Transcriptional Regulators and the Evolution of Plant Form

The modern or neo-Darwinian theory of evolution was formulated well over half a century ago at a time when genes were abstractions and development was understood only as the transition between a series of phenotypic states. Since that time, our knowledge of the gene, the diversity of types of genes, the interactions among genes, and the organization of genes into the networks and hierarchies that regulate development has grown enormously. Although still incomplete, this burgeoning knowledge of the molecular mechanisms that regulate ontogeny compels a continual reassessment of the question: How did the diversity of forms of life on earth evolve? Darwin resolved this question at the ecological level by explaining how natural selection would favor the “fittest” forms within a specific environmental context. The neo-Darwinian synthesis resolved this question at the genetic level by showing how the mechanisms of mutation, selection, and drift produce the directional changes in gene frequencies that underlie the evolution of novel forms (Huxley, 1963). The challenge facing the current generation of evolutionary biologists is to resolve this question at the molecular and developmental levels. In this article, we explore how recent advances in the understanding of gene structure, function, and interactions have fueled the creation of new ideas on the molecular mechanisms underlying the evolution of plant form. We summarize evidence for our view that change in the cis-regulatory elements of transcriptional regulators provides a predominant mechanism for the generation of novel phenotypes. We end with a discussion of the opportunities for testing this idea empirically by establishing causal linkages between specific changes in the molecular structure of genes and the production of novel phenotypes in naturally occurring plant species. After the synthesis of genetics and evolution in the 1920s, several authors hypothesized that genes regulating the timing or rate of development must be required for the evolution of form (Haldane, 1932; Huxley, 1932; Goldschmidt, 1940). Exactly how genes could regulate development was a mystery. The decoding of DNA and illumination of the regulation of gene expression (Jacob and Monod, 1961) set the stage for another round of hypothesis development about the underlying molecular mechanisms for both physiological and morphological evolution. Around 1970, Britten and Davidson (1969, 1971) presented a compelling theory for the molecular basis of evolutionary change. Although their terminology and some of the specifics of their model have not survived, the logic and the general outlines of their model have. Britten and Davidson distinguished between “producer” (or structural) genes and “receptor” (or regulator) genes, which are now called coding regions and promoters, respectively. They further identified “integrator” genes whose product, “activator RNA,” binds to receptor genes and thereby regulates the transcription of the producer gene. Integrator genes are equivalent in spirit to transcriptional regulators. The model proposed a hierarchical structure, with single integrators providing for the coordinate expression of multiple downstream producers. With this model of gene regulation, Britten and Davidson speculated that evolution would principally involve changes in the “regulatory genes,” or what we now call cis-regulatory elements. The concept that the regulatory regions of genes have a predominant role in evolution won early but inconclusive support. Ohno (1971) argued that because the same set of structural genes are apt to be involved in the formation of the digits of a horse and a human, functionally and anatomically very different structures, changes in the regulation of these genes can best explain the differences in these structures. Wilson and his colleagues (Wilson et al., 1974; Wilson, 1976) noted the uncoupling of protein and morphological evolution and argued that extreme morphological divergence in the absence of a corresponding level of protein divergence could be reconciled if regulatory gene mutations (i.e., in cis-regulatory regions) rather than structural gene mutations (i.e., in protein coding regions) were the principal mechanism of evolutionary change. For example, these authors note that there is relatively little protein variation among mammals, despite considerable anatomical diversity, whereas considerable protein variation exists among frogs with little corresponding morphological diversity. The neutral theory of molecular evolution, as formulated by Kimura (1968, 1983), also proposes that the vast majority of protein evolution does not affect the morphology of organisms. Because changes in morphology are often adaptive, neutral theory therefore implies that most protein evolution will not lead to adaptive changes in the phenotype. More recently, Endo et al. (1996) followed up on Kimura's conclusions by performing a global search for the signature of positive selection on proteins. Their logic was that for neutral evolution, the ratio of nonsynonymous substitutions (per nonsynonymous site) to synonymous substitutions (per synonymous site) will equal one (or less if deleterious amino acid changes are selected against). However, if positive selection is causing the fixation of advantageous amino acid substitutions, this ratio should be greater than one. For 3595 groups of proteins examined, only 17 showed evidence for positive selection. Again, this result is concordant with the view that the portion of morphological evolution that is adaptive in nature is not driven by changes in protein function. This early evidence for the importance or predominance of regulatory evolution over protein evolution is at best suggestive. For example, although the level of protein differentiation between humans and chimpanzees may not be commensurate with their behavioral and morphological differences, a small number of effective amino acid substitutions could be sufficient to control these differences. In fact, under the model that protein evolution explains most morphological evolution, a reasonable expectation would be that a small number of effective changes in protein function should exist amid a much larger number of ineffective (or neutral) changes. If most proteins are not evolving in a positive way under the guidance of natural selection, then their functions are apt to be conserved over long periods of evolutionary time. This, in fact, appears to be the case when one considers the common observation that a protein from one species can often complement a mutant or produce a similar phenotype in a second organism, even when the two species have been separated for long evolutionary periods. One of the most striking cases is the ability of the mouse Small-eye (Pax-6) gene, which controls eye formation, to induce ectopic eye formation in Drosophila, indicating that the function of this protein has been conserved for the 500 million years since the divergence of arthropods and vertebrates (Halder et al., 1995). In another example, the chicken gHoxb-1 gene can substitute for a Drosophila Hox mutant (lab), which functions in head involution, when linked to the appropriate Drosophila regulatory sequences (Lutz et al., 1996). The Drosophila paired (prd) gene can rescue a gooseberry (gsb) mutant when the gsb control region is fused with prd coding regions, despite the fact that these two homeobox genes serve distinct roles in Drosophila development (Li and Noll, 1994). Similarly, the cdc2 homolog of maize complements the cdc28 mutant of Saccharomyces, despite a divergence time of at least 600 million years (Colasanti et al., 1991). Finally, the DEF protein of snapdragon fully complements an ap3-1 mutation and partially complements the ap3-3 mutation of Arabidopsis (Irish and Yamamoto, 1995). Evidence for the conservation of protein function over long evolutionary periods notwithstanding, proteins do evolve in function, and change in protein function is a crucial component of evolution. Indeed, there are a few cases in which the functional evolution of proteins involved in morphogenesis has been documented. For example, among members of the MADS box gene family of Arabidopsis, differences in three separate regions of the protein are involved in conferring the functional specificities that distinguish family members (Krizek and Meyerowitz, 1996). In this case, homologous proteins cannot substitute for other members of the same protein family. Because these different functions are required for normal flower development, one might infer that floral evolution has involved changes in protein function. When one observes full complementation of function by a protein from another species, the interpretation is clear. Function has been conserved. When partial or noncomplementation is observed, the interpretation is more complex. One possibility is that the protein has specifically evolved to interact with a different set of partners or bind to a different set of downstream genes in ways that alter the phenotype. A second possibility is that the protein has changed in “specificity” but not in “function as related to the phenotype.” For example, a pair of interacting proteins in one species may have coevolved on a particular path, whereas in a second species the orthologous pair of proteins has evolved independently along a different path. These pairs may fulfill the same function (e.g., they may activate the same set of downstream genes) in both species, but the individual proteins of one species may not be able to complement the “function” of their orthologs in the other species because of changes that specify the interaction with the particular partner. Function is contextual, and one cannot infer a change in function (as related to the phenotype) from failure of a protein to complement its ortholog in another species. The view that evolution often proceeds via changes in the spatial–temporal patterns of gene expression has been embraced by diverse authors (see Raff, 1996). Evolutionary tinkering with promoters and enhancers could readily generate variation in the pattern of gene expression and thereby produce novel phenotypes upon which natural selection could act. The fact that plants readily tolerate insertions, rearrangements, and other forms of sequence evolution in cis-regulatory regions implies that these would be likely “hot spots” for the generation of novelty (Wessler et al., 1995). Two lines of evidence support the view that changes in gene expression are important evolutionary mechanisms. The first line of evidence comes from studies of natural species that reveal correlations between the spatial patterns of gene expression and phenotype. For example, the formation of eye spots on butterfly wings is correlated with the domains of Distal-less expression (Brakefield et al., 1996), so a change in the spatial pattern of Distal-less expression could confer a change in the size and distribution of eye spots. Comparison of Hox gene expression in primitive wingless and advanced winged insects suggests that genes involved in wing formation acquired cis-regulatory elements that brought them under the negative control of Hox genes in some body segments (Carroll et al., 1995). Changes in Hox gene expression patterns are also associated with the transformation of thoracic limbs into feeding appendages in crustaceans (Averof and Patel, 1997). In tomato, expression of a homeobox-containing gene in the leaf primordia is associated with the formation of compound rather than simple leaves (Hareven et al., 1996). There are also several cases in the evolution of physiological traits in which specific adaptations are correlated with changes in the pattern of gene expression (Dudareva et al., 1996; Sinha and Kellogg, 1996). For some physiological traits, shifts in gene expression have been mapped to 5′ regulatory regions (Schulte et al., 1997; Stockhaus et al., 1997). It seems reasonable that morphological evolution could occur by the same mechanism. A second line of evidence comes from mutations that reveal how rearrangements in the nontranscribed flanking regions of genes can produce novel patterns of gene expression and/or novel phenotypes. In tomato, Chen et al. (1997) demonstrated how a chimeric gene resulting from the insertion of a duplicate copy of a metabolic enzyme into the 5′ end of the homeobox-containing gene LeT6 resulted in overexpression of the homeodomain protein and the conversion of unipinnately compound leaves into three- or fourfold pinnately compound leaves. Thus, changes in regulation associated with alterations of 5′ regulatory sequences can produce novel phenotypes. More importantly, the novel phenotype is not simply a monstrous disruption of normal development but a well-integrated form of the pinnately compound leaf not uncommon among flowering plants. In another example, transposable element–induced alterations in the promoter of the maize alcohol dehydrogenase gene resulted in quantitative, organ-specific changes in its expression (Kloeckener-Gruissem and Freeling, 1995). Modular organization of promoters may contribute to their evolutionary flexibility (Kirchhamer et al., 1996). Distinct promoter elements that drive expression in different cell, tissue, or organ types can be added in a stepwise fashion during evolution without the disruption of preexisting promoter elements. Britten and Davidson's model for regulatory evolution was developed at a time when little was known about the diversity of gene functions and the interactions among genes. Although this knowledge is still incomplete, a far better catalog of the components is now available as well as substantial, if imperfect, knowledge of the interactions between these components. Some systems, such as vulval development in Caenorhabditis elegans and the pheromone response pathway in yeast, are understood at impressive levels of detail (Kayne and Sternberg, 1995; Schultz et al., 1995). Thus, one can now ask: Are some classes of proteins more apt to contribute to the evolution of form than are others? In broad outline, developmental pathways are composed of individual modules that are arranged in a semihierarchical system (Raff, 1996). We define individual modules as consisting of signal(s), transducers, transcriptional regulator(s), and targets (Figure 1). The signals may be environmental (e.g., photoperiod) or internal (e.g., auxin). The transducers compose the signal transduction pathway and include those gene products that participate in signal production, perception, transmission, and modification (e.g., ligands, receptor kinases, phytochromes, and proteins/enzymes involved in hormone synthesis/deactivation). Transcriptional regulators would be the genes involved specifically in the regulation of gene expression through direct binding to cis-regulatory elements. Finally, target genes can be any type of gene. For modules placed near the endpoints of development, the targets are apt to be genes whose actions do not affect other genes directly (e.g., metabolic enzymes). For modules higher up in the cascade, the targets would include genes involved in signaling and in transcriptional regulation. Viewed this way, plant development is a cascade of events that begins with the internal signals that initiate embryogenesis, followed by continual external and internal input directing the orderly activation of the hierarchically arranged modules. The modules themselves are visible in the processes, cell types, tissues, organs, and organ systems that we see. For example, one module may regulate trichome formation, another petal development, and a third seed maturation. Simplified Representation of the Interactions among Genes in a Developmental Pathway. Either an internal or external signal activates the receptor. The signal is transmitted by a transducer gene, which turns on the transcriptional regulators. The transcriptional regulators activate the target genes to produce a specific phenotype. As diagrammed, changes in the ligand, receptor, transducers, or target genes will be pleiotropic, altering multiple phenotypes (A, B, C, and D). However, alterations in a transcriptional regulator would affect only one phenotype. The yellow area represents a single developmental module. We compiled a list of cloned genes from Arabidopsis with mutant phenotypes that altered the morphology of the plant. We asked a colleague to categorize the severity of the mutant phenotype as (1) affecting a single organ or organ system (not pleiotropic), (2) affecting several related organ systems (weakly pleiotropic), (3) affecting multiple organ systems (moderately pleiotropic), or (4) strongly disrupting overall development (strongly pleiotropic). We also categorized the genes as transcriptional regulators or signaling system genes. Tallying the data revealed (Figure 2) that mutations in transcriptional regulators typically have less severe phenotypic effects (mostly classes 1 and 2) than do mutations in signaling genes (all classes 3 and the transcriptional regulators should be more to evolutionary modification because their would be less likely to have a on Genes involved in on the other to be into diverse of development, that alterations in their function and/or expression would be to The that transcriptional regulators a role in evolution has been by one of and appears to the of For systems, most evolutionary has on transcriptional regulators and 1995). For the MADS transcriptional regulators have been the of evolutionary studies et al., 1995; et al., 1997). The for this are clear. Transcriptional regulators as between phenotypic that duplicate or those that distinguish natural species or higher level Transcriptional regulators could as evolutionary by providing for the coordinate expression of their targets in new spatial or patterns as required to generate functional and phenotypes et al., et al., 1996; et al., 1996). of Arabidopsis Genes with genes to have broad whereas transcriptional regulators are more The structure of two known developmental pathways provides some support for our (1) that transcriptional regulators are apt to have less mutant phenotypes than their signaling genes and (2) that signaling pathways to be among multiple developmental modules. et al. (1996) the role of signaling in the of floral in These authors proposed that a transcription to the of floral is a target of for a mutation can a from floral to under some of this as does a mutation at a signaling gene whose function is to the signal transduction Their that which has of which has more If nature should to with floral would a more target than does The fact that mutations a of floral also that is only one of multiple developmental modules that to the A that transcription are apt to have more phenotypic effects than their signaling genes comes from the signal transduction pathway in elegans (Kayne and Sternberg, 1995; 1997). The pathway vulval development, development, and other The pathway is via a and a receptor and the transcription and are known to be by the in genes at the of this such as or are and produce multiple developmental However, with in the transcription are and the specific of Similarly, are also but or with Thus, multiple developmental modules to the pathway so that mutations affecting are pleiotropic, whereas the transcription by the pathway have more We have two molecular mechanisms by which plant form may (1) changes in cis-regulatory elements of genes and (2) changes in transcriptional regulators. We to these two mechanisms and that in the cis-regulatory regions of transcriptional regulators a predominant for the evolution of novel This model has a over a number of cases in which the molecular mechanisms of new have been cis-regulatory changes in transcriptional regulators will be How can this model be of the molecular mechanisms underlying the morphological differences between species is not a simple provides one Distinct pairs of species that in morphological traits are common among flowering plants (Figure The genes these differences can be in and mapped molecular the most systems for this are among species related to model such as or gene between model and related species would one to genes morphological differences between natural species to cloned genes in the model species 1995). a gene is or genetic could be to that controls the and to the molecular underlying the phenotypic differences. This will be the sequence of the of the model species is known 1997). among the of a of Two of the flowering there are of and species. One striking is to The the and their The has an phenotype. to from the and and and phenotypes are has identified a small number of of that control these phenotypic differences. by from et al., 1995). When Darwin proposed the theory of evolution by natural selection, genetics did not and although was understood that the mechanism of and therefore the mechanism of evolution was a mystery. The of genetics in the early of this century identified the gene as the of and the for modern evolutionary theory on how the of mutation, selection, and drift produce evolutionary change via their on genes. However, still from the was an understanding of genes, their structure, and the diversity of their was an understanding of the interactions among genes and their organization into the networks and hierarchies that control the several of these elements in the have been the for both more of morphological evolution and of In this we have presented some that we and have from the modern understanding of gene structure, function, and These include (1) that development is in (2) that protein function to be conserved over long evolutionary (3) that plant promoters are in (4) that signaling genes to have strongly and that transcriptional regulators as in plant The we from these is that the evolution of plant form will be most readily by changes in the cis-regulatory regions of transcriptional regulators. The challenge now is to this model by the molecular mechanisms underlying specific events in the evolution of plant form among natural species. and plants will the We for his of Arabidopsis and an for on the and for to related to this is by from the and to and a from the to