Anatomical and ontogenetic reassessment of the Ediacaran frond Arborea arborea and its placement within total group Eumetazoa

Organisms in possession of a frondose body plan are amongst the oldest and most enigmatic members of the soft‐bodied Ediacaran macrobiota. Appraisal of specimens from the late Ediacaran Ediacara Member of South Australia reveals that the frondose taxon Arborea arborea probably possessed a fluid‐filled holdfast disc, the size and form of which could vary within populations. Mouldic preservation of internal anatomical features provides evidence for tissue differentiation, and for bundles of tubular structures within the stalk of the organism. These structures connect in a fascicled arrangement to individual lateral branches, before dividing further into individual units housed on those branches. The observed fascicled branching arrangement, which seemingly connects individual units to the main body of the organism, is consistent with a biologically modular construction for Arborea, and raises the possibility of a colonial organization. In conjunction with morphological characters previously recognized by other authors, including apical‐basal and front‐back differentiation, we propose that to the exclusion of all alternative known possibilities, Arborea can be resolved as a total group eumetazoan.

F O S S I L S of macroscopic, soft-bodied organisms are found globally in late Ediacaran rocks of~570-541 million years in age. These fossils are considered to document a polyphyletic assemblage of diverse and morphologically complex marine organisms (Fedonkin et al. 2007 . This unit documents a variety of shallow-marine and deltaic depositional environments (Gehling 2000;Gehling & Droser 2013;Callow et al. 2013;Tarhan et al. 2017) and contains the impressions of thousands of organisms representing at least 30 distinct macrofossil taxa. Although the precise mechanism by which these fossils are preserved is a matter of considerable debate (Gehling 1999;Retallack 2007;Tarhan et al. 2016Tarhan et al. , 2018Bobrovskiy et al. 2019;Liu 2019), there is a general consensus that Ediacara Member palaeoenvironments were reasonably high-energy marine settings, and that the seafloor upon which the organisms lived was covered by benthic microbial mat communities (Gehling & Droser 2009;Tarhan et al. 2017;Droser et al. 2019).
Fossil assemblages of the Ediacara Member are perhaps most widely known for possessing some of the oldest candidate bilaterian animals (Gold et al. 2015;Cunningham et al. 2017) Sperling & Vinther 2010). Alongside these taxa, frondose organisms (Glaessner 1971) assigned to the unranked morphogroups Rangeomorpha and Arboreomorpha (Erwin et al. 2011) represent a comparatively little-studied component of the Australian Ediacaran assemblages. Frondose taxa are more typically known from older, deep-marine Ediacaran palaeoenvironments in Newfoundland (Canada) and England (Liu et al. 2015), but in the Ediacara Member they occur in shallow-marine facies interpreted to reflect deposition in delta front, sheet-flow and mass-flow depositional environments (Gehling & Droser 2013; see also Tarhan et al. 2016). Frondose taxa represented in the Ediacara Member include Charnia (Gehling & Droser 2013), Bradgatia sp. (Droser & Gehling 2015) and Pambikalbae (Jenkins & Nedin 2007), and their facies distributions contrast with the shoreface and wave-base sand settings in which non-frondose taxa are most abundant (Gehling & Droser 2013). However, numerous discoidal impressions, initially interpreted as medusoids (Glaessner 1984) but more recently reinterpreted as holdfast structures of frondose organisms (Tarhan et al. 2015), may indicate that frondose taxa were reasonably abundant within all Ediacara Member palaeoenvironments. Taphonomic variation in disc expression currently precludes identification of original taxa in situations where the frond is absent (Gehling et al. 2000;Burzynski & Narbonne 2015;Tarhan et al. 2015).
The most common frondose taxon in the Ediacara Member is Arborea arborea (Glaessner & Daily 1959) fig. S2), making this one of the largest known Ediacaran macro-organisms. A detailed reassessment of frondose taxa in South Australia synonymized specimens previously assigned to Charniodiscus oppositus, Charniodiscus arboreus, Rangea arborea, A. arborea, and even some Charnia sp. within A. arborea, following determination of the three-dimensional structure of Arborea branches (Laflamme et al. 2018). That study diagnosed Arborea as a bifoliate frond with second order branches that lack rangeomorph sub-divisions (consistent with Laflamme & Narbonne 2008;Erwin et al. 2011;Brasier et al. 2012;Laflamme et al. 2018): an arrangement that is distinct from that observed in the type Charniodiscus material from the UK. We concur with these opinions, but to avoid confusion we resist drawing morphological comparison to arboreomorph taxa described from outside of Australia in this study. Whereas rangeomorph taxa have historically been assigned to multiple, often contradictory, phylogenetic positions within the eukaryotes (summarized in Dunn et al. 2018), Arborea has only seriously been proposed to fall within either the hypothetical phyla Petalonamae (Pflug 1970(Pflug , 1972Hoyal-Cuthill & Han 2018) or Vendobionta (formerly Kingdom Vendozoa, more recently considered to be a class or order of rhizoid protists; Seilacher 1989Seilacher , 2007Buss & Seilacher 1994;Seilacher et al. 2003), or the Cnidaria (Jenkins & Gehling 1978). We here reassess the morphology of multiple Arborea specimens from South Australia, and build upon recent studies (Laflamme et al. 2018) to propose a new model for Arborea anatomy.

METHOD
We assessed 56 specimens that have either been historically assigned to Arborea, or recently synonymized with that taxon ( fig. S1). Many of the studied specimens are incomplete, and when originally catalogued by their discoverers (who include M. Wade, M. Glaessner, W. Sun, R. Jenkins and J. Gehling), they were assigned to several different taxa. We follow recent synonymization (Laflamme et al. 2018) of these specimens, but note that we cannot categorically reject the possibility that some specimens may derive from a different taxon. Care has been taken to base the principal findings of this study only on specimens we are confident derive from a single taxon conforming to the most recent diagnosis of A. arborea (Laflamme et al. 2018).
Most of the studied specimens are preserved as positive hyporelief impressions on the bases of sandstone beds, but some reflect composite impressions of original external as well as internal anatomy. A small number of specimens are preserved in three dimensions, as sandfilled casts typically documenting external morphology preservation of internal structures, and the ability for sediment to become incorporated within the specimens. We confirm those findings but interpret several additional anatomical observations to be biologically informative. We refrain from using phylogenetically loaded terminology in our description of Arborea, for reasons discussed in previous publications (Dunn et al. 2018).

RESULTS
Arborea arborea is composed of a holdfast, a stem, and an ovate, leaf-like frond comprising two rows of lateral branches (following Runnegar 1995) emanating from either side of a central stalk (Fig. 1A). Each branch within the frond comprises smaller sub-divisions (here called units, previously referred to as second order branches) that appear to lie behind a covering structure, or 'pod'  fig. S2). The frond outline transitions from tapering (in terms of branch length) at both tips in smaller specimens (fusiform), to tapering primarily at the apical tip. In a speci-men~4.5 cm in length (Fig. 3E) the basal-most branches are~40% of the length of the longest branches, whereas in a specimen~30 cm in length (Fig. 1A) the basal-most branches are~78% of the length of the longest branch.
The following description provides a model of the anatomy of Arborea (Fig. 6).
Arborea possesses a holdfast structure that may variously exhibit a small number of concentric rings (Fig. 1A, D), a prominent but smooth central boss (  fig. S3), or multiple radial grooves (Fig. 1F). Such structures have, when found in isolation, previously been referred to discoidal taxa such as Aspidella or Eoporpita (Wade 1972; Tarhan et al. 2017), but those are now largely interpreted as organ taxa, with much of the observed variation in discoid morphology asserted to be taphonomic in origin (Tarhan et al. 2015;Burzynski et al. 2017). The holdfast connects at its centre to a single stem ( Fig. 1), and varies in size relative to the width of the stem within the studied population, being of roughly equal diameter in some specimens (Fig. 1C, D), or 3-4 times larger in others (Fig. 1F). This variation does not appear to be directly correlated to specimen size (here measured as frond length), with a specimen of~30 cm in length (SAM P19690a; Fig. 1A) possessing a holdfast of 108.6 mm diameter, while another ≫74.45 cm (SAM P40858) possesses a holdfast of only 82.2 mm diameter. In one specimen, a holdfast is associated with an arcuate fan of sandy material (Fig. 1E). This fan does not exhibit any of the morphological characters typical of frond holdfasts (e.g. a central boss, or radiating striations), and a narrow projection of sand associated with the holdfast margin appears to connect the base of this disc to the 'arcuate fan' that lies stratigraphically above it. This relationship would be highly unusual in two overlapping discs. Together with its distinct morphology, this leads us to F I G . 3 . The 'sidedness' of Arborea. A, SAM P40785, the smallest specimen studied, with no visible sub-division of lateral branches. B, SAM P19690b, the tip of the frond is over-folded revealing the two sides of the organism; the bottom of the frond shows 'pods' and units, and the tip of the frond (over-folded section) shows undifferentiated rectangular branches with no visible 'pods' or units. C-D, SAM P34499 and SAM P35704b respectively, exhibiting smooth rectangular panels interpreted as the 'back' of the organism. E, SAM P48727 with lateral branches visible in one of the smallest described specimens. F, SAM P42686, 'pods' and units clearly visible (interpreted as the 'front' of the organism), with rectangular undifferentiated branches absent. All scale bars represent 10 mm. Colour online.
postulate that this fan does not reflect the impression of a second holdfast. We instead suggest that the sediment fan represents fluidized sediment emanating from a break in the wall of the large holdfast. The sediment fan is similar in morphology to lobate structures produced by fluid escape in other late Ediacaran mat-bound sedimentary units (e.g. the Longmyndian Supergroup of the UK; Menon et al. 2016).
Within the studied population, the stems can exhibit variable relative lengths (see Fig. 1A for a very short example), an observation that in other taxa has been considered functionally significant in terms of ecological tiering (Laflamme et al. 2012) or reproduction (Mitchell & Kenchington 2018). Stem length shows no clear relationship to frond size. Stems can be smooth and featureless (Fig. 1D), finely wrinkled ( Fig. 1C) or composed of numerous grooves and ridges that run parallel to their length into the stalk (Figs 3F, 4). These structures distally taper in width, and do not branch or amalgamate within the stalk. They do not continue into the holdfast in any studied specimen, and appear to record tubular structures extending up the stalk (Fig. 4). Along the length of the frond, individual tubes successively exit the stalk and become the primary axis for individual lateral branches (e.g. Fig. 4A). The tubes can connect to branches either at the margin of the stalk (Figs 2E, 4A, C), or closer to its centre (Fig. 4D).
The frond itself is composed of two rows of lateral branches (one on either side of the central stalk; Laflamme et al. 2018), which appear either bilaterally or alternately arranged across the midline. The longest branches are present in the middle of the frond, with branch lengths diminishing both apically and basally (Fig. 1A). Arborea has previously been described as possessing branches resembling 'pea pods' (Laflamme et al. 2018), with two sheet-like structures representing a continuation of the stalk wrapping up and around the serially-arranged units. Observed fronds typically show one of two possible branch variants. The first comprises solid, almost featureless rectangular blocks, which can occasionally exhibit transverse linear ornament. These abut one another to form a continuous smooth impression (e.g. Figs 1A, 3C). The second variant exhibits branches with a lenticular 'pod', partially covering a row of finely divided units along the length of the lateral branch (Fig. 2). In such cases, each lateral branch attaches to the central stalk via a single tubular structure (e.g. Figs 2D-E, 4). The distal end of each branch can also attach to the frond margin in some specimens, along what has previously been termed an undivided or marginal rim (Glaessner & Daily 1959;Jenkins & Gehling 1978). The secondary units within individual lateral branches can be oriented either apically or basally even within individual specimens (compare Fig. 2A, B), suggesting that they could pivot along the branch axis. In the smallest specimens, lateral branches appear bulbous, with no units visible (Fig. 3A). Each unit is rectangular to tear-shaped and may exhibit one order of transverse sub-F I G . 5 . The backing sheet and lateral margin of Arborea. A, SAM P40786, with lateral branches splitting off the stalk (at left), but also connecting to the lateral margin; linear striations running apico-basally between and seemingly beneath the lateral branches may indicate the presence of a wrinkled backing sheet underlying the branches. B, SAM P40772, exhibiting a striated surface, interpreted as the backing sheet, in between the lateral branches. C, SAM P40369, individual branches connecting to a lateral margin (at right). D, SAM P40773, revealing a striated backing sheet between the relatively smooth lateral branches. All scale bars represent 10 mm. Colour online. divisions along its length (Figs 2A, 3B; termed striations by Hoyal-Cuthill & Han 2018). These subdivisions appear to emanate in a single direction, suggesting a comb-like morphology for individual units.
The tubular structures running along the stalk connect to individual lateral branches in a one-to-one, fascicled, arrangement (Fig. 4). They then divide and orient themselves perpendicular to the lateral branch, before branching further, or debundling, at regular intervals ( Fig. 4A-C). Specimens only rarely exhibit both tubular structures and branch units. The tubular structures run up the lateral branches to their distal margin, dividing/debundling as they go to correspond, in a one-for-one relationship, with the expected positions of individual units that sit within the 'pod' (Figs 4A, C; 5A).
The lateral branches may additionally be underlain by a set of unidirectional linear striations arranged parallel (e.g. Fig. 5A, D) or oblique (Fig. 5B) to the marginal rim. These can be present across the entire width of the frond between the stalk and the lateral margin. This striated fabric may reflect a continuous sheet-like structure.

Model of anatomy
Holdfasts are rarely preserved in association with complete Arborea fronds, most likely due to both the large size of Arborea specimens and because in life much of F I G . 6 . An anatomical reconstruction of the Ediacaran frondose taxon Arborea arborea, based on the features discussed in this study. The 'back' (left) and 'front' (right) faces of the organism are shown. The right-hand side of the front shows the organism with the 'pods' and units (i.e. the branches) removed to reveal the underlying backing sheet. Inset: fine-scale arrangement of units within the 'pod'. Units are each connected to their own tubular, stolon-like structure running into the stalk. Note that pods (green) are free to pivot about the lateral branch axis. the holdfast may have been located beneath the sediment-water interface, and thus in a different plane of preservation (although preservation varies between beds; see Fig. 1A and Dunn et al. (2019, fig. S3) for examples of fronds and holdfasts preserved in the same plane). In the three clearest examples within the studied collection, where the complete frond and holdfast disc are articulated, there is no relationship between the size of the frond and the size of its associated holdfast, although the smallest specimen does possess the smallest holdfast structure. Laflamme et al. (2018) referred to one specimen (their fig. 2.2) as 'deflated'. Our observation of variable holdfast size is consistent with this interpretation. The ability of holdfasts to deflate, either during life or upon burial, is consistent with the organism being able to control and modify its shape. This interpretation is supported by the specimen with a fan of what appears to be escaping sediment (Fig. 1E), which may imply fluid fill within such holdfasts, and thus a potential ability to hydrostatically control holdfast size. An ability to actively modulate holdfast shape and size would imply the presence of contractile (muscular?) tissue (Jenkins & Gehling 1978), though in the absence of further data, contraction due to dehydration could represent an alternative possibility. An absence of contraction rims or disturbed sediment surrounding the specimens may suggest that this is unlikely.
The stalk of Arborea was likely to have originally been cylindrical (Laflamme et al. 2018), as supported by observed variation in the position of branch connection points, and the presence of both alternating and bilaterally symmetrical branch arrangements amongst the studied population. We consider at least some of this variation to result from rotation of the branch connection points out of the plane of preservation prior to compression of the cylindrical stalk, followed by their composite moulding on to the stalk in their 'rotated' positions. It is difficult to determine whether lateral branches were originally arranged in an alternating or bilaterally symmetrical manner, since these two branching arrangements are observed in almost equal numbers within the studied population.
The fascicled arrangement of tubular structures in the stalk and within the lateral branches (Fig. 4) appears to document the connection of individual units along each branch to the central part of the organism. These tubular structures extend into the stalk beyond the position expected of branches, and since Arborea is only known to possess two rows of branches, we do not consider the tubes to represent overprints of other lateral branches. The consistent one-for-one relationship of the tubes with individual lateral branches in multiple specimens precludes taphonomic interpretations such as wrinkling of an epithelium or a similar soft-tissue structure. It is not currently possible to determine whether these tubes were originally hollow or solid structures.
Since the tubular structures are most commonly observed when the pods and units assumed to reflect the exterior surface of the lateral branches are not preserved, we interpret the tubes as internal anatomical features. The relatively sharp boundary between these tubular structures and the smooth stem in some specimens (e.g. Fig. 4A) indicates that this difference is unlikely to be taphonomic in origin. Differential preservation of the smooth exterior of the stalk and these internal structures (Figs 3C; 4A, C) implies that they originally comprised different anatomical structures, suggestive of 'tissue' differentiation.
The tubular structures we report were documented and termed spicules by Glaessner & Wade (1966;see also Jenkins & Gehling 1978), an interpretation focusing on their sharp outlines and straight trajectories. However, their preservation as impressions rather than as biomineralized structures, the observation that they bend to extend into the branches, the presence of examples that curve and are clearly not straight within the stalk, and their ability to divide within the lateral branches (Fig. 4), lead us to question this hypothesis. True spicules in extant poriferans and cnidarians exhibit a variety of form. In cnidarians, calcitic spicules represent a derived condition, being present only in the Octocorallia. They are secreted by the mesoglea and are largely concentrated in the base of the colony, but may also be present in polyp leaves, or on anthocodia (Hyman 1940). In siliceous sponges, spicules are generally classified as either microscleres (smaller 'flesh' spicules) or megascleres (the main skeletal support elements). Megascleres are known to reach sizes of up to 3 m (and be up to 8.5 mm in diameter) in the basalia of Monorhapis chuni, where they function as a stalk (M€ uller et al. 2007). More commonly, microscleres are on the order of 1-60 lm, whereas megascleres are between 60-200 lm, and both can bundle and inter-weave (e.g. in the order Halichondrida; Hooper & van Soest 2004). The continuation of tubular structures up the stalk of Arborea and into its individual branches and units is an arrangement not seen in any extant spicular organism.
An alternative possibility, favoured here, is that the tubular structures in A. arborea represent non-mineralizing, stolon-like projections, consistent with their length, seemingly flexible nature, and one-to-one relationship with individual lateral branches and then units (Fig. 4). Stolons or stolon-like projections represent a derived condition in the Bilateria, but are nevertheless possessed by several invertebrate groups (e.g. the Bryozoa (Osborne 1984) and Entoprocta (Nielsen 2012, p. 201)) as well as many plants (de Kroons & Hutchings 1995)  If the holdfast of Arborea was hydrostatically regulated, some form of hydraulic system would be expected. We find no firm evidence for any such system, but note that some extant hydraulic systems, such as the inhalant and exhalent siphonozooids of pennatulaceans (Williams et al. 2012) are unlikely to be expressed in known specimens of Ediacaran frondose taxa due to their position beneath branch attachment points along the stalk. Alternatively, the fascicled tubes may have been involved in hydraulic regulation, particularly if the individual units to which they connect were open to the water column.
The 'backing sheath' in Arborea (the apparent connective structure that joins the stalk with the marginal rim) may have anchored the lateral branches in place, though Laflamme et al. (2018) proposed that the rim could alternatively reflect folding of the distal tips of the lateral branches. The Russian frondose taxon Charniodiscus yorgensis has also been interpreted as having first-order branches that are constrained along their horizontal axes, but unlike A. arborea, C. yorgensis is reconstructed as exhibiting full branching units on both sides of the organism (Ivantsov 2016). No fascicled branching arrangement has been noted in C. yorgensis despite the pyritization of internal anatomical features.
The observation that 'pods' and units can be present or absent in Arborea specimens, even within individual specimens (Fig. 3B), is consistent with the suggestion that they are only present on one side of the organism, conferring front-back differentiation (Fig. 6;Jenkins & Gehling 1978;Laflamme et al. 2018). The 'back' of the organism comprises the backing sheath, subdivided into rectangular blocks defined by lateral seams. The linear striations observed running behind lateral branches in certain specimens (e.g. Fig. 5A) are interpreted to reflect either the inner surface of the backing sheath, or a distinct layer within the organism. In addition to the clear apico-basal differentiation of the organism, this character could potentially assist in constraining phylogenetic affinities.
Lateral branches were attached to the stalk by both a tubular continuation of external tissue, and by the internal tubular projections (leading to apparent pairing of connections in some specimens; Gehling 1991). Lateral branches consist of two main elements: the 'pod', which was constructed of two lens-shaped sheets (not bound to each other at either their apical or basal margins) and the sub-rounded to comb-shaped units (Fig. 6, inset), which lay within the pod. Previous studies have considered subdivisions within second order units to reflect wrinkling of a soft tissue structure (Laflamme et al. 2018) but their consistent morphology both within and across specimens leads us to consider them biological features. We note that the first order branches of Arborea, being comprised of a lenticular 'pod' and subdivided units housed therein, differ fundamentally in architecture from the linear subdivisions seen in second and third order units. This distinction does not fit the 'self-similar' branching definition of the Rangeomorpha, and we therefore follow previous workers (e.g. Laflamme & Narbonne 2008) in considering branching arrangements in Arborea to be distinct.
If the pod does indeed surround the units, this has potentially interesting implications for the production of micro-eddies and flow disturbance around the units (which have previously been hypothesized to explain community dynamics in Ediacaran fronds; Singer et al.

Growth
The anatomical organization described above permits inference of the morphogenetic strategy of Arborea, which is informative when considering organismal affinities. The smallest, assumed to be youngest, specimens of A. arborea possess fewer branches than larger specimens. This suggests that branch growth and differentiation actively occurred during the frondose stage of the organism's life cycle, with new tubular structures presumably developing and terminally differentiating as the frondose organism grew (rather than undergoing a single event of terminal differentiation). We find no upper size limit to Arborea, and thus suggest that it may reasonably be interpreted to have displayed indeterminate (size) growth, with no known maximum number of branches. Significant branch differentiation appears to have occurred in small specimens, with the smallest known specimens (~3.5 cm) possessing~19 lateral branches. Arborea also shows a determinate (i.e. consistent and predictable) form within the studied population, with no evidence for aberrant branches (branches that are unusually long or short, or do not conform to the expected branching architecture; e.g. Kenchington et al. 2018). That the frond outline appears to change as specimen size increases, with the basal-most branches becoming relatively larger despite continued branch differentiation, suggests that new branches in Arborea differentiated from a (sub)apical generative zone (as indirectly inferred by Hoyal-Cuthill & Han 2018). We find no evidence for further, lateral generative zones.
An ordered fascicled branching arrangement requires a unidirectional guidance and pathfinding system along both the apico-basal and front-back axes. Pathfinding refers to the ability of a cell or group of cells to locate their final destination: neurons, for example, are able to find their destination by growing in permissive substrates and binding to adhesive cues (Raper & Mason 2010). Differentiation of the tubular structures (fascicles) into both branches and units occurs only after they emerge from the stalk wall, suggesting either the removal of an inhibitory signal within the stalk, or the presence of a positive differentiation signal in the stalk wall. The strategy outlined above is consistent with morphogenesis of branches in Arborea having occurred by localized outgrowth, as opposed to regional apoptosis (from an undifferentiated sheet). This is in line with many other forms of branching growth in extant eukaryotes, for example that seen in the alga Ectocarpus

Phylogenetic placement of Arborea
It is reasonable to assume that the anatomical complexity and large size of some Arborea specimens (~2 m in length) demonstrate that it was a multicellular organism, dwarfing even the largest multinucleate protists (xenophyophores). Indeterminate growth is compatible with several non-metazoan (e.g. Peterson et al. 2003) and metazoan (Sebens 1987) hypotheses of affinity, and is thus not considered an informative character here.
Arborea lacks the serially quilted arrangement that has been considered diagnostic of the Vendobionta, and inferred in some rangeomorph taxa (Seilacher et al. 2003;Seilacher 2007). The constrained form of Arborea within populations exhibits no aberrant branches, a lateral margin bounding the branches, and determinate changes in form (i.e. a transition from a fusiform to a distally tapering frond outline). This is inconsistent with the growth pattern of many extant modular groups (e.g. plant or algal groups), and some multifoliate rangeomorphs, which are characterized by a lack of constrained form (Kenchington et al. 2018). The differentiation of new branches as Arborea grew is also incompatible with a fungal affinity, where a fruiting body undergoes one round of terminal differentiation (Umar & Van Griensven 1997). We therefore consider that to the exclusion of extant non-metazoan comparators, A. arborea was a total group metazoan.
The constrained form, presence of two main body axes, and extensive body regionalization is incompatible with a poriferan affinity, but such an axial arrangement is compatible with a eumetazoan affinity. We recognize differential preservation of anatomical features in Arborea, with structures in the interior of the organism being preserved, and external structures being entirely or partially missing in different specimens. This implies that these structures were distinct, and potentially composed of different original materials, and could indicate tissue differentiation: a eumetazoan character. Possession of a fluid-filled holdfast, potentially indicating a capacity for hydrostatic regulation, is also compatible with, but not unique to, a eumetazoan affinity. On the basis of all available evidence, we therefore propose that A. arborea lies within the Eumetazoa. Such a phylogenetic position has been presented previously (Buss & Seilacher 1994;Hoyal-Cuthill & Han 2018; though we disagree with the monophyletic clade of Ediacaran organisms favoured by these authors) but this reassessment of Arborea provides developmental and anatomical support. Our current knowledge of anatomical characters in Arborea is insufficient to permit further constraint of its phylogenetic position.
The fascicled internal anatomy of Arborea suggests that each lateral branch grew independently of its neighbours, implying developmental independence and thus conforming to the definition of biological modularity. Such an arrangement is comparable with extant taxa that possess colonial organization (e.g. various hydrozoans; Hyman 1940) and it is therefore entirely feasible that Arborea could represent an Ediacaran colonial eumetazoan (contra Landing et al. 2018). Coloniality has previously been predicted to be the plesiomorphic condition for the Cnidaria, with A. arborea itself (then termed Charniodiscus) proposed to lie at the base of the cnidarian tree (Dewel 2000; see also putative stem-group colonial cnidarians from Cambrian Series 3; Park et al. 2011). However, more recent work (Zapata et al. 2015;Kayal et al. 2018) would suggest that this scenario is unlikely, with coloniality only being known in derived cnidarian positions. Ctenophores are not known to be colonial (we favour the view that Porifera represents the earliest diverging animal clade; Simion et al. 2017;Fueda et al. 2017) suggesting that the Ur-eumetazoan was a unitary organism. Coloniality is also noted as a derived condition within the Bilateria, with the only truly colonial phylum being the Bryozoa. If our interpretation of Arborea as a potentially colonial organism is correct, this may suggest that coloniality in eumetazoans was present in early-diverging groups. With no current evidence to tie Arborea to any crown group, this character could feasibly be present in early-branching positions of the eumetazoan stem-lineage, suggesting further (perhaps derived) excursions into the colonial state were possible, thus broadening the possible permutations of the eumetazoan ancestor.

CONCLUSION
Reconstruction of the anatomy and developmental biology of Arborea arborea leads us to conclude that it represents a total-group eumetazoan. In addition to previously recognized morphological characters (Laflamme et al. 2018) we note a distinctive fascicled internal branching arrangement and a fluid-filled holdfast. The different taphonomic expressions of structures within the studied Arborea collection imply the possible presence of different tissue types, and thus tissue differentiation. We conclude that Arborea was a modular organism, and note that it displays characters consistent with (but not exclusive to) a colonial body-plan, something previously argued to have emerged in eumetazoans only in the Ordovician (Landing et al. 2018). Key differences between Arborea and rangeomorphs support morphological distinction between these frondose organisms, hinting at multiple independent excursions into frondose morphospace amongst early diverging animal groups.