Introduction

Cancer has been described as a multistep process requiring six essential alterations in cell physiology: (i) self-sufficiency in growth signals, (ii) insensitivity to antigrowth signals, (iii) evasion of apoptosis, (iv) limitless replicative potential, (v) sustained angiogenesis, and (vi) tissue invasion and metastasis (reviewed in Hanahan and Weinberg, 2000). To study these principles of malignant growth, numerous models are being used. The mouse model is one of the best because of the similarities between mouse and human tumours. However, it has disadvantages such as cost and space, and transgenic technologies are time consuming and tedious. An alternative is to use other model organisms such as Xenopus, Drosophila, or Caenorhabditis elegans. These model organisms have contributed importantly to our understanding of basic biological phenomena that are at the heart of the six alteration principles proposed by Hanahan and Weinberg.

In this review, we will focus on C. elegans and how it can contribute to cancer gene discovery. Although C. elegans does not form tumours per se, other types of mutant phenotypes are clearly relevant for cancer research. As a few examples, there are mutants where cell division is unregulated (e.g. cki-1, a cycline kinase inhibitor (Kostic et al., 2003) and glp-1, a Notch-like gene (Berry et al., 1997)), where cells that should undergo apoptosis fail to die (e.g. ced-3 (Ellis and Horvitz, 1986), ced-9 (Hengartner et al., 1992), and cep-1 TP53 (Derry et al., 2001; Schumacher et al., 2001)), and where the genome is unstable, generating spontaneous mutations (e.g., msh-2 and msh-6, mismatch repair genes; Tijsterman et al., 2002). For each of these phenotypes and for many others, some of the mutations causing the defects identify genes linked to cancer in humans. Screening for C. elegans genes disrupting processes relevant to cancer and concentrating research in these areas should further our understanding of the basic processes underlying cancer.

C. elegans as a model in the post-genomic era

Among multicellular model organisms, C. elegans can be considered a pioneer of the post-genomic era (Sternberg, 2001). C. elegans was the first metazoan to have its genome sequenced, its genome annotation is of remarkable accuracy, and the C. elegans database Wormbase presents data to the world in a user-friendly format. About 60% of C. elegans genes have similarity to a human gene (Harris et al., 2004). A real breakthrough for increasing the usefulness of C. elegans as a model for identifying and studying conserved gene functions was the discovery of RNA interference (RNAi). Previously, classical genetics was the most efficient way of identifying gene functions in C. elegans and it did so in a remarkable manner, being acknowledged by the 2002 Nobel Prize in Physiology and Medicine to Sydney Brenner, Robert Horvitz, and John Sulston. Nevertheless, gene identification through forward genetics is still labour intensive.

Fire and Mello discovered that injection of double-stranded RNA (dsRNA) into worms leads to specific degradation of the corresponding mRNA (Fire et al., 1998). Soon afterwards, it was shown that either soaking worms in dsRNA solution or feeding worms bacteria engineered to produce dsRNA could also elicit a robust RNAi response (Tabara et al., 1998; Timmons and Fire, 1998). Of these three methods (injection, soaking, feeding), RNAi by feeding has the advantage that, once a bacterial strain is made, it is a permanent reagent that can be reused and replicated. The availability of the genome sequence, together with the discovery of RNAi and the advances in dsRNA delivery, made possible the building of an RNAi feeding library to inactivate most C. elegans genes (Fraser et al., 2000; Kamath et al., 2003).

This RNAi feeding library consists of 16 757 bacterial strains, kept as glycerol stocks, each designed to produce dsRNA to target an individual C. elegans gene; there is currently a bacterial strain for 86% of genes. Using this library, C. elegans was the first metazoan to have most of its genes subjected to a loss-of-function analysis. This tripled the number of genes associated with loss-of-function phenotypes, underlying the power of the RNAi library for rapidly finding gene functions and encouraging users of other model organisms to generate similar tools. The library can also be used for candidate-based approaches where only subsets of genes (e.g. 100–2000) are subjected to analysis. RNAi by feeding is versatile and can be performed in liquid culture in 96-well format, reducing the screening labour (Nollen et al., 2004). The bottleneck of genome-wide RNAi screens is the assessment of phenotypes. In many cases this will require a human scorer, and it is worth investing time to make the assay as easy and rapid as possible. Automated visual systems are starting to be used for screens where the assay is life versus death, and these should be amenable for assays of altered reporter expression as well (e.g. GFP). A genome-wide RNAi screen in liquid culture where the assay is rapid and unambiguous (life versus death or GFP+ versus GFP-) can be carried out manually by one person in about a month.

Apoptosis

Apoptosis is a genetically regulated form of cell death that is triggered by different stimuli. It is used to discard misbehaving cells, maintain homeostasis, and eliminate cells during development. Misregulation of apoptosis can lead to a range of pathological conditions, including cancer. Therefore, understanding the process and regulation of apoptosis is important for medical research.

During C. elegans development, 131 cells undergo programmed cell death to produce the final 959 somatic cells (Sulston and Horvitz, 1977). In addition, during oogenesis, about half of all germ cells are eliminated by apoptosis (Gumienny et al., 1999). As in other systems, apoptosis in C. elegans requires a caspase (CED-3) and an Apaf-1 homologue (CED-4), and is blocked by a Bcl-2-like protein (CED-9) (reviewed in Liu and Hengartner, 1999). DNA damage (such as -radiation) causes an increase in germline cell deaths, and this requires the C. elegans p53 orthologue cep-1 and the checkpoint gene hus-1 (Derry et al., 2001; Schumacher et al., 2001). Two RNAi screens for genes involved in apoptosis have already been carried out. In the first, the authors screened for an increase in germline apoptosis; germline corpses are easily detectable by acridine orange staining followed by fluorescent microscopy (Figure 1c–f). After screening 16 757 genes, 21 were discovered where an RNAi knockdown causes excess deaths (Lettre et al., 2004). Of these, 16 were previously unknown, including pmk-3, a p38 MAPK homologue, and bmk-1, a homologue of the BimC kinesin-like motor protein involved in spindle formation. Even though the screening procedure identified just a few of the expected known genes, it succeeded in finding new candidates. Of the 21 genes identified, 16 require p53 and a functional DNA damage response pathway to activate germ cell death; these genes might act in DNA repair or genome stability. In fact, the screen identified homologues of RAD50 and RAD51 which are required for DNA repair and meiotic recombination (de Jager and Kanaar, 2002; Sung et al., 2003). Of the five p53 independent genes found, one is ced-9, a regulator of physiological germ cell death (Gumienny et al., 1999), suggesting that the other four might also play a role in that process.

Figure 1.

Examples of C. elegans phenotypes used for genome-wide RNAi screening. Growth, apoptosis, and vulval phenotypes are shown, with wildtype on the left and mutant on the right. (a, b) Growth /Lethality assay. Wells from an RNAi screen for lethality carried out in 96-well liquid culture format. (a) Wildtype adult worms put into the bacterial culture and their progenies are visible. (b) In a well where embryonic lethality was induced by RNAi, no progeny are visible. (c–f) Germline apoptosis assay. Germline corpses are visible after live staining with acridine orange. Sections of gonads viewed with DIC optics (c and d) or under fluorescence (e and f). (c and e) Wildtype adult hermaphrodite gonad with one corpse (arrow). (d and f) ced-9 mutant gonad with excess corpses (arrows). (g and h) Multivulval assay. (g) A single wildtype vulva or (h) multiple vulvae protrusions that can be detected under the dissecting scope; this animal has a gain-of-function mutation in let-60 Ras. (g and h) White arrows point to the normal vulva and in (h) the black arrows indicate the ectopic vulvae

Full figure and legend (111K)

A second group used a candidate-based functional genomic screen for apoptotic DNA degradation in C. elegans (Parrish and Xue, 2003). Using INTERPRO and PFAM motifs to find genes corresponding to deoxyribonucleases, ribonucleases, cyclophillins, and topoisomerases, they found 77 candidates that were analysed by a terminal deoxyribonucleotidyl transferase (TdT)-mediated biotin-16-dUTP nick-end labelling (TUNEL) assay. RNAi of nine of these candidates were positive in that assay and all of them have a human homologue. Candidate-based approaches that combine bioinformatics with RNAi technology, such as those described above, are very powerful. They can also be very rapid, since most of the selected candidates can be provided by the existing RNAi library, avoiding the step of generating reagents. Past studies of apoptosis have shown that C. elegans research has played a leading role, and future investigation into this process looks promising.

C. elegans cancer genes

The screens described above took advantage of a well-characterized biological process relevant to cancer (apoptosis). An alternative method for finding genes relevant to cancer is to study C. elegans homologues of human cancer genes. Futreal et al. (2004) recently compiled a list of 291 genes causally implicated in human cancer by conducting a census of the literature (listed at http://www.sanger.ac.uk/genetics/CGP/Census/). We took the subset of 61 of these cancer genes where germline mutations have been found and searched for C. elegans homologues. Strikingly, 80% have a putative worm homologue (Table 1), and a mutant or RNAi phenotype has been reported for 53% of these (the total set of cancer genes yielded similar results). These genes can be used as starting points for conducting RNAi screens for new cancer genes. For example, mutants of the mismatch repair genes msh-2 and msh-6 have mutator activity in C. elegans and an assay for genome instability was developed where these mutants scored positive (Tijsterman et al., 2002). Pothof et al. (2003) then used this assay in a genome-wide RNAi screen and identified 61 genes required for genome stability (van Haaften, this issue); these are excellent cancer gene candidates. Indeed, C. elegans homologues of two of the human cancer genes in Table 1, the DNA mismatch repair proteins PMS2 and MLH1, were identified in that screen. Similar approaches could be used for other genes on this list. For example, screens could be designed to find genes with similar knockdown phenotypes or to find genes that show a synthetic phenotype with the C. elegans cancer genes. Mammalian counterparts of these genes could then be studied in human cells or mice. This type of approach will potentially be a powerful and fruitful way to systematically identify new cancer genes.

Table 1 - C. elegans homologues of germline mutated human cancer genes.

Full table

Synthetic interactions

Most if not all oncogenes require cooperation with other factors. For example, activated forms of Ras cause cell-cycle arrest or apoptosis; however, if p53 is simultaneously inactivated, uncontrolled proliferation occurs. Hence, cancer could be classified as a type of synthetic disease. In a classical synthetic genetic interaction, single mutants of gene A or gene B do not show a phenotype, but an A;B double mutant does. Applying this to gene discovery, carrying out a synthetic genetic screen is an effective way to look for factors that cooperate together. As an example, one could carry out RNAi screens with a strain mutant for p53 (cep-1). After RNAi of each gene in this background, a number of synthetic phenotypes could be assayed (e.g., death, proliferation of the germline, over- or underexpression of a reporter, etc.). Homologues of positive candidates could then be tested for synthetic effects with p53 in human cells or transgenic mice. The value of this type of approach is illustrated by a recent screen for RNAi knockdowns that cause synthetic lethality with a mutant that generates double-strand breaks; several of the genes identified are needed for resistance to ionizing radiation (van Haaften et al., personal communication).

The Ras and synthetic Multiple vulvae (synMuv) pathways

The RTK/Ras/Raf/MAPK signalling pathway is highly conserved and is used in many contexts during animal development; unregulated Ras signalling can lead to cancer. Work in model organisms has greatly contributed to our understanding of this pathway and its regulation (Wassarman et al., 1995; Tan and Kim, 1999; Moghal and Sternberg, 2003). In C. elegans, Ras signalling has been most studied in the development of the vulva, the egg-laying organ, where Ras activity is needed for adoption of the vulval cell fate. In mutants where Ras signalling is hyperactivated, nonvulval cells are transformed to the vulval fate. This condition leads to ectopic vulval development, visualized by the presence of multiple vulvae (Figure 1g and h; Moghal and Sternberg, 2003). To prevent ectopic vulval development, Ras signalling is antagonized through the synMuv genes, which include components of Rb histone deacetylase (Lu and Horvitz, 1998; Ceol and Horvitz, 2001) and NuRD chromatin remodelling complexes (Solari and Ahringer, 2000; von Zelewsky et al., 2000). In other systems, these complexes have transcriptional repressor activity (Harbour and Dean, 2000; Feng and Zhang, 2003), suggesting that Ras signalling is antagonized by transcriptional repression of vulval development genes (Fay and Han, 2000). The C. elegans vulva is an excellent system for studying the regulation of the Ras signal transduction cascade and of its antagonism by chromatin remodelling proteins.

Classical forward genetic screens have identified numerous regulators of Ras signalling (reviewed in Sternberg and Han, 1998). Many of these screens involved looking for enhancers or suppressors of altered Ras signalling. For example, an activating mutation in let-60 RAS causes ectopic vulval development (Figure 1g and h); several genetic screens for suppressors of this phenotype identified positive regulators of the pathway (Kornfeld et al., 1995; Sundaram and Han, 1995; Sieburth et al., 1998). Repeating the classic screens (e.g., suppressor or enhancer of activated Ras) using RNA interference will likely prove effective in identifying additional regulators, since, in a pilot RNAi screen of 1000 genes, three new suppressors of activated let-60 Ras were found (AG Fraser and JA, unpublished). An advantage of the RNAi screening approach is that it can easily identify genes where loss-of-function mutants are unviable; these can be difficult to isolate in a standard genetic screen. Indeed, an RNAi screen for synMuv mutants identified 11 new genes, most of which are needed for viability, whereas only about 30% of synMuv genes identified by genetics are essential; several of the new genes encode components of the sumoylation pathway, suggesting that sumoylation of one or more proteins has a role in antagonizing Ras signalling (GP and JA, unpublished results).

Cell-invasive behavior

Some cells have an invasive activity that allows them to travel through basement membranes, and this is important in several developmental contexts including blastocyst implantation and organogenesis. Metastatic cancer is associated with a loss of regulation over invasive activity. Recently, a C. elegans model system for studying cell invasiveness was developed (Sherwood and Sternberg, 2003). In making the connection between the gonad and the vulva, a gonadal cell called the anchor cell (AC) invades the basement membranes of vulval cells. The authors identified mutant conditions where only 20% animals showed invasion by the AC. By screening for reduced or increased invasiveness in this background, one might be able to identify genes that control invasive activity.

Conclusion

Cancer is a complex illness involving many pathways conserved in C. elegans. The development of RNAi technology as a tool for systematic screening in C. elegans has already been fruitful for identifying new gene functions relevant to cancer. Careful investigation of the new phenotypes observed can provide a deeper understanding of the underlying biology and give ideas for the design of new RNAi screens. As human cancer genes are more highly conserved than other genes on average, the new genes identified are also expected in many cases to have human counterparts. A drawback to RNAi screening is that it is a loss-of-function method – it cannot be used to find gain-of-function mutants, which in the past have been very informative in understanding cancer processes. Nevertheless, taken together, RNAi screening and phenotypic analysis tools in C. elegans are extremely powerful for uncovering new gene functions and driving new mechanistic hypotheses. It will be important to investigate these new hypotheses in both C. elegans and mammalian systems; collaborations between laboratories with different and complementary expertise should lead to a better understanding of the biological processes disrupted in cancer cells.

References

1. Berry LW, Westlund B and Schedl T. (1997). Development, 124, 925–936. | PubMed | ISI | ChemPort |
2. Boulton SJ, Martin JS, Polanowska J, Hill DE, Gartner A and Vidal M. (2004). Curr. Biol., 14, 33–39. | Article | PubMed | ISI | ChemPort |
3. Ceol CJ and Horvitz HR. (2001). Mol. Cell, 7, 461–473. | Article | PubMed | ISI | ChemPort |
4. de Jager M and Kanaar R. (2002). Genes Dev., 16, 2173–2178. | Article | PubMed | ISI | ChemPort |
5. Derry WB, Putzke AP and Rothman JH. (2001). Science, 294, 591–595. | Article | PubMed | ISI | ChemPort |
6. Ellis HM and Horvitz HR. (1986). Cell, 44, 817–829. | Article | PubMed | ISI | ChemPort |
7. Fay DS and Han M. (2000). Genesis, 26, 279–284. | Article | PubMed | ISI | ChemPort |
8. Feng Q and Zhang Y. (2003). Curr. Top. Microbiol. Immunol., 274, 269–290. | PubMed | ISI | ChemPort |
9. Fire A, Xu S, Montgomery MK, Kostas SA, Driver SE and Mello CC. (1998). Nature, 391, 806–811. | Article | PubMed | ISI | ChemPort |
10. Fraser AG, Kamath RS, Zipperlen P, Martinez-Campos M, Sohrmann M and Ahringer J. (2000). Nature, 408, 325–330. | Article | PubMed | ISI | ChemPort |
11. Futreal PA, Coin L, Marshall M, Down T, Hubbard T, Wooster R, Rahman N and Stratton MR. (2004). Nat. Rev. Cancer, 4, 177–183. | Article | PubMed | ISI | ChemPort |
12. Gumienny TL, Lambie E, Hartwieg E, Horvitz HR and Hengartner MO. (1999). Development, 126, 1011–1022. | PubMed | ISI | ChemPort |
13. Hanahan D and Weinberg RA. (2000). Cell, 100, 57–70. | Article | PubMed | ISI | ChemPort |
14. Harbour JW and Dean DC. (2000). Curr. Opin. Cell Biol., 12, 685–689. | Article | PubMed | ISI | ChemPort |
15. Harris TW, Chen N, Cunningham F, Tello-Ruiz M, Antoshechkin I, Bastiani C, Bieri T, Blasiar D, Bradnam K, Chan J, Chen CK, Chen WJ, Davis P, Kenny E, Kishore R, Lawson D, Lee R, Muller HM, Nakamura C, Ozersky P, Petcherski A, Rogers A, Sabo A, Schwarz EM, Van Auken K, Wang Q, Durbin R, Spieth J, Sternberg PW and Stein LD. (2004). Nucleic. Acids Res., 32 (database issue), D411–D417. | Article | PubMed | ISI | ChemPort |
16. Hengartner MO, Ellis RE and Horvitz HR. (1992). Nature, 356, 494–499. | Article | PubMed | ISI | ChemPort |
17. Kamath RS, Fraser AG, Dong Y, Poulin G, Durbin R, Gotta M, Kanapin A, Le Bot N, Moreno S, Sohrmann M, Welchman DP, Zipperlen P and Ahringer J. (2003). Nature, 421, 231–237. | Article | PubMed | ISI | ChemPort |
18. Kornfeld K, Hom DB and Horvitz HR. (1995). Cell, 83, 903–913. | Article | PubMed | ISI | ChemPort |
19. Kostic I, Li S and Roy R. (2003). Dev. Biol., 263, 242–252. | Article | PubMed | ISI | ChemPort |
20. Lettre G, Kritikou EA, Jaeggi M, Calixto A, Fraser AG, Kamath RS, Ahringer J and Hengartner MO. (2004). Cell Death Differ., in press.
21. Liu QA and Hengartner MO. (1999). Ann. N. Y. Acad. Sci., 887, 92–104. | PubMed | ISI | ChemPort |
22. Lu X and Horvitz HR. (1998). Cell, 95, 981–991. | Article | PubMed | ISI | ChemPort |
23. Moghal N and Sternberg PW. (2003). Exp. Cell Res., 284, 150–159. | Article | PubMed | ISI | ChemPort |
24. Nollen EA, Garcia SM, van Haaften G, Kim S, Chavez A, Morimoto RI and Plasterk RH. (2004). Proc. Natl. Acad. Sci. USA, 101, 6403–6408. | Article | PubMed | ChemPort |
25. Parrish JZ and Xue D. (2003). Mol. Cell, 11, 987–996. | Article | PubMed | ISI | ChemPort |
26. Pothof J, van Haaften G, Thijssen K, Kamath RS, Fraser AG, Ahringer J, Plasterk RH and Tijsterman M. (2003). Genes Dev., 17, 443–448. | Article | PubMed | ISI | ChemPort |
27. Schumacher B, Hofmann K, Boulton S and Gartner A. (2001). Curr. Biol., 11, 1722–1727. | Article | PubMed | ISI | ChemPort |
28. Sherwood DR and Sternberg PW. (2003). Dev. Cell, 5, 21–31. | Article | PubMed | ISI | ChemPort |
29. Sieburth DS, Sun Q and Han M. (1998). Cell, 94, 119–130. | Article | PubMed | ISI | ChemPort |
30. Solari F and Ahringer J. (2000). Curr. Biol., 10, 223–226. | Article | PubMed | ISI | ChemPort |
31. Sternberg PW and Han M. (1998). Trends Genet., 14, 466–472. | Article | PubMed | ISI | ChemPort |
32. Sternberg PW. (2001). Cell, 105, 173–176. | Article | PubMed | ISI | ChemPort |
33. Sulston JE and Horvitz HR. (1977). Dev. Biol., 56, 110–156. | Article | PubMed | ISI | ChemPort |
34. Sundaram M and Han M. (1995). Cell, 83, 889–901. | Article | PubMed | ISI | ChemPort |
35. Sung P, Krejci L, Van Komen S and Sehorn MG. (2003). J. Biol. Chem., 278, 42729–42732. | Article | PubMed | ISI | ChemPort |
36. Tabara H, Grishok A and Mello CC. (1998). Science, 282, 430–431. | Article | PubMed | ISI | ChemPort |
37. Tan PB and Kim SK. (1999). Trends Genet., 15, 145–149. | Article | PubMed | ISI | ChemPort |
38. Tijsterman M, Pothof J and Plasterk RH. (2002). Genetics, 161, 651–660.
39. Timmons L and Fire A. (1998). Nature, 395, 854. | Article | PubMed | ISI | ChemPort |
40. Van Haaften et al., Personal communication.
41. von Zelewsky T, Palladino F, Brunschwig K, Tobler H, Hajnal A and Muller F. (2000). Development, 127, 5277–5284. | PubMed | ISI | ChemPort |
42. Wassarman DA, Therrien M and Rubin GM. (1995). Curr. Opin. Genet. Dev., 5, 44–50. | Article | PubMed | ChemPort |
43. Wong AK, Pero R, Ormonde PA, Tavtigian SV and Bartel PL. (1997). J. Biol. Chem., 272, 31941–31944. | Article | PubMed | ISI | ChemPort |

Acknowledgements

We are extremely grateful to Peter Askjaer, Guillaume Lettre, and David Welchman for providing pictures for Figures 1A, B, 1C–F, and 1G–H, respectively. We also thank David Rivers, Nathalie Le Bot, and Verena Wolfram for their helpful comments on the manuscript. GP is supported by the Wellcome Trust and JA is a Wellcome Trust Senior Research Fellow (054523).