Interplay between eutrophication and climate warming on bacterial communities in coastal sediments differs depending on water depth and oxygen history

Coastal aquatic systems suffer from nutrient enrichment, which results in accelerated eutrophication effects due to increased microbial metabolic rates. Climate change related prolonged warming will likely accelerate existing eutrophication effects, including low oxygen concentrations. However, how the interplay between these environmental changes will alter coastal ecosystems is poorly understood. In this study, we compared 16S rRNA gene amplicon based bacterial communities in coastal sediments of a Baltic Sea basin in November 2013 and 2017 at three sites along a water depth gradient with varying bottom water oxygen histories. The shallow site showed changes of only 1.1% in relative abundance of bacterial populations in 2017 compared to 2013, while the deep oxygen-deficient site showed up to 11% changes in relative abundance including an increase of sulfate-reducing bacteria along with a 36% increase in organic matter content. The data suggested that bacterial communities in shallow sediments were more resilient to seasonal oxygen decline, while bacterial communities in sediments subjected to long-term hypoxia seemed to be sensitive to oxygen changes and were likely to be under hypoxic/anoxic conditions in the future. Our data demonstrate that future climate changes will likely fuel eutrophication related spread of low oxygen zones.


Scientific Reports
| (2021) 11:23384 | https://doi.org/10.1038/s41598-021-02725-x www.nature.com/scientificreports/ decrease in nitrogen concentration since 1990 but less improvement in phosphorus reflecting its higher degree of storage in sediments 16 . Various mitigation attempts have been performed to reduce nutrient intake and to recover 'dead zones' , such as biomanipulation, lowering growth of phytoplankton by precipitation of phosphorus, and artificial re-oxygenation of bottom waters 17 . However, how the interplay between nutrient backlog, mitigation of nitrate and phosphate release, and climate change will alter coastal zones is not well understood. Microorganisms in sediment use a range of electron acceptors for energy conservation, from oxygen with the highest energy efficiency to alternative electron acceptors including nitrate, manganese (IV), iron (III), and sulfate under hypoxic/anoxic conditions 18 that favors the growth of anaerobic microorganisms 19 . Previous studies within False Bay, Washington show that organic matter degradation is slower under anaerobic conditions 20 , and it has been estimated in laboratory incubation studies of Baltic Sea sediments that organic matter is consumed upon oxygenation to reach levels similar to long-term oxic sediments after approximately three months at 8 °C 14 . Due to sulfate reduction in oxygen depleted sediments, microbial communities are rich in Gamma-and Deltaproteobacteria 21 while Planctomycetes and Betaproteobacteria are more abundant in oxygen rich sediments 22 . In addition, the Sulfurimonas and Arcobacter genera have been shown to increase in relative abundance upon oxygenation of anoxic sediment 23 , suggesting they are key players in the microbial community upon re-oxygenation. However, this was observed under laboratory incubation experiments 14 and has not been investigated under field conditions. Microorganisms are key to nutrient cycling and changes in their community structure and ability to respond to eutrophication in combination with climate change might have large effects on the benthic ecosystem.
In this study, we compared coastal sediments in a Baltic Sea basin in November 2013 and November 2017 from three sites with varying oxygen conditions, oxygen histories, and depth 14,23,24 . These include a shallow site that has been under long-term, constant oxic conditions as shown in previous studies 14 ; an intermediate site with seasonal hypoxia; and a deep site that was in previous studies all year anoxic/hypoxic 14,23 . Collected data from the area around the sampling sites showed a history of long-term oxygen deficiency 14,23 . The aim was to investigate spatiotemporal differences in sediment nutrient contents and bacterial communities. We hypothesized that the response to possible ongoing climate change warming effects and eutrophication in coastal sediments was mainly dependent on depth and oxygen histories. More specifically, that bacterial communities in shallower coastal sediments are highly affected but better adapted due to sufficient oxygen concentrations to environmental changes, whereas deeper coastal sediments with long-term hypoxic conditions are sensitive to even small environmental perturbations.

Results
Sediment and bottom water chemistry. Within the last two decades data collected by HELCOM and ICES (International Council for the Exploration of the Sea) 25 showed trends of increasing temperatures on coastal bottom waters (1-30 m below sea surface (mbs)) across the Baltic Sea (Western Gotland-, Eastern Gotland-, and Bornholm-Basins). The temperature increased by several degrees independent of the depth of the coastal bottom waters (Supplemental Fig. S1) along with a trend in oxygen decline (Supplemental Fig. S1). The data within our study showed similar trends ( Table 1). The observation station closest to the investigated sampling sites showed a general temperature increase with increasing phosphate and decreasing nitrate concentrations between the years 2001 and 2012 (Supplemental information). Geochemical data from the three sites in 2013 and 2017 in this study are provided in Fig. 1 and Table 1 (statistical support in Supplemental Tables S1;  S2) and support the HELCOM data temperature and oxygen concentration trends for the Baltic Sea as a whole.
The shallow site bottom water oxygen concentration in November in 2013 was 11.3 mg/L and was lower in 2017 by approximately 50% with 5.88 mg/L (Table 1). At the same time, the bottom water pH dropped from 7.85 to 7.04 ( Table 1). The 6.5 m deep shallow site showed a strong geochemical response within the 0-1 cm sediment layer with significantly lower nitrate + nitrite (in combination) concentrations measured in November 2017 (pore water, 2013, n = 3, 33.73 ± 8.08 µM compared to 2017, n = 3, 19.19 ± 5.81 µM, ANOVA, pairwise comparison, p < 0.01) and high variation in total iron as measured in 2017 (pore water, 2013, 0.55 ± 0.21 µM and 2017, 2.47 ± 2.18 µM). Pore water sulfate concentrations were lower in 2017 with 4.33 ± 0.16 mM compared to 4.72 ± 0.11 mM in 2013 while organic matter sediment content was significantly higher (2013, n = 3, 11.7 ± 0.95% and 2017, n = 3, 13.83 ± 0.45%, p < 0.05). The low concentration of pore water phosphate stored in the shallow site had a small increase in 2017 (2013, 2.39 µM ± 0.8 and 2017, 7.86 µM ± 1.20). Table 1. Oxygen, temperature, and pH in 2013 and 2017. Overview of in situ oxygen and temperature in bottom waters in 2013 and 2017 at each sampling site as well as pH measured for bottom waters for both years and all sites in triplicates; mean (n = 1 or 3 as indicated) ± standard deviation. a n = 1.
In contrast to the intermediate site, the deep site sediments had strong responses. Despite that the site was partially oxic in 2017 (Supplemental Fig. S2), it has previously showed long-term hypoxic conditions. In November 2013 the bottom water oxygen concentration was hypoxic with 0.85 mg/L and decreased to 0.49 mg/L (Table 1) in November 2017. The bottom water pH was 8.  Table 1). The pore water (0-1 cm) nitrate + nitrite concentration remained low with 3.19 ± 0.3 µM in 2013 compared to 3.54 ± 1.57 µM in 2017 as it was likely continually reduced in this long-term hypoxic site. Pore water total iron was 4.  Microbial diversity, community composition, and taxonomy. Rarefying the data to the lowest sample size (Supplemental Table S3 for comparison with rarefied data) confirmed that the main bacterial diversity was covered and therefore, to retain as much data as possible the unrarefied data was used. The unrarefied average sample size was 31,474 reads (n = 18; Supplemental Table S4 for Table S1 for statistical support). The shallow site community composition did not change significantly (PERMANOVA, n = 6, p > 0.05) from 2013 to 2017 and showed small shifts in the community (  Table S1). The bacterial community composition in the deep site significantly (n = 6, p < 0.05; Supplemental Table S1)  The bacterial communities between the different sites along the depth gradient have been shown to be significantly different in a previous study 14 that was confirmed for the year 2017 (PERMANOVA, p < 0.01, Supplemental Table S1). Based on SIMPER analysis, the major contributors to the differences between the sampling sites in 2013 were the CG2-30-66-27, UBA6092, and Desulfobacula; while in 2017 the top contributors between sites shifted (Supplementary Tables S4 and Table S5). The top contributors between the shallow and intermediate stayed the same with the MBNT15 genera CG2-30-66-27 (6.61%, average contribution of dissimilarity) and UBA6092 (2.72%), while between the shallow and deep site the main contributor now included the Cyanobacteria Nodularia (7.73%). The differences between the deep and intermediate site were CG2-30-66-27 (10.85%) and Nodularia (7.76%) (Supplementary Tables S4 and Table S5).
A general trend could be observed when comparing the bacterial community composition of different sites between the years 2013 and 2017 where the highest dissimilarity were in the deep site (79.29%) with increasing similarity with decreasing depth (intermediate, 60.8%; shallow, 56.66%). Additional constrained ordination analysis (RDA) showed the environmental variables most explaining the changes between the sampling sites and years were oxygen (ANOVA, p < 0.05, Supplemental Table S1 and Supplemental Fig. S5), temperature (p < 0.05), and organic matter (p < 0.01).
The number of detected phyla within each site differed between years, showing a generally higher diversity among the total sum of amplicon sequence variants (ASVs) (Supplemental Fig. S3) but a lower number of abundant ASVs (> 0.5% relative abundance) from 2013 to 2017. The most abundant phyla within each site were the Proteobacteria, Desulfobacterota, Bacteroidetes, and Cyanobacteria.

Discussion
Global coastal oceans are continuously influenced by anthropogenic induced changes such as accelerated eutrophication that can result in decreased oxygen concentrations in bottom waters 4 or the steady influence of ongoing climate change related effects such as rising temperatures that exacerbate existing eutrophication effects 26 . In situ bottom water temperature and oxygen data from March 2017 to December 2017 showed a decline in oxygen concentrations along with a temperature increase in November in coastal bottom waters at all the three sampling sites. These data match that collected by HELCOM and ICES 25 showing trends of increasing temperatures and declining oxygen concentrations in coastal bottom waters (1-30 mbs) independent of the depth across the Western Gotland-, Eastern Gotland-, and Bornholm Basins (Supplemental Fig. S1). A previous study showed significantly different microbial communities along this coastal water depth gradient in 2013 14 , which could also be shown for the communities of the same gradient in the year 2017. The focus of this study was to investigate Baltic Sea sediment changes between 2013 and 2017 at different coastal water depths to study oxygen deficient zones common in the Baltic Sea 27 . The deepest investigated coastal sediment (~ 30 mbs) was previously shown to be under consistent oxygen deficient conditions while the area around the bay has a history of long-term oxygen deficiency 14 . However, this area was found to have changed to seasonal varying hypoxic-oxic conditions in 2017 that was potentially due to the reduction of nutrient intake into a close-by bay (~ 4 km) from a nearby sewage-treatment plant 28 . The data suggested that most changes within the bacterial communities happened at the deep oxygen deficient site (11% significant different ASVs comparing 2013 and 2017). Cyanobacteria, especially Nodularia, went from being undetectable in 2013 to being the most abundant genus in 2017. This increase in Nodularia may also have caused the significant increase of organic matter due to deposition of pelagic Cyanobacteria on the sediment 29 . A further potential cause of increase in OM could have been due to primary production sources such as green algae or dinoflagellates that bloom in spring 30 along with Diatoms that bloom in spring 30 and autumn 31 . This coupled to the generally higher OM content within the deep site may also have been due to decreased OM degradation rates in deeper, less well mixed, and low oxygen conditions 20 . This increase in organic matter likely led to an increase in sulfate reducing bacteria like Desulfobacula and Desulfatiglans in 2017. Such bacteria have been shown to participate in the mineralization of organic matter 32 , and sulfate reduction can account for as much as up to half of organic matter mineralization 33 . Furthermore, the abundance of sulfate reducers has been shown to be positively correlated with organic nitrogen and phosphorus in the surface layer of sediments 34 , and Desulfobacula has been found to be associated with Baltic Sea organic-rich sediments 35 . The higher relative abundance of sulfate reducers may also have caused the significant increase of phosphate as sulfate-reducing bacteria may indirectly contribute to dissolved phosphate release during organic matter mineralization 35 by the production of sulfide that binds ferric iron and hinders the formation of phosphate binding ferric oxyhydroxides 36 . Another potential contribution to the increased pore water phosphate concentrations in November 2017 was the raised oxygen concentration from March to July 2017 leading to iron oxide (Fe-P) that binds phosphate 37 . As the oxygen concentrations became hypoxic in November, Fe-P dissolution may have increased the release of phosphate into the pore water 38 . The general increase in temperature within the last years could have led to a shift within seasonal changing oxygen states, with possible higher regional river-runoff in spring 39 feeding even long-term hypoxic deeper coastal sediment with oxygen-rich water. This can lead to greater bacterial metabolism rates, followed by lower oxygen concentrations. Similar results have also been reported from arctic lakes with increasing runoff 40 . The data confirmed that with increasing temperature, an increase of cyanobacterial deposition in already hypoxic sediments is likely expected 41 . In addition, more organic matter leads to an increased sulfur cycling close to the sediment surface 34,42 , that might accelerate oxygen decline with subsequent increased release of phosphates 43 . Due to prolonged seasonal stagnant water with less mixing of the bottom water, deeper sites are more affected by eutrophication and can turn long-term hypoxic 44 , and an increase in temperature is likely to accelerate eutrophication induced hypoxia of benthic waters.
In comparison, intermediate coastal sediments (intermediate site, ~ 20 mbs) exposed to already long-term fluctuating hypoxic-oxic conditions may already have had a bacterial community that was adapted to oxygen fluctuations. Compared to 2013 less dominant bacteria taxa were found in 2017. These few dominant taxa include CG2-30-66-27 from the candidate phylum MBNT15, which might be obligate anaerobes that couple H 2 and acetate oxidation to nitrate reduction 45 . The genus UBA1847 from the family Woeseiaceae within the Proteobacteria that are among the most abundant microorganism in coastal sediments. They cover a broad spectrum ranging from facultative sulfur-and hydrogen-based chemolithoautotrophic to obligate chemoorganoheterotrophic bacteria 46 (Fig. 4). The significantly increased relative abundance of ASVs aligning within the candidate phylum MBNT15 that thrives in anoxic deep sediments 45  www.nature.com/scientificreports/ environmental variation between 2013 and 2017 occurred or the proportion of anoxic sediment within the sliced 0-1 cm increased compared to 2013. In general, the data based on environmental variables in 2017 could suggest a potential reversal of eutrophication effects when compared to 2013, with for example an increase in oxygen concentrations and a decline in phosphorus 23 . In addition to the minor changes in the microbial communities, the data suggested that oxygen histories may play an important role in the extent microbial communities react to environmental changes. This importance was already shown in previous studies, where long-term anoxic coastal Baltic Sea sediment has been re-oxygenated 14 .
Shallow coastal areas are in general more exposed to anthropogenic influenced changes related to climate change or eutrophication, such as input of nutrients from land 47 , increased acidification 48 , and faster warming 49 of the water column reaching the sediments. However, these sediments have due to their nature a constant oxygen input from e.g. mixing of the water column by the wind and previous studies show that the shallow area has been under long-term oxic conditions 14 . Despite that the studied coastal bay has not undergone any artificial changes, beside decreased nutrient intake of a wastewater treatment plant of a close-by bay (~ 4 km) in the intervening years 28 , seasonally changing oxygen conditions were observed between March and December 2017 (Supplemental Fig. S2). In addition, a previous study shows significant differences between bacterial communities at all three sites with the shallow site being most distinct from the other two 14 . Within the shallow site, only 1.1% of the dominant taxa (i.e. comprising > 0.5% relative abundance) were significantly different comparing 2013 and 2017 indicating the highly abundant bacteria were still present (Fig. 3). Sinkko et al. 32 showed that the most abundant ASVs in shallow coastal sediments (~ 4 mbs) are also the most stable fraction of the community during oxygen deficiency events. This indicated that a large part of this community were likely facultative anaerobes. Members of this group include the phyla Planctomycetota and Proteobacteria that are adapted to temporally varying oxygen concentrations and short-term low oxygen events. The data on one hand showed only minor changes on the bacterial community of the shallow coastal sediments within the Baltic Sea basin in 2017 compared to 2013. Changes in environmental factors such as temperature, pH, oxygen, and organic matter could be either the result of temporal differences or be a first hint of a changing environment. The environmental conversion could be related to climate change with temperature increase, acidification 50 , organic matter accumulation leading to CO 2 release via decomposition 51 followed by decreasing oxygen concentration. On the other hand, the intermediate site, already adapted to seasonal oxygen changes, could either show temporal differences over the time or could have potentially showed first hints of a reversed eutrophication. Nevertheless, shallow coastal ecosystems and their bacterial communities will be directly affected by environmental changes related to future climate change effects in combination with already increased nutrient load from the land, and how bacterial communities will adapt to changes needs to be further investigated.

Conclusions
No simple, single response within bacterial communities of changes within coastal sediments was discerned and how bacterial communities react to environmental changes like eutrophication and/or climate change are an interplay of different spatiotemporal factors including water depth and oxygen supply. Sediments exposed to long-term hypoxic conditions were sensitive to even small changes in e.g. oxygen concentration and temperature increase, probably resulting in accelerated bacterial activity in combination with increased organic matter leading to potential future expansion of already existing hypoxia zones. Bacterial communities of oxygen rich shallow coastal sediments did not show strong responses to environmental changes, showing that these communities were most likely stable under potential transitory oxygen depletion. Continuous exposure to eutrophication, warming, and other climate change related effects has the potential to expand and aggravate already existing hypoxic zones.

Materials and methods
Study site. The sampling location was a coastal Baltic Sea bay near the town of Loftahammaer, Sweden. The three sampling sites (Supplemental Fig. S9  Sampling. Sampling was conducted at each site on 2 November 2017 (n = 3). In addition, published data 14 from the same sites (also n = 3 per site) from 2013 were used for comparison 14 . Sampling was conducted as previously described 14 using acrylic cores (internal diameter 7 cm, length 60 cm) to sample the sediment surface with a Kajak gravity corer at each location. Oxygen and temperature were measured in situ (Multiline™ sensor, WTW™) in the bottom waters (~ 10 cm above sediment; Supplemental Table S2). The sediment surface (0-1 cm) was sliced and transferred into a sterile 50 mL polypropylene centrifuge tube (Thermo Scientific™) and mixed until homogenized. A subsample of 15 mL homogenized sediment was transferred into an acid washed 15 mL polypropylene centrifuge tube (Thermo Scientific™) for pore water analysis as well as an additional subsample of 2 mL for organic matter analysis. The sediment remaining in the initial tube was used for DNA extraction. Samples for chemistry analysis and DNA extraction were stored cooled until transported back to the laboratory the same day and then stored until further analysis at − 20 °C and − 80 °C, respectively 14 . Additionally, oxygen data were measured on bottom water in situ from March 2017 to December 2017 (n = 9) on all three sites (Supplemental Fig. S2, Table S2). Chemistry analysis. Chemistry analyses were conducted on pore waters and the sediment itself. Pore waters were sampled by centrifuging the 15 mL sediment tubes at 2200 × g for 15 min and the supernatant transferred into a new 15 mL acid washed centrifuge tube (Thermo Scientific™). The pore water was filtered

Statistical approaches.
To determine the sequencing depth, rarefaction curves were calculated with the 'vegan´ package version 2.5-6 54 based on the number of ASVs and final read counts after DADA2 analysis. The dataset showed differences in sequencing depth between the years. To confirm our results based on scaling with ranked subsampling and relative abundance normalization, the data have been compared to results using rarefied samples that showed the methods were suitable (Supplemental Table S3). Diversity of the microbial communities was estimated by calculating the alpha diversity (Shannon's H index) based on counts normalized using scaling with ranked subsampling 55 . To test for significant differences in diversity between years and sites a linear regression model was used due to the nested nature of the data with three replicates per site and the interaction between the sites and years. Different models were fitted and the model that showed the best fit for the data structure was chosen based on Akaike information criterion (AIC). All factors (years and sampling sites as interaction and replicates nested within sites) were used as fixed effects in the final model ('lme()' function, 'stats' package) 57 due to the small sample size. The statistical differences between the factors were tested using the 'anova' function from the 'stats' package in R 53 with pairwise comparison for years at each site using the 'emmeans' package (version 1.5.4) 56 . For testing on differential abundances between the years at each sampling site, taxa not seen in at least 20% of the samples were removed and a zero-inflated negative binominal model was used to reduce the excess of zeros in the dataset. Then, differential abundance analysis on counts was done with the 'DESeq2´5 7 package to analyze statistical differences between communities based on all ASVs. The cutoff of 0.5% was based on an ASV having > 0.5% relative abundance in at least one sample. For further analysis (if not otherwise stated), relative abundances were calculated from all ASV count data. Beta-diversity was investigated via non-metric multidimensional scaling (NMDS) based on Bray-Curtis distances. Variance inflation factor analysis showed multi-collinearity between oxygen and nitrate (Supplemental Table S1). The focus of this study was to investigate changes based on oxygen and temperature. Therefore, the variable nitrate has been removed for the redundancy analysis (RDA), which resulted in VIF < 5 for the remaining variables. The RDA analysis was performed using the 'RDA()' function within the 'vegan' package plus an ANOVA permutation test, with a direct model using marginal terms to investigate if significant environmental variables influenced the differences between the bacterial communities. Bray-Curtis dissimilarities were calculated using the 'vegdist' function within the 'vegan' package. To check if microbial community composition in 2013 and 2017 was significantly different, a PERMANOVA within the 'adonis' function in the 'vegan' package with Benjamin-Hochberg p-value correction was applied. Additionally, PERMANOVA testing was used to compare the microbial communities along the depth gradient (comparing the different sites) for the years 2013 and 2017. Similarity of percentages (SIMPER) analysis from the `vegan' package was used to discriminate species responsible for dissimilarities between the year 2013 and 2017 at each site, as well as between sites within each year (Supplemental Table S4). To determine whether differences in environmental variables between the years were significant, the same model that was used for diversity testing was constructed and tested. The best fit (best AIC) for the data structure was reached using a linear regression model ('lme()' function, 'stats' package) with years and sampling sites as interaction and replicate nested within site as fixed effects.

Data availability
16S rRNA gene sequencing data are available on the NCBI database under Bioproject PRJNA721576 and PRJNA322450.

Code availability
The code used is available at Supplemental Methods S1.