From worm genetic networks to complex human diseases

Preview:

Citation preview

NEWS AND V IEWS

effects than those found in heterogeneous stocks of mice. Consequently, experiments in those populations will probably require extremely large sample sizes for the success-ful analysis of complex traits.

The great similarity in the genetic archi-tecture of 101 different quantitative traits is remarkable. Most traits were found to be affected by a large number of QTLs, each explaining a small proportion of the phe-notypic variance. It was impressive to see, nonetheless, that Valdar et al. succeeded in identifying, on average, 75% of the additive genetic variance. Finding that about 75% of the genetic effects fall in the category of small but detectable effects is encouraging, at least for mouse studies.

It is clear, however, that the task of dis-secting the genetic basis of complex traits is by no means complete. Neither gene-gene nor gene-environment interactions were analyzed in the current study, yet their importance is increasingly being recognized. As Valdar et al. point out, they already have evidence that gene-environment interac-tions are at work and may contribute as much as the additive heritability to variation

in some phenotypes. This means that there is another genetic country still to map.

Getting closer to the underlying genes?Valdar et al. present a great step forward toward the identification of genes at each QTL, but additional challenges remain. Even on the statistical analysis front, we do not have a robust method to analyze data of the sort generated here. Valdar et al. had to develop an ad hoc solution (the boot-strap posterior probability, or BPP) to deal with the complex correlation structure of the data. With large-scale biology becom-ing more and more common, the need to develop new analytical tools will only increase.

Probably the most challenging problem of all is how best to move from fine map-ping to gene identification. Even at the high mapping resolution achieved by Valdar et al., the 95% confidence intervals of the QTL map locations contain, on average, 27 genes. Consequently, identifying the underlying causal genetic variants is not trivial. This challenge has been recognized, and some new resources and technologies, such as a

complete knockout collection of all genes in the mouse6,7, as well as other advances, may help bridge the gap5.

Whether we are getting closer in identify-ing the underlying genes at each QTL is a matter of perspective. Here I have stressed the difficulties and challenges, but looking back at the advances made in the last decade, I would say that yes, we are definitely getting closer. Furthermore, in a wild effort to specu-late about future progress (in the next decade or two), I believe that at least the detectable 75% of the genetic effects will be brought down to the level of single genes, thereby considerably elucidating the biological basis of complex traits and eventually (in the next century or two) revolutionizing science and medicine.

1. Valdar, W. et al. Nat. Genet. 38, 879–887 (2006).2. Darvasi, A. & Soller, M. Genetics 141, 1199–1207

(1995).3. Couzin, J. Science 309, 81 (2005).4. Weiss, K.M. & Terwilliger, J.D. Nat. Genet. 26, 151–

157 (2000).5. Flint, J., Valdar, W., Shifman, S. & Mott, R. Nat. Rev.

Genet. 6, 271–286 (2005).6. Austin, C.P. et al. Nat. Genet. 36, 921–924

(2004).7. Auwerx, J. et al. Nat. Genet. 36, 925–927 (2004).

From worm genetic networks to complex human diseasesHoward Bussey, Brenda Andrews & Charles Boone

Systematic mapping of genetic interactions for Caenorhabditis elegans genes involved in signaling pathways implicated in human disease reveals a network of 350 interactions. The topology of this network resembles that mapped previously in yeast, reinforcing the idea that similar networks may underlie the genetic basis of complex human disease.

Howard Bussey is in the Biology Department, McGill University, Montreal, Quebec, Canada, and Brenda Andrews and Charles Boone are in the Banting and Best Department of Medical Research and Department of Medical Genetics and Microbiology, University of Toronto, Toronto, Canada. e-mail: charlie.boone@utoronto.ca

Genes do not act in isolation; rather, par-ticular genes interact with one another to modulate cellular systems and generate specific phenotypes. Consequently, map-ping of genetic networks is central to our understanding of the function of biologi-cal systems and how they can go wrong (for example, in complex inherited dis-eases). The large-scale analysis of synthetic enhancement phenotypes using pairwise

combinations of defined gene deletion or knockdown alleles in model organisms should provide us with a global view of genes that buffer one another functionally and thereby map genetic interaction net-works1. However, moving from this realiza-tion to the experimental mapping of genetic interactions is no simple business; indeed, large-scale genetic interaction networks have so far been made only in the single-cell bud-ding yeast Saccharomyces cerevisiae. Now, a report by Lehner et al.2 on page 896 of this issue promises the extension of wholesale mapping of genetic interactions to an ani-mal, the nematode worm C. elegans.

Mapping worm genetic networksThe current work builds on earlier studies in C. elegans3 using the surrogate genetics

afforded by RNA interference (RNAi) tech-nology. Here, in the presence of the RNAi for a gene, the activity of that gene is reduced, often giving a phenocopy of what a mutant allele in the gene would display. The start-ing point was a collection of worms, each of which harbored a mutation within a differ-ent ‘query’ gene in a known pathway, such as that involving epidermal growth factor (EGF), which determines vulval develop-ment. Lehner et al. arrayed such ‘query’ mutant worms in 96-well chambers and fed each with bacterial RNAi feeding strains har-boring double-stranded RNA (dsRNA) for every member of a library of some ~1,750 genes known or suspected to be involved in biological processes or pathways associ-ated with various disease states. Feeding the worms with dsRNA-laced bacteria allows

862 VOLUME 38 | NUMBER 8 | AUGUST 2006 | NATURE GENETICS

©20

06 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

ureg

enet

ics

NEWS AND V IEWS

the RNAi to permeate the worm. The effi-cacy and simplicity of this RNAi delivery enables a high-throughput approach. After the worms’ growth and development, one can examine the progeny in a given well for an altered phenotype indicative of a pair-wise interaction between the query gene and the known RNAi target. By doing this systematically for ~31 different query genes related to the EGF and other signaling path-ways and ~1,750 library genes predicted to function in signal transduction, transcrip-tional regulation and chromatin remodel-ing, some ~65,000 pairwise combinations were tested and ~350 genetic interactions were scored.

Conservation of network propertiesWhat has emerged from all this work? Lehner et al. were able to recapitulate much that was known about EGF signaling, but they also identified and substantiated the role of previously unknown components of this developmental pathway. Importantly, the resulting genetic networks seem to resemble the topology of those found in yeast4,5, displaying a similar network com-plexity. In particular, ~0.6% of interactions gave a phenotype, and the network exhibited a so-called ‘scale-free’ structure, with most genes having few interactions but with an exponentially smaller group of genes hav-ing increasingly many interactions and forming major hubs (Fig. 1). Interestingly, in the study of one hub centered around chromatin remodeling, Lehner et al. point out that the consequence of a central

process being involved in genetic buffering is that it will be highly pleiotropic, affecting many seemingly unrelated processes. Such hub-like properties are known to be asso-ciated with genes coding for Hsp90 chap-erones6 and those encoding prefoldins5 or other essential proteins7, which are highly connected in the yeast network. The list of genes displaying these core buffering properties is likely to grow, and similar genes may be implicated in the genetics of numer-ous human diseases.

The C. elegans network is still small, con-taining just a few hundred of the expected hundreds of thousands of interactions among the ~20,000 worm genes. Thus, if EGF signaling is atypical, these findings could be biased and not reflect the worm network as a whole. However, compara-tive network analysis can help here: the topological similarity between the yeast and worm networks, despite different sets of query genes, engenders confidence that these early worm subnetworks are likely typical. Further, the work serves to validate yeast network studies, which, although more advanced, have sampled fewer than 10% of the predicted 200,000 genetic interactions in S. cerevisiae7.

Because many genes are highly conserved, the topology and even specific interactions within genetic networks may also be con-served. The analysis by Lehner et al. of genetic interactions within the context of complex developmental phenotypes is an important expansion of the field into ani-mal development, extending to phenotypes

where yeast cannot reach. This latest work is also very important because it strengthens the idea that not only are eukaryotic genes conserved, but so too are the fundamental properties of genetic interaction networks.

Comprehensive comparative geneticsThese findings make a strong case for more comprehensive, model system–based com-parative genetics. Indeed, the field of genetic interactions seems poised for a major role in the post-genomic era. For example, there is a compelling case for building a complete genetic interaction network in an organism such as yeast, even if this remains a tech-nically formidable undertaking. In other organisms, the technical hurdles are even higher. Although it has been widely cham-pioned as an alternative to conventional genetics, RNAi has so far proved difficult to use in a high-throughput way. The method-ological improvements used here by Lehner et al., while helping greatly, still seem far from capable of analyzing the 20,000 × 20,000 gene array needed to map the worm genetic interaction space. However, efforts to get global sets of worm mutants, another key resource necessary to do such work, are well under way by the C. elegans gene knockout consortium (http://celeganskoconsortium.omrf.org). Work to expand this type of ana-lysis to fruit flies, zebrafish, mice and human cells seems as inevitable as the value of the scientific rewards that await those who carry out such studies.

Complex genetic interactions may under-lie most inherited phenotypes, and the consequences of defects in such genetic buffering can contribute to the etiology of various human diseases8. However, mov-ing from the study of monogenic diseases to a systematic examination of the basis for complex inherited diseases remains a major unsolved problem. Comparative genetic net-work analysis using high-throughput RNAi methods for pairwise gene knockdown in model systems may provide the extended reach we need to predict candidates for genes involved in complex human diseases1,9.

1. Hartman, J.L., Garvik, B. & Hartwell, L. Science 291, 1001–1004 (2001).

2. Lehner, B., Crombie, C., Tischler, J. Fortunato, A. & Fraser, A.G. Nat. Genet. 38, 896–903 (2006).

3. Kamath, R.S. et al. Nature 421, 231–237 (2003).4. Tong, A.H. et al. Science 303, 808–813 (2004).5. Pan, X. et al. Cell 124, 1069–1081 (2006).6. Rutherford, S.L. & Lindquist, S. Nature 396, 336–

342 (1998).7. Davierwala, A.P. et al. Nat. Genet. 37, 1147–1152

(2005).8. Badano, J.L. & Katsanis, N. Nat. Rev. Genet. 3,

779–789 (2002).9. Hartwell, L.H., Szankasi, P., Roberts, C.J., Murray,

A.W. & Friend, S.H. Science 278, 1064–1068 (1997).

Figure 1 Most genes have few interactions, but an exponentially smaller group of genes have increasingly many interactions and form major hubs.

Num

ber

of g

enes

Number of interactions

Disease gene modifier?

NATURE GENETICS | VOLUME 38 | NUMBER 8 | AUGUST 2006 863

©20

06 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

ureg

enet

ics

Recommended