Evolvability of Animal Developmental Systems: Remarks on their Modularity and Robustness
Riddhi Benani
Department of Life Sciences Imperial College London, UK

Supervisor: Prof Armand Leroi June 11, 2012
Abstract The ultimate aim of Evolutionary developmental biology (Evo-Devo) is to answer questions about evolvability of animal developmental systems. Evolvability or the ability to evolve is a ubiquitous property of living organisms. More specifically, it is the capacity to generate heritable, non-lethal phenotypic variation. Biologists have long recognized that evolvability of developmental programs in animals is key to their complex morphological architecture. However there is an increasing need to synthesize known facts about the developmental phenomena into mechanistic descriptions of complex systems. This ambition demands the need to understand the underlying determinants of evolvable developmental systems. I aim to review the dynamics of two systems-level phenomena: modularity and robustness and their evolutionary implications. Despite a plethora of literature, these terms have remained very ambiguous. Modularity reduces interdependence of components and confers robustness. Robustness, which is broadly understood, as the insensitivity of a biological systems functionalities to perturbations is another design principle in itself. Such robustness could enhance the potential for future evolutionary innovations. Both these properties therefore affect evolvability of a lineage. In this essay I aim to articulate my way through this hierarchy of modularity, robustness and evolvability, elucidating mechanisms that reveal their interplay to maximize functionality. I further discuss whether evolution of evolvability itself is possible.

1

Acknowledgements
Firstly, I would like to sincerely thank my supervisor, Professor Armand Leroi for his guidance and support throughout the project, while also encouraging independent work. Secondly, I would like to thank my friend, Isha Kotecha, for her help and support with Latex. I would like to dedicate this project to my parents and grandparents, who are my idols and who have provided me with an opportunity to study at Imperial College London, one of the best universities of the world.

2

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Modularity . . . . . . . . . . . . . . . 2.1 Modules in Developmental Systems 2.2 More Questions than Answers . . . 2.3 Modularity facilitates Evolvability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 4 5 6 6 9 9 10 10 11 13 15 16 18

3 Robustness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Genetic Robustness: An evolutionary dead end? . . . . . . . . 3.2 Robustness facilitates Evolvability . . . . . . . . . . . . . . . . 3.2.1 Neutrality and Cryptic Genetic Variation . . . . . . . . 3.2.2 Genetic Redundancy, Gene duplications and Innovation 3.2.3 Innovation Case Studies . . . . . . . . . . . . . . . . . 3.2.4 Distributed Robustness . . . . . . . . . . . . . . . . . . 4 Is Evolvability evolvable?

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5 Discussion and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3

1

Introduction

Novel structures such as the eye exhibit ”extreme perfection”, and this deeply worried Darwin (1859). Not only are they incredibly complex, but also they are very well suited to their environment in a multiplicity of ways. In a rather fascinating attempt, Neilson and Pelger (1994) have proposed a simulation about how exquisitely an organ as complex as the eye has evolved under natural selection in a surprisingly short period of time. Initially what was only ”a sheet of photoreceptors sandwiched between transparent tissue layers”, then transforms into a well-adapted eye. Organs like eyes and wings are ”character complexes” that are generated by interactions between many developmental genes, their gene products and the changing physical conditions during development to produce a well-adapted structure. A key question then is: What makes these animal developmental systems evolvable? And what exactly is evolvability? In simple terms, evolvability or evolutionary adaptability is the capacity of a lineage to evolve. Kirschner and Gerhart (1998) propose that a ”biological system is evolvable if its properties show heritable genetic variation and if natural selection can thus act on these properties.” However there is another aspect of evolvability, which is in regards to a biological systems ability to ”innovate” or acquire novel functions through genetic change, such that it contributes positively to the organisms survival and reproduction (Wagner, 1996). The concept of evolvability has been approached by several authors and formulated many times, be it from theoretical models or via drawing conclusions through morphological examples (such as tooth diversification). It has increasingly become clearer that the laws and processes that affect the generation of phenotypic variation govern the systems evolvability (Kirschner and Gerhart, 1998). Darwin’s ”Origin of Species” was inspired by the concepts of natural selection and heritable variation. Following that, Mendelian genetics provided the first widely noticed answers to the mysteries of inheritance and genetic variance. However we have yet to understand the relationship between this genetic variation and how it relates to selectable phenotypic variation (Mayr, 1982). In this essay I aim to explain how modularization and robustness of animal developmental systems can facilitate evolvability.

2

Modularity

In the early nineteenth century Georges Cuvier proposed the ”Principle of Correlation” which was based on the idea that living organisms were coordinated functional wholes. However we now know that organisms are not equally integrated throughout like Cuvier envisioned, rather they are organized into distinct parts or entities called modules. Topologically speaking, a module is defined as an assemblage of parts that is tightly integrated internally by strong interactions, whilst having relatively weak interactions externally, with other modules as shown in Figure 1. (Raff, 1996, and Wagner and Altenberg, 1996). It is a hierarchical concept, which means that there can be modules within modules. For instance, in Figure 1, modules 1 and 2 together constitute module 4, which is at a higher-level of organisation.

4

Figure 1: Modules and their developmental interactions. A module is a group of parts/traits (solid circles), which is highly coherent internally with many interactions (arrows) amongst its constituent traits, but is semi-independent from the rest of the system, having relatively fewer interactions with other modules. (Adapted from Klingenberg, 2005)

2.1

Modules in Developmental Systems

The concept of modularity has been with us for a long time and modular thinking, although very ambiguously, is deeply entrenched in evolution of developmental systems. For example, the concept of characters or traits as ’evolutionary units’ from an adaptionistic point of view presupposes a degree of modularity and hence an independent evolutionary potential of the modules or characters involved. Going back to Nilsson and Pelgers eye example - at the beginning only a sheet of photoreceptors exists between transparent tissue layers. Then the shape and form of tissue layers to form a pinhole-resembling eye. Thereafter, the refractive index of the vitreous body changes, followed by the formation of a graded-index lens and so on. The authors have very optimistically and conveniently assumed a continuous gradation of genetic variation in a number of defined traits such as the size, shape and optical properties of the involved tissues, which is presented to natural selection in a step-by-step manner. In essence, what this really resembles is the evolution of a one-dimensional quantitative character with unlimited variability, which is slightly misleading (Hansen, 2003). Evolutionists usually take it for granted that this variation is presented in the right manner and at the right time for natural selection to act on. It is not hard to identify that the immense evolvability of their model stems from dividing the eye into different modules. Modularity helps in a better understanding of the variational properties of developmental systems in order to understand their evolvability. Furthermore, classical accounts of animal form have been predicated on a modular view, by identifying morphological characters and tracing their individual evolution. As Von Dassow and Munro (1999) point out, notions of homology and analogy also carry an implicit assumption that one may meaningfully isolate, both in the physical as well as the conceptual context individual processes in animal design. Such a logical separability can also be seen in embryogenesis where the developmental processes like gastrulation, establishment of body axes or morphogenesis of 5

individual organs retain their individuality. Putative modules could span the entire hierarchy of development ranging from regulatory genes affecting developmental mechanisms all the way to morphological features of the whole organism. At the most basic or molecular level, cis- and trans- regulatory elements of transcription molecules exhibit a modular structure. Also, Gene Regulatory Networks (GRNs) and signalling cascades (such as Notch, Wnt and Hedgehog) - all act as candidate modules in development. Gerhart and Kirschner (1998) claimed that there are only 16 basic signal transduction pathways in eukaryotes that underlie complex animal developmental programs. They are used in multiple ways in various body parts, which is reflected in the conservation of their core network components. Moreover, the downstream effects of these GRN or signalling cascade may vary, depending on the presence of other tissue-specific factors. Moving up the hierarchy, cell types such as myocytes and neurons are modules. And finally, organs such as limb buds or insect segments all are also potential modules.

2.2

More Questions than Answers

It is true that there is no precise definition of what constitutes an ’evolutionary developmental module’. We have more questions about the term than we have answers for it. The general idea of modularity however, is definitely useful but only so long as the level of developmental hierarchy is made clear. What may be a meaningful, coherent module at one level might not necessarily be considered a module at the more basic or molecular level. One may think of it like the concept of homology, which also exhibits itself at multiple levels. However when thinking about the evolution of development, the big question isn’t to define modularity; rather it is to know ways to approach the concept and how, if at all it relates to evolvability.

2.3

Modularity facilitates Evolvability

In this section I will explore on how modularity as a principle can facilitate evolvability, providing examples wherever necessary. 1. Autonomy leads to Robustness: The idea is that a modular organization favours evolvability of complex developmental systems by permitting changes in one module without any interference from the rest of the organism (Wagner, 1996). From a developmental perspective, this modularity results in groups of traits, which possess what Lewontin (1978) termed ”quasi-independent” evolutionary potential. It is known that the probability of a random mutation being advantageous is a sharply decreasing function of the number of traits it is affecting (Fisher, 1958) Hence in a nonmodular system which is highly integrated, simultaneous random mutations are very less likely to introduce functional improvements. This is because adverse effects in other parts will most likely overcome chance improvements in one part. On the other hand, a modular organization allows for a degree of individuality and robustness in its components. Hence selection will be able to optimize and fine-tune each one without any interference. Robustness further enhances evolvability. (Refer section 3, ”Robustness”, of this dissertation) 2. Channels Variation: Modularity permits individuality in its components, which means it reduces the number of genes that can affect a character or trait. Some evolutionists have 6

claimed that this reduces its mutational target size and therefore hampers evolvability of the developmental system. This reasoning is partly true. However, I argue that the notion of evolvability means having the ability to respond to a selective challenge (Hansen, 2003), which necessitates the organism to produce the ”right” kind of variation for natural selection to act upon. Modularity offers a way of presenting small but ordered and facilitated variation and therefore increases the likelihood of character improvements. 3. Reduces Pleiotropic Entrenchment: The intuitive appeal of modularity means that we sometimes underestimate the significance of the inter-modular interactions. The developmental modules enjoy their individuality to a certain extent, however this is an oversimplified definition. Modular integrity is just as important. Interactions between different modules are very crucial. In fact, this very connectedness between modules gives birth to the hierarchical concept of modularity. If we consider the eye, the lens is a distinct module, but it has a number of critical inter-relationships with other characters. In amphibians the lens is formed by inductive interactions with various tissues such as the heart mesoderm and the retina (Raff, 2006). As characters grow towards more complexity, combinatorial control and interaction of parts become more apparent. As a result, a major problem that arises in the evolution of complex adaptions is the avoidable increasing pleiotropy, where several genes can simultaneously affect numerous traits. Such a pleiotropic load could be an absolute constraint to producing heritable non-lethal variation. How would such complex characters evolve? The answer is through simultaneous, on-going evolution of modularity. Wagner (1996) describes a phenomenon known as parcellation or dissociation (Raff and Raff, 2000), which introduces modularity in complex character adaptations and is responsible for a continuous de-coupling of pleitropic effects. 4. Co-option and Innovation: The above three explore different variational properties of modularity, which maximizes the possibility of presenting the ”right” kind of variation to selection, one can potentially lead to evolvability. At a macro-evolutionary scale, developmental modules have another role to play in evolvability - pre-existing modules can be re-used as building blocks in development, with certain modifications at various instances in evolution. This phenomenon of re-usability of existing modular units is termed co-option. A point in case is the broad usage of the suite of genes for the development of the eye. Halder et al., (1995) have discussed the role of the transcription factor-encoding gene Pax-6 (popularly known amongst developmental biologists as the ”master control” gene). As a result of its versatility and conserved structure, the gene is part of the core regulatory network that builds the eye and has been used multiple times in a vast majority of metazoans, including mammals, flies and molluscs. Although ”co-option” was originally recognized in relation to evolution of transcription factors, it could also apply to signalling pathways, which are higher-level modules per se. Co-option of pre-existing signalling pathways has generated new structures and morphologies. This is exemplified by the formation of wing eyespots in butterflies - an evolutionary novelty amongst insects. Hedgehog signalling cascade, which is used to establish the anterior-posterior patterning of Drosophila melanogaster wing-discs has been redeployed and in the butterfly Precis Coenia (Beldade and Brakefield, 2002). In addition to retaining this ancestral function, the Hedgehog regulatory circuitry has also been recruited to cells that form the butterfly wing eyespots, as depicted by Figure 2 below.

7

Another example of a morphological novelty generated by co-option is the turtle shell. It is a bony exoskeleton comprising of a dorsal carapace and a ventral plastron. It is suggested that the expression of shell outgrowth is correlated with that of the fibroblast growth factor 10, which is normally implicated in the limb outgrowth (Loredo, et al., 2001, Pires-daSilva and Sommer, 2003). Yet another example of re-usability is the invention of cell types - bodies of metazoans are mostly made up of the same cellular toolkits. Evolutionary innovations like new cell types are rare, and guarding them is only sensible as they could serve as ad hoc and effective solutions for future use. The snake lineage for example, did not have to rely on new ways for building vertebrae; they only had to add extra vertebrae. Jacob’s (1982) famously known metaphor ”Natural Selection is not an engineer”, applies very well in this context, where he highlights the fact that evolution does not think ahead but operates more like a tinkerer, and brings short-sighted, efficient solutions with whatever it has available. From this perspective on evolution, modularity may indeed provide a very logical way to achieving evolvability. But it does not therefore mean that it is the sole determinant of evolvability. I will now discuss another major principle of complex systems, robustness and explore different ways in which it relates to evolvability.

Figure 2: Co-option of the Hedgehog signalling pathway for the induction of eyespots in butterflies. Hedgehog signalling expression pattern the in Drosophila melanogaster (Top-left), redeployed in the lepidopteran wing to form 2 eyespots (Top-right). Dorsal side of the butterfly hindwing with 2 eyespots, with the anterior-posterior partitioning (blue line) (Bottom). (Adapted from Pires-daSilva and Sommer, 2003)

8

3

Robustness

Robustness, just like modularity, is a very prevalent principle in complex biological systems. It is what provides the resilience to organismal functions from failing disastrously when confronted by a barrage of internal or external disturbances. Wagner (2005) describes robustness as ”the invariance of certain phenotypic characteristics in the face internal and external perturbations”. It is important to remember that robustness implies invariance of function, not structure. In principle, a system that flexibly re-structures its components to maintain its functionalities, and still be highly robust. Robustness is not a generic attribute, which means that it is always defined in the context of certain phenotypic characteristics over a given set of perturbations. It therefore acts as a buffering mechanism to shield phenotypic constancy. Ultimately however, robustness of only one feature matters - evolutionary fitness, i.e., an organism’s ability to survive and reproduce. (Wagner, 2005) There are two types of robustness, depending on the nature of the perturbation: 1. Environmental robustness, e.g. to external perturbations such as changes in the physical environment (such as temperature, salinity, nutrient availability etc.) 2. Genetic or mutational robustness, e.g. to internal perturbations resulting from genetic variability (either through de novo mutation or through recombination) Environmental robustness could come in various forms encompassing almost all homeostatic phenomena, namely regulation of osmotic balance, metabolite concentrations, thermoregulation and so on. Biologists have studied these phenomena for decades and there is a long list of literature on each of these phenomena. Many developmental traits, such as fate patterns of cells in the vulva of the nematode Caenorhaditis elegans are found to be extremely robust to environmental changes (Felix and Wagner, 2006). On the other hand, genetic robustness, in all its naivety, lacks a comprehensive framework. While both types of robustness are equally important for the proper functioning and development of organisms, my primary aim would be to focus on genetic robustness, and how it relates to evolvability, given that mostly genetic changes are heritable and permanent. Therefore they have relatively more severe and long-term evolutionary consequences on an organismal lineage, and hence to its evolvability, as compared to environmental changes (Wagner, 2005).

3.1

Genetic Robustness: An evolutionary dead end?

The connection between genetic robustness and evolvability is very crucial. At first sight both of these phenomena imply an antagonistic relationship. By definition, robustness acts as a buffer to reduce the degree to which phenotypic variation is affected by the underlying genetic variation (Arjan, et al., 2003). However phenotypic variation is what provides the raw material for natural selection to act on, for future evolutionary change. From such an adaptionistic view, robustness acts a variational constraint, leading to an evolutionary dead end. (Ancel & Fontana 2000; Sumedha et al. 2007). But is there really such a stark negative correlation? The answer is No. I will go on to review many real life scenarios where we can see how evidently and yet so subtly robustness acts to foster evolvability.

9

3.2

Robustness facilitates Evolvability

It is true that evolution favours robust characteristics and developmental systems are no exception to this rule. Genetic robustness is the ability to withstand mutations and determines a biological system’s persistence to genetic variation. But how is genetic robustness achieved in developmental systems? It is achieved by extensive usage of modularity and hierarchical design, which is an entire design principle per se (Section on Modularity). It has two main mechanistic causes - redundancy of parts and distributed architecture of a system otherwise also known as distributed robustness (Wagner, 2005). I will now explore the evolutionary consequences of robustness, in all its different forms - neutrality, gene redundancy and distributed robustness, all of which affect evolvability. 3.2.1 Neutrality and Cryptic Genetic Variation

Wagner (2005) proposes that robustness causes many mutations to be neutral. A neutral mutation, by definition is genetic change that has no effect on an organism’s fitness, which implies that it is invisible to natural selection. Such a definition of neutrality however is operationally irrelevant to innovation or evolvability. Wagner argues that this is a rather incomplete and short-term definition of neutrality and subtly points to our ignorance of the fact that neutrality depends not only on the mutation itself but also on its interactions with other genes and its environment - both of which may change over time and could potentially render it non-neutral. Equally important is the fact that selection pressures on mutations may vary over time, which means that majority of the mutations do become visible to natural selection eventually. He thereby further proposes a new, ”non-essentialist” perspective of neutrality, which he suggestscould harbour seeds of new adaptations and become key to innovation. Neutral variation, owing to its nature, accumulates quickly and much safely that deleterious variation. Hence it is ubiquitously present in populations. Indeed, the central theme of the ”neutral theory” asserts that a high proportion of genetic variation in populations is neutral and has minimal effect on phenotype (Rutherford et al., 2000). Mutational robustness gives the ability to withstand mutations. However if a system is robust to mutations, it does not therefore follow that selection prevents them altogether. In fact, the immediate consequence of this is the accumulation of cryptic or ”hidden” genetic variation (CGV), which can become exposed, in the appropriate context - for example when the genetic or environmental background is changed. A remarkable example is that of the gene Antennapedia in D. Melanogaster. A loss-of-function mutation in this gene causes transformation of antennae into legs in the flies. When placed in different wild-type genetic backgrounds, the same Antennapedia mutation was shown to produce a range of phenotypes - practically a continuum from roughly perfect antenna to almost perfect legs, where the antenna should be (Emerald and Cohen 2004). CGV is the dark matter of biology - hidden but ubiquitously present virtually all the time. Evolutionary capacitance is a phenomenon that describes the storage and release of CGV. There are many scenarios, which could trigger the revelation of CGV. Genetic or environmental perturbations that affect robustness could reveal the evolutionary potential of CGV. An example can be conditions of stress. If an organism is under stress that means that current phenotype is not well - adapted to those conditions. In response to stress, which can be correlated to an opportunity for adaptation, CGV offers an evolutionary significant resource. It could act like 10

a capacitor acts, like a switch and could change its state to modulate the quantity of heritable phenotypic variation, therefore promoting evolvability (Massel and Trottter, 2010). Robustness is a matter to degree - absolute robustness of 100% is not achievable in living systems.It has been suggested that under less robust conditions of the systems, CGV gathers under weakened selection, which may lead to accumulation of strongly deleterious alleles, swamping away effects of any adaptive alleles, which might surface in a future environment. But this can prove harmful in the future for the organism and its evolvability. Masel (2008) addressed this problem quantitatively, and to his surprise he found that in fact, CGV could be substantially enriched with potentially adaptive alleles. Moreover, this enrichment was found markedly stronger when experiencing a strong directional selection, i.e., when several simultaneous adaptations are vital to generation of a potentially adaptive phenotype. It is suggested that under conditions of incomplete robustness, CGV might undergo a weak purifying selection by a process called pre-adaptation to eliminate deleterious alleles while they are still in a hidden state. As a result, it is of much higher quality and ”refined”, and likely to be an effective source of useful adaptations compared to random new mutations. 3.2.2 Genetic Redundancy, Gene duplications and Innovation

Genetic Redundancy is one of the major mechanistic causes of genetic robustness (Wagner 2005). It is not surprising, given that genes encode the most important parts of a biological system, proteins. Gene Redundancy refers to the existence of multiple genes in a genome, which encode proteins that have similar roles. Such gene redundancy has received great attention by developmental biologists. There is an exhaustive list of systemic gene-deletion experiments, which have demonstrated that genes thought to have been very central could often be removed without affecting the organism’s development. For instance, the BarH2 gene encodes a transcription factor with a homeobox DNA-binding domain in the fruit fly Drosophila melanogaster. It is involved in the cell fate determination of neural structures in embryos and yet its deletion does not cause any appreciable morphological defects in the sensory organ development (Higashijima et al., 1992). Tenascin, which is an extracellular matrix glycoprotein in mice, is implicated in a variety of morphological phenomena during embryogenesis. And yet, disrupting the gene encoding this protein leads to no detectable morphological abnormalities (Saga et al, 1992). And finally, the Drosophila gap gene knirps is a key transcriptional regulator, which is responsible for the establishment of anterior head structures. This gene can be eliminated without affecting the head development (Gonzalez-Gaitan et al, 1994). A recurrent theme here is that all of the above genes have duplicates - BarH2 has a gene duplicate BarH1 and knirps has a gene duplicate knirps - related. Redundant gene duplicates produce polypeptides with very similar functions, which serves to justify the weak gene knockout effects in the above experiments. Gene duplications are very abundant - up to 50% of the genes present in the eukaryotes have at least one duplicate within the same genome (Rubin et al., 2000). Subsequent to a gene duplication event, these paralogs accumulate mutations independently and could either lead to elimination of one or divergence of the duplicates functions (Wagner, 2005). Such a functional divergence results in multifunctional proteins with overlapping or redundant functions, hence they arent fully identical in function eventually, as depicted in Figure 3. Like for instance, although knirps is dispensable for the development of head structures, it is necessary for abdominal development (Wagner, 2005; Gonzalez-Gaitan et al, 1994). 11

Figure 3: Functional Divergence of genes following a gene duplication event. After gene duplication, the products P and P* of the two duplicated genes will interact with the same proteins on the short run. Eventually, new interactions will be formed and the common ones will be lost. Gene redundancy is very similarly linked to neutrality and therefore to evolvability (because of accumulation of CGV). As long as one copy of the duplicated gene retains its ancestral key function, mutations in the other copy are less very likely to be deleterious, since the normal copy most likely will compensate as a back upfor the negative effects. The more gene duplications that occur in a genome, the higher is the number of neutral mutations it can afford to accommodate. Gene redundancy provides an excellent justification for the high occurrence of neutral mutations, since it masks the effects. The higher the neutrality, the higher are the chances of new gene properties, which could eventually lead to new adaptations and foster innovation. In summary, gene redundancy in the form of gene duplications leads to robustness (which allows neutrality), which in turn facilitates evolvability. Syllogistically therefore, gene duplications facilitate evolvability. This is a not something new - In fact it is a relatively old idea. Ohno (1970) was amongst the first few to recognise and illustrate the evolutionary implications of gene duplications. The only difference now is that we classify gene duplications under the broad principle of robustness. Put differently, following the gene duplication event, the genes experience very relaxed selective constraints, which can justify an elevated rate of evolution (Wagner, 2002). Systemic gene deletions have shown us how gene duplications cause robustness and could lead to innovations. Many laboratory-based comparative analyses have revealed multiple occurrences in the past where gene duplication has resulted in functional innovation. A point in case is the evolution of the butterfly photoreceptor genes. Briscoe (2001) reported that after a gene duplication event, the newly found photoreceptor genes encoded opsins with sensitivity to an increased fraction of the visible spectrum of the electromagnetic spectrum, as compared to that of the previously categorised lepidopteron opsins. Indeed, this is innovation caused by gene duplication. However, such experiments work on a modest small-laboratory scale. We could therefore ask - How effective is robustness, in the form of gene duplications, at producing innovations on larger geological time scales? I will now review two further examples of innovations and their associations with gene duplications on much larger time scales.I will first explore the spectacular vertebrate diversification and how Hox gene duplications have a role in that. Next I will reflect upon the evolution of the four-chambered mammalian heart, and its relation to gene duplications.

12

3.2.3

Innovation Case Studies

Case Study 1: Vertebrate diversification No evolutionary account of the role of gene duplications in evolvability is better illustrated than that of Hox genes. They are essential metazoan genes, implicated in the regionalization and patterning of structures along the anterior-posterior body axis. They show a characteristic tightly - linked or clustered organization (McGinnis and Krumlauf, 1992). Their chromosomal alignment within the cluster is colinear with their spatiotemporal expression pattern along the cephalo-caudal (head-to-tail) axis. Invertebrate such as insect and nematode genomes possess only a single cluster of Hox genes, which seems to have duplicated at least twice during the evolution of vertebrates. This means that vertebrates should have 4 Hox gene clusters, which is indeed the case. For instance in mammals, there are 38 Hox genes arranged in four clusters on different chromosomes. The most recent common vertebrate ancestor is thought to have 14 Hox genes, while vertebrates have retained only 13 Hox genes. Therefore, vertebrates possess 4 Hox clusters, designated A-D, which are subdivided into 13 different paralogous groups labelled 1 - 13 (Lemons and McGinnis, 2006). Vertebrate diversification is characterized by so many spectacular innovations as compared to their chordate ancestors. To name a few, they have a very complex and specialized brain (with forebrain, midbrain and hindbrain regions), cartilage and other mineralized structures such as bone and teeth. The invention of the bone in vertebrates further led on to many new innovations that are unique to them only, such as a protective skull, a differentiated vertebral column, paired appendages and so on and so forth. The list goes on. The point is to illustrate the powerful and elaborate roles that Hox genes play in the proper embryonic development in all of these traits. It is not hard to identify an underlying redundancy in function since between the 13 different Hox genes, they perform so many developmental roles. It therefore doesn’t come with a surprise that despite so many million years of duplication, many of these Hox genes have retained a few redundant functions, or as Wagner (2008) likes to call them - ”remnants of mutational robustness” characteristic of gene duplications events. For example, it was shown using mutational analysis that in mice, genes from the Hox paralogous group 8 exhibited functional redundancy in the positioning of their hind limbs (van den Akker et al., 2001). In another instance, Hunter and Prince (2002) reported 2 genes from the zebra fish Hox paralogue group 2, namely Hoxa2 and Hoxb2, which function redundantly in the patterning of the second pharyngeal arch. Overall, we can see how 2 Hox gene duplications events early in vertebrate radiation have played a critical role in producing myriad evolutionary innovations. Case Study 2: Evolution of the Heart Apart from being associated with large-scale radiations, gene duplications have been implicated in evolutionary innovations at a smaller scale - at the level of individual trait as exemplified by the evolution of heart. The heart, an ancient organ has evolved with increasing sophistication by addition of novel structures and functions to what was once a primitive pump (Olson, 2007). However with increasing body size and the need for an efficient distribution of oxygen and nutrients, a powerful pump became an imperative. The invertebrate heart, much like that of the ancestral chordates 13

is a single-layered contractile tube that can pump blood in both directions. In contrast, hearts of fishes are two-chambered consisting of a single atrium connected to a ventricle, whereas that of amphibians are three-chambered with two atria separated by a septum and a single ventricle. A transition from an aquatic to a terrestrial habitat meant that the oxygenated and deoxygenated blood had to be kept separated. The amniotes (i.e. reptiles, birds and mammals) therefore have the most advanced heart of all, with two atria and two ventricles. Such a four-chambered heart, separated by valves and septae allows a unidirectional blood flow and accommodates the need for separation of oxygenated and de-oxygenated blood. An evolutionary conserved network of genes, which act as transcription factors, controls the development of heart in vertebrates and invertebrates. Five such core cardiac transcription factors include NK2, MEF2, GATA, Tbx and Hand (Oslon, 2007; Wagner, 2008). It has been found that most of these genes have more number of duplicates in vertebrates as compared to their chordate ancestor, as shown in Figure 4. For instance there is only one single copy of MEF2 gene in Drosophila, which functions to produce muscle contractile proteins. On the other hand, in vertebrates there are four MEF2 duplicates, all of which show overlapping or redundant functions, which is thought to be a remnant of robustness following their ancient duplication events. Loss-of-function mutations in one of MEF2 duplicates, MEF2c, results in the malformed heart with no right ventricle. Moreover, the right ventricle is made up from a subset of cardiac cell population which unique to amniotes. These two facts, put together suggest that following gene duplication, MEF2c possibly adopted a specialized and novel role paving way for morphological innovation.

Figure 4: Role of Gene Duplications in the Evolution of the Heart. 1. Number of gene duplicates for each of the key transcription factors of the cardiac developmental network 2. A simplified version of the vertebrate phylogeny, also depicting the increasing complexity of the heart. (Adapted from Wagner, 2008). The Hand (heart and neural crest derivatives expressed transcript) gene illustrates yet an14

other striking example of a new function acquisition. A single copy of the Hand gene is expressed in amphibians and zebra fish, both of which have one ventricle. In mice however, which have 2 ventricles, there are 2 copies of the Hand gene - Hand1 and Hand2. While Hand1 mutants are defective in the formation of the left ventricle, Hand2 mutants are defective in the formation of the right ventricle (Buckingham et al., 2005). It appears that after a Hand gene duplication event, the functions of the resulting genes have become partitioned such that both the duplicates have acquired new roles associated with the partitioning of a morphological organ. The sheer number of genes with no phenotypic effects upon deletion is humongous - many thousands, which means that gene duplications cannot be the exclusive cause behind the observed genetic robustness. Besides, even if the gene did have a duplicate, the pair may have substantially diverged in function to account for such retention of function that could justify the robustness (Wagner, 2005). In fact results from a genome-wide RNA interference study on the nematode worm Caenorhabditis elegans reveal that as high as 89% of single-copy genes and 96% of duplicate genes showed no detectable phenotypic effects. The scientists further showed that proportion of weak gene deletion effects caused by gene duplications is ∼3-36% which is much lower than expected (Conant and Wagner, 2003). However this doesn’t imply that gene redundancy is unimportant for mutational robustness. For example, knockouts of genes, which have a duplicate, have a much lower likelihood to cause a lethal effect (Conant and Wagner, 2003 and Gu et al., 2003). Moreover, the greater the similarity of function between the duplicates, the less severe is the effect observed on phenotype when silencing one of them. Hence, recent gene duplicates are more likely to be efficient compensators for each other. Furthermore, the higher the number of duplicates a gene has, the faster it evolves owing to an increased tolerance for mutations (Wagner, 2005). Thus, gene duplications clearly do play a role to a certain extent in determining the outcomes of gene deletions and therefore do relate to mutational robustness, although they don’t account for the overwhelming majority of absent effects of gene deletions. Then the question arises: If gene redundancy is not the sole key player responsible for weak perturbation effects, then what else is? The answer lies in phenomenon commonly known as distributed robustness. 3.2.4 Distributed Robustness

Distributed Robustness, or degeneracy is another mechanistic cause of robustness and emerges as a result of the distributed architecture of biological systems. Note here that without modularization, robustness couldn’t be distributed to maximize and maintain functionalities in the face of perturbations. Distributed Robustness is not synonymous to the concept of redundancy. Here, many components of the system contribute towards the overall system functioning, however they all have different roles. This is shown in demonstrated in Figure 5. Therefore, a mutation in one of the components can be compensated for, by others in the system.

15

Figure 5: Difference between Distributed Robustness and Redundancy of Parts. Scientists argue that distributed robustness is more prevalent in biological systems than redundancy. Classic examples of distributed robustness are those involving genetic regulatory networks. A point in case is the widely discussed and the most explicitly studied Drosophila segment polarity network, which is responsible for subdividing the embryo into multiple segments (Von Dassow et al., 2000). This gene regulatory network is considered to be a very robust module. That is because of its distributed architecture - No two genes have the same function, yet they all work in a harmonious to produce a robust whole. If we move up the hierarchy, on higher levels of biological organisation it is hard to determine with accuracy the exact effect of individual genes on phenotype as many hierarchical levels of organisation are likely to mild down its effects. For example, during embryonic development, so many genes are responsible are responsible to form different phenotypic characters. Most of them docks of mutations. Such hidden variation is not due to redundancy.

4

Is Evolvability evolvable?

We have explored how core processes underlying animal developmental systems, namely their modular organisation and robustness affect evolvability.We can therefore ask if there can be selection for evolvability itself. Possible evolution of evolvabilityhasbeen discussed several timesin the past (Kirschner and Gerhart, 1998, Wagner and Altenberg, 1996 and Wake and Roth, 1989). Such a question however, is rather difficult to address. Kirschner and Gerhart (1998) have proposed that evolvability is more of an organismal clade level trait. An evaluation of how natural selection works at such a larger scale could be challenging. Moreover, unlike other normal traits under selection, which confer present benefit, evolvability seems to confer a future advantage. Our view of the underlying dynamics of developmental processes offers a nice opportunity to better understand the phenotypic variation aspect of evolvability. However, to be able to fully appreciate the deeper selected function of evolutionary change, there are some fundamental issues that will have to be investigated further.Nevertheless, there are some generalisations that can be made regarding selection for evolvability (Kirschner and Gerhart, 1998). It has three main components to it, each relating to a different level. 1. Selection for evolvability at an individual level: At the most basic individual level, there may be selection for various flexible and robust processes that can tolerate variability occurring from internal stochastic noise, for example variations in the cell numbers or cell positions. Such robust processes could contribute positively to the biological fitness and 16

complex development of the individual. As a result they will therefore be selected for in evolving complex organisms either via a). Spontaneous or de novo modification or b). Modification of pre-existing robust processes. A non-selected by-product alongside selection of robust processes would be of phenotypic variation and therefore evolvability itself. Furthermore, individual with such processes would have a higher potential to generate phenotypic novelty with lower number of mutations. 2. Selection for evolvability at an individual + clade level: More robust, flexible processes would buffer individuals against deleterious and lethal mutations. Therefore, populations are likely to accumulate and carry non-lethal mutations. As a result, more of such non-lethal genetic variety would be available directly to individuals or through mating in the population with other clade-members (for gene recombinations). There might be a clade-level selective preference forclade members with more robust processes, because they have a higher potential to diversify under selective conditions and exhibit greater evolvability. 3. Selection for evolvability at a clade level: Superior evolvability of clade-members would be advantageous for survival through clade-level selection for example when rapid radiations occured, either after ecological niches were vacated (through mass extinctions) or when new niches were entered. An example of latter would have been the exploration of new wetlands for the first amphibia. Therefore in times of ecological upheavals, clades are likely to be selectively favoured due to their ability for rapid radiations into new and emptied ecological domains, i.e., based on their evolvability. Lineages that have experienced such a history include chordates and arthropods. Additionally, lineages that are exposed frequently to such radiations might have been solely selected for their greater ability to evolve. Also, if we look back in history, starting from the pre-Cambrian period, and closely inspect the breakthroughs in metazoan design we can identify many attributes that suggest the evolution of evolvability (Kirschner and Gerhart, 1998). There have been numerous rounds of ”variation, selection and fixation” that have occurred in different traits in the metazoans. At the beginning of the metazoan evolutionary scale, when multicellularity was new to them, variation, selection and conservation of the most basic traits such as the extracellular milieu and epithelia occurred. Thereafter came the first signalling pathways and simple gene regulatory circuits, followed by invention of new cell types and body organisation plan, then the appendages and so on. All of these went through a round of variation, selection and fixation, and in each of these rounds, only processes that were robust and versatile enough were conserved and fixed. As explained before, these properties could have been the source of phenotypic variation for the next round ahead. In a way, by promoting phenotypic variation, such processes ensure their own conservation and which could further be selected upon, along with the novel traits. Organisms that lackedrobust processes will be at a selective disadvantage, as they cannot produce as much non-lethal heritable phenotypic variation, i.e. they have a lower evolvability. This historical perspective exemplifies how evolvability itself can evolve.

17

5

Discussion and Conclusions

This dissertation has explored the ways in which evolvability of developmental systems in animals are influenced by 1) the underlying architectural feature of modularity 2) robustness that emerges out of such modularity. It has also looked upon how various mechanisms such as mechanisms such as gene redundancy and degeneracy confer robustness and enhance evolvability. I have summarised their interplay in a schematic model below, Figure 6. Although, I have focussed heavily on the variational aspects of organisms and their relation to evolvability, it is important to remember that for evolutionary change to occur, both -variation and natural selection are crucial. Only variation but no selection doesn’t lead anywhere. A noteworthy example is that of the nematode C. elegans vulva. As I already mentioned, the gene regulatory network underlying the vulva development is very robust to genetic change and has changed significantly in the past 200 million years of its life. However, the structure of the vulva itself remains almost the same. This depicts how robustness can change the genetic composition and gene network architecture, yet does not lead to morphological evolution in the absence of the key impetus - that of natural selection. Similarly, the ”right” kind of variation, such as that produced by gene duplications, needs to be produced for innovation. Robustness facilitates evolvability. Like for any other biological generalisation this doesn’t come without caveats. What happens if there is an overload of robustness in a system? Can there be an excessive amount of robustness, for example by too many gene duplications, which could impede innovation? If so, what are the factors that decide the correct threshold? We do not have answers to questions like these. Terms like fitness and modularity - we still don’t have concise definitions for them. There is no consensus reached amongst developmental biologists as to how isolated a character must be to be counted as a module. All these provide fertile grounds for future investigations.

18

Figure 6: Interplay between Modularity, Robustness and Evolvability.

19

References
Ancel, L.W. & Fontana, W. (2000) Plasticity, evolvability and modularity in RNA Journal of Experimental Zoology, 288 (3), 242-283. Beldade, P. and Brakefield P.M. (2002) The Genetics and Evo-Devo of Butterfly Wing Patterns. Genetics, 3, 442-452. Briscoe, A. D. (2001) Functional diversification of lepidopteran opsins following gene duplication. Molecular Biology and Evolution, 18(12), 2270-2279. Buckingham, M. Meilhac, S., and Zaffran, S. (2005) Building the mammalian heart from two sources of myocardial cells. Conant, G.C., and Wagner, A. (2003) Duplicate genes and robustness to transient gene knockouts in Caenorhabditis elegans. Proceedings of the Royal Society of London Series B 271, 89-96. Darwin, C. (1859) On the Origin of Species by Means of Natural Selection or the Preservation of Favoured Races in the Struggle for Life, Oxford: Oxford University Press. Emerald, B.S and Cohen, S.M. (2004) Spatial and temporal regulation of the homeotic selector gene Antennapedia is required for the establishment of leg identity in Drosophila. Developmental Biology, 267(2), 462-472. Felix, M.A. and Wagner, A. (2006) Robustness and evolution: concepts, insights and challenges from a developmental model system. Heredity, 100(2), 132-140. Fisher, R.A. (1958) The Genetical Theory of Natural Selection, 2nd ed. New York: Dover. Gerhart, J. and Kirschner, M. (1998) Cells, embryos, and evolution.Boston, MA: Blackwell. Gonz´lez-Gait´n, M., Rothe, M., Wimmer, E.A., Taubert, H. and J¨ckle, H. (1994) Rea a a dundant functions of the genes knirps and knirps-related for the establishment of anterior Drosophila head structures. Proceedings of the National Academy of Sciences of the United States of America 91: 8567-8571. Gu, Z., Nicolae, D., Lu, H. and Li, W. (2002) Rapid divergence in expression between duplicate genes in genetic robustness against null mutations. Nature, 421, 63-66. 20

Halder, G., Callaerts, P., Gehring, W.J. (1995). Induction of ectopic eyes by targeted expression of the eyeless gene in Drosophila. Science, 267 (5205), 1788-1792. Hansen, T.F. (2003) Is modularity necessary for evolvability?: Remarks on the relationship between pleiotropy and evolvability. Biosystems, 69 (2-3), 83-94. Higashijima ,S., Michiue, T., Emori, Y. and Saigo, K.(1992) Subtype determination of Drosophila embryonic external sensory organs by redundant homeo box genes BarH1 and BarH2. Genes and Development, 6:1005-1018. Hunter, M.P. and Prince V.E. Zebra fish Hox paralogue group 2 genes function redundantly as selector genes to pattern the second pharyngeal arch. Developmental Biology, 247(2), 367-389. Jacob, F. (1982) The Possible and the Actual. University of Washington press. Klingenberg, C.P. (2005) Variations. Manchester: Elsevier Inc. Lemons, D. and McGinnis W. (2006) Genomic evolution of Hox gene clusters. Science, 313(5795), 1918-1922. Lewontin, R. (1978) Adaptation. Scientific American, 239, 212-231. Loredo, G. A., Brukman, A., Harris, M.P., Kagle, D., LeClair, E. E., Gutman, R., Denney, E., Henkelman, E., Murray, B.P., Fallon, J.F., Tuan, R.S., Gilbert, S.F. (2001) Development of an evolutionarily novel structure: fibroblast growth factor expression in the carapacial ridge of turtle embryos. Journal of Experimental Zoology, 291(3), 274-281. Masel, J. (2006) Cryptic genetic variation is enriched for potential adaptations. Genetics, 172 (3), 1985-1991. Masel, J., & Trotter, M. V. (2010). Robustness and Evolvability. Trends in Genetics, 26(9), 406-414. Mayr, E. (1982) The Growth of Biological Thought: Diversity, Evolution, and Inheritance Mcginnis, W. and Krumlauf, R. (1992) Homeobox genes and axial patterning. Cell, 68(2), 283-302.

21

Nilsson, D.E. and Pelger, S. (1994) A pessimistic estimate of the time required for an eye to evolve. Proceedings of the Royal Society, B256 (1994), pp. 53-58. Olson, E.N. (2006) Gene Regulatory Networks in the Evolution and Development of the Heart. Science, 313 (5795), 1922-1927. Oslon, E.N. (2007) Gene Regulatory Networks in the Evolution and Development of the Heart. Science, 313 (5795), 1922-1927 Pires-daSilva, A. and Sommer, R.J. (2003) The Evolution of Signalling Pathways in Animal Development. Genetics 4 (1), 39-49. Raff, R.A. (1996) The Shape of Life: Genes, Development, and the Evolution of Animal Form. Chicago: University of Chicago Press. Raff, E.C. and Raff R.A. (2000) Dissociability, modularity, evolvability. Evolution & Development, 2(5), 235-237. Rubin, G.M., Yandell, M.D., Wortman, J.R., Miklos, G.L.G., Nelson C.R. et al. (2000) Comparative genomics of the eukaryotes. Science. 287, 2204-2215. Rubin, G.M., Yandell, M.D., Wortman, J.R., Gabor Miklos, G.L., Nelson, C. R., Hariharan, I.K., Fortini, M.E., Li, P.W., Apweiler, R., Fleischmann, W., et al. (2000). Comparative genomics of the eukaryotes. Science, 287(5461), 2204-2215. Saga, Y.,Yagi, T., Ikawa, Y., Sakakura, T. and Aizawa, S. Mice develop normally without tenascin. Genes and Development, 6: 1821-1831. Von Dassow, G. and Munro, E (1999) Modularity in Animal Development and Evolution: Elements of a Conceptual Framework for EvoDevo. Journal of Experimental Zoology, 285(4), 307-325. Wagner, G.P. (1996) Homologues, Natural Kinds and the Evolution of Modularity Integrative and Comparative Biology (formerly: American Zoologist), 36(1), 36-43. Wagner, G.P. and Altenberg, L. (1996) Complex Adaptations and evolution of evolvability. International Journal of Organic Evolution, 50 (3), 967-976.

22

Wagner, A. (2002) Selection and gene duplication: a view from the genome. Genome Biology, 3(5), 1012.1-1012.3. Wagner (2005). Robustness and Evolvability in Living Systems. Princeton, New Jersey: Princeton University Press. 12-45. Wagner, A. (2008) Gene duplications, robustness and evolutionary innovations. Bioessays, 30(4), 367-373.

23