Global marine redox changes drove the rise and fall of the Ediacara biota

Abstract The role of O2 in the evolution of early animals, as represented by some members of the Ediacara biota, has been heavily debated because current geochemical evidence paints a conflicting picture regarding global marine O2 levels during key intervals of the rise and fall of the Ediacara biota. Fossil evidence indicates that the diversification the Ediacara biota occurred during or shortly after the Ediacaran Shuram negative C‐isotope Excursion (SE), which is often interpreted to reflect ocean oxygenation. However, there is conflicting evidence regarding ocean oxygen levels during the SE and the middle Ediacaran Period. To help resolve this debate, we examined U isotope variations (δ238U) in three carbonate sections from South China, Siberia, and USA that record the SE. The δ238U data from all three sections are in excellent agreement and reveal the largest positive shift in δ238U ever reported in the geologic record (from ~ −0.74‰ to ~ −0.26‰). Quantitative modeling of these data suggests that the global ocean switched from a largely anoxic state (26%–100% of the seafloor overlain by anoxic waters) to near‐modern levels of ocean oxygenation during the SE. This episode of ocean oxygenation is broadly coincident with the rise of the Ediacara biota. Following this initial radiation, the Ediacara biota persisted until the terminal Ediacaran period, when recently published U isotope data indicate a return to more widespread ocean anoxia. Taken together, it appears that global marine redox changes drove the rise and fall of the Ediacara biota.


| INTRODUC TI ON
After life first emerged more than three billion years ago, single-celled organisms dominated the planet for most of its history. It is not until the Ediacaran Period (635-541 Ma) that large and morphologically complex multicellular eukaryotes became abundant and diverse (Yuan, Chen, Xiao, Zhou, & Hua, 2011). The Ediacara biota, which characterizes the second half of the Ediacaran Period, arose in the middle Ediacaran Period (Xiao & Laflamme, 2009), reached their maximum taxonomic diversity and morphological disparity about 560 Ma, and then declined in the terminal Ediacaran Period (~550-541 Ma) (Darroch, Smith, Laflamme, & Erwin, 2018;Laflamme, Darroch, Tweedt, Peterson, & Erwin, 2013;Shen, Dong, Xiao, & Kowalewski, 2008;Xiao & Laflamme, 2009). Although the phylogenetic affinities of members of the Ediacara biota remain controversial, it is clear that some of them represent mobile macrometazoans, including putative cnidarian-grade animals (Liu, McLlroy, & Brasier, 2010) and bilaterians (Gehling, Runnegar, & Droser, 2015). Importantly, most taxa of the Ediacara biota, and certainly the White Sea and Nama assemblages, appear to be bracketed by two negative carbon isotope excursions (Darroch et al., 2018), raising the intriguing possibility that the rise and fall the Ediacara biota may have been related to environmental and ecological events.
Recent studies provide evidence that an episode of extensive marine anoxia during the terminal Ediacaran Period may have contributed to the decline of the Ediacara biota (Tostevin et al., 2019;Wei et al., 2018;Zhang, Xiao, et al., 2018). However, the cause of the rise of the Ediacara biota during the middle Ediacaran Period remains a subject of intensive debate. A temporal correlation with evidence for a major redox transition suggests that a profound ocean oxygenation event may have sparked this evolutionary event (Canfield, Poulton, & Narbonne, 2007;Fike, Grotzinger, Pratt, & Summons, 2006;McFadden et al., 2008;Shi et al., 2018). However, others have argued that the diversification of bilaterians may have been enabled by the evolution of key developmental toolkits (Erwin, 2009) and/or that the rise of eumetazoans was driven by positive ecological feedbacks (Butterfield, 2007).
The oxygenation hypothesis is attractive because aerobic metabolic pathways provide much more energy than anaerobic ones, and so the presence of free O 2 is often regarded as a prerequisite for the evolution of macroscopic animals, particularly those involved in energetically expensive lifestyles such as mobility, burrowing, and predation . Given the importance of O 2 for animal physiology, researchers have combed Neoproterozoic successions to determine when there were significant changes in the proportion of oxic to anoxic waters in the deep ocean (Canfield et al., 2008(Canfield et al., , 2007Fike et al., 2006;Johnston et al., 2013;McFadden et al., 2008;Sperling et al., 2015).
Carbonate sedimentary rocks from the middle Ediacaran Period have attracted special attention (Fike et al., 2006;Grotzinger, Fike, & Fischer, 2011;Li et al., 2017;McFadden et al., 2008), because they offer an opportunity to clarify the relationship between a redox event and the rise of the Ediacara biota. Middle Ediacaran carbonates in many parts of the world (including South China, Siberia, western United States, Oman, and South Australia) record the largest negative δ 13 C carb excursion (<−12‰) in Earth history, termed the "Shuram excursion" (SE) after its initial discovery in the Shuram Formation of Oman (Burns & Matter, 1993;Grotzinger et al., 2011). When the Ediacara biota and the SE are recorded in the same succession, the former always postdate the latter (Xiao et al., 2016), with only one possible exception in the southeastern Mackenzie Mountains where rangeomorph, arboreomorph, and erniettomorph Ediacara fossils predate a negative δ 13 C carb excursion (−2‰) interpreted as a putative equivalent of SE Narbonne et al., 2014).
Thus, the rise of the Ediacara biota, particularly the appearance of large, mobile, and morphologically complex animals, may have occurred either during (Darroch et al., 2018) or immediately following the SE (e.g., McFadden et al., 2008;Xiao et al., 2016). As such, it has been proposed that the SE represents an unprecedented ocean oxygenation event, which sparked the diversification of complex animals (Fike et al., 2006;McFadden et al., 2008;Shi et al., 2018;Wood et al., 2015). However, the extent of global ocean redox change across this critical interval is poorly constrained (Bristow and Kennedy, 2008).
For instance, proxies for tracking local or regional Fe-S-C systematics and iodine chemistry have been used to infer oxygenation of the deep ocean in some locations during or after the SE (Fike et al., 2006;Hardisty et al., 2017;McFadden et al., 2008;Wood et al., 2015).
However, similar data from other localities have been used to argue for a persistence of redox-stratified and ferruginous marine environments during this critical interval (Canfield et al., 2008;Li et al., 2010;Sahoo et al., 2016;Sperling et al., 2015).
These contrasting views likely arise because these paleoredox proxies are inherently local or indirect tracers of oxygenation. The U isotope system ( 238 U/ 235 U, denoted as δ 238 U) measured in carbonate sedimentary rocks is a more direct probe of global ocean redox conditions and can be used to place quantitative constraints on the extent of global redox changes (Brennecka, Herrmann, Algeo, & Anbar, 2011;Clarkson et al., 2018;Elrick et al., 2017;Lau et al., 2016;Tostevin et al., 2019;Wei et al., 2018;Zhang, Algeo, Cui, et al., 2019; terminal Ediacaran period, when recently published U isotope data indicate a return to more widespread ocean anoxia. Taken together, it appears that global marine redox changes drove the rise and fall of the Ediacara biota.

K E Y W O R D S
early animals, Neoproterozoic, ocean oxygenation, Shuram negative carbon isotope excursion, uranium isotopes Zhang, Xiao, et al., 2018). The power of U isotopes as a global proxy derives from the fact that in the modern ocean U has a long residence time, ~500 kyr (Dunk, Mills, & Jenkins, 2002), and hence, δ 238 U is uniform in the open ocean (e.g., Tissot & Dauphas, 2015). Although the concentration and residence time of U in seawater would both be reduced during times of expanded marine anoxia, studies suggest that the U isotope composition of open ocean seawater was likely uniform even during periods of expanded anoxia Zhang, Xiao, et al., 2018). Seawater δ 238 U varies with redox conditions because the reduction of dissolved U(VI) to U(IV), which is immobilized in anoxic sediments, results in a large and detectable change in δ 238 U (0.6‰-0.85‰), favoring the 238 U over 235 U in the reduced species (Andersen et al., 2014). Thus, δ 238 U of U(VI) dissolved in seawater decreases as the global areal extent of bottom water anoxia increases (Brennecka et al., 2011). Marine carbonate sediments have been demonstrated to record the δ 238 U of seawater, subject to a 0.2‰-0.4‰ offset, which likely reflects incorporation of U(IV) into shallow sediments from anoxic porewaters Romaniello, Herrmann, & Anbar, 2013;Tissot et al., 2018). Studies comparing the trends and absolute values of δ 238 U in coeval Permian-Triassic carbonate sediments from around the world have shown excellent agreement, demonstrating that carbonates may provide a robust record of variations in seawater δ 238 U (Brennecka et al., 2011;Elrick et al., 2017;Lau et al., 2016;. To obtain new constraints on the extent of global redox change across the SE event, we applied the U isotope proxy and associated major and trace elements to carbonates across the SE from three widely separated sections: the Jiulongwan section in South China; the Bol'shoy Patom section in Siberia; and the Death Valley section (the Johnnie Formation) in the western United States (Figure 1). Though there are some debates about the detailed depositional environments (Jiang, Shi, Zhang, Wang, & Xiao, 2011;McFadden et al., 2008;Zhu, Zhang, & Yang, 2007) The SE at the Bol'shoy Patom section is represented by the Kholychskaya Formation, the Alyanchskaya Formation, and the Nikol'skaya Formation, which are ~200, ~530, and ~390 m thick, respectively, and are composed of well-preserved high-Sr limestone (Melezhik, Pokrovsky, Fallick, Kuznetsov, & Bujakaite, 2009).

| S TUDY S EC TI ON S
Sedimentary facies associations suggest deposition on a shallow carbonate platform that was well connected to the open ocean with neither basin isolation nor chemical or physical stratification (see Melezhik et al., 2009 for details). Forty-five samples from the Bol'shoy Patom section were analyzed for U isotopes.
The SE in the Death Valley region, California, comes from Saddle Peak Hills (GPS: N 35°45.439′, W 116°20.936′) and is represented by the Rainstorm Member of the Johnnie Formation, which is >100 m thick in the study section and is composed of interbedded siltstone, sandstone, and conglomerate, with locally abundant dolostone. Sedimentary features suggest deposition in distal-fluvial and shallow-marine (above storm wave base) conditions (Verdel, Wernicke, & Bowring, 2011). The Shuram δ 13 C carb excursion occurs primarily in dolomitic siltstone, but begins in an ~2 m thick dolomitic oolite member known as the Johnnie Oolite. The Johnnie Oolite is a consistent marker bed across the Death Valley region and has been characterized and discussed in many previous studies (Bergmann, Zentmyer, & Fischer, 2011;Corsetti & Kaufman, 2003;Kaufman, Corsetti, & Varni, 2007;Verdel et al., 2011). Fifteen samples from the Death Valley section were analyzed for U isotopes.
F I G U R E 1 Paleogeography at ~565 Ma (modified after Meert & Lieberman, 2008 The precise stratigraphic/temporal correlation between different Shuram sections is difficult because of the lack of radiometric dates to directly constrain the initiation and termination of the Shuram excursion. Previous studies have variously suggested that the Shuram excursion is either a brief event occurring at ca. 560-550 Ma or a protracted event at ca. 580-550 Ma (see summary in Xiao et al., 2016).
Thus, it is uncertain whether the initiation of the Shuram excursion temporally coincides or postdates the ca. 580 Ma Gaskiers glaciations (Pu et al., 2016), and it is also unclear whether the first appearance of diverse Ediacara-type fossils in the Avalon assemblage at ca. 571 Ma (Pu et al., 2016) coincides or postdates the initiation of the Shuram excursion. Recent studies, however, suggest that the Shuram excursion was initiated around 580 Ma (Witkosky & Wernicke, 2018) and ended significantly earlier than 551 Ma Xiao, Bykova, Kovalick, & Gill, 2017;Zhou et al., 2017). On the basis of a subsidence analysis of the Johnnie Formation that hosts the Shuram excursion, Witkosky and Wernicke (2018)  suggest that the Shuram excursion at these localities is broadly synchronous over a duration of 8-10 Myr (Gong et al., 2017;Minguez & Kodama, 2017;Minguez et al., 2015). Ediacaran succession at the Bol'shoy Patom section in Siberia, one of the three sections in this study, is not constrained by radiometric and paleomagnetic data, and the largest negative δ 13 C carb excursion is regarded as equivalent to the Shuram excursion found at other localities (e.g., Grotzinger et al., 2011;Melezhik, Fallick, & Pokrovsky, 2005;Melezhik et al., 2009).

| ANALY TIC AL ME THODS
We have carefully selected fresh rock specimens to avoid veins and cleaned the specimens using 18.2 MΩ Milli-Q water. The cleaned specimens were then dried and powdered to ~200 mesh using an agate ball mill. Approximately 5 g of each sample was dissolved in 1 M hydrochloric acid (HCl) for 24 hr at room temperature. Digests were centrifuged, and the supernatant was separated.
Major, minor, and trace element concentrations were measured on a Thermo iCAP™ quadrupole inductively coupled plasma mass spectrometer (Q-ICP-MS) at the W. M. Keck Laboratory for Environmental Biogeochemistry at Arizona State University (ASU) on splits from each supernatant. Typical precision was better than 3% and 5% for major and trace elements, respectively, based on repeated analysis of in-run check standards.
Prior to U isotopes column chemistry, appropriate amounts of the 236 U: 233 U double spike were added to each sample (e.g., Brennecka et al., 2011;Weyer et al., 2008;Zhang, Xiao, et al., 2018). The spikesample mixtures were evaporated to dryness and taken up in 3N HNO 3 . Uranium was purified using the UTEVA method for isotopic analysis (Brennecka et al., 2011;Chen et al., 2018;Kendall et al., 2015;Romaniello et al., 2013;Weyer et al., 2008;Zhang, Algeo, Cui, et al., 2019;Zhang, Xiao, et al., 2018). All samples were put through UTEVA resin twice in order to completely remove matrix ions. The final purified U was dissolved in 0.32 M HNO 3 and diluted to a U concentration of 50 ppb. Uranium isotopes were measured at ASU on a Thermo-Finnigan Neptune multi-collector ICP-MS at low mass resolution. The standard solution CRM145 (50 ppb U) was analyzed every two samples. Two secondary standards CRM129a and Ricca ICP solution were measured after every fifteen measurements. Sample δ 238 U values were normalized by the average of the bracketing standards.

| UR ANIUM ISOTOPE RE SULTS
An extremely negative δ 13 C carb excursion that characterizes the SE is observed at each of the three sections studied here (

| Post-depositional diagenetic alteration
We compared our U isotope data to standard carbonate diagenetic indicators, such as Mn/Sr ratios and O isotope compositions, to evaluate the influence of post-depositional burial diagenesis.
We note that these traditional diagenetic proxies are not explicitly developed for U isotopes but carbonate C, O, and Sr isotope systematics (e.g., Chen et al., 2018). Although these proxies have their limitations and may not be directly relevant to evaluate diagenesis for carbonate U isotopes proxy, numerical modeling of diagenetic rock-fluid interactions suggests that δ 238 U should be more robust against diagenetic fluid exchange than δ 18 O and 87 Sr/ 86 Sr Lau, Macdonald, Maher, & Payne, 2017). Thus, these traditional carbonate diagenetic indicators may be useful in identifying samples with diagenetically altered δ 238 U signatures.
Mn/Sr ratios in carbonate precipitates have commonly been used as indicators of post-depositional alteration (e.g., Gilleaudeau, Sahoo, Kah, Henderson, & Kaufman, 2018;Jacobsen & Kaufman, 1999;Veizer, 1989), with a cutoff of 3-10 suggested for Precambrian carbonate sedimentary rocks (e.g., Gilleaudeau et al., 2018;Jacobsen & Kaufman, 1999) relative to the other two study sections. The Johnnie Formation is mainly comprised of dolomitic sandstones, which have a low capacity to reserve Sr but a high capacity to reserve Mn, thus having relatively high Mn/Sr ratios (Gilleaudeau et al., 2018;Veizer, 1983). We further investigated the geochemical correlations of Mn/Sr-δ 238  argue against a meteoritic diagenetic origin for the observed uranium isotope trends and instead strongly favor a primary seawater origin.
In carbonates that underwent extensive recrystallization, δ 238 U may be offset from primary depositional values, and therefore, petrographic studies and duplication in different sections are necessary when studying carbonate δ 238 U (e.g., Hood, 2016 Importantly, our key interpretation is built on the average of the three study sections rather than the Jiulongwan section alone.

| Evaluation of detrital contamination
Changes in the extent of detrital input might also cause a δ 238 U offset. Our samples were dissolved in 1 M hydrochloric acid (HCl) prior to extraction of U, which will minimize dissolution of any non-carbonate minerals (e.g., silicates) and organic matter. This expectation is supported by the overall high U/Al ratios in our analyses. The aver- is mainly comprised of well-preserved limestone with high-Sr concentrations (Melezhik et al., 2009 (Lau et al., 2016), Guandao (Lau et al., 2016), Dawen (Brennecka et al., 2011), and Daxiakou (Elrick et al., 2017) sections in South China; the Taşkent section in Turkey (Lau et al., 2016); the Zal section in Iran ; and the Kamura section in Japan ]. All of these sections show strikingly similar trends in δ 238 U across the Permian-Triassic boundary, which is remarkable because they span 1,000s of km-even in  , the three widely separated sections with very different lithologies yielded largely identical δ 238 U records, which strongly argue against anything but primary oceanographic trends.

| Isotopic offset induced from syndepositional diagenesis
Modern carbonate sediments have a δ 238 U composition that is 0.2-0.4‰ higher than that of the contemporaneous seawater (0.27‰ by average; Chen et al., 2018;Romaniello et al., 2013;Tissot et al., 2018), which likely reflects incorporation of 238 Romaniello et al., 2013;Tissot et al., 2018). However, this process does not operate at greater burial depths as the mobility of U is severely restricted in anoxic porewaters, as shown by near-zero porewater U concentrations in deep Bahamian drillcores (Henderson, Slowey, & Haddad, 1999). On this basis, we applied a diagenetic correction factor of 0.2‰-0.4‰ to the measured δ 238 U values prior to U isotope mass balance calculations presented below. Considering this range of diagenetic offset, our best estimates of δ 238 U for the pre-SE and SE seawaters are −0.94‰ to −1.14‰ and −0.46‰ to −0.66‰, respectively.

| S TR ATI G R APHI C VARIATI ON OF U CON CENTR ATI ON S
Several previous U isotope studies suggested that in unaltered rocks, changes to the extent of global seafloor oxygenation will affect the dissolved seawater reservoir of U, and in return the abundance of U incorporated into marine carbonates (Brennecka et al., 2011;Elrick et al., 2017;Lau et al., 2016). Under ideal conditions, stratigraphic variation in U concentrations can record meaningful seawater redox variations, but this relationship can be easily masked by other sources of variation (e.g., Lau et al., 2017 (Romaniello et al., 2013).
Therefore, changing carbonate mineralogy can result in large differences in uranium concentrations but only small changes in the isotopic composition (Lau et al., 2017).  (Table 2), further confirming a mineralogical control on the U concentration. We thus hypothesize that the decoupling of U concentration from δ 238 U in the study sections can be attributed to mineralogical/lithological shifts that affect only the reliability of the carbonate U concentration paleoredox proxy and but not δ 238 U. We therefore focus on the δ 238 U data as a paleoredox proxy, although we acknowledge that the mechanisms that led to the differences in U concentration merit further investigation.

| NE AR-MODERN LE VEL S OF O CE AN OX YG ENATI ON DURING THE S H UR AM E VENT
Because the duration of the SE (>8 Myr; Minguez & Kodama, 2017) is significantly longer than the residence time of U in the SE ocean ) is sensitive to Δ anoxic values (Lau et al., 2017;Zhang, Xiao, et al., 2018). Assuming Δ anoxic = 0.6‰-an average value that is representative of modern anoxic basins like the Saanich Inlet (Holmden, Amini, & Francois, 2015) and the Black Sea (Andersen et al., 2014), the δ 238 U data imply that nearly 100% of the total U ocean sink in the pre-Shuram ocean was accounted for by removal into an- Zheng, 2017; Zhang, Xiao, et al., 2018), representing the range of estimates inferred from modern analogs and microbial U reduction experiments, the estimated area of anoxic seafloor in the pre-SE ocean ranges from 26% to 100% (Figures 2 and 7). Thus, compared with the modern ocean which has ~0.11% anoxic seafloor (e.g., Sheen et al., 2018), widespread anoxia in the pre-SE ocean is implicated for all plausible values of Δ anoxic .
During the SE, the marked positive shift in δ 238 U seawater to values of −0.46‰ to −0.66‰ corresponds to a dramatic expansion of seafloor oxygenation. The extent of ocean anoxia inferred from these values is also sensitive to Δ anoxic values. However, the majority of seafloor needed to be oxic to drive SE seawater δ 238 U to higher values between −0.46‰ and −0.66‰. Under all circumstances, the calculated anoxic seafloor area in the SE ocean is <6% (Figures 2 and 7). Thus, the SE represents a significant ocean oxygenation event, and such a rapid increase in global ocean oxygenation likely occurred within 0.4 Myr if we accept a duration of the SE of ~8 Myr and assume a constant sedimentation rate during the SE event (Minguez & Kodama, 2017 Hardisty et al., 2017;McFadden et al., 2008). Organic-rich mudrocks deposited near the end of the SE have high δ 238 U values that point to an episode of extensive oceanic oxygenation ca. 560-551 Myr ago (Kendall et al., 2015), consistent with the δ 238 U data presented here suggesting that the SE represents a significant ocean oxygenation event.
Furthermore, seemingly conflicting results from Fe-S-C data suggesting local oxygenation and local sustained anoxia as well as local redox stratification (Canfield et al., 2008;Johnston et al., 2013;Li et al., 2010;Sahoo et al., 2016;Sperling et al., 2015) can be reconciled if the ocean redox regime during the SE was similar to or slightly more reducing than the present day. Specifically, this would imply generally oxic global ocean conditions coexisting with anoxia in some local shelf settings (such as oxygen minimum zones) and semi-enclosed basins (such as the modern Cariaco Basin). The combined U proxy from this study and Fe-S-C proxies from previously published studies ultimately provide a more detailed illustration of the redox state of the ocean on global and local scales.

| G LOBAL MARINE REDOX CHANG E DROVE THE RIS E AND FALL OF THE ED IAC AR A B I OTA
The U isotope data from this study combined with previously pub-  (Tostevin et al., 2019;Wei et al., 2018;Zhang, Xiao, et al., 2018). Therefore, this and previous studies confirm that the oceanic redox evolution from the Neoproterozoic to the Paleozoic was not a history of simple and unidirectional oxygenation, but one with rapid perturbations in the relative proportions of anoxic versus oxic waters (Figure 8) Sahoo et al., 2016;Wood et al., 2015;Zhang, Xiao, et al., 2018). F I G U R E 9 Correlation of marine redox evolution and the temporal distribution of macroscopic Ediacaran fossils. Correlations 1 and 2 are modified from Xiao et al. (2016). The anoxic seafloor area estimates shortly after the Shuram excursion (during the Doushantuo Member IV stage) and during the terminal Ediacaran Period (551-541 Ma) are based on δ 238 U data from Kendall et al. (2015) and from Zhang, Xiao et al. (2018), respectively [Colour figure can be viewed at wileyonlinelibrary.com] The possible causal relationship between Ediacaran redox events and the evolution of the Ediacara biota is intriguing. The Ediacara biota contains three temporally successive assemblages that are reasonably constrained by radiometric dates (see summary in Xiao et al., 2016). These are the Avalon (~570-560 Ma ago), White Sea (~560-550 Ma ago), and Nama (~550-540 Ma ago) assemblages, which are named after representative geographic regions where they occur (Waggoner, 2003). However, as discussed above, the age and duration of the Shuram excursion are poorly constrained, and the temporal relationship between the SE and the Ediacara biota is uncertain. Given these uncertainties, we consider two end-member scenarios (Figure 9) and triradialomorphs) whose morphologies are consistent with higher minimum oxygen requirements compared with the majority of taxa from either the Avalon or Nama assemblages (Evans, Diamond, Droser, & Lyons, 2018). These taxa are prominently absent (and may have disappeared) from the Nama assemblage.
Therefore, an episode of pervasive ocean oxygenation across the SE may have been an extrinsic driver either for the emergence of the Ediacara biota during the Avalon assemblage or its diversification in the White Sea assemblage. The subsequent shift to extensive anoxic conditions during the terminal Ediacaran Period coincides with the decline and extinction of the Ediacara biota (Tostevin et al., 2019;Wei et al., 2018;Zhang, Xiao, et al., 2018).
Thus, although genetic, environmental, and ecological factors may have played a role in shaping the evolutionary history of early animals, our data suggest that the rise and fall of the Ediacara biota is, on the first order, coupled with the wax and wane of global ocean oxygenation.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest and no competing financial interests.