Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Transcriptional changes associated with breast cancer occur as normal human mammary epithelial cells overcome senescence barriers and become immortalized

Transcriptional changes associated with breast cancer occur as normal human mammary epithelial... Background: Human mammary epithelial cells (HMEC) overcome two well-characterized genetic and epigenetic barriers as they progress from primary cells to fully immortalized cell lines in vitro. Finite lifespan HMEC overcome an Rb-mediated stress-associated senescence barrier (stasis), and a stringent, telomere-length dependent, barrier (agonescence or crisis, depending on p53 status). HMEC that have overcome the second senescence barrier are immortalized. Methods: We have characterized pre-stasis, post-selection (post-stasis, with p16 silenced), and fully immortalized HMEC by transcription profiling and RT-PCR. Four pre-stasis and seven post-selection HMEC samples, along with 10 representatives of fully immortalized breast epithelial cell lines, were profiled using Affymetrix U133A/B chips and compared using both supervised and unsupervised clustering. Datasets were validated by RT-PCR for a select set of genes. Quantitative immunofluorescence was used to assess changes in transcriptional regulators associated with the gene expression changes. Results: The most dramatic and uniform changes we observed were in a set of about 30 genes that are characterized as a "cancer proliferation cluster," which includes genes expressed during mitosis (CDC2, CDC25, MCM2, PLK1) and following DNA damage. The increased expression of these genes was particularly concordant in the fully immortalized lines. Additional changes were observed in IFN-regulated genes in some post-selection and fully immortalized cultures. Nuclear localization was observed for several transcriptional regulators associated with expression of these genes in post-selection and immortalized HMEC, including Rb, Myc, BRCA1, HDAC3 and SP1. Conclusion: Gene expression profiles and cytological changes in related transcriptional regulators indicate that immortalized HMEC resemble non-invasive breast cancers, such as ductal and lobular carcinomas in situ, and are strikingly distinct from finite-lifespan HMEC, particularly with regard to genes involved in proliferation, cell cycle regulation, chromosome structure and the DNA damage response. The comparison of HMEC profiles with lines harboring oncogenic neu changes (e.g. overexpression of Her-2 , loss of p53 expression) identifies genes involved in tissue remodeling as well as proinflamatory cytokines and S100 proteins. Studies on carcinogenesis using immortalized cell lines as starting points or "normal" controls need to account for the significant pre-existing genetic and epigenetic changes inherent in such lines before results can be broadly interpreted. Page 1 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 immortalized using several different pathologically rele- Background Genetic and epigenetic changes that occur early in the vant agents, e.g., chemical carcinogens, over-expression of process of carcinogenesis may enable the survival and the breast cancer-associated oncogenes c-myc and/or growth of cells that subsequently acquire oncogenic muta- ZNF217, and/or inactivation of p53 function [8,9,11]. tions. One early alteration in the development of human Fully immortal HMEC maintain telomeres at short, stable carcinomas is the acquisition of an immortal potential, lengths, but do not necessarily express malignancy-associ- associated with reactivation of endogenous hTERT expres- ated properties; overexpression of specific oncogenes can sion and maintenance of stable telomere lengths. [1]. We confer malignant properties [20-22]. have employed an in vitro HMEC model system to exam- ine gene expression changes during the process of trans- Transcriptional profiling has proven to be a valuable tech- formation of normal finite cells to immortality and nology for describing the differences between cell types malignancy [2-11]. Two mechanistically distinct barriers and experimental treatments for many disease models, to unlimited proliferation have been described. The first particularly cancer [23]. One of the most well-developed barrier, stasis (stress-associated senescence) is associated stratifications of human cancers has been for breast cancer with elevated levels of the cyclin-dependent kinase inhib- [24,25]. These and other studies have shown that a com- INK4A itor (CKI) p16 [6]. Stasis appears to be Rb-mediated mon set of genes is consistently overexpressed in most and not directly dependent on telomere length. Cells cancers [26], including many cell cycle regulated genes arrested at this barrier exhibit a viable G1 arrest with a low and genes required for mitosis (e.g. MKI67, PCNA, BIRC5, labeling index (LI), normal karyotypes, expression of MYBL2, TOP2A, PLK1, MCM2-MCM6, CDC20). The fre- senescence -associated ß-galactosidase (SA-ß-gal) activity, quent identification of these genes in cancer cells suggests and a senescent morphology [7,12]. HMEC can undergo a that they represent a common characteristic of cancers, variable number of population doublings (PD), depend- irrespective of the cell type from which the cancers origi- ing upon culture conditions, prior to encountering stasis. nate. Multiple types of single changes that prevent Rb-mediated The data described here examines the changes that occur growth inhibition will overcome stasis. Loss of CDKN2A as HMEC overcome the barriers to indefinite prolifera- ink4a ) expression, from methylation- (encoding p16 tion. We show that pre-stasis and post-selection HMEC induced CDKN2A promoter silencing, or mutations, is are profoundly different from fully immortalized HMEC one alteration frequently observed in human breast can- lines, despite the fact that the immortalized lines may cers and cultured HMEC [6,13,14]. HMEC cultured in a retain normal growth factor requirements, lack anchor- serum-free medium can produce rare cells that spontane- age-independent growth or invasiveness, and are not tum- ously silence the p16 promoter and resume growth, a origenic in animal models [4]. Rather, the non-malignant process termed selection, with the resulting post-stasis immortalized lines display the cancer-associated prolifer- population called post-selection [3]. In the HMEC, no ation cluster of genes frequently identified in transcrip- ARF increase in p53, p21, or p14 levels have been seen at tional profiling studies of cancer cells and tissues [26]. stasis [7] and p53 function is not required for the stasis barrier (J.G. and M.S., unpublished). Rare HMEC with Materials and methods silenced p16 are also observed in vivo and have been Reagents and supplies MEBM serum-free medium was purchased from the called variant HMEC (vHMEC) [15,16]. Clonetics division of Cambrex BioScience (Walkersville, HMEC that have overcome or bypassed stasis encounter a MD), and was supplemented with EGF, hydrocortisone, second barrier as a consequence of telomere dysfunction. insulin, and BPE using Singlequot reagent packs from Ongoing proliferation in the absence of telomerase Clonetics, as well as 5 µg/ml transferrin (Clonetics) and expression leads to critically shortened telomeres, and 10 nM isopeterenol (Sigma). Hams F-12/DMEM (50:50) chromosomal aberrations [7,17]. In post-selection HMEC was purchased from Invitrogen or prepared by Core Tech- with functional p53, these aberrations induce a mostly nical Services (Wyeth Research), and supplemented to viable G1 and G2 arrest (termed agonescence); if p53 is contain 5% FBS (Invitrogen), 2 mM pyruvate (Invitro- non-functional, massive cell death (crisis) ensues (J.G. gen), 2 mM glutamine (Invitrogen), 20 ng/ml EGF and M.S., unpublished) [18]. Telomere dysfunction poses (Clonetics), 200 µg/ml cholera toxin (Sigma), 1× ITS an extremely stringent barrier to human cellular immor- (Clonetics), 500 ng/ml hydrocortisone (Sigma or Clonet- talization; in post-selection HMEC multiple errors appear ics), and 20 mg/ml gentamycin (Invitrogen). MM to be necessary for telomerase reactivation, and immortal- medium was prepared as described [2]. Antibodies and ization [4,8]. Since this barrier is dependent upon tel- fluorescent dyes used in High Content Screening (HCS, or omere length, ectopic overexpression of hTERT readily quantitiative immunofluorescence) were obtained from immortalizes post-selection HMEC [19]. HMEC can be Cell Signaling Technologies (Beverly, MA), Upstate Bio- Page 2 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 technologies (Lake Placid, NY), and Molecular Probes/ value was then computed for each gene by multiplying its Invitrogen (Carlsbad, CA), as described in the supplemen- original Signal intensity with the scale factor (100/ tary material. Antibodies were screened by Western blot trimmed-mean). Subsequently, genes were filtered to prior to immunofluorescence studies to verify that they remove those with uninformative or noisy expression recognize a single specific antigen of the expected molec- changes across the entire samples. A gene is selected for ular size. downstream analysis if its expression exceeds 50 (scaled) Signal unit in at least one sample. Analysis of variance Cell culture (ANOVA) was performed with log2 transformation on the Pre-stasis and post-selection HMEC, from specimens 48, scaled Signals of several cell lineage groups (see details 161, 184, 191, 195 and 239, as well as the immortally below). Data was analyzed using several analytical transformed lines 184A1, 184AA2, 184AA3, 184B5 were approaches, including unsupervised clustering [30], developed and characterized at LBNL, starting with reduc- supervised clustering [31,32], and principal components tion mammoplasty tissues; an additional post-selection analysis. For the unsupervised clustering, genes that are HMEC strain was obtained from Clonetics. Remaining filtered based on the Pvalues from one-way analysis of lines, as well as additional samples of 184A1 and 184B5 variance (ANOVA) on four cell lineage groups as well as were obtained from ATCC (Manassas, VA). 184B5ME was greater than 2 fold difference among the four groups. +/ derived from immortal 184B5 following stable expression These groups consist of 1) all finite lifespan cells, 2) p53 + -/- of ERBB2/Her2 and selection for anchorage independent immortalized 184A1 and 184B5, 3) p53 immortalized growth (Stampfer, unpublished). Pre-stasis cells were 184AA2 and 184AA3, and 4) immortalized non-184 maintained in MM media [2], and post-selection cells derived cells (including MCF10A, MCF10A-2, and were maintained in MEBM prior to this study. Pre-stasis MCF12A). HMEC display 15–25 PD in MM, and 10–15 PD in Promoter analysis MEBM, prior to growth arrest at stasis. For transcriptional profiling studies, all lines maintained at LBNL (listed Genes identified as unique classes in a subset of post- above), as well as the post-selection HMEC purchased selection HMEC were examined in detail (see Results for a from Clonetics, were revived in MEBM media and cul- complete list of genes). Initially, the 500 bp upstream of . Consequently, the pre-stasis tured at 37°C with 1% CO the transcription start site for each gene was examined for HMEC were studied as they neared stasis. Pre-stasis HMEC well-characterized transcription binding sites using two used in HCS were cultured in MM medium. Fully immor- algorithms, Match and Clover [33,34]. For most of the talized cell lines obtained from ATCC (184A1, 184B5, groups, strong assignments of specific promoter binding MCF10A, MCF10A-2 and MCF12A) were cultured in sites could be identified using both algorithms. One class DMEM/Ham's F-12 medium, at 37°C with 10% CO , as (Class B in the Results) was less definitive, so the region they were maintained prior to crypreservation. was extended to 2 kb prior to the transcription start site for those genes. RNA labeling, GeneChip hybridizations and expression analysis Taqman™ quantitative PCR Cells to be prepared for RNA extraction were revived from Primer sets for 15 genes analyzed by Taqman™ analysis cryopreservation and cultured to 80% confluence in a sin- were obtained from Applied Biosystems (Foster City, CA) gle T-75 flask, trypsinized under conditions appropriate and used according to standard protocols. Genes tested for each line, and split 1:4 into four new T-75 flasks. When are listed in the Results section. cells reached 80% confluence three of the flasks were trypsinized, lysed and total RNA isolated using the High content screening Midiprep RNA isolation kit from Qiagen, according to Cells were seeded at 5000 cells/well in a 96-well black manufacturers instructions. wall, clear bottom Packard ViewPlate, and incubated in MM, MEBM or DMEM/F-12 medium for pre-stasis, post- An 11-point standard curve of bacterial cRNA control selection and immortalized HMEC, respectively, for 48 samples was added prior to hybridization as described hours. Cells were washed with PBS, and fixed with pre- [27,28]. Three independent replicates were generated per warmed 4% paraformaldehyde for 10 minutes. Cells were cell type at the indicated stage. Affymetrix's MAS5 algo- washed 2× with PBS, permeabilized with 0.2% Triton X- rithm was used to generate expression measures including 100 for 3–5 minutes, and washed 2× with PBS again. Cells Signal values and Absent/Present calls (Affymetrix (2001) were stained with primary antibodies in 1% BSA/PBS. Pri- Microarray Suite User Guide, Version 5. [29]. A global scal- mary antibodies were used as follows: E2F1 (BD/Phar- ing normalization was applied to the raw signal intensity. magin, 1:200 dilution), E2F4 (Abcam, 1:400), Rb (Cell Briefly, a 2% trimmed-mean was calculated per chip, and Signaling Technologies, 1:400), p107 (Santa Cruz, was scaled to an arbitrary value of 100. A scaled Signal 1:200), BRCA1 (Abcam, 1:200), p53 (Cell Signaling Tech- Page 3 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 nologies, 1:200), SP1 (Upstate Biotechnologies, 1:400), types, we compared several samples of these cultures by NF-κB (Cellomics, 1:200). Cells were washed 3× with transcriptional profiling; the HMEC samples character- PBST (0.05% Tween-20), and stained with DAPI and sec- ized are described in Table 1. The finite lifespan pre-stasis ondary antibodies of appropriate species/isotype specifi- and post-selection HMEC are referred to as strains or cell city and conjugated to either Alexa-488 or Alexa-594. types from a specific source, and culture conditions Cells were washed again 3× with PBST; 100 µl of PBS was (including stage) are noted for each particular sample. The added and plates were sealed with an adhesive cover. relationships between samples in this study, their origins, are indicated graphically in Figure 1. Triplicate cultures for Quantitative immunofluorescence was performed using a each sample were grown under the conditions indicated ti Cellomics ArrayScan V . Images were taken using a 20× in the Methods, and in Table 1, following which the total objective and data was collected for a minimum of 1000 RNA was isolated, labeled and hybridized to the Affyme- valid cells per well. Valid cells are defined as having nuclei trix U133A/B GeneChips. with expected DNA content (defined by DAPI fluores- cence intensity), nuclei size and shape typical for the cell Principal Component Analysis (PCA) was used to visual- line/type, and well-separated from neighboring cells, such ize the gross relationships among the cell types, as shown that cytoplasmic regions could be clearly resolved. DNA in Figure 2A. The first three components, which explains content and antigen intensity were quantitated for each about 60% of the total variation, are displayed in a three cell, and the nuclear-cytoplasmic ratio for each antigen dimensional graph. The pre-stasis HMEC (in red) and was determined by a mask derived from the DAPI stain- post-selection HMEC (in pink) are clearly separated from ing, which was used to define the nucleus, and a region the immortalized lines (in blue, black and green) along surrounding the nucleus (which was specific for each cell the first principal component axis. Thus, transcriptional line/type) was used to define the cytoplasm. Quantitation profiling defines the transition from finite lifespan to fully was performed using either the Compartmental Analysis immortalized HMEC as the most significant change in or Nuclear Translocation BioApplications, from Cel- HMEC progression. The pre-stasis and post-selection lomics. HMEC are also well segregated within their unique space. In addition, the fully immortalized lines that either do not Results express p53 or are transduced with ERBB2/Her2 (green Transcriptional profiling of pre-stasis, post-selection and and blue, respectively) are distinguished from the rest of immortalized HMEC the immortalized lines (black). According to the PCA, To better understand the extent to which pre-stasis, post- there are no significant differences between the fully selection and immortalized HMEC represent distinct cell immortalized lines derived from various methods of Table 1: Cell Types and Lines Used in This Study Cell Name Source Stage Growth Media 48L LBNL Pre-stasis, finite lifespan strain MM (MEBM)*** 161 LBNL Pre-stasis, finite lifespan strain MM (MEBM) 184 LBNL Pre-stasis, finite lifespan strain MM (MEBM) 195L LBNL Pre-stasis, finite lifespan strain MM (MEBM) 48R LBNL Post-selection, finite lifespan strain MEBM 161 LBNL Post-selection, finite lifespan strain MEBM 184 LBNL Post-selection, finite lifespan strain MEBM 195L LBNL Post-selection, finite lifespan strain MEBM 191 LBNL Post-selection, finite lifespan strain MEBM 239 LBNL Post-selection, finite lifespan strain MEBM HMEC-1001-13 Clonetics Post-selection, finite lifespan strain** MEBM 184A1 LBNL Fully immortal cell line MEBM 184B5 LBNL Fully immortal cell line MEBM 184AA2 LBNL Fully immortal cell line MEBM 184AA3 LBNL Fully immortal cell line MEBM 184B5ME LBNL Fully immortal cell line MEBM 184A1* ATCC Fully immortal cell line DMEM/F-12 184B5* ATCC Fully immortal cell line DMEM/F-12 MCF-10A ATCC Fully immortal cell line DMEM/F-12 MCF-10A-2 ATCC Fully immortal cell line DMEM/F-12 MCF-12A ATCC Fully immortal cell line DMEM/F-12 *designated as 184A1(a) and 184B5(a) in other tables and figures **stage defined by transcriptional profile ***cells isolated from reduction mammoplasty tissues and expanded in MM media to passages 2–3, then cultured in serum-free MEBM media for transcriptional profiling Page 4 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 pd50-100 pd50-100 pd15-25 pd15-25 pd1 pd1 Post-stasis: Population doubling: Pre-Stasis Post-Selection or Conversion Fully Immortalized HMEC Extended Lifespan HMEC stage: Proliferation Block: Stasis Agonesence/Crisis (Epi)Genetic Changes: p16 loss Telomerase reactivation Cell lines used 184A1 p46 184A1 in this study: 48R p9 48L p2 161 p2 161 p7 184B5 p43 184B5 195L p2 195L p6 184 p3 184 p7 MCF10A 184AA2 p44 MCF10A-2 HMEC-1001-13 184AA3 p43 239 p7 MCF12A 191 p7 184B5ME p54 184Aa 184 p1 184Be Comparisons Analyzed: Selection clonal deriviative Immortalization BaP treatment p53 Status retroviral transduction neu transfection and selection for anchorage Her2 Expression independence not analyzed Graphic relationship of cell li Figure 1 nes profiled in this study Graphic relationship of cell lines profiled in this study. Cell lines characterized in this study are shown with reference to their stage in transformation. The pre-stasis HMEC used were cultured for 2–3 passages before analysis, and reach stasis by passages 3–5. Rare isolates of cells grown in serum-free media (MEBM) emerge spontaneously from stasis, associated with the absence of p16 expression due to promoter silencing, and continue growing as post-selection HMEC until reaching a second, proliferation barrier (telomere dysfunction). This barrier is highly stringent, and spontaneous immortalization has never been observed in cells that were not mutagenized or virally transduced during pre-stasis or post-selection growth. HMEC grown in MM do not spontaneously give rise to post-selection cells, however primary populations exposed to the chemical carcinogen benzo(a)pyrene (BaP) have produced rare clonal isolates with post-stasis growth, associated with absence of p16 expression due to mutation or promoter silencing. These non-spontaneously arising post-stasis cells are referred to as extended lifespan, and may harbor additional errors due to the carcinogen exposure. Overcoming the telomere dysfunction barrier is associated with reactivation of telomerase activity. The fully immortalized lines 184A1 and 184B5 were derived from extended lifespan post-stasis cells grown in MM and exposed to BaP in primary culture. Exposure of extended lifespan 184Aa cells to retroviral infection resulted in two cell lines that had lost both copies of the TP53 gene. The cell lines profiled in this study are shown rel- ative to the profiling analyses performed. Comparisons used to analyze selection and immortalization, as well as the influence of p53 and ERBB2/Her2 status are shown by colored boxes and identified in the key at the lower left of the figure. Page 5 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 immortalization, or from lines maintained at LBNL versus Cancer-Testis Antigen 2 (CTAG-2) is very strongly expressed those obtained from ATCC. Unsupervised (or Eisen) clus- (30-fold according to the GeneChip data), as are ARH- tering of the genes that change following selection and GDIB/Ly-GDI, and IGFBP6. The cytokine induced genes immortalization for most of the samples is shown in Fig- [35] include a set previously reported as increasing in ure 2B. These data reflect the 1 342 genes that are filtered post-selection HMEC, such as IFIT1, IFITM1, G1P2 and based on the Pvalues from one-way analysis of variance OAS1 [36]. The genes that are unique to 48 HMEC (Group (ANOVA), as described in the supplementary material. B) include several transcription factors and cell cycle pro- teins whose roles in cancer or breast tissue development Gene expression changes following selection have not been well characterized to date, including Gene expression changes that distinguish pre-stasis from NUCKS, SON and HOXB2. Group C includes many genes post-selection cells were identified using GeneCluster previously associated with cancer cell proliferation. [31], and the results are shown in Figure 3A. The figure characterizes a large set of concordantly-regulated genes in Since these geneset classes were comprised of a relatively the pre-stasis strains, and a high level of concordance in small number of genes, we performed promoter analyses, four of the six post-selection HMEC (200 genes for each to see if these sets are linked in specific pathways. Pro- class). Among these top-200 genes in the pre-stasis cell moter binding sites we were able to identify are listed in types, the largest number of genes we identified are Table 2. For Group A, interferon-responsive elements involved in the extra-cellular matrix (ECM), including were found for most of the genes, but not the cancer/ structural proteins and matrix remodeling enzymes (listed metastasis-associated genes (BST2 is an exception), con- in supplementary Additional file 1). Examples include sistent with previous studies that did not identify these many collagen and kallikrein genes. Genes that increase genes as IFN-regulated [35]. Instead, several genes in this expression level in post-selection HMEC include a large group have been shown to be direct or indirect targets of number of genes associated with proliferation and the cell p53 and Myc. A common element in the regulation of cycle. These genes are strongly associated with cancer cell both p53/Myc and IFN-regulated genes is BRCA1, and in growth, and increase in expression directly with tumor particular, BRCA1 is essential for the activation of stress grade. Specific examples include BIRC5, A and B type cyc- and inflammatory response genes following treatment lin genes, CDC2, and the MCM chromosomal proteins. with interferons [37]. Group B was less well-defined by The increased expression of these genes is dependent on specific binding sites near the promoter, but an extended E2F transcription factors and reflects the proliferative state analysis (2 kb) identified SP1, E2F, MAZ and NF-Y bind- a cell. Since the pre-stasis cells were nearing stasis, the ing sites for many genes. These binding sites were also increased expression of the genes in the post-selection identified in the genes of Group C, especially the E2F, NF- HMEC may reflect either a loss of Rb repression (consist- Y and SP1 sites, which is consistent previous work [38,39]. ent with a loss of p16), or could reflect the relative prolif- Group D, genes significantly repressed in post-selection erative state of these pre-stasis and post-selection cells. HMEC, may be under the control of MAZ (Myc-associated zinc finger protein), as binding sites were found in 19 of The two discordant post-selection HMEC we observed in 22 genes examined, which is consistent with previous Figure 3A (195L and 1001-13), suggest that additional observations that increased Myc can repress ECM genes molecular events can occur during selection; these sam- [40-42]. In conclusion, although distinct gene expression ples also show a loss of p16 expression (results not patterns could be observed for each of the pre-stasis/post- shown), a definitive event for post-selection HMEC. In selection HMEC pairs we have characterized, in each case order to probe further into the changes that occur during strong associations could be made between the promoters selection, we compared the four sets of HMEC studied as of each class and the proliferation and cell cycle transcrip- pre-stasis and post-selection samples. For this analysis, we tion factors, particularly E2F, SP-1, NF-Y and the Myc- identified genes that increase expression in post-selection related MAZ. The distinguishing features for each of these HMEC, as compared to the corresponding pre-selection expression classes is likely to be found in additional, sample. Four patterns were observed. The genes we iden- unique pathways such as BRCA1-mediated regulation. tified in each group are listed in Table 2, and the expres- sion changes we observe for three of the groups are shown Gene expression changes that distinguish finite life span in Figure 3B. The group not explicitly shown in Figure 3B HMEC from immortally transformed HMEC is uniformly down-regulated in all four pairs. Genes The most significant transition observed in this study is expressed exclusively in post-selection 195L HMEC that of immortalization. Genes whose expression are (Group A) fall into two categories: genes previously iden- reduced in the immortalized lines include a significant tified as cancer-associated (including several antigens pro- number that suppress angiogenesis, contribute to the posed as cancer biomarkers), and genes induced by ECM, or regulate the actin cytoskeleton. Many of these interferons [35]. Among the cancer-associated genes, the genes were identified as down-regulated in HMEC follow- Page 6 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 Pr Pre e- -s st ta as si is s Po Pos st t- -s se el le ec ct ti io on n I Im mm mo o r rt ta al li iz ze ed, p53- d, p53- I Im mm mo o rt rta al li iz ze ed d, , H He er2 r2+ ++ + Im Imm mo or rtta alliiz ze ed d pre-stasis post-selection immortalized Re Figure 2 lationship of HMEC as determined by transcriptional profiles Relationship of HMEC as determined by transcriptional profiles. A. Data from 2319 genes were used to determine the number of principal components of the data. Three components were identified, and the contribution of the components to the transcription profile of each cell line samples are shown in the figure. Individual replicates for each cell line are shown. Cell lines grouped in Figure 1 are shown in Figure 2A as shown in the legend. Vertical axis is PC1, the first, and therefore the strongest. principal component. B. Unsupervised clustering of HMEC. All genes that change expression in one or more sam- ples were used to cluster the cell types and lines by overall similarity. Cell types and lines are identified by color under the des- -/- ignations: pre-stasis HMEC: light green; post-selection HMEC: light blue; fully immortalized HMEC: dark blue; p53 fully immortal HMEC: burgundy, and lines not formally characterized: black. Samples of 184A1 and 184B5 designated by (a) were obtained from ATCC. Page 7 of 17 (page number not for citation purposes) 195L 48L 48R 1001-13 195L 184A1 184B5 184A1(a) 184B5(a) MCF-12A MCF-10A MCF-10A-2 184AA2 184AA3 Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 A B CTAG2 CTAG1, CTAG2 CTAG1, CTAG2 IFIT1 BST2 C1orf29 ARHGDIB G1P2 G1P3 MMP7 NUCKS TRAP150 HOXB2 CA12 LZ16 IF2 SON BRD4 PIST BX TOP2A RRM2 RRM2 KIF20A TOP2A BIRC5 ANKT CCNA2 CDC2 MKI67 161 184 195 48 pre-stasis post-selection Supervised Cluster Figure 3 ing of Pre-stasis, and Post-selection HMEC Supervised Clustering of Pre-stasis, and Post-selection HMEC. A. Gene expression values were normalized and character- ized for the significance of overexpression in one group relative to other groups in the comparison. The top 50 genes that are significantly overexpressed in one group are shown. All pre-stasis and post-selection cell types have been used. Analysis was performed in GeneCluster, and the color bar describing how normalized values are depicted is shown at the bottom of the fig- ure. B. Distinct classes of genes over-expressed in post-selection HMEC. Genes showing one of three specific patterns of expression in the four pairs of pre-stasis and post-selection samples are diagramed. The top 10 qualifiers (based on fold change) are shown (some genes are represented by more than one qualifier). ing selection as well; some are further down-regulated in commonly observed "proliferation cluster" described the immortalized lines, as shown in Figure 4A. These com- above. These genes were also observed to be up-regulated parisons include multiple independent samples from in the post-selection, compared to pre-stasis HMEC. Fewer each stage, including four distinct fully immortalized cell of these "proliferation genes" are identified in the fully lines, and three additional samples from either different immortalized samples following a three-way comparison, sources (184A1 and 184B5 from ATCC) or two separate but this is because GeneCluster identifies the most defin- isolates from the same experiment (MCF-10A and MCF- itive group of genes for each class, and since some of the 10A-2) [43]. The genes identified in each group are post-selection samples express increased levels of genes described in Additional file 2. Collectively, the pre-stasis such as MCM2 and STK12, they are not unique to either and post-selection samples are distinguished most the post-selection or the fully immortalized HMEC. strongly by changes to the ECM and cell-cell communica- tion genes, particularly collagens, kallikrein, matrix metal- We have examined the expression of the cancer cell prolif- loproteinase and serpin proteinases; genes that affect the eration class of genes directly in Figure 4B. In this exam- actin cytoskeleton are also noted (both actin and actin- ple, the absolute expression levels of each gene listed in interactors, such as actinin, nidogen, transgelin, and pal- the figure are displayed directly (rather than the ratio of ladin, genes). Several well-recognized classes of genes are post-selection over pre-stasis expression levels in Figure up-regulated in fully immortalized lines, including the 3B). These genes are compared to equal subsets of genes Page 8 of 17 (page number not for citation purposes) Group C Group B Group A 48L 195L 48R 195L 1001-13 Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 Table 2: Genes and Promoter Elements That Define Post-Selection HMEC Gene Expression Classes Geneset Classes Genes Promoter Elements Group A IFN genes IFIT1, BST2, G1P2, G1P3, IFIT2, OAS1, IFI44, IFIT4 IRF, ILR, IRL Non-IFN genes CTAG2, ARHGDI-B/Ly-GDI, MMP7, PLAU, CALB1, SLC1A6, MDA5, FXYD5, HMOX1 Group B NUCKS, HDAC3, TRAP150, HOXB2, SON, IF2, LZ16, ANLN, BBX, TOP1, H4FG, SFRP1, KTN1, SP1 GTAR, BAZ1A, PK428, FALZ, TTC3, DNCL12, RBM9 Group C TOP2A, RRM2, KIF20A, BIRC5, ANKT, CCNA2, CDC2, MKI67, CDC20, MCM5, HMMR, IL-1B, E2F, NF-Y, B-Myb PRC1, PMSCL1, MADL1, DLG7 Group D H11, COL11A1, IGFBP5, CNN1, COMP, LGALS7, CLDN7, KLK6, KLK7, KLK10, KLK11 KRT23, MAZ, MAZR, MEF-3 LOXL4, THY1, FLJ21841 pre-stasis post-selection immortalized COL1A2 COL2A1 COL4A1 COL6A1 COL6A2 COL11A1 ID4 IGFBP3 IGFBP5 ITGB3 KLK6 KLK7 KLK10 KLK11 KRT8 KRT16 KRT19 KRT23 PMP22 SERPIN G1 APP ACTN4 C COL4A5 OL4A5 COL4A6 COL8A1 COL12A1 COL16A1 COL17A1 HLA-A HLA-B HLA-C HLA-F HLA-G IGTA3 ITGAV ITGB6 MMP2 SERPIN E1 SERPIN E2 SERPIN F1 BIRC5 BUB1 BUB1B CCNA2 CCNB1 CCNB2 CDC2 CDC20 CDC25B CDC6 CENPA CENPF MCM2 MCM4 MPHOSH1 NEK2 RRM2 STK6 STK12 TOP2A pre-stasis post-selection immortalized Supervised Cluster Figure 4 ing of Pre-stasis, Post-selection and Immortalized HMEC Supervised Clustering of Pre-stasis, Post-selection and Immortalized HMEC. A. Gene expression values were normalized and characterized for the significance of over-expression in one group relative to other groups in the comparison. The top 200 genes (of 1342) that are significantly over-expressed in one group are shown. All pre-stasis, post-selection and immortalized -/- HMEC (except the p53 and ERBB2/Her2 transfected variants) have been grouped. The top 100 genes (of 1440) that are over- expressed in one group relative to the other two are presented. Analysis was performed in GeneCluster. B. Expression of a subset of highly concordant genes in pre-stasis, post-selection and fully immortalized HMEC. Gene-normalized expression of 60 genes identified in the figure are shown for four representatives each for the three groups of HMEC. Samples are (left to right): 48L, 161, 195L and 184; 48R, 161, 195L, and 184; 184A1, 184B5, MCF-10A and MCF12A. Page 9 of 17 (page number not for citation purposes) 48L 195L 48R 195L 1001-13 184A1 184B5 184A1(a) 184B5(a) MCF-10A MCF-10A-2 MCF-12A cancer proliferation cluster genes Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 that show maximal levels of expression in the pre-stasis genicity [21]. Gene expression changes seen for 184B5ME and post-selection HMEC samples. As can be observed in that are distinct from its parent are listed in the supple- the figure, genes showing maximal expression in the pre- mentary Additional file 6. Genes showing increased stasis samples are robust, whereas those showing maximal expression include many that were down-regulated in expression in the post-selection are less strongly definitive post-selection HMEC, including kallikreins KLK6 and of post-selection cells. The "proliferation cluster" genes KLK7, and cystatin E/M. These phenotypic reversions may show strongest expression in the fully immortalized play a role in the transition to invasive cancer [44]. Addi- HMEC lines, however expression of these genes is hetero- tional gene expression changes include a dramatic geneous for both the post-selection and fully immortal- increase in the expression of IL24 and significant changes ized sets. Increased expression can be observed for the in BIRC3, HRASLS3, and PTGES. Genes showing down- post-selection 48R and 184 samples (as was seen for some regulation as a consequence of ERBB2/Her2 overexpres- of these genes in Figure 3B), and lesser expression is seen sion include many of the IFN genes that showed increased for MCF-12A. However, the rise in expression of this expression following selection (in 195L) or immortaliza- group of genes as HMEC progress from pre-stasis through tion (in 184A1, 184B5 and others). fully immortalized stages is clear. Real-time PCR measurement for selected genes identified -/- Gene expression changes observed in p53 cell lines in this study HMEC lines that have lost p53 during immortalization The results presented comprise a large study of human show distinctive changes in transcriptional profiles when mammary cell samples that have not been characterized compared to closely related lines that have retained p53 by transcriptional profiling previously, and the gene function. The complete list of genes is presented in the expression patterns are either new or not previously asso- supplementary Additional file 3. When we explicitly look ciated with non-cancerous cell lines. As such we wished to for genes whose expression changes are common to the validate the findings by corroborating the gene expression p53 status of the lines derived from specimen 184 cells, changes observed by genechips with an independent +/ several genes showing concordant changes between p53 method. 15 genes were chosen from the data to be vali- + -/- 184A1 and 184B5 versus p53 184AA2 and 184AA3 are dated by Taqman™ quantitative PCR. Genes that change observed. SIAH2, Lipocalin 2, Asparagine synthase and Ker- following selection (PMP22/GAS3 and several insulin-like atin 15 are all upregulated in both 184AA2 and 184AA3, growth factor binding protein (IGFBP) genes: IGFBP2, relative to both 184A1 and184B5. Genes down-regulated IGFBP3, IGFBP4, IGFBP5, IGFBP6, and IGFBP7), as well as -/- in the p53 lines include several that are explicitly regu- genes that change in immortalized lines (CCNB1, CDC2, lated by p53 (including RRM2 and TP53INP1). A compar- CDC25B, HDAC3, MYC, and STK6) were evaluated by RT- +/+ -/- ison of the two p53 and the two p53 lines shows that PCR in 17 cell types, comprising pre-stasis, post-selection additional gene expression changes unique to each line and fully immortalized samples, and the results compared have occurred. Examples include DUSP1 and BIRC3, to expression data from the oligonucleotide arrays. The expressed at significantly higher levels 184AA3 than in concordance between expression of a gene as measured by 184AA2, and FABP4, IFI27, HRASLS3, and Fibulin 1, oligonucleotide array and Taqman™ assays were generally expressed much more robustly in 184A1 than in 184B5. quite good; in 14 cases, only minor discordances can be The complete list of genes is presented in the supplemen- observed (see Figure 5). HDAC3 was an exception. The tary Additional file 4 and Additional file 5. expression level changes of three probes sets for HDAC3 on the Affymetrix U133 GeneArrays, and the Taqman™ Gene expression changes resulting from ectopic expression primer set, were highly discordant, so we were not able to of Her2 validate the expression changes of this gene by RT-PCR, The events characterized thus far in this study concern however were able to show significant changes in HDAC3 HMEC immortalization; however, additional events are protein expression and localization by immunofluores- critical to malignancy. To connect these studies directly to cence microscopy (described below). changes that occur following an oncogenic event, we have Transcriptional regulatory factors are localized to the compared one immortalized HMEC line, 184B5, with a derivative that ectopically expresses the ERBB2/Her2 nucleus following selection and immortalization oncogene, 184B5ME. ERBB2/Her2 is frequently over- We explored the changes that occur in several critical reg- expressed in breast cancer, and is transforming simply by ulators of cell cycle progression and chromosomal stabil- being over-expressed, so this line models clinically rele- ity by quantitative fluorescence microscopy, or High vant features of breast cancer. Over-expression of ERBB2/ Content Screening (HCS). These factors were chosen Her2 in 184B5 results in anchorage independent growth, based on patterns observed in the transcription profiling a malignancy-associated property, while over-expression data as ones that would be expected to change as HMEC of oncogenic ERBB2/Her2 in 184B5 can confer tumori- progress past senescence barriers, based on the gene Page 10 of 17 (page number not for citation purposes) MCF-12A MCF-10A-2 MCF-10A 184AA3 184AA2 184B5ME 184B5 184A1 MCF-12A MCF-10A-2 MCF-10A 184AA3 184AA2 184B5ME 184B5 184A1 Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 48 161 184 195L 48 161 184 195L pre-stasis post-selection immortalized Real-time PCR me Figure 5 asurements of gene identified in transcriptional profiling analyses Real-time PCR measurements of gene identified in transcriptional profiling analyses. Representative genes from groups identified as changing expression during selection or immortalization were characterized by real-time PCR analysis (Taq- Man™). Genes were selected as representative of classes were described in this study. Each gene is presented as a separate graph, as identified in the figure. Cell lines are presented in the same order in each graph, as listed in the bottom left panel. The finite lifespan samples are shown as pairs, with the pre-stasis sample on the left and the post-selection sample on the right. For each cell line, expression data from Affymetrix GeneChips are shown as blue bars, according to the scale at the left of the graphs. Expression data from real-time PCR of the same samples are shown as yellow bars, according to the scale at the right of the graphs. Page 11 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 expression patterns we observe. Example images are HMEC). Levels of IGFBP4 were significantly reduced in shown in Figure 6A. For these images, Rb is shown in red 184B5ME relative to 184B5. IGFBPs are frequently and DNA is shown in blue. In the pre-stasis 184 HMEC, observed to be reduced in breast cancers, and these reduc- Rb is punctate and is evenly distributed between the tions are associated with increased sensitivity to IGF-I and nucleus and cytoplasm. In post-selection 184 HMEC and IGF-II [49,50]. in immortalized lines such as 184A1 (shown in the figure) and 184B5 (not shown), Rb is very strongly localized to (B) BRCA1, a gene deleted in about 5% of women with the nucleus, and the staining is no longer punctate. The breast cancer, encodes a protein that interacts with many nuclear/cytoplasmic ratio (determined using least 1000 other proteins [51]. These complexes recognize and cells per sample for three samples each) are shown in Fig- orchestrate the repair of DNA damage. Many genes that ure 6B for Rb and 8 other proteins. The ratio for Rb in pre- encode proteins that interact with BRCA1 were identified stasis cells is 0.5–2, whereas for post-selection and in this study as genes that increase expression following immortalized HMEC it is greater than 100. Similar dra- either selection or immortalization. BAP, RAD51, CSE1L matic changes are observed for HDAC3, BRCA1, p53 and and RFC4 all increased expression following selection in a the general transcription factor SP1. BRCA1 and c-Myc are pattern similar to the E2F-regulated genes identified as localized in the cytoplasm in pre-stasis HMEC, but to the Group C in Figure 3B. MYC, RAD50 and RFC3 increased nucleus in post-selection and immortalized HMEC. For expression in fully immortalized lines, including the p53 /- other proteins associated with G1 progression (E2F1, lines. These changes suggest the possibility that BRCA1- E2F4 and p107), the differential is in the range of two to mediated functions are affected by overcoming stasis and/ four-fold. or immortalization, which is supported by the significant change in localization of BRCA1 to the nucleus in post- Discussion selection HMEC. Transcriptional profiles and quantitative immunofluoresence of HMEC reveal significant cancer- (C) The increased expression of a well-characterized clus- associated changes following both selection and ter of IFN-regulated genes was observed in some lines in immortalization this study, as well as in other studies of HMEC [36], and The effect of malignant transformation (oncogenesis) on in a taxol-resistant MCF-7 line [52]. The IFN-dependent gene expression has been studied extensively in both cell stress response is mediated by BRCA1 [37,53]. Therefore, lines and tissues in an effort to characterize the causes of since we have noted expression changes in many genes cancer at the molecular level [45]. Gene signatures com- associated with BRCA1 function, as well as in BRCA1 monly found in breast and other human cancers include abundance and localization in post-selection HMEC, IFN those critical for the cell cycle, chromosomal stability and gene signature may reflect changes in BRCA1-mediated proliferation; the extent of the increase in the expression functions. of this signature correlates with tumor grade and poorer prognosis [26,46]. A separate signature of IFN-regulated (D) Inhibitors of Differentiation (ID) genes are important genes has also been observed in ductal carcinoma in situ regulators of differentiation by dominantly interfering (DCIS) [47] and has been associated with metastasis to with the function of bHLH proteins during embryogene- the lymph nodes in aggressive breast cancers [48]. We sis, neurodevelopment and cancer. Part of their function have observed both of these signatures in non-malignant, is through the repression of CKIs, including p16. Some immortally transformed, HMEC lines that had overcome functions have been attributed to specific members, the two senescence barriers to immortalization, despite including the interaction of ID2 with Rb [54], and the these lines retaining many characteristics of finite lifespan expression of BRCA1 by ID4 [55], which is in turn epithelial cells. repressed by BRCA1 [56]. In this study, ID1 is expressed at higher levels in the immortalized lines (184AA2 is an Transcriptional changes in gene families associated with exception), while ID4 is repressed in post-selection HMEC mammary epithelial biology or breast cancer in post- and all of the immortalized lines. selection and fully immortalized HMEC There are several gene families that we identified in this (E) S100 proteins comprise a large family of calcium-acti- study which have direct connections to breast epithelial vated proteins that function in homo- and hetero-dimers biology and breast cancer, which we can summarize: to regulate many intra- and extra-cellular targets [57]. Their increased expression in cancer and inflammatory (A) Several IGFBPs show reduced expression in post-selec- diseases has provoked interest in this family as potential tion HMEC and immortalized lines, including IGFBP2 drug targets and clinical biomarkers. We observe increases -/- (minor decreases overall, but larger in the p53 lines), in the expression of S100A8 and S100A9, which comprise IGFBP3 and IGFBP5 (very large decreases in immortal the heterodimer Calprotectin, following selection and fur- Page 12 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 Rb BRCA1 p53 1.2 2.5 1.0 2.0 2.0 0.8 1.5 0.6 1.5 0.4 1.0 1.0 0.2 0.5 0.5 0 -0.2 184 161 184 161 184A1 184B5 184 161 184 161 184A1 184B5 184 161 184 161 184A1 184B5 p107 E2F1 E2F4 1.0 1.0 1.0 0.9 0.9 0.9 0.8 0.8 0.8 0.7 0.7 0.7 0.6 0.6 0.6 0.4 0.4 0.4 0.3 0.3 0.3 0.2 0.2 0.2 0.1 0.1 0.1 184 161 184 161 184A1 184B5 184 161 184 161 184A1 184B5 184 161 184 161 184A1 184B5 HDAC3 Myc SP1 2.0 0.6 2.5 0.5 1.5 2.0 0.4 0.3 1.5 1.0 0.2 1.0 0.1 0.5 0.5 -0.1 -0.2 184 161 184 161 184A1184B5 184 161 184 161 184A1184B5 184 161 184 161 184A1184B5 pre-stasis post-selection immortalized H Figure 6 igh Content Screening of proteins associated with cell cycle progression and chromosomal stability High Content Screening of proteins associated with cell cycle progression and chromosomal stability. (A) Immunofluo- rescent images of Rb (red) and DNA (blue) obtained using a Cellomics ArrayScan Vti are shown for pre-stasis 184 HMEC (left), post-selection 184 HMEC (center) and the 184A1 cell line (right). (B) Quantitation of the nuclear/cytoplasmic ratio is shown for pre-stasis 184 and 161 HMEC, post-stasis 184 and 161 HMEC and the cell lines 184A1 and 184B5, as indicated in the figure panels. Antigens quantitated in each panel are identified above the panel. Page 13 of 17 (page number not for citation purposes) log (nuclear/cytoplasmic ratio) log (nuclear/cytoplasmic ratio) log (nuclear/cytoplasmic ratio) 10 10 log (nuclear/cytoplasmic ratio) log (nuclear/cytoplasmic ratio) log (nuclear/cytoplasmic ratio) 10 10 log (nuclear/cytoplasmic ratio) log (nuclear/cytoplasmic ratio) log (nuclear/cytoplasmic ratio) 10 10 Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 ther dramatic increases following immortalization. regulatory factor localizations we observe are concordant. Increased expression of S100P is seen in DCIS [58], and Proteins directly responsive to p16/CDK4 activation, par- was also observed in several of the immortalized lines, ticularly Rb, show striking changes in cytoplasmic/nuclear particularly 184B5ME, the ERBB2/Her2 transduced line. distribution in both post-selection and fully immortalized Increased expression of S100A7, also known as psoriasin, HMEC, compared to pre-stasis HMEC. Additional pro- is seen in both DCIS and IDC, particularly ER negative teins also showing strong changes in localization are breast cancers [59]; increased expression was observed in BRCA1, p53, HDAC3, Myc and SP1. Each of these pro- several immortalized lines, most strongly in 184AA3. teins have well characterized roles in oncogenesis and in the regulation of hTERT [63-66], a critical event in immor- Transcriptional changes that occur following genetic talization [1,5]. These changes are consistent with both changes associated with invasive cancer the transcriptional profiles we have generated of post- p53 imposes a cell cycle arrest when chromosomal break- selection and fully immortal HMEC, as well as with what age or damage is detected, and its loss in breast cancer is is known about the role of these factors on telomerase reg- associated with increased chromosomal instability and a ulation. -/- lines we have more aggressive subtype [60]. The two p53 characterized show a number of transcriptional changes The relationship between immortalized HMEC and DCIS -/- that are expected of p53 cell lines, as well as changes Taken together, these data support a classification of unique to the two lines. Of note is expression of the IFN- immortalized breast epithelial cell lines as in vitro models induced genes observed in post-selection 195L cells and of highly dysregulated epithelial cells, rather than as per- in the 184AA3 line. This may indicate a common molec- petually growing models of normal breast epithelia. Gene ular event occurred following selection of the 195L cells expression patterns we have identified in the comparison and the immortalization of the 184AA3 cells. Further of finite-lifespan and immortalized HMEC lines are -/- studies on the changes common and unique to p53 highly similar to changes observed in DCIS and invasive HMEC lines may be important in understanding differ- human breast cancers [47,67,68], and are consistent with +/+ -/- ences between p53 and p53 cell lines and breast can- other similarities between immortal HMEC lines and cers in overcoming senescence barriers and DCIS. Specifically, short telomeres and moderate chromo- immortalizing. somal instability, as well as telomerase re-activation, are common to many early-stage tumors [69], including the In data presented here, transfection of an immortalized breast [17]. In addition, p16 expression is lost in post- line with a clinically-relevant oncogene, ERBB2/Her2, selection, as it is in vHMEC [15,16], which are proposed showed fewer transcriptional changes than were observed to be premalignant breast cancer precursors in vivo. In con- following selection or immortalization, and these changes trast, we observe that a cell line, 184B5ME, which grows were generally limited to genes involved in invasive invasively in tissue culture and in in vivo models, shows growth and motility. Specifically, expression of the prolif- fewer changes. eration geneset was not dramatically altered, but there was increased expression of genes encoding the secreted pro- DCIS is a complex disease [70], often requiring no imme- teases Cystatin E/M, and Kallikrein 6, as well as tissue diate treatment in the strict sense, however it is not cur- plasminogen activator. Such changes could enable these rently possible to forecast when, or if, progression to IDC cells to grow invasively in breast tissue. will occur. This necessitates an aggressive strategy, even in cases where it may be effectively managed by substantially Activation of transcriptional regulators associated with simpler, cheaper, and less emotionally challenging modes gene expression changes in post-selection and [71]. The ability to characterize DCIS, and to target it immortalized HMEC, telomerase reactivation and cancer explicitly when it manifests invasive potential, is a critical In quiescent or unstimulated cells, many transcription fac- need with regard to effective breast cancer treatment strat- tors are excluded from the nucleus and localize to the egies. In particular, established markers for breast cancer, neu nucleus upon activation [61]. In the case of BRCA1, including Ki-67, p53, Her-2 and ER expression are very nuclear retention has been shown to suppress its pro- effective for identifying aggressive, invasive cancers, and apoptotic functions [62]. The proliferation, cell cycle and for determining the most effect treatment strategy in these DNA damage response genes identified in the gene cases, but are less informative about the likelihood that a expression signatures we observe are supported by the well-contained DCIS will progress to invasive cancer. Cur- changes in the localization of several associated regulatory rently, some of the best indicators of DCIS progression proteins and transcription factors, as determined by quan- risk are cytological, including grade, necrosis and architec- titative immunofluorescence. Based on previous studies tural patterns [72]. Additional molecular markers, partic- linking regulatory pathways to gene expression, the rela- ularly those that correlate strongly (or better, explain) the tionship between the gene expression signatures and the histological patterns used to stage DCIS would be very val- Page 14 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 uable. Some additional molecular markers are emerging. Abbreviations COX-2 has been identified as a marker of vHMEC [15,16], LI, labeling index; HMEC, human mammary epithelial and expression levels have been correlated with DCIS cells; CKI, cyclin-dependent kinase inhibitor, DCIS, duc- grade, as well [73]. For these reasons, recognizing immor- tal carcinoma in situ; IDC, invasive ductal carcinoma; PD, talized HMEC as resembling early-stage cancers would population doubling; ANOVA, analysis of variance; CIN, facilitate a formal interrogation of their genetics and phys- chromosomal instability; IGFBP; insulin-like growth fac- iology for clues to how DCIS occurs, and to the factors tor binding protein; SA-b-gal, senescence associated beta- that can enable DCIS to progress. galactosidase; ECM, extracellular matrix; HCS, high con- tent screening. Use of post-selection and immortalized HMEC to study normal mammary cell biology and breast cancer Competing interests Immortalized cell lines have been used to address com- The author(s) declare that they have no competing inter- plex problems in cancer [74] and epithelial cell biology ests. [75] precisely because they allow for controlled experi- ments to be performed and theories of breast cancer to be Authors' contributions tested. In studies of oncogenesis, the non-malignant sta- JP, JHL, CT, and KS performed experiments and analyzed tus of immortalized lines allows for the specific steps in primary data. YL, J-JL, MW, SJ and SH analyzed normal- full malignant transformation to be examined, such as by ized data and interpreted results. JG and MS developed the introduction of activated oncogenes [76,77]. How- cell lines and analyzed normalized data. JP performed ever, in many cases immortalized cell lines are referred to cell-based assays on the transcription factors and regula- and used as "normal" cells. This inaccurate characteriza- tory proteins. JP and SH analyzed data from the cell-based tion may obscure understanding of the multiple errors assays. YL, MS and SH wrote the manuscript. All authors that permit immortal transformation, and thus aspects of read and approved the final version of the manuscript. early stage carcinogenesis. While established breast cancer cell lines are usually derived from advanced, metastatic Additional material tumors (particularly pleural effusions), and therefore are quite different from immortalized cell lines, immortal- Additional file 1 ized lines themselves have undergone extensive genetic Table s1: Genes Expressed Concordantly in Pre-Stasis and Post-Selection and epigenetic changes, especially in frequently studied Cell Types. Compilation of genelists that distinguish the two classes of aspects of oncogenesis, such as G1 checkpoint function finite-lifespan HMEC strains. and the DNA damage response. The use of immortalized Click here for file [http://www.biomedcentral.com/content/supplementary/1476- HMEC as "normal" controls for tumor-derived lines can 4598-6-7-S1.doc] impede our ability to understand early stages of carcino- genesis, and obscure the potential of treating DCIS-stage Additional file 2 changes as additional targets for clinical benefit. Table s2. Genes Concordantly Expressed in Pre-stasis, Post-selection or Fully Immortalized HMEC. Compilations of genelists that define the three Conclusion classes of non-malignant HMEC cell strains and lines. Gene expression profiles and cytological changes in Click here for file related transcriptional regulators indicate that immortal- [http://www.biomedcentral.com/content/supplementary/1476- 4598-6-7-S2.doc] ized HMEC resemble non-invasive breast cancers, such as ductal and lobular carcinomas in situ, and are strikingly Additional file 3 distinct from finite-lifespan HMEC, particularly with +/+ Table s3: Genes Expressed Concordantly in p53 (184A1 and 184B5) regard to genes involved in proliferation, cell cycle regula- -/- or p53 (184AA2 and 184AA3) HMEC. Compilations of commonly tion, chromosome structure and the DNA damage expressed genes in multiple wild type and p53 immortalized HMEC cell response. The comparison of HMEC profiles with lines lines. harboring oncogenic changes (e.g. overexpression of Her- Click here for file neu [http://www.biomedcentral.com/content/supplementary/1476- 2 , loss of p53 expression) identifies genes involved in 4598-6-7-S3.doc] tissue remodeling as well as proinflamatory cytokines and S100 proteins. Studies on carcinogenesis using immortal- Additional file 4 ized cell lines as starting points or "normal" controls need Table s4. Gene Expression changes of p53 cell lines 184A1 versus 184B5. to account for the significant pre-existing genetic and epi- Compilations of genelists and expression statistics of genes expressed genetic changes inherent in such lines before results can uniquely in two p53 wild type HMEC cell lines. be broadly interpreted. Click here for file [http://www.biomedcentral.com/content/supplementary/1476- 4598-6-7-S4.doc] Page 15 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 13. Geradts J, Wilson PA: High frequency of aberrant p16INK4A expression in human breast cancer. American Journal of Pathology Additional file 5 1996, 149:15-20. Table s5. Gene Expression changes of p53 cell lines 184A1 versus 184B5. 14. Baylin SB, Herman JG, Graff JR, Vertino PM, Issa JP: Alterations in DNA methylation: a fundemental aspect of neoplasia. Advances Compilations of genelists and expression statistics of genes expressed in Cancer Research 1998, 72:141-196. uniquely in two p53 HMEC cell lines. 15. Crawford YG, Gauthier ML, Joubel A, Mantei K, Kozakiewicz K, Afshar Click here for file CA, Tlsty TD: Histologically normal human mammary epithelia with silenced p16(INK4a) overexpress COX-2, promoting a [http://www.biomedcentral.com/content/supplementary/1476- premalignant program. Cancer Cell 2004, 5:263-273. 4598-6-7-S5.doc] 16. Holst CR, Nuovo GJ, Esteller M, Chew K, Baylin SB, Herman JG, Tlsty TD: Methylation of p16(INK4a) promoters occurs in vivo in histologically normal human mammary epithelia. Cancer Additional file 6 Research 2003, 63:1596-1601. Table s6. Gene Expression Changes Resulting from Expression of ERB-B2/ 17. Chin K, de Solorzano CO, Knowles D, Jones A, Chou W, Rodriguez neu Her2 in 184B5. Compilations of genelists and expression statistics of EG, Kuo WL, Ljung BM, Chew K, Myambo K, Miranda M, Krig S, Garbe neu genes that change expression following overexpression of the Her-2 J, Stampfer M, Yaswen P, Gray JW, Lockett SJ: In situ analyses of genome instability in breast cancer. Nature Genetics 2004, oncogene. 36:984-988. Click here for file 18. Goldstein JC, Rodier F, Garbe JC, Stampfer MR, Campisi J: Caspase- [http://www.biomedcentral.com/content/supplementary/1476- independent cytochrome c release is a sensitive measure of low-level apoptosis in cell culture models. Aging Cell 2005, 4598-6-7-S6.doc] 4:217-222. 19. Stampfer MR, Garbe J, Levine G, Lichtsteiner S, Vasserot AP, Yaswen P: Expression of the telomerase catalytic subunit, hTERT, induces resistance to transforming growth factor beta growth inhibition in p16INK4A(-) human mammary epithelial cells. Proceedings of the National Academy of Sciences, USA 2001, 98:4498-4503. Acknowledgements 20. Clark R, Stampfer MR, Milley R, O'Rourke , Walen KH, Kriegler M, Work performed at LBNL was supported by the Department of Defense Kopplin JMC F.: Transformation of human mammary epithelial BCRP grant W81XWH-04-1-0580 to M.R.S and contract AC03-76SF00098 cells by oncogenic retroviruses. Cancer Research 1988, to the University of California from the US Department of Energy. 48:4689-4694. 21. Pierce JH, Arnstein P, DiMarco E, Artrip J, Kraus MH, Lonardo F, Di Fiore PP, Aaronson SA: Oncogenic potential of erbB-2 in human References mammary epithelial cells. Oncogene 1991, 6:1189-1194. 22. Frittitta L, Vigneri R, Stampfer MR, Goldfine ID: Insulin receptor 1. Meyerson M, Counter CM, Eaton EN, Ellisen LW, Steiner P, Caddle SD, overexpression in 184B5 human mammary epithelial cells Ziaugra L, Beijersbergen RL, Davidoff MJ, Liu Q, Bacchetti S, Haber DA, induces a ligand-dependent transformed phenotype. Journal of Weinberg RA: hEST2, the putative human telomerase catalytic Cellular Biochemistry 1995, 57:666-669. subunit gene, is up-regulated in tumor cells and during 23. Slonim DK: From patterns to pathways: gene expression data immortalization. Cell 1997, 90:785-795. analysis comes of age. Nature Genetics 2002, 32(s502-s508.):. 2. Stampfer M: Cholera toxin stimulation of human mammary epi- 24. Perou CM, Sorlie T, Eisen MB, van de Rijn M, Jeffrey SS, Rees CA, Pol- thelial cell in culture. In vitro 1982, 18:531-537. lack JR, Ross DT, Johnsen H, Akslen LA, Fluge O, Pergamenschikov A, 3. Hammond SL, Ham RG, Stampfer MR: Serum-free growth of Williams C, Zhu SX, Lonning PE, Borresen-Dale AL, Brown PO, Bot- human mammary epithelial cells: Rapid clonal growth in stein D: Molecular portraits of human breast tumours. Nature defined medium and extended serial passage with pituitary 2000, 406:747-752. extract. Proceedings of the National Academy of Sciences, USA 1984, 25. Sorlie T, Perou CM, Tibshirani R, Aas T, Geisler S, Johnsen H, Hastie T, 81:5435-5439. Eisen MB, van de Rijn M, Jeffrey SS, Thorsen T, Quist H, Matese JC, 4. Stampfer MR, Yaswen P: Human epithelial cell immortalization Brown PO, Botstein D, Eystein Lonning P, Borresen-Dale AL: Gene as a step in carcinogenesis. Cancer Letters 2003, 194:199-208. expression patterns of breast carcinomas distinguish tumor 5. Stampfer MR, Bodnar A, Garbe J, Wong M, Pan A, Villeponteau B, Yas- subclasses with clinical implications. Proceedings of the National wen P: Gradual phenotypic conversion associated with immor- Academy of Sciences, USA 2001, 98:10869-10874. talization of cultured human mammary epithelial cells. 26. Whitfield ML, George LK, Grant GD, Perou CM: Common markers Molecular and Cellular Biology 1997, 8:2391-2405. of proliferation. Nature Reviews Cancer 2006, 6:99-106. 6. Brenner AJ, Stampfer MR, Aldaz CM: Increased p16 expression 27. Byrne MC, Whitley MZ, Follettie MT: Preparation of mRNA for with first senescence arrest in human mammary epithelial expression monitoring. In Current Protocols in Molecular Biology John cells and extended growth capacity with p16 inactivation. Wiley and Sons; 2000:22.2.1-22.2.13. Oncogene 1998, 17:199-205. 28. Hill AA, Brown EL, Whitley MZ, Tucker-Kellogg G, Hunter CP, Slonim 7. Romanov S, Kozakiewicz K, Holst C, Stampfer MR, Haupt LM, Tlsty DK: Evaluation of normalization procedures for oligonucle- TD: Normal human mammary epithelial cells spontaneously otide array data based on spiked cRNA controls. Genome Biology escape senescence and aquire genomic changes. Nature 2001, 2001, 2:0055.1-55.13. 409:633-637. 29. Affymetrix I: http://www.affymetrix.com/support/technical/ 8. Nonet G, Stampfer MR, Chin K, Gray JW, Collins CC, Yaswen P: The manuals.affx. . ZNF217 gene amplified in breast cancers promotes immortal- 30. Eisen MB, Spellman PT, Brown POB D.: Cluster analysis and display ization of human mammary epithelia cells. Cancer Research of genome-wide expression patterns. Proceedings of the National 2001, 61:1250-1254. Academy of Sciences, USA 1998, 95:14863-14868. 9. Stampfer MR, Garbe J, Nijjar T, Wiginton D, Swisshelm K, Yaswen P: 31. Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP, Loss of p53 function accelerates acquisition of telomerase Coller H, Loh ML, Downing JR, Caligiuri MA, Bloomfield CD, Lander ES: activity in indefinite lifespan human mammary epithelial cell Molecular classification of cancer: class discovery and class lines. Oncogene 2003, 22:5238-5251. prediction by gene expression monitoring. Science 1999, 10. Strunnikova M, Schagdarsurengin U, Kehlen A, Garbe JC, Stampfer MR, 286:531-537. Dammann R: Chromatin inactivation precedes de novo DNA 32. Reich M, Ohm K, Angelo M, Tamayo P, Mesirov JP: GeneCluster 2.0: methylation during the progressive epigenetic silencing of the an advanced toolset for bioarray analysis. Bioinformatics 2004, RASSF1A promoter. Molecular and Cellular Biology 2005, 20:1797-1798. 25:3923-3933. 33. Frith M, Fu Y, Yu L, Chen JF, Hansen U, Weng Z: Detection of func- 11. Stampfer MR, Bartley JC: Induction of transformation and contin- tional DNA motifs via statistical over-representation. Nucleic uous cell lines from normal human mammary epithelial cells Acids Research 2004, 32:1372-1381. after exposure to benzo(a)pyrene. Proceedings of the National 34. Kel AE, Gossling E, Reuter I, Cheremushkin E, Kel-Margoulis OV, Win- Academy of Sciences, USA 1985, 82:2394-2398. gender E: MATCH: A tool for searching transcription factor 12. Garbe J, Wong M, Wigington D, Yaswen P, Stampfer MR: Viral onco- binding sites in DNA sequences. Nucleic Acids Research 2003, genes accelerate conversion to immortality of cultured 31:3576-3579. human mammary epithelial cells. Oncogene 1999, 18:2169-2180. 35. Der SD, Zhou A, Williams BR, Silverman RH: Identification of genes differentially regulated by interferon alpha, beta, or gamma Page 16 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 using oligonucleotide arrays. Proceedings of the National Academy of 58. Poola I, DeWitty RL, Marshalleck JJ, Bhatnagar R, Abraham J, Leffall LD: Sciences, USA 1998, 95:15623-15628. Identification of MMP-1 as a putative breast cancer predictive 36. Zhang H, Herbert BS, Pan KH, Shay JW, Cohen SN: Disparate effects marker by global gene expression analysis. Nature Medicine 2005, of telomere attrition on gene expression during replicative 11:481-483. senescence of human mammary epithelial cells cultured 59. Emberley ED, Niu Y, Njue C, Kliewer EV, Murphy LC, Watson PH: under different conditions. Oncogene 2004, 23:6193-6198. Psoriasin (S100A7) expression is associated with poor out- 37. Andrews HN, Mullan PB, McWilliams S, Sebelova S, Quinn JE, Gilmore come in estrogen receptor-negative invasive breast cancer. PM, McCabe N, Pace A, Koller B, Johnston PG, Haber DA, Harkin DP: Clinical Cancer Research 2003, 2627-2631:. BRCA1 regulates the interferon gamma-mediated apoptotic 60. Miller LD, Smeds J, George J, Vega VB, Vergara L, Ploner A, Pawitan Y, response. Journal of Biological Chemistry 2002, 277:26225-26232. Hall P, Klaar S, Liu ET, Bergh J: An expression signature for p53 38. Muller H, Bracken AP, Vernell R, Moroni MC, Christians F, Grassilli E, status in human breast cancer predicts mutation status, tran- Prosperini E, Vigo E, Oliner JD, Helin K: E2Fs regulate the expres- scriptional effects, and patient survival. Proceedings of the National sion of genes involved in differentiation, development, prolif- Academy of Sciences, USA 2005, 102:13550-13555. eration, and apoptosis. Genes and Developement 2001, 15:267-285. 61. Ziegler EC, Ghosh S: Regulating inducible transcription through 39. Ren B, Cam H, Takahashi Y, Volkert T, Terragni J, Young RA, Dynlacht controlled localization. SciSTKE 2005, 284:re6. BD: E2F integrates cell cycle progression with DNA repair, 62. Fabbro M, Schuechner S, Au WW, Henderson BR: BARD1 regulates replication, and G(2)/M checkpoints. Genes and Developement BRCA1 apoptotic function by a mechanism involving nuclear 2002, 16:245-256. retention. Experimental Cell Research 2004, 298:661-673. 40. Coller HA, Grandori C, Tamayo P, Colbert T, Lander ES, Eisenman RN, 63. Won J, Yim J, Kim TK: Sp1 and Sp3 recruit histone deacetylase Golub TR: Expression analysis with oligonucleotide microar- to repress transcription of human telomerase reverse tran- rays reveals that MYC regulates genes involved in growth, cell scriptase (hTERT) promoter in normal human somatic cells. cycle, signaling, and adhesion. Proceedings of the National Academy Journal of Biological Chemistry 2002, 277(38230-383238.):. of Sciences, USA 2000, 97:3260-3265. 64. Cong YS, Bacchetti S: Histone deacetylation is involved in the 41. Lawlor ER, Soucek L, Brown-Swigart L, Shchors K, Bialucha CU, Evan transcriptional repression of hTERT in normal human cells. GI: Reversible kinetic analysis of Myc targets in vivo provides Journal of Biological Chemistry 2000, 275:35665-35668. novel insights into Myc-mediated tumorigenesis. Cancer 65. Kyo S, Takakura M, Taira T, Kanaya T, Itoh H, Yutsudo M, Ariga H, Research 2006, 66:4591-4601. Inoue M: Sp1 cooperates with c-Myc to activate transcription of 42. Watson JD, Oster SK, Shago M, Khosravi F, Penn LZ: Identifying the human telomerase reverse transcriptase gene (hTERT). genes regulated in a Myc-dependent manner. Journal of Biological Nucleic Acids Research 2000, 28:669-677. Chemistry 2002, 277:36921-36930. 66. Li H, Lee TH, Avraham H: A novel tricomplex of BRCA1, Nmi, 43. Paine TM, Soule HD, Pauley RJ, Dawson PJ: Characterization of epi- and c-Myc inhibits c-Myc-induced human telomerase reverse thelial phenotypes in mortal and immortal human breast transcriptase gene (hTERT) promoter activity in breast can- cells. International Journal of Cancer 1992, 50:463-473. cer. Journal of Biological Chemistry 2002, 277:20965-20973. 44. Zhang J, R. S, Dai Q, Song J, Barlow SC, Yin L, Sloane BF, Miller FR, 67. Porter D, Lahti-Domenici J, Keshaviah A, Bae YK, Argani P, Marks J, Meschonat C, Li BD, Abreo F, Keppler D: Cystatin m: a novel can- Richardson A, Cooper A, Strausberg R, Riggins GJ, Schnitt S, Gabrielson didate tumor suppressor gene for breast cancer. Cancer E, Gelman R, Polyak K: Molecular markers in ductal carcinoma in Research 2004, 64:6957-6964. situ of the breast. Molecular Cancer Research 2003, 1:362-375. 45. Rhodes DR, Kalyana-Sundaram S, Mahavisno V, Barrette TR, Ghosh D, 68. Zhao H, Langerod A, Ji Y, Nowels KW, Nesland JM, Tibshirani R, Chinnaiyan AM: Mining for regulatory programs in the cancer Bukholm IK, Karesen R, Botstein D, Borresen-Dale AL, Jeffrey SS: Dif- transcriptome. Nature Genetics 2005, 37:579-583. ferent gene expression patterns in invasive lobular and ductal 46. Dai H, van't Veer L, Lamb J, He YD, Mao M, Fine BM, Bernards R, van carcinomas of the breast. Molecular Biology of the Cell 2004, de Vijver M, Deutsch P, Sachs A, Stoughton R, Friend S: A cell prolif- 15:2523-2536. eration signature is a marker of extremely poor outcome in a 69. Meeker AK, Hicks JL, Iacobuzio-Donahue CA, Montgomery EA, subpopulation of breast cancer patients. Cancer Research 2005, Westra WH, Chan TY, Ronnett BM, De Marzo AM: Telomere 65:4059-4066. length abnormalities occur early in the initiation of epithelial 47. Seth A, Kitching R, Landberg G, Xu J, Zubovits J, Burger AM: Gene carcinogenesis. Clinical Cancer Research 2004, 10:3317-3326. expression profiling of ductal carcinomas in situ and invasive 70. Burstein HJ, Polyak K, Wong JS, Lester SC, Kaelin CM: Ductal Carci- breast tumors. Anticancer Research 2003, 23:2043-2051. noma in Situ of the Breast. New England Journal of Medicine 2004, 48. Huang E, Cheng SH, Dressman H, Pittman J, Tsou MH, Horng CF, Bild 350:1430-1441. A, Iversen ES, Liao M, Chen CM, West M, Nevins JR, Huang AT: Gene 71. Baxter NN, Virnig BA, Durham SB, Tuttle TM: Trends in the Treat- expression predictors of breast cancer outcomes. Lancet 2003, ment of Ductal Carcinoma In Situ of the Breast. Journal of the 361:1590-1596. National Cancer Institute 2004, 96:443-448. 49. Pollak MN, Schernhammer ES, Hankinson SE: Insulin-like growth 72. Tsikitis VL, Chung MA: Biology of ductal carcinoma in situ classi- factors and neoplasia. Nature Reviews Cancer 2004, 4:505-518. fication based on biologic potential. American Journal of Clinical 50. Yu H, Rohan T: Role of the insulin-like growth factor family in Oncology 2006, 29:305-310. cancer development and progression. Journal of the National Can- 73. Boland GP, Butt IS, Prasad R, Knox WF, Bundred NJ: COX-2 expres- cer Institute 2000, 92:1472-1489. sion is associated with an aggressive phenotype in ductal car- 51. Deng CX, Brodie SG: Roles of BRCA1 and its interacting pro- cinoma in situ. British Journal of Cancer 2004, 90:423-429. teins. Bioessays 2000, 22:728-737. 74. Adler AS, Lin M, Horlings H, Nuyten DS, van de Vijver MJ, Chang HY: 52. Luker KE, Pica CM, Schreiber RD, Piwnica-Worms D: Overexpres- Genetic regulators of large-scale transcriptional signatures in sion of IRF9 confers resistance to antimicrotubule agents in cancer. Nature Genetics 2006, 38:421-430. breast cancer cells. Cancer Research 2001, 61:6540-6547. 75. Debnath J, Mills KR, Collins NL, Reginato MJ, Muthuswamy SK, Brugge 53. Gilmore PM, McCabe N, Quinn JE, Kennedy RD, Gorski JJ, Andrews JS: The role of apoptosis in creating and maintaining luminal HN, McWilliams S, Carty M, Mullan PB, Duprex WP, Liu ET, Johnston space within normal and oncogene-expressing mammary PG, Harkin DP: BRCA1 interacts with and is required for paclit- acini. Cell 2002, 111:29-40. axel-induced activation of mitogen-activated protein kinase 76. Thompson EW, Torri J, Sabol M, Sommers CL, Byers S, Valverius EM, kinase kinase 3. Cancer Research 2004, 64:4184-4154. Martin GR, Lippman ME, Stampfer MR, Dickson RB: Oncogene- 54. Lasorella A, Noseda M, Beyna M, Yokota Y, Iavarone A: Id2 is a retin- induced basement membrane invasiveness in human mam- oblastoma protein target and mediates signalling by Myc mary epithelial cells. Clinical and Experimental Metastasis 1994, oncoproteins. Nature 2000, 407:592-598. 12:181-194. 55. Beger C, Pierce LN, Kruger M, Marcusson EG, Robbins JM, Welcsh P, 77. Wang B, Soule HD, FR. M: Transforming and oncogenic potential Welch PJ, Welte K, King MC, Barber JR, Wong-Staal F: Identification of activated c-Ha-ras in three immortalized human breast epi- of Id4 as a regulator of BRCA1 expression by using a thelial cell lines. Anticancer Research 1997, 17:4387-4394. ribozyme-library-based inverse genomics approach. Proceed- ings of the National Academy of Sciences, USA 2001, 98:130-135. 56. Welcsh PL, Lee MK, Gonzalez-Hernandez RM, Black DJ, Mahadevappa M, Swisher EM, Warrington JA, King MC: BRCA1 transcriptionally regulates genes involved in breast tumorigenesis. Proceedings of the National Academy of Sciences, USA 2002, 28:7560-7565. 57. Donato R: S100: a multigenic family of calcium-modulated pro- teins of the EF-hand type with intracellular and extracellular functional roles. International Journal of Biochemistry and Cell Biology 2001, 33:637-668. Page 17 of 17 (page number not for citation purposes) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Molecular Cancer Springer Journals

Transcriptional changes associated with breast cancer occur as normal human mammary epithelial cells overcome senescence barriers and become immortalized

Download PDF
17 pages

Loading next page...
 
/lp/springer-journals/transcriptional-changes-associated-with-breast-cancer-occur-as-normal-W0xfsXRg1E

References (88)

Publisher
Springer Journals
Copyright
Copyright © 2007 by Li et al; licensee BioMed Central Ltd.
Subject
Biomedicine; Cancer Research; Oncology
eISSN
1476-4598
DOI
10.1186/1476-4598-6-7
pmid
17233903
Publisher site
See Article on Publisher Site

Abstract

Background: Human mammary epithelial cells (HMEC) overcome two well-characterized genetic and epigenetic barriers as they progress from primary cells to fully immortalized cell lines in vitro. Finite lifespan HMEC overcome an Rb-mediated stress-associated senescence barrier (stasis), and a stringent, telomere-length dependent, barrier (agonescence or crisis, depending on p53 status). HMEC that have overcome the second senescence barrier are immortalized. Methods: We have characterized pre-stasis, post-selection (post-stasis, with p16 silenced), and fully immortalized HMEC by transcription profiling and RT-PCR. Four pre-stasis and seven post-selection HMEC samples, along with 10 representatives of fully immortalized breast epithelial cell lines, were profiled using Affymetrix U133A/B chips and compared using both supervised and unsupervised clustering. Datasets were validated by RT-PCR for a select set of genes. Quantitative immunofluorescence was used to assess changes in transcriptional regulators associated with the gene expression changes. Results: The most dramatic and uniform changes we observed were in a set of about 30 genes that are characterized as a "cancer proliferation cluster," which includes genes expressed during mitosis (CDC2, CDC25, MCM2, PLK1) and following DNA damage. The increased expression of these genes was particularly concordant in the fully immortalized lines. Additional changes were observed in IFN-regulated genes in some post-selection and fully immortalized cultures. Nuclear localization was observed for several transcriptional regulators associated with expression of these genes in post-selection and immortalized HMEC, including Rb, Myc, BRCA1, HDAC3 and SP1. Conclusion: Gene expression profiles and cytological changes in related transcriptional regulators indicate that immortalized HMEC resemble non-invasive breast cancers, such as ductal and lobular carcinomas in situ, and are strikingly distinct from finite-lifespan HMEC, particularly with regard to genes involved in proliferation, cell cycle regulation, chromosome structure and the DNA damage response. The comparison of HMEC profiles with lines harboring oncogenic neu changes (e.g. overexpression of Her-2 , loss of p53 expression) identifies genes involved in tissue remodeling as well as proinflamatory cytokines and S100 proteins. Studies on carcinogenesis using immortalized cell lines as starting points or "normal" controls need to account for the significant pre-existing genetic and epigenetic changes inherent in such lines before results can be broadly interpreted. Page 1 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 immortalized using several different pathologically rele- Background Genetic and epigenetic changes that occur early in the vant agents, e.g., chemical carcinogens, over-expression of process of carcinogenesis may enable the survival and the breast cancer-associated oncogenes c-myc and/or growth of cells that subsequently acquire oncogenic muta- ZNF217, and/or inactivation of p53 function [8,9,11]. tions. One early alteration in the development of human Fully immortal HMEC maintain telomeres at short, stable carcinomas is the acquisition of an immortal potential, lengths, but do not necessarily express malignancy-associ- associated with reactivation of endogenous hTERT expres- ated properties; overexpression of specific oncogenes can sion and maintenance of stable telomere lengths. [1]. We confer malignant properties [20-22]. have employed an in vitro HMEC model system to exam- ine gene expression changes during the process of trans- Transcriptional profiling has proven to be a valuable tech- formation of normal finite cells to immortality and nology for describing the differences between cell types malignancy [2-11]. Two mechanistically distinct barriers and experimental treatments for many disease models, to unlimited proliferation have been described. The first particularly cancer [23]. One of the most well-developed barrier, stasis (stress-associated senescence) is associated stratifications of human cancers has been for breast cancer with elevated levels of the cyclin-dependent kinase inhib- [24,25]. These and other studies have shown that a com- INK4A itor (CKI) p16 [6]. Stasis appears to be Rb-mediated mon set of genes is consistently overexpressed in most and not directly dependent on telomere length. Cells cancers [26], including many cell cycle regulated genes arrested at this barrier exhibit a viable G1 arrest with a low and genes required for mitosis (e.g. MKI67, PCNA, BIRC5, labeling index (LI), normal karyotypes, expression of MYBL2, TOP2A, PLK1, MCM2-MCM6, CDC20). The fre- senescence -associated ß-galactosidase (SA-ß-gal) activity, quent identification of these genes in cancer cells suggests and a senescent morphology [7,12]. HMEC can undergo a that they represent a common characteristic of cancers, variable number of population doublings (PD), depend- irrespective of the cell type from which the cancers origi- ing upon culture conditions, prior to encountering stasis. nate. Multiple types of single changes that prevent Rb-mediated The data described here examines the changes that occur growth inhibition will overcome stasis. Loss of CDKN2A as HMEC overcome the barriers to indefinite prolifera- ink4a ) expression, from methylation- (encoding p16 tion. We show that pre-stasis and post-selection HMEC induced CDKN2A promoter silencing, or mutations, is are profoundly different from fully immortalized HMEC one alteration frequently observed in human breast can- lines, despite the fact that the immortalized lines may cers and cultured HMEC [6,13,14]. HMEC cultured in a retain normal growth factor requirements, lack anchor- serum-free medium can produce rare cells that spontane- age-independent growth or invasiveness, and are not tum- ously silence the p16 promoter and resume growth, a origenic in animal models [4]. Rather, the non-malignant process termed selection, with the resulting post-stasis immortalized lines display the cancer-associated prolifer- population called post-selection [3]. In the HMEC, no ation cluster of genes frequently identified in transcrip- ARF increase in p53, p21, or p14 levels have been seen at tional profiling studies of cancer cells and tissues [26]. stasis [7] and p53 function is not required for the stasis barrier (J.G. and M.S., unpublished). Rare HMEC with Materials and methods silenced p16 are also observed in vivo and have been Reagents and supplies MEBM serum-free medium was purchased from the called variant HMEC (vHMEC) [15,16]. Clonetics division of Cambrex BioScience (Walkersville, HMEC that have overcome or bypassed stasis encounter a MD), and was supplemented with EGF, hydrocortisone, second barrier as a consequence of telomere dysfunction. insulin, and BPE using Singlequot reagent packs from Ongoing proliferation in the absence of telomerase Clonetics, as well as 5 µg/ml transferrin (Clonetics) and expression leads to critically shortened telomeres, and 10 nM isopeterenol (Sigma). Hams F-12/DMEM (50:50) chromosomal aberrations [7,17]. In post-selection HMEC was purchased from Invitrogen or prepared by Core Tech- with functional p53, these aberrations induce a mostly nical Services (Wyeth Research), and supplemented to viable G1 and G2 arrest (termed agonescence); if p53 is contain 5% FBS (Invitrogen), 2 mM pyruvate (Invitro- non-functional, massive cell death (crisis) ensues (J.G. gen), 2 mM glutamine (Invitrogen), 20 ng/ml EGF and M.S., unpublished) [18]. Telomere dysfunction poses (Clonetics), 200 µg/ml cholera toxin (Sigma), 1× ITS an extremely stringent barrier to human cellular immor- (Clonetics), 500 ng/ml hydrocortisone (Sigma or Clonet- talization; in post-selection HMEC multiple errors appear ics), and 20 mg/ml gentamycin (Invitrogen). MM to be necessary for telomerase reactivation, and immortal- medium was prepared as described [2]. Antibodies and ization [4,8]. Since this barrier is dependent upon tel- fluorescent dyes used in High Content Screening (HCS, or omere length, ectopic overexpression of hTERT readily quantitiative immunofluorescence) were obtained from immortalizes post-selection HMEC [19]. HMEC can be Cell Signaling Technologies (Beverly, MA), Upstate Bio- Page 2 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 technologies (Lake Placid, NY), and Molecular Probes/ value was then computed for each gene by multiplying its Invitrogen (Carlsbad, CA), as described in the supplemen- original Signal intensity with the scale factor (100/ tary material. Antibodies were screened by Western blot trimmed-mean). Subsequently, genes were filtered to prior to immunofluorescence studies to verify that they remove those with uninformative or noisy expression recognize a single specific antigen of the expected molec- changes across the entire samples. A gene is selected for ular size. downstream analysis if its expression exceeds 50 (scaled) Signal unit in at least one sample. Analysis of variance Cell culture (ANOVA) was performed with log2 transformation on the Pre-stasis and post-selection HMEC, from specimens 48, scaled Signals of several cell lineage groups (see details 161, 184, 191, 195 and 239, as well as the immortally below). Data was analyzed using several analytical transformed lines 184A1, 184AA2, 184AA3, 184B5 were approaches, including unsupervised clustering [30], developed and characterized at LBNL, starting with reduc- supervised clustering [31,32], and principal components tion mammoplasty tissues; an additional post-selection analysis. For the unsupervised clustering, genes that are HMEC strain was obtained from Clonetics. Remaining filtered based on the Pvalues from one-way analysis of lines, as well as additional samples of 184A1 and 184B5 variance (ANOVA) on four cell lineage groups as well as were obtained from ATCC (Manassas, VA). 184B5ME was greater than 2 fold difference among the four groups. +/ derived from immortal 184B5 following stable expression These groups consist of 1) all finite lifespan cells, 2) p53 + -/- of ERBB2/Her2 and selection for anchorage independent immortalized 184A1 and 184B5, 3) p53 immortalized growth (Stampfer, unpublished). Pre-stasis cells were 184AA2 and 184AA3, and 4) immortalized non-184 maintained in MM media [2], and post-selection cells derived cells (including MCF10A, MCF10A-2, and were maintained in MEBM prior to this study. Pre-stasis MCF12A). HMEC display 15–25 PD in MM, and 10–15 PD in Promoter analysis MEBM, prior to growth arrest at stasis. For transcriptional profiling studies, all lines maintained at LBNL (listed Genes identified as unique classes in a subset of post- above), as well as the post-selection HMEC purchased selection HMEC were examined in detail (see Results for a from Clonetics, were revived in MEBM media and cul- complete list of genes). Initially, the 500 bp upstream of . Consequently, the pre-stasis tured at 37°C with 1% CO the transcription start site for each gene was examined for HMEC were studied as they neared stasis. Pre-stasis HMEC well-characterized transcription binding sites using two used in HCS were cultured in MM medium. Fully immor- algorithms, Match and Clover [33,34]. For most of the talized cell lines obtained from ATCC (184A1, 184B5, groups, strong assignments of specific promoter binding MCF10A, MCF10A-2 and MCF12A) were cultured in sites could be identified using both algorithms. One class DMEM/Ham's F-12 medium, at 37°C with 10% CO , as (Class B in the Results) was less definitive, so the region they were maintained prior to crypreservation. was extended to 2 kb prior to the transcription start site for those genes. RNA labeling, GeneChip hybridizations and expression analysis Taqman™ quantitative PCR Cells to be prepared for RNA extraction were revived from Primer sets for 15 genes analyzed by Taqman™ analysis cryopreservation and cultured to 80% confluence in a sin- were obtained from Applied Biosystems (Foster City, CA) gle T-75 flask, trypsinized under conditions appropriate and used according to standard protocols. Genes tested for each line, and split 1:4 into four new T-75 flasks. When are listed in the Results section. cells reached 80% confluence three of the flasks were trypsinized, lysed and total RNA isolated using the High content screening Midiprep RNA isolation kit from Qiagen, according to Cells were seeded at 5000 cells/well in a 96-well black manufacturers instructions. wall, clear bottom Packard ViewPlate, and incubated in MM, MEBM or DMEM/F-12 medium for pre-stasis, post- An 11-point standard curve of bacterial cRNA control selection and immortalized HMEC, respectively, for 48 samples was added prior to hybridization as described hours. Cells were washed with PBS, and fixed with pre- [27,28]. Three independent replicates were generated per warmed 4% paraformaldehyde for 10 minutes. Cells were cell type at the indicated stage. Affymetrix's MAS5 algo- washed 2× with PBS, permeabilized with 0.2% Triton X- rithm was used to generate expression measures including 100 for 3–5 minutes, and washed 2× with PBS again. Cells Signal values and Absent/Present calls (Affymetrix (2001) were stained with primary antibodies in 1% BSA/PBS. Pri- Microarray Suite User Guide, Version 5. [29]. A global scal- mary antibodies were used as follows: E2F1 (BD/Phar- ing normalization was applied to the raw signal intensity. magin, 1:200 dilution), E2F4 (Abcam, 1:400), Rb (Cell Briefly, a 2% trimmed-mean was calculated per chip, and Signaling Technologies, 1:400), p107 (Santa Cruz, was scaled to an arbitrary value of 100. A scaled Signal 1:200), BRCA1 (Abcam, 1:200), p53 (Cell Signaling Tech- Page 3 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 nologies, 1:200), SP1 (Upstate Biotechnologies, 1:400), types, we compared several samples of these cultures by NF-κB (Cellomics, 1:200). Cells were washed 3× with transcriptional profiling; the HMEC samples character- PBST (0.05% Tween-20), and stained with DAPI and sec- ized are described in Table 1. The finite lifespan pre-stasis ondary antibodies of appropriate species/isotype specifi- and post-selection HMEC are referred to as strains or cell city and conjugated to either Alexa-488 or Alexa-594. types from a specific source, and culture conditions Cells were washed again 3× with PBST; 100 µl of PBS was (including stage) are noted for each particular sample. The added and plates were sealed with an adhesive cover. relationships between samples in this study, their origins, are indicated graphically in Figure 1. Triplicate cultures for Quantitative immunofluorescence was performed using a each sample were grown under the conditions indicated ti Cellomics ArrayScan V . Images were taken using a 20× in the Methods, and in Table 1, following which the total objective and data was collected for a minimum of 1000 RNA was isolated, labeled and hybridized to the Affyme- valid cells per well. Valid cells are defined as having nuclei trix U133A/B GeneChips. with expected DNA content (defined by DAPI fluores- cence intensity), nuclei size and shape typical for the cell Principal Component Analysis (PCA) was used to visual- line/type, and well-separated from neighboring cells, such ize the gross relationships among the cell types, as shown that cytoplasmic regions could be clearly resolved. DNA in Figure 2A. The first three components, which explains content and antigen intensity were quantitated for each about 60% of the total variation, are displayed in a three cell, and the nuclear-cytoplasmic ratio for each antigen dimensional graph. The pre-stasis HMEC (in red) and was determined by a mask derived from the DAPI stain- post-selection HMEC (in pink) are clearly separated from ing, which was used to define the nucleus, and a region the immortalized lines (in blue, black and green) along surrounding the nucleus (which was specific for each cell the first principal component axis. Thus, transcriptional line/type) was used to define the cytoplasm. Quantitation profiling defines the transition from finite lifespan to fully was performed using either the Compartmental Analysis immortalized HMEC as the most significant change in or Nuclear Translocation BioApplications, from Cel- HMEC progression. The pre-stasis and post-selection lomics. HMEC are also well segregated within their unique space. In addition, the fully immortalized lines that either do not Results express p53 or are transduced with ERBB2/Her2 (green Transcriptional profiling of pre-stasis, post-selection and and blue, respectively) are distinguished from the rest of immortalized HMEC the immortalized lines (black). According to the PCA, To better understand the extent to which pre-stasis, post- there are no significant differences between the fully selection and immortalized HMEC represent distinct cell immortalized lines derived from various methods of Table 1: Cell Types and Lines Used in This Study Cell Name Source Stage Growth Media 48L LBNL Pre-stasis, finite lifespan strain MM (MEBM)*** 161 LBNL Pre-stasis, finite lifespan strain MM (MEBM) 184 LBNL Pre-stasis, finite lifespan strain MM (MEBM) 195L LBNL Pre-stasis, finite lifespan strain MM (MEBM) 48R LBNL Post-selection, finite lifespan strain MEBM 161 LBNL Post-selection, finite lifespan strain MEBM 184 LBNL Post-selection, finite lifespan strain MEBM 195L LBNL Post-selection, finite lifespan strain MEBM 191 LBNL Post-selection, finite lifespan strain MEBM 239 LBNL Post-selection, finite lifespan strain MEBM HMEC-1001-13 Clonetics Post-selection, finite lifespan strain** MEBM 184A1 LBNL Fully immortal cell line MEBM 184B5 LBNL Fully immortal cell line MEBM 184AA2 LBNL Fully immortal cell line MEBM 184AA3 LBNL Fully immortal cell line MEBM 184B5ME LBNL Fully immortal cell line MEBM 184A1* ATCC Fully immortal cell line DMEM/F-12 184B5* ATCC Fully immortal cell line DMEM/F-12 MCF-10A ATCC Fully immortal cell line DMEM/F-12 MCF-10A-2 ATCC Fully immortal cell line DMEM/F-12 MCF-12A ATCC Fully immortal cell line DMEM/F-12 *designated as 184A1(a) and 184B5(a) in other tables and figures **stage defined by transcriptional profile ***cells isolated from reduction mammoplasty tissues and expanded in MM media to passages 2–3, then cultured in serum-free MEBM media for transcriptional profiling Page 4 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 pd50-100 pd50-100 pd15-25 pd15-25 pd1 pd1 Post-stasis: Population doubling: Pre-Stasis Post-Selection or Conversion Fully Immortalized HMEC Extended Lifespan HMEC stage: Proliferation Block: Stasis Agonesence/Crisis (Epi)Genetic Changes: p16 loss Telomerase reactivation Cell lines used 184A1 p46 184A1 in this study: 48R p9 48L p2 161 p2 161 p7 184B5 p43 184B5 195L p2 195L p6 184 p3 184 p7 MCF10A 184AA2 p44 MCF10A-2 HMEC-1001-13 184AA3 p43 239 p7 MCF12A 191 p7 184B5ME p54 184Aa 184 p1 184Be Comparisons Analyzed: Selection clonal deriviative Immortalization BaP treatment p53 Status retroviral transduction neu transfection and selection for anchorage Her2 Expression independence not analyzed Graphic relationship of cell li Figure 1 nes profiled in this study Graphic relationship of cell lines profiled in this study. Cell lines characterized in this study are shown with reference to their stage in transformation. The pre-stasis HMEC used were cultured for 2–3 passages before analysis, and reach stasis by passages 3–5. Rare isolates of cells grown in serum-free media (MEBM) emerge spontaneously from stasis, associated with the absence of p16 expression due to promoter silencing, and continue growing as post-selection HMEC until reaching a second, proliferation barrier (telomere dysfunction). This barrier is highly stringent, and spontaneous immortalization has never been observed in cells that were not mutagenized or virally transduced during pre-stasis or post-selection growth. HMEC grown in MM do not spontaneously give rise to post-selection cells, however primary populations exposed to the chemical carcinogen benzo(a)pyrene (BaP) have produced rare clonal isolates with post-stasis growth, associated with absence of p16 expression due to mutation or promoter silencing. These non-spontaneously arising post-stasis cells are referred to as extended lifespan, and may harbor additional errors due to the carcinogen exposure. Overcoming the telomere dysfunction barrier is associated with reactivation of telomerase activity. The fully immortalized lines 184A1 and 184B5 were derived from extended lifespan post-stasis cells grown in MM and exposed to BaP in primary culture. Exposure of extended lifespan 184Aa cells to retroviral infection resulted in two cell lines that had lost both copies of the TP53 gene. The cell lines profiled in this study are shown rel- ative to the profiling analyses performed. Comparisons used to analyze selection and immortalization, as well as the influence of p53 and ERBB2/Her2 status are shown by colored boxes and identified in the key at the lower left of the figure. Page 5 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 immortalization, or from lines maintained at LBNL versus Cancer-Testis Antigen 2 (CTAG-2) is very strongly expressed those obtained from ATCC. Unsupervised (or Eisen) clus- (30-fold according to the GeneChip data), as are ARH- tering of the genes that change following selection and GDIB/Ly-GDI, and IGFBP6. The cytokine induced genes immortalization for most of the samples is shown in Fig- [35] include a set previously reported as increasing in ure 2B. These data reflect the 1 342 genes that are filtered post-selection HMEC, such as IFIT1, IFITM1, G1P2 and based on the Pvalues from one-way analysis of variance OAS1 [36]. The genes that are unique to 48 HMEC (Group (ANOVA), as described in the supplementary material. B) include several transcription factors and cell cycle pro- teins whose roles in cancer or breast tissue development Gene expression changes following selection have not been well characterized to date, including Gene expression changes that distinguish pre-stasis from NUCKS, SON and HOXB2. Group C includes many genes post-selection cells were identified using GeneCluster previously associated with cancer cell proliferation. [31], and the results are shown in Figure 3A. The figure characterizes a large set of concordantly-regulated genes in Since these geneset classes were comprised of a relatively the pre-stasis strains, and a high level of concordance in small number of genes, we performed promoter analyses, four of the six post-selection HMEC (200 genes for each to see if these sets are linked in specific pathways. Pro- class). Among these top-200 genes in the pre-stasis cell moter binding sites we were able to identify are listed in types, the largest number of genes we identified are Table 2. For Group A, interferon-responsive elements involved in the extra-cellular matrix (ECM), including were found for most of the genes, but not the cancer/ structural proteins and matrix remodeling enzymes (listed metastasis-associated genes (BST2 is an exception), con- in supplementary Additional file 1). Examples include sistent with previous studies that did not identify these many collagen and kallikrein genes. Genes that increase genes as IFN-regulated [35]. Instead, several genes in this expression level in post-selection HMEC include a large group have been shown to be direct or indirect targets of number of genes associated with proliferation and the cell p53 and Myc. A common element in the regulation of cycle. These genes are strongly associated with cancer cell both p53/Myc and IFN-regulated genes is BRCA1, and in growth, and increase in expression directly with tumor particular, BRCA1 is essential for the activation of stress grade. Specific examples include BIRC5, A and B type cyc- and inflammatory response genes following treatment lin genes, CDC2, and the MCM chromosomal proteins. with interferons [37]. Group B was less well-defined by The increased expression of these genes is dependent on specific binding sites near the promoter, but an extended E2F transcription factors and reflects the proliferative state analysis (2 kb) identified SP1, E2F, MAZ and NF-Y bind- a cell. Since the pre-stasis cells were nearing stasis, the ing sites for many genes. These binding sites were also increased expression of the genes in the post-selection identified in the genes of Group C, especially the E2F, NF- HMEC may reflect either a loss of Rb repression (consist- Y and SP1 sites, which is consistent previous work [38,39]. ent with a loss of p16), or could reflect the relative prolif- Group D, genes significantly repressed in post-selection erative state of these pre-stasis and post-selection cells. HMEC, may be under the control of MAZ (Myc-associated zinc finger protein), as binding sites were found in 19 of The two discordant post-selection HMEC we observed in 22 genes examined, which is consistent with previous Figure 3A (195L and 1001-13), suggest that additional observations that increased Myc can repress ECM genes molecular events can occur during selection; these sam- [40-42]. In conclusion, although distinct gene expression ples also show a loss of p16 expression (results not patterns could be observed for each of the pre-stasis/post- shown), a definitive event for post-selection HMEC. In selection HMEC pairs we have characterized, in each case order to probe further into the changes that occur during strong associations could be made between the promoters selection, we compared the four sets of HMEC studied as of each class and the proliferation and cell cycle transcrip- pre-stasis and post-selection samples. For this analysis, we tion factors, particularly E2F, SP-1, NF-Y and the Myc- identified genes that increase expression in post-selection related MAZ. The distinguishing features for each of these HMEC, as compared to the corresponding pre-selection expression classes is likely to be found in additional, sample. Four patterns were observed. The genes we iden- unique pathways such as BRCA1-mediated regulation. tified in each group are listed in Table 2, and the expres- sion changes we observe for three of the groups are shown Gene expression changes that distinguish finite life span in Figure 3B. The group not explicitly shown in Figure 3B HMEC from immortally transformed HMEC is uniformly down-regulated in all four pairs. Genes The most significant transition observed in this study is expressed exclusively in post-selection 195L HMEC that of immortalization. Genes whose expression are (Group A) fall into two categories: genes previously iden- reduced in the immortalized lines include a significant tified as cancer-associated (including several antigens pro- number that suppress angiogenesis, contribute to the posed as cancer biomarkers), and genes induced by ECM, or regulate the actin cytoskeleton. Many of these interferons [35]. Among the cancer-associated genes, the genes were identified as down-regulated in HMEC follow- Page 6 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 Pr Pre e- -s st ta as si is s Po Pos st t- -s se el le ec ct ti io on n I Im mm mo o r rt ta al li iz ze ed, p53- d, p53- I Im mm mo o rt rta al li iz ze ed d, , H He er2 r2+ ++ + Im Imm mo or rtta alliiz ze ed d pre-stasis post-selection immortalized Re Figure 2 lationship of HMEC as determined by transcriptional profiles Relationship of HMEC as determined by transcriptional profiles. A. Data from 2319 genes were used to determine the number of principal components of the data. Three components were identified, and the contribution of the components to the transcription profile of each cell line samples are shown in the figure. Individual replicates for each cell line are shown. Cell lines grouped in Figure 1 are shown in Figure 2A as shown in the legend. Vertical axis is PC1, the first, and therefore the strongest. principal component. B. Unsupervised clustering of HMEC. All genes that change expression in one or more sam- ples were used to cluster the cell types and lines by overall similarity. Cell types and lines are identified by color under the des- -/- ignations: pre-stasis HMEC: light green; post-selection HMEC: light blue; fully immortalized HMEC: dark blue; p53 fully immortal HMEC: burgundy, and lines not formally characterized: black. Samples of 184A1 and 184B5 designated by (a) were obtained from ATCC. Page 7 of 17 (page number not for citation purposes) 195L 48L 48R 1001-13 195L 184A1 184B5 184A1(a) 184B5(a) MCF-12A MCF-10A MCF-10A-2 184AA2 184AA3 Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 A B CTAG2 CTAG1, CTAG2 CTAG1, CTAG2 IFIT1 BST2 C1orf29 ARHGDIB G1P2 G1P3 MMP7 NUCKS TRAP150 HOXB2 CA12 LZ16 IF2 SON BRD4 PIST BX TOP2A RRM2 RRM2 KIF20A TOP2A BIRC5 ANKT CCNA2 CDC2 MKI67 161 184 195 48 pre-stasis post-selection Supervised Cluster Figure 3 ing of Pre-stasis, and Post-selection HMEC Supervised Clustering of Pre-stasis, and Post-selection HMEC. A. Gene expression values were normalized and character- ized for the significance of overexpression in one group relative to other groups in the comparison. The top 50 genes that are significantly overexpressed in one group are shown. All pre-stasis and post-selection cell types have been used. Analysis was performed in GeneCluster, and the color bar describing how normalized values are depicted is shown at the bottom of the fig- ure. B. Distinct classes of genes over-expressed in post-selection HMEC. Genes showing one of three specific patterns of expression in the four pairs of pre-stasis and post-selection samples are diagramed. The top 10 qualifiers (based on fold change) are shown (some genes are represented by more than one qualifier). ing selection as well; some are further down-regulated in commonly observed "proliferation cluster" described the immortalized lines, as shown in Figure 4A. These com- above. These genes were also observed to be up-regulated parisons include multiple independent samples from in the post-selection, compared to pre-stasis HMEC. Fewer each stage, including four distinct fully immortalized cell of these "proliferation genes" are identified in the fully lines, and three additional samples from either different immortalized samples following a three-way comparison, sources (184A1 and 184B5 from ATCC) or two separate but this is because GeneCluster identifies the most defin- isolates from the same experiment (MCF-10A and MCF- itive group of genes for each class, and since some of the 10A-2) [43]. The genes identified in each group are post-selection samples express increased levels of genes described in Additional file 2. Collectively, the pre-stasis such as MCM2 and STK12, they are not unique to either and post-selection samples are distinguished most the post-selection or the fully immortalized HMEC. strongly by changes to the ECM and cell-cell communica- tion genes, particularly collagens, kallikrein, matrix metal- We have examined the expression of the cancer cell prolif- loproteinase and serpin proteinases; genes that affect the eration class of genes directly in Figure 4B. In this exam- actin cytoskeleton are also noted (both actin and actin- ple, the absolute expression levels of each gene listed in interactors, such as actinin, nidogen, transgelin, and pal- the figure are displayed directly (rather than the ratio of ladin, genes). Several well-recognized classes of genes are post-selection over pre-stasis expression levels in Figure up-regulated in fully immortalized lines, including the 3B). These genes are compared to equal subsets of genes Page 8 of 17 (page number not for citation purposes) Group C Group B Group A 48L 195L 48R 195L 1001-13 Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 Table 2: Genes and Promoter Elements That Define Post-Selection HMEC Gene Expression Classes Geneset Classes Genes Promoter Elements Group A IFN genes IFIT1, BST2, G1P2, G1P3, IFIT2, OAS1, IFI44, IFIT4 IRF, ILR, IRL Non-IFN genes CTAG2, ARHGDI-B/Ly-GDI, MMP7, PLAU, CALB1, SLC1A6, MDA5, FXYD5, HMOX1 Group B NUCKS, HDAC3, TRAP150, HOXB2, SON, IF2, LZ16, ANLN, BBX, TOP1, H4FG, SFRP1, KTN1, SP1 GTAR, BAZ1A, PK428, FALZ, TTC3, DNCL12, RBM9 Group C TOP2A, RRM2, KIF20A, BIRC5, ANKT, CCNA2, CDC2, MKI67, CDC20, MCM5, HMMR, IL-1B, E2F, NF-Y, B-Myb PRC1, PMSCL1, MADL1, DLG7 Group D H11, COL11A1, IGFBP5, CNN1, COMP, LGALS7, CLDN7, KLK6, KLK7, KLK10, KLK11 KRT23, MAZ, MAZR, MEF-3 LOXL4, THY1, FLJ21841 pre-stasis post-selection immortalized COL1A2 COL2A1 COL4A1 COL6A1 COL6A2 COL11A1 ID4 IGFBP3 IGFBP5 ITGB3 KLK6 KLK7 KLK10 KLK11 KRT8 KRT16 KRT19 KRT23 PMP22 SERPIN G1 APP ACTN4 C COL4A5 OL4A5 COL4A6 COL8A1 COL12A1 COL16A1 COL17A1 HLA-A HLA-B HLA-C HLA-F HLA-G IGTA3 ITGAV ITGB6 MMP2 SERPIN E1 SERPIN E2 SERPIN F1 BIRC5 BUB1 BUB1B CCNA2 CCNB1 CCNB2 CDC2 CDC20 CDC25B CDC6 CENPA CENPF MCM2 MCM4 MPHOSH1 NEK2 RRM2 STK6 STK12 TOP2A pre-stasis post-selection immortalized Supervised Cluster Figure 4 ing of Pre-stasis, Post-selection and Immortalized HMEC Supervised Clustering of Pre-stasis, Post-selection and Immortalized HMEC. A. Gene expression values were normalized and characterized for the significance of over-expression in one group relative to other groups in the comparison. The top 200 genes (of 1342) that are significantly over-expressed in one group are shown. All pre-stasis, post-selection and immortalized -/- HMEC (except the p53 and ERBB2/Her2 transfected variants) have been grouped. The top 100 genes (of 1440) that are over- expressed in one group relative to the other two are presented. Analysis was performed in GeneCluster. B. Expression of a subset of highly concordant genes in pre-stasis, post-selection and fully immortalized HMEC. Gene-normalized expression of 60 genes identified in the figure are shown for four representatives each for the three groups of HMEC. Samples are (left to right): 48L, 161, 195L and 184; 48R, 161, 195L, and 184; 184A1, 184B5, MCF-10A and MCF12A. Page 9 of 17 (page number not for citation purposes) 48L 195L 48R 195L 1001-13 184A1 184B5 184A1(a) 184B5(a) MCF-10A MCF-10A-2 MCF-12A cancer proliferation cluster genes Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 that show maximal levels of expression in the pre-stasis genicity [21]. Gene expression changes seen for 184B5ME and post-selection HMEC samples. As can be observed in that are distinct from its parent are listed in the supple- the figure, genes showing maximal expression in the pre- mentary Additional file 6. Genes showing increased stasis samples are robust, whereas those showing maximal expression include many that were down-regulated in expression in the post-selection are less strongly definitive post-selection HMEC, including kallikreins KLK6 and of post-selection cells. The "proliferation cluster" genes KLK7, and cystatin E/M. These phenotypic reversions may show strongest expression in the fully immortalized play a role in the transition to invasive cancer [44]. Addi- HMEC lines, however expression of these genes is hetero- tional gene expression changes include a dramatic geneous for both the post-selection and fully immortal- increase in the expression of IL24 and significant changes ized sets. Increased expression can be observed for the in BIRC3, HRASLS3, and PTGES. Genes showing down- post-selection 48R and 184 samples (as was seen for some regulation as a consequence of ERBB2/Her2 overexpres- of these genes in Figure 3B), and lesser expression is seen sion include many of the IFN genes that showed increased for MCF-12A. However, the rise in expression of this expression following selection (in 195L) or immortaliza- group of genes as HMEC progress from pre-stasis through tion (in 184A1, 184B5 and others). fully immortalized stages is clear. Real-time PCR measurement for selected genes identified -/- Gene expression changes observed in p53 cell lines in this study HMEC lines that have lost p53 during immortalization The results presented comprise a large study of human show distinctive changes in transcriptional profiles when mammary cell samples that have not been characterized compared to closely related lines that have retained p53 by transcriptional profiling previously, and the gene function. The complete list of genes is presented in the expression patterns are either new or not previously asso- supplementary Additional file 3. When we explicitly look ciated with non-cancerous cell lines. As such we wished to for genes whose expression changes are common to the validate the findings by corroborating the gene expression p53 status of the lines derived from specimen 184 cells, changes observed by genechips with an independent +/ several genes showing concordant changes between p53 method. 15 genes were chosen from the data to be vali- + -/- 184A1 and 184B5 versus p53 184AA2 and 184AA3 are dated by Taqman™ quantitative PCR. Genes that change observed. SIAH2, Lipocalin 2, Asparagine synthase and Ker- following selection (PMP22/GAS3 and several insulin-like atin 15 are all upregulated in both 184AA2 and 184AA3, growth factor binding protein (IGFBP) genes: IGFBP2, relative to both 184A1 and184B5. Genes down-regulated IGFBP3, IGFBP4, IGFBP5, IGFBP6, and IGFBP7), as well as -/- in the p53 lines include several that are explicitly regu- genes that change in immortalized lines (CCNB1, CDC2, lated by p53 (including RRM2 and TP53INP1). A compar- CDC25B, HDAC3, MYC, and STK6) were evaluated by RT- +/+ -/- ison of the two p53 and the two p53 lines shows that PCR in 17 cell types, comprising pre-stasis, post-selection additional gene expression changes unique to each line and fully immortalized samples, and the results compared have occurred. Examples include DUSP1 and BIRC3, to expression data from the oligonucleotide arrays. The expressed at significantly higher levels 184AA3 than in concordance between expression of a gene as measured by 184AA2, and FABP4, IFI27, HRASLS3, and Fibulin 1, oligonucleotide array and Taqman™ assays were generally expressed much more robustly in 184A1 than in 184B5. quite good; in 14 cases, only minor discordances can be The complete list of genes is presented in the supplemen- observed (see Figure 5). HDAC3 was an exception. The tary Additional file 4 and Additional file 5. expression level changes of three probes sets for HDAC3 on the Affymetrix U133 GeneArrays, and the Taqman™ Gene expression changes resulting from ectopic expression primer set, were highly discordant, so we were not able to of Her2 validate the expression changes of this gene by RT-PCR, The events characterized thus far in this study concern however were able to show significant changes in HDAC3 HMEC immortalization; however, additional events are protein expression and localization by immunofluores- critical to malignancy. To connect these studies directly to cence microscopy (described below). changes that occur following an oncogenic event, we have Transcriptional regulatory factors are localized to the compared one immortalized HMEC line, 184B5, with a derivative that ectopically expresses the ERBB2/Her2 nucleus following selection and immortalization oncogene, 184B5ME. ERBB2/Her2 is frequently over- We explored the changes that occur in several critical reg- expressed in breast cancer, and is transforming simply by ulators of cell cycle progression and chromosomal stabil- being over-expressed, so this line models clinically rele- ity by quantitative fluorescence microscopy, or High vant features of breast cancer. Over-expression of ERBB2/ Content Screening (HCS). These factors were chosen Her2 in 184B5 results in anchorage independent growth, based on patterns observed in the transcription profiling a malignancy-associated property, while over-expression data as ones that would be expected to change as HMEC of oncogenic ERBB2/Her2 in 184B5 can confer tumori- progress past senescence barriers, based on the gene Page 10 of 17 (page number not for citation purposes) MCF-12A MCF-10A-2 MCF-10A 184AA3 184AA2 184B5ME 184B5 184A1 MCF-12A MCF-10A-2 MCF-10A 184AA3 184AA2 184B5ME 184B5 184A1 Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 48 161 184 195L 48 161 184 195L pre-stasis post-selection immortalized Real-time PCR me Figure 5 asurements of gene identified in transcriptional profiling analyses Real-time PCR measurements of gene identified in transcriptional profiling analyses. Representative genes from groups identified as changing expression during selection or immortalization were characterized by real-time PCR analysis (Taq- Man™). Genes were selected as representative of classes were described in this study. Each gene is presented as a separate graph, as identified in the figure. Cell lines are presented in the same order in each graph, as listed in the bottom left panel. The finite lifespan samples are shown as pairs, with the pre-stasis sample on the left and the post-selection sample on the right. For each cell line, expression data from Affymetrix GeneChips are shown as blue bars, according to the scale at the left of the graphs. Expression data from real-time PCR of the same samples are shown as yellow bars, according to the scale at the right of the graphs. Page 11 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 expression patterns we observe. Example images are HMEC). Levels of IGFBP4 were significantly reduced in shown in Figure 6A. For these images, Rb is shown in red 184B5ME relative to 184B5. IGFBPs are frequently and DNA is shown in blue. In the pre-stasis 184 HMEC, observed to be reduced in breast cancers, and these reduc- Rb is punctate and is evenly distributed between the tions are associated with increased sensitivity to IGF-I and nucleus and cytoplasm. In post-selection 184 HMEC and IGF-II [49,50]. in immortalized lines such as 184A1 (shown in the figure) and 184B5 (not shown), Rb is very strongly localized to (B) BRCA1, a gene deleted in about 5% of women with the nucleus, and the staining is no longer punctate. The breast cancer, encodes a protein that interacts with many nuclear/cytoplasmic ratio (determined using least 1000 other proteins [51]. These complexes recognize and cells per sample for three samples each) are shown in Fig- orchestrate the repair of DNA damage. Many genes that ure 6B for Rb and 8 other proteins. The ratio for Rb in pre- encode proteins that interact with BRCA1 were identified stasis cells is 0.5–2, whereas for post-selection and in this study as genes that increase expression following immortalized HMEC it is greater than 100. Similar dra- either selection or immortalization. BAP, RAD51, CSE1L matic changes are observed for HDAC3, BRCA1, p53 and and RFC4 all increased expression following selection in a the general transcription factor SP1. BRCA1 and c-Myc are pattern similar to the E2F-regulated genes identified as localized in the cytoplasm in pre-stasis HMEC, but to the Group C in Figure 3B. MYC, RAD50 and RFC3 increased nucleus in post-selection and immortalized HMEC. For expression in fully immortalized lines, including the p53 /- other proteins associated with G1 progression (E2F1, lines. These changes suggest the possibility that BRCA1- E2F4 and p107), the differential is in the range of two to mediated functions are affected by overcoming stasis and/ four-fold. or immortalization, which is supported by the significant change in localization of BRCA1 to the nucleus in post- Discussion selection HMEC. Transcriptional profiles and quantitative immunofluoresence of HMEC reveal significant cancer- (C) The increased expression of a well-characterized clus- associated changes following both selection and ter of IFN-regulated genes was observed in some lines in immortalization this study, as well as in other studies of HMEC [36], and The effect of malignant transformation (oncogenesis) on in a taxol-resistant MCF-7 line [52]. The IFN-dependent gene expression has been studied extensively in both cell stress response is mediated by BRCA1 [37,53]. Therefore, lines and tissues in an effort to characterize the causes of since we have noted expression changes in many genes cancer at the molecular level [45]. Gene signatures com- associated with BRCA1 function, as well as in BRCA1 monly found in breast and other human cancers include abundance and localization in post-selection HMEC, IFN those critical for the cell cycle, chromosomal stability and gene signature may reflect changes in BRCA1-mediated proliferation; the extent of the increase in the expression functions. of this signature correlates with tumor grade and poorer prognosis [26,46]. A separate signature of IFN-regulated (D) Inhibitors of Differentiation (ID) genes are important genes has also been observed in ductal carcinoma in situ regulators of differentiation by dominantly interfering (DCIS) [47] and has been associated with metastasis to with the function of bHLH proteins during embryogene- the lymph nodes in aggressive breast cancers [48]. We sis, neurodevelopment and cancer. Part of their function have observed both of these signatures in non-malignant, is through the repression of CKIs, including p16. Some immortally transformed, HMEC lines that had overcome functions have been attributed to specific members, the two senescence barriers to immortalization, despite including the interaction of ID2 with Rb [54], and the these lines retaining many characteristics of finite lifespan expression of BRCA1 by ID4 [55], which is in turn epithelial cells. repressed by BRCA1 [56]. In this study, ID1 is expressed at higher levels in the immortalized lines (184AA2 is an Transcriptional changes in gene families associated with exception), while ID4 is repressed in post-selection HMEC mammary epithelial biology or breast cancer in post- and all of the immortalized lines. selection and fully immortalized HMEC There are several gene families that we identified in this (E) S100 proteins comprise a large family of calcium-acti- study which have direct connections to breast epithelial vated proteins that function in homo- and hetero-dimers biology and breast cancer, which we can summarize: to regulate many intra- and extra-cellular targets [57]. Their increased expression in cancer and inflammatory (A) Several IGFBPs show reduced expression in post-selec- diseases has provoked interest in this family as potential tion HMEC and immortalized lines, including IGFBP2 drug targets and clinical biomarkers. We observe increases -/- (minor decreases overall, but larger in the p53 lines), in the expression of S100A8 and S100A9, which comprise IGFBP3 and IGFBP5 (very large decreases in immortal the heterodimer Calprotectin, following selection and fur- Page 12 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 Rb BRCA1 p53 1.2 2.5 1.0 2.0 2.0 0.8 1.5 0.6 1.5 0.4 1.0 1.0 0.2 0.5 0.5 0 -0.2 184 161 184 161 184A1 184B5 184 161 184 161 184A1 184B5 184 161 184 161 184A1 184B5 p107 E2F1 E2F4 1.0 1.0 1.0 0.9 0.9 0.9 0.8 0.8 0.8 0.7 0.7 0.7 0.6 0.6 0.6 0.4 0.4 0.4 0.3 0.3 0.3 0.2 0.2 0.2 0.1 0.1 0.1 184 161 184 161 184A1 184B5 184 161 184 161 184A1 184B5 184 161 184 161 184A1 184B5 HDAC3 Myc SP1 2.0 0.6 2.5 0.5 1.5 2.0 0.4 0.3 1.5 1.0 0.2 1.0 0.1 0.5 0.5 -0.1 -0.2 184 161 184 161 184A1184B5 184 161 184 161 184A1184B5 184 161 184 161 184A1184B5 pre-stasis post-selection immortalized H Figure 6 igh Content Screening of proteins associated with cell cycle progression and chromosomal stability High Content Screening of proteins associated with cell cycle progression and chromosomal stability. (A) Immunofluo- rescent images of Rb (red) and DNA (blue) obtained using a Cellomics ArrayScan Vti are shown for pre-stasis 184 HMEC (left), post-selection 184 HMEC (center) and the 184A1 cell line (right). (B) Quantitation of the nuclear/cytoplasmic ratio is shown for pre-stasis 184 and 161 HMEC, post-stasis 184 and 161 HMEC and the cell lines 184A1 and 184B5, as indicated in the figure panels. Antigens quantitated in each panel are identified above the panel. Page 13 of 17 (page number not for citation purposes) log (nuclear/cytoplasmic ratio) log (nuclear/cytoplasmic ratio) log (nuclear/cytoplasmic ratio) 10 10 log (nuclear/cytoplasmic ratio) log (nuclear/cytoplasmic ratio) log (nuclear/cytoplasmic ratio) 10 10 log (nuclear/cytoplasmic ratio) log (nuclear/cytoplasmic ratio) log (nuclear/cytoplasmic ratio) 10 10 Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 ther dramatic increases following immortalization. regulatory factor localizations we observe are concordant. Increased expression of S100P is seen in DCIS [58], and Proteins directly responsive to p16/CDK4 activation, par- was also observed in several of the immortalized lines, ticularly Rb, show striking changes in cytoplasmic/nuclear particularly 184B5ME, the ERBB2/Her2 transduced line. distribution in both post-selection and fully immortalized Increased expression of S100A7, also known as psoriasin, HMEC, compared to pre-stasis HMEC. Additional pro- is seen in both DCIS and IDC, particularly ER negative teins also showing strong changes in localization are breast cancers [59]; increased expression was observed in BRCA1, p53, HDAC3, Myc and SP1. Each of these pro- several immortalized lines, most strongly in 184AA3. teins have well characterized roles in oncogenesis and in the regulation of hTERT [63-66], a critical event in immor- Transcriptional changes that occur following genetic talization [1,5]. These changes are consistent with both changes associated with invasive cancer the transcriptional profiles we have generated of post- p53 imposes a cell cycle arrest when chromosomal break- selection and fully immortal HMEC, as well as with what age or damage is detected, and its loss in breast cancer is is known about the role of these factors on telomerase reg- associated with increased chromosomal instability and a ulation. -/- lines we have more aggressive subtype [60]. The two p53 characterized show a number of transcriptional changes The relationship between immortalized HMEC and DCIS -/- that are expected of p53 cell lines, as well as changes Taken together, these data support a classification of unique to the two lines. Of note is expression of the IFN- immortalized breast epithelial cell lines as in vitro models induced genes observed in post-selection 195L cells and of highly dysregulated epithelial cells, rather than as per- in the 184AA3 line. This may indicate a common molec- petually growing models of normal breast epithelia. Gene ular event occurred following selection of the 195L cells expression patterns we have identified in the comparison and the immortalization of the 184AA3 cells. Further of finite-lifespan and immortalized HMEC lines are -/- studies on the changes common and unique to p53 highly similar to changes observed in DCIS and invasive HMEC lines may be important in understanding differ- human breast cancers [47,67,68], and are consistent with +/+ -/- ences between p53 and p53 cell lines and breast can- other similarities between immortal HMEC lines and cers in overcoming senescence barriers and DCIS. Specifically, short telomeres and moderate chromo- immortalizing. somal instability, as well as telomerase re-activation, are common to many early-stage tumors [69], including the In data presented here, transfection of an immortalized breast [17]. In addition, p16 expression is lost in post- line with a clinically-relevant oncogene, ERBB2/Her2, selection, as it is in vHMEC [15,16], which are proposed showed fewer transcriptional changes than were observed to be premalignant breast cancer precursors in vivo. In con- following selection or immortalization, and these changes trast, we observe that a cell line, 184B5ME, which grows were generally limited to genes involved in invasive invasively in tissue culture and in in vivo models, shows growth and motility. Specifically, expression of the prolif- fewer changes. eration geneset was not dramatically altered, but there was increased expression of genes encoding the secreted pro- DCIS is a complex disease [70], often requiring no imme- teases Cystatin E/M, and Kallikrein 6, as well as tissue diate treatment in the strict sense, however it is not cur- plasminogen activator. Such changes could enable these rently possible to forecast when, or if, progression to IDC cells to grow invasively in breast tissue. will occur. This necessitates an aggressive strategy, even in cases where it may be effectively managed by substantially Activation of transcriptional regulators associated with simpler, cheaper, and less emotionally challenging modes gene expression changes in post-selection and [71]. The ability to characterize DCIS, and to target it immortalized HMEC, telomerase reactivation and cancer explicitly when it manifests invasive potential, is a critical In quiescent or unstimulated cells, many transcription fac- need with regard to effective breast cancer treatment strat- tors are excluded from the nucleus and localize to the egies. In particular, established markers for breast cancer, neu nucleus upon activation [61]. In the case of BRCA1, including Ki-67, p53, Her-2 and ER expression are very nuclear retention has been shown to suppress its pro- effective for identifying aggressive, invasive cancers, and apoptotic functions [62]. The proliferation, cell cycle and for determining the most effect treatment strategy in these DNA damage response genes identified in the gene cases, but are less informative about the likelihood that a expression signatures we observe are supported by the well-contained DCIS will progress to invasive cancer. Cur- changes in the localization of several associated regulatory rently, some of the best indicators of DCIS progression proteins and transcription factors, as determined by quan- risk are cytological, including grade, necrosis and architec- titative immunofluorescence. Based on previous studies tural patterns [72]. Additional molecular markers, partic- linking regulatory pathways to gene expression, the rela- ularly those that correlate strongly (or better, explain) the tionship between the gene expression signatures and the histological patterns used to stage DCIS would be very val- Page 14 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 uable. Some additional molecular markers are emerging. Abbreviations COX-2 has been identified as a marker of vHMEC [15,16], LI, labeling index; HMEC, human mammary epithelial and expression levels have been correlated with DCIS cells; CKI, cyclin-dependent kinase inhibitor, DCIS, duc- grade, as well [73]. For these reasons, recognizing immor- tal carcinoma in situ; IDC, invasive ductal carcinoma; PD, talized HMEC as resembling early-stage cancers would population doubling; ANOVA, analysis of variance; CIN, facilitate a formal interrogation of their genetics and phys- chromosomal instability; IGFBP; insulin-like growth fac- iology for clues to how DCIS occurs, and to the factors tor binding protein; SA-b-gal, senescence associated beta- that can enable DCIS to progress. galactosidase; ECM, extracellular matrix; HCS, high con- tent screening. Use of post-selection and immortalized HMEC to study normal mammary cell biology and breast cancer Competing interests Immortalized cell lines have been used to address com- The author(s) declare that they have no competing inter- plex problems in cancer [74] and epithelial cell biology ests. [75] precisely because they allow for controlled experi- ments to be performed and theories of breast cancer to be Authors' contributions tested. In studies of oncogenesis, the non-malignant sta- JP, JHL, CT, and KS performed experiments and analyzed tus of immortalized lines allows for the specific steps in primary data. YL, J-JL, MW, SJ and SH analyzed normal- full malignant transformation to be examined, such as by ized data and interpreted results. JG and MS developed the introduction of activated oncogenes [76,77]. How- cell lines and analyzed normalized data. JP performed ever, in many cases immortalized cell lines are referred to cell-based assays on the transcription factors and regula- and used as "normal" cells. This inaccurate characteriza- tory proteins. JP and SH analyzed data from the cell-based tion may obscure understanding of the multiple errors assays. YL, MS and SH wrote the manuscript. All authors that permit immortal transformation, and thus aspects of read and approved the final version of the manuscript. early stage carcinogenesis. While established breast cancer cell lines are usually derived from advanced, metastatic Additional material tumors (particularly pleural effusions), and therefore are quite different from immortalized cell lines, immortal- Additional file 1 ized lines themselves have undergone extensive genetic Table s1: Genes Expressed Concordantly in Pre-Stasis and Post-Selection and epigenetic changes, especially in frequently studied Cell Types. Compilation of genelists that distinguish the two classes of aspects of oncogenesis, such as G1 checkpoint function finite-lifespan HMEC strains. and the DNA damage response. The use of immortalized Click here for file [http://www.biomedcentral.com/content/supplementary/1476- HMEC as "normal" controls for tumor-derived lines can 4598-6-7-S1.doc] impede our ability to understand early stages of carcino- genesis, and obscure the potential of treating DCIS-stage Additional file 2 changes as additional targets for clinical benefit. Table s2. Genes Concordantly Expressed in Pre-stasis, Post-selection or Fully Immortalized HMEC. Compilations of genelists that define the three Conclusion classes of non-malignant HMEC cell strains and lines. Gene expression profiles and cytological changes in Click here for file related transcriptional regulators indicate that immortal- [http://www.biomedcentral.com/content/supplementary/1476- 4598-6-7-S2.doc] ized HMEC resemble non-invasive breast cancers, such as ductal and lobular carcinomas in situ, and are strikingly Additional file 3 distinct from finite-lifespan HMEC, particularly with +/+ Table s3: Genes Expressed Concordantly in p53 (184A1 and 184B5) regard to genes involved in proliferation, cell cycle regula- -/- or p53 (184AA2 and 184AA3) HMEC. Compilations of commonly tion, chromosome structure and the DNA damage expressed genes in multiple wild type and p53 immortalized HMEC cell response. The comparison of HMEC profiles with lines lines. harboring oncogenic changes (e.g. overexpression of Her- Click here for file neu [http://www.biomedcentral.com/content/supplementary/1476- 2 , loss of p53 expression) identifies genes involved in 4598-6-7-S3.doc] tissue remodeling as well as proinflamatory cytokines and S100 proteins. Studies on carcinogenesis using immortal- Additional file 4 ized cell lines as starting points or "normal" controls need Table s4. Gene Expression changes of p53 cell lines 184A1 versus 184B5. to account for the significant pre-existing genetic and epi- Compilations of genelists and expression statistics of genes expressed genetic changes inherent in such lines before results can uniquely in two p53 wild type HMEC cell lines. be broadly interpreted. Click here for file [http://www.biomedcentral.com/content/supplementary/1476- 4598-6-7-S4.doc] Page 15 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 13. Geradts J, Wilson PA: High frequency of aberrant p16INK4A expression in human breast cancer. American Journal of Pathology Additional file 5 1996, 149:15-20. Table s5. Gene Expression changes of p53 cell lines 184A1 versus 184B5. 14. Baylin SB, Herman JG, Graff JR, Vertino PM, Issa JP: Alterations in DNA methylation: a fundemental aspect of neoplasia. Advances Compilations of genelists and expression statistics of genes expressed in Cancer Research 1998, 72:141-196. uniquely in two p53 HMEC cell lines. 15. Crawford YG, Gauthier ML, Joubel A, Mantei K, Kozakiewicz K, Afshar Click here for file CA, Tlsty TD: Histologically normal human mammary epithelia with silenced p16(INK4a) overexpress COX-2, promoting a [http://www.biomedcentral.com/content/supplementary/1476- premalignant program. Cancer Cell 2004, 5:263-273. 4598-6-7-S5.doc] 16. Holst CR, Nuovo GJ, Esteller M, Chew K, Baylin SB, Herman JG, Tlsty TD: Methylation of p16(INK4a) promoters occurs in vivo in histologically normal human mammary epithelia. Cancer Additional file 6 Research 2003, 63:1596-1601. Table s6. Gene Expression Changes Resulting from Expression of ERB-B2/ 17. Chin K, de Solorzano CO, Knowles D, Jones A, Chou W, Rodriguez neu Her2 in 184B5. Compilations of genelists and expression statistics of EG, Kuo WL, Ljung BM, Chew K, Myambo K, Miranda M, Krig S, Garbe neu genes that change expression following overexpression of the Her-2 J, Stampfer M, Yaswen P, Gray JW, Lockett SJ: In situ analyses of genome instability in breast cancer. Nature Genetics 2004, oncogene. 36:984-988. Click here for file 18. Goldstein JC, Rodier F, Garbe JC, Stampfer MR, Campisi J: Caspase- [http://www.biomedcentral.com/content/supplementary/1476- independent cytochrome c release is a sensitive measure of low-level apoptosis in cell culture models. Aging Cell 2005, 4598-6-7-S6.doc] 4:217-222. 19. Stampfer MR, Garbe J, Levine G, Lichtsteiner S, Vasserot AP, Yaswen P: Expression of the telomerase catalytic subunit, hTERT, induces resistance to transforming growth factor beta growth inhibition in p16INK4A(-) human mammary epithelial cells. Proceedings of the National Academy of Sciences, USA 2001, 98:4498-4503. Acknowledgements 20. Clark R, Stampfer MR, Milley R, O'Rourke , Walen KH, Kriegler M, Work performed at LBNL was supported by the Department of Defense Kopplin JMC F.: Transformation of human mammary epithelial BCRP grant W81XWH-04-1-0580 to M.R.S and contract AC03-76SF00098 cells by oncogenic retroviruses. Cancer Research 1988, to the University of California from the US Department of Energy. 48:4689-4694. 21. Pierce JH, Arnstein P, DiMarco E, Artrip J, Kraus MH, Lonardo F, Di Fiore PP, Aaronson SA: Oncogenic potential of erbB-2 in human References mammary epithelial cells. Oncogene 1991, 6:1189-1194. 22. Frittitta L, Vigneri R, Stampfer MR, Goldfine ID: Insulin receptor 1. Meyerson M, Counter CM, Eaton EN, Ellisen LW, Steiner P, Caddle SD, overexpression in 184B5 human mammary epithelial cells Ziaugra L, Beijersbergen RL, Davidoff MJ, Liu Q, Bacchetti S, Haber DA, induces a ligand-dependent transformed phenotype. Journal of Weinberg RA: hEST2, the putative human telomerase catalytic Cellular Biochemistry 1995, 57:666-669. subunit gene, is up-regulated in tumor cells and during 23. Slonim DK: From patterns to pathways: gene expression data immortalization. Cell 1997, 90:785-795. analysis comes of age. Nature Genetics 2002, 32(s502-s508.):. 2. Stampfer M: Cholera toxin stimulation of human mammary epi- 24. Perou CM, Sorlie T, Eisen MB, van de Rijn M, Jeffrey SS, Rees CA, Pol- thelial cell in culture. In vitro 1982, 18:531-537. lack JR, Ross DT, Johnsen H, Akslen LA, Fluge O, Pergamenschikov A, 3. Hammond SL, Ham RG, Stampfer MR: Serum-free growth of Williams C, Zhu SX, Lonning PE, Borresen-Dale AL, Brown PO, Bot- human mammary epithelial cells: Rapid clonal growth in stein D: Molecular portraits of human breast tumours. Nature defined medium and extended serial passage with pituitary 2000, 406:747-752. extract. Proceedings of the National Academy of Sciences, USA 1984, 25. Sorlie T, Perou CM, Tibshirani R, Aas T, Geisler S, Johnsen H, Hastie T, 81:5435-5439. Eisen MB, van de Rijn M, Jeffrey SS, Thorsen T, Quist H, Matese JC, 4. Stampfer MR, Yaswen P: Human epithelial cell immortalization Brown PO, Botstein D, Eystein Lonning P, Borresen-Dale AL: Gene as a step in carcinogenesis. Cancer Letters 2003, 194:199-208. expression patterns of breast carcinomas distinguish tumor 5. Stampfer MR, Bodnar A, Garbe J, Wong M, Pan A, Villeponteau B, Yas- subclasses with clinical implications. Proceedings of the National wen P: Gradual phenotypic conversion associated with immor- Academy of Sciences, USA 2001, 98:10869-10874. talization of cultured human mammary epithelial cells. 26. Whitfield ML, George LK, Grant GD, Perou CM: Common markers Molecular and Cellular Biology 1997, 8:2391-2405. of proliferation. Nature Reviews Cancer 2006, 6:99-106. 6. Brenner AJ, Stampfer MR, Aldaz CM: Increased p16 expression 27. Byrne MC, Whitley MZ, Follettie MT: Preparation of mRNA for with first senescence arrest in human mammary epithelial expression monitoring. In Current Protocols in Molecular Biology John cells and extended growth capacity with p16 inactivation. Wiley and Sons; 2000:22.2.1-22.2.13. Oncogene 1998, 17:199-205. 28. Hill AA, Brown EL, Whitley MZ, Tucker-Kellogg G, Hunter CP, Slonim 7. Romanov S, Kozakiewicz K, Holst C, Stampfer MR, Haupt LM, Tlsty DK: Evaluation of normalization procedures for oligonucle- TD: Normal human mammary epithelial cells spontaneously otide array data based on spiked cRNA controls. Genome Biology escape senescence and aquire genomic changes. Nature 2001, 2001, 2:0055.1-55.13. 409:633-637. 29. Affymetrix I: http://www.affymetrix.com/support/technical/ 8. Nonet G, Stampfer MR, Chin K, Gray JW, Collins CC, Yaswen P: The manuals.affx. . ZNF217 gene amplified in breast cancers promotes immortal- 30. Eisen MB, Spellman PT, Brown POB D.: Cluster analysis and display ization of human mammary epithelia cells. Cancer Research of genome-wide expression patterns. Proceedings of the National 2001, 61:1250-1254. Academy of Sciences, USA 1998, 95:14863-14868. 9. Stampfer MR, Garbe J, Nijjar T, Wiginton D, Swisshelm K, Yaswen P: 31. Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP, Loss of p53 function accelerates acquisition of telomerase Coller H, Loh ML, Downing JR, Caligiuri MA, Bloomfield CD, Lander ES: activity in indefinite lifespan human mammary epithelial cell Molecular classification of cancer: class discovery and class lines. Oncogene 2003, 22:5238-5251. prediction by gene expression monitoring. Science 1999, 10. Strunnikova M, Schagdarsurengin U, Kehlen A, Garbe JC, Stampfer MR, 286:531-537. Dammann R: Chromatin inactivation precedes de novo DNA 32. Reich M, Ohm K, Angelo M, Tamayo P, Mesirov JP: GeneCluster 2.0: methylation during the progressive epigenetic silencing of the an advanced toolset for bioarray analysis. Bioinformatics 2004, RASSF1A promoter. Molecular and Cellular Biology 2005, 20:1797-1798. 25:3923-3933. 33. Frith M, Fu Y, Yu L, Chen JF, Hansen U, Weng Z: Detection of func- 11. Stampfer MR, Bartley JC: Induction of transformation and contin- tional DNA motifs via statistical over-representation. Nucleic uous cell lines from normal human mammary epithelial cells Acids Research 2004, 32:1372-1381. after exposure to benzo(a)pyrene. Proceedings of the National 34. Kel AE, Gossling E, Reuter I, Cheremushkin E, Kel-Margoulis OV, Win- Academy of Sciences, USA 1985, 82:2394-2398. gender E: MATCH: A tool for searching transcription factor 12. Garbe J, Wong M, Wigington D, Yaswen P, Stampfer MR: Viral onco- binding sites in DNA sequences. Nucleic Acids Research 2003, genes accelerate conversion to immortality of cultured 31:3576-3579. human mammary epithelial cells. Oncogene 1999, 18:2169-2180. 35. Der SD, Zhou A, Williams BR, Silverman RH: Identification of genes differentially regulated by interferon alpha, beta, or gamma Page 16 of 17 (page number not for citation purposes) Molecular Cancer 2007, 6:7 http://www.molecular-cancer.com/content/6/1/7 using oligonucleotide arrays. Proceedings of the National Academy of 58. Poola I, DeWitty RL, Marshalleck JJ, Bhatnagar R, Abraham J, Leffall LD: Sciences, USA 1998, 95:15623-15628. Identification of MMP-1 as a putative breast cancer predictive 36. Zhang H, Herbert BS, Pan KH, Shay JW, Cohen SN: Disparate effects marker by global gene expression analysis. Nature Medicine 2005, of telomere attrition on gene expression during replicative 11:481-483. senescence of human mammary epithelial cells cultured 59. Emberley ED, Niu Y, Njue C, Kliewer EV, Murphy LC, Watson PH: under different conditions. Oncogene 2004, 23:6193-6198. Psoriasin (S100A7) expression is associated with poor out- 37. Andrews HN, Mullan PB, McWilliams S, Sebelova S, Quinn JE, Gilmore come in estrogen receptor-negative invasive breast cancer. PM, McCabe N, Pace A, Koller B, Johnston PG, Haber DA, Harkin DP: Clinical Cancer Research 2003, 2627-2631:. BRCA1 regulates the interferon gamma-mediated apoptotic 60. Miller LD, Smeds J, George J, Vega VB, Vergara L, Ploner A, Pawitan Y, response. Journal of Biological Chemistry 2002, 277:26225-26232. Hall P, Klaar S, Liu ET, Bergh J: An expression signature for p53 38. Muller H, Bracken AP, Vernell R, Moroni MC, Christians F, Grassilli E, status in human breast cancer predicts mutation status, tran- Prosperini E, Vigo E, Oliner JD, Helin K: E2Fs regulate the expres- scriptional effects, and patient survival. Proceedings of the National sion of genes involved in differentiation, development, prolif- Academy of Sciences, USA 2005, 102:13550-13555. eration, and apoptosis. Genes and Developement 2001, 15:267-285. 61. Ziegler EC, Ghosh S: Regulating inducible transcription through 39. Ren B, Cam H, Takahashi Y, Volkert T, Terragni J, Young RA, Dynlacht controlled localization. SciSTKE 2005, 284:re6. BD: E2F integrates cell cycle progression with DNA repair, 62. Fabbro M, Schuechner S, Au WW, Henderson BR: BARD1 regulates replication, and G(2)/M checkpoints. Genes and Developement BRCA1 apoptotic function by a mechanism involving nuclear 2002, 16:245-256. retention. Experimental Cell Research 2004, 298:661-673. 40. Coller HA, Grandori C, Tamayo P, Colbert T, Lander ES, Eisenman RN, 63. Won J, Yim J, Kim TK: Sp1 and Sp3 recruit histone deacetylase Golub TR: Expression analysis with oligonucleotide microar- to repress transcription of human telomerase reverse tran- rays reveals that MYC regulates genes involved in growth, cell scriptase (hTERT) promoter in normal human somatic cells. cycle, signaling, and adhesion. Proceedings of the National Academy Journal of Biological Chemistry 2002, 277(38230-383238.):. of Sciences, USA 2000, 97:3260-3265. 64. Cong YS, Bacchetti S: Histone deacetylation is involved in the 41. Lawlor ER, Soucek L, Brown-Swigart L, Shchors K, Bialucha CU, Evan transcriptional repression of hTERT in normal human cells. GI: Reversible kinetic analysis of Myc targets in vivo provides Journal of Biological Chemistry 2000, 275:35665-35668. novel insights into Myc-mediated tumorigenesis. Cancer 65. Kyo S, Takakura M, Taira T, Kanaya T, Itoh H, Yutsudo M, Ariga H, Research 2006, 66:4591-4601. Inoue M: Sp1 cooperates with c-Myc to activate transcription of 42. Watson JD, Oster SK, Shago M, Khosravi F, Penn LZ: Identifying the human telomerase reverse transcriptase gene (hTERT). genes regulated in a Myc-dependent manner. Journal of Biological Nucleic Acids Research 2000, 28:669-677. Chemistry 2002, 277:36921-36930. 66. Li H, Lee TH, Avraham H: A novel tricomplex of BRCA1, Nmi, 43. Paine TM, Soule HD, Pauley RJ, Dawson PJ: Characterization of epi- and c-Myc inhibits c-Myc-induced human telomerase reverse thelial phenotypes in mortal and immortal human breast transcriptase gene (hTERT) promoter activity in breast can- cells. International Journal of Cancer 1992, 50:463-473. cer. Journal of Biological Chemistry 2002, 277:20965-20973. 44. Zhang J, R. S, Dai Q, Song J, Barlow SC, Yin L, Sloane BF, Miller FR, 67. Porter D, Lahti-Domenici J, Keshaviah A, Bae YK, Argani P, Marks J, Meschonat C, Li BD, Abreo F, Keppler D: Cystatin m: a novel can- Richardson A, Cooper A, Strausberg R, Riggins GJ, Schnitt S, Gabrielson didate tumor suppressor gene for breast cancer. Cancer E, Gelman R, Polyak K: Molecular markers in ductal carcinoma in Research 2004, 64:6957-6964. situ of the breast. Molecular Cancer Research 2003, 1:362-375. 45. Rhodes DR, Kalyana-Sundaram S, Mahavisno V, Barrette TR, Ghosh D, 68. Zhao H, Langerod A, Ji Y, Nowels KW, Nesland JM, Tibshirani R, Chinnaiyan AM: Mining for regulatory programs in the cancer Bukholm IK, Karesen R, Botstein D, Borresen-Dale AL, Jeffrey SS: Dif- transcriptome. Nature Genetics 2005, 37:579-583. ferent gene expression patterns in invasive lobular and ductal 46. Dai H, van't Veer L, Lamb J, He YD, Mao M, Fine BM, Bernards R, van carcinomas of the breast. Molecular Biology of the Cell 2004, de Vijver M, Deutsch P, Sachs A, Stoughton R, Friend S: A cell prolif- 15:2523-2536. eration signature is a marker of extremely poor outcome in a 69. Meeker AK, Hicks JL, Iacobuzio-Donahue CA, Montgomery EA, subpopulation of breast cancer patients. Cancer Research 2005, Westra WH, Chan TY, Ronnett BM, De Marzo AM: Telomere 65:4059-4066. length abnormalities occur early in the initiation of epithelial 47. Seth A, Kitching R, Landberg G, Xu J, Zubovits J, Burger AM: Gene carcinogenesis. Clinical Cancer Research 2004, 10:3317-3326. expression profiling of ductal carcinomas in situ and invasive 70. Burstein HJ, Polyak K, Wong JS, Lester SC, Kaelin CM: Ductal Carci- breast tumors. Anticancer Research 2003, 23:2043-2051. noma in Situ of the Breast. New England Journal of Medicine 2004, 48. Huang E, Cheng SH, Dressman H, Pittman J, Tsou MH, Horng CF, Bild 350:1430-1441. A, Iversen ES, Liao M, Chen CM, West M, Nevins JR, Huang AT: Gene 71. Baxter NN, Virnig BA, Durham SB, Tuttle TM: Trends in the Treat- expression predictors of breast cancer outcomes. Lancet 2003, ment of Ductal Carcinoma In Situ of the Breast. Journal of the 361:1590-1596. National Cancer Institute 2004, 96:443-448. 49. Pollak MN, Schernhammer ES, Hankinson SE: Insulin-like growth 72. Tsikitis VL, Chung MA: Biology of ductal carcinoma in situ classi- factors and neoplasia. Nature Reviews Cancer 2004, 4:505-518. fication based on biologic potential. American Journal of Clinical 50. Yu H, Rohan T: Role of the insulin-like growth factor family in Oncology 2006, 29:305-310. cancer development and progression. Journal of the National Can- 73. Boland GP, Butt IS, Prasad R, Knox WF, Bundred NJ: COX-2 expres- cer Institute 2000, 92:1472-1489. sion is associated with an aggressive phenotype in ductal car- 51. Deng CX, Brodie SG: Roles of BRCA1 and its interacting pro- cinoma in situ. British Journal of Cancer 2004, 90:423-429. teins. Bioessays 2000, 22:728-737. 74. Adler AS, Lin M, Horlings H, Nuyten DS, van de Vijver MJ, Chang HY: 52. Luker KE, Pica CM, Schreiber RD, Piwnica-Worms D: Overexpres- Genetic regulators of large-scale transcriptional signatures in sion of IRF9 confers resistance to antimicrotubule agents in cancer. Nature Genetics 2006, 38:421-430. breast cancer cells. Cancer Research 2001, 61:6540-6547. 75. Debnath J, Mills KR, Collins NL, Reginato MJ, Muthuswamy SK, Brugge 53. Gilmore PM, McCabe N, Quinn JE, Kennedy RD, Gorski JJ, Andrews JS: The role of apoptosis in creating and maintaining luminal HN, McWilliams S, Carty M, Mullan PB, Duprex WP, Liu ET, Johnston space within normal and oncogene-expressing mammary PG, Harkin DP: BRCA1 interacts with and is required for paclit- acini. Cell 2002, 111:29-40. axel-induced activation of mitogen-activated protein kinase 76. Thompson EW, Torri J, Sabol M, Sommers CL, Byers S, Valverius EM, kinase kinase 3. Cancer Research 2004, 64:4184-4154. Martin GR, Lippman ME, Stampfer MR, Dickson RB: Oncogene- 54. Lasorella A, Noseda M, Beyna M, Yokota Y, Iavarone A: Id2 is a retin- induced basement membrane invasiveness in human mam- oblastoma protein target and mediates signalling by Myc mary epithelial cells. Clinical and Experimental Metastasis 1994, oncoproteins. Nature 2000, 407:592-598. 12:181-194. 55. Beger C, Pierce LN, Kruger M, Marcusson EG, Robbins JM, Welcsh P, 77. Wang B, Soule HD, FR. M: Transforming and oncogenic potential Welch PJ, Welte K, King MC, Barber JR, Wong-Staal F: Identification of activated c-Ha-ras in three immortalized human breast epi- of Id4 as a regulator of BRCA1 expression by using a thelial cell lines. Anticancer Research 1997, 17:4387-4394. ribozyme-library-based inverse genomics approach. Proceed- ings of the National Academy of Sciences, USA 2001, 98:130-135. 56. Welcsh PL, Lee MK, Gonzalez-Hernandez RM, Black DJ, Mahadevappa M, Swisher EM, Warrington JA, King MC: BRCA1 transcriptionally regulates genes involved in breast tumorigenesis. Proceedings of the National Academy of Sciences, USA 2002, 28:7560-7565. 57. Donato R: S100: a multigenic family of calcium-modulated pro- teins of the EF-hand type with intracellular and extracellular functional roles. International Journal of Biochemistry and Cell Biology 2001, 33:637-668. Page 17 of 17 (page number not for citation purposes)

Journal

Molecular Cancer – Springer Journals

Published: Jan 18, 2007

There are no references for this article.