small rna sequencing and functional characterization ... · for retroperitoneal disease, the most...

Integrated Systems and Technologies

Small RNA Sequencing and Functional CharacterizationReveals MicroRNA-143 Tumor Suppressor Activity inLiposarcoma

Stacy Ugras1, Elliott Brill1, Anders Jacobsen2, Markus Hafner5, Nicholas D. Socci2, Penelope L. DeCarolis1,Raya Khanin2, Rachael O’Connor1, Aleksandra Mihailovic5, Barry S. Taylor2, Robert Sheridan2,Jeffrey M. Gimble6, Agnes Viale4, Aimee Crago1, Cristina R. Antonescu3,Chris Sander2, Thomas Tuschl5, and Samuel Singer1

AbstractLiposarcoma remains the most common mesenchymal cancer, with a mortality rate of 60% among patients

with this disease. To address the present lack of therapeutic options, we embarked upon a study of microRNA(miRNA) expression alterations associated with liposarcomagenesis with the goal of exploiting differentiallyexpressed miRNAs and the gene products they regulate as potential therapeutic targets. MicroRNA expressionwas profiled in samples of normal adipose tissue, well-differentiated liposarcoma, and dedifferentiatedliposarcoma by both deep sequencing of small RNA libraries and hybridization-based Agilent microarrays.The expression profiles discriminated liposarcoma from normal adipose tissue and well differentiated fromdedifferentiated disease. We defined over 40 miRNAs that were dysregulated in dedifferentiated liposarcomas inboth the sequencing and the microarray analysis. The upregulated miRNAs included two cancer-associatedspecies (miR-21 and miR-26a), and the downregulated miRNAs included two species that were highly abundantin adipose tissue (miR-143 and miR-145). Restoring miR-143 expression in dedifferentiated liposarcoma cellsinhibited proliferation, induced apoptosis, and decreased expression of BCL2, topoisomerase 2A, proteinregulator of cytokinesis 1 (PRC1), and polo-like kinase 1 (PLK1). The downregulation of PRC1 and its dockingpartner PLK1 suggests that miR-143 inhibits cytokinesis in these cells. In support of this idea, treatment with aPLK1 inhibitor potently induced G2–M growth arrest and apoptosis in liposarcoma cells. Taken together, ourfindings suggest that miR-143 re-expression vectors or selective agents directed at miR-143 or its targets mayhave therapeutic value in dedifferentiated liposarcoma. Cancer Res; 71(17); 5659–69. �2011 AACR.

Introduction

Liposarcoma, the most common soft tissue sarcoma,accounts for 20% of adult sarcoma cases (1). It is classifiedinto 5 subtypes constituting 3 biological groups, the mostcommon of which consists of well-differentiated liposarcomas

(WDLS) and dedifferentiated liposarcomas (DDLS). WDLS/DDLS is characterized by chromosome 12q amplification inapproximately 90% of cases (2). DDLSs are thought to arisefrom WDLS because the dedifferentiated component of thesetumors always coexists with an adjacent region of WDLS andbecause WDLSs often progress and recur as DDLS. Surgicalresection is the primary treatment for WDLS/DDLS, but localrecurrence is common and more than 60% of patients even-tually die from these tumors. For retroperitoneal disease, themost common anatomic location, DDLS has higher rates oflocal and distant recurrence and a 6-fold higher mortality thanWDLS (3). WDLS/DDLS is largely resistant to conventionalchemotherapy and radiotherapy, so there is a pressing need todevelop new targeted therapies for patients with recurrentdisease.

The search for oncogenes and tumor suppressor genes hasrecently expanded to include small RNAs, including miRNAs.These are RNAs of 20 to 24 nucleotides that bind to the 30

untranslated region (UTR) of target mRNAs to inhibit transla-tion and decrease mRNA stability (4–6). An individual miRNAmay regulate hundreds of transcripts directly or indirectly(7), affording each miRNA extensive control over cellularfunctions. MicroRNA (miRNA) expression is dysregulated in

Authors' Affiliations: 1Department of Surgery, Sarcoma Biology Labora-tory, Sarcoma Disease Management Program; 2Computational BiologyCenter; 3Department of Pathology; 4Genomics Core Lab, Memorial Sloan-Kettering Cancer Center; 5Laboratory of RNA Molecular Biology, TheRockefeller University, New York, New York; and 6Stem Cell BiologyLaboratory, Pennington Biomedical Research Center, Louisiana StateUniversity System, Baton Rouge, Louisiana

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

Corresponding Authors: Thomas Tuschl, Laboratory of RNA MolecularBiology, The Rockefeller University, New York, NY 10065. Phone: 212-327-7651; Fax: 212-327-7652; E-mail: [email protected] or SamuelSinger, Gastric and Mixed Tumor Service, Department of Surgery, HowardBuilding, H1205, Memorial Sloan-Kettering Cancer Center, 1275 YorkAvenue, New York, NY 10021. Phone: 212-639-2940; Fax: 646-422-2300; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-11-0890

�2011 American Association for Cancer Research.

CancerResearch

www.aacrjournals.org 5659

on February 2, 2020. © 2011 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst June 21, 2011; DOI: 10.1158/0008-5472.CAN-11-0890

http://cancerres.aacrjournals.org/

multiple human cancers, including lung, breast, prostate, andgastrointestinal cancers (8, 9). Some of the dysregulatedmiRNAs seem to be tumor suppressors, and therefore miRNAre-expression vectors could be considered as a therapeutic fortumors that underexpress the miRNA. Other miRNAs seem tobe oncogenes (10), so anti-miRNAs (11) could be considered asa therapeutic for tumors that overexpress the miRNA. There-fore, we sought to identify dysregulated miRNAs and theirtarget genes in liposarcoma. MicroRNA expression in WDLS,DDLS, and normal adipose tissue was analyzed by small RNAsequencing and hybridization-based Agilent microarrays. Wethen characterized the role in liposarcoma for a stronglydownregulated miRNA, miR-143, through phenotypic analysisof liposarcoma cells re-expressing miR-143, profiling of genesregulated by miR-143, and pharmacologic inhibition of one ofthe pathways implicated.

Materials and Methods

Sequence data analysis, quantitative real-time reverse tran-scription PCR, in situ hybridization, proliferation assays, fluor-escence-activated cell-sorting (FACS) analysis of DNA contentusing 40,6-diamidino-2-phenylindole, immunohistochemicalstaining, immunoblots, luciferase reporter assays, apoptosisanalysis, and cell-cycle analysis with propidium iodide aredescribed in Supplementary Data.

Patient sample acquisitionInformed consent was obtained and the project was

approved by the MSKCC Institutional Review Board. Patientcharacteristics are shown in Supplementary Table S1. Tumorand normal adipose tissue samples obtained during surgicalresection were snap-frozen in liquid nitrogen and embeddedin cryomolds.

Cell cultureLiposarcoma cell lines were established from tissue samples

obtained from consenting patients: LPS141 and DDLS8817from DDLS samples and WD0082 from a WDLS sample.Comparative genomic hybridization array studies confirmedthat these cell lines contain the 12q amplification. Primaryhuman adipose tissue-derived stromal/stem cells (ASC) wereisolated from subcutaneous fat tissue samples from consent-ing patients as described (12). The cell lines were maintainedin Dulbecco's modified Eagle's medium HG:F12 supplementedwith 10% heat-inactivated fetal bovine serum and 1% peni-cillin/streptomycin and kept at 37�C in 5% carbon dioxide.Media promoting differentiation included the above plus 100nmol/L insulin, 1 mmol/L dexamethasone, 250 mmol/L 3-isobutyl-1-methylxanthine, 33 mmol/L biotin, 77 mmol/L pan-tothenic acid, and 5 mmol/L rosiglitazone. All cultures wereMycoplasma free.

RNA isolationCryomolds (0.5 � 1 � 1 cm) were macrodissected under

the supervision of a pathologist as described (13) to ensuresubtype uniformity and to eliminate necrotic/normal tissue.Tumor samples were then lysed with QIAzol lysis reagent

and homogenized using Mixer Mill MM 300 (Retsch). Cellline samples were trypsinized, washed once in phosphate-buffered saline (PBS), then lysed in QIAzol. For all samples,total RNA was purified using standard phenol-chloroformextraction and ethanol precipitation. Cell line miRNA usedin quantitative PCR was purified with the miRNeasy Mini Kit(Qiagen).

Microarray analysisThe Agilent HumanmiRNAMicroarray was used to compile

gene expression profiles of 34 DDLS, 32 WDLS, and 17 normalfat samples. The miRNA array images were quantitated usingAgilent's Feature Extraction program, which provides inte-grated gene-level signals for each miRNA. The gene-levelvalues were then transformed with the generalized log func-tion (to handle negative values) and normalized using the VSNpackage from Bioconductor.

Illumina arrays were used to generate triplicate geneexpression profiles in LPS141 and DDLS8817 cell lines at2.5 and 3.5 days following infection with miR-143 or scramblelentiviruses, and in untreated controls. The expression profileswere processed using the LUMI package (Illumina microarraydata analysis, Bioconductor) with the default options: back-ground subtraction, normalization with the quantilemethod, and log2 transformation of expression values. Datawere normalized using the variance-stabilization method(lumiExpresso).

For both Agilent and Illumina data, LIMMA (linear modelsfor microarray data; Bioconductor) was used to computedifferential expression, and the false discovery rate (FDR)method was used for multiple testing corrections.

Solexa sequencingSmall RNA cDNA libraries were prepared from 22 DDLS, 22

WDLS, and 11 normal fat tissue samples as described (14). In20-mL reactions, 2 mg total RNA was ligated to 100 pmoladenylated 30 adapter containing a unique pentamer barcodeat the 50 end (Supplementary Table S2) using 1 mg Rnl2(1-249)K227Q [purified from Escherichia coli containing pET16b-Rnl2(1-249)K227Q (Addgene)] in 50 mmol/L Tris-HCl, pH 7.6; 10mmol/L MgCl2; 10 mmol/L 2-mercaptoethanol; 0.1 mg/mLacetylated bovine serum albumin (Sigma-Aldrich); and 15%dimethyl sulfoxide for 16 hours on ice. After ligation, up to 20samples bearing unique barcodes were pooled and purified ona 15% denaturing polyacrylamide gel. RNAs of 45 and 50nucleotides were excised from the gel, eluted, and ligatedto 100 pmol 50 oligoribonucleotide adapter (guucagaguucua-caguccgacgauc) as described above for the 30 adaptors, exceptthat reactions contained 0.2 mmol/L ATP and RNL1 instead ofRNL2(1-249)K227Q and were incubated for 1 hour at 37�C.Ligated small RNAs were purified on a 12% polyacrylamide gel,reverse transcribed using SuperScript III Reverse Transcrip-tase (Invitrogen), and amplified by PCR. The forward primerwas AATGATACGGCGACCACCGACAGGTTCAGAGTTCTA-CAGTCCGA; reverse transcription and reverse primer wasCAAGCAGAAGACGGCATACGA. On average 200,000 (range9,000–680,000) sequence reads of miRNAs were obtained persample.

Ugras et al.

Cancer Res; 71(17) September 1, 2011 Cancer Research5660




In situ hybridizationDDLS cells and ASCs were grown on chamber slides and

treated as described in Supplementary Methods (15).

miRNA re-expressionRe-expression was based on the miRexpress Lentiviral

microRNA system (Open Biosystems) The vectors consistedof pLKO.1 with inserts encoding precursors for miR-143, miR-145, or Scr, a scramble control sequence not targeting anyknown human genes. These constructs were produced bytransient cotransfection of 293T cells (American Type CultureCollection) with 10 mg of pLKO.1-derived plasmid, 9 mg ofpackaging plasmid psPAX2, and 1 mg of envelope plasmidpMD2.G. Cells (in 10-cm2 Petri dishes) were transfected usingLipofectamine 2000 (Invitrogen). Infectious viral supernatantswere collected at 48, 72, and 96 hours after transfection,pooled, and concentrated by centrifugation using an AmiconUltra-15 100K cutoff filter device (Millipore). LPS141 andDDLS8817 cells were then infected with lentivirus containingmiR-143, miR-145, or Scr sequence. Puromycin (Sigma) wasused to select infected cells.

Proliferation assaysProliferation of LPS141 and DDLS8817 cells was evaluated

in triplicate samples by estimation of DNA content using theCyQuant Cell Proliferation Kit (Molecular Probes) as detailedin Supplementary Methods. The results per condition werenormalized to the earliest time point (day 2 for lentiviralinfection, day 0 for drug treatment).

Analysis of gene expression changes after miR-143re-expressionTo analyze correlations between motifs in mRNA 30 UTRs

and expression changes upon miR-143 re-expression, wesubjected Illumina microarray results (for miR-143 vs.untreated and for miR-143 vs. Scr in DDLS8817 and LPS141cells at 2.5 and 3.5 days postinfection) to 2 different methods, alinear-regression–based analysis, miReduce (16), and a non-parametric motif correlation analysis (17). We identified genesconsistently upregulated or downregulated at 2.5 days aftermiR-143 re-expression by requiring significant differentialexpression (P < 0.05, unadjusted LIMMA moderated t test;ref. 18) in at least 3 out of the 4 comparisons (miR-143 re-expression in LPS141 and DDLS8817 each compared withuntreated and Scr controls). The combined set of dysregulatedgenes was analyzed for enrichment of known molecularinteractions using NetBox, which uses known pathway andprotein–protein interactions extracted from Pathway Com-mons (19) to identify connected genes in the input list (19).NetBox settings were shortest-path threshold ¼ 2, p-valuecutoff¼ 0.05. Statistical significance of the size of the resultinggene module and degree of connection within it was tested bycomparison with 1,000 randomly selected sets of 268 genes.

Measurement of DNA replicationDNA replication was quantified using incorporation of

bromodeoxyuridine (BrdU). Briefly, cultured cells were incu-bated with fresh medium containing 10 mmol/L BrdU (Sigma)

for 1 hour. Cells were washed with PBS and fixed in 70%ethanol at–20�C. Following fixation, cells were rinsed with PBSand DNA was denatured with hydrochloric acid and Triton X-100. Cells were then resuspended in 0.1 mol/L sodium borateand were stained with fluorescein isothiocyanate-conjugatedmouse anti-BrdU antibody (BD Pharmingen) for 1 hour atroom temperature. A 488-nm laser was used for excitation andfluorescence was measured (FACSCalibur, Becton Dickinson).Percentage of BrdU incorporation was measured usingFLOWJO Flow Cytometry Analysis Software.

Results

MicroRNA profiles discriminate adipose tissue, WDLS,and DDLS

We profiled miRNA expression in 83 samples of WDLS,DDLS, and normal adipose tissue (hereafter normal fat) usingAgilent microarrays. In 52 of these samples and in 1 additionalWDLS sample and 2 additional normal fat samples, we alsoprofiled miRNA expression using deep sequencing of smallRNA libraries. To compare the 2 profiling platforms, we firstfound those miRNAs that were significantly differentiallyexpressed in DDLS compared with normal fat in the sequen-cing analysis. Among those miRNAs, fold changes in sequencecount were highly correlated with fold changes in microarrayexpression value (Supplementary Fig. S1), indicating thatstatistically significant ratios are consistent in the 2 technol-ogies.

Unsupervised consensus clustering of the miRNA profilesfrom deep sequencing revealed 3 distinct clusters (Fig. 1A).These clusters corresponded exactly to DDLS, WDLS, andnormal fat, except for one WDLS sample that clustered withthe DDLS samples. This WDLS sample was isolated from alocally recurrent retroperitoneal tumor that, within a year ofcomplete resection, subsequently recurred as a DDLS. ThesemiRNA profiles therefore discriminate liposarcomas fromnormal fat tissues, and WDLS from DDLS.

The miRNAs that were strongly dysregulated in WDLS andDDLS are listed in Tables 1 and 2. A number of them hadrelatively high expression levels (in terms of the percentage ofmicroRNA sequence reads) in either normal fat or in DDLScells, suggestive of functional significance.

miR-143 and miR-145 are underexpressed in WDLS andDDLS

The miRNA most frequently sequenced in normal fatlibraries was miR-143, with 7.75% of the total miRNA reads(Tables 1 and 2). This miRNA was strongly downregulated inliposarcomas compared with normal fat (3.3- and 7.9-foldchange in WDLS and DDLS, respectively; Fig. 1B). miR-145,which is transcribed from the same bicistronic primary tran-script (20), followed a similar pattern of downregulation: 2.8-fold inWDLS (though this was not statistically significant) and6.6-fold in DDLS.

In addition to their dysregulation in human tumors, bothmiR-143 andmiR-145 were downregulated in DDLS andWDLScell lines relative to ASCs. Both quantitative PCR (Fig. 1C) anddeep sequencing (Fig. 1D) showed downregulation in the two

MicroRNA-143 as a Tumor Suppressor in Liposarcoma

www.aacrjournals.org Cancer Res; 71(17) September 1, 2011 5661




DDLS cell lines (DDLS8817 and LPS141) and the WDLS line(WD0082). In most cases, expression of these miRNAs in theliposarcoma cell lines was less than 5% of the level in ASCs.Furthermore, in situ hybridization showed substantiallyweaker perinuclear miR-143 staining in DDLS cells than inASCs (Fig. 1E).

Re-expression of miR-143 inhibits proliferation andinduces apoptosis

We next re-expressed miR-143 and miR-145 in the 2 DDLScell lines using a lentiviral expression system. QuantitativePCR showed increased expression in cells re-expressing eachmiRNA compared with cells expressing Scr control (all P <0.001; Fig. 2A). By 2 days after lentivirus infection, levels of

miR-143 and miR-145 were similar to or somewhat higherthan the levels in ASCs. In an analysis of cell proliferation, re-expressing miR-143 was antiproliferative relative to control byday 6 in DDLS cells (P < 0.001). Re-expressing miR-145,however, had no effect on proliferation (Fig. 2B). To assessapoptosis, we were unable to use propidium iodide stainingbecause our lentivirus-infected cells express turbo red fluor-escent protein. Instead, we used FACS to analyze DNA con-tent, and found that re-expression of miR-143 induced a 2.9-fold increase in the frequency of cells with subG1 DNAcontent, whereas miR-145 had little effect (SupplementaryFig. S2). We also measured apoptosis using immunohisto-chemical staining of cleaved caspase-3 and immunoblot ana-lysis of cleaved caspase-3 and cleaved PARP. By these assays as

Figure 1.MicroRNA expression analysis of WDLS and DDLS. A, deep sequencing analysis of miRNA expression inWDLS, DDLS, and normal fat samples. Theheat map was generated by unsupervised clustering of miRNA levels. Darker shading represents more frequent clustering among matched pairs. B,dysregulation of miR-143 and miR-145. The plots show normalized miRNA read frequencies from the deep sequencing analysis. Boxes show the median andinterquartile range (IQR) and whiskers showing the range of data within 1.5 IQR units of the first and third quartiles. The tables show expressionratios and statistical significance (FDR). C and D, expression of miR-143 and miR-145 in DDLS cells lines (DDLS8817 and LPS141) and a WDLS cell line(WD0082) compared with ASCs. C, expression levels from quantitative PCR, normalized to the expression level in undifferentiated ASCs. D, expressionlevels from deep sequencing, including levels in ASCs after 0 to 10 days of culture in differentiation media. E, in situ hybridization for miR-143 inLPS141 and ASCs. Perinuclear miR-143 staining (red) was greater in ASCs than in LPS141. The negative controls were probes specific for miR-124,which is known to be expressed in neural tissue and not adipose tissue, and no probe (blank).

Ugras et al.





well, re-expression of miR-143, but not miR-145, inducedapoptosis (Fig. 2C and D; Supplementary Fig. S3). Re-expres-sion of miR-143 resulted in a 26-fold or more increase in thepercentage of liposarcoma cells with caspase-3 cleavage (P <0.001) and an accompanying increase in PARP cleavage.

Re-expression of miR-143 decreases entry into S phaseand inhibits mitosisTo assess the effects of miR-143 re-expression on the cell

cycle, we assessed BrdU incorporation and phosphorylatedhistone H3 immunohistochemistry. BrdU incorporation wassignificantly decreased with miR-143 re-expression relative toScr and miR-145 re-expression, indicating decreased entryinto S phase (Fig. 2E). Phosphorylated histone H3 was alsosignificantly decreased with miR-143 re-expression relative toScr and miR-145 (Fig. 2F, Supplementary Fig. S4), signifyingdecreased mitosis.

miR-143 downregulation has no clear role in blockingadipocyte differentiationTo determine the mechanism for the proliferative and

apoptotic effects of miR-143, we first examined whethermiR-143 downregulation is involved in the block of adipogen-esis in DDLS cells. Recently, miR-143 levels were found to beincreased in differentiating adipocytes, and antisense inhibi-tion of miR-143 in preadipocytes blocked differentiation (21).Similarly, we found that miR-143 expression increased 2-fold

during adipocyte differentiation (Fisher test, P < 1 � 10–15;Fig. 1D). Nevertheless, miR-143 re-expression in DDLS cellsgrown in media promoting differentiation did not cause anyincrease in the expression of adipocyte differentiationmarkers(CEBPA, FABP4, and PPARG; Supplementary Fig. S5).

miR-143 regulates a gene module involved in apoptosis,DNA replication, and cytokinesis

We continued our investigation into how miR-143 affectsproliferation and apoptosis with a systematic exploration ofmRNA expression changes driven by miR-143 in both DDLScell lines. Microarray analysis showed 437 genes significantlydifferentially expressed (P < 0.05, Limma moderated t test) inthe DDLS8817 cell line and 819 genes in the LPS141 cell lineafter miR-143 re-expression (measured relative to both Scrand untreated controls). Of these, 268 genes were significantlydifferentially expressed in both cell lines (125 downregulatedand 143 upregulated genes).

In both DDLS cell lines, the expression changes after miR-143 expression correlated more strongly with the miR-143seed sequence than with any other 30UTR sequence motif(see Materials and Methods), validating miR-1430s effect inour experiments. Similarly, in DDLS8817 cells re-expressingmiR-143 versus Scr, mRNAs with miR-143 target sites (pre-dictions from TargetScan) in their 30UTRs were significantlydownregulated compared with mRNAs lacking seeds at both2.5 days (P < 0.001, Wilcoxon 1-sided test) and 3.5 days

Table 1. Differentially expressed microRNAs in WDLS compared with normal fat tissue samples (FDR <0.05, fold change �3)

MicroRNA Mean clone count Frequency of cloning Fold change(WDLS versusnormal fat)

FDR

WDLS Normal fat WDLS (%) Normal fat (%)

MicroRNAs overexpressedin tumor

miR-26a 34,639 9,416 13.36 3.63 3.7 3.3E-04miR-199a-3pa 8,672 2,316 3.34 0.89 3.7 2.1E-05miR-199b-3p 3,870 1,225 1.49 0.47 3.2 6.3E-06miR-199a-5pa 3,264 878 1.26 0.34 3.7 5.6E-06miR-127-3p 738 205 0.28 0.08 3.6 2.3E-04miR-376a 592 155 0.23 0.06 3.8 7.0E-05miR-34a 540 176 0.21 0.07 3.1 5.0E-06MicroRNAs underexpressed in tumormiR-143 6,098 20,102 2.35 7.75 –3.3 1.5E-05miR-451 1,623 14,550 0.63 5.61 –9.0 4.6E-06miR-486-5p 49 458 0.02 0.18 –9.3 4.1E-06miR-144 46 411 0.02 0.16 –8.9 1.3E-04miR-195 25 404 0.01 0.16 –16.0 8.6E-07

NOTE: MicroRNAs are listed in order of mean clone count (in DDLS for overexpressed species; in normal fat for underexpressedspecies).amiR-199a-3p andmiR-199a-5p are both encoded by 2 genes, miR-199a-1 andmiR-199a-2. miR-199a-2 is in a cistronwithmiR-214,which was not upregulated in WDLS (or in DDLS). Therefore, most of the upregulation likely originates from miR-199a-1, which ismonocistronic.






(P < 10–8). In LPS141 cells, miR-143 mRNA targets showed asimilar trend of downregulation (2.5 days: P ¼ 0.09; 3.5 days:P < 1.0 � 10–5).

To explore functional programs that may be controlled bymiR-143 in these cells, we evaluated the 268 differentiallyexpressed genes for enrichment of molecular interactionsusing a computational network-based approach (NetBox) that

takes into account known signaling pathways and protein–protein interactions. This analysis revealed a module of 24connected genes (Fig. 3), 22 of which were among the differ-entially expressed genes. The number of affected genes andinteractions was significantly higher than expected by chance(P¼ 0.04). This result suggests that the phenotypic changes ofmiR-143 loss may be driven through the pathway and the

Table 2. Differentially expressed microRNAs in DDLS compared with normal fat tissue samples (FDR <0.05, fold change �3)

MicroRNA Mean clone count Frequency of cloning Fold change(DDLS versusnormal fat)

FDR

DDLS Normal fat DDLS (%) Normal fat (%)

MicroRNAs overexpressedin tumor

miR-26a 29,703 9,416 11.45 3.63 3.2 4.8E-04miR-21 28,956 7,002 11.17 2.70 4.1 1.3E-05miR-199a-3p 13,787 2,316 5.32 0.89 6.0 8.6E-09miR-199a-5p 6,754 878 2.60 0.34 7.7 1.5E-11miR-199b-3p 6,655 1,225 2.57 0.47 5.4 1.2E-10miR-199b-5p 2,925 501 1.13 0.19 5.8 3.8E-09miR-424 1,835 263 0.71 0.10 7.0 2.5E-07miR-376c 1,160 285 0.45 0.11 4.1 7.4E-06miR-376a 910 155 0.35 0.06 5.9 6.9E-08miR-127-3p 900 205 0.35 0.08 4.4 5.8E-06miR-34a 639 176 0.25 0.07 3.6 4.1E-08miR-136 406 44 0.16 0.02 9.1 1.5E-08miR-377 384 76 0.15 0.03 5.1 6.7E-06miR-376b 289 25 0.11 0.01 11.5 8.6E-09miR-409-3p 255 46 0.10 0.02 5.6 1.3E-07MicroRNAs underexpressed

in tumormiR-143a 2,544 20,102 0.98 7.75 –7.9 3.9E-12miR-451b 321 14,550 0.12 5.61 –45.4 3.7E-13miR-145a 1,008 6,624 0.39 2.55 –6.6 3.2E-10miR-378 320 6,480 0.12 2.50 –20.3 3.8E-12miR-335 164 1,693 0.06 0.65 –10.4 7.3E-08miR-193a-3p 120 818 0.05 0.32 –6.8 1.6E-06miR-452 148 697 0.06 0.27 –4.7 4.7E-06miR-193b 98 660 0.04 0.25 –6.7 2.9E-09miR-193a-5p 42 527 0.02 0.20 –12.4 6.3E-12miR-497 133 526 0.05 0.20 –4.0 5.6E-03miR-29c 120 476 0.05 0.18 –4.0 1.6E-05miR-486-5p 6 458 0.00 0.18 –72.9 6.1E-15miR-146a 90 423 0.03 0.16 –4.7 5.1E-04miR-223 136 422 0.05 0.16 –3.1 5.5E-03miR-144b 15 411 0.01 0.16 –26.6 8.8E-09miR-195 33 404 0.01 0.16 –12.2 8.1E-07miR-652 32 312 0.01 0.12 –9.9 1.0E-11miR-365 62 271 0.02 0.10 –4.4 1.3E-08miR-190 43 254 0.02 0.10 –5.9 6.3E-10

amiR-143 and miR-145 cocistronic.bmiR-451 and miR-144 are cocistronic.

Ugras et al.





protein–protein interactions implied by this gene set. Thisgene module contained 9 predicted miR-143 targets based onsequence complementarity: CENPM, PRC1, BCL2, PTPN2,HMGA2, TOP2A, CDC2, CDC25B, and TSC1 (SupplementaryTable S3).Importantly, the gene module includes genes involved in

apoptosis (BCL2), DNA replication (TOP2A), and cytokinesis(PRC1, PLK1, CDC25B, ECT2, and CDC2; refs. 22–24). Of theseBCL2, TOP2A, PRC1, CDC25B, and CDC2 are predicted directtargets of miR-143. We used PCR andWestern blots to validatethe downregulation of BCL2, TOP2A, PRC1, and PLK1 in DDLScells re-expressing miR-143 (Fig. 4, Supplementary Fig. S6).Compared with Scr, miR-143 expression reduced mRNA levelsof all 4 genes in both LPS141 and DDLS8817 cell lines (P < 0.001for all except BCL2 in DDLS8817, which was not significant).miR-143 re-expression reduced BCL2, topoisomerase 2A(TOP2A), polo-like kinase 1 (PLK1), and protein regulator ofcytokinesis 1 (PRC1) protein levels in DDLS cells. In addition,luciferase reporter assays showed that miR-143 directly reg-ulates the 30UTRs of TOP2A and PRC1 (Supplementary Fig. S7).

PLK1 inhibition in DDLS cells is antiproliferative andproapoptotic

Because miR-143 regulates key members of the centralspindle regulatory pathway, we tested whether selective inhi-bition of this pathway would have antiproliferative or proa-poptotic effects in DDLS cells. PLK1, a regulator of thispathway, is abundant in DDLS cells relative to ASCs(Fig. 4C). We utilized a pharmacologic inhibitor of PLK1, BI2536 (Selleck Chemicals; ref. 25), which is currently in phase IIclinical trials and has shown limited success in the treatmentof multiple solid tumors and chronic myeloid leukemia (26,27). Treatment with BI 2536 at 20 nmol/L resulted in completegrowth inhibition of LPS141 cells and an 83% reduction in cellnumber of DDLS8817 cells, without affecting proliferation ofASCs (Fig. 5A). A 2-day treatment induced a 3- to 4.5-foldincrease in apoptosis in DDLS8817 and LPS141 cells, respec-tively, compared with untreated controls (P < 0.001), butinduced no change in apoptosis in ASCs (Fig. 5B). Further-more, the 2-day treatment induced G2–M cell-cycle arrest(Fig. 5C).

miR-143 miR‐145 DDLS8817

Day 2 Day 4 Day 6 Day 2 Day 4 Day 6

LPS141

cleaved

Scr miR

-143

Scr

Scr miR

-143

Scr miR

-143

Scr miR

-143

Caspase-3

Tubulin

LPS141

PARP

cleaved

Tubulin

cleaved

Caspase-3

Tubulin Tubulin

PARP

cleaved

DDLS8817

LPS141 DDLS8817

A B

C

D

E

F

Change in c

ell

count

rela

tive to D

ay 2

Expre

ssio

n

rela

tive to A

SC

2

1.5

1

0.5

2.5

2

1.5

1

0.5

miR-143

miR-145

Scr

miR-143

miR-145

Figure 2. Effects of re-expression of miR-143 and miR-145 on proliferation and apoptosis in liposarcoma cell lines. A, quantitative PCR to assess re-expression of miR-143 and miR-145 in LPS141 and DDLS8817 cells at 2.5 days after infection with lentiviral vectors for miR-143, miR-145, or Scr,and in untreated cells (Untx). Expression data are shown relative to the level in ASCs. B, cell proliferation assessed by a DNA quantitation assay at 2 to 6 daysafter lentivirus infection. C, proportion of cells staining for caspase-3 by immunohistochemistry 3 days after lentiviral infection. D, immunoblots forcaspase-3 and PARP as well as their cleaved forms. Caspase-3 was assessed on day 3 after lentiviral infection and PARP on day 3 in LPS141 andday 4 in DDLS8817. E, percentage of cells incorporating BrdU on day 3. F, proportion of cells staining for phosphorylated histone H3 on day 4.






Discussion

We have showed that sequencing-based miRNA expressionprofiles discriminate normal fat, WDLS, and DDLS. Micro-RNA-143, which is abundant in normal adipose tissue, isunderexpressed in WDLS, and its expression decreases furtheras tumors progress to DDLS. These findings imply that miR-143 underexpression is an early event in liposarcomagenesis,but that further loss may play a role in dedifferentiation. Esauand colleagues previously showed that miR-143 levels increasein differentiating adipocytes and that miR-143 is required fordifferentiation (21). We found as well that miR-143 levelsincrease in differentiating adipocytes; however, re-expressionof miR-143 is insufficient to restore differentiation in thecellular context of DDLS. Therefore, while miR-143 lossmay produce effects through inhibition of differentiation,miR-143 restoration alone is unable to reverse this phenotypein DDLS cells.

Re-expression of miR-143 but not its co-cistronic partnermiR-145 inhibits DDLS cell proliferation and induces apop-tosis. miR-143 re-expression also decreases entry into Sphase and inhibits mitosis. The effects of miR-143 on pro-liferation, apoptosis, and cell-cycle progression may beexplained by a gene network targeted by miR-143, includingdirect miR-143 targets BCL2, TOP2A, and PRC1 and the

indirect target PLK1. This network includes genes involvednot only in cell proliferation and apoptosis, but also incytokinesis. Pharmacologic inhibition of PLK1, a regulatorof cytokinesis, induces apoptosis and cell-cycle arrest inDDLS cells.

Prior work has implicated miR-143 as a tumor suppressorin other cancers. In bladder cancer, miR-143 inhibits cancergrowth through antagonizing the expression of RAS (28).Both miR-143 and miR-145 are downregulated in coloncancer and in B-cell malignancies, such as chronic lympho-cytic leukemia and Burkitt's lymphoma (29). miR-143 inparticular seems to be a tumor suppressor in colorectalcancer, acting through downregulation of KRAS, a mediatorof the mitogen-activated protein kinase cascade (30). Inthese other cancer types, the main effect of miR-143 re-expression is inhibition of proliferation without induction ofapoptosis. In contrast, in liposarcoma cells we detect sig-nificant induction of apoptosis after only 48 hours of miR-143 re-expression.

We have shown that miR-143 downregulates BCL2, TOP2A,and PRC1 in DDLS cells. BCL2 encodes a mitochondrialmembrane protein that blocks apoptosis and has been impli-cated in multiple human cancers. It was recently identified asa direct target of miR-143 in osteosarcoma (31) and bears atleast 2 predicted miR-143 target sites (SupplementaryTable S3). The repression of BCL2 may contribute to miR-1430s proapoptotic properties, but, given miR-1430s lack ofsignificant effect on BCL2 in one of the two DDLS cell lines,this may not apply to all DDLS tumors.

TOP2A is an enzyme controlling the topologic state ofDNA and involved in transcription and DNA replication. Itis upregulated in multiple human cancers such as breastand prostate cancer, and its expression is associated withandrogen resistance and decreased survival in prostatecancer. Our group has found that TOP2A expression isupregulated 43-fold in DDLS relative to normal fat (Sup-plementary Table S4). Furthermore, TOP2A upregulation inliposarcomas was associated with decreased distant recur-rence-free survival. Short hairpin RNA knockdown ofTOP2A causes induction of apoptosis and decreased inva-sive ability of DDLS cells (32), underscoring its importancein this disease.

To date, the regulation of TOP2A in DDLS and other tumorshas been poorly defined. Our study provides the first evidencethat miR-143 regulates TOP2A and that this regulation maypartially explain miR-1430s antiproliferative and proapoptoticeffects. In addition, miR-1430s regulation of TOP2A, a keyenzyme of DNA replication, may explain the decreased pro-gression into S phase after miR-143 restoration.

TOP2A is a target of anthracycline-based chemotherapeuticagents, such as doxorubicin, which alter TOP2A catalysis,generating high levels of DNA breaks that then trigger celldeath pathways. These agents, however, are not completelyTOP2A-specific, and their use is limited by high toxicity andpoor efficacy in DDLS. The present results suggest thatrational screening for more potent and specific TOP2A-tar-geted agents may lead to more effective therapy for patientswith liposarcoma.

Figure 3. Gene module identified through network analysis of geneexpression changes induced by miR-143 re-expression and targetvalidation. Downregulation is represented by blue and upregulation by red.Bolded borders represent predicted direct targets of miR-143. Circlesrepresent genes consistently down- or upregulated, and diamondsrepresent linker genes. CDC2 was downregulated to an extent just belowthe cutoff for being included in the input list and is therefore a linker gene.

Ugras et al.





PRC1 and PLK1 are crucial regulators of cytokinesis.PRC1, a docking partner of PLK1, recruits PLK1 to thecentral spindle during anaphase (22). Suppression of PRC1expression (33, 34) or chemical inhibition of PLK1 (35)results in mitotic failure, and both proteins are requiredfor proper cytokinesis. Our group has previously found thatPRC1 is upregulated 24-fold in DDLS relative to normal fat(13). We showed here that miR-143 negatively regulatesPRC1 and PLK1 expression, that miR-143 restoration inhibitsmitosis in DDLS cells, and that PLK1 inhibition via drugtreatment leads to mitotic arrest and cell death. Theseobservations, together with the microarray and networkanalysis implicating miR-1430s regulation of other genesinvolved in cytokinesis (ECT2, CDC2, and CDC25B), suggestthat miR-143 plays an important role in regulating cytokin-esis in liposarcoma cells. Furthermore, they imply that PLK1inhibitors may have potential for the treatment of DDLS.

Key to our study has been elucidation of the complexregulatory network of miR-143. Consistent with the observa-tion that microRNAs target a multitude of genes, each ofwhich has a small but significant effect on the overall phe-notype conferred by the microRNA, we identified a stronglyconnected network regulated by miR-143. Interestingly,expression levels of 10 of the 24 genes in the network areassociated with decreased distant recurrence-free survival(ref. 32; Supplementary Table S4). In addition, 9 of these 10genes are upregulated significantly in DDLS relative to normalfat (13). The 24-gene network controls several importantcellular functions ultimately converging in cell growth andsurvival. Although agents targeting individual miR-143 targetsmay provide useful therapy for DDLS, our results suggesttherapy targeting the entire dysregulated gene network viadelivery of miR-143 to DDLS tumors may provide moreeffective therapy.

Figure 4. Downregulation of A,TOP2A, B, PRC1, and C, PLK1after miR-143 re-expression.Levels of mRNAs (left) weremeasured by quantitative PCR 2.5days after lentiviral infection andare shown relative to ASCs.Protein levels of TOP2A, PRC1,and PLK1 were assessed byWestern blot analysis at day 2after lentivirus infection in LPS141and at day 4 in DDLS8817.






Disclosure of Potential Conflicts of Interest

T. Tuschl is a cofounder and scientific advisor to Alnylam Pharmaceuticalsand an advisor to Regulus Therapeutics. The other authors disclosed nopotential conflicts of interest.

Acknowledgments

We thank Marc Ladanyi and Matthew Meyerson, Co-PIs of the SarcomaGenomic Project, for support and advice. We appreciate the technical assistanceof W. Zhang, S. Dewell, and C. Zhao (Genomics Resource Center) for SOLEXAsequencing. We also thank the members of the MSKCC Genomics CoreLaboratory and N.H. Moraco for clinical data support.

Grant Support

This work was supported in part by the Soft Tissue Sarcoma Program Project(P01 CA047179, S. Singer, C.R. Antonescu, and C. Sander), the Starr FoundationCancer Consortium, and by a donation from Richard Hanlon and Mortimer B.Zuckerman. M. Hafner was supported by the Deutscher Akademischer Aus-tauschdienst and is a fellow of the Charles H. Revson, Jr. Foundation.

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicate thisfact.

Received March 11, 2011; revised June 14, 2011; accepted June 16, 2011;published OnlineFirst June 21, 2011.

References1. Mack TM. Sarcomas and other malignancies of soft tissue, retro-

peritoneum, peritoneum, pleura, heart, mediastinum, and spleen.Cancer 1995;75:211–44.

2. Barretina J, Taylor BS, Banerji S, Ramos AH, Lagos-Quintana M,Decarolis PL, et al. Subtype-specific genomic alterations define newtargets for soft-tissue sarcoma therapy. Nat Genet 2010;42:715–21.

3. Singer S, Antonescu CR, Riedel E, Brennan MF. Histologic subtypeand margin of resection predict pattern of recurrence and survival forretroperitoneal liposarcoma. Ann Surg 2003;238:358–70; discussion70–1.

4. Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and func-tion. Cell 2004;116:281–97.

5. Bartel DP. MicroRNAs: target recognition and regulatory functions.Cell 2009;136:215–33.

6. Guo H, Ingolia NT, Weissman JS, Bartel DP. Mammalian microRNAspredominantly act to decrease target mRNA levels. Nature 2010;466:835–40.

7. Friedman RC, Farh KK, Burge CB, Bartel DP. Most mammalianmRNAs are conserved targets of microRNAs. Genome Res 2009;19:92–105.

8. Croce CM. Causes and consequences of microRNA dysregulation incancer. Nat Rev Genet 2009;10:704–14.

9. Ventura A, Jacks T. MicroRNAs and cancer: short RNAs go a longway. Cell 2009;136:586–91.

10. Li M, Lee KF, Lu Y, Clarke I, Shih D, Eberhart C, et al. Frequentamplification of a chr19q13.41 microRNA polycistron in aggressiveprimitive neuroectodermal brain tumors. Cancer Cell 2009;16:533–46.

11. Krutzfeldt J, Rajewsky N, Braich R, Rajeev KG, Tuschl T, ManoharanM, et al. Silencing of microRNAs in vivo with ‘antagomirs’. Nature2005;438:685–9.

12. Gimble J, Guilak F. Adipose-derived adult stem cells: isolation, char-acterization, and differentiation potential. Cytotherapy 2003;5:362–9.

13. Singer S, Socci ND, Ambrosini G, Sambol E, Decarolis P, Wu Y, et al.Gene expression profiling of liposarcoma identifies distinct biological

Figure 5. Effects of the PLK1inhibitor BI 2536 on liposarcomacell lines. A, cell proliferationassessed by DNA quantitationprior to treatment (day 0) and atdays 2 and 4 after treatment with20 nmol/L BI 2536. B, percentageof cells that are Annexin(þ) 7-AAD(–) by flow cytometric analysis inuntreated versus treated LPS141and DDLS8817 cells at day 2 aftertreatment. C, flow cytometryanalysis of untreated cells andcells subjected to 2-day treatmentwith BI 2536. The smooth curvesshow the numbers of cells in G1,G2, and S phases.

Ugras et al.





types/subtypes and potential therapeutic targets in well-differentiatedand dedifferentiated liposarcoma. Cancer Res 2007;67:6626–36.

14. Hafner M, Renwick N, Pena J, Mihailovi�c A, Tuschl T. BarcodedcDNA libraries for miRNA profiling by next-generation sequencing.In:Hartmann RK, Bindereif A, Sch€on A, Westhof A, editors. Handbookof RNA Biochemistry, 2nd ed. Wennheim, Germany: Wiley-VCHVerlag GmbH; 2010. In press.

15. Pena JT, Sohn-Lee C, Rouhanifard SH, Ludwig J, Hafner M,MihailovicA, et al. miRNA in situ hybridization in formaldehyde and EDC-fixedtissues. Nat Methods 2009;6:139–41.

16. Sood P, Krek A, Zavolan M, Macino G, Rajewsky N. Cell-type-specificsignatures of microRNAs on target mRNA expression. Proc Natl AcadSci U S A 2006;103:2746–51.

17. Jacobsen A, Wen J, Marks DS, Krogh A. Signatures of RNA bindingproteins globally coupled to effective microRNA target sites. GenomeRes 2010;20:1010–9.

18. Smyth GK. Linear models and empirical bayes methods for assessingdifferential expression in microarray experiments. Stat Appl GenetMolBiol 2004;3:Article3. Epub 2004 Feb12.

19. Cerami E, Demir E, Schultz N, Taylor BS, Sander C. Automatednetwork analysis identifies core pathways in glioblastoma. PLoSOne 2010;5:e8918.

20. Cordes KR, Sheehy NT, White MP, Berry EC, Morton SU, Muth AN,et al. miR-145 and miR-143 regulate smooth muscle cell fate andplasticity. Nature 2009;460:705–10.

21. Esau C, Kang X, Peralta E, Hanson E, Marcusson EG, RavichandranLV, et al. MicroRNA-143 regulates adipocyte differentiation. J BiolChem 2004;279:52361–5.

22. Neef R, Gruneberg U, Kopajtich R, Li X, Nigg EA, Sillje H, et al. Choiceof Plk1 docking partners during mitosis and cytokinesis is controlledby the activation state of Cdk1. Nat Cell Biol 2007;9:436–44.

23. Schmidt A, Durgan J, Magalhaes A, Hall A. Rho GTPases regulatePRK2/PKN2 to control entry into mitosis and exit from cytokinesis.EMBO J 2007;26:1624–36.

24. Wolfe BA, Takaki T, Petronczki M, Glotzer M. Polo-like kinase 1 directsassembly of the HsCyk-4 RhoGAP/Ect2 RhoGEF complex to initiatecleavage furrow formation. PLoS Biol 2009;7:e1000110.

25. Wasch R, Hasskarl J, Schnerch D, Lubbert M. BI_2536–targeting themitotic kinase Polo-like kinase 1 (Plk1). Recent Results Cancer Res2010;184:215–8.

26. Schoffski P, Blay JY, De Greve J, Brain E, Machiels JP, Soria JC, et al.Multicentric parallel phase II trial of the polo-like kinase 1 inhibitor BI2536 in patients with advanced head and neck cancer, breast cancer,ovarian cancer, soft tissue sarcoma and melanoma. The first protocolof the European Organization for Research and Treatment of Cancer(EORTC) Network Of Core Institutes (NOCI). Eur J Cancer 2010;46:2206–15.

27. Gleixner KV, Ferenc V, Peter B, Gruze A, Meyer RA, Hadzijusufovic E,et al. Polo-like kinase 1 (Plk1) as a novel drug target in chronic myeloidleukemia: overriding imatinib resistance with the Plk1 inhibitor BI2536. Cancer Res 2010;70:1513–23.

28. Lin T, Dong W, Huang J, Pan Q, Fan X, Zhang C, et al. MicroRNA-143 as a tumor suppressor for bladder cancer. J Urol 2009;181:1372–80.

29. Akao Y, Nakagawa Y, Naoe T. MicroRNA-143 and -145 in coloncancer. DNA Cell Biol 2007;26:311–20.

30. Chen X, Guo X, Zhang H, Xiang Y, Chen J, Yin Y, et al. Role of miR-143targeting KRAS in colorectal tumorigenesis. Oncogene 2009;28:1385–92.

31. Zhang H, Cai X, Wang Y, Tang H, Tong D, Ji F. microRNA-143,down-regulated in osteosarcoma, promotes apoptosis and sup-presses tumorigenicity by targeting Bcl-2. Oncol Rep 2010;24:1363–9.

32. Gobble RM, Qin LX, Brill ER, Angeles CV, Ugras S, O’Connor RB, et al.Expression profiling of liposarcoma yields a multigene predictor ofpatient outcome and identifies genes that contribute to liposarcoma-genesis. Cancer Res 2011;71:2697–705.

33. Mollinari C, Kleman JP, Saoudi Y, Jablonski SA, Perard J, Yen TJ, et al.Ablation of PRC1 by small interfering RNA demonstrates that cyto-kinetic abscission requires a central spindle bundle in mammaliancells, whereas completion of furrowing does not. Mol Biol Cell 2005;16:1043–55.

34. Mollinari C, Kleman JP, Jiang W, Schoehn G, Hunter T, Margolis RL.PRC1 is a microtubule binding and bundling protein essential tomaintain the mitotic spindle midzone. J Cell Biol 2002;157:1175–86.

35. Santamaria A, Neef R, Eberspacher U, Eis K, Husemann M, MumbergD, et al. Use of the novel Plk1 inhibitor ZK-thiazolidinone to elucidatefunctions of Plk1 in early and late stages of mitosis. Mol Biol Cell2007;18:4024–36.






2011;71:5659-5669. Published OnlineFirst June 21, 2011.Cancer Res Stacy Ugras, Elliott Brill, Anders Jacobsen, et al. MicroRNA-143 Tumor Suppressor Activity in LiposarcomaSmall RNA Sequencing and Functional Characterization Reveals

Updated version

10.1158/0008-5472.CAN-11-0890doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2011/06/21/0008-5472.CAN-11-0890.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/71/17/5659.full#ref-list-1

This article cites 33 articles, 10 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/71/17/5659.full#related-urls

This article has been cited by 4 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://cancerres.aacrjournals.org/content/71/17/5659To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-11-0890

http://cancerres.aacrjournals.org/content/suppl/2011/06/21/0008-5472.CAN-11-0890.DC1

http://cancerres.aacrjournals.org/content/71/17/5659.full#ref-list-1

http://cancerres.aacrjournals.org/content/71/17/5659.full#related-urls

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

mailto:[email protected]

http://cancerres.aacrjournals.org/content/71/17/5659


small rna sequencing and functional characterization ... · for retroperitoneal disease, the most...

Documents