Erratum: Genetic variability in a frozen batch of MCF-7 cells invisible in routine authentication affecting cell function

Scientific Reports 6: Article number: 28994; published online: 26 July 2016; updated: 14 September 2016

character of MCF-7 7 . Later, these findings were confirmed by others [8][9][10] and it is now recognized that MCF-7 is heterogeneous with respect to both the expression of hormone receptors and to the utilization of the signaling pathways linked to these receptors, differences that result in phenotypic heterogeneity 11 . Sub-clones vary in estrogen and progesterone receptor expression, as well as epidermal growth factor. However, genotyping analysis shows that all sub-clones are related to the parental MCF-7 cell line 12 . Nonetheless, even though questions have been raised about the reproducibility of results with MCF-7 cells 13 , many laboratories assume that by using cells obtained from a cell bank, standardizing protocols, limiting the number of passages, and employing SNP or STR cell authentication techniques would ensure that "their sub-clone" will behave with sufficient stability and reproducibility.
Our experience is that this may not be necessarily sufficient. Based on data from our Human Toxome Project 14,15 , we demonstrate by various techniques that there can be marked cellular and phenotypic heterogeneity in a single batch of cells from a cell bank that are invisible with the usual STR cell authentication protocols, and that this heterogeneity has serious consequences for reproducibility and primary outcomes of experiments.

Results
As part of our "Mapping the Human Toxome" project, two laboratories (Brown University [BU] and Johns Hopkins University [JHU]) used MCF-7 cells from the same ATCC lot (lot number 59388743, passage 147) combined with strict adherence to standards for validation (standard operations protocols, formal training, and transfer) for cell culture and analytic methods 14 including the recommendations for Good Cell Culture Practice 16 .
In a first step, this work included expansion of the cells from the original ATCC vials using three passages for BU and eight passages for JHU, respectively to create vials for use in experiments, each of which were then passaged up to 10 times after which another vial was thawed for continuing experiments. Recommended genomic typing of short tandem repeat markers (STR) showed that all MCF-7 cell markers were the same lengths as provided by the reference ATCC genotyping panel for all 9 typed markers (Table 1). Nonetheless, significant differences were observed between the two laboratories in terms of phenotype, gene expression patterns, metabolomics, and (most crucially) sensitivity to estradiol-driven proliferation. To exclude any possible inter-laboratory and/or inter-operator effects, the JHU cells were shipped to BU to verify the results. The results of the morphological, phenotypical and gene expression differences that have been performed in one laboratory (BU) by one individual are given in Fig. 1: Morphologic assessment of the MCF-7 cells showed that BU cells grow in large aggregations while JHU cells grow flat, with cobblestone morphology (Fig. 1A). Following 72 hours of exposure to estradiol (E2), BU MCF-7 cells displayed significant increases in proliferation (cell count) at concentrations of 0.1, 1 and 10 nM, while JHU cells did not have a significant change in cell count (Fig. 1B, left). Exposure to the estrogen receptor alpha agonist propyl pyrazole triol (PPT) for 72 hours resulted in a significant increase in proliferation at concentrations of 0.1, 1.0 and 10 nM in the BU subline, while JHU cells did not exhibit significant changes (right). Gene expression analysis using quantitative PCR of the estrogen receptor target progesterone receptor (PgR) following 6 hours of exposure to E2 indicated that BU MCF-7 cells are responsive to low levels of estrogen (Fig. 1C). To exclude possible laboratory/technician influences, cells were exchanged between the laboratories, but the differences persisted (data not shown).
To verify whether these striking differences can be generalized, two omics technologies, i.e. gene expression microarrays and untargeted LC-MS metabolomics, have been performed as well. The experiments included different concentrations of E2 and PPT at various time points. Principal Component analysis (PCA) of the obtained 1048 features from the untargeted metabolomics experiments indicates that the two laboratories responded differentially to the E2 treatment (Fig. 2), which was more pronounced than any agonist effect. The QC samples (i.e. the pool of all samples in each experiment per lab) group together for each experiment but are well separated in the PCA. This demonstrates good technical reproducibility, while a clear distinction between the two sublines is observed. Moreover, all experimental conditions from an individual experiment group together, but separately  D5S818  11,12  11,12  11,12  11,12  11,12   D13S317  11  11  11  11  11   D7S820  8,9  8,9  8,9  8,9  8,9   D16S539  11,12  11,12  11,12  11,12  11,12   vWA  14,15  14,15  14,15  14,15  14,15   TH01  6  6  6  6  6 AMEL X X X X X TPOX 9,12 9,12 9,12 9,12 9,12 CSF1PO 10 10 10 10 10 from the other subline. Very similar effects can be seen in the PCA of the gene expression microarray data which indicate that the transcriptional state of the cells is already different in the control untreated cells and this difference persists in treated samples (Fig. 3). This can also be seen in a heat map and unsupervised hierarchical clustering from the same experiment of 84 genes selected from the literature 17 covering estrogen receptor signaling, breast and ductal morphogenesis, cellular growth and differentiation, proliferation, tumor progression and epithelial to mesenchymal transition (Fig. 4). These results suggested that MCF-7 cells of the same batch from a cell bank include subpopulations with different genomic backgrounds, and that this could explain the phenotypic differences. To explore this possibility, (C) Gene expression analysis of the estrogen receptor target progesterone receptor (PgR) following 6 hours of exposure to E2 indicated that BU MCF-7 cells are responsive to low levels of estrogen. All experiments have been performed in one laboratory (BU) by one individual to exclude any possible inter-laboratory and/or inter-operator effects on reproducibility. * p < 0.05, * * p < 0.01, * * * p < 0.001, * * * * p < 0.0001.
we looked for possible genomic differences using genome-wide Comparative Genomic Hybridization (CGH) in the two MCF-7 subpopulations. We also included genomic DNA directly prepared from second untouched never thawed ATCC vials with the same lot number that had been continuously stored since 2011 in liquid nitrogen at both laboratories (ATCC BU and ATCC JHU). Technical replicates of MCF-7 genomic DNA from MCF-7 cells being used for experiments after a few passages hybridized against human Caucasian female reference DNA showed only small differences of short length, which are simply expected by statistical chance because of the > 100,000 involved CGH probes (Fig. 5). However, the direct comparative hybridization of DNA from genomic DNA directly prepared from two original ATCC vials and the cell cultures derived from them provided surprising results: MCF-7 is not just a mixture of cells with heterogeneous genetic backgrounds from the same donor so they will have similar STR patterns, -in fact, almost the same significant genomic differences on chromosomes 1, 3, 4, 5, 7, 8, 9, 10, 13, 15, 20 and X as seen after a few passages in cell culture can already be detected from genomic DNA directly prepared from the two original ATCC vials from the same lot number (Fig. 6). This observation  indicates that the original vials from the same ATCC lot already showed most of the genetic differences, and the genetic heterogeneity introduced by passaging at the two laboratories was minimal.

Discussion
As our data demonstrate, seemingly similar cells from the same ATCC batch, that show the same STR genotypes for cell line authentication and applied current recommendations on Good Cell Culture Practice, produced starkly different results for key outcomes -proliferation, expression of estrogen responsive genes, estrogen dose-response curve, metabolomics and transcriptomics. Geno-or karyotyping profiling just fulfills the purpose of authentication testing, which is to detect significant cross-contamination and cell line misidentification as defined by The International Cell Line Authentication Committee (ICLAC): "The aim of authentication is to confirm or verify the identity of a cell line, ensuring that it is derived from the correct species and donor" 18 (page 4) . In other words, the current standard ways that are used to "ensure" cell authentication do not preclude the possibility of genetic heterogeneity at a level that could fundamentally compromise reproducibility within and between laboratories and heterogeneity needs to be examined using additional characterization methods. Even simple outcomes, such as estradiol dose-response, can be divergent between two laboratories using what appear to be the "same" cells and the same protocol. Similar results for MCF-7 have already been demonstrated in the past by other studies e.g. refs 9 and 19, however, none of the reported studies showed that significant genomic differences can already be present within the same lot from a cell bank. Since the genomic DNA was directly prepared from the original ATCC vials by one operator at the same time point possible thawing, cell-culture, inter-laboratory and/ or inter-operator effects can be excluded. Given the widespread use of cancer-derived cell lines not only for basic science but also for drug development and regulatory decision-taking, ensuring that such cells are adequately standardized represents a challenge going forward. These findings underscore the need for additional quality assurance in Good Cell Culture Practice and cell characterization, especially using CGH or deep sequencing, to reveal possible genomic heterogeneity and drifts within cell lines, which are not detected by geno-or karyotyping and other current recommendations. CGH compared to deep sequencing offers advantages as to costs and bioinformatics efforts needed to analyze such differences.
In this respect, MCF-7 are likely not unique. For example, studies with T47D -the second most commonly used cell line as an in vitro breast cancer model -as well as others 20 suggest similar genotypic and phenotypic heterogeneity [21][22][23] . However, these have not been shown for cells from the same batch without further cultivation. Cell model heterogeneity could be an advantage under some circumstances, and MCF-7 has been proposed as "an interesting model for genetic evolution of breast tumors" 10 . But such heterogeneity can most certainly be a curse when it comes to other common applications of such cells, as it not only weakens "the direct relevance of such cultures as models of human cancer" 8 but also "makes inter-laboratory reproduction of experimental findings difficult" 24 . It is noteworthy that, in work parallel to our own, the MCF-7 estrogen disruptor assay failed international validation by the US National Toxicology Program Interagency Center for the Evaluation of Alternative  Toxicological Methods (NICEATM), mainly because of concerns about inter-laboratory reproducibility 25 . These issues are of profound concern when an in vitro cell model is used for investigations designed to inform patient care or hazard assessment. The combination of omics technologies does not overcome the limitations of tumor cell lines but makes their problems more evident.

Materials and Methods
Experimental Design. A schematic overview of the whole process from cell acquisition to experiments performed is given in Fig. 7.  point, and plated at 5,000 cells/well for the 72 hour time point. Cells were allowed to grow in 10% complete medium for 72 hours, and then placed in 5% DCC treatment media for 48 hours. Cells were then exposed to propyl pyrazole triol (PPT, Tocris, Minneapolis, MN, USA, no. 1426), 17β -estradiol or vehicle control DMSO for 24, 48 or 72 hours. Cells were fixed in formalin, washed with 1× PBS and permeabilized with 0.25% Triton X-100 for 15 minutes. Samples were stained for 20 minutes using rhodamine phalloidin (Molecular Probes, Life Technologies, no. R415, 1:2000 dilution) and DAPI (Molecular Probes, Life Technologies, no. D1306, 1:2000 dilution). Cells were imaged using the 20× water immersion objective on an Opera Phenix High-Content Imaging System (Perkin Elmer). Morphologic analysis was performed using Harmony Software (V 4.1, Perkin Elmer) to assess nuclear count, cell area, cell roundness, nuclear area and nuclear roundness. Border cells were excluded. Data was exported into GraphPad Prism (v5.01, GraphPad Software) and two-way analysis of variance and post-hoc test was performed.
RNA Isolation and Quantitative Real-Time PCR. Cells were scraped into TRI Reagent (Sigma Aldrich, no. T9424) and stored at − 80 °C until RNA isolation and qPCR analysis. RNA was isolated using the RNEasy Mini Kit (Qiagen) per manufacturer's instructions, and cDNA was made using the Superscript II First Strand Synthesis System per manufacturer's instructions. qRT-PCR was performed using primers for progesterone receptor (PGR) and ribosomal protein, large P0 (RPLP0) described previously 17 . Changes in mRNA were determined using the Δ Δ -C T method, data was plotted in GraphPad Prism Software and two-way ANOVA used to determine statistical significance.
RNA isolation and microarray experimental design. Total RNA from MCF7 cells was extracted using TRizol Reagent according to manufacturer's instruction, and purified using RNeasy Mini Kit (Qiagen). Purified RNA was quantified by using NanoDrop ND-1000 spectrophotometer (Thermo Scientific) and the quality of RNA was assessed by using Agilent Bioanalyzer (Agilent).
100 ng of total RNA from treated and control cells were converted into cDNA and then into labeled cRNA using Agilent LowInput QuickAmp Labeling Kit (Agilent). The resulting cRNA was labeled with Cy3. Labeled cRNAs were then purified, and RNA concentration and dye incorporation were measured using NanoDrop ND-1000 spectrophotometer. Hybridization to Agilent SurePrint G3 human whole genome 8 × 60 K microarray (Agilent) was conducted following manufacture's protocol. Microarrays were scanned with an Agilent DNA microarray scanner. Feature Extraction (11.5.1.1 version, Agilent) was used to calculate the signal intensity and ratios. After deleting non-detected probes and quantile-normalization, Principal Component Analyses have been performed with activated mean centering and scaling option (GeneSpring V13.1, Agilent).
Untargeted metabolomics analysis. The cell culture media was removed by gentle vacuum suction and the cells were washed two times with 1 mL of pre-warmed PBS. Any residue of PBS was removed from the wells. A solution of 700 uL dry-ice cold 80:20 (v/v) methanol/water was immediately added, and the cells were scraped and collected in a 1.5 ml Eppendorf tube. The wells were washed again with an additional 700 uL solution of methanol/water and this solution was combined with the previous one. The solution was vortexed for 1 min and then stored at − 80 °C for 2 h to allow for protein precipitation. For metabolite extraction, tubes were placed on dry ice for 15 min and centrifuged at 14 000 × g for 5 min at 4 °C. The supernatant was transferred to a new 1.5 ml tube and placed on dry ice. Then, 300 μl of 80:20 methanol/water was added to the pellet and a second extraction was performed. The combined supernatants were evaporated overnight to dryness at room temperature in a Speedvac concentrator (Savant, Thermo Fisher Scientific, Waltham, MA, USA). The dried samples were reconstituted with 60 μL of 60% methanol with 0.1% formic acid and clarified by centrifugation at 14000 × g for 5 min. The clarified samples were transferred to HPLC vials for LC-MS measurements.
Chromatographic separations were performed using an Agilent 1260 high-performance liquid chromatography system with a well-plate autosampler (Agilent, Santa Clara, CA, USA). For aqueous normal phase (ANP) separation, a Cogent Diamond Hydride (MicroSol, Eatontown, NJ, USA) column (150 × 2.1 mm i.d., 4 μm particle size, 100 Å pore size) was used for separation of metabolites. The LC parameters were as follows: autosampler temperature, 4 °C; injection volume, 5 μL; column temperature, 35 °C; and flow rate, 0.4 mL/min. The solvents and optimized gradient conditions for LC were: Solvent A, 50% methanol/50% water/0.05% formic acid; Solvent B, 90% acetonitrile with 5 mM ammonium acetate; elution gradient: 0 min 100% B; 20-25 min 40% B; post-run time for equilibration, 10 min in 100% B. The LC system was coupled directly to the Q-TOF mass spectrometer. A 6520 accurate-mass Q-TOF LC-MS system (Agilent) equipped with a dual electrospray (ESI) ion source was operated in negative-ion mode for metabolic profiling. The optimized ESI Q-TOF parameters for MS experiments were: ion polarity, negative; gas temperature, 325 °C; drying gas, 10 l/min; nebulizer pressure, 45 psig; capillary voltage, 4000 V; fragmentor, 140 V; skimmer, 65 V; mass range, 70-1100 m/z; acquisition rate, 1.5 spectra/s; instrument state, extended dynamic range (1700 m/z, 2 GHz). Spectra were internally mass-calibrated in real-time by continuous infusion of a reference mass solution (standards with known mass at specific concentrations, which are introduced into the ion-source throughout the sample run to perform dynamic calibration) using an isocratic pump connected to a dual sprayer feeding into an electrospray ionization source. Data were acquired with MassHunter Acquisition software from Agilent.
For the data processing and chemometric analysis of the LC-MS untargeted data, the acquired raw data files (.d files) were first checked for quality in MassHunter Qualitative Analysis software (Agilent, version 6.0). Reproducibility of chromatograms was visually inspected by overlaying the Total Ion Chromatograms (TICs) of all samples. Data files that exhibit outlier peaks, i.e. replicates with very dissimilar chromatograms, were excluded for further processing. The raw data files were then converted to mzXML using ProteoWizard 3.0 26 . Raw LC-MS data were analyzed by the MZmine 2 software 27 for chromatogram deconvolution, peak detection and alignment. The putative identification was achieved by online searching for the accurate m/z values of the peaks against Scientific RepoRts | 6:28994 | DOI: 10.1038/srep28994 HMDB and KEGG databases 28,29 . Those peaks were manually inspected for the quality of the EIC (extracted ion chromatograms) and also for remaining duplicate compounds names. Principal Component Analysis has been performed with GeneSpring V13.1 with activated mean centering and scaling option.
Data were normalized and analyzed with CytoGenomics 3.0.4.1 (64 bit) and implemented default analysis method CGH v2. Briefly, after filtering for saturated and non-uniform probes, data were normalized by GC correction with a window size of 2 kb and Diploid Peak Centralization. Aberrations were detected by the Aberration Detection Method 2 (ADM-2) with a threshold of 10 and applied mosaic aberration filter with standard build-in parameters. For the graphical visualization a moving average smoothing function of + /− 50 probes has been applied and shown in the figures.
For the direct Comparative Genomic Hybridization analysis, the JHU cells have been defined as Cy3-labeled reference samples (positive log ratio defined comparative gain in BU versus JHU cells). For the indirect CGH sample, DNA was hybridized against human female reference DNA as supplied by Agilent SureTag Complete DNA labeling kit.
All CGH experiments have been performed in one laboratory (JHU) by one individual to exclude any possible inter-laboratory and/or inter-operator effects on reproducibility.
Short Tandem Repeat profiling. STR profiling was carried out following the ANSI/ATCC ASN-0002-2011 guidance, Authentication of Human Cell Lines: Standardization of STR Profiling. Briefly, a Promega GenePrint 10 Kit was used to polymerase chain (PCR) amplify eight short tandem repeat (STR) loci plus a gender-determining marker, Amelogenin. The PCR product was electrophoresed on an ABI Prism ® 3730xl Genetic Analyzer using an Internal Lane Standard 600 (Promega). Data was analyzed using GeneMapper version 4.0 software (Applied Biosystems). Appropriate positive and negative controls were used.
All STR experiments have been performed in one laboratory (JHU) by one individual to exclude any possible inter-laboratory and/or inter-operator effects on reproducibility.