A Functionally Robust Phenotypic Screen that Identifies Drug Resistance-associated Genes Using 3D Cell Culture

Background

Despite significant progress in therapeutic strategies, almost 30% of breast cancer patients develop chemotherapy resistance following initial treatment leading to relapse (Brewster et al., 2008; Holohan et al., 2013). Genes or pathways that are responsible for chemoresistance have been identified through traditional screening approaches including gain-of-function and loss-of-function tools (Grimm, 2004; Arnoldo et al., 2014). More recently, a high-throughput screening technique of CRISPR/Cas9-based system has been used to identify molecular targets and their related pathways (Hu and Zhang, 2016; Luo, 2016). Nevertheless, most assays fail to provide a mechanism of resistance and alternative possibilities of treatment.

In a tissue, dynamic and reciprocal signaling between cells and their local microenvironment is required for maintenance of tissue architecture and homeostasis (Bissell and Hines, 2011; Furuta et al., 2018). In particular, biochemical and biophysical cues from the local microenvironment are recognized by cell surface receptors such as growth factor receptors and integrins that initiate a signaling cascade transduced into the nucleus to change gene expression (Bissell et al., 1999; Ghajar and Bissell, 2008). In addition to mutations that drive gene expression involved in drug resistance, the microenvironment plays a significant role in multidrug resistance (Weaver et al., 2002; Correia and Bissell, 2012). Therefore, a physiologically relevant assay system–one that permits screening while recapitulating the microenvironment in vivo–is critical for identifying new targets for breast cancer therapy and for elucidating mechanisms underlying the tissue specificity of drug resistance.

HMT-3522 progression series of human mammary epithelial cells was derived from a reduction mammoplasty of a woman with benign fibrocystic breast disease (Petersen et al., 1990; Briand et al., 1996). The epithelial tissue was cultured in a chemically defined culture medium and the phenotypically normal S1 cells were established (Briand et al., 1987). After over 100 passages of S1 cells in epidermal growth factor-free medium, growth factor-independent and tumorigenic T4-2 cells were obtained (Briand et al., 1996). Non-malignant S1 cells in 3D culture of Laminin-rich extracellular matrix (lrECM) recapitulate an organized polar unit of breast structure referred to as acinus, in culture we refer to them as mammary organoids (Petersen et al., 1992; Weaver et al., 1995; Lee et al., 2007). In the same medium and culture conditions, malignant T4-2 cells form disorganized colonies in 3D culture and tumors in mice. We have shown that malignant T4-2 cells in the presence of lrECM become quiescent and can be reverted to a structure partially resembling non-malignant counterparts, despite still containing the cancer genome (Figure 1). This “phenotypic reversion” is achieved by adjusting extracellular and intracellular signaling pathways, such as inhibitors of receptor tyrosine kinase or the humanized blocking antibody hampering oncogenic signaling which otherwise leads cells to lose architecture and grow out of control (Weaver et al., 1997; Wang et al., 1998; Weaver and Bissell, 1999; Muschler et al., 2002; Wang et al., 2002; Kenny and Bissell, 2007; Itoh et al., 2007; Beliveau et al., 2010; Onodera et al., 2014). Other breast cancer cell lines, such as metastatic MDA-MB-231, can be reverted by treatment with a combination of inhibitors of integrin beta1, PI3K, or MAPK (Wang et al., 2002).

Using the three conditions described above, non-malignant, malignant and reverted malignant cells in our 3D cell culture models, we demonstrated that the efficacy of the chemotherapeutic response was strictly dependent on cell and tissue polarity and tissue architecture (Weaver et al., 2002). By combining the ability to “phenotypically” revert the malignant cells with a gain-of-function genetic screen, we developed a phenotype-based functional genetic screen that allowed the identification of novel genes associated with EGFR -TKI resistance (Lee et al., 2012).

In this protocol, viral cDNA library generated from malignant breast epithelial cells is transduced into the same malignant cells under 2D (tissue culture on plastic) followed by subsequent transfer to treatment with EGFR-TKI (AG1478) or PI3K inhibitor (LY294002), on 3D lrECM culture in the presence of antibiotics for selection of integrated genes. The 3D cultures are maintained for 5-7 days to allow the malignant T4-2 cells to undergo phenotypic reversion in the presence of either inhibitor. Then colonies that fail to revert and continuously grow are isolated from the cultures and subjected to sequencing for identifying the integrated cDNA that conferred resistance to EGFR-TKI or PI3K inhibitor (Figure 2).

Using this technique, we identified five candidates (Table 2) and further characterized one of the genes that conferred EGFR-TKI resistance in breast cancer cells when overexpressed. This gene was identified as FAM83A, one of the eight members of a previously unidentified oncogene family. We observed that FAM83A expression was increased when cells were treated with tyrosine kinase inhibitor, lapatinib; the cells that survived became resistant to the treatment. FAM83A-associated molecular mechanism of drug resistance serves as an example of the utility of this screening method and provides a basis for predictive diagnosis and possible molecular targets for chemotherapy.

This assay can be applied for identifying novel therapeutic targets by modulation of gene expression such as systemic gain-of-function genetic screens with a cDNA library or loss-of-function genetic screens with shRNA (short hairpin RNA) or sgRNA (single guide RNA) libraries. Additionally, this assay is also a useful tool for evaluation of drug response and efficacy of pharmacological candidates in different cancer cell lines allowing us to predict therapeutic outcomes.


Figure 1. Fluorescence microscopy images of HMT 3522 progression series. A. Non-malignant S1 cells organized into acini structure resembling alveoli in the mammary gland in vivo. B. Malignant T4-2 cells formed large disorganized colonies. C. “Phenotypically reverted” malignant T4-2 cell showed basal polarity and growth arrest by modulation of signal pathways. All cells were grown in 3D culture of lrECM, and stained for α6 integrin (green), β-catenin (red), and nuclei (blue). Adapted from Figure 3 in Lee et al. (2007).


Figure 2. Workflow for phenotype-based genetic screen. (1-2) Amphotropic retroviral cDNA library was generated and transduced into HMT3522 malignant T4-2 cells. (3) The cDNA expressing T4-2 cells were plated on 3D lrECM culture and treated with EGFR-TKI AG1478. Colonies that fail to revert (circled in blue) were isolated from the culture using a flame-drawn glass pipette under a phase-contrast microscope in a Biosafety cabinet. (4) Each isolated colony was transferred to one well of a 48-well tissue culture plate and subsequently expanded to a larger multi-well plate until the cell number reached one million cells. (5) Cells were re-plated on 3D lrECM culture and treated with EGFR-TKI AG1478, and the isolation procedure was repeated as described in (3). (6) Genomic DNA was isolated from selected clones grown in 2D and cDNA inserts were recovered by PCR using the flanking sequences of retroviral vector used for cloning. (7) The amplified PCR products were analyzed for sequencing to determine the identity of inserted cDNA.