Introduction: Despite success in hematological malignancies, immunotherapies are less effective in treating solid tumors. Solid tumors comprise a dense extracellular matrix (ECM) that mediates soluble signaling and is a physical barrier for immune cell infiltration and cancer cell recognition. Commonly, studying immune cell-cancer cell interactions in vitro is limited to 2D systems, which lack a relevant and controllable ECM; as a result, the mechanisms employed by the ECM to avoid immune detection remain largely unknown. By contrast, in vivo murine models are capable of better replicating the tumor microenvironment but limit the mechanistic understanding of immune cell inhibition. A 3D in vitro model that replicates the tumor microenvironment in a controllable manner is critical for understanding the mechanisms involved in ECM-related immunosuppression in solid tumors. The Sharma Lab has developed biochemical and mechanical properties of the tumor microenvironment [1]. Here, we characterize the extent to which our in vitro lung tumor model recapitulates the gene expression profile of other conventional tumor systems.
Materials and
Methods: RNA was isolated from human A549 luciferase-expressing cancer cells which were grown in systems of increasing complexity. Briefly, cancer cells were either cultured on 2D tissue-treated plastic, encapsulated in hydrogels by photo-polymerizing a mixture of PEG-diacrylate, PEG functionalized with MMP degradable sites, and 5 mM of PEG functionalized with integrin binding sites, or subcutaneously injected into the flank of a male NOD-SCID mouse. Tumors were grown in vitro for 1 week and in vivo for 3 weeks with their growth being monitored either using calipers. Library preparation was done using Illumina RNA-Seq Libraries PolyA directional/stranded and samples were subjected to paired-end sequencing (50 million reads/sample) using the Illumina Novaseq at the University of Florida Health Cancer Center. The data was filtered using NOISeq and a principal component analysis (PCA) and a pair plot were generated to visualize the pairwise relationship between tumor culture systems.
Results, Conclusions, and Discussions: Exploratory data analysis reveals similarities in gene expression data between tumor culture systems. Over 91% of the data variance was explained by the first two principal components in PCA (Figure 1). The first principal component arranged the samples with the 3D PEG-based A549 hydrogel models and the subcutaneous tumors on one end, and the 2D cell culture on the other. Interestingly, the second principal component clusters 2D samples and subcutaneous tumors on one end and 3D hydrogel models on the other. This suggests that the 3D hydrogel tumor model mimics the subcutaneous tumor more than 2D, but may be incompletely replicating an important aspect of data variance that 2D culture captures better in principal component 2. While all culture conditions had relatively high Spearman correlation coefficients, 3D tumor models and subcutaneous tumors had the highest correlation (0.87) and 2D cell culture and subcutaneous tumors had the lowest correlation (0.76). This initial analysis demonstrates that the gene expression profile of our in vitro lung tumor more closely correlates with in vivo models than traditional 2D cell culture systems. This suggests that the 3D PEG-based hydrogels offer a platform to bridge the gap between 2D cell culture and in vivo studies for lung cancer research. Moreover, after just 1 week in culture, our in vitro tumor model replicated aspects of subcutaneous tumors grown for 3 weeks, offering a time-saving alternative to animal studies. Future work will investigate the differential gene expression profiles and the downstream molecular pathways that our in vitro lung tumor model is replicating. The described study provides the foundation to understand what aspects of solid tumors are necessary to be engineered into in vitro systems to model the molecular mechanisms of tumor progression in vivo.
Acknowledgements (Optional): We would like to thank Kalyanee Shirlekar and Jason Orr Brandt at the Department of Biostatistics, UF for performing quality control and filtering on the RNA sequencing data.
