCR-14-3141doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2015/04/21/1078-0432.CCR-14-3141.DC1

Access the most recent supplemental material at:

Cited articles

http://clincancerres.aacrjournals.org/content/21/15/3501.full#ref-list-1

This article cites 47 articles, 12 of which you can access for free at:

Citing articles

http://clincancerres.aacrjournals.org/content/21/15/3501.full#related-urls

This article has been cited by 11 HighWire-hosted articles. Access the articles at:

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

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://clincancerres.aacrjournals.org/content/21/15/3501To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-14-3141

http://clincancerres.aacrjournals.org/content/suppl/2015/04/21/1078-0432.CCR-14-3141.DC1

http://clincancerres.aacrjournals.org/content/21/15/3501.full#ref-list-1

http://clincancerres.aacrjournals.org/content/21/15/3501.full#related-urls

http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://clincancerres.aacrjournals.org/content/21/15/3501


clinically relevant molecular subtypes in leiomyosarcoma · biology of human tumors clinically...

Documents