Challenges and Opportunities in Defining the Essential Cancer Kinome

See allHide authors and affiliations

Science Signaling  24 Mar 2009:
Vol. 2, Issue 63, pp. pe15
DOI: 10.1126/scisignal.263pe15


Signaling pathways controlled by protein kinases underlie a large fraction of human diseases and participate in the development and progression of all forms of cancer. Targeted therapeutic strategies to treat cancer and other diseases are focused almost exclusively on protein kinases, with a strong bias toward a small subset of the entire human kinome. RNA interference (RNAi)–based screens for protein kinase requirements have revealed a surprisingly high degree of diversity between cancer cell lines in their dependence on specific protein kinases. These screens also demonstrate that some of the most critical protein kinases for the proliferation and survival of cancer cell lines are also the least studied. Although the concept of oncogene addiction is powerful in designing therapeutic strategies to treat cancer, unbiased kinome-specific and genome-wide RNAi screens are revealing unexploited areas of potential therapeutic intervention.

The Therapeutic Potential of the Human Kinome

Signal transduction pathways composed of protein kinases are the primary lines of cellular communication, relaying cues from external and internal sources to affect cell physiology. Nearly half of the 518 protein kinases encoded by the human genome are expressed from loci associated with specific diseases or regions amplified in human cancers (1). In combination with the “druggability” of these enzymes, it is not surprising, therefore, that the development of pharmacological inhibitors of specific protein kinases is at the forefront of the targeted therapeutics revolution. Rewiring of cellular signaling networks underlies the development and progression of cancer, where multiple protein kinase pathways can be corrupted within a single tumor. The complexity of this reprogramming makes it difficult to predict which kinases are most strongly contributing to tumor cell growth, proliferation, survival, metastasis, and therapeutic response. In principle, unbiased RNA interference (RNAi)–based screens could help identify points of vulnerability within these rewired signaling networks and, therefore, define drug targets specific to particular oncogenic settings. The results of several such screens on specific cancer cell lines have now been published, using both kinome-specific (26) and genome-wide short-hairpin RNA (shRNA) libraries (713). The goal here is not to review the specific technical strengths and weaknesses of various RNAi-based screening approaches, but rather to discuss surprising findings in regard to cell-line–specific requirements for protein kinases and the challenges that lie ahead in identifying the kinases underlying the pathophysiology of human cancers.

Differential Protein Kinase Requirements for Human Cell Lines

Four recent studies have sought to define the protein kinases essential for the proliferation, survival, or both of various human cell lines (36). These studies use a previously described lentiviral-based shRNA collection assembled by the RNAi Consortium at the Broad Institute of MIT and Harvard (7), from which they screened shRNAs targeting ~85% of the human kinome. A well-based approach introducing single shRNA constructs into both HeLa and 293T cells was used (5). Required kinases were identified in a viability assay 5 days after transfection, scored with an assay for mitochondrial metabolism (alamarBlue) that does not distinguish between cell cycle arrest, senescence, autophagy, or apoptosis. Strikingly, these two cell lines, which are among the most commonly used lines for in vitro mechanistic studies, were found to have little overlap in their sensitivities to kinase-targeted shRNAs. Only ~25% (19 of 80) of protein kinases scoring positive in this assay decreased the viability of both the HeLa and 293T cell lines. To determine whether these vast differences in kinase requirements are specific to this pair of cell lines, which have very different origins, four non–small cell lung cancer (NSCLC) cell lines were similarly screened with the kinase-targeted shRNAs. Surprisingly, there was only a 5% overlap (14 of 278) among all four NSCLC cell lines for kinase requirements (5). Finally, this analysis was expanded to 21 different human cell lines, including tumor-derived and immortalized cell lines, as well as primary keratinocytes and fibroblasts. Once again, very little overlap was observed between the different tumor-derived cell lines. However, the primary cells isolated from different patients showed substantial cell-type–specific overlap in kinase requirements. This study demonstrates that defining a core set of essential human kinases shared by all cells may be impossible and suggests that variation between cells of different origins profoundly affects their reliance on specific protein kinases.

The results of this kinome-specific shRNA screen suggests that the wiring of protein kinase pathways regulating key cellular processes required for proliferation and survival varies substantially between different types of cells. The differential dependence on specific protein kinases may be especially true of cancer cells, even if derived from the same origin. The findings with primary human cells isolated from different patients on different days demonstrate the existence of hard-wired, cell-type–specific kinase requirements in normal cells. This suggests that, at their onset, tumor cells arising from the same origin are dependent on a nearly identical set of protein kinases. This set of “essential” kinases then diverges dramatically from its origins as the cell acquires genetic and epigenetic alterations over the course of tumor progression. Furthermore, the myriad of oncogenic paths from primary cell to cancer yields tumor cells with very little resemblance to either the cell of origin or cells from independent tumors derived from this origin. Of course, one must also take into account the adaptation to culture conditions when comparing tumor-derived cell lines. However, it is surprising that, with few exceptions (6), these particular studies did not find substantial overlap in the kinase requirements between tumor cell lines derived from the same tissue of origin, as one might expect them to at least cluster more closely with one another than they do with cell lines of other origins (5). In contrast, genome-wide, pool-based shRNA screens against panels of tumor-derived cell lines have found origin-specific signatures of essential genes, as well as pan-cancer sets of essential genes (1113). Finally, the vast differences in kinase requirements between some of the most commonly used cell lines (for example, HeLa and 293Ts) questions our tendency to broadly extrapolate from findings in one cell system to the next. Whereas it is likely that the basic molecular features of major proliferation and survival pathways (for example, the Ras and phosphoinositide 3-kinase pathways) are shared from cell line to cell line, variations triggered by genetic and epigenetic events, as well as by cell culture conditions, may dramatically alter a cell’s dependence on a specific pathway. Such variations include the presence or absence of critical kinase substrates, redundant kinases for the phosphorylation of these substrates, and other parallel compensatory mechanisms.

Challenges in Defining “Essential” Kinases

It is clear that defining a protein kinase as “essential” to a given cell type is a matter of perspective and is dependent on the conditions in which the requirement was assayed. This is best illustrated with viability screens in yeast. Whereas only 28 of the 110 kinases encoded by the Saccharomyces cerevisiae genome are essential for vegetative growth in rich medium, most of the remaining kinases are essential under one or more specific nutrient or stress condition (14). This is also true of the Schizosaccharomyces pombe kinome (15). Therefore, the growth conditions and end points chosen for a particular screen will profoundly affect the “hits” identified. In addition to differences in RNAi-based libraries, differences in growth parameters and assays make it difficult to compare the results of one screen to the next. For instance, the kinase requirements defined for HeLa cells in the study described above (5), although having some overlap, vary substantially with those determined by an siRNA-based screen with apoptosis as an end point (2). Underlying these points is the likelihood that the collection of kinases defined as essential for the viability of cancer cells under culture conditions will differ substantially from those essential within the setting of a tumor. However, in vitro screens such as these can be powerful in defining unexpected points of vulnerability specific to tumor cells or, as discussed below, specific to a common oncogenic lesion.

Protein Kinase Requirements Dictated by Single Oncogenic Events

Screens for essential protein kinases may be particularly powerful when comparing two otherwise isogenic cell lines that differ only in the presence of the products of a single oncogene or tumor suppressor gene. Such screens offer the potential to define oncogene-induced and tumor suppressor–induced resistance or sensitivity to the loss of specific kinases, thereby identifying points of resistance and sensitivity to therapeutics. To this end, two screens were reported that targeted a subset of human protein kinases in pairs of isogenic lines differing in either a single oncogene (3) or tumor suppressor (4). Baldwin et al. determined the effects of kinase-targeted shRNAs on the viability of a colon carcinoma cell line (RKO) compared with the same cells expressing the E7 oncogene, derived from the high-risk human papillomavirus HPV16 (3). Interestingly, expression of the E7 oncogene rendered the RKO cells more resistant to the loss of 19 different protein kinases, of which 5 were more rigorously confirmed, whereas it did not substantially sensitize them to any kinase-targeted shRNAs. In a similar study, Bommi-Reddy et al. examined the effects of the von Hippel-Lindau (VHL) tumor suppressor gene, which is mutated in the majority of clear cell renal cell carcinomas (RCC), on the sensitivity of two RCC lines to protein kinase–targeted shRNAs (4). The viability of the VHL−/− RCC lines 786-O and RCC4 was compared to those stably reconstituted with exogenous VHL. In this case, the VHL-null cell lines were sensitive to more kinase knockdowns than the reconstituted lines, and three kinases were further confirmed as selectively essential to both cell lines lacking VHL. The identification of these kinase requirements specific to VHL loss was aided by the fact that these two RCC lines, as seen with other pairs of tumor-derived lines from shared origins (5, 6), displayed little overlap in their dependence on specific kinases. Therefore, isogenic screens such as these can identify both points of resistance and vulnerability induced by either oncogene activation or tumor suppressor loss.

It is the hope that kinome-specific and genome-wide shRNA screens will reveal the Achilles’ heel, or heels, of specific oncogenic events. Although single genetic lesions, in isolation, do not give rise to tumors and cancer, they can profoundly affect the dependence of a tumor cell on specific protein kinases and pathways. These requirements can fall under the category of classic “oncogene addiction” (16), but can also extend to unexpected parts of the larger signaling network (Fig. 1). It is these secondary kinase requirements (for example, Kinase 3 in the figure) that we have yet to fully tap into in the development of targeted therapeutics. The collective work from shRNA screens and pathway-specific mechanistic studies should continue to expose these important candidate drug targets.

Fig. 1

Model of a hypothetical signaling network demonstrating how a single oncogenic event might alter the dependence of a cell on specific protein kinases. (A) In this example of a typical protein kinase signaling network from a normal cell, three nonessential protein kinases cooperate, in a partially redundant manner, to regulate two downstream substrates. The phosphorylation of both of these substrates is essential for the proliferation and survival of this particular cell. (B) Loss of a single tumor suppressor can render two nonessential protein kinases essential. Disruption of a tumor suppressor that normally inhibits Kinase 1 results in aberrantly elevated Kinase 1 activity that rewires the signaling network such that Kinase 2 activity is attenuated. This single genetic lesion renders Kinase 1 essential to the tumor cell, an example of “oncogene addiction.” At the same time, a seemingly unconnected kinase, Kinase 3, also becomes essential to this cell.

Scratching the Surface of the Human Kinome

One underlying message from the “hit” lists in all of these RNAi-based screens for essential genes is that, as signal transduction researchers, we are focused on a rather narrow set of protein kinase pathways that neglects a large number of kinases essential for tumor cell proliferation and survival. In fact, we know very little regarding the biochemical and cell biological properties of the majority of human protein kinases. In the study by Grueneberg et al. (5), the 53 protein kinases found to be essential across the largest number of cell lines represent every major branch, as well as many poorly characterized side branches, of the human kinome (1). Furthermore, many of these frequently essential kinases are members of larger kinase families that might be expected to exhibit functional redundancy. It is estimated that over half of the human protein kinases would be found to be essential in at least one cancer cell line (5). Delineating how these understudied protein kinases interact with well-characterized signaling subnetworks is critical to our understanding of their role in cancer progression and their potential as therapeutic targets.


View Abstract

Navigate This Article