is ~1% per gene per million years, and because small fragments of DNA are tandemly duplicated at much higher rates than entire genes (Katju and Lynch, 2003), variation in the regulatory modules of genes must arise at least as rapidly as single-nucleotide polymorphisms. However, because of the mutational cost of allelic complexity, the likelihood of completion of semineutral modularization processes becomes negligible once 1/Ng becomes smaller than the excess mutational burden (Force et al., 2005). Thus, contrary to popular belief, natural selection may not only be an insufficient mechanism for the origin of genetic modularity, but population-genetic environments that maximize the efficiency of natural selection may actually promote the opposite situation, alleles under unified transcriptional control. Under this view, the reductions in Ng that likely accompanied both the origin of eukaryotes and the emergence of the animal and land-plant lineages may have played pivotal roles in the origin of modular gene architectures on which further developmental complexity was built.
Despite the initial invariance of phenotypic expression patterns during this type of gene-architectural repatterning, the emergence of independently mutable subfunctions in modularized alleles can contribute to adaptive evolution in significant ways. For example, if the ancestral allele under unified control was subject to pleiotropic constraints associated with shared regulatory regions, modularization may open up previously inaccessible evolutionary pathways. Relief from pleiotropy can be even further facilitated following the duplication of the entire gene (bottom of Fig. 5.2), as complementary degenerative mutations partition cellular tasks among paralogous copies (Force et al., 1999). This process of subfunctionalization is known to be a frequent fate of duplicate genes in multicellular species (Prince and Pickett, 2002; Lynch, 2007), and theory suggests that it too is most likely to occur in populations with small Ng, again because of the mutational burden of distributing a fixed number of subfunctions over multiple genes (Lynch et al., 2001). Thus, the joint operation of both processes (the emergence of gene subfunctions and their subsequent partitioning among paralogs) in the small to moderate population-size environment that exists in multicellular species provides a powerful mechanism for the passive remodeling of entire developmental genetic pathways (Lynch, 2007).
Another peculiar aspect of developmental pathways that has defied explanation is their seemingly baroque structure (Wilkins, 2002, 2005). It is common for linear pathways to consist of a series of genes whose products are essential to the activation/deactivation of the next downstream member, with only the expression of the final component in the series having an immediate phenotypic effect. For example, the product of gene D may be necessary to turn on gene C, whose product turns on gene B, whose