Significance Phylogenomic extrapolations indicate the last common ancestor of sponges and eumetazoans existed deep in the Cryogenian, perhaps 200 million years (Myr) before the Cambrian (541 Ma). This inference implies a long Precambrian history of animals phylogenetically allied with sponges. However, there is yet little unequivocal paleontological evidence of Precambrian sponges. Here, we present a newly discovered 600-Myr-old fossil preserved at cellular resolution, displaying multiple poriferan features. The animal was covered with a dense layer of flattened cells resembling sponge pinacocytes, displaying a hollow tubular structure with apparent water inflow and outflow orifices. Although requiring additional specimens of similar form for confirmation, this finding is consistent with phylogenomic inference, and implies the presence of eumetazoan ancestors by 60 Myr before the Cambrian.

Abstract An extraordinarily well preserved, 600-million-year (Myr)-old, three-dimensionally phosphatized fossil displaying multiple independent characters of modern adult sponges has been analyzed by SEM and synchrotron X-ray tomography. The fossilized animal (Eocyathispongia qiania gen. et sp. nov.) is slightly more than 1.2 mm wide and 1.1 mm tall, is composed of hundreds of thousands of cells, and has a gross structure consisting of three adjacent hollow tubes sharing a common base. The main tube is crowned with a large open funnel, and the others end in osculum-like openings to the exterior. The external surface is densely covered with flat tile-like cells closely resembling sponge pinacocytes, and this layer is punctuated with smaller pores. A dense patch of external structures that display the form of a lawn of sponge papillae has also survived. Within the main funnel, an area where features of the inner surface are preserved displays a regular pattern of uniform pits. Many of them are surrounded individually by distinct collars, mounted in a supporting reticulum. The possibility cannot be excluded that these pits are the remains of a field of choanocytes. The character set evinced by this specimen, ranging from general anatomy to cell type, uniquely indicates that this specimen is a fossil of probable poriferan affinity. So far, we have only this single specimen, and although its organized and complex cellular structure precludes any reasonable interpretation that its origin is abiogenic, confirmation that it is indeed a fossilized sponge will clearly require discovery of additional specimens.

Extensive phylogenomic analyses have converged on a distant Precambrian vintage for the origin of Metazoa, an evolutionary event that must have long predated the last common ancestor of Porifera (sponges) and Eumetazoa (cnidarians + bilaterians), given the organismal and genomic complexity of these major groups. In modern phylogenomic analyses, elegant computational algorithms are applied to sequence divergence data from very large sets of genomic protein and other sequences so as to establish evolutionary branching order, and thus patterns of relatedness among animal clades. Key branch points can then be referred to real time by use of extrapolations from calibrations afforded by the dated fossil record. Various analyses of this type have been published, as well as an authoritative recent review by Erwin and Valentine (1). All calibrated phylogenomic analyses agree on a remotely Precambrian evolutionary origin of sponges.

Although earlier studies portrayed the four extant sponge classes as polyphyletic, technically sophisticated recent analyses indicate Porifera to be monophyletic, that is, to have had a single common evolutionary origin (2, 3). The result implies a last common ancestor in metazoan phylogeny that gave rise to the sponges [and possible metazoan sister groups (3)] on the one hand and to Eumetazoa [plus possible sister groups (3)] on the other. From the enormous and detailed overlap in the gene toolkit of sponges and eumetazoans (4, 5), there can be little doubt that these two lineages indeed derived from a common ancestor. This sponge/eumetazoan ancestor is predicted by phylogenomics to have lived deep in the Cryogenian (6), the geological period that terminated with the end of the Snowball Earth glaciations at 635 Ma, although how deep is not clear at present (7). The Cryogenian origin of this last common ancestor does not tell us, however, when animals displaying the character complexes that identify modern sponges first arose, only when their stem group ancestors diverged from the stem group ancestors of the eumetazoans.

Tantalizing paleontological claims of the Precambrian fossils that display one or another sponge-like characteristic have appeared in recent years, as critically reviewed by Antcliffe et al. (8). We leave aside isolated objects that are claimed to be fossilized siliceous spicules, because these objects have little morphological information content, and also the numerous claims of fossilized sponge-grade organisms from several parts of the world dating to about 555 Ma, close to the Precambrian/Cambrian boundary. There remain, however, several proposed sponge-like fossils of sufficiently deep temporal provenance to be relevant in principle to the phylogenomic predictions, although it must be noted that none are accepted as fossil sponges in the study by Antcliffe et al. (8). Among these putative fossilized sponges are thin-walled, hollow microfossils, perforated by various openings, found in rocks as old as 760 million years (Myr) old (9); asymmetrical ellipsoid forms that appear to be perforated with interconnecting canals of pre-Marinoan Cryogenian age when tomographically reconstructed from images of serial surfaces (10); and a macroscopic triangular impression fossil interpreted as the remains of a conical sponge-like form of Ediacaran age, about 575 Myr old (11). In addition, the recovery from Cryogenian-aged rocks of an organic sterol biomarker that is synthesized by marine demosponges suggested the early presence of this poriferan clade (12), although the conclusion that the provenance of this fossil sterol indicates the existence of Cryogenian sponges is now also questioned (7). Despite phylogenomic extrapolations that indicate divergence of sponge lineages from eumetazoan lineages deep in Precambrian time, lack of substantial paleontological evidence directly supporting this prediction has thus remained frustrating. Such evidence would require recovery and analysis of much better preserved fossils, displaying identifiable sponge characters but dating to the earlier part of the period following Snowball Earth (i.e., earlier Ediacaran) or even before that, to the Cryogenian period. Because sponges do not display tissue-grade anatomical characters or organs beyond a general gross, sponge-like structure, this requirement would essentially demand preservation of recognizable cell types in the fossil. Unfortunately, none of the currently proposed deep Ediacaran or Cryogenian fossils display any evidence of particular cell types.

Here, we present a new, remarkably preserved, phosphatized fossil, recovered from rocks of the early Ediacaran Doushantuo Formation in South China, which are dated to about 600 Myr old (stratigraphy and geological context are presented below). The fossilized animal, about 2–3 mm3 in size, was composed of easily recognizable cells. It displays the unmistakable gross anatomy of an adult sponge-grade animal, but beyond this finding, several distinct cell types and cellular structures can be clearly recognized, as displayed in the figures accompanying this report. The Doushantou Formation has been the source of a remarkable series of fossils that similarly reveal cellular level preservation. These phosphatized microfossils include metazoan as well as other forms, such as acritarchs (13), multicellular algae, and Volvox-like forms displaying cellular detail (14⇓⇓–17). In the present context, the most important Doushantuo microfossils are cleavage stage and later embryonic forms that apparently represent a variety of different animal lineages at diverse developmental stages (14, 18⇓⇓⇓⇓–23). Adult forms have been reported only rarely, which has increased the difficulty of interpreting the putative fossilized embryos. However, reported Doushantuo microfossils include small tubular cnidarians (24, 25) and a small bilaterian form, Vernanimalcula (26), of which multiple fossils have been recovered (27). Alternative interpretations have been proffered for virtually all of the Doushantuo microfossils (28⇓⇓–31). The present report, which describes an unmistakable adult animal form, will alter the structure of this debate, although the full force of the implications will not be realized until more than this single specimen becomes available.

Discussion In its overall construction, the remarkably well-preserved fossil organism, of which Fig. 1A provides the best overview, displays definitive poriferan characteristics. The fossilized animal was asymmetrical, lacking body axes but presenting a differentiated basal-to-apical structure. Its basic organization is tubiform, consisting of three relatively thick reflexed tubes all emerging from a common basal apparatus in this case, and all opening apically to the outside. The largest tube possessed a prominent funnel-like outflow structure, and each of the others terminates in specialized oscula. Additionally, as in modern sponges, the fossilized organism had numerous pore-like openings in the main body walls of the tubes, through which water could flow in. The animal was apparently sessile with a specialized, dense basal tissue, perhaps functioning like a sponge holdfast, and all of the walls of body tubes were covered with a sponge-like, multicellular external wall. The cellular structure of this wall is strikingly similar to the cellular structure of sponges of some modern classes in displaying distinct, plate-like pinacocytes. These pinacocytes, as in modern sponges, lacked continuous epithelial connections; cross-sectional tomographic observations do not reveal any signs of basement membranes, which are also lacking in modern sponges. In the funnel area and also the holdfast, the inner and outer layers of the body were distinct, again similar to the basic diploblastic organization of sponges. In addition, this animal possessed a sharply delimited external patch of differentiated, hollow tubular structures, perhaps highly modified pinacocytes, resembling the papillae of demosponges. There is no sign of mesohyl, but the gelatinous nature of this material in modern sponges is unlikely to have provided a sufficiently structured substrate for mineralization. More problematic is the absence of clear evidence of internal choanocytes, although the arrays of collared pits seen within the funnel render absence of choanocytes a moot point, as noted above. Even aside from this last issue, the characteristic complex displayed by this fossil does not resemble in detail any one of the modern crown group sponge classes. Thus, for example, hexactinellid sponges, but not demosponges, mount their choanocytes in syncytial reticular structures such as portrayed in Fig. 3D (36), whereas demosponges generate papillae, and the overall essentially simple, smooth-walled tubular structure of this organism also resembles simpler demosponges, such as asconoids. Phylogenetically, this animal is perhaps to be placed on the stem lineage of siliceous poriferans (hexactinellids plus demosponges). A formal description of this fossil, which is named Eocyathispongia qiania, is presented in SI Text (Systematic Paleontology). Due to its state of preservation, the multiplicity of its recognizable features, and the fortunate circumstance that these features extend to cell-type level characteristics, the informational content of this fossil is unique with respect to any putative sponge of the Ediacaran vintage that has been previously presented. Its age (600 Ma) torques the skeptical evidential landscape of Precambrian sponge evolution. Thus, the discovery of this adult form in the early Ediacaran Doushantuo strata could be said to fulfill the prediction, on the basis of uniquely poriferan cleavage stage embryo forms described earlier, that adult sponges must have existed in that environment (21). Similarly, this specimen provides an independent argument in evaluating the discovery of putative demosponge biomarkers of earlier as well as coeval age (12), and perhaps the same can be said of the paleontological claims of the Cryogenian sponge-like body fossils (9, 10). More to the point with which we began, the age of the elegantly structured poriferan described here provides independent support, for the first time to our knowledge, for the phylogenomic conclusion that the last common poriferan/eumetazoan ancestor must have existed much earlier, deep in the Cryogenian (1). However, important as they are, these conclusions rest on the single specimen described here, and they will require confirmation from additional Doushantuo fossils with the same characteristics. For example, discovery of additional specimens of Eocyathispongia would render inescapable the remote antiquity of the common ancestor of sponges and eumetazons. Recently, transcriptomes from eight species of sponge representing all four extant poriferan classes were analyzed by Riesgo et al. (4). Taken together with the genomes now available for two sponges [Amphimedon, a demosponge (5, 37) and Oscarella, a homoscleromorph sponge (37)], these data show overwhelmingly that the genetic heritage of all four classes is very similar in regard to several sets of metazoan genes of developmental importance. This similarity includes genes encoding the intracellular apparatus needed for Wnt, Hedgehog, and Tgfβ signal transduction; for epithelial junction and focal adhesion formation; and for innate immune response. This result is obviously consistent with the conclusion of poriferan monophyly (2, 3). Because the fossil sponge cannot be assigned uniquely to any of the extant poriferan classes but rather displays characters distributed among the modern sponge classes, it is to be regarded as a stem group descendant of the common ancestor of all sponges. It is of great interest that free-living choanoflagellates (of two different species) share very few of the genes that define the pan-metazoan gene set revealed in the transcriptome and genome sequence studies, whereas, on the other hand, in respect to this portion of their toolkit, the sponges lack only occasional members of the complete eumetazoan repertoire. The fossil stem group sponge described here must lie relatively far up in the phylogenetic tree that encompasses the four sponge classes and had to have shared this same pan-metazoan genomic repertoire. It is interesting to consider the implications of the 600-Myr-old fossil sponge from the vantage point of our own eumetazoan genomic heritage. Sponge genomes totally lack essential eumetazoan developmental regulatory apparatuses, such as hox genes, and yet share with us an enormously detailed pan-metazoan gene set. Therefore, this plesiomorphic genomic toolkit was being used in the last common ancestor of sponges and eumetazoans, whereas after the divergence of these lineages, additional derived eumetazoan-specific genomic regulatory apparatuses augmented the developmental capacities that would characterize the bilaterians and their sister groups. Thus, just as implied by the current temporal extrapolations of phylogenomics, the “calibration point” afforded by this fossil suggests that the shared metazoan genetic toolkit must have originated in the Cryogenian. Furthermore, if a relatively advanced sponge existed 600 Ma, then so did coeval animals of the eumetazoan lineage that also descended from the same last common poriferan/eumetazoan ancestor. Thus, it is a clear prediction that fossilized organisms of eumetazoan affinity from similarly deep in time await paleontological discovery, and some such may already have been seen in the Doushantuo animal microfossils cited above.

Materials and Methods The fossil specimen for this study was collected from the gray oolitic, phosphatized dolostone of the Upper Phosphate Member of the Doushantuo Formation at Weng’an phosphate mining area in Guizhou Province, southwest China (Fig. S1). We used acetic acid digestion to liberate the microfossils from the rock samples. All of the microfossils were examined by SEM, and well-preserved specimens were scanned at the European Synchrotron Radiation Facility (Grenoble, France) with high resolution. We used an undulator source, which can deliver a single harmonic X-ray with energy of 17.68 keV, at beamline ID19. The relative monochromaticity of the beam is so good that use of a monochromator was unnecessary. A CCD-based high-resolution detector with isotropic voxel sizes of 0.75 μm was used, and 1,800 projections over 180° were obtained for each scan. To get a phase-contrast effect, 10 mm was adopted as the propagation distance. In addition to the simple edge detection mode, we applied a single distance phase retrieval process for the fossils. This process permits retrieval of high-quality differential contrast very similar to the high-quality differential contrast achieved by holotomography but requiring far more simple acquisition and reconstruction protocols. Three-dimensional volume data processing was performed using VGstudio Max 2.1 software (Volume Graphics). The specimen described in this paper is housed at the Nanjing Institute of Geology and Paleontology, Chinese Academy of Sciences.

Acknowledgments We thank beamlines ID19, BM5, and ID22 of the European Synchrotron Radiation Facility for providing beam time. We thank Gang Li from the Institute of High Energy Physics, Chinese Academy of Sciences, for help in the synchrotron scanning experiments. This work was supported by funding from the Ministry of Science and Technology of China (Grant 2013CB835000), the National Natural Science Foundation of China (Grant 41302003), and the Chinese Academy of Sciences (Grant KZZD-EW-02-2). E.H.D. and D.J.B. were supported by US National Science Foundation Grant IOS1240626.

Footnotes Author contributions: M.Z. and E.H.D. designed research; Z.Y., M.Z., E.H.D., D.J.B., and F.Z. performed research; Z.Y. and P.T. conducted the synchrotron X-ray microtomographic experiments and analyzed volume data; and Z.Y., M.Z., E.H.D., and D.J.B. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1414577112/-/DCSupplemental.

Freely available online through the PNAS open access option.