Three-Dimensional Reconstructions of the Putative Metazoan Namapoikia Show That It Was a Microbial Construction Akshay Mehraa,B,C,1 , Wesley A

Home , Namapoikia, Nama Group

Three-dimensional reconstructions of the putative metazoan Namapoikia show that it was a microbial construction Akshay Mehraa,b,c,1 , Wesley A. Wattersd, John P. Grotzingere, and Adam C. Maloofa

aDepartment of Geosciences, Princeton University, Princeton, NJ 08544; bNeukom Institute, Dartmouth College, Hanover, NH 03755; cDepartment of Earth Sciences, Dartmouth College, Hanover, NH 03755; dDepartment of Astronomy, Wellesley College, Wellesley, MA 02481; and eDivision of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125

Edited by Andrew H. Knoll, Harvard University, Cambridge, MA, and approved June 23, 2020 (received for review May 12, 2020)

Strata from the Ediacaran Period (635 million to 538 million built wave-resistant frameworks (9, 10), revealed that aggregates years ago [Ma]) contain several examples of enigmatic, puta- comprise transported and deformed individuals (11), furthering tive shell-building metazoan fossils. These fossils may provide the idea that Cloudina produced weakly-to-non-biomineralized insight into the evolution and environmental impact of biomin- tubes. Sinotubulites, another tubular organism, also had plastic eralization on Earth, especially if their biological affinities and walls, but ones that were made up of a predominately organic modern analogs can be identified. Recently, apparent morpho- matrix (12). logical similarities with extant coralline demosponges have been Recently, researchers have proposed that Namapoikia, a used to assign a poriferan affinity to Namapoikia rietoogen- labyrinthine encrusting construction, produced skeletal mate- sis, a modular encrusting construction that is found growing rial by rapidly calcifying an organic scaffold (13). Studies of between (and on) microbial buildups in Namibia. Here, we polished, two-dimensional (2D) transverse and longitudinal sec- present three-dimensional reconstructions of Namapoikia that tions of Namapoikia suggest a complex interplay between the we use to assess the organism’s proposed affinity. Our mor- construction and surrounding microbial growths, with the two phological analyses, which comprise quantitative measurements life forms competing and, in some cases, repeatedly encrusting of thickness, spacing, and connectivity, reveal that Namapoikia over one another (13). Workers have proposed that Namapoikia produced approximately millimeter-thick meandering and branch- shares morphological characteristics with Chaetetid sponges and ing/merging sheets. We evaluate this reconstructed morphol- have inferred a biomineralizing pathway that is like that of the ogy in the context of poriferan biology and determine that extant demosponges Vaceletia and Acanthochaetetes (13, 14). On Namapoikia likely is not a sponge-grade organism. the basis of these similarities, Namapoikia has been assigned a poriferan affinity. 3D reconstruction | Ediacaran | early life Molecular clock and phylogenetic estimates (15) suggest that poriferans evolved during the Cryogenian Period (720 Ma to 635 Ma). Indeed, the Precambrian fossil record is replete with exam- n the Late Ediacaran (∼550 Ma), microbe-dominated reefs ples of purported sponge remains. Spicules, biomarkers, and Ibore witness to the arrival of putative biomineralizing meta- even full body fossils, all older than the onset of the Ediacaran zoans. By the Cambrian radiation [beginning 538.6 Ma to 538.8 Period, have been described—and debated—by researchers (for Ma (1)], a time period during which most modern animal phyla a complete review, see ref. 16; also see refs. 17 and 18 for more first emerged, skeletal reef dwellers were producing framework recent examples of debate). Additionally, by the early Cambrian, constructions and effectively engineering their surroundings (2). Today, biomineralizing organisms are responsible for building some of Earth’s largest organic constructions (e.g., the Great Significance Barrier Reef), which is indicative of the outsize impact that biomineralization has had on the planet’s sedimentological, Animals that build skeletons have an outsized impact on Earth’s biological, and geochemical makeup. biological, geochemical, and sedimentological cycles. To deter- To understand when, where, and why animals began to mine when, where, and why metazoan biomineralization first biomineralize, as well as to determine the environmental, eco- emerged, it is necessary to study the earliest record of skele- logical, and evolutionary ramifications associated with the first tal animals. This record is made up of four genera from the biomineralizers, it is necessary to study the earliest skeletal Ediacaran period: Namacalathus, Cloudina, Sinotubulites, and metazoan fossil record. This record comprises four genera from Namapoikia. Here, we measure three-dimensional reconstruc- Ediacaran shallow water settings: Namacalathus, Cloudina, Sino- tions of Namapoikia to test the hypothesis that it is a cal- tubulites, and Namapoikia (3). Although morphologically simple, Namapoikia these organisms have proven to be enigmatic, and their growth cifying sponge. We find that lacks the physical habits, biological affinities, and environmental impacts are the characteristics expected of a sponge, or, for that matter, an subject of ongoing debate. animal. With respect to early biomineralization, modes of shell build- Author contributions: A.M., W.A.W., J.P.G., and A.C.M. designed research; A.M. per- ing appear to have varied among the Ediacaran putative biomin- formed research; A.M. and A.C.M. analyzed data; and A.M. and A.C.M. wrote the eralizers. Exactly how, and to what degree, each organism made paper.y hard parts remains unresolved. Workers have suggested that The authors declare no competing interest.y Namacalathus, a flexible, goblet-shaped organism, produced a This article is a PNAS Direct Submission.y foliated calcitic ultrastructure (4). Conversely, Namacalathus also has been shown to have been lightly calcified (5, 6). Cloud- Published under the PNAS license.y ina, a tubular organism made up of a “cup in cup” morphology, Data deposition: The computational source code used in this paper is available in GitHub was thought to have precipitated carbonate on an organic matrix at https://github.com/giriprinceton/namapoikia.y (5, 7). More recent work, however, has demonstrated that phos- See online for related content such as Commentaries.y phatized Cloudina share a nanoparticulate fabric with extant 1 To whom correspondence may be addressed. Email: [email protected] biomineralizers, suggesting that the organism formed skeletons This article contains supporting information online at https://www.pnas.org/lookup/suppl/ in the same way as modern animals (8). That said, reconstruc- doi:10.1073/pnas.2009129117/-/DCSupplemental.y tions of Cloudina, made to test the assertion that the organism First published August 3, 2020.

19760–19766 | PNAS | August 18, 2020 | vol. 117 | no. 33 www.pnas.org/cgi/doi/10.1073/pnas.2009129117 Downloaded at California Institute of Technology on August 19, 2020 Downloaded at California Institute of Technology on August 19, 2020 er tal. et Mehra fteEryCmra,myhv rteovdteaiiyto ability whether the determine this test to evolved To seek first Ediacaran. Late we have the idea, during may engineers skeletons Cambrian, calcified dominant build Early the the during the becoming appeared Archaeocyathids, of eventually having (2). poriferans, the and reefs that framework Cryogenian reason namely first to world’s stands sponges, the It for some responsible even and were present were Earth, affinity poriferan on definitive of organisms calcifying ok r ato h ubsSbru OkkMme)o the of Member) (Omkyk Subgroup Kuibis the of 1 part Fig. are Namibia 7358951N; rocks Reitoog, 667792E 33K near UTM Farm (WGS84 Driedoornvlakte on of outcropping examples situ Reconstructions In and Observations, the Field Setting, assess Geologic quantitatively to us of combined paleoecology reconstructions, enable and 3D affinity observations, surrounding resulting the field The and 19). with differentiate fossils) to imag- 11, (e.g., texture (6, and and interest matrix grinding color of on serial features relies utilize between which we method a problem, ing, this address techniques. To density-sensitive dis- traditional, acid with imaging via outcrop) or isolation solution precluding rock, in carbonate within analysis. planes minerals (11), accurate error for bedding required measurement Unfortunately, on are and data or (3D) misinterpretation three-dimensional analogs, to (e.g., slabs possible subject measurements polished structures, are identify 2D on of Since habit, affinity. distribution made growth biological and describe evaluate shape, and to size, order the in of as such characteristics mens, morphological analyze to gross necessary is it alter- the result, diagenetic a of As signs recrystallization). (e.g., exhibit ation and preservation tissue soft lack organism. sponge-grade C A iemn te daaa osl,seiesof specimens fossils, Ediacaran other many Like Namapoikia Namapoikia B kltn r rsre scarbonate as preserved are skeletons Namapoikia. r on napnal reef pinnacle a in found are Namapoikia Namapoikia A and s nfc,a fact, in is, Namapoikia .These B). D E speci- a o endsrbdi te,cneprnosrcsin rocks contemporaneous Earth. on other, else anywhere in matter, that knowledge, described for our or, Namibia been of best not the has To Driedoornvlakte 1C). at (Fig. reef locations pinnacle the that of tabulae reveals survey perpendicular by detailed the intersected A be to par- (13). to reference specimens, said certain been visual In have work). a titions this for and in used 2 transverse conventions Fig. in spatial (see merge directions and split longitudinal as elements to These referred the here “partitions.” elements, to structural 23). millimeter-thick opened (13, spaced, that tall m reef 0.25 the in fractures 1 shallow seafloor. (Fig. are dykes neptunian which decimeter-wide grainstones. in clinoformal buildups in and buildups Namapoikia microbial thrombo- between massive fill or columnar Both columnar of and combination lites. a stromatolites com- of encrusting pinnacles up or These made mounds (21). buildups microbial margins pinnacle prise platform Member, the Urikos the on of prior of developed course just shales stage, by the final drowning the over to in created cycles; accommodation was distinct ramp three 40 carbonate to The 25 quartzite. dips and long, 545.41 of age deposition minimum a subgroup provides the Schwarzrand (22) overlying constrain the in date Kuibis zircon uranium–lead the 548.8 to from direct age depositional ages no maximum are zircon How- stratigraphy. There uranium–lead Driedoornvlakte 21). ever, the (20, from orogens dates Gariep radiometric and along Damara convergence with the coincident subbasin Zaris ramp northern carbonate the a in on formed They Group. Nama Neoproterozoic Individual km 10 thick, m 500 is which Farm, Driedoornvlakte at reef The Cloudina sfudecutn h al n oso microbial of tops and walls the encrusting found is Namapoikia Namapoikia PNAS eeec oWS4UM33K. UTM WGS84 to reference in coordinates Map Namapoikia. xiiigafaue(rsneo etr denoted (D feature of black). (presence locations Pie in observed feature (Bottom) of a (11). percent exhibiting the model depicting surface charts imagery- digital drone a derived from generated were contours m; of occurrence Namapoikia the showing data ( C survey rectangle. Interpolated yellow the the by with marked area Farm, study Driedoornvlakte at complex reef the am hl h ih ryfildpcstegeo- the (B depicts rocks. fill Group Nama gray of light extent the graphic Driedoornvlakte is while square red Farm, The Namibia. in area 1. Fig. Namapoikia and | ◦ uut1,2020 18, August otesuhat iso Precambrian on sits southeast, the to srr,ocrigin occurring rare, is oaino h study the of Location (A) observations. Field a eu o1mwd n pto up and wide m 1 to up be can Namacalathus ntere.Tecnoritra s5 is interval contour The reef. the in and pcmn opieregularly comprise specimens ± E il htgah fi situ in of photographs Field ) a(0.Adtoal,a Additionally, (20). Ma 1 | o.117 vol. a efudi the in found be can 3% | B ± and o 33 no. Namapoikia fobserved of Ma. 1 D C iwof View ) and r with are | (Top) ) 19761 E),

EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES We reconstruct and measure two specimens of Namapoikia Kolmogorov–Smirnov test of partition thicknesses rejects the (referred to as sample A and sample B) from the pinnacles null hypotheses that the populations are from a single distri- at Driedoornvlakte Farm. Both samples comprise white, cal- bution (P < 0.001, D = 0.19; D refers to the magnitude of the cified partitions surrounded by a matrix of fine-grained, black Kolmogorov–Smirnov statistic). The same test on interpartition micrite fill (Fig. 2 A and B). The fill contains no evident syn- thicknesses also rejects the null hypothesis (P < 0.001, D = 0.41). depositional sedimentary structures. Blocky calcite spar occurs Notably, both samples A and B contain large, irregularly dis- throughout both samples and appears to replace fill between tributed voids. To test whether the presence of these voids the partitions (Fig. 2 A and B). Sample A includes a small contributes to the differences in partition and interpartition amount of yellow dolomite (approximately 1.6% by volume, thicknesses between the two samples, representative sub- occurring predominately along fracture planes). Sample B is volumes—comprising regularly spaced partitions and referred to bounded by microbial textures (Fig. 2B). A petrographic thin sec- as subvolume A and subvolume B—are chosen and measured tion of sample A reveals that the fossil and matrix phases are (Fig. 3D). Within these selected regions, disparities between the recrystallized. two samples persist. In subvolume A, partitions have a thickness When reconstructed, partitions meander, branch, and merge of 430/856/1,204 µm, and interpartition voids have a thick- in transverse and longitudinal sections with no evidence of tabu- ness of 1,285/1,811/2,377 µm, while, in subvolume B, partitions lae (Fig. 2 C and D) (SI Appendix, Fig. S1F and Movies S1 and have a thickness of 566/781/1,020 µm, and interpartition voids S2). Partition thickness and spacing vary between the two sam- have a thickness of 671/1,005/1,435. A two-sample Kolmogorov– ples. In sample A, partitions have a thickness of 640/1,030/1,393 Smirnov test of partition thicknesses rejects the null hypothesis µm (25th/50th/75th percentiles; this convention is used through that the populations are from the same distribution (P < 0.001, the remainder of the text), while the interpartition voids have a D = 0.14), as does a test of interpartition thicknesses (P < 0.001, thickness of 1,559/2,402/3,547 µm (Fig. 3B). In sample B, parti- D = 0.44). tions have a thickness of 539/831/1,122 µm, and the interpartition The median thickness of partitions varies with respect to voids have a thickness of 735/1,221/1,881 µm. A two-sample height in each sample (Fig. 3E). When applying a best-fit line

A B

Fig. 2. Reconstructions of Namapoikia samples. (A) Single slice of Namapoikia sample A, processed using GIRI. The circle marker labeled 1 shows blocky calcite that is distributed throughout the sample, while the circle marker labeled 2 is pointing to dolomite ﬁlling a fracture within the rock. (B) Single slice of Namapoikia sample B, processed using a manual serial grinding and imaging procedure (after ref. 6). As the Namapoikia specimen is bounded by thrombolite fabrics (denoted using a yellow dotted line), a subregion, marked by the outlined red square, is used for morphological analyses. The circle marker labeled 1 denotes an example of blocky calcite. (C) Rendering of the sample A reconstruction. (Inset) Diagram illustrating the terms transverse, longitudinal, and latitudinal. (D) Rendering of the sample B reconstruction. (Scale bar at the bottom left of each panel, 0.5 cm.) The direction of stratigraphic up is denoted by the red arrow on the axis ﬁgure above the scale bar in each panel.

19762 | www.pnas.org/cgi/doi/10.1073/pnas.2009129117 Mehra et al. Downloaded at California Institute of Technology on August 19, 2020 Downloaded at California Institute of Technology on August 19, 2020 ,teaslt hneo hs rnsi ml rela- small is determination al. trends et Mehra these of B. of coefficients change sample low absolute in the and exhibiting up 3E), A to with (Fig. sample stratigraphic trends, addition in toward different In up exhibit thickening stratigraphic specimens partitions toward two thinning the excluding after partitions data, even similar the are Trends to outliers. trends excludes best-fit fit two the sample, blue, each light For in growth. while, of data, direction samples. all the both account be in into to height) interpreted takes sampled r is fit the also the only which black, so height, in between (E outliers, increasing lines: spacing of text). dashed direction to as the the equivalent shown in in considered are described is be up (as can Stratigraphic in partitions voids B. shown interpartition spaced and region (B of regularly values. the thickness their median The depicts for respectively. depict area selected B, dots highlighted and white blue A the The samples while partitions. in specimen, of voids single case interpartition a the and from In partitions data specimens. measured multiple of from bounds sometimes percentile respectively), 40, For and study. 31, this in discussed sponges calcified 3. Fig. esrmnsmd nrcntutosof reconstructions on made Measurements 2 ftedte lc iei eoe.(Inset denoted. is line black dotted the of .crypta, V. Namapoikia and seunesi, A. pcmn.( specimens. o ltsoigterneo einpriintikess(.. scluae teach at calculated as (i.e., thicknesses partition median of range the showing plot box A ) oprsno attosaditratto odtikessfo subvolumes from thicknesses void interpartition and partitions of comparison A (D) A. asrpeetterneo ausa eotdi ieaue(es 14, (refs. literature in reported as values of range the represent bars discoforma, G. einpriintikes ensbrce,vru egti ohsmlsA samples both in height versus subtracted, mean thickness, partition Median ) h ieo aiu kltleeet advis rmfu ifrn hyper- different four from voids) (and elements skeletal various of size The A) .perforata A. iey,wie nsml ,tecag s131 is change the B, 117 respec- sample 13%). thicknesses, is in and interpartition change while, and the partition tively), median A, interpartition of and sample 6% thickness partition In median spacing. sample’s a to tive i.S2 Fig. Appendix, (SI PNAS | and uut1,2020 18, August A C and itgasilsrtn h hcns of thickness the illustrating Histograms ) ,br ersn h 5had75th and 25th the represent bars B), | o.117 vol. o 1 and 11% (or µm | o 16% (or µm o 33 no. | 19763

EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES Discussion irregular voids (Fig. 3 B–D). In contrast, sponges (and indeed, Researchers increasingly believe sponges to be a monophyletic all metazoa) produce regularly spaced and sized structures. group that—in terms of evolution and phylogeny—should be For example, measurements of voids in V. Crypta and Acan- placed at or near the base of the metazoan tree (24). This thochaetetes seunesi (another proposed Namapoikia analog), group diverged in the Cryogenian, and the last common ancestor both made on single specimens, show a total range of only 300 (LCA) of modern sponges likely was thin walled, with a single µm (∼11% of the variation seen in Namapoikia; Fig. 3A). Taken layer of spicules (24) (this prediction is debated, however; see together, our observations of scale suggest that Namapoikia ref. 25 for a counterpoint). From a morphological standpoint, likely created structures too large to be choanocyte bearing (and Namapoikia is unlike the proposed LCA or even the thin-walled therefore capable of actively pumping and filter feeding) and too sponges of the early Cambrian [e.g., the Archaeocyathids (18)]. variable to be a regular metazoan construction. Namapoikia does bear a passing resemblance to hypercalcified Given our reconstructions, which lack the regularity expected sponges in the rock record, such as Vaceletia crypta and various of sponges or, more generally, animals, we suggest that Inozoa (SI Appendix, Figs. S1 and S2). However, 3D reconstruc- Namapoikia was not a metazoan. Namapoikia’s morphological tions of Namapoikia, with its sheet-like partitions and lack of expression, which can be summarized as widely spaced, mean- tabulae or chambers (SI Appendix, Fig. S1 C–F), make it clear dering partitions that split and merge both transversely and that any apparent morphological similarities are superficial (see longitudinally (SI Appendix, Fig. S1F), likely lacked the struc- SI Appendix for a quantitative comparative analysis). tural integrity to stay upright without external support (i.e., much Given the lack of morphological similarities between like trying to stand playing cards up on their sides). As a result, Namapoikia and other described sponges, it may be argued Namapoikia probably had low emergent synoptic relief (Fig. 4 that the fossil represents a stem group poriferan. While exact illustrates this proposed characteristic). We suggest that such poriferan synapmorphies are debated, traits unique to Porifera a morphological expression can be explained by the growth of include a branching aquiferous system that moves water from partially or totally microbially mediated structures. pores known as ostia, through choanocyte-bearing chambers, Microbially mediated sedimentary constructions, including and out one or more osculii (26). Specimens of Namapoikia lack the stromatolites and thrombolites that make up the reef at any remnants of a clear aquiferous system, ostia, an osculum, Driedoornvlakte Farm, are the result of incremental growth, and/or any spongin, tissues, or fibers. As a result, a poriferan aggregation, and calcification. The morphologies of such bio- assignment for Namapoikia is doubtful. constructions partly are controlled by environmental conditions, An additional challenge to a poriferan affinity for Namapoikia including water depth, light levels, and sediment flux (32). In comes from the scale of the partitions and voids in observed the case of Namapoikia, the observed intersample differences specimens. While the skeletons of calcareous sponges are not in partition and interpartition thicknesses could be explained necessarily canal systems, their dimensions are controlled by by variations in local environmental conditions through time the scale of living tissue. Both the size and spacing of skeletal and space on the reef. Consistent with our hypothesis is the elements may be impacted by diagenetic processes (e.g., thicken- observation that morphological expressions of microbial con- ing of calcareous elements at the expense of interpartition void structions often are regional (33). This property would account space), the effects of which can be difficult to determine in recon- for why Namapoikia is not found cooccurring with Cloud- structions. That said, the combined thickness of partition and ina and Namacalathus assemblages on other paleocontinents interpartition void in Namapoikia—a metric which negates the (5, 34, 35). effects of postmortem diagenesis—speaks to a structure that is Both stromatolites—comprising fine laminations—and throm- anomalously large when compared to other poriferans. bolites—made up of clotted fabrics—can produce branching Sponges typically produce small-diameter, high-density canal forms. Certain thrombolites, such as Favosamaceria cooperi (33), systems to deal with the diffusion processes required for gas also create calcified vertical curtains (referred to as “mace- exchange and nutrient capture (27). In order to effectively pump riae” for their likeness to the walls of garden mazes; see SI water (and overcome resistance/frictional losses), small, densely Appendix, Fig. S3 and Movie S4 for our reconstruction). It fol- populated choanocyte chambers are thought to function as peri- lows that a microbially mediated construction could produce staltic pumps within sponges (28, 29) (see also ref. 30). Structures the morphologies expressed by Namapoikia. Fabrics in throm- built by calcifying sponges generally have features that are sub- bolites characteristically are millimeter-to-centimeter thick and millimeter to a millimeter thick. For example, in the recent much less regular in size than metazoans. Thrombolites also hypercalcified demosponge V. crypta, which has been proposed to contain large, irregularly distributed voids like Namapoikia share morphological and biological properties with Namapoikia (36). The scale and variance of partitions and interpartition (13), walls and void spaces have a combined thickness of 650 µm thicknesses, especially within individual Namapoikia specimens, to 950 µm (with the walls being 50 µm thick and tissue-bearing are consistent with these attributes of thrombolites. chambers ranging from 600 µm to 900 µm in diameter; Fig. 3A There is, to the best of our knowledge, no other exactly equiv- and ref. 14). Gigantospongia discoforma, an exceptionally large alent microbially mediated structure to reconstruct and compare hypercalcified sponge (31), has a combined skeletal and canal to Namapoikia. Given the influence that environment has on thickness ranging from 1,000 µm to 2,200 µm (with walls between morphological expression of microbial construction, this lack of 300 µm and 1,000 µm thick and tubular canals between 700 µm formal twin is not surprising. and 1,200 µm in diameter; Fig. 3A). The fractures in which Namapoikia is found provided a unique In our reconstructions, Namapoikia partition thicknesses ecological niche in which Namapoikia could grow with a distinct range from 200 µm to 1,650 µm (5th to 95th percentiles for both morphological expression (Fig. 4). In our model, Namapoikia samples combined), and, even when only examining subvolumes would, at any given moment in time, appear as a series of pro- that exclude large voids, Namapoikia interpartition thicknesses truding ridges, which would respond to changing conditions (i.e., range from 200 µm to 2,905 µm (5th to 95th percentiles for both light, nutrient, and/or sediment flux) via migration and branch- samples combined). The combined partition and interpartition ing (Fig. 4 B, i–iii). Namapoikia would exhibit low synoptic relief, thickness of Namapoikia (considering all data: 483 µm to 6,425 and the space between partitions would be filled with baffling µm; considering subvolumes only: 300 µm to 4,452 µm) is up to cement and/or sediment (Fig. 4), so as to provide support to 2.9 times as large as the combined skeletal and canal thickness of microbial structures that otherwise would deform or topple eas- G. discoforma, one of the largest known sponges. ily. Thus, the character of longitudinal and transverse branching In addition to producing large and widely spaced structures, is indicative of the shape evolution of microbial ridges (Fig. 4B) Namapoikia samples also exhibit significant variance in both and not the result of coalescing metazoan skeletal walls. Simply partition and interpartition void thicknesses that cannot be put, the final expression of Namapoikia represents a collection of explained by selective calcification, even when controlling for multiple, incremental events through time.

19764 | www.pnas.org/cgi/doi/10.1073/pnas.2009129117 Mehra et al. Downloaded at California Institute of Technology on August 19, 2020 Downloaded at California Institute of Technology on August 19, 2020 i.4. Fig. eerdt sAadB eecletda reorvat amfr3D for Farm Driedoornvlakte al. at et Mehra collected were B, and A as Grinding. to Serial referred and Collection Sample in of resulted distributions which spatial kriging, 0.14, the indicator = of perform sill map to 127.69, used a = was best-fit range 0.10) = with The nugget exponential dataset. and an (culled) case, empirical this resulting (in the semivariogram to fit were variograms of absence or of evidence presence the for examined were accuracy vertical mean m). a 0.256 and = m) (SD 0.126 station, m = 0.706 base (SD of mean m Springbok a 0.596 have TrigNet of data accuracy corrected the The horizontal Farm. from Driedoornvlakte from data km GPS 654 with located the made using cor- were Differential corrected University. rections Princeton differentially at were package characteristics— software data Office physical Pathfinder GPS keywords—corresponding and/or The discrete recorded. lithological, of were set sedimentological, a visible point, 10 to each with grid At orthogonal collected. an total on were a arrayed antenna), points, external GPS hand- the 1,254 a (excluding of Using unit conducted. GPS was GeoXH6000 Farm Trimble Driedoornvlakte held at reef pinnacle the of Data. Survey Methods and Materials func- observations and 2D form using extracted organismal be alone. confidently basic cannot about that sugges- tion insights the test- provide (e.g., (e.g., morphology data functional identification that about in tion chambers; clues use and/or give for and tabulae structures for presence of ing the elucidate absence both measure- can or Quantitative reconstructions mor- on required. on made are based ments observations fossils 3D problematic of phology, affinity organismal assign vertical. perfectly be not may partitions sheet-like that such cross-section, longitudinal in drift to lead partitions (iii of meanders and migrations Transverse red. (i (ii illustrated. relief. partitions synoptic several only emergent with growth, incremental of model Namapoikia attoshv ohmre n pl.Telcto ftemre smre nyellow. in marked is merger the of location The spilt. and merged both have Partitions ) et h orce aaaogwt soitdfil observations— field associated with data—along corrected the Next, of analyses Our igasilsrtn h rpsdmdlfor model proposed the illustrating Diagrams Namapoikia Namapoikia A (B) reef. the up made that stromatolites encrusting and thrombolites the with interacting fissures, syndepostional in incrementally grew omptesailetn fvrosfce,adtie survey detailed a facies, various of extent spatial the map To Namapoikia eefitrdot n hnmlil oee semi- modeled multiple then and out, filtered were rwwt o yotcrle) h 3D The relief). synoptic low with grew eodmmn ntm,weeapriinhsgonuwr n pi n irtd h oain fslt r akdin marked are splits of locations The migrated. and split and upward grown has partition a where time, in moment second A ) Namapoikia. eosrt ht nodrto order in that, demonstrate Two i.S1 Fig. Appendix, SI baigsamples, Namapoikia-bearing onswt no with Points Namapoikia. atosdpcigtepeec of presence the depicting Cartoons (A) Namapoikia. × 0mspacing, m 20 C–F lutaino attosadvi l tsm ie0 ihtepriin aiglow having partitions the with 0, time some at fill void and partitions of Illustration ) ) eue sfrsml wt tpsz f30 of size step a (with the pro- A of imaging sample and grinding for same as the cedures at following GIRI, redeposited with and processed The History. was preserved Natural also was of sample Museum National the Smithsonian of the mm 1 approximately 20 and (i.e., size step A, smaller sample a 30 for being as difference procedures only imaging the and with grinding same the an following GIRI, and The reconstruction. (37)] 3D for Senowbari-Daryan selected and Rigby of by resolution recorded per-pixel were a data to and (corresponding scanner, resolution inch flatbed 42.33 per EPSON removing dots an 600 After on at procedure. down) imaging side ished and grinding serial 100 manual a grinding the using repeated then took 5) times. 3) 658 and of coolant, quality, 5.73 total excess image a of any 30 process evaluated off resolution away 4) wiped a ground 2) image, at 1) surface, an programmatically sample ratio the GIRI reproduction from A, 1:1 material verti- sample a positioned For attain 120-mm of pixel. is to a per system up as with imaging made so equipped This is back cally lens. with stage digital macro retrofitted imaging IQ180 Kreuznach been One Schneider The Phase stages. has 80-megapixel com- imaging that an a and grinder comprises wiping, GIRI surface misting, (11). control University numerical Princeton puter at (GIRI) Instrument tion 41.0 B: imaged and and A 126.9 sectioned samples serially for was dimensions ground then (final sample Each adhesive. epoxy using multiple on markings arrow recorded with was (i.e., denoted faces. orientation was location field up) rock’s its direction and the plane unit, bedding GPS A, GeoXH6000 sample Trimble handheld of a with case the In reconstruction. o oprtv nlss an analysis, comparative For (MIT) Technology of Institute Massachusetts at processed was B Sample Reconstruc- and Imaging, Grinding, the using processed was A Sample plates steel to on mounted then and slabbed were both B and A Samples ) oa f164iae fthe of images 1,624 of total A µm). fmtra ihasraegidr h apewspae (pol- placed was sample the grinder, surface a with material of µm ) hsmto a eetd39times. 319 repeated was method This µm). × .cooperi F. 116.8 × 94m,respectively). mm, 29.4 apewr collected. were sample PNAS Namapoikia | uut1,2020 18, August .perforata A. .perforata A. .perforata A. ntere tDidonlkeFarm. Driedoornvlakte at reef the in pcmn[rgnlycollected [originally specimen | ) oa f31images 391 of total A µm). o.117 vol. apewspoesdwith processed was sample .cooperi F. pcmnwr collected, were specimen × 57.9 sopsdto opposed as µm | .cooperi F. × o 33 no. pcmnwere specimen 97m and mm 19.7 | sample of µm 19765 µm

EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES Image Processing. GIRI outputs data in a proprietary raw image file for- gradient descent with momentum) with the intent of improving network mat (.IIQ), which must be converted to 16-bit RGB TIFF files before further accuracy. processing. With the exception of applying the same white balance value Following training, images of sample A, sample B, A. perforata, and F. to all images, the raw data were not adjusted before conversion. Sample cooperi were run through their respective neural network to produce prob- B images, which were created as 8-bit JPEG files, required no additional ability maps, which then were thresholded to create classified TIFFs. These conversion before processing. TIFF files then were loaded into Avizo, a software package designed for visu- Prior to 3D visualization and analysis, images were segmented into dis- alization and analysis of volumetric datasets. In particular, thickness values tinct classes (e.g., matrix, calcified elements, blocky calcite, and dolomite). were generated using the Thickness Map module within Avizo. The mod- Two different neural networks were leveraged for this classification task. In ule, which implements local thickness as defined by ref. 39, calculates, at the case of sample A, a hidden layer neural network—operating on super- each volumetric pixel (or voxel), the diameter of the largest sphere that pixels, or pixel clusters made on the basis of color and texture (38)—was both is contained in the object (i.e., partition or void) and includes that used. For sample B, A. perforata, and F. cooperi, a convolutional neural voxel. network was applied. All raw image data are available upon request. The computational source In all instances, the neural network had to be trained prior to classifi- code used to process data in this paper is located in a public repository at cation. First, a number of representative images—three for sample A, A. https://github.com/giriprinceton/namapoikia. perforata, and F. cooperi and, due to variations in image quality throughout the grinding process, five for sample B—were selected. Next, training ACKNOWLEDGMENTS. We thank C. Husselmann for granting us access to data were compiled using a series of custom scripts written in Matlab. For Driedoornvlakte Farm. At the Geological Survey of Namibia, G. Schneider and J. Eiseb granted us permits for working in Namibia and assisted us with sample A, superpixels were calculated for each image, after which a user the export permitting process, respectively. At MIT, B. Ren performed grind- selected and assigned superpixels to one of a set of predefined classes via a ing and imaging, and T. Mason created an early reconstruction, of sample graphical user interface. Upon completion, a collection of statistics (i.e., the B. All of Situ Studio, but especially B. Samuels, were instrumental in the mean, SD, and covariance of Red, Green, and Blue channel values, as well as development of GIRI. A. Tasistro Hart and R. Bartolucci provided invaluable an entropy term) for each chosen superpixel were calculated and stored in assistance in the field. At the Smithsonian, M. Florence dedicated his time to a data structure. In the case of sample B, A. perforata, and F. cooperi, pixels help locate Inozoan specimens for study, while D. Erwin kindly gave permis- were painted by a user, thereby marking them as belonging to a given class. sion to destructively analyze the A. perforata sample. R. Shapiro generously For each painted pixel, a square neighborhood (11 × 11 for sample B and provided us with a sample of F. cooperi for reconstruction. Our morphological analyses benefited from discussions with A. Getraer, B. Howes, A. perforata and 33 × 33 for F. cooperi) was extracted and then stored as R. Manzuk, and E. Geyman. We thank J. Strauss for feedback on the a TIFF in a directory corresponding to its assigned class. For both networks, manuscript and A. Knoll and two anonymous reviewers for their thought- training was accomplished by 1) initializing each neuron within the net- ful critique and input. This work was supported by NSF Earth Sciences Grant work with a random weight, 2) running training data through the network 1028768 to A. Maloof and by funding from the Princeton Tuttle Invertebrate to produce a prediction, and 3) updating neuron weights (via stochastic Fund.

1. Ulf. Linnemann et al., New high-resolution age data from the Ediacaran–Cambrian 21. E. W. Adams, S. Schroder,¨ J. P. Grotzinger, D. S. McCormick, Digital reconstruction boundary indicate rapid, ecologically driven onset of the Cambrian explosion. Terra. and stratigraphic evolution of a microbial-dominated, isolated carbonate plat- Nova 31, 49–58 (2019). form (terminal Proterozoic, Nama Group, Namibia). J. Sediment. Res. 74, 479–497 2. S. M. Rowland, R. A. Gangloff, Structure and paleoecology of lower Cambrian reefs. (2004). Palaios 3, 111–135 (1988). 22. J. P. Grotzinger, S. A. Bowring, B. Z. Saylor, A. J. Kaufman, Biostratigraphic 3. Y. Cai, S. Xiao, G. Li, H. Hong, Diverse biomineralizing animals in the terminal and geochronologic constraints on early animal evolution. Science270, 598–604 Ediacaran Period herald the Cambrian explosion. Geology 47, 380–384 (2019). (1995). 4. A. Y. Zhuravlev, R. A. Wood, A. M. Penny, Ediacaran skeletal metazoan interpreted as 23. R. A. Wood, J. P. Grotzinger, J. A. D. Dickson, Proterozoic modular biomineralized a lophophorate. Proc. Biol. Sci. 282, 20151860 (1818). metazoan from the Nama Group, Namibia. Science 296, 2383–2386 (2002). 5. J. P. Grotzinger, W. A. Watters, A. H. Knoll, Calcified metazoans in thrombolite- 24. J. P. Botting, L. A. Muir, Early sponge evolution: A review and phylogenetic stromatolite reefs of the terminal Proterozoic Nama Group, Namibia. Paleobiology framework. Palaeoworld 27, 1–29 (2018). 26, 334–359 (2000). 25. C. Luo, F. Zhao, Z. Han, The first report of a vauxiid sponge from the Cambrian 6. W. A. Watters, J. P. Grotzinger, Digital reconstruction of calcified early metazoans, Chengjiang Biota. J. Paleontol. 94, 28–33 (2020). terminal Proterozoic Nama Group, Namibia. Paleobiology 27, 159–171 (2001). 26. C. W. Dunn, S. P. Leys, S. H. D. Haddock, The hidden biology of sponges and 7. S. W. Grant, Shell structure and distribution of Cloudina, a potential index fossil for ctenophores. Trends Ecol. Evol. 30, 282–291 (2015). the terminal Proterozoic. Am. J. Sci. 290, 261–294 (1989). 27. J. U. Hammel et al., The non-hierarchical, non-uniformly branching topology of a 8. P. U. P. A. Gilbert et al., Biomineralization by particle attachment in early animals. leuconoid sponge aquiferous system revealed by 3D reconstruction and morpho- Proc. Natl. Acad. Sci. U.S.A. 116, 17659–17665 (2019). metrics using corrosion casting and X-ray microtomography. Acta Zool. 93, 160–170 9. A. M. Penny et al., Ediacaran metazoan reefs from the Nama Group, Namibia. Science (2012). 344, 1504–1506 (2014). 28. P. S. Larsen, H. U. Riisgad,˚ The sponge pump. J. Theor. Biol. 168, 53–63 (1994). 10. A. Shore, R. Wood, A. Curtis, F. Bowyer, Multiple branching and attachment structures 29. H. U. Riisgard,˚ P. S. Larsen, Comparative ecophysiology of active zoobenthic filter in cloudinomorphs, Nama Group, Namibia. Geology, 10.1130/G47447.1 (2020). feeding, essence of current knowledge. J. Sea Res. 44, 169–193 (2000). 11. A. Mehra, A. C. Maloof, Multiscale approach reveals that Cloudina aggregates are 30. S. P. Leys et al., The sponge pump: The role of current induced flow in the design of detritus and not in situ reef constructions. Proc. Natl. Acad. Sci. U.S.A. 115, E2519– the sponge body plan. PloS One 6, e27787 (2011). E2527 (2018). 31. J. Keith Rigby, B. Senowbari-Daryan, Gigantospongia, new genus, the largest known 12. Z. Chen, S. Bengtson, C. M. Zhou, H. Hong, Y. Zhao, Tube structure and original Permian sponge, Capitan limestone, Guadalupe Mountains, New Mexico. J. Paleontol. composition of Sinotubulites: Shelly fossils from the late neoproterozoic in southern 70, 347–355 (1996). Shaanxi, China. Lethaia 41, 37–45 (2008). 32. T. Bosak, A. H. Knoll, A. P. Petroff, The meaning of stromatolites. Annu. Rev. Earth 13. R. Wood, A. Penny, Substrate growth dynamics and biomineralization of an Ediacaran Planet Sci. 41, 21–44 (2013). encrusting poriferan. Proc. Biol. Sci. 285, 20171938 (2018). 33. R. S. Shapiro, S. M. Awramik, Favosamaceria cooperi new group and form: A widely 14. J. Vacelet, “Recent “Sphinctozoa”, order Verticillitida, family Verticillitidae Stein- dispersed, time-restricted thrombolite. J. Paleontol. 80, 411–422 (2006). mann, 1882” in Systema Porifera, J. N. A. Hooper, R. W. M. Van Soest, Eds. (Springer, 34. H. J. Hofmann, E. W. Mountjoy, Namacalathus-Cloudina assemblage in Neoprotero- 2002), pp. 1097–1098. zoic Miette Group (Byng Formation), British Columbia: Canada’s oldest shelly fossils. 15. E. A. Sperling, J. M. Robinson, D. Pisani, K. J. Peterson, Where’s the glass? Biomarkers, Geology 29, 1091–1094 (2001). molecular clocks, and microRNAs suggest a 200-myr missing Precambrian fossil record 35. J. E. Amthor et al., Extinction of Cloudina and Namacalathus at the Precambrian- of siliceous sponge spicules. Geobiology 8, 24–36 (2010). Cambrian boundary in Oman. Geology 31, 431–434 (2003). 16. J. B. Antcliffe, R. H. T. Callow, M. D. Brasier, Giving the early fossil record of sponges 36. L. C. Kah, J. P. Grotzinger, Early Proterozoic (1.9 Ga) thrombolites of the Rocknest a squeeze. Biol. Rev. 89, 972–1004 (2014). Formation, Northwest Territories, Canada. Palaios 7, 305–315 (1992). 17. B. J. Nettersheim et al., Putative sponge biomarkers in unicellular Rhizaria question 37. J. Keith Rigby, B. Senowbari-Daryan, Upper Permian Inozoid, Demospongid, and an early rise of animals. Nature Ecol. Evol. 3, 577–581 (2019). Hexactinellid Sponges from Djebel Tebaga Tunisia (University of Kansas, 1995). 18. J. P. Botting, B. J. Nettersheim, Searching for sponge origins. Nature Ecol. Evol. 2, 38. R. Achanta et al., SLIC superpixels compared to state-of-the-art superpixel methods. 1685–1686 (2018). IEEE Trans. Pattern Anal. Mach. Intell. 34, 2274–2282 (2012). 19. A. C. Maloof et al., Possible animal-body fossils in pre-Marinoan limestones from 39. T. Hildebrand, P. Ruegsegger,¨ A new method for the model-independent assessment South Australia. Nat. Geosci. 3, 653–659 (2010). of thickness in three-dimensional images. J. Microsc. 185, 67–75 (1997). 20. J. Grotzinger, E. W. Adams, S. Schroder,¨ Microbial–metazoan reefs of the terminal 40. K. Rutzler,¨ J. Vacelet, “Family Acanthochaetetidae Fischer, 1970” in Systema Porifera, proterozoic Nama Group (c. 550–543 Ma), Namibia. Geol. Mag. 142, 499–517 (2005). J. N. A. Hooper, R. W. M. Van Soest, Eds. (Springer, 2002), pp. 275–278.

19766 | www.pnas.org/cgi/doi/10.1073/pnas.2009129117 Mehra et al. Downloaded at California Institute of Technology on August 19, 2020