Biology Reference
In-Depth Information
concentration between different cells. Strong positive
autoregulation can result in a bimodal distribution of TF
expression in different cells where either high or low TF
expression can be maintained. The latter can be useful to
generate populations of cells with different properties
during development or in response to different environ-
mental (stress) cues. Strikingly, it was observed in early
constructions of the E. coli GRN that ~40% of TFs nega-
tively regulate their own expression
[134]
. However, only
6% autoactivate, suggesting that positive autoregulation is
much less beneficial to the organism than autorepression.
The E. coli GRN is still incomplete, and it is not clear which
regulatory interactions correspond to physical interactions.
Thus, it is possible that some autoregulation involves
multiple TFs, rather than being caused by direct binding of
TFs to their own promoter. Surprisingly, only 10 of 141
yeast TFs tested by ChIP bind their own promoter region
(7%), and it is not clear how many of these events have
a regulatory consequence
[100,101]
. This observation
suggests that autoregulation may be more prevalent in
prokaryotes, whereas more complex circuitry has evolved in
eukaryotes. Comprehensive physical and regulatory data-
sets in more complex multicellular organisms will provide
further insights into the prevalence of autoregulation in the
different kingdoms of life.
TF cascades are the second type of building block that
occurs in GRNs. They are defined as the sequential regu-
lation of TFs by other TFs, followed by the regulation of
downstream, non-TF target genes. Numerous TF cascades
have been discovered in many systems, for instance in the
context of animal development
[135]
. TF cascades function
through a time delay because each step depends on the
accumulation of sufficient amounts of the upstream TF. It is
intuitive that this can be useful in development and
differentiation, where new cellular states need to be
sequentially acquired
[136]
. Whereas TF cascades do not
occur frequently in E. coli, the yeast GRN contains 188
potential cascades with lengths between 3 and 10 regulators
[101]
, suggesting that this gene regulatory mechanism is
more prevalent in eukaryotes. The vast body of TF cascades
that has been identified by more conventional studies
suggests that TF cascades may occur even more frequently
as complexity increases. Comprehensive data in more
complex metazoans and plants will no doubt shed further
light on TF cascades and their functions.
The third building block is called a single-input motif
and is defined as a set of genes that are bound or regulated
by the same TF (e.g., a gene module or gene battery). The
bifan and multi-input motifs are variations of the single-
input motif. A bifan is defined as two TFs binding or
regulating the same two target genes. A multi-input motif
corresponds to a gene battery that is co-regulated by
multiple TFs. In the latter, not every TF necessarily controls
each gene in the battery. This class of GRN circuits is
highly useful in development, when sets of genes need to be
turned on or off in cellular or organismal differentiation.
Feedforward loops are another important type of
regulatory circuit in GRNs. These are defined by the
regulation of a gene by two TFs, one of which also
regulates the other
[137]
. Each of the edges in feedforward
loops can be either activating or repressing. As a result,
there are eight possible types of feedforward loop. When
the overall outcome of the direct and indirect paths to the
downstream target gene is the same (repressing or acti-
vating), the loops are called coherent; when the signs of
direct and indirect edges are opposite, the loops are called
incoherent. There are four loops of each type. However,
each loop can behave differently, depending on the
underlying logic structure. For instance, in an 'AND-logic
gate' both TFs are required for expression of the down-
stream target gene, whereas in an 'OR-logic gate' one TF
is sufficient. Each type of feedforward loop can confer
a particular transcriptional response upon activation of the
most upstream TF by an outside signal or by another TF
[137]
. For instance, in the simplest feedforward loop
where all edges are activating, AND-logic provides
a delay in expression of the target gene because sufficient
amounts of both TFs need to accumulate, whereas OR-
logic provides a delay in the reduction of target gene
expression upon removal of the upstream signal because
both TFs need to be inactivated. Although all types of
feedforward loops can occur in GRNs, not all are a true
network motif; i.e., only some occur more frequently than
expected by chance
[130]
. Most frequently occurring in
E. coli and S. cerevisiae are the coherent feedforward loop
in which all edges are activating, and the incoherent
feedforward loops where the upstream TF activates both
the target gene and the downstream TF, and the latter
represses the downstream target
[101,138]
. It remains an
open question whether similar types of feedforward loop
are enriched in the GRNs of complex multicellular
systems and which type of logic (AND vs. OR) is most
prevalent.
The final circuit that occurs in GRNs is the feedback
loop. Feedback can involve a single TF (in the case of
autoregulation discussed above), two TFs, or more
complicated circuits comprising larger numbers of TFs.
An example of a three-node feedback loop is shown in
Figure 4.3
E. Interestingly, although feedback was found in
E. coli GRNs, it was not found to be enriched. Because
feedback is so abundant throughout biology, this suggests
that other types of biomolecule may be involved. Indeed,
both protein
e
protein interactions and microRNAs were
found to contribute
[106,139]
. However, since it has
become clear that different types of network building
blocks can be enriched in different species, transcriptional
feedback loops might also occur in the GRNs of more
complex systems such as animals and plants.
