The central idea of evolutionary developmental biology (evo-devo) is that including information about developmental processes provides for a more complete explanation of observed evolutionary patterns, because evolutionary change is shaped not only by different processes of sorting of standing variation, such as natural selection, but also by the variation that arises at each generation, which depends on developmental systems (Fusco, 2001, 2015).

The comparative approach is one of the main instruments of analysis in evo-devo studies (Müller, 2007), serving in both reconstructing the evolution of developmental systems and in exploring how developmental processes can themselves affect organismal evolution (Moczek, 2015). Gaining access to the state of developmental characters in extinct organisms, through the study of so-called “fossilized ontogenies,” increases the power of the comparative analysis, by introducing novel, diachronic observational data (Fusco et al., 2016).

The study of fossil organisms with a focus on their development, aimed at the study of the evolution of more inclusive groups to which they belong, has been dubbed “palaeo-evo-devo” (Minelli and Fusco, 2008, p. 216), and since applied in several articles, books, symposia and websites (e.g., Haug et al., 2013; Wilson, 2013; Kliman, 2016; http://palaeo-evo-devo.info/). Paleo-evo-devo studies rest, however, on the availability of detailed records of development in fossil species, which is not at all common (Sánchez, 2012). Among the few exceptions there is a major arthropod clade, the Trilobita, which has a record of fossilized ontogenies that is among the most comprehensive for any extinct group (Hughes, 2003).

Herein, based on the results of an extended series of morphometric analysis we conducted during the last 25 years, we review the development of the middle Silurian trilobite A. koninckii (Barrande, 1846) (Figure 1) and present the first reconstruction of its post-embryonic ontogeny. The extraordinary abundance of well-preserved complete exoskeletons of this species provides an unparalleled opportunity to explore post-embryonic development in an early arthropod, so that it has recently become one of the most intensely studied trilobite species.

Current understanding of the development of this 429 million year old arthropod goes beyond descriptive patterns of growth and segmentation. Considerable information has been discovered about how specific developmental pathways were taken and how different mechanisms of developmental control operated in this animal. In this respect, the A. koninckii developmental system reveals several features that are found in modern organisms, and whose occurrence in this animal represents their most ancient evidence.

Taking advantage of these findings, the novel reconstruction of A. koninckii ontogeny presented here is not simply a series of drawings representing the main features of sequential developmental. Rather, the drawings are based on morphometric character estimates that were derived by applying the growth and segmentation dynamics revealed in our prior published work (see Supplementary Appendix). This distinguishes this reconstruction of a trilobite ontogeny from any other that we know of published to date. In order to explain and justify these estimates, this paper illustrates sequentially different aspects of A. koninckii development, in which the previous findings provide the basis for the present ontogeny reconstruction. Combining different developmental dynamics together permits us to visualize patterns of relative and absolute growth and segmentation as never before possible for a fossilized arthropod ontogeny.

Although, the development of A. koninckii cannot be claimed to represent the whole clade Trilobita (more than 20,000 species in the whole group, spread across a 250 Myr time interval), its study shows that it is possible to use data from fossils to address questions of high interest for evolutionary developmental biology, demonstrating that paleo-evo-devo is a well-defined research program (sensu Müller, 2007) in the study of evolutionary change.

In order to contextualize the investigations of A. koninckii development, firstly we outline some general features of trilobite post-embryonic development.

The onset of exoskeletal biomineralization in trilobites presumably initiated shortly after hatching, allowing fossil preservation since the earliest post-embryonic stages. Development was direct, lacking drastic metamorphosis, and the gradually changing forms of the progressive developmental stages were separated by episodes of molting, as in extant arthropods. Quantitative data on successive molt stages are available for about 80 species of trilobites, including A. koninckii (Fusco et al., 2012).

Standard descriptions of trilobite ontogeny are based on the development of articulating joints between dorsal segments (Figure 2; see Hughes et al., 2006). An anterior set of conjoined segments comprised the head region or cephalon that seemingly had a fixed number of segments throughout ontogeny. The trunk region lay posterior to the cephalon and changed during ontogeny both in the number of segments expressed and in the number of articulating joints between segments.

The earliest widely recognized phase of trilobite ontogeny is the protaspid period, during which all body segments (both cephalic and trunk) formed an undivided dorsal shield (Chatterton and Speyer, 1997). This period typically embraced a small number of stages. Later stages were characterized by the appearance of a series of articulations between dorsal segments, the first of which occurred at the cephalic-trunk boundary.

The appearance of the cephalic-trunk articulation marked entry into the meraspid period, which divided the dorsal exoskeleton into two components, the cephalon and a set of conjoined trunk segments called the pygidium. During subsequent meraspid molts, new articulations developed sequentially at the posterior of the leading pygidial segment (in most species, one per molt), resulting in a set of articulating trunk segments, collectively called the thorax, that were located in front of the pygidium. In parallel, new trunk segments appeared sequentially in a subterminal zone within the pygidium. The rate at which segments were released into the thorax relative to the rate at which segments were expressed in the subterminal growth zone determined the number of segments allocated to the meraspid pygidium, and varied among species (see Hughes et al., 2006). The meraspid pygidium thus comprised a dynamically changing complement of segments (Minelli et al., 2003). Progressive release of trunk segments into the thorax continued until the animal entered the final, holaspid period of development which was characterized by a stable number of thoracic segments.

Like many extant arthropods (including several myriapod lineages, Minelli and Fusco, 2013), trilobites displayed hemianamorphic development (Minelli et al., 2003), i.e., they added new trunk segments during post-embryonic life until a fixed adult number was reached at a given stage, after which growth and molting continued but without any further increase in segment number. Trilobite ontogeny can thus be subdivided into an anamorphic phase, during which new trunk segments appeared at the rear of the trunk, and a subsequent epimorphic phase, during which the number of trunk segments remained constant (Figure 2). There was variation among (and in some cases, possibly within) species in whether onset of epimorphosis preceded (protomeric development), occurred synchronously with (synarthromeric development), or succeeded (protarthrous development) onset of the holaspid phase (Hughes et al., 2006).

Large numbers of articulated exoskeletons of A. koninckii covering a broad span of juvenile and adult ontogeny occur on multiple bedding plane surfaces within an 1.4 m interval of mudstone on Na Černidlech Hill near Loděnice in the Czech Republic. This interval likely accumulated over a period of a few thousand years (Hughes et al., 2014). Both physical evidence preserved in the sedimentary rocks hosting the fossils and in the form and occurrence of the biota itself (which includes many species other than A. koninckii) suggest an environment of fluctuating oxygen availability in which numbers of A. koninckii, which had a morphology commonly associated with oxygen-poor settings, periodically bloomed (Hughes et al., 2014). The locality was first excavated by the French-Czech paleontologist Joachim Barrande, likely in the 1840's, and almost all of the samples studied come from that excavation.

Aulacopleura koninckii belongs to the order Aulacopleurida, a long-lived and morphologically quite conservative group with origins within the wave of diversification of relatively derived trilobite clades early in the Ordovician (Hong et al., 2014). It has long been noted that A. koninckii is homeomorphic with trilobites from a variety of other clades, several of which were particularly common in low-oxygen settings in the Cambrian (Hughes et al., 1999), sharing a narrow axis (the central lobe running the length of the trilobite body), multiple homonomous trunk segments, and a small holaspid pygidium. This similarity is convergent, because more basal members of Aulacopleurida do not share this form.

Adult A. koninckii showed variation in the number of thoracic segments, from 18 to 22, which partitioned the species into five distinct morphotypes. Intraspecific variation in adult thoracic segment number is a relatively rare condition among trilobites, especially among those younger to the Cambrian (Hughes et al., 1999).

Our investigations of A. koninckii during the last three decades have been based on three datasets, none of which was independent of the others. Early studies (e.g., Hughes and Chapman, 1995; Hughes et al., 1999) focused on a sample of 86 specimens available in museums in the USA and UK (dataset 1), but this dataset was expanded in later studies (e.g., Fusco et al., 2004) to 391 specimens that included those within Czech collections (dataset 2). Following a later comprehensive review of all materials available in the Czech National Museum and Czech Geological Survey, a third dataset (dataset 3) was compiled (Fusco et al., 2014, 2016; Hong et al., 2014) consisting of 352 specimens judged to include only the very best preserved specimens (more than 10,000 were inspected, Hong et al., 2014). This dataset, that included 163 of the same specimens included within the earlier, 391 specimen dataset 2, is the one which provided the morphometric data used in the present study (Figure 3, Supplementary Data Sheet).

Protaspid stages are not known for A. koninckii, although comparison with congeneric species suggest up to three or more (Yuan et al., 2001). The three datasets include specimens that range from 4 to 22 thoracic segments. Specimens with 4–17 thoracic segments were certainly meraspid, but of unknown morphotype (assuming that the morphotype was already determined at that stage, see below), and specimens with 22 thoracic segments were certainly holaspids of the corresponding morphotype. However, small specimens with 18 to 21 thoracic segments represent a mix of late stage meraspids (of an undetermined morphotype with more segments) and early holaspids (of the morphotype with the same number of segments), while larger specimens with 18–21 thoracic segments were holaspids of the corresponding morphotype. Unfortunately, no associated characters have been found permitting assignment of any individual to a given morphotype when it is below a minimal size above which it can be confidently labeled holaspid (Hong et al., 2014). Post-protaspid stages are designated s_n, where n is a progressive natural number, starting from s₀ for the first meraspid stage.

Although papers describing trilobite ontogeny are commonly accompanied by drawings that represent ontogenetic series, the reconstruction presented herein (Figures 4–7) differs in that morphometric character estimates were derived by applying previously inferred growth and segmentation dynamics, as discussed in sections 4 and 5 above. The morphologies of this reconstruction thus come from an understanding of the way in which growth and segmentation were executed in this animal, rather than simply being an attempt to graphically represent mean form at a particular stage. This also allowed us to estimate the morphology of some instars for which measurable specimens are not presently available. When similar studies of the development of other trilobites are completed, this method will provide an effective way to illustrate visually the morphological consequences of systematic differences in developmental dynamics and control. Details of how this reconstruction was made are provided in the Supplementary Appendix.

Aulacopleura koninckii was among the first species for which trilobite growth stages were described, in a work that pioneered the reporting of fossilized ontogenies (Barrande, 1852). By utilizing large numbers of specimens Barrande showed that forms previously recognized by Hawle and Corda (1847) as different trilobite taxa were, in fact, ontogenetic variants of a significantly smaller number of species. The description of ontogenetic changes, where accessible, has since become a standard part of paleontological analysis and increasingly has employed quantitative methods to describe and differentiate modes of development. This is particularly true for those groups, generally skeletonized, for which significant numbers of fossilized specimens are available. Beside other fossilized arthropod taxa, such as Rehbachiella (Walossek, 1993) and Bredocaris admirabalis (Müller and Walossek, 1988), examples include echinoderms (Smith, 2005; Rahman et al., 2015), ammonoid molluscs (Gerber et al., 2007; De Beats et al., 2013), fish (Johanson et al., 2010), dinosaurs (Erickson et al., 2017), and early mammals (Chinsamy-Turan and Hurum, 2006). Among trilobites, analyses have included testing whether differences between the ontogenies of close relatives reveal strict heterochronic change or other types of developmental repatterning (e.g., Hunda and Hughes, 2007), and in patterns of ontogenetic and static character covariation (Gerber and Hopkins, 2011; Webster and Zelditch, 2011a,b). Likewise, progress has also been made in the establishment of ontogenetic series for a range of soft-bodied fossils including fascinating, if sometimes controversial, forms from the Cambrian (e.g., Yue and Bengtson, 1999; Harvey et al., 2010; Haug and Haug, 2015) and even earlier (Hoyal Cuthill and Conway Morris, 2014).

Most paleontological ontogenetic studies (including some that are quantitative) remain descriptive in that they characterize and compare developmental patterns, but provide limited information on the mechanisms that controlled them. However, where fossil ontogenies are unusually rich both in terms of the span of developmental stages represented and in the number of individuals available for each stage, as here in A. koninckii, it is possible to frame and test specific hypotheses about how aspects of development were controlled. Among trilobites A. koninckii offers an unusually wide range of insights, as we have seen, but increasingly others are being recognized. For example, evidence of compensatory growth has recently been reported in Olenellus gilberti, a species living approximately 80 million years before A. koninckii (Webster, 2015).

A sound assessment of the relevance of the paleo-evo-devo approach to arthropod evolution in general requires that the investigation undertaken into A. koninckii is expanded to a variety of other trilobites including those showing different morphotypes and developmental modes, and to other extinct arthropod taxa. This should permit identification of aspects of developmental control conserved among all members of the clade, and those that varied within it. In this regard, a focus on the early Cambrian is likely to bear particular fruit. Firstly, trilobites of this age are phylogenetically basal within the clade and might thus yield insights into the original developmental condition of the clade from which derived forms, such as A. koninckii, evolved. Secondly, several relatively complete ontogenies of articulated species are already known from this time, mostly from parts of China (e.g., Dai et al., 2014, 2016; Hou et al., 2015, 2017), offering potential insights into the nature of variation in developmental controls among close relatives.

In summary, looking beyond the taxonomic boundaries of trilobites and arthropods, it seems clear that a paleo-evo-devo approach to the study of evolution can convey two main types of insight. Firstly, through a phylogenetic comparative approach, developmental data about extinct taxa can contribute to elucidating both the state of developmental characters at ancient nodes, and the state transitions of the same characters along the branches of an evolutionary tree. Moreover, because the geological age of extinct taxa makes the branches connecting them to the rest of the tree relatively short (in terms of evolutionary time), uncertainties about the exact branching topology affect inferences about primitive character state conditions less strongly (Donoghue et al., 1989; Fusco et al., 2014). Secondly, the possibility of directly investigating the mechanisms controlling developmental processes allows us to explore the evolution of developmental mechanisms, rather than simply the evolution of the developmental patterns that they produce. Both kinds of data and investigation can contribute significantly to the study of the evolvability of developmental patterns and processes in the total (stem + crown) group of interest.

NH conducted initial investigations and data collection for A. koninckii, later aided by PH who produced the current landmark dataset. Analyses were conceived and performed by GF, PH, and NH. GF conceived of the full reconstruction presented herein and JH drew the reconstructions in Figures 4, 5.

This work was funded by NSF EAR-0616574 to NH and Basic Research Project (GP17-3111-1) of KIGAM to PH.