The perfection of a fly's eye and the chaotic nature of tumours provide eloquent examples of the need to coordinate cell death and proliferation. The intricacies of the underlying mechanism are now being uncovered.
Just as the number of working parts in a machine is crucial, so too is the number of cells in a tissue. This means that the creation of new cells (by proliferation) and the elimination of excess ones (by programmed cell death, or 'apoptosis') must be tightly coordinated. Although many genes have been shown to regulate either proliferation or apoptosis, little is known about how the two are coupled. Recently, however, geneticists have begun to uncover some of the genes involved in this coordination, and, writing in Cell, Wu et al.1 and Harvey et al.2 describe another such gene, hippo. Their findings significantly advance our understanding of this fundamental problem in organ development and cancer biology.
Both groups1,2 looked at the imaginal discs of fruitflies. Imaginal discs are packets of cellular monolayers (hence the name disc) that are set aside during larval development, and differentiate during metamorphosis (the pupal stage) to give rise to most of the tissues of the adult fly. In the case of the eye, on which the two groups focused, undifferentiated retinal imaginal disc cells proliferate rapidly throughout much of development. Later, waves of differentiation sweep across the eye disc to pattern the tissue (that is, to determine the fates and arrangements of the cells) and to halt cell division. This leads to the formation of a highly organized retina consisting of about 750 regularly spaced 'seeing units', called ommatidia (Fig. 1a). During this process, the imaginal disc has an excess of cells between the ommatidia. Most of these leftover 'interommatidial' cells (some 2,000) are later eliminated by a strictly controlled wave of apoptosis, leaving a thin layer. The end result is a retina so perfectly organized that it has been called a “neurocrystalline lattice”3. Its perfection provides a sensitive assay for probing the link between proliferation and apoptosis.
Previous work had already indicated that cell proliferation and death in the developing retina are tightly coupled. First, inducing apoptosis early on in a group of retinal precursors — a manipulation that should dramatically reduce the size of the eye if cell death and proliferation were completely independent and inflexible — does not result in a much smaller eye4. Second, driving extra proliferation causes the imaginal disc to grow too large only if cell death is simultaneously inhibited5. Third, several genes, including salvador (also known as shar-pei)6, have been shown to coordinately regulate both proliferation and apoptosis7,8.
This coupling of proliferation and apoptosis seems to be a general rule in multicellular organisms, and the dysregulation of these processes can lead to cancer. Indeed, tumour development is thought to require both an increase in proliferation and a decrease in cell death9. Thus, a gene that promotes both apoptosis and exit from the cell-division cycle should be a tumour-suppressor gene. In fact, salvador is such a gene, and its human relative hWW45 has been implicated in several cancers8.
Quite how salvador and other such genes couple cell proliferation and death was unknown. But our understanding of this is now advanced by the characterization of hippo1,2, a gene that functions with salvador and another tumour suppressor called warts (also known as lats)10. Wu et al.1 and Harvey et al.2 embarked on genetic screens to identify genes that negatively regulate tissue growth. They found hippo, a gene that belongs to a family of kinases — enzymes that add phosphate groups to other proteins. The Hippo protein is 60% identical to the human kinase Mst2, a close relative of which, Mst1, has been implicated in apoptosis11.
The authors then found1,2 that tissues lacking hippo contain too many cells. Within those mutant tissues, cell size and many aspects of patterning and cell-fate determination are remarkably normal — but the number of cells is dramatically increased. For instance, in the mutant eye, early patterning is not affected and photoreceptor cells differentiate and assemble into normal ommatidia. Yet there are many more interommatidial cells than normal, resulting in overgrown and folded eyes (Fig. 1b).
What causes the increase in cell number? First, there is increased proliferation. After the wave of differentiation has swept across the imaginal disc, normal cells stop dividing (or divide only once more). But cells lacking hippo continue to cycle for some time. At least one reason is that these cells upregulate Cyclin E, a limiting factor for entry into the DNA-replication phase of proliferation in imaginal disc cells5,12. Both groups agree that transcription of the Cyclin E gene is increased, although Harvey et al. suggest that a post-transcriptional mechanism might also contribute. So why are these extra cells not simply pruned away by apoptosis? The reason is that hippo is also required for apoptosis. Cells lacking hippo have increased levels of a key inhibitor of apoptosis, DIAP1, and are strongly resistant to treatments that should induce apoptosis.
The defects produced by a lack of hippo are remarkably similar to those produced when warts or salvador is missing, and indeed Wu et al. observed strong, specific genetic interactions between the three genes. These experiments, as well as other studies described in the papers, show that Hippo can associate with and phosphorylate Salvador, and that this interaction helps Hippo to phosphorylate Warts, another kinase. This suggests a model in which Salvador functions as an adaptor protein that brings Hippo and Warts together in a ternary complex. Presumably, phosphorylated Warts — as the major (but not the only) output of this complex — then goes on to regulate Cyclin E, DIAP1 and other downstream effectors. This work has thus begun to define a genetic pathway for the coordinated regulation of cell-cycle exit and apoptosis.
So what remains to be discovered? First, there is the question of mechanism. What is the exact nature of the physical, genetic and biochemical interactions between Hippo, Salvador and Warts? What is the significance of the phosphorylation of Warts by Hippo? And when, where and how does phosphorylation occur in this pathway in vivo? Also, given that any pathway that includes kinases also has regulatory phosphatases (proteins that remove phosphate groups), are phosphatases also involved in coordinating proliferation and death?
Second, there is the question of downstream targets. Tissues lacking salvador, hippo or warts show reduced cell death and more proliferation — but the cells also divide faster, yet remain of normal size. So this pathway also regulates cell growth (mass accumulation), and downstream targets in addition to DIAP1 and Cyclin E must exist. Might Hippo, as a member of the Ste20 family of kinases13, be connected to signalling modules that contain mitogen-activated protein kinases — modules that regulate cell proliferation, death and growth elsewhere?
Third, there is the question of upstream control. How do cells receive and process the signals that work through the Hippo–Salvador–Warts pathway to control proliferation, death and growth? As all three proteins are ubiquitously expressed, it is likely that their activity (or the ability of cells to respond to them) is regulated, rather than their expression. Perhaps, as in other pathways involving Ste20-family kinases, regulatory enzymes such as members of the Ras family will be found upstream.
Finally, as a first step towards testing whether Hippo's role in coupling proliferation to apoptosis is evolutionarily conserved, Wu et al.1 investigated whether the human Mst2 protein could restore the status quo to fruitflies that lack Hippo. They found that it could, raising the exciting possibility that Mst1 and Mst2 are tumour suppressors. This question could be investigated in several ways: by knocking out the genes in mice; by analysing them in tumour cell lines; and by searching for human families with heritable cancer syndromes that are caused by mutations in the genes for Mst1 or Mst2.
Wu, S., Huang, J., Dong, J. & Pan, D. Cell 114, 445–456 (2003).
Harvey, K. F., Pfleger, C. M. & Hariharan, I. K. Cell 114, 457–467 (2003).
Ready, D. F., Hanson, T. E. & Benzer, S. Dev. Biol. 53, 217–240 (1976).
Stowers, R. S. & Schwarz, T. L. Genetics 152, 1631–1639 (1999).
Neufeld, T. P., de la Cruz, A. F., Johnston, L. A. & Edgar, B. A. Cell 93, 1183–1193 (1998).
Kango-Singh, M. et al. Development 129, 5719–5730 (2002).
Brennecke, J., Hipfner, D. R., Stark, A., Russell, R. B. & Cohen, S. M. Cell 113, 25–36 (2003).
Tapon, N. et al. Cell 110, 467–478 (2002).
Green, D. R. & Evan, G. I. Cancer Cell 1, 19–30 (2002).
Xu, T., Wang, W., Zhang, S., Stewart, R. A. & Yu, W. Development 121, 1053–1063 (1995).
Graves, J. D. et al. EMBO J. 17, 2224–2234 (1998).
Richardson, H. et al. Development 121, 3371–3379 (1995).
Dan, I., Watanabe, N. M. & Kusumi, A. Trends Cell Biol. 11, 220–230 (2001).