Abstract
Chromosomal instability—a hallmark of epithelial cancers—is an ongoing process that results in aneuploidy and karyotypic heterogeneity of a cancer cell population. Previously, we stratified cancer cell lines in the NCI-60 drug discovery panel based on their karyotypic complexity and heterogeneity. Using this stratification in conjunction with drug response data for the cell lines allowed us to identify classes of chemical compounds whose growthinhibitory activity correlates with karyotypic complexity and chromosomal instability. In this article, we asked the question: What are the biological processes, pathways, or genes associated with chromosomal instability of cancer cells? We found that increased instability of the chromosomal content in a cancer cell population, particularly, persistent gains and losses of chromosomes, is associated with elevated expression of genes involved with aggressive cellular behavior, including invasion- and metastasis-associated changes in cell communication, adhesion, motility, and migration. These same karyotypic features are negatively correlated with the expression of genes involved in cell cycle checkpoints, DNA repair, and chromatin maintenance.
Introduction
Most cancers have an abnormal chromosomal content characterized by changes in chromosomal structure and number. Chromosomal aberrations tend to be more numerous in malignant tumors than in benign ones [1,2], and karyotypic complexity is associated with aggressive clinical behavior of tumors and poor prognoses [3–5]. Cancers contain cells that not only possess an abnormal number of chromosomes but also often show population heterogeneity with regard to the exact chromosomal complement. That heterogeneity is a marker of ongoing chromosomal instability in cancers—accelerated rates of gains and losses of whole chromosomes or large portions of cancer cell genomes [6,7]. Chromosomal instability is a process that facilitates cancer cell evolution under selection pressure, and because of instability and selection, tumors most frequently acquire an aneuploid chromosomal content.
The NCI-60 cancer cell line panel was assembled by the Developmental Therapeutics Program of the National Cancer Institute for in vitro anticancer drug screening. It includes human cancer cell lines of lung, renal, colorectal, ovarian, breast, prostate, central nervous system (CNS), melanocyte, and hematological origins [8]. Since 1990, the NCI-60 cells have been exposed to more than 100,000 compounds in short-term cytotoxicity assays (http://dtp.nci.nih.gov) [9], and they have been profiled more extensively at the DNA, RNA, protein, and functional levels than any other set of cell lines [10]. Previously, we defined and studied karyotypic complexity and heterogeneity (as markers of ongoing instability) in the NCI-60 [11] and used this stratification to identify groups and classes of chemical compounds with higher growth-inhibitory activity in the more karyotypically complex and chromosomally unstable lines [12–14]. In the present work, we ask the question: What biological processes, pathways, and genes are associated with karyotypic instability of cancer cells? To address that question, we analyzed the correlation between karyotypic parameters and expression of genes across the NCI-60 panel.
Materials and Methods
Karyotypic Parameters of the NCI-60 Panel of Cancer Cell Lines
The NCI-60 panel of cancer cell lines was developed by the National Cancer Institute for in vitro anticancer drug screening, and it included cell lineages derived from different tissues (lung, renal, colorectal, ovarian, breast, prostate, CNS, melanoma, and hematological malignancies). Every cell line in the NCI-60 panel carries karyotypic abnormalities with notable individual variations among the cell lines at the level of karyotypic complexity and heterogeneity [11]. Complexity of the karyotype was defined on the basis of numerical abnormalities and structural rearrangements. Numerical chromosomal complexity (NC) was expressed in relation to the cell line ploidy level, according to the International System for Human Cytogenetic Nomenclature convention, and was calculated as a sum of the number of deviations of each specific chromosome from the designated ploidy level. Structural chromosomal complexity (SC) was expressed as the number of different structurally rearranged chromosomes present in two or more metaphases. Numerical chromosomal heterogeneity (NH) corresponds to the number of cell-to-cell variations of similar chromosomes. Loss of a chromosome in only one or two cells or gain in only one cell was not considered in the calculation of numerical heterogeneity because of the possibility of mechanical loss or gain during preparation of the metaphase spreads. Any specific chromosome displaying a higher number of gains or losses was considered to show variability and tallied as “1 point” in the numerical heterogeneity index. Structural chromosomal heterogeneity (SH) was estimated as the number of nonclonal (i.e., present in only one metaphase, according to the International System for Human Cytogenetic Nomenclature convention) structurally abnormal chromosomes per metaphase.
NCI-60 Gene Expression Data
We used mRNA expression databases for the NCI-60 cell lines from the HG-U95 and the HG-U133 Affymetrix sets. HG-U95A data set includes ∼12,000 array features corresponding to 8978 genes. HG-U133 data set has ∼22,000 array features corresponding to 13,032 genes [15]. Gene expression data for NCI-60 cell lines were generated by hybridization to U133A and U133B arrays (Affymetrix, Santa Clara, CA), and CEL files were downloaded from http://discover.nci.nih.gov/cellminer. Three cell lines were excluded from analysis for different reasons (MDA-N was not available for karyotypic analysis, NCI/ADR-RES was a derivative of OVCAR-8, and lung cancer cell line H23 had no gene expression data available).
Data Analysis
CEL files for 57 cell lines were imported into R-2.4.0 language and statistical computing environment [16] using the Affy package of BioConductor-1.8 [17]. Probe set summarization, convolution background correction, and quantile normalization were conducted using a robust multichip average method [18] separately for U133A and U133B array sets. Data were filtered to have expression measurements to be above 16 in at least 25% of the samples and the interquartile range across the samples to be at least 0.5 on log2 scale. This nonspecific filtering leaves 7500 probe sets for the U133A array set and 4036 probe sets for the U133B array set. Probe set expression measurements (log2-transformed) were combined into one data matrix (11,536 probe sets) that was transposed and aligned with karyotypic characteristics of cell lines for analysis of Spearman's rank correlation between gene expression and karyotypic parameters.
To calculate the false discovery rate (FDR) for each Spearman correlation coefficients (SCCs), 1000 random permutations of the karyotypic characteristic value (NC, NH, SC, or SH) were performed, and SCCs were recalculated for all 11,536 probe sets for each permutation. The number of probe sets having an SCC with a P value ≤ .001 per random permutation was used to estimate FDR as the number of such probe sets at 0.95 or 0.99 quantiles (corresponding percentiles were considered as confidence levels for FDR estimates).
Genes differentially expressed in cell lines depending on tissue of origin were identified separately for U133A and U133B arrays using univariate F test as it is implemented in BRB-Array Tools [19]. Genes were considered differentially expressed if their F test parametric P value was less than .001.
To identify the most relevant GO terms associated with gene lists, gene-GO term enrichment analysis was performed using the DAVID2007 Functional Annotation Tool with customized gene backgrounds [20,21].
Results
Correlation of Expression Profiles with Karyotypic Parameters
We obtained SCCs between karyotypic parameters (SC, SH, NC, and NH) and each of the 11,536 probe sets from HG-U133A and HG-U133B data set obtained after filtering. The highest number of statistically significant (P < .001) correlations was identified for NH (454 array elements corresponding to 360 genes), followed by SH (318 array elements corresponding to 264 genes), NC (278 array elements corresponding to 224 genes), and SC (54 array elements corresponding to 50 genes; Table W1). To estimate the proportion of false-positives, 1000 random permutations of NH, NC, SH, and SC values were performed, and for each permutation, SCCs were calculated for all 11,536 probe sets. The density distributions of the number of SCCs with P value ≤ .001 for permuted sample were then analyzed. With 99% confidence, among 454 array elements correlated with NH for which SCCs have P values .001 or less, there were no more than 73 false-positive probe sets for which a positive or negative correlation of expression measurements with NH may be due to random variation in gene expression or NH (Table W1).
Because NH correlates with the highest number of gene expression profiles, and the number of possible false-positive results at P < .001 is the lowest for this parameter, correlations with gene expression profiles obtained for NH were chosen for further investigation.
Data for 454 probe sets for which the SCCs between NH and gene expression had P value .001, are presented in Table W2. Variability in SCCs for these 454 array elements was estimated by drawing 1000 bootstrap replicates and calculating the average bootstrap SCCs and corresponding quantiles as confidence intervals. For all 454 array elements, 99% bootstrap confidence intervals of SCCs do not include zero, remaining either positive or negative throughout.
Top Genes Associated with NH
Table 1 shows the top 50 Affy IDs positively associated with NH. These 50 Affy IDs correspond to only 39 gene transcripts because 8 genes (ABL2, ATP6V0E, CRTAP, FN1, GNG12, IL13RA1, TIMP2, and TNPO1) had more than one Affy ID selected in the top 50 positive list. On the basis of the SCC and P value, the three top genes with expression positively associated with NH are VEGFC, leprecan-like 1, and Ras-related GTP-binding C (Table 1). The list of 50 positive top genes is significantly enriched with genes that control cell communication (15 genes) and signal transduction (14 genes), response to wounding and cell motility (6 genes), cytokine-cytokine receptor interaction (5 genes), and cell migration (4 genes).
Table 1.
Top 50 Genes with Expression Profiles Correlated Positively (A) or Negatively (B) with NH across the NCI-60 Panel of Cancer Cell Lines.
| Affy ID | Gene Symbol | Gene Name | ρ | P |
| (A) | ||||
| 209946_at | VEGFC | Vascular endothelial growth factor C | 0.5817 | 2.08e-06 |
| 218717_s_at | LEPREL1 | Leprecan-like 1 | 0.5763 | 2.72e-06 |
| 218088_s_at | RRAGC | Ras-related GTP binding C | 0.5750 | 2.89e-06 |
| 204140_at | TPST1 | Tyrosylprotein sulfotransferase 1 | 0.5549 | 7.48e-06 |
| 226656_at | CRTAP | Cartilage-associated protein | 0.5501 | 9.31e-06 |
| 235086_at | THBS1 | Thrombospondin 1 | 0.5460 | 1.12e-05 |
| 212464_s_at | FN1 | Fibronectin 1 | 0.5428 | 1.29e-05 |
| 220092_s_at | ANTXR1 | Hypothetical protein FLJ10601 | 0.5425 | 1.30e-05 |
| 201426_s_at | VIM | Vimentin | 0.5377 | 1.61e-05 |
| 202859_x_at | IL8 | Interleukin 8 | 0.5349 | 1.82e-05 |
| 201172_x_at | ATP6V0E | ATPase, H+ transporting, lysosomal 9 kDa, V0 subunit E | 0.5341 | 1.88e-05 |
| 211719_x_at | FN1 | Fibronectin 1 | 0.5331 | 1.96e-05 |
| 210495_x_at | FN1 | Fibronectin 1 | 0.5330 | 1.97e-05 |
| 201828_x_at | CXX1 | CAAX BOX 1 | 0.5306 | 2.18e-05 |
| 201380_at | CRTAP | Cartilage-associated protein | 0.5301 | 2.22e-05 |
| 237444_at | Unknown | 0.5288 | 2.36e-05 | |
| 205743_at | STAC | SH3 and cysteine-rich domain | 0.5265 | 2.60e-05 |
| 201105_at | LGALS1 | Lectin, galactoside-binding, soluble, 1 (galectin 1) | 0.5257 | 2.67e-05 |
| 200096_s_at | ATP6V0E | ATPase, H+ transporting, lysosomal 9 kDa, V0 subunit E | 0.5257 | 2.68e-05 |
| 204214_s_at | RAB32 | RAB32, member ras oncogene family | 0.5248 | 2.78e-05 |
| 200885_at | RHOC | Ras homolog gene family, member C | 0.5242 | 2.85e-05 |
| 231907_at | ABL2 | V-abl Abelson murine leukemia viral oncogene homolog 2 (ARG, Abelson-related gene) | 0.5225 | 3.05e-05 |
| 226955_at | FLJ36748 | Hypothetical protein FLJ36748 | 0.5195 | 3.46e-05 |
| 216442_x_at | FN1 | Fibronectin 1 | 0.5172 | 3.79e-05 |
| 212509_s_at | MXRA7 | FLJ46603 protein | 0.5144 | 4.26e-05 |
| 222834_s_at | GNG12 | Guanine nucleotide binding protein (G protein), gamma 12 | 0.5130 | 4.50e-05 |
| 224560_at | TIMP2 | TIMP metallopeptidase inhibitor 2 | 0.5127 | 4.56e-05 |
| 211612_s_at | IL13RA1 | Interleukin 13 receptor, alpha 1 | 0.5114 | 4.80e-05 |
| 226939_at | CPEB2 | Cytoplasmic polyadenylation element binding protein 2 | 0.5110 | 4.88e-05 |
| 202378_s_at | LEPROT | Leptin receptor overlapping transcript | 0.5107 | 4.93e-05 |
| 235072_s_at | Unknown | 0.5104 | 4.99e-05 | |
| 209226_s_at | TNPO1 | Transportin 1 | 0.5086 | 5.36e-05 |
| 202733_at | P4HA2 | Procollagen-proline, 2-oxoglutarate 4-dioxygenase (proline 4-hydroxylase), alpha polypeptide II | 0.5083 | 5.43e-05 |
| 220407_s_at | TGFB2 | Transforming growth factor, beta 2 | 0.5074 | 5.62e-05 |
| 212294_at | GNG12 | Guanine nucleotide binding protein (G protein), gamma 12 | 0.5073 | 5.64e-05 |
| 202377_at | LEPR | Leptin receptor | 0.5064 | 5.85e-05 |
| 209278_s_at | TFPI2 | Tissue factor pathway inhibitor 2 | 0.5056 | 6.02e-05 |
| 201887_at | IL13RA1 | Interleukin 13 receptor, alpha 1 | 0.5048 | 6.23e-05 |
| 209225_x_at | TNPO1 | Transportin 1 | 0.5047 | 6.23e-05 |
| 208924_at | RNF11 | Ring finger protein 11 | 0.5044 | 6.31e-05 |
| 212658_at | LHFPL2 | Lipoma HMGIC fusion partner-like 2 | 0.5044 | 6.31e-05 |
| 213696_s_at | MED8 | Mediator of RNA polymerase II transcription, subunit 8 homolog (yeast) | 0.5043 | 6.33e-05 |
| 229465_s_at | PTPRD | Protein tyrosine phosphatase, receptor type, D | 0.5043 | 6.34e-05 |
| 203262_s_at | FAM50A | Family with sequence similarity 50, member A | 0.5037 | 6.48e-05 |
| 209013_x_at | TRIO | Triple functional domain (PTPRF interacting) | 0.5034 | 6.56e-05 |
| 207657_x_at | TNPO1 | Transportin 1 | 0.5026 | 6.78e-05 |
| 231579_s_at | TIMP2 | TIMP metallopeptidase inhibitor 2 | 0.5013 | 7.12e-05 |
| 200998_s_at | CKAP4 | Cytoskeleton-associated protein 4 | 0.5009 | 7.22e-05 |
| 229307_at | ANKRD28 | Ankyrin repeat domain 28 | 0.5001 | 7.45e-05 |
| 226893_at | ABL2 | V-abl Abelson murine leukemia viral oncogene homolog 2 (ARG, Abelson-related gene) | 0.5000 | 7.48e-05 |
| (B) | ||||
| 226148_at | BTBD15 | BTB (POZ) domain containing 15 | -0.6236 | 2.20e-07 |
| 212316_at | NUP210 | Nucleoporin 210 kDa | -0.6040 | 6.56e-07 |
| 224784_at | MLLT6 | Myeloid/lymphoid or mixed-lineage leukemia (trithorax homolog, Drosophila); translocated to, 6 | -0.5733 | 3.15e-06 |
| 200069_at | SART3 | Squamous cell carcinoma antigen recognized by T cells 3 | -0.5686 | 3.94e-06 |
| 212315_s_at | NUP210 | Nucleoporin 210 kDa | -0.5666 | 4.35e-06 |
| 226482_s_at | F11R | F11 receptor | -0.5620 | 5.38e-06 |
| 206687_s_at | PTPN6 | Protein tyrosine phosphatase, non-receptor type 6 | -0.5518 | 8.60e-06 |
| 213947_s_at | NUP210 | Nucleoporin 210 kDa | -0.5473 | 1.06e-05 |
| 212482_at | FLJ13910 | Hypothetical protein FLJ13910 | -0.5423 | 1.32e-05 |
| 227134_at | SYTL1 | Synaptotagmin-like 1 | -0.5405 | 1.42e-05 |
| 204142_at | ENOSF1 | Enolase superfamily member 1 | -0.5376 | 1.62e-05 |
| 224428_s_at | CDCA7 | Cell division cycle-associated 7 | -0.5354 | 1.78e-05 |
| 201969_at | NASP | Nuclear autoantigenic sperm protein (histone-binding) | -0.5354 | 1.78e-05 |
| 225887_at | Chromosome 13 open reading frame 23 | -0.5305 | 2.19e-05 | |
| 218491_s_at | THYN1 | Thymocyte nuclear protein 1 | -0.5300 | 2.24e-05 |
| 227560_at | SFXN2 | Sideroflexin 2 | -0.5288 | 2.35e-05 |
| 204798_at | MYB | V-myb myeloblastosis viral oncogene homolog (avian) | -0.5250 | 2.76e-05 |
| 212978_at | LRRC8B | Leucine-rich repeat containing 8 family, member B | -0.5218 | 3.15e-05 |
| 212446_s_at | LASS6 | LAG1 longevity assurance homolog 6 (Saccharomyces cerevisiae) | -0.5215 | 3.18e-05 |
| 202107_s_at | MCM2 | MCM2 minichromosome maintenance deficient 2, mitotin (S. cerevisiae) | -0.5202 | 3.36e-05 |
| 212873_at | HMHA1 | Histocompatibility (minor) HA-1 | -0.5197 | 3.43e-05 |
| 227378_x_at | MGC13114 | Hypothetical protein MGC13114 | -0.5192 | 3.49e-05 |
| 213149_at | DLAT | Dihydrolipoamide S-acetyltransferase (E2 component of pyruvate dehydrogenase complex) | -0.5189 | 3.55e-05 |
| 227586_at | LOC124491 | LOC124491 | -0.5188 | 3.55e-05 |
| 204767_s_at | FEN1 | Flap structure-specific endonuclease 1 | -0.5170 | 3.83e-05 |
| 201038_s_at | ANP32A | Acidic (leucine-rich) nuclear phosphoprotein 32 family, member A | -0.5166 | 3.90e-05 |
| 208901_s_at | TOP1 | Topoisomerase (DNA) I | -0.5163 | 3.94e-05 |
| 219378_at | NARG1L | NMDA receptor-regulated 1-like | -0.5155 | 4.07e-05 |
| 225179_at | HIP2 | Huntingtin interacting protein 2 | -0.5151 | 4.13e-05 |
| 225716_at | BRI3BP | BRI3 binding protein | -0.5139 | 4.34e-05 |
| 217980_s_at | MRPL16 | Mitochondrial ribosomal protein L16 | -0.5125 | 4.59e-05 |
| 225845_at | BTBD15 | BTB (POZ) domain containing 15 | -0.5116 | 4.75e-05 |
| 213251_at | LOC441046 | Hypothetical LOC 441046 | -0.5101 | 5.04e-05 |
| 220035_at | NUP210 | Nucleoporin 210 kDa | -0.5091 | 5.26e-05 |
| 223268_at | C11ORF54 | Chromosome 11 open reading frame 54 | -0.5091 | 5.26e-05 |
| 210206_s_at | DDX11 | Dead/H (Asp-Glu-Ala-Asp/His) box polypeptide 11 (CHL1-like helicase homolog, S. cerevisiae) | -0.5079 | 5.52e-05 |
| 202778_s_at | ZNF198 | Zinc finger protein 198 | -0.5058 | 5.99e-05 |
| 201202_at | PCNA | Proliferating cell nuclear antigen | -0.5053 | 6.09e-05 |
| 212943_at | KIAA0528 | KIAA0528 | -0.5048 | 6.21e-05 |
| 219067_s_at | C10ORF86 | Chromosome 10 open reading frame 86 | -0.5044 | 6.31e-05 |
| 219188_s_at | LRP16 | LRP16 protein | -0.5020 | 6.94e-05 |
| 201401_s_at | ADRBK1 | Adrenergic, beta, receptor kinase 1 | -0.5014 | 7.10e-05 |
| 203375_s_at | TPP2 | Tripeptidyl peptidase II | -0.4994 | 7.65e-05 |
| 200091_s_at | RPS25 | Ribosomal protein S25 | -0.4992 | 7.72e-05 |
| 224944_at | TMPO | Thymopoietin | -0.4982 | 8.01e-05 |
| 228992_at | MED28 | Mediator of RNA polymerase II transcription, subunit 28 homolog (yeast) | -0.4970 | 8.39e-05 |
| 213626_at | CBR4 | Carbonic reductase 4 | -0.4966 | 8.54e-05 |
| 231887_s_at | KIAA1274 | KIAA1274 | -0.4926 | 9.91e-05 |
| 202163_s_at | CNOT8 | CCR4-NOT transcription complex, subunit 8 | -0.4906 | 1.07e-04 |
| 202163_s_at | CNOT8 | CCR4-NOT transcription complex, subunit 8 | -0.4906 | 1.07e-04 |
The top 50 Affy IDs (corresponding to 46 genes) expression of which negatively correlated with NH are shown in the Table 1B. In this top 50 negative list, NUP210 is represented by four Affy IDs and BTBD16 is represented by two. Both are among the three top genes negatively associated with NH. The third is MLLT6 (Table 1). Among all 46 gene transcripts, 26 are significantly associated with intracellular organelles, 22 have membrane-bound products, and 18 are related to the nucleus.
Identification of Origin-Independent Transcriptional Changes Associated with NH
The NCI-60 panel includes cell lines derived from different tissues. It seems possible, therefore, that the differences in gene expression profiles that correlate with karyotype are actually determined by tissue of origin and that correlations with karyotypic parameters simply reflect that fact epiphenomenally. Indeed, cluster analysis (Supplementary Materials and Methods) shows that the main factor that affects pattern of gene expression in the NCI-60 cell lines is the tissue of origin of the cell line (Figures W1–W3). Even when we selected for this analysis the 454 probe sets whose expression correlates with NH, cell lines still tend to cluster according to the tissue of origin (Figure W4). However, in this case, there is also a correlation of the clustering pattern with NH (Supplementary Results).
To address this question of origin-independent transcriptional changes associated with NH, we used two approaches. First, to estimate possible biases imposed on karyotype correlations by tissue-of-origin groupings, we used a “jackknife” procedure: each tissue-of-origin was omitted in turn, and SCCs for the 454 probe sets that had SCCs between NH and gene expression with P values ≤ .001 were calculated for the excluded subset of cell lines. For all 454, 99% jackknife confidence intervals of SCCs excluded zero and remained positive or negative throughout, confirming that correlation of the expression of these transcripts with NH is not associated with any specific tissue-of-origin (data not shown).
Second, univariate F test identifies 1579 and 1098 probe sets in U133A and U133B arrays, respectively, as differentially expressed between cell lines grouped according to their 9 tissues of origin. Among them, 241 array features (corresponding to 172 genes) overlap with the previously identified ones significantly associated with NH (P < .001). Consequently, among the 454 Affymetrix IDs associated with NH at P < .001, 213 IDs corresponding to 188 genes are not differentially expressed based on their tissue of origin (129 features/121 genes negative, 84 features/67 genes positive; Table W3).
Discussion
Aneuploidy and chromosomal instability are common conditions for most epithelial cancer cells, but the relationships between those factors and cellular functions are not clear. We, therefore, used correlations between NH and gene expression profiles of the NCI-60 cell lines, followed by Gene Ontology (GO) categorization, to identify genes and cellular processes associated with the increase of chromosomal instability in cancer cells.
Gene Ontology analysis of the distribution of 360 genes correlated with NH (P < .001) indicated that cell communication and signal transduction, cell adhesion, motility, and migration, response to wounding and inflammatory response, negative regulation of cell proliferation, and DNA replication are the main biological processes associated with numerical heterogeneity of the chromosomal content in the cancer cells (Table 2). Moreover, when these genes were divided into two groups based on their positive or negative correlation coefficients, we saw a striking difference between these two groups. Genes, expression of which was positively correlated with NH, fell into GO categories such as cell communication and signal transduction, including cell surface receptor-linked signal transduction, cell adhesion, locomotion, motility, and migration, development, morphogenesis, and differentiation, response to wounding, and inflammatory response (Figure 1A). Products of these genes were associated with extracellular matrix and extracellular space, plasma membrane, and cytoskeleton and were involved in the focal adhesion pathway (HSA04510), cytokine-cytokine receptor interaction (HSA04060), regulation of actin cytoskeleton (HSA04810), JAK-STAT signaling pathways (HSA04630), cell communication (HSA01430), and ECM-receptor interaction pathways (HSA04512; Table 2).
Table 2.
Gene Ontology Categories Associated with Genes Whose Expression Profiles Correlated with NH (P < .001) Across the NCI-60 Panel of Cancer Cell Lines.
| Category | GO Term | GO Terms Associated with 360 Gene Transcripts Correlated with NH, P < .001 | GO Terms Associated with 189 Gene Transcripts Positively Correlated with NH, P < .001 | GO Terms Associated with 171 Gene Transcripts Negatively Correlated with NH, P < .001 | ||||||
| Count | % | P | Count | % | P | Count | % | P | ||
| GOTERM_BP_ALL | homophilic cell adhesion | 22 | 5.76 | 1.14e-13 | 22 | 10.48 | 9.44e-19 | |||
| GOTERM_BP_ALL | cell-cell adhesion | 23 | 6.02 | 4.25e-11 | 23 | 10.95 | 3.08e-16 | |||
| GOTERM_BP_ALL | cell adhesion | 36 | 9.69 | 1.16e-08 | 36 | 17.14 | 2.35e-15 | |||
| GOTERM_BP_ALL | cell surface receptor-linked signal transduction | 32 | 8.38 | 9.20e-04 | 27 | 12.86 | 3.65e-06 | |||
| GOTERM_BP_ALL | cell communication | 70 | 18.32 | 4.48e-02 | 54 | 25.71 | 8.16e-05 | |||
| GOTERM_BP_ALL | signal transduction | 65 | 17.02 | 6.08e-02 | 50 | 23.81 | 2.30e-04 | |||
| GOTERM_BP_ALL | response to wounding | 13 | 3.40 | 7.31e-03 | 11 | 5.24 | 9.33e-04 | |||
| GOTERM_BP_ALL | response to external stimulus | 16 | 4.19 | 7.84e-03 | 13 | 6.19 | 1.11e-03 | |||
| GOTERM_BP_ALL | development | 41 | 10.80 | 2.80e-01 | 36 | 17.14 | 1.14e-03 | |||
| GOTERM_BP_ALL | locomotion | 15 | 3.93 | 1.39e-03 | 11 | 5.24 | 1.33e-03 | |||
| GOTERM_BP_ALL | localization of cell | 15 | 3.93 | 1.39e-03 | 11 | 5.24 | 1.33e-03 | |||
| GOTERM_BP_ALL | cell motility | 15 | 3.93 | 1.39e-03 | 11 | 5.24 | 1.33e-03 | |||
| GOTERM_BP_ALL | cell migration | 8 | 2.09 | 1.20e-02 | 7 | 3.33 | 2.88e-03 | |||
| GOTERM_BP_ALL | G-protein-coupled receptor protein signaling pathway | 14 | 3.66 | 3.38e-03 | 10 | 4.76 | 4.18e-03 | |||
| GOTERM_BP_ALL | inflammatory response | 9 | 2.36 | 5.25e-03 | 7 | 3.33 | 4.26e-03 | |||
| GOTERM_BP_ALL | cellular morphogenesis | 10 | 4.76 | 1.75e-02 | ||||||
| GOTERM_BP_ALL | cell differentiation | 12 | 5.71 | 2.68e-02 | ||||||
| GOTERM_BP_ALL | response to other organism | 13 | 3.40 | 8.48e-02 | 10 | 4.76 | 3.16e-02 | |||
| GOTERM_BP_ALL | neurogenesis | 5 | 2.38 | 3.17e-02 | ||||||
| GOTERM_BP_ALL | acute-phase response | 3 | 0.79 | 8.75e-02 | 3 | 1.43 | 3.18e-02 | |||
| GOTERM_BP_ALL | morphogenesis | 15 | 3.90 | 4.70e-01 | 14 | 6.67 | 4.30e-02 | |||
| GOTERM_BP_ALL | negative regulation of cell proliferation | 10 | 2.62 | 2.49e-02 | 6 | 2.86 | 9.36e-02 | |||
| GOTERM_BP_ALL | neutrophil chemotaxis | 3 | 0.79 | 2.30e-02 | ||||||
| GOTERM_BP_ALL | regulation of chemotaxis | 3 | 0.79 | 2.30e-02 | ||||||
| GOTERM_BP_ALL | positive regulation of chemotaxis | 3 | 0.79 | 2.30e-02 | ||||||
| GOTERM_BP_ALL | immune cell migration | 3 | 0.79 | 3.33e-02 | ||||||
| GOTERM_BP_ALL | immune cell chemotaxis | 3 | 0.79 | 3.33e-02 | ||||||
| GOTERM_CC_ALL | membrane | 108 | 28.42 | 5.18e-02 | 85 | 40.48 | 1.83e-06 | |||
| GOTERM_CC_ALL | intrinsic to membrane | 78 | 20.53 | 7.70e-02 | 64 | 30.48 | 1.77e-05 | |||
| GOTERM_CC_ALL | plasma membrane | 29 | 13.81 | 6.32e-03 | ||||||
| GOTERM_CC_ALL | extracellular region | 25 | 6.58 | 6.30e-03 | 25 | 11.90 | 1.75e-06 | |||
| GOTERM_CC_ALL | intrinsic to plasma membrane | 20 | 9.52 | 1.24e-02 | ||||||
| GOTERM_CC_ALL | cytoskeleton | 19 | 9.05 | 5.25e-02 | ||||||
| GOTERM_CC_ALL | extracellular matrix | 10 | 2.63 | 4.36e-02 | 10 | 4.76 | 1.66e-03 | |||
| GOTERM_CC_ALL | extracellular space | 8 | 3.81 | 3.35e-02 | ||||||
| GOTERM_CC_ALL | heterotrimeric G-protein complex | 3 | 1.43 | 4.23e-02 | ||||||
| GOTERM_CC_ALL | extrinsic to plasma membrane | 4 | 1.05 | 2.63e-02 | 3 | 1.43 | 5.77e-02 | |||
| GOTERM_MF_ALL | calcium ion binding | 35 | 9.21 | 6.86e-06 | 35 | 16.67 | 8.96e-13 | |||
| GOTERM_MF_ALL | metal ion binding | 66 | 31.43 | 2.57e-06 | ||||||
| GOTERM_MF_ALL | cation binding | 59 | 28.10 | 1.48e-05 | ||||||
| GOTERM_MF_ALL | transmembrane receptor activity | 16 | 4.21 | 2.62e-02 | 16 | 7.62 | 7.32e-05 | |||
| GOTERM_MF_ALL | signal transducer activity | 39 | 18.57 | 1.52e-04 | ||||||
| GOTERM_MF_ALL | protein binding | 135 | 35.53 | 3.29e-03 | 81 | 38.57 | 1.45e-03 | |||
| GOTERM_MF_ALL | receptor activity | 22 | 10.48 | 2.28e-03 | ||||||
| GOTERM_MF_ALL | actin binding | 12 | 3.16 | 6.81e-02 | 10 | 4.76 | 1.06e-02 | |||
| GOTERM_MF_ALL | growth factor binding | 4 | 1.90 | 2.60e-02 | ||||||
| GOTERM_MF_ALL | carbohydrate binding | 6 | 2.86 | 3.63e-02 | ||||||
| GOTERM_MF_ALL | l-ascorbic acid binding | 3 | 1.43 | 4.02e-02 | ||||||
| GOTERM_MF_ALL | enzyme regulator activity | 15 | 7.14 | 4.06e-02 | ||||||
| GOTERM_MF_ALL | cytoskeletal protein binding | 11 | 5.24 | 4.17e-02 | ||||||
| GOTERM_MF_ALL | extracellular matrix structural constituent | 4 | 1.90 | 4.64e-02 | ||||||
| GOTERM_MF_ALL | oncostatin-M receptor activity | 2 | 0.53 | 9.61e-02 | 2 | 0.95 | 5.34e-02 | |||
| KEGG_PATHWAY | HSA04060: CYTOKINE-CYTOKINE RECEPTOR INTERACTION | 10 | 2.63 | 2.17e-03 | 10 | 4.76 | 6.75e-05 | |||
| KEGG_PATHWAY | HSA04510: FOCAL ADHESION | 12 | 3.16 | 3.43e-02 | 11 | 5.24 | 3.78e-03 | |||
| KEGG_PATHWAY | HSA04630: JAK-STAT SIGNALING PATHWAY | 8 | 2.11 | 1.68e-02 | 7 | 3.33 | 6.57e-03 | |||
| KEGG_PATHWAY | HSA01430: CELL COMMUNICATION | 6 | 1.58 | 4.99e-02 | 5 | 2.38 | 3.85e-02 | |||
| KEGG_PATHWAY | HSA04810: REGULATION OF ACTIN CYTOSKELETON | 8 | 3.81 | 4.37e-02 | ||||||
| KEGG_PATHWAY | HSA04512: ECM-RECEPTOR INTERACTION | 5 | 2.38 | 5.13e-02 | ||||||
| GOTERM_BP_ALL | DNA-dependent DNA replication | 12 | 3.14 | 1.67e-04 | 12 | 7.02 | 3.70e-08 | |||
| GOTERM_BP_ALL | DNA replication | 15 | 3.93 | 1.80e-03 | 15 | 8.77 | 1.08e-07 | |||
| GOTERM_BP_ALL | DNA replication initiation | 7 | 1.83 | 1.74e-04 | 7 | 4.09 | 1.25e-06 | |||
| GOTERM_BP_ALL | nucleobase, nucleoside, nucleotide and nucleic acid metabolism | 57 | 33.33 | 3.65e-06 | ||||||
| GOTERM_BP_ALL | DNA metabolism | 22 | 12.87 | 1.09e-05 | ||||||
| GOTERM_BP_ALL | regulation of cellular metabolism | 38 | 22.22 | 1.03e-04 | ||||||
| GOTERM_BP_ALL | regulation of nucleobase, nucleoside, nucleotide and nucleic acid metabolism | 36 | 21.05 | 1.21e-04 | ||||||
| GOTERM_BP_ALL | transcription | 36 | 21.05 | 2.14e-04 | ||||||
| GOTERM_BP_ALL | regulation of transcription | 34 | 19.88 | 3.73e-04 | ||||||
| GOTERM_BP_ALL | macromolecule metabolism | 61 | 35.67 | 4.92e-03 | ||||||
| GOTERM_BP_ALL | cellular metabolism | 85 | 49.71 | 5.44e-03 | ||||||
| GOTERM_BP_ALL | response to DNA damage stimulus | 10 | 5.85 | 6.34e-03 | ||||||
| GOTERM_BP_ALL | response to endogenous stimulus | 10 | 5.85 | 7.73e-03 | ||||||
| GOTERM_BP_ALL | cell cycle | 24 | N/A | N/A | 18 | 10.53 | 8.62e-03 | |||
| GOTERM_BP_ALL | DNA unwinding during replication | 3 | 0.79 | 4.51e-02 | 3 | 1.75 | 8.85e-03 | |||
| GOTERM_BP_ALL | DNA repair | 9 | 5.26 | 9.74e-03 | ||||||
| GOTERM_BP_ALL | DNA packaging | 7 | 4.09 | 3.44e-02 | ||||||
| GOTERM_BP_ALL | phosphoinositide-mediated signaling | 4 | 2.34 | 3.78e-02 | ||||||
| GOTERM_BP_ALL | base-excision repair | 3 | 1.75 | 3.96e-02 | ||||||
| GOTERM_BP_ALL | chromosome organization and biogenesis | 8 | 4.68 | 4.19e-02 | ||||||
| GOTERM_BP_ALL | DNA ligation during DNA repair | 2 | 0.52 | 9.85e-02 | 2 | 1.17 | 4.23e-02 | |||
| GOTERM_BP_ALL | base-excision repair, DNA ligation | 2 | 0.52 | 9.85e-02 | 2 | 1.17 | 4.23e-02 | |||
| GOTERM_CC_ALL | nucleus | 71 | 41.52 | 6.28e-10 | ||||||
| GOTERM_CC_ALL | intracellular organelle | 92 | 53.80 | 2.32e-07 | ||||||
| GOTERM_CC_ALL | organelle | 92 | 53.80 | 2.37e-07 | ||||||
| GOTERM_CC_ALL | intracellular membrane-bound organelle | 85 | 49.71 | 4.34e-07 | ||||||
| GOTERM_CC_ALL | intracellular | 100 | 58.48 | 1.88e-06 | ||||||
| GOTERM_CC_ALL | chromosome | 15 | 3.68 | 4.65e-02 | 15 | 8.77 | 2.62e-06 | |||
| GOTERM_CC_ALL | chromatin | 8 | 4.68 | 4.68e-04 | ||||||
| GOTERM_CC_ALL | intracellular non-membrane-bound organelle | 25 | 14.62 | 1.11e-02 | ||||||
| GOTERM_CC_ALL | synaptic vesicle | 3 | 1.75 | 2.53e-02 | ||||||
| GOTERM_CC_ALL | nuclear envelope | 5 | 2.92 | 3.50e-02 | ||||||
| GOTERM_CC_ALL | condensed chromosome | 3 | 1.75 | 4.36e-02 | ||||||
| GOTERM_MF_ALL | nucleic acid binding | 55 | 32.16 | 4.48e-06 | ||||||
| GOTERM_MF_ALL | DNA binding | 37 | 21.64 | 7.16e-05 | ||||||
| GOTERM_MF_ALL | DNA-dependent ATPase activity | 6 | 1.32 | 2.09e-02 | 6 | 3.51 | 9.65e-05 | |||
| GOTERM_MF_ALL | RNA binding | 16 | 9.36 | 9.44e-03 | ||||||
| GOTERM_MF_ALL | transcription regulator activity | 22 | 12.87 | 9.78e-03 | ||||||
| GOTERM_MF_ALL | ATPase activity | 9 | 5.26 | 2.13e-02 | ||||||
| GOTERM_MF_ALL | ATPase activity, coupled | 8 | 4.68 | 3.19e-02 | ||||||
| KEGG_PATHWAY | HSA04110: CELL CYCLE | 10 | 2.63 | 3.13e-02 | 10 | 5.85 | 1.64e-05 | |||
Figure 1.
Biological processes associated with expression of genes positively (A) and negatively (B) correlated with NH.
Genes whose expression negatively correlated with NH fell into totally different GO categories: cellular metabolism, nucleic acid metabolism, regulation of transcription, DNA replication, response to DNA damage stimulus, DNA repair, chromosome organization and biogenesis, DNA packaging, unwinding and replication initiation, and base-excision repair (Figure 1B and Table 2). Products of those genes serve as transcription regulators, involved in nucleic acid binding, linked to ATP-ase activity, and associated with the cell cycle regulation pathway (HSA04110). They localize on intracellular organelles and are, for the most part, found in the nucleus, chromosome/chromatin, or nuclear envelope.
Among the 454 Affy IDs associated with NH at P < .001, 213 of them (corresponding to 188 genes) are not differentially expressed based on their tissue of origin (see the Results section). When we examined distribution of 188 genes across GO categories, genes whose expression is negatively associated with NH (121 genes) again were significantly enriched within such categories as intracellular membrane-bound organelle, nuclear protein, nucleic acid, or protein binding, regulation of transcription, DNA replication initiation, cell cycle, chromatin-related protein, and DNA repair. Genes whose expression correlated positively with NH (67 genes) again showed significant enrichment in such categories as cell communication and signal transduction, extracellular region, development and morphogenesis, response to wounding, cell migration, and motility. In summary, then, the GO classifications of genes positively and negatively associated with NH did not seem to depend on tissue of origin of the cancer cell.
Evaluation of 1122 genes whose expression correlated with NH at P < .01 (Table W4) revealed further enrichment of the same GO categories identified under stricter conditions (P < .001; data shown in the Table 2). That observation suggests that the essential findings are not sensitive to the P value cutoff used. Also reassuring with respect to robustness of the findings, analysis of the HG-U95A data showed similar GO category enrichments (data not shown).
Focusing on the negative correlations of expression with NH, we found enrichment of the following biological processes:
Cellular metabolism
Nucleic acid metabolism
Regulation of transcription
DNA replication
Response to DNA damage stimulus
DNA repair
Chromosome organization and biogenesis
DNA replication initiation
DNA packaging
Cell cycle regulation
The CIN phenotype has been associated with very rare mutations in the checkpoint genes [22–27] and with decreased protein levels of mitotic checkpoint components [25,27,28]. Deletion in mice of one allele of Mad2, Bub1, or Bub3 compromises the mitotic checkpoint, yielding higher rates of chromosome missegregation [29–32]. In our study, regulators of mitotic cell cycle checkpoint (MAD2 and BUB3), as well as a component of APC/C (APC4), are found among the genes whose expression is negatively correlated with NH. Correlations do not imply causative relationships; however, it would not be unreasonable to suggest that the decreased level of mitotic checkpoint components could be the basis of mitotic checkpoint relaxation leading to increased gains and losses of chromosomes. This supports already existing assumptions that a compromised mitotic checkpoint leads to accelerated rates of chromosomal instability in cancer cells [33].
The expression of genes involved in DNA damage checkpoints (CHK1, CHK2, H2AX, RAD21, XRCC5, DDB1) and DNA re-replication prevention (BCCIP, BRCA2, CDT1, MCM2-7, cyclin B2) negatively correlates with NH as well. The expression levels of genes involved in DNA packaging, chromosome condensation, and kinetochore formation (H3 histone, H1FX, H2AX, H2AZ, TOP1, RCC1, RCC2, SMARCA5, RCBTB1, CENPC1, ZWINT) are also relatively down-regulated in cancer cells with higher level of chromosomal instability compared to cancer cells with a lower level of instability.
Compromised cell cycle checkpoints give cancer cells an advantage in that they may be able to proliferate in a stressful environment with uncompleted DNA repair, perhaps with some level of unfinished chromatin condensation and/or individualization of chromatids, and perhaps with defects of mitotic chromosome organization. As a result, on-going gains and losses of chromosomes or their fragments can occur while cells are proliferating. We also found that chromosomal instability is associated with less effective cellular metabolism, DNA replication and transcription, DNA repair and packaging, weakness in proper chromatin condensation, and mitotic chromosome structural organization possibly owing to extensive imbalances in cellular protein composition of cells that undergo continuous gains and losses of parts of genome.
Gene Ontology's biological processes positively correlated with higher NH include the following:
Cell communication and signal transduction
Cell adhesion, including cell-cell adhesion
Cell surface receptor-linked signal transduction
Development and differentiation
Cell motility and migration
Response to wounding
Inflammatory response
G-protein-coupled receptor protein signaling
Cellular morphogenesis
A collective molecular portrait of numerical chromosomal heterogeneity in cancer cells includes relative up-regulation of genes that are associated with increased motility and migration, epithelialmesenchymal transition (EMT), and are critical for tumor invasion and metastasis: RhoC, fibronectin, LOX, TWIST, SNAI2, EGFR, laminins, integrins, collagens, CDC42 effector protein (Rho GTPase binding), Rho family GTPase 3, RAB, CXCL2, TGF-b2, VEGFC, IL-6, IL-8, CTGF, vimentin, N-cadherin, CD44, BCAR3, protocadherins, MMP2 and MMP14, NOTCH2, SERPINE1, 2, and 8, IGFBP3 and 7, TNFAIP3, TNFRSF12A and 19, PLAUR, and SPARC. Expression of several genes that promote cell proliferation and G1 entry into cell cycle (CCD1, EGFR, VEGFC) correlate positively with the higher NH as well.
Advances in the molecular profiling of cancer using genome-wide approaches have revealed genes whose expression levels in primary tumors correlate strongly with the likelihood of metastatic recurrence [34–37]. In particular, genes that are involved in physiological programs of cellular response to tissue damage (inflammation, wound healing, tissue remodeling, and regeneration) and genes that participate in differentiation and morphogenesis during epithelial tissue development are involved in tumor invasiveness and metastasis [38–41]. Important to those processes is an EMT, which enables cells to undergo major changes in morphology, lose cellular contacts, and acquire motility and ability to migrate. Also an important part of those programs is stimulation of proliferation and differentiation of epithelial cells themselves, as well as angiogenesis and lymphangiogenesis. Our findings suggest a link, at least indirectly, between chromosomal instability and cancer proliferation, invasion, and metastasis.
Chromosomal instability can create both advantages and disadvantages for cancer cells. For instance, severe genomic damage due to losses or gains of whole chromosomes or their essential fragments can prevent cancer cells from further proliferation. At the same time, the program of response to tissue damage represents an advantageous feature for overall cancer progression. Release of chemokines and cytokines promotes a response from surrounding tumor and stromal cells leading to the EMT in cancer cells, changes in cell adhesion and motility, promoting epithelial cell proliferation, as well as angiogenesis and lymphangiogenesis. We suggest that gross instability of cancer cell genomes leads to evolution of cancer cell populations, which “internalize” various inducers of a tissue damage response, gradually making cancer cells more and more independent of environmental stimuli. These changes in the gene expression pattern persist in cancer-derived cell lines.
Supplement: Cluster Analysis
Materials and Methods
For cluster analysis, U133A and U133B array data were processed separately as described (Materials and Methods, main text) with modifications. First, robust multichip average-transformed data were filtered to have expression measures higher than 64 at least in two cell lines, which is the minimum number of cell lines in a group. Two gene sets were selected: one for which the interquartile range in expression values was higher than 0.5 on log2 scale (11,848 probe sets) and the second for which the interquartile range in expression values was higher than 1 on log2 scale (this left 1667 and 682 probe sets for U133A and U133B array data sets, respectively). Gene expression matrixes were then additionally normalized by subtracting log2(median) expression value for each probe set. Gene expression matrixes for differentially expressed genes selected according to the F test (Materials and Methods, main text) and for 454 probe sets that show correlation with numerical heterogeneity (NH) level were also normalized by subtracting log2(median) expression value for each probe set. This normalization makes results more comparable when different distance measures are used. For Figures W2 to W4, clustering was done by Hierarchical Ordered Partitioning And Collapsing Hybrid (HOPACH) method as implemented in R package hopach [1] using cosine angle (uncentered correlation) and Euclidean distances for genes and arrays, respectively. For Figure W1, hierarchical clustering was performed by using function hclust (R package cluster), Euclidean distance, and average linkage. In the HOPACH algorithm, minimization of the median split silhouette value is used to choose the cluster number and cluster ordering. Compared with other clustering algorithms, the HOPACH method produces consistent cluster ordering regardless of the original ordering of the arrays in gene expression matrix [2]. Bootstrap resampling was done by selecting 10,000 sets of 57 arrays one array at a time with replacement. The proportion of resampled data sets in which each cell line falls into each of the clusters serves as an estimate of the membership of that cell line in each cluster. Bigger proportion of cell lines with consistent membership in a cluster indicates that cluster is well differentiated from other clusters and stable. Because we wanted to analyze mainly cell line characteristics (particularly karyotypic instability), the results are presented as ordered distance matrixes for cell lines together with bar plots of the bootstrap reappearance proportions for each cell line and each cluster. Additional details and results of cluster analysis are available on request.
Results
Hierarchical clustering of the 57 cell lines from NCI-60 collection shows that cell lines are mostly grouped according to the presumptive tissue of origin when using 11,848 probe sets with moderate level in expression variability (Figure W1).
We observed such correspondence after using for hierarchical clustering of cell lines 2349 genes that were selected also nonspecifically as displaying a higher (more than twofold difference between 25th and 75th percentiles) level of expression variability across cell lines (Figure W2). However, it can be seen that there are some similarities between cell lines unrelated to common tissue of origin: colon cell lines are similar to some prostate, renal, and ovarian cell lines and some colon cell lines are also similar to leukemia cell lines (Figure W2). Thus, though the tissue of origin seems to be a critical factor that influences gene expression pattern in the NCI-60 cell lines, there are additional variables that can be used to stratify cell lines if the gene set is selected appropriately (see also Ross et al. [3]).
For example, the correspondence of cell line gene expression patterns to tissue of origin can be enhanced if clustering is performed using a subset of genes that were selected according to an F test as differentially expressed between cell lines of different tissues of origin (Figure W3). Cell lines derived from leukemias, melanomas, colon tumors, renal tumors, CNS tumors, and, to a lesser extent, lung tumors form clusters that can be seen as separate blocks on the diagonal of the ordered dissimilarity matrix. This confirms the results obtained using spotted cDNA arrays [3] and illustrates that selection of genes changes the cluster analysis results: the clusters defined by differentially expressed genes are not only better separated but also are more stable than clusters obtained using nonspecifically selected genes, as follows from a bootstrap resampling analysis (compare Figures W2 and W3).
Impact of the tissue of origin of cell lines on gene expression pattern can still be traced after clustering using 454 probe sets whose expression correlates with NH index. However, the ordered dissimilarity matrix is quite different from the one based on genes differentially expressed among cell lines of different tissues of origin (Figure W4). Particularly, all cell lines derived from leukemia and colon tumors (together with two cell lines from breast tumors and one cell line from a lung tumor) form a tight cluster that is remarkably stable upon bootstrap resampling. Cell lines in this cluster have an average NH index of 0.27, whereas 21 cell lines in the biggest cluster that show maximal dissimilarity (denoted by the presence of a white color in the matrix) with the cluster of leukemia and colon cell lines have an average NH index 0.65.
Clustering analysis demonstrates that gene expression is influenced mainly by the tissue of origin, and at the same time, clustering of genes selected based on their correlation with NH is different, demonstrating an influence by NH as well.
Acknowledgments
The authors thank W. Michael Kuehl for helpful discussion and suggestions.
Abbreviations
- EMT
epithelial-mesenchymal transition
- NH
numerical chromosomal heterogeneity
- NC
numerical chromosomal complexity
- SH
structural chromosomal heterogeneity
- SC
structural chromosomal complexity
- GO
Gene Ontology
- SCC
Spearman correlation coefficient
Footnotes
This article refers to supplementary materials, which are designated by Tables W1 to W4 and Figures W1 to W4 and are available online at www.neoplasia.com.
References
- 1.Mitelman F, Johansson B, Mandahl N, Mertens F. Clinical significance of cytogenetic findings in solid tumors. Cancer Genet Cytogenet. 1997;95:1–8. doi: 10.1016/s0165-4608(96)00252-x. [DOI] [PubMed] [Google Scholar]
- 2.Ried T, Heselmeyer-Haddad K, Blegen H, Schrock E, Auer G. Genomic changes defining the genesis, progression, and malignancy potential in solid human tumors: a phenotype/genotype correlation. Genes Chromosomes Cancer. 1999;25:195–204. doi: 10.1002/(sici)1098-2264(199907)25:3<195::aid-gcc1>3.0.co;2-8. [DOI] [PubMed] [Google Scholar]
- 3.Pandis N, Idvall I, Bardi G, Jin Y, Gorunova L, Mertens F, Olsson H, Ingvar C, Beroukas K, Mitelman F, et al. Correlation between karyotypic pattern and clincopathologic features in 125 breast cancer cases. Int J Cancer. 1996;66:191–196. doi: 10.1002/(SICI)1097-0215(19960410)66:2<191::AID-IJC9>3.0.CO;2-Y. [DOI] [PubMed] [Google Scholar]
- 4.Bardi G, Fenger C, Johansson B, Mitelman F, Heim S. Tumor karyotype predicts clinical outcome in colorectal cancer patients. J Clin Oncol. 2004;22:2623–2634. doi: 10.1200/JCO.2004.11.014. [DOI] [PubMed] [Google Scholar]
- 5.Duesberg P, Li R, Rasnick D. Aneuploidy approaching a perfect score in predicting and preventing cancer: highlights from a conference held in Oakland, CA in January, 2004. Cell Cycle. 2004;3:823–828. [PubMed] [Google Scholar]
- 6.Lengauer C, Kinzler KW, Vogelstein B. Genetic instability in colorectal cancers. Nature. 1997;386:623–627. doi: 10.1038/386623a0. [DOI] [PubMed] [Google Scholar]
- 7.Lengauer C, Kinzler KW, Vogelstein B. Genetic instabilities in human cancers. Nature. 1998;396:643–649. doi: 10.1038/25292. [DOI] [PubMed] [Google Scholar]
- 8.Shoemaker RH, Monks A, Alley MC, Scudiero DA, Fine DL, McLemore TL, Abbott BJ, Paull KD, Mayo JG, Boyd MR. Development of human tumor cell line panels for use in disease-oriented drug screening. Prog Clin Biol Res. 1988;276:265–286. [PubMed] [Google Scholar]
- 9.Weinstein JN, Pommier Y. Connecting genes, drugs and diseases. Nat Biotechnol. 2006;24:1365–1366. doi: 10.1038/nbt1106-1365. [DOI] [PubMed] [Google Scholar]
- 10.Weinstein JN. Spotlight on molecular profiling: “integromic” analysis of the NCI-60 cancer cell lines. Mol Cancer Ther. 2006;5:2601–2605. doi: 10.1158/1535-7163.MCT-06-0640. [DOI] [PubMed] [Google Scholar]
- 11.Roschke AV, Tonon G, Gehlhaus KS, McTyre N, Bussey KJ, Lababidi S, Scudiero DA, Weinstein JN, Kirsch IR. Karyotypic complexity of the NCI-60 drug-screening panel. Cancer Res. 2003;63:8634–8647. [PubMed] [Google Scholar]
- 12.Roschke AV, Lababidi S, Tonon G, Gehlhaus KS, Bussey K, Weinstein JN, Kirsch IR. Karyotypic “state” as a potential determinant for anticancer drug discovery. Proc Natl Acad Sci USA. 2005;102:2964–2969. doi: 10.1073/pnas.0405578102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Wallqvist A, Huang R, Covell DG, Roschke AV, Gelhaus KS, Kirsch IR. Drugs aimed at targeting characteristic karyotypic phenotypes of cancer cells. Mol Cancer Ther. 2005;4:1559–1568. doi: 10.1158/1535-7163.MCT-05-0224. [DOI] [PubMed] [Google Scholar]
- 14.Roschke AV, Kirsch IR. Targeting cancer cells by exploiting karyotypic complexity and chromosomal instability. Cell Cycle. 2005;4:679–682. doi: 10.4161/cc.4.5.1687. [DOI] [PubMed] [Google Scholar]
- 15.Shankavaram UT, Reinhold WC, Nishizuka S, Major S, Morita D, Chary KK, Reimers MA, Scherf U, Kahn A, Dolginow D, et al. Transcript and protein expression profiles of the NCI-60 cancer cell panel: an integromic microarray study. Mol Cancer Ther. 2007;6:820–832. doi: 10.1158/1535-7163.MCT-06-0650. [DOI] [PubMed] [Google Scholar]
- 16.Ihaka R, Gentleman R. R: a language for data analysis and graphics. J Comput Graph Stat. 1996;5:299–314. [Google Scholar]
- 17.Gentleman R, Carey V, Huber W, Irizarry R, Dudoit S. Bioinformatics and Computational Biology Solutions Using R and Bioconductor. New York: NY: Springer; 2005. [Google Scholar]
- 18.Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U, Speed TP. Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics. 2003;4:249–264. doi: 10.1093/biostatistics/4.2.249. [DOI] [PubMed] [Google Scholar]
- 19.Simon R, Lam A. BRB-ArrayTools. Version 3.6. 2007 http://linus.nci.nih.gov/brb.
- 20.Dennis G, Jr, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC, Lempicki RA. DAVID: database for annotation, visualization, and integrated discovery. Genome Biol. 2003;4:P4. [PubMed] [Google Scholar]
- 21.Hosack DA, Dennis G, Jr, Sherman BT, Lane HC, Lempicki RA. Identifying biological themes within lists of genes with EASE. Genome Biol. 2003;4:R70. doi: 10.1186/gb-2003-4-10-r70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Cahill DP, Lengauer C, Yu J, Riggins GJ, Willson JK, Markowitz SD, Kinzler KW, Vogelstein B. Mutations of mitotic checkpoint genes in human cancers. Nature. 1998;392:300–303. doi: 10.1038/32688. [DOI] [PubMed] [Google Scholar]
- 23.Imai Y, Shiratori Y, Kato N, Inoue T, Omata M. Mutational inactivation of mitotic checkpoint genes, hsMAD2 and hBUB1, is rare in sporadic digestive tract cancers. Jpn J Cancer Res. 1999;90:837–840. doi: 10.1111/j.1349-7006.1999.tb00824.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Sato M, Sekido Y, Horio Y, Takahashi M, Saito H, Minna JD, Shimokata K, Hasegawa Y. Infrequent mutation of the hBUB1 and hBUBR1 genes in human lung cancer. Jpn J Cancer Res. 2000;91:504–509. doi: 10.1111/j.1349-7006.2000.tb00974.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Shichiri M, Yoshinaga K, Hisatomi H, Sugihara K, Hirata Y. Genetic and epigenetic inactivation of mitotic checkpoint genes hBUB1 and hBUBR1 and their relationship to survival. Cancer Res. 2002;62:13–17. [PubMed] [Google Scholar]
- 26.Ru HY, Chen RL, Lu WC, Chen JH. hBUB1 defects in leukemia and lymphoma cells. Oncogene. 2002;21:4673–4679. doi: 10.1038/sj.onc.1205585. [DOI] [PubMed] [Google Scholar]
- 27.Wang X, Jin DY, Ng RW, Feng H, Wong YC, Cheung AL, Tsao SW. Significance of MAD2 expression to mitotic checkpoint control in ovarian cancer cells. Cancer Res. 2002;62:1662–1668. [PubMed] [Google Scholar]
- 28.Wang X, Jin DY, Wong YC, Cheung AL, Chun AC, Lo AK, Liu Y, Tsao SW. Correlation of defective mitotic checkpoint with aberrantly reduced expression of MAD2 protein in nasopharyngeal carcinoma cells. Carcinogenesis. 2000;21:2293–2297. doi: 10.1093/carcin/21.12.2293. [DOI] [PubMed] [Google Scholar]
- 29.Michel LS, Liberal V, Chatterjee A, Kirchwegger R, Pasche B, Gerald W, Dobles M, Sorger PK, Murty VV, Benezra R. MAD2 haplo-insufficiency causes premature anaphase and chromosome instability in mammalian cells. Nature. 2001;409:355–359. doi: 10.1038/35053094. [DOI] [PubMed] [Google Scholar]
- 30.Michel L, Benezra R, Diaz-Rodriguez E. MAD2 dependent mitotic checkpoint defects in tumorigenesis and tumor cell death: a double edged sword. Cell Cycle. 2004;3:990–992. [PubMed] [Google Scholar]
- 31.Dai W, Wang Q, Liu T, Swamy M, Fang Y, Xie S, Mahmood R, Yang YM, Xu M, Rao CV. Slippage of mitotic arrest and enhanced tumor development in mice with BubR1 haploinsufficiency. Cancer Res. 2004;64:440–445. doi: 10.1158/0008-5472.can-03-3119. [DOI] [PubMed] [Google Scholar]
- 32.Kalitsis P, Fowler KJ, Griffiths B, Earle E, Chow CW, Jamsen K, Choo KH. Increased chromosome instability but not cancer predisposition in haploinsufficient Bub3 mice. Genes Chromosomes Cancer. 2005;44:29–36. doi: 10.1002/gcc.20215. [DOI] [PubMed] [Google Scholar]
- 33.Schmidt M, Medema RH. Exploiting the compromised spindle assembly checkpoint function of tumor cells: dawn on the horizon? Cell Cycle. 2006;5:159–163. doi: 10.4161/cc.5.2.2309. [DOI] [PubMed] [Google Scholar]
- 34.Chen JJ, Peck K, Hong TM, Yang SC, Sher YP, Shih JY, Wu R, Cheng JL, Roffler SR, Wu CW, et al. Global analysis of gene expression in invasion by a lung cancer model. Cancer Res. 2001;61:5223–5230. [PubMed] [Google Scholar]
- 35.van 't Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M, Peterse HL, van der Kooy K, Marton MJ, Witteveen AT, et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature. 2002;415:530–536. doi: 10.1038/415530a. [DOI] [PubMed] [Google Scholar]
- 36.Kluger HM, Kluger Y, Gilmore-Hebert M, DiVito K, Chang JT, Rodov S, Mironenko O, Kacinski BM, Perkins AS, Sapi E. cDNA microarray analysis of invasive and tumorigenic phenotypes in a breast cancer model. Lab Invest. 2004;84:320–331. doi: 10.1038/labinvest.3700044. [DOI] [PubMed] [Google Scholar]
- 37.Rhodes DR, Yu J, Shanker K, Deshpande N, Varambally R, Ghosh D, Barrette T, Pandey A, Chinnaiyan AM. Large-scale meta-analysis of cancer microarray data identifies common transcriptional profiles of neoplastic transformation and progression. Proc Natl Acad Sci USA. 2004;101:9309–9314. doi: 10.1073/pnas.0401994101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Gupta PB, Kuperwasser C, Brunet JP, Ramaswamy S, Kuo WL, Gray JW, Naber SP, Weinberg RA. The melanocyte differentiation program predisposes to metastasis after neoplastic transformation. Nat Genet. 2005;37:1047–1054. doi: 10.1038/ng1634. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Gupta GP, Massague J. Cancer metastasis: building a framework. Cell. 2006;127:679–695. doi: 10.1016/j.cell.2006.11.001. [DOI] [PubMed] [Google Scholar]
- 40.Huber MA, Kraut N, Beug H. Molecular requirements for epithelialmesenchymal transition during tumor progression. Curr Opin Cell Biol. 2005;17:548–558. doi: 10.1016/j.ceb.2005.08.001. [DOI] [PubMed] [Google Scholar]
- 41.Kopfstein L, Christofori G. Metastasis: cell-autonomous mechanisms versus contributions by the tumor microenvironment. Cell Mol Life Sci. 2006;63:449–468. doi: 10.1007/s00018-005-5296-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
References
- 1.Pollard KS, van der Laan MJ. Cluster analysis of genomic data. In: Gentlemen R, et al., editors. Bioinformatics and Computational Biology Solutions Using R and Bioconductor. New York, NY: Springer; 2005. pp. 209–228. [Google Scholar]
- 2.van der Laan M, Pollard K. A new algorithm for hybrid hierarchical clustering with visualization and bootstrap. J Stat Plan Inference. 2003;117:275–303. [Google Scholar]
- 3.Ross DT, Scherf U, Eisen MB, Perou CM, Rees C, Spellman P, Iyer V, Jeffrey SS, Van de Rijn M, Waltham M, et al. Systematic variation in gene expression patterns in human cancer cell lines. Nat Genet. 2000;24(3):227–235. doi: 10.1038/73432. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.