Main results‎ > ‎evo-devo‎ > ‎Nature‎ > ‎nrg Focus‎ > ‎


Written in stone: fossils, genes and evo-devo
Rudolf A. Raff
Nature Reviews Genetics 8, 911-920 (December 2007)


Nature Reviews Genetics 8911-920 (December 2007) | doi:10.1038/nrg2225

Focus on: Evo–devo

Written in stone: fossils, genes and evo–devo

Rudolf A. Raff1  About the author

Top of page


Fossils give evo–devo a past. They inform phylogenetic trees to show the direction of evolution of developmental features, and they can reveal ancient body plans. Fossils also provide the primary data that are used to date past events, including divergence times needed to estimate molecular clocks, which provide rates of developmental evolution. Fossils can set boundaries for hypotheses that are generated from living developmental systems, and for predictions of ancestral development and morphologies. Finally, although fossils rarely yield data on developmental processes directly, informative examples occur of extraordinary preservation of soft body parts, embryos and genomic information.

Evolutionary developmental biology (evo–devo) is the study of how developmental processes evolve to produce new patterns of development, new developmental gene regulation, new morphologies, new life histories and new behavioural capabilities1. It also is about how established developmental processes influence what is evolutionarily possible under selection. Much of the current evo–devo research centres on the evolution of developmental genetic mechanisms. The first reason for this focus is the exciting discoveries in developmental biology that show that a restricted set of developmental regulatory genes, such as the Hox genes, are phylogenetically widely shared in patterning development in each generation across the diversity of the animal kingdom2. The second reason is the availability of powerful tools in developmental genetics that make it possible to investigate the roles of genes in development and to define developmental mechanisms. However, evo–devo is multidimensional. In its history, the mechanistic link between evolution and ontogeny arose from several disciplinary strands, including comparative embryology, morphology, genetics, developmental biology, evolutionary theory and palaeontology, which together defined the research agenda of evo–devo3, 4. Further, evo–devo is tied to geological time through the necessity of placing developmental evolution in its context within geological history. This involves finding when and how rapidly developmental features evolved, and raises the need to integrate developmental evolution into major transitions in the evolution of body plans.

However, even conceding that evolutionary time is important, how does studying the fossil record contribute to what has become a highly molecular and genomic subject? Why won't molecular clocks and comparisons of living diversity allow us to estimate the timing of evolutionary events? We would like to describe the common ancestor of protostomes and deuterostomes, the major superphyla of bilaterian animals. Can we not do this by comparing the morphologies, developmental processes, developmental gene networks and genomes of two living exemplars, Drosophila melanogaster, the genetically and developmentally best-understood protostome, and mouse, the genetically and developmentally best-understood deuterostome? The protostome–deuterostome common ancestor should exhibit the shared features of these two living protostome and deuterostome organisms. Could this logic, using living model systems, be applied at other levels of evolutionary innovation — for example, the origins of tetrapod limbs from fish fins, the evolution of mammalian middle ear bones from the reptilian jaw, the loss of limbs by snakes among the tetrapods or the gain of wings by insects among the arthropods — all without need for fossils? Although it might be a tempting thought, without fossils there would be no dating of evolutionary events, nor any direct evidence of past developmental features.

The afterlives of fossils

Developmental genetic comparisons are unsurprisingly confined to living organisms. What information might we expect non-living extinct creatures to add to what we can learn from living beings? Fossils preserve only a fraction of the structural information of live animals (contrast a skeleton or shell with the entire soft anatomy of a living creature) and, except for Pleistocene fossils of 100,000 years old or younger, no DNA is preserved5, 6. Alas, we will never have any trilobite or dinosaur genomes, although at least one extinct human genome — of Neanderthals, who became extinct about 30,000 years ago — is in progress. These data will allow comparisons between genomes that diverged approximately 500,000 years ago to allow an understanding of something about the genome of the common ancestor of us and our Neanderthal cousins7. Fossils contribute a suite of data about evolution that are otherwise inaccessible (Box 1), and give us our only direct information about extinct organisms, including especially features of body design that are now lost. We would not otherwise know about the existence of extinct major taxa such as trilobites, ammonites or non-avian dinosaurs, all animals with spectacular evolutionary histories that are discernable only from the fossil record. Fossils also give us vitally important direct data on primitive character states in extinct lineages of morphological features that are homologous to those in related clades living today. This knowledge has emerged as crucial as we try to reconstruct ancient homologous developmental and genomic features and their subsequent evolution8.

If the suggestion made in the introduction is correct, we should be able to project the features of living model taxa backward in time and successfully infer ancestral states without data from fossils. However, new fossil discoveries of exceptionally preserved early to mid-Cambrian faunas show how far off morphological extrapolations that are based only on living taxa are likely to be. Members of the recently discovered Cambrian fossils of extinct animals called halwaxiids make this point brilliantly9, 10, 11. These creatures share features from now distinct phyla — molluscs (radula and foot), annelids (setae) and brachiopods (setae and shell) (Fig. 1). But the halwaxiids bear little resemblance to living members of any of these phyla, and their precise placement in phylogenetic trees is still in discussion.

Figure 1 | Example of a metazoan stem taxon.
Figure 1 : Example of a metazoan stem taxon. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.comThe halwaxiid, Orthrozanclus, is a basal lophotrochozoan from the mid-Cambrian Burgess Shale. The body is covered by sclerites and an anterior shell. This and other halwaxiid taxa combine features of now distinct phyla. Such forms provide a concrete look at the early stages of the evolutionary history of a group at combinations of developmental features that no longer exist. Reproduced with permission from Ref. 11 © (2007) American Association for the Advancement of Science.

The basal forms of other groups whose ancestry has been obscure are also emerging as more Cambrian fossils are discovered and a better understanding of their features is gained. Another spectacular example is the case of echinoderms, the familiar if weird starfish, sea urchins, crinoids and others. Echinoderms are deuterostomes, and they are present with other deuterostome clades, notably chordates and hemichordates, in the lower Cambrian. No fossil of a deuterostome basal to these is known12crown group echinoderms are allpentameral, but some stem group echinoderms, although they share the porous calcarious stereom skeleton with crown group echinoderms, are bilateral in symmetry, and some possess features, notably the presence of gills, which are primitive to deuterostomes and present in chordates13. Gills are absent in later echinoderms. Without fossils like these we would not be able to comprehend the unique combination of primitive features that have now been lost, ancestral states of features found in living crown group descendants of major phyla, and how transitions between clades might occur.

Fossils can be analysed and their relationship to evo–devo established only in light of phylogeny and dating of their places in geological time.

Phylogeny: evolution of characters and organisms

Evolution is descent with modification. The pattern that this process yields is a tree-like structure of relationships among all living beings in which branches join as one goes deeper in time (Fig. 2). Branch points (nodes) within phylogenetic trees reflect speciation events that produce separate evolving gene pools (branches). The separating branches carry a record of diverging homologous genes, developmental pathways and morphologies. Phylogenetic trees have to be inferred using various algorithms applied to data sets that are generally derived from morphology or gene-sequence data from living organisms, or even from combinations of data sets. Trees now are commonly cladograms, in which branching patterns of trees are inferred explicitly from derived features that are shared by taxa, rather than from all available features14.

Figure 2 | Fossils can reveal evolutionary pathways in the evolution of body features.
Figure 2 : Fossils can reveal evolutionary pathways in the evolution of body features. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.comScenario a shows three living taxa, 1, 2 and 3, and a phylogeny linking them. There are two character states for a particular feature, represented as white and red circles. Rectangles represent the evolution of a new character state. X indicates an extinct fossil form. Several hypotheses might explain the evolution of the pattern. Scenarios bc and d show how the character states of extinct taxa from the fossil record can help to resolve the problem. In scenario b, the basal taxon is seen to have the red character state. It can be suggested that taxon a changed from the red character state to the white, and that the shared red states of taxa 2 and 3 are due to inheritance of a shared primitive feature of the clade (symplesiomorphy). In scenario c, the fossil ancestor has the white character state. The presence of an extinct sister clade to 2 and 3 that bore the red state suggests that red is a shared transformation of the clade (snyapomorphy). In scenario d, the fossil basal taxon has the white character state. Descendant taxa 2 and 3 share the red state, but the presence of an extinct fossil form lying basal to taxon 3 is consistent with the possibility of independent transformation to the red state in the 2 and 3 lineages (homoplasy).

Fossils can be used to help establish phylogenetic trees (Fig. 2), which are crucial to evo–devo because they allow the mapping of the distribution of features among lineages of related organisms. This permits the direction of evolution of features to be determined, and allows us to infer patterns of evolution of organismal features15. The utility of fossil data is shown in Fig. 2. Scenario a shows a simple cladogram for three living taxa — 1, 2 and 3. A single feature of the organism 'character' is shown. The taxa have either the white or red state of the character. Having only the living forms allows hypotheses of evolutionary changes in character to be made, but none can be chosen conclusively from the facts shown. The possession of fossils of extinct members of the clade (indicated by an X), as in scenarios b, c and d, shows three distinct possible evolutionary histories for the distribution of the red character state in taxa 2 and 3. In scenario b, red is primitive and changes to white in taxon a. In scenario c, white is primitive and red is gained in the basal 2 + 3 lineage. In scenario d, white is primitive and red might have been gained twice convergently (or lost in the extinct white lineage that is basal to taxon 3 — additional data are needed to decide).

If the reconstructed phylogeny is sufficiently robust, we often discover that apparently homologous features shared between crown groups were acquired independently in distinct lineages by convergent evolution, as suggested in scenario d of Fig. 2. Convergences that are identified in this way are called 'homoplasies'. Cladograms allow the important definition of stem and crown lineages as shown in Fig. 3.

Figure 3 | Finding pattern and time estimates from a phylogenetic tree.
Figure 3 : Finding pattern and time estimates from a phylogenetic tree. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.coma | This tree contains two living crown groups and two extinct stem groups. The red oval marks a node on the tree, in this case, the basal node for crown group A. The trees in parts b and c contain the same topology, but represent two attempts to date the divergence time of crown groups A and B from fossil data. b | The fossil (represented by an asterisk) is a basal member of the B clade, and gives a good minimal estimate. c | In this case, the fossil is actually a member of the stem group A–B that is basal to both crown groups, but possessed only primitive features. It is here mistakenly thought to represent a basal member of the B clade as no actual member of the basal B lineage has been found. Note the longer divergence time that results in the inferred time of divergence (tA–B). A detailed discussion of trees in the radiation of bilaterians is given in Ref. 70.

Despite claims that the relative incompleteness of fossil remains reduces their information content, features from fossils have been important in inferring phylogenetic relationships that could otherwise be reconstructed only on the basis of living forms, which might have lost features of past extinct members of the clades involved in the analysis16, 17, 18, 19. Further, data from fossil taxa provide a higher density of species sampling, thereby improving the accuracy of phylogenetic trees, because data from fossils can add informative new features from primitive clades. These features can polarize the direction of evolution of homologous features in living clades; that is, they allow us to determine the order of gain, loss or modification of features. Significant errors would result without these data because it would be more difficult to distinguish important primitive features from derived features. Fossils also reveal ancient relationships among clades that have since been severely pruned of relatives by extinction, and break long, naked branches in phylogenies that are based only on living taxa.

For example, the tetrapods represent an enormous radiation of terrestrial vertebrates. Many tetrapod clades are still living; however, the closest living relatives of tetrapods are sparse in number and diversity, comprising only three species of lungfish and a single species of coelacanth. The three-taxon tree that includes only tetrapods, lungfish and coelacanth branches is simplistic, and shows, for instance, that lungfish and tetrapods are sister groups and share homologies such as the palate fused to the braincase. This appealingly simple phylogenetic result is just plain misleading, and shows how important fossils can be. It is only when the enormous number of extinct fossil clades of sarcopterygians are added to the analysis that lungfish become separated from tetrapods and the apparent homology of the fused palate (and internal palatal nostrils) is shown to have evolved independently in lungfish and tetrapods within a far richer tree of now extinct lineages18, 20.

The ticking of the clocks

It is crucial in the study of developmental evolution to know the time of divergence of taxa, the time of appearance of evolving features within taxa and their rates of evolutionary change. This information is derived from geochronology combined with the fossil record of living taxa. In some cases of taxa with good preservation and high prevalence (that is, fossils are common in a number of strata), the record allows relatively direct dating of fossils, in a 'fossil clock'. However, in many cases of interest, the fossil record is poor or incomplete. It is in these instances that the 'molecular clock' has come into prominence as a calibration tool to be applied to dating evolutionary events in the past for which fossils are unavailable (for example, the origin of metazoans or the evolution of novel larval forms). However, molecular clocks are only as good as the fossil data on which they are calibrated, as they are based on divergence times that have been determined from taxa with strong fossil records. Although reasonably well calibrated divergence dates have been determined, there are four kinds of potentially confounding problems associated with deriving clocks from palaeontology21, 22. These can lead to over- or underestimation of divergence times, which for a molecular clock translates into corresponding errors in the dating of events to which the clock is applied.

First, the fossil record is incomplete, so the chance of finding or correctly identifying the earliest member of a clade is low, and statistical treatment is necessary to estimate the error limits on the time of divergence. Maximum and minimum constraints must be estimated on calibration dates22. Denser sampling over several strata gives tighter confidence intervals. Second, the phylogeny must be correct; that is, do we have the node we think we have (Fig. 3b,c)? The effects of an erroneous phylogenetic tree topology on time estimates can be large. Third, uncertainties arise through the need for rock units to be correlated. This uncertainty arises from the fact that fossil-bearing strata are generally not directly dateable by radiometric dating. Igneous rocks (commonly granite, lava or volcanic ash) must be used, and these must be correlated with sedimentary rocks that lie above or below the dateable igneous rocks — it is the sedimentary rocks that actually bear fossils. Finally, the choice of gene that is used is important. A molecular clock is derived from determining the number of nucleotide changes in a homologous gene that is present in two living species for which the time of divergence from each other can be estimated from the fossil record. A rate of change is then estimated for the gene in the two lineages since divergence. Correct identification oforthologues in cases of gene duplication and divergence is vital, as is the fact that not all molecular clocks 'tick' (accumulate nucleotide substitutions) at the same rate23, 24.

The importance of choosing the right genes and taxa for clock estimations is illustrated by the problem of estimating the time of animal origins25. The origin of metazoans from a protistan (probably resembling the living choanoflagellate sister group of metazoans) ancestor has not yet been detected in the fossil record. In this case, an estimate of timing will have to derive from molecular clock estimates, which might suggest a more constrained time interval for seeking traces of this event in the fossil record.

Previous estimates of the times of divergence of metazoan phyla made using molecular clocks have varied widely, and have tended to give dates far earlier than those that are consistent with the fossil record of the Cambrian radiation and its precursors in the late Precambrian12. Peterson et al.25 recognized that rates of molecular evolution in vertebrates were significantly slower than molecular evolution in invertebrate clades, and suggested that previous use of vertebrate-derived clocks therefore erroneously dated animal origins. Use of molecular clock rates that are derived from well documented divergences of invertebrate groups in the fossil record yielded a time of divergence of bilaterians to between 656 million years ago (mya) and 573 mya. This is consistent with the existence of animal fossils (apparently including cnidarians and the earliest potential fossil bilaterians) of the Precambrian Ediacara fauna, which date from about 570 mya to the start of the Cambrian (544 mya)26. The conformity of these new molecular clock dates with the fossil record of this period, if correct, suggests that the fossils are accurately recording a geologically rapid radiation of basal animal clades in the Ediacaran and early to mid-Cambrian. However, the molecular clock dating of the metazoan radiation remains unsettled. More recent methods of analysis are being introduced to various molecular clock problems and there is disagreement about proper models27. Blair and Hedges, who used a bayesian estimation method with a large multiprotein database to estimate the divergence times of deuterostome clades, calculated that chordates diverged from hemichordates plus echinoderms about 900 mya27. This is a very deep time estimate and, startlingly, would suggest that metazoan crown-group lineages diverged long before any metazoan fossil record — an result that is unlikely but illustrates the uncertainties involved.

Molecular clocks are also crucial in estimating rates of relatively recent and rapid fossil-free evolution in evo–devo. In such cases, molecular clock rates can be derived from other kinds of geological dating. For example, in the case of the evolution of novel larval forms in starfish and sea urchins, molecular clock data (derived from the timing of separation between geminate species pairs lying on either side of the Isthmus of Panama) show that radical changes in larval form arose in time ranges from about 0.5 mya to 4 mya, confirming a rapid evolution of early developmental processes in recent geological time28, 29.

The dance of genes and fossils

Several spectacular and diverse examples have emerged that illustrate the mutually supporting roles of the fossil record and knowledge of the genic basis for the evolution of important developmental and morphological features. These come from studies in which fossils are crucial for information on basal forms and for constraining the hypotheses that are derived from evo–devo30, some of which are listed in Table 1. However, there are systematic limitations to what fossils can tell us directly about evo–devo. There are relatively few instances of the preservation of developing stages in the fossil record, and none of developmental genetic processes themselves. Nonetheless, fossils can set boundaries for hypotheses that are generated from living developmental systems. One of the strongest of these is the case of homology of digits in the bird wing. Developmental data have been used to suggest that these are digits 2, 3 and 4; however, the ever more complete fossil record of bird ancestry shows that birds are theropod dinosaurs (thus cousins of the famed Velociraptor) and constrains these identities to digits 1, 2 and 3. Molecular genetic studies of hand development in birds are bearing out the fossil-constrained prediction31.

Fossils have an important role in helping to determine which living organisms related to developmental genetic model species32 are most suitable for use in comparative evo–devo studies of developmental genetic systems, which seek living proxies for extinct basal fossil taxa. Deciding which taxa, and what they can model, must be based on an analysis of phylogenetic information as well as morphological similarity to important fossil taxa. The interplay of information among levels of investigation is shown in Fig. 4. These studies form the basis for evolutionary comparisons in which living relatives in more basal clades are studied to show less derived states of developmental genetic features, giving insights into developmental features of the now extinct ancestral basal taxa — albeit with the caveat that the further back in time the extrapolation reaches, the less similar a living proxy might be to extinct forms. Finally, fossils give phylogenetic information that is crucial to the choice of living basal taxa as experimental models, and allow testing of hypotheses derived from living species.

Figure 4 | Relationships between different levels of investigation.
Figure 4 : Relationships between different levels of investigation. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.comThe flow of information among studies of model developmental genetic systems, comparative studies of living non-model organisms that stand in for basal fossil taxa and the study of features of fossil taxa that give a view of the probable ancestral features and the direction of evolution of features.

Fossils in testing evo–devo hypotheses

In 1837, the anatomist Carl Reichert revealed the peculiar observation that the posterior part of meckel's cartilage in the developing mammalian jaw ossifies and detaches from the jaw to enter the region of the ear and become the middle ear bone, the malleus. This observation became the basis for the recognition that bones that formed the articulation of the reptilian jaw to the skull are homologous to the bones of the mammalian middle ear. This of course led to the hypothesis that the events seen in ear development reflect actual evolutionary events, a conclusion that was tested in the fossil record. There it was found that a series of jaws of animals in the transition between reptiles and mammals exhibited changes in the anatomy of the lower jaw that showed the development of a new articulation joint with the skull and the co-option of elements of the former articulation into structures connected with sound transmission33. The existence of this confirmation of developmental homology has allowed the investigation of how regulatory genes associated with jaw articulation have changed their functions. It has been shown that the transcription factor gene Bapx1 (also known as Nkx3.2) is involved in the development of the jaw joint in non-mammalian vertebrates34. Tucker et al. have further shown that Bapx1 is expressed in the primordia of middle ear bones that are evolutionarily derived from the jaw joint35. In mutant mice, the width of the malleus is reduced, but Bapx1 surprisingly does not affect the articulation between the malleus and the incus (the ancestral homologues of these — the articular and quadrate — form the joint of the non-mammalian jaw). Tucker et al. show that this change results from a loss of Bapx1 regulation of Gdf5 and Gdf6 genes, which are needed for formation of the joint.

Fossils of primitive mammals have produced a new twist to this story. Rich et al. discovered that the lower jaw of a Cretaceous monotreme still possessed an internal mandibular trough, which housed elements of the reptilian jaw that gave rise to mammalian ear elements36. The extant monotremes, the platypus and the echidna, lay eggs and belong to a clade that is distinct from the therians (marsupial plus placental mammals). The bones of the inner ear were thought to be a shared homologue (synapomorphy) of mammals; however, here is a fossil monotreme that still possesses the ancestral layout after the split from therians. Therefore, the co-option of middle ear bones has been convergent in mammal groups. Unfortunately, monotreme embryos are likely to be hard to come by for any experimental studies of gene expression in the living forms.

A recent model derived from an experimental study of the roles of inhibitors and activators of growth of molars along the tooth row gives the prospect of a molecular developmental hypothesis that can be tested in the mammalian fossil record. The inhibitory cascade model successfully predicted relative molar sizes in rodents37. When it was separately applied to various mammalian orders, it explained most of the molar size proportions of mammals — including several extinct fossil forms38.

Limbs from fins

The case of the origins of tetrapod limbs provides one of the best-documented evolutionary transitions in the fossil record, and one that can be correlated with a growing knowledge of the roles of patterning genes in the evolution of the developmental system39, 40. Studies of mice and chicks have revealed many of the regulatory genes and interactions that control the development of the three axes of the limb — the proximal–distal, anterior–posterior and dosal–ventral axes are shared between the two organisms. However, major unsolved problems lie further back in evolutionary time in tracing the origin of paired appendages and the transition from fins to legs. Fossil data show that the most primitive fossil vertebrates found in the approximately 520 million year old Cambrian rocks of China had midline unpaired fins but no paired appendages41. Paired appendages appear later in lower Palaeozoic fossil fish. Comparisons of gene expression were made in non-paired medial fins of the catshark, including Hoxd and Tbx18, which are important in specifying paired fin positions. These results suggest that the mechanisms of paired-fin development were co-opted from the development of non-paired fins early in vertebrate history42.

Figure 5 presents the anatomy of the forelimbs of several fossil sarcopterygians and two living ones (the coelacanth Latimeria chalumnae and the lungfish, Neoceratodus forsteri). The tree in Fig. 5 shows that the lobe fins, which were ancestral to hands, had to be asymmetrical. Living coelacanth and lungfish have reduced symmetrical fins. Thus, the best living proxy taxon should be not a living lobe-fin fish, but a primitive ray-fin fish with an asymmetrical fin similar to that of the ray-fin and lobe-fin common ancestor40. The paddlefish,Polyodon spathula, seems to provide good model43, 44Hoxa11 and Hoxa13 expression in P. spathula shows two phases. The second phase is in the distal mesenchyme and suggests that the distal polarity of Hoxa13 in the fin bud is more ancient than the distal expression in the tetrapod hand bud. Exclusion of Hoxa11 by Hoxa13 from the distal mesenchyme is not seen in paddlefish, unlike tetrapods. Intriguingly, late-phase expression of HoxD genes resembles the posterior- and distal-restricted nested pattern that is seen in the tetrapod limb44.

Figure 5 | Phylogeny of the forelimb in lobe-finned vertebrates.
Figure 5 : Phylogeny of the forelimb in lobe-finned vertebrates. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.comThe forelimb skeletons of extinct and living representatives of the three surviving clades of lobe fins, coelacanths, tetrapodomorphs and lungfish, are shown. The colour key refers to the three main domains of evolutionary change in the forelimb. The novelty of a hand with fingers appears in the most primitive known fossil tetrapod, Acanthostega. The living lungfish (Neoceratodus) and coelacanth (Latimeria) have reduced and highly symmetrical forelimbs. This reduces their value as living models for the events of limb evolution. Modified with permission from Ref. 40 © (2007) Blackwell Science.

Comparisons with the zebrafish, a model teleost fish, show that aspects of regulatory gene expression are conserved with tetrapods. Teleosts are highly derived ray-fin fish that diverged from the lobe-fin ancestors of tetrapods more than 420 mya. They are not primitive, but have evolved their own derived developmental features. Zebrafish fin buds, like tetrapod limb buds, exhibit the early phases of Hoxd13 expression45 — conserved use of sonic hedgehog (Shh) in anterior–posterior patterning46 and of fibroblast growth factors (FGFs) in the apical ectodermal ridge47. However, Mabee has pointed out that zebrafish and tetrapod appendages share no skeletal homologues48. Teleosts have lost the posterior portion (metapterygium) of the ancestral fin, which makes up the entire tetrapod limb. Only fossil forms and basal living fish taxa have fins that retain all primitive skeletal elements. Davis et al. suggest a phylogenetic history in which the late-phase HoxD expression pattern was present in primitive bony fish and lost along with the metapterygium in teleosts, but was retained in tetrapod-like sarcopterygians like Tiktaalik and co-opted into the tetrapod limb44.

Finally, the polydactylous feet of the earliest tetrapods, such as Acanthostega, pose a conundrum. Before the fossils of the earliest tetrapods were known, the morphology of living tetrapods was taken to show that five digits are primitive. We now know from the fossil record that the earliest tetrapods in fact had more toes — Acanthostega has eight49. The five HoxD genes expressed in distal limb buds regulate formation of the five digits in living tetrapods. Tabin suggested that the extra digits of the first tetrapods represent duplications of digits with identities that were defined by the five sectors of a primitive pattern of HoxD-gene expression50. However, the digits in these primitive hands are similar. What if Acanthostega digits were generated by a distinct ancestral Hox-gene pattern that produced eight sectors? Fossils here are suggestive, but the answer can be sought only in studies of gene expression in living basal fish like P. spathula. Without fossils, we would never have known that such a question even existed.

Evo–devo at the limits of the fossil record

Development is difficult to study in the fossil record. Sufficient developmental stages are not often preserved and, of course, the 'developing' creatures are dead and the development of no individual can be followed — only inferred stages in the fossil 'population'. Some descriptions of spectacular series of fossil ontogenies have been made (for example, of basal Cambrian arthropods)51, and there are some fossil instances where evo–devo studies of the process can be made. This has been particularly true with trilobites, for which a rich fossil record including developmental stages has been used to comparatively study modes of segment addition and analyse heterochronic processes in developmental evolution through stratigraphic time52, 53.

In most instances, even the limited direct evo–devo approaches that are possible with trilobites cannot be applied. Yet, there are cases in which it is vital to extrapolate back from living taxa to reclaim past events and complement what might be obscure in the fossil record. As the fossil record is nearly silent about the origin of metazoans and the protostome–deuterostome ancestor in bilaterian animal evolution, reconstruction of plausible ancestral features depends on phylogenetic reconstruction, and the projection of genome composition and developmental mechanisms from living taxa into the past. Living proxy taxa have been crucial in reconstruction of features of basal forms at crucial divergence points in metazoan evolution before the Cambrian explosion. Unusually well-preserved animals from the Cambrian explosion, notably from the soft-bodied mid-Cambrian Chengjiang and Burgess Shale faunas, provide an amazingly clear picture of the morphological features of animals that existed approximately 520 mya and 505 mya. Most of these fossils belong to recognizable stem taxa related to living phyla. The primitive Cambrian arthropods, vertebrates, lobopods, echinoderms, molluscs and others are all products of a prior radiation of animal forms that can so far be seen only dimly in the fossil record.

It has been suspected that the rapid evolution of body plans just before the Cambrian explosion might have been in part due to less constrained development in early metazoans. Suggestive data connecting developmental variability to the Cambrian explosion have recently emerged54. Trilobites were a major and diverse part of the arthropod radiation, appearing in the early Cambrian and persisting for another 270 million years before becoming extinct. Morphological variation within individual species is highest in the early and mid-Cambrian fossils. After that, individual species variability falls, suggesting that ecological or developmental constraints increased as a hypothetical phase of early metazoan body plan 'experimentation' drew to a close54.

As discussed above in relation to the timing of early metazoan evolution, as yet no fossil traces have been found of the very earliest animals. Some evidence about the appearance, rates of evolution and ecological effects of the first bilaterian animals derives from geochronology and geochemistry55, 56, 57. Among the key steps that are apparently missing from the fossil record of life in the period before the Cambrian are the origins of the first multicellular animals, the first two-layered (diploblastic) forms, the first bilaterally symmetrical animals, and the last common ancestor of protostomes (annelids, molluscs, arthropods and others) and deuterostomes (hemichordates, echinoderms and chordates, which include vertebrates). These organisms were probably small and had poor potential for preservation as fossils, although trace fossils recording the movement of likely bilaterians on the sea floor occur for the first time in the fossil record. Body fossils of some late Precambrian bilaterian animals are preserved along with apparent diploblastic Ediacaran age forms. Thus, tracing the early steps of metazoan evolution depends on conclusions drawn from phylogenies based on gene sequences, molecular clocks and the use of proxies to stand in for the genomes of basal animal groups.

Two approaches to proxy extrapolations are used. One is to exploit data on genes and developmental organization from genetic model systems, such as D. melanogaster and mice, in the reconstruction of the developmental genetics of the protostome–deuterostome ancestor. Two important insights have been gained by this approach. A 'toolkit' of genes used across the bilateria was identified by Erwin and Davidson58. Some show conserved tissue use, for example tinman, which is associated with heart development. The authors note that genes associated with structures in this way are likely to represent general features of development, rather than the derived structures of living forms. Two concepts have also been merged. The first is that developmental modules are seats of genetically discrete organization that define domains within a developing organism, and that these modules undergo evolution59. The second is the hypothesis that the most basic modular gene expression networks, called 'kernels', provide the crucial upstream regulation of body-plan development at the phylum level60. It is suggested that conserved kernels arose in the evolution of phylum and superphylum characteristics, with subsequent evolution lying in lower-level elements in the gene-network hierarchy60.

The alternative approach to extrapolating from model systems is the use of members of living basal taxa as proxies to provide more direct approximations of major ancient branch points in metazoan evolution. Currently, sponges are used to represent the most basal metazoans, sea anemones to represent ancient diploblastic metazoans and acoel flatworms for the most basal bilaterians. For example, in seeking the root of metazoans, molecular phylogeny indicates that sponges are basal, and there is an active and successful search in progress of genes encoding important proteins of animal development, such as transcription, adhesion and signalling proteins61, 62. Thus, it should be possible to find precursors to genes that are important in the development of more advanced metazoans, and to map such features as gene-family expansion and specialization of functions. The issue that will arise, of course, is how well the modern sponge proxy represents the features of the body plan and development of the last common ancestor of living animals.

Small fossils, big surprises

Occasionally, the rocks surprise us by yielding unexpected fossils that exceed the expected limits of fossilization. Feathered dinosaurs and Cambrian soft-bodied animals are among these, along with, against all odds, well-preserved marine embryos and larvae from the late Precambrian and Cambrian63, 64. Thus, finally, fossil evidence can even extend to embryos and larvae, and is starting to contribute to the unravelling of ancient developmental patterns and the evolution of life histories. These fossils show egg sizes, patterns of cleavage and, in some cases, transformations to larvae. In addition, advanced imaging reveals internal features such as cleavage planes, cell numbers and apparent cytoplasmic structures65, 66, 67. Information is for the first time becoming available on the development and the life-history evolution of the first animals that produced the complex feeding marine larvae of today68, 69.

Where to next?

New fossil discoveries expand our knowledge of crucial problems, as has been very much the case for the evolution of tetrapod limb origins. Strong possibilities also exist where the fossil record is currently weak, for example, in the early steps of the metazoan radiation. Existing fossils and improving molecular clocks can suggest crucial time windows that guide the exploration for appropriate new fossil-bearing strata. The introduction of new approaches from other disciplines will also have a pronounced role on the uses of fossil record data. The experimental study of mechanisms of preservation of embryos is one such direction. Another will arise from the use of genomic sequences from organisms selected on the basis of phylogenetic position. Comparative genomics reveals a kind of molecular fossil record of events such as genome duplication, gene-family expansion and evolution of new genes. Concepts from genomics and gene networks are likely to be used to reconstruct hypothetical developmental transitions in evolution. New fossils with an unusual preservation of soft tissue will allow tests of these hypotheses, and will generate further hypotheses that will in turn be tested by developmental genetic studies of living proxies.


Top of page


I thank M. Friedman and M. I. Coates for the figure used in Figure 5, S. Conway Morris and J. B. Caron for providing Figure 1, A. Moczek and P. C. J. Donoghue for helpful advice and M. I. Coates, E. C. Raff, J. W. Valentine and two anonymous reviewers for their critical reading and suggestions.

Top of page


  1. Raff, R. A. Evo–devo: the evolution of a new disciplineNature Rev. Genet. 1, 74–79 (2000).

  2. Lemons, D. & McGinnis, W. Genomic evolution of Hox gene clustersScience 313, 1918–1922 (2006).

  3. Love, A. C. & Raff, R. A. Knowing your ancestors: themes in the history of evo–devoEvol. Dev. 5, 327–330 (2003).

  4. Amundson, R. The Changing Role of the Embryo in Evolutionary Thought. Roots of Evo–Devo. (Cambridge University Press, Cambridge, 2005).

  5. Austin, J. J., Smith, A. B. & Richard H. Thomas, R. H. Palaeontology in a molecular world: the search for authentic ancient DNATrends Ecol. Evol. 12, 303–306 (1997).

  6. Marota, I. & Rollo, F. Molecular paleontologyCell. Mol. Life Sci. 59, 97–111 (2002).

  7. Green, R. E. et alAnalysis of one million base pairs of Neanderthal DNANature 444, 330–336 (2006).

  8. Wagner, G. P. The developmental genetics of homologyNature Rev. Genet. 8, 473–479 (2007).

  9. Butterfield, N. J. Hooking some stem-group "worms": fossil lophotrochozoans in the Burgess ShaleBioessays 28, 1161–1166 (2006).

  10. Conway Morris, S. & Peel, J. S. Articulated halkieriids from the lower Cambrian of north Greenland and their role in early protostome evolutionPhil. Trans. R. Soc. B. Biol. Sci. 347, 305–358 (1995).

  11. Conway Morris, S. & Caron, J.-B. Halwaxiids and the early evolution of the lophotrochozoansScience 315, 1255–1258 (2007).

  12. Swalla, B. J. & Smith, A. B. Deciphering deuterostome phylogeny: molecular, morphological and palaeontological perspectivesProc. R. Soc. London B Biol. Sci. (in the press).
    These two papers show how Cambrian animal fossils can be reconstructed to give information about features of taxa that are basal to crown group phyla.

  13. Smith, A. B. The pre-radial history of echinodermsGeol. J. 40, 255–280 (2005).

  14. Sereno, P. C. The logical basis of phylogenetic taxonomySyst. Biol. 54, 595–619 (2005).

  15. Smith, A. B. Systematics and the Fossil Record. (Blackwell Scientific, Oxford, 1994).

  16. Gauthier, J., Kluge, A. G. & Rowe, T. Amniote phylogeny and the importance of fossilsCladistics 4, 105–209 (1988).

  17. Donoghue, M. J., Doyle, J. A., Gauthier, J., Kluge, A. G. & Rowe, T. The importance of fossils in phylogeny reconstructionAnnu. Rev. Ecol. Syst. 20, 431–460 (1989).

  18. Smith, A. B. What does paleontology contribute to systematics in a molecular worldMol. Phylogenet. Evol. 9, 437–447 (1998).

  19. Wiens, J. J. Can incomplete taxa rescue phylogenetic analyses from long-branch attraction? Syst. Biol. 54, 731–742 (2005).

  20. Clouthier, R. & Ahlberg, P. E. Sarcopterygian interrelationships: how far are we from a phylogenetic consensus? Geobios 19, 241–248 (1995).

  21. Donoghue, P. C. J. & Benton, M. J. Rocks and clocks: calibrating the tree of life using fossils and moleculesTrends Ecol. Evol. 22, 424–431 (2007).

  22. Benton, M. J. & Donoghue, P. C. J. Paleontological evidence to date the tree of lifeMol. Biol. Evol. 24, 26–53 (2007).
    These two papers discuss in detail the use of fossils in generating minimum divergence times for taxa, and using fossils and molecules in dating the tree of life.

  23. Bromham, L. & Penny, D. The modern molecular clockNature Rev. Genet. 4, 216–224 (2003).

  24. Welch, J. J. & Bromham, L. Molecular dating when rates varyTrends Ecol Evol. 20, 320–327 (2005).

  25. Peterson, K. J. et alEstimating metazoan divergence times with a molecular clockProc. Natl Acad. Sci. USA 101, 6536–6541 (2004).

  26. Knoll, A. H. Learning to tell Neoproterozoic timePrecambrian Res. 100, 3–20 (2000).

  27. Blair, J. E. & Hedges, S. B. Molecular phylogeny and divergence times of deuterostome animalsMol. Biol. Evol. 22, 2275–2284 (2005).

  28. Hart, M. W., Byrne, M. & Smith, M. J. Molecular phylogenetic analysis of life-history evolution in asterinid starfishEvolution 51, 1848–1861 (1997).

  29. Zigler, K. S., Raff, E. C., Popodi, E., Raff, R. A. & Lessios, H. A. Adaptive evolution of bindin in the genus Heliocidaris is correlated with the shift to direct developmentEvolution 57, 2293–2302 (2003).

  30. Peterson, K. J., Summons, R. E. & Donoghue, P. C. J. Molecular paleobiologyPaleontology 50, 775–809 (2007).

  31. Vargas, A. O. & Fallon, J. F. The digits of the wing of birds are 1, 2, and 3. A reviewJ. Exp. Zool. Mol. Dev. Evol. 304B, 206–219 (2005).
    This paper discusses how fossil data can constrain hypotheses about developmental evolution that are based on living forms. The discussion focuses on the identities of bones in the modified hands of birds.

  32. Bolker, J. A. Model systems in developmental biologyBioessays 17, 451–455 (1995).

  33. Crompton, A. W. & Jenkins, F. A. Jr. in Mesozoic Mammals (eds J. A. Lillegraven, Z. Kielen-Jaworowska & W. A. Clemens) 59–73 (Univ. California. Press, Berkeley, 1979).

  34. Wilson, J. & Tucker, A. S. Fgf and Bmp signals repress the expression of Bapx1 in the mandibular mesenchyme and control the position of the developing jaw jointDev. Biol. 266, 138–150 (2004).

  35. Tucker, A. S., Watson, R. P., Lettice, L. A., Yamada, G. & Hill, R. E. Bapx1 regulates patterning in the middle ear: altered regulatory role in the transition from the proximal jaw during vertebrate evolutionDevelopment 131, 1235–1245 (2004).

  36. Rich, T. H., Hopson, J. A., Musser, A. M., Flannery, T. F. & Vickers-Rich, P. Independent orgins of middle ear bones in monotremes and theriansScience 307, 910–914 (2005).

  37. Kavanagh, K. D., Evans, A. R. & Jernvall, J. Predicting evolutionary patterns of mammalian teeth from developmentNature 449, 427–433 (2007).

  38. Polly, P. D. Development with a biteNature 449, 413–415 (2007).

  39. Tanaka, M. & Tickle, C. in Fins into Limbs. Evolution, Development, and Transformation (ed. B. K. Hall) 65–78 (Univ. Chicago Press, Chicago, 2007).

  40. Friedman, M., Coates, M. I. & Anderson, P. First discovery of a primitive coelacanth fin fills a major gap in the evolution of lobed fins and limbsEvol. Dev. 9, 329–337 (2007).

  41. Zhang, X. G. & Hou, X. G. Evidence for a single median fin-fold and tail in the Lower Cambrian vertebrate, Haikouichthys ercaicunensisJ. Evol. Biol. 17, 1162–1166 (2004).

  42. Freitas, R., Zhang, G. & Cohn, M. J. Evidence that mechanisms of fin development evolved in the midline of early vertebratesNature 442, 1033–1037 (2006).

  43. Metscher, B. D. et alExpression of Hoxa-11 and Hoxa-13 in the pectoral fin of a basal ray-finned fish, Polyodon spathula: implications for the origin of tetrapod limbsEvol. Dev. 7, 186–195 (2005).

  44. Davis, M. C., Dahn, R. D. & Shubin, N. H. An autopodial-like pattern of Hox expression in the fins of a basal actinopterygian fishNature 447, 473–477 (2007).

  45. Sordino, P., van der Hoeven, F. & Duboule, D. Hox gene expression in teleost fins and the origin of vertebrate digitsNature 375, 678–681 (1995).

  46. Neumann, C. J., Grandel, H., Gaffield, W., Schulte-Merker, S. & Nüsslein-Volhard, C. Transient establishment of anteroposterior polarity in the zebrafish pectoral fin bud in the absence of sonic hedgehog activityDevelopment 126, 4817–4826 (1999).

  47. Nomura, R. et alFgf16 is essential for pectoral fin bud formation in zebrafishBiochem. Biophys. Res. Commun. 347, 340–346 (2006).

  48. Mabee, P. M. Developmental data and phylogenetic systematics: evolution of the vertebrate limbAm. Zool. 40, 789–800 (2000).

  49. Coates, M. I. & Clack, J. A. Polydactyly in the earliest known tetrapod limbsNature 347, 66–69 (1990).

  50. Tabin, C. J. Why we have (only) five fingers per hand: Hox genes and the evolution of paired limbsDevelopment 116, 289–296 (1992).

  51. Walossek, D. The Upper Cambrian Rehbachiella and the phylogeny of Branchiopoda and CrustaceaFossils Strata 32, 1–202 (1993).

  52. Hughes, N. C., Minelli, A. & Fusco, G. The ontogeny of trilobite segmentation: a comparative approachPaleobiology 32, 602–627 (2006).

  53. Hunda, B. R. & Hughes, N. C. Evaluating paedomorphic heterochrony in trilobites: the case of the diminutive trilobite Flexicalymene retrorsa minuens from the Cincinnatian Series (Upper Ordovician), Cincinnati regionEvol. Dev. 9, 483–498 (2007).

  54. Webster, M. A Cambrian peak in morphological variation within trilobite speciesScience 317, 499–502 (2007).

  55. Bowring, S. A. et alCalibrating rates of early Cambrian evolutionScience 261, 1293–1298 (1993).

  56. Martin, M. W. et alAge of Neoproterozoic bilaterian body and trace fossils, White Sea, Russia: implications for metazoan evolutionScience 288, 841–845 (2000).

  57. Rothman, D. H., Hayes, J. M. & Summons, R. E. Dynamics of the Neoproterozoic carbon cycleProc. Natl Acad. Sci. USA. 100, 8124–8129 (2003).

  58. Erwin, D. H. & Davidson, E. H. The last common bilaterian ancestorDevelopment 129, 3021–3032 (2002).

  59. Raff, R. A. The Shape of Life. Genes, Development, and the Evolution of Animal Form. (Univ. Chicago Press, Chicago, 1996).

  60. Davidson, E. H. & Erwin, D. H. Gene regulatory networks and the evolution of animal body plansScience 311, 796–800 (2006).

  61. Larroux, C. et alDevelopmental expression of transcription factor genes in a demosponge: insights into the origin of metazoan multicellularityEvol. Dev. 8, 150–173 (2006).

  62. Nichols, S. A., Dirks, W., Pearse, J. S. & King, N. Early evolution of animal cell signaling and adhesion genesProc. Natl Acad. Sci. USA 103, 12451–12456 (2006).

  63. Donoghue, P. C. J. et alFossilized embryos are widespread but the record is temporally and taxonomically biasedEvol. Dev. 8, 232–238 (2006).

  64. Steiner, M., Zhu, M., Li, G., Qian, Y. & Erdtmann, B.-D. New Early Cambrian bilaterian embryos and larvae from ChinaGeology 32, 833–836 (2004).

  65. Briggs, D. E. G. The role of decay and mineralization in the preservation of soft-bodied fossilsAnnu. Rev. Earth Planet Sci. 31, 275–301 (2003).

  66. Raff, E. C., Villinski, J. T., Turner, F. R., Donoghue, P. C. & Raff, R. A. Experimental taphonomy shows the feasibility of fossil embryosProc. Natl Acad. Sci. USA 103, 5846–5851 (2006).

  67. Hagadorn, J. W. et alCellular and subcellular structure of neoproterozoic animal embryosScience 314, 291–294 (2006).

  68. Dunn, E. F. et alMolecular paleoecology: using gene regulatory analysis to address the origins of complex life cycles in the Late PrecambrianEvol. Dev. 9, 10–24 (2007).
    References 63 and 68 examine two approaches to the evolution of animal larvae and life history in the metazoan radiation. Fossil data and comparative developmental molecular data are complementary.

  69. Nützel, A., Lehnert, O. & Frýda, J. Origin of planktotrophy — evidence from early mollusksEvol. Dev. 8, 325–330 (2006).

  70. Valentine, J. W. Ancestors and urbilateriaEvol. Dev. 8, 391–393 (2006).

  71. Coates, M. & Ruta, M. Nice snake, shame about the legsTrends Ecol. Evol. 15, 503–507 (2000).

  72. Cohn, M. J. & Tickle, C. Developmental basis of limblessness and axial patterning in snakesNature 399, 474–479 (1999).

  73. Xu, X. et alFour-winged dinosaurs from ChinaNature 421, 335–340 (2003).

  74. Lin, C. M., Jiang, T. X., Widelitz, R. B. & Chuong, C. M. Molecular signaling in feather morphogenesisCurr. Opin. Cell Biol. 18, 730–741 (2006).

  75. Kukalova-Peck, J. Origin of the insect wing and wing articulation from the arthropodal legCan. J. Zool. 61, 1618–1669 (1983).

  76. Jockusch, E. L. & Ober, K. A. Hypothesis testing in evolutionary developmental biology: a case study from insect wingsJ. Hered. 95, 382–396 (2004).

  77. Sun, G., Ji, Q., Dilcher, D. L., Zheng, S., Nixon, K. C. & Wang, X. Archaefructaceae, a new basal angiosperm familyScience 296, 899–904 (2002).

  78. Hernández-Hernández, T., Martínez-Castilla, L. P. & Alvarez-Buylla, E. R. Functional diversification of B MADS-box homeotic regulators of flower development: adaptive evolution in protein–protein interaction domains after major gene duplication eventsMol. Biol. Evol. 24, 465–481 (2007).

  79. Purugganan, M. D. The MADS-box floral homeotic gene lineages predate the origin of seed plants: phylogenetic and molecular clock estimatesJ. Mol. Evol. 45, 392–396 (1997).

Author affiliations

  1. Department of Biology, Indiana University, Bloomington, Indiana 47401, USA, and School of Biological Sciences, University of Sydney, Sydney 2006, New South Wales, Australia.