Tumour cells tend to carry many gene mutations, but at a potential cost to their overall fitness. Studying the interactions between genes on a large scale could be a way of identifying the chinks in the tumour cell armour.
Cancer arises when the right combination of genes is mutated in a susceptible cell, much like lock tumblers falling into place. These mutations bestow various properties of malignancy upon the cell, such as independence from growth control and the ability to colonize other tissues. The fact that such cells contain multiple gene mutations tends to be viewed as problematic by developers of anticancer treatments — the typical thinking goes that each mutation is beneficial to the cancer cell and makes it hardier. But the mutations probably come at a cost to the cell with respect to its ability to respond to certain situations. Unfortunately, we cannot yet predict how cancer-associated mutations might make a cell vulnerable, and there have been no unbiased methods for systematically discovering these weaknesses. On page 106 of this issue, however, Louis Staudt and colleagues1 describe an approach for identifying genes that become essential for the survival of cancer cells.
The idea that the deleterious consequences of a particular mutation might be revealed only under specific conditions is an old one. Not surprisingly, it has been most extensively explored in organisms such as yeast and fruitflies, in which the genome can be manipulated easily and the consequent changes seen rapidly. Such investigations uncovered a gene– gene interaction, called ‘synthetic lethality’, that has potential significance for cancer drug discovery. Two genes (A and B, say) are said to be synthetically lethal if the cell containing them dies when both genes are mutated, but it can survive if either gene alone is mutated. In perhaps the simplest case, A and B perform the same function and are uniquely redundant with respect to one another. Alternatively, B might be part of a pathway that can ‘rescue’ the pathway that is damaged by mutation of A. Studies in yeast, however, suggest that these possibilities are the tip of the iceberg — synthetic-lethal interactions seem to be unexpectedly common, although they are often hard to predict2,3.
Hartwell et al.4 proposed that synthetic-lethal interactions could be used to develop safer and more effective anticancer drugs. In particular, they considered the case of tumour-suppressor genes, the protein products of which inhibit tumour growth, and which are often inactivated in cancers by mutations. A drug that inhibits the protein product of a gene that is synthetically lethal to a tumour-suppressor gene would, by definition, kill those cells in which the tumour-suppressor gene was inactivated, but not their normal counterparts.
This approach is conceptually appealing because it turns cancer-specific mutations into a liability for the cells that contain them. It also tackles two vexing problems in cancer drug discovery: how to kill cancer cells without harming normal cells, and how to tackle ‘loss- of-function’ mutations pharmacologically — most effective drugs cause a loss of function, rather than restoring a damaged function. Synthetic lethality typically involves two loss-of-function mutations, but it also applies to gain-of-function mutations — those making an enzyme permanently active, say. For example, a gain of function in A might be synthetically lethal if it is not tolerated when B is inactive. In the jargon of cancer biology, genes that contribute to cancer when carrying gain-of-function mutations are termed proto-oncogenes, and one can envision other targets that might become vital to the cell specifically in the context of such proto-oncogene mutations.
But where does this leave the study of human cancer? So far, there are only a few reports of synthetic-lethal interactions involving human cancer genes, and we do not know enough about the genetic networks in cancer cells to predict such interactions. Although discovering synthetic-lethal interactions is fairly simple in yeast, most human cancer genes do not have relatives in yeast cells and the analogous experiments are much trickier, or nigh-on impossible, in human cells. So Staudt and colleagues developed a much-needed, high-throughput approach to discovering synthetic-lethal interactions in human cancer cells.
Their method is based on a technique called RNA interference, which involves the RNA molecules that act as a bridge between a gene and its product. In this technique, a ‘short interfering RNA’ (siRNA) molecule is used to target a specific messenger RNA for destruction within the cell — halting production of its encoded protein. It is possible to perform large-scale genetic screens in human cells using libraries of genes that encode numerous siRNAs or ‘short hairpin RNAs’ (shRNAs, which are processed to siRNAs in vivo)5. Certain human shRNA libraries contain genes that are tagged with a unique DNA sequence, or ‘bar code’6,7. These bar codes can be used to identify the shRNAs that are selectively enriched under certain experimental conditions8,9. However, identifying shRNAs that are depleted under certain conditions — which would be required for synthetic-lethal screens — has been problematic in human cells.
Staudt and colleagues1 created a human shRNA library of DNA ‘vectors’ representing around 2,500 genes. Each vector contained a unique 60-base-pair bar code, and was designed such that shRNA expression could be switched on with a compound called doxycycline. The authors introduced the libraries into cells from two different types of diffuse large B-cell lymphoma — called activated B-cell-like (ABC) and germinal-centre B-cell-like (GBC) diffuse large B-cell lymphoma — on the assumption that these two types of cancer are driven by different mutations. The cells were then divided into two groups and either treated with doxycycline or left untreated. The abundance of the individual shRNA vectors was monitored using a DNA microarray consisting of probes that could detect the different bar codes in the library (Fig. 1). Vector abundance gives an indication of its effect on cell viability, or ‘fitness’, as abundance will increase as the cells divide and will decrease as cells die. So once a particular shRNA is induced by doxycycline, the effects of the reduction in the specific protein it targets can easily be monitored.
As expected, many shRNAs did not affect cellular fitness, or affected fitness negatively in both lymphomas. However, shRNAs targeting the genes IKBKB, CARD11, MALT1 and BCL10 were depleted after doxycycline addition only in the ABC cells. These shRNAs inhibited ABC proliferation selectively in further experiments, suggesting that IKBKB, CARD11, MALT1 and BCL10 are synthetically lethal to one or more mutations responsible for ABC. Alternatively, the normal cells that are the precursors of ABC might rely more heavily on these genes than do the progenitors of GBC because of heritable, non-genetic differences that persist after malignant conversion. These genes all encode agonists for the gene regulatory factor NF-κB, so the results are consistent with an earlier finding that ABC is particularly sensitive to small-molecule NF-κB antagonists10.
The approach described by Staudt and colleagues could be used to examine pairs of cell lines that differ only with respect to a mutation in a particular proto-oncogene or tumour-suppressor gene. In such a screen, a ‘hit’ would be an shRNA vector that was depleted after shRNA induction in the mutant cells, but not their unmutated counterparts. Libraries of shRNAs should continue to improve, increasing genomic coverage and the degree of target inhibition, and so becoming ever more powerful tools in this endeavour. Sociologists have noted that some technological advances enable scientists to become ‘communal harvesters’ rather than ‘hunter-gatherers’. With luck, the approach described by Staudt and colleagues will yield a bountiful harvest of cancer-drug targets that exploit weaknesses created by cancer-associated mutations.