Conflicting hypotheses on the nature of megaevolution

wallace arthur

Here is a question of the utmost importance for our understanding ofwhat has been called the 'big picture' of evolution (Simpson 1944,1953): are the divergences that lead ultimately to high-level sister groups, such as those that would typically be labelled as orders, classes and phyla, qualitatively or quantitatively different from those that lead to low-level sister groups, such as races, species and genera? In other words, is megaevolution more than just accumulated micro/macro-evolution, or alternatively is evolution effectively 'scale-independent' (Leroi 2000)?

This question can be approached in three ways. We can choose to compare the magnitude of changes involved in high- and low-level divergences, the type of changes, or the timing (in development) of changes. Here, I argue that previous work on the first of these has been unproductive and has generated more heat than light; but that the second and third offer better prospects for shedding light on this important issue. However, in an unusual strategy, I also play devil's advocate with my own argument at the end of the chapter. This helps to take us in an interesting, final (for now) direction.

Because the designation of high-level sister groups as, for example, orders or classes, is a subjective rather than an objective process, I will, wherever possible, use specific examples rather than general levels of taxon.

magnitudes of change

Throughout the history of evolutionary biology, there have been assertions that large-scale evolution involves individually bigger changes in

Evolving Pathways: Key Themes in Evolutionary Developmental Biology, ed. Alessandro Minelli and Giuseppe Fusco. Published by Cambridge University Press. # Cambridge University Press 2008.

the developmental trajectory, and hence in the adult phenotype, than does its small-scale, micro-evolutionary, equivalent. Such a view has been labelled macromutational or saltational; it had several well-known proponents (Bateson 1894, De Vries 1910, Goldschmidt 1940). Although macromutational theories of evolution have been rejected by the majority of biologists (most prominently, perhaps, by Dawkins 1986), they still occasionally surface (Whiting and Wheeler 1994) and are worthy of brief consideration.

At the phenotypic level, most change that occurs in the divergence of intraspecific variants, and of congeneric species, is continuous rather than discrete. A classic example is the divergence of beak size and shape in Darwin's finches, which is now beginning to be understood at the molecular level (Abzhanov et al. 2006). However, when higher-level divergences are considered, these often seem to be explicable only by invoking qualitative changes instead of, or as well as, quantitative ones. For example, in the Diptera there is only one pair of wings, in contrast to the four-wing condition from which this clade emerged. This famously led Goldschmidt to propose that the divergence of taxa at this level involved homeotic mutations (Goldschmidt 1952).

One of the reasons that Goldschmidt's ideas were not accepted is that homeotic mutations usually confer a massive loss in fitness; another is that the reduction of a pair of wings to halteres (Diptera) or their specialisation for some other purpose (e.g. the evolution of the elytra from the forewing of coleopteran ancestors) can be achieved by a series of small changes. In other words, the invoking of evolution by homeotic mutation in these cases would appear to be both problematic and unnecessary. However, it is possible that the argument of major fitness decrease being necessarily associated with homeotic mutation is context-dependent. For example, the discovery of an apparently healthy adult homeotic centipede in a natural population (Kettle et al. 1999, 2000) contrasts with the Drosophila situation. Also, a role for homeotic mutations in the evolution of angiosperm flowers now seems possible (De Craene 2003).

Although segment identity (the realm of homeotic mutations) can change gradually, segment number, as a meristic character, cannot. In this realm, we can only have one integer value or another. So perhaps there is a stronger case to be made here for a macromutational theory of the divergence of orders and classes, especially within the Arthro-poda? Actually this is not so. There are indeed some high-level arthropod sister-groups characterised by a segment number difference between them but a lack of (or negligible) segment number variation within either of them, especially in the Crustacea (Schram 1986). However, there are also well-known cases of variation in segment number within arthropod species, such as the geophilomorph centipedes (Minelli and Bortoletto 1988, Arthur and Kettle 2001). There is no reason why normally invariant groups should not exhibit transient periods of intraspecific variability, perhaps environmentally induced but partially heritable, which provide the source of a later high-level difference in segment number.

Although segmentation has provided fertile ground for controversies about possible saltational evolution, we should not forget that it is itself an evolutionarily derived condition - regardless of whether it has been derived on one, two or three occasions (Davis and Patel 1999). A feature of broader phylogenetic scope than segmentation is symmetry/asymmetry. This again provides an opportunity for salta-tional theories of mega-evolution, perhaps most importantly in relation to the reversal of dorso-ventral orientation that appears to have accompanied the divergence of protostomes and deuterostomes (Geoffroy Saint-Hilaire 1822, Holley et al. 1995).

As with segment number, however, there are examples of intras-pecific variation in the direction of asymmetries, not just rare cases such as the medical condition known as situs inversus in humans, but also polymorphisms in natural populations, such as for dextrality/sinis-trality in the gastropod Partula suturalis (Murray and Clarke 1966).

The most reasonable conclusion at present with respect to the magnitudes of changes at the phenotypic level that contribute to evolutionary divergence is as follows. All levels of divergence typically involve accumulation of minor continuous variants; and all levels of divergence involve the occasional incorporation of changes that are 'large' and discrete. Some of these latter changes are based on a single big-effect mutation - as in chirality, but perhaps with some complications (Gould et al. 1985). Others probably have a more complex genetic basis. In any event, consideration of the magnitude of effect of changes in development and in the adult phenotype provides us with no clear evidence for a qualitative difference between mega-evolution and its smaller-scale (micro/macro) equivalent.

types of change

In the long term, this may be the best approach to the issue of scale-dependence versus scale-independence in evolution, but we still lack adequate information to come up with a rational and complete classification of types.

There is a connection here with the concept of an evolutionary novelty - an exciting and important concept but one that has been, and continues to be, hard to define. Different authors have adopted very different approaches (Mayr 1963, Müller and Wagner 1991). Given the problems with defining novelties, it is probably best to proceed by considering examples.

Most high-level clades are characterised by possession of a novelty -so novelties can be considered as a subset of the more general category of synapomorphies. Examples include chelonians (shells), mammals (hair), centipedes (forcipules) and dipterans (halteres). At the highest taxo-nomic levels, we can perhaps equate novelties with body plans (vertebrate endoskeleton, arthropod exoskeleton, and so on); while at middle levels (as in Diptera), the novelties appear to be less deep-rooted in the body architecture, although whether this subjective notion can be quantified is another matter. In fact, 'novelties' can be thought of not as a discrete category but rather as a hierarchy that ranges from the very conspicuous (e.g. the origin of skeletons) to the much less conspicuous (e.g. the redeployment of certain bones of the skeleton, as in the case of the origin of mammalian ear ossicles), thus intergrading with more minor evolutionary changes that should probably not be described as novelties at all (e.g. the segment number increase that characterises centipedes of the order Geophilomorpha when compared with their sister-order, the Scolopendromorpha).

Given this intergradation, there would seem to be no strong argument arising from consideration of novelties that mega-evolution is fundamentally different from its lower-level counterparts. It is true, perhaps, that some novelties would appear to involve a very specific and rare sort of variant - e.g. the chelonian shell with its fused ribs and internal shoulder girdle (Gilbert et al. 2001) - but this is hardly the basis for a general theory of the origin of novelties.

timing of change

This may, at least for the moment, represent the best opportunity for making headway with the issue of the relationship between megaevolution on the one hand and micro/macro-evolution on the other.

The key question now becomes: are those divergences that ultimately lead to the origin of higher taxa characterised by a different distribution of changes, in developmental time, from those which lead only to the origin of congeneric species? More specifically, are 'big' divergences characterised by earlier (on average) ontogenetic changes than 'small'

ones? This is one of those so-called deceptively simple questions - i.e. it is not simple at all. I will approach it from a historical angle.

The obvious starting point is von Baer (1828) and his idea of divergence in embryonic trajectories. The most famous of the many comparisons he made are those involving different vertebrate classes and orders. For example, the ontogenies of birds and mammals start off being rather similar, but end up being very different (Figure 3.1A). Although von Baer did not interpret such patterns in evolutionary terms, Haeckel (1866) subsequently did, and he combined von Baerian divergence with a more enlightened version of recapitulation than that of the earlier 'nature-philosophers,' as can be seen in the following quote from the English translation of his Anthropogenie (Haeckel 1896):

examination of the human embryo in the third or fourth week of its evolution shows . . . that it exactly corresponds to the undeveloped embryo-form presented by the Ape, the Dog, the Rabbit and other Mammals, at the same stage of their Ontogeny.

It has subsequently been emphasised that the very earliest developmental stages do not fit into this apparently neat picture. For example, the ontogenies of birds and mammals have very different starting points before converging to their point of maximum similarity (the pharyngula stage), beyond which von Baerian divergence does indeed occur. Thus we have come to recognise the 'phylotypic stage', and this applies not just to vertebrates but to other groups as well. In fact, the phrase was first used to describe the germ-band stage in insect development (Sander 1983).

The recognition of the phylotypic stage led to the egg-timer or hour-glass model of comparative embryology (Duboule 1994); see Figure 3.1B. And a further refinement of this was to acknowledge that, when enough species were compared, the idea of a phylotypic 'stage' was too neat, and a better model was based on the idea of an extended phylotypic period (Richardson 1995). In any event, it must be stressed that the 'egg-timer' of comparative embryology is a very asymmetric one: the point of constriction is much closer to the start of ontogeny, the fertilised egg, than it is to the end-point, the adult. Ofcourse, I use 'start' and 'end' here in a pragmatic way; it is not my intention to take an overly 'adultocentric' (Minelli 2003) view of development.

Another complication to our overall picture is the existence, indeed the prevalence, of complex life-cycles, such as those of insects, amphibians, parasitic platyhelminthes and many marine invertebrates with assorted types of larvae, including trochophores and pluteuses.

Parasitic Platyhelminthes

Figure 3.1 A, von Baerian divergence, as illustrated by the early similarity of four vertebrate embryos giving way to later differences. B, The egg-timer or hour-glass model of comparative embryology, showing convergence from different starting points to a point of maximum similarity prior to von Baerian divergence from that point onwards. Note that the time axis is distorted. The point ofmaximum similarity, or phylotypic stage, occurs very soon after the start of development. Reproduced, with permission, from two sources (Raff and Kaufman 1983, Richardson et al. 1997).

Figure 3.1 A, von Baerian divergence, as illustrated by the early similarity of four vertebrate embryos giving way to later differences. B, The egg-timer or hour-glass model of comparative embryology, showing convergence from different starting points to a point of maximum similarity prior to von Baerian divergence from that point onwards. Note that the time axis is distorted. The point ofmaximum similarity, or phylotypic stage, occurs very soon after the start of development. Reproduced, with permission, from two sources (Raff and Kaufman 1983, Richardson et al. 1997).

These complex life cycles clearly cannot be described in the same terms as the simple ones of direct developers such as birds and mammals. However, they do not require a fundamentally different model than von Baerian divergence, or its more sophisticated (and more accurate) egg-timer equivalent. Rather, they require that such a model be applied separately to each life stage rather than to the whole of ontogeny throughout the life-cycle. Because the questions at issue here are difficult enough without having to think simultaneously about different life stages, in what follows I will base the discussion largely on direct developers; the argument can be extended later to deal with indirect developers.

There is something that so far has remained implicit; now is the time to make it explicit. The comparisons of two or more ontogenies that we have been making have been of phylogenetically distant animals: that is, those belonging to different classes or orders (Figure 3.1A). What picture of similarity and difference through developmental time would we see if, instead, we made a comparison between the ontogenies of a 'typical' pair of congeners, in as much as there is such a thing?

We are hindered here because of a shortage of comparative developmental studies at this level, particularly those that include an embryonic component rather than merely comparing post-embryonic growth patterns (Andersson 1990). So it may be useful to begin with a 'thought experiment', as follows. Suppose that a research project is undertaken over the next few months, specifically to make a series of comparisons of the ontogeny of pairs of direct-developing congeners, sampling from a range of different phyla. For example, in the molluscs, we might compare the ontogenies of the land-snails Cepaea hortensis and Cepaea nemoralis that are well known from an ecological genetics perspective (Jones et al. 1977) but not from a developmental perspective. And we would make similar inter-congener comparisons within other phyla. What overall pattern would we see if many such studies were undertaken?

My suspicion is that, after the appropriate phylotypic stage (where there is one) we would see von Baerian divergence, with the point at which the two ontogenies begin to diverge being much later than in the case of comparisons between species from different orders or classes.

Suppose that this result - congeners diverging later than more distantly related animals (say from different classes) - is a general one. This leads us to an interesting argument about the possibility (or otherwise) of extrapolating from micro/macro-evolution of development to its mega-evolutionary counterpart. The argument goes like this. When we think only of magnitudes of morphological change in evolution, the accumulation of many small differences (as typically found between congeners) over long periods of time, will, with an inescapable logic, produce the much larger differences that we observe between more phylogenetically distant forms. This view of evolution has been disseminated to a wide audience (Dawkins 1986) and is generally accepted. However, when we think in explicitly developmental terms, and especially in terms of the timing of changes in ontogeny, the equivalent argument does not work: lots of late changes in development do not add up to an early one.

At this juncture, it is worth stressing the contrast between absolute and statistical differences, for example in the 'evolvability' (Kirschner and Gerhart 1998) of different developmental stages, because this point has been neglected in some previous discussions.For example, consider the following (Scholtz 2005):

The comparative view of developmental processes is that of a series of potentiallly independent patterns (stages) which can be altered individually ... The recognition of this potential freedom of developmental stages results in a refutation of a special importance of any stage.

In my view, Scholtz is stating the obvious when he says that the various stages of development can all be altered in the course of evolution. But the relative probabilities of altering different stages are not the same. The fact that different phylotypic stages, such as vertebrate pharyngula and insect germband, are manifestly different from each other attests to the former; while the fact that these stages are recognizable within higher taxa attests to the latter.

These considerations lead us in the following direction: both early (e.g. phylotypic) and late (e.g. allometric) developmental stages can change in evolution, but the latter change more often, and so are typically found in all comparisons of taxa at whatever level; while the former are rarer and so are not often found in comparisons involving congeneric or confamilial species. Although this argument might be a circular one in the case of congeneric species defined as such solely on developmental grounds, it would not be so if the species concerned were defined by other means, such as DNA sequence data.

Note, however, that 'not often' and 'never' are different. Just as we saw earlier that big-effect changes sometimes occur between congeners, early changes sometimes do likewise. Indeed, sometimes these are the same changes looked at in a different way. Not only do chirality reversals in gastropods invert the symmetry of the whole animal, but they do so by modifying, through the maternal-effect genes involved, the very earliest stage of development, namely cleavage (Verdonk and van den Biggelaar 1983). Also, evolutionary switches between the planktotrophic and lecithotrophic larvae of congeneric sea urchin species have major morphological effects (loss of feeding arms) and involve early developmental stages (Wray and Raff 1989).


Consideration of the timing of changes in development rather than their magnitudes of effect on the adult phenotype is helpful in one way but not in another. It is helpful because there are no temporal equivalents of the philosophically loaded terms micro- and macromutation, and no history of ill-tempered debate equivalent to that between Goldschmidt and the architects of the 'Modern Synthesis.' However, it is unhelpful because although it suggests that mega-evolution is unique in its temporal distribution of changes through developmental time, it is not conclusive in this respect.

Let us approach this problem in two ways: first, by playing devil's advocate with the 'lots of late changes do not add up to an early one' argument; and second, by re-examining the concept of an evolutionary novelty.

The devil's advocate argument can be quite a simple one, if the right context is chosen, as follows. Consider two extant high-level sister groups of direct developers, for example two mammalian orders, chosen so that their phylotypic stages are very similar. Ignore the pre-phylotypic phase of their development. With regard to the post-phyloty-pic phase, ask the following question. Is there any time-point, tx, between the phylotypic stage, tp, and the completion of development, tc, beyond which no inter-congener embryonic divergences (within either order) go back, while the inter-order divergence itself does? If there is such a time-point, then accumulated 'ordinary' speciations cannot account for the origin of the two orders concerned, because the argument 'lots of late changes do not add up to an early one' applies. On the other hand, if there is not such a time-point, scale-independence is possible. In most cases, we lack enough relevant comparative developmental data to know which of these alternative possibilities is true.

Another problem of an argument based solely on timing is that, while mega-evolutionary divergences usually, perhaps always, include early changes in the developmental pathways in one or both of the diverging lineages, earliness is not in itself a sufficient descriptor of what is going on. A novelty characterising a major clade often involves a change in development that is in some sense 'special' as well as 'early'. Returning to the example of the chelonian shell: getting the pectoral girdle to develop inside the ribcage rather than outside, as in the (unknown) chelonian ancestor and in all other tetrapod groups, would appear to involve a quite distinct form of developmental reprogramming (Arthur 2000) from the quantitative changes in the size, shape and position of this girdle within either chelonians or typical tetrapods.

This is just one specific example of the general difference between qualitative and quantitative changes. Another angle on this difference was taken by DArcy Thompson in relation to his method of geometric transformations (Thompson 1917). This quantitative method could adequately (and beautifully) describe the evolutionary changes that produced the different members of a particular group. But it could not be extended beyond a certain level of divergence. DArcy Thompson himself pointed this out, and argued 'that discontinuous variations are a natural thing, that "mutations" or sudden changes, greater or less - are bound to have taken place, and new "types" to have arisen now and then.'

So, for now we must be satisfied with the following tentative conclusion about conflicting hypotheses on the nature of the small minority of divergences that lead ultimately to the production of high-level sister-clades characterised by different novelties - i.e. to mega-evolutionary change. There are not just two such hypotheses but three. The first, that mega-evolutionary changes are something quite apart from 'routine' ones, in the sense of non-overlapping sets, can in my view be rejected on the basis of the available evidence. The second, that mega-evolutionary divergences are statistically different from their lower-level counterparts, cannot. The third, that all levels of evolution are the same in both absolute and statistical senses, also cannot be rejected. Therefore, the question of whether evolution is a scale-dependent process, or a scale-independent one, remains open.

0 0

Post a comment