Description
Cancer tissue-like structures were developed by using established human tumor cell lines in perfusion-based bioreactor systems. In colorectal cancer (CRC) cell lines, perfusion allowed more homogeneous scaffold seeding than tri-dimensional (3D) static cultures and significantly (13.7 fold, p<0.0001) higher proliferation. Resulting tissues exhibited morphology and phenotypes similar to xenografts generated in immunodeficient mice. Whole transcriptome analysis of 2D, 3D static and 3D perfusion cultures revealed the highest correlation between xenografts and 3D perfusion cultures (r=0.985). Clinically relevant concentrations of 5-FU, used in neo- and adjuvant CRC treatment, had no effect on numbers of HT-29 CRC cells cultured in 3D perfusion or xenografts, as compared with a 55.8% reduction in 2D cultures. Treatment induced apoptosis in 2D cultures, but only “nucleolar stress” in perfused cells and xenografts, consistent with partial responsiveness. In 3D perfusion cultures BCL-2, TRAF1, and FLIP gene expression was marginally affected, as compared with significant down-regulation in 2D cell cultures. Accordingly, ABT-199 BCL-2 inhibitor, induced cytostatic effects in 3D perfusion but not in 2D cell cultures (p=0.003). Tumor cells from partially responsive (Dworak 2) patients undergoing neo-adjuvant treatment, typically (10/11) expressed BCL-2, as compared with 0/3 highly (Dworak 3-4) responsive and 4/15 fully resistant CRC (Dworak 0/1, p=0.03), closely matching 3D perfusion cultures data. These results indicate that 3D perfusion cultures efficiently mimic phenotypic and functional features observed in xenografts and clinical specimens. These models may be of critical translational relevance to address fundamental human tumor cell biology issues and to develop predictive pre-clinical tests of novel compounds. Overall design: Expression profiles of colorectal cancer cell lines cultured in 2D, 3D static, 3D perfusion or growing as xenografts were generated by deep sequencing, in triplicates, using Illumina HiSeq2000.