The origin and early evolution of arthropods

The rise of arthropods is a decisive event in the history of life. Likely the first animals to have established themselves on land and in the air, arthropods have pervaded nearly all ecosystems and have become pillars of the planet's ecological networks. Forerunners of this saga, exceptionally well‐preserved Palaeozoic fossils recently discovered or re‐discovered using new approaches and techniques have elucidated the precocious appearance of extant lineages at the onset of the Cambrian explosion, and pointed to the critical role of the plankton and hard integuments in early arthropod diversification. The notion put forward at the beginning of the century that the acquisition of extant arthropod characters was stepwise and represented by the majority of Cambrian fossil taxa is being rewritten. Although some key traits leading to Euarthropoda are indeed well documented along a diversified phylogenetic stem, this stem led to several speciose and ecologically diverse radiations leaving descendants late into the Palaeozoic, and a large part, if not all of the Cambrian euarthropods can now be placed on either of the two extant lineages: Mandibulata and Chelicerata. These new observations and discoveries have altered our view on the nature and timing of the Cambrian explosion and clarified diagnostic characters at the origin of extant arthropods, but also raised new questions, especially with respect to cephalic plasticity. There is now strong evidence that early arthropods shared a homologous frontalmost appendage, coined here the cheira, which likely evolved into antennules and chelicerae, but other aspects, such as brain and labrum evolution, are still subject to active debate. The early evolution of panarthropods was generally driven by increased mastication and predation efficiency and sophistication, but a wealth of recent studies have also highlighted the prevalent role of suspension‐feeding, for which early panarthropods developed their own adaptive feedback through both specialized appendages and the diversification of small, morphologically differentiated larvae. In a context of general integumental differentiation and hardening across Cambrian metazoans, arthrodization of body and limbs notably prompted two diverging strategies of basipod differentiation, which arguably became founding criteria in the divergence of total‐groups Mandibulata and Chelicerata. The kinship of trilobites and their relatives remains a source of disagreement, but a recent topological solution, termed the ‘deep split’, could embed Artiopoda as sister taxa to chelicerates and constitute definitive support for Arachnomorpha. Although Cambrian fossils have been critical to all these findings, data of exceptional quality have also been accumulating from other Palaeozoic Konservat‐Lagerstätten, and a better integration of this information promises a much more complete and elaborate picture of early arthropod evolution in the near future. From the broader perspective of a total‐evidence approach to the understanding of life's history, and despite persisting systematic debates and new interpretative challenges, various advances based on palaeontological evidence open the prospect of finally using the full potential of the most diverse animal phylum to investigate macroevolutionary patterns and processes.


I. INTRODUCTION
Arthropods constitute a central and colossal component of Earth's biosphere, at both the macroscopic and microscopic levels. Since the beginning of the Phanerozoic, these hyperdiverse, articulated animals have shaped most terrestrial and marine ecosystems, and their pivotal roles in trophic networks often have a direct and considerable impact on our industries and economywhether vital or detrimental.
In turn, the war on insects, waged in the name of a wasteful and profit-driven agriculture, has led to catastrophic consequences for the survival of these animals worldwide, and the loss of pollinators in particular is likely to lead to cascading ecosystem breakdown (S anchez-Bayo & Wyckhuys, 2019). The direct agro-economic peril to arthropods, in conjunction with other environmental crises caused by unbridled resource exploitation and consumption, such as global warming, threatens irremediably to pauperize the planet's landscapes (Dirzo et al., 2014;Newbold et al., 2016). This waning and fragility stand in stark contrast to more than half a billion years of exceptional resilience to mass extinctions. Although trilobites, which vanished at the end of the Permian, are a notable exception, the body plans (which can be broadly defined based on morphoanatomy, see e.g. Aria, 2020) of the other four largest and traditional arthropod groupschelicerates, myriapods, 'crustaceans' and insectswere all present by at least by the Late Devonian (Garrouste et al., 2012;Siveter et al., 2014b;Waddington, Rudkin & Dunlop, 2015;Suarez et al., 2017) and diversified through all of the five major pre-Anthropocene biodiversity crises. Likely since the Jurassic (Labandeira & Sepkoski, 1993), insects have become by a large margin the most diversified and abundant arthropods (Grimaldi & Engel, 2005), but all the main lineages have characteristically experienced explosive radiations and have shown extended stability of their families and genera (Gould, 1989;Labandeira & Sepkoski, 1993;Dunlop, 2010;Lee, Soubrier & Edgecombe, 2013;Edgecombe et al., 2020).
The search for the causes and mechanisms surrounding the origin of the highly modular architecture that has certainly been determinant of the success and expansion of the arthropod phylum has therefore focused on the earliest Phanerozoic, and specifically the Cambrian explosion (Erwin & Valentine, 2013). Starting with the Burgess Shale, a variety of Cambrian Fossil-Lagerstätten across the world have yielded a wealth of non-biomineralizing species informing early diversity and character transitions leading to arthropods and to their ramifications (Budd & Telford, 2009;Edgecombe & Legg, 2014). In this context, arthropods have famously initiated discussions about shifts in evolutionary tempo and mode at the macroevolutionary scale [that is, at the inter-specific taxonomic level and above (Jablonski, 2017)], and in particular the heterogeneity of disparity patterns and their possible meaning for body plan evolution (Gould, 1989;Briggs, Fortey & Wills, 1992;Lee et al., 2013).
The insights and developments following these studies have been marked by debates about the deep phylogenetic relationships among arthropods, to which fossil taxa have contributed in increasingly significant ways (Budd, 2002;Cotton & Braddy, 2004;Scholtz & Edgecombe, 2006), in the context of a seemingly intractable phylogeny of extant lineages. In the last 10 years, broad-scale combined morphological and molecular phylogenetics, as well as phylogenomics, have broken the systematic deadlock by achieving strong branch support and topological convergence for major extant clades (Regier et al., 2010;Rota-Stabelli et al., 2011;Giribet & Edgecombe, 2019;Edgecombe, 2020), although the resolution of certain internal nodes remains a salient issue (Sharma et al., 2014). There is now robust evidence that all extant arthropods can be divided into two main lineages: Chelicerata and Mandibulata, the latter including Myriapoda as well as Pancrustacea (also called Tetraconata), a broad grouping according to which Hexapoda (including insects) arose from a paraphyletic crustacean group (Regier et al., 2010;Rota-Stabelli et al., 2011;Schwentner et al., 2017).
The inclusion of fossils to a total-evidence morphological data setkey to a contextualized macroevolutionary Biological Reviews 97 (2022) 1786-1809 © 2022 Cambridge Philosophical Society.
The origin and early evolution of arthropods perspectivehas been shown to be consistent with these topologies (Edgecombe & Legg, 2014;Legg, Sutton & Edgecombe, 2013). While summarizing certain solid advances in this field, this result did not mean that the palaeontological understanding itself was complete, and recent findings, catalysed in part by the discovery of new fossil sites (Caron et al., 2010(Caron et al., , 2014Yang et al., 2013) or the use of new technologies Zhai et al., 2019), have since rewritten the significance of many extinct taxa. This, in turn, has changed our perspective on early body plan evolution in these animals, introducing new fundamental questions to current research . Conversely, a series of exceptional discoveries involving preserved neural tissues (Strausfeld, Ma & Edgecombe, 2016a) have led to new scenarios describing the evolution of arthropod heads (Ortega-Hern andez, Janssen & Budd, 2017), broad interpretative questioning (Scholtz, 2016;Liu et al., 2018;Aria et al., 2020) and emphasis on the importance of maintaining consistency with the information provided by external morphology . Beyond genes and morphoanatomy, an integrated palaeobiological and palaeoecological picture and its role in the early radiation of arthropods is also starting to take shape (Zacai, Vannier & Lerosey-Aubril, 2016;Caron & Aria, 2017;Bicknell et al., 2018;Lerosey-Aubril & Pates, 2018;Ou et al., 2020;.
We are at a decisive point at which an unprecedented amount of often seemingly conflicting evidence from revised fossils, new fossils, new types of preserved tissues, genes, development, genetic networks, new technologies, and new phylogenetic methods is converging. This review aims to provide a simple but critical guide to current knowledge, and to lay out a synthesis of persisting or emerging challenges in early arthropod evolution, to serve as a foundation for future studies. The stakes are high, for the elucidation of the early diversification of the largest animal phylum may also provide the richest insights into the biological principles governing macroevolution.
I begin, in Sections II and III, by narrating known evolutionary steps leading to the origin of arthropods and euarthropods, following the phylogenetic framework; then, in Sections IV and V, I focus specifically on the origin and early evolution of extant arthropod lineages, and briefly discuss the issue of the kinship of trilobites; finally, in the remaining sections, I place the consequences of these finds in a broader macroevolutionary context.

II. THE PANARTHROPOD CRADLE AND A 'CAMBRIAN PLANKTONIC REVOLUTION'
Arthropoda is now recognized as a monophyletic phylum within Ecdysozoa, the moulting animals (Budd & Telford, 2009;Edgecombe & Legg, 2014;Giribet & Edgecombe, 2019) (see Table 1 for a glossary of key terms used in this review). Ecdysozoa is composed of the cycloneuralian 'worms', including priapulids and nematodes, which may be a mono-or paraphyletic grouping, and Panarthropoda, an expanded systematic definition of Arthropoda also including, among extant forms, onychophorans (velvet worms) and tardigrades (water bears) (Giribet & Edgecombe, 2017).
There is ongoing debate about whether onychophorans or tardigrades are the closest sister taxon to Arthropoda. Evidence from neuroanatomy (Mayer et al., 2013) and other internal organs favours either a sister-group relationship with tardigrades [e.g. presence of metameric ganglia along the ventral nerve cord; a grouping also called Tactopoda (Budd, 2001a;Smith & Ortega-Hernandez, 2014)] or with onychophorans (e.g. presence of sacculus and podocytes on metanephridia), and in a number of cases is ambiguous, with their presence or absence varying also among arthropods (e.g. presence of a peritrophic membrane or Malpighian tubules) (Edgecombe et al., 2000). A recent fossil-inclusive analysis found Tardigrada to be the sister group to Onychophora + Arthropoda (Caron & Aria, 2017), consistent with most other phylogenetic studies (Giribet & Edgecombe, 2017). This result is influenced by the fact that, despite their dramatic developmental contraction (Smith et al., 2016), tardigrades display the plesiomorphic condition of a truncated posterior termination bearing a limb pair with claws pointing anteriorly. This was possibly inherited from the pool of adaptations acquired by suspension-feeding lobopodians, which includes an anchoring function of the posterior lobopods (Caron & Aria, 2020). The plesiomorphic presence of more trunk somites and the presence of several elongate, curved claws on their limbs are consistent with a sister-group relationship between tardigrades and the Burgess Shale lobopodian Aysheaia pedunculata (Caron & Aria, 2017). However, while providing detailed genetic and phenotypic information, tardigrades and onychophorans also possess a number of outstanding autapomorphic characters [onychophorans appear to have independently evolved a ventral mouth opening and internalized mouthparts (Caron & Aria, 2017 and, like many tardigrades, are largely terrestrial], which urges caution when attempting to extrapolate shared derived conditions. By contrast, fossils called lobopodians, mostly found in Cambrian rocks, have more directly enriched our understanding of the early evolution of panarthropods (Liu & Dunlop, 2014) (Figs 1G, L, 2). Although specimens are characteristically rare across assemblages, these worm-like taxa bearing paired metameric, annulated and lightly sclerotized limbs (the lobopods) have revealed that a broad diversity of organisms had in fact initially evolved from cycloneuralian ancestors. Fossil-inclusive phylogenetic analyses find tardigrades to be well nested within Panarthropoda (Smith & Ortega-Hernandez, 2014;Smith & Caron, 2015;Caron & Aria, 2017), and show that the surviving Onychophora and Tardigrada are indeed offshoots with highly autapomorphic traits from this initial radiation associated with the Cambrian explosion. However, other lobopodians survived through the Silurian (Siveter et al., 2018) up to at least the Carboniferous (Haug et al., 2012a), showing that they represented The articulation of two sclerotized cuticular elements by an arthrodial membrane. The word 'arthropodization' is sometimes used to apply specifically to limb podomeres, structurally and developmentally different from body segments (see below). Arthropod An ecdysozan protostome with arthrodized appendages. Basipod The proximal unit of the biramous limb, to which are connected its two defining rami: endopod and exopod. The basipod is commonly modified as a feeding device in euarthropods, either through its development into a masticatory gnathal plate (a characteristic of arachnomorphs) or its subdivision into endite-bearing units (a characteristic of mandibulates). Whether the basipod originated from a single limb Anlage inherited from early panarthropods or formed by fusion of two separate appendicular branches corresponding to endopod and exopod is a matter of debate. Boxshall (2004) proposed to generalize the term 'protopodite' from the crustacean jargon to represent the whole basal portion of the biramous limb in euarthropods, which would include the basipod (as the distalmost, undivided part of the protopodite) as well as other coxal and sub-coxal elements. I strictly keep to this definition here by not using the term protopodite to designate a basipod from which the coxa has not yet been fully differentiated. Carapace Cephalic sclerite formed by the fusion of a few anterior head tergites and largely overlapping posterior cephalic and anterior trunk segments. This is mostly a functional definition: shields and carapaces show a range of structural and developmental overlap, see main text. Cheira (pl. cheirae) Frontalmost, usually active but sometimes passive raptorial arthrodized appendage of the first arthropods. Typically bearing at least three well-developed endites differentiated according to various feeding functions. Shortened and directed upward in megacheiran euarthropods, multichelate, suited to a prehensile predatory role. Originally coined 'great appendage' from Leanchoilia, then expanded to radiodontans, and subsequently called 'short great appendage' in megacheirans by some authors. Chelicera Frontalmost (deutocerebral) arthrodized appendage with chelate or sub-chelate termination characteristic of Chelicerata. Commonly considered homologous to the 'chelifores' of sea spiders. Coxa A proximalmost podomere usually fulfilling a masticatory function in pancrustaceans, and from which mandibles and other specialized head apparatus are thought to be derived. Coxae would be derived from the proximal endite of subdivided basipods in early members of the mandibulate lineage.

Endopod
The adaxial/inner ramus of the two rami defining the biramous arthropod appendage, usually stenopodous and used for locomotion. Endite Outgrowth on the ventral side of a limb, usually associated with a particular podomere. Commonly bearing or expanded into spines or setae. Euarthropod Arthropod with arthrodized body segments and biramous arthrodized appendages. Exite Outgrowth on the dorsal side of a limb, usually associated with a particular podomere. Common among crustaceans, especially on the coxae and basipods. Developmentally distinct from the exopod by developing as a growth axis secondary to the main cell lineage forming the arthrodized limb. Exopod The abaxial/outer ramus of the two rami defining the biramous arthropod appendage, often used for swimming and breathing.

Frontalmost appendage
The anteriormost pair of arthrodized appendages in arthropods, regarded here as being innervated by the deutocerebrum across euarthropods. This is distinct from certain other appendicular structures that could originate more frontally but are not arthrodized and considered to belong to the ocular somite (Ortega-Hernandez & Budd, 2016). In lobopodians, the term simply refers to the anteriormost, fully functional and differentiated pair of appendages, sometimes also placed posterior to smaller filamentous structures (Park et al., 2018). Gnathobasipod A basipod differentiated into a large masticatory gnathal plate, often fringed with teeth. Great appendage See 'cheira'.

Mandible
Masticatory device whose gnathal portion is fundamentally derived from the coxal podomere (proximal to the basipod) of the fourth head segment (including ocular segment). Mandibles commonly encompass the basipod and associated elements, including reduced endopods with sensory and manipulatory functions termed palps.

Multipodomerous
Composed of a great number of podomeres, usually 15 or more. The origin and early evolution of arthropods much more than an 'experimental' body plan and eventually reached a relatively stable (if cryptic) adaptive zone within Palaeozoic marine ecosystems. Perhaps the most important aspect of lobopodians put forward in recent years is their arguably common adaptation, to various extents, to suspension-feeding (Yang et al., 2015;Caron & Aria, 2017). Most striking among the luolishaniids, which possess stout anchoring back limbs and frontal arms adorned with pairs of thin spinules (Ma, Hou & Bergström, 2009;Yang et al., 2015;Caron & Aria, 2017) (Fig. 1L), but sometimes lacking posterior limbs altogether (Howard et al., 2020a), this ecology also possibly characterizes the famed hallucigeniids (Smith & Caron, 2015;Caron & Aria, 2017), and would thereby apply to a majority of lobopodians with diagnostically elongate appendages. These are distinct from a series of other taxa, including much larger and stouter forms (Dzik, 2011;Vannier et al., 2014) (termed herein 'xenusiids'), that exclusively bear stout and conical lobopods, similar to those in onychophorans and tardigrades, and which would lie closer to the common arthropod ancestor (Fig. 2). A basal position of luolishaniids and hallucigeniids (Caron & Aria, 2017 implies that arthropods and their closest relatives arose from a paraphyletic lineage of suspensionfeeders: the 'long-legged' lobopodians. A separate line of thinking (e.g. Smith & Ortega-Hernandez, 2014;Yang et al., 2015;Howard et al., 2020a) groups luolishaniids, hallucigeniids and most of their relatives into a monophyletic total-group Onychophora. The data set used when recovering this topology, however, broadly homologizes lobopodian appendages with those of extant panarthropods based on purported affinities with the tripartite arthropod brain, while lacking certain critical characters for lobopod types, including the peculiar 'tentacles' of hallucigeniids (Caron & Aria, 2017. The total-group Onychophora node itself would only be supported by the imbricated type of sclerotic growth (Smith & Ortega-Hernandez, 2014;Caron & Aria, 2017). Regardless of the phylogenetic scenario considered, however, the distinction between an ambulatory or semi-sessile feeding lifestyle was determinant in the primordial diversification of panarthropods.
Parallel studies on the first arthropods, the radiodontans (I use here a definition of Arthropoda based on the presence of an arthrodized appendage (Aria, 2019; see also Table 1), add even greater significance to suspension-feeding, broadly defined, in the rise of this phylum. A filter-feeding strategy, more precisely using a filtration structure to capture food particles below a precise size threshold , has been shown to be present in several relatives of the iconic predator Anomalocaris, having evolved multiple times within the group, and led to gigantism in the Ordovician Van Roy, Daley & Briggs, 2015;Lerosey-Aubril & Pates, 2018;De Vivo, Lautenschlager & Vinther, 2021) (Fig. 1H, I). Contrary to long-legged lobopodians, however, filter-feeding in radiodontans was entirely carried out by the extensive modification of a single pair of appendages, the frontalmost, arthrodized appendages, characteristic of this group which otherwise lacks any body or limb arthrodization. These appendages are coined here 'cheirae' (see Table 1).
This evidence further emphasizes the central role of small macro-to microscopic organisms in Cambrian sea waters, and notably larvae. It should be questioned, in fact, whether the 'planktonic revolution' thought to characterize the Great Ordovician Biodiversification Event (Servais et al., 2008) should be placed within the Cambrian instead, correlated to a massive and sudden expansion of suspension-feeding strategies that arguably originated during the Ediacaran (Wood & Curtis, 2015;Gibson et al., 2019). Cases of suspension-feeding cited above among Cambrian panarthropods have been encountered in more derived fossil taxa (Izquierdo-L opez & Caron, 2019), and are documented throughout Metazoa, in sponges, cnidarian polyps, echinoderms, brachiopods and a variety of other animals (Nanglu et al., 2016;Moysiuk, Smith & Caron, 2017). Certain localities, like Marble Canyon, were arguably even built on suspension-feeding (Nanglu, Caron & Gaines, 2019). Although the fossil evidence for small meso-to microplankton is still largely indirect (Lerosey-Aubril & Pates, 2018), small carbonaceous fossils (SCFs) (Harvey, Velez & Butterfield, 2012) and 'Orsten' taxa from different localities around the world (Walossek & Müller, 1998;

Multisegmented
Composed of a great number of segments, usually 20 or more. Podomere Unit of an arthropod limb as defined externally by an arthrodized sclerotic ring and internally by intrinsic muscular attachment. Segment Sclerotized metameric unit (somite) separated from adjoining units by clear margins.

Shield
Cephalic sclerite formed by the fusion of most head tergites and articulating with the first trunk segment, or with limited overlap over anteriormost trunk segments. This is mostly a functional definition: shields and carapaces show a range of structural and developmental overlap, see main text. Somite Constitutive body unit containing an arrangement of organs serially repeated in other somites.

Stenopodous
An arthrodized appendage articulated by a series of well-defined, relatively elongate podomeres. Typically opposed to annulate appendages, composed of annuli, which, following Boxshall (2004), lack individual muscle insertions. Tergite Dorsal segmental sclerite, usually arthrodized. The origin and early evolution of arthropods et al., 2007, 2010) directly document the existence of abundant, planktonic crustaceomorph faunas that may be related to known mandibulate macrofossils (see Section IV). Burgess Shale-type (BST) deposits do not preserve micrometric fossils well simply because of grain resolution limit, but there is plentiful evidence of tiny arthropod-like fossils. For the most part, those are undescribable (personal observations), but several post-hatchling stage larvae have already been published (Liu et al., , 2020, some of them differing somewhat in morphology from their adult counterparts, pointing to ontogenetic niche differences. As arthropods were already the most diverse and abundant animals by the Cambrian Stage 3, their larvae must have constituted an important part of planktonic life forms, even if there were many benthic direct developers . Radiodontans also radiated based on a much broader diversity of specializations involving the cheirae , which also possibly includes sediment sifting (Moysiuk & Caron, 2019a;but see De Vivo et al., 2021), that translated into a significantly greater disparity for these appendages De Vivo et al., 2021). Similar observations can be made about the variety of shapes realized by other arthrodized limbs and arthropod body segments, by comparison with the rather conservative morphology of lobopodians. Arthrodization, as a structural innovation, was therefore arguably decisive in the early success of arthropods by providing a modular medium with both developmental flexibility and structural rigidity, a 'sculpting material' that worked particularly well as a rapid driver of phenotypic evolution, notwithstanding the morphoanatomical and genetic trade-offs that later stabilized a number of well-defined body plans (Aria, 2020).
From a xenusiid-like ancestor emerged the so-called 'gilled lobopodians', namely Kerygmachela (Fig. 1G) and Pambdelurion, long thought endemic to the early Cambrian Greenland locality of Sirius Passet, but possibly present elsewhere (Vinther et al., 2016), described as bearing flap-like swimming appendages in addition to lobopods. In contrast to xenusiid ancestors, Pambdelurion displays a circumoral sclerotic mouth apparatus clearly placed on the ventral side of the body, although it is argued that the animal also retained an eversible pharynx (Vinther et al., 2016). The rotation of the mouth opening, which in extant lineages is characteristically ventral with postero-ventral orientation and connected to an anteriorly looped oesophagus, therefore occurred during the xenusiid-radiodontan transition (Fig. 2). Kerygmachela may document a transitional morphological state in which the mouth opening is ventral but directed anteriorly (Park et al., 2018).
A circumoral sclerotized apparatus, giving its name to radiodontans [radius (Latin)odoús (Greek) meaning radialteeth], exemplified by Anomalocaris (Daley & Edgecombe, 2014) or Hurdia (Daley, Budd & Caron, 2013), is therefore not exclusive to this group. A 'peytoia' type of outer sclerotized ring, or its derivatives (Daley & Bergström, 2012), composed of size-differentiated plates, also commonly called the 'oral cone', would unite radiodontans, but resemblances with Pambdelurion are such that some isolated radiodontanlike mouthparts from the Chengjiang biota were proposed to belong to a relative of Pambdelurion (Vinther et al., 2016). The numerous inner teeth of Pambdelurion are found in radiodontans in the form of an inner row of smaller dented plates, which could be derived from the symplesiomorphic pharyngeal teeth. Interestingly, a comparable set of elements are also found in dissociation in amplectobeluid radiodontans from the Chengjiang biota, never forming the typical oral cone (Cong et al., 2017(Cong et al., , 2018. A single specimen of Amplectobelua symbrachiata shows overlapping gnathobase-like differentiated plates associated with alleged reduced flaps (Cong et al., 2017). The lack of fully formed peytoia mouthparts in this material begs the question of whether these elements could instead represent an additional differentiated set associated with the buccal apparatus, but the apparent variability of features otherwise considered diagnostic in different parts of the arthropod tree , including gnathobasipods known in Parapeytoiaperhaps even along peytoia-like mouthparts (see Budd, 2021)but not in other megacheirans, leaves open the possibility of convergence or parallelism.
In summary, the priapulid-like and frontal to ventral position of the mouth opening is a well-documented evolutionary transition from lobopodians to arthropods (Fig. 2), although it evolved convergently in onychophorans and certain tardigrades. It is accompanied by the formation and sclerotization of a circumoral apparatus, which led to a diversification of oral and pharyngeal masticatory devices. This mouth-related disparity, possibly associated with the expansion of feeding strategies (Daley & Bergström, 2012;Zeng et al., 2018a), was likely synergistically involved in the pivotal adaptions of the frontalmost pair of appendages.
(2) Arthrodization, compound eyes and trait mosaic in Opabinia As mentioned previously, the emergence of radiodontans as the earliest members of Arthropoda is fundamentally characterized by the evolution of an arthrodized pair of appendages (see below and Fig. 3A for considerations regarding the somitic identity of the frontalmost appendage). In contrast to the oral complex, there is no known sequence of character change leading to the arthro(po)dization of the frontalmost appendage. This condition seems to appear rather suddenly in radiodontans, even if the cheirae themselves are likely homologous to the similar stout and purportedly raptorial appendages of Pambdelurion, Kerygmachela and xenusiids, which lack subdivisions into podomeres or even externally defined pseudo-podomeres, and despite the fact that topological homology seems at odds with brain homology (Ortega-Hern andez, Janssen & Budd, 2017;Park et al., 2018;Aria et al., 2020;Zeng et al., 2020; see Section VI). As mentioned above, the appearance of an arthrodized frontalmost appendage, the cheira, is associated with a burst of disparity De Vivo et al., 2021) rapidly expanding feeding niches through the specialization of podomere endites, which also seem to originate from lobopodian ancestors.
Recent studies have confirmed previous observations (Fu et al., 2014) that the trunk appendages in isoxyids likely covered various levels of articulation, some endopods being segmented but not necessarily articulating via arthrodial membranes (Fu et al., 2022), while in certain species the whole biramous appendage morphology seems close to that of hymenocarines (Zhang et al., 2021). According to these data, arthrodization of post-frontal appendages could have originally arisen from inter-specific variations. While it can be hypothesized that post-frontal limb arthrodization coopted existing developmental pathways expressed first in the frontalmost appendage (Setton et al., 2017), selective pressures underlying the acquisition of one or other type of arthrodized appendages were probably not entirely overlapping. The strengthening and articulation of the frontalmost appendage evolved rapidly as a much more efficient and versatile prehensile apparatus, and thus as a feeding system for larger and more energy-demanding organisms. Likewise, strengthening as an improvement of mastication efficiency seems to have played a role in the arthrodization of trunk limbs, but it has to be considered alongside ventral migration [compared to laterally placed flaps in dinocaridids (i.e. radiodontans and opabiniids)], thereby involving different swimming biomechanics, as well as walking (Zhang et al., 2021)a topological repositioning which itself Biological Reviews 97 (2022) 1786-1809 © 2022 Cambridge Philosophical Society.
The origin and early evolution of arthropods correlates with the development of a laterally extended dorsal carapace.
Another important trait associated with the appearance of radiodontans is the presence of well-developed compound eyes, which in these taxa are stalked (Paterson et al., 2011; but see Paterson, Edgecombe & García-Bellido, 2020), whereas lobopodians only possess simple ocelli (Ma et al., 2012a), when present (Fig. 2). Compound eyes are thought to have evolved from increasingly larger faceted visual organs (Ma et al., 2012a;Park et al., 2018), and were present in the great majority of Cambrian arthropod lineages, in stark contrast with the many reversals to simpler ocular structures known among extant taxa (Strausfeld et al., 2016b). Depending on the phylogenetic placement of Opabinia regalis, stalked compound eyes would in fact be plesiomorphic to Arthropoda. Despite having come a long Relationships and characteristics of the main fossil panarthropod groups. Summarized phylogenetic framework of panarthropod relationships, with main fossil groups highlighted. Boxes contain combinations of defining but overlapping character states for these fossil groups, which can be para-or monophyletic. Insets (A, B, C) represent crucial steps of ocular, mouth and external protocerebral evolution at the onset of the arthropod radiation: (A) mouth is ventralized and accommodates circumoral plates, frontal sensory organs derived from protocerebrum (yellow); (B) stalked compound eyes, arthrodization, complex of frontal organs protected by a sclerite (yellow); (C) later, formation of the hypostome-labrum complex, with pre-oral sternal plate (green) protecting the mouth instead of circumoral plates, and a fleshy extension (red) possibly derived from the same Anlage as those of the anteriormost sclerotic/sensory complex (yellow; see Fig. 3). Yellow stars on tree mark important morphological innovations or evolutionary events. Coloured branches indicate the frontalmost appendage type (red, cheira; green, chelicera; blue, antennula). The arrow and question mark associated with the hymenocarine morphogroup represent the uncertainty as to whether some of these taxa lay closer to pancrustaceans. way in the understanding of its significance since it was first redescribed by Whittington (1975a), to this day, the iconic 'weird wonder' from the Burgess Shale remains a tantalizing evolutionary puzzle (Briggs, 2015;Pates et al., 2022). Like radiodontans, Opabinia possesses soft swimming flaps with dorsal gills, a tail fan, and five stalked eyes, but its unique, unpaired proboscis-like 'nuzzle' carrying a terminal dented claw as well as a lack of sclerotized buccal structures would place it a step away from the origin of arthropods. On the other hand, the mouth opening and oesophagus are oriented postero-ventrally, as in euarthropods (the mouth opening is fully ventral in radiodontans), the body shows clear external segmentation when generally in radiodontans segment boundaries are poorly defined, and there are both lateral The robust, raptorial frontalmost appendage of xenusiids and radiodontans (α) is known to transition to the megacheiran cheira (β), adopting a dorsal orientation, and sometimes coupling differentiated grasping and sensory functions (as in leanchoiliids). The cheira supposedly diversifies into an exclusively sensory (the antennula, γ) or predatory, manipulating form (the chelicera, δ) in extant taxa, but these transitions are not yet documented clearly by the fossil record, except perhaps in Kiisortoqia. (B) Labrum. The black square represents the mouth, the black circle is the eye. The labrum presumably originates in early panarthropods from a protocerebral or pre-protocerebral Anlage that could have formed originally an organ serving a sensory function (α), then forming an externalized sensory organ commonly covered by an 'anterior' or 'ocular' sclerite (β). In more derived forms, frontal sensory features co-exist with the hypostome-labrum complex (γ), in which a pre-oral sclerite also bears a fleshy protrusion; the latter is the labrum in the traditional sense. As the labrum of the hypostome-labrum is protocerebral in origin, the question remains whether it derives evolutionarily from the frontal preoral organs seen in some fossils. fa, frontalmost appendage; hy, hypostome; la, labrum; lc, labral complex; os, ocular sclerite; pc, protocerebrum; so, sensorial organ. (C) Head tagma. The head tagma is poorly defined in early arthropods, but in some cases appendage differentiations seem to delimit a five-somitic head (α); in megacheirans, this five-somitic configuration is clearly delimited by the head shield in cheiromorphs, but the ancestral jianfengiids appear to display variability in the length of the cephalon and a possible decoupling between the tergal and appendicular head tagmata (β). From the plesiomorphic fivesomitic head arose the diagnostic six-somitic mandibulate cephalon (δ, although the tritocerebral somite independently became limbless in some groups, and, beyond the larval stage, crustaceans evolved the more inclusive cephalothorax), but also the more variable cephala of arachnomorphs. In these taxa, the five-somitic tagma transitions directly to possibly six-, seven-and even eight-somitic heads, the latter representing the ancestral condition of panchelicerates (γ). (D) Biramous appendage. There exist two scenarios for the origin of biramicity, both supported by different fossil evidence: the split of the main limb axis, as suggested by isoxyids (α), and the fusion of separate limb axes, as interpreted in radiodontans with double rows of swimming flaps (β). Either of these initial conditions led to the archetypal biramous appendage with basipod, endopod and exopod (γ), as expressed in megacheirans. The differentiation of the basipod plays a critical role in the emergence of cenocondylans. The arachnomorphs are distinguished by a gnathobasipod (δ), while early members of the mandibulate lineage evolved subdivisions of the basipod that later gave rise to coxal features, including the mandible (ε). From a developmental point of view, the 'true' exopod could a priori be recognized by attaching to the original basipod, that is, the distalmost segment of the entire basipod complex, or protopodite; by contrast, exites arise from other basal segments (ε). Early members of the chelicerate lineage possess 'semidetached' stenopodous exopods whose affinity as exopods or exites is unclear (δ). b, basipod; c/m, coxa/mandible; df, dorsal flap; en, endopod; ex, exopod; exi, exite; vf, ventral flap. The origin and early evolution of arthropods and median eyes as in some early euarthropods. Midgut glands are typically stacked with radial folds, identical to those of Kerygmachela and Pambdelurion (Budd, 1997) on the one hand, and Isoxys (Vannier et al., 2009), leanchoiliid euarthropods (Butterfield, 2002;Aria, Caron & Gaines, 2015) and even the trilobite-like Kiisortoqia (Stein, 2010), on the other hand, showing a broad evolutionary contiguity of this feature across these taxa. Quasi-identical digestive glands with radial folds or diverticulate pattern are found in xenusiids (Vannier et al., 2014), yet are not stacked.
In summary, the evolution of limb arthrodization in arthropods occurred first in frontalmost appendages, coined here the cheirae, before arising in more derived taxa as a feature of biramous trunk appendages (Fig. 2). Limb arthrodization, consequently, first catalysed feeding-based diversification before being developmentally co-opted as an improvement of mastication and locomotion efficiency. Compound eyes are arguably ancestral for arthropods, derived from simpler faceted visual organs, and acquired structural complexity comparable to that of modern taxa already in the first arthropods (Fig. 2). Five-eyed Opabinia might have been the first panarthropod with compound eyes, already displaying lateral and median eyes. Many morphological traits of Opabinia are at odds with the most parsimonious sequence of their acquisition along the euarthropod stem, and the single frontal clawed proboscis is unique to this taxon, arguably making it still one of the most enigmatic Cambrian animals.

(3) Articulation of body segments and biramous limbs
Opabinia also serves as a point of reference for the two most critical lines of discussion pertaining to the origin of 'true' arthropods (Euarthropoda), as defined by the presence of arthrodized body segments and biramous limbs (Aria, 2019). While irregularities in length between visible somites suggest that there were no articulating tergites, all authors who have studied Opabinia have recognized the presence of some form of external segmentation (Briggs, 2015), which is, by contrast, more difficult to detect in complete radiodontan specimens (Chen, Ramskold & Zhou, 1994;Daley & Edgecombe, 2014;Moysiuk & Caron, 2019a), perhaps due to the general absence of lateral preservation. There does not seem to be any comparable form of externalization of somite boundaries in more basal lobopodians, although there exist differentiations at limb insertions and different annulation patterns (Budd, 2001b;Caron & Aria, 2017). Body arthrodization is unclear in isoxyids (arthropods with bivalved carapaces sharing affinities with radiodontans; Figs 1F, 2), but it would seem that tergites are not expressed under the carapace (Zhang et al., 2021). This would place Kylinxia  and megacheiranshistorically, the so-called 'great appendage' euarthropodsas the earliest unambiguous euarthropod representatives   (Figs 1M, 2). Details about the formation of tergite articulation are not documented.
It has been debated whether Opabinia combined both lateral flaps and lobopods (Budd, 1996;Zhang & Briggs, 2007;Briggs, 2015). Although any developmental remnants of lobopodous limbs in Opabinia seem fully internalized and associated with the circum-intestinal haemocoelic cavity , such combination is arguably well evidenced at least in Pambdelurion (Budd, 1997). Two separate rows of lateral flaps were otherwise described in the massive Aegirocassis from the Lower Ordovician Fezouata Lagerstätte in Morocco (Fig. 1H, I), and as possibly present in other radiodontans (Van Roy et al., 2015). This evidence would suggest that the typical biramous limbs of euarthropods formed by fusion of separate limb Anlagen (Fig. 3D). This is at odds with some other fossil evidence, such as in the isoxyid Surusicaris , which shows broadly attached and morphologically similar endopods and exopods (Fig. 3D), as well as with developmental data also supporting that both rami originated by splitting of a single limb axis (Wolff & Scholtz, 2008). Complicating this matter, early members of the chelicerate lineage (Fig. 1C) display an intriguing separation of the exopod branch from the main basipod-endopod limb axis (Sutton et al., 2002;Briggs et al., 2012;Aria & Caron, 2017b, 2019) ( Fig. 3D; see Section IV), likely related to the derived loss of exopods in the euchelicerate head (the prosoma), and also supporting the view that the exopod might belong to a separate limb Anlage. Further developmental data could help shed light on this issue, but we must be cautious about our interpretation of extant models, for their external morphology may sometimes hide derived developmental complexity. Olesen et al. (2011) have shown for instance that, in certain branchiopods, podomeres in stenopodous limbs were likely derived from the endites of phyllopodous appendages, and therefore that using developmental data from these stenopodous limbs to extrapolate podomere origin in crustacean stenopodous appendages in general would be misleading.
In summary, the exact timing of body arthrodization remains unclear, but external segmentation would be ancestral to arthropods. The appearance of tergo-pleurae quickly led to long, multi-segmented bodies among the earliest euarthropods. Body arthrodization developed alongside biramy, for the origin of which there are currently two main competing scenarios (Fig. 3D): either both rami originated from the fusion of separate limb Anlagen on the same somites (e.g. dorsal and ventral sets of flaps), or they arose by the splitting of a single growth axis.

(4) Head sclerotization
The journey towards Euarthropoda also involves the formation of a broad sclerite protecting the head, taking the form of a carapace or head shield. Various head sclerites are known in long-legged lobopodians, but a basal phylogenetic position of these taxa speaks against any direct homology with arthropod tergites (Caron & Aria, 2017). A variety of antero-dorsal and paired ventro-lateral sclerites mark the early evolution of arthropods, and unique lateral elements ('p-elements') may even constitute one of the strongest apomorphies of Radiodonta (Van Roy et al., 2015;Moysiuk & Caron, 2019a;Cong et al., 2017;Zeng et al., 2018b). The continuity of these distinct sclerites with arthropod carapaces and head shields is not entirely clear, but there is some evidence  to posit that at least the antero-dorsal element, despite spanning a very large size range in radiodontans (Moysiuk & Caron, 2019a), corresponds to the so-called 'anterior/ocular sclerite' identified across early arthropods (Ortega-Hern andez, 2015; Aria & Caron, 2017a), including megacheirans   (Fig. 3B).
Isoxyids (Fig. 1F), now retrieved by different large phylogenetic data sets as sister taxa to all other euarthropods Aria & Caron, 2017a) (Figs 2, 4; although this is partly dependent on uncertainties, including body arthrodization) bear bivalved carapaces, as defined by tergites of the anteriormost somites extending dorsally over other tergites and thus having a free posterior range of motion (Table 1). By comparison, euarthropods such as arachnomorphs are typically identified by the presence of a head shield, which represents the fusion of all cephalic tergites and has limited posterior overlap over trunk tergites. This tagma evolved into the chelicerate prosoma (Aria & Caron, 2017b, 2019. In reality, the morphological ranges of these structures overlap, as is clearly documented by crustaceans (Olesen, 2013). Shields and carapaces could be seen as different phases of an evolutionary continuity based on the integration of additional segments into the head tagma, although their homology also depends on the precise somite from which they originate. For instance, the lateral 'carapace-like' p-elements of radiodontans are likely analogous with euarthropod carapaces because they arguably originate from the protocerebral somite Moysiuk & Caron, 2019a), instead of a deutocerebral origin as seems to be case in isoxyids and Kylinxia based on the post-ocular location of the carapace/shield Zeng et al., 2020), or more posterior affinity as seen in crustaceans, with the maxillary somite (Olesen, 2013). Yet bivalved carapaces enclosing a part or the entire body laterally are easily recognizable in a wide range of closely related Cambrian taxa, despite showing shape variations (Izquierdo-L opez & Caron, 2019), and possibly being modified into a flat 'shield' in fuxianhuiids (Fig. 1A). Hence, these carapaces are arguably topologically homologous and may constitute an ancestral diagnostic feature of mandibulates, at least in adults, contrasting with the more restrictive head shield of arachnomorphs (Fig. 2). The lack of broad protecting carapaces in arachnomorphs is further associated with greater cuticular developments of post-cephalic segments, in particular in the form of pleural extensions, fusion of posterior segments [for instance, the pygidium, but see Izquierdo-L opez & Caron (2021)] and other ornamentations.
In summary, sets of frontal sclerites in the first arthropods, questionably with deeper panarthropod origins, are regarded as possible precursors of cephalic sclerites in euarthropods, or, like the ocular sclerite, inherited homologous features (Fig. 3B). Although the distinction between head shields and carapaces can be tenuous, with developmental characteristics usually unavailable in fossils, each broadly characterize separate lineages during the early radiation of euarthropods (Fig. 2). Despite widespread convergences and great variability later on, these structures therefore played a central role in setting unique evolutionary trajectories for these lineages, impacting critical traits such as brood care (Caron & Vannier, 2016), vision (Scholtz, Staude & Dunlop, 2019) or respiration (Vannier, Williams & Siveter, 1997).

(5) Origin of euarthropods
Euarthropoda Lankester, 1904 is defined within the modern phylogenetic framework by the combined presence of an arthrodized body and arthrodized appendages (Aria, 2019). As discussed above, the significance of isoxyids in this context is likely pivotal, but its full understanding requires further investigation. The reported presence of dinocaridid-like tail flaps in isoxyids (Legg & Vannier, 2013) would favour an incomplete tergal sclerotization (Zhang et al., 2021), but certain species otherwise possess trunk endopods with distinct podomere boundaries, in some cases with arguably limited articulation (Fu, Zhang & Shu, 2011;Fu et al., 2014Fu et al., , 2022. A form of post-frontal metameric limb arthrodization may therefore have appeared in these animals, prior to taking a The origin and early evolution of arthropods more conventional leg-like aspect in megacheirans, but this depends on whether isoxyids are basal to euarthropods or an autapomorphic clade possibly basal within total-group Mandibulata ; see Section IV).
Megacheirans therefore essentially relied on their cheirae for feeding morpho-functionality, sometimes cumulating both raptorial and differentiated sensory functions on this single limb (Fig. 1M), a unique morpho-functional combination of the frontalmost appendage among all adult arthropods, and likely an evolutionary solution to the lack of division of labour across other limbs . Jianfengiids, the earliest members of this paraphyletic group, show decoupling of dorsal and ventral cephalization, possibly reflecting a greater initial variability of the cephalon in euarthropods . Decoupling of dorsal and ventral (specifically, appendicular) tagmatization is an important phenomenon to consider across euarthropod groups (Lamsdell, 2013;Scholtz, 2016), and its mechanism has been documented developmentally (Janssen, Prpic & Damen, 2004), but will not be covered in detail here.
The recently described Kylinxia beautifully documents the homologous continuity of the cheirae across arthropods and euarthropods and anchors the basal position of megacheirans in the euarthropod tree . This animal bears two large lateral as well as three smaller median eyes, thus also importantly shedding light on the longpuzzling quintet of eyes in Opabinia, now possibly present in the common euarthropod ancestor. However, owing to the numerous characters (dinocaridid-like tailfan, nonarthrodized head limbs in Surusicaris, absence of clear body arthrodization) that still indicate a basal position of isoxyids, Kylinxia is here resolved either simply as the basalmost euarthropod in the more classic topology (Fig. 2), or as sister to megacheirans and Arachnomorpha under a 'deep split' scenario: the early separation of the total-group Mandibulata and Chelicerata close to the origin of Euarthropoda itself (Fig. 4B, C).
Rare Cambrian arthropods with bivalved carapaces have also been described bearing cheirae. It would therefore appear that the presence of these elaborate frontalmost appendages was contiguous across two separate lineages, one of them also possibly retaining the bivalved carapace of isoxyids ( Fig. 2; see also Zeng et al., 2020). Although relatively simple in principle, the plausibility of the 'deep split' evolutionary scenario (Aria, 2020; Fig. 4B), which, in its original form, would also settle the lengthy dispute about the phylogenetic position of trilobites (but see Fig. 4C and Sections IV and V), is only made possible by recent reassessments of critical Cambrian taxa and, in particular, their relation to extant clades.
In summary, Kylinxia and megacheirans represent archetypes of the first 'true' arthropods, with both arthrodized bodies and limbs, and, like radiodontans, relied on morphological differentiations between their cheirae to radiate and perform eco-morphologically. Other limbs had limited antero-posterior and proximo-distal variations, but were already distinctly biramous, with separate basipod, endopod and exopod, and thus likely serving locomotory, breathing and crude masticatory functions. Decoupling between dorsal and ventral cephalization in some megacheirans suggests an ancestral plasticity in head development. Specific morphoanatomical similarities between Kylinxia and radiodontans as well as Opabinia conflict with other lines of evidence placing isoxyids as sister taxa to euarthropods, and may imply a deep rooting of both extant arthropod lineages, perhaps even prior to the appearance of isoxyids themselves.

IV. DEEP CAMBRIAN ORIGINS OF EXTANT LINEAGES
Arthropod phylogenies, with or without fossils, have long presented seemingly intractable problems and never-ending debates. In the last 10 years or so, however, considerable progress has been made towards a consensus, in no small part due to the improvement and expansion of molecular analyses, although difficulties remainfor instance internal chelicerate relationships (Giribet, 2018). Some disagreements persist regarding the placement of early fossil groups (Edgecombe, 2020), but cumulative evidence from redescriptions and new discoveries has recently constrained the broad panarthropod topology as presented in Fig. 4A: lobopodians, radiodontans, isoxyids and megacheirans forming the stem of a clade containing both extant lineages (Chelicerata and Mandibulata) as well as trilobites and their relatives (Artiopoda), which is called Cenocondyla (Aria, 2019). This scenario is challenged in part by early euarthropods suggesting close continuity across both total-group Mandibulata (e.g. Occacaris, bearing both carapace and cheirae) and an expanded Arachnomorpha (e.g. the five-eyed Kylinxia, or Parapeytoia, a megacheiran bearing gnathobasipods). An alternative topology accommodating these issues has been presented in recent work (Aria, 2020;Aria, Zhao & Zhu, 2021) and is called the 'deep split' topology, due to the early branching of total-groups Mandibulata and Chelicerata (Fig. 4B). In this scenario, megacheirans are closer to chelicerates than they are to mandibulates, while hymenocarines are brought closer to the common euarthropod ancestor. In part, this view reconciles hypotheses previously seen as conflicting, in which authors posited a chelicerate affinity of megacheirans (Haug et al., 2012b) or a basal position for bivalved taxa . Some authors  have recently proposed a variant of the 'deep split' scenario grouping artiopodans with mandibulates, according to the Antennulata scenario (see Section V). This alternative tree configuration is represented in Fig. 4C, with isoxyids placed basally but with uncertainty (dashed line), consistent with the results of Zeng et al. (2020) recovering these taxa as derived. As the stability of any 'deep split' topology requires further testing, the more consensual topology of Fig. 4A is used in Fig. 2. In any case, however, extant euarthropod clades are recovered as forming two separate monophyletic lineages, Panchelicerata and total-group Mandibulata, with very deep roots into the origin of Euarthropoda itself.
(1) Total-group Mandibulata Numerous morphotypes from the Burgess Shale have long been included in or compared to crustaceans, even following their redescriptions from Walcott's 'shoehorned' assessments (Briggs, 1978(Briggs, , 1992Bergström, 1992;Hou & Bergström, 1997). In parallel, and before the phylogenetic concepts of Mandibulata and Pancrustacea/Tetraconata had gained wider support, a more cryptic yet rich diversity of Cambrian 'crustaceomorphs' was being revealed from two main fossil sources. First, the significant 'Orsten' biotas from the early Cambrian, discovered from Swedish outcrops (Maas et al., 2006), but now known more generally around the world   (Fig. 1E), as a type of exceptional threedimensional preservation by secondary phosphatisation, have yielded a wealth of micro-to meso-planktonic taxa which have been associated with the origin of 'crustaceans' (Walossek & Müller, 1998;Siveter, Williams & Waloszek, 2001). These fossil taxa were particularly remarkable for "recapitulating" the plausible origin of mandibles through near-complete ontogenetic sequences (Walossek, 1993;Müller, 1988), and in particular its development from the diversification of a proximalmost endite on a groundpattern biramous limb. Second, SCFs from western Canada revealed disarticulated assemblages of decidedly modern-looking appendages, including mouthparts, found mostly nowadays in anostracans and copepods (Butterfield, 1994;Harvey et al., 2012), in certain cases reaching adult sizes (Harvey & Butterfield, 2008). The crustaceomorph fossilsthe 'Orsten' ones in particular, with their plesiomorphic-looking mandibular conditionacquired by default the role of stem mandibulates as Burgess-Shale-type bivalved arthropods were given a very reductive cephalic interpretation (Budd, 2002;Legg et al., 2012) and as Mandibulata stabilized as a robust clade (Regier et al., 2010;Rota-Stabelli et al., 2011).
Recently, new palaeontological evidence, made available notably thanks to the discovery of the new Burgess Shale locality of Marble Canyon (Caron et al., 2014), provided support for the mandibulate affinity of a Branchiocaris relative, Tokummia (Aria & Caron, 2017a), and also shed light on the affinities of Cambrian bivalved arthropods as a whole, coined (with the exclusion of isoxyids, ostracods and bradoriids) the hymenocarines (Fig. 1B). With the help of careful fossil dissection, it was shown that the anterior section of the carapace in fact concealed large branchiopod-like mandibles, part of a mandibulate-like head, which in Tokummia and its relatives specifically was bound by large maxillipeds. The redescription of one of the first-found and best-preserved Burgess Shale arthropods, Waptia fieldensis , largely corroborated and completed these observations. Apart from the presence of mandibles and a six-somitic head, these arthropods illustrated and clarified hypotheses concerning arthropod limb evolution and the origin of proximal features in mandibulates (Walossek & Müller, 1998;Boxshall, 2004), namely the role of subdivided basipods bearing multiple differentiated endites in the formation of the coxa, sub-coxa and features derived from them, such as the mandibles themselves (Popadi c et al., 1998;Scholtz, Mittmann & Gerberding, 1998), or articulating pleurites around limb insertions in terrestrial arthropods (Coulcher, Edgecombe & Telford, 2015) (Fig. 3D).
Hymenocarines would minimally resolve as early mandibulates (Aria & Caron, 2017a), but the presence, notably, of five-podomerous endopods in Waptia pointed to an affinity closer to pancrustaceans . Problematically, however, these taxa did not show any trace of an appendage pair between the antennules and mandibleswhich would be expected if the mandibles as described were indeed homologous to those of mandibulates, and if hymenocarines were early pancrustaceans. This configuration implies instead that the post-antennular pair is reduced, and its corresponding somite would therefore be equivalent to the intercalary segment of terrestrial mandibulate taxa, a convergent characteristic thought so far to have been associated with the functional repositioning of the head following terrestrialization. It would also be particularly unparsimonious to consider the second limb pair to be reduced in hymenocarines, then reacquired in crustaceans, then lost again in hexapods (Edgecombe, 2017). Such a condition also strongly contrasts with Cambrian Orsten crustaceomorphs, in which antennae are well developed as biramous limbs. Thanks to unprecedented quality of computed tomographic rendering for this type of fossil, a small hymenocarine, Ercaicunia (Fig. 1B), was subsequently documented with three-dimensional preservation of appendages, including a pair of post-antennular 'hooks' interpreted as differentiated antennae (Zhai et al., 2019), therefore showing that at least some hymenocarines possessed developed post-antennular limbs. Given the challenging preservation and the very anterior location of the mandibles in Waptia , there is also a possibility that these appendages are in fact mandibular palps. In any case, the limbless post-antennular segment clearly remains a characteristic of other hymenocarines (Izquierdo-L opez , and is accompanied by other unusual appendicular reductions in Odaraia and its allies, which seemingly also lack antennules altogether. More than simply showing a diagnostic absence Biological Reviews 97 (2022) 1786-1809 © 2022 Cambridge Philosophical Society.
The origin and early evolution of arthropods or presence of development of the post-antennular somite and limbs, hymenocarines are arguably characterized by an uncommon variability in expression of these frontal appendages (Izquierdo-L opez & Caron, 2021), which may relate to early lability in the formation of the mandibulate head.
The revised morphoanatomical significance of hymenocarines also imposes a re-evaluation of Cambrian crustaceomorphs, either Orsten-type or as small carbonaceous isolated remains. Owing to their small size, the larval nature of Orsten forms and a number of SCFs opens the possibility that some are related to adult hymenocarines, with the observed differences being developmental. Because ontogeny-based phylogenetic analyses have retrieved them nested among diverse extant crustacean lineages (Wolfe & Hegna, 2014;Aria & Caron, 2017a), it would appear more plausible that the expansion of distinct planktonic niches exploited by juveniles underlies these differences, and possible heterochronic effects on the early evolution of mandibulates (Aria & Caron, 2017a), as opposed to showing a fully realized pancrustacean radiation by the early Cambrian. Larger disarticulated SCF mouthparts could even belong to adult hymenocarines, since both show similar anostracan characteristics. Nonetheless, fundamental morphological differences between juvenile components of both Orsten and SCFs suggests the co-existence during the Cambrian of two separate planktonic crustacean-like faunas, one of them perhaps more closely related to the origin of pancrustaceans. The diversification of larvae in the water column with potentially different phenotypes should itself be considered important to the early evolution of arthropods for two main reasons. First, larvae reasonably constituted immediate evolutionary feedback on the radiation of suspension-feeders ('larval explosion feedback', Fig. 2). Second, from an evo-devo perspective, the creation of larval niches different from adult ones likely served as a catalyst for the emergence of new morphological features (Aria & Caron, 2017a;Wolfe, 2017), potentially accelerating evolution in a way similar to that of the emergence of holometaboly in insects (Rainford et al., 2014).
Resolving the relationships of early mandibulate fossils, and especially their placement on either the pancrustacean or myriapod lineages, has also been made difficult by the fact that terrestrial Myriapoda lacked a documented marine stem group, which is necessary to constrain apomorphies of both mandibulate clades at their time of divergence. Without such evidence, neither the presence of antennae (Zhai et al., 2019) nor epipods Zhai et al., 2019) can be unambiguously considered pancrustacean synapomorphies; five-podomerous endopods represent in this regard a stronger character because myriapod endopods likely share a plesiomorphic heptapodomerous condition . This evolutionary gap has recently been filled by re-evaluations of certain fossil groups. Heterocrania rhyniensis, first, from the renowned Devonian Rhynie Chert Konservat-Lagerstätte, has illuminated the morphoanatomy of the enigmatic euthycarcinoids, confirming their placement on the myriapod lineage, thanks in part to the use of confocal microscopy . In addition to tergo-sternal decoupling (with tergites encompassing several somites), uniramous trunk appendages and rod-like coxal apodemes, these animals possessed a mandible-derived lower lip called the hypopharynx as well as many associated structures as they are known among myriapods. Although known from aquatic settings, euthycarcinoids likely made occasional excursions to land (see Section VIII), and their morphology accounts for this transition. Second, also corroborating previous phylogenetic results Aria et al., 2020), the emblematic fuxianhuiids of the Chengjiang fauna (Figs 1A, 2), long presented as and used as models for the first euarthropods (Chen et al., 1995;Bergström et al., 2008;Yang et al., 2013Yang et al., , 2018, were redescribed as mandibulates sharing affinities with both euthycarcinoids and myriapods, in particular with respect to their 'polypodous' trunks and the configuration of the head (Aria et al., 2021). Of particular interest is that the cephala of both fuxianhuiids and euthycarcinoids are also arguably characterized by the presence of intercalary segments (Aria et al., 2021), like those of hymenocarines, which further testifies to the prevalence of this trait in marine taxa at the origin of mandibulate lineages, even if the causes of this segmental reduction remain unexplained.
(2) Panchelicerata In contrast to the mandibulates, the Burgess Shale fossil Sanctacaris had long represented the first and only relative of chelicerates from the Cambrian (Briggs & Collins, 1988;Legg, 2014), although megacheirans have also been considered by some authors as members of this lineage (Haug et al., 2012b;Tanaka et al., 2013;Liu et al., 2020). The formerly unclassified Burgess Shale Habelia optata, originally described by Charles D. Walcott, showed that Sanctacaris was not a lonely offshoot, and that, although numerically rare, chelicerate precursors had already diversified in Cambrian seas (Aria & Caron, 2017b). Wisangocaris from the Australian Emu Bay Shale displaying the same type of large mandible-like gnathobases was also revealed to be part of this worldwide radiation (Jago, Garcia-Bellido & Gehling, 2016;see Fig. 1D). Habelia clarified the thought-provoking complexity of the head of Sanctacaris, with both taxaand likely all of their close relatives, including Wisangocarisdisplaying an unparalleled alignment of seven fully developed cephalic appendage pairs, which form the basis of the extant chelicerate prosoma. The five core pairs in these taxa are multifunctional appendages combining sensory, grasping and crushing abilities. Although stemming from a different appendicular architecture, this evolutionary solution arguably mimicked the morphofunctional head of mandibulates, combined into single appendages (Aria & Caron, 2017b). It appears that this adaptation enabled the predation of small crawling animals with hard integuments --in essence, trilobite juveniles. No gut content has so far been found to verify this hypothesis, but biomechanical modelling (Bicknell et al., 2018) coupled with digestive remains in other taxa (Zacai et al., 2016) supports the view that shell and hard-cuticle crushing was a function acquired by early euarthropods with gnathobasipods, and upon which they radiated.
Habelia, Sanctacaris, Wisangocaris and other possible relatives (Lerosey-Aubril, Skabelund & Ortega-Hern andez, 2020), now grouped in Habeliida, also allowed a direct connection with horseshoe crab-like taxa from the Silurian thought to represent basal euchelicerates (Sutton et al., 2002;Briggs et al., 2012) through a particularly unwieldy character. Cephalic exopods in these taxa are leg-or antenna-like and seem to be somehow 'detached' from the basipod (Legg, 2014;Aria & Caron, 2017b). The alternative location of attachment of these exopods to the body is not known, but there is evidence that they moved independently from the rest of the main limb axis. This condition would hence be intermediary to the later loss of this limb ramus in chelicerates and would provide support for the developmental hypothesis that the 'exopod' of basal euarthropod taxa developed in fact as a separate limb axis (Van Roy et al., 2015), which would be called an exite instead of an exopod (Wolff & Scholtz, 2008) (Fig. 3D).
Given that Chelicerata are diagnosed by the eponymous chelicerae (Table 1), it is not clear whether habeliidans are strictly chelicerates, because the frontalmost appendages potentially homologous to chelicerae in these taxa are very small and not evidently chelate or sub-chelate. However, Mollisonia, another typical problematic taxon first introduced by Walcott, recently grounded the origination of chelicerates per se from at least the middle Cambrian, again from new material found at Marble Canyon   (Fig. 1C). The interpretation was criticized, based on the preservation of the chelicerae and their forward orientation (Budd, 2021). The chelicerae are indeed frontally directed and preserved both dorsally and laterally in two different specimens, consistently showing enlarged peduncles with opposing clawed tips, exactly as would be expected from a chelicera in these orientations (see extended data fig. 3 in . Preservation of frontally directed chelicerae is well known in pterygotid eurypterids (Tetlie & Briggs, 2009) and xiphosurids from the Fezouata biota (Van Roy et al., 2010), but is otherwise common among both extinct and extant arachnids (Pocock, 1900). In addition to chelicerae, Mollisonia sports sets of overlapping 'gills' reminiscent of the merostome book gills, albeit with a much-reduced number of constitutive elements. Because of this, Mollisonia and probably other mollisoniids Lerosey-Aubril et al., 2020) resolve as the sister group to Euchelicerata, further pointing to an early Cambrian origination of extant lineages.
The combination of the fossil and contemporaneous morphological evidence constrains xiphosurans and other merostomes as well as scorpions as basal within Chelicerata, which is at odds with the majority of molecular phylogenetic results retrieving both xiphosurans and scorpions as derived, the latter forming with other book-gill-bearing arachnids the Arachnopulmonata clade (Nolan, Santib añez-L opez & Sharma, 2020). The morphological data are also supported by the timing of the fossil record (Howard et al., 2020b), with the well-documented early radiation of xiphosurans (Bicknell & Pates, 2020), and fossil scorpions known since the middle Silurian, long before tetrapulmonates (Dunlop, 2010), while molecular results are supported by developmental gene expression (Nolan et al., 2020). Even if early tetrapulmonates were particularly cryptic, this would not explain the Cambrian polarization of the seventh prosomal limb pair (reduced but present as the 'chilaria' in horseshoe crabs) or book gills as plesiomorphic for euchelicerates. Alternatively, it must be considered that the molecular data are unexpectedly biased.

V. THE TRILOBITES OF BURIDAN
In order to illustrate the paradox of indecision and the human ability to choose without motive, French philosopher Jean Buridan used a fable in which a donkey let itself starve to death, incapable of choosing between two identical buckets filled with oats. An equally staggering indecision has long affected the placement of trilobites in the arthropod phylogeny (Edgecombe & Ramsköld, 1999;Cotton & Braddy, 2004;Scholtz & Edgecombe, 2006;Stein & Selden, 2012;Aria & Caron, 2017b;Zeng et al., 2017;Scholtz et al., 2019). Trilobites possess antennules, an a priori strong character to associate them with mandibulates (forming the eponymous Antennulata), especially since the ancestral euarthropod appendage is the cheira. They can also have setae on their exopods, like crustaceans often do, have large plate-like hypostomes, and it was shown recently that their eyes had a crystalline cone comparable to that of mandibulates (Scholtz et al., 2019). However, trilobites also bear gnathobasipods, sets of fully developed cephalic endopods and, importantly, tripartite apoteles (i.e. claws) that constitute robust apomorphies of Arachnomorpha (Aria & Caron, 2017b). The retrieval of the 'deep split' topology (see Fig. 4B) in which artiopodans are deeply nested within total-group Chelicerata suggests that the mandibulate-like characters can reasonably be interpreted as convergences, and in extreme, autapomorphicperhaps heterochroniccases could be related to a more pelagic lifestyle (Moysiuk & Caron, 2019b). The eye structure of trilobites would thus illustrate the problem of extrapolating evolutionary scenarios based on the association of an extant character with a few fossils, without considering that the absence of information in most other fossil forms could in fact hide polarization of this character as plesiomorphic. Zeng et al. (2020) recovered a 'deep split' topology preserving Antennulata, but the exceptionally high support values both at the Antennulata node and within Artiopodastill very difficult to resolve in other recent studies (Lerosey-Aubril, Zhu & Ortega-Hern andez, 2017;Mayers, Aria & Caron, 2019) is concerning. It may indicate that optimizations based on secondary homology led to the removal of important characters and the loss of phylogenetic signal that could be locally important within the phylogram (Congreve & Lamsdell, 2016;.

VI. HEAD PROBLEMS AND FOSSIL BRAINS
A series of groundbreaking studies interpreting neurological and other rare internal remains in Cambrian fossils, at first Biological Reviews 97 (2022) 1786-1809 © 2022 Cambridge Philosophical Society.
The origin and early evolution of arthropods from the Chengjiang biota (Ma et al., 2012bTanaka et al., 2013;Cong et al., 2014), have attracted much recent attention and delivered thought-provoking new evidence in the context of early arthropod evolution . One of these studies revealed the existence of complex visual systems in the iconic Chinese arthropod Fuxianhuia (Ma et al., 2012b), a find recently supplemented by the arguably distantly related Mollisonia from the Burgess Shale (Aria & Caron, 2017a). These fossils suggest that the presence of multiple optical neural centres originated early in euarthropods and were later repeatedly simplified in more derived taxa, for instance in arachnids and myriapods (Strausfeld et al., 2016b). This scenario provides an example that even complex and a priori generally advantageous structures such as efficient eyes remain governed by evolutionary trade-offs (Alexander, 1996).
Other studies also attempted to use neural remains in order to elucidate historical disputes about appendage homology in both extinct and extant arthropods (Tanaka et al., 2013;Cong et al., 2014). Central to this debate is the labrum, a generally pre-oral ventral structure found in a variety of shapes across extinct and extant arthropods (Scholtz & Edgecombe, 2006), typically associated with a sclerotic plate called a hypostome, and shown by some studies to originate from an appendicular protocerebral Anlage (Haas, Brown & Beeman, 2001;Kimm & Prpic, 2006). A series of recent investigations have provided renewed support and nuance to the concept of 'acron', according to which the labrum would in fact originate from an apicalmost, asegmental region of the forebrain with its own, distinct neuropils (Urbach & Technau, 2003;Posnien, Bashasab & Bucher, 2009;He et al., 2019).
The Cambrian palaeoneurological evidence was considered as supportive of the hypothesis that the frontalmost appendage of radiodontans was not homologous to the frontalmost appendage of early euarthropods, being instead reduced to form the labrum (Budd, 2002(Budd, , 2021Ortega-Hern andez et al., 2017). This view was contested, however, based on the alignment of anterior panarthropod metameres and the homologization of onychophoran antennae with similar protocerebral structures in euarthropods (Scholtz, 2016), but also on the argument that external morphoanatomy and phylogenetic analyses strongly support a continuous evolutionary history of the cheirae across early arthropods Zeng et al., 2020) (Fig. 3A).
It follows that the labrum more likely has a complex history across early arthropods, potentially involving the posterior migration of part, but not all the ancestral 'labral complex' (Fig. 3B). It has been shown that leanchoiliid juveniles possessed a well-developed labral protrusion (Liu et al., 2016(Liu et al., , 2020, confirming the predicted presence of this structure in megacheirans based on a reinterpretation of Oelandocaris oelandica from the Swedish 'Orsten' deposit (Stein et al., 2008;. This suggests that the ostracod-like frontal complex (including sensory organs and the labrum) observed in hymenocarines (Aria & Caron, 2017a;Vannier et al., 2018;Izquierdo-L opez & Caron, 2021) may have already dissociated from a posterior labrum, or perhaps that the individualization and posterior migration of the labrum occurred convergently in totalgroup Mandibulata and Panchelicerata/total-group Arachnomorpha (Fig. 3B). A developmental and evolutionary dissociation between protocerebral organs and the labrum itself is compatible in particular with the hypothesis of a separate labral origin.
Perhaps the directed effort in homologizing the tripartite brain (protocerebrum, deutocerebrum, tritocerebrum) in fossil taxa (Ortega-Hern andez et al., 2017) is misguided by the assumption that this brain is visibly tripartite in all fossils. It should be considered that the morphoanatomy of the brain itself has evolved, and therefore that brain subdivisions in fossils (in the form of fused and emerging ganglia) could mislead topological alignments based on extant taxa (Scholtz, 2016). A recent investigation may provide evidence to support this view by showing that pre-gnathal segments have different developmental properties compared to trunk segments, which is extrapolated into considering that the former originated from a single somite during the rise of euarthropods (Lev & Chipman, 2021). This is by far the best and most innovative explanation to the conundrum of 'prototo deutocerebral transition' of the cheira in panarthropods , and implies that single anterior connectives to the cheirae (Park et al., 2018) are not protocerebral in an extant sense, but 'metaprotocerebral', as they connect in fact to an undivided neural mass that later is separated into the proto-, deuto-and tritocerebrum as defined among crown euarthropods.
Some authors have also generally rejected palaeoneurological evidence based on the frailty of such internal tissues as ganglia and nerves and their high susceptibility to decay (Liu et al., 2018). Taphonomic and decay patterns these authors document seem to show convincingly that the published reconstruction of a vascular system in Fuxianhuia is dubious, and in general that peri-intestinal and haemocaelic structures are often neglected yet occupy a central importance in the understanding of arthropods from BST deposits Vannier et al., 2018;Mayers et al., 2019). This cautionary approach is not trivial because the general taphonomic shrinking of the peri-intestinal cavity as well as possible appendicular hemocoelic cavities has led some authors to misinterpret these long-known and common remains or even the gut as part of the central nervous system (e.g., Ortega-Hern andez, Lerosey-Aubril & Pates, 2019b; Zeng et al., 2020).
The presence of neural tissues in Cambrian fossils, however, remains supported by a solid line of evidence, as these also occur in areas not overlapping with other body parts and away from the gut, such as eye stalks, and where they are known to constitute a large portion of the organic mass (Ma et al., 2012b;Vannier et al., 2018;. The selective resistance of nerves to decay has also been demonstrated experimentally (Edgecombe, Ma & Strausfeld, 2015). In general, a temporal decay-based approach in experimental taphonomy is not applicable to fossils of BST deposits, because the selective taphonomy of tissues is based on idiosyncratic environmental and diagenetic conditions leading to this mode of preservation, as is generally the case for all Konservat-Lagerstätten (Parry et al., 2018). Nonetheless, we still lack a full causal understanding of specific tissue preservation in these deposits, which is why reports of this kind must remain particularly cautious (Scholtz, 2016).

VII. WEIRD WONDERS OF THE POST-CAMBRIAN
Although existing collections and further discoveries from the Cambrian certainly hold more surprises (Kylinxia being a recent example), it is also evident that the majority of Cambrian arthropods now fall within definite lineages, whether radiodontans, isoxyids, megacheirans, fuxianhuiids, hymenocarines, artiopodans, or the stem of extant groups (Fig. 2). In parallel, Silurian fossils from the Herefordshire biota in Wales, have, for a number of years, and alongside very modern-looking forms (Siveter, Sutton & Briggs, 2004;Siveter et al., 2010), revealed many arthropods with challenging morphologies, despite being three-dimensionally preserved animals generally yielding an impressive amount of morphological detail (Fig. 1J). Enalikter, for instance, was presented as a megacheiran (Siveter et al., 2014a), but this interpretation was not straightforward, for Enalikter arguably lacks any megacheiran apomorphy, and its frontalmost appendages are not clearly distinct from some tripartite crustacean antennules; yet, this is also clearly not a crustacean, and some authors went as far as interpreting it as a polychaete (Struck et al., 2015). This is a similar story to those of the 'weird wonder' days of the Burgess Shale. Taxa such as Cascolus, Aquilonifer, Tanazios or Xylokorys are similar in this regard: although they possess characters linking them with some known extinct or extant arthropod groups, their morphoanatomies also show significant differences affecting their stable phylogenetic position and leading to placement in their own group. This issue may result from a combination of the fact that they are Silurian, with much less soft-bodied data on arthropods from this period than there is available from the Cambrian BST deposits, and from discrepancies between types of preservation: with information provided by the Herefordshire material that a BST deposit lacks, and vice versa, differences between fossils may appear greater than they are. The renewed effort made by attempting to dissect fossils (Whittington, 1975b;Briggs, Bruton & Whittington, 1979;Aria & Caron, 2017a) and to obtain three-dimensional information from typically two-dimensional preservation (Zhai et al., 2019) promises to harmonize our morphoanatomical understanding. As a synthesis emerges and these data are better integrated, 'oddities' from the Herefordshire biota, but also from other exceptional Palaeozoic deposits yielding stem-group arthropods and euarthropods, such as the Hunsrück slate in Germany (Kühl & Rust, 2009, may prove to be more significant contributions to our understanding of the arthropod tree of life, and provide another dimension to the breadth of arthropod body plans after the Cambrian explosion.

VIII. TEMPORAL CONSTRAINTS
The accumulation of evidence in recent years that the origination of both mandibulates and chelicerates occurred deep within the Cambrian necessarily represents a strong timing constraint on the arthropod evolutionary tree. On the other hand, the first appearance datum (FAD) of trilobites is well constrained to the base of Cambrian Stage 3 (Zhang et al., 2017;Paterson, Edgecombe & Lee, 2019), and is documented also by the distribution of trilobite and lobopodian (Microdictyon) fragments among small shelly fossils (SSFs), which show relatively few discontinuities and have stratigraphic significance across the Lower Cambrian (Steiner et al., 2007). A wealth of traces that arguably only arthropod appendages could produce have been described from older sediments, deep into the Fortunian, but there is no solid evidence to date that would suggest the presence of arthropods before the Cambrian . Most of the panarthropod basic phenotypic pool would thus have appeared within 20 million years, with the presence of mineralized elements from Stage 3 implying that mineralization was an accelerating evolutionary factor in the specialization of masticatory appendages. The palaeontological evidence therefore points to an even more dramatic radiative event than has been assumed thus far, as is corroborated by wellcalibrated molecular clocks (Lee et al., 2013;Paterson et al., 2019). This necessarily has important implications on the genetic and phenotypic levels early in this group (Lee et al., 2013), but also for methodology, insofar as the evolutionary models guiding phylogenetic analyses require the flexibility to allow for rate heterogeneity among characters to be taken into account, alongside a process of stabilization of developmental pathways (Aria, 2020). In this context, parsimony is likely to be an oversimplistic approach to reconstruct relationships between basal taxa compared to likelihood-based methods (O'Reilly et al., 2016), potentially explaining historical conflicts in the reconstruction of early arthropod relationships using parsimony . The conceptual superiority of likelihood models to parsimony is defendable despite the fact that, ideally, a better model than the one currently in practice (Mkv + Г) would avoid establishing a general correlation between overall character rates of change and changes in branch lengths (Goloboff et al., 2019).
Very recent studies have also completed the ichnological record  with morphological evidence to constrain the timing of terrestrialization in both total-group mandibulates, through euthycarcinoids , and total-group chelicerates, or arachnomorphs, through aerial breathing in eurypterids (Lamsdell et al., 2020). The oldest known euthycarcinoids are from the middle Cambrian (Collette & Hagadorn, 2010) and the first eurypterids are from the Middle Ordovician (Lamsdell et al., 2015). Although both groups remained primarily aquatic, this suggests that excursions onto land were well underway by the end of the Cambrian for both of the extant euarthropod lineages, raising the question of what advantages these first land dwellers might have found on supposedly barren grounds.

IX. MACROEVOLUTIONARY PERSPECTIVES
The manifestations of evolutionary mechanisms on long timescales and among species involve asymmetrical patterns of morphological and taxonomic diversity due in part to the latency between genetic regulation and phenotypic expression (Jablonski, 2017). This asymmetry between disparity and diversity is particularly obvious in arthropods, emphasising that their high-level systematics have been shaped by evolutionary constraints and trade-offs at least as much as by phenotypic innovations (Aria, 2020). In that sense, the 'sculpting material' metaphor of the arthropod body is approached more interestingly from the perspective of what composed, and what can deform this material rather than simply the wealth of its possible shapes. Researchers have long worked on quantifying heterogeneous disparity patterns using arthropods as models, especially early disparity bursts among lineages Wills, Briggs & Fortey, 1994;Hughes, Gerber & Wills, 2013). They have also explored the promising and far-reaching avenue of defining persistent evolutionary trends, in terms of increasing morpho-functional complexity (Adamowicz, Purvis & Wills, 2008). This is of particular interest to our understanding of arthropod evolution, because this phenomenon appears early on as a driver of selection . Updating these analyses in the context of Mandibulata and expanding them to other groups could reveal a powerful explanatory factor for overall phenotypic evolution in Arthropoda.
Owing to their strong biomineralization, which correlates with their known abundance and diversity through the Palaeozoic, trilobites have generally been pioneering models to identify macroevolutionary patterns in arthropods and the fossil record as a whole (Eldredge, 1971). They also have been used to investigate early burst models of high Cambrian disparity preceding phenotypic canalization (Hughes, 1991;Webster, 2007), a view that was later refined to point out the variable relaxation of constraints on segment number across lineages, often associated with the co-evolution of adaptive features on a large scale (Hughes, Chapman & Adrain, 1999;Hughes, 2003;Webster & Zelditch, 2011). Reference ontogenetic work on trilobites, especially to reconstruct heterochronic trends (Hughes, 2007), should inspire research on soft-bodied larvae (Zhang et al., 2010;Liu et al., 2016;Fu et al., 2018), because heterochrony is another potentially highly significant explanatory variable of arthropod morphoanatomy over time.
In general, however, comparative studies on fossil arthropods are lacking. A preliminary top-down approach investigating disparity in euarthropods as a whole (Aria, 2020) found evidence that a 'displaced-optimum model' of evolution [that is, with swift but increasingly smaller translations from one adaptive peak to another (Hendry, 2007)] characterizes the rise of body plans in these animals, and that this phenomenon was associated with the rapid build-up of genetic regulatory networks, as suggested by others for all metazoans (Deline et al., 2018). The next step is to link these patterns to morphological characters, notably through studying co-variation in the context of heterochrony and developmental plasticity, as was done for trilobites. Although this integrated information will serve to refine our evolutionary models for phylogenetic analyses, now that a phylogenetic framework appears to be broadly stabilizing for fossil and extant arthropods (Aria, 2020;Edgecombe, 2020;Zeng et al., 2020), we should look beyond the sole genealogy and use these uniquely rich data to elucidate the many persistent unknowns of macroevolution.

X. CONCLUSIONS
(1) Research on early arthropod evolution is experiencing a turnover motivated by rich new fossil localities, powerful new imagery techniques and improved phylogenetic approaches.
(2) Suspension-feeding was a key and commonly exploited niche during the rise of panarthropods, and, together with other lines of evidence such as Cambrian crustaceomorph larvae, suggests an earlier radiation of arthropod-based plankton and possibly the 'planktonic revolution' as a whole.
(3) In comparison to the evolution of the mouth and its related structures, steps leading to the arthrodization of both podomeres and external segments, if they exist, are not yet clearly documented. In this context, isoxyids and other taxa basal to megacheirans are crucial to understanding the acquisition of features leading to Euarthropoda, as recently shown with Kylinxia.
(4) The majority of former 'stem euarthropods' have now been moved within the total groups of the extant lineages, Chelicerata and Mandibulata. Under the 'deep split' scenario, even megacheirans and Kylinxia could in fact represent basalmost members of total-group Chelicerata. This in turn implies that the origin of extant phyla goes much deeper within the Cambrian than previously thought, and therefore that the construction of body plans was even more dramatically rapid during that time.
(5) While mandibulates were characterized early on by a greater division of tasks between their appendages, both mandibulates and chelicerates had their early radiation marked by the diversification of the proximal elements of their limbs for mastication, likely associated with the spread of harder cuticles in arthropods and other animals.
(6) The phylogenetic position of trilobites and their allies is one of the oldest and arguably the last persisting uncertainty among the main early branches of Euarthropoda. (7) Soft-bodied post-Cambrian taxa with unusual morphologies, particularly from the Silurian Herefordshire and Devonian Hunsrück biotas, but also the Ordovician Fezouata, could expand and reshape our understanding of arthropod diversity and disparity during the early Phanerozoic.
(8) The growing stabilization of a fossil-inclusive panarthropod tree finally provides us with a trustable framework to use the largest amount of fossil and extant data among animals to explore macroevolutionary patterns and understand the deterministic factors behind the rise of these now threatened yet tremendously successful body plans.

XI. ACKNOWLEDGEMENTS
I thank Joe Moysiuk for various discussions and comments. I also thank Gregory Edgecombe and an anonymous reviewer for their contributions, and Alison Cooper for editorial suggestions. With respect to the first, peer-reviewed pre-print version of this work (PaleorXiv, 4zmey), I am particularly grateful to Tae-Yoon Park for assuming editorial responsibility and to both Jean Vannier and Gerhard Scholtz for providing fair and constructive reviews. In this context, the evaluation followed the peer-review transparency requirements of Peer Community In Paleontology, which are conducive to fair and objective evaluations, in agreement with the guidelines of the Committee on Publication Ethics. Additionally, this work could not have been completed without the support of Jean-Bernard Caron, Dongjing Fu, Xingliang Zhang, Fangchen Zhao, Maoyan Zhu and the Chinese Academy of Sciences. This work was supported by a President's International Fellowship Initiative grant (#2018PC0043) and a China Postdoctoral Science Foundation Grant (#2018 M630616). The author declares having no competing interests.