Introduction—development and phylogeny of the arthropods: Darwin’s legacy
- First Online:
- Cite this article as:
- Deutsch, J.S. Dev Genes Evol (2006) 216: 357. doi:10.1007/s00427-006-0089-0
- 98 Views
In the present essay, I first recall the genealogical concept of classification settled by Charles Darwin in the Origin of Species. Darwin tightly linked what we now call phylogeny and development. He often insisted to take into account embryonic and larval characters, most often using as examples his favourite animals, the cirripedes. Then I discuss remaining problems, and also perspectives, to address the link between phylogeny and development in the modern terms of molecular and cladistic phylogenetics and of molecular and genetic developmental biology.
Darwin and phylogeny
Darwin’s theory of evolution is based on two concepts: natural selection and, not less important, descent with modification, without which the mere effects of selection could not lead to evolution. Darwin’s views on classification are developed in the 13th chapter of the Origin of Species. He wrote: ‘Naturalists try to arrange species, genera and families in each class on what is called the Natural System. But what is meant by this system? [...] Many naturalists believe that it reveals the plan of the Creator. I believe that propinquity of descent—the only cause of similarity of organic beings—is the bound, hidden as it is by various degrees of modification, which is partially revealed by our classifications’ (pp. 398–399 in Darwin 1859). This is, in clear terms, the direct application of the descent with modification principle to the question of classification. This idea is so important to Darwin that he repeated it twice in almost the same words a few pages further (pp. 414 and 427 in Darwin 1859).
Darwin and development
The word development was not used during Darwin’s time. Somewhat surprisingly for modern biologists, the word evolution was used for what we call development until the end of the 19th century, and even often during the first half of the 20th. The evolution of a frog thus meant what we now call the development of a frog, from the egg to the adult, through the metamorphosis. Darwin mainly used the word embryology where he included post-embryonic development. He did not make much difference between embryonic and post-embryonic development, but rather between embryo and larva on the one hand and adult on the other. He wrote: ‘Larvae are active embryos’ (p. 434 in Darwin 1859).
In several instances, the Origin is a plea for integrating embryological and larval characters in the practice and thinking of naturalists: ‘The structure of the embryo is even more important for classification than that of the adult.’ (p. 427 in Darwin 1859). To illustrate the point, Darwin often used the examples of the cirripedes.
Cirripedes are strange animals. The most commonly known cirripedes are the barnacles, can be found settled on the rocks on the seashore. They were clearly recognised as crustaceans by the scientific community only when a British army surgeon, J. Vaughan Thomson, showed that the larvae preceding the settlement of the barnacle Balanus are nauplii, a typical crustacean larval form (Thomson 1830, cited in p. 9 in Darwin 1854; p. 2 in Anderson 1994). Evidence was brought by the larva. Darwin wrote: ‘Even the illustrious Cuvier did not perceive that a barnacle was, as it certainly is, a crustacean; but a glance at the larva shows this to be the case in an unmistakable manner’ (p. 420 in Darwin 1854).
Without hesitation, cirripedes can be qualified as Darwin’s favourite animals. Indeed, he spent no less than 8 years studying them and he published two books on living barnacles (Darwin 1851, 1854), and two shorter monographs on fossil ones (for accounts of Darwin’s work on cirripedes, see Newman 1987; Love 2002; Deutsch et al. 2004). The period of these publications is worth considering. Indeed, it lies between the first drafts of his theory (1842 and 1909, published by his son Francis in 1909) and the famous letter read at the Linnean Society the same day as that of Wallace that led to the publication of the first edition of The Origin of Species in 1859. Darwin’s books are still considered as references in cirripede studies (Newman 1987). Darwin wrote in his autobiography: ‘my work [on the cirripedes] was of considerable importance to me when I had to discuss in The Origin of Species the principles of a natural classification’ (Darwin 1892). Indeed, what could be a better example to illustrate the emphasis that Darwin put on the importance of developmental characters in systematics?
After Darwin, the link between development and evolution was perhaps over-amplified by Ernst Haeckel with his Biogenetic Law summarised by the famous phrase ‘Ontogeny recapitulates phylogeny.’ The excesses of Haeckel led to the reactive attitude of Wilhelm Roux and others that launched the mechanistic approach of developmental biology (Entwickelungsmechanik), willingly solely experimental, rejecting any philosophical and too speculative link between development and evolution. Some great figures, such as Richard Goldschmidt and Conrad Waddington, were able to stand on both legs, having accomplished remarkable work in the field of experimental developmental biology without rejecting genetics and evolutionary thinking. Without doubt, they were forerunners of the evo–devo present thinking, but remained out of the main stream of the biology of their times. Then the mechanistic approach to development became genetical, originating from Drosophila developmental genetics exemplified by the analysis of homeotic genes by Lewis (1978) and of segmentation genes by Nüsslein-Volhard and Wieschaus (1980). A step further, developmental genetics became molecular, leading to the discovery of the homeobox (McGinnis et al. 1984; Scott and Weiner 1984); surprisingly, a set of genes, similar by their homeotic function, share a common motif. This means that they are issued from a common ancestor gene by descent and modification, as predicted by Lewis (1978). It was rapidly followed by the discovery of Hox genes in many other animals, including vertebrates, and at the general surprise of that time, that mouse Hox genes’ function is homeotic as in the fly. This was the decisive shot that launched evolutionary developmental genetics. It is now familiar to think that a common genetic toolkit (Carroll et al. 2001) inherited from a common ancestor is used in building the body during development in animals, at least in all Bilateria. In the early 19th century, on a comparative anatomy basis, Geoffroy Saint-Hilaire (1830 in p. 236 in Le Guyader 1998) provokingly hypothesised homology between all animal body plans, thus predating on a solely structural ground Darwin’s evolutionary views. At the dawn of the 21st century, our genetic and molecular evo–devo approach vindicates these insightful ideas.
Arthropod development and phylogeny
In present times, developmental biology and phylogenetic studies run closer to each other. It is clear that comparative developmental biology needs reliable phylogenies to correctly interpret developmental data in an evolutionary framework. In this respect, the arthropod situation is far from satisfactory. At present, with some exceptions (Willmer 1990; Fryer 1997), the monophyly of Arthropoda is generally recognised, rejecting previous Mantonian polyphyletic views (Manton and Anderson 1979). Onychophora and Tardigrada are regarded as the sister groups of this clade, but their relative position is not clear. Other questions are: Are the Pycnogonida the sister group of all other extant arthropods or are they included within the Chelicerata as sister group of Euchelicerata (horseshoe crabs and arachnids)? Within the Arthropoda, the relative groupings between the four classical classes or sub-phyla, namely, Chelicerata, Crustacea, Hexapoda and Myriapoda, were recently reconsidered. Based on morphological, molecular and evo–devo data, the traditional grouping of Myriapoda + Hexapoda as Atelocerata has been demised (Telford and Thomas 1995) in favour of a grouping between Hexapoda + Crustacea as Pan-Crustacea or Tetraconata (Dohle 2001; Richter 2002). As first proposed by Nielsen (1995) and Averof and Akam (1995), the Crustacea could even be paraphyletic, with the Hexapoda included within. If true, the question on which is the sister group of Hexapoda amongst Crustacean sub-classes remains. The monophyly of Myriapoda (Dohle 1997) and of Hexapoda (Nardi et al. 2003) were challenged. An even more unexpected phylogenetic hypothesis was recently proposed on molecular grounds, challenging the traditional Mandibulata assemblage (i.e. Myriapoda + Hexapoda + Crustacea) in favour of a closer link between Chelicerata and Myriapoda in the well-named Paradoxopoda (Mallatt et al. 2004) or Myriochelata (Pisani et al. 2004). This situation of intense phylogenetic debate is the same at a shorter evolutionary distance with insufficient (morphological and molecular) modern phylogenetic analysis in some groups and hot debates about others, such as within Hexapoda, the position of the Strepsiptera amongst Holometabola and recurrent problems about the base of the hexapod tree; and within Crustacea, monophyly or not of Maxillopoda, amongst others.
Molecular phylogenies are mainly based on so-called housekeeping genes such as ribosomal RNAs and/or proteins involved in the mechanics of transcription and translation. This was a wise choice at the dawn of molecular phylogenetics because the homology of these molecules leaves no doubt throughout the whole living world. In addition, it could reasonably be assumed that there is no differential selective bias that may lead to artefacts, considering that these basic cellular functions are equally necessary in all metazoans. However, this assumption turned to be wrong: Differential rates of evolution were found because of differences in rates of reproduction or other unknown causes, leading to the now familiar long branch attraction artefact in phylogenetic tree re-constructions (e.g. Philippe et al. 1994). We now know that metazoans share a common set of developmental genes. Considering that characters shaping the body plans were instrumental in building the morphology-based phylogeny of metazoans, why not turn to use developmental genes to build an accurate phylogeny? Mutatis mutandis, this would be a way to apply on molecular data Darwin’s advice to take into account developmental characters (see Deutsch 1997). At the large scale of the whole Metazoa (de Rosa et al. 1999) and on Arthropoda (Cook et al. 2001), this approach was tried with some success using Hox genes, although many problems still remain (Telford 2000). As a net result, these analyses supported previous molecular analyses based on housekeeping genes, which is worth something, but one would expect more. Adding other developmental genes might help to clarify some unresolved or contradictory nodes of arthropod phylogeny. With more and more complete genomes available, despite a wretched bias towards holometabolous insects in the taxon sampling, it might be possible in the near future to take into account not only the sequences, but also duplications and losses of developmental genes that may be critical in animal body plan evolution (Ohno 1970; see discussion in Nam and Nei 2005).
An even more direct link between comparative developmental genetics (evo–devo) and phylogeny would be to take into account in phylogenetic analyses not only gene sequences, but also gene expression patterns (see Cracraft 2005). How can we handle a phylogenetic analysis of gene expression patterns? The starting point of many phylogenetic analyses is a two-dimensional matrix, one dimension being the character list (or its molecular equivalent, sequence alignment), the other the taxon sample. If we have a look at the taxon file, data are scarce; within the Arthropoda, the number of species where evo–devo studies were undertaken is not more than a dozen or so. Then, we shall see that for character A, a certain panel of species was studied, but not the same for character B. Thus, the chessboard of the data matrix will resemble an Emmental cheese: more holes than material. This is a severe drawback. However, we can be rather optimistic. Indeed, molecular techniques are going faster and cheaper, allowing more genes (cDNAs) to be cloned faster in a wider panel of species—and help is provided by comparison with already known sequences. Automatic devices are developed to perform in situ hybridisations. We can thus hope to fill in the blanks in a not so far future.
Let us now consider the character file. Phylogenetic systematics distinguishes character (e.g. second antenna) from character state (e.g. presence vs absence). Dealing with gene expression patterns, character could be expression of gene G, character states expression vs no expression. The definition of the character expression pattern of gene G is critical. A pre-requisite to character definition is that character states should be homologous (primary homology in cladistic meaning) or the data matrix would be meaningless. For instance, in the data matrix, if the character state absence referred to the second antenna in taxon T1, it should not correspond in the same column to the character state presence of the first antenna in taxon T2. This implies that antenna is not a valid character; character definition must ensure what we call antenna in taxon T1 is homologous (as a primary hypothesis) to antenna of taxon T2. We have to distinguish two characters here, namely, first antenna and second antenna, but there is not always a very obvious simple solution. Paralogy, i.e. when several homologous genes are present in the same genome due to gene duplication, is the molecular equivalent to serial homology in morphology. A phylogenetic analysis of gene expression patterns must ensure that orthologous genes are concerned.
More problematically, we need to precisely specify what expression pattern means, i.e. where and when the gene G is expressed. The wingless (wg) gene for example, is expressed in various structures and at various times during Drosophila development (pleiotropy). It would be meaningless to compare wg expression during segmentation in Drosophila to that of a wg homologue during appendage formation, for instance, in a crustacean. It follows that the domain of expression has to be a priori specified in the definition of the character itself. Primary homology between the domains of interest is thus implicitly or explicitly hypothesised before any phylogenetic analysis. It follows, to avoid circular reasoning that primary homology between organs or any parts of the body taken as expression domains must be defined independently from gene expression patterns. In practice, primary homology can only be based on structural, i.e. morphological, grounds.
This problematic link between genes and developing structures was raised many years ago by a forerunner of evo–devo, one of the rare scientists who made a link between genetics and development during the first half of the 20th century, Gavin de Beer. Even before the publication of his most famous book Embryos and ancestors, he wrote as early as in 1938, in conclusion of an astonishingly insightful analysis of the scarce data available at his time on the impact of genes on development: ‘It is clear that characters controlled by identical [i.e. homologous] genes are not necessarily homologous. [...] It is clear that homologous characters need not to be controlled by identical genes. [...] The homology of phenotypes does not imply the similarity of genotypes. [...] It follows, therefore, that the best criterion for homology is comparative anatomy’ (pp. 66–67 in de Beer 1938).
The first consequence is that comparative developmental biology (evo–devo) cannot replace comparative anatomy and descriptive embryology, but is rather dependent on them. A large body of thorough work in the field of comparative anatomy has accumulated in the past. However, going back to the good old books may not be sufficient. Indeed, the issues we are now addressing are sometimes different from those addressed during the 19th and 20th centuries, such that we cannot find answers to all our present questions. Moreover, we now have novel technical means, e.g. antibodies and/or fluorescent transgenic proteins as in vivo markers, confocal microscopy and new 3D software that enable us to describe anatomy in more detail and more completely during the whole development. We thus need to re-visit comparative anatomy of the taxa and the organs that we are interested in. This necessary new comparative anatomy is a huge task, but is also quite an exciting one.
The second consequence is of greatest importance. In some cases, the development of homologous structures is foreshadowed by the expression of homologous genes, but it is not always true. We cannot hypothesise homology on a firm basis on gene expression patterns only. As already pointed by several authors, we are not allowed to state: ‘expression domains of homologous developmental genes give rise to homologous structures’ (Bolker and Raff 1996; Abouheif et al. 1997; Scholtz 2005).
Discrepancies between morphological/structural homology and expression patterns of homologous genes (called homocracy by Nielsen and Martinez 2003) are observed: (1) when non-homologous genes are expressed in otherwise seemingly homologous structures and (2) when homologous genes (and sometimes entire gene networks) are expressed in similar but non-homologous (homoplastic) structures. The mere statement of discrepancy is not satisfactory. I think that such discrepancies could in fact be very informative on developmental and evolutionary processes.
As an example of the first case, Damen (2002) showed that during segment determination in the spider Cupiennius salei, the Wnt family gene expressed in some cells that are homologous to that of wg-expressing cells in Drosophila is not the orthologue Cs-wg but a paralogous gene, Cs-Wnt-5-1. Thus, during a critical developmental process, similar, but non-homologous, genes are expressed in homologous cells. It is worth noting that it involves members of a gene family that may present functional redundancy (see Nei 2005). This is indicative of some flexibility in the genetic basis of the segmentation process, leaving open some kind of evolvability (Kirschner and Gerhart 1998).
As an example of the second case, Telford and Thomas (1998) and Damen et al. (1998) showed that Hox genes are not expressed in chelicerates in the way predicted if their expression domains were homologous to that of insects along the classical hypothesis of absence of a deutocerebral segment. Then Mittmann and Scholtz (2003) showed without ambiguity, the presence of a deutocerebrum homologous to that of other arthropods in the chelicerate Limulus. In that case, the discrepancy was solved by re-interpretation of structural homologies, thanks to new data. Gene expression patterns data were instrumental in re-considering an old debated issue.
More intriguing issues have to be considered when homologous genes are expressed in functionally similar (analogous) but obviously non-homologous structures. The eye/Pax6 case is the most famous and heavily debated example (Gehring and Ikeo 1999, Pichaud et al. 2001). We have to take it seriously to determine where the level of homology is (Dickinson 1995; Abouheif 1997; Scholtz 2005), which parts of the process are homologous and which are not, and, if there is repeated recruitment, whether there is any underlying molecular mechanism that may favour or explain it.
From the examples above, it appears to me that phylogenetic treatment of gene expression patterns would be a real achievement of the evo–devo approach. From the phylogenetic point of view, a number of question marks are left in our present knowledge of arthropod phylogeny. It is hoped that gene expression patterns during development would provide new and phylogenetically informative data that would help resolve present uncertainties of the trees. As for the developmental biology approach, the benefits could be even more important. Indeed, we now know that a relatively limited developmental genetic toolkit (transcription factors and signalling molecules) is common to all Bilateria, and maybe all Metazoa (Carroll et al. 2001). The present challenge is to understand how such a limited genetic panel is used during development to generate the huge morphological diversity we observe. The still unlocked black box in evo–devo, and maybe the most important problem in the whole present biology, is between genotype and phenotype. Given the homology of genes, what is the balance between descent with modification, i.e. homology, and recruitment or evolutionary tinkering (Jacob 1977), i.e. homoplasy, in the generation of new morphological characters? To me, this crucial issue cannot be resolved, or even addressed, out of the framework of a rigorous evolutionary thinking, that is, without the help of phylogenetic analysis.
As a conclusion of the present essay, I warmly advocate to view the construction of phylogenetic matrices including evo–devo data as a goal. This will need endeavour from both the systematic side and the developmental side. Systematists rarely include developmental characters and developmental genes in their analyses. On the other side, in most instances, developmental biologists seldom take into account the power of interpretation provided by the cladistic approach.
To reach that goal, I would suggest for future co-operative work that (1) that teams would include both morphologists and developmental biologists; and (2) projects would not focus on a very small number of species, erected as model species, as in the 20th century fashion, but rather focus on biological problems, that is, on characters.
I thank Eric Quéinnec for his comments on a previous draft of this manuscript and for signalling Cracraft’s paper. I am grateful to Gerhard Scholtz and to an anonymous referee for careful reading of this manuscript and helpful suggestions.