Biology Reference
In-Depth Information
Innovations that involve new combinations of existing
system parts have many guises. Students of embryonic
development, for example, have coined the term co-option,
the use of an existing regulator or an existing regulatory
interaction for new purposes
[113]
. Examples include the
regulator Distal-less mentioned earlier. It is involved in the
development of insect legs and wings, but it has been co-
opted to form eyespots
[44,114]
. Distal-less does not act
alone in these processes. It is part of regulatory circuit that
involves other molecules, some of which may also have
changed their interactions and expression in helping form
a new body structure. One can view the co-option of Distal-
less as a special case of a more general principle, in which
new combinations of regulators and their interactions
specify new body parts.
In sum, evidence from three very different kinds of
systems can help answer several related questions about the
origins of new phenotypes. It can help us grasp how life can
preserve old phenotypes while exploring many new
phenotypes. It can help us understand how many evolu-
tionary innovations have originated multiple times in the
history of life. And it can help us appreciate that innovation
will generally involve combinations of old parts to achieve
new purposes. The fact that the properties described exist in
very different kinds of system suggests that they apply to
multiple different kinds of innovation. They are suitable to
form the basis of a general innovability theory.
extent robust in this sense
[39,115
123]
. This robustness
has been estimated experimentally in systems such as
proteins
e
through
random mutagenesis
experiments
[115
118]
, in metabolic networks through knockout
mutations of enzyme-coding genes, and in regulatory
circuits through circuit rewiring
[39,115
e
123]
. Computer
modeling confirms that such robustness is a generic feature
of these three system classes
[54,84,124]
. Typically,
between 10% and more than 50% of a genotype's neighbors
have the same phenotype as itself, depending on the system
and the individual genotype
[10]
. It can be shown mathe-
matically that this property is both necessary and sufficient
to bring forth genotype networks that are astronomically
large, and that extend far through genotype space
[10]
.
From this vantage point, one could argue that genotype
networks are a consequence of robustness. (Their diverse
phenotypic neighborhoods emerge from the fact that many
more phenotypes exist than the neighborhood of any one
genotype can contain
[10]
.)
These observations raise a further question. What is the
ultimate cause of this robustness? Although multiple
answers have been proposed, the current best candidate
emerges from the observation that living systems need to
operate in different environments
[119,125
e
129]
.The
notion of an environment should be broadly defined in this
context, and include the biotic, chemical, and physical
environment outside an organism, as well as inside its cells.
For example, it includes the changing chemical environ-
ments that provide nutrients to a metabolic network, the
different regions of a developing embryo in which a regu-
latory circuit is exposed to different chemical signals, and
the intracellular chemical environment that macromole-
cules need to operate in.
The role of changing environments for robustness has
been most thoroughly studied in the context of metabolic
networks
[56, 119,129
e
Robustness, Genotype Networks
and Environmental Change
A question so far left open is why genotype networks exist
in metabolism, regulatory circuits, and macromolecules. At
first sight this may seem difficult to answer, because these
system classes are so different. However, it can be shown
that this commonality emerges from a very simple property
that they share: the robustness of their phenotypes to
mutational changes in individual system parts.
In the genotype space framework such robustness can
be thought of as a property of individual genotypes.
Mutations often change any one genotype into one of its
neighbors. A loss-of-function mutation in an enzyme-
coding gene may eliminate one reaction from a metabolic
network and transform the network into one of its neigh-
bors; a mutation-changing regulatory DNA may eliminate
a transcription factor's binding to this DNA, and hence its
regulatory interaction with a target gene, transforming the
circuit into one of its neighbors; a nucleotide change in
a protein-coding gene often transforms the protein into one
of its neighbors. One way to quantify the robustness of
a genotype is through the proportion of its neighbors that
have the same phenotype as itself. Metabolic networks,
regulatory circuits and macromolecules are all to some
133]
. A free-living organism such
as E. coli, which encounters multiple different environment
containing different nutrients, can sustain life on dozens of
different nutrients. It also has a large metabolic network
that comprises more than 900 reactions. In any one such
environment, it is also robust to the removal of individual
reactions
[134]
. For example, more than 70% of its reac-
tions are dispensable in a minimal environment with
glucose as the sole carbon source. Such robustness is not
a peculiarity of the E. coli metabolic network: it is a general
property of metabolic networks that can sustain life in
multiple environments
[57,119]
. (Note that any reaction
that is dispensable in one environment may be essential in
a different environment
[119]
.) If E. coli lived for many
generations in an environment that did not vary in its
nutrient composition, its robustness would slowly disap-
pear. This is what happened in endosymbiotic organisms
such as Buchnera aphidicola, a relative of E. coli that has
lived for millions of years inside its host organism, an aphid
e
