This study has several important implications for analysis of tumor cell growth and drug responses. First, using a novel fibroblast-derived 3D matrix system, we segregated tumor cell lines into different classes based on the magnitude of their responsiveness to matrix and the degree to which matrix induces proliferative versus morphological changes in growth. The morphological responses we documented included increased acquisition of a spindle morphology by some cells, and rounded or amoeboid-like morphology by others. In vivo
, as tumors become invasive, some neoplastic cells leave the tumor mass and undergo epithelial to mesenchymal transition (EMT), in which single cells (as opposed to tumor aggregates) assume a spindle shape or a mesenchymal fibroblast-like appearance. Conversely, metalloproteinase inhibitors, loss of p53, inhibition of the E3 ubiquitin ligase Smurf1, and other conditions which activate RhoA, trigger a mesenchymal to amoeboid transition (MAT), wherein tumor cells become rounded, thus modifying their survival and invasive behaviors (Condeelis and Segall, 2003
; Gadea et al., 2007
; Sahai et al., 2007
; Sahai and Marshall, 2003
). The 3D culture system we described here is well designed to analyze the behavior of isolated single cells that have undergone EMT or MAT and are invading through mesenchymal stroma.
As context, other recent studies have also explored 3D substrates as screening platforms to study tumor behavior. For example, Kenny and co-workers have analyzed spheroids, acini and cell aggregates, which mimic tumor masses that interact with basement membrane-like 3D substrates (Kenny et al., 2007
). These 3D laminin gel-based cultures induced four distinct tumor-mass morphologies, specific to the 3D environment. Moreover, these cell-aggregate morphologies were each associated with specific gene expression profiles, which correlated with increased invasive tumor cell potential (Kenny et al., 2007
). Together, platforms addressing single cell and aggregate cell growth in 3D have the potential to parse the different signaling conditions induced by tumor microenvironment.
We suggest that characterizing the morphological changes in isolated tumor cells on fibroblast-derived 3D matrices may allow rapid testing of how a specific tumor cell type will react to the microenvironment, thus facilitating the development of strategies for its specific treatment. For example, in our study PANC-1 adopts a rounded, amoeboid-like, morphology on 3D matrices. Amoeboid-like motility is not dependent on the activity of matrix metalloproteinases, but instead requires the action of RhoA GTPase (Friedl, 2004
; Friedl and Wolf, 2003
; Sahai and Marshall, 2003
). This matrix-induced specific morphological behavior may predict that MMP inhibitors will not affect invasion of PANC-1 cells through tumor stroma, but that treatment of these cells with Rho inhibitors may successfully block their migration.
Second and importantly, our data suggest that the tendency of a cell to undergo morphological change (rounded or spindle) following plating onto stroma-like matrix may predict a tendency of matrix to influence drug response by that cell line (compare data with PANC-1, HS 578T and PA-1 versus COLO 205 and SW620). The profile of behavior of PA-1 indicates that the fact that 3D matrix can influence cell morphology does not predict whether change in drug response will be positive or negative; rather, this needs to be assessed for individual cell lines. Why 3D matrix induced response to taxol but not to other drugs remains to be determined. Nevertheless, it is intriguing that among those tested this agent alone targets the cytoskeleton. In this context, it has been reported that in breast cancer cells seeded onto isolated ECM proteins (e.g., fibronectin and type I collagen), β1-integrin is essential to developing resistance to taxol-induced apoptosis (Aoudjit and Vuori, 2001
). It is also interesting to note that taxol has been reported to induce focal adhesion disorganization, with the associated taxol-dependent cell death rescued by integrin linked kinase (Deschesnes et al., 2007
). Integrin overexpression frequently depresses levels of E-cadherin expression (Thiery and Sleeman, 2006
); it has also been reported that loss of E-cadherin causes resistance to taxol (Ferreira et al., 2005
). We hypothesize that a combination of specific ECM proteins, presented to tumor cells in association with a 3D architecture and β1-integrins, particularly enhances taxol responses by altering adhesion structure organization and signaling.
Other groups have also identified 3D matrix-dependent regulation of drug responses, using different culture systems. For example, Fischbach and colleagues used engineered polymeric scaffolds with carcinoma cells as 3D human tumor models, and have recently shown these synthetic matrices to effectively mimic many aspects of in vivo
tumor behavior including some resistance to drugs (Fischbach et al., 2007
). As another example, Weaver et. al. have demonstrated that Matrigel generates a β4-integrin-regulated cell polarization, which induces drug resistance in tumor cells derived from epithelial cells (Weaver et al., 2002
). This significant study also showed that for some drugs, disrupting hemidesmosome formation perturbed matrix-directed tissue polarization, and subsequently inhibited NFκB activation, thus promoting apoptosis. We have also tested integrin-dependent NFκB activities in the presence of taxol using the 3D culture system described here. Our results indicate that although NFκB activity was induced by 3D fibroblast matrix, these changes were not reversed by integrin inhibition (data not shown), indicating activation of NFκB does not influence drug resistance in this system.
We propose that a 3D matrix system prepared from stable cell lines such as NIH3T3 can provide a convenient and highly physiological assay system to measure and correlate cell morphology, proliferation, tendency to apoptosis, and drug response in high throughput formats. Further, we have recently described a “progressive” stromal system in which we compared the properties of 3D matrices prepared from normal, carcinogen-primed, and tumor-associated fibroblasts (Amatangelo et al., 2005
; Castelló-Cros and Cukierman, 2008
). The data presented in the present study indicate that most of the cell lines evaluated proliferated better on plastic or 2D matrix substrates than on the in vivo
-like 3D matrix. Notably, the 3D matrix used in this study represents an early stromagenic stage in tumor development, which may still impose the restrictive natural barrier of the host’s stroma (Amatangelo et al., 2005
). Based on the results described here, it is likely that these staged, progressive matrices will each differentially interact with tumor cell lines, allowing further parsing of the tumor-microenvironment dialog. We believe that in the future, specific fibroblast-derived matrices representing different stages of tumor progression, from different tumor types, would enhance the screening for drug response in discrete classes of tumor cell lines, i.e. pairing tumor-associated fibroblast-derived 3D matrices with tissue-specific neoplastic cells should better predict matrix-induced drug responses for the cancer in question (e.g., breast or pancreas). Carefully selecting cells, drugs and microenvironmental settings is important, as many drugs that are potent in in vitro
biochemical assays and/or in conventional assays performed on cultured cells fail at the stage of in vivo
assays tested in animals. Extended work in this area may help develop more robust in vitro
drug screening regimens that will better mimic in vivo
settings, thus reducing the failure rate of drug leads in pre-clinical development.