**Linking Stoichiometric Organic Carbon–Nitrogen Relationships to planktonic Cyanobacteria and Subsurface Methane Maximum in Deep Freshwater Lakes**

**Santona Khatun 1,\*, Tomoya Iwata 2,\*, Hisaya Kojima 3, Yoshiki Ikarashi 2, Kana Yamanami 2, Daichi Imazawa 1, Tanaka Kenta 4, Ryuichiro Shinohara <sup>5</sup> and Hiromi Saito <sup>6</sup>**


Received: 29 November 2019; Accepted: 31 January 2020; Published: 2 February 2020

**Abstract:** Our understanding of the source of methane (CH4) in freshwater ecosystems is being revised because CH4 production in oxic water columns, a hitherto inconceivable process of methanogenesis, has been discovered for lake ecosystems. The present study surveyed nine Japanese deep freshwater lakes to show the pattern and mechanisms of such aerobic CH4 production and subsurface methane maximum (SMM) formation. The field survey observed the development of SMM around the metalimnion in all the study lakes. Generalized linear model (GLM) analyses showed a strong negative nonlinear relationship between dissolved organic carbon (DOC) and dissolved inorganic nitrogen (DIN), as well as a similar curvilinear relationship between DIN and dissolved CH4, suggesting that the availability of organic carbon controls N accumulation in lake waters thereby influences the CH4 production process. The microbial community analyses revealed that the distribution of picocyanobacteria (i.e., *Synechococcus*), which produce CH4 in oxic conditions, was closely related to the vertical distribution of dissolved CH4 and SMM formation. Moreover, a cross-lake comparison showed that lakes with a more abundant *Synechococcus* population exhibited a greater development of the SMM, suggesting that these microorganisms are the most likely cause of methane production. Thus, we conclude that the stoichiometric balance between DOC and DIN might cause the cascading responses of biogeochemical processes, from N depletion to picocyanobacterial domination, and subsequently influence SMM formation in lake ecosystems.

**Keywords:** dissolved inorganic nitrogen; dissolved organic carbon; phosphonate; subsurface methane maximum; stoichiometry; *Synechococcus*

#### **1. Introduction**

In recent decades, considerable efforts have been made to elucidate the source of methane (CH4) emissions, because the atmospheric concentration of this greenhouse gas has been increasing [1–3].

Among the natural sources of CH4 emissions, freshwater lakes have recently been recognized as an important contributor, accounting for 6–16% of total natural CH4 emissions [4,5]. Traditionally, CH4 production in lakes was believed to occur via anaerobic methanogenesis in oxygen-depleted environments [6,7]. However, recent studies have identified the occurrence of the subsurface methane maximum (SMM) in the oxygenated water columns of many deep freshwater lakes [8–15], as well as in marine ecosystems (known as the methane paradox) [16–18]. As subsurface CH4 in upper water column layers is closer to the atmosphere than sedimentary CH4, the formation of the SMM may have a significant impact on CH4 emissions from freshwater lakes [19].

It has been proposed that SMM might develop as a result of the tributary inflow, and the transport and diffusion of CH4 from hypolimnetic and littoral sediments [4,9,20–22]. Moreover, anaerobic methanogenesis by archaea present in anoxic microenvironments within algal cell aggregates, zooplankton guts or other particles was also proposed as a possible solution of the methane paradox [10,23]. However, recent studies have added empirical evidence that SMM formation is likely via phosphonate metabolism by planktonic bacteria that carry C-P lyase genes [12,13,15,16,24,25] or via unknown photosynthesis-related processes [26]. For example, planktonic autotrophic and heterotrophic bacteria (e.g., *Trichodesmium*, *Pseudomonas*, *SAR11*, *Synechococcus*) can utilize phosphonate compounds when they are P-starved and produce CH4 aerobically as a byproduct of phosphonate decomposition. Other methylated compounds, such as dimethylsulfoniopropionate (DMSP), trimethylamine (TMA), and methionine (Met), can also serve as a possible precursor of aerobic CH4 production by planktonic bacteria and microalgae [17,19,27]. However, most of these studies focused only on specific microbes linked with CH4 production in oxic water columns. Moreover, few studies have clarified the relationship among biogeochemical processes, microbial communities, and aerobic CH4 production in lakes.

Our previous studies have shown that SMM starts to develop in early summer in association with the development of stratification, and that it peaks in midsummer around the metalimnion in a deep oligotrophic lake [15]. Then, the SMM disappears in the winter season. It is hypothesized that the seasonal predominance of cyanobacterial populations, such as *Synechococcus*, during the thermal stratification period may be responsible for the development of SMM through organophosphonate metabolism [15]. However, the hypothesis has never been tested in other lake ecosystems. Although a few studies have examined the role of the microbial community in lake SMM formation [12–15], knowledge of the relationship between the vertical profile of dissolved CH4 concentrations and microbial communities is also lacking. If planktonic cyanobacteria are responsible for SMM formation, then biogeochemical processes other than CH4 production in metalimnetic waters may also be affected through their metabolic activities, such as primary productivity, nutrient uptake, and the extracellular release of DOC [28,29]. Furthermore, the biogeochemical processes that control SMM development have remained unidentified, even though the degree of the SMM development varies greatly among lakes [15].

The objective of the present study was to clarify the pattern and mechanisms of SMM formation in the aerobic environments of lake ecosystems. We surveyed nine deep freshwater lakes in Japan to determine the pattern of CH4 supersaturation in the metalimnion and to identify the environmental variables that affect such SMM formation. Moreover, we performed microbial community analyses of bacterioplankton and algae to identify the microbes responsible for aerobic CH4 production across the lakes. Finally, we analyzed lake physicochemical environments linked with both the microbial community and SMM formation to expand our knowledge of aerobic CH4 production and emission from oxic layers in deep freshwater lakes.

#### **2. Materials and Methods**

#### *2.1. Study Area and Field Survey*

The field survey was conducted in nine freshwater lakes of Japan (Figure S1, Tables S1 and S2) during the summer stratification period (from July to September) in 2016–2017. The study lakes

are located in temperate and cool-temperate regions (35◦11' N–42◦36' N latitudes) and are mainly classified as caldera lakes, dammed lakes, or tectonic lakes (Table S1). The study lakes are deep stratified lakes, with maximum depths (31–363 m) deeper than the depth of 1% of the irradiance at the surface (i.e., a surrogate for the compensation depth) during the study period. The lake trophy was classified as mesotrophic, oligotrophic, or ultraoligotrophic conditions according to chlorophyll *a* and TP concentrations (Table S2, [30]). Most of the lakes are monomictic or dimictic and rarely freeze over; except Lake Hibara, where the surface water generally freezes during the December–March period. The watersheds of the study lakes are mostly covered with deciduous and coniferous forests (range = 67–99%). Agricultural lands such as rice paddies and farmlands occupy 20% and 11% of the watershed areas of Lake Toya and Lake Nojiri, respectively. Grassland vegetation accounts for 11% of the watershed area of Lake Hibara, while in Lake Ashinoko, residential areas (9%) contribute to the land cover to some extent.

At the deepest point of each study lake, we established a sampling site and measured the vertical profiles of environmental variables such as water temperature (◦C), dissolved oxygen (DO) concentration (DO, mgO2/L) and saturation (%), pH, and photosynthetically active radiation (PAR, μmol m−<sup>2</sup> s−1) by using a multi-parameter sonde (YSI ProDSS, YSI-Nanotech, Tokyo, Japan), a DO meter (HQ40d, Hach, Loveland, CO, USA), a pH meter (Orion3-Star, Thermo Scientific, Chelmsford, MA, USA), and a submersible spherical quantum sensor (LI-193, LI-COR, Inc., Lincoln, NE, USA), respectively. For each sampling site, lake water samples were collected at 10 depths, from the surface to hypolimnetic waters, by using a 6 L Van Dorn sampler. For determination of the dissolved CH4 concentration, the collected lake water was siphoned into two 30 mL glass vials and sealed with a butyl rubber stopper and an aluminum crimp. Then, 0.2 mL of saturated HgCl2 was added to each vial as a preservative and stored in a dark place at room temperature until analysis. For water quality analyses, the collected lake water was filtered with a 100 μm nylon mesh to remove large particles (e.g., zooplankton), and the filtrates were stored in polypropylene bottles for total nitrogen (TN) and total phosphorus (TP) analyses. The remaining filtrate water was further filtered through pre-combusted GF/F glass fiber filters (Whatman plc., Maidstone, Kent, UK) for dissolved nutrient (i.e., SRP, NH4, NO2, and NO3) and dissolved organic carbon (DOC) analyses; the former samples were stored in polypropylene bottles, while the latter samples were preserved in pre-combusted amber glass bottles. All of the water samples for nutrient and DOC analyses were kept at −20 ◦C until analyzed. For analyzing dissolved manganese (Mn) concentrations, the 100 μm mesh-filtered lake waters were further filtered with cellulose acetate disposable filters (pore size 0.45 μm), and nitric acid (final concentration = 0.1 M HNO3) was added to prevent the precipitation and adsorption of metals. Then, the acidified water samples were stored in a refrigerator.

For determination of the vertical profile of phytoplankton biomass and planktonic microbe density, we also collected lake waters from the 10 sampling depths and performed chlorophyll *a* (Chl *a*) measurements and CARD-FISH (catalyzed reporter deposition fluorescence in situ hybridization) analyses. We collected phytoplankton pigment samples by filtering 50−500 mL of lake water with GF/F glass fiber filters, and then stored the samples at −20 ◦C until analysis. For the CARD-FISH analyses, an aliquot of lake water samples was fixed in a 0.2 μm-filtered paraformaldehyde solution (final concentration = 1%, v/v) and stored at 4 ◦C for <24 h. To analyze the community structure of planktonic eukaryotic algae and prokaryotic bacteria, we collected additional lake water samples from each epilimnion (0 m depth), metalimnion (7−15 m depth), and hypolimnion (26−60 m depth). The samples for the enumeration of eukaryotic algae were fixed with a Lugol solution (final concentrations, 0.2%−0.4%) and stored in the dark at room temperature, while the samples for bacterial community structure were collected on 0.22 μm filters (Sterivex filter cartridges, Millipore, Billerica, MA, USA) and stored at −20 ◦C until analysis.

To identify the presence of phosphonic compounds in microbial cells, a two-dimensional NMR ( 1H–31P heteronuclear multiple bond correlation, 1H–31P HMBC) analysis was performed for the suspended particles in lake water samples according to the previous study [31]. A large amount of lake water (68−201 L) was collected from the metalimnion of each lake with the Van Dorn sampler and filtered with a 100 μm nylon mesh to remove large particles. Then, the suspended particulate matter was collected onboard by filtering the collected lake water through a 2 μm cartridge filter (MCP-HX-C10S, Advantec, Inc., Tokyo, Japan) by using a peristaltic pump (Masterflex, L/S VFP002, Cole-Parmer, Vernon Hills, IL, USA). The filter samples were stored at −20 ◦C until analysis.

We also surveyed all of the major tributary streams identifiable on topographical maps for each study lake (Table S3), except for two inaccessible tributaries (Yanagisawa Stream of Lake Chuzenji and Okotanpe Stream of Lake Shikotsu). Near the downstream end of each tributary inflow, we collected river water samples to analyze the same environmental variables (water temperature, DO, pH, dissolved CH4, nutrients, DOC, and Mn) as those in the lake survey. In addition, the discharge (m3/s) of tributary streams was measured by the midsection method using a portable current meter (CR-7WP; Cosumo Riken, Inc., Kashiwara, Japan).

#### *2.2. Chemical Analyses*

Dissolved CH4 concentrations of the collected lake waters were quantified by the headspace equilibration method using a gas chromatograph with an FID detector (GC-FID, GC-8A, Shimadzu Corp., Kyoto, Japan). The headspace phase created by 3 mL of high-purity He gas (>99.9999%) in the vial was equilibrated with the aqueous phase. The *p*CH4 in the equilibrated headspace was then analyzed by GC-FID. The dissolved CH4 concentration (nmol/L) in the water samples were determined from the amounts of gasses in both the equilibrated headspace and the aqueous phase; the latter was calculated by using *p*CH4 in the headspace and Henry's law constant.

The DOC concentration of lake water samples was measured using a total organic carbon analyzer (TOC-L CPN, Shimadzu Corp., Kyoto, Japan). The ammonium concentration (NH4, μM) was quantified spectrophotometrically using the indophenol method [32]. Nitrate (NO3) and nitrite (NO2) concentrations were determined through the second-derivative UV spectrophotometric method [33] and Bendschneider and Robinson's method [34], respectively. Dissolved inorganic nitrogen (DIN) concentrations were determined as the sum of NH4, NO2, and NO3. The total nitrogen (TN) concentration of the lake water samples was determined by the alkaline persulfate digestion method, followed by the absorbance measurement (at 220 nm) for the resultant nitrate by UV spectrophotometry. Soluble reactive phosphorus (SRP) concentrations were measured spectrophotometrically using the molybdenum blue method. Likewise, the dissolved total phosphorus (DTP) and total phosphorus (TP) concentrations were quantified using the molybdenum blue method after persulfate digestion for the GF/F-filtered lake waters and unfiltered lake waters, respectively. For the phosphomolybdic acid measurements for SRP, DTP, and TP analyses, we increased the sulfuric acid concentration by a factor of two to minimize the effect of silicate interference [35]. The dissolved organic phosphorus (DOP) concentration was determined as the difference between DTP and SRP concentrations. Dissolved Mn concentrations were analyzed using inductively coupled plasma optical emission spectroscopy (SPS 3520 UV-DD, Hitachi High-Tech Science Corp., Tokyo, Japan).

#### *2.3. Analyses for Phytoplankton Biomass and Community Structure*

The Chl *a* sample pigments collected on GF/F glass fiber filters were extracted with N, N-dimethylformamide (6-mL) in the dark for 24 h. The Chl *a* content in lake water (μg-chl *a*/L) was then measured by the Welschmeyer method using a fluorometer (Trilogy, Turner Designs, San Jose, CA, USA). To identify the community structure of eukaryotic algae, we identified phytoplankton species to the lowest taxonomic level and enumerated the cell numbers of each algal species for the epilimnion, metalimnion and hypolimnion of each study lake by microscopic enumeration. We also enumerated red autofluorescent algal cells on the CARD-FISH filter (see below) under blue excitation (470 nm, Filter 09, Carl Zeiss Microscopy GmbH, Jena, Germany) to estimate algal cell density (cells/mL).

#### *2.4. Analyses for Planktonic Bacterial Density and Community Structure*

To analyze the distribution of specific microbes related to SMM formation, the CARD-FISH analysis was performed for the water samples collected from each sampling depth. The aliquot of lake water samples fixed with paraformaldehyde solution was filtered through a white polycarbonate membrane filter (type GTTP, pore size 0.2 μm, Millipore, Billerica, MA, USA) and a brown polycarbonate membrane filter (type GTBP, pore size 0.2 μm, Millipore, Billerica, MA, USA) for the enumeration of CARD-FISH- and DAPI-stained bacterial cells, respectively. We applied the oligonucleotide probes (labeled with horseradish peroxidese) specific for eubacteria (EUB338), type I MOB (Mg84 + Mg705), *Synechococcus* (405\_Syn) and archaea (ARCH915) for the filters at 46 ◦C for > 8h for hybridization (Table S4) because these microbial groups have been reported to influence the dissolved methane profile in lakes [15,36–38]. FITC-labeled tyramide solution and DAPI solution were then applied to the filter sections for the enumeration of cells on a fluorescence microscope (Zeiss Axio Lab.A1, Carl Zeiss Microscopy GmbH, Jena, Germany) under blue excitation (470 nm, Filter 10) and UV excitation (365 nm, Filter 01), respectively. We also enumerated the autofluorescent cyanobacterial cells under green excitation (545 nm, Filter 43) to estimate the density of cyanobacteria. At least 1000 cells were counted to estimate the cell density of the target bacteria (cells/mL).

To analyze the community structure of planktonic bacteria, 16S rRNA gene amplicon sequencing was performed for the filter samples collected at the epilimnion, metalimnion and hypolimnion of each lake. From the Sterivex filter samples, DNA was extracted using the PowerWater Sterivex DNA Isolation Kit (Qiagen, Hilden, Germany). From the DNA samples, sequencing libraries were prepared according to the Illumina 16S Metagenomic Sequencing Library Preparation protocol and then subjected to Illumina MiSeq sequencing. The resulting reads were processed with QIIME, an open-source bioinformatics platform for microbial community analysis [39], as follows. Quality-filtered sequences were clustered using the UCLUST algorithm with an identity threshold of 97%, to generate operational taxonomic units (OTUs). The representative sequences of the OTUs were subjected to chimera-checking to eliminate chimeric OTUs. The representatives were then used for taxonomic assignment of the OTUs via the Greengenes database, version 13\_8. Greengenes annonation includes eukaryotic organelles (i.e., chloroplasts and mitochondria) as a taxonomic group in bacterial phyla. As we focused on bacterioplankton communities in the 16S rRNA gene analyses, we removed the chloroplast and mitochondria reads from the amplicon library for data analyses. The nucleotide sequences of the 16S rRNA gene obtained in the present study are available in the NCBI/SRA with the accession number of PRJNA599317.

#### *2.5. 1H–31P NMR Analyses for Suspended Particles*

For NMR analysis, we extracted the organic phosphorus fraction from the suspended particles collected on the 2 μm cartridge filter [31]. We separated the pleated filter paper from the polypropylene filter cartridge by using a cutter and tweezers. The separated filter papers were then placed into a Teflon vial with 50 mL Milli-Q water, and the vials were subsequently immersed for approximately 1 h in a water bath set at 60 ◦C for the extraction [40]. The extracts were then filtered through a GF/F glass fiber filter followed by membrane filter filtration (PVDF: nominal pore size 0.45 μm, Millex, Millipore, Billerica, MA, USA). The filtrates were frozen (−20 ◦C) and lyophilized to concentrate the extracts for NMR analysis. Just before the NMR analyses, the lyophilized extracts were re-dissolved in 0.9 mL of EDTA solution (20 mmol L<sup>−</sup>1) with D2O (0.1 mL, Fujifilm Wako Pure Chemical Corp., Osaka, Japan) and shaken at room temperature for 10 minutes. The solution was then filtered through a membrane filter (nominal pore size: 0.45 μm, Millex, Millipore, Billerica, MA, USA) and transferred into a 5 mm NMR tube. The 1H*–*31P NMR spectrum of the extracts was recorded at 500.2MHz (for 1H) and 202.5 MHz (for 31P) via a JNM-ECA 500 spectrometer (JEOL Ltd., Tokyo, Japan), equipped with a 5 mm auto-tune probe. These spectra were externally referenced to 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS) and D3PO4 (δ = 0 ppm), respectively. The datasets of 1H–31P HMBC spectra were acquired with 2048

and 256 data points and sweep widths of 9384 and 20,259 Hz for the *x* and *y* dimensions, respectively. The obtained spectra were processed with a 2 Hz line broadening using Delta, version 5.

#### *2.6. Data Analyses*

A generalized linear model (GLM) was developed to identify environmental variables that affect the dissolved CH4 concentrations in the study lakes. We included physical (depth, water temperature, and percentage of surface irradiance), chemical (DIN, SRP, DOC, DOP, and TN/TP), and biological (Chl *a*) variables as the explanatory variables. Two variables (i.e., SRP and Chl *a*) were transformed as log10 (*x* + 1) because the data were not normally distributed by visual examination. A GLM analysis with a gamma error distribution and an inverse link function was used to analyze the factors that influenced the response variable (dissolved CH4). We built candidate models containing all possible combinations of the nine explanatory variables (29 <sup>−</sup> 1 = 511 models) and selected the best model based on the AIC value. The relative importance of an explanatory variable was evaluated by the sum of Akaike weights (*w*) over all models, including that variable [41]. To identify the indirect causal relationships of the environmental variables with SMM formation, we also constructed GLMs for the environmental variables that most strongly affected the dissolved CH4 concentrations. When we constructed the GLM for DOP values, we transformed the response variable as DOP + 1 because the gamma regression analysis did not allow for the inclusion of variables with negative or zero values. The GLM analyses and model-selection procedures were performed using the MuMIn package, version 1.40.0 [42] and the MASS package, version 7.3-48 [43] of the statistical R software 3.3.3 (R Development Core Team 2017).

For the community analyses of planktonic bacteria, 16S rRNA gene sequence read counts were used to calculate the relative abundance of each OTU for eplimnetic, metalimnetic, and hypolimnetic bacterial communities, by assuming that the amplicon reads could be used as a surrogate for the abundance of each bacterial population (total number of amplicon reads = 8848–29,331 reads per sample). In this calculation, we removed the taxonomically unassigned OTUs from the analyses, as the percentage of their amplicon reads accounted only for <1.6% of the total reads. We also confirmed a significant positive correlation between the total number of bacterial read counts per unit of lake volume and the cell density of eubacteria obtained from the CARD-FISH analyses (EUB338, *r* = 0.50, *p* = 0.015). Finally, the taxonomic composition data were combined into the phylum level for the analyses. Non-metric multidimensional scaling (NMDS) analysis was performed to identify the relationships between bacterioplankton community composition and the environmental variables of the epilimnion, metalimnion, and hypolimnion of each study lake. To obtain an NMDS ordination with a low stress value at <0.2 (an indication of well-represented data by a two-dimensional representation [44]), the phyla with a relative abundance of <0.4% of the total read counts when all of the samples were combined were grouped into a category as "Others". For the environmental variables, the average values of 10 variables (i.e., depth, temperature, percentage of surface irradiance, CH4, Chl *a*, DIN, SRP, DOC, DOP, TN/TP) were determined for each of epilimnion, metalimnion and hypolimnion, and were then incorporated in the analyses. The NMDS analysis (based on the Bray–Curtis dissimilarity) was performed by using the vegan 2.4-5 package [45] of the R software 3.3.3.

Finally, we examined the relationship between the degree of SMM development and lake environmental variables to identify the factors controlling the oxic CH4 peak. According to our previous study [15], the SMM is herein defined as the layer of the local CH4 maximum that forms below the surface water. The degree of SMM development for each study lake was quantified by peak SMM, namely the maximum CH4 concentration in the SMM minus the atmospheric equilibrium CH4 concentration at that depth. The peak SMM for the study lakes was regressed against the environmental variables responsible for SMM formation.

#### **3. Results**

#### *3.1. SMM Formation in Lakes*

The vertical profiles of dissolved CH4 concentrations revealed that the SMM developed in aerobic layers in all of the nine study lakes (Figure 1). Further, peak SMM concentrations (range = 62.5–592.6 nM) showed the highest values in the vertical CH4 profiles for all of the study lakes. The SMM peak tended to occur within the metalimnion, or in depths adjacent to the metalimnion where DO saturation levels were high (range = 85–131%). In addition, the peak depth and vertical profile of dissolved CH4 (i.e., subsurface peak of CH4) were similar to those of DO saturation, except in Lake Nojiri.

**Figure 1.** Vertical profiles of dissolved CH4 concentrations in the nine study lakes of Japan during the July to September in 2016–2017 period. AS: Lake Ashinoko, MO: Lake Motosu, SA: Lake Saiko, AO: Lake Aoki, NO: Lake Nojiri, CH: Lake Chuzenji, HI: Lake Hibara, TO: Lake Toya, SH: Lake Shikotsu. Red and gray circles refer to the dissolved methane (nM) and oxygen (%) concentrations, respectively. Dotted lines denote the atmospheric equilibrium concentration of dissolved CH4 in lake water (3.7–7.2 nM). Shaded areas denote the range of the thermocline.

Regarding the other environmental variables, water temperature and percentage of surface irradiance were highest at the lake surface and decreased with depth; this pattern did not correspond with the vertical profile of the dissolved CH4 concentration for any of the study lakes (Figures S2 and S3). Although the overall pattern of Chl *a* and DOP profiles did not match the CH4 profile, their peaks were located at depths similar to the CH4 peaks (Figures S4 and S5). In most of the study lakes, DIN concentrations tended to be higher in the hypolimniion and lower in both the epilimnion and metalimnion, the latter of which exhibited high CH4 concentrations (Figure S6). Although SRP did not show a clear overall trend, the peak depth and vertical profile of SRP were similar to those of CH4 in Lake Motosu, Lake Saiko, Lake Toya and Lake Shikotsu (Figure S7). Similarly, the vertical patterns of DOC were variable, but its concentrations tended to be higher in the metalimnion or nearby layers (except in Lake Nojiri), as observed in the CH4 profile (Figure S8). The observed values of lake TN/TP ratio (median = 151, range = 32–875) were high relative to the Redfield ratio (N:P = 16:1) for all of the study lakes (Figure S9). The vertical TN/TP profile tended to show a pattern opposite to that of the CH4 profile—except in Lake Shikotsu, where a close association was observed for both variables.

The cross-lake comparison showed that the development of SMM varied among the study lakes (Figure 1). The highest value of peak SMM was observed in Lake Saiko (an oligotrophic lake), while small CH4 peaks were observed in three ultraoligotrophic lakes (Lake Motosu, Lake Toya, and Lake Shikotsu).

#### *3.2. Relationships between Dissolved CH4 and Physicochemical Variables*

GLM analyses were performed to identify the physicochemical variables affecting the water-column CH4 concentrations in lakes. In the analyses, six environmental variables (i.e., depth, water temperature, percent of surface irradiance, DIN, DOP, and TN/TP ratio) were retained in the best model selected by AIC (Figure 2). The sum of Akaike weights (*w*) further identified that depth, DIN, DOP, and TN/TP ratio, were the most important variables explaining the variability of dissolved CH4 concentrations in the lakes (Table S5).

**Figure 2.** Relationships between environmental variables and dissolved CH4 concentrations in the study lakes. Univariate regression lines are shown for the explanatory variables that were retained in the best generalized linear models (GLMs) with a gamma distribution of the response variable. -: Lake Ashinoko, -: Lake Motosu, : Lake Saiko, : Lake Aoki, : Lake Nojiri, : Lake Chuzenji, : Lake Hibara, -: Lake Toya, : Lake Shikotsu.

We observed a strong negative nonlinear relationship between DIN and dissolved CH4 concentrations when we combined the data from all nine study lakes (Figure 2 and Table S6). The CH4 concentration increased rapidly with a decrease in the DIN concentration. Similar curvilinear patterns were observed for depth and the TN/TP ratio; the lake waters with shallower depths and a lower TN/TP ratio increased the water-column CH4 concentrations. On the other hand, a positive effect of DOP on dissolved CH4 concentrations was observed for the study lakes.

To explore the potential factors controlling the chemical variables that strongly affect dissolved CH4, we further constructed GLMs for DIN, DOP, and TN/TP (Table S5). The results showed that DOC concentration was the best explanatory variable for all three influential variables (Figure 3 and Table S6). DOC exhibited negative nonlinear relationships with both DIN and the TN/TP ratio. In contrast, DOC showed a positive relationship with DOP concentrations.

**Figure 3.** Relationships between dissolved organic carbon (DOC) and chemical variables in the study lakes. Univariate regression lines estimated by the best GLMs with a gamma distribution of the response variable (*y*) are shown. See the legend for Figure 2 for a description of the symbols.

#### *3.3. Community Analyses of Planktonic Algae and Bacteria*

Microscopic enumeration for eukaryotic algae identified six phytoplankton classes present in the study lakes (Table S7). Among these phytoplanktonic algal groups, Chlorophyceae or Bacillariophyceae predominated in the epilimnion of most of the study lakes except for Lake Toya, where Dinophyceae (53.6%) was the dominant group. Similarly, Chlorophyceae and/or Bacillariophyceae tended to predominate in the algal communities in the metalimnion, although Cryptophyceae and Dinophyceae were relatively abundant in the metalimnion of Lake Ashinoko. In the hypolimnion, Bacillariophyceae accounted for 54–97% of the relative abundance of algal communities, although Chlorophyceae was relatively abundant in the hypolimnion community of Lake Chuzenji (45.8%). There was no phytoplanktonic algal group that consistently exhibited a metalimnetic peak in terms of relative abundance for any of the study lakes.

The 16S rRNA gene amplicon sequencing analyses detected 6510 OTUs of bacteria (except unassigned OTUs), including 47 phyla, from the epilimnetic, metalimnetic, and hypolimnetic layers of all the study lakes. Actinobacteria, Verrucomicrobia, Proteobacteria, Cyanobacteria, Bacteroidetes and Armatimonadetes accounted for, respectively, 1.8–47.2%, 2.7–45.0%, 10.4–41.1%, 0.02–29.1%, 5.7–26.1%, and 0.1–5.8% of the total bacterial community in the epilimnion, metalimnion, and hypolimnion (Figure 4).

The NMDS analyses showed that the community composition of epilimnetic and metalimnetic microbes differed from that of the hypolimnetic community (Figure 4). The results revealed that the relative abundance of Proteobacteria, Verrucomicrobia, Actinobacteria, Bacteroidetes, and Cyanobacteria tended to be high in the epilimnion and metalimnion, whereas Planctomycetes, Acidobacteria, Chloroflexi, and Chlorobi were relatively abundant in the hypolimnion. We further observed the association of community structure and environmental variables. In epilimnetic and metalimnetic layers with high CH4 and DOP concentrations, Cyanobacteria, and Verrucomicrobia tended to be relatively abundant.

The results of epifluorescent microscopic analyses (CARD-FISH, DAPI, and autofluorescent enumeration) showed that both Cyanobacteria and *Synechococcus* cell density showed significant positive relationships with water-column dissolved CH4 concentrations (Figure 5). *Synechococcus* accounted for 21–96%, 64–96%, and 22–75% of the total cyanobacterial density in the epilimnion, metalimnion, and hypolimnion of the study lakes, respectively. In contrast to the patterns of Cyanobacteria and *Synechococcus*, phytoplankton biomass (Chl *a*), as well as the cell density of eubacteria (EUB338), type I MOB (Mg84 + Mg705), archaea (ARCH915), and algae, showed no significant relationships with dissolved CH4 concentrations.

**Figure 4.** Taxonomic composition of bacterioplankton in the study lakes revealed by 16S rRNA amplicon sequencing analyses (**A**) and the result of non-metric multidimensional scaling (NMDS\_ analysis for the relationship between lake physicochemical variables and the community composition of planktonic bacteria (**B**). See the legend for Figure 2 for the abbreviations of lake names appearing in panel (A). Red, blue and orange symbols refer to the bacterioplankton samples obtained from the epilinmion, metalimnion, and hypolinmion, respectively.

**Figure 5.** Relationships between the dissolved CH4 concentration and the cell density of microbial communities in study lakes. EUB338, bacteria; Mg84 + Mg705, type-I methane-oxidizing bacteria; 405\_Syn, *Synechococcus*; ARCH915, Archaea (see Table S4). GLM with a Gaussian error distribution and an identity link function was used for the variables, while regression lines shown when significant (*p* < 0.05). See the legend for Figure 2 for a description of the symbols.

We then analyzed the vertical profile of *Synechococcus* density in relation to the CH4 profiles, which showed that both profiles were closely associated with each other (Figure 6). In addition, a negative curvilinear pattern was found for the relationship of *Synechococcus* cell density with both depth and DIN concentration, as was observed in the patterns of dissolved CH4 concentrations (Figure S10). In contrast, a positive nonlinear relationship between *Synechococcus* cell density and DOP concentrations was found.

**Figure 6.** Vertical distribution of *Synechococcus* density (red circles) in relation to the dissolved CH4 profile (blue circles) in the study lakes. Shaded areas denote the range of the thermocline.

#### *3.4. Relationship between Peak SMM and Environmental Variables*

A correlation analysis between the peak SMM and environmental variables revealed that the peak SMM was positively and linearly correlated with lake DOC, DOP, and *Synechococcus* cell density observed at the same depth (Figure 7 and Table S6). Lakes with higher concentrations of both DOC and DOP and higher *Synechococcus* cell density tended to develop a larger SMM peak. Moreover, a negative nonlinear relationship of peak SMM was observed for both DIN concentration and the TN/TP ratio, implying that lakes with lower DIN and TN/TP values experienced a rapid increase in peak SMM.

**Figure 7.** Relationships between lake environmental variables and the peak subsurface methane maximum (SMM) for the study lakes. For the relationships of both dissolved inorganic nitrogen (DIN) and TN/TP, the regression lines were estimated using generalized linear models (GLMs) with a gamma error distribution and an inverse link function for each variable, while the GLM with a Gaussian error distribution and an identity link function was used for the other variables. For explanatory variables, the values at the peak SMM depths were used for the analyses.

#### *3.5. Identification of Phosphonates in Suspended Particles*

Particulate suspended matter collected from the metalimnion was characterized by a two-dimensional 1H*–*31P NMR analysis. The spectral region of 20–31 ppm in 31P NMR analysis is known to represent the chemical shifts of phosphonate compounds [25,46]. Our mass spectra at 1.2 ppm (1H) and 25 ppm (31P) revealed the presence of phosphonate in the suspended particles of the metalimnion in Lake Toya (Figure 8).

**Figure 8.** Two-dimensional (1H-31P) NMR microscopy for suspended particles collected from the metalimnion showed the presence of phosphonate compounds (black spot inside the red circles) in the metalimnion of Lake Toya.

#### **4. Discussion**

In the present study, we observed the formation of the CH4 maximum in oxic subsurface layers (i.e., SMM) within or near the metalimnion in all of the study lakes (Figure 1). The seasonal development of metalimnetic CH4 peaks during the stratification period was also reported by previous studies in deep freshwater lakes [8–10,15] as well as in pelagic marine ecosystems [16,17]. These results suggest that the development of the CH4 peak in aerobic subsurface waters may be a common phenomenon in aquatic ecosystems.

SMM formation was unlikely to couple with the dissolution of atmospheric CH4 because the observed CH4 peak concentrations (67–597 nM, Figure 1) were one or two orders of magnitude higher than the dissolved CH4 equilibrated with the atmosphere (range = 3.4–7.4 nM). Tributary inflow is known to contribute to SMM development in lakes [9]. However, we observed apparent SMM

development even in lakes with no tributary streams (i.e., Lake Motosu and Lake Aoki). For the other lakes, moreover, the discharge-weighted average of the dissolved CH4 concentrations of tributary streams (range = 8.8–88 nM) was generally lower than the peak CH4 concentrations—except in Lake Ashinoko, where there is a small tributary (discharge < 0.02 m3/s) with a high CH4 concentration (Table S3). Another potential source of SMM formation might be the transportation of CH4 from anoxic littoral and profundal sediments [20,21]. However, the vertical profile of dissolved Mn concentrations, a tracer of water mass transported from anoxic environments [9,15], showed a pattern unrelated to the CH4 profiles (Figure S11). Moreover, the vertical gradient of CH4 profiles suggests no substantial contribution of hypolimnetic CH4 to the SMM of any of the study lakes (Figure 1). Therefore, we argue that the diffusion of anoxic CH4 from anoxic littoral and profundal sediments was also unlikely to be the source of the CH4 supersaturation observed in our study.

Recent studies have suggested that the microbial degradation of phosphonates (such as MPn and 2-AEP) can explain aerobic methane production in oligotrophic waters [12,13,15,16,24,25]. The marine and freshwater microbial community (e.g., *Trichodesmium*, *Pseudomonas*, *Synechococcus*, SAR11) can express the C-P lyase operon (*phn* genes) to utilize phosphonic compounds under phosphate-starved conditions [15,16,24,25]. As a result of the cleavage of C-P bonds of phosphonates, these microbes can produce CH4. In fact, the batch-culture experiments of lake water amended with MPn confirmed in situ aerobic CH4 production by planktonic bacteria in one of the study lakes (Lake Saiko) [15]. Considering such circumstantial evidences, as well as the fact that the DO saturation profile resembled the dissolved CH4 pattern (Figure 1), we argue that photosynthesis-related biogeochemical processes mediated by planktonic microbes may be relevant to the development of SMM.

The relationships between dissolved CH4 and limnological variables suggest that DOC controls the availability of DIN, DOP, and the TN/TP ratio, thereby influencing the water-column CH4 concentration in the study lakes (Figures 2 and 3). In particular, the remarkable negative nonlinear relationships of DOC with DIN and the subsequent nonlinear effects of DIN on dissolved CH4 concentrations were found. The nonlinear relationship between DOC and nitrate (NO3) has recently been reported from a wide variety of aquatic environments along a hydrologic continuum from soils to streams and lakes to coastal and pelagic ocean ecosystems [47–50]. It has been hypothesized that NO3 accumulation may occur in aquatic environments with low DOC concentrations due to an organic C limitation for assimilation and/or denitrification by heterotrophic microbes [50]. In contrast, NO3 depletion may occur in aquatic environments with high DOC concentrations due to the sufficient supply of organic C for heterotrophs. These patterns imply that stoichiometric controls of DOC–NO3 relationships over microbial activity may regulate the fate of N in aquatic ecosystems. In fact, NO3 is the dominant species of DIN in the study lakes (median = 85%, range = 0–98%). Moreover, even when we separately analyzed individual lakes, a significant negative correlation of DOC and DIN was found for each of them (*r* = −0.69–−0.90, *p* < 0.05), except Lake Shikotsu (*r* = −0.44, *p* > 0.05). Furthermore, a strong nonlinear effect of DOC on the TN/TP ratio was also evident (Figure 2), suggesting that the availability of organic carbon controls N accumulation in the water column, thereby influencing the elemental stoichiometric balance of nitrogen and phosphorus in lakes.

Previous studies explained that aerobic CH4 production from phosphonate decomposition by planktonic microbes occurs under P-starved environments because the organisms utilize phosphonates as an alternative P source [12,13,15,16,24]. However, the present study showed that higher CH4 concentrations tended to occur in lower DIN concentrations; no clear pattern was found for the relationship between SRP and CH4 (Figure 2). These patterns seemed to be intuitively paradoxical, as a lower N availability was likely to stimulate aerobic CH4 production. However, the observed TN/TP ratio in the study lakes (median = 151, range = 32–875, Table S2) was much higher than the Redfield ratio (N/P = 16), suggesting that lake productivity may be more P-limited than N-limited in these lakes. Therefore, we predict that P-starved planktonic microbes utilize phosphonate compounds to form SMM in the oxic waters of the study lakes. The positive relationship between DOP, which may contain dissolved organic phosphonate compounds, and CH4 concentrations, also supports this argument

(Figure 2). Moreover, the NMR analysis revealed the presence of particulate phosphonates in the metalimnion of Lake Toya, where SMM formation was observed (Figure 8). A previous study also detected 2-aminoethylphosphonate (2-AEP), a possible precursor of aerobic methane production [51], from the suspended particles in the epilimnion of Lake Saiko [31]. Therefore, we hypothesized that planktonic bacteria in deep unproductive lakes may store P as phosphonate molecules and utilize such intracellular P compounds via the enzymatic liberation of C-P bonds under P-limited environments [15]. Although we were unable to detect phosphonates in the suspended fractions from the other study lakes, this is probably due to the insensitivity of NMR; prolonged data acquisition and/or a larger volume of lake water filtration may be necessary to obtain the spectra. In addition, further development of quantitative analyses for the dissolved species of phosphonate compounds [52] is necessary to understand the phosphonate and CH4 dynamics in lakes.

The 16S rRNA gene amplicon sequencing analyses and the CARD-FISH analyses revealed that cyanobacteria, especially *Synechococcus*, were related to dissolved CH4 concentrations and SMM formation. The composition of cyanobacteria estimated from amplicon reads was associated with the gradient of the CH4 concentration (Figure 4). Moreover, a close association of the vertical CH4 profile with *Synechococcus* distribution was observed across the lakes (Figure 6), as found in a previous study [15]. The batch-culture experiments conducted in previous studies also showed that *Synechococcus* strains have the ability to produce CH4 in oxic conditions as a result of phosphonate (such as MPn and 2-AEP) utilization [15,51] or unknown photosynthesis-related processes [26]. Although other heterotrophic bacteria, such as *Limnohabotans* and *Pseudomonas*, are also known to carry the C-P lyase (*phnJ*) gene [13,15], the strong correlation between the CH4 concentration and *Synechococcus* cell density (Figure 5), as well as the association of the DO saturation profile with the CH4 peak (Figure 1), suggested that *Synechococcus* (photosynthetic cyanobacteria) may be one of the potential drivers of SMM formation in deep freshwater lakes.

*Synechococcus*, a group of ubiquitous freshwater picocyanobacteria, often predominate in the planktonic communities in the epilimnion and metalimnion of unproductive lakes during the summer stratification period, where nutrient depletion occurs due to the prevention of vertical water mixing and nutrient exhaustion by competitors, such as eukaryotic algae [53–55]. Our previous studies also indicated that the development of nutrient-depleted environments by stratification during midsummer (i.e., low N and P conditions) may favor the growth of *Synechococcus* because their small cell size is advantageous for nutrient uptake under oligotrophic conditions due to efficient nutrient diffusion per unit of cell volume [15,55]. The present results add a further potential scenario for picocyanobacterial dominance in the metalimnion; that is, the reduced nutrient availability associated with the increase in DOC loading in lakes may promote the population growth of *Synechococcus*. This may be the reason for the cascading negative responses of biogeochemical elements from DOC to DIN, and from DIN to both *Synechococcus* and CH4, that were observed in our study. In fact, a cross-lake comparison revealed that lakes with relatively high DOC and low DIN concentrations enhanced the degree of SMM development (Figure 7). Likewise, lakes with more abundant *Synechococcus* populations and a higher DOP pool promoted the development of SMM (Figure 7). Therefore, we conclude that the stoichiometric balance between DOC and DIN in the water column may regulate nutrient availability for bacterioplankton communities (e.g., *Synechococcus*) and may subsequently control CH4 production and SMM formation in deep freshwater lakes (Figure 9). Although the variability of DOC and DIN concentrations could also be explained by the extracellular release of DOC and DIN uptake by phytoplankters [28,29], the relationship between phytoplankton biomass (Chl *a*) and DOC concentration, as well as that between Chl *a* and DIN, was unclear (Figure S12). Moreover, the vertical profiles of DOC and *Synechococcus* density were not closely associated with each other for the study lakes (Figure S13), except Lake Saiko and Lake Hibara. Therefore, we argue that phytoplankton (including *Synechococcus*) per se may not be a major factor controlling the DOC and DIN variability, even though further studies are necessary to clarify the causal relationships between microbial activities and lake biogeochemical conditions.

**Figure 9.** Effects of the stoichiometric balance between DIN and DOC on nitrogen availability, *Synechococcus* density and dissolved CH4 concentration in the study lakes. See the legend for Figure 2 for a description of the symbols.

Although the present studies are conducted in monomictic or dimictic lakes where stratification occurs seasonally, enhanced SMM development might be predicted in meromictic lakes. This is because the lack of vertical mixing in such permanently stratified lakes may reduce the up-welling supply of nutrients from hypolimnion [56], which might result in a low DIN/DOC ratio in shallow water, as observed in the present study (Figure 9). If this is the case, the projected alteration in the lake mixing regime from dimictic to monomictic, or monomictic to meromictic, under climate change might increase the aerobic CH4 production and emission from the nutrient-depleted oxic lake waters. Further studies are required to precisely predict the effect of lake-mixing regime on SMM formation.

#### **5. Conclusions**

The formation of SMM might have a significant impact on methane emissions from deep freshwater lakes because methanotrophs generally consume the majority of CH4 and prevent fugitive methane emissions from deep profundal sediments or hypolimnetic zones in lakes [36,37]. In contrast, CH4, produced by aerobic methanogenesis within the SMM of the metalimnetic layer, might easily leak into the atmosphere because of the proximity of CH4 production sites to the lake surface. Therefore, the present study predicted that DOC loading triggers the cascading responses of biogeochemical processes, from N depletion to picocyanobacterial domination, which may promote CH4 production and emission to the atmosphere from the SMM layer. Although previous studies also reported that DOC might influence the CH4 emissions of freshwater ecosystems [57,58], these studies considered DOC as a direct substrate for anaerobic methanogenesis in sediments. In contrast, our study suggested that organic C exerts stoichiometric control over CH4 production in the oxic layer of lakes. DOC concentrations are now expected to increase due to ongoing climate change, which may result in significant changes to the structure and functioning of lake ecosystems [48]. For example, increasing concentrations of DOC often change the light and thermal environments in water columns, thereby affecting the primary and secondary productivity of lake food webs, as well as the thermal stratification regime in lakes. An increased DOC supply can also provide energy subsidies to heterotrophic consumers and consequently influence the metabolic balance of lake ecosystems. Moreover, DOC can change the chemical environments of lake ecosystems, such as their pH and dissolved iron concentrations. We argue that, in addition to these phenomena, increased organic carbon loading under changing environments may promote aerobic CH4 production and emission from the oxic layer of deep freshwater lakes.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4441/12/2/402/s1, supplementary data accompany this paper with Figure S1: Geographic location of nine study lakes in Japan, Figure S2: Vertical profiles of dissolved CH4 concentration (nM, red circles) and water temperature (◦C, gray circles) in nine study lakes during the period from July to September in 2016–2017. AS: Lake Ashinoko, MO: Lake Motosu, SA: Lake Saiko, AO: Lake Aoki, NO: Lake Nojiri, CH: Lake Chuzenji, HI: Lake Hibara, TO: Lake Toya, SH: Lake Shikotsu. Shaded areas denote the range of the thermocline, Figure S3: Vertical profiles of dissolved CH4 concentration (nM, red circles) and the percentage of surface irradiance (%, gray circles) in nine study lakes

during the period from July to September in 2016–2017. See legend of Figure S2 for the abbreviation of lake names and the descriptions of shaded areas, Figure S4: Vertical profiles of dissolved CH4 (nM, red circles) and chlorophyll *a* (μg/L, gray circles) concentrations in nine study lakes during the period from July to September in 2016–2017. See legend of Figure S2 for the abbreviation of lake names and the descriptions of shaded areas, Figure S5: Vertical profiles of dissolved CH4 (nM, red circles) and DOP (μM, gray circles) concentrations in nine study lakes during the period from July to September in 2016–2017. See legend of Figure S2 for the abbreviation of lake names and the descriptions of shaded areas, Figure S6: Vertical profiles of dissolved CH4 (nM, red circles) and DIN (μM, gray circles) concentrations in nine study lakes during the period from July to September in 2016–2017. See legend of Figure S2 for the abbreviation of lake names and the descriptions of shaded areas, Figure S7: Vertical profiles of dissolved CH4 (nM, red circles) and SRP (μM, gray circles) concentrations in nine study lakes during the period from July to September in 2016–2017. See legend of Figure S2 for the abbreviation of lake names and the descriptions of shaded areas, Figure S8: Vertical profiles of dissolved CH4 (nM, red circles) and DOC (μM, gray circles) concentrations in nine study lakes during the period from July to September in 2016–2017. See legend of Figure S2 for the abbreviation of lake names and the descriptions of shaded areas, Figure S9: Vertical profiles of dissolved CH4 concentration (nM, red circles) and TN/TP (ratio, gray circles) concentrations in nine study lakes during the period from July to September in 2016–2017. See legend of Figure S2 for the abbreviation of lake names and the descriptions of shaded areas, Figure S10: Relationships between *Synechococcus* cell density (cells/mL) and lake physicochemical variables in study lakes. Univariate regression lines were shown for the explanatory variables that were retained in the best model of GLMs with gamma distribution of response variable. See the legend of Figure S10 for symbol description, Figure S11: Vertical profiles of dissolved CH4 concentration (nM, red circles) and Mn (μM, gray circles) concentrations in nine study lakes during the period from July to September in 2016–2017. See legend of Figure S2 for the abbreviation of lake names and the descriptions of shaded areas, Figure S12: Relationships between phytoplankton biomass (log10 (Chl *a*), μg/L) and two variables (DIN and DOC, μM) in study lakes, Figure S13: Vertical distribution of *Synechococcus* density (red circles) in relation to the DOC profile (gray circles) in the study lakes. Shaded areas denote the range of the thermocline. See legend of Figure S2 for the abbreviation of lake names, Table S1: Watershed and lake morphological variables of the nine study lakes in Japan, Table S2: Water quality variables of the nine study lakes in Japan, Table S3: Water quality variables of tributary streams for study lakes, Table S4: The CARD-FISH probes used in the study, Table S5: Relative importance for limnological explanatory variables used in the generalized linear models (GLMs) with a gamma error distribution and an inverse link function for each response variable. Relative importance was evaluated by the sum of Akaike weights (*w*) for each model (CH4, DIN, DOP and TN/TP), Table S6: Results of the generalized linear models (GLM) assessing the univariate relationship between environmental variables and dissolved CH4 concentrations (as shown in Figure 2), the univariate relationship between DOC and chemical variables (Figure 3), and the univariate relationship between lake environmental variables and the peak subsurface maximum (SMM) in nine study lakes (Figure 7). GLMs were constructed with a gamma error distribution and an inverse link function for each variable (1/*y* = *ax* + *b*), except for the variables with an asterisk whose relationships were estimated with GLMs with a Gaussian error and an identity link function (*y* = *ax* + *b*), Table S7: Relative abundance (%) of phytoplanktonic algae in the epilimnion, metalimnion and hypolimnion of Table S6. Relative abundance (%) of phytoplanktonic algae in the epilimnion, metalimnion and hypolimnion of study lakes.

**Author Contributions:** T.I., Y.I., K.Y., D.I., T.K. and H.S. performed the field survey and laboratory analyses. S.K., T.I., Y.I. and K.Y. performed the CARD-FISH analyses. H.K. and T.K. performed the genome analyses and database research with S.K. and T.I. R.S. performed the 1H–31P NMR analyses. S.K., Y.I., K.Y., and T.I. performed the data analyses and presentation. S.K. and T.I. wrote the manuscript with the assistance of the other co-authors. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Grants-in-Aid (B) (16H02935) and by the Grant-in-Aid for Young Scientists (A) (23681003) from the Japan Society for the Promotion of Science, the Grant for Joint Research Program of the Institute of Low Temperature Science, Hokkaido University in 2016, and the Grant from the Nippon Life Insurance Foundation in 2015.

**Acknowledgments:** We thank the laboratory members for their support of our study. We also thank Denboh and the staff of Lake Toya Field Station at Hokkaido University and the staff of Shikotsuko Fishermen's Cooperative for their support during our field survey.

**Conflicts of Interest:** The authors declare no competing financial interests.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Protected Freshwater Ecosystem with Incessant Cyanobacterial Blooming Awaiting a Resolution**

**Nada Tokodi 1, Damjana Drobac Backovi´c 1,\*, Jelena Luji´c 2, Ilija Š´ceki´c 2, Snežana Simi´c 3, Nevena Đorđevi´c 3, Tamara Duli´c 4, Branko Miljanovi´c 1, Nevena Kitanovi´c 2, Zoran Marinovi´c 2, Henna Savela 5, Jussi Meriluoto 1,4 and Zorica Svirˇcev 1,4**


Received: 27 November 2019; Accepted: 29 December 2019; Published: 31 December 2019

**Abstract:** For 50 years persistent cyanobacterial blooms have been observed in Lake Ludoš (Serbia), a wetland area of international significance listed as a Ramsar site. Cyanobacteria and cyanotoxins can affect many organisms, including valuable flora and fauna, such as rare and endangered bird species living or visiting the lake. The aim was to carry out monitoring, estimate the current status of the lake, and discuss potential resolutions. Results obtained showed: (a) the poor chemical state of the lake; (b) the presence of potentially toxic (genera *Dolichospermum, Microcystis, Planktothrix*, *Chroococcus*, *Oscillatoria*, *Woronichinia* and dominant species *Limnothrix redekei* and *Pseudanabaena limnetica*) and invasive cyanobacterial species *Raphidiopsis raciborskii*; (c) the detection of microcystin (MC) and saxitoxin (STX) coding genes in biomass samples; (d) the detection of several microcystin variants (MC-LR, MC-dmLR, MC-RR, MC-dmRR, MC-LF) in water samples; (e) histopathological alterations in fish liver, kidney and gills. The potential health risk to all organisms in the ecosystem and the ecosystem itself is thus still real and present. Although there is still no resolution in sight, urgent remediation measures are needed to alleviate the incessant cyanobacterial problem in Lake Ludoš to break this ecosystem out of the perpetual state of limbo in which it has been trapped for quite some time.

**Keywords:** cyanobacteria; blooms; microcystin; Lake Ludoš

#### **1. Introduction**

In the very north of Serbia there is an old and unusual lake, Lake Ludoš, with beautiful open water landscapes surrounded by reeds, wetlands and steppe. The environment is rich in many plant and animals species. European pond turtles, various amphibians, otters, moles, rabbits, foxes and roe deer have found their home there. What makes Lake Ludoš especially famous, and validates the name originating from the Hungarian word "lud" meaning goose, is that there are more than 200 bird species, including rare and endangered species, nesting or resting during their migration. Because of all this, Lake Ludoš is recognized and protected as a special nature reserve on the list of Ramsar sites.

One part of the lake is often visited by fishermen, but their catch mostly consists of the very resilient and adaptable Prussian carp (*Carassius gibelio* (Block, 1782)) which are quite small in size. This may be related to the fact that the water of Lake Ludoš has an intense green color throughout the year caused by cyanobacteria (e.g., [1]). Their extensive growth and blooming causes many problems in freshwater ecosystems, including this one. Cyanobacteria can produce cyanotoxins that affect other organisms, including valuable flora and fauna, especially aqueous organisms such as fish [2]. Cyanobacteria also present a threat to humans, such as fishermen, who may be exposed to cyanotoxins through contaminated food, inhalation and direct contact [3,4]. In addition to human health, the health of this important ecosystem is also jeopardized. Furthermore, this "disease" could also be transmissible, since there is a possibility that water birds visiting the lake during their migration path can disseminate toxic cyanobacteria [5].

Lake Ludoš is only one of many aquatic ecosystems in Serbia where cyanobacteria are present and blooming [6]. What sets this ecosystem apart from many others is that it has been known for perpetual blooming in the last 50 years. Previous research has shown that the lake is in a poor ecological state which leads to the question of whether the protection of this natural habitat in a bad ecological state is justified. The cyanobacterial problem, which can potentially affect every living being in the proximity of this ecosystem, has also been preserved as measures to improve the water quality have not been undertaken on the lake [5]. The problem of cyanobacteria in Lake Ludoš has been addressed during our previous research when potentially toxic cyanobacterial species, cyanotoxins in water, macrophytes and fish tissues were detected, as well as histological alterations and DNA damage in fish tissue (see [5]). Six years later further monitoring was carried out in order to estimate the current state of this ecosystem. Therefore, several investigative steps have been taken: monitoring of physical and chemical parameters of the lake; assessment of qualitative and quantitative analyses of cyanobacteria; the first survey of the cyanotoxin coding genes; determination of cyanotoxins in water and fish; analysis of histopathology of different fish organs; and discussion of potential health risks and resolutions.

Freshwater ecosystems throughout the world have similar problems in connection to cyanobacterial blooming and cyanotoxin production. Recent publication of a global geographical and historical overview of cyanotoxin distribution demonstrated the presence of well-known cyanotoxins in each continent (including 520 lakes) and their harmful consequences on human health [4]. Hence, this issue is of global concern. The present investigation aims to assist in making appropriate decisions and measures for the remediation of not only this, but many other old and rapidly aging and protected lakes.

#### **2. Materials and Methods**

#### *2.1. Sampling Site and Sampling of Water and Fish*

Lake Ludoš (Figure 1) is a one of the few preserved shallow lakes in the region. It has a maximum depth of 2.25 m, is 4.5 km long, and it represents a remnant of the Pannonian Sea. In most places the depth does not exceed 1 m, and it may be frozen for more than three months a year. It is located in the north part of Serbia, near the city of Subotica. The lake and the associated wetland ecosystem is highly valued due to the great biological diversity, and as such the area is classified among wetlands of international significance. The quality of the lake's water is of great importance for the preservation of the flora and fauna connected to this marshland ecosystem.

The lake is supplied with water from aquifers and Kereš River. However, in the northern part, Lake Ludoš receives water from the canal Pali´c-Ludoš which is the recipient of wastewaters from the Pali´c settlement. Water treatment of these wastewaters is still inadequate and the canal water is characterized by a high level of organic pollution, high concentrations of salt and very high nutrient concentrations. The inflow of untreated and partially purified waters in Lake Ludoš contributes to the deterioration of the water quality and the increase of the sludge quantity [7–11].

**Figure 1.** Pier view of the blooming Lake Ludoš in July 2018.

Water samples were collected from the surface water layer within the littoral zone (pier next to the visitor center) (46.103207 N, 19.821360 E) and from the center of the lake (46.102159 N, 19.821149 E) in March, May, July and September of 2018. Samples of Prussian carp were collected from the center of the lake before and after the summer (March and September 2018) with gillnets of various mesh sizes and a standard electrofishing device.

#### *2.2. Analyses of Physical and Chemical Parameters*

Multi-parameter WTW probes were used for carrying out in situ measurements in the field and the following physical and chemical parameters were determined: temperature, pH, conductivity, O2 concentration and O2 saturation. TSS (total suspended solids), TOC (total organic carbon), NO3, detergents, COD (chemical oxygen demand) and BOD (biological oxygen demand) were measured in the laboratory conditions with a Pastel Ultraviolet (UV) Secomam.

#### *2.3. Qualitative and Quantitative Analyses of Cyanobacteria*

The phytoplankton samples for the cyanobacterial qualitative analysis were collected by sweeping a plankton net (netframe 25 cm ø, net mesh 23 μm). All samples were immediately preserved in a Lugol solution. Taxonomic identifications of cyanobacteria were made according to several taxonomic keys [12–15] and were done under a light microscope Motic BA310 using a Bresser (9MP) digital camera and Micro Cam Lab software. For the quantitative analysis of phytoplankton, the 15 L of water was collected by sweeping a plankton net at the depth of 0.3 m in March, while in May, July and September only 200 mL of water were collected directly from the lake as a result of high bloom density. The quantitative analysis was made by using the Utermöhl method [16] under a Motic AE 2000 inverted microscope. The phytoplankton individuals were sedimented and cyanobacteria quantified on the chamber (i.e., transects) with an inverted microscope at different magnifications depending on their size (100×, 400×) and expressed as the number of cells per mL.

#### *2.4. Cyanotoxin Coding Gene Analyses*

2.4.1. Samples—Reference Strains for Polymerase Chain Reaction (PCR) Analysis

Reference strains were obtained from Pasteur Culture Collection (PCC), National Institute for Environmental Studies Microbial Culture Collection (NIES), Australian National Algae Culture Collection (CS), and Finnish Environment Institute (SYKE). They consisted of:

• microcystin (MC) producers: PCC7820 (*Microcystis aeruginosa*), NIES-107 (*Microcystis wesenbergii*);


#### 2.4.2. DNA Extraction

Depending on the bloom density, 30–400 mL of water samples were filtered (pore size 2–3 μm), and filtride was freeze-dried. Approximately 10 mg of freeze-dried biomass was used for DNA extraction from reference strains. Genomic DNA from biomass of the reference strains and filtrides was extracted with the DNeasy Plant Mini Kit (QIAGEN, Hilden, Germany) according to manufacturer's instructions, with minimal modifications for the extraction from filtrides (double amount of Buffer AP1, RNase A and Buffer P3 was added to fully suspend the samples). During the initial steps of extraction, samples were homogenized using zirconia/silica disruption beads (0.5 mm) and by vortexing for 1 min. The quality was assessed spectrophotometrically (NanoDrop ND-1000, Thermo Scientific, Waltham, MA, USA), where A260/A280 ratio varied between 1.22 and 2.04.

#### 2.4.3. Qualitative PCR

Qualitative PCR was run to analyze samples for the presence of MC (*mcyE*), CYN (*cyrJ*), STX (*sxtA, sxtG, sxtS*) and ATX (*anaC*) synthetase genes. PCR reaction mixtures were prepared in a total volume of 20 μL containing 1× Phire Reaction Buffer, 0.4 μL Phire II HotStart polymerase (Thermo Scientific), 0.2 mM deoxyribonucleotide triphosphates (dNTPs) (Thermo Scientific), 0.5 μL forward and reversed primers (Table 1), 2 μL of template and sterile deionized water. PCRs were run on a C1000 Touch Thermal Cycler (Bio-Rad, Helsinki, Finland) according to the following protocols: initial denaturation for 30 s at 98 ◦C; 40 cycles of 5 s at 98 ◦C, 5 s at 61 ◦C (HEPF, HEPR, sxtA1480\_R stxA855\_R), or 62 ◦C (cyrJ\_F, cyrJ\_R, sxtG432\_F, sxtG928\_R, sxtS205\_F, sxtS566\_R), or 52 ◦C (anaC-genF, anaC-genR) and 10 s at 72 ◦C; and a final extension of 1 min at 72 ◦C. To examine the potential inhibition of PCRs, an exogenous amplification control template was prepared containing 1 μL:1 μL (reference:sample). Following strains were used as a reference in the control template: PCC7820 for *mcyE*, CS-506 for *cyrJ*, CS-537/13 for *sxtA, sxtG, sxtS*, and ANA123 for *anaC*. Visualization of PCR products was performed on a 1.5% Top Vision agarose gel (Thermo Scientific) dyed with SYBR® Safe DNA gel stain. The observed bands were documented on Gel Doc™ XR (Bio-Rad) using Quantity One software (v. 4.6.9).


**Table 1.** List of primers used for qualitative polymerase chain reaction (PCR).

#### *2.5. Cyanotoxin Analyses*

2.5.1. Preparation of Water Samples for Liquid Chromatography–Tandem Mass Spectrometry (LC–MS/MS)

Depending on the bloom density, 30–100 mL of water samples were filtered (pore size 2–3 μm). Biomass on the filter was then freeze-dried. Afterwards, filtrides were placed in glass tubes and the toxin was extracted with 3 mL of 75% MeOH and ultrasonication. The extracts were centrifuged for 10 min at 10,000× *g* and two 1 mL aliquots of the supernatant were evaporated to dryness (50 ◦C nitrogen flow) in glass tubes. For MC analysis the sample was redissolved in 75% MeOH in 200 μL, and for CYN analysis the sample was redissolved in 200 μL H2O. The samples were then filtered (0.2 μm GHP ACRODISC 13 Pall Life Sciences, Ann Arbot, MS, USA) into inserts and were ready for liquid chromatography–tandem mass spectrometry (LC–MS/MS) analysis.

The extracellular MC content was concentrated by solid-phase extraction (SPE) on Waters Oasis HLB (30 mg). The samples eluted with 5 mL 90% MeOH were placed into glass tubes, evaporated using nitrogen flow and redissolved in 200 μL of 75% MeOH. Subsequently, they were filtered (0.2 μm GHPACRODISC 13 Pall Life Sciences) into inserts and were ready for LC–MS/MS analysis.

#### 2.5.2. Preparation of Fish Tissue Samples for LC–MS/MS

Prussian carp from Lake Ludoš was analysed for MCs in fish liver, gills, kidney, intestine, gonads (testis and ovaries), spleen, and muscle samples by LC–MS/MS. A total of 30 individuals (TL: 16.63 ± 2.24 cm, SL: 14.58 ± 1.68 cm, 18 female/12 male) were collected during two separate sampling surveys—spring (March) and autumn (September) of 2018. Different fish tissues were separately homogenized and freeze-dried. Samples of the same organ of all the individuals were pooled together. Before further preparation, several fish tissues were spiked in order to test the preparation method. Freeze-dried fish tissue samples (100 mg) were placed into glass tubes and 5 mL of 75% MeOH was added for extraction of cyanotoxins. Homogenization was performed on ice for 30 second, and the samples were then ultrasonicated in a bath sonicator for 15 min and further extracted with a probe sonicator. Samples were then centrifuged for 10 min at 10,000× *g* followed by an addition of 1 mL of hexane to 2 mL of the supernatants obtained. The hexane (lipid) layer was removed using glass pipettes and the remaining samples were evaporated (50 ◦C nitrogen flow) in glass tubes. Finally, samples were redissolved in 300 μL 75% MeOH and filtered (0.2 μm GHPACRODISC 13 Pall Life Sciences) into inserts. The fish tissue samples were then ready for LC-MS/MS analysis.

#### 2.5.3. LC–MS/MS

Toxin analyses were performed by LC–MS/MS [22]. The analytical targets consisted of nine MC variants (MC-dmRR, MC-RR, MC-dmYR, MC-YR, MC-dmLR, MC-LR, MC-LY, MC-LW and MC-LF) and CYN.

#### *2.6. Analyses of Fish Histology*

Captured Prussian carp from Lake Ludoš used for cyanotoxin detection were also used for histological analyses. Liver, kidney, gill, intestine, spleen, gonad and muscle samples were dissected from each fish and fixed in 4% formaldehyde. Additionally, 6 individuals of common carp (*Cyprinus carpio* L.) (TL: 18.35 ± 6.24 cm, 3 females/3 males) were obtained from the Department of Aquaculture, Szent István University, Hungary. These fish were kept in a recirculation system (Sentimento Kft., Hungary), under a 12 h light/12 h dark cycle at 24 ± 0.2 ◦C and served as a control group in this study.

After fixation of at least three days, samples were processed by standard histological procedure. Gill and muscle samples were decalcified beforehand. For tissue processing, samples were dehydrated in graded series of ethanol, cleared in xylene and subsequently embedded in paraffin wax blocks. Three five-μm-thin sections per tissue per individual were cut and placed onto glass slides, and

stained with haematoxylin and eosin (H&E) dyes. Sections were examined under a Nikon Eclipse 600 microscope and photographed using a QImaging Micro Publisher 3.0 digital camera.

#### **3. Results**

#### *3.1. Physical and Chemical Parameters of Water Samples*

The recent investigations of Lake Ludoš during 2018 corroborated the continuously poor chemical state of the lake. pH levels, saturation with O2 as well as electrical conductivity were high during the investigated period (Table 2).


**Table 2.** Physical and chemical parameters of water from Lake Ludoš in 2018.

#### *3.2. Presence of Cyanobacterial Species in Water Samples*

Most dominant cyanobacterial species were *Limnothrix redekei* (Van Goor) Meffert and *Pseudanabaena limnetica* (Lemmermann) Komárek (Table 3, Figure 2). Furthermore, both *Microcystis* species, *Microcystis aeruginosa* (Kützing) Kützing and *Microcystis wesenbergii* (Komárek) Komárek, were numerous during the whole investigated period. At the end of the summer, an invasive *Raphidiopsis raciborskii* (Woloszynska) Aguilera, Berrendero Gómez, Kastovsky, Echenique and Salerno (basionym *Cylindrospermopsis raciborskii* (Woloszynska) Seenayya and Subba Raju) also started to occur. Usually, more cells per mL were found in the pier samples compared to the center samples; however, the same species were present in both sampling sites.

**Figure 2.** Dominant cyanobacterial species from Lake Ludoš in 2018: (**a**) *Limnothrix redekei*; (**b**) *Pseudanabaena limnetica*; (**c**) *Raphidiopsis raciborskii*; (**d**) *Microcystis aeruginosa*; (**e**) *Microcystis wesenbergii*; Scale bars: 10 μm.



75

#### *Water* **2020** , *12*, 129

#### *3.3. Presence of Cyanotoxin Coding Genes in Biomass Samples*

Biomass samples were tested for the presence of cyanotoxin coding genes, including MCs, STX, CYN and ATX (Table 4).


**Table 4.** The prevalence of *mcyE, sxtA, sxtG, sxtS, cyrJ* and *anaC* PCR products in Lake Ludoš.

Legend: (+) amplified; (−) not amplified; (/) not analysed.

MC coding gene *mcyE* (472 bp, Figure 3a) was amplified in all samples. STX coding genes *sxtG* (519 pb, Figure 3c) and *sxtS* (382 bp, Figure 3d) were amplified in a total of 6 and 4 samples, respectively. The *sxtG* gene was observed in all sampling seasons, while *sxtS* gene was observed only in samples collected in August and September. Saxitoxin coding gene *sxtA* (648 bp, Figure 3b), CYN coding gene *cyrJ*, and ATX coding gene *anaC* were not amplified in this study. Significant inhibition of the PCR reaction was not observed in exogenous amplification control templates.

**Figure 3.** Visualization of PCR products on agarose gel: (**a**) *mcyE*; (**b**) *sxtA*; (**c**) *sxtG*; (**d**) *sxtS*. Legend: L—ladder; B—blank; R1, R2—reference strains; C—exogenous amplification control; S1—sample September—center; S2—sample September—pier; S3—sample July—center; S4—sample July—pier; S5—sample May—center; S6—sample May—pier; S7—sample March—center; S8—sample March—pier.

#### *3.4. Presence of Cyanotoxins in Water and Fish Samples*

Biomass and extracellular content of water samples were tested for the presence of cyanotoxins. Several MC variants were noted in the total content (Table 5) including most commonly occurring MC-LR and MC-RR, however, their concentrations were rather low during the whole investigated period. CYN was not detected during the investigated period in Lake Ludoš, even though there were some cyanobacterial species that could potentially produce this cyanotoxin.


**Table 5.** Presence and highest concentrations of microcystin (MC) variants in Lake Ludoš during 2018.

Legend: (+) present in sample; (−) not present in sample.

Analyses of several fish tissue samples did not show a presence of investigated MC variants in spring nor autumn samples.

#### *3.5. Histological Alterations in Fish Samples*

During both investigated periods, spring (before) and autumn season (during and after the bloom), most affected fish organs were liver, kidneys and gills. Liver samples of individuals from the control group displayed typical organization of the hepatic parenchyma, with cord-like formations of hepatocytes interspaced with sinusoids and radially arranged around blood vessels (Figure 4a). Hepatocytes were polygonal, with a clearly visible cell membrane and large round nuclei with distinguishable nucleoli. In contrast, microscopic examination of fish from Lake Ludoš revealed severe alterations of liver histology which were observed in both March and September sampling groups. Livers of these fish showed changes in architectural structure, with less prominent cord-like organization and sinusoid capillaries no longer clearly distinguishable (Figure 4b). Loss of shape and rounding of hepatocytes was most characteristic in the March group, with large groups of cells displaying ball-or onion-like shape (Figure 4c). Altered hepatocytes typically had darker nuclei with condensed chromatin and no discernable nucleoli. Signs of kariopyknosis, predominantly in the September sampling group, were indicative of necrosis (Figure 4d). Most prominent alterations present in all examined samples were glycogen depletion and vacuolization of hepatocytes. Many cells had a completely clear cytoplasm, which suggest hypervacuolization (Figure 4e).

**Figure 4.** Histopathological alterations in the liver of Prussian carp *Carassius gibelio* from Lake Ludoš, 2018: (**a**) control fish; (**b**) loss of the cord-like parenchymal structure; (**c**) rounding of hepatocytes; (**d**) necrotic fields (asterisk) with hepatocytes displaying karyopyknosis; (**e**) hepatocytes displaying vacuolization and glycogen depletion. Haematoxylin and eosin (H&E) staining. A, B and D–50 μm; C and E–20 μm.

Renal corpuscles in the kidneys of the control individuals were round in shape and had relatively large glomeruli. The Bowman's capsule was continuous with thin intercapsular space. Both proximal and distal renal tubules had one layer of columnar epithelial cells with proximal segments having basal nuclei, and distal segments having central nuclei and less intensive cytoplasmic stain (Figure 5a). Even in controls, slight clogging of tubules and slight vacuolization was observed.

**Figure 5.** Histopathological alterations in the kidney of Prussian carp *Carassius gibelio* from Lake Ludoš, 2018: (**a**) control fish; (**b**) degeneration of tubules including vacuolization and separation of epithelial layer from the basal lamina; (**c**) reduction of glomeruli size and intense dilatation of Bowman's capsule. H&E staining. A and B–50 μm; C–20 μm.

Only individuals sampled in March had significant pathological alterations in the kidney. These included degeneration and loss of nephron formation, as well as interstitium structure. Renal corpuscles showed a reduction of glomeruli size, accompanied with dilatations of intercapsular space of the Bowman's capsule (Figure 5c). In tubules, epithelial cells showed intense vacuolization and in some cases the epithelial layer was separated from the basal lamina (Figure 5b). The number of tubules appeared clogged and in the process of necrosis. Cells in the necrotic area had pyknotic nuclei and displayed signs of cell membrane lysis, such as no discernable boundary between cells. In some fish, necrosis, hyalization of the interstitium and the presence of macrophage aggregates were evident. Certain alterations, such as vacuolization and tearing of the tubular epithelium, were also present in individuals from the September group; however, this was not as frequent and severe compared to the March group.

Gills of the control group showed no pathological changes (Figure 6a). Secondary lamellae regularly lined both sides of the primary lamellae (filament) and were covered with one layer of squamous epithelial cells. Contrarily, individuals from Lake Ludoš in both sampling periods (March and September) had noticeably altered gill structure. Several individuals displayed signs of hyperplasia, as well as hypertrophy of interlamellar cell mass, mainly epithelial and mucous cells. Such swelling of secondary lamellae and proliferation of interepithelial cells has led to a complete fusion of the secondary lamellae, especially noticeable in the September sampling group (Figure 6b). Other observed lesions included epithelial lifting and oedema, accompanied with hypertrophy of interepithelial chloride cells (Figure 6c). In some individuals, endothelial cells of the capillaries showed signs of telangiectasia (aneurysm), along with epithelial rupture and hemorrhage (Figure 6d).

Other examined organs of the bloom-exposed fish in the present study did not display histopathological alterations (not shown). Intestines of all groups showed normal histology, with villi regularly lining the lumen. A single layer of enterocytes with basal nuclei lined the surface of the vili, along with fewer goblet cells. Furthermore, sections of muscle tissue in all groups also showed no structural alterations. Spleen of all examined individuals displayed a normal structure, with aggregates of erythroid and lymphoid cells, surrounding blood vessels. Neither male nor female gonads had histopathological changes. Testes and ovaries had normal structural organization with germline cells present in different stages and numbers, which is dependent on the season and age of the fish.

**Figure 6.** Histopathological alterations in the gills of Prussian carp *Carassius gibelio* from Lake Ludoš, 2018: (**a**) control fish; (**b**) fusion of the secondary lamellae; (**c**) intensive proliferations of chloride cells (arrowhead) and oedema (arrow); (**d**) telangiectasia. H&E staining. A and D–50 μm; B–100 μm; C–20 μm.

#### **4. Discussion**

#### *4.1. Monitoring of the Water*

4.1.1. Physical and Chemical Parameters in Lake Ludoš

The measurements of physical and chemical parameters showed several important findings:


It is assumed that cyanobacteria are more resistant to solute increases compared to other phyla e.g., Chlorophyta [23]. Additionally, conductivity changes lead to decreased zooplankton—predators of phytoplankton, thus it is possible that a reduction in zooplankton would potentiate phytoplankton increase.

Regular monitoring showed similar findings in recent years (2013–2017) [7–11]. pH levels were between 8 and 9, O2 saturation during the year was uneven with high values and supersaturation during summer and sometimes during autumn months, while high values of electric conductivity of the water seemed to be rising in the recent years. Additionally, COD was extremely high, similar to wastewaters, indicating a poor status of the lake and based on the measured parameters it is recommended that the lake should not to be used for any purpose [24]. BOD was uneven during the year, demonstrating the problem of instability of the system. The water of the Lake Ludoš is very rich in nitrogen compounds (mostly organically bound nitrogen), which led to increased biological production. Nitrates were uneven during the year and seemed to depend on the inflow from the Pali´c-Ludoš canal. Phosphates showed a similar trend as nitrates and their concentration seemed to be on the rise thus contributing to the deterioration of the water quality. Sediment analysis suggested rich deposits of nutrients which will contribute to the continual hypertrophic state of the lake [7–11].

The status of surface waters in terms of general quality can be shown by the Serbian Water Quality Index (SWQI). SWQI is based on 10 quality parameters (temperature, pH, conductivity, O2 saturation, BOD for 5 days, suspended matter, total nitrogen oxides, orthophosphates, ammonia ions, coliform bacteria) that are aggregated into a composite indicator of the quality of surface waters, leading them to one index number [24]. SWQI of the water from Lake Ludoš through 2018 found it to be of either very low quality or low quality [25]. Similar findings were also noted for the period from 2013 to 2017 [7–11].

#### 4.1.2. Cyanobacterial Community in Lake Ludoš

During 2018, the most dominant cyanobacterial species were *L. redekei* and *P. limnetica* (Table 3). Furthermore, both *Microcystis* species, *M. aeruginosa* and *M. wesenbergii*, were numerous during the whole investigated period. Same species were present in pier and center samples, although more cells were noted in the pier samples possibly due to lower depth, higher temperature, wave, and wind effects. In the recent years (2013–2017), several other cyanobacterial species were also frequently found: *Microcystis delicatissima*, *Oscillatoria agardhii* (current *Planktothrix agardhii*), *Oscillatoria putrida*, *Lyngbya limnetica* and *Anabaena spiroides* [7–11]. Furthermore, in our previous research during 2011 and 2012, similar cyanobacterial species were abundant: *L. redekei*, *P. limnetica*, *P. agardhii* and *Microcystis* spp. [5]. Most of the present species and genera are known as potential cyanotoxin producers.

Additionally, in the autumn of 2018, the invasive species *Raphidiopsis raciborskii* was also found. This invasive and potentially toxic species was first noted in Serbia in 2006 [26] and soon in Lake Ludoš as well [27], and since then it has been frequently found and blooming in the lake. The *R. raciborskii* presence is of particular concern due to its ability to expand its distribution rapidly (see [28]). Differences in toxin production are known among the strains: South American strains produce STX, Australian and Chinese strains produce CYN, and European and North American strains are considered to be non-toxic [28]. Furthermore, non-toxic and toxic strains can co-exist, and even co-occurring strains exhibit genomic variability [29]. Bearing in mind that this species is expanding its distribution in Europe, and in Serbia as well, it is necessary to establish whether this species is a greater threat than previously assumed, and how it succeeds in spreading so rapidly.

#### 4.1.3. Presence of Cyanotoxin Genes and Cyanotoxins in Lake Ludoš

For the first time, Lake Ludoš was assessed for the presence of genes coding several frequently occurring cyanotoxins. During investigated months, the MC-coding gene *mcyE* was amplified in all the biomass samples. This is in accordance with the cyanobacterial species composition observed in the lake, and confirms the uniform distribution of MC-producing species throughout the year. Amplification of STX coding genes, *sxtG* and *sxtS* occurred in some of the samples. The *sxtG* gene was observed in all sampling seasons, while *sxtS* gene was observed only in samples collected in August and September. The STX *sxtA* coding gene was not amplified in this study. Since the specificity of *sxtA*, *sxtG* and *sxtS* primers has been previously validated by Savela et al. [19,20], the lack of targeted genes may be due to unexpected sequence dissimilarities between primers and target that can occur in natural populations. Large-scale gene mutations such as deletions and insertions resulting in non-functional gene clusters may also cause a lack of PCR amplification. Major deletions events in MC-coding gene cluster were previously observed in non-toxic *Planktothrix* strains [30], suggesting that similar events may occur in STX-producing strains. Similarly to this study, the lack of the some of the targeted STX genes was previously observed by Savela et al. (2017) in Polish lakes. As STXs were not measured directly, conclusions on correlation between gene presence/absence and toxin production cannot be made. However, since STXs have been detected in the lake Ludoš before [5], and a potential STX producer, *R. raciborskii,* was detected, additional studies are warranted to confirm potential production of STXs in the lake.

There are several publications that show MC presence in various water bodies in Serbia [6,31]. However, only a few publications demonstrate presence of cyanotoxins in Lake Ludoš. MCs were first

noted in 2006 by Simeunovi´c [32,33], thereafter the presence of MCs and low concentrations of STXs in the lake was observed during our previous investigation in 2011 and 2012 [5]. In 2018, analyses of water and biomass samples for the presence of MCs and CYN were performed by LC–MS/MS and several MC variants were detected in low concentrations. CYN was not detected, although some investigations have shown that some of the present cyanobacterial species potentially could produce this toxin (e.g., [34,35]).

#### *4.2. Uptake of Cyanotoxins and Bloom E*ff*ects*

Fish are directly influenced by water blooms, which can cause health impairments and mortality [36]. Primary route of exposure is by ingestion, either directly or through the food web, and indirectly via epithelial absorption [37]. Consequently, liver and gills are often the most affected organs during cyanobacterial blooms [38,39]. Furthermore, there is evidence of cyanotoxin accumulation in fish tissue, which ultimately endangers the health of animals and even humans that use them as a food source (e.g. [40,41]). Multiple other stress factors co-occuring with the blooms can additionally threaten fish, damaging tissues and impairing their development and survival [42].

A previous investigation of Lake Ludoš by the authors showed cyanotoxin uptake by macrophytes and fish. Accumulation of MC-RR was detected in the rhizome of *Phragmites australis*, cattail *Typha latifolia* and royal blue water lily *Nymphaea elegans* [5]. Furthermore, the same investigation demonstrated accumulation of two MC variants (MC-LR and MC-RR) in the tissue samples of the Prussian carp from Lake Ludoš. During 2011 and 2012, MCs were found in the intestine (MC-LR) and muscle samples (MC-RR), however, no MCs were found in the liver samples. In 2011, the fish gills were also positive for MC-RR, and in 2012, kidneys and gonads were found to accumulate MC-LR [5]. In 2011, histopathological changes were observed in liver, kidney, gills and intestine samples of the tested individuals. Furthermore, DNA damage was detected in Prussian carp as revealed via comet assay in 2012. Three tissues were selected and assessed: blood had the lowest level of DNA damage, while liver and gills showed more damage [5]. In another study from fishponds in Serbia, MC-RR accumulation in muscle tissues was recorded and histopathological changes were noted in liver, kidneys, gills, intestines and muscles of the *Cyprinus carpio* tissues [41]. Many histological changes in different fish species and tissues after exposure to cyanobacterial bloom or cyanotoxins were presented in a review by Svirˇcev et al. [2].

In addition to histological changes and accumulation, cyanotoxins can affect fish growth, development, reproduction, and survival. Embryos and larvae are especially sensitive to the effects of MCs, and exposure to these cyanotoxins in the early life stages can disrupt embryos, reduce survival and growth rate, and cause disorders such as: small head, curvature of the body and tail, enlarged and opaque yolk sachet, hepatobiliary abnormalities and cardiac disturbances [43–45]. Furthermore, cyanobacteria and their metabolites can have a wide range of negative effects on the adults of fish. They affect locomotor activity and fish behavior, impair ingestion rate and growth, disrupt heart function, ion regulation, cause changes in serum composition, trigger oxidative stress and disturb the reproduction of fish. Alongside direct harmful effects caused by cyanotoxins, cyanobacteria can adversely affect fish through the alteration of environmental conditions during blooming. Decreased dissolved oxygen, increase in ammonia concentration, changes in pH value and water temperature, or a combination of all of these factors can have detrimental changes on fish [46].

#### 4.2.1. Cyanotoxin Accumulation in Fish Tissues from Lake Ludoš

Several fish tissues of *Carassius gibelio* were examined for the presence of MCs. As already stated, in our previous study accumulation of two microcystin variants (MC-LR and MC-RR) was noted in several fish tissues: muscle, intestines, kidneys, gonads and gills of *Carassius gibelio* [5], however, accumulation in tissues was not found during this investigation. Although spiked samples demonstrated that the method for the preparation of the samples is adequate, it is possible that cyanotoxins remained covalently bound to protein phosphatases in the tissue [47–50] resulting in

negative results. Furthermore, it is possible that cyanotoxins were excreted and concentrations lowered by detoxification processes which vary between different species, organ, MC congener and metabolism [40,51–53], or even that concentrations in water were not high enough to be accumulated in high concentrations for detection.

#### 4.2.2. Bloom Effects on Histopathology of Fish from Lake Ludoš

Histopathology has an important place as a biomarker of fish health status, as histological changes often occur in response to acute or chronic exposure of an organism to a pollutant or a hazardous chemical, as well as adverse conditions present in the water. Even though no accumulation of cyanotoxins were detected during this investigation, histological observation of tissues from Prussian carp taken from Lake Ludoš suggest serious health implications often attributed to water blooming and cyanobacteria-rich aquatic environments. In general, most affected organs during both the spring and autumn seasons were liver, kidneys and gills. Furthermore, presence of alterations in both sampling periods (spring and autumn) suggests that chemical state of the lake was poor and that harmful blooms probably occurred in the previous year.

Structural alterations of the liver observed in this study are similar to those found by other authors after the exposure of fish to cyanotoxins or extracts of cyanobacterial strains and blooms [54–60]. Most prominent changes were loss of parenchymal structure, rounding of cells, glycogen depletion, vacuolization and pyknosis, all of which were so far reported after exposure of fish to MCs [2]. These alterations, specifically vacuolization and increase of lipid content, were also observed in mammalian livers in reaction to MC [61–63]. Hepatotoxicity of MC might be attributed to its affinity to bind and inhibit eukaryotic serine-threonine protein phosphatases 1 and 2A (PP1 and PP2), enzymes that are important in maintaining cell homeostasis and tumor suppression signaling pathways [64–66]. Studies in mammalian models have shown that inhibition of PP1 and PP2 can cause cytoskeletal damage [67,68] and may lead to loss of cell-to-cell contact, rounding of hepatocytes, condensation of chromatin and nuclear pyknosis.

Kidneys may be good indicators of environmental stress since they receive the majority of postbranchial blood. Additionally, kidney tubule cells possess a transport mechanism similar to that of hepatocytes (the multispecific bile acid transport system) which is responsible for MC uptake into the cell [69]. A study undertaken by Fischer and Dietrich [56] has shown that due to this efficient uptake of toxins in carp, MC-induced kidney pathologies in carp develop rapidly and at lower toxin concentrations. Such a finding supports the results of this study, as the kidneys exhibited the strongest histopathological changes, particularly in the fish that were taken during the spring when the levels of cyanotoxins are lower. Histological changes observed in the present study supports previous research on the effects of MCs on fish [39,55,56]. Glomerulopathy, with dilatations of the Bowman's capsule, and vacuolization of tubules are the most common alterations observed after exposure of fish to MCs [70,71]. Necrosis and impairment of renal tubules can affect ion and water regulation in the kidney, thus damaging the survival capabilities of fish [72].

Fish gills constitute over 50% of the total surface area of the animal and are in direct contact with water, which makes them sensitive to pollutants and toxic chemicals. A number of histopathological alterations were detected in fish from Lake Ludoš, which can be associated with conditions during water blooms, mostly the presence of cyanotoxins [54,57,73]. Lifting of the lamellar epithelial cells caused by the fluid penetration and epithelial hyperplasia were the most common lesions. Along with epithelial hyperplasia, these are considered defensive mechanisms of gills, both of which reduce the uptake of xenobiotics [74,75]. Other persistent alterations, such as oedema and telangiectasia from the secondary lamellae could be attributed directly to MC toxicity. Gill ion pumps (Na<sup>+</sup> and K<sup>+</sup> ATPases) could be inhibited by MCs [76,77], leading to a decline in blood Na+ and Cl− concentration and ion exchange imbalance. This could result in swelling of secondary lamellae as well as proliferation of interepithelial chloride cells. Intensive hyperplasia decreases the space between lamellae and causes fusion, which increases the thickness of the water–blood barrier and decreases the oxygen uptake. These lesions can cause capillary hemorrhage and significantly hinder gill functions, such as respiration, ion regulation, acid-base regulation and nitrogenous waste excretion [78]. The molecular actions of cyanotoxins in fish liver and gills involve increased generation of reactive oxygen species (ROS) which randomly attack all cell components, including proteins, lipids and nucleic acids [79,80]. An imbalance between the generation and removal of ROS in these tissues results in oxidative stress and extensive cellular tissue damage in these organs [81,82].

Previous research on fish from Lake Ludoš by the authors showed similar findings. The liver showed a loss of cordlike parenchymal structure; presence of onion-shaped hepatocytes with clear cytoplasm as well as pyknosis and fields of anucleated cells. Changes in the kidney were glomerulopathy with intense dilatation of Bowman's capsule as well as vacuolization of tubules and macrophage infiltration. Furthermore, histopathological alterations observed in the gills showed fusions of lamellae; oedema and epithelial lifting and intensive proliferations of chloride cells. Alterations were also found in intestines where intensive oedematous alteration in the lamina propria; desquamation of enterocytes; and hypertrophies of goblet cells was noted [5]. Although more than five years have passed since the previous research, problems found in fish tissues remain present and could be a result from the constant cyanobacterial blooming in this lake. The aforementioned alterations can severely impact the life quality of fish and consequently disturb the whole food chain.

It should also be noted that fishing at Lake Ludoš by local fisherman is frequent, and therefore *Carassius gibelio* is sometimes included in the human diet. It is necessary to monitor concentrations of cyanotoxins in water and fish to make sure that ingested concentrations of MCs are below tolerable daily intake of 0.04 μg MC-LR equiv./kg body weight/day [83], so that any health consequences could be prevented.

#### *4.3. Potential Health Risks Caused by Cyanobacterial Blooms*

#### 4.3.1. Transfer to Protected Water Birds

Cyanotoxins can be transmitted through the food web to different consumers, such as various animals living in and around the lake, including birds. In the case of Lake Ludoš, this is particularly important since one of the main reasons for protection of this wetland is because it is a habitat for water birds. In addition to the contaminated food, birds can come into contact with cyanotoxins via direct contact with the blooming water. There are few papers dealing with the effect of cyanotoxins on birds. Early reports, such as that from Storm Lake in Iowa associated with *Anabaena flos–aquae* blooms, include estimated deaths of 5–7000 gulls, 560 ducks, 400 coots, 200 pheasants, 50 squirrels, 18 muskrats, 15 dogs, 4 cats, 2 hogs, 2 hawks, 1 skunk, 1 mink, plus "numerous" songbirds. It seems that neurotoxicity resulted in prostration and convulsions preceded death; milder cases displayed restlessness, weakness, dyspnoea and tonic spasms. Furthermore, 57 weak and partially paralyzed mallards were recovered following gastric lavage [84,85]. In Denmark in 1993, two grebes and a coot died during cyanobacterial bloom and *Anabaena lemmermannii* was found in the stomach contents all three birds together with low levels of anatoxin-a(S)-like compounds [86,87]. Death of 20 ducks in Shin–ike pond (Japan) were described after evident *M. aeruginosa* bloom and presence of MC was confirmed, indicating MC intoxication as a possible cause of death [88]. Cyanobacteria-related mortalities have been reported in three flamingo species, both wild and captive [89]. Four MC congeners and ATX found in cyanobacterial mats and stomach contents of dead lesser flamingos as well as faecal pellets collected from shorelines of Lake Bogoria (Kenya) [90]. Similar Lesser Flamingo poisonings have been reported from alkaline lakes in Tanzania, with toxic *Arthrospira fusiformis* implicated [91]. The presence of MCs in several organs was detected in the domestic duck (*Anas platyrhynchos*) and in the black-crowned night heron (*Nycticorax nycticoraxs*) (alongside fish and turtle) from Lake Taihu (China) during toxic *Microcystis* blooms [92]. After experimental exposure to *Microcystis* biomass containing MCs, histopathological changes were observed in the form of cloudy swelling of hepatocytes (shrunken nuclei with ring-like nucleoli, cristolysis within mitochondria and vacuoles with pseudomyelin

structures), vacuolar dystrophy, steatosis, hyperplasia of lymphatic centers and vacuolar degeneration of the testicular germinative epithelium in male Japanese quails (*Coturnix coturnix japonica*) [93].

#### 4.3.2. Transfer to Humans as a Result of Fish Consumption

Humans at the top of the food chain may also be endangered. Lake Ludoš is regularly frequented by the local fishermen, and year-long and day-long fishing permits are available for this protected freshwater ecosystem. Based on the available data it was found that fishermen with prolonged and cumulative exposure through contaminated water (direct contact, inhalation, oral) and food may suffer a type of poisoning with a long-term complications. Such an outcome was demonstrated through biochemical alterations of liver damage biomarkers in fishermen and children that consumed (in addition to water) ducks, fish, shrimps, snails, and other aquatic organisms grown in blooming lakes in China [94,95]. Serum analysis of 35 fishermen who worked and lived on fishing ships on the blooming Lake Chaohu (China) for over 5 years, drank the lake water and ate fish, shrimps, and snails, showed the presence of MC [94]. The values of serum enzymes were significantly higher in the children that were exposed to MCs for over 5 years through drinking MC contaminated water and food e.g., carp and duck in the Three Gorges Reservoir Region in China. Additionally, 0.9% of parents of high-exposed children self-reported a cancer diagnosis (9 of 994, including four with hepatic carcinoma) compared to 0.5% of parents of low-exposed children (1 of 183), and none of the parents of children in the unexposed group [94]. To confirm this, in south-west China where MC-LR was detected in water, fish, and ducks, people with abnormal indicators of renal function had a much higher mean level of MC-LR exposure than those with normal indicators [96]. The foregoing indicates that the blooming phenomenon should not be taken lightly, but rather should receive more scientific attention.

#### *4.4. A Potential Resolution?*

By some estimates, we lost 50% of the world's wetlands in the 20th century [97]. So far an effective and long-term solution to reduce cyanobacterial blooms worldwide has not been attained. Cyanobacterial blooms have been recorded in the wetlands of the Perth region (Australia), with *Microcystis aeruginosa* and *M. flos-aquae* being the most ubiquitous bloom-forming cyanobacteria. Furthermore, for the first time *Nodularia* blooms have been recorded in such low salinity waters in Australia, and hepatotoxins, microcystin and nodularin, were associated with the analysed blooms [98]. Lake Ludoš has been notoriously known for consistent cyanobacterial blooming. Wetland ecosystems need help in dealing with this issue.

Mitigation of the global expansion of cyanobacterial harmful blooms, together with a variety of traditional (e.g., nutrient load reduction) and experimental (e.g., artificial mixing and flushing, omnivorous fish removal) approaches has been presented in a review by Pearl et al. [99]. Virtually all mitigation strategies are influenced by climate change, which may require setting new nutrient input reduction targets and establishing nutrient-bloom thresholds for impacted waters. Physical-forcing mitigation techniques, such as flushing and artificial mixing, will need adjustments to deal with the ramifications of climate change. Current mitigation strategies need to be examined and the potential options for adapting and optimizing them in a world facing increasing human population pressure and climate change [99]. A notable approach on large-scale wetland restoration has been proposed in Ohio (USA). Through restoration of the Great Black Swamp, cyanobacterial blooming in Lake Erie should be mitigated [97]. The goal would be to restore rare and declining plant communities and species and to provide nutrient load reduction to Lake Erie, specifically total phosphorous. With the creation of the 20,000 of wetlands in "hot spots" of the former Great Black Swamp, removal of 18% of the 2617 metric tons/year of phosphorus loading by the Maumee River to Lake Erie would be expected [100]. This would lead to a cleaner lake and a more sustainable landscape in that region [97].

In order to reduce cyanobacterial population in Lake Ludoš, the potential efficiency of hydrogen peroxide treatment was examined in vitro. Although further research is needed, the initial laboratory results showed that this method may not be readily applicable, since the dense cyanobacterial

population and the high load of organic matter (that consumes hydrogen peroxide) would require the use of harmfully high doses of hydrogen peroxide in order to fight the cyanobacteria [5]. Another recent study performed in Lake Ludoš demonstrated the use of H2O2 and the MC-degrading capacity of the enzyme MlrA. Results showed that the treatment decreased the abundance of the dominant cyanobacterial taxa and reduction of the intracellular concentration of MC by H2O2, but the reduction of the extracellular MC was not accomplished in combination with MlrA. Since H2O2 was found to induce the expression of *mcyB* and *mcyE* genes involved in MC biosynthesis, the use of H2O2 as a safe cyanobacteriocide still requires further investigation [101]. Several other authors have also suggested measures for water-quality improvement of this lake, including the identification of source water with lower nutrient content for maintaining the volume of the lake, sediment removal [102,103], and sediment phytoremediation [104]. Even if restoration of the endangered ecosystem were to be realized, continuous monitoring would still be preferable. Such system (South Florida Wetland Monitoring Network (SFWMN)) has been created in the Everglades (USA), with three real-time hydrologic, water-quality, and meteorological field stations. Besides research, data from these monitoring stations assist in a better understanding of wetlands dynamics and function [105].

Lake Ludoš, a significant ecosystem and a habitat for several endemic and relict plant species, is preserved at the moment, however, the problematic ecological state of the lake is also perpetuated, so justification for such action was questioned. Current research strongly supports the earlier findings that the ecological balance of this Ramsar site is impaired. By preserving the lake and its cyanobacterial problem the water birds and their habitat are not really protected and, quite the contrary, the lake and nearby ecosystems are put at risk. There is a likely probability that the birds visiting the contaminated lake during their migrations carry viable cyanobacterial cells on their feet, feathers, bills, gullets, and faecal material, thus contributing to the spreading of cyanobacteria and hence expanding the problem [5].

The future of Lake Ludoš is still unclear and a potential resolution is still not in sight. How long will this vital ecosystem stay in a state of limbo? Based on this research, and as an extension and confirmation of the previous investigations, it is possible to predict the continuation of cyanobacterial blooms in Lake Ludoš, degradation of protected habitat and negative effects on aquatic organisms. Food items derived from Lake Ludoš also present one path of human exposure to cyanotoxins, which together with the direct contact and ingestion of contaminated water, as well as inhalation, signifies a health risk. Therefore, it is necessary to continue the monitoring of this lake, and work on finding an effective treatment that will help this ecosystem, but also many others that suffer from the same problem. Recently, it was published that there are over 1000 recorded identifications of major cyanotoxins in more than 800 aquatic ecosystems from over 60 countries worldwide [4].

It is time to solve this problem but most if not all the options being considered are limited and/or inconsequential [97]. This paper should be seen as an invitation to scientists, engineers, competent authorities, policy makers and anyone else who can contribute to solving the problems of Lake Ludoš and finally ending the lake's perpetual state of limbo. Potential solutions should be comprehensive and holistic, and lead to a sustainable management of this marshland ecosystem and its services.

#### **5. Conclusions**

This investigation regarding the presence of cyanobacteria and cyanotoxins, together with the observations of effects on aquatic organisms, in Lake Ludoš during 2018 has resulted in the following findings:


Additional emphasis has to be placed on the detection of MC and STX coding genes, which was performed for the first time in Lake Ludoš indicating these toxins as the main threat, as well as identification of demetylated forms of well-known toxins (MC-dmLR, MC-dmRR), and a more rare variant MC-LF in the water. Although present MC variants have a different toxicity, nonetheless, they contribute to possible adverse effects.

The results presented indicate that the potential threat to many organisms in the ecosystem, including birds and humans, is real and present. The persistent alarming condition of Lake Ludoš poses a great health risk and a "ticking (cyanobacterial) bomb" that could lead to the collapse of this special nature reserve. Urgent remediation measures are needed to alleviate the incessant cyanobacterial problem in Lake Ludoš and to break this ecosystem out of the perpetual state of limbo in which it has been trapped for quite some time.

**Author Contributions:** Funding acquisition, Z.S., J.M. and J.L.; conceptualization and design of the experiments, N.T., D.D.B., J.L. and Z.M.; investigation and field sampling, B.M. and I.Š.; phytoplankton analyses, S.S. and N.Đ.; histopatological analyses, N.K., J.L. and Z.M.; cyanotoxin analysis, N.T., D.D.B., T.D. and J.M.; cyanotoxin coding genes analysis, T.D. and H.S.; writing—original draft preparation, N.T.; writing—review and editing, N.T., Z.S. and J.M.; project administration, N.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Ministry of Education, Science and Technological Development of the Serbian Government (project number: ON-176020, TR-31011, III-43002), Bilateral project Hungary-Serbia Invasive and blooming cyanobacteria in Serbian and Hungarian waters (451-03-02294/2015-09/3) and the Erasmus+ programme of the European Union (agreement number: 2017-1-FI01-KA107-034440).

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

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
