<<

. 52
( 66 .)



>>

on this mechanism would be unreliable. Perturbation of the diffu-
sive ¬elds in the early Drosophila embryo, however, leaves the forma-
tion of even-skipped stripes remarkably unchanged (Lucchetta et al.,
2005).
If the striped expression of pair-rule genes in the ancestor of mod-
ern Drosophila was generated by a reaction--diffusion mechanism, un-
der the assumption that there was selective pressure to conserve body
architecture from one generation to the next, this inherently variable
developmental system would have been a prime candidate for canaliz-
ing evolution. The elaborate systems of multiple promoter elements
responsive to preexisting nonuniformly distributed molecular cues
264 BIOLOGICAL PHYSICS OF THE DEVELOPING EMBRYO


(e.g., maternal and gap-gene products), seen in Drosophila is therefore
not inconsistent with the origination of this pattern as a reaction--
diffusion process.

Evolution of developmental gene networks: the model
of Salazar-Ciudad et al.
We asked above why modern-day Drosophila does not use a reaction--
diffusion mechanism to produce its segments. In the light of the dis-
cussion in the previous paragraphs, we can consider the following
tentative answer: pattern-forming systems based on reaction--
diffusion are inherently unstable to environmental and random ge-
netic changes and would therefore, under the pressure of natural
selection, have been replaced, or at least reinforced, by more hierar-
chically organized genetic-control systems. A corollary of this hypo-
thesis is that the patterns produced by modern, highly evolved,
pattern-forming systems would be highly robust in the face of fur-
ther genetic change.
The ¬rst part of this evolutionary hypothesis has been examined
computationally in a simple physical model by Salazar-Ciudad and
coworkers (2001a). The model consists of a ¬xed number of commu-
nicating cells in a line or of nuclei arranged in a row within a syn-
cytium. Each nucleus has the same genome (i.e., the same set of genes)
and the same epigenetic system (i.e., the same activating and inhibitory
relationships among these genes). The genes in these networks specify
either receptors or transcription factors that act within the cells or
syncytial nuclei that produce them or paracrine factors that diffuse
between cells or syncytial nuclei (Fig. 10.5). These genes interact with
each other according to a set of simple rules that embody unidirec-
tional interactions in which an upstream gene activates a downstream
one, like those represented in Figs. 10.2 and 10.3, as well as recipro-
cal interactions in which genes feed back (via their products) on each
other™s activities (Fig. 10.5). This formalism was based on a similar one
devised by Reinitz et al. (1995), who considered the speci¬c problem
of segmentation in the Drosophila embryo.
The effect of gene i on gene j (i, j = 1, 2, . . . , N; N is the total num-
ber of genes in the model) is characterized by the parameter Wi j (the
gene--gene coupling constant). These constants are formal represen-
tations of cases in which a gene product directly affects another™s
activity by binding to or chemically modifying it or a gene changes
the amount of another gene™s product by acting as a transcription
factor (as in the Keller model described in Chapter 3). In general the
Wi j are not symmetrical, Wi j = W ji : the effect of gene i on gene j may
be different from the effect of gene j on gene i and could be zero
(representing no effect). Some of the genes specify diffusible factors
and these factors are assigned diffusion constants.
One may ask whether a system of this sort, with particular values
of the gene--gene coupling and the diffusion coef¬cients, can form
a spatial pattern of differentiated cells. Salazar-Ciudad and cowork-
ers performed simulations on systems containing 25 nuclei and a
10 EVOLUTION OF DEVELOPMENTAL MECHANISMS 265




Fig. 10.5 Schematic illustration of the gene“gene interactions in the model of
Salazar-Ciudad et al. (2001a). A line of cells is represented at the top. (The model
equvralently represents a row of nuclei in a syncytium.) Below, types of genes are
illustrated. Genes whose products act solely on or within the cells that produce them
(r, receptors; f, transcription factors) are represented by squares; diffusible factors
(h, paracrine factors or hormones) that pass between cells and enable genes to affect
one another™s activities, are represented by circles. Activating and inhibitory interactions
are denoted by small arrows and by lines terminating in circles respectively. The
double-headed green arrows denote diffusion. (After Salazar-Ciudad et al., 2001a.)




variety of arbitrarily chosen but ¬xed parameters W. A pattern was
considered to arise if the system™s dynamics lead to a stable state in
which different nuclei stably expressed one or more of the genes at
different levels. The systems were isolated from external in¬‚uences
(zero-¬‚ux boundary conditions were used, that is, the boundaries of
the domain were impermeable to diffusible morphogens) and initial
conditions were set such that at t = 0 the levels of all gene products
had zero value except for that of an arbitrarily chosen gene, which
for the nucleus at the middle position had a nonzero value.
Isolated single nuclei, or isolated patches of contiguous nuclei
expressing a given gene, are the one-dimensional analogue of iso-
lated stripes of gene expression in a two-dimensional sheet of nu-
clei, such as those in the Drosophila embryo prior to cellularization
(Salazar-Ciudad et al., 2001a). It had earlier been determined that
the core mechanisms responsible for all stable patterns fell into two
non-overlapping topological categories (Salazar-Ciudad et al., 2000).
These are the hierarchical and emergent categories referred to ear-
lier in this chapter. Hierarchical mechanisms form patterns by virtue
of the unidirectional in¬‚uence of one gene on the next in an or-
dered succession, as in the ˜˜maternal gene product induces gap gene
product induces pair-rule gene” scheme described above for the early
stages of Drosophila segmentation. In contrast, in emergent mecha-
nisms reciprocal positive and negative feedback interactions give rise
266 BIOLOGICAL PHYSICS OF THE DEVELOPING EMBRYO


to the pattern. As noted above, emergent systems are equivalent to
self-organizing dynamical systems, as seen for the transcription-factor
networks discussed in Chapter 3 and the reaction--diffusion systems
discussed in Chapters 7 and 8.
This is not to imply that an entire embryonic patterning event --
the formation of a pair-rule stripe in Drosophila, for example -- must be
wholly hierarchical or emergent. As discussed below, many aspects of
development are ˜˜modular” -- decomposable into semi-autonomous
functional or structural units. What Salazar-Ciudad and coworkers
observed was that each patterning event in the classes of genetic net-
works they studied could be decomposed into modules that are unam-
biguously of one or the other topology (Salazar-Ciudad et al., 2000).
Regardless of whether a particular system is capable of giving rise
to a stable pattern, one may ask how its ability to do so would change
if it evolved in a fashion analogous to real developmental systems.
The evolution of organisms is re¬‚ected in permanent changes (not
just changes in expression) of their genes. In terms of the model of
Salazar-Ciudad and coworkers, evolution of a developmental system
occurs if its gene--gene coupling constants Wi j change. Since gene
change in biological systems is typically undirected, it is of inter-
est to determine whether these model systems gain or lose pattern-
forming capacity over successive ˜˜generations,” where a new gen-
eration consists of replicas of a set of developmental systems with
random changes introduced in the coupling constants Wi j . Salazar-
Ciudad and coworkers therefore used such systems to perform a large
number of computational ˜˜evolutionary experiments.” Here we will
describe a few of these that are relevant to the questions raised by
the evolution of segmentation in insects, discussed above.
In their computational studies of the evolution of developmen-
tal mechanisms the investigators found that an arbitrary emergent
network was much more likely than an arbitrary hierarchical net-
work to generate complex patterns, i.e., patterns with three or more
(one-dimensional) ˜˜stripes”. This was taken to suggest that the evolu-
tionary origination of complex forms, such as segmented body plans,
would have been more readily achieved in a world of organisms
in which self-organiging (e.g., multistable, reaction--diffusion, oscil-
latory) mechanisms were also present (Salazar-Ciudad et al., 2001a)
and not just mechanisms based on genetic hierarchies.
Following up on these observations, Salazar-Ciudad and coworkers
performed a set of computational studies on the evolution of devel-
opmental mechanisms after a pattern had arisen. First they identi-
¬ed emergent and hierarchical networks that produced a particu-
lar pattern, e.g., three ˜˜stripes” (Salazar-Ciudad et al., 2001a). (Note
that patterns themselves are neither emergent nor hierarchical --
these terms apply to the mechanisms that generate them.) They next
asked whether given networks would ˜˜breed true” phenotypically,
despite changes to their underlying circuitry. That is, would their
genetically altered ˜˜progeny” exhibit the same pattern as the unal-
tered version? Genetic alterations in these model systems consisted
of point mutations (i.e., changes in the value of a gene--gene coupling
10 EVOLUTION OF DEVELOPMENTAL MECHANISMS 267


constant), duplications, recombinations (i.e., the interchange of
coupling-constant values between pairs of genes), and the acquisition
of new interactions (i.e., a coupling constant that was initially equal
to zero was randomly assigned a small positive or negative value).
To evaluate the consequence of such alterations it was necessary
to de¬ne a metric of ˜˜distance” between different patterns. This was
done by specifying the state of each nucleus in a model syncytium in
terms of the value of the gene product forming a pattern. Two pat-
terns were considered to be equivalent if all nuclei in corresponding
positions were in the same state. The distance between two patterns
was one unit if they differed in the state of one of their similarly
positioned nuclei, and so forth. The degree of divergence between
patterns could then be quanti¬ed (Salazar-Ciudad et al., 2001a).
It was found that hierarchical networks were much less likely to
diverge from the original pattern (after undergoing simulated evo-
lution as described) than emergent networks (Salazar-Ciudad et al.,
2001a). That is to say, a given pattern would be more robust (and thus
evolutionarily more stable) under genetic mutation if it were gener-
ated by a hierarchical, rather than an emergent network. Occasionally
it was observed that networks that started out as emergent were con-
verted into hierarchical networks with the same number of stripes.
The results of Salazar-Ciudad and coworkers on how network topol-
ogy in¬‚uences evolutionary stability imply that these ˜˜converted” net-
works would continue to produce the original pattern in the face of
further genetic evolution. Recall that this is precisely the scenario
that was hypothesized above to have occurred during the evolution
of Drosophila segmentation (see also Salazar-Ciudad et al., 2001b).
Subject to caveats about what is obviously a highly schematic ana-
lysis, the possible implications of these computational experiments
for the evolution of segmentation in long-germ-band insects are the
following: (i) if the ancestral embryo indeed generated its seven-stripe
pair-rule protein patterns by a reaction--diffusion mechanism and (ii)
if this pattern were suf¬ciently well adapted to provide a premium on
breeding true then (iii) organisms resulting from genetic changes that
preserved the pattern but converted the underlying network from an
emergent one to a hierarchic one (as seen in present-day Drosophila)
would have come to represent an increasing proportion of the popu-
lation.
We can now return to the question raised at the beginning of
this chapter, what is it about the construction of organisms and em-
bryos that permits a useful and relatively coherent body of knowledge
to be generated that primarily considers causal chains acting at the
level of gene interactions? An answer can be suggested based on the
preceding discussion. Let us take pattern formation as an example,
since we have just been considering it in detail. (We could equally
well apply these ideas to cell differentiation or to the shaping of ep-
ithelial or mesenchymal tissues). A given pattern-forming mechanism
may have originated as a physical process -- the differential-adhesion-
based sorting of cells, the diffusion-based generation of nonunifor-
mity, reaction--diffusion coupling. But the more hierarchical genetic
268 BIOLOGICAL PHYSICS OF THE DEVELOPING EMBRYO


control acquired by this mechanism over the course of evolution,
the less its outcome (a cell pattern, in this case) will be based on
generic physical processes. Thus, hierarchically organized genetic
mechanisms, by a process of canalizing selection, may have become
superimposed upon, and in some cases have superseded, the original
pattern-forming and morphogenetic processes. The early phase of the
Drosophila segmentation process, in this view, is a genetic hierarchy
with only a vestige of its original self-organizational capability (i.e.,
the positive autoregulation of the pair-rule genes).
It is clear, however, from the earlier chapters of this book (which
addressed the biology of contemporary, not ancient, organisms) that
notwithstanding any tendency of genetic hierarchies to replace con-
ditional physical mechanisms, for many key developmental processes
physical mechanisms remain central. In the following section we
show that, in contrast with the hierarchical genetic programming
of the early stages of Drosophila segmentation, the later stages of this
developmental process, and their robust character, are unequivocally
emergent (as de¬ned above).

The Drosophila segment polarity network as a robust
developmental module: the analysis of von Dassow et al.
We conclude this chapter, and this book, with a discussion of the
establishment of segment polarity in Drosophila, a phenomenon that
touches on many themes previously discussed in this and the earlier
chapters: the dynamics of cell differentiation, the creation of tissue
interfaces, pattern formation based on juxtacrine and paracrine sig-
naling, and molecular circuits that evolved for maintaining develop-
mental robustness. This system and the model for it upon which we
focus elaborate upon an important theme in development, mentioned
brie¬‚y below, that of modularity, i.e., the capacity of a complex whole
to be decomposed into independently operable, but interacting, struc-
tural or functional units (˜˜modules”) (von Dassow and Munro, 1999;
Winther, 2001; Schlosser and Wagner, 2004).
There are clearly enormous differences in short-, intermediate-,
and long-germ-band insects in the processes and patterns of gene ex-
pression leading up to the point at which the segment primordia
are made asymmetric, or ˜˜polar,” by the expression of the transcrip-
tion factor Engrailed in a column of cells on the anterior side and
the morphogen Wingless in a column of cells on the posterior side of
each parasegment border. In contrast with the variability of the early
stages of segment formation in insects with respect to the expression
of maternal, gap, and, to some extent, pair-rule genes, the molecular
circuitry involved in generating the facing bands of engrailed (en) and
wingless (wg) expression is highly conserved (Patel, 1994).
von Dassow and coworkers inferred from these observations that
insect segmentation may be a functionally decomposable process. In
their words, ˜˜perhaps segmentation is modular, with each module
autonomously expressing a characteristic intrinsic behavior in re-
sponse to transient stimuli. If so, evolution could rearrange inputs to
10 EVOLUTION OF DEVELOPMENTAL MECHANISMS 269




en
EN en

<<

. 52
( 66 .)



>>