Water Column Microbial Communities Vary along Salinity Gradients in the Florida Coastal Everglades Wetlands

Planktonic microbial communities mediate many vital biogeochemical processes in wetland ecosystems, yet compared to other aquatic ecosystems, like oceans, lakes, rivers or estuaries, they remain relatively underexplored. Our study site, the Florida Everglades (USA)—a vast iconic wetland consisting of a slow-moving system of shallow rivers connecting freshwater marshes with coastal mangrove forests and seagrass meadows—is a highly threatened model ecosystem for studying salinity and nutrient gradients, as well as the effects of sea level rise and saltwater intrusion. This study provides the first high-resolution phylogenetic profiles of planktonic bacterial and eukaryotic microbial communities (using 16S and 18S rRNA gene amplicons) together with nutrient concentrations and environmental parameters at 14 sites along two transects covering two distinctly different drainages: the peat-based Shark River Slough (SRS) and marl-based Taylor Slough/Panhandle (TS/Ph). Both bacterial as well as eukaryotic community structures varied significantly along the salinity gradient. Although freshwater communities were relatively similar in both transects, bacterioplankton community composition at the ecotone (where freshwater and marine water mix) differed significantly. The most abundant taxa in the freshwater marshes include heterotrophic Polynucleobacter sp. and potentially phagotrophic cryptomonads of the genus Chilomonas, both of which could be key players in the transfer of detritus-based biomass to higher trophic levels.


Introduction
Coastal environments, which are among the most diverse and productive habitats in the world, provide ecosystem services worth trillions of dollars annually [1,2]. Over the last century, freshwater diversion and the conversion of wetlands to agricultural land has reduced the area of global coastal ecosystems by more than two-fold [3]. These developments have made the ecological well-being of coastal ecosystems a pressing global issue, and especially conservation of unique and iconic sites that are inscribed on the UNESCO World Heritage List and considered in 'critical' status, like the Everglades National Park (ENP), are in the focus of local and national environmental policies. Consequently, the Comprehensive Everglades Restoration Plan, with an approximate 50-year construction schedule, was initiated in the 1990s to restore the quantity, timing and distribution of the pre-drainage water flow through the Everglades that was changed profoundly due to urban development, agriculture and extensive drainage [4,5]. The size and color of site markers indicate bacterial abundances and bacterial production, respectively.

Sample Collection
Two liters of surface water samples were obtained from 14 FCE LTER core sampling sites during the subtropical wet season in August of 2017 (Table 1; Figure 1; https://fcelter.fiu.edu/research/sites/, accessed on 18 January 2022). Samples were kept on ice until filtration and were processed within 12 h. To determine bacterial abundances (BA), 9 mL subsamples were separated for flow cytometry and fixed with paraformaldehyde (final concentration 1%; pH = 7.4), incubated at room temperature (RT) for 60 min, and stored at −20 °C until analyses. The microbial communities were separated during filtration into two size fractions using 5 μm (>5 μm fraction) and 0.22 μm (0.22 μm-5 μm fraction) nitrocellulose membranes (MF-Millipore, Darmstadt, Germany). The physicochemical background data and bacterial secondary production values were obtained via standard FCE-LTER protocols (https://fcelter.fiu.edu/data/protocols/index.html, accessed on 18 January 2022). Stations 1-3 in SRS and stations TS/Ph 1-3 are in freshwater marshes. The remaining three stations in SRS (4-6) and stations 6 and 7 of TS/Ph are considered ecotones. Stations TS/Ph 9-11 are situated in Florida Bay. Details on the sites, including coordinates, can be found at: https://fce-lter.fiu.edu/research/sites/index.php (accessed on 18 January 2022). Table 1. Station locations, bacterial abundances (BA) and bacterial productivities (BP) and selected environmental data (Chlorophyll a-Chl a; total nitrogen-TN; total phosphorous-TP).  The size and color of site markers indicate bacterial abundances and bacterial production, respectively.

Sample Collection
Two liters of surface water samples were obtained from 14 FCE LTER core sampling sites during the subtropical wet season in August of 2017 (Table 1; Figure 1; https://fce-lter. fiu.edu/research/sites/, accessed on 14 December 2021). Samples were kept on ice until filtration and were processed within 12 h. To determine bacterial abundances (BA), 9 mL subsamples were separated for flow cytometry and fixed with paraformaldehyde (final concentration 1%; pH = 7.4), incubated at room temperature (RT) for 60 min, and stored at −20 • C until analyses. The microbial communities were separated during filtration into two size fractions using 5 µm (>5 µm fraction) and 0.22 µm (0.22 µm-5 µm fraction) nitrocellulose membranes (MF-Millipore, Darmstadt, Germany). The physicochemical background data and bacterial secondary production values were obtained via standard FCE-LTER protocols (https://fcelter.fiu.edu/data/protocols/index.html, accessed on 14 December 2021). Stations 1-3 in SRS and stations TS/Ph 1-3 are in freshwater marshes. The remaining three stations in SRS (4-6) and stations 6 and 7 of TS/Ph are considered ecotones. Stations TS/Ph 9-11 are situated in Florida Bay. Details on the sites, including coordinates, can be found at: https://fce-lter.fiu.edu/research/sites/index.php (accessed on 14 December 2021).

Flow Cytometry
Samples for flow cytometry were incubated with SYBR Green I nucleic acid stain for 30 min at RT. Flow cytometry analyses were performed on a Guava easyCyte HT (Luminex, Austin, TX, USA) at a flow rate of 0.24 µL s −1 . Cell populations were discriminated via green fluorescence (532 nm) and side scatter channels using a blue laser (488 nm). High nucleic acid and low nucleic acid content bacterial cell counts were pooled together to obtain total bacterial abundances (BA, [22]). Samples were analyzed using Guava's InCyte software (Luminex, Austin, TX, USA).

Molecular Methods
DNA extractions were carried out with the Qiagen PowerWater Kit following the manufacturer's recommended protocol (Qiagen, Hilden, Germany). DNA sequence data was generated using Illumina paired-end sequencing (151 bp × 12 bp × 151 bp MiSeq run) at the Environmental Sample Preparation and Sequencing Facility at Argonne National Laboratory (Lemont, IL, USA). DNA extracts were used as templates for the amplification of the V4 hypervariable region of the 16S rRNA gene (515F-806R primer pair, [23]) and V9 hypervariable region of the 18S rRNA gene (1389F-EukB primer pair, [24]). In addition, primers contained sequencer adapter sequences and the reverse amplification primer also contained a twelve base barcode sequence for multiplexing. Each 25-µL PCR reaction . Once quantified, volumes of each of the products were pooled into a single tube so that each amplicon is represented in equimolar amounts. This pool was then cleaned up using AMPure XP Beads (Beckman Coulter, Brea, CA, USA), and afterwards quantified using a fluorometer (Qubit, Invitrogen). After quantification, the molarity of the pool was determined and diluted to 2 nM, denatured, and diluted to a final concentration of 6.75 pM with a 10% PhiX spike for sequencing on an Illumina MiSeq.

Bioinformatics
The QIIME 2 microbiome analysis package [25], was used for sequence analyses. Quality filtering, chimera identification and merging of paired-end reads was carried out using the DADA2 plugin [26], as implemented in QIIME2. SILVA release 132 (Ref NR 99) taxonomy and q2-feature-classifier were used for classification of 16S rRNA gene sequences [27,28]. Data filtering and statistical analyses were carried out with R version 3.2.0 (R Core Team 2014). Vegan package was used to carry out permutational multivariate analysis of variance using distance matrices (vegdist and adonis functions, with 999 permutations) and perform detrended correspondence analyses (decorana function) with environmental fitting (envfit function) [29]. Sequence variants (SVs) classified as chloroplasts or mitochondria were discarded from the dataset. Demultiplexed raw data was submitted to the Sequence Read Archive under accession number PRJNA525456.

Preprocessing of the Dataset
A total of 15,698 sequence variants (SVs) of microbial eukaryotes and bacteria were obtained from the analyses of 14 water samples ( Table 2). Amplicon sequencing of the 16S rRNA gene (V4 hypervariable region) was used to analyze bacterioplankton community composition (BCC) within the 0.22-5.0 µm size fraction. After quality control and the removal of mitochondrial and chloroplast ribosomal rRNA sequences, a total of 1,207,635 partial 16S rRNA gene sequences were utilized in this study (average of 83,890 sequences per sample, Table 2). All samples were rarefied to 50,000 sequences for statistical analyses. A total of 4755 prokaryotic sequence variants (SVs) were observed, which could be assigned to 562 genus-level taxa (prokaryotic genera, PG, as defined by classification level 6 in QIIME2), out of which 75 contributed more than 0.1% of the sequences in the entire dataset. Table 2. Total number of analyzed sequences, observed sequence variants (SV) and genera-level taxa (Qiime2 classifier, level 6), as well as Shannon diversity indices (H') for 16S rRNA (bac.) and 18S rRNA (euk.) gene data.

Station
Sequences (bac.) Taxa  The composition of eukaryotic communities was determined via 18S rRNA gene amplicon sequencing (>5 µm size fraction). After discarding all sequences classified as Bacteria and Metazoa, a total of 3,201,328 sequences (average 218,606 per sample, Table 2) was used for further analyses. A relatively large fraction of eukaryotic SVs (10,943) remained classified only as Eukaryota (11.0%). To improve these classifications, eukaryotic sequence variants were clustered into operational taxonomic units (OTUs, with 95% sequence similarity threshold), which lowered the fraction of unclassified eukaryotes to 5.4%. There were 55 OTUs (out of a total of 5020) that contributed to at least 1% of sequences in at least one of the samples.

General Overview of the Datasets
In both transects, bacterial and microbial eukaryotic communities were clearly different among freshwater, ecotone, and marine wetlands. Freshwater communities of both transects were similar to each other, while microbial communities in the more saline samples (ecotone, mangroves), and in Florida Bay, were distinctly different ( Figure 2). The bacterial community composition (BCC) was most significantly correlated with salinity (R 2 = 0.92; p < 0.001), concentrations of total nitrogen (TN; R 2 = 0.66; p < 0.01), and chlorophyll a (Chl a, R 2 = 0.63; p < 0.01; Figure 2). Permutational multivariate analysis using distance matrices was used to quantify the effects of environmental factors, transects (SRS, TS/Ph and FB) and ecosystem type (freshwater marshes, ecotone and marine). The ecosystem type better explained variability in BCC (R 2 = 0.44; p < 0.001) than salinity (R 2 = 0.10; p < 0.05), which was the only environmental variable with significant effect (Table S1). When the ecosystem types in different transects were split into separate groups, even more variability in BCC was explained (R 2 = 0.58; p < 0.001).

General Overview of the Datasets
In both transects, bacterial and microbial eukaryotic communities were clearly different among freshwater, ecotone, and marine wetlands. Freshwater communities of both transects were similar to each other, while microbial communities in the more saline samples (ecotone, mangroves), and in Florida Bay, were distinctly different ( Figure 2). The bacterial community composition (BCC) was most significantly correlated with salinity (R 2 = 0.92; p < 0.001), concentrations of total nitrogen (TN; R 2 = 0.66; p < 0.01), and chlorophyll a (Chl a, R 2 = 0.63; p < 0.01; Figure 2). Permutational multivariate analysis using distance matrices was used to quantify the effects of environmental factors, transects (SRS, TS/Ph and FB) and ecosystem type (freshwater marshes, ecotone and marine). The ecosystem type better explained variability in BCC (R 2 = 0.44; p < 0.001) than salinity (R 2 = 0.10; p < 0.05), which was the only environmental variable with significant effect (Table  S1). When the ecosystem types in different transects were split into separate groups, even more variability in BCC was explained (R 2 = 0.58; p < 0.001).  At a higher taxonomic level, BCC followed patterns documented for other estuarine and wetland ecosystems with a strong salinity gradient [30,31], with Betaproteobacteria and Actinobacteria dominating low salinity environments, and Alphaand Gammaproteobacteria dominating coastal marine environments. Relative abundances of the Bacteroidetes phylum stayed relatively constant in both fresh and saltwater environments (Figure 3). Although the dataset contained 101 different class-level bacterial taxa, 94.4% of sequences could be assigned to just ten higher taxa: Betaproteobacteria (39.2%), Alphaproteobacteria (14.4%), Actinobacteria (13.6%), Gammaproteobacteria (10.3%), Flavobacteriia (5.7%), Sphingobacteriia (3.2%), Spartobacteria (3.2%), Cyanobacteria (2.3%), Proteobacteria Incertae Sedis (1.9%) and Acidimicrobiia (0.6%) (Figure 3). In total, 562 genus-level taxa were identified, out of which 75 contributed to more than 0.1% of all the sequences and can therefore be considered 'common' in the ecosystem (Figure 4). At a higher taxonomic level, BCC followed patterns documented for other estuarine and wetland ecosystems with a strong salinity gradient [30,31], with Betaproteobacteria and Actinobacteria dominating low salinity environments, and Alpha-and Gammaproteobacteria dominating coastal marine environments. Relative abundances of the Bacteroidetes phylum stayed relatively constant in both fresh and saltwater environments (Figure 3). Although the dataset contained 101 different class-level bacterial taxa, 94.4% of sequences could be assigned to just ten higher taxa: Betaproteobacteria (39.2%), Alphaproteobacteria (14.4%), Actinobacteria (13.6%), Gammaproteobacteria (10.3%), Flavobacteriia (5.7%), Sphingobacteriia (3.2%), Spartobacteria (3.2%), Cyanobacteria (2.3%), Proteobacteria Incertae Sedis (1.9%) and Acidimicrobiia (0.6%) (Figure 3). In total, 562 genus-level taxa were identified, out of which 75 contributed to more than 0.1% of all the sequences and can therefore be considered 'common' in the ecosystem (Figure 4). . Bacterioplankton community composition on order-level taxa (top) and eukaryotic microbial community composition represented by variable higher taxa (bottom). In both cases, only taxa that contributed to more than 1% of the respective sequence datasets were included.
Like the BCC, the composition of the eukaryotic microbial communities (EMC) also varied along the physicochemical gradients ( Figure 2). Permutational multivariate analysis demonstrated that in the case of the EMC the ecosystem type (R 2 = 0.27; p < 0.001; Table S2) also outperformed salinity (R 2 = 0.11; p < 0.05) and Chl a (R 2 = 0.09; p < 0.05) which there were only two environmental variables with a significant effect. Differences between Figure 3. Bacterioplankton community composition on order-level taxa (top) and eukaryotic microbial community composition represented by variable higher taxa (bottom). In both cases, only taxa that contributed to more than 1% of the respective sequence datasets were included.
Most sequences that could be assigned to eukaryotic taxa included members o Ochrophyta (27.3%), Cryptophyta (15.0%), Ciliophora (14.3%), Dinoflagellata (13.2% Chlorophyta (2.3%) and Fungi (2.3%) (Figure 3). Clear shifts in community compositio were evident even for higher taxonomic levels in both BCC and EMC. Therefor freshwater, ecotone and marine estuary ecosystems are discussed separately below for more comprehensive overview.  Like the BCC, the composition of the eukaryotic microbial communities (EMC) also varied along the physicochemical gradients ( Figure 2). Permutational multivariate analysis demonstrated that in the case of the EMC the ecosystem type (R 2 = 0.27; p < 0.001; Table S2) also outperformed salinity (R 2 = 0.11; p < 0.05) and Chl a (R 2 = 0.09; p < 0.05) which there were only two environmental variables with a significant effect. Differences between the transects were not significant. However, discriminating ecosystem types between transects constrained more variability than ecosystem types alone (R 2 = 0.41; p < 0.005).

Freshwater Marsh Communities
Stations 1-3 in SRS and stations TS/Ph-1-2 are in freshwater marshes. TS/Ph-3 is heavily impacted by water flow (tides and freshwater input), had a very low salinity at the time of sampling (Table 1), and was thus clustered with the other freshwater sites. The lowest bacterial cell abundances were found in freshwater stations compared to the other stations, in both transects (in TS/Ph-1-3 around 1.2 × 10 6 cells L −1 and 1.0 × 10 6 up to 1.4 × 10 6 cells L −1 in SRS 1d-3; Figure 1). Bacterial secondary production (BP), a proxy for the integration of DOM into bacterial biomass, did not exhibit a clear trend and varied between the freshwater stations ( Figure 1). In TS/Ph, BP rates decreased about five times from TS/Ph-1 (36.5 µg C L −1 d −1 ) to TS/Ph-3, while the highest values for the entire dataset were observed in SRS3 (95.8 µg C L −1 d −1 ) (Figure 1).

Ecotone
Stations 4-5 in SRS are situated within mangrove forests and TS/Ph-6 and TS/Ph are located in an estuarine ecotone with mangrove islands [32]. The highest bacteri abundances in our datasets were found in ecotone stations bordering marine estuarie reaching 3.9 × 10 6 cells L −1 in TS/Ph-7 (Table 2). In the TS/Ph transect, BP peaked in TS/Ph

Ecotone
Stations 4-5 in SRS are situated within mangrove forests and TS/Ph-6 and TS/Ph-7 are located in an estuarine ecotone with mangrove islands [32]. The highest bacterial abundances in our datasets were found in ecotone stations bordering marine estuaries, reaching 3.9 × 10 6 cells L −1 in TS/Ph-7 (Table 2). In the TS/Ph transect, BP peaked in TS/Ph-6 (71.9 µg C L −1 d −1 ) and decreased nearly 20-fold towards the estuary. A similar trend was observed in SRS, where rates dropped from 42.9 to 22.6 µg C L −1 d −1 within the ecotone (Figure 1).
The major differences in the BCC of the ecotones compared to the freshwater stations were decreases of the relative abundances of Betaproteobacteria (to 33.4%), Sphingobacteriia (to 3.6%) and Spartobacteria (to 0.1%), and concomitant increase in the relative abundances of Actinobacteria (to 20.6%), Gammaproteobacteria (to 13.6%), Alphaproteobacteria (to 7.7%), Proteobacteria Incertae Sedis (to 5.2%) and Cyanobacteria (to 3.6%). In parallel, increases in the relative abundances of eukaryotic taxa were found among Ochrophyta (to 34.6%) and Ciliophora (to 21.7%). Significant differences were also detected among both the BMC and EMC of the two distinct ecotones. Members of Micrococcales (Actinobacteria) became the second most abundant prokaryotic group in the TS/Ph ecotone (30.4%), while contributing only to 4.9% in waters of the SRS ecotone. Similarly, Synechococcales and Rhizobiales were found to be more abundant in the ecotone of Taylor Slough (on average 7.7% and 7.6%, respectively), while they both contributed to less than 1% in the Shark River Slough ecotone. A large fraction of BCC at the SRS ecotone were identified as Ectothiorhodospirales (17.4%), Thiotrichales (8.6%) and Frankiales (7.9%), while these three orders were absent from the TS/Ph-ecotone. The relative abundance of Flavobacterales were similar between the ecotones, contributing 5.1% at TS/Ph and 7.4% at SRS.

Freshwater Marsh Communities
Long-term monitoring data indicated that abrupt and sustained increases in TP and DOC from marine storm surges and severe low-temperature events increase bacterioplankton productivity for extended periods, and that these responses are more pronounced in SRS than in the TS/Ph transect [36]. Despite these striking differences in BP, the composition of the freshwater microbial communities was not significantly different between the two transects.
The abundance of Betaproteobacteria in numerous freshwater habitats is heavily affected by salinity [37][38][39][40]. A similar trend is apparent in this study as well. The most abundant Betaproteobacteria were composed of two genus-level taxa: Polynucleobacter (PG002, 26.0%) and MWH-UniP1 aquatic group (PG001, on average 21.6%). Both taxa belong to the order Burkholderiales, together with less abundant PG016 (unclassified) and PG050 (Limnobacter), which made up 3.9% and 0.4% of all bacterial sequences in the freshwater marshes, respectively.
Polynucleobacter-related sequences were divided between two free-living species P. cosmopolitanus and P. asymbioticus, which are ubiquitous and frequently abundant members of freshwater bacterioplankton [41][42][43] in habitats that vary in chemical and climatic conditions [44][45][46]. As 16S rRNA data does not provide sufficient resolution for the identification of Polynucleobacter species [47], the composition and ecological function of these highly abundant organisms in the marshes of the Everglades warrants further detailed studies. Cultured strains of P. asymbioticus originate mainly from humic-rich habitats, where they can utilize products of photodegradation of humic substances [42,48]. The high relative abundance of Polynucleobacter in the Everglades watershed is likely explained by the prevailing high concentrations of humic substances, which compose about 50% of the DOC in this environment and are in part mineralized by solar radiation [49].
Verrucomicrobia, which were mostly represented by the order Chtioniobacterales in the class Spartobacteria, had a relative abundance of 7.3% in the ENP freshwater marshes (Figure 3). These bacteria are present in a large variety of terrestrial and aquatic ecosystems and are also a dominant group in many humic lakes, composing up to 19% of the respective BCCs [50][51][52]. Most Verrucomicrobia are specialized in the degradation of algal polymers, specifically polysaccharides, such as cellulose and chitin [53]. Therefore, their occurrence is usually correlated with the biomass dynamics of phytoplankton, including Chrysophyceae [54], a group that was also present in the freshwater stations in ENP. There were many similarities between the BCC of the ENP transects and that of the Brazos and Mississippi Rivers that also flow into GoM, including high relative abundances of Limnohabitans, Polynucleobacter, acI clade, LD28 and others [19,55,56].
Cryptophyta have mostly been considered autotrophic, but exceptions were reported for cryptophytes in ice-covered lakes in Antarctica, where mixotrophic behavior for survival under light-limited conditions was observed during winter [57,58]. Williams and Trexler [59] demonstrated the importance of 'detrital' carbon flow in the Everglades, indicating that the microbial loop provides a major route of energy flow to higher trophic levels. The most abundant group of cryptophytes in the marshes, Chilomonas, are heterotrophs that can feed on detritus in the form of particulate organic matter [60] and could therefore be a key taxon in the carbon cycle of the Everglades.
Fungi are considered important heterotrophic degraders in periphytic communities [15,61] and could contribute to same processes in the water column. The highest relative abundances of Fungi were observed in the freshwater marshes (8.3% in SRS1d and 10.6% at TS/Ph-2). The most abundant fungal OTU (OTU29) could only be classified as a member of the division Glomeromycota, which includes all species involved in arbuscular mycorrhizal symbioses. We cannot exclude the possibility that these sequences come from spores rather than metabolically active cells.

Ecotone
Ecotones are transitions between ecosystems that are characterized by steep environmental gradients [62]. They are critical in regulating the transport of DOM and nutrients into coastal waters [63][64][65]. The hydrographic changes in the Everglades over the last 100 years have had the most impact on its ecotones [66].
Mangrove plant roots excrete organic compounds and release oxygen to the rhizosphere, thus changing the chemical characteristics in the sediment area around the roots [67][68][69]. Nevertheless, mangrove sediments are primarily anaerobic with a thin aerobic sediment layer on top [69,70], thus providing chemical characteristics which encourage anoxygenic photolithotrophic and chemolithotrophic sulfur-oxidizing bacteria, which in this study were also found in the water column of the SRS ecotone (e.g., Ectothiorhodospirales, Thiotrichaceae and SUP05 cluster). Our results indicate that anaerobic processes in the rhizosphere have a high impact on microbial communities in overlaying water column.
Diatoms showed increased diversity and relative abundance towards the ecotone in our dataset, most notably along the SRS transect ( Figure 5). Concomitant with the increase of relative abundance in diatoms, an increase in the abundances of bacterial taxa that have been shown to be associated with diatom blooms, like Flavobacteriales and Rhodobacterales [71] was also observed ( Figure 4). Representatives of Flavobacteriales are well-known degraders and consumers of high-molecular-mass organic matter [53,72,73], and Rhodobacterales are known to utilize exopolymer particles [74] and could therefore play an important role in the degradation of these compounds in the ecotones in the Everglades.

Florida Bay
FB is a large and shallow estuary, with an average depth of only 1.5 m [75]. In our dataset, the samples from the FB are represented by stations TS/Ph-9, TS/Ph-10 and TS/Ph-11 ( Figure 1). The bay receives freshwater runoff from the Everglades marsh mainly through the C-111 Canal and Taylor Slough. Its west side opens to the GoM, the main source of phosphorous for FB, resulting in a phosphorous gradient between the eastern and western parts of FB [76]. Drainage canals are a major source of contamination to the local reef environments, and the nutrients that leak into the estuaries in South Florida lead to occasional/regular algal blooms. Florida Bay waters have been divided into six segments based on their biogeochemical characteristics [77,78]. Our data are in line with some of these general features, as the northern part of the bay (TS/Ph-9) has higher nutrient concentrations, and the central part of the bay (TS/Ph-10) has higher Chl a level ( Table 1). The highest bacterial cell counts and BP within the FB were also observed in TS/Ph-10 (2.8 × 10 6 cells L −1 ; 13.8 µg-C L −1 d −1 ), surpassing the corresponding values of TS/Ph-9 (1.5 × 10 6 cells L −1 ; 4.0 µg-C L −1 d −1 ).
The highly variable abundance of certain taxa within the FB sites indicates that a higher spatiotemporal resolution is needed to accurately describe the composition of microbial plankton communities in these habitats. The northeastern part of the bay (TS/Ph-9) had higher inorganic nutrient concentrations (Table 1) and autotrophic diatoms dominated the EMC; Cyclotella (OTU4, 27.4%) and Chaetoceros (OTU11, 17.8%) were the most abundant genus-level taxa. The accompanying BCC contained a large fraction of SAR11 clade (22.2%) and Rhodospirillales (15.5%) that specialize in the active uptake of low-molecular weight monomers during diatom dominated phytoplankton blooms [71,79]. Flavobacteriales made up 9% of BCC at TS/Ph-9 and might be key players in the initial degradation of organic matter derived from the observed algae [53,72,73]. The closest station to the Gulf of Mexico, TS/Ph-11, exhibited a high abundance of the NS5 marine group (Flavobacteriaceae, 23.7%), Roseobacter (12.2%) and Oceanibaculum (SAR116 clade, 8.3%), which are also abundant in northern parts of the Gulf of Mexico [19,80].

Conclusions
Pelagic microbial communities differed significantly in habitats along the salinity gradient in the Florida Coastal Everglades. In the freshwater marshes, detrital 'brown' carbon flow is essential for the food web and, accordingly, the most abundant organisms were heterotrophs that are presumably capable of degradation of complex organic carbon. In these habitats, solar radiation generates dissolved organic matter via photo-dissolution of flocculent, detrital material and terrestrial humic-like components that can contribute up to 70% of the chromophoric DOM [81]. Polynucleobacter, the most abundant prokaryotic group that we detected in the marsh samples, have been shown to utilize products of photodegradation humic substances-a capability that might explain their abundance in this ecosystem. Potentially phagotrophic cryptomonads of the genus Chilomonas were identified as predominant eukaryotic microorganisms, indicating that they could hold a key position in this ecosystem by transferring detritus-based biomass to higher trophic levels. In wetland ecotones, oxygen production and excretion of organic matter by mangrove roots, as well as the increased concentration of sulfate from marine waters, create niches for aerobic as well as anaerobic microbial communities, even in the water column. We identified photolithotrophic and chemolithotrophic sulphur-oxidizing bacteria (Ectothiorhodospirales and Thiotrichales) as predominant members of BCC at the SRS ecotone. Similarly, marine microbial communities at the three sampling sites in Florida Bay were heterogeneous, indicating the presence of (micro)niches and the need of higher resolution of sampling sites. Further data is necessary to pinpoint these trends and to analyze the potential seasonality of microbial communities in these systems. Nevertheless, the datasets presented here will provide valuable baseline information for environmental monitoring in these habitats.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/microorganisms10020215/s1, Table S1. Quantitative effects of environmental parameters, ecosystem type (freshwater marshes, ecotone, marine) and transect (SRS, TS/Ph, FB + ) on the bacterial community variation by permutational multivariate analysis of variance (using 'adonis' function of vegan package) based on weighted UniFrac distance. R 2 values present the proportion of variation constrained by factors. Table S2. Quantitative effects of environmental parameters, ecosystem type (freshwater marshes, ecotone, marine) and transect (SRS, TS/Ph, FB + ) on the eukaryotic microbial community variation by permutational multivariate analysis of variance (using 'adonis' function of vegan package) based on weighted UniFrac distance. R 2 values present the proportion of variation constrained by factors.

Conflicts of Interest:
The authors declare no conflict of interest.