Biology Reference
In-Depth Information
machineries regulating early embryogenesis in the worm
[87,89,91]
, and these are described in more detail below.
Second, coverage is far easier to assess with RNAi than
with random mutagenesis screens, with their unknown
biases. The typical RNAi screen uses the Ahringer library
[63,66]
which covers ~80% of predicted genes; most
screens have an estimated false negative rate of a little
under 50%
[66]
. Thus coverage can be readily assessed,
making systems analysis far easier.
Identifying synthetic genetic interactions by classical
genetics is hard without a starting point: if two independent
genes must both be mutated in order for a phenotype to
become detectable, this is extremely improbable in any
screen unless they are chromosomal neighbors, for example
lin-15A and lin-15B
[36]
or (in an example from the fly) rpr,
hid and grim
[92
and transcription factors
[63,66]
. Second, one can also
begin to see evidence of the role of domain innovation in
the evolution of novel biological processes in a way that is
statistically relevant and not anecdotal. For example, genes
that have RNAi phenotypes that only affect complex
multicellular processes such as tissue organization, body
size regulation, or the coordination of movement are
enriched for domains that are only present in animal
genomes, suggesting that the evolution of complex
behavior and multicellular architecture has in part been
driven by the evolution of novel protein folds and functions,
rather than simply through the rewiring of existing
components and domains
[66]
. Third, systematically
derived phenotypic datasets such as those from RNAi
screens can be easily compared to analogous datasets in
other organisms: for example, over 60% of the 1:1 ortho-
logs of S. cerevisiae 'essential' genes have lethal or sterile
RNAi phenotypes in the worm
[98]
, showing that the core
functions of the eukaryotic cell are highly conserved and
carried out by identical machineries across long evolu-
tionary periods. RNAi-based genetic interaction screens in
the worm indicated that ~4% of genetic interactions in the
worm are non-additive
[39]
; similar analysis of synthetic
lethal interactions in yeast
[99]
gives a similar proportion of
non-additive interactions, suggesting that the underlying
architectures of genetic interaction networks are main-
tained over long evolutionary timescales.
One can also use the data deriving from systematic
RNAi screens to re-examine some of the basic implications
of the neutral theory of evolution and natural selection.
Some genes have extremely strong RNAi phenotypes:
worms with these genes knocked down either die or are
completely sterile. Others have much weaker defects, such
as mild deviations from normal vulval development or
subtle body length changes. Finally, there are a large
number of genes that have no detected phenotype in any of
the 50 genome-scale screens to date. Both intuitively and
based on theoretical predictions, it seems reasonable that
the stronger the loss-of-function phenotype of a gene
(that is, the stronger the effect of inheritance of a mutation
in that gene), the stronger the strength of negative selection
acting on that gene over evolutionary timescales. With the
systematic datasets deriving from RNAi screens, one can
begin to examine these predictions. At a crude level, there
are indications that this is true: a higher proportion of genes
with strong (i.e., lethal or sterile) RNAi phenotypes have
a yeast ortholog than genes with weak or no detectable
RNAi phenotypes
[63,66]
. Looking at more sensitive tests
for strength of negative selection, one can indeed see
a difference in ka/ks for genes with sterile phenotypes
compared to other genes
[100]
. Most recently, assays were
developed to examine the effects of targeting genes by
RNAi on the overall reproductive fitness of a population of
worms
[101]
. Strikingly, these studies indicate that the
94]
. By RNAi, however, this raises few
problems aside from those of scale. Genetic interactions
can be mapped de novo either by carrying out an RNAi
screen in a mutant background and comparing the pheno-
types observed with those seen in wild-type worms
[39,95,96]
, or by carrying out combinatorial RNAi
[97]
in
an RNAi hypersensitive strain like rrf-3
[69]
or lin-35
[74,79]
. This not only allows the direct and unbiased
identification of synthetic genetic interactions, but can also
address directly the issue of specificity of any genetic
interaction detected. For example, we previously inves-
tigated genetic interactions between genes involved in
C. elegans signal transduction and transcriptional networks
and identified a set of ~350 genetic interactions out of
~65 000 tested pairwise interactions
[39]
. Unlike in clas-
sical genetic screens, we know immediately which inter-
actions yielded synthetic phenotypes without any need for
cloning, we know the number of tested interactions
precisely (and thus can estimate the rate of non-additive
genetic interactions in vivo), and we can also estimate the
specificity of any interaction with a component or pathway
tested. Whereas most genes show synthetic genetic inter-
actions with a single signaling pathway, a handful of genes,
all encoding chromatin regulators, have synthetic interac-
tions with every signaling pathway tested. Similar 'hubs' in
the genetic networks in humans may have key disease
importance, as inherited variation in such hubs would be
likely to affect the phenotypic outcome of a large propor-
tion of other personal variant alleles.
Taken together, then, what have these RNAi screens
taught us about gene function in the worm in a way that
would have been inaccessible from classical genetics?
First, it is possible to examine the relationship between the
molecular functions of a gene and its in vivo role in a way
that was never previously possible. Crude trends are
immediately obvious
e
almost half the genes giving sterile
phenotypes are involved in protein translation, whereas
genes that only affect more complex post-embryonic tissue
development are highly enriched for signaling molecules
e
