Early animal evolution and highly oxygenated seafloor niches hosted by microbial mats

The earliest unambiguous evidence for animals is represented by various trace fossils in the latest Ediacaran Period (550–541 Ma), suggesting that the earliest animals lived on or even penetrated into the seafloor. Yet, the O2 fugacity at the sediment-water interface (SWI) for the earliest animal proliferation is poorly defined. The preferential colonization of seafloor as a first step in animal evolution is also unusual. In order to understand the environmental background, we employed a new proxy, carbonate associated ferrous iron (Fecarb), to quantify the seafloor oxygenation. Fecarb of the latest Ediacaran Shibantan limestone in South China, which yields abundant animal traces, ranges from 2.27 to 85.43 ppm, corresponding to the seafloor O2 fugacity of 162 μmol/L to 297 μmol/L. These values are significantly higher than the oxygen saturation in seawater at the contemporary atmospheric pO2 levels. The highly oxygenated seafloor might be attributed to O2 production of the microbial mats. Despite the moderate atmospheric pO2 level, microbial mats possibly provided highly oxygenated niches for the evolution of benthic metazoans. Our model suggests that the O2 barrier could be locally overcome in the mat ground, questioning the long-held belief that atmospheric oxygenation was the key control of animal evolution.

www.nature.com/scientificreports www.nature.com/scientificreports/ what exactly were the oxygen conditions under which the animals, especially benthos that live on the SWI or even penetrate into the sediment, survived and radiated in the terminal Ediacaran ocean.
In order to understand the environmental conditions experienced by the earliest benthic animals, we propose a method of retrieving the seafloor O 2 fugacity from sections where the earliest animals evolved. The Ediacaran Dengying Formation (551-541 Ma), Yangtze Block, South China, contains abundant trace fossils together with both canonical and atypical Ediacaran fossils 4,26,27 . These fossils are particularly well known from in the bituminous limestone of the Shibantan Member (Fig. 1). In addition, the earliest biomineralized organism, Cloudina, has been discovered in the Baimatuo Member of the uppermost Dengying Formation [28][29][30][31] . As one of the few terminal Ediacaran fossiliferous carbonate successions in the world, the Dengying limestone may provide a unique window to view the environmental and ecological background of the latest Ediacaran evolution. In this study, we developed a new proxy, carbonate associated ferrous iron (Fe carb ), to constrain the seafloor O 2 fugacity during the emergence of the earliest benthic animals in the latest Ediacaran Dengying Formation in South China.

fe carb As A Proxy of Seafloor O 2 Fugacity
Ferrous Fe [Fe(II)] is thermodynamically unstable in oxic conditions, and would be oxidized to ferric Fe [Fe(III)], resulting in the precipitation of iron oxides or iron oxyhydroxides at neutral to alkaline pH conditions. The redox equilibrium between Fe 2+ and O 2 could be expressed as:  Thus, modern oxic seawater have extremely low dissolved Fe content (0~1.5 nmol·L −1 ) 32 . Ferric Fe can be reduced by iron reducing microbes (IRM) in suboxic-anoxic sediments by process collectively known as the dissimilatory iron reduction (DIR) 33 . The chemical reaction of DIR can be expressed as: Fe  HCO  OH  2  3  4  7  (2)   2 3  2  2  2  3 DIR generates Fe 2+ , increasing Fe 2+ concentration in sediment porewater (0~500 μmol·L −1 in modern ocean) 34,35 . As such, high porewater Fe 2+ content and low seawater Fe 2+ content generate a Fe 2+ concentration gradient, which results in the upward diffusion of dissolved Fe 2+ and a benthic Fe 2+ flux. The concentration gradient of Fe 2+ (∇ Fe ) can be expressed by the following simplified equation: The benthic flux of Fe 2+ (Flux Fe ) can be expressed as: where D Fe is the diffusivity coefficient of Fe 2+ in porewater. At equilibrium, Eq. 1 can be written as: Thus, in theory, when pH and temperature remain unchanged, there is an inversely exponential relationship between benthic iron flux (Flux Fe ) and bottom water O 2 (O 2-BW ) adjacent to the SWI, which can be expressed as an empirical equation: In the modern ocean, there exists a negative correlation between O 2-BW and Flux Fe 36-38 ( Supplementary  Fig. 6). Instead of using in situ fluxes, we collected the benthic flux data measured by non-invasive benthic chambers 39 . Strong benthic bioturbation related with water depth in the modern ocean can elevate the iron flux ( Supplementary Fig. 6). In order to recede the influence of benthos, we choose the data collected from locations with water depth greater than 500 m. Thus, the best fitted power function can be expressed as follows (see supplementary text; Supplementary Fig. 7; the units of Flux Fe and O 2-BW are mol · m −2 · Myr −1 and mol · L −1 ): We suggest that Flux Fe could be recorded by carbonate precipitating on the seafloor. Because Fe 2+ and Ca 2+ have similar ionic radii and charge, Fe 2+ has the tendency to replace Ca 2+ in carbonate minerals 40 . Fe carb is determined by Fe 2+ concentration in solution that is related to the redox condition (or oxygen fugacity) and the partitioning coefficient that is temperature, pH, Eh and mineralogy dependent 41  where K Fe is the partitioning coefficient of the benthic Fe 2+ flux into the carbonate lattice. M Fe is the molecular weight of Fe (56 g/mol), s is the sedimentation rate, and ρ is the density of carbonates (2.71 g/cm 3 ).
Notably, although Fe carb content is determined by the seafloor O 2 fugacity, which is controlled-although not uniquely-by the atmospheric pO 2 level, the quantitative reconstruction of atmospheric pO 2 level by using Fe carb is not directly applicable. On one hand, Fe speciation in seawater is complex. In addition to free Fe 2+ , the (2019) 9:13628 | https://doi.org/10.1038/s41598-019-49993-2 www.nature.com/scientificreports www.nature.com/scientificreports/ dissolved ferrous Fe species also include various Fe-organic complexes, which accounts for the majority of the dissolved Fe in the modern ocean 32 . On the other hand, atmospheric pO 2 level is not the only control of bottom seawater redox. Instead, both organic matter influx and ocean circulation also play important roles 42 . If the water column above the SWI enriches organic matter or ocean circulation is stagnant, there can be decoupling between atmospheric pO 2 level and bottom water O 2 content. Therefore, Fe carb can only constrain the redox conditions on the seafloor, and not in the atmosphere.
Furthermore, using Fe carb to reconstruct seafloor O 2 fugacity can only be applied to carbonate that precipitated on the seafloor. Before the evolution of skeletonizing organisms, i.e. Ca-carbonate biomineralization in the latest Ediacaran 43,44 , marine carbonate precipitation derive from biotically induced carbonate precipitation and direct abiotic precipitation from seawater or porewater 45 . The inorganic precipitation, identified by crystal fan and herringbone structures in carbonate, was common in Archean and decreased in abundance after the late Paleoproterozoic 46 . By contrast, biotically induced precipitation is driven by an elevation of carbonate saturation resulting from releasing of microbial metabolite into carbonate producing micro-environments 47,48 . The Shibantan limestone contains abundant organic-rich filaments, and is composed of crinkled microlaminae that have been interpreted as microbial mats 7,27,49 (Fig. 2). It is reasonable to argue that the Dengying carbonate precipitation was triggered by benthic microbes on the seafloor, warranting the application of Fe carb to reconstruct seafloor O 2 fugacity. Furthermore, before the evolution of pelagic planktonic carbonate secreting organisms in Mesozoic, nearly all marine carbonate in the Paleozoic ocean was generated by benthic calcifiers, such as brachiopods, corals, and echinoderms 50 . Although carbonates precipitation from the water column cannot be completely ruled out, physical and biological abrasions of biogenic carbonate should be the major source of carbonate mud (i.e. micrite) in the Paleozoic ocean 51 . Therefore, Fe carb of micrite from the Paleozoic carbonate can be used to reconstruct the seafloor O 2 fugacity as well. In this study, we use Fe carb of the late Paleozoic (late Devonian and early Carboniferous) limestone (see supplementary text; Supplementary Fig. 3) as references. It is reasonable to speculate that the concentration of dissolved oxygen in the late Paleozoic shallow water was in equilibrium with the atmosphere, whose pO 2 levels were at least comparable to or even higher than that of the modern atmosphere 9 .
It should be noted that there are limitations and assumptions when applying Fe carb to reconstruct the seafloor O 2 fugacity. First, considering the short residence time of dissolved iron in seawater (on the order of 100~200 yr) 52 , Fe carb only reflects the local seafloor redox rather than the global state which can be estimated by uranium isotopes 53 . Second, Eq. 10 is based on the assumption of unlimited benthic Fe 2+ flux. However, benthic Fe 2+ flux would be finite when there are deficient supplies of reactive Fe or organic matter. In this case, low Fe carb could also be generated at low seafloor O 2 fugacity with insufficient supplies of reactive Fe or organic matter. Thus, the interpretation of Fe carb data should also consider siliciclastic and TOC contents in carbonate so as to guarantee sufficient Fe flux from sediment porewater. Third, we suggest that Fe carb can only be applied to limestone rather than dolostone. Generally higher Fe carb of dolostone may not only be the consequence of Fe-enriched dolomitization fluid, but also result from higher miscibility between Mg and Fe in carbonate lattice than between Ca and Fe 40 . Possibly multiple stages and fluid origins of dolomitization also interfere Fe carb as a seafloor redox indicator. Thus, we recommend samples with low Mg/Ca (<0.05) should be selected. Finally, authigenic carbonate precipitation within DIR zone that has high Fe 2+ content could also contribute to higher Fe carb . Therefore, Fe carb represents the minimum estimation of the seafloor O 2-BW .

Geological Setting and Sample Description
The Dengying Formation in the Yangtze Gorges area can be divided into, in ascending order, the Hamajing, Shibantan and Baimatuo members 30 (Supplementary Fig. 2). The Shibantan Member, sandwiched between the peritidal dolostone of the Hamajing and Baimatuo members, is composed of dark, laminated, organic-rich limestone. The Shibantan limestone contains a variety of fossils, including trace fossils, Ediacara fossils (e.g. Wutubus annularis), algal fossils (e.g. Vendotaenia), as well as benthic cyanobacteria (e.g. Oscillatoria) (Fig. 1). The absence of subaerial exposure structures as well as the occasional occurrences of hummocky cross bedding suggests the deposition in a deep subtidal environment, probably below the fair-weather wave base but above the storm-wave base ( Fig. 2; see Supplementary Text).
Samples were collected from the Sixi and Huangniuyan sections in the Yangtze Gorges area, South China ( Fig. 2a-d; Supplementary Fig. 1). The fine laminae are confined by organic-rich microbial filaments and are composed of alternating micritic and calcspar layers (Fig. 2e-h). The micritic layer normally has higher organic and siliciclastic contents, while the calcspar layer is composed of subhedral-anhedral calcite crystals of up to 100 μm in size. The calcspar crystals usually have fuzzy boundaries and contain abundant remaining micrites, suggesting the calcspar might derive from recrystallization of micrite in the early stage of diagenesis.

Discussion
The Shibantan limestone (Mg/Ca molar ratio < 0.05) samples have extremely low Fe carb values both in micritic and calcspar layers, with little difference between these two types of layers, supporting the petrological observation that the calcspar mainly derives from recrystallization of micrite (Fig. 3). In addition, the Mn content in the Shibantan limestone is extremely low or even undetectable, suggesting little alteration by meteoric fluids 54 (Supplementary Table 1). Other diagenetic processes, which dominantly occur in anoxic conditions and cause more Fe 2+ incorporation into the carbonate lattice, would most likely elevate Fe carb . Therefore, low Fe carb of the Dengying limestone may not result from diagenetic processes.
Low Fe carb of the Dengying limestone cannot be attributed to oceanic euxinia as well (i.e. H 2 S enriched but Fe 2+ depleted), because abundant trace fossils and macroscopic Ediacara fossils strongly argue for a non-sulfidic environment 4,26,49 . Neither could the low Fe carb be attributed to low [Fe] pw resulting from insufficient supply of organic matter and reactive Fe. Firstly, the Dengying limestone has high siliciclastic contents (average = 14.34%, n = 47; Supplementary Table 4), suggesting sufficient reactive Fe in sediments ( Supplementary Fig. 9). Secondly, high organic carbon content (average = 2.47%, n = 47, Supplementary Table 4) in bituminous limestone warrants DIR in sediment porewater. In addition, low Fe carb of the Shibantan limestone (mean = 22.48 ppm, n = 24) is close to, or even lower than that of the late Paleozoic shallow water carbonates (mean = 63.44 ppm, n = 31; Fig. 3). Considering the sedimentation rates of the Shibantan limestone (at least 24 m/Myr) and shallow marine carbonates in the Late Paleozoic sections (12.5 m/Myr for the Madao section, 6.4 m/Myr for Panlong section and 28.6 m/Myr for Dazhai section), the Ediacaran seafloor O 2 fugacity should be comparable to that of the well ventilated seafloor in Late Paleozoic (Eq. 10). Thus, low Fe carb of the Shibantan limestone can only be explained by high seafloor O 2 fugacity.
To quantify the seafloor O 2 fugacity by Eq. 10, K Fe should be determined in advance. However, K Fe that specifically represents partition coefficient of benthic Fe 2+ incorporation into calcite has not been directly determined for modern limestone, although the factors affecting Fe 2+ incorporation into calcite in aqueous solution at equilibrium state have been investigated 55,56 . To constrain this unknown, we use Fe carb data of late Devonian-early Carboniferous shallow marine carbonate samples (Madao, Panlong and Dazhai sections), to calculate the K Fe . The calculated K Fe value is then validated by the equivalent deep water carbonate samples (Duli, Xiada and Daposhang sections), which were collected from the beddings that contains abundant benthic animal fossils and thus were also inferred to represent oxic conditions ( Fig. 3 55 . To reconstruct O 2-BW during the deposition of carbonates by Eq. 10, we use K Fe = 2.32 and s = 5, 10, 20, 40 m/Myr. O 2-BW of 6.25 μmol/L is set as the upper bound of anoxic and euxinic conditions (i.e. microbial sulfate reduction occurs below this threshold) 17 and 68 μmol/L as the cutoff for the suboxic and oxic conditions 57,58 . The modeling result displays a negatively exponential relationship between Fe carb and O 2-BW (Fig. 4). Indeed, we cannot completely exclude carbonate precipitation from the water column, which will ultimately decrease Fe carb due to low Fe 2+ content in the seawater. Assuming that 80% of carbonate originally precipitates on the seafloor, and the other 20% precipitates from the water column, the colored area in Fig. 4  www.nature.com/scientificreports www.nature.com/scientificreports/ (162 μmol/L) is equivalent to the dissolved O 2 content at 60% PAL pO 2 . If we assume that measured Fe carb of the Dengying Formation is 20% lower than original benthic carbonates because of seawater carbonate mixing, the minimum seafloor O 2 fugacity is still 143 μmol/L. Thus, even if the Ediacaran atmospheric pO 2 level reaches the maximum estimate of 40% PAL 9 , the seafloor O 2 fugacity during the deposition of the Shibantan limestone was likely oversaturated with respect to the atmospheric pO 2 level.
High seafloor O 2 fugacity during the deposition of the Shibantan limestone seems contradictory with the modest atmospheric pO 2 levels and extensive seafloor anoxia in Ediacaran and Cambrian 8,59 . Local seafloor oxygenation must require a continuous O 2 supply to maintain oxic status under such predominately anoxic condition. Inspired by a modern analog, Los Roques lagoon in Venezuela 60 has low O 2 concentration in water column and is colonized by O 2 -generating cyanobacteria mat in floor, but the O 2 concentration in the mat ground could be four times higher than that in the water column. Here, we suggest that the seafloor oxygenation might result from the development of microbial mats on the seafloor during the precipitation of Shibantan limestone (Figs 1a,e, 5 and 6). Microbial mats generate O 2 that is ready to be emitted into the water column, resulting in partial oxygenation of the adjacent bottom water. In addition, the downward diffusion of O 2 produced by microbial mats would lower the redox boundary of DIR zone, reducing the intensity of benthic flux of Fe 2+ (Fig. 6). This model is consistent with the widespread microbial structures, e.g. warty textures 26 , microbial laminae 4 , and benthic cyanobacteria (Fig. 1e) in the Dengying limestone.
It has been proposed that microbial mats might have played a key role in the preservation of Ediacara fossils (the death mask hypothesis) 61 , and the disappearance of Ediacara fossils at the Ediacaran-Cambrian boundary might be related to the disappearance of microbial mats after the evolution of metazoans 2 . Here, our new model proposes an alternative, but not mutually exclusive, interpretation that microbial mats might also provide a more locally oxygenated environment in the context of generally low atmospheric pO 2 level and widespread seafloor anoxia. Thus, microbial mats on shallow marine seafloor may generate an oxygenated oasis that might have stimulated the diversification of metazoans, even when the atmospheric pO 2 level was only 10-40% PAL, barely meeting the threshold for animal evolution 16 . Therefore, it is plausible that the earliest animals would refrain from floating in the ocean that is primarily anoxic and is characterized by dramatic redox oscillation, and prefer utilizing O 2 and food on and within the microbial mat. Such a hypothesis is also supported by the widespread late Ediacaran trace fossils associated with microbial sedimentary structures 4,5 , some of which may indicate activities under microbial mats 49 . Our hypothesis could also support that the majority of the earliest animals were evidenced by trace fossils which record benthic instead of pelagic ethology. The lack of pelagic body fossil records may reflect the ecological constraint of the terminal Ediacaran communities driven by the ocean redox, not just a result of taphonomy or poor preservation.
Finally, the rise of atmospheric pO 2 level, which was thought to provide the upper constraint on the redox of the ocean, has been regarded as the priori for the animal evolution. However, our study suggests that local seafloor O 2 fugacity might significantly exceed the saturated O 2 content at a given atmospheric pO 2 level, and the local seafloor oxygenation might be attributed to the development of microbial mats. If this is the case, seafloor oxidation and atmospheric oxygenation might be decoupled. It is highly probable that seafloor might have long been locally oxidized when atmospheric pO 2 level was still low, because oxygenic microbial mats, e.g. stromatolites, are believed to cover the shallow marine seafloor since Archean time 62 . Therefore, we suggest that atmospheric or oceanic oxygenation may not be the crucial control on the emergence of animals; instead, life may have played the central role in the evolution of habitable planet.

Methods
Elemental compositions measurement. Mirrored thin and thick sections were prepared for micro-mill sampling. Sample powders were micro-drilled from the thick section under the guide of thin section observation under optical microscopy. Based on the character of laminated limestone, two types of carbonate fabrics, micrite and calcispar, were sampled from the same specimen. About 50 mg of limestone powder was collected approximately in each sample and placed into a 15 ml centrifuge tube.
The sample preparation followed the sequential extraction procedure for carbonated associated Fe designed by Poulton and Canfield (2005) 63 . A buffer solution mixed by acetic acid (HAc) and ammonium acetate (NH 4 Ac) was prepared, and the pH of 4.5 was adjusted accurately before use. For each sample, about 50 mg of sample powder was weighed and was dissolved in 10 ml buffer solution in a centrifuge tube. In order to ensure the solution has full contact with the sample, tubes were placed in a shaking table at 50 °C for 48 hours. After centrifugation, 0.5 ml supernatant was taken out and was mixed with 4.5 ml 2% nitric acid (HNO 3 ) in a new centrifuge tube. Finally, elemental compositions were measured with a Spectra Blue Sop Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES) at Peking University. All analyses were calibrated by a series of gravimetric standards with different concentrations (ranging from 0.1 ppm to 10 ppm) that were run before sample measurements. TOC measurement. The limestone of the Dengying Formation was smashed into sample powder and about 100 mg of powder for each sample was weighed and was placed into a 50 ml centrifuge tube. To fully remove the inorganic carbon, 20 ml hydrochloric acid (HCl, 3N) was added to each centrifuge tube, which was then placed in an ultrasonic bath for 1 hour. The reaction was allowed for 12 hours. Then Milli-Q water (18.2 MΩ) was used to rinse the powders until pH reaches 4-5. After that, samples were dried overnight and loaded into capsules for TOC analysis at the Stable Isotope Research Facility (SIRF) at Louisiana State University, USA. Elemental analyzer (Micro Vario Cube, Isoprime Ltd., Cheadle, UK) flash-combust the samples in Tin capsule in a 950 °C furnace. Isoprime 100 (Isoprime 100, Cheadle, UK) gas source mass spectrometer can analyze the resulting CO 2 by continuous flow. The analyzed precision for TOC data is within 0.3%.

Data Availability
All data is available in the main text or the supplementary materials.