Main

Morphologically diverse animal fossils have been recognized in fossil assemblages of the late-Ediacaran period (around 575–539 million years ago (Ma)) and include examples of total-group eumetazoans14, cnidarians15,16 and bilaterians17. These fossils, along with molecular clock estimates2,3 and contentious biomarker data18,19,20, demand an Ediacaran existence of sponges, which are probably the most basal animal phylum1. However, few sponge fossils have been found from the Ediacaran period or earlier5,7,21. The absence of Precambrian sponge fossils has been attributed to the low preservation potential of siliceous sponge spicules due to low Al3+ concentrations in Precambrian porewaters22 or to the possibility that early sponges were aspiculate and entirely non-biomineralizing animals9. Here we report a crown-group sponge fossil, Helicolocellus cantori gen. et sp. nov., from the late-Ediacaran Shibantan limestone in South China (Extended Data Fig. 1). This new fossil is characterized by an organic latticework skeleton that is compositionally different from, but architecturally similar to and probably related to, spiculate hexactinellid sponges. It thus fills the late Neoproterozoic gap in sponge evolution and indicates that Precambrian sponges may have been aspiculate and non-biomineralizing animals, particularly if biomineralized skeletons evolved independently among sponge classes10,11,12,13.

Systematic palaeontology

Phylum Porifera Grant, 1836

Helicolocellus cantori gen. et sp. nov.

Etymology. Genus name from Greek/Latin helix, helix; and Latin locellus, small box. Species epithet in honour of the mathematician Georg Cantor (1845–1918), with reference to the Cantor set, which describes the regular, self-similar pattern of subdivided rectangular lattices as observed in this fossil.

Holotype. NIGP-176531 (Figs. 1 and 2) part and counterpart, deposited in the Nanjing Institute of Geology and Palaeontology (NIGP).

Fig. 1: Holotype of H. cantori gen. et sp. nov., NIGP-176531.
figure 1

a,b, Positive relief on bed top: photographed under reflected light directed from the upper right (a) and topographic elevation map from laser scanning microscopy (b). White arrows mark transition from bottom of conical body to the stem. The specimen is broken into two pieces along the yellow dashed line in a. c, Fracture surface along the breakage in a exposes a cross-section through the holotype specimen, showing a three-dimensional outline (dashed ellipse) infilled with coarser sparry calcite cement, in contrast to fine-grained micritic matrix. f, fringe; ff, first fringe; sf, second fringe. Scale bars, 50 mm.

Full size image
Fig. 2: Hierarchical rectangles on the surface of H. cantori gen. et sp. nov., holotype, NIGP-176531.
figure 2

a, Upper part of the body preserved in positive epirelief, viewed under reflected light directed from the upper right. Arrow marks fringe-like structure. b–d, Negative hyporelief counterpart impression with a different lighting angle. b, First-order rectangle in b corresponds to box in a. c, Second-order rectangle in c corresponds to box in b. d, Third-order rectangle in d corresponds to box in c, with dashed box marking a faintly preserved fourth-order rectangle. e, Schematic diagram of hierarchical rectangles. Black box marks first-order rectangle. Yellow, blue and red lines represent grooves that divide first-, second- and third-order rectangles, respectively. Scale bars, 10 mm (a), 5 mm (b), 2 mm (c), 1 mm (d).

Full size image

Referred material. Paratype: NIGP-176532. Other specimens: NIGP-176533–176538 (Fig. 3 and Extended Data Fig. 2).

Fig. 3: More specimens of H. cantori gen. et sp. nov.
figure 3

a, Paratype, NIGP-176532, with the body preserved in negative epirelief and the associated stem and disc in positive epirelief. b, NIGP-176533, body preserved in positive relief, with irregular arrangement of rectangular boxes towards the base of the specimen. Stratigraphic orientation uncertain. Arrows indicate two regions where the stem width expands from 32 mm to 48 mm. c, Presumed juvenile specimen, NIGP-176536, showing a truncated distal end. The body is preserved in positive relief, whereas the stem is a negative relief impression. Stratigraphic orientation uncertain. Scale bars, 50 mm (a,b), 20 mm (c).

Full size image

Locality and horizon. From the upper-Ediacaran Shibantan Member of the Dengying Formation at Wuhe, Yangtze Gorges area, Hubei Province, South China.

Diagnosis. A conical to subcylindrical body connected by means of a stem to a basal discoidal structure. The outer surface of the conical body is characterized by regularly arranged rectangles each subdivided into at least three orders of smaller rectangles, forming a hierarchical quadrate reticulation. Rectangles are orthogonally arranged in the upper portion of specimens and become helically twisted around the longitudinal axis towards the base. Rectangles are outlined by grooves which represent an organic cruciform skeletal structure.

Description. The holotype is preserved in positive relief on a bed top (Fig. 1), with a fragmentary impression of the corresponding negative relief preserved on a bed sole (Fig. 2b–d). H. cantori gen. et sp. nov. has a goblet-like morphology with a holdfast, a stem and a conical body, indicating a probable erect benthic lifestyle. The conical body is 291 mm in length (or height in reconstructed life orientation), with a maximum width of 108 mm, tapering basally to a connection with the stem. The stem is partially preserved as a flat and smooth impression, with a width of 23 mm, but the basal disc of the holotype is not preserved. The positive-relief specimen is broken into two parts (along the yellow dashed line in Fig. 1a and perpendicular to the bedding plane), allowing the observation of an elliptical cross-section through the partially flattened body, which is infilled with sparry calcite cement (Fig. 1c), implying an originally conical body with a central cavity.

The body wall exhibits a hierarchical latticework of rectangles, each of which bulges outwardly with a convex surface on the top bedding surface (Fig. 1b). The largest (first-order) rectangles (Fig. 2a,b) are subdivided into smaller rectangles, here termed second- and third-order rectangles (Fig. 2b–e). There are eight first-order rectangles per half-circumference along the distal end of the body. First-order rectangles are 12 ± 1.8 (1σ) mm long and 8.9 ± 1.2 (1σ) mm wide on average (n = 41 rectangles). They are separated from each other by grooves which are 1.1 mm wide and approximately 0.4 mm deep. Finer grooves (0.61 mm wide and about 0.2 mm deep) subdivide the first-order rectangles into four equal second-order rectangles, which have an average length and width of 5.3 ± 0.9 (1σ) mm and 3.8 ± 0.7 (1σ) mm, respectively (n = 164 rectangles). The second-order rectangles are further subdivided into third-order rectangles, which are 2.55 mm long and 1.8 mm wide (n = 146 rectangles), by successively finer grooves (0.3 mm wide and about 0.1 mm deep). Fourth-order rectangles are faintly preserved (Fig. 2d,e).

In the distal region, rectangles are oriented in transverse rows parallel to the arched top edge of the body. Rectangles become diagonally aligned in the middle part of the body. The basal part of the conical body is often poorly preserved, consisting of vaguely defined but poorly aligned rectangles without fine structure, possibly resulting from postmortem distortion, as evidenced by a separate specimen showing regularly and diagonally arranged rectangles at the basal part (Extended Data Fig. 2d). The holotype exhibits two fringes on one side of the body (the right-hand margin in Fig. 1a,b). The fringes each consist of a single row of rectangles, although one of them (labelled ‘ff’ in Fig. 1a,b; hereafter, first fringe) has more sharply defined rectangles than the other (labelled ‘sf’ in Fig. 1a,b: hereafter, second fringe). It is possible that the first fringe represents a longitudinal furrow or suture in the body wall. Alternatively, the two fringes may be attributed to the conical body splitting along a single seam due to compaction. The alignment of neighbouring rectangles in the main body seems to be congruent with those that comprise the fringes (Figs. 1a,b and 2a).

In addition to the holotype, seven other specimens are known (Fig. 3 and Extended Data Fig. 2), two of which are complete (Fig. 3a,c). The paratype NIGP-176532 (Fig. 3a) is 447 mm long and up to 93 mm wide. It possesses a basal disc (Fig. 3a; 57 mm in diameter), which is connected to the conical body (284 mm long) by means of a stem (163 mm long, 30 mm wide). The stem is smooth, with two broader regions where its width expands from 32 mm to 48 mm (arrows in Fig. 3a). The basal disc and stem are both preserved in positive relief on the bed top, whereas the conical body is a negative relief impression. The first-order rectangles have an average length of 12.2 mm (σ = 1.7) and a width of 9 mm (σ = 1.8) (n = 15 measurements). They are subdivided into second-order rectangles, which are approximately 5.2 mm long and 4 mm wide. The second complete specimen, NIGP-176536 (Fig. 3c), is only 113 mm long. It comprises a conical body that is 65 mm long and 38 mm wide, and a stem that is 48 mm long and 11 mm wide. It has a truncated distal end possibly representing an artefact of breakage. A basal disc is not present. The poorly defined and irregularly arranged first-order rectangles are about 7.3 mm long and 6.6 mm wide. The irregular arrangement of rectangles is also observed in specimens NIGP-176536 (Fig. 3b) and NIGP-176534 (Extended Data Fig. 2a). By contrast, specimen NIGP-176538 exhibits well-preserved, diagonally arranged rectangles in the lower part of the conical body (Extended Data Fig. 2d). One of the incomplete specimens, NIGP-176535, shows a single marginal fringe (Extended Data Fig. 2c) rather than double fringes as observed in the holotype.

The preservation style of Helicolocellus is identical to that of other non-biomineralized macrofossils in the Shibantan Member, including Arborea23, Flabellophyton24 and Wutubus25. Specifically, the distinction between the fossils and the sedimentary matrix is defined by lithological contrast, with proportionally more carbonate cement in the sediment filling the central cavity of the conical body (compare Fig. 1c with Fig. 7.2 of ref. 23, Fig. 9E of ref. 24 and Fig. 6g,h of ref. 25). The fossils are preserved as casts and moulds, but the body walls and their constituent elements are not preserved. This taphonomic style differs markedly from that of biomineralized tubular fossils such as Cloudina26 and Sinotubulites27 from the Shibantan Member, which preserve well-defined although secondarily replaced tests. This difference is probably because organic walls are more easily degraded and partially compacted (Fig. 1c), whereas a biomineralized test would be expected to retain its original three-dimensional morphology.

Discussion

The presence of a stem and a discoidal holdfast suggests an erect benthic lifestyle (Fig. 4a), but these features are not phylogenetically informative because they are also present in several distantly related benthic Ediacaran taxa, such as the putative cnidarian Haootia15, the macroalga Discusphyton28, arboreomorphs29 and some rangeomorphs (for example, Charnia14 and Primocandelabrum30), as well as many extant sponges31, cnidarians32, fungi and various algae. Instead, the stem and holdfast probably represent a convergent adaptation to a benthic lifestyle on microbially bound, firm substrates, which were prevalent during the Ediacaran period and are inferred to be present in the Shibantan assemblages on the basis of dark, organic-rich crinkly laminations observed in thin sections from this unit33. Common anchored benthic Ediacaran taxa such as Charnia34 and Arborea35 are leaf-like and constructed by branching modules, whereas H. cantori has a conical body consisting of a regular hierarchical latticework, evidencing a very different body plan and functional morphology.

Fig. 4: Morphological reconstruction and phylogenetic position of H. cantori gen. et sp. nov.
figure 4

a, Reconstructed life position of H. cantori on the Ediacaran seafloor. Artwork by D. Yang. b, Bayesian phylogenetic analysis of metazoans, based on character matrix modified from refs. 14,54,68. H. cantori is resolved as a stem-group hexactinellid along with Eiffelia, Diagoniella, Protospongia, Cyathophycus and Minitaspongia. The total-group hexactinellid clade shows a basal polytomy of three branches: 1, Helicolocellus and reticulosans; 2, Eiffelia; 3, crown-group or extant hexactinellids. Numbers are posterior probabilities for nodes. See Supplementary Information for morphological data matrix.

Full size image

The main body of Helicolocellus shares a conical to cylindrical morphology with late-Ediacaran taxa, which have previously been likened to sponges, but possesses sufficient morphological differences to justify the establishment of a new taxon based on the Shibantan material. The putative sponge Thectardis36,37 from Newfoundland is inferred to have possessed a conical body which is similar to Helicolocellus, but Thectardis typically shows a featureless external surface and lacks a stem or basal disc. Another cylindrical fossil, the putative sponge Ausia from the Nama Group of Namibia38, has millimetre-scale pore-like structures on its surface, albeit notably smaller than the first-order rectangles observed in H. cantori and lacking second-order divisions. Palaeophragmodictya, a discoidal fossil found in South Australia and the White Sea region of Russia, was initially interpreted as a sponge39 and shares reticulate surface patterns which are smaller than those of H. cantori. However, it differs markedly in shape, being discoidal rather than conical, and more recent research indicates that it may represent an attachment disc of frondose organisms or a microbially produced texture5. The Ediacaran fossil Gibbavasis40,41 from Iran and Moldova shares a lattice-like appearance with Helicolocellus but possesses only a single order of orthogonally arranged ‘boxes’. Furthermore, Gibbavasis is much smaller (4–14 mm long, 2–7 mm wide) than Helicolocellus. Perpendicular cross-hatched structures have also been described in an unnamed fossil from Nevada (Fig. 3a,b in ref. 42). However, its surface ornamentation can be readily distinguished from the reticulation observed in H. cantori by the absence of secondary latticework.

The box-like latticework pattern of Helicolocellus is superficially similar to the transverse and longitudinal furrows on the body walls of some living hexacorallians (Cnidaria: Anthozoa), such as stony corals (Scleractinia) and sea anemones (Actiniaria). In these hexacorallians, the furrows correspond to a latticework of internal muscle fibres known as a muscle field43. However, the box-like structures in the Hexacorallia are not subdivided into finer units, nor are they arranged diagonally as in Helicolocellus and sponges. Furthermore, the irregular arrangement of boxes observed in some Helicolocellus specimens (for example, Figs. 1a and 3b) cannot be easily explained by disturbance of the contractile muscle field seen in the Hexacorallia. Contractile muscle fields are typically highly variable in length-to-width ratio, inconsistent with the relatively stable ratio of Helicolocellus specimens (around 3.0, measured on three specimens) in Helicolocellus. More importantly, tentacles are considered a synapomorphy (shared derived characteristic) of the Anthozoa44 but are not present in Helicolocellus, thus excluding Helicolocellus from the total-group Hexacorallia. Similarities with the longitudinal and transverse muscle bundles of extant tunicates can also be refuted by the ability of Helicolocellus rectangles to behave as discrete individual blocks, which can result in irregular arrangement, particularly in the lower part of the conical body (Figs. 1a and 3b). The enigmatic Cambrian fossil Paramackenzia, which has previously been compared to modular Ediacaran organisms and exhibits potential three-dimensional box-like compartmental structures45, has passing similarity, but its compartmental structures neither subdivide nor show a helical arrangement.

The body plan of Helicolocellus, with its goblet-like shape and especially the hierarchical rectangular ornament, is highly similar to the overall morphology and skeletal grid of Palaeozoic hexactinellid sponges, in which the pattern results from the presence of regularly arranged biomineralized spicules. For example, stem-group hexactinellids such as the Protospongiidae (Extended Data Fig. 3a,b) and the Dictyospongiidae (Extended Data Fig. 3c,d) are characterized by similar regular and uniformly divided meshes of spicules46. Dictyospongiids, which are funnel-shaped or cylindrical sponges, can reach considerable sizes, with genera such as Hydnoceras growing larger than 250 mm (ref. 46). Clathrospongia and Minitaspongia exhibit first-order meshes that are 1.5–15 mm wide and 2–15 mm long47,48, which are comparable in size to those of Helicolocellus. In some cases, the orthogonal rays of spicules that form the skeletal grids and outline the ‘boxes’ in Clathrospongia may appear as deep impressions as a result of the dissolution of spicules during diagenesis48, resulting in a preservational style virtually identical to Helicolocellus. The protrusion of sediment through the spaces between the orthogonal rays may also result in a rectangular pattern47 similar to that of Helicolocellus. Some other Cambrian sponges, for example, the demosponge Vauxia and the ascosponge Leptomitus, also exhibit a grid pattern49. However, these patterns differ significantly from the regular and hierarchical grid pattern found in Helicolocellus and other Palaeozoic hexactinellid sponges in having less hierarchy in their organization and less regularity in the spicule arrangements.

The hierarchical skeletal grid in these Palaeozoic hexactinellid sponges consists of multiscale bundles of spicules, typically either fused together or loosely assembled to form a latticework. Although Helicolocellus lacks direct evidence for a biomineralized skeleton, it probably had a somewhat rigid skeleton consisting of discrete cruciform elements, considering the presence of both regularly and irregularly arranged reticulate patterns in observed specimens. It is thus taphonomically similar to the Devonian fossil Pontagrossia50 (Extended Data Fig. 3f), which has been compared to a sponge and is inferred to have possessed a largely organic skeleton characterized by a reticulate pattern. The skeleton of Helicolocellus could similarly have been originally organic, considering its taphonomic style similar to other non-biomineralized macrofossils such as Arborea23, Flabellophyton24 and Wutubus25 but different from the biomineralized tubular fossil Cloudina26 in the Shibantan Member. Additionally, the outwardly bulging upper surface of the rectangles in Helicolocellus (as evidenced by convex-up positive epirelief preservation, seen in the holotype) is consistent with a flexible wall or membrane which was pushed outward during sediment infilling and compaction and impressed against a more rigid external framework.

Some clusters of rectangles in Helicolocellus are irregularly arranged, particularly in the lower part of the conical body (Figs. 1a and 3b). These irregularly arranged rectangles may result from the dislocation of some rectangular elements during either degradation or compaction. Similar irregular arrangement is also observed in protospongiid sponges (Extended Data Fig. 3b), which are constructed of loosely articulated cruciform units known as stauractine spicules which are prone to postmortem dislocation51. This observation indicates that the skeleton of Helicolocellus was constructed of unfused skeletal elements. Considering the possibility that early Palaeozoic sponge spicules were weakly biomineralized and contained large proportions of organic matter, with full biomineralisation only seen in later sponges9, we might expect early sponges such as Helicolocellus to have organic, unfused skeletal elements prone to dislocation.

The attachment strategy of Helicolocellus can also find analogues among younger sponges. The Jurassic protospongiid-like sponge Ammonella, for example, shows a regular and hierarchical meshwork of stauractines49, resembling the pattern observed in Helicolocellus, and was also anchored to a potentially microbially stabilized substrate by a discoid root plate52. This attachment strategy bears similarity to the way Helicolocellus adheres to the (presumably microbially stabilized53) Ediacaran substrate.

To more rigorously test the possible hexactinellid affinity of Helicolocellus, a Bayesian phylogenetic analysis was conducted using a dataset consisting of 79 taxa (including 8 fossil taxa) and 235 characters, with 67 of these characters scored for Helicolocellus (see Supplementary Information for methods and details of the character matrix). Notably, Helicolocellus was assigned to have a highly regular hierarchical reticulate skeleton (character 176). The analysis recovers Helicolocellus as a crown-group sponge and a stem-group hexactinellid (Fig. 4b and Extended Data Fig. 4). This result is unsurprising given that all crown-group hexactinellids have biomineralized spicules54 and Helicolocellus does not. Moreover, the Bayes factor (15.82) indicates a strong statistical support for a crown-group sponge placement of Helicolocellus over a stem-group Porifera alternative (Supplementary Information). The results remain stable in several sensitivity tests: when ctenophores are constrained as the sister-group to all other animals55 (Extended Data Fig. 5a) and when the relationships of the classes in Porifera and the internal relationships in Hexactinellida are constrained to follow recent molecular phylogenies10,54 (Extended Data Fig. 5b).

The latticework of Helicolocellus may have facilitated feeding or mechanical stability. As in modern sponges and diploblastic animals, early sponges probably lacked specialized internal organs and may have depended on diffusion for gas exchange and osmotrophic or filter feeding32,56. A simple and effective strategy to achieve these functions is to increase the ratio of surface area to volume, as occurs in several extant shallow-water sponges, in which organic particles and dissolved organic matter serve as a chief food source57. Modern hexactinellid sponges possess a syncytial pinacoderm and a water canal system lined with choanocytes58 (branched choanoderm). The various forms (for example, folding, branching and anastomosis) of hexactinellid skeletons have been interpreted as an adaptation to maximize the choanoderm surface area in the limited space occupied by the sponge59. The repeatedly divided box pattern in Helicolocellus may represent a similar strategy to increase surface area. This pattern resembles a self-similar fractal known as the Cantor dust set, which is generated through a recursive process of repeatedly inserting a central cross in rectangles, resulting in successively smaller rectangles which are geometrically similar to one another (Fig. 2). Fractals are ubiquitous in biology but appear almost exclusively in the form of branching tubes60, such as in the lungs, leaf veins and plant roots. Non-branching fractals are exceedingly rare in organisms. The Cantor dust set is unique to Helicolocellus and some Palaeozoic sponges. This multiscale hierarchical structure may also provide mechanical benefits; broadly similar skeletal systems in modern hexactinellid sponges, such as Euplectella, have been shown to contribute to mechanical stability61.

The overall morphology and regular grid-like pattern on the body wall of Helicolocellus are consistent with an affinity with sponges, particularly hexactinellids. However, there are two key features of modern hexactinellids that are not observed in Helicolocellus. First, the reticulation of modern hexactinellid sponges is typically constructed by cruciform stauract spicules bounded by soft tissue or by fused spicules, whereas no mineralized spicules have been found in Helicolocellus. This apparent absence is probably original, with any cruciform skeletal elements in Helicolocellus being non-biomineralized. The last common ancestor of sponges may have been aspiculate11,12,13, and the siliceous spicules of the modern sponge classes—Hexactinellida, Demospongiae and Homoscleromorpha—have been shown not to be homologous13. Those sponge classes may have independently acquired mineralized skeletons along with other metazoan lineages in the early Cambrian10,11,12,13. Palaeontological and molecular phylogenetic analyses have not arrived at a conclusive resolution with regard to the origin(s) of biomineralization in sponges12, with some studies advocating the origin of biomineralized spicules in the last common ancestor of the Silicea62 or even Porifera6 and others entertaining the possibility of independent origins of siliceous spicules in the Hexactinellida, Demospongiae and Homoscleromorpha9,13,63. Our phylogenetic placement of Helicolocellus as a stem-group hexactinellid that possesses some (for example, a reticulate skeleton) but not all (for example, biomineralized spicules) features of the crown-group Hexactinellida is consistent with either independent origins or secondary loss of siliceous spicules. Second, the surface of Helicolocellus lacks evidence for ostia (inhalant pores). It is possible that the minute size of ostia, such as those observed in extant hexactinellid sponges (for example, 4–30 µm; ref. 64) may not have been preserved in this deposit. The smallest resolvable features preserved in the Shibantan limestone are tertiary branches of Charnia, which have a submillimeter minimum dimension (ref. 65), an order of magnitude larger than would be predicted for ostia. Our ability to determine the presence of an osculum-like structure in Helicolocellus is hampered by the lateral compression of all specimens at hand, but the three-dimensional cement-filled cross-section through the holotype (Fig. 1c) suggests the likely presence of a central cavity in Helicolocellus, which could be homologous to the spongocoel in modern sponges.

It is also worth commenting on the phylogenetic placement of the hexactine-bearing Reticulosa, which has traditionally been assigned to the hexactinellids66 but has since been proposed to be paraphyletic to the Hexactinellida, Demospongiae, Calcarea and Homoscleromorpha and even the entire Porifera6. Our phylogenetic analysis indicates that at least some extinct reticulosan taxa, that is, Diagoniella, Protospongia, Eiffelia, Cyathophycus and Minitaspongia, along with the heteractinid Eiffelia, are grouped with extant hexactinellids (Fig. 4b). Therefore, reticulosan and heteractinid sponges may represent stem-group hexactinellids.

To conclude, H. cantori represents an Ediacaran crown-group sponge with an organic skeleton that is architecturally similar to the Hexactinellida. If siliceous biomineralization evolved independently in the Hexactinellida, Demospongiae and Homoscleromorpha13, then a pre-existing organic scaffold with a regular hierarchical reticulate skeleton (as present in H. cantori) may have served as a template for subsequent acquisition of biomineralized spicules. An important ramification is that we should broaden our search image of Precambrian sponge fossils, not only because they may have been aspiculate if sponge biomineralization evolved several times (for example, Tonian candidate keratose sponge material from Canada7) but also because stem-group representatives necessarily lacked some features diagnostic of their crown-group counterparts67. This emphasizes the phylogenetic importance of the fossil record in the search of the evolutionary root of sponges and indeed all animals.

Methods

Fossil specimens were collected from a single stratigraphic horizon about 2–2.5 m above the base of the Shibantan Member (around 551–543 Ma) at Wuhe village in the Yangtze Gorges area, South China (Extended Data Fig. 1). All specimens are preserved on limestone bedding surfaces and are deposited at the NIGP, Nanjing, China. Photographs were taken using a Nikon D850 DSLR camera and a Zeiss Axio Zoom V16 microscope. Measurements were carried out on fossil images using ImageJ 1.52a and analysed using Microsoft Excel 2013. Laser scanning data (Fig. 1b) were obtained using the Faro Design ScanArm and processed using the software Geomagic Warp 2017 and CloudCompare 2.13 to capture surface details and to generate elevation maps.

Phylogenetic analysis was conducted on the basis of a previously published character matrix for metazoans14,54,68 with modifications. No statistical methods were used to predetermine sample size. A total of 235 morphological characters were coded for 79 taxa. Only characters that are parsimony-informative were included in the analysis, to accentuate shared derived characters (synapomorphies). We scored H. cantori for 67 of these morphological characters (see Supplementary Information for details). Bayesian phylogenetic analysis was run using MrBayes 3.2.7 (ref. 69) on the CIPRES Science Gateway70. Analyses were run for 6,000,000 generations, sampled with a frequency of every 1,000 generations, discarding the first 25% samples as burn-in. The average standard deviation of split frequencies was about 0.01 in all runs. The effective sample size, calculated using Tracer 1.7 (ref. 71), indicated that all parameters had effective sample size scores above 200.

We compared the Mk model with gamma and lognormal distributions, considering both symmetric and asymmetric transition frequencies. Additionally, we evaluated the topological hypotheses, specifically whether Helicolocellus is a stem-group sponge or a stem-group hexactinellid. To assess the strength of support for different models and hypotheses, we calculated marginal likelihoods, which were computed using stepping-stone sampling with 50 steps and 20,000,000 generations. The marginal likelihoods for each model were used to calculate Bayes factors and to determine the best-fit model72. Hard constraints were applied to all nodes during the stepping-stone sampling analysis for the two hypotheses. The results revealed that the model with a lognormal distribution and asymmetrical transition frequencies and the topology of the stem-group hexactinellid hypothesis were better supported (Supplementary Information and Supplementary Table 1). Therefore, the model with a lognormal distribution and asymmetrical transition frequencies was used in downstream analyses, as presented in Fig. 4b and Extended Data Fig. 4.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.