Influence of methane seepage on isotopic signatures in living deep-sea benthic foraminifera, 79° N

Fossil benthic foraminifera are used to trace past methane release linked to climate change. However, it is still debated whether isotopic signatures of living foraminifera from methane-charged sediments reflect incorporation of methane-derived carbon. A deeper understanding of isotopic signatures of living benthic foraminifera from methane-rich environments will help to improve reconstructions of methane release in the past and better predict the impact of future climate warming on methane seepage. Here, we present isotopic signatures (δ13C and δ18O) of foraminiferal calcite together with biogeochemical data from Arctic seep environments from c. 1200 m water depth, Vestnesa Ridge, 79° N, Fram Strait. Lowest δ13C values were recorded in shells of Melonis barleeanus, − 5.2‰ in live specimens and − 6.5‰ in empty shells, from sediments dominated by aerobic (MOx) and anaerobic oxidation of methane (AOM), respectively. Our data indicate that foraminifera actively incorporate methane-derived carbon when living in sediments with moderate seepage activity, while in sediments with high seepage activity the poisonous sulfidic environment leads to death of the foraminifera and an overgrowth of their empty shells by methane-derived authigenic carbonates. We propose that the incorporation of methane-derived carbon in living foraminifera occurs via feeding on methanotrophic bacteria and/or incorporation of ambient dissolved inorganic carbon.

Fossil benthic foraminifera are used to trace past methane release linked to climate change. However, it is still debated whether isotopic signatures of living foraminifera from methane-charged sediments reflect incorporation of methane-derived carbon. A deeper understanding of isotopic signatures of living benthic foraminifera from methane-rich environments will help to improve reconstructions of methane release in the past and better predict the impact of future climate warming on methane seepage. Here, we present isotopic signatures (δ 13 C and δ 18 O) of foraminiferal calcite together with biogeochemical data from Arctic seep environments from c. 1200 m water depth, Vestnesa Ridge, 79° N, Fram Strait. Lowest δ 13 C values were recorded in shells of Melonis barleeanus, − 5.2‰ in live specimens and − 6.5‰ in empty shells, from sediments dominated by aerobic (MOx) and anaerobic oxidation of methane (AOM), respectively. Our data indicate that foraminifera actively incorporate methane-derived carbon when living in sediments with moderate seepage activity, while in sediments with high seepage activity the poisonous sulfidic environment leads to death of the foraminifera and an overgrowth of their empty shells by methane-derived authigenic carbonates. We propose that the incorporation of methane-derived carbon in living foraminifera occurs via feeding on methanotrophic bacteria and/or incorporation of ambient dissolved inorganic carbon.
One of the consequences of the ongoing climate warming is an increase in ocean temperature 1 . The Arctic is already warming about twice as fast as the global average, because of a process called 'the polar amplification' caused by decline in sea-ice cover and increased atmospheric heat transport from the equator to the Arctic. As large amounts of methane are stored on Arctic continental margins in the form of gas hydrates (pressure-temperature sensitive methane captured in ice [2][3][4] ), concern has increased that ongoing ocean warming will trigger destabilization of the gas hydrate reservoirs and cause further release of methane in the future 1,3,5,6 . Because methane is a ~ 25 times more potent greenhouse gas than CO 2 , a significant increase in the atmosphere can cause further amplification of the global warming. In the geological past, methane released from marine reservoirs has been suggested to be linked to paleoclimatic and palaeoceanographic changes during the Quaternary 7 , Late Paleocene 8 , the Cretaceous 9 and also been linked to the Permian-Triassic extinction event 10 . In methane-rich environments such as cold seeps, the carbon pool available for benthic foraminifera is enriched in inorganic methane-derived CO 2 and HCO 3 − , and organic carbon in the form of methane-related microbial communities characterized by low δ 13 C values.
It has been hypothesized that benthic foraminifera are able to record past methane seepage events by incorporating the low δ 13 C values derived from methane into their shells (called tests), and that they thus have a high potential to record variations in past methane release from the seabed 11,12 . Although the δ 13 C signatures of benthic foraminifera are a widely used proxy in paleoceanography to reconstruct past ocean circulation and productivity [13][14][15] , it is still disputed how methane-derived carbon enters foraminiferal shells, which might be via consumption of 13 C-depleted microbes, the presence of microbial symbionts 16,17 , active incorporation of dissolved

Results
Sediment biogeochemistry. Sediments of the Siboglinidae field MUC 10 showed strong indications for bio-irrigation by the tubeworms (Fig. 2a,b). Sulfate (~ 28 mmol L −1 ), total alkalinity (~ 3 mmol L −1 ), sulfide (< 0.2 mmol L −1 , except 3 mmol L −1 at 3-4 cm), and methane concentrations (< 0.1 mmol L −1 ) remained relatively unchanged in the topmost 4-6 cm. Irrigation was further suggested by the bright brown coloring of the top  23 ). Red dots indicate multicorer locations: Siboglinidae field (MUC 10), bacterial mats field (MUC 12), and control site (MUC 11). This figure is original, made using ArcMap v10.6. https:// www. esri. com/ en-us/ arcgis/ produ cts/ arcgis-deskt op/ overv iew".  S1) indicative of oxidized sediment. Below 4-6 cm, sulfate decreased while total alkalinity, sulfide, and methane increased (Fig. 2a,b). Sulfate declined to a minimum of 16.8 mmol L −1 at the bottom of the core, while total alkalinity and methane increased to 18.5 and 1.1 mmol L −1 , respectively. Sulfide peaked with 7 mmol L −1 at 9 cm and then declined with depth to reach 1.5 mmol L −1 at the bottom of the core. Accordingly, sediment color changed to black and (deeper in the core) grey indicating reducing conditions (Fig. S1a,b). In all three replicates, the majority of methane oxidation occurred in the top 4 cm of the sediment with rates up to 196 nmol cm −3 d −1 in the top (0-1 cm) sediment layer (Fig. 2c). This activity showed no match with sulfate reduction (Fig. 2d), neither in the profile, nor in magnitude, and suggests that it was coupled to MOx. Methane oxidation reached a minimum (~ 0.4 nmol cm −3 d −3 ) at 5-6 cm, below which rates increased again (see insert in Fig. 2c) reaching a maximum of 4.8 nmol cm −3 d −1 at 7-8 cm (Fig. 2c). The double peaking of methane oxidation suggests a change from an aerobic to an anaerobic methane oxidation pathway likely coupled to sulfate reduction below 6 cm, i.e., below the bio-irrigation activity of the tubeworms. Methane oxidation declined below the second peak to values ~ 1 nmol cm −3 d −3 at the bottom of the core. Sulfate reduction was low (< 3 nmol cm −3 d −3 ) in the top 0-1 cm, but steeply increased in all three replicates reaching values between 11 and 23 nmol cm −3 d −3 at 2-3 cm (Fig. 2d). Below 3 cm, sulfate reduction steadily declined reaching values ~ 1 nmol cm −3 d −3 at 10 cm, which remained consistently low down to the bottom of the core. The decoupling of methane oxidation and sulfate reduction in the surface sediment suggest that sulfate reduction was coupled to organic matter degradation in the top 6 cm, while part of it was likely also coupled to anaerobic methane oxidation (AOM) below 6 cm. The sediment from the bacterial mat field (MUC 12) showed steep geochemical gradients in the top 3-4 cm of the sediment: pore water sulfate and sulfide concentration declined from 28 to 2 and from 6.5 to 0.8 mmol L −1 , respectively, while total alkalinity increased from 2.5 to 35 mmol L −1 (Fig. 2e,f). Methane peaked with concentrations ~ 11 mmol L −1 at 2-4 and 28.5 cm and varied between 2 and 5 mmol L −1 in other depths with no clear trend (Fig. 2f). It is likely that measured concentrations were below in-situ levels and that the true methane profile was blurred due to degassing after sample recovery from depth. Degassing was clearly noticeable during core handling (Fig. S1c,d). Methane oxidation was low at the surface (< 13 nmol cm −3 d −1 ) and steeply increased in all three replicates to a maximum of up to 181 nmol cm −3 d −1 between 2 and 5 cm (Fig. 2g). Below the peaks, www.nature.com/scientificreports/ methane oxidation in all three replicates sharply declined and reached values around 1-4 nmol cm −3 d −1 below 7 cm. Profiles of all three sulfate reduction samples showed a general alignment with methane oxidation (Fig. 2h), suggesting a coupling to AOM. However, sulfate reduction was about two times higher than methane oxidation in the surface sediment (maximum 408 nmol cm −3 d −1 ) and therefore likely also coupled to other processes, most reasonably organic matter degradation.

Discussion
Siboglinidae field MUC 10-moderate methane seepage. Biogeochemical data of the pore water from the Siboglinidae field (MUC 10) indicate moderate methane seepage activity 19,35,36 . The bio-irrigation activity of the Siboglinidae tubeworms might cause oxidation of the top layer of the sediment 37 and potentially presence of free oxygen, resulting in the consumption of methane by aerobic methanotrophic microorganisms (top 4 cm of the sediment; Fig. 2). As a result of aerobic methane oxidation (MOx), the pore water is likely enriched in methane-derived CO 2 and HCO 3 -38 and in microbial biomass, which provides a carbon source for the benthic foraminifera during construction of their tests by biocalcification. To build their tests, benthic foraminifera use carbon from both the ambient DIC pool and intracellular storage (i.e., resulting from respiration and diet 39,40 ). Consequently, isotopically light carbon is likely incorporated by the benthic foraminifera, not only as an inorganic carbon from pore water, but also via nutrition (i.e., by consumption of methanotrophic microbes). Presumably, during biocalcification CO 2 is preferred as CO 2 diffuses more efficiently across cell membranes compared to HCO 3 − and/or CO 3  . Irrespective of that, benthic foraminifera have at times been shown to reach low δ 13 C values (down to − 6.9‰ in Brizalina pacifica) even in non-seep environments 41 , the fact that δ 13 32 . In this very active 'Lomvi' pockmark, studies have shown that methane transport often occurs via mini-fractures, and it is speculated that the gas can escape without affecting the foraminifera 43 . The δ 13 C of RB-stained C. neoteretis from the Siboglinidae field showed fairly negative values (from − 1.4‰ to − 1.8‰; MUC 10A and -B) compared to the non-seep site MUC 11A, and -B (− 0.3‰) (Fig. 4), and to other non-seep sites from the Arctic Ocean (from − 0.3‰ to − 1‰) 42 . The δ 13 C values of C. neoteretis from this study are also considerably lower compared to previously published values (− 0.3‰) from the active 'Lomvi' pockmark 32 (see above).
The δ 13 C of calcareous benthic foraminifera is determined by vital effects i.e., species-specific intracellular metabolic processes 34,44 and biogeochemical conditions of their microhabitat including organic matter and dissolved inorganic carbon content 45,46 . Vital effects can cause differences in the δ 13 C values of ~ 1-2‰ between specimens from the same species. In the present study, δ 13 C vary from − 5.2 to − 1.3‰ between specimens of RB-stained M. barleeanus from the Siboglinidae field MUC 10 and from the control site MUC 11. The differences between RB-stained specimens of C. neoteretis range from − 0.3 to − 1.8‰. These large differences in δ 13 C values between methane-influenced sites and non-methane sites greatly exceeds values of vital effects and thus cannot be attributed to vital effects alone. We suggest that the major factor controlling the δ 13 C in the foraminiferal tests of RB-stained M. barleeanus and C. neoteretis in the seep samples from the Siboglinidae field MUC 10 comes from microhabitat effects related to presence or absence of methane.
The δ 13 C of RB-stained C. wuellerstorfi from the Siboglinidae field vary between 0.1 and 1.1‰ (MUC 10A and -B), which is similar to the δ 13 C values of their conspecifics from the control site (from 1.1 to 1.2‰; MUC 11A and -B), and within the range of 'normal' values for the species 21 . Thus, there is no considerable influence of carbon-derived methane on their isotopic signature. This 'normal' carbon isotopic signature is probably related to the epibenthic lifestyle of C. wuellerstorfi. The species tends to attach itself to structures extending above the www.nature.com/scientificreports/ seafloor, e.g., tubes of Siboglinidae worms 21,47,48 . They do so to avoid hostile environmental conditions, such as oxygen depletion and toxicity of sulfide, both common at cold seeps 48,49 (Fig. 2). In the Siboglinidae field, samples showed specimens of C. wuellerstorfi attached to Siboglinidae tubes (Fig. 8). However, due to the absence of hostile environmental conditions, we suggest that C. wuellerstorfi is more likely attached to the tubes to support its filter-feeding behavior.
A laboratory culturing experiment performed by Wollenburg et al. 50 showed that artificially injecting 13 C-enriched methane to the water altered not only the δ 13 C signatures of the ambient DIC pool, but also the δ 13 C of the foraminiferal offspring of the epifaunal species C. wuellerstorfi and the shallow infaunal species C. neoteretis. These findings indicate that the δ 13 C values of C. wuellerstorfi can become more negative with low δ 13 C DIC in the ambient water. The experiments resulted in mean δ 13 C values of − 1.4‰ for C. wuellerstorfi and − 2.2‰ for C. neoteretis under controlled culturing conditions 50 . Since the δ 13 C measured in C. neoteretis (offspring) has values similar to those obtained from an in-situ reference site from the Håkon Mosby Mud Volcano 21 , the authors suggested that the δ 13 C of this shallow infaunal species mainly reflect their dietary preferences, i.e., feeding on bacteria 50 , whereas the epifaunal species C. wuellerstorfi reflects the δ 13 C of the bottom water DIC 50 . Thus, the δ 13 C signatures of M. barleeanus, C. neoteretis, and C. wuellerstorfi from the same sample might reflect the different microhabitat preferences of these species. Foraminifera that calcify deeper in the sediment (intermediate infaunal and deep-infaunal species), as for example M. barleeanus, often have low isotopic values when compared to shallow infaunal or epifaunal species, such as C. neoteretis or C. wuellerstorfi, respectively 44,45 . Our results from the Siboglinidae field (MUC 10) indicate that the proportion of carbon from methane in the ambient bottom water was not sufficient to considerably affect the isotopic signature of the epifaunal C. wuellerstorfi in comparison to the species that live deeper in the sediment, i.e., M. barleeanus and C. neoteretis that are more susceptible to the effects of methane. These infaunal species are probably affected by feeding on the 13 C-depleted methanotrophic microbial communities within the sediment and from incorporating 13 C-depleted DIC from the pore water during calcification. Food sources of 13 C-depleted microbes (archaea, bacteria) can contribute to up to a 5 to 6‰ decrease of the δ 13 C values of foraminiferal tests at seep sites 22 . Bacterial mat field MUC 12-strong methane seepage. It has been suggested that benthic foraminifera do not calcify in environments influenced by strong methane seepage with hostile concentrations of H 2 S 51 and consequently their δ 13 C values do not record methane seepage. Our results on the pore water biogeochemistry of the bacterial mat field MUC 12 indicate a strong methane seepage regime, which is dominated by AOM and sulfate reduction and features high concentrations of H 2 S just below the sediment surface (Fig. 2). Although foraminifera have a high tolerance to short-term exposure to H 2 S (up to 21 days), the prolonged exposure to H 2 S  Table S2), and even − 6.2‰ for C. wuellerstorfi (Fig. 5). These values are considerably lower compared to δ 13 C values of empty tests from its conspecifics from the control site MUC 11A and -B, which show 'normal' values, and are also much lower than in empty foraminiferal shells from the Siboglinidae field MUC 10A and -B (Figs. 3, 4, 5). We assume the low δ 13 C values are related to the formation of MDAC related to AOM, which may severely overprint the initial isotopic signatures of the foraminiferal tests and further indicates that the process is of minor importance or even absent at the Siboglinidae field MUC 10B. In support of this hypothesis, the SEM investigation of the planktonic foraminiferal species N. pachyderma from the bacterial mats field MUC 12B revealed signs of authigenic precipitation of carbonate on the outer surface (Fig. 7c,d), and the δ 13 C signature is considerably more negative (− 0.9 to − 4.2‰) (Table S2) compared to 'normal' values of the species in surface water environments (− 0.5‰) 53 . AOM is a strong contributor to authigenic carbonate overgrowth due to its production of HCO 3 − and increase in alkalinity 12,18 (and references therein). Overprinting by authigenic carbonate on foraminiferal shells can cause a lowering of the δ 13 C values of > 10‰ 12,18 . In the bacterial mat field MUC 12 samples, overgrowth was detected at relatively shallow sediment depth, i.e., 2-3 cm below the sediment surface (MUC 12A). Hence, the measured low values are most likely the result of a minor degree of a very recent overgrowth and are therefore less depleted in 13 C compared to values previously recorded in Vestnesa Ridge studies 12,18 .
The δ 18 O ratio in calcareous foraminiferal tests is influenced by several factors, including bottom water temperature, isotopic composition of the ambient seawater, and vital effects 54 , but can also be changed by diagenetic There was no such pattern in C. neoteretis from the Siboglinidae field MUC 10 or the control site MUC 11 (Fig. 6). We suggest that the higher 18 (Table S2). In accordance with these observations, authigenic carbonates from seep sites in the 'Lomvi' pockmark also displayed relatively high δ 18 O values (4.5 to 5.9‰) 12 . Given the more positive values, C. neoteretis might have a high predisposition for authigenic overgrowth, likely due to its test structure 18 , which may explain why only this species showed higher δ 18 O values compared to other species from the same samples.

Conclusion
The δ 13 C values measured in both RB-stained benthic foraminifera and empty tests of both planktonic and benthic foraminifera from Vestnesa Ridge together with biogeochemical datasets of pore water conditions showed a large degree of variation between different habitats (Siboglinidae field, bacterial mat field, and control site). At the Siboglinidae field MUC 10 with moderate seepage of methane, dominance of aerobic methane oxidation (MOx), and low concentrations of sulfide, live benthic foraminifera (RB-stained) incorporate methane-derived carbon. We propose that methane derived carbon was incorporated via feeding on methane-oxidizing bacteria and/or by direct intake of CO 2 in dissolved inorganic carbon produced from MOx. The effect, however, differed between species: the epifaunal species Cibicidoides wuellerstorfi appeared to be less susceptible to methane influence, while the intermediate infaunal species Melonis barleeanus responded more strongly by reaching δ 13 C values down to − 5.2‰. In sediments from the bacterial mat field MUC 12 with strong methane seepage, high activity of anaerobic oxidation of methane and sulfate reduction produced high levels of sulfide and total alkalinity, which killed living specimens and lead to the lowest δ 13 C values recorded in dead specimens due to postmortem MDAC overgrowth, respectively. Overgrowth may have started the coating of the fossil foraminiferal tests at relatively shallow depth in the sediment (2-3 cm), causing δ 13 C signature shifts of tests towards low values (down to − 6.5‰ for fossil M. barleeanus). Higher δ 18 O values in fossil C. neoteretis (5.1‰) from the bacterial mat field MUC 12 combined with low δ 13 C values (− 6.2‰) also indicate MDAC coatings of their tests. Fossil records derived from benthic foraminifera thus reflect the cumulative history of methane seepage covering the lifespan of the organisms, during which methane-derived carbon may be incorporated, as well as postmortem processes, such as shell overgrowth by MDAC. Therefore, in the context of palaeoceanographic studies, the use of δ 13 C signatures from foraminiferal shells as a paleo-methane indicator requires the consideration of MDAC coatings to separate between processes occurring during and after the lifetime of a benthic foraminifera.

Methods
Sediment sampling. Sediment samples were collected from a pockmark on Vestnesa Ridge, NW Svalbard margin in August 2011 during the POS419 expedition of the RV Poseidon. Using a TV-guided multicorer, sediment samples were taken from a Siboglinidae field (i.e. sediments covered by chemosymbiotic tubeworms), from www.nature.com/scientificreports/ a sulfur-bacterial mat field, and from far outside of the pockmark as a control site, where no methane seepage occurs (Table S1). The TV-guided multicorer system enables visual localization of active methane seeps based on the presence of cold-seep related structures, such as bacterial mats and authigenic carbonate crusts for targeted, designated sampling spots. The multicorer collected 6 cores of 10 cm in diameter at each location. After recovery of the multicorer, two cores (labelled A, B) were selected from each site for the study of foraminifera and subsampled onboard into 1-cm thick horizontal slices down to 10 cm core depth. The samples were transferred into plastic containers, and stained with Rose Bengal-ethanol solution following the FOBIMO protocol (2 g/L) 57 . Samples were kept onboard in a dark cool room at + 4 °C until further processing. A third core was sectioned in 1, 2, 3, and 5 cm increments (from top to bottom) for sediment pore water analyses. A fourth core was sectioned in 1, 2, 3, and 5 cm increments (from top to bottom) for sediment methane analyses. A fifth core was subsampled with a total of six mini polycarbonate cores (inner diameter 26 mm, length 30 cm) for the determination of methane concentration, methane oxidation, and sulfate reduction. All sediment sampling procedures were conducted at + 4 °C inside an environmental room.
Pore water analyses. Pore water was extracted onboard at + 4 °C using a low-pressure squeezer (argon at 1-5 bar). While squeezing, pore water was filtered through 0.2 μm cellulose acetate nuclepore filters and collected in argon-flushed recipient vessels. Pore water samples were subjected to geochemical analyses for total sulfides, total alkalinity, sulfate-and methane concentrations.
Sulfide, total alkalinity and sulfate measurements. Onboard, the collected pore water samples were analyzed for their content of dissolved total sulfides (in the following referred to as "sulfide") 58 . A 1 mL sample was added to 50 μL of zinc acetate solution. Subsequently, 10 μL of N,N-dimethyl-1,4-phenylenediamine-dihydrochloride color reagent solution and 10 μL of the FeCl 3 catalyst were added and mixed. After 1 h of reaction time, the absorbance was measured at 670 nm. Total alkalinity (TA) was determined by direct titration of 1 mL pore water with 0.02 M HCl using a mixture of methyl red and methylene blue as an indicator and bubbling the titration vessel with argon gas to strip CO 2 and hydrogen sulfide. The analysis was calibrated using IAPSO seawater standard, with a precision and detection limit of 0.05 mmol L −1 . Pore water samples for sulfate (SO 4 2− ) analyses were stored in 2-mL glass vials at + 4 °C and analyzed onshore. Sulfate was determined by ion chromatography (Metrohm, IC Compact 761). Analytical precision based on repeated analysis of IAPSO standards (dilution series) was < 1%.
Methane measurements. According to Sommer et al. 59 methane concentrations in sediment cores were determined in 1-cm intervals down to a depth of 6 cm followed by 2-cm intervals down to 12 cm, 3-cm intervals down to 18 cm and 5-cm intervals deeper than 18 cm. From each depth horizon, a 2-mL sub-sample was transferred into a septum-stoppered glass vial (21.8 mL) containing 6 mL of saturated NaCl solution and 1.5 g of NaCl in excess. The volume of headspace was 13.76 mL. Within 24 h, the methane concentration in the headspace was determined using a Shimadzu GC 14A gas chromatograph fitted with a flame ionization detector and a 4-m × 1⁄8-in. Poraplot Q (mesh 50/80) packed column. Prior to the measurements the samples were equilibrated for 2 h on a shaking table. Precision to reproduce a methane standard of 9.98 ppm was 2%.
Microbial methane oxidation rates. On board, radioactive methane ( 14 CH 4 dissolved in water, injection volume 15 µL, activity ~ 5 kBq, specific activity 2.28 GBq mmol −1 ) was injected into three replicate mini cores at 1-cm intervals according to the whole-core injection method 60 . The mini cores were incubated at in-situ temperature for ~ 24 h in the dark. To stop bacterial activity, the sediment cores were sectioned into 1-cm intervals and transferred into 50-mL crimp glass vials filled with 25 mL sodium hydroxide (2.5% w/w). After crimpsealing, glass vials were shaken thoroughly to equilibrate the pore-water methane between the aqueous and gaseous phase. Control samples were first terminated before addition of tracer. In the home laboratory, methane oxidation rates and methane concentrations in the sample vials were determined according to Treude et al. 61 .
Microbial sulfate reduction rates. Sampling, injection, and incubation procedures were identical to methane oxidation samples. The injected radiotracer was carrier-free 35 SO 4 2− (dissolved in water, injection volume 6 µL, activity 200 kBq, specific activity 37 TBq mmol −1 ). To stop bacterial activity after incubation, sediment cores were sectioned into 1-cm intervals and transferred into 50 mL plastic centrifuge vials filled with 20 mL zinc acetate (20% w/w) and frozen. Control sediment was first terminated before addition of tracer. In the home laboratory, sulfate reduction rates were determined according to the cold-chromium distillation method 62 .
Foraminiferal analyses. Rose-Bengal stained samples were sieved over a 100-µm sieve. The > 100-µm fraction was kept wet and further examined under reflected-light microscopy. All benthic foraminiferal individuals that stained dark magenta and were fully filled with cytoplasm were considered to be 'living' foraminifera i.e., live + recently dead individuals, still containing cytoplasm. Foraminifera showing no colorization were considered as unstained, empty (dead) individuals. The foraminifera were wet picked, sorted by species and placed on micropaleontology slides.
Isotope analyses. For carbon (δ 13 C) and oxygen (δ 18 O) stable isotope analyses, both Rose Bengal stained and unstained (empty) specimens of benthic foraminiferal species Melonis barleeanus, Cassidulina neoteretis and Cibicidoides wuellerstorfi, and empty specimens of the planktic foraminiferal species Neogloboquadrina pachyderma were picked. When present, approximately 10 specimens of each species were taken from each www.nature.com/scientificreports/ sample. Only pristine and transparent, and clean tests were picked. Empty tests were obtained from the same samples as the Rose-Bengal stained foraminifera. No replicate measurements for isotope ratios were made due to low amounts of foraminiferal material available. Samples were cleaned using pure ethyl alcohol in an ultrasonic bath, following the protocol from Sztybor and Rasmussen 12 Isotopic measurements were performed at the Isotope Geochemistry Facility at Woods Hole Oceanographic Institution (WHOI). Data are reported in standard notation (δ 13 C, δ 18 O), according to the Pee Dee Belemnite (PDB) standard. Reported precision was estimated to be ± 0.07‰ for δ 13 C and ± 0.15‰ for δ 18 63 . Neogloboquadrina pachyderma, two specimens from intervals 0-1 cm and 4-5 cm from each core were selected and investigated using Scanning Electron Microscopy (SEM). To make our data comparable with other studies on live foraminifera, we deliberately used the most widely used staining method, the Rose Bengal. Since Rose Bengal indicates both live and dead cytoplasm, even weeks to months after the death of an individual 32,64 , we nevertheless refer here to Rose Bengal stained foraminifera as 'live' specimens, and empty, unstained tests as dead specimens.