Seasonal phytoplankton and geochemical shifts in the subsurface chlorophyll maximum layer of a dimictic ferruginous lake

Abstract Subsurface chlorophyll maxima layers (SCML) are ubiquitous features of stratified aquatic systems. Availability of the micronutrient iron is known to influence marine SCML, but iron has not been explored in detail as a factor in the development of freshwater SCML. This study investigates the relationship between dissolved iron and the SCML within the dimictic, ferruginous lake Grosses Heiliges Meer in northern Germany. The occurrence of the SCML under nonferruginous conditions in the spring and ferruginous conditions in the fall are context to explore temporal changes in the phytoplankton community and indicators of primary productivity. Results indicate that despite more abundant chlorophyll in the spring, the SCML sits below a likely primary productivity maximum within the epilimnion, inferred based on colocated dissolved oxygen, δ13CDIC, and pH maxima. The peak amount of chlorophyll in the SCML is lower in the fall than in the spring, but in the fall the SCML is colocated with elevated dissolved iron concentrations and a local δ13CDIC maximum. Cyanobacteria and Chlorophyta have elevated abundances within the SCML in the fall. Further investigation of the relationship of iron to primary productivity within ferruginous SCML may help to understand the environmental controls on primary productivity in past ferruginous oceans.

the SCML can also be a dynamic motility response to diel fluctuations in light (Gervais, 1997), and its formation can be influenced by zooplankton grazing (Moeller et al., 2019). SCML can also be a photoacclimation response of phytoplankton where the amount of chlorophyll is enhanced without a change in biomass density. This usually occurs in oligotrophic waters where nutrients are most available at the base of the photic zone (Fennel & Boss, 2003).
Nutrient availability within an SCML is generally discussed in terms of nitrogen or phosphorus (i.e., N or P). But iron (Fe) is one of the most abundant metals in phytoplankton biomass (Ho et al., 2003), indicating that low environmental availability could also limit phytoplankton growth and/or primary productivity. It is a ubiquitous redox carrier in cytochromes, ferredoxin, and Fe-S proteins involved in electron transport during photosynthesis (Twining & Baines, 2013).
Iron is the limiting nutrient in the high nitrate low chlorophyll (HNLC) regions of the ocean (Boyd et al., 2000), and a (co)-limiting nutrient in some lakes (Havens et al., 2011;North et al., 2007;Vrede & Tranvik, 2006). Iron limitation has been described within marine SCML (Hogle et al., 2018), but iron bioavailability is generally not considered in investigations of SCML in lakes. This may be because iron is not frequently measured in lakes as it is not often found to be a limiting nutrient in lakes .
In oxygenated waters, the thermodynamically stable form of iron is Fe 3+ (ferric), which is poorly soluble at circumneutral pH and rapidly precipitates as Fe 3+ (oxyhydr)oxides. Removal of this mineralized iron to the sediments results in low iron bioavailability in the photic zone.
While eukaryotic phytoplankton generally requires unchelated Fe 3+ , some cyanobacteria produce siderophores, which chelate Fe 3+ and facilitate acquisition (Morrissey & Bowler, 2012). Due to its requirement as a nutrient by phytoplankton, the presence of ligand-bound forms, and scavenging into a particle, iron generally displays a hybrid-type element profile in marine systems (Bruland & Lohan, 2003). While many of the same processes control iron distribution in lakes, iron concentrations are generally higher in lakes than in oceans due to smaller basins and shorter water residence times that allow the iron to achieve higher concentrations. In lakes where thermal stratification gives rise to anoxic hypolimnia, sedimenting Fe 3+ (oxyhydr)oxides can be reductively dissolved to form the much more soluble Fe 2+ species. "Ferruginous" lakes are those with abundant dissolved iron, predominantly as Fe 2+ , in the hypolimnion enabled by seasonal or permanent stratification, which can result in a redoxcline between dissolved iron and oxygen.
Although ferruginous conditions are generally not found in modern marine basins, they were a common feature of the deepwater of Precambrian oceans (Poulton & Canfield, 2011;Swanner et al., 2020).
SCML are common features of ferruginous lakes (Baker & Brook, 1971;Boehrer et al., 2017;Savvichev et al., 2017;Swanner et al., 2020), and can overlap with the dissolved iron-oxygen redoxcline. The dominant phytoplankton within ferruginous SCML can be cyanobacteria (Baker & Brook, 1971;Savvichev et al., 2017), or diatoms (Swain, 1984), but this has also not been determined for many ferruginous lakes (Boehrer et al., 2017). Nevertheless, the abundance of dissolved iron and its predominantly ferrous form within the anoxic conditions of the hypolimnion has potential implications for phytoplankton composition and primary productivity in lakes. For instance, Cyanobacteria growth in an SCML overlapping a redoxcline in a ferruginous lake was linked to enhanced Fe 2+ availability (Dillon & Molot, 2005;Molot et al., 2014). If enhanced availability of Fe 2+ is associated with elevated biomass or primary productivity, it could have implications for estimates of primary productivity from lacustrine systems with anoxic hypolimnia. Lake hypoxia is potentially expanding due to land-use changes and climate change (Jenny et al., 2016;North et al., 2014). As ferruginous conditions may be common in lake-rich areas in northern boreal and deglaciated regions (Schiff et al., 2017;Swanner et al., 2020), the role of hypolimnetic dissolved iron on aquatic primary productivity in these regions is worth investigating.
The questions we address are which phytoplankton predominate within a ferruginous SCML, if their presence is related to the abundance of dissolved iron, and whether a ferruginous SCML is a locus of biomass and/or primary productivity. In this study, we pose these questions through an investigation of the geochemistry and phytoplankton community composition of a seasonally ferruginous dimictic lake with a persistent SCML, the Grosses Heiliges Meer (GHM). The SCML occurs within nonferruginous conditions in the spring but is colocated to the dissolved iron-oxygen redoxcline under ferruginous conditions that develop in the fall, allowing for an investigation of the potential effects of dissolved iron on the SCML.

| Study site
The GHM is located near Hopsten, Germany at 52°21′06.73″ N and 7°38′03.22″ E (Figure 1). It has a surface area of 7.9 ha, a maximum depth of 10.8 m, and an average depth of 4.4 m based on a survey in 1993 by the Institut für Angewandte Ökologie und Gewässerkunde, Niederzier, Germany. It becomes thermally stratified from about mid-April to late October. The GHM is part of an established field station in a protected area of the Landschaftsverband Westfalen-Lippe (LWL) Museum für Naturkunde (Museum of Natural History). The GHM is one of several sinkhole lakes in the area, whose water chemistry varies as a function of the age of the lake, due to interaction with the groundwater of varying chemical properties (Pott, 2009). There are no natural surface water inlets or outflows, although until 1968 an industrial drainage ditch, the Meerbecke, flowed through the lake. This drainage has since been relocated to flow next to the lake. Iron in the lake is presumed to have been supplied directly by the Meerbecke in the past, or through the current Meerbecke-groundwater system, as groundwater iron concentrations are highest in wells lying between the lake and the Meerbecke (Pust, 1993).
Samples were taken from a rowboat at a fixed mooring in the middle of the lake, with a water depth of approximately 10 m. There exists a record of physicochemical profiles that extend to when the field station was established in the 1930s (Chmiel, 2010;Terlutter, 2009;  | 3 of 20 data are presented as the average of triplicates with two standard deviations. To remove background fluorescence attributed to dissolved organic carbon rather than photosynthetic pigments, water was filtered through 0.22 μm polyethersulfone (PES) filters, analyzed, and the results were subtracted from the raw water signals (i.e., a Z off correction).
Samples were agitated continuously with a magnetic stirrer during analysis. The Phyto-PAM II has a detection limit of 0.1 μg L −1 chlorophyll.
Taxon-specific maximum quantum yields were either zero or had very high errors among triplicate measurements when chlorophyll concentrations for that taxon were <1 μg L −1 . Therefore, maximum quantum yields were not reported when the corresponding taxon-specific chlorophyll values for that sample were <1 μg L −1 . The significance of maximum quantum yields between depths was assessed with a non-parametric Kruskal-Wallis test, as not all taxa produced maximum quantum yields at all depths.
Similar measurements were made in September 2014 and May 2015 using a Water-PAM (Heinz Walz GmbH). However, this instrument was only sensitive to fluorescence from Cyanobacteria.
Also, measurements were made on samples returned to the laboratory more than 24 h after collection, and so are not appropriate to assess maximum quantum yields. In September 2018, we had the opportunity to make measurements on-site for multiple taxa using the Phyto-PAM, justifying the additional sampling campaign.

| Most probable number (MPN)
MPN incubations were carried out in September 2014 and May 2015 to quantify active Cyanobacteria. The major salts of "Holy Medium" were modified from BG11 (Stanier et al., 1971) to better represent the major element chemistry of the GHM. The 100× stock solution of macronutrients contained 1 g NaNO 3 , 1.25 g MgSO 4 × 7H 2 O, 2 g CaCl 2 × 2H 2 O, and 0.05 g Na 2 -EDTA in 500 ml of ultrapure water. No organic carbon sources were added to prevent heterotrophic growth, and cycloheximide was added at a concentration of 100 μg ml −1 to inhibit eukaryotic growth (Ferris & Hirsch, 1991). Dilution series were made in "Holy Medium" nongrowth buffer, which lacked trace metals, K 2 HPO 4 , ferric ammonium citrate, Na 2 CO 3 , and vitamins. For the MPN plates, 1% Noble agar was added to liquid media before autoclaving. Each well contained 0.9 ml of medium and 0.1 ml of diluted sample. Plates were incubated for 6-8 weeks at 20°C (depths of 0-5.5 m) and 10°C (depths 6-10 m). Wells were scored as positive growth based on turbidity, which was in most cases accompanied by visible green coloration. MPN and confidence intervals were calculated using MPNcalc v1.2.0 (https://mpncalc.galaxytrakr.org/). Between 75 and 100 ml was filtered onto 0.22 μm filters per depth.
DNA was extracted with a DNEasy Powersoil DNA Isolation kit. The Powersoil kit is not generally applied to algae but it is effective for prokaryotic and eukaryotic DNA from aquaculture samples (Pearman et al., 2020). DNA was quantified with a Nanodrop. In September 2018, water was collected from every meter depth, plus 5.8 and 6.2 m, bracketing the 6 m chlorophyll maximum detected with the in situ probe. Water was filtered in series through 11 µm nylon and 0.22 μm PES filters to capture eukaryotic algae and bacteria (and archaea), respectively. DNA was extracted from filters using a DNeasy PowerBiofilm kit (Mäki et al., 2017), and DNA was quantified with a Qubit.
The 16S and 18S rRNA gene data were processed using the standard protocol in Mothur (version 1.39; Schloss et al., 2009). Out of 990,084 total 16S rRNA gene sequences, 745,127 sequences were retained after they were checked for quality (quality score greater than 25, error <1%, ambiguities [N] removed) and assembled.
Any 16S rRNA gene sequences less than 299 bp or greater than 375 bp were removed and the sequences were aligned to the V4 region of the SILVA database (version 138; Quast et al., 2013).
Overhangs were removed to ensure sequences overlapped in the same region. The final alignment was determined to be 658 columns wide. Next, chimeras were checked with VSEARCH version 2.13.3 (Rognes et al., 2016), and then preclustered up to two nucleotide differences between sequences. As such, 115,696 unique sequences were identified. Representative operational taxonomic units (OTU) for 16S rRNA gene sequences were classified (Bayesian classifier) to the SILVA database and clustered at 97% similarity with an 80% confidence threshold and rarefied to the lowest sequencing depth.
The OTUs assigned to the phylum Cyanobacteria were blasted (NCBI blastn) to validate taxonomic assignment.
Similarly, 18S rRNA gene sequences were subjected to a processing method much like the 16S rRNA gene sequences, where 915,958 sequences were recovered from 1,067,170 total sequences after quality checks were applied. 18S rRNA gene sequences with less than 250 bp or more than 375 bp in length were removed.
Sequences were aligned to the V9 region of interest (created by aligning Saccharomyces cerevisiae, downloaded from NCBI) to the SILVA database. The final alignment was determined to be 816 columns wide after overhangs and chimeras were removed (by VSEARCH) and this resulted in 35,231 unique sequences identified.
The 18S rRNA gene sequences were taxonomically classified using a Bayesian classifier with the PR 2 database (version 4.72; Guillou et al., 2013) and OTUs were clustered at 97% similarity and 80% confidence threshold. Two samples contained only a few hundred sequences; these were removed before rarefaction in Phyloseq (version 1.36.0; McMurdie & Holmes, 2013). All sequencing data was visualized using Phyloseq.

| Calculations and statistical analysis
Secchi disk depths were converted to K d (m −1 ), the light attenuation coefficient of photosynthetically active radiation (PAR), by dividing the depth by 1.7 m. The depth of 1% of surface PAR (Z 1% ) was estimated by dividing ln(100) by K d (Idso & Gilbert, 1974;Poole & Atkins, 1929). The depth of the thermocline was calculated using rLakeAnalyzer (Winslow et al., 2019). Dissolved oxygen saturation was determined from measured dissolved oxygen, water temperature, and specific conductance values and published partition coefficients for oxygen in equilibrium with the atmosphere (Benson & Krause, 1980, 1984. Data analysis and graphing utilized R v. 3.5.3 and the package tidyverse (Wickham, 2017). To determine for relationships between environmental variables to cyanobacterial 16S rRNA gene orders and photosynthetic eukaryotic 18S rRNA gene classes, Spearman correlation was performed using R packages hmisc (Harrell, 2020) and corrplot (Wei & Simko, 2017). Only environmental variables that had data for all depths were assessed. For this analysis, 16S and 18S rRNA gene abundances were normalized to the maximum abundances of that taxon from all depths at the date of sampling to ensure a normal distribution of values, and physicochemical data were normalized so that values ranged from 0 to 1.  Table 1). The euphotic depth (1% PAR) calculated from the Secchi disk readings was always below the thermocline and was shallowest in September 2014 (8.4 m) compared to May 2015 and September 2018 (9.2 m; Table 1). Chemical stratification, as evidenced by a gradient in specific conductance within the water column is most pronounced in fall, as both September dates have sharper chemoclines than May 2015, and the hypolimnion has a higher maximum specific conductance ( Figure 2). Chemical changes throughout the water column are shown in Figure 3. The DIC profiles follow closely the specific conductance profiles. The magnitude of DIC was in the mM-range, indicating that DIC is likely a major chemical constituent contributing to the increase in specific conductance with depth.

| Seasonal chemical and biological trends
Other major ions are likely to be Na + , Ca 2+ , K + , Cl − , and SO 4 2− , which are all abundant in local groundwater (Pust, 1993). During late summer stagnation, dissolved iron is typically present in the deep water column up to 5 m below the surface (Terlutter, 2009) (Table A2), suggesting it is a persistent feature during summer stratification that is likely to be laterally extensive. The term SCML is therefore applied. The was 9.2 m in September 2018 (Table 1). The SCML is centered at a depth below where optimal light conditions exist (Figure 4), 1-2 m below the Secchi disk depth but well within the calculated photic zone. It is, therefore, possible that Chlorophyta in the SCML have increased chlorophyll to adapt to low light. There is also a possibility that the Chlorophyta living below the optimal photic depth could utilize anaerobic or fermentative metabolisms within the anoxic SCML, such as are known from sequenced Chlorophyta genomes (Atteia et al., 2013). Such a lifestyle was suggested for Chlorophyta 18S rRNA gene sequences recovered from the aphotic monimolimnion of meromictic and ferruginous Lake Pavin (Lepère et al., 2016). Oxygenic cyanobacterial 16S rRNA gene sequences were detectable at all depths sampled ( Figure 6). Sequences represented the following orders, with the average 16S rRNA gene copy number and standard deviation of that order in parenthesis, where known (Table A3) Only those correlations that were significant (i.e., p < 0.05) are denoted with colored circles. Blue is a positive correlation and red is a negative correlation. rRNA, ribosomal RNA.

| Relationship of iron to organisms
The seasonal variation in iron abundance within the SCML is useful to pose the question of whether the abundance of dissolved iron has any effect on the identity of the phytoplankton within the ferruginous SCML. A limitation of this system for answering the question of whether iron abundance influenced the phytoplankton community is that other nutrients and physicochemical conditions also vary seasonally.

| Seasonal indicators of primary productivity
Another hypothesis regarding the origin of SCMLs is that they are biomass maxima that result from optimal conditions for primary productivity. Optimal conditions could include light, temperature, or nutrient availability. If the SCML were a primary productivity maximum, a prediction would be that indicators of primary productivity should co-occur with the chlorophyll maximum. Neither gross nor net primary productivity was measured in this study, but several chemical species were assessed that are expected to vary because of photosynthesis and/or carbon fixation. These include oxygen, pH, and DIC/δ 13 C DIC. In addition, the maximum quantum yield determined on samples in September 2018 indicates a capacity for photosynthesis and carbon fixation.
Oxygen can be introduced to the epilimnion from the atmosphere or produced in situ via oxygenic photosynthesis. If dissolved oxygen exceeds the value predicted to be due to air saturation, it indicates production that is outpacing consumption due to respiration. In May 2015, the lake had a deep oxygen maximum centered around 4 m (145% air saturation; Figure 3), and dissolved oxygen was above air saturation from 0 to 5.1 m (SCML ; Table 1). Dissolved oxygen and pH were correlated throughout the water column, and elevated pH also occurred at 4 m depth in May 2015 (Figure 7). Elevated pH can result during the growth of phytoplankton due to uptake of bicarbonate (Schultze-Lam et al., 1997), and is an indicator of carbon fixation in phytoplankton.
Trends in DIC and δ 13 C DIC in aquatic systems can reflect carbon fixation, but they also incorporate signals from processes such as respiration and chemical equilibration (i.e., CO 2 dissolution/exsolution, (bi) carbonate (de)protonation, and mineral precipitation). In the GHM, DIC concentrations increased below the oxycline, as might be expected for photosynthetic carbon fixation in the euphotic zone and remineralization of DIC from exported organic carbon (Figure 3). In May 2015, a DIC minimum occurred at 5 m. The CO 2 concentrations determined by titration had a minimum at 4 m in May 2015 . In general, the δ 13 C DIC was consistently lighter in the bottom 4 m of the GHM at all sampling times (−12 to −16‰), and heavier in the upper 4 m of the water column (−2 to −5‰). In May 2015 there was a δ 13 C DIC excursion at 4 m to the heaviest values observed on that date (Figure 4).
The shift to heavier δ 13 C DIC from surrounding antecedent conditions can result from biological carbon fixation, which preferentially consumes 12 C and leaves the residual DIC pool enriched in 13 C. DIC and δ 13 C DIC were anticorrelated, while δ 13 C DIC was correlated with pH, and dissolved oxygen (Figure 7). Chemical interactions or dependencies of these environmental variables are not expected, other than through photosynthetic carbon fixation. Although particulate organic carbon (POC) quantities and δ 13 C POC data (which were not collected) would be necessary to quantify primary productivity, the chemical data are consistent with elevated photosynthetic primary productivity at 4 m in May 2015, above the SCML at 5 m.
Dissolved oxygen was never above air saturation in September 2018 but exceeded air saturation from 0 to 3 m in September 2014 (Table 1) (Lambrecht et al., 2018;Michiels et al., 2017), and are an expected product of anaerobic remineralization.
Past lake studies have observed that Cyanobacteria dominate lakes later in the season when ammonium is more available (Glibert et al., 2016). The  (Magnusson, 1997). Due to the absence or low abundance of Chlorophyta chlorophyll elsewhere in the water column, no other yield measurements were recorded for the taxon. This high maximum quantum yield may indicate that nutrient and/or light availability was optimal for Chlorophyta within the SCML (Rattan et al., 2012).

| Insights into phytoplankton dynamics in ferruginous oceans
Ferruginous lakes such as the GHM have become important analogs to past ferruginous oceans because the modern oceans are generally oxygenated and have extremely low concentrations of iron. While the physical factors driving mixing are very different between lakes and oceans, some of the chemical and biological features of modern ferruginous lakes, such as their lower sulfate content, can approximate aspects of past ferruginous oceans (Swanner et al., 2020). This study provides data on the distribution of phytoplankton and indicators of primary productivity and gives some insight into what factors might be important to the relationship between the SCML and primary production in ferruginous conditions.
In this study, indicators of biomass and productivity maxima were spatially separated from the SCML under non-ferruginous conditions, but under ferruginous conditions, these coincided with the SCML.
The phytoplankton with the highest iron demands, Cyanobacteria and Chlorophyta, had elevated abundance within the SCML, suggesting a ferruginous SCML could alleviate iron demands, although this was not supported by simple correlations, likely because reverse gradients of other required resources, such as light.
It has also been proposed oxygen production would have been limited by the toxicity of reactive oxygen species (ROS) produced in the presence of Fe 2+ in ferruginous oceans (Swanner et al., 2015).
However, experiments investigating this phenomenon are generally carried out in monoculture under high-light conditions (Herrmann et al., 2021). It is possible that the low-light conditions likely for an SCML could help to diminish the production of ROS. Another possibility is that ROS-mitigating bacteria within a redoxcline also help to decrease oxidative stress (Szeinbaum et al., 2021). Exploring these questions within ferruginous lakes should provide snapshots into how photoautotrophic processes are regulated by the different phytoplankton and microbial populations, as well as chemical and light conditions across a range of ferruginous lakes.
Much of the investigation into the ecology of primary producers in ferruginous lakes as analogs to past ferruginous oceans has focused on anoxygenic photosynthetic bacteria, particularly those that can use Fe 2+ as a photosynthetic electron donor (Crowe et al., 2008;Llirós et al., 2015;Walter et al., 2014). This type of primary productivity is commonly invoked for past ferruginous oceans (Johnston et al., 2009;Kappler et al., 2005). Despite the potential for Fe 2+ -based anoxygenic photosynthesis to fix significant carbon within portions of modern ferruginous lakes (Morana et al., 2016), there is generally still abundant oxygenic phytoplankton within these systems if the light is available (Savvichev et al., 2017;Swanner et al., 2020). Thus, oxygenic photosynthesis could still be a major photoautotrophic pathway within ferruginous water columns past and present, especially as anoxygenic photosynthetic bacteria can be confined to discrete layers dictated by electron donor and light availability. Significant chlorophyll, turbidity, and/or biomass maxima attributable to phytoplankton occur in ferruginous lakes, sometimes corresponding to the depth of the redoxcline (Baker & Brook, 1971;Swanner et al., 2020). Elucidating whether these layers are hot spots for primary productivity, whether oxygenic photosynthesis predominates over anoxygenic photosynthesis in these layers, and what environmental or ecological factors allow anoxygenic photosynthesis to flourish should be goals of future research into the ecology of ferruginous systems.

| CONCLUSIONS
The SCML is a seasonally persistent feature of the GHM, a dimictic ferruginous lake. However, the biological and geochemical characteristics of the SCML vary strongly between the spring and fall seasons.
In the spring, the SCML does not appear to be a zone of enhanced primary productivity. Several indicators of enhanced primary productivity, including dissolved oxygen, pH, and a heavier δ 13 C DIC excursion occur at a shallower depth than the SCML. In the fall, the magnitude of the SCML diminishes, but it co-occurs with a redoxcline between dissolved iron and oxygen. There are enhanced cultivatable Cyanobacteria that the SCML in the fall when dissolved iron was elevated. There is also a slight positive δ 13 C DIC excursion within the SCML, and the highest maximum quantum yields  (Cullen, 2015), investigating the relationship between dissolved iron and phytoplankton and photosynthesis in a range of ferruginous systems can help to identify the environmental triggers for the development and maintenance of a ferruginous SCML, as well as factors influencing primary productivity and oxygen production.
With further study, such insights might provide context for the locus and mode of primary productivity in past ferruginous oceans.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT
All sequencing data are available in the NCBI repository under

APPENDIX
The multi-wavelength fluorescence observations documented the SCML, although the magnitude and depth of the maximum were systematically lower than from the in situ fluorometer ( Figure A1).
Light exposure upon collection may have quenched fluorescence in the samples analyzed by the multi-wavelength fluorometer. These samples were dark acclimated for at least 20 min before measurement which should have minimized this effect (Huot & Babin, 2010).
Fluorescence from organic matter might also produce higher chlorophyll a reading in the in situ fluorometer. However, the Z off correction applied by analyzing filtered samples in the multiwavelength fluorometer was smaller than the variation between the two analyses, suggesting this interference had a minor effect. The differences in chlorophyll values most likely arise from the different calibrations between the two fluorometers. No attempt was made to validate the calibrations here, that is, with chlorophyll extractions.
The in situ fluorometer is expected to be internally consistent and is used to compare trends in absolute quantities of chlorophyll from different sampling times and depths, while the multi-wavelength fluorometer data are used to make comparisons between sampling depths in September 2018 only ( Figure A2).
These data show the development of seasonal stratification (i.e., dimixis) and the persistence of an SCML in the GHM from May to September. BDL = below detection limit; ND = not determined.  The copy number of the 16S rRNA gene was retrieved from all available cyanobacterial genomes in NCBI and is summarized by order in Table A3.
Lake water for fluorescent cell counts was preserved with 2.5% paraformaldehyde and stored at 4°C. Samples were washed, sonicated, iron minerals were digested if necessary, and stained with Sytox according to procedures described in Wu et al. (2014).