global microrna analysis of the nci-60 cancer cell panel · spotlight on molecular profiling global...

Spotlight on Molecular Profiling

Global microRNA Analysis of the NCI-60 Cancer Cell Panel

Rolf Søkilde1, Bogumil Kaczkowski2, Agnieszka Podolska3, Susanna Cirera3, Jan Gorodkin4,Søren Møller1, and Thomas Litman1

AbstractMicroRNAs (miRNA) are a group of short noncoding RNAs that regulate gene expression at the

posttranscriptional level. They are involved in many biological processes, including development, differ-

entiation, apoptosis, and carcinogenesis. Because miRNAsmay play a role in the initiation and progression of

cancer, they comprise a novel class of promising diagnostic and prognostic molecular markers and potential

drug targets. By applying an LNA-enhanced microarray platform, we studied the expression profiles of 955

miRNAs in the NCI-60 cancer cell lines and identified tissue- and cell-type–specific miRNA patterns by

unsupervised hierarchical clustering and statistical analysis. A comparison of our data to three previously

publishedmiRNA expression studies on the NCI-60 panel showed a remarkably high correlation between the

different technical platforms. In addition, the current work contributes expression data for 369 miRNAs that

have not previously been profiled. Finally, by matching drug sensitivity data for the NCI-60 cells to their

miRNA expression profiles, we found numerous drug–miRNAs pairs, for which the miRNA expression and

drug sensitivity profiles were highly correlated and thus represent potential candidates for further investiga-

tion of drug resistance and sensitivity mechanisms. Mol Cancer Ther; 10(3); 375–84. �2011 AACR.

Introduction

Twenty years have passed since the NCI-60 cell panelfor anticancer drug screening was established in 1990 (1),and today, these cell lines represent the most extensivelystudied and best characterized set of tumor cells available.The panel comprises 59 human cancer cell lines derivedfrom 9 different tissues of origin: melanoma, leukemia,breast, central nervous system (CNS), colon, lung, ovary,prostate, and kidney. [One breast cancer cell line,MB-468,has high identity to MDA-MB-231 (http://www.sanger.ac.uk/genetics/CGP/Genotyping/nci60.shtml) and wasexcluded from the NCI-60 panel that we received for ouranalysis.] Chemosensitivity profiles of theNCI-60 cells formore than 100,000 compounds have been generated, andtogether with genomic, transcriptomic, proteomic, epige-nomic, andmetabolomic data, this information is publiclyavailable via the NCI Developmental Therapeutics Pro-

gramwebsite (2)or canbemineddirectlyby theCellMinerquery tool (3). One type of data, which could be importantfor understanding the link between genomic and chemo-sensitivity data, particularly for which there seems to be alack of correlation betweenmRNAandprotein levels,wasrecently added to the NCI-60 database, namely, micro-RNA (miRNA) expression information (4–6).

miRNAs constitute a group of short, noncoding RNAswhich repress gene expression at the posttranscriptionallevel by binding to complementary sequences in the30-untranslated region of their target mRNAs (7). Theyare involved in the regulation of numerous biologicalprocesses, including differentiation, proliferation, andapoptosis. Moreover, aberrant miRNA expression hasbeen associated with many diseases, including cancer,in which miRNAs can function both as tumor suppres-sors and as oncogenes (8). Because miRNAs play a role incancer pathogenesis, they may have potential as mole-cular biomarkers, not only for classification purposes butalso for prediction of treatment response and for discov-ery of new targets. Today, 1,100 miRNAs and severalthousand putative miRNA targets have been identified inthe human genome and are accessible via miRBase, themain miRNA database (9). It is the expression of thesemiRNAs and a number of viral and novel sequencesidentified by high-throughput sequencing (10) that wehave been studying with the application of an LNA-enhanced microarray platform.

We present here a comprehensive analysis of theNCI-60 cell line panel miRNA landscape, where theexpression of 955 miRNAs has been correlated to drugsensitivity, mRNA and protein level, and tissue of origin

Authors’ Affiliations: 1Department of Biomarker Discovery, Exiqon A/S,Vedbæk; 2Department of Biology, Bioinformatics Centre, University ofCopenhagen, CopenhagenN; 3Department of Animal and Veterinary BasicSciences, and 4Center for Non-coding RNA in Technology and Health,Division of Genetics and Bioinformatics, Faculty of Life Sciences, Uni-versity of Copenhagen, Frederiksberg C, Denmark

Note: Supplementary material for this article is available at MolecularCancer Therapeutics Online (http://mct.aacrjournals.org/).

R. Søkilde and B. Kaczkowski contributed equally to this work.

Corresponding Author: Thomas Litman, Department of Biomarker Dis-covery, Exiqon A/S, Bygstubben 9, DK-2950 Vedbæk, Denmark. Phone:4545650933; Fax: 4545661888. E-mail: [email protected]

doi: 10.1158/1535-7163.MCT-10-0605

�2011 American Association for Cancer Research.

MolecularCancer

Therapeutics

www.aacrjournals.org 375

on June 11, 2020. © 2011 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst January 20, 2011; DOI: 10.1158/1535-7163.MCT-10-0605

http://mct.aacrjournals.org/

of the cells. In addition, our meta-analysis of the 4 avail-able NCI-60 miRNA data sets [i.e., real-time PCR(RT-PCR) study ofGaur and colleagues (5), custommicro-array data of Blower and colleagues (4),Agilent array dataof Liu and colleagues (6), and the current LNAmicroarrayanalysis] cross-validates the individual platforms wherethey have overlapping and concordant targets. (At thetime of submission of this article, the complete data setwas not yet publicly available. Therefore, our analysis islimited to the 365 miRNAs reported to be expressed in atleast 10% of the cell lines by Liu and colleagues.)

Materials and Methods

Cell linesThe NCI-60 cell stocks were kindly provided by Dr.

Susan Holbeck at the NCI Developmental TherapeuticsProgram in May 2006, except for the SR lymphoma cellline, which was purchased from American Type CultureCollection (CRL-2262) in 2007. Supplementary Table S1lists the origin of the cell lines and some of their char-acteristics. All cells were grown at 37�C in 5% CO2 inRPMI-1640 medium (Invitrogen), supplemented with 2mmol/L L-glutamine and 10% FBS (Invitrogen). Thetissue origin of each cell line is denoted as follows: BR,breast; CNS, glioma; CO, colon; LE, leukemia; LU, lung;ME, melanoma; OV, ovary; PR, prostate; RE, renal. Thecells were grown to approximately 80% confluence andpassaged twice before RNA extraction.

RNA extraction and quality controlTotal RNAwas extracted by Trizol reagent (Invitrogen)

according to the manufacturer’s instructions. RNA qual-ity was assessed with Bioanalyzer 2100 Nano chips (Agi-lent) and did not show sign of degradation as judged bythe RNA integrity number, which was above 8 for mostsamples. RNA concentrations were measured on a Nano-Drop ND-1000 (Thermo Scientific).

RNA labelingOur microarray experimental setup used a common

reference design as described in detail by Søkilde andcolleagues (11). One microgram of total RNA from eachsample was labeled using themiRCURY LNAmicroRNAPower labeling Kit (Exiqon), following the man-ufacturer’s recommendations. Cell line–specific RNAwas labeled with Hy3 fluorophore, whereas the commonreference RNA pool was labeled with Hy5.

Microarray hybridization and scanningFor miRNA profiling, we applied the LNA-enhanced

miRCURY Dx 9.2 microarray platform, containing quad-ruplicate probes for 955 mature miRNAs (Exiqon).Labeled RNA samples were hybridized with the arraysfor 16 hours at 65�C in a Tecan HS 4800 hybridizationstation. After washing and drying, the microarrays werescanned in a DNA Microarray Scanner (Agilent) and theresulting images were quantified using ImaGene v. 8.0(BioDiscovery). Both automatic quality control (i.e., flag-

ging of poor spots) and manual, visual inspection wasdone to ensure the highest possible data quality.

Data preprocessing and normalizationAll low-level analyses, such as importing and prepro-

cessing of the data, were done in the R environment byusing the LIMMA package (12). After excluding flaggedspots from the analysis, the "normexp" background cor-rection method, plus offset¼ 50, was applied, after whichintensities were log2 transformed and quantile normal-ized as implemented in LIMMA. Both log2 intensities(single-channel analysis) and log2 ratios (dual channelanalysis) of 4 intraslide replicates were averaged. Allexpression data were deposited in the Rosetta Resolver(Rosetta Biosoftware) data management system.

High-level data analysisTo identify miRNAs that were differentially expressed,

we applied the empirical Bayes, moderated t-statisticsimplemented in LIMMA. Linearmodels were fitted to thedata, and comparisons of interest were extracted as con-trasts. Unsupervised hierarchical clustering with com-plete linkage, using both Euclidean and (1 � Pearsoncorrelation) as distancemetrics, was applied to cluster theNCI-60 cell lines according to their miRNA expressionlevels.

Mapping probes from prior data setsThe quantitative PCR (qPCR) data generated by Gaur

and colleagues (5) and the array data by Blower andcolleagues (4) and Liu and colleagues (6) werematched tothe current miRBase release 15.0, in which annotationwas kept at the level of MIMAT names (the identifier formature miRNA sequences; see Supplementary Table S2).Probe designs for the Blower array were downloadedfrom theArrayExpress Archive (13), and all probeswith aperfect match to a mature miRNA in miRBase 15.0 wereused for the subsequent analysis. Probes with only partialoverlap were excluded from the analysis.

Drug sensitivity correlation analysisTo examine the relationship between drug sensitivity

and miRNA expression, we took 3 approaches: First, wedownloaded GI50 data (July 2007 release) for the NCI-60panel from the NCI DTP web site (14). The raw dataconsisted of GI50 values for 45,857 drugs, which afterapplying a variance filter (variance > 0; ref. 15) werereduced to 6,081 drugs. After minimum range filtering(GI50_max-GI50_min> 2), 4,442drugswere left in thedata set.

Second, in addition to the aforementioned raw GI50data, we included a preprocessed data set from theCellMiner database (3). These data consist of 118 drugswith known mechanism of action and are described byBussey and colleagues (15).

Correlation between miRNA expression and drug sen-sitivity was calculated in a pairwise manner, for eachmiRNA–drug combination, wherein each drug wasrepresented by a vector of GI50 measurements and each

Søkilde et al.

Mol Cancer Ther; 10(3) March 2011 Molecular Cancer Therapeutics376




miRNA was represented by a vector of log2 Hy3 inten-sities. Both Spearman and Pearson correlation coefficientswere calculated, and the P values of the correlationsignificance were adjusted for multiple testing by theBenjamini and Hochberg false discovery rate (FDR) cor-rection method (16).Finally, a COMPARE analysis (October 2009 release)

was run to search for correlations with unfiltered GI50values and resulted in 36,553 correlations with P < 0.005.

mRNA expression correlation analysisGlobal mRNA expression profiles of NCI-60 cells mea-

sured on the Affymetrix HG U133A and U133B micro-array platform were downloaded from Gene ExpressionOmnibus (GEO; accession number GSE5720; ref. 17) andnormalized using Rosetta Resolver. Correlations betweenlog2 intensities of overlapping miRNA–mRNA pairs [i.e.,miRNAs that reside within introns of protein codinggenes; the pairs are represented by miRCURY (formiRNA) and Affymetrix (for mRNA) probes] were cal-culated using both Spearman and Pearson correlationcoefficients, and the P values were adjusted for multipletesting by the Benjamini andHochberg method with FDRof 0.01 or less. The genomic coordinates of all knownhuman miRNAs were extracted from the human miRNAannotation file downloaded from miRBase release 15.0(18). Pairs of miRNA and mRNA transcripts were iden-tified by matching the genomic location: chromosome,start, end, and strand.

Molecular targets correlation analysisA correlation analysis of our miRNA expression data

against the Molecular Targets database (September 2008release; ref. 19) included 43,981 Pearson correlations witha cutoff value of P < 0.005 (Supplementary Table S3). TheMolecular Targets data sets compile information on pro-tein levels, RNA expression, DNA mutation status, andenzyme activity for the NCI-60 cell lines.

Quantitative RT-PCRThe expression levels of 20 selected miRNAs and 10

reference genes were validated by quantitative RT-PCRapplying the miRCURY LNA microRNA PCR systemand SYBR Green master mix, following the man-ufacturer’s instructions (Exiqon). The results are shownin Supplementary Fig. S3.

Northern blot analysisTo visualize the level and size of 7 candidate miRNAs,

we conducted Northern blot analysis on RNA from theSNB-19 (glioma), HCT-15 (colon), and SR (leukemia) celllines. The method and results are described in the Sup-plementary Fig. S4.

Results and Discussion

We studied global miRNA expression in the NCI-60cell line panel by applying a recently described (20),

highly sensitive and specific miRCURY microarray plat-form, which contains LNA-enhanced and melting tem-perature (Tm-normalized capture probes for quantitationof 599 mature human miRNAs, annotated in miRBasev.15.0, and 345 proprietary miRNAs, 29 viral miRNAs,and 10 snoRNAs.

Cross-platform comparisonThree previous studies have reported miRNA expres-

sion patterns from the NCI-60 cell lines: Blower andcolleagues (4) and Liu and colleagues (6) used oligonu-cleotide microarrays, whereas Gaur and colleagues (5)applied real-time quantification by TaqMan stem-loopRT-PCR. A comparison of the different platform cover-age and the latest miRBase version 15.0 shows that thecurrent work contributes with expression data for 369miRNAs that have not previously been profiled(Fig. 1A), including 345 miRPlus sequences not yetannotated in miRBase.

To assess correlation between platforms, the 125miRNAs common to all 4 analyses were identified andthe 6 pairwise (interplatform) Pearson correlation coeffi-cients were computed, as summarized in Table 1 (valuesin bold). (Only 365/727 miRs were reported as expressedby Liu and colleagues. Therefore, the number ofexpressed miRNAs in common between the 4 studiesis 125, and not 158, which is the total number of commonmiRNAs analyzed.) Taking into consideration that com-pletely independent RNA preparations and assay tech-nologies have been applied, the concordance between the4 miRNA platforms is remarkably high compared with asimilar mRNA analysis on the NCI-60 cell line panel (17).The miRNA correlation values (the same miRNA mea-sured across platforms) ranged from highly correlated(r ¼ 0.97) to negative (r ¼ �0.29) and with an overallaverage of 0.48 (SD ¼ 0.28), which is far better than whathas been reported previously in an mRNA correlationstudy comparing oligonucleotide and cDNA microarraymeasurements on the NCI-60 panel (the Pearson correla-tion coefficient between the 2 array platformswas close tozero; ref. 21).

Besides obvious methodologic and biological inter-study differences (such as differences in growth condi-tions, RNA extraction, and analysis method), we havenoticed that the 3 previously published studies applyvarious types of flooring to the data. Although flooringmay serve to eliminate negative expression values due tobackground subtraction from low-intensity values, suchan ad hoc threshold does not reflect the experimentalmeasurements and can even lead to artifacts (22). There-fore, we chose to apply the normexp background correc-tion method for stabilizing the variance of low-intensityprobes after log2 transformation.

Although there is a good overall consistency betweenplatforms as assessed by the distribution of miRNAcorrelations (Fig. 1B), we observed an even highercorrelation between the cell line–specific signatures(Table 1, values in italics; and Fig. 1C). This suggests

NCI-60 miRNA Analysis

www.aacrjournals.org Mol Cancer Ther; 10(3) March 2011 377




that cell line–specific miRNA profiles are robust and canbe identified regardless of platform. Only 2 cell linesshowed very low correlation between the differentstudies, namely, the CNS cell lines SNB-19 and SNB-75, with an average correlation coefficient of 0.18 to 0.31.This could indicate that these cell lines have a pheno-type which is sensitive to culturing conditions. Inter-estingly, the LNA microarray was the only platform tocorrelate the SNB-19 and U251 cell lines closely

together, which agrees with their similar genotypes,as they are derived from the same individual (23).

For further comparison of the platforms, the best cor-relating of the 125 common miRNAs were used in anunsupervised hierarchical cluster analysis which showedthat most of the tissue- and cell-specific expression pat-terns are retained between the 4 assays (Fig. 2). Weselected 62 miRNAs with the highest average correlation(R > 0.5) between all combinations of platforms to obtain a

Table 1. Correlation between studies

Study Gaur Blower LNA Liu

Gaur 1 0.43 0.51 0.59Blower 0.58 1 0.47 0.55LNA 0.56 0.55 1 0.47Liu 0.74 0.64 0.69 1

NOTE: The table shows pair wise Pearson correlation coefficients between the differentmiRNAprofiling datasets on the basis of eithermiRNAs (125 in common, in bold) or on the 59 cell lines studied (italics).

A B

C

Gaur (181)

LNA (944)

Blower (391)

345

2 0 0

0.25

0.20

0.15

Fre

quen

cy

0.10

0.05

0.00–1 –0.8–0.6–0.4–0.2 0 0.2 0.4 0.6 0.8 1

Matched miRNA correlationsAll miRNA correlations

0.25

0.20

0.15

Fre

quen

cy

0.10

0.05

0.00–1 –0.8–0.6–0.4–0.2 0 0.2 0.4 0.6 0.8 1

Matched miRNA correlationsAll miRNA correlations

1

438158179

221 15 0 103

3460024

Liu (727)

miRBase 15(1,100)

Figure 1. A, four-way Venn diagram showing the overlap of the miRNAs studied in the 4 profiling studies (Liu, Blower, Gaur, and LNA) and the miRNAsannotated in miRBase version 15.0. B, a comparison of the near-random distribution of all possible combinations of miRNA correlations (left peak around zero)and the matched miRNA correlations (right peak). C, a similar comparison as in B, but for cell line expression profiles, all cell line combinations give a randomdistribution profile (left peak around zero), whereas matched cell line–specific miRNA profiles correlate well (right peak).

Søkilde et al.





core set of "high confidence" miRNAs for clustering withresampling (Fig. 2A). Strong tissue-specific miRNA sig-natures were identified for cell lines of colon, melanoma,leukemia, and kidney origin (the latter included a singleoutlier, SN12C). These cell lines clustered together inde-pendently of platform and cell line origin, suggesting astrong and consistent tissue-specific signature in thesetissues, such as miR-211 expression in melanoma, miR-142 expression in leukemia, miR-192 expression in colon,and miR-138 in kidney. However, we also identified celllines representing tissue groups, including lung, ovary,and breast, with a more diverse expression pattern.Therefore, we chose to cluster these tissues separatelyto identify miRNA patterns unique to the individual cellline (Fig. 2B–D).Again, we found good concordance between the plat-

forms on the basis of "high confidence" set of miRNAs,showing that cell lines, though not displaying tissue-specific expression, retain their miRNA profile betweendifferent laboratories and studies. This is promising forthe research of miRNA function, because related cell linesstudied in different experimental settings will likely dis-play similar miRNA profiles. However, there are excep-tions; for example, the lung cancer cell line HOP-62 on the

Gaur platform showed a miRNA profile that differedmarkedly from that of the 3 other platforms (Fig. 2B).

We found that the lung cancer cell lines (Fig. 2B)showed the most intratissue diversity among all the celllines studied, as these cell lines scattered throughout thefull cluster (Fig. 2A). From the clustering of the lungcancer cell lines, the non–small cell lung carcinoma largecell HOP-92, which is extremely slow growing, showedthe most distinct miRNA profile.

Two of the ovarian cancer cell lines, OVCAR-8 andNCI/ADR-RES, have similar karyotypes (24). This isreflected in Fig. 2C, as all 4 platforms cluster these celllines together. Other cell line pairs with high identity, andthus most likely, a common origin, include the CNS celllines SNB-19 and U251 and the melanoma cell lines M14and MDA-MB-435.

All platforms agreed on separating the breast cancercell lines into 2 distinct clusters, representing miRNAexpression profiles specific for basal and luminal typebreast cancer (Fig. 2D).

In summary, our correlation analysis shows that mostmiRNA expression profiles are retained across NCI-60data sets and therefore the data can potentially be usedin combination, for example, in a meta-analysis. The

A All cell lines

B Lung C Ovary D Breast

Figure 2. Combined hierarchical clustering of the 4 miRNA data sets, using z-scaled expression data from the common probe set. A, all 59 cell lines (236expression profiles across 4 platforms). B, the 9 lung cancer cell lines. C, the 7 ovarian cancer cell lines. D, the 6 breast cancer cell lines. The selected set ofmiRNAs represents the best correlating miRNAs across the 4 platforms. The clustering is based on resampling analysis of 62 miRs and 59 cell lines per dataset. The color code at the very top designates the tissue origin of the cell lines. Pink, breast; black, CNS; green, colon; blue, lung; red, leukemia; brown,melanoma; orange, ovary; dark pink, prostate; yellow, renal.






observed interstudy variation may be due to both tech-nology and biology: different assay and probe designs arelikely to give different results, besides the expected varia-tion due to differences in growth conditions and handlingof the cell lines. For the rest of the article, we focus on thedata generated by the use of LNA-enhanced microarrays.

Tissue- and cell line–specific miRNA expressionThe distribution of tissue-specific miRNAs (i.e., those

miRNAs that were preferentially expressed in cell linesoriginating from one tissue compared with the rest of

the cell lines) is summarized in the heatmap shown inFig. 3.

Breast cancerBecause of the heterogeneous nature of the breast

cancer cell lines, only 2 miRNAs were identified as breastcancer specific compared with the rest of the cell lines:miR-195 and miR-196a. Both these miRNAs are upregu-lated in breast cancer patients, and recently, miR-195 hasbeen identified as a serum biomarker diagnostic forbreast cancer (25). When looking at the "breast cancer

A B

Breast

CNS

Colon

COLO-205C

LungHCT-116

Leukemia

Ovary

Prostate

Renal

Melanoma

Figure 3. Tissue-specific miRNAs. A, miRNAs with significantly higher expression level in cell lines from one tissue compared with the rest of the cell lines. Theheatmap shows the log2 expression for the miRNAs (mean centered data), which are enriched in each tissue. B, miRNAs are significantly different betweenluminal and basal breast cancer cell lines. C, high expression of miR-142 in cell lines lacking cell–cell adhesion properties.

Søkilde et al.





only" miRNA cluster (Fig. 2D), it seems that the cell linesfall in 2, clinically distinct clusters, representing luminal(MCF7 and T47D) and basal (HS578T, BT549, and MDA-MB-231) breast cancer, in which the latter, basal triple-negative subtype, is the prognostically more unfavorable(26). The luminal cell lines are characterized by hormonereceptor expression [estrogen receptor (ESR1), progester-one receptor (PGR), and HER2] and luminal cytokeratins(e.g., CK7), whereas the basal type cancer cell lines arenegative for the hormone receptors and have expressionof basal cytokeratins (e.g., CK5 and CK6).Recently, miR-22 was found to target ESR1, which is in

agreementwith our data, inwhich lowmiR-22 expressionis observed when ESR1 levels are high (Fig. 3B). A similarinverse relationship between miR-222–miR-221 and ESRexpression was reported by Zhao and colleagues in bothbreast cancer cell lines and patients (27), consistent withour observations.Among the upregulated miRNAs in ESR-positive cell

lines are miR-200a–miR-200b–miR-429 (Fig. 3B), a clusterlinked to the regulation of E-cadherin via ZEB1/2 (28).miR-193b has been associated with the regulation ofurokinase-type plasminogen activator (uPA), an impor-tant prognostic factor in breast cancer development (29).We found that the E-cadherin mRNA level follows thelevel of miR-200, whereas the uPA transcript levels arenegatively correlated with miR-193b in support of thereported associations. On the other hand, miR-23b, whichwas found to correlate positively with uPA, has beenshown to downregulate uPA and MET in hepatocellularcarcinomas, although this interaction has not beenreported in breast cancer.Scott and colleagues showed that miR-125a/b regu-

lates ERRB2 at both transcript and protein levels (30),which is in agreement with our observations: miR-125b isexpressed at higher levels in the ERRB2-negative breastcancer cell lines than in ERBB2-positive cells (Fig. 3B).Unexpectedly, we found positive correlation between

miR-146 expression and epidermal growth factor recep-tor mRNA and protein levels (Supplementary Fig. S1).This is unlike the canonical model for miRNA action, inwhich high levels of miRNA repress their correspondingtargets protein levels.

CNS–gliomaThe CNS-specific miR-376, miR-377, and miR-654-3p

localize to a large cluster of miRNAs on chromosome 14,often referred to as themiR-379–miR-656 cluster, found inan imprinted region of transcripts. A functional role ofthe cluster has been shown in neuronal development (31);interestingly, the 3 candidates are expressed at higherlevels in the 2 estrogen receptor–positive cell lines, sug-gesting a role in hormone regulation (Fig. 3A).Two miRNA clusters, miR-100–let-7a-2–125b-1 and

miR-23b–27b–24-1, are expressed significantly higher inthe brain-derived cells than in the rest of the cell lines. ThemiR-100 cluster, along with the let-7 family, has beendescribed as brain specific in mammals and conserved

expression was also found in zebrafish (32). Interestingly,there are no reports on specific expression of the miR-23bcluster in the brain. The brain-specific miR-124 was unde-tected in all CNS-derived cell lines. This miRNA is sup-posedly important for the maturation of neurons and thelack of its expression may be explained by the lack ofdifferentiation of the CNS tumor cell lines. A miRNAreported to be involved in neurogenesis is miR-9; weobserved a distinct pattern of miR-9 and miR-9* expres-sion in the cell lines SF-295, SNB-19, and U251, whereas 3other CNS cell lines, namely, SF-268, SF-539, and SNB-75,seem to be mir-9 and miR-9* negative, suggesting differ-ences in differentiation between these 2 groups.

Colon cancerMost notably, the cells of colon origin gave rise tomany

colon-specific miRNAs, due to the homogenous expres-sion pattern observed in these cell lines. Interestingly,SW-620 and COLO-205 do not form tight epithelial cellclusters, which are characteristic of the other colon cancercell lines in the panel. Instead, they form round bipolarcells with lower cell–surface and cell–cell adhesion (33).Recently, the miR-200 family was reported to be con-nected to epithelial cluster formation (28), which is whywe decided to investigate this circuit in the cell lines.Although we did not identify any differences that couldaccount for the lower cell–cell adhesion properties of SW-620 and COLO-205 (Fig. 3C), we found thatmiR-142 (both5p and 3p)-–a miRNA characteristic for leukemia andinvolved in hematopoiesis—was highly expressed inthese 2 cell lines. We then looked for possible pathwaystargeted by miR-142 by applying DIANA-mirPath (34)and found that many adhesion pathways were enrichedfor target sites for miR-142 (3p and 5p), including TGF-bsignaling, adherence junction, tight junction, regulationof actin cytoskeleton, and extracellular matrix–receptorinteraction. Many of the proteins targeted by miR-142 areinvolved in several of these pathways and are importantregulators of adhesion (Supplementary Fig. S2).

We also detected differential expression of miRplussequences, among which we looked closer at hsa-miR-Plus-C1018, a sequence variant of hsa-miR-20a*. We wereunable to detect hsa-miR-20a* on the array, and becausethe miRplus-C1018 and miR-20a signals seem highlycorrelated (R2 ¼ 0.82), this could indicate a maturemiRNA-20a* different from the one annotated in miR-Base.

Lung cancerThe lung cancer cell lines displayed very diverse

expression profiles. From the clustering of the lung cancercell lines alone (Fig. 2B), HOP-92 andNCI-H460 clusteredfurthest away from the other cell lines and have pre-viously been described as large cell undifferentiatedcarcinomas (35). Another subcluster contained NCI-H322M and EKVX; both are epithelial cell lines with highexpression of miR-200 and miR-203. On the other hand,the squamous miRNA marker miR-205 was only high in






NCI-H322M, indicating that this cell line is of squamousorigin. The other cell line originally described as a squa-mous cell carcinoma, NCI-H226, had more resemblanceto a mesothelioma in our study, supporting the notion ofNutt and colleagues (36).

Recently, the metastatic potential of the NCI-60 lungcancer cells was measured in Matrigel and an mRNAsignature was found that correlated to overall survival(37). By applying the same split (into a low and highmetastatic potential group), we identified a short list ofmiRNAs (Table 2) which could separate high from lowmetastatic potential cell lines. We found miR-452 to beupregulated in the group with high metastatic potential,which is in agreement with the observation that miR-452is upregulated in lymph node positive bladder cancerpatients (38).

LeukemiaIn general, the miRNA expression pattern in the leuke-

mia cell lines is distinct from the other cell lines, which areall derived fromsolid tumors. Thus,weare able to confirmthe "hematologic" signature differentiating nonsolid fromsolid tumors, as suggested by Gaur and colleagues (5).However, we also observed more specific expressionpatterns. For example, miR-451 is exclusively expressedin theK562 cell line,which is a bipotential cell line, as it candifferentiate into either erythrocytes or megakaryocytes(39).miR-451 clusters togetherwithmiR-144,which is alsohighly expressed only in this cell line.

We also noticed high miR-223 expression in a subset ofthe leukemia cell lines (particularly HL-60 had extremelyhigh miR-223 levels), which have been shown to beactivated by the transcription factor C/EBP alpha. Themyeloma cell line RPMI_8226 and the lymphoma cell lineSR are both negative for miR-223, which is in agreementwith their lineage.

MelanomaThe melanoma cell lines displayed a distinct clustering

pattern (Fig. 2A), due to their unique expression of miR-204 and miR-211. These miRNAs have a similar seed

sequence but differ in 2 nucleotides on positions 17 and18. miR-211 is located in an intron of transient receptorpotential cation channel subfamilyMmember 1 (TRPM1)and shows high correlation to the expression of TRPM1mRNA (data not shown). TRPM1 is known to be tran-scriptionally regulated by microphthalmia-associatedtranscription factor (MITF), but the function of TRPM1in melanocytes is still unclear. The role of miR-204 andmiR-211 has been investigated in fetal human retinalpigment epithelium, in which they regulate the epithelialbarrier functions and target TGF-bR2 and SNAIL2 (40).We also observed high levels of miR-146a and miR-146bin the melanoma cell lines; these miRNAs are involved ininflammation and psoriasis (41), but their function inmelanocytes is still unclear. In addition, the melanomacell lines showed high expression of the testis- and ovary-specific miR-509–miR-514 cluster, which is located in agenomic region (Xq27) known to be highly expressed inmelanoma. We investigated (array-based comparativegenomic hybridization data; not shown)whether the highmiR-509–miR-514 expression could be explained by asimilar increase in copy number in the melanoma celllines, but this was not the case.

We found that theLOX-IMVIcell linedifferedmarkedlyfrom the other melanoma cell lines in its miRNA profile.This cell line is amelanotic and thus does not expresstypical melanocyte genes. LOX-IMVI is characterized bylowMITF expression and, in parallel, diminished expres-sion of melanocyte-specific genes and miRNAs, such asmiR-211, which is located within TRPM1.

Ovarian cancerAs the only cell line in the NCI-60 panel, IGROV1

expresses the miRNA cluster miR-302b–miR-302c–miR-302d–miR-367 (miR-302–miR-367), which is mainlyexpressed in pluripotent stem cells. Morphologic studiesof IGROV1 cells suggest that they are derived from anovarian carcinoma with multiple differentiations, endo-metroid to a large extent, together with some serous clearcells and undifferentiated foci (42). We find it likely thatsuch compartments of undifferentiated cells posses apluripotent precursor, which is reflected in the expressionof miR-302–miR-367. In addition, the IGROV1 cell line isonly in the NCI-60 panel carrying the BRCA1 mutation,which gives a very strong correlation (r ¼ 1, P ¼ 4E-68)between BRCA1 and the miR-302–miR-367cluster, butwhether this is a causal relationship, or just a coincidence,remains to be determined. The 2 closely related NCI/ADR-RES and OVCAR-8 cell lines both have a deletion in1p36.33 (15), which is reflected in the lack of expression ofthe miR-200a-200b-429 cluster, situated in this region.

Prostate cancerThe2prostate cancer cell lines bothhavehigh expression

of miR-31, and 2 members of the miR-96–182–183 clusterhave all been found to be differentially expressed in arecent analysis of prostate cancer and normal adjacentprostate (43). It has been debated whether the 2 prostate

Table 2.miRNAs significantly different betweenhigh and low metastatic potential lung cancercell lines

microRNA P log2 fold(high/low)

miR-342-3p 0.035 �1.3miR-374b 0.047 �0.7miR-7 0.009 �0.6miR-452* 0.031 0.5miR-452 0.011 1miRPlus-E1013 0.045 0.6miR-760 0.007 0.6miR-31 0.032 1.7

Søkilde et al.





cancer cell lines PC-3 and DU-145 are androgen receptorpositive or negative and whether or not they respond tohormone depletion (44). We therefore surveyed the litera-ture for evidence of miRNA-guided differentiationbetween androgen-dependent or -independent prostatecancer. Lee and colleagues (45) showed that miR-125bwas critical in cell proliferation and expressed at higherlevels in PC-3 than in DU-145, whereas Shi and colleagues(46) showed thatmiR-125b could beupregulatedby andro-gen receptor signaling. We also detect miR-125b at higherlevels in PC-3 than in DU-145, indicating that PC-3 couldhave active androgen signaling. The same pattern is seenfor miR-100, which was found to be upregulated in thestudy by Shi and colleagues (46). Interestingly, Shi andcolleagues speculate that theexpressionofmiR-125bcomesfrommir-125b-2, a precursor located on chromosome 21 inthe miR-99a–let-7c–125b-2 cluster. We found that expres-sion from PC-3 originates from another cluster, namely,miR-100–let-7a-2–125b-1 on chromosome 11, as the miR-125b expression data correlate better with this cluster. Wedetected only low levels ofmiR-99a, althoughmiR-100 andlet-7a are upregulated in PC-3 compared with DU-145.

Renal cancerIn the clustering between platforms (Fig. 2A), the renal

cancer cell lines all group together except SN12C, whichhas a distinct miRNA profile. Stinson and colleaguesobserved that SN12C is the only renal cell line positivefor neuronal system–specific enolases (35), which areexpressed in neurons but with cross-reactivity to cancersof other origins. The cell line is negative for many of therenal-specific miRNAs, including miR-30a and miR-886.We detected an interesting pattern in the expression ofmiR-200 in the renal cell lines: All of themiR-200–positiverenal cell lines were negative for miR-200c–141, whichhave been shown to be regulated by DNA methylation(47), suggesting that tissue-specific methylation may playa role in kidney development.

Correlation with protein expression and with drugsensitivitySee the Supplementary Results section and Figs. S5

and S6.

Conclusion

Several conclusions can be drawn from this study:First, despite major technical differences between theassays used to study miRNA expression (qRT-PCR,spotted arrays, Agilent arrays, LNA-enhanced arrays,total RNA vs. fractionated RNA), the correlation between

the 4 methods is high and serves to cross-validate theassays. Where there is a lack of correlation, this can beattributed to the data-trimming methods previouslyapplied, in which low values have been floored, whichrenders the correlation analysis imperfect.

Second, the biology of the NCI-60 cell lines is reflectedin their miRNA landscape, as tissue-specific miRNAsignatures are retained across all cell lines. For eachhistology, we also identified a number of characteristicmiRNAs that may explain the regulation of tissue-spe-cific markers; this conclusion, however, is in some casesbased on only a few cell lines and should therefore beinterpreted with caution.

Third, our miRNA–protein correlation analysisrevealed several known (e.g., the miR-200 regulatorypathway) and potentially novel miRNA–protein interac-tions, most of which are associatedwith cell adhesion andmotility.

Fourth, the correlation between miRNA expressionand drug sensitivity (GI50) data can help in the identifica-tion of new molecular targets and miRNA-based biomar-kers for personalized medication.

In conclusion, the miRNA landscape of the NCI-60 celllines assayed by our LNA-enhanced microarray platformis in good overall agreement with previous miRNAstudies. The miRNA expression data are available inGEO (accession number GSE26375). We envisage thatfuture studies on the NCI-60 cell lines will includeChIP-Seq data, next-generation sequencing, and func-tional follow-up studies on associated targets.

Disclosure of Potential Conflicts of Interest

This study was carried out when R. Søkilde, S. Møller, and T. Litmanwere employees of Exiqon A/S.

Acknowledgments

We thank Dr. SusanHolbeck for providing the NCI-60 cell lines and forthe supplementary correlation analysis data. We also thank Gitte Friis,Jesper Salomon, Hanni Willenbrock, Nana Jacobsen, and Ditte Andreasenfor excellent technical assistance.

Grant Support

This research was supported by the EU Sixth Framework Program "Onco-miR."

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 indicatethis fact.

Received June 29, 2010; revisedNovember 17, 2010; accepted January 5,2011; published OnlineFirst January 20, 2011.

References1. Monks A, Scudiero D, Skehan P, Shoemaker R, Paull K, Vistica D, et al.

Feasibility of a high-flux anticancer drug screen using a diverse panelof cultured human tumor cell lines. J Natl Cancer Inst 1991;83:757–66.

2. Available from: http://dtp.cancer.gov. [accessed 2011 Feb 11].3. Available from: http://discover.nci.nih.gov/cellminer/. [accessed

2011 Feb 11].

4. Blower PE, Verducci JS, Lin S, Zhou J, Chung JH, Dai Z, et al.MicroRNA expression profiles for the NCI-60 cancer cell panel. MolCancer Ther 2007;6:1483–91.

5. Gaur A, Jewell DA, Liang Y, Ridzon D, Moore JH, Chen C, et al.Characterization of MicroRNA expression levels and their biologicalcorrelates in human cancer cell lines. Cancer Res 2007;67:2456–68.






6. Liu H, D’Andrade P, Fulmer-Smentek S, Lorenzi P, Kohn KW,Weinstein JN, et al. mRNA and microRNA expression profiles ofthe NCI-60 integrated with drug activities. Mol Cancer Ther 2010;9:1080–91.

7. Olena AF, Patton JG. Genomic organization of microRNAs. J CellPhysiol 2010;222:540–5.

8. Esquela-Kerscher A, Slack FJ. Oncomirs—microRNAs with a role incancer. Nat Rev Cancer 2006;6:259–69.

9. Griffiths-Jones S, Saini HK, van DS, Enright AJ. miRBase: tools formicroRNA genomics. Nucleic Acids Res 2008;36:D154–8.

10. Nygaard S, Jacobsen A, LindowM, Eriksen J, Balslev E, Flyger H, et al.Identification and analysis of miRNAs in human breast cancer andteratoma samples using deep sequencing. BMC Med Genomics2009;2:35.

11. Søkilde R, Kaczkowski B, Barken KB, Mouritzen P, Møller S, Litman T.MicroRNA expression analysis by LNA enhanced microarrays.In:MicroRNA Profiling in Cancer: A Bioinformatics Perspective. Hack-ensack NJ: World Scientific Publishing Co. 2009; p. 23–46.

12. Available from: http://bioconductor.org. [accessed 2011 Feb 11].13. Available from: http://www.ebi.ac.uk/microarray-as/ae/. [accessed

2011 Feb 11].14. Available from: http://dtp.nci.nih.gov. [accessed 2011 Feb 11].15. Bussey KJ, Chin K, Lababidi S, Reimers M, Reinhold WC, Kuo WL,

et al. Integrating data on DNA copy number with gene expressionlevels and drug sensitivities in the NCI-60 cell line panel. Mol CancerTher 2006;5:853–67.

16. Benjamini Y, Hochberg Y. Controlling the false discovery rate: apractical and powerful approach to multiple testing. J R Stat Soc1995;57:289–300.

17. ShankavaramUT, ReinholdWC, Nishizuka S,Major S, Morita D, CharyKK, et al. Transcript and protein expression profiles of the NCI-60cancer cell panel: an integromic microarray study. Mol Cancer Ther2007;6:820–32.

18. Available from: http://www.mirbase.org. [accessed 2011 Feb 11].19. Available from: http://dtp.nci.nih.gov/mtargets/mt_index.html.

[accessed 2011 Feb 11].20. Castoldi M, Benes V, Hentze MW, Muckenthaler MU. miChip: a

microarray platform for expression profiling of microRNAs basedon locked nucleic acid (LNA) oligonucleotide capture probes. Meth-ods 2007;43:146–52.

21. Kuo WP, Jenssen TK, Butte AJ, Ohno-Machado L, Kohane IS.Analysis of matched mRNA measurements from two different micro-array technologies. Bioinformatics 2002;18:405–12.

22. Yang Y, Zhu M, Wu L, Zhou J. Assessment of data processing toimprove reliability of microarray experiments using genomic DNAreference. BMC Genomics 2008;9 Suppl 2:S5.

23. Lorenzi PL, Reinhold WC, Varma S, Hutchinson AA, Pommier Y,Chanock SJ, et al. DNA fingerprinting of the NCI-60 cell line panel.Mol Cancer Ther 2009;8:713–24.

24. Roschke AV, Tonon G, Gehlhaus KS, McTyre N, Bussey KJ, LababidiS, et al. Karyotypic complexity of the NCI-60 drug-screening panel.Cancer Res 2003;63:8634–47.

25. Heneghan HM, Miller N, Lowery AJ, Sweeney KJ, Newell J, Kerin MJ.Circulating microRNAs as novel minimally invasive biomarkers forbreast cancer. Ann Surg 2010;251:499–505.

26. Perou CM, Serlie T, Eisen MB, Van de Rijn M, Jeffrey SS, Rees CA,et al. Molecular portraits of human breast tumors. Nature2000;406:747.

27. Zhao JJ, Lin J, Yang H, Kong W, He L, Ma X, et al. MicroRNA-221/222 negatively regulates estrogen receptor alpha and is associatedwith tamoxifen resistance in breast cancer. J Biol Chem 2008;283:31079–86.

28. Gregory PA, Bert AG, Paterson EL, Barry SC, Tsykin A, Farshid G,et al. The miR-200 family and miR-205 regulate epithelial to mesench-ymal transition by targeting ZEB1 and SIP1. Nat Cell Biol 2008;10:593–6.

29. Li XF, Yan PJ, Shao ZM. Downregulation of miR-193b contributes toenhance urokinase-type plasminogen activator (uPA) expression andtumor progression and invasion in human breast cancer. Oncogene2009;28:3937–48.

30. Scott GK, Goga A, Bhaumik D, Berger CE, Sullivan CS, Benz CC.Coordinate suppression of ERBB2 and ERBB3 by enforced expres-sion of micro-RNA miR-125a or miR-125b. J Biol Chem 2007;282:1479–86.

31. Glazov EA, McWilliam S, Barris WC, Dalrymple BP. Origin, evolution,and biological role of miRNA cluster in DLK-DIO3 genomic region inplacental mammals. Mol Biol Evol 2008;25:939–48.

32. Wienholds E, KloostermanWP, Miska E, varez-Saavedra E, BerezikovE, de BE, et al. MicroRNA expression in zebrafish embryonic devel-opment. Science 2005;309:310–1.

33. Leibovitz A, Stinson JC, McCombs WB III, McCoy CE, Mazur KC,Mabry ND. Classification of human colorectal adenocarcinoma celllines. Cancer Res 1976;36:4562–9.

34. Available from: http://diana.cslab.ece.ntua.gr/pathways/. [accessed2011 Feb 11].

35. Stinson SF, Alley MC, Kopp WC, Fiebig HH, Mullendore LA, PittmanAF, et al. Morphological and immunocytochemical characteristics ofhuman tumor cell lines for use in a disease-oriented anticancer drugscreen. Anticancer Res 1992;12:1035–53.

36. Nutt JE, O’Toole K, Gonzalez D, Lunec J. Growth inhibition by tyrosinekinase inhibitors in mesothelioma cell lines. Eur J Cancer 2009;45:1684–91.

37. Hsu YC, Yuan S, Chen HY, Yu SL, Liu CH, Hsu PY, et al. A four-genesignature from NCI-60 cell line for survival prediction in non-small celllung cancer. Clin Cancer Res 2009;15:7309–15.

38. Veerla S, Lindgren D, Kvist A, Frigyesi A, Staaf J, Persson H, et al.miRNA expression in urothelial carcinomas: important roles of miR-10a, miR-222, miR-125b, miR-7 and miR-452 for tumor stage andmetastasis, and frequent homozygous losses of miR-31. Int J Cancer2009;124:2236–42.

39. Lu J, Guo S, Ebert BL, Zhang H, Peng X, Bosco J, et al. MicroRNA-mediated control of cell fate in megakaryocyte-erythrocyte progeni-tors. Dev Cell 2008;14:815–6.

40. Wang FE, Zhang C, Maminishkis A, Dong L, Zhi C, Li R, et al.MicroRNA-204/211 alters epithelial physiology. FASEB J 2010;24:1552–71.

41. Sonkoly E, Stahle M, Pivarcsi A. MicroRNAs and immunity: novelplayers in the regulation of normal immune function and inflammation.Semin Cancer Biol 2008;18:131–40.

42. Benard J, Da SJ, De Blois MC, Boyer P, Duvillard P, Chiric E, et al.Characterization of a human ovarian adenocarcinoma line, IGROV1, intissue culture and in nude mice. Cancer Res 1985;45:4970–9.

43. Schaefer A, JungM, Mollenkopf HJ, Wagner I, Stephan C, Jentzmik F,et al. Diagnostic and prognostic implications of microRNA profiling inprostate carcinoma. Int J Cancer 2010;126:1166–76.

44. Marchiani S, Tamburrino L, Nesi G, Paglierani M, Gelmini S, OrlandoC, et al. Androgen-responsive and -unresponsive prostate cancer celllines respond differently to stimuli inducing neuroendocrine differen-tiation. Int J Androl 2010;33:784–93.

45. Lee YS, Kim HK, Chung S, Kim KS, Dutta A. Depletion of humanMicro-RNA miR-125b reveals that it is critical for the proliferationof differentiated cells but not for the down-regulation of putativetargets during differentiation. J Biol Chem 2005;280:16635–41.

46. Shi XB, Xue L, Yang J, Ma AH, Zhao J, Xu M, et al. An androgen-regulated miRNA suppresses Bak1 expression and induces andro-gen-independent growth of prostate cancer cells. Proc Natl Acad SciU S A 2007;104:19983–8.

47. Vrba L, Jensen TJ, Garbe JC, Heimark RL, Cress AE, Dickinson S,et al. Role for DNA methylation in the regulation of miR-200c andmiR-141 expression in normal and cancer cells. PLoS One 2010;5:e8697.

Søkilde et al.





2011;10:375-384. Published OnlineFirst January 20, 2011.Mol Cancer Ther Rolf Søkilde, Bogumil Kaczkowski, Agnieszka Podolska, et al. Global microRNA Analysis of the NCI-60 Cancer Cell Panel

Updated version

10.1158/1535-7163.MCT-10-0605doi:

Access the most recent version of this article at:

Material

Supplementary

http://mct.aacrjournals.org/content/suppl/2011/01/21/1535-7163.MCT-10-0605.DC1

Access the most recent supplemental material at:

Cited articles

http://mct.aacrjournals.org/content/10/3/375.full#ref-list-1

This article cites 37 articles, 15 of which you can access for free at:

Citing articles

http://mct.aacrjournals.org/content/10/3/375.full#related-urls

This article has been cited by 5 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] at

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

Permissions

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

.http://mct.aacrjournals.org/content/10/3/375To request permission to re-use all or part of this article, use this link



http://mct.aacrjournals.org/lookup/doi/10.1158/1535-7163.MCT-10-0605

http://mct.aacrjournals.org/content/suppl/2011/01/21/1535-7163.MCT-10-0605.DC1

http://mct.aacrjournals.org/content/10/3/375.full#ref-list-1

http://mct.aacrjournals.org/content/10/3/375.full#related-urls

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

mailto:[email protected]

http://mct.aacrjournals.org/content/10/3/375


global microrna analysis of the nci-60 cancer cell panel · spotlight on molecular profiling global...

Documents