An Integrative Study

Published OnlineFirst January 12, 2010; DOI: 10.1158/1535-7163.MCT-09-0321

Molecular Spotlight on Molecular Profiling Cancer Therapeutics Multifactorial Regulation of E-Cadherin Expression: An Integrative Study

William C. Reinhold1, Mark A. Reimers1,2, Philip Lorenzi1,3, Jennifer Ho1, Uma T. Shankavaram1,4, Micah S. Ziegler1, Kimberly J. Bussey1,5, Satoshi Nishizuka1,6, Ogechi Ikediobi1,7, Yves G. Pommier1, and John N. Weinstein1,3

Abstract E-cadherin (E-cad) is an adhesion molecule associated with tumor invasion and metastasis. Its down- regulation is associated with poor prognosis for many epithelial tumor types. We have profiled E-cad in the NCI-60 cancer cell lines at the DNA, RNA, and protein levels using six different microarray platforms plus bisulfite sequencing. Here we consider the effects on E-cad expression of eight potential regulatory factors: E-cad promoter DNA methylation, the transcript levels of six transcriptional repressors (SNAI1, SNAI2, TCF3, TCF8, TWIST1, and ZFHX1B), and E-cad DNA copy number. Combined bioinformatic and pharmacological analyses indicate the following ranking of influence on E-cad expression: (1) E-cad promoter methylation appears predominant, is strongly correlated with E-cad expression, and shows a 20% to 30% threshold above which E-cad expression is silenced; (2) TCF8 expression levels correlate with (−0.62) and predict (P < 0.00001) E-cad expression; (3) SNAI2 and ZFHX1B expression levels correlate positively with each other (+0.83) and also correlate with (−0.32 and −0.30, respectively) and predict (P =0.03and 0.01, respectively) E-cad expression; (4) TWIST1 correlates with (−0.34) but does not predict E-cad expression; and (5) SNAI1 expression, TCF3 expression, and E-cad DNA copy number do not correlate with or predict E-cad expression. Predictions of E-cad regulation based on the above factors were tested and ver- ified by demethylation studies using 5-aza-2′-deoxycytidine treatment; siRNA knock-down of TCF8, SNAI2, or ZFHX1B expression; and combined treatment with 5-aza-2′-deoxycytidine and TCF8 siRNA. Finally, levels of cellular E-cad expression are associated with levels of cell-cell adhesion and response to drug treatment. Mol Cancer Ther; 9(1); 1–16. ©2010 AACR.

Introduction tions (2). Down-regulation of E-cad has been described in multiple carcinoma types during tumor progression E-cadherin (E-cad) is a transmembrane glycoprotein (3–6). Its down-regulation, a sign of poor prognosis for that functions to maintain stable cell-cell contacts in epi- multiple types of epithelial carcinomas (7–9), is asso- thelial cell types (1). It forms Ca+2-dependent homodi- ciated with increases in both invasion (3, 10, 11) and me- mers that bind to their counterparts in adjacent cells, tastasis (8, 12). In melanocytes, E-cad down-regulation resulting in the formation of intercellular adherens junc- and a concurrent up-regulation of N-cadherin lead to altered cell-cell relationships; whereas normal melanocytes interact primarily with keratinocytes, melanoma cells interact more strongly with melanocytes and fibro- Authors' Affiliations: 1Laboratory of Molecular Pharmacology, Center for blasts (13, 14). Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, Maryland; 2Virginia Commonwealth University, Richmond, Multiple single factors have been reported to regulate Virginia; 3Department of Bioinformatics and Computational Biology and E-cad expression in one or another cancer type (3–6, 15–23). Department of Systems Biology, M. D. Anderson Cancer Center, Houston, Texas; 4Radiation Oncology Branch, National Cancer However, those factors have not been studied together Institute, National Institutes of Health, Bethesda, Maryland; 5Clinical in combination as a system and across the spectrum of Translational Research Division, Translational Genomics Research cancers. Accordingly, to provide an integrative portrait Institute, Phoenix, Arizona; 6Department of Surgery, Iwate Medical University, School of Medicine, Uchimaru, Japan; and 7Helen Diller of E-cad regulation within and across cancer cell types, Family Comprehensive Cancer Center, San Francisco, California we used six different microarray platforms and bisulfite Corresponding Authors: William C. Reinhold, 9000 Rockville Pike, sequencing to assess eight potential E-cad regulatory fac- Building 37, Room 5056, Bethesda, MD 20892-4255. Fax: 301-496- tors in the NCI-60 human cancer cell line panel at the 9571. E-mail: [email protected] and John N. Weinstein, M.D. Anderson Cancer Center Systems Biology, 1515 Holcombe Boulevard, Houston, DNA, RNA, protein, and epigenetic levels. One very TX 77030-4009. Fax: 301-496-9571. E-mail: [email protected] practical motivation for understanding the complexities doi: 10.1158/1535-7163.MCT-09-0321 of E-cad regulation is the potential for reversing down- ©2010 American Association for Cancer Research. regulation of E-cad and restoring its function. That

www.aacrjournals.org 1

Reinhold et al.

might, in principle, be achieved through the use of agents (18).Inbrief,genomicDNA(5μg) from each cell line that reverse promoter region methylation or by knocking was treated with sodium bisulfite, amplified using down relevant transcriptional repressors. nested polymerase chain reaction primers, and se- The NCI-60 panel consists of 60 diverse human quenced. cancer cell lines used by the National Cancer Insti- tute's Developmental Therapeutics Program to screen Quantitation of Transcript Expression Using Four compounds for anticancer activity (24). The panel in- Microarray Platforms cludes leukemias, melanomas, and cancer cells of breast, We have previously described our processing and nor- central nervous system (glioma), colon, non-small cell malization of NCI-60 transcript expression data from pin- lung, ovarian, prostate, and renal origin. It constitutes the spotted cDNA arrays (Incyte, Inc., Palo Alto, CA; refs. most comprehensively profiled set of cells in existence, hav- 27, 28), Affymetrix Hu-6800 arrays (Affymetrix, Sunny- ing been analyzed at the DNA, RNA, protein, chromosom- vale, CA; ref. 29), Affymetrix HG-U95 arrays (30), and al, metabolomic, and pharmacological levels (25). Profiling Affymetrix HG-U133 (30). The data from those and the of the NCI-60 has been considered a forerunner of The Can- other molecular profiling studies are available in a cer Genome Atlas Project8, which is restricted to the nucleic queryable relational database (CellMiner).10 Further in- acid level but in the more difficult context of clinical tu- formation for these microarrays is available at http:// mors. www.broadinstitute.org/mpr/NCI60/NCI60.html for To test whether the correlative relationships uncovered Hu-6800, and at the Gene Expression Omnibus11 for are causal at the molecular level and whether they pro- the cDNA array, HG-U95, and HG-U133 (identifiers vide the basis for strategies to up-regulate E-cad on a cell GDS1761, GSE5949, and GSE5720, respectively). type-specific basis, we followed up with siRNA knock- down and 5-azacytidine demethylation experiments. Quantitation of E-cad Protein Expression Using This overall integromic (26) approach, supported by Reverse-Phase Lysate Arrays functional data, yields a picture of the multifactorial reg- Our methods for quantitation of proteins using re- ulation of E-cad expression. It provides the ability to pre- verse-phase lysate arrays have been described previously dict rationally and prospectively, independent of cancer (31). Further information for this array is available at the tissue of origin type, whether E-cad will be successfully Gene Expression Omnibus, identifier GSE5501. up-regulated by a given treatment. Comparative Genomic Hybridization Array The arrays comprised 450 cancer-related BAC, PAC, Materials and Methods and P1 clones printed in quadruplicate (32). The clones and their genomic locations have been defined prev- 12 Cell lines and Cell Culture iously. Further information for this array is available at The NCI-60 cells were obtained from the National Array Express at http://www.ebi.ac.uk/microarray-as/ Cancer Institute Frederick Cancer DCTD Tumor/Cell ae/browse.html?keywords=e-geod-5720. 9 Line Repository andculturedasdescribedpreviously 5-Aza-2′-deoxycytidine Inhibition of DNA (18, 27). All culture flasks were examined by microscope ∼ Methylation for anomalies, and the cells were harvested at 80% We modified a previously described protocol (33) to confluence. It is important to note that all cell culture, study inhibition of DNA methylation by 5-AC. Briefly, harvests, and purifications were performed by a single re- 4,000 to 5,000 cells were seeded in 96-well cell culture searcher to maximize interoperability of the data. clusters on day 0, treated with 0.1 to 2 mg/mL 5-aza- RNA and DNA Isolation 2′-deoxycytidine (5-AC) (Sigma, St. Louis, MO) on days RNA was isolated as we have described previously (18, 1, 3, and 5, and washed to remove the drug on days 2 27). Briefly, total RNA was purified using the RNeasy and 4. The cells were lysed on day 6 using Lysis mixture Midi Kit (Qiagen Inc., Valencia, CA) according to the (Panomics, Inc., Fremont, CA). manufacturer's instructions. Genomic DNA was purified using either the QIAamp DNA Blood Maxi Kit or RNA Interference the Blood & Cell Culture DNA Maxi Kit (Qiagen) ac- We used the synthetic siRNAs siTCF8.1, siTCF8.2, cording to manufacturer's instructions (18). siSNAI2.1, siSNAI2.2, siZFHX1B.1, and siZFHX1B.2 (Qiagen) to inhibit expression of the transcription fac- Sodium Bisulfite DNA Modification, Polymerase tors TCF8, SNAI2, and ZFHX1B, as described previous- Chain Reaction Amplification, and Sequencing ly (34). Our bisulfite-sequencing protocol for the minimal promoter region of E-cad has been described previously

10 See http://discover.nci.nih.gov/. 8 http://cancergenome.nih.gov/ 11 http://www.ncbi.nlm.nih.gov/geo/ 9 See http://dtp.nci.nih.gov/. 12 See http://cc.ucsf.edu/gray/public.

2 Mol Cancer Ther; 9(1) January 2010 Molecular Cancer Therapeutics

E-cadherin Regulation in NCI-60

Branched DNA Assay for Quantitation of RNA were generally higher for epithelial cell types (breast, colon, Expression non-small cell lung, ovarian, prostate, and renal) than for E-cad, TCF8, ZFHX1B, SNAI2, and PPIB mRNA levels nonepithelial ones (gliomas, leukemias, and melanomas). were assayed in 5-AC and siRNA studies using a branched Table 1 also contains our array comparative genomic DNA assay (Panomics) as described previously (34). hybridization DNA copy number data (ploidy-relative) for E-cad (32) and mRNA expression levels from HG- Statistical Analyses U133 Affymetrix arrays for the E-cad transcriptional We determined 95% confidence intervals for the cor- regulatoryfactorsSNAI1,SNAI2,TCF3,TCF8,TWIST1, relations between E-cad methylation, expression, and and ZFHX1B (30). Gray blocks indicate DNA methyla- DNA copy number via bootstrap analysis with 10,000 tion levels (18) and transcriptional repressor transcript resamplings. Unless otherwise stated, all calculations levels associated at statistically significant levels (two- were done using R.13 Confidence intervals and P values tailed P < 0.05 by bootstrap with 10,000 iterations) with for prediction of relationships between expression of E-cad down-regulation. E-cad and the transcriptional repressors were calculated via multivariate linear regression for the cell lines with E-cad Methylation Is Inversely Correlated with ≥30% mean methylation. P values for all correlations were E-cad mRNA and Protein Expression but not calculated to test the null hypothesis of no association. with E-cad DNA Copy Number Phase Contrast Microscopy The mean methylation pattern (Table 1) for E-cad cor- Cell images (at 25× magnification) were obtained using related inversely with E-cad transcript levels measured a Zeiss Axiovert 25 inverted phase contrast microscope using all four microarray types (r = −0.49, −0.45, −0.52, with a halogen bulb and a Canon DS126071, Rebel XL and −0.38 for the Table 1 Hu6800, cDNA, HG-U95, and digital camera. In some cases, the contrast was increased HG-U133 microarrays, respectively). It also correlated for better visibility. inversely with the protein (r = −0.49) pattern (data from Table 1). All of those correlations were statistically sig- Drug Activity Data nificant (bootstrap two-tailed P < 0.05; data not shown). The mechanism-of-action drug set used in this study The ploidy-relative DNA copy number (Table 1), how- has been described previously (32). It comprises 118 com- ever, showed no significant association with mean pounds whose mechanism is presumptively known. The methylation, transcript expression, or protein expres- activities are growth inhibition 50% concentrations, ex- − sion. Transcript level data from the four array types pressed as log(GI50), and are determined as described were correlated with each other at statistically signif- previously (27, 28). icant levels, with a mean r =+0.86-thatis,thefour transcript expression platforms corroborated each other, Results indicating that the measurements were generally robust. The protein and transcript levels also correlated with r Multiple Factors Can Regulate E-cad Expression each other ( = +0.72, +0.81, +0.86, and +0.75 for the Table 1 summarizes the molecular profiles used in this Hu6800, cDNA, HG-U95, and HG-U133 microarrays, study. We classify MDA-MB-435 and its ERBB2-transfec- respectively) at statistically significant levels. tant MDA-N as melanomas based on multiple molecular mRNA lLevels of Some Transcriptional Regulators and pharmacological profiles from our laboratory and Are Inversely Correlated with E-cad mRNA Levels others (18, 27–29, 35–37). Similarly, we classify OV- CAR8/ADR-RES (previously called MCF7/ADR-RES Pearson's correlation coefficients relating E-cad and and NCI/ADR-RES) as ovarian based on compelling ev- E-cad transcriptional repressor expression (Table 1) in the NCI-60 are shown in Table 2. Bold type indicates cor- idence from our spectral karyotyping (38), comparative P genomic hybridization (32), single-nucleotide polymorphism relations with < 0.05 (without multiple comparisons analysis (39), and microsatellite fingerprinting (40) that it correction). Significant negative correlations, consistent is a (drug-resistant) derivative of OVCAR-8. across the five E-cad expression platforms, were seen The E-cad promoter methylation levels in Table 1 are ex- for the NCI-60 for SNAI2, TCF8, TWIST1, and ZFHX1B. pressed as the mean methylation percentage for all 29 CpG SNAI2 and E-cad expression were negatively correlated cytosines in the minimal promoter region (18). The E-cad for breast cancer. TCF8 and E-cad were negatively corre- mRNA levels were determined using our data from lated for breast, lung, and ovarian cancers. TWIST1 and 9,706-clone cDNA arrays (27, 28) and three different Affy- E-cad expression were negatively correlated for the mel- metrix chip types: HU-6800 (29), HG-U95 (30), and HG- anomas. All of the subpanel relationships, however, were U133 (30). E-cad protein levels were determined using based on small numbers of cell types, and they must be reverse-phase lysate arrays (31). E-cad expression levels considered in light of multiple comparisons issues. Figure 1 shows the statistically significant relationships between E-cad and transcriptional repressor transcript expression levels (Table 1). We previously noted a thresh- 13 See www.r-project.org. old of 20% to 30% mean E-cad methylation above which

www.aacrjournals.org Mol Cancer Ther; 9(1) January 2010 3

Reinhold et al.

Table 1. E-cadherin methylation, transcript, protein, and DNA copy number, and six E-cad transcriptional repressor transcript profiles Cell line* E-cad profiles Mean Transcript levels Protein Array comparative methylation‡ levels§ genomic hybridization∥ Hu6800¶ 9706 cDNA** U95†† U133‡‡

BR:BT-549 54 4.91 0.51 4.32 4.82 −1.46 −0.15 BR:HS578T 47 4.58 0.47 4.32 4.91 −0.69 NA∥∥∥ BR:MCF7 21 9.39 2.82 9.01 8.39 1.63 0.11 BR:MDA-MB-231 13 4.32 0.21 4.32 4.87 −2.62 −0.37 BR:T47D 6 8.96 2.96 9.44 8.46 1.66 −0.04 CNS:SF-268 48 4.70 0.40 4.32 4.56 −2.06 −0.26 CNS:SF-295 43 4.32 0.66 4.32 4.99 −1.37 0.33 CNS:SF-539 73 4.32 0.75 4.32 4.85 −1.73 −0.13 CNS:SNB19 86 5.17 0.46 4.30 4.92 −1.16 0.40 CNS:SNB-75 16 4.39 0.62 4.32 4.81 −1.92 −0.03 CNS:U251 93 5.29 0.39 4.30 4.83 −1.74 −0.10 CO:COLO205 8 9.52 2.34 9.50 7.77 1.08 −0.14 CO:HCC-2998 11 7.81 1.46 8.58 6.13 1.56 0.33 CO:HCT-116 9 6.29 0.60 5.90 4.97 −0.16 0.03 CO:HCT-15 11 8.08 2.22 8.99 6.47 1.10 0.01 CO:HT29 8 5.75 1.53 8.50 6.05 1.51 −0.01 CO:KM12 6 7.55 1.27 8.04 6.33 0.32 0.03 CO:SW-620 23 6.75 0.80 7.02 5.20 0.07 0.12 LC:A549-ATCC 14 6.89 0.88 6.90 5.12 −0.52 0.15 LC:EKVX 7 7.72 1.13 6.90 5.58 −0.03 0.33 LC:HOP-62 21 4.32 0.41 4.32 4.88 −1.40 −0.15 LC:HOP-92 7 4.32 0.30 4.32 4.71 −1.76 0.21 LC:NCI-H226 12 6.11 0.46 4.30 4.86 −1.44 0.00 LC:NCI-H23 12 4.46 0.27 4.30 4.69 −2.31 0.11 LC:NCI-H322M 9 8.08 2.36 9.25 6.63 1.80 NA∥∥∥ LC:NCI-H460 41 4.32 0.58 4.30 4.90 −0.55 0.46 LC:NCI-H522 15 4.46 0.34 4.32 4.94 −1.17 NA∥∥∥ LE:CCRF-CEM 88 4.75 0.39 4.32 4.99 −2.58 0.22 LE:HL-60 85 4.46 0.35 4.32 5.24 −2.59 0.44 LE:K-562 98 5.83 0.36 4.32 4.97 −3.29 0.06 LE:MOLT-4 97 4.32 0.47 4.32 5.03 −2.13 0.00 LE:RPMI-8226 39 5.39 0.84 4.32 5.06 −1.48 0.40 LE:SR 95 4.32 0.36 4.32 4.84 −1.51 −0.02 ME:LOXIMVI 85 4.32 0.52 4.32 4.79 −1.64 −0.17 ME:M14 20 4.32 0.69 4.32 5.03 −2.52 −0.05 ME:MALME-3M 12 7.28 1.00 9.04 6.23 −0.28 −0.44 ME:MDA-MB-435¶¶¶ 84 4.32 0.33 4.32 5.14 −1.44 −0.42 ME:MDA-N¶¶¶ 84 4.32 0.35 4.32 4.80 −1.49 −0.46 ME:SK-MEL-2 26 4.32 0.40 4.32 5.11 −1.74 −0.03 ME:SK-MEL-28 25 5.25 0.33 4.32 4.76 −2.37 −0.47 ME:SK-MEL-5 80 6.89 0.58 5.13 5.23 −2.13 −0.23 ME:UACC-257 7 7.34 1.17 7.48 5.58 −0.64 −0.04 ME:UACC-62 59 4.32 0.45 4.32 4.98 −3.05 −0.29 OV:IGROV1 24 4.95 0.61 4.32 4.93 −1.97 0.03 OV:OVCAR-3 8 7.20 0.99 6.83 5.47 −1.70 −0.60 OV:OVCAR-4 5 7.22 1.20 7.43 5.56 1.80 −0.48 OV:OVCAR-5 13 5.70 0.41 4.32 4.81 −1.91 −0.20

(Continued on the following page)

4 Mol Cancer Ther; 9(1) January 2010 Molecular Cancer Therapeutics

E-cadherin Regulation in NCI-60

Table 1. E-cadherin methylation, transcript, protein, and DNA copy number, and six E-cad transcriptional repressor transcript profiles (Cont'd)

E-cad transcriptional repressors transcript levels† SNAI1§§,∥∥ SNAI2∥∥,¶¶ TCF3∥∥,*** TCF8∥∥,††† TWIST1‡‡‡ ZFHX1B§§§

5.75 7.09 6.28 7.59 6.82 3.68 5.59 10.11 5.95 8.32 6.27 4.42 5.74 4.38 6.35 3.63 3.84 3.02 5.41 8.16 5.91 7.42 3.49 3.32 5.48 4.22 6.22 3.63 3.86 3.09 5.36 6.98 5.86 7.04 6.15 5.11 6.33 7.32 5.89 7.18 6.30 3.37 6.61 9.91 5.97 7.42 5.78 4.06 5.64 6.00 6.18 6.22 6.26 3.49 6.57 7.94 6.37 7.01 5.65 4.93 5.51 6.39 5.98 5.61 5.43 3.84 5.97 4.26 6.11 3.81 3.74 3.09 5.98 4.37 6.50 3.65 4.02 2.98 6.19 4.44 6.42 4.52 4.16 3.12 5.65 4.21 6.30 3.57 3.66 3.17 5.72 4.36 6.26 3.70 3.64 3.19 5.83 4.56 6.36 4.17 3.72 3.09 6.17 4.97 6.18 5.18 3.61 3.12 5.98 4.38 6.58 5.77 3.99 3.13 5.93 4.50 6.34 6.05 3.72 3.08 6.14 5.35 6.14 6.14 5.79 3.50 5.89 7.75 6.57 6.95 3.99 3.12 5.69 7.32 6.14 6.37 5.55 3.74 5.61 5.20 5.93 6.67 5.85 3.59 5.55 4.84 6.16 3.51 4.36 3.07 6.09 4.95 6.50 5.66 3.64 3.42 6.07 4.32 6.24 6.74 5.09 3.17 5.62 4.39 6.39 8.17 4.08 3.19 6.07 4.29 6.50 6.12 4.37 4.43 6.15 4.32 6.06 4.73 3.80 5.19 5.61 4.48 6.36 7.42 3.97 3.16 5.94 5.62 6.38 5.34 3.69 3.15 5.41 4.73 6.03 6.43 5.54 4.35 5.66 9.98 6.11 7.54 5.06 4.00 5.74 8.90 6.22 3.72 5.97 4.93 5.81 8.14 6.32 3.88 4.48 4.78 5.95 9.57 6.16 5.08 5.70 4.60 5.73 9.65 5.92 5.71 6.03 5.69 6.06 6.46 6.57 4.44 6.32 4.09 5.51 9.86 5.96 4.07 5.71 5.85 5.81 8.90 6.43 4.36 5.64 4.31 5.76 10.00 6.29 3.52 4.70 5.07 6.47 8.55 6.44 5.36 5.74 4.10 5.86 4.49 6.61 5.87 3.60 3.06 6.23 4.42 6.28 3.85 4.98 3.06 6.19 4.27 6.48 3.75 5.09 3.19 5.69 5.42 6.37 6.40 3.67 3.29

(Continued on the following page)

www.aacrjournals.org Mol Cancer Ther; 9(1) January 2010 5

Reinhold et al.

Table 1. E-cadherin methylation, transcript, protein, and DNA copy number, and six E-cad transcriptional repressor transcript profiles (Cont'd)

Cell line* E-cad profiles Mean Transcript levels Protein Array comparative methylation‡ levels§ genomic hybridization∥ Hu6800¶ 9706 cDNA** U95†† U133‡‡

OV:OVCAR-8 85 5.21 0.32 4.32 4.82 −1.78 0.22 OV:OVCAR8/ADR-RES**** 89 6.19 0.32 4.32 4.96 −2.03 0.24 OV:SKOV3 8 5.36 0.72 4.32 5.32 −0.92 0.15 PR:DU-145 14 6.74 0.47 5.58 5.19 −0.16 0.04 PR:PC-3 15 6.75 0.69 4.83 5.21 −2.18 −0.07 RE:786-0 87 4.58 0.44 4.32 4.69 −1.87 0.10 RE:A498 93 4.46 0.53 4.32 4.61 −1.61 0.34 RE:ACHN 10 5.81 0.58 5.51 5.01 −2.16 0.35 RE:CAKI-1 33 6.74 0.78 4.64 4.97 −2.30 0.66 RE:RXF-393 25 4.32 0.33 4.32 4.73 −1.62 −0.14 RE:SN12C 74 5.55 0.29 4.32 4.96 −1.25 0.72 RE:TK-10 99 4.32 0.27 4.32 5.02 −2.25 −0.03 RE:UO-31 12 5.78 0.43 4.32 4.70 −1.73 0.29

(Continued on the following page)

E-cad expression was not detected (18). Four of the repressive. An F-test was used to check whether addition of sors,SNAI2,TCF8,TWIST1,andZFHX1B(Fig.1A-D), SNAI2 or ZFHX1B leads to a significant decrease in showed L-shaped relationships with E-cad expression. the unexplained variance of E-cad expression. Because Brackets indicate the approximate levels of repressor the levels of SNAI2 and ZFHX1B were significantly cor- (>5.16, 5.18, 5.25, and 3.30, respectively) above which related (+0.83) over the cell lines with <30% methyla- E-cad expression was predominately at background tion, effects of the two on E-cad expression were not levels. SNAI1 and TCF3 (data not shown) did not reliably linearly separable. This significant (P < 0.0001) correla- predict E-cad repression. The SNAI2, TCF8, and ZFHX1B tion suggests coordinate regulation. The statistical calcu- E-cad down-regulation regions (as demarcated by brack- lations were based on standard parametric t tests for ets) had measurable E-cad expression in two, one, and the null hypothesis that the true regression slope was two cell lines, respectively. Those three E-cad-expressing 0. The P values were not significant for the other tran- cell lines (MALME-3M and UACC-257 in Fig. 1A and D; scriptional regulator combinations. The relationships EKVX in Fig. 1B) indicate a lack of complete repression of between E-cad expression and TCF8, SNAI2, and E-cad by the transcriptional repressors when expressed ZFHX1B (with SNAI2 and ZFHX1B not being separable) within the (bracketed) ranges. were also maintained qualitatively when all cell lines, regardless of mean methylation level, were included TCF8, ZFHX1B, and SNAI2 Predict E-cad Expression (data not shown). For reasons considered in the Discussion section, it appears likely that the transcription factor effects are 5-AC Can Up-regulate E-cad Expression secondary to methylation status in regulating E-cad Because promoter methylation of E-cad is strongly expression.Hence,forthefollowingcalculation,we associated with its down-regulation, we choose the DNA adopted a model in which the transcriptional repressors demethylation agent 5-AC to determine if we could up- cause differences in transcription rates only when mean regulate E-cad expression. Figure 2A shows the expression methylation is low enough to permit E-cad expression levels of E-cad transcript in 10 cell lines as measured via (<30%) (ref. 18). When we then tested for a linear rela- branched DNA assay, with or without 5-AC treatment. tionship between TCF8 and E-cad transcript levels for The lines are ordered based on a combination of their ac- the cell lines with permissive methylation levels, we tual and predicted levels of E-cad up-regulation following found P < 0.00001 and a slope (effect size) of 0.50 when 5-AC treatment. E-cad was up-regulated by 5-AC in the both variables were represented on a log2 scale. After three cell lines (SW-620, TK-10, and IGROV1) that exhib- fitting that model, we tested the residuals for additional ited a combination of repressive levels of methylation effects of SNAI2 (P= 0.03; effect size 0.16) and ZFHX1B (≥20%) and zero or one transcriptional repressor at levels (P = 0.01; effect size 0.4) when methylation is permis- associated with low levels of E-cad (Fig. 1). E-cad was not

6 Mol Cancer Ther; 9(1) January 2010 Molecular Cancer Therapeutics

E-cadherin Regulation in NCI-60

Table 1. E-cadherin methylation, transcript, protein, and DNA copy number, and six E-cad transcriptional repressor transcript profiles (Cont'd) (Cont'd)

† E-cad transcriptional repressors transcript levels† §§,∥∥ ∥∥,¶¶ ∥∥,*** ∥∥,††† ‡‡‡ §§§ SNAI1§§,∥∥ SNAI2∥∥,¶¶ TCF3∥∥,*** TCF8∥∥,††† TWIST1‡‡‡ ZFHX1B§§§

5.81 4.97 6.18 6.53 3.69 3.57 6.14 5.95 6.24 5.29 5.77 3.46 5.86 5.95 6.46 5.50 5.09 3.13 5.98 4.60 6.34 5.41 3.73 3.22 5.89 7.23 6.52 5.48 5.67 3.33 6.25 6.57 6.18 5.97 3.58 3.40 6.49 4.73 6.43 5.50 3.64 3.52 5.61 4.73 6.20 5.76 3.56 3.70 6.27 4.48 6.34 5.81 5.10 3.36 6.06 7.00 6.12 7.04 3.50 4.61 5.76 5.42 6.43 7.03 5.58 3.39 6.25 5.41 6.52 4.68 3.83 3.30 5.86 5.16 6.19 5.95 4.68 3.46

*Tissues of origin are breast (BR), central nervous system (CNS), colon (CO), non-small cell lung cancer (LC), leukemia (LE), melanoma (ME), ovarian (OV), prostate (PR), and renal (RE). †U133 transcript measurement, log2-transformed. Data processed using the RMA algorithm. Boldface indicates levels of transcriptional repressors associated with low levels of E-cad expression (Fig. 1). ‡Mean methylation profile of E-cad promoter. Boldface and boldface/underscore indicate levels above or within the 20% to 30% E-cad expression threshold, respectively. §Protein levels expressed as log2 of relative protein concentration from reverse phase protein lysate microarrays. ∥Ploidy-relative DNA content. Values are log2-transformed. Array spot number 1485. BAC number RMC16P004. ¶Gene accession number Z35402 (Hu6800 oligonucleotide arrays). **Clone ID 416386 (cDNA microarrays). Values are ratios from cohybridization of an individual cell type and a 12-cell line pool. ††Fragment name 977_s_at (Affymetrix HG-U95 oligonucleotide microarray, log2-transformed). ‡‡Affymetrix fragment identifier 201130_s_at. §§Common name Snail, Affymetrix identifier 219480_at, Exemplar Sequence Accession Number NM_005985. ∥∥Transcriptional repressors chosen as being representative from multiple probe sets. ¶¶Common name SLUG, Affymetrix identifier 213139_at, Exemplar Sequence Accession Number AI572079. ***Common name E12/E47 gene product, Affymetrix identifier 216647_at, Exemplar Sequence Accession Number AL117663. HG- U133 transcript measurements, log2-transformed. †††Common name ZEB1, Affymetrix identifier 212764_at, Exemplar Sequence Accession Number AI806174. ‡‡‡Common name ACS3, Affymetrix identifier 213943_at, Exemplar Sequence Accession Number NM_000474. §§§Common name ZEB2, Affymetrix identifier 203603_s_at, Exemplar Sequence Accession Number NM_014795. ∥∥∥Data not available. ¶¶¶Cell lines considered to be melanomas. See Results. ****Cell line considered to be a doxorubicin-selected, resistant derivative of OVCAR-8. Previously named MCF7/ADR-RES and NCI/ADR-RES. See Results.

up-regulated by 5-AC in any of the four cell lines (A498, Effects of TCF8-siRNA on E-cad Expression NCI-H460, OVCAR-8, and SK-MEL-28) with repressive The transcript expression of TCF8 had the highest cor- levels of methylation plus two or more transcriptional re- relation with the E-cad expression profiles (Table 2), a pressors at levels associated with low E-cad expression striking L-shaped data distribution in the plot of E-cad (Fig. 1). It was also not up-regulated in any of the three cell versus TCF8 expression (Fig. 1B), and the strongest pre- lines with nonrepressive levels of E-cad methylation dictive value among the six transcriptional repressors for (A549-ATCC, HCT-116, and HT29). The latter three cells E-cad expression (P < 0.00001; parametric t test). We linesservedascontrolsforsecondaryeffectsduetothe therefore used siRNAs (siTCF8.1 and siTCF8.2) to test nonspecificity of 5-AC's demethylating activity. the effect of TCF8 down-regulation on E-cad expression

www.aacrjournals.org Mol Cancer Ther; 9(1) January 2010 7

Reinhold et al.

Table 2. Correlations between E-cad expression and methylation parameters, and transcript factor expression levels from HG-U133 DNA microarray analyses

NCI-60 BR CNS CO LC LE ME OV RE

SNAI1 Methylation 0.07 0.62 −0.49 0.50 0.51 −0.27 0.12 −0.05 0.62 Hu6800 −0.12 0.21 −0.81 −0.13 −0.43 0.69 −0.25 0.85 −0.40 9706 −0.13 0.15 0.96 −0.68 −0.45 0.14 −0.15 0.63 0.17 U95 −0.13 0.07 0.59 −0.69 −0.40 NA* −0.13 0.76 −0.54 U133 −0.17 0.09 0.47 −0.48 −0.43 0.62 0.05 0.73 −0.41 Protein −0.11 0.27 0.02 −0.64 −0.21 −0.58 −0.34 0.42 −0.18 SNAI2 Methylation 0.17 0.59 −0.26 0.76 −0.23 −0.90 0.31 0.37 0.11 Hu6800 −0.39 −0.91 −0.77 −0.39 −0.28 0.26 0.07 −0.39 −0.63 9706 −0.32 −0.90 0.84 −0.71 −0.35 0.93 0.06 −0.63 −0.60 U95 −0.34 −0.90 0.68 −0.58 −0.42 NA* −0.06 −0.69 −0.43 U133 −0.32 −0.90 0.01 −0.60 −0.37 −0.08 −0.28 −0.38 −0.37 Protein −0.35 −0.76 −0.33 −0.63 −0.44 0.75 0.05 −0.39 0.34 TCF3 Methylation −0.17 0.00 −0.30 −0.29 0.21 −0.40 −0.26 −0.66 0.69 Hu6800 0.19 0.72 0.03 −0.29 0.04 −0.28 0.24 −0.18 −0.07 9706 0.10 0.69 0.12 −0.37 −0.02 0.30 0.27 0.43 −0.14 U95 0.15 0.64 −0.16 −0.30 0.08 NA* 0.21 0.09 −0.24 U133 0.12 0.65 0.22 −0.42 −0.09 0.84 0.41 0.24 0.39 Protein 0.14 0.67 0.08 0.05 0.14 0.03 −0.05 0.36 −0.05 TCF8 Methylation 0.36 0.69 −0.64 0.70 −0.04 0.32 0.77 0.44 −0.35 Hu6800 −0.66 −0.98 −0.94 −0.37 −0.67 −0.70 −0.55 −0.90 0.11 9706 −0.62 −0.98 0.74 −0.77 −0.94 −0.27 −0.48 −0.88 −0.19 U95 −0.70 −0.99 0.94 −0.77 −0.84 NA* −0.46 −0.91 −0.13 U133 −0.62 −0.99 −0.11 −0.65 −0.91 −0.07 −0.52 −0.92 −0.19 Protein −0.53 −0.87 −0.24 −0.85 −0.92 0.13 −0.15 −0.62 0.71 TWIST1 Methylation 0.20 0.97 −0.12 −0.26 −0.14 0.38 0.29 0.06 −0.10 Hu6800 −0.35 −0.52 −0.17 −0.10 −0.37 −0.61 −0.72 0.61 0.68 9706 −0.32 −0.52 0.02 −0.41 −0.38 −0.47 −0.78 0.34 0.10 U95 −0.40 −0.59 0.17 −0.48 −0.43 NA* −0.84 0.37 −0.19 U133 −0.32 −0.59 0.12 −0.33 −0.33 −0.46 −0.70 0.60 0.35 Protein −0.30 −0.30 0.51 −0.20 −0.52 0.41 −0.69 0.33 0.26 ZFHX1B Methylation 0.28 0.77 −0.57 −0.02 0.33 0.45 −0.26 0.87 −0.44 Hu6800 −0.35 −0.71 −0.25 −0.42 −0.43 0.32 0.13 −0.23 −0.33 9706 −0.34 −0.68 −0.15 0.07 −0.51 −0.58 0.01 −0.69 −0.18 U95 −0.31 −0.72 0.49 −0.08 −0.62 NA* 0.09 −0.41 0.04 U133 −0.30 −0.71 −0.86 −0.14 −0.53 −0.10 −0.09 −0.60 −0.28 Protein −0.42 −0.44 −0.90 −0.13 −0.57 −0.53 0.14 −0.23 0.25

NOTE: Significance of correlations at P < 0.05, without multiple comparisons correction, for blocks with bold type. Tissues of origin are breast (BR), central nervous system (CNS), colon (CO), non-small cell lung cancer (LC), leukemia (LE), melanoma (ME), ovarian (OV), prostate (PR), and renal (RE). All data profiles are from Table 1. *Correlation not available due to an absence of pattern for E-cad expression in HG-U95 LEs.

(Fig. 2B). The first two cell lines in the figure (A549-ATCC OVCAR-8 and TK-10, which both have E-cad DNA and DU-145) were among the best candidates for E-cad methylation levels above the 30% threshold and levels up-regulation, and both displayed E-cad up-regulation of SNAI2 and ZFHX1B associated with low-level E-cad when TCF8 expression was knocked down. As predicted, expression (Fig. 1), failed to up-regulate E-cad when

8 Mol Cancer Ther; 9(1) January 2010 Molecular Cancer Therapeutics

E-cadherin Regulation in NCI-60

TCF8 was knocked down. T47D and HCT-116 both had vels of ZFHX1B or SNAI2 associated with low E-cad levels of TCF8 not associated with low E-cad expression, expression. We therefore tested the effects on E-cad ex- and thus provided controls for nonspecific effects. As pression of ZFHX1B and SNAI2 knock-down by siRNA predicted, neither cell line showed any change in either using the next best possibilities: MALME-3M, UACC-257, TCF8 or E-cad expression levels. and ACHN for ZFHX1B (Fig. 2C), and MALME-3M and UACC-257 for SNAI2 (data not shown). Each of those Effects of ZFHX1B-siRNA and SNAI2-siRNA on cell lines (for both ZFHX1B and SNAI2) has one additional E-cad Expression transcription factor in the range associated with low E-cad The transcript expression levels of ZFHX1B and SNAI2 expression. MALME-3M and UACC-257, but not ACHN, were significantly correlated with E-cad expression displayed E-cad up-regulation in the presence of ZFHX1B (Table 2), L-shaped in distribution when plotted against down-regulation (Fig. 2C). MALME-3M and UACC-257 E-cad expression (Fig. 1D and A, respectively), signifi- failed to up-regulate E-cad when SNAI2 expression was cantly correlated with one another in expression (r = knocked down. 0.83), and predictive of E-cad expression (P =0.03and 0.01, respectively; parametric two-tailed t test). Howev- Combining 5-AC with siTCF8 on TCF8 and E-cad er, when we tried to identify candidate cell lines in our Expression set with which to test whether ZFHX1B or SNAI2 down- We tested the effects on E-cad expression of combina- regulation was independently sufficient for up-regulation treatment directed on the two factors at the top of tion of E-cad expression, there were none. That is, none our proposed ranking of influence on E-cad: DNA meth- of the cells showed ZFHX1B or SNAI2 expression in the ylationandTCF8(Fig.2D).IGROV1,whichhasboth range associated with low E-cad expression (Fig. 1, E-cad methylation and TCF8 at levels associated with Table 1), E-cad DNA methylation levels <20%, and le- low E-cad expression, displayed E-cad up-regulation

Figure 1. Distribution of E-cad expression as a function of four transcriptional represser expression levels for the NCI-60. E-cad expression vs. A, SNAI2 expression, B, TCF8 expression, C, TWIST1 expression, and D, ZFHX1B expression. In all panels, the x axis is the log2 transcriptional repressor transcript level obtained from Affymetrix HG-U133 microarrays; the y axis is the log2 E-cad expression level from the same arrays (data from Table 1). Diamonds, circles, and squares represent cell lines with <20%, 20% to 30%, and >30% methylation of the E-cad promoter region, respectively. Pearson's correlation coefficients (R, from Table 2), P values (P), and confidence intervals (CI) appear in the upper right for each dataset.

www.aacrjournals.org Mol Cancer Ther; 9(1) January 2010 9

Reinhold et al.

Figure 2. Branched DNA assay measurement of the effect of 5-AC or siRNA treatment on E-cad transcript expression. In all cases, the bar graphs depict branched DNA assay measurements of transcript levels. The y axis of each bar is a cyclophilin-normalized expression. The error bars indicate one standard deviation for duplicate measurements. Factors that are proposed as suppressive of E-cad expression are shown in tabular form (log2 data from Table 1), with grayed or cross-hatched blocks indicating proposed repressive levels (as described in Table 1). Because of variations in the number of cells harvested, we do not consider the branched-DNA assay expression levels to be accurate for cross-cell line comparisons. A, the effect of 5-AC on E-cad expression levels. The open bars indicate E-cad expression in the absence of drug. The black bars indicate E-cad expression after 5-AC treatment for 24 h at 2, 0.5, 0.5, 2, 2, 0.5, 2, 2, 0.1, and 1 mg/mL (for SW-620, TK-10, IGROV1, A498, NCI-H460, OVCAR-8, SK-MEL-28, A549-ATCC, HCT-116, and HT29, respectively). B, the effect of TCF8 down-regulation by siRNA on E-cad expression levels. Six cell lines were treated with siNeg (white bars), siTCF8.1 (vertically striped bars), or siTCF8.2 (black bars), and their TCF8 and E-cad expression levels were determined.

following either 5-AC or siTCF8 treatment. With 4 also showed vertical cell stacking. Stacking occurred combined treatment, the level of E-cad up-regulation in none of the cells in Fig. 3A. An exception to the gen- was more than additive. eral trend was the unusual cell line COLO205, which expresses detectable levels of E-cad (transcript and pro- Epithelial Cell Lines with Elevated Levels of E-cad tein) but has an almost leukemic appearance and is only Show Increased Cell-Cell Adhesion weakly adherent to plastic (image not shown). Using the HG-U133 E-cad expression data in Table 1, Fig. 3A shows that cell lines with undetectable (<5.40) E-cad Expression in the 54 Attached NCI-60 Cell E-cad expression displayed lower levels of cell-cell Lines Is Correlated with Drug Activities adhesion than did those with detectable levels (Fig. We next examined the relationship between growth in- 3B). In addition to the obvious increase in cell-cell con- hibitory 50% measurements of drug activity for 118 drugs tact in Fig. 3B, T47D, KM12, NCI-H322M, and OVCAR- (27, 28) and E-cad measurements of mean methylation,

10 Mol Cancer Ther; 9(1) January 2010 Molecular Cancer Therapeutics

E-cadherin Regulation in NCI-60

Figure 2 Continued. C, the effect of ZFHX1B down-regulation by siRNA on E-cad expression levels. Three cell lines were treated with siNeg (white bars), siZFHX1B.1 (vertically striped bars), or siZFHX1B.2 (black bars), and their ZFHX1B and E-cad expression levels were determined. D, the effect of combined 5-AC treatment and TCF8 down-regulation by siRNA on E-cad expression levels. IGROV1 was treated with no siRNA (diagonally striped bars), siNeg (white bars), siTCF8.1 (vertically striped bars), or siTCF8.2 (black bars). The cells were treated with 5-AC treatment (+) or no drug (−) for 24 h at 2 μg/mL. transcript level, protein level, and DNA content (Table 3). olite. Of those 10 drugs, all but the DNA antimetabolite The mean of the correlation coefficients of the five E-cad showed negative correlations, indicating that as E-cad ex- expression measurements for the 54 attached (nonleuke- pression increases, drug activity decreases. Adding back mic) cell lines reached significant levels (P < 0.02 without the leukemias to the correlation calculations (Table 3) multiple comparisons correction) for six alkylating agents, yielded consistent or slightly stronger correlations for all three topoisomerase I inhibitors, and one DNA antimetab- but the DNA antimetabolite.

Figure 3. Variations in cell-cell adhesion as observed by phase contrast microscopy. A, cells with undetectable levels of E-cad expression (<5.40 units on HG-U133 arrays; Table 1). B, cells with detectable levels of E-cad (≥5.4 units on HG-U133 arrays; Table 1).

www.aacrjournals.org Mol Cancer Ther; 9(1) January 2010 11

Reinhold et al.

Table 3. Correlations of drug activities versus E-cad transcript, protein, and DNA copy numbers

Mechanism of Alkylating agents action type Carmustine Chlorozotocin PCNU Cisplatin DABQ Spiromustine

NSC 409962 178248 95466 119875 182986 172112 Fifty-six attached cell lines‡ Mean methylation 0.36 0.44 0.53 0.13 0.17 0.15 Hu6800§ −0.39 −0.43 −0.45 −0.33 −0.33 −0.41 cDNA§ −0.22 −0.24 −0.25 −0.29 −0.26 −0.23 U95§ −0.28 −0.33 −0.35 −0.31 −0.35 −0.37 U133§ −0.25 −0.27 −0.26 −0.34 −0.28 −0.23 Protein levels −0.34 −0.32 −0.33 −0.29 −0.28 −0.32 DNA copy number −0.07 −0.04 0.05 0.02 0.31 0.24 NCI-60 Mean methylation 0.53 0.48 0.59 0.26 0.29 0.33 Hu6800§ −0.42 −0.45 −0.48 −0.38 −0.39 −0.46 cDNA§ −0.27 −0.26 −0.28 −0.31 −0.30 −0.27 U95§ −0.32 −0.35 −0.37 −0.33 −0.37 −0.38 U133§ −0.24 −0.27 −0.27 −0.32 −0.28 −0.22 Protein levels −0.40 −0.35 −0.35 −0.31 −0.32 −0.38 DNA copy number 0.09 −0.03 0.07 0.08 0.32 0.29

(Continued on the following page)

Discussion of the mean −0.47; P < 0.05). In modeling the influences on E-cad expression, we treated methylation as an inde- Several factors generally studied in isolation in previ- pendent variable because of two lines of evidence. First, ous studies have been shown to influence E-cad expres- there are multiple prior examples of transcription factor sion (4, 15, 16, 18–23).Inthepresentstudy,weassessed binding being blocked by the presence of DNA methyla- regulation of E-cad expression levels in the NCI-60 cell tion (41–43). Second, our data from multiple microarray panel by eight of the potential effectors (Table 1 and Fig. platforms, as well as the branched DNA assay, indicated 2A-D) using our E-cad promoter DNA methylation da- that E-cad is not expressed in any of the NCI-60 if the ta), data from six different microarray platforms, and methylation level is greater than approximately 30% data from pharmacological assays employing RNA in- (18). That is, DNA methylation (but not the other factors terference and 5-AC treatment. These factors were studied) appeared sufficient by itself for down-regulation assessed in combination, yielding a coherent picture of of E-cad. We therefore define a nonpermissive level of multifactorial network regulation of E-cad expression. methylation as >30% for the purposes of regulatory Included were the effects of DNA methylation of the and statistical modeling. Down-regulation of E-cad in E-cad promoter region, DNA copy number, and expres- cancer by promoter region methylation (but not the sion levels of the transcriptional repressors SNAI1, threshold at which it occurs) has been well documented SNAI2, TCF3, TCF8, TWIST1, and ZFHX1B. In addition, (3–6, 17, 18, 44, 45). levels of E-cad were shown to be associated with cell- Three additional factors found to be associated with cell adhesion and to correlate to the potencies of alkylat- and statistically predict E-cad expression in the NCI-60 ing agents, topoisomerase 1 inhibitors, and a DNA an- were SNAI2, TCF8, and ZFHX1B. All showed significant timetabolite. negative correlations across the panel (P < 0.05; two- Because transcript and protein levels often do not corre- tailed, no multiple comparisons correction) with the five late over diverse cell types (32), it was not obvious that they measurements of E-cad expression (Table 2). Additionally, would in this case. However, we found that the E-cad tran- Fig. 1A, B, and D show that high expression levels of script and protein expression data do correlate with each at SNAI2, TCF8, or ZFHX1B are associated with low levels statistically significant levels (mean r =0.83;P <0.05). of E-cad expression. Although high levels of TWIST1 There was no correlation for DNA copy number, hence were also associated with low E-cad expression (P = no evidence that copy number regulates E-cad expression. 0.01), they were not independently predictive once the Next, we found that E-cad expression correlates nega- confounding effects of methylation, TCF8, SNAI2, and tively with E-cad DNA promoter region methylation in a ZFHX1B had been removed. The bracketed expression consistent and significant manner (correlation coefficient ranges of SNAI2, TCF8, and ZFHX1B (Fig. 1A, B, and

12 Mol Cancer Ther; 9(1) January 2010 Molecular Cancer Therapeutics

E-cadherin Regulation in NCI-60

Table 3. Correlations of drug activities versus E-cad transcript, protein, and DNA copy numbers (Cont'd)

Topoisomerase 1 inhibitors DNA antimetabolite* CPT,9-MeO† CPT,10-OH† CPT,11-HOMe (RS)† Inosine-glycodialdehyde

176323 107124 606173 118994

0.13 0.42 0.28 −0.20 −0.36 −0.35 −0.35 0.31 −0.39 −0.39 −0.32 0.36 −0.37 −0.49 −0.39 0.35 −0.47 −0.36 −0.35 0.41 −0.35 −0.41 −0.32 0.28 0.29 0.22 0.17 −0.27

0.21 0.41 0.44 0.10 −0.39 −0.35 −0.41 0.13 −0.40 −0.39 −0.35 0.19 −0.40 −0.45 −0.40 0.15 −0.47 −0.33 −0.34 0.27 −0.38 −0.38 −0.34 0.04 0.32 0.17 0.14 −0.06

NOTE: Drugs were selected from a 118-drug set of known mechanism of action based on level of correlation to the 54 attached cell lines. Bold numbers indicate statistically significant correlations at P < 0.02 without multiple comparisons correction. Abbreviations: PCNU, Polycarbonate-based polyurethane; DABQ, Diaziridinylbenzoquinone; NSC, National Service Center. *Classification as a DNA antimetabolite by the National Cancer Institute's Developmental Therapeutics Program. †Camptothecin. ‡National Service Center numbers. §The nonleukemic subset of the NCI-60 that grows in tissue cultures as attached cells.

D, respectively) associated with low E-cad expression all at E-cad methylation levels >30% and is repressed at show exception cell lines with detectable E-cad expression SNAI2, ZFHX1B, and TCF8 HG-U133 intensity levels (MALME-3M and UACC-257 for SNAI2 and ZFHX1B; >5.16, >3.30, and >5.18, respectively (Fig. 1A, D, and B, re- EKVX for TCF8), indicating that those ranges do not spectively). In this model, DNA methylation is sufficient absolutely prevent E-cad expression. SNAI1 and TCF3 for down-regulation of E-cad, whereas the transcriptional expression were uncorrelated with E-cad expression. repressors are conditionally sufficient (i.e., sufficient in The correlations shown in Table 2 between the six some cell types but not in others). E-cad measurements and the six transcriptional repressor To test this regulatory ranking and also assess our expression levels for the NCI-60 and the eight tissues of ability to predict treatment combinations that would re- origin (excluding the prostate, for which there are only sult in E-cad up-regulation, we designed functional as- two lines) provides additional information that confirms says (Fig. 2A-D and data for SNAI2 not shown) in prior relationships and indicates novel relationships. The which we manipulated the four statistically predictive significant correlations with E-cad expression for SNAI2 factors. In each case, the optimal test case for E-cad in breast and TCF8 in non-small cell lung cancer support up-regulation was a cell line with that repressor in its prior findings (46, 47). The correlations between E-cad proposed repressive range, levels of the other three re- expression and TCF8 in breast and ovarian lines and pressive factors that were not, and undetectable levels TWIST1 in melanomas are novel. The correlations with of E-cad expression (i.e., <5.40 by HG-U133 array in Ta- SNAI1, TCF3, and ZFHX1B were not consistently statis- ble 1). Quantitation of RNA expression was provided by tically significant for any of the tissue-of-origin types. the branched DNA assay, which proved to be more sen- Based on this bioinformatic and statistical analysis, we sitive than the HG-U133 microarrays. propose for the regulatory factors analyzed that the top When choosing cells in which we predict that E-cad three levels of influence on E-cad expression in the NCI- would be up-regulated by DNA demethylation, none 60 are E-cad promoter methylation, expression of TCF8, fit our optimal test case parameters. The flaws in the and expression of ZFHX1B or SNAI2 (whose profiles are three best candidates were that TK-10 expresses SNAI2 not linearly separable statistically). E-cad is not expressed in its repressive range, and SW-620 and IGROV1 have

www.aacrjournals.org Mol Cancer Ther; 9(1) January 2010 13

Reinhold et al.

methylation levels slightly below the 30% cutoff. Never- in E-cad expression of 4.8-fold for 5-AC alone, a mean theless, up-regulation of E-cad followed treatment with of 2.3-fold following siTCF8 down-regulation, and a 5-AC (for SW-620, TK-10, and IGROV1) by 2.2-, 22-, mean of 12.0-fold for the combination of 5-AC with and 6.1-fold, respectively (Fig. 2A), providing functional siTCF8. Those results support our predictive criteria evidence that the 23% and 24% E-cad DNA methylation for identifying cells that can be made to re-express E- levels (in SW-620 and IGROV1, respectively), although cad by down-regulating both DNA methylation and repressive, are insufficient to silence E-cad expression TCF8 expression; indicate that both associations are po- completely (see Fig. 2A, no drug) and that the association tentially causal; and suggest that, in the presence of between E-cad methylation and expression might be multiple repressive factors, a targeted combination of causal. E-cad was re-expressed in TK-10 in the presence treatments is likely the optimal approach for up-regula- of repressive levels of SNAI2. We were unable to cause tion of E-cad expression. E-cad up-regulation in the presence of multiple transcrip- Both TCF8 and ZFHX1B have previously been de- tional repressors in their repressive ranges (for A498, scribed as being targeted by the miR-200 family (48, NCI-H460, OVCAR-8, and SK-MEL-28). E-cad is not up- 49). Consistent with these results, we found significant regulated in the secondary affect control cell lines, A549, negative correlations between TCF8 expression and hsa- HCT-116, or HT29, in the presence of low (baseline) levels miR-200a and 200b of −0.49, −0.55, respectively, as well of methylation. as between ZFHX1B and hsa-miR-200a, 200b, and 200c To test E-cad up-regulation following TCF8 down- of −0.56, −0.67, and −0.34, respectively. Levels for the regulation (Fig. 2B), we used two cell lines, A549-ATCC miRNA are from our prior study (50) and are available and DU-145, with optimal profiles. E-cad was success- at http://discover.nci.nih.gov/cellminer/queryLoad.do. fully up-regulated 2.4- and 2.8-fold, respectively, by In the present study, we also defined two additional TCF8 knock-down in those lines. Those findings sup- important factors that correlate with E-cad expression port our predictive criteria for identifying cells that will in the NCI-60: cell-cell adhesion and the potencies of a up-regulate E-cad expression following TCF8 down- variety of drugs. Within the subset of 54 (nonleukemic) regulation and indicate that the E-cad/TCF8 expression attached cell lines, the high expressers of E-cad tend association is potentially causal. Our model predicted to exhibit higher levels of cell-cell adhesion (Fig. 3A successfully that E-cad up-regulation would fail in and B), consistent with prior reports (3, 10, 11, 19, 20, OVCAR-8 because of high DNA methylation. E-cad 22). The leukemias, with no detectable E-cad, grow as expression is not re-expressed in TK-10, T47D, and spherical detached cells and exhibit limited cell-cell ad- HCT-116 in the absence of repressive TCF-8 levels. hesion (data not shown). The drug activity patterns There were no cell lines ideal for testing E-cad up-reg- (Table 3) have significant negative correlations to E-cad ulation following down-regulation of ZFHX1B. Still, by expression in 9/10 drugs for both the attached cell subset using our criteria to select the “next best” cell lines, we and the NCI-60. were able to up-regulate E-cad following ZFHX1B In conclusion, we report an integromic analysis of mul- down-regulation in MALME-3M and UACC-257 by 2.4- tiple factors with the potential, either individually or in and 1.7-fold (mean values), respectively (Fig. 2C). Those combination, to regulate E-cad expression. We relate findings support our predictive criteria for identifying those factors to E-cad expression at the transcript and cells that can re-express E-cad following ZFHX1B protein levels. Statistical analysis of this data allows the down-regulation and indicate that the E-cad/ZFHX1B prediction of which cell lines would show up-regulation expression association is potentially causal. E-cad is not of E-cad expression in pharmacological assays using re-expressed in ACHN, in the presence of repressive lev- 5-AC or siRNAs against TCF8, ZFHX1B, or SNAI2. els of TCF8. Among those regulatory factors, methylation status is pro- There were also no cell lines ideal for testing E-cad up- posedtobenonpermissiveandsufficientbyitselfto regulation following SNAI2 down-regulation. We were down-regulate E-cad expression when above a 30% unable to re-express E-cad in either of our next best cell threshold (18). TCF8 expression, SNAI2 and ZFHX1B are lines, MALME-3M or UACC-257, following SNAI2 proposed to be conditionally sufficient, and repressive down-regulation in the presence of repressive levels of ranges are proposed for each of those factors. TWIST1 is ZFHX1B (data not shown). correlated with E-cad down-regulation while not being Most of the cell lines that showed no detectable E-cad shown to be predictive. SNAI1, TCF3, and DNA copy by U133 array have multiple repressive factors in their number show no obvious effect on E-cad. The functional proposed repressive ranges, implying that multiple in- assays done either confirmed or extended the proposed terventions might be required for substantial E-cad regulatory ranking, leading us to conjecture causality for up-regulation in many cancer cell lines. Therefore, we E-cad regulation for promoter methylation, TCF8, and did an initial test of the effects of siRNA and 5-AC treat- ZFHX1B. The data thus provide a rational basis for pro- ment in combination in IGROV1, a cell line that com- spectively predicting what pharmacological combinations bines the presence of the top two repressive factors in of DNA demethylation and down-regulation of transcrip- our regulatory ranking, E-cad DNA methylation and tional repressors would yield E-cad up-regulation in par- TCF8 expression. The results (Fig. 2D) were increases ticular cancer cell types. The findings thus have

14 Mol Cancer Ther; 9(1) January 2010 Molecular Cancer Therapeutics

E-cadherin Regulation in NCI-60

implications for strategies to suppress cancer invasion and who led development of the NCI-60, and Kenneth metastasis associated with E-cad loss. Paull, who pioneered the associated informatics.

Disclosure of Potential Conflicts of Interest Grant Support

No potential conflicts of interest were disclosed. Intramural Research Program of the National Institutes of Health, National Cancer Institute, Center for Cancer Acknowledgments Research. The costs of publication of this article were defrayed in We are grateful to the many members of the National part by the payment of page charges. This article must Cancer Institute's Developmental Therapeutics therefore be hereby marked advertisement in accordance Program for their work on the screen and Molecular with 18 U.S.C. Section 1734 solely to indicate this fact. Targets Program. We particularly acknowledge the Received 6/8/09; revised 11/5/09; accepted 11/15/09; contributions of Bruce Chabner and Michael Boyd, published OnlineFirst 1/6/10.

References 1. Huber O, Bierkamp C, Kemler R. Cadherins and catenins in develop- (CDH1) mutations predispose to familial gastric cancer and colorec- ment. Curr Opin Cell Biol 1996;8:685–91, PubMed doi:10.1016/ tal cancer. Hum Mol Genet 1999;8:607–10, PubMed doi:10.1093/ S0955-0674(96)80110-4. hmg/8.4.607. 2. Overduin M, Harvey T, Bagby S, et al. Solution structure of the epi- 16. Hiraguri S, Godfrey T, Nakamura H, et al. Mechanisms of inactivation thelial cadherin domain responsible for selective cell adhesion. Sci- of E-cadherin in breast cancer cell lines. Cancer Res 1998;58:1972– ence 1995;267:386–9, PubMed doi:10.1126/science.7824937. 7, PubMed. 3. Matsumura T, Makino R, Mitamura K. Frequent down-regulation of 17. Tsutsumida A, Hamada J, Tada M, et al. Epigenetic silencing of E-cadherin by genetic and epigenetic changes in the malignant pro- E- and P-cadherin gene expression in human melanoma cell lines. gression of hepatocellular carcinomas. Clin Cancer Res 2001;7: Int J Oncol 2004;25:1415–21, PubMed. 594–9, PubMed. 18. Reinhold WC, Reimers M, Maunakea A, et al. Detailed DNA methyl- 4. Li LC, Zhao H, Nakajima K, et al. Methylation of the E-cadherin gene ation profiles of the E-cadherin promoter in the NCI-60 cancer cells. promoter correlates with progression of prostate cancer. J Urol 2001; Mol Cancer Ther 2007;6:391–403, PubMed doi:10.1158/1535-7163. 166:705–9, PubMed doi:10.1016/S0022-5347(05)66047-8. MCT-06-0609. 5. Nojima D, Nakajima K, Li L, et al. CpG methylation of promoter 19. Batlle E, Sancho E, Franci C, et al. The transcription factor snail is a region inactivates E-cadherin gene in renal cell carcinoma. Mol repressor of E-cadherin gene expression in epithelial tumour cells. Carcinog 2001;32:19–27, PubMed doi:10.1002/mc.1060. Nat Cell Biol 2000;2:84–9, PubMed doi:10.1038/35000034. 6. Yoshiura K, Kanai Y, Ochiai A, Shimoyama Y, Sugimura T, Hirohashi 20. Bolos V, Peinado H, Perez-Moreno M, Fraga M, Esteller M, Cano A. S. Silencing of the E-cadherin invasion-suppressor gene by CpG The transcription factor Slug represses E-cadherin expression and methylation in human carcinomas. Proc Natl Acad Sci U S A 1995; induces epithelial to mesenchymal transitions: a comparison with 92:7416–9, PubMed doi:10.1073/pnas.92.16.7416. Snail and E47 repressors. J Cell Sci 2003;116:499–511, PubMed 7. Siitonen SM, Kononen J, Helin H, Rantala I, Holli K, Isola J. Reduced doi:10.1242/jcs.00224. E-cadherin expression is associated with invasiveness and unfavor- 21. Comijn J, Berx G, Vermassen P, et al. The two-handed E box binding able prognosis in breast cancer. Am J Clin Pathol 1996;105:394– zinc finger protein SIP1 downregulates E-cadherin and induces inva- 402, PubMed. sion. Mol Cell 2001;7:1267–78, PubMed doi:10.1016/S1097-2765 8. Paul R, Ewing C, Jarrard D, Isaacs W. The cadherin cell-cell adhesion (01)00260-X. pathway in prostate cancer progression. Br J Urol 1997;79:37–43, 22. Perez-Moreno MA, Locascio A, Rodrigo I, et al. A new role for PubMed. E12/E47 in the repression of E-cadherin expression and epithelial- 9. Dunsmuir WD, Gillett C, Meyer L, et al. Molecular markers for predict- mesenchymal transitions. J Biol Chem 2001;276:27424–31, PubMed ing prostate cancer stage and survival. BJU Int 2000;86:869–78, doi:10.1074/jbc M100827200. PubMed doi:10.1046/j.1464-410x.2000.00916.x. 23. Yang J, Mani SA, Donaher JL, et al. Twist, a master regulator of mor- 10. Berx G, Cleton-Jansen A, Nollet F, et al. E-cadherin is a tumour/in- phogenesis, plays an essential role in tumor metastasis. Cell 2004; vasion suppressor gene mutated in human lobular breast cancers. 117:927–39, PubMed doi:10.1016/j cell.2004.06.006. EMBO J 1995;14:6107–15, PubMed. 24. Stinson SF, Alley M, Kopp W, et al. Morphological and immuno- 11. Christofori G, Semb H. The role of the cell-adhesion molecule E-cad- cytochemical characteristics of human tumor cell lines for use in a herin as a tumour-suppressor gene. Trends Biochem Sci 1999;24: disease-oriented anticancer drug screen. Anticancer Res 1992;12: 73–6, PubMed doi:10.1016/S0968-0004(98)01343-7. 1035–53, PubMed. 12. Nass SJ, Herman J, Gabrielson E, et al. Aberrant methylation of the 25. Weinstein JN. Spotlight on molecular profiling: 'integromic' analysis estrogen receptor and E-cadherin 5′ CpG islands increases with ma- of the NCI-60 cancer cell lines. Mol Cancer Ther 2006;5:2601–5, lignant progression in human breast cancer. Cancer Res 2000;60: PubMed doi:10.1158/1535-7163.MCT-06-0640. 4346–8, PubMed. 26. Weinstein JN. 'Omic' and hypothesis-driven research in the molecu- 13. Hsu M, Andl T, Li G, Meinkoth J, Herlyn M. Cadherin repertoire de- lar pharmacology of cancer. Curr Opin Pharmacol 2002;2:361–5, termines partner-specific gap junctional communication during mel- PubMed doi:10.1016/S1471-4892(02)00185-6. anoma progression. J Cell Sci 2000;113:1535–42, PubMed. 27. Scherf U, Ross D, Waltham M, et al. A gene expression database for 14. Li G, Fukunaga M, Herlyn M. Reversal of melanocytic malignancy by the molecular pharmacology of cancer. Nat Genet 2000;24:236–44, keratinocytes is an E-cadherin-mediated process overriding β-cate- PubMed doi:10.1038/73439. nin signaling. Exp Cell Res 2004;297:142–51, PubMed doi:10.1016/j 28. Ross DT, Scherf U, Eisen M, et al. Systematic variation in gene ex- yexcr.2004.03.012. pression patterns in human cancer cell lines. Nat Genet 2000;24: 15. Richards FM, McKee S, Rajpar M, et al. Germline E-cadherin gene 227–35, PubMed doi:10.1038/73432.

www.aacrjournals.org Mol Cancer Ther; 9(1) January 2010 15

Reinhold et al.

29. Staunton JE, Slonim D, Coller H, et al. Chemosensitivity prediction four known cancer genes in the NCI-60 cell line set. Mol Cancer Ther by transcriptional profiling. Proc Natl Acad Sci U S A 2001;98:10787– 2006;5:2606–12, PubMed doi:10.1158/1535-7163.MCT-06-0433. 92, PubMed doi:10.1073/pnas.191368598. 40. Lorenzi PL, Reinhold W, Varma S, et al. DNA fingerprinting of the 30. Shankavaram UT, Reinhold W, Nishizuka S, et al. Transcript and pro- NCI-60 cell line panel. Mol Cancer Ther 2009;8:713–24, PubMed tein expression profiles of the NCI-60 cancer cell panel: an integro- doi:10.1158/1535-7163.MCT-08-0921. mic microarray study. Mol Cancer Ther 2007;6:820–32, PubMed doi: 41. Hark AT, Schoenherr C, Katz D, Ingram R, Levorse J, Tilghman S. CTCF 10.1158/1535-7163.MCT-06-0650. mediates methylation-sensitive enhancer-blocking activity at the H19/ 31. Nishizuka S, Charboneau L, Young L, et al. Proteomic profiling of the Igf2 locus. Nature 2000;405:486–9, PubMed doi:10.1038/35013106. NCI60 cancer cell lines using new high-density 'reverse-phase' ly- 42. Kim J, Kollhoff A, Bergmann A, Stubbs L. Methylation-sensitive bind- sate microarrays. Proc Natl Acad Sci U S A 2003;100:14229–34, ing of transcription factor YY1 to an insulator sequence within the PubMed doi:10.1073/pnas.2331323100. paternally expressed imprinted gene, Peg3. Hum Mol Genet 2003; 32. Bussey KJ, Chin K, Lababidi S, et al. Integrating data on DNA copy 12:233–45, PubMed doi:10.1093/hmg/ddg028. number with gene expression levels and drug sensitivities in the NCI- 43. Mancini DN, Singh S, Archer T, Rodenhiser D. Site-specific DNA 60 cell line panel. Mol Cancer Ther 2006;5:853–67, PubMed doi: methylation in the neurofibromatosis (NF1) promoter interferes with 10.1158/1535-7163.MCT-05-0155. binding of CREB and SP1 transcription factors. Oncogene 1999;18: 33. Gallagher WM, Bergin OE, Rafferty M, et al. Multiple markers for mel- 4108–19, PubMed doi:10.1038/sj onc.1202764. anoma progression regulated by DNA methylation: insights from 44. Kawakami T, Okamoto K, Ogawa O, Okada Y. Multipoint methylation transcriptomic studies. Carcinogenesis 2005;26:1856–67, PubMed and expression analysis of tumor suppressor genes in human renal doi:10.1093/carcin/bgi152. cancer cells. Urology 2003;61:226–30, PubMed doi:10.1016/S0090- 34. Lorenzi PL, Reinhold W, Rudelius M, et al. Asparagine synthetase as 4295(02)02110-6. a causal, predictive biomarker for L-asparaginase activity in ovarian 45. Ribeiro-Filho LA, Franks J, Sasaki M, et al. CpG hypermethylation of pro- cancer cells. Mol Cancer Ther 2006;5:2613–23, PubMed doi: moter region and inactivation of E-cadherin gene in human bladder can- 10.1158/1535-7163.MCT-06-0447. cer. Mol Carcinog 2002;34:187–98, PubMed doi:10.1002/mc.10064. 35. Annereau JP, Szakacs G, Tucker CJ, et al. Analysis of ATP-binding 46. Hajra KM, Chen D, Fearon E. The SLUG zinc-finger protein represses cassette transporter expression in drug-selected cell lines by a E-cadherin in breast cancer. Cancer Res 2002;62:1613–8, PubMed. microarray dedicated to multidrug resistance. Mol Pharmacol 2004; 47. Ohira T, Gemmill R, Ferguson K, et al. WNT7a induces E-cadherin in 66:1397–405, PubMed doi:10.1124/mol.104.005009. lung cancer cells. Proc Natl Acad Sci U S A 2003;100:10429–34, 36. Huang Y, Anderle P, Bussey K, et al. Membrane transporters and PubMed doi:10.1073/pnas.1734137100. channels: role of the transportome in cancer chemosensitivity and 48. Gregory PA, Bert AG, Paterson EL, et al. The miR-200 family and miR- chemoresistance. Cancer Res 2004;64:4294–301, PubMed doi: 205 regulate epithelial to mesenchymal transition by targeting ZEB1 and 10.1158/0008-5472.CAN-03-3884. SIP1. Nat Cell Biol 2008;10:593–601, PubMed doi:10.1038/ncb1722. 37. Ellison G, Klinowska T, Westwood R, Docter E, French T, Fox J. Fur- 49. Park SM, Gaur AB, Lengyel E, Peter ME. The miR-200 family de- ther evidence to support the melanocytic origin of MDA-MB-435. termines the epithelial phenotype of cancer cells by targeting the Mol Pathol 2002;55:294–9, PubMed doi:10.1136/mp.55.5.294. E-cadherin repressors ZEB1 and ZEB2. Genes Dev 2008;22:894– 38. Roschke AV, Tonon G, Gehlhaus K, et al. Karyotypic complexity of 907, PubMed doi:10.1101/gad.1640608. the NCI-60 drug-screening panel. Cancer Res 2003;63:8634–47, 50. Blower PE, Verducci JS, Lin S, et al. MicroRNA expression profiles PubMed. for the NCI-60 cancer cell panel. Mol Cancer Ther 2007;6:1483–91, 39. Ikediobi ON, Davies H, Bignell G, et al. Mutation analysis of twenty- PubMed doi:10.1158/1535-7163.MCT-07-0009.

16 Mol Cancer Ther; 9(1) January 2010 Molecular Cancer Therapeutics

Multifactorial Regulation of E-Cadherin Expression: An Integrative Study

William C. Reinhold, Mark A. Reimers, Philip Lorenzi, et al.

Mol Cancer Ther 2010;9:1-16. Published OnlineFirst January 12, 2010.

Updated version Access the most recent version of this article at: doi:10.1158/1535-7163.MCT-09-0321

Cited articles This article cites 50 articles, 25 of which you can access for free at: http://mct.aacrjournals.org/content/9/1/1.full#ref-list-1

Citing articles This article has been cited by 2 HighWire-hosted articles. Access the articles at: http://mct.aacrjournals.org/content/9/1/1.full#related-urls

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

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

Permissions To request permission to re-use all or part of this article, use this link http://mct.aacrjournals.org/content/9/1/1. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.