Heterotrophic Nitrogen Fixation at the Hyper-Eutrophic Qishon River and Estuary System

Planktonic heterotrophic diazotrophs (N2-fixers) are widely distributed in marine and freshwater systems, yet limited information is available on their activity, especially in environments with adverse conditions for diazotrophy (e.g., N-rich and oxygenated). Here, we followed the localization and activity of heterotrophic diazotrophs in the hyper-eutrophic N-rich Qishon River—an environment previously considered to be unfavorable for diazotrophy. Our results indicate high heterotrophic N2 fixation rates (up to 6.9 nmol N L–1 d–1), which were approximately three fold higher at an upstream location (freshwater) compared to an estuary (brackish) site. Further, active heterotrophic diazotrophs were capture associated with free-floating aggregates by a newly developed immunolocalization approach. These findings provide new insights on the activity of heterotrophic diazotrophs on aggregates in environments previously considered with adverse conditions for diazotrophy. Moreover, these new insights may be applicable to other aquatic regimes worldwide with similar N-rich/oxygenated conditions that should potentially inhibit N2 fixation.

It is generally accepted that N 2 fixation occurs by cyanobacterial diazotrophs in N-poor, sunlit environments, and that high levels of DIN can inhibit this process (Karl et al., 2002;Gruber, 2008;Knapp, 2012). However, several studies reported unexpectedly high N 2 -fixation rates in N-rich environments such as aphotic waters (Hamersley et al., 2011;Rahav et al., 2013;Benavides et al., 2015), estuaries (Subramaniam et al., 2008;Bentzon-Tilia et al., 2014;Bombar et al., 2016;Pedersen et al., 2018), nutrient-rich coastal waters (Mulholland et al., 2019), and in laboratory settings using N-amended cultures of diazotrophic cyanobacteria (Fu and Bell, 2003;Knapp, 2012). In many of these environments, a high diversity of heterotrophic diazotrophs, rather than phototropic cyanobacterial N 2 -fixers, is often found (Riemann et al., 2010), suggesting other controls and possibly mechanisms that enable N 2 fixation by these heterotrophic microorganisms. While numerous studies have investigated the limiting factors for cyanobacterial diazotrophs (reviewed in Moore et al., 2013), much less is known about the nutrient controls of aquatic heterotrophic diazotrophs. One of the ways heterotrophic diazotrophs could flourish in the abovementioned aquatic environments is by adopting a particle-associated lifestyle rather than being free-living (Rahav et al., 2013;Bombar et al., 2016;Farnelid et al., 2019).
To date, several studies suggested that free-floating aggregates may be favorable loci for heterotrophic N 2 fixation due to its unique physicochemical characteristics (Paerl and Carlton, 1988;Rahav et al., 2013Rahav et al., , 2016Pedersen et al., 2018 and see more details below). Free-floating aggregates often referred to as marine, lake, or river snow (Grossart et al., 1997;Simon et al., 2002;Lundgreen et al., 2019) are ubiquitous throughout the aquatic environment and play a central role in marine and freshwater food webs as micro-islands with intense microbial activity (Azam and Malfatti, 2007;Bar-Zeev and Rahav, 2015;Arnosti et al., 2016). In comparison to the surrounding waters, these aggregates usually contain high levels of organic and inorganic compounds, trace elements, as well as detrital matter and fecal pellets held together by a sticky scaffold comprised of proteins and polysaccharides such as transparent exopolymer particles (TEP) (Ploug and Grossart, 2000;Passow, 2002). These aggregates usually have a high C:N ratio compared to the typical ∼6.6:1 Redfield ratio, thereby inducing N-limiting conditions for different microbes, including heterotrophic diazotrophs. Indeed, once formed, aggregates are often heavily colonized by various microorganisms (del Giorgio and Cole, 1998;Simon et al., 2002;Bar-Zeev et al., 2012), including heterotrophic diazotrophic representatives affiliated with nifH Cluster III (e.g., Desulfovibrio putealis) and Cluster 1 (e.g., Azoarcus sp., Dechloromonas aromatic, and Rhodobacterales) (Pedersen et al., 2018;Farnelid et al., 2019). Bacterial abundances associated with aggregates are several orders of magnitudes higher compared to the surrounding water (Grossart and Simon, 1993;Turley and Mackie, 1994;Kiørboe et al., 2002). Bacteria/archaea associated with aggregates solubilize and re-mineralize organic matter at higher rates compared to free-living bacteria (Grossart and Simon, 1998), resulting in "hotspots" of intense microbial activity (Long and Azam, 2001;Bar-Zeev and Rahav, 2015;Arnosti et al., 2016), potentially also to heterotrophic diazotrophs (Rahav et al., 2013;Bombar et al., 2016;Farnelid et al., 2019). Further, the high aerobic respiration by bacteria colonizing aggregates (including heterotrophic diazotrophs), combined with slow diffusion rates, may lead to reduced oxygen levels toward the aggregate's center of ≤80% air saturation and occasionally even to anoxia conditions (Paerl and Prufert, 1987;Klawonn et al., 2015). Such oxygen-reduced micro-zones may greatly benefit diazotrophs since the nitrogenase enzyme may be irreversibly damaged by O 2 (Gruber, 2008). Moreover, O 2 -protective mechanisms used by different diazotrophs have been shown to consume much of the energy required for N 2 fixation (Großkopf and LaRoche, 2012). Currently, it is unknown how the nitrogenase in heterotrophic diazotrophs is protected from O 2 , and it is possible that aggregates/fecal pellets provide low O 2 microenvironment that "enable" N 2 fixation. There is currently limited information on the activity of planktonic heterotrophic diazotrophs and their association with aggregates in eutrophic freshwater environments.
In this study, we measured heterotrophic N 2 fixation in a DIN-rich aquatic environment and established a direct association between active heterotrophic diazotrophs (that synthesized nitrogenase) and polysaccharide aggregates. To this end, we sampled the eutrophic Qishon River (SE Mediterranean Sea) at two sites with different C:N and N:P ratios that may affect heterotrophic diazotrophy. Specifically, heterotrophic N 2 fixation, bacterial abundance, bacterial activity, and TEP concentrations were measured along with a suite of different physicochemical variables during summer and winter. Since the focus of this study was to examine heterotrophic N 2 fixation, water collected was incubated in the dark and with the addition of a photosynthetic inhibitor. Further, we used an immunolabeling approach to specifically localize diazotrophs actively expressing nitrogenase on aggregates comprising polysaccharides. Our results suggest that particles could be loci for N 2 fixation by heterotrophic bacteria in eutrophic environments traditionally considered to be unfavorable for diazotrophy.

Study Sites and Sampling Strategy
Surface water (∼0.3 m) was collected from the Qishon River at two sites; one located downstream at the entrance of the estuary to the coast (hereafter referred to as "estuary, " Lat: 32 • 48 44.42"N, Lon. 35 • 02 00.6 E), and one upstream (hereafter referred to as "stream, " Lat: 32 • 43 34.5 N, Lon.  Table S1). The collected water was divided into four pre-cleaned (10% HCl and autoclaved) 1-L Nalgene bottles and amended with artificial estuary water enriched in ultra-pure 15 N 2 (Mohr et al., 2010). Specific focus on the activity of heterotrophic diazotrophs (rather than cyanobacterial diazotrophs) was achieved by incubating the samples for 48 h in the dark with 3-(3,4-dichlorophenyl)-1,1-dimethylurea (DCMU, 50 µM final concentration, Sigma-Aldrich D2425) to impair phototrophic activity (Clavier and Boucher, 1992). An additional bottle was collected and filtered immediately to measure the natural abundance of dissolved 15 N 2 (no tracer addition) (Montoya et al., 1996). Previous studies showed that prolonged dark incubation (48 h) and addition of DCMU can halt autotrophic diazotrophy, thereby providing valuable information specifically on heterotrophic N 2 fixation (Rahav et al., , 2016Benavides et al., 2018). In addition, the association between aggregates and diazotrophs was visualized using a recently developed confocal microscopy approach at the conclusion of the incubations (under dark+DCMU conditions as well as under ambient light as controls). Natural aggregates collected from the stream and estuary are shown in Geisler et al. (2019).

Physicochemical Characteristics
Light penetration through the water column was measured using a Secchi disk (Holmes, 1970). Salinity was measured using an electrical conductivity meter (EC-30 cond, Phoenix Instruments), dissolved oxygen (DO) using an oxygen meter (ProODO, YSI), pH using a pH meter (Cybercan pH 11, Eutech), and turbidity using a turbidity meter (Tu-2016, Lutron). All sensors were calibrated based on the manufacturer's instructions. Total phosphorus (TP), orthophosphate (PO 4 ), and nitrate+nitrite (NO 3 +NO 2 ) were measured using a segmented flow Technicon Auto-Analyzer II (AA-II) system. The detection limit for P was 2.5 × 10 −4 mg L −1 , while for N it was 5.6 × 10 −4 mg L −1 (Sisma-Ventura and Rahav, 2019). Total organic carbon (TOC) and total nitrogen (TN) were analyzed with a TOC analyzer (Multi N/C, Analytik-Jena, Germany, detection limit; 0.3 mg L −1 ) and calibrated using a fivepoint calibration procedure (Kowalski et al., 2009). These physicochemical measurements were carried out to characterize the initial environmental characteristics of the water at both sites and were measured during September 2017 and January 2018 (same as the visualization approach described below).

Transparent Exopolymer Particles (TEP) Concentration
Subsamples (<100 mL) were filtered through a 0.4-µm polycarbonate filter (GE Water & Process Technologies) using low pressure (<150 mbar) to avoid breakdown of the aggregates at the end of the incubation (T 48 ). Filters were stained with a 0.2% Alcian blue solution and washed three times with deionized water. TEP was extracted using sulfuric acid (80%) for 2 h. The supernatant (1 ml) absorbance was measured using a spectrophotometer at 787 nm wavelength (Thermo GENESYTM). Alcian-blue dye was calibrated against known concentrations of the purified polysaccharide gum-xanthan (GX) (Passow and Alldredge, 1995).

Bacterial Abundance (BA)
Water samples (1.7 mL) were fixed with glutaraldehyde (final concentration 0.02%, Sigma-Aldrich G7651) at T 48 , flash frozen in liquid nitrogen, and stored at −80 • C until analyses within a few days. Prior to analyses, the samples were thawed at room temperature, sonicated for 30 s (Qsonica, Q55 Sonicator), and EDTA (2.5 µM) was added to release the cells associated/attached to the aggregates to the water media and stained with 0.5 nM SYBR Green for 10 min in the dark. Subsamples were analyzed with an Attune NEXT flow cytometer (Applied Biosystems) using a blue laser signal. Samples were run at a constant flow rate of 25 µL min −1 and cells were differentiated using green fluorescence and side scatter (Vaulot and Marie, 1999). Cell size range in an Attune NEXT system is ∼0.5-70 µm. Beads (0.93 µm, Polysciences) were run in parallel as a size standard. Blank samples of stained sterile river water (0.2 µm) and deionized water were also run and their reads were removed from the total bacterial counts.

Bacterial Production (BP)
Triplicate samples (1.7 mL) were spiked at T 48 with 2 nM of 3 H-leucine (123 Ci mmol −1 , Perkin Elmer) and incubated for 3-4 h under ambient temperature in the dark. Incubations were terminated by adding 100% trichloroacetic acid, followed by applying the micro-centrifugation technique (Simon and Azam, 1989;Smith et al., 1992). Disintegration per minute (DPM) was measured using a Tri-CARB 4810 TR (Packard) liquid scintillation counter. Incorporation of leucine was converted to carbon by a conservative factor of 3.1 kg C mol −1 with an isotope dilution factor of 2.0.

Heterotrophic Dinitrogen (N 2 ) Fixation
Measurements were taken by adopting the 15 N 2 -enriched water method (Mohr et al., 2010), without degassing pre-step. The enriched medium was prepared before each sampling event (summer or winter) by injecting 15 N 2 gas (99%, Cambridge Isotopes, lot # NLM-363-PK) into pre-filtered (0.2 µm) artificial estuary water at a 1:100 (v:v) ratio (i.e., 1 cc of 15 N 2 gas per 100 ml of artificial water). The enriched stock was vigorously shaken to completely dissolve the 15 N 2 gas bubble. The same 15 N 2 tank was used to prepare the enriched water for both the summer and winter sampling campaigns. We used the atom % value provided previously using similar methodology and recipe for preparing the enriched seawater (Wilson et al., 2012) as no membrane-introduction mass spectrometry (MIMS) measurements were available. The percent enrichment can vary between batches when preparing 15 N 2 -enriched water. However, since we used the same 15 N 2 -enriched water in each sampling event, a comparison between the estuary and stream sites at the Qishon River reflects the measured differences between the sites. 15 N 2 stock was added to the 1 L experimental bottles (5% of total sample volume, Rahav et al., 2015) and incubated for 48 h in the dark with DCMU to impair phototrophic diazotrophy Benavides et al., 2018) and focus specifically on heterotrophic diazotrophs. Preliminary experiment from the Northern Red Sea (and not the Qishon River) showed that N 2 fixation rates were insignificantly different following dark alone or dark+DCMU incubations, thereby refuting any concern that the addition of DCMU may result in dissolved organic matter supply derived from autotrophs death/lysing which may favors heterotrophic microbial activity (including diazotrophy) (Supplementary Figures S1, S2). The samples were filtered through pre-combusted (450 • C, 4.5 h) glass fiber filters and dried overnight at 60 • C. A minimum of 10 µg particulate N (PN) per filter ensured adequate sample mass to resolve small differences in N isotope ratios (White et al., 2020). The samples were analyzed on a CE Instruments NC2500 elemental analyzer interfaced to a Thermo-Finnigan Delta Plus XP isotope ratio mass spectrometer (IRMS). Heterotrophic N 2 fixation was calculated according to Mulholland et al. (2006) using N solubility factors described by Weiss (1970). One bottle without 15 N enrichment was used for natural abundance of N 2 at each station. Standard curves were processed with each sample run to determine N for isotope ratio mass spectrometry. Samples were run only when standard curves had R 2 > 0.99. At masses >4.7 µg N, precision for the atom percent 15 N measurement was ±0.0001% based on daily calibrations made in association with sample runs and calibrations averaged over runs made over several years. Based on natural abundance, N mass on the filters, incubation times, and the precision of mass spectrometer, the detection limit for 15 N uptake was ∼0.02 nmol N L −1 d −1 . This detection limit was lower by ∼70-95% of the heterotrophic N 2 fixation rates measured in this study, thus providing credibility to the results.

Visualization of Active Diazotrophs on Polysaccharide-Based Aggregates
The staining protocol is described in detail in Geisler et al. (2019). Briefly, sub-samples (∼20 mL) were collected from the microcosm bottles at the end of incubations (T 48 ) during the September 2017 and January 2018 sampling campaigns (Supplementary Table S1), gently filtered (>150 mbar) on a 0.4 µm filter and stained for the following: (i) diazotrophs were tagged with an anti-nitrogenase (Agrisera Antibodies AS01 021A) solution (6 µg mL −1 ) followed by an anti-chicken antigen conjugated to a FITC fluorophore (Thermo Fisher Scientific A-11039) for 45 min, (ii) Total bacteria were stained with DAPI for 45 min, (iii) The polysaccharide matrix of the aggregates was stained with the fluorescent lectin concanavalin A (ConA, 200 µg mL −1 , Ex. 630 nm, Em. 647 nm, Thermo Fisher Scientific C11252) for 40 min, and (iv) Heterotrophic bacteria were distinguished from phototrophic cyanobacteria by subtracting the auto-fluorescence of phycoerythrin (Ex. 490 nm and Em. 580 nm). Although indicative to most cyanobacteria (not necessarily diazotrophs), phycoerythrin is not found in all strains, and thus, it is possible that some phototrophic diazotrophs were not identified. The stained samples were imaged with a Zeiss confocal laser scanning microscope (CLSM 510 Meta) equipped with 405 nm diode, 488 nm Aragon, and 633 nm helium-neon lasers. Captured CLSM images were processed using ZEN (blue edition). We surmised that most of the bacteria that have translated the nitrogenase enzyme would actively fix N 2 , resulting in a positive FITC fluorophore staining (Geisler et al., 2019). Additional controls for the visualization approach are provided in Supplementary Figures  S2-S4 and include (i) testing if the addition of DCMU introduced carbon-rich substrates stimulated heterotrophic N 2 fixation (Supplementary Figure S2); (ii) incubations under ambient light that capture the activity of both autotrophic and heterotrophic diazotrophic activity (Supplementary Figure S3); and (iii) testing unspecific staining of the secondary antibody (Supplementary Figure S4).

Statistical Analyses
The statistical significance between the values of heterotrophic N 2 fixation, bacterial production (BP), bacterial abundance (BA) Frontiers in Microbiology | www.frontiersin.org and TEP at the "estuary" or "stream" stations was determined using a student's t-test using XLSTAT 2016 (Microsoft, New York, United States) with a confidence level of 95% (α = 0.05).
The identification of active diazotrophs in association with aggregates comprising polysaccharides such as TEP in the Qishon River was visualized using a recently developed immunolabeling approach (Geisler et al., 2019). This approach enabled direct visualization of active diazotrophs that synthesized the nitrogenase enzyme on aggregates comprising a polysaccharides matrix, along with cyanobacteria and other (not necessarily diazotrophs) prokaryotic/eukaryotic microorganisms (Figure 3 and Supplementary Figure S2). Using this direct visualization approach, we demonstrated that polysaccharidebased aggregates collected from the Qishon River (estuary and stream) were colonized by dense communities of active heterotrophic diazotrophs (Figure 3 and Supplementary Figure S5). Additional microscopic analyses taken after 48 h incubation at ambient light conditions clearly show that cyanobacteria colonized most of the aggregates area but only few were also diazotrophs (Supplementary Figure S3). Additionally, incubation for 48 h under dark+DCMU conditions of the same water indicated that only few unicellular cyanobacteria have synthesized the nitrogenase enzyme (i.e., were "active"). We cannot rule out that some of the colonizing phototrophic (cyanobacteria) diazotrophs were mixotrophs, namely bacteria that can "switch" between heterotrophic metabolism to carbon fixation via photosynthesis, rather than obligatory phototrophs. Recent studies demonstrated that the cyanobacterium Trichodesmium, previously characterized as phototrophs, may, in fact, use mixotrophic metabolism (Benavides et al., 2017(Benavides et al., , 2020. Similarly, the unicellular cyanobacterial diazotroph Cyanothece have been shown to take up carbohydrates and amino acids (Feng et al., 2010). Thus, it is possible that under dark+DCMU conditions, mixotrophic diazotrophs could also be captured, hence the phycoerythrin signal on our aggregates.
Our immunolocalization images from the Qishon River (Figure 3 and Supplementary Figure S5) suggest that these microenvironments are active hubs for heterotrophic diazotrophs. These images therefore support previous   Table S1). Whiskers indicate the interquartile range (25th to 75th percentile) of the dataset. The mean values are shown as a solid line. Asterisks indicate the statistical significant differences between the Qishon stream and the estuary stations (t-test) where *P < 0.05; **P < 0.01. The averaged value in each sampling campaign and location is shown in Supplementary Table S2, and the actual measured values used to generate the box-plots are shown in gray.
Colonization of aggregates by diazotrophs, especially in N-rich ecosystems, may provide microenvironments favoring heterotrophic N 2 fixation due to the following reasons: (i) The polysaccharide matrix of the aggregate can be degraded by the inhabiting microbes, thus providing additional energy sources to heterotrophic diazotrophs in the form of labile carbon (Bar-Zeev and , (ii) Aggregates comprising a polysaccharides matrix such as TEP would often exhibit high C:N ratio (∼20:1) compared to the Redfield ratio (∼6.6:1) in the surrounding water (Passow, 2002;Engel et al., 2004), thereby "inducing" N-limiting conditions that favor diazotrophs (Gruber, 2008), Frontiers in Microbiology | www.frontiersin.org (iii) The aggregate matrix may provide microenvironments with reduced O 2 concentration as it is respired toward the center of the aggregate (Paerl and Prufert, 1987;Klawonn et al., 2015;Bombar et al., 2016), thus minimizing oxidative damages to the nitrogenase enzyme (Milligan et al., 2007). This may reduce the cellular energetic cost tunneled to segregate the nitrogenase from O 2 in oxygenated environments (Großkopf and LaRoche, 2012), and (iv) Aggregates often comprise high levels of trace metals such as Fe and Mo that are adsorbed to the particles from the environment (Vicente et al., 2009) and are essential to the nitrogenase function (Berman-Frank et al., 2001).

CONCLUSION
The chemical conditions that prevailed at the eutrophic Qishon River should potentially inhibit heterotrophic N 2 fixation, especially given the high concentrations of DIN and the high N:P ratio. Nonetheless, we measured high rates of heterotrophic N 2 fixation, which were several folds higher compared to rates reported from many oligotrophic marine environments, including the neighboring eastern Mediterranean Sea. Generally, the highest heterotrophic N 2 fixation rates were measured in the Qishon stream, where lower C:N and higher N:P ratios were measured compared to the estuary sampling site. These results are in agreement with the few studies that report on unexpectedly high diazotrophic activity in N-rich aquatic environments. We also show evidence of dense colonies of active heterotrophic diazotrophs associated with polysaccharidesbased aggregates in N-rich environments. We suggest that heterotrophic diazotrophs are supported by aggregates and may gain different biochemical advantages compared to planktonic diazotrophs in environments with unfavorable conditions for diazotrophy such as the eutrophic Qishon River. It is likely that the close physical proximity of diazotrophs (as well as other bacteria) to each other on aggregates (including TEP) may enhance the hydrolysis efficiency of carbohydrates and uptake of low-molecular weight organic compounds. In turn, energy gained through this C-rich hydrolysis can be tunneled to nitrogenase synthesis and heterotrophic N 2 fixation. We stress that future studies should use methods that couple visualization of individual cells and heterotrophic N 2 -fixation rates supported by aggregates such as TEP (e.g., immunolocalization and NanoSIMS). These studies should also investigate the diversity, activity, and biomass of particleassociated diazotrophs as a function of the environmental and particle's characteristics. Finally, the contribution of cyanobacterial and heterotrophic diazotrophy by aggregate associated cells to the bulk N 2 fixation should be quantified via dedicated studies in different marine and freshwater environments. These will enable a better understanding of the role of aggregate-associated diazotrophy throughout the marine and freshwater environments, which can later be implemented in N budget studies and models.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
EG, ER, and EB-Z conceived and designed the experiments, performed the samplings, and wrote the manuscript. EG, AB, ER, and EB-Z analyzed the data. ER and EB-Z contributed reagents, materials, and analysis tools. All authors contributed to the article and approved the submitted version.

FUNDING
This study was partly supported by the Israeli Monitoring Program (Ministry of Environmental Protection and Ministry of Energy), the Roy J. Zuckerberg Foundation, and the Israel Science Foundation (grant #1211/17).