Resource Partitioning Between Phytoplankton and Bacteria in the Coastal Baltic Sea

Eutrophication coupled to climate change disturbs the balance between competition and coexistence in microbial communities including the partitioning of organic and inorganic nutrients between phytoplankton and bacteria. Competition for inorganic nutrients has been regarded as one of the drivers affecting the productivity of the eutrophied coastal Baltic Sea. Yet, it is unknown at the molecular expression level how resources are competed for, by phytoplankton and bacteria, and what impact this competition has on the community composition. Here we use metatranscriptomics and amplicon sequencing and compare known metabolic pathways of both phytoplankton and bacteria co-occurring during a summer bloom in the archipelago of Åland in the Baltic Sea to examine phytoplankton bacteria resource partitioning. The expression of selected pathways of carbon (C), nitrogen (N), and phosphorus (P) metabolism varied over time, independently, for both phytoplankton and bacteria, indicating partitioning of the available organic and inorganic resources. This occurs regardless of eukaryotic plankton growth phase (exponential or stationary), based on expression data, and microbial community composition. Further, the availability of different nutrient resources affected the functional response by the bacteria, observed as minor compositional changes, at class level, in an otherwise taxonomically stable bacterial community. Resource partitioning and functional flexibility seem necessary in order to maintain phytoplankton-bacteria interactions at stable environmental conditions. More detailed knowledge of which organisms utilize certain nutrient species are important for more accurate projections of the fate of coastal waters.


INTRODUCTION
Coastal marine ecosystems have a key role of filtering nutrients and pollutants coming from land, before entering the pelagic environment (Bonsdorff et al., 1997;Rabalais et al., 2009;Carstensen et al., 2019). In coastal oceans, as along the coast of the Baltic Sea, allochthonous organic matter will get incorporated into the microbial food web, which may have large consequences for the nutrient balance between phytoplankton and bacteria (Andersson et al., 2015;Diner et al., 2016). During summer, low levels of inorganic nutrients (Nausch et al., 2018;Savchuk, 2018) limits phytoplankton growth (Granéli et al., 1990;Kivi and Setäl, 1995), while high levels of organic nutrients (Nausch et al., 2018;Savchuk, 2018) are present.
Effects of global warming including increased freshwater flow and allochtonous input of dissolved nutrients are likely to aggravate the long-term consequences of eutrophication in coastal ecosystems, such as archipelagic systems common to the brackish Baltic Sea, making it a suitable study system for the impacts of these effects. The consequences may include higher phytoplankton growth, oxygen deficiency, and a transition into a less productive ecosystem (Rönnberg and Bonsdorff, 2004;Rabalais et al., 2009). A complex interplay between autotrophic and heterotrophic marine microbes will determine how the accumulation and turnover of organic matter will be affected in a future climate scenario (Andersson et al., 2015;Zhou Y. et al., 2018). An understanding of how organic and inorganic nutrients are partitioned between phytoplankton and bacteria, and the dependence of phytoplankton on bacterial remineralization, is therefore important for predictions of status and knowledge of how to manage coastal ecosystems (Heiskanen et al., 2019).
One of the ways to investigate phytoplankton and bacteria interactions is by looking for co-occurrence patterns in the natural environment (Elifantz et al., 2005;Teeling et al., 2012;Landa et al., 2016). Several studies have investigated interactions during open water phytoplankton bloom conditions, and have shown that the different growth phases of blooms cause compositional changes in the co-occurring bacterial community, with successions of different bacterial groups assigned to Alpha-, Beta-, and Gamma-proteobacteria, Flavobacteria, Cytophagia, Actinobacteria, and Firmicutes (Teeling et al., 2012;Landa et al., 2016;Sison-Mangus et al., 2016;Camarena-Gómez et al., 2018). These changes have been attributed to the succession of exudates released by phytoplankton, differing both in quality (Cottrell and Kirchman, 2000;Elifantz et al., 2005;Camarena-Gómez et al., 2018;Mühlenbruch et al., 2018) and quantity (Sarmento et al., 2016) during their lifespan (Grossart et al., 2005;Grossart and Simon, 2007;Mühlenbruch et al., 2018). This suggests that a diverse bacterial community is important for the degradation of amino acids, proteins, extracellular polysaccharides and carbohydrates released by phytoplankton. This is also seen in the taxonomical difference among bacterial groups found either free-living or particle attached (Grossart et al., 2005;Rink et al., 2007;Sapp et al., 2007;Bagatini et al., 2014). Compositional shifts are often assumed to involve functional changes (Hunt et al., 2008;Williams et al., 2012), especially for bacteria, as different taxa are attributed with different functions, e.g., being copiotroph or oligotroph (Lauro et al., 2009), or generalists or specialists (Sarmento et al., 2016). However, composition studies alone limit the conclusions about which underlying processes and functions are selected for.
Using metatranscriptomics, capturing the actively expressed mRNA pool, with the inclusion of polyA and total mRNA, both the eukaryotic and prokaryotic transcriptional repertoire of a microbial community can be investigated (Dupont et al., 2015). Metatranscriptomics provide a glimps into the possible functions within a community at the time of sampling, while amplicon sequencing, using the ribosomal gene pool (16S and 18S rRNA gene), can be used for taxonomic identification. By combining the two methods both functional and compositional characterization of a microbial community can be made (McCarren et al., 2010;Teeling et al., 2012;Gifford et al., 2013;Alexander et al., 2015), thereby allowing for analysis of the connection between taxonomy and function. Functional analyses using transcriptomic data have suggested that bacterial species may avoid competition by ecological niche formation by having preferences for different types of algal substrates (Teeling et al., 2012) and of marine dissolved organic matter (McCarren et al., 2010). Taken together, this emphasizes the importance of resource partitioning and microbial community dynamics, encompassing both phytoplankton and bacteria, in the global carbon cycle.
In this study planktonic eukaryotes (PE <90 µm) and bacterioplankton (BAC) of an archipelagic bay in the Baltic Sea were sampled at two timepoints during late summer, with the aim of investigating carbon and nutrient (nitrogen and phosporus) utilization and partitioning between the two groups. The PE, consisting of phytoplankton (auto-and heterotrophic flagellated forms) and ciliates, and the BAC, dominated by heterotrophic bacteria with some cyanobacteria, were characterized separately, taxonomically, through both microscopic observation and amplicon sequencing, and functionally, focusing on the transcription of genes associated with uptake and assimilation of carbon, nitrogen, and phosphorus. The combination of morphological and molecular methods for characterization of the PE allows a more accurate identification of microbes, in light of known weaknesses such as high copy numbers for some protists (Gong and Marchetti, 2019) and brittleness when fixed for others (Mironova et al., 2009). Few, if any, studies have previously characterized the functional repertoire or resource partitioning of both phytoplankton and bacteria together, in such a coastal environment dominated by photosynthetic and often mixotrophic dinoflagellates.
For DNA and RNA collection at each sampling day, 3 replicates of 1 L of seawater were prefiltered through a 90 µm mesh, followed by filtration through a 3 µm membrane filter (Versapor, 47 mm), capturing both phytoplankton and the attached (AT) bacterial fraction. Subsequently, 300 mL of each filtrate were filtered through a 0.2 µm Supor filter (47 mm), capturing the free-living (FL) bacterial fraction. Filtrations were done using a hand operated vacuum pump. All filters were

Morphological Identification and Enumeration
For phytoplankton, subsamples (50 mL) were fixed using pH neutral Lugol's solution and stored in darkness. Phytoplankton cells sedimented in Utermöhl chambers were counted at 100x magnification using an inverted microscope (Olympus CKX 41). Phytoplankton, identified according to Hällfors (2004), were grouped as planktonic eukaryotes (PE) and cyanobacteria (filamentous). Phytoplankton biomass was calculated from cell biovolume and carbon content according to Olenina et al. (2006). Bacterial abundance (BAC) was measured by flow cytometry in 1.8 mL water subsamples fixed using 200 µL formaldehyde (1.5% final concentration) stored at−20 • C for transport to the laboratory, were stored at −80 • C until further processing. Samples were stained with SYBR green (Life Technologies) and bacterial cells were counted using a Cube8 flow cytometer (Partec). The cytograms were analyzed using FCS Express 4 flow cytometry data analysis software.

DNA Extraction and Preparation for Sequencing
Filters were cut into 4 quarters, of which one quarter was used for DNA extraction and the remaining 3 for RNA extraction. DNA was extracted using a phenol:chloroform protocol according to Boström et al. (2004). The V3-V4 region of the 16S rRNA gene was amplified, from both the AT and the FL size fractions, using primers 341F (CCTACGGGNGGCWGCAG) and 805R (GACTACHVGGGTATCTAATCC), and the V4-V5region of the 18S rRNA gene was amplified, from the AT size fraction, using primers 574 * F (CGGTAAYTCCAGCTCYV) and 1132R (CCGTCAATTHCTTYAART). Primers were connected to Illumina adapters, in accordance with Herlemann et al. (2011) and Hugerth et al. (2014)

RNA Extraction and Sequencing for Metatranscriptomes
Filters were placed in petri dishes, on ice, and cut into smaller wedges with scissors, then placed in MatrixE tubes (MPbio), along with RLT-buffer (Qiagen), TE-buffer, βmercaptoethanol (0.1 %) and Lysozyme (0.04 mg mL −1 ). For the BAC samples, two vectors (pTZ19R, 546 bp and pFN18A, 970 bp) (Supplementary Material 1) at 0.1 ng µL −1 each, were added at this step to be used as internal standards, according to Satinsky et al. (2013). Cells were lysed using a FastPrep-24 instrument with a QuickPrep adaptor (MPbio), 3 rounds at 6 m s −1 for 40 s, with 1 min on ice in between, after which the Qiagen RNeasy mini kit was used for the extraction, according to protocol. The extracted RNA was treated with DNase to remove DNA (AMBIONTurbo DNA free, Invitrogen). The 6 samples intended for the BAC metatranscriptomes (0.2-3 µm size fraction) were rRNA depleted (RiboMinus Transcription isolation kit, Invitrogen) with a RiboMinus Concentration module, followed by cDNA to aRNA protocol (MessageAmp II-Bacteria RNA amplification kit, Invitrogen). To each of the 6 DNase treated samples intended for PE metatranscriptomes (3-90 µm size fraction) 0.02 pg spike-in ERCC Mix1 (Ambion) was added before they were sent for poly-A selection, which was followed by mRNA fragmentation and synthesis of cDNA, at SciLifeLab/NGI (Stockholm, Sweden). All 12 samples were then sequenced on one lane with Illumina HiSeq 2500 High Output mode v4, PE 2 × 125 bp, at SciLifeLab (Stockholm, Sweden).

Analysis of Sequencing Data
For the 16S and 18S gene amplicons, paired end sequences were analyzed in two separate rounds through the DADA2 pipeline, version 1.4.0 (Callahan et al., 2016) (available at https:// github.com/benjjneb/dada2), implemented as a package in R, version 3.4.3. After the initial quality filtering 40% of the 16S reads and 25% of the 18S remained, resulting in 11 345 unique error corrected amplicon sequence variants (ASVs) for the 16S data and 29 642 unique error corrected ASVs for the 18S data (Supplementary Table 1). These were taxonomically assigned using Silva-128 db (Quast et al., 2013). To analyze the libraries, packages in R (version 3.4.1) were used, vegan (nMDS, betadiversity) (Oksanen et al., 2008(Oksanen et al., , 2019 and tidyR to sort ASVs by annotation, and ggplot2 for all plotting (Wickham, 2016).

Statistics and Internal Standards
Differential expression analyses were made for both the PE and BAC metatranscriptome data using the DESeq2 (Love et al., 2014) package in R (version 3.4.4), with the internal standards (controlGenes) in the estimateSizeFactors function to normalize counts between samples (Supplementary Table 3) (Beier et al., 2018). Transcriptional changes of enzymes refer to relative log2 fold changes between contrasted timepoints (including triplicates), at an adjusted p < 0.01, unless otherwise stated (Supplementary Tables 5, 7). PE transcript raw counts were normalized using transcripts per kilobase million (TPM) for relative abundance analysis (Wagner et al., 2012). Bacterial absolute counts were calculated from raw counts using the formula from Satinsky et al. (2013) and used for abundance analyses. The vegan package (Oksanen et al., 2008) was used for permanova analysis.

Data Availability
Raw amplicon and metatranscriptome sequencing reads in our study were deposited at the European Nucleotide Archive under study accession numbers ERP107850 and EPR109469, with sample accession numbers ERS2571542-ERS2571562 and ERS2572272-ERS2572283, respectively.

Environmental Parameters
Samples were taken during a 2-week peak in sea surface temperature (SST), general for the whole Åland archipelago, Northern Baltic Sea (Supplementary Figure 1), with similar, warm temperatures at both sampling occasions (Table 1). Salinity, Chl-a and nitrate/nitrite levels were similar, while levels FIGURE 1 | Patterns of phytoplankton and autotrophic ciliate Mesodinium-like community composition (>3 µm) and contribution to total biomass (µg Carbon L −1 ) at T1 and T2. Taxonomic composition was assessed by total microscopy counts and converted into biomass, using C conversion from cell to biomass as factors 0.17 for Cyanophycea (filamentous) and 0.15 for small phototrophic flagellates (Olenina et al., 2006). Carbon unit (pg C cell −1 ) were calculated based on cell volume for the genera Alexandrium (7,723 pg C cell −1 ), Gonyaulax (1,300 pg C cell −1 ), Heterocapsa (160 pg C cell −1 ), et al., 2006). denote taxa assigned to dinophycea (dinoflagellates). Twelve milliliter was counted in total at each date (T1 and T2).
of inorganic phosphate decreased from T1 to T2 along with an increase of Si, and pH ( Table 1). The bacterial abundance was slightly lower at T2 compared to T1. In all, the environmental parameters indicated stable environmental conditions during the studied period.

Microbial Community Functions
Metatranscriptome data from the PE and BAC parts of the community were analyzed to identify functional processes associated with uptake and assimilation of carbon, nitrogen and phosphorus, being the primary elements and nutrients utilized by both trophic levels. Differential expression (DE) analyses were made by contrasting the two timepoints (T1 and T2, in triplicates). For many of the ORFs assigned to the same functional annotation (SEED Rank1 and Rank3), the DE analysis resulted in a range of significant (padj < 0.01) log2 fold change values, spanning from positive to negative [Supplementary Tables 5 (PE), 7 (BAC)].
The taxonomic annotations of the PE metatranscriptome data did not correspond well with either the 18S rRNA gene amplicon annotations (Supplementary Figure 4) nor the microscopy counts at the family/genus level as only photosynthetic or potentially mixotrophic organisms were counted (Figure 1). Thus, no attempt was made to connect the identified PE functional processes with taxonomy. The taxonomy assigned to the BAC metatranscriptome data were more in accordance with the 16S rRNA gene amplicon results (Supplementary Figure 5), thus the assigned taxonomy from the BAC metatranscriptome was used together with that of the 16S rRNA gene ampliconbased annotation, to couple functional processes with taxonomy.

DISCUSSION
In this study we show that resources (organic and inorganic forms of C, N, and P) were partitioned between planktonic eukaryotes (PE) and heterotrophic bacteria (BAC). This indicates a broad functional ability for resource acquisition and utilization within a coastal microbial community at stable environmental conditions. Our findings also reveal that changes in resource partitioning between PE and BAC does not necessarily result in large-scale taxonomic shifts in the bacterial fraction. Previously, a decoupling between function and taxonomy among bacteria has been proposed (Louca et al., 2017(Louca et al., , 2018, suggesting that the environmental conditions select for necessary metabolic functions leading to a specific, but unpredictable, taxonomic composition of the bacterial population. Partly contradicting this, our results indicate that both environmental (nutrient conditions) and biotic (plankton species present) conditions select which bacterial taxa, with suitable metabolic functions are found at the study site. This is exemplified by that members (Flavobacteriaceae and Comamonadaceae) of the previously suggested A. ostenfeldii core microbiome, isolated from the brackish Baltic Sea (Sörenson et al., 2019), being the dominating dinoflagellate species among the PE, were found to be frequent among the BAC, both in the free-living and the attached fractions. Further, our findings indicate that while only minor BAC taxonomic shifts occurred, in both size fractions, during the dinoflagellate bloom (T1: exponential growth, T2: stationary phase), the expression of functions both related to resource acquisition and to metabolic pathways varied in relation to the availability of carbon, nitrogen and phosphorus. This can be seen in the conceptual model in Figure 8, illustrating the relevant transcripts that has been more highly expressed at either T1 or T2, or remained at similar expression levels. When compared to systems with supposed weaker biotic interactions, such as bacteria found in bromeliad cups (Louca et al., 2017), our study suggests that the interactions between PE and BAC are of importance for which plankton and bacteria are found within the community and that resource partitioning and a functional flexibility are necessary in order to maintain these interactions at stable environmental conditions, such as those seen in this coastal bay. Resource partitioning, allowing different species to coexist both temporally and spatially by the division of available resources, has previously been shown to exist among organisms within planktonic microbial communities. Hunt et al. (2008) show that within a single bacterial family there are signs of broad resource partitioning, likely dependent on nutrient utilization and season. Also, among coexisting diatom species, different inorganic and organic nitrogen and phosphorus resources have been suggested to be partitioned (Alexander et al., 2015). Previous studies have focused either on the partitioning of carbon, between primary and secondary producers on a broad scale (encompassing viruses to fish) in the Baltic Sea (Sandberg et al., 2004), and among bacteria associated with different phases of a phytoplankton bloom (Landa et al., 2016;Berg et al., 2018), or of nitrogen and phosphorus (Alexander et al., 2015).
In this field study of a natural archipelagic environment, the microbial plankton community was dominated by mainly phototrophic dinoflagellates (Gonyaulacales, Peridiniales, Suessiaceae) including mixotrophic species, and heterotrophic FIGURE 7 | Taxonomical distribution of bacterial metatranscriptome ORFs grouped into SEED Rank3. Absolute counts mL −1 were log2 transformed and then relative abundances were calculated per timepoint (T1 and T2, in triplicates). Taxonomical groups are given at phylum or class level. (A) Carbon, nitrogen and phosphorous uptake and assimilation, (B) carbon cycling, n, signify the number of ORFs per Rank3 category.
ciliates (Oligotrichia, Choreotrichia) along with primarily heterotrophic bacteria [Actinobacteria (Microbacteriaceae, Sporichthyaceae), Bacteroidetes (Flavobacteriaceae, Cytophagaceae), Proteobacteria (Pseudomonadaceae, Pelagibacter, and Comamonadaceae)], in line with taxa frequently occurring in the Baltic Sea Riemann et al., 2008;Kremp et al., 2009;Herlemann et al., 2011;Wohlrab et al., 2018). Discrepancies in phytoplankton genus/taxa composition between light microscopy observations and 18S rRNA amplicon sequencing have been reported previously (Xiao et al., 2014) and illustrate the benefit of a combined approach. In our study, the discrepancies in PE community structure between observations using microscopy (dominance of Dinophycea) and those resulting from 18S rRNA (dominated by Intramacronucleata) could depend on that sequences in public database are still insufficient for identifying diverse eukaryotic microbes (Kataoka and Kondo, 2019) and that there are large differences in copy number of the 18S rRNA gene between ciliates (high copy number) and other protists such as dinoflagellates (Gong and Marchetti, 2019). Taking this into consideration, we did not pursue a function to taxonomy analysis of the PE community based on molecular results. FIGURE 8 | Model of PE and BAC C, N, and P processes at T1 and T2, with inferred nutrient transport of DOC and Pi, based on metatranscriptome data analysis. Uptake/transport by either high-affinity ABC-transporters, low-affinity transporters or permeases. Processes more highly expressed at T1 (blue), at T2 (red), or at similar levels (white), according to TPM (PE) and absolute counts (BAC) (Supplementary Tables 4, 6). GS, glutamine synthetase; MEP, methylammonium permease; DOC, dissolved organic carbon (allochtonous and autochtonous); Gln, glutamine; complex III and V, of oxidative phosphorylation/electron transport chain; Gln, glycine; Val, valine; Leu, leucine; Ile, isoleucine; His, histidine; ROS, reactive oxygen species; Mn-SOD, manganese sodium dismutase; NorQ, nitric oxide reductase Q; PE, planktonic eukaryotes; BAC, bacteria.

Carbon Partition
The functional analysis of the metatranscriptome indicated a partitioning in the acquisition and uptake of organic and inorganic carbon between the PE and BAC, which contribute to explaining the coexistence of both eukaryotes and prokaryotes (Figure 8). The microplankton PE part of the community was constituted of both phototrophic and heterotrophic taxa. Dinoflagellates were mainly phototrophic, including mixotrophic and primarily heterotrophic genera with kleptochloroplasts or algal endosymbionts (Stoecker, 1999;Gómez, 2012), while the ciliates mostly belonged to heterotrophic taxa  and the mixotrophic Mesodinium rubrum (Lips and Lips, 2017). This explains the occurrence of both carbon fixation and assimilation of amino acids and uptake of sugars observed within this part of the community. During bloom development (T1), carbon uptake associated processes were highly expressed, indicating a larger need for both inorganic and organic carbon, suggesting a more rapidly growing community. This trend is supported by the higher expression of carbonic anhydrase at T1, known to accelerate the HCO − 3 conversion to CO 2 upon inorganic carbon limitation, caused by high levels of aerobic metabolism (Rost et al., 2006). The expression of carbonic anhydrase indicates an increased efficiency of carbon fixation leading to a reduction in oxygen production (Foyer and Harbinson, 1994). This would result in the formation of reactive oxygen species and their scavenger, manganesesuperoxide dismutase. Both carbonic anhydrase and manganesesuperoxide dismutase were expressed at higher levels at T1 compared to T2, an expression pattern previously seen during a dinoflagellate bloom peak (Zhuang et al., 2015). In addition, the number of 18S reads after quality filtration were lower at T1 (avg.) = 213,024, std = 15,201, compared to T2 (avg.) = 222,557, std = 3,330, pointing to a higher richness among the PE during T2, which indicate that the community had progressed through the more homogenous bloom phase (Sison-Mangus et al., 2016;. Taken together, this suggested that the PE community at T1 was sampled during an exponential growth phase while the community at T2 represents a more stationary growth phase. In comparison, BAC had similar expression levels of uptake and assimilation of organic carbon at T1 and T2 (Figure 4C), indicating an actively growing community at both timepoints. Possibly, the difference between growth phases of the phytoplankton is reflected in the variety of organic carbon sources available to the heterotrophic bacteria. During active PE growth (T1), the BAC enzymes were associated with the utilization of the monosaccharides maltose and xylose and the disaccharide trehalose, while during PE slower growth (T2) the expression levels of organic carbon utilization mechanisms encompassed mono-, di-and tri-saccharides, two polyols, of which erythritol was especially highly expressed, and chitinase. The initial bacterial uptake of simple sugars appeared to go with the uptake of lipids and amino-acids, through the expression of phospholipid-lipopolysaccharide ABC transporter and branched-chain amino acid transport permease protein LivM. These carbon sources were likely channeled through gluconeogenesis and the glyoxylate bypass, more costly pathways that allow the conversion of proteins and lipids into glucose, which had higher expression levels at T1. Concomitantly, there was also a high occurrence of transcripts of components in the tricarboxylic acid cycle (TCA), likely providing glucose for growth, via pyruvate metabolism. This implies that during PE rapid growth, the need for carbon was higher also for the BAC, leading to a functional diversification to utilize several different variants of carbon.
While the PE, containing phototrophs, mixotrophs, and heterotrophs, utilizes both inorganic and simple organic carbon sources (amino-acids and sugars), the BAC has a wider functional repertoire to acquire varied forms of organic carbon, including amino-acids and a wide range of sugars, at both timepoints. During bloom development and active PE growth, when there are indications of a higher total energy demand (T1), there are however indications that the PE and BAC possibly compete for the same types of organic carbon (sugars and amino-acids), likely derived from both allochthonous sources and phytoplankton, though the BAC primarily utilizes other sugars than the PE. Taken together, this suggest that the PE and BAC partition the available organic carbon resources, primarily during PE active growth, in order to sustain growth and avoid competition for this valuable resource.

Nitrogen and Phosphorus Partition
Nutrient conditions in the Föglö archipelago indicated inorganic nitrogen (N) limitation and low phosphate concentrations, reflecting summer conditions in the shallow coastal Baltic Sea. In these conditions, phytoplankton and bacteria have to compete for limiting nutrients (Mayali and Doucette, 2002), and coexist despite competition for essential nutrient sources. Limiting levels of inorganic nutrients promote competition between trophic levels (Tilman, 1982), and the conditions in the bay, including the presence of mixotrophic eukaryotes, suggest that nutrientbased competition between eukaryotes and prokaryotes could be expected.
During bloom development (T1), along with a high carbon demand, the actively growing PE showed a high demand for the, likely limited resource, nitrogen. The PE expressed highaffinity N transporters for the inorganic N sources ammonium and nitrate/nitrite (Morey et al., 2011). This type of expression is indicative of low nutrient availability and these transporters have previously been seen to be upregulated in N starved diatoms (Alipanah et al., 2015). Another indication of the high N demand is the expression of cyanate hydratase (Figure 5A), shown to be related to limited nitrogen access (Alipanah et al., 2015). This expression coincides with expression of components for reduction of nitrite and N assimilation via the urea cycle. The expression of cyanate hydratase, primarily at T1, indicates that cyanate was also utilized as an organic N source. Cyanate has previously been observed to be taken up by dinoflagellates (Hu et al., 2012;Zhuang et al., 2015). This broad repertoire for accessing both inorganic and organic N has previously been reported for a coastal dinoflagellate bloom (Zhuang et al., 2015). In our study, the N uptake coincided with the expression of glutamine synthetase, both located in plastids and mitochondria, which is important for both nitrate and ammonium assimilation (Glibert et al., 2016). Expression of glutamine synthetase indicate limiting levels of internal nitrogen, and has previously been shown to be highly expressed in dinoflagellates during N stress (Parker and Armbrust, 2005;Morey et al., 2011;Zhuang et al., 2015). During PE slow growth (bloom stationary phase, T2), lower expression levels of these N assimilation enzymes with the addition of ammonium/methylammonium permease point to a lower N demand. As the PE community diversity increases at T2, a broadening of PE functions is likely, which could lead to a wider N uptake repertoire (Cardinale et al., 2006;Ptacnik et al., 2008). Nitrogen limited conditions seem to enforce the use of many strategies by the PE to access both organic and inorganic N, at both timepoints (Figure 8).
Nitrogen utilization mechanisms by aquatic bacteria in Nlimited conditions have been much studied (Voss et al., 2013). Recent findings highlight the diverse N utilization strategies used by microbial communities to adapt to different N conditions and the ecological importance of the urea metabolism in Nlimited regimes (Li et al., 2018). Our results confirm these findings as both the PE and BAC utilize a variation of strategies to acquire N. While the bacterial N demand was similar at both timepoints, the BAC expressed different N uptake mechanisms than the PE. Early in the bloom, the enzyme allantoicase was actively expressed, meeting the N demand by the intracellular formation of urea from purines (Solomon et al., 2010), as an organic N source. This process is known to occur in cyanobacteria and belongs to purine catabolism (Pomati et al., 2001;Solomon et al., 2010). Nitric oxide reductase activation protein NorQ, along with manganese superoxide dismutase, were also expressed in parallel, which suggests that both ammonium and nitrate transporters were inhibited by nitric oxide (Glibert et al., 2016). During late bloom (T2), direct acquisition of inorganic ammonium, organic urea, and amino acid histidine prevailed in the BAC N uptake strategy, of which all represent nitrogen species frequently associated with heterotrophic bacterial nutrient uptake (Kirchman, 1994;Fouilland et al., 2007).
Altogether, these results support a clear partition of the available N resources between PE and BAC at the different PE growth phases (Figure 8). Where the actively growing PE preferably utilize the scarce inorganic N sources and/or organic N-cyanate to sustain their growth, while the BAC is restricted to degrade organic purines in order to obtain nitrogen. In comparison, the PE, in a more stationary growth phase, utilize less inorganic N while the BAC switch to acquiring N from both inorganic and diverse organic extracellular resources. This illustrates how PE and BAC partition available N resources, in different ways depending on both the PE growth phase and likely the availability of different nitrogen species.
Phosphate is the preferred P substrate for photosynthetic and heterotrophic microbial growth in aquatic systems, and ambient phosphate concentrations were relatively high in the Föglö archipelago compared to the offshore Baltic Sea during summer (Savchuk, 2018). The rapid uptake of phosphate by phytoplankton, and to a minor extent by large macrophytes (e.g., Phragmites in this study) in the coastal zone, the subsequent production of dissolved organic phosphorus (DOP), and the P uptake (adsorption) by organic matter regulate the P availability for the microbial community (White and Dyhrman, 2013). At low phosphate levels, eukaryotic phytoplankton and heterotrophic bacteria can efficiently scavenge P from DOP through hydrolysis using alkaline phosphatase (AP) enzymes, which has been observed both in the coastal Baltic Sea and elsewhere (Chróst, 1991;Tanaka et al., 2006;Traving et al., 2017). Though, during bloom development (T1) in Föglö, high levels of AP were expressed among both BAC and in PE, suggesting limiting levels of internal Pi despite relatively high external Pi (PO 4 ) levels (>0.2 µM). This could be reflecting high carbon and nitrogen turn-over (Baltar et al., 2016) in the coastal bay, rich in dissolved organic matter (DOM). Alternatively, as dissolved organic carbon (DOC) has a contributory effect on microbial AP response (Stewart et al., 1982), the release of simple sugars during PE active growth may trigger higher expression of AP among BAC (Anderson, 2018). Whether it is to alleviate bacterial P limitation or to satisfy BAC organic C demand is unclear. Dinoflagellates, that dominate the PE fraction, have a high diversity of AP genes that can confer diverse strategies to access and utilize DOP (Lin et al., 2012). Whether actively growing mixotrophic dinoflagellates would increase their AP response in relation to organic C sources and not P limitation is unknown. In addition, the PE expressed mechanisms for uptake of several inorganic and organic P sources during active growth, which was not seen in the BAC (Figure 8). This supports that some of the PO 4 transformed by BAC-AP at T1 may be directly used by the PE, and that the BAC mainly used the resulting carbon products (Benitez-Nelson and Buesseler, 1999;Santos-Beneit, 2015). During late bloom (T2), the assumed lower eukaryotic C and N demand may link to the lower AP expression by PE and their expression of a low-affinity sodium dependent Pi transporter. The fact that BAC expressed uptake of two types of high-affinity organic P transporters, along with the expression of the Pi uptake repressor PhoU, suggests that there is a higher demand for organic P, but a lower demand for Pi at this timepoint (Santos-Beneit, 2015). Our findings demonstrate partitioning of the macronutrient P between PE and BAC, over bloom phase, by type (organic or inorganic) and nutrient species, beyond the direct coupling of composition and functional shifts.

Function and Community Composition
In the archipelago of Föglö, northern Baltic Proper, the taxonomic composition, based on amplicon sequencing analyses of both the PE and the BAC, did not vary significantly during the summer bloom while significant differences in resource acquisition were observed. The decoupling between bacterial community taxonomy and function previously suggested by Louca et al. (2016) has recently been evaluated experimentally (Louca et al., 2020) and showed that taxonomy shifts do not systematically affect metabolic rates. Our findings also point toward a decoupling between taxonomic composition and function in the field. For uptake of organic C, the same BAC taxa (Bacteroidetes, Actinobacteria, Alpha-and Beta-proteobacteria) initially utilize trehalose, xylose and maltose while during stationary bloom phase they instead utilize the more complex organic C compounds tagatose/galactitol, erythritol, inositol, raffinose, and chitin. The exception is gammaproteobacterial expression associated with uptake of inositol seen only during the later bloom stage. Traditionally, Bacteroidetes and Actinobacteria are correlated to algal blooms conditions (Riemann et al., 2008) and are often associated to late bloom phase due to their metabolic capacity to degrade HMW carbon compounds (Pinhassi et al., 1997) or polyaromatic compounds such as chitin (Beier and Bertilsson, 2013). This traditional view may perhaps not be so clear cut at this coastal location. Regarding BAC uptake of N, the results suggest that distinct organic and inorganic N resources were acquired and utilized during the two bloom phases. While these transcripts were assigned to several different BAC taxa, the same groups were represented throughout the bloom (Bacteroidetes, Actinobacteria, Cyanobacteria, Alpha-, Beta-, and Gamma-proteobacteria). Similarly, the transcripts associated with BAC acquisition and metabolism of P partly varied between the bloom phases, primarily with higher expression of alkaline phosphatase during the initial stage. These processes were annotated to a variety of BAC taxa, which were all represented during the whole bloom (Acidimicrobiia, Actinobacteria, Bacteroidetes, Cyanobacteria, Alpha-, Beta-, and Gamma-proteobacteria). Taken together, these results suggest that the community composition of BAC in association with a PE bloom is stable despite distinct metabolic and functional variations in C, N, and P resource usage. Thus, a decoupling between taxonomy and function is clear, it does however seem to be dependent on both the abiotic (availability of nutrients) and biotic (PE and BAC interactions) conditions in the bay.

CONCLUSIONS
In this eutrophied archipelagic environment, with high organic:inorganic nutrient ratio, interactions in the form of competition for highly desirable inorganic (nitrate, nitrite, ammonium, Pi) and organic (sugars, amino acids, oligo-peptides, phosphonates, urea, cyanate) nutrients, was characterized by an observed resource partitioning between PE (primarily phototrophs and mixotrophs) and BAC (primarily heterotrophs). This pattern was evident both when the PE were in exponential and stationary phase, implying that uptake and assimilation of resources may be partitioned regardless of growth stage. This resulted in a range of metabolic functions, dependent on the resources available, which did not result in taxonomic shifts, suggesting a decoupling of function from taxonomy among BAC. This study shows the importance of BAC remineralization of allochthonous organic nutrients and of how PE and BAC divide available C, N, and P species, both in type and over time. At unbalanced nutrient conditions, these mechanisms likely enable phytoplankton growth, and even the development of blooms. This study reveals the potential importance of mixotrophic nutrition of PE in brackish coastal systems, which is still not taken into account in biogeochemical models. The results provide insight into how phytoplankton and bacteria in natural communities interact, and opens up for new questions such as how are resources partitioned at low organic:inorganic nutrient ratio, and what effect would a reduction of land run-off have on the microbial community and the nutrient distribution?

AUTHOR CONTRIBUTIONS
ES designed the study, performed the sampling together with EL, handled all bioinformatic work, and led the writing of the manuscript, to which all authors contributed. ES, EL, and HF performed the laboratory work. All authors contributed to the interpretation of the results and the methods used for analysis.

FUNDING
The study was supported by funding from the Strategic Research Environment-ECOCHANGE funded by the Swedish Research Council Formas (CL), the Marie Skłodowska-Curie grant, Grant/Award Number: 659453 (EL), the Linnaeus University Centre of Ecology and Evolution in Microbial Model Systems (EEMiS) (CL and HF).