Mutagenesis and selection was initiated in five parallel tissue-culture plates seeded with 1 million MCF-10A cells in each plate (subsequently referred to as MCF10Aα, MCF10Aβ, MCF10Aγ, MCF10Aδ and MCF10Aε). As a result, five independent sets of MCF-10A cultures with various degrees of transformation were generated. Two consecutive steps of in vitro selection were used to transform the MCF-10A cells. The first selective step promoted clonal expansion of cells that had acquired tolerance to DNA damage-signaling which was achieved by cell cultivation in a complete medium containing 500 ng/ml of ICR191. This concentration of drugs allowed cell proliferation despite the DNA damage-response activation manifested as an increased level of TP53 protein, alterations in cell-cycle progression and changes in cell morphology in tissue culture, which was manifested as a slightly larger cell size and more flat appearance (Figure 1A–C).

After six months of selection for the resistance to DNA damage-signaling, all five cultures had a higher than control untreated cells rate of replication in the presence of mutagen (not shown). Also, all five cultures had acquired signs of cell transformation, such as the loss of contact inhibition at high density and the ability to form growing colonies in soft agar (Figures 2A and 2B). One culture (MCF10Aε) has acquired a higher rate of cell proliferation in the absence of mutagen (Figure 2C). Following subcutaneous injections of 5 million cells into the flanks of nude mice, three out of five cultures (MCF10Aα, MCF10Aγ and MCF10Aδ) produced small (~2 mm in diameter) palpable tumors. These tumors were localized at the site of inoculation for more than four months without an increase in size.

The second step of selection used to isolate more aggressive variants of the MCF-10A cells involved selection for either the ability of anchorage-independent growth or the ability to grow in a medium with either reduced growth-factor supply or resistance to hypoxia. Selection for anchorage-independent growth was applied to the MCF10Aα and MCF10Aβ cultures. Cells from these cultures had been cloned in soft agar and, after two weeks, four of the fastest-growing colonies of MCF10Aα cells, and one colony of MCF10Aβ cells, were isolated and expanded in culture. After subcutaneous injections of 5 million of these cells into the flanks of nude mice, the soft agar-derived clone of MCF10Aβ cells (MCF10Aβ1) produced a large (~1 cm in diameter), progressively growing tumor four months after inoculation. Two soft agar-derived clones from the MCF10α culture, MCF10α3 and MCF10α4, produced large and small tumors, respectively.

After selecting for the ability to grow in the medium with reduced growth-factor supply, two of the five cultures (MCF10Aα and MCF10Aβ) acquired the ability to form large, progressively growing tumors, and three cultures (MCF10Aγ, MCF10Aδ and MCF10Aε) produced small tumors four months after inoculation. After selecting for resistance to hypoxia, three cultures (MCF10α, MCF10Aβ and MCF10Aγ) produced large tumors, and the remaining two cultures (MCF10δ and MCF10ε) produced small tumors. Large tumors produced by transformed cultures were shown to have undergone extreme angiogenesis (Figure 3A) and showed different degrees of differentiation and invasiveness (Figure 3C–H).

Individual cell-lines were established from both large and small tumors and expanded in culture. The names of cell-lines derived from large and small tumors are subsequently denoted with the letters L and S, respectively. Inoculations of cell-lines derived from large tumors resulted in the formation of large tumors (1 cm in diameter) in five out of five mice after one month. In contrast, inoculations of cell-lines derived from the small tumors produced either small tumors or none at all.

To identify chromosomal changes associated with the transformation of MCF-10A cells, we compared the karyotype of the transformed MCF10AαL culture derived from the tumor grown in nude mice with that of control MCF-10A cells using SKY. Control MCF-10A cells are almost diploid cells with several chromosomal abnormalities. All MCF-10A cells from our lab have a reciprocal t(3;9)(p14;p21) translocation and an unbalanced translocation from chromosome 5q to the derivative 910. Loss of the p16 locus associated with t(3;9) translocation is believed to contribute to the immortalization of MCF-10A cells17. Also, a t(6;19)(p25;q12) rearrangement can be seen in all control MCF-10A cells used in our lab (Figure 4A). In addition, trisomies of chromosomes 20 and 16, as well as an additional structurally abnormal derivative of chromosome 8, were present in all control MCF-10A cells. Few additional chromosomal alterations have occurred in the transformed cells. Figure 4A shows the duplication of a 16q arm and the loss of one copy of a 16p arm in MCF10AαL cells. The other alteration that has occurred in transformed MCF10AαL cells was the loss of one copy of chromosome 19. Also, one derivative, which includes the translocation of chromosomes 9, 3 and 5, has been lost in the transformed MCF10AαL. A comparison of MCF10AαL cells versus control MCF-10A cells using array comparative genomic hybridization (aCGH) has also identified the duplication of 16q, the loss of part of 16p and the loss of t(9;3;5) derivative in transformed cells (Figure 4B).

To identify mutations that might have contributed to MCF-10A cell-transformation, we used GINI analysis first on the MCF10Aβ1 cells, as described in Material and Methods. These cells, which were isolated from the colony growing in soft agar, produce large, continuously growing tumors in nude mice, indicating that breast cancer-related tumor-suppressors could be mutated and selected during transformation. Analyzing levels of mRNA transcripts in the MCF10Aβ1 and control MCF1-10A cells using Affymetrix U133Plus2.0 GeneChips before and after inhibition of NMD with emetine, we have selected 14 genes that satisfied the requirements for candidates for sequencing analysis (see Material and Methods). The candidates included the genes coding for TP53, smoothelin (SMTN) and ubiquitin specific peptidase 10 (USP10), which contained insertions of one cytosine in short coding poly-cytosine repeats, which are the characteristic mutations induced by ICR191. The TP53 gene in the MCF10Aβ1 cells contained two heterozygous mutations, resulting in the absence of p53 protein expression (Figure 5A), while the SMTN (Figure 5C) and UPS10 (not shown) genes had homozygous and heterozygous mutations, respectively.

The TP53 gene is known to be frequently mutated in more than half of all human cancers, including breast cancer. Since the TP53 gene is the main regulator of DNA damage-response18 and the transformed derivatives of MCF-10A cells had been selected for the ability to replicate despite DNA damage-response activation, we analyzed all the transformed MCF-10A cultures for the mutations in the p53 gene. We found that all MCF-10A cultures that can produce large, continuously growing tumors in nude mice contained p53 gene mutations. The MCF10AγL culture had two heterozygous mutations that were different from those in the MCF10β1 cells (Figure 5A), and the MCF10AαL culture had one heterozygous mutation that was identical to one mutation found in the MCF10 β1 cells. Although the mutation had been identified only in one allele, the MCF10AαL cells, similarly to the MCF10AγL and MCF10β1, do not express p53 protein (Figure 5B). The DNA sequencing analysis of the p53 gene in MCF10Aα3 and MCF10Aα4 cells generated from soft agar growing clones has identified the mutation identical to that in MCF10AαL cells. Western blot analysis has shown that MCF10Aα3 cells, which can produce large tumors after inoculations into nude mice, do not express p53 protein, while the MCF10Aα4, which cannot produce large tumors, do. Similarly, MCF10Aδ and MCF10Aε cultures, which, after all rounds of selection, could form only small tumors in nude mice, had a wild-type allele of the p53 gene and showed p53 protein-accumulation in response to treatment with ICR191 (Figure 5B). On the other hand, the MCF10AγS cells, which do not express p53 protein due to mutations identical to that in MCF10AγL cells, did not produce large tumors in nude mice, suggesting that p53 inactivation was necessary but not sufficient for in vivo tumor growth.

In contrast to the p53 gene mutations, the mutation in the SMTN gene had been found only in the MCF10Aβ1 cells. All other transformed MCF-10A cultures, including the related MCF10AβL cell-line derived from a large tumor produced by MCF10Aβ cells selected for resistance to hypoxia, contained wild-type alleles of this gene. To identify other genes mutated in the process of in vitro transformation, we analyzed, using GINI, the MCF10Aα4 and MCF10αL cultures. Sequencing of selected candidate genes has identified heterozygous frameshift mutations in the Ras association (RalGDS/AF-6) domain family 6 (RASSF6) (Figure 5C) in the MCF10Aα4 cells and in the kinesin family member 22 (KIF22) in the MCF10αL cells.

The immortalized, non-transformed human mammary epithelial cell-line, MCF-10A, was grown on standard tissue-culture plastic in a 5% CO₂-humidified incubator at 37°C in Complete MCF-10A Growth Medium composed of DMEM/F12 (Mediatech, Inc., Herndon, VA) supplemented with 5% donor horse serum, 20 ng/ml epidermal growth factor (EGF), 10 μg/ml insulin, 0.5 μg/ml hydrocortisone (Sigma, St. Louis, MO), 100 ng/ml cholera toxin (Cambrex, Westborough, MA), and 100 units/ml penicillin and 100 μg/ml streptomycin (Invitrogen, Carlsbad, CA). A Growth Factor-Reduced MCF-10A Medium for the second step of transformation and culturing transformed clones was composed of DMEM/F12 supplemented with 5% donor horse serum (Sigma, St. Louis, MO) and 100 units/ml penicillin and 100 μg/ml streptomycin (Invitrogen, Carlsbad, CA), without the addition of supplements.

The acridine mutagen ICR191 (Sigma Chemical Co., St. Louis, MO) was dissolved at 1 mg/mL in 0.01 M HCl and was used as a stock solution before diluting in the tissue-culture medium at 500 ng/ml concentration.

Two steps of mutagenesis and selection had been used to achieve the transformation of MCF-10A cells. The first step of selection was for the ability to replicate despite DNA damage-response activation. In this step, 1 million MCF-10A cells were seeded in a 100 mm tissue-culture plate and cultivated for six months in Complete MCF-10A Growth Medium containing 500 ng/ml of ICR191. During this process, as cells reached confluence, they were trypsinized, and half the cells were frozen in liquid nitrogen. The remaining cells were returned to the culture. In the second step, cells were selected either for the ability to grow in the medium with reduced growth-factor supply, for the ability to grow in the hypoxic condition, or for anchorage-independent growth. To select cells resistant to a reduced growth-factor supply, cells were cultivated in the Growth Factor-Reduced MCF-10A Medium containing 500 ng/ml of ICR191 for one to three months. To select cells resistant to hypoxia, cells were cultivated in the Complete MCF-10A Growth Medium containing 100 μM of cobalt chloride, which is known to generate hypoxic environment 22, and 500 ng/ml of ICR191 for one month. To select for the ability for anchorage-independent growth, cells were challenged to grow in soft agar.

1 × 10⁵ cells in Complete MCF-10A Growth Medium were seeded into six-well plates with 2 ml 0.33% (w/v) low-melting agarose, which was overlaid onto 2 ml of 0.6% (w/v) agarose and fed every other day by exchanging 1.5 ml medium on top of the soft agar. Soft-agar cultures were maintained at 37°C and observed microscopically for the appearance of colonies. After three to four weeks in culture, the colonies were counted. In cases where cultures were to be established from anchorage-independent colonies, the top layer of low-melting agarose was plated into tissue-culture dishes and kept in Complete MCF-10A Growth Medium. Clones were picked using cloning cylinders and developed into cell-lines.

Cells in phosphate-buffered saline were mixed 1:1 with Matrigel basement membrane matrix (BD Biosciences, Franklin Lakes, NJ) and 3–4 × 10⁶ cells in 100–150 μl were injected subcutaneously into the flanks of 5-week-old athymic female Ncr -nu/nu mice (National Cancer Institute). The animals were maintained in accordance with institutional guidelines at the Laboratory Animal Facility at RPCI for six to eight months and palpated weekly for tumor appearance. When tumor nodules reached 1–2 cm in size, the mice were euthanized by exposure to carbon dioxide. The tumors were excised and divided into three pieces for developing into cell-lines to be frozen for future microarray analysis, fixed in neutral buffered formalin and embedded in paraffin for a routine histological evaluation with 5 μm hematoxylin and eosin (H&E) stained sections. Histological procedures were performed at the RPCI Pathology Core Facility.

Identifying mutant genes in the transformed MCF-10A cells using inhibition of nonsense-mediated decay and Affymetrix oligonucletide U133Plus2.0 GeneChip analysis was performed as described previously for prostate cells 16. Briefly, control untransformed MCF-10A cells and the MCF-10A transformed derivative cells were grown in Complete MCF-10A Growth Medium until they reached 70–80% of confluency. To inhibit NMD in cells, emetine (100 μg/ml) was added to the tissue-culture medium and, eight hours later, the total RNA was prepared with TRIZOL reagents (Invitrogen, Carlsbad, CA) according to manufacturer's specifications from emetine-treated as well as untreated cells. The integrity and purity of the total RNA were analyzed using a microfluidic electrophoretic system (Agilent 2100 Bioanalyzer; Agilent Technologies). RNA samples were processed and then hybridized to Human Genome U133APlus2.0 GeneChip (Affymetrix Inc, Santa Clara, CA), according to the manufacturer's instructions, by the Gene Expression Core facility at Roswell Park Cancer Institute. GeneChip expression array analysis was performed using Affymetrix Microarray Suite software, version 5 (GeneChip Analysis Suite). This software uses two independent sets of algorithms, a quantitative algorithm (robust estimator of the mean difference in probe intensities) that computes the raw signal and the SLR directly from the hybridization intensities of the probes, as well as a confidence algorithm (non-parametric) that provides P-values to estimate the confidence in detection (absent or present) or change (increase, decrease or no change) of a specific target.

To select candidate genes out of hundreds genes that show mRNA increase following emetine treatment in the transformed cells for sequencing analysis, we used the following cut-off criteria:1) the level of expression of a gene in the emetine-untreated transformed cells had to be more than threefold lower than that in emetine-untreated control MCF-10A cells; 2) the gene expression in control untreated MCF-10A cells had to be detected as present; 3) fold-change increase of mRNA level following emetine treatment in the transformed cells had to be at least threefold higher than that in the untransformed control cells.

One microgram of total RNA from emetine-treated transformed MCF-10A cells was reverse transcribed using the SuperScript II protocol (Invitrogen, Carlsbad, CA). Overlapping PCR primer sets were used to generate products spanning entire open-reading frames for candidate genes. Primers for sequencing analysis were designed using Primer3 software available online 23. To sequence the p53 gene from genomic DNA, the sequences of p53 gene exons with the flanking introns were retrieved from the genomic databases, and the primers for PCR amplification of genomic DNA fragments of p53 gene were designed using Primer3 software. These are available upon request. The PCR products were gel-purified and sequenced using Applied Biosystems' PRISM 3100 Genetic Analyzer at the DNA Sequencing section of the Biopolymer Facility at RPCI.

Whole-cell extracts were prepared in a RIPA buffer, and protein samples were denatured, separated by SDS/PAGE gel, and transferred onto a Hybond-P nitrocellulose membrane (Amersham Pharmacia) by using the MiniPROTEAN 3 apparatus (Bio-Rad). Membranes were rinsed and blocked with 0.05% Tween 20 (Sigma) and 1% skimmed milk in PBS for one hour with three changes of this blocking buffer every 20 minutes at room temperature. Membranes were then incubated with DO1 antibody conjugated with horseredish peroxidase (Santa Cruz Biotechnology, Santa Cruz) in a blocking buffer. After washing the membranes in blocking buffer for one hour with a change of the buffer every 15 minutes at room temperature, membranes were visualized using ECL plus the Western blotting detection system (Amersham Pharmacia). The anti-actin monoclonal antibody (Sigma) was hybridized to membranes to serve as a measure of loading control.

After rinsing with a phosphate-buffered saline solution, the cells were trypsinized from culture dishes, rinsed with PBS, fixed in 70% ice-cold ethanol, and stained with 10 μg/ml propidium iodide (PI) for 30 minutes. Flow cytometry analysis was performed on the Coulter FACS analyzer equipped with an air-cooled argon laser, producing 488 nm light in which fluorescence light-emission is collected with a 620BP filter. An extended analysis of the DNA content and calculations of the percentage of cells in each phase of the cell-cycle were performed on ModFit Lt software (Verity Software House, Topsham, ME).

Cell cultures usually received treatments with Colcemid (0.03–0.06 μg/ml). However, nocodazole (0.5 μg/ml) was found to give the best results with some cells. Chromosome spreads were prepared using air-drying methods. After sequential digestion with RNase and pepsin according to the procedure recommended by Applied Spectral Imaging, Inc. (ASI: Vista, CA 92081), the chromosomal DNA on the slides was denatured in 70% formamide and then hybridized with a cocktail of human SKY paint probes tagged with various nucleotide analogues (i.e., a mixture of individual chromosome DNA prepared by flow-sorting and PCR amplification)24. Thirty to 50 mitoses were chosen at random, and the images developed by combinations of five different flouorophores, such as Rhodamine, Texas-Red, Cy5, FITC and Cy5.5, were captured with a Spectral cube and Interferometer module installed on a Nikon microscope. Spectra-Karyotypes were carried out using SKY View software (Version 1.62).

The preparation of Roswell Park Cancer Institute (RPCI) custom CGH arrays using RPCI-11 BACs has been described previously2526. Briefly, the RPCI array contains ~6,000 RPCI-11 BAC clones that provide an average resolution across the genome of 420 kilobases. BACs were printed in duplicate on amino-silanated glass slides (Schott Nexterion, type A) using a MicroGrid ll TAS arrayer (Apogent Discoveries, Hudson, NH) to generate an array of ~12,000 elements. Genomic control MCF-10A DNA and the transformed MCF10AαL DNA were fluorescently labeled by random priming and hybridized, as described 25. The hybridized slides were scanned using a Genepix 4200A scanner (Axon, Inc., Union City, CA) to generate high-resolution (10 μm) images for both Cy3 and Cy5 channels, and image analysis was performed using ImaGene (version 4.1) software (BioDiscovery, Inc., El Segundo, CA).