Foraminiferal community response to seasonal anoxia in Lake Grevelingen (the Netherlands)

. Over the last decades, hypoxia in marine coastal environments has become more and more widespread, prolonged and intense. Hypoxic events have large consequences for the functioning of benthic ecosystems. In severe cases, they may lead to complete anoxia and presence of toxic sulphides in the sediment and bottom-water, thereby strongly affecting biological compartments of benthic marine ecosystems. Within these ecosystems, benthic foraminifera show a high diversity of 20 ecological responses, with a wide range of adaptive life strategies. Some species are particularly resistant to hypoxia/anoxia and consequently, it is interesting to study the whole foraminiferal community as well as species specific responses to such events. Here we investigated

, both in the water column and in the sediment, with a very strong pelagic/benthic coupling (de Vries and Hopstaken, 1984). The benthic environment is characterised by the presence of two antagonistic groups of bacteria, with contrasting seasonal population dynamics (i.e. cable bacteria in winter/spring and Beggiatoaceae in autumn/winter), which have a profound impact on all biogeochemical cycles in the sediment column Sulu-Gambari et al., 2016a. 70 The combination of hypoxia/anoxia with sulphidic conditions, which is rather unusual in coastal systems without external nutrient input, and the activity of antagonistic bacterial communities make Lake Grevelingen a very peculiar environment. In the Den Osse Basin, seasonal anoxia coupled with the presence of H2S at or very close to the sediment-water interface occurs in summer (i.e. between July-September). However, euxinia (i.e. diffusion of free H2S in the water column) does not occur, because of cable bacterial activity . 75 Although the tolerance of foraminifera to low DO contents and long term anoxia (from weeks to 10 months) has been well documented for many species from different types of environments in laboratory culture (e.g. Moodley and Hess, 1992;Alve and Bernhard, 1995;Bernhard and Alve, 1996;Moodley et al., 1997;Duijnstee et al., 2003;Geslin et al., 2004;Ernst et al., 2005;Pucci et al., 2009;Koho et al., 2011; as well as in field studies (e.g. Piña-Ochoa et al., 2010b ;Langlet et al., 2013;2014), their tolerance to free H2S is still debated. In the vast majority of previous 80 studies, no decrease in the total abundances of living foraminifera (i.e. strongly increased mortality) was observed during anoxic events. Unfortunately, studies on foraminiferal response in systems affected by seasonal hypoxia/anoxia with sulphidic conditions are still very sparse. The few available observations are not conclusive, but suggest that H2S could be toxic for foraminifera even on fairly short time scales (Bernhard, 1993;Moodley et al., 1998b;Panieri and Sen Gupta, 2008;. 85 To our knowledge, all earlier studies show that the foraminiferal response to hypoxia/anoxia is species-specific (e.g. Bernhard and Alve, 1996;Ernst et al., 2005;Bouchet et al., 2007;Langlet et al., 2014). However, this species-specific response generally follows the same scheme (usually decrease in density, reduction of growth and/or reproduction), with different response intensities. Duijnstee et al. (2005) suggested that oxic stress leads to an increased mortality and an inhibited growth and reproduction. The suggestion of inhibited growth is supported by LeKieffre et al. (2017) who observed that the 90 morphospecies Ammonia tepida (probably Ammonia sp. T6) showed minimal or no growth under anoxia. Conversely, Geslin et al. (2014) and Nardelli et al. (2014) suggested that, in the same morphospecies, reproduction was strongly reduced, but growth would not be affected by hypoxic and/or short anoxic events. Additionally, under low-oxygen conditions, some species are able to shift to anaerobic metabolism (i.e. denitrification, Risgaard-Petersen et al., 2006;Piña-Ochoa et al., 2010a), to sequester chloroplast (i.e. kleptoplastidy, Jauffrais et al., 2018), to associate with bacterial symbionts (Bernhard et al., 2010) 95 or to enter into a state of dormancy (Ross and Hallock, 2016;LeKieffre et al., 2017).
The highly peculiar environmental context of Lake Grevelingen offers an excellent opportunity to study this still poorly known aspect of foraminiferal ecology.
The conventional method to discriminate between live and dead foraminifera uses Rose Bengal, a compound which stains proteins (i.e. organic matter). This method was proposed for foraminifera by Walton (1952) and is based on the assumption 100 that "the presence of protoplasm is positive indication of a living or very recently dead organism". The author already noted that this assumption implied that the rate of degradation of organic material should be relatively high. Previous studies of living benthic foraminifera in environments subjected to hypoxia/anoxia were almost all based on Rose Bengal stained samples (e.g. Nordberg, 1999, 2000;Duijnstee et al., 2004;Panieri, 2006;Schönfeld and Numberger, 2007;Polovodova et al., 2009;Papaspyrou et al., 2013). However, foraminiferal protoplasm may remain stainable from several weeks to months 105 after their death (Corliss and Emerson, 1990), especially under low dissolved oxygen concentrations where organic matter degradation may be very slow (Bernhard, 1988;Hannah and Rogerson, 1997;Bernhard et al., 2006). The Rose Bengal staining method is therefore not suitable for studies in environments affected by hypoxia/anoxia. Consequently, the results of foraminiferal studies in low-oxygen environments based on this method have to be considered with reserve. In order to avoid this problem, we used CellTracker™ Green (CTG) to recognise living foraminifera. CTG is a fluorescent probe which marks 110 only living individuals with cytoplasmic (i.e. enzymatic) metabolic activity (Bernhard et al., 2006). Since metabolic activity stops after the death of the organism, CTG should give a much more accurate assessment of the living assemblages at the various sampling times, and thereby avoid over-estimation of the live foraminiferal abundances.
In this study, samples were collected in August and November 2011 and then every month through the year 2012, at two different stations in the Den Osse Basin, with two replicates dedicated to foraminifera. The two stations were chosen in 115 contrasted environments regarding water depth (34 m and 23 m, respectively) and duration of seasonal hypoxia/anoxia and sulphidic conditions. Living foraminiferal assemblages were studied in the uppermost sediment and size distributions were determined in order to get insight into the possible moment(s) of reproduction or accelerated growth in test size. The seasonal variability study of the foraminiferal community allows us (1) to better understand the foraminiferal tolerance to seasonal hypoxia/anoxia with presence of free H2S in their microhabitat and (2) to obtain information about the responses of the various 120 species to adverse conditions. This knowledge will be useful for the development of indices assessing environmental quality (i.e. biomonitoring) and may also improve paleoecological interpretations of coastal records (e.g. Murray, 1967;Gustafsson and Nordberg, 1999).

Studied areaenvironmental settings in the Den Osse Basin. 125
Lake Grevelingen is a part of the former Rhine-Meuse-Scheldt estuary, in the southwestern Netherlands. This former estuarine branch was turned into an artificial saltwater lake during the Delta Works project. In Lake Grevelingen, the water circulation is strongly limited by the construction of dams (in the early 1970s) and only a small sluice allows water exchanges with open sea waters (i.e. very weak hydrodynamics). In the Lake, development of bottom-water hypoxia/anoxia occurs in the deepest part of the basin in summer (i.e. July-September) to early autumn (i.e. October-December, Bannink et al., 1984;Hagens et 130 al., 2015). In the literature, the terminology and threshold values used to describe oxygen depletion are highly variable (e.g., oxic, dysoxic, hypoxic, suboxic, microxic, postoxic; see Jorissen et al., 2007;Altenbach et al., 2012). In this study we defined hypoxia as a concentration of oxygen <63 µmol L -1 (1.4 mL L -1 or 2 mg L -1 ) whereas anoxia is defined as no detectable oxygen (following Rabalais et al., 2010).
In Den Osse Basin, the nutrient input from external sources is very low and pelagic/benthic coupling is essential, as already 135 noted by de Vries and Hopstaken (1984). In 2012, phytoplankton blooms occurred in April-May and July (Hagens et al., 2015) in response to the increasing solar radiation and nutrient availability in the water column following organic matter recycling in winter. This led to an increased food availability in the benthic compartment in the same periods. In general, Chl a concentrations in Den Osse Basin are below 10 µg L -1 , excluding very short peaks during blooms in April-May and July which did not exceed 30 µg L -1 in 2012 (Hagens et al., 2015). Thermal stratification of the water column and increased oxygen 140 consumption due to organic matter input (i.e. from phytoplankton blooms) both are responsible for the development of seasonal bottom-water hypoxia/anoxia in summer (i.e. July-September). Although euxinia (i.e. the presence of free H2S in the water column) does not occur in the Den Osse Basin due to cable bacterial activity in winter, free H2S is present in the uppermost layer of the sediment in summer . Summarising, in the benthic ecosystem, increased food availability in summer is counterbalanced by strongly decreasing oxygen contents, sometimes accompanied by the presence of free sulphides 145 in the topmost sediment.

Field Sampling
The two studied sites are located along a depth gradient in the Den Osse Basin of Lake Grevelingen. Both station 1 (51°44.834' N, 3°53.401' E) and station 2 (51°44.956' N, 3°53.826' E) are located in the main channel, at 34 and 23 m depth, respectively ( Fig. 1). 150 Measurements of bottom-water oxygen (BWO) concentrations were performed at 2 m above the sediment-water interface and are from Donders et al. (2012), whereas the data for 2012 were published in Hagens et al. (2015). Sediment cores were collected monthly in 2012 using a single core gravity corer (UWITEC, Austria) using PVC core liners (6cm inner diameter, 60cm length). All cores were inspected upon retrieval and only visually undisturbed sediment cores were used for further analysis (Seitaj et al., 2017). Oxygen penetration depth (OPD) and depth of free H2S detection were determined by Seitaj et al., (2015) 155 using profiling microsensors for station 1. The data for station 2 (Supplementary Table 1) were acquired similarly and during the same cruises but never published, for further details about the sampling method, see Seitaj et al. (2015).
Two replicate sediment cores dedicated to the foraminiferal study were sampled in August and November 2011 using the same gravity corer (UWITEC, Austria) and then monthly throughout the year 2012 at the same sampling time as for BWO concentration and OPD and H2S measurements in the sediment (see Seitaj et al., 2015). Consequently, for 2012 at station 1 160 and 2, OPD and H2S were measured in the sediment column at the same time as foraminifera were sampled .
For each replicate, the uppermost centimetre (0-1 cm) of the core was then transferred on board in a vial of 250 mL, and 30 mL of seawater (at the same temperature than in situ) was added in the vial. Then we labelled the samples with CellTracker™ Green CMFDA (CTG, 5-chloromethylfluorescein diacetate, final concentration of 1µmol L -1 following Bernhard et al., 2006) and slowly agitated manually to allow the CTG diffusion in the whole sample. Samples were then fixed in 5 % sodium borate 165 buffered formalin after 24 h of incubation in the dark.

Sample Treatment
All samples were sieved over 315, 150, 125 µm meshes, and foraminiferal assemblages were studied in all three size fractions.
Individuals were picked wet under an epifluorescence stereomicroscope (Olympus SZX12, light fluorescent source Olympus URFL-T, excitation/emission wavelengths: 492 nm/517 nm) and placed on micropalaeontological slides. Only specimens that 170 fluoresced brightly green were considered as living and were identified to the (morpho-)species level when possible. Since picking foraminifera under an epifluorescence stereomicroscope is particularly time-consuming, we decided to study samples only every two months for the year 2012. At a later stage, in view of the large differences in foraminiferal abundances between the samples of September and November 2012 at station 2, we decided to study the October and December 2012 samples as well for this station. The sampling dates investigated in this study are listed in Table 1. 175 Abundances were then standardised to a volume of 10 cm 3 . The abundances of living foraminifera for each sampling time and replicate are listed in Supplementary Tables 2 and 3. The mean abundance and standard deviation ( ± ) for the two replicates for each sampling date were calculated both for the total living assemblage and the individual species, as an indication of spatial patchiness.
Concerning the genus Ammonia, two living specimens collected at Grevelingen station 1 were molecularly identified (by DNA barcoding) as phylotype T6 by Bird et al. (2019). At the same site, we genotyped seven other living Ammonia specimens, 185 which were all T6. Their sequences were deposited on GenBank (accession numbers MN190684 to MN190690) and Supplementary Figure 1 shows Scanning Electron Microscope (SEM) images of the spiral side and of the penultimate chamber at 1000x magnification for four individuals. A morphological screening based on the criteria proposed by Richirt et al. (2019) confirmed that T6 accounts for the vast majority (>98 %) of Ammonia individuals, whereas phylotypes T1, T2, T3 and T15 are only present in very small amounts (Supplementary Table 3). 190 The specimens of Elphidium magellanicum were identified exclusively on the basis of morphological criteria, as there are no molecular data available yet. This morphospecies, although rare, is regularly recognised in Boreal and Lusitanean provinces of Europe (e.g. Gustafsson and Nordberg, 1999;Darling et al., 2016;Alve et al., 2016). However, as the type species was described from the Magellan strait (Southern Chile), the European specimens may represent a different species and further studies involving DNA sequencing of both populations are needed to confirm or infirm this taxonomic attribution (see Roberts 195 et al., 2016). Elphidium selseyense has often been considered as an ecophenotype of Elphidium excavatum (Terquem, 1875) and has been identified as E. excavatum forma selseyensis (e.g. Feyling-Hanssen, 1972;Miller et al., 1982). Four living specimens were already sampled for DNA analysis at station 1 and were all identified as the species E. selseyense (phylotype S5, . We only observed minor morphological variations in our material, especially concerning the number of small bosses 200 in the umbilical region, which we considered as intraspecific variability. Consequently, we identified all our specimens as E. selseyense.
The specimens attributed to Trochammina inflata were also identified exclusively on the basis of morphological criteria, as no molecular data are available yet.

Size distribution measurement 205
In order to detect periods of increased growth and/or reproduction, size measurements were performed on all samples of 2012.
The measurements were made for all species (4176 individuals for station 1 and 19624 individuals for station 2) and trochospiral species were all orientated spiral side up prior to measurements. High-resolution images (3648*2736 pixels) of all micropalaeontological slides were taken with a stereomicroscope (Leica S9i, 10x magnification) and individual measurements were processed using ImageJ software (Schneider et al., 2012, Supplementary Figure 2). 210 Each individual was isolated (Supplementary Figure 2) and its maximum diameter was measured (i.e. Feret's diameter). We represented all size distributions using histograms with 20 µm classes (the best compromise between the total number of individuals and the size range (Supplementary Figure 3). As we only examined the size fractions >125 µm, our analysis mainly concerns adult specimens, and does not include juveniles. This limitation should be kept in mind when interpreting the results.
Assuming that the size distribution was a sum of Gaussian curves, each of them representing a cohort, we tried to identify the 215 approximate mode for the Gaussian curves (i.e. cohorts) using the changes in slope (i.e. inflexion points) of the second-order derivative of the total size distribution (Gammon et al., 2017). Unfortunately, this tentative to distinguish cohorts by using a deconvolution method was not conclusive. The main problem was the lack of information concerning individuals smaller than 125 µm, so that our size distributions were systematically skewed on the left side (i.e. toward small individuals). Because the identification of individual cohorts was not successful, a study of population dynamics was not possible. For this reason, the 220 data are only shown in the supplementary material ( Supplementary Figures 2 and 3). Nevertheless, the size distribution data give some clues concerning the possible moment(s) of reproduction or intensified test growth for the different species.

Encrusted forms of E. magellanicum
In our samples, we found abundant encrusted forms of E. magellanicum at station 1 (May 2012) and station 2 (May, July, September and December 2012, Fig. 8). Most individuals were totally encrusted (Fig. 8a), others only partly (Fig. 8b). These 225 crusts were hard, firmly stuck to the shell (difficult to remove with a brush), thin (Fig. 8c-e) and rather coarse. In order to determine if the crust matrix is constituted of carbonate, we placed some specimens in microtubes and exposed them to 0.1 M of EDTA (EthyleneDiamineTetraacetic Acid) diluted in 0.1 M cacodylate buffer (acting as a carbonate chelator). After an exposition of 24h, we checked under a stereomicroscope if the crust was still cohesive (no carbonate in the crust) or was disaggregated (crust contains carbonate). 230

Total abundances of foraminiferal assemblages
Averaged total abundances varied between 1.1 ± 1.5 and 449.9 ± 322.1 ind. 10 cm -3 for station 1, and between 91.1 ± 25.0 and 604.8 ± 3.5 ind. 10 cm -3 for station 2 ( Fig. 2 and Table 2). For every studied month, the total density was higher at station 2 than at station 1. The seasonal succession is very different between the two sites (Fig. 2). Station 1 shows very low total 235 foraminiferal abundances for most months, contrasting with much higher densities in May and July. Conversely, station 2 shows high total foraminiferal abundances throughout the year, with somewhat lower values in November 2011, and October and November 2012 (Fig. 2).

Dominant Species
At station 1, the major species were, in order of decreasing abundances, Elphidium selseyense ( Fig. 3a- Table 2). At station 2, the dominant species, in order of decreasing abundances, were E. selseyense, Ammonia sp. T6, E. magellanicum and T. inflata (Table 2). Here, "Other species" account only for 2.6 % of the total assemblage. Whereas E. selseyense and E. magellanicum were dominant species at both stations, both Ammonia sp. T6 and T. inflata were present in much higher abundances at station 2 compared to station 255 1, where the latter species was almost absent (Fig. 4-5).
At station 1, only some very scarce individuals of E. selseyense were observed in August and November 2011 ( Fig. 4 and Table 2). In 2012, E. selseyense abundances were very low in January started to increase in March ( = 23.9 ± 6.8) to reach maximal values in May ( = 336.5 ± 275.8). In July, values for E. selseyense were still high ( = 162.0 ± 121.5) and further decreased until an almost total absence in November 2012. No specimen of E. magellanicum was observed in 2011 ( Fig. 4 and 260 Table 2). The abundance of E. magellanicum was very low in January 2012, started to increase in March ( = 21.6 ± 11.0) to reach maximal values in May ( = 96.4 ± 47.3), then strongly decreased in July ( = 3.7 ± 0.3). The species was absent from samples in September and November 2012. Ammonia sp. T6 was almost absent in August and November 2011 and present with very few specimens in January 2012 ( = 3.2 ± 3.5). Maximum abundances were reached between March and July 2012 (ranging between = 9.2 ± 6.5 and = 12.9 ± 1.3). Then abundances rapidly decreased until the species was 265 almost absent in November. Trochammina inflata was absent in 2011 and was only present with very low abundances from January to May and in September 2012.
At station 2, the two dominant species were E. selseyense and Ammonia sp. T6, which together always represented at least 70 % of the total assemblage ( Fig. 5 and Table 2 in December ( = 25.5 ± 13.0).

Encrusted forms of Elphidium magellanicum
After exposition to 0.1 M of EDTA diluted in 0.1 M cacodylate buffer, the crusts remained cohesive, indicating that it does not consist of carbonate, and suggesting that it is composed of sediment particles cemented by an organic matrix.

Tolerance of foraminiferal communities to anoxia and free sulphide
At station 1, bottom-waters were hypoxic in July 2012 and became anoxic in August (Fig. 10). Both in July and August, 290 oxygen penetration into the sediment was null, whereas it was 0.7 ± 0.1 mm depth in September. In all three months (July to September 2012), sulphidic conditions were observed very close to the sediment-water interface (1 mm or less, Fig. 10 and Supplementary Table 1). In view of these results, the duration of anoxic and sulphidic conditions in the uppermost sediment layer can be estimated as one to two months (in July and August, Fig. 10).
After the strong increase of foraminiferal densities in May 2012, there was a decrease starting in July, leading to a near-absence 295 of foraminifera at station 1 in November (Fig. 10). The most probable cause of the strong decline of the foraminiferal community appears to be a prolonged presence of sulphides in the foraminiferal microhabitat. However, the fact that foraminiferal abundances reached almost zero only in September (about two months after the first occurrence of anoxic and sulphidic conditions in the upper sediment, in July) suggests that the presence of H2S did not cause instantaneous mortality, but that the disappearance of the foraminiferal community was a delayed response, probably caused by inhibited reproduction 300 and, eventually, increased mortality. Inhibited reproduction has previously been suggested as a response to hypoxic/short anoxic  and sulphidic conditions (Moodley et al., 1998b).

Such a time lag between a change in foraminiferal abundances and changes in environmental parameters affecting reproduction
and/or growth of foraminifera has been suggested previously by Duijnstee et al. (2004). These authors highlighted that the density patterns of some foraminiferal species showed a higher correlation with measured environmental parameters (e.g., 305 oxygenation or temperature) when a time lag of about three months was applied.
For 2011, at station 1, no pore-water O2 and H2S measurements are available. However, severe hypoxia was observed in the bottom-waters from May to August, with anoxia in June 2011 (Fig. 10). We therefore assume that like in 2012, anoxic and probably co-occurring sulphidic conditions were responsible for the very low standing stocks in August and November 2011 and January 2012. 310 Our observations confirm the suggestion in previous studies that the foraminiferal community is severely affected by a longterm presence of H2S in its habitat, but does not show instant mortality. In fact, after a 66-day incubation in euxinic conditions (a maximum of 11.9 ± 0.4 µmol L -1 of H2S in the overlying water) of foraminiferal assemblages collected at a 19 m deep site in the Adriatic Sea, Moodley et al. (1998a) found a strong decrease of the total density of Rose Bengal stained foraminifera.
After 21 days, living specimens were still observed, whereas after 42 and 66 days, the live checks (based on protoplasm 315 movement) gave only negative results. Langlet et al. (2013Langlet et al. ( , 2014, performed an in situ experiment with closed benthic chambers at a 24 m deep site in the Gulf of Trieste, in the Adriatic Sea. They observed a decrease of living foraminiferal density (labelled with CTG), but also found that almost all species survived after 10 months of anoxia and periodically cooccurring H2S in the sediment and overlying water. However, the duration of sulphidic conditions, which was estimated to several weeks, could not be assessed precisely . The suggestion that short-time exposure to euxinic 320 conditions is not directly lethal for foraminifera is confirmed by the experimental results of Bernhard (1993), who found that foraminiferal activity (as determined by ATP content) was not significantly affected after 30-day exposure to euxinia (32.6 ± 8.6 % of active individuals, n=174. in control conditions versus 29.5 ± 6.2 %, n= 173 in sulphidic conditions).
After the 2011 hypoxia/anoxia, standing stocks at station 1 only started to increase in March 2012, indicating a very long recovery time (about 6 months) of the foraminiferal faunas after a temporary near-extinction due to anoxic and sulphidic 325 conditions. This confirms observations of relatively long recovery times in the literature (e.g. Alve, 1995Alve, , 1999Gustafsson and Nordberg, 2000;Hess et al., 2005). For instance, Gustafsson & Nordberg (1999) showed that in the Koljö Fjord, at comparable water depths, foraminiferal populations responded with increased densities only three months after a renewal of sea-floor oxygenation following hypoxic conditions in the bottom-waters. However, in that case, the disappearance of the foraminiferal population was only partial, and not nearly complete as in our study. 330 At station 2, in 2012, hypoxia was only observed in August, when the OPD was zero, and sulphidic conditions were observed in the superficial sediment (i.e. from 0.4 ± 0.2 mm downwards, Fig. 11, Supplementary Table 1). Both in July and September, oxygen penetrated more than one millimetre into the sediment (1.3 ± 0.4 mm and 1.2 ± 0.2 mm, respectively). However, free H2S was still detected at about one millimetre depth in the sediment (1.1 ± 0.8 mm in July and 0.8 ± 0.2 mm in September). 335 Although the sampling plan does not allow us to be very precise about the duration of anoxic and sulphidic conditions, we can estimate their duration to be 1 month or less (Fig. 11).
Foraminiferal abundances showed a strong decrease in October and November 2012, about two months after the presence of anoxic and sulphidic conditions in the topmost part of the sediment (Fig. 11). Like at station 1, this temporal offset between the presence of anoxia/sulphidic conditions at station 2 (in August) and the strong decrease of faunal densities may be explained 340 as a delayed response, mainly due to inhibited reproduction during the anoxic/sulphidic event. If true, the mortality of adults did not strongly increase in the months following the H2S production in the uppermost sediment. Nevertheless, there was no replacement in the >125 µm fraction by growing juveniles, probably because reproduction was interrupted when H2S was present in the foraminiferal microhabitat. A renewed recruitment after the last stage of sulphidic conditions somewhere in September would then explain why the faunal density in the >125 µm fraction increased again in December 2012 345 (Supplementary Figure 3).
In 2011, at station 2, bottom-waters oscillated between hypoxic and oxic conditions between May and August (Fig. 11).
Although we have no measurements of H2S in the pore waters for this year, it seems probable that bottom-water hypoxia was accompanied by the presence of free H2S very close to the sediment surface, strongly affecting the foraminiferal communities.
If we assume that, like in 2012, rich foraminiferal faunas were present in May-July 2011 at both stations, the low faunal 350 densities observed in August and November 2011 could suggest that foraminifera may have also shown a delayed response to sulphidic conditions in 2011.
It is interesting to note that the foraminiferal densities observed at station 2 were lower in August 2011 than in July or September 2012. This may be a consequence of the repetition of short hypoxic events in the bottom-water between May and the foraminiferal community more substantially in 2011 than in 2012, when a hypoxic event was recorded in August only.
The important decrease of total standing stocks at station 2 in October and November 2012 (Fig. 11) suggests that, in spite of the shorter duration of anoxia and sulphide conditions (compared to station 1; one month or less compared to one to two months), the foraminiferal faunas were still strongly affected. However, at station 2, foraminiferal abundances increased again in December 2012, suggesting a recovery time of about two months, which is likely much shorter than at station 1, where 360 standing stocks in the >125 µm fraction only increased 6 months after the presence of anoxia and free sulphides.
Summarising, the foraminiferal communities of both stations 1 and 2 seem strongly impacted by the anoxic and sulphidic conditions developing in the uppermost part of the sediment in summer (i.e. July-September). However, at station 1, where anoxic and sulphidic conditions lasted for one to two months, the response is much stronger, leading ultimately (in November) 365 to almost complete disappearance of the foraminiferal fauna. The delayed response at both stations shows that instantaneous mortality was limited, and suggests that the decreasing standing stocks might rather be the result of inhibited reproduction, and eventually, increased mortality. Recovery is much faster at station 2 (about two months) than at station 1 (about six months), probably because at station 1 (in contrast to station 2) the foraminiferal extinction was nearly complete, and the site had to be recolonised (e.g. possibly by nearby sites or by the remaining few individuals) after reoxygenation of the sediment. 370 At station 2, a reduced but significant foraminiferal community remained present, explaining the faster recovery.

Species-specific response to anoxia, sulphide and food availability in Lake Grevelingen
The comparison of the different seasonal patterns of the major species at the two investigated stations allows us to draw some conclusions about interspecific differences in the response to seasonal anoxic and sulphidic conditions. First, there is a clear faunal difference between the two stations. Station 1 is dominated by E. selseyense and E. magellanicum 375 while at station 2, these two taxa are accompanied by Ammonia sp. T6 and T. inflata. The latter species is almost absent at station 1, where Ammonia sp. T6 is present with low densities. At first view, the dominance of the two Elphidium species at station 1, would suggest that they have a greater tolerance to the seasonal anoxic and sulphidic conditions, which lasted much longer there. It is interesting to note that the temporal evolution of standing stocks at station 1 is different for the two Elphidium species. Elphidium magellanicum shows a strong drop in absolute density in July 2012, at the onset of H2S presence in the 380 uppermost part of the sediment, whereas the diminution of E. selseyense is more progressive and the species disappears almost completely only in November (Fig. 4). This strongly suggests that E. magellanicum is more affected by increased mortality than E. selseyense in response to the combined effects of anoxic and sulphidic conditions. This hypothesis is confirmed by the patterns observed at station 2, where the drop in standing stocks in October-November is also more drastic in E. magellanicum than in E. selseyense (Fig. 5). 385 et al., 2006;Piña-Ochoa et al., 2010a), sequester chloroplasts (i.e. kleptoplastidy, Jauffrais et al., 2018), host bacterial symbiont (Bernhard et al., 2010) or enter in dormancy (Ross and Hallock, 2016;LeKieffre et al., 2017) to deal with low-oxygen conditions. Concerning the species found in this study, although the presence of intracellular nitrate was shown for Ammonia, 390 denitrification tests yielded negative results (Piña-Ochoa et al., 2010a;Nomaki et al. 2014). Similarly, the presence of active symbionts was previously suggested for Ammonia but never confirmed (Nomaki et al., 2016;Bernhard et al., 2018). To our knowledge, denitrification or the presence of bacterial symbionts was never shown for Elphidium either. In conclusion, a shift to an alternative anaerobic metabolism or an association with bacterial symbionts has never been shown conclusively for the dominant foraminiferal species found in Lake Grevelingen. 395 The greater tolerance of E. selseyense to low-oxygen conditions could be explained by the fact that it is able to sequester chloroplasts from ingested diatoms, and to keep them active for several days to weeks, conversely to Ammonia sp. T6 (Jauffrais et al., 2018). These active chloroplasts could serve as an alternative source of oxygen and/or food through photosynthesis (Bernhard and Alve, 1996) or another metabolic pathway (Jauffrais et al., 2019), and thereby increase the capability of this 400 species to survive anoxic events. Although sequestration of chloroplasts was never investigated for E. magellanicum, its abundant spinose ornamentation in the umbilical region and in the vicinity of the aperture (Fig. 3c-d) suggests that this species is capable to crush diatom frustules as some kleptoplastic species (Bernhard and Bowser, 1999;Austin et al., 2005). As Hagens et al. (2015) observed that the light penetration depth in the Den Osse Basin never exceeded 15 m in 2012, and therefore photosynthesis by kleptoplasts (Bernhard and Alve, 1996) appears unlikely for both our aphotic stations (34 and 23 m depth). 405 However, other foraminifera from aphotic and anoxic environments such as deep fjords are kleptoplastic and use these kleptoplasts for a yet unknown purpose (Jauffrais et al. 2019).
Rather surprisingly, the drop in foraminiferal densities at station 2 in October-November, which we interpreted as a delayed response to sulphidic conditions, is less strong for Ammonia sp. T6 than for the two Elphidium species, suggesting that this 410 species is less affected. However, this does not agree with our previous suggestion that the two Elphidium species would be more tolerant to anoxic and sulphidic conditions. As already proposed by LeKieffre et al. (2017), Ammonia seems to be able to deal with anoxia (up to 28 days, but with no sulphide) by reducing its metabolic activity, but this ability was never shown for Elphidium species. If E. selseyense and E. magellanicum are indeed unable to resist to anoxia by reducing their metabolism or by entering a dormancy state, this could explain their stronger decrease in densities at station 2 compared to Ammonia sp. 415 T6. Nevertheless, further studies about the ability and mechanisms of the two Elphidium species to resist to anoxic/sulphidic conditions are necessary.
Another remarkable observation is that Ammonia sp. T6 (and T. inflata) shows maximum densities in January-March, contrasting with the two Elphidium species, which have their density maxima later in the year (May-September). This temporal 420 offset could possibly be explained by a difference in preferential food source, with food particles available in winter (January-March) being more suitable for Ammonia sp. T6 (and T. inflata), versus food particles available later in the year, resulting from phytoplankton blooms, being more favourable for E. selseyense and E. magellanicum.
In our study, for E. selseyense (and E. magellanicum), the continuous presence of a high proportion of small sized specimens and progressively increasing densities between January and September 2012 strongly suggest ongoing and continuous 425 reproduction (Supplementary Figure 3A). Continuous reproduction during the year has been described earlier for different foraminiferal genera, such as Elphidium, Ammonia, Haynesina, Nonion and Trochammina (e.g. Jones and Ross, 1979;Murray, 1983;Cearreta, 1988;Murray, 1992;Basson and Murray, 1995;Gustafsson and Nordberg, 1999;Murray and Alve, 2000).
Conversely, for Ammonia sp. T6, a decrease in densities coupled with a rapid increase of overall test size between March and May 2012 (small sized specimens remain present but in smaller proportions) could be indicative of a period of reduced 430 recruitment (Supplementary Figure 3B).
In fact, foraminifera exhibit a large range of feeding strategies, with several species showing selective feeding with specific food particles (Muller, 1975;Suhr et al., 2003;Chronopoulou et al., 2019). Hagens et al. (2015) reported that in Lake Grevelingen the phytoplankton composition was different between April-May and July 2012. In April-May, the phytoplankton bloom was mainly composed of the haptophyte Phaeocystis globose (Scherffel, 1899), whereas it was dominated by the 435 dinoflagellate Prorocentrum micans (Ehrenberg, 1834) in July. Elphidium was reported to be able to feed on various food sources (e.g. diatoms, dinoflagellates, green algae; Correia and Lee, 2002;Pillet et al., 2011). However, diatoms are a major food source for kleptoplastic species (Bernhard and Bowser, 1999), such as E. selseyense (Jauffrais et al., 2018;Chronopoulou et al., 2019). Ammonia spp. seems able to feed on very diverse food sources including microalgae, diatoms, bacteria or even metazoans (Lee et al., 1969;Moodley et al., 2000;Dupuy et al., 2010;Jauffrais et al., 2016;Chronopoulou et al., 2019). 440 Recently, Chronopoulou et al. (2019) showed different feeding preferences for Ammonia sp. T6 and E. selseyense in intertidal environments in the Dutch Wadden Sea. Although diatoms are ingested by both species (but much more by E. selseyense), dinoflagellates were consumed by E. selseyense but not by Ammonia sp. T6. The latter species is also capable to feed on metazoans by active predation (Dupuy et al., 2010).
These observations suggest that at station 2, the different seasonal density patterns of Ammonia sp. T6 and the two Elphidium 445 species are not the consequence of a large difference in tolerance to anoxia/sulphides, but rather a different adjustment to the seasonal cycle of food availability. At station 1, the very low densities of Ammonia sp. T6 could putatively be explained by a recolonization starting in January, when food conditions were favourable for this taxon (as testified by the strong density increase in January 2012 at station 2). However, once a more abundant pioneer population had developed (in March-May), food conditions may have been no longer favourable for Ammonia sp. T6, explaining why its density did not show a further 450 increase. Conversely, the food conditions may have become optimal for the two Elphidium species, explaining their strong density increase between March and May 2012. If true, this would mean that the lower densities of Ammonia sp. T6 would not be due to a lower resistance to anoxia and free sulphides, but rather due to an unfavourable seasonal succession of food availability. Previous studies already suggested that hypoxic/anoxic conditions coupled with increased food input from (Gustafsson and Nordberg, 1999). The fact that also at station 2, this species was mainly observed between March and September 2012 corroborates our conclusion of its dependence on a specific food regime.
Finally, encrusted forms of E. magellanicum were observed at both stations from May until the end of the year, but were absent in the samples of March 2012. In view of the fact that the crusts consist mainly of organic matter, the encrusted individuals appear to be specimens with preserved feeding cysts. The precise functions of cysts observed around foraminifera are not clear, 460 and include feeding, reproduction, chamber formation, protection or resting (Cedhagen, 1996;Heinz et al., 2005). Concerning the cysts of E. magellanicum described here, very similar observations have been made for Elphidium incertum at different locations (Norwegian Greenland Sea and Baltic Sea in Linke and Lutze, 1993;Koljö Fjord in Gustafsson and Nordberg, 1999; Kiel Bight in Polovodova et al., 2009). If we assume that encrusted specimens indeed present remains of feeding cysts, the observation of abundant encrusted specimens corroborates our conclusion that the surface water phytoplankton bloom in May 465 2012 (i.e. probably mainly Phaeocystis globosa) provided a food source particularly well suited to the nutritional preferences of this species.

Conclusion
In this study we examined the foraminiferal community response to different durations of seasonal anoxia coupled with the presence of sulphide in the uppermost layer of sediment at two stations in Lake Grevelingen. In both stations investigated, 470 foraminiferal communities are highly impacted by the combination of anoxia and H2S in their habitat. The foraminiferal response varied depending on the duration of adverse conditions, and led to a near extinction at station 1, where anoxic and sulphidic conditions were present for one to two months, compared to a drop in standing stocks at station 2, where these conditions lasted for one month or less. At both sites, foraminiferal communities showed a two-month delay in the response to anoxic and sulphidic conditions, suggesting that the presence of H2S inhibited reproduction, whereas mortality was not 475 necessarily increased. The duration of the subsequent recovery depended on whether the foraminiferal community was almost extinct (station 1) or remained present with reduced numbers (station 2). In the former case, six months were needed for faunal recovery, whereas in the latter case, it took only two months. We hypothesize that the dominance of E. selseyense and E. magellanicum at station 1 is not due to a lower tolerance of Ammonia sp. T6 to anoxic and sulphidic conditions, but is rather the consequence of a different adjustment between the two Elphidium species and Ammonia sp. T6 with respect to the seasonal 480 cycle of food availability.

Data availability
Raw data are available in Supplementary Material. The authors declare that they have no conflict of interest.