358
Genes are expressed and regulated in a structured cascade of events, thus form-
ing a network with well-defined switching patterns of activation and inhibition. The
complexity and differences observed during development are a function of the total
number of gene patterns (rather than the total number of genes), and the interactions
between those genes. As the number of gene patterns increase, the likelihood of
interaction between protein products increase as well, therefore increasing the com-
plexity of the network of gene interactions.
The first step in understanding a GRN is to characterize the regulation of gene pat-
terns. The transcription of genes involved in morphogenesis is regulated by noncoding
DNA regions known as cis-regulatory elements (CREs). CREs perform their function
by acting as binding sites for transcription factors. A major development function of
CREs is to respond to spatially diverse driver inputs so that the output pattern differs
from the individual inputs, and thus CREs allow different genes to respond in multiple
ways to similar regulatory states (Davidson 2010 ). A dramatic example of this is the
comparison of Drosophila melanogaster with its approximately 14,000 genes, versus
the significantly morphologically less complex Caenorhabditis elegans, which never-
theless has a greater number of genes, approximately 20,000. The difference is that D.
melanogaster has between two and three CREs per gene, producing about 50,000
gene expression patterns, while C. elegans has between one and two CREs per gene,
producing about 40,000 gene expression patterns (Markstein and Levine 2002 ).
Experimental validation of CREs is laborious process; the alternative is to use
computational methods. An approach to determine CREs computationally is through
the identification of clusters of high-affinity motifs within a given window of a deter-
mined gene; for example, in the case of the zen gene, the high-affinity consensus
sequences representing 106 binding sites GGGWWWWCCM and
GGGWDWWWCCM (W = A or T, M = C or A, D = A or T or G) were found within
a window of 400–660 bp of zen. A survey of the Drosophila genome informed 15
novel clusters in addition to the zen cluster. This approach has been successfully used
to study regulation of other genes (Markstein and Levine 2002 ; Markstein et al. 2002 ).
Given a set of gene patterns, as informed by CREs regulation, the value of under-
standing a GRN is to reveal the sequence of events that originate complexity. This
knowledge can be used to formulate hypothesis about the outcomes of perturbations
to the gene network, and even to control development to reach a desired outcome (or
avoid an undesirable one). The study of GRNs during development is confounded
by the large number of changes that occur simultaneously. Traditional network
reconstruction approaches can only elucidate structure in the presence of a single or
a few simultaneous changes (Laubenbacher and Stigler 2004 ; Kholodenko et al.
2005 ); when successful, these methods can represent a GRN as a systems of ordi-
nary differential equations (ODEs). However, the use of ODEs is often thwarted by
lack of data to determine kinetic parameters, and absence of mechanistic details; an
alternative is to use Boolean networks, i.e., a set of nodes with binary states deter-
mined by other nodes in the network (Wang et al. 2012 ).
In the case of vertebrate embryos, the GRN for the establishment of the germ layers
is complex. Morphogens are extracellular signal ligands propagated via diffusion, cells
are development units, and the activation of regulatory responses is a function of the
intensity of signaling thus activating different genes in different parts of the embryo.
W. Tseng et al.