Dominant eukaryotic export production during ocean anoxic events reflects the importance of recycled NH₄⁺

Abstract

The Mesozoic is marked by several widespread occurrences of intense organic matter burial. Sediments from the largest of these events, the Cenomanian–Turonian Oceanic Anoxic Event (OAE 2) are characterized by lower nitrogen isotope ratios than are seen in modern marine settings. It has remained a challenge to describe a nitrogen cycle that could achieve such isotopic depletion. Here we use nitrogen-isotope ratios of porphyrins to show that eukaryotes contributed the quantitative majority of export production throughout OAE 2, whereas cyanobacteria contributed on average approximately 20%. Such data require that any explanation for the OAE nitrogen cycle and its isotopic values be consistent with a eukaryote-dominated ecosystem. Our results agree with models that suggest the OAEs were high-productivity events, supported by vigorous upwelling. Upwelling of anoxic deep waters would have supplied reduced N species (i.e., ) to primary producers. We propose that new production during OAE 2 primarily was driven by direct -assimilation supplemented by diazotrophy, whereas chemocline denitrification and anammox quantitatively consumed and . A marine nitrogen reservoir dominated by , in combination with known kinetic isotope effects, could lead to eukaryotic biomass depleted in ¹⁵N.

Mid-Cretaceous episodes of deposition of organic-rich sediments in the proto-Atlantic and Western Tethys basins known as Oceanic Anoxic Events (OAEs) (1) are attributed to high productivity and/or enhanced organic-matter preservation resulting from increases in nutrient supply and/or decreases in the ventilation of deep waters (2–6). Because OAEs are thought to be associated with enhanced CO₂ outgassing during emplacement of large igneous provinces (7, 8), understanding the feedbacks between CO₂, anoxia, and nutrient availability may help us understand better the effects of anthropogenic climate change on ocean circulation, oxygen balance, and marine ecology (9).

Basinal anoxia during OAEs would have promoted loss of fixed nitrogen through the processes of denitrification and anammox. The resulting nitrogen deficits in waters returning to the surface via upwelling would have been amended by nitrogen-fixing cyanobacteria, assuming iron and other micronutrients were adequately available (10). Indeed, enhancement of cyanobacterial production during many episodes of ocean anoxia has been proposed based on increased burial of 2-methylhopanoids (11–13), as these compounds are thought to be markers for cyanobacteria (14). Because such biomarker indices are only qualitative indicators of change and cannot provide quantitative estimates of export flux, complementary data generally include isotope ratios of total sedimentary nitrogen () (11, 15–17), as diazotrophy also affects the nitrogen isotopic budget of the ocean (18–20).

The modern ocean has several localized regions of anoxia, and in these regions, values of generally are higher than the present deep-water average value of +5‰ because of the isotopic fractionation of denitrification expressed in the water column (19, 21, 22). In contrast, sediments from OAE 2 record striking nitrogen isotopic depletion. They are characterized by values of consistently < -1‰, and often < -3‰ (11, 16, 17, 23). Expression of these negative values of varies consistently by depositional location, with the average value of for OAE 2 horizons of the Bonarelli section (Gubbio and Furlo, Italy) being -3.3‰ and the South Ferriby formation (England) being -2.8‰ (16); whereas the average value for the South Atlantic is -1.9‰ (23), for the proto-North Atlantic is -1.8‰ (11, 17, this work), and for the Tarfaya Basin, Morocco is -1.7‰ (between 45–60 m in section) (16). Such differences thus reflect regional heterogeneity of water masses, phototroph ecology, and/or nutrient biogeochemistry; and it has been suggested that patterns of intrabasinal upwelling intensity and nutrient concentrations correspond directly to regional patterns of sedimentation (8, 23, 24).

When viewed alongside the elevated 2-methylhopanoid ratios, negative values of δ¹⁵N have been interpreted as evidence for diazotrophic rebalancing of the nitrogen budget and cyanobacterial dominance of the nitrogen supply for new (export) primary production (6, 11, 15, 25, 26). However, the minimum value of δ¹⁵N for the biomass of marine diazotrophs () should be on average approximately -1.3‰, based on the fractionation associated with nitrogenase (ε_fix = 0-2‰) and the δ¹⁵N value of dissolved N₂ in seawater (approximately +0.7‰). This number is supported by data that consistently show N-fixing cyanobacteria to have values of between 0.5‰ and -2‰ (average -1.4 ± 0.9‰) (25, 27–35). Reports of values significantly < -2‰ are from a cultured Trichodesmium sp. (-3.5‰) that was more negative than field samples collected in situ by the same investigators (32) and from experiments on N₂-sparged Anabaena spp. (-2.4‰) grown in an artificial-seawater medium (ASP-2) that also contained (31). Non-N₂-derived N in the culture media may explain these outliers. Given the likelihood that in situ values of would average approximately -1‰, the prevalence of sedimentary values lower than -2‰ in many OAE sections cannot be explained solely by N supplied via N fixation. These patterns require that additional N-cycling processes be invoked to explain the source of nitrogen driving primary production during OAEs.

Nitrogen Isotopic Records of Sediments, Porphyrins, and Kerogen

Chlorophyll-derived sedimentary porphyrins can be used to generate records of δ¹⁵N values of eukaryotic and prokaryotic phytoplankton that are unaffected by diagenesis (36, 37), as well as to estimate the contribution of cyanobacteria to burial flux (38). We examined nitrogen cycling during OAE 2 using measurements of coeval bulk and porphyrin nitrogen isotopes in sediments from the Ocean Drilling Program Leg 207, Site 1258A (Demerara Rise). A well-defined positive carbon isotope excursion in this section is contained within the high total organic carbon (TOC) interval commonly associated with the OAE (39) (Fig. 1A). Values of (Fig. 1B) from before the OAE through the first two-thirds of the OAE (428.5–423 m composite depth; mcd) decrease from approximately -0.1‰ to approximately -1.4‰ and fluctuate with greater variance than in the middle and top intervals (p < 0.05). The middle of the analyzed section (423 to 419.5 mcd), which spans the end of the OAE as defined by , is characterized by stable values of -2.1 ± 0.3‰. The top section (419.5 to 415.5 mcd) is characterized by an increase in values, returning to -0.7‰ in the uppermost samples. All of these values are lower than the δ¹⁵N minima that are observed in modern sediments, even in regions underlying zones of water-column anoxia or intense nitrogen fixation (40, 41). Because diagenesis and interstitial in clays can shift values of (18, 42), we also measured δ¹⁵N values of kerogen. They show an average negative offset of -0.4‰ relative to bulk sediment (Fig. 1B), suggesting the original primary producers had even lower δ¹⁵N values than what remains recorded by values of .

Fig. 1.

Elemental and isotopic data for site 1258A. The shaded bar represents the excursion interval that defines the OAE. (A) %TOC (squares) and (circles) from (39). (B) δ¹⁵N values of bulk sediment (triangles) and kerogen (diamonds), and their 1-m averaged trends. (C) Porphyrin δ¹⁵N values, and their 3- (green), 5- (tan), and 9- (blue) point averaged trends. (D) The isotopic offset ε_por between bulk sediment and porphyrins (triangles), and kerogen and porphyrins (diamonds), as well as the corresponding fraction of cyanobacterial export based on the endmember values described in the text. The solid vertical line represents a typical algal value of ε_por, and the dotted vertical line represents a marine cyanobacterial value of ε_por (38). All error bars represent 1σ, and preparative and analytical errors are compounded when possible. Raw data are shown in Table S1.

Porphyrin values of δ¹⁵N () also show patterns that are similar to bulk N isotopes, although they exhibit more scatter than the and values. The large error ranges are due to full propagation of analytical uncertainty associated with preparation and analysis by the denitrifier method (43). To overcome the scatter, we plotted 3-, 5-, and 9-point moving averages (Fig. 1C). These different resolutions all show similar patterns, indicating that temporal trends in the results are not sensitive to the degree of smoothing and are not dependent on data density. The data are spaced relatively uniformly (0.5 m), although sampling resolution is higher in some horizons surrounding the excursion interval, from 421.9 to 427.5 mcd (0.2 m). The duration of OAE 2 has been estimated to be approximately 400–800 ka (44, 45), corresponding to sampling resolution for the porphyrin data of approximately 20,000–100,000 y per sample, or significantly longer than present-day estimates of N residence time in the ocean (2,000–5,000 y; ref. 9). Trends observed in the smoothed data thus reflect persistent, potentially steady-state perturbations of the marine N cycle. The top and bottom sections of the core have identical values of δ¹⁵N _por: -5.6 ± 0.7‰ above 419.5 mcd and -5.4 ± 0.7‰ below 423 mcd. However, between 423 and 419.5 mcd, values of average -6.4 ± 0.6‰ and decrease sharply to -7.5‰ approaching and just after the termination of the OAE. This shift correlates with the phasing observed for and , but in both cases the N isotopes lag the excursion observed in .

The ε_por Proxy for Eukaryotic vs. Cyanobacterial Burial

The relative fraction of eukaryotic vs. cyanobacterial export production can be estimated from δ¹⁵N values of porphyrins and their associated sediments. In previous work we examined the biochemical and physiological basis for fractionation of nitrogen isotopes between biomass and chloropigments (38). The offset, known as ε_por (), differs systematically between eukaryotes and cyanobacteria. Values of ε_por for eukaryotes are around 5 ± 2‰; i.e., chlorophyll 5‰ more depleted in ¹⁵N than biomass (33, 38, 46). In contrast, cyanobacteria have values of ε_por between 0 and -10‰ (i.e., chlorophyll equal to or up to 10‰ enriched in ¹⁵N relative to biomass). One example of a value of ε_por near -10‰ had been observed previously for Anabaena cylindrica (33). To expand on this finding, we recently reported data showing that among the seven species of cyanobacteria tested to date, freshwater ecotypes cluster around the ε_por = -10 ± 2‰ endmember, whereas marine ecotypes cluster around the ε_por = 0 ± 2‰ endmember (38).

Because ε_por reflects intracellular partitioning of N isotopes downstream of the amino acid glutamate, it is independent of the nitrogen substrate utilized by the organism (N₂, , or ; ref. 38). Thus, we proposed that ε_por would be an excellent proxy for calculating the relative contributions of eukaryotes and cyanobacteria to marine export production. Measured values of ε_por would be 5 ± 2‰ in a 100% eukaryotic system and would be 0 ± 2‰ in a 100% marine cyanobacterial system, regardless of the proportion of diazotrophic species among the latter (not all marine cyanobacteria are diazotrophs). Moreover, influx of terrigenous biomass and/or cyanobacteria from fresh waters would lead to values of ε_por < 0‰. Indeed, to date the only in situ value of cyanobacterial ε_por from the environment is from a freshwater lake in Japan in which ε_por was determined to be -13 to -16‰ (47).

At Site 1258A the observed value of ε_por throughout the section averages 4.3 ± 0.8‰ (if calculated vs. ) or 4.0 ± 0.8‰ (if calculated vs. ) (Fig. 1D). The values of ε_por reach minima—reflecting maximum burial of cyanobacterial biomass—toward the end of the OAE. There appears to be a qualitative trend of decreasing magnitude of ε_por from the beginning of the OAE until approximately 1 mcd below the termination. At this point, ε_por increases and fluctuates repeatedly until approximately 4 mcd after termination of the OAE, after which ε_por then returns to the starting value near 5‰. This again shows that changes in nitrogen cycle processes are out of phase with the changes in the carbon cycle that define the OAE. Such differences may be expected: in multiple OAE 2 sections the local primary productivity is known to be variable within the overall record of the OAE as defined by δ¹³C values (24, 48).

If we assume that only eukaryotic algae and marine cyanobacteria contribute significantly to the burial flux of photosynthetic pigments (i.e., eliminating the possibility of freshwater cyanobacteria), a value of ε_por consistently > 4‰ during the OAE indicates that the export flux remained on average ≥80% eukaryotic throughout the event. In contrast with previous interpretations invoking N fixers as the primary source of nutrient N (11, 15, 26), these results indicate that the abundance of cyanobacteria contributing directly to export production during OAE 2 was not large; and it suggests that another N source would have been required to sustain such high rates of eukaryotic export production. However, the data for ε_por are consistent with a large relative change in the cyanobacterial population, as observed values of ε_por approximately 4‰ within the OAE indicate approximately 20% cyanobacterial biomass, whereas pre- and post-OAE values of ε_por nearer 5‰ indicate less burial of cyanobacterial biomass (certainly < 5–10%). The ε_por data thus indicate at minimum a doubling to quadrupling of cyanobacterial production, but within a system consistently and overwhelmingly dominated by eukaryotic primary producers. If Site 1258A is representative of OAE 2 in general, the widespread negative values of sedimentary throughout OAE 2 deposits must be attributed to burial of eukaryotes having significant ¹⁵N-depletion in their biomass.

Nitrogen Cycle in Anoxic Oceans: A Paradox of Nitrification and Denitrification

A modest increase in cyanobacterial production is consistent with expected changes to the nitrogen cycle. During OAE 2, anoxic deep waters of the proto-Atlantic and Western Tethys would have contained nitrogen predominantly in the form of . Upwelling rates were high (8), and upwelled into oxic surface waters either was assimilated by phytoplankton or oxidized to and/or . Reducing conditions impinging on the photic zone likely meant that a greater fraction of this and subsequently was reduced to N₂ via denitrification and anammox, causing a modestly greater fixed-nitrogen deficit. Such widespread N deficits suggest it is unlikely that negative values of in sediments of OAE 2 could be due to the expression of isotopic discrimination during nutrient uptake, which occurs for eukaryotes only in nutrient-replete systems in which the nitrogen supply is in excess of biological demand (49, 50). The extent to which fixed N was used to completion in OAE 2 surface waters would have determined the ecological niche for N-fixing cyanobacteria, either as free-living cells or as symbionts; but the overall system was N limited, as generally is the case in the marine photic zone (10). Complete utilization of the available nutrient N implies that the total flux must have been isotopically negative.

A deficit in fixed N during OAEs is not surprising, as anoxia promotes denitrification. What may be surprising is that the deficit was not larger. We suggest that counterintuitively, rates of denitrification may decrease under conditions of extreme basin-wide anoxia. Denitrification and anammox depend on sufficient availability of and . Because these oxidized N species are produced aerobically, extreme oxygen limitation in the water-column may decrease their rate of formation, leaving a greater fraction of remineralized organic nitrogen to cycle throughout these regionally isolated basins and reenter the photic zone as . This in turn would limit the need for compensating N fixation. Evidence for photic-zone sulfide oxidation during OAEs suggests that indeed was completely absent beneath the photic zone, at least episodically (5, 51), and that fixed N in these deep waters would have remained in reduced form. We propose that the values of found in OAE sediments reflect severe diminishment of the deep-water component of the marine N cycle, implying that the deep ocean was a reservoir of . Upwelled , rather than newly fixed N, was the main N source for primary production. Chemocline impingement on the photic zone would have driven nitrification, denitrification, and anammox into competition with -assimilation. The balance between these processes—which varied regionally—would have set the loss rate of N from the ocean and the compensatory rates of N fixation.

To explain the observed values of , isotopic mass balance would then require that the newly fixed N (), plus the upwelled supply, together can yield new production that has values of δ¹⁵N < -2‰ (e.g., Bonarelli and South Ferriby sections; ref. 16). This is different from a modern-ocean scenario, in which denitrification associated with the spreading of anoxic zones leads to progressively higher (positive) values of that are then propagated to (21, 22). The modern-ocean endmembers are thus near-zero (diazotrophs) and more positive (nitrate assimilation and/or recycling), whereas the OAE endmembers must be near-zero (diazotrophs) and more negative ( assimilation and recycling). Although required to explain the data, such a scenario is far from intuitive: it requires that the fixed N lost from the ocean by the processes of denitrification plus anammox have a net positive value of δ¹⁵N. Below we explore how such a system might be possible.

-Upwelling Model

To yield a marine system in which the burial flux of has a negative value, we assume that (and ) are produced only in the aerobic photic zone and are reduced quantitatively to N₂ in the chemocline by denitrification and/or anammox. This loss is analogous isotopically to sedimentary denitrification in the modern ocean, which is considered to impart zero fractionation because it proceeds to completion, and by mass balance, (19).

The following additional conditions then would be sufficient to achieve a denitrifying flux of N₂ that is net isotopically positive. To yield surface waters in which and N₂ are the most important bioavailable sources of N, we assume that nitrification of the upwelling flux to followed by phytoplanktonic assimilation is much less significant than direct assimilation of concomitant upwelling . Where is available, is a less favorable nutrient for phytoplankton growth due to the higher energetic costs associated with its reduction (52). Nitrite generally is not believed to be an important source of nutrient N (53), and thus we assume it also is removed by denitrification or, more likely, by anammox. The dominant fractionations and fluxes in the N cycle are then ε_fix and φ_fix (N₂ fixation), ε₁and φ₁ (-assimilation), and ε₂ and φ₂ (ammonium oxidation, ), whereas the burial flux is small relative to these internal cycles (Fig. 2A). We also specify the flux associated with remineralization of sinking phytoplankton N (φ₃), and assume no fractionation for this process. As stated above, all oxidations and reductions downstream of φ₂ are quantitative and do not impart further fractionation.

Fig. 2.

Conceptual model for sedimentary values of in an ocean in which is the dominant fixed N species. (A) System in which the δ¹⁵N values of exported eukaryotic biomass depend on the fractional fluxes to ammonium assimilation (φ₁), oxidation (φ₂), and recycling (φ₃), as well as the difference between the associated fractionation factors ε₁ and ε₂. (B–D) Calculated δ¹⁵N values of sedimentary organic matter as a function of percent export from diazotrophs and fractional fluxes φ₁ and φ₂ for three sets of fractionation factors: (B) ε₁ - ε₂ = -5‰; (C) ε₁ - ε₂ = 5‰; (D) ε₁ - ε₂ = 10‰. (E) Data for for OAE 2 from the literature [11 (red); 16 (blue, Italy; green, England); 17 (yellow); 23 (orange), and this study (gray)], plotted relative to the range of paired values of ε₁ and ε₂ solved with the model, assuming 20% export of diazotrophic biomass (solid lines), as well as an ε₁/ε₂ offset of 10‰ assuming 10% export of diazotrophic biomass (dashed line).

In N-limited surface waters, new production reflects the isotopic signature of the integrated nitrogen budget. The resulting value of will reflect a weighted average of the δ¹⁵N values of diazotrophic cyanobacteria (δ_diazo) and of -consuming phytoplankton (δ_phyto). The former will be equal to minus ε_fix, whereas the latter will be equal to the δ¹⁵N value of their source minus a fractionation factor ε₁. Isotopic mass balance dictates that the source is set by the relative flux of that is utilized (φ₁) vs. nitrified (φ₂), as well as the flux (φ₃) that returns remineralized to the surface via upwelling. Ratios of these fluxes, the fractionations associated with utilization and oxidation (ε₁ and ε₂, respectively), and the value of δ_diazo, together set :

Assuming δ_diazo = -1‰ and substituting and enables the system to be solved for a range of combinations of ε₁ and ε₂ (full derivation in SI Text). This model generates negative values for δ_phyto when ε₁ > ε₂, and it produces an ocean system in which the major reservoir of dissolved inorganic nitrogen (DIN) accumulates as ¹⁵N-depleted .

Biomass having a negative value of δ¹⁵N results from the co-occurrence of ammonium oxidation and ammonium assimilation in the photic zone, the competing effects of fractionations associated with these processes on a single pool, and the upwelling of recycled, ¹⁵N-depleted . Both assimilation and oxidation fractionate such that their products are more ¹⁵N-depleted than the source , and therefore the pool in surface waters becomes more ¹⁵N-enriched as it is consumed. If the fractionation associated with assimilation exceeds the enrichment of the pool that is caused by nitrification/denitrification (i.e., ε₁ > ε₂), the resulting biomass (φ₁) is isotopically negative. Regenerated in deep waters isotopically resembles the sinking biomass from which it is remineralized. As this upwells into the photic zone, it again becomes ¹⁵N-enriched and the system maintains steady-state.

The resulting value for total buried organic matter () is tempered by the percent contribution of diazotrophic biomass (Fig. 2 B–D) such that values of approach -1‰ when there is greater burial of diazotrophs, but decrease as the ratio φ₂/φ₁ increases and diazotrophic burial decreases. This is consistent with records showing the most negative values of in pelagic locations with lesser apparent bacterial biomass burial (16) and more positive values of in epicontinental environments with higher apparent bacterial flux (16).

The model thus depends on the relative magnitudes of ε₁ and ε₂ compared to the N deficit and resulting diazotrophic contribution. It is possible that the fractionation associated with -assimilation (ε₁) by the enzyme glutamine synthetase (GS) may exceed that of -oxidation (ε₂) by the enzyme ammonium monooxygenase (AMO) under some circumstances. The observed value of ε₁ (4–27‰) will depend on concentration, with larger fractionations expressed under -rich conditions (54). In the modern ocean, concentrations are low and ε₁ is confined to the lower end of this range. Under the -replete conditions that we propose for OAE 2, assimilation using different enzymatic controls may lead to expression of ε₁ with a larger magnitude, although to date very little information is available about fractionation during assimilation by natural planktonic assemblages (55).

The value of ε₂ also remains poorly constrained. The relative fraction of aerobic ammonia oxidation by archaea vs. bacteria during OAE 2 is not known, but δ¹³C and archaeal biomarker data measured in black shales deposited during the Albian OAE1b (approximately 112 Ma) suggest that Crenarchaeota (now called Thaumarchaeota; ref. 56) that are believed to be responsible for most ammonium oxidation in the modern ocean (57), were abundant in the Cretaceous (58). Values of ε₂ for bacterial AMO are approximately 14–38‰, for a variety of species grown on 1–2 mM (59). Recent measurements of isotope effects associated with archaeal ammonia oxidation show a similar range of values, from 10–37‰ (60). In all cases, the relative contributions of fractionations associated with transport of or diffusion of NH₃ through membranes and equilibrium of are uncertain. It is thus difficult to extrapolate these cultures to natural systems, except to suggest that bacterial and archaeal AMO results are similar.

If ε₁ was large due to elevated concentrations (54) upwelling to the base of the photic zone from a large, deep pool, the condition of ε₁ > ε₂ could be met. For example, if ε₂ = 22‰ (average archaeal value) and ε₁ = 27‰ (maximum enzymatic effect on -assimilation), then ε₁ - ε₂ = 5‰. This results in values of for export production that will be < -2‰ (Fig. 2C) if oxidation consumes at least one-tenth of the upwelling flux (φ₂ > 0.1) and the burial contribution of diazotrophs is 20%, the upper limit based on our data for ε_por. Other versions of the model that impose larger differences between ε₁ and ε₂ (e.g., ε₁ - ε₂ = 10‰, Fig. 2D) also are compatible with some of the data from OAE 2, in particular a few of the very negative values of for the sections from Italy and England (Fig. 2 D and E) (16). Analogous models with ε₁ < ε₂ can produce only positive values of , as would be seen in the modern ocean (Fig. 2B). Using our conceptual model, most data for compiled from OAE 2 (11, 16, 17, 23, this paper) fall within isotope space corresponding to ranges of ε₁ - ε₂ = 5‰ (Fig. 2E).

We further tested the plausibility of our conceptual framework using a simplified steady-state model that calculates δ¹⁵N values of biomass N, , , and in a two-box (surface and deep) ocean. The model was optimized to reproduce known modern values using estimates of fluxes and fractionation factors from the literature. To run the model subsequently for the OAE, we modified nitrogen-redox partitioning (more , less ) and changed the magnitude of associated fluxes proportionally. Rates of upwelling and the total N inventory remained the same in both cases. By changing these parameters, the model generated sedimentary values of -4.4‰ for the OAE and +4.9‰ for modern sediments. For a complete model description, results, and sensitivity analysis, see Supplementary Information.

Implications

Our model implies a widespread and well-mixed “ammonia ocean” for the proto-Atlantic and Western Tethys because it requires a sustained source of upwelling that can be used for biological assimilation. This can be achieved if nitrate production is limited by severe demands on , possibly through enhanced anammox. In such an ocean, ammonia assimilators and N fixers both could out-compete assimilatory reducers due to the dominance of and a limited rate of generation. Postulated high rates of upwelling, combined with nutrient trapping under estuarine circulation in the North Atlantic (8), may explain why these negative δ¹⁵N signals are widespread during OAEs, yet are regionally variable (16). The trapping of quantitatively significant levels of in deep waters during OAEs also helps preserve the total pool of marine N, alleviating the need for excessive rates of nitrogen fixation. Extreme anoxia may therefore exert a natural, negative feedback on the nitrogen cycle by preventing the ocean from denitrifying completely.

Our proposed model for the N cycle during OAE 2 also helps to explain why extreme N isotopic depletion is not seen in modern anoxic basins like the Black Sea and the Cariaco Trench, where δ¹⁵N values of particulate organic nitrogen are > 0‰ throughout the water column (41). The nutrient sources and circulation patterns in these two systems are not analogous to anoxic oceans. The Cariaco Trench is a silled basin that receives from the Atlantic, and sedimentary organic nitrogen in the Cariaco basin carries an isotopic signature that reflects a mass balance between Atlantic that has been influenced by N₂ fixation (approximately 3‰) and N₂ (local nitrogen fixation) (61). In the Black Sea, a commonly used analog for anoxic oceans, the supply of N to surface waters is largely sourced from continental rivers, whereas the intense salinity stratification limits the upwelling of deep and promotes formation of followed by nearly quantitative loss via the anammox process (62). The nutrient N cycle of the modern Black Sea, therefore, primarily is analogous to a large lacustrine system with severe stratification. In contrast, we envision OAE 2 as a time of sustained upwelling.

The ammonia ocean scenario also may help to explain the temporal evolution of N isotope patterns seen in our data. Values of and are out of phase with carbon isotopes. They do not begin to decrease until the middle of the OAE interval, and their minimum persists past the traditionally defined termination of the event. This phase lag may reflect the balance of oxidants in the marine system. Enhanced burial of organic carbon during OAEs should be associated with accumulation of oxygen in the ocean and atmosphere. This in turn would increase the rates of ammonium oxidation and nitrification, eventually suppressing anammox and allowing to accumulate. Indeed, our predicted values of decrease as φ₂ increases (Fig. 2 C–E). The predicted isotopic trajectory, therefore, is that values will decrease during the early stages of ocean reoxidation. Values of only would “flip” to positive values when the nitrification flux (φ₂) was sufficiently high to accumulate excess , allowing subsequent denitrification to enrich ¹⁵N in the accumulating reservoir. These results highlight the importance and promise of using temporal records of ε_por in conjunction with values to examine both the succession of marine ecosystems and the redox state of the ocean.

In sum, a mid-Cretaceous deep ocean dominated by reduced rather than oxidized nitrogen species, normal rates of ocean circulation (63), and enhanced input of nutrients (5, 6, 8) together could yield negative values of biomass δ¹⁵N and sustain a primary producer community that remained rich in eukaryotes. Although the oxidation state and temperature of OAE oceans was very different from the modern ocean, the persistent dominance of eukaryotes and dependence of primary producers on upwelled nutrients suggests that the balance between gross and net production was not greatly dissimilar from the present-day. Our results imply that additional feedbacks act under oxygen-limited conditions to maintain nitrogen balance, thereby limiting the extent of denitrification and the compensatory expansion of diazotrophy during OAEs.

Materials and Methods

Sediments were obtained from Ocean Drilling Program Leg 207, Site 1258A, from the Demerara Rise, offshore from modern Surinam. Samples spanned 415–428 m composite depth (mcd). Forty samples were analyzed for bulk , , and at approximately 0.5-m spacing. Sampling resolution was higher leading into and coming out of the OAE, which spanned approximatley 422–426 mcd (Table S1. Sample preparation and isotopic analysis followed established methods (43); details are given in SI Text.