Response to trastuzumab, erlotinib, and bevacizumab, alone and in combination, is correlated with the level of human epidermal growth factor receptor-2 expression in human breast cancer cell lines.

Human epidermal growth factor receptor-2 (HER2) and epidermal growth factor receptor (EGFR) heterodimerize to activate mitogenic signaling pathways. We have shown previously, using MCF7 subcloned cell lines with graded levels of HER2 expression, that responsiveness to trastuzumab and AG1478 (an anti-EGFR agent), varied directly with levels of HER2 expression. HER2 and EGFR up-regulate vascular endothelial growth factor (VEGF), a growth factor that promotes angiogenesis and participates in autocrine growth-stimulatory pathways that might be active in vitro. Here, we show that trastuzumab, erlotinib, and bevacizumab, individually and in combination, inhibit cell proliferation in a panel of unrelated human breast cancer cell lines, in proportion to their levels of HER2 expression. The combination of all three drugs provided a greater suppression of growth than any single drug or two-drug combination in the high HER2-expressing cell lines (P < 0.001). Combination index analysis suggested that the effects of these drugs in combination were additive. The pretreatment net level of VEGF production in each cell line was correlated with the level of HER2 expression (r = 0.883, P = 0.016). Trastuzumab and erlotinib each reduced total net VEGF production in all cell lines. Multiparameter flow cytometry studies indicated that erlotinib alone and the triple drug combination produced a prolonged but reversible blockade of cells in G1, but did not increase apoptosis substantially. These studies suggest that the effects of two and three-drug combinations of trastuzumab, erlotinib, and bevacizumab might offer potential therapeutic advantages in HER2-overexpressing breast cancers, although these effects are of low magnitude, and are likely to be transient.

A suitable method for identifying cell aggregates in laser scanning cytometry listmode data for analyzing disaggregated cell suspensions obtained from human cancers.

BACKGROUND: The presence of cell aggregates in cell suspensions obtained from human solid tumors can interfere with the measurement of cell DNA content of cell singlets, and can confound multiparameter analysis of other measurements on the same cells. Flow cytometric corrections for cell aggregates based on signal pulse shape have not proven to be reliable. Mathematical models have been developed to correct for cell aggregates in binned DNA histogram data, but they are not suitable for the correction of correlated non-DNA measurements obtained on the same cells. METHODS: A total of 21 samples representing a variety of normal and malignant human cell types, including normal lymphocytes, normal sputum, human breast cancer cell lines, and mechanically disaggregated cell suspensions from primary breast cancers and nonsmall cell lung cancers, were studied by laser scanning cytometry (LSC) using the CompuCyte laser scanning cytometer (Cambridge, MA). Nuclear area, nuclear perimeter, and an LSC-based cell texture parameter were measured on approximately 400 cells in each sample, using an air-cooled, violet laser emitting at a wavelength of 405 nm for DAPI excitation, and each cell was classified as a singlet or aggregate by its appearance under direct observation. A "saddle function" provided by CompuCyte was used, together with an algorithm based on the measurements of nuclear area, perimeter, and cell texture (the APT algorithm), to identify cell aggregates and exclude them from the listmode data file. RESULTS: Proportions of cell aggregates in the uncorrected samples ranging from 6 to 56% (mean, 20%) were reduced to proportions ranging from 0 to 7% (mean, 2.4%) after correction. The discriminant function was "tuned" to maintain both average cell singlet purity and average cell singlet yield at >70% over a broad range of cell DNA contents. CONCLUSIONS: A combined approach to cell aggregate detection, which utilizes both the saddle function and the APT algorithm, produces list mode data files that exclude >80% of cell aggregates from samples of disaggregated cell suspensions of human tumors and other sources of clinical material. Such data files are suitable for multiparameter analysis.