Cancer Proliferation Gene Discovery through Functional Genomics

http://www.sciencemag.org/cgi/content/abstract/319/5863/620

Schlabach MR, Luo J, Solimini NL, Hu G, Xu Q, Li MZ, Zhao Z, Smogorzewska A, Sowa ME, Ang XL, Westbrook TF, Liang AC, Chang K, Hackett JA, Harper JW, Hannon GJ, Elledge SJ.

Science.2008 Apr 18;320(5874):316.

It is well-known that because of the wide variety of cellular mechanisms used by different cancers to achieve tumorigenesis and metastasis, identifying genes involved in each of these mechanisms to be used as potential drug targets is a nontrivial problem. Schlabach et al. have developed shRNA libraries that can be easily screened using DNA barcodes and microarray deconvolution. These libraries can be used to rapidly perform dropout screens to identify genes involved in cancer proliferation. In order to efficiently and rapidly quantify the pool of shRNAs, the authors develop a methodology they call half-hairpin (HH) barcoding. Essentially, while shRNAs are comprised of hairpin loops with both sense and anti-sense strands, only the anti-sense strand of each hairpin is amplified and used to produce oligonucleotides for the microarray. 19-nucleotide HH barcodes are advantageous compared with 60-nucleotide full-hairpin barcodes becausethey allow more rapid construction and screening and eliminate probe self-annealing, and they have been shown to be highly reproducible and highly specific. Genes with proliferative or anti-proliferative phenotypes can be detected through changes in levels of the hairpins over time, which can be identified through the competitive hybridization of barcodes of shRNA populations before and after they have been propagated for a few weeks.

The authors constructed a pool of 8203 distinct shRNAs targeting 2924 genes that are cancer-related or are known to play roles in well characterized pathways involving phosphorylation, dephosphorylation, or ubiquitination. They used their screening platform to identify genes with proliferative and anti-proliferative phenotypes in DLD-1 and HCT116 colon cancer cells, human HCC1954 breast cancer cells, and normal human mammary epithelial cells (HMECs). Comparing different cancers with distinct origins allows identification of growth regulatory pathways that are common to both types of cancer as well as cancer-specific growth regulatory pathways. The shRNAs were placed in a murine stem cell virus (MSCV)-derived retroviral vector, and cells were infected with an average representation of 1000 independent integrations per shRNA. Hairpins from initial and end samples were PCR-recovered and dyed red and green, respectively. Red probes indicate a gene necessary for growth, and green probes indicate a tumor suppressor. This method was shown to be highly reproducible, as the results of independent trials were very highly correlated.

Most shRNAs showed unappreciable changes in abundance over time, but some showed depletion. Many of these showed depletion in all four cell lines, but others showed selective depletion in certain cell lines and not in others. Upon hierarchical clustering of the cell types based on similarity of gene dropout and enrichment, the two colon cancer cell lines were the most similar to one another, as expected, and the normal HMECs were the most segregated from the other three. Genes that were selectively silenced in the cancer cell lines and not in HMECs, such as the cell cycle regulator and spindle checkpoint kinase BUB1, reflect underlying differences between the reproductive and proliferative machinery of cancer cells and normal cells and should be further investigated as potential drug targets in the future. Additionally, knowledge of which genes are necessary for cancer cell survival may lead to a better understanding of important signaling pathways common in cancer cells.

The authors do a good job of thoroughly explaining and testing a method that promises to be very useful in better characterizing pathways essential in cancer cell proliferation. The half-hairpin barcoding methodology described has the ability to uncover a large number of potential drug targets for the treatment of cancer. However, this technique should not be counted on for a thorough analysis of all genes selectively required for survival of tumor cells in various cancers. Many shRNAs seem not to show up at all on the microarray, probably because they were never successfully integrated into the genome of the human cell. Another interesting point is that the initial samples were taken after the cells had been sitting for over 48 hours, probably because before this time, a large portion of the viral vectors were not integrated into the host genome. It is possible that cells may undergo important changes during this waiting period. For example, if a shRNA targets an anti-apoptosis gene before two days pass and causes cell death, this phenomenon would go unnoticed. While this method may uncover many genes that are important in cancer tumorigenesis and lead to the discovery of new drugs for cancer treatment, it cannot be counted on to generate a comprehensive library of genes with proliferative phenotypes in cancer cells that are potential drug targets.

Michel Nofal
19384585