Role of trace metal co-limitation in cyanobacterial blooms of Maumee Bay (Lake Erie) and Green Bay (Lake Michigan)

Abstract The open waters of large lakes can sometimes become so depleted in important metals that phytoplankton communities become either growth limited or limited in some metabolic function. Metals such as Fe, Ni, Mo, and Zn are used as co-factors for enzymes by phytoplankton in core metabolic functions, as well as metabolic pathways that allow phytoplankton to use less preferred forms of N and P (e.g. nitrates, urea, and organic phosphorus). In the Laurentian Great Lakes, metal limitation has been observed primarily in waters that are isolated from tributary inputs and sediment exchange. These are situations where the supply of metals is very low relative to demand. We hypothesized that another situation where metal limitation could occur is within algal blooms, where the demand for metals is high because preferred forms of N and P are often low or absent and the phytoplankton biomass is extremely high. As a preliminary test of this hypothesis, we performed seven laboratory incubation experiments on naturally occurring phytoplankton communities from two nearshore habitats that frequently experience blooms (Green Bay in Lake Michigan and Maumee Bay in Lake Erie). Metals and labile nutrients (inorganic N and P) were often present at low concentrations or below the method detection limit. Amendments of inorganic N (5 experiments) and P (1 one experiment) resulted in increased chlorophyll in laboratory incubations, but metal amendments alone never appeared to stimulate growth. Although we attempted to sample during conditions when we hypothesized metal limitation would be most likely, we cannot rule out the possibility that metal limitation is occurring at other times in these eutrophic nearshore areas. Further, metal availability could affect other aspects of the phytoplankton community, such as the production of cyanotoxins or the interactions between different phytoplankton taxa. KEY POLICY HIGHLIGHTS Although harmful algal blooms are more likely to develop when external nutrient loads are high, while a bloom is occurring, the concentrations of key nutrients (nitrogen and phosphorus) can be very low. When key nutrients are in low abundance or absent, algae require additional metals to take advantage of poorer quality versions of these key nutrients. We hypothesized that metals might become growth-limiting in blooms, which might in turn have implications for predicting bloom toxicity. Seven experiments on phytoplankton from nearshore Great Lakes areas experiencing algal blooms did not show evidence for growth-limitation of algae, although the spatial and temporal extent of these experiments is limited.


Introduction
The external inputs of nitrogen (N) and phosphorus (P) have been identified as major drivers of increasingly frequent and intense eutrophication in lakes (Howarth 2008, Heisler et al. 2008, Stumpf et al. 2016, Newell et al. 2019).As a result, the abundance, distribution, form, and dynamics of N and P attract considerable research attention, whereas the role of other elements is sometimes less well understood (Oliver et al. 2012, Schindler 2012, Paerl et al. 2016).Beyond N and P, many other elements are required for growth and metabolism in phytoplankton (Kaspari andPowers 2016, Facey et al. 2019).Often metals are needed at active sites of enzymes that facilitate uptake of certain forms of N and P (McKay et al. 2001, Twiss et al. 2005, Merchant et al. 2006, North et al. 2007).This creates a situation where metals can become co-limiting with N and P (North et al. 2007).For example, cyanobacteria prefer reduced inorganic N (i.e.NH 4 +), but when only nitrate, urea, or N 2 gas is available, cyanobacteria require iron (Fe), nickel (Ni), and molybdenum (Mo) to acquire and use these forms of N (Rees andBekheet 1982, Glass et al. 2012).If the appropriate metals are unavailable, then the cyanobacteria could be considered N limited (i.e.adding NH 4 + would increase growth) or limited by the relevant metal cofactor, since it would allow the cyanobacteria to use those other N forms.Using the co-limitation terminology developed by Harpole et al. (2011), this would be considered independent co-limitation.
Within the Laurentian Great Lakes, several studies have explored the spatial and temporal scope of metal limitation to naturally occurring phytoplankton communities.Sterner et al. (2004) performed experiments with several trace metals (Fe, zinc [Zn], manganese [Mn]) on phytoplankton and concluded that Lake Superior taxa were P limited, but alleviating P limitation even slightly induced Fe limitation.Twiss et al. (2005) performed 13 experiments on Lake Erie phytoplankton, and reviewed many more, and concluded Lake Erie phytoplankton are infrequently limited by trace metal bioavailability.North et al. (2007), working in Lake Erie, found support for independent co-limitation as described above, whereas Fe limited the ability of the phytoplankton to take up NO 3 -.Other studies in Lake Erie and Lake Superior have also demonstrated Fe limitation of NO 3 -uptake or phytoplankton growth (Twiss et al. 2000, Ivanikova et al. 2007, Havens et al. 2012).Collectively, these studies found metal limitation was most likely in offshore, stratified waters, and the presumption is that this was because surface waters in these areas were isolated from both sediment and tributary inputs that resupply metals (North et al. 2007).
Another area where metal limitation might become important is within harmful algal blooms (HABs).Evidence shows that bioavailable forms of iron and other metals can become depleted within blooms, indicating they are important for sustaining the biomass of blooms (Facey et al. 2019, Leung et al. 2021).Metal demand per cell will increase when labile N (NH 4 ) and P (measured as soluble reactive phosphorus; SRP) are absent, but other forms of N and P are present.We hypothesize that this high demand per cell coupled with the very dense accumulations of cells could cause metals to become co-limiting with N or P. To test these predictions, we collected naturally occurring phytoplankton communities from Green Bay (Lake Michigan, Wisconsin) and Maumee Bay (Lake Erie, Ohio) during late-summer blooms and measured their response to nutrient and metal amendments.Green Bay and Maumee Bay experience recurrent annual algal blooms and are the focus of major management efforts (EPA 2011, De Stasio et al. 2014, Great Lakes Interagency Task Force 2014, Scavia et al. 2014, Annex 4 Task Team 2015).

Study locations
Green Bay is a major embayment of Lake Michigan that drains the predominantly agricultural Fox River watershed.Green Bay is typically dominated by cyanobacteria during the late-summer and fall periods, with Microcystis spp.often being among the most dominant taxa (De Stasio et al. 2008, 2014).We sampled at two locations in Green Bay: (1) Green Bay just outside the mouth of the Fox River at approximate coordinates 41.7265 N, 83.4095 W at a location that is usually <3 m deep (GBM) and (2) 4.7 km offshore from the mouth (open waters) at approximate coordinates 44.5781 N, 87.9823 W at a location that is usually ∼4 m deep (GBO).Phytoplankton were collected from Green Bay on July 10, July 31, and August 21, 2018, and experiments were completed on July 12, August 2 and August 23, 2018 (referred to hereafter as the July, early August and late August experiments).GBO was not sampled on July 10 due to hazardous conditions.
Maumee Bay is at the mouth of the Maumee River, which is the largest direct tributary to Lake Erie.The Maumee River drains a predominantly agricultural watershed, and western Lake Erie (and Maumee Bay) is typically dominated by Microcystis spp.during the late summer and fall periods (Allinger andReavie 2013, Scavia et al. 2014).Our sampling location for Maumee Bay was located just outside the mouth of the Maumee River at approximate coordinates 41.7103 N, 83.4535 W, at a location that is usually <2.5 m deep (MB).Phytoplankton were collected from Maumee Bay on July 11 and August 21, 2018, and experiments were completed 2 days later (referred to hereafter as the July and August experiments).A third sampling event for Maumee Bay was not possible due to dangerous weather conditions at the time when we attempted sampling.

Experimental design
At each location, water was collected by lowering a pump to mid-depth and collecting ∼40 L of water into acid-washed and pre-rinsed 10-L containers.Water temperatures at each sampling event were measured using a dissolved oxygen probe (YSI ProODO probe; Yellow Springs, Ohio) lowered to the depth at which sample water was collected.Sample containers were immediately placed in coolers and either driven from the sampling location to the U.S. Geological Survey (USGS) Upper Midwest Environmental Sciences Center (UMESC; La Crosse, Wisconsin) for samples from Green Bay or shipped via overnight shipping to UMESC (samples from Maumee Bay).Samples were always collected in late afternoon to minimize holding times.The following morning, the sample containers were mixed together in a single large container to homogenize the community.This large container remained continuously mixed during experimental setup.From this homogenized material, triplicate samples of whole water were collected for nutrients (listed below), chlorophyll a, trace metals (listed below) and DNA/RNA (details below).We attempted to minimize external metal contamination via use of low density polyethylene collection vessels, thorough acid-washing of all materials prior to use (including the pump used for collection), and avoiding contact with any non-clean surfaces (Benoit 1994).Samples were never in contact with metal surfaces.We also used trace-metal grade chemicals for creating nutrient stock solutions.
After the initial samples were collected, 24 borosilicate bottles were filled with 1 L of homogenized lake water and randomly assigned a treatment combination.Treatments included additions of concentrated P (K 2 PO 4 ), N (NH 4 Cl).or a mixture of trace metals (T; FeCl 3 , MnCl 2 , NiCl 2 , ZnSO 4 , NaMoO 4 ) in a full factorial design with a total of 3 replicates for each treatment combination.Phosphorus and trace metal concentrations were targeted at the final concentrations present in COMBO media, which was chosen because it is known to support very high algal growth rates and would therefore alleviate any limitation by those elements (Kilham et al. 1998).The final target concentration for P was 1.55 mg L −1 .Final concentration targets for the T treatment were as follows: 0.998 mg Fe L −1 , 0.131 mg Mn L −1 , 0.0128 mg Zn L −1 , and 0.0217 mg Mo L −1 .Mn was added to the experiment as it is a trace metal included in COMBO media.Ni is not present in COMBO media but it was added at a concentration similar to that of Zn (0.0190 mg Ni L −1 ).COMBO media includes N as nitrate, but because many phytoplankton use trace metals to assimilate nitrate, we used NH 4 at a target concentration of 2.7 mg N L −1 (well above what we expected to see in field conditions).
Once bottles were filled and spiked with nutrient treatments, they were inverted several times and placed in an incubation chamber set to the water temperature at the time of collection.Light was provided on a 15:9 day-night cycle (consistent with the day-night interval for July 9 th at Maumee Bay) by a bank of LED grow lamps (Durolux DLED824W LED Grow Light).Treatments were arranged in a stratified random pattern from front to back to minimize the effect of location within the experimental chamber on potential treatment effects.Each day, bottles were mixed by inversion at least twice (at approximately 8 a.m. and 5 p.m.).Experiments were run for 48 h, which has been shown to be sufficient to observe changes in phytoplankton during other experiments studying blooms (Davis et al. 2015).At the end of the experiment, each bottle was thoroughly mixed, and samples were collected for chlorophyll a, volatile suspended solids (VSS), and DNA.Only two of three replicates were sampled for DNA due to the high cost of that analysis, and no DNA analysis was measured at site GBO.

Nutrient and trace metal analysis
Nutrient samples were processed following methods described in Soballe and Fischer (2004, Sections 5.3-5.5, 5.7-5.8) at the UMESC Water Quality Laboratory.Collection methods were identical to those previously reported in other studies (e.g.Larson et al. 2019).Briefly, whole water samples were analyzed for total N (TN) and total P (TP), and filtered (0.45-µm glass fiber filter) samples were analyzed for soluble reactive P (SRP), nitrate + nitrite (NO X ), and ammonium (NH 4 + ).Dissolved inorganic nitrogen (DIN) was calculated by adding the NO X and NH 4 + .Detection limits were as follows: SRP − 0.001 mg L −1 , TP − 0.001 mg L −1 , NO X − 0.01 mg L −1 , NH 4 + − 0.008 mg L −1 , and TN − 0.01 mg L −1 (Soballe and Fischer 2004).
Trace metal samples were collected by filtering whole water through a pre-rinsed 0.45-µm glass fiber filter, then acidifying with sulfuric acid and freezing until samples could be delivered to the UMESC Water Quality Laboratory.Samples were analyzed for Fe, Zn, Mo, and Ni using an Agilent 5110 Inductively coupled Plasma-Optical Emission Spectrometer (U.S. Environmental Protection Agency 1994).Detection limits were as follows: Fe − 1.2 µg L −1 , Ni − 1.8 µg L −1 , Mo − 2.0 µg L −1 , and Zn − 1.8 µg L −1 .

Chlorophyll a and volatile suspended solids
Chlorophyll a (chl a) and VSS concentrations were measured on initial water samples and after experimental treatments.The collected water was filtered onto 0.45-µm glass fiber filters that were held frozen until they were processed following methods described in Larson et al. (2016).Filtering for chl a took place in the dark, and chl a samples were processed within 30 days of collection.We report VSS here, rather than total suspended solids, because we have observed in rivermouth habitats like these that inorganic suspended solids can sometimes contribute substantially to the total suspended solids (Larson et al. 2016).

DNA analysis
Samples for DNA analysis were collected by filtering up to 100 mL of sample water onto a 0.2-µm filter using sterile components (filter funnels were decontaminated using a bleach rinse between uses).Filters were then folded using sterile forceps into bead tubes and immediately frozen.Frozen samples were sent to the USGS Ohio Water Microbiology Laboratory where DNA-based quantitative polymerase chain reaction (qPCR) was used to quantify Microcystis 16S genes.DNA extraction and qPCR methods are described in Stelzer et al. (2013).Quality control samples for all qPCR assays included process blanks (deionized water filtered instead of sample water), extraction blanks, qPCR blanks, and positive control standards.All blank results were several orders of magnitude below sample values, if detected at all.

Statistical analysis
For initial samples, we occasionally observed non-detects.If only 1 of the 3 samples was below the detection limit, we used a left-censored method for estimating mean and standard deviation (Kaplan-Meier) using the NADA package in R (Lee 2020), as recommended by Helsel (2005).For all other descriptive statistics, means and standard deviation were calculated as usual.To estimate correlation coefficients between individual variables, we used the non-parametric Kendall's tau (τ), which does not require distributional assumptions about the variable being correlated and handles non-detects appropriately (Helsel 2005).To relate treatments to response variables in the experiments, we used generalized linear models in R (Version 4.2.2,R Core Team 2022) with interaction terms to estimate effect sizes and interaction effects.Models to estimate effect sizes on Microcystis DNA copies had insufficient sample size for proper parameterization and are not reported here.Once models were parameterized, we estimated a 95% confidence interval around the effect sizes using a bootstrap method.We considered effect size 95% confidence intervals that did not include zero to have the strongest support.

Initial conditions
The preferred form of N (NH 4 ) was present at the start of all experiments performed here, although often at concentrations very close to the detection limit in Green Bay (0.008 mg N L −1 ; Table 1).NO X was present in both Maumee Bay sampling events, although it decreased from 5.78 mg N L −1 to 0.31 mg N L −1 from July to late August (Table 1).SRP was present at the start of both Maumee Bay experiments, and the July Green Bay experiment, but by August the Green Bay experiments had initial concentrations of SRP that were below the detection limit (Table 1).TN:TP molar ratios were always above 16:1, but varied from 17.9 in the July Green Bay experiment, to 115.1 in the July Maumee Bay experiment (Table 2).Chl a concentration was relatively high during all experiments (about 43-69 µg chl a L −1 ; Table 2).Fe and Zn were always present above the detection limit in these experiments, but Mo and Ni were below the detection limit at the GBM site in July and early August (Table 2).Ni was also below the detection limit in the July MB sample and the GBO site in early August (Table 2).Microcystis DNA was present initially in all locations and sites where it was measured, indicating the presence of cyanobacteria in these samples (Table 2).The copies of DNA were not strongly correlated to chl a concentration in initial samples (Kendall's τ = 0.03), but was moderately related to VSS (Kendall's τ = 0.47).

Experimental results
Chl a was our primary response variable for evaluating N, P, or T limitation.Effect sizes for P treatments were positive in just 2 of 7 experiments: MB in July and GBO in late August (Table 3, Figure 1A, 1G).N treatment effect sizes were positive in 5 of 7 experiments, and in two experiments adding P + T to N seemed to further increase chl a accumulation (MB in late August and GBO; Table 3; Figure 1).There were no occasions when T-only treatments had positive effect sizes (Table 3).
The other response variable examined here was VSS, which unlike chl a is not specific to phytoplankton or even living tissues.VSS did not indicate P limitation in any of the experiments, although in one experiment (Maumee Bay, late August) VSS decreased in the treatments (Table 4, Figure 2).VSS appeared to be N-limited in 5 of 7 experiments (Table 4, Figure 2), but N seemed to be serially limited with P and T in the GBO site in late August.N and P were simultaneously co-limiting in MB in July (Table 4).The T treatment had a negative effect in the GBO site in late August, but there were no occasions where T treatments had positive effect sizes.In both GBO experiments, P + T treatments had less VSS than would be expected based on the individual effect sizes, indicating a negative interaction (Table 4).
Table 1.environmental conditions at the start of the nutrient incubation experiments.Mean and standard deviation are estimated using Kaplan-Meier methods due to the presence of non-detects (helsel 2005).units for all are mg l −1 .no X -nitrates plus nitrite; nh 4 -ammonium; srP -soluble reactive phosphorus; tP -total phosphorus; tn -total nitrogen; BDl -below detection level.

Discussion
We hypothesized that trace metals might become growth-limiting in algal blooms when three conditions occur simultaneously: (1) low availability of labile N and P (NH 4 and SRP), ( 2) the presence of other forms of N and P, and (3) high densities of phytoplankton.These conditions were met in some of our experiments, especially in Green Bay during August.However, we never observed strong statistical support for trace metal effects in our experiments.
Among the experiments we performed here, the Green Bay (July) phytoplankton community was probably the best chance to observe metal limitation.In the initial conditions both Ni and Mo were below the detection limit and these elements are needed for the acquisition of nitrate and urea when NH 4 is low (Rees andBekheet 1982, Glass et al. 2012).Although NH 4 was present at the start of the experiment, it was present at low concentrations and other forms of N were also present.Furthermore, the addition of NH 4 stimulated chl a and VSS (relative to the control), indicating the community was N limited.Even in this situation, metals did not stimulate growth.All of experiments we performed indicate that the metals studied here were available at sufficient concentrations to support phytoplankton growth, consistent with predictions about the spatial distribution of potential metal limitation made by earlier studies (North et al. 2007, Havens et al. 2012).
If we focus on the chl a data, which is more specific to phytoplankton and thus the purpose of this study, nutrient limitation was observed in 6 of 7 experiments we performed.Maumee Bay appeared to be P-limited in July, but N-limited in late August, consistent with similar bioassay experiments in the western basin of Lake Erie (Chaffin et al. 2018, Barnard et al. 2021).The N:P ratios decreased between those sampling events (DIN:SRP molar ratio from about 1,072 to 67; TN:TP from about 115 to 21; Table 2).The GBM site near the Fox rivermouth was N-limited in July and late August when TN:TP ratios were 17.9 and 21.7, respectively.Although the classic Redfield ratio (N:P, 16:1) is often used to indicate the transition from N to P limitation (Sterner and Elser 2002), here we experimentally observed N limitation at TN:TP and DIN:SRP ratios that would have indicated P limitation using the 16:1 ratio.This may indicate that some of the N present was not available to the phytoplankton community.However, this lack of availability was not alleviated by metal amendment in this study as we had hypothesized.Other studies on Lake Erie and Lake Superior nutrient uptake have shown that uptake of less bioavailable N (nitrate) can be limited by metals, but presumably that was not the case here (Ivanikova et al. 2007, Havens et al. 2012).
Table 2. nutrient ratios (molar), chlorophyll a (chl a), volatile suspended solids (Vss), and trace metal concentrations (fe, Mo, ni, Zn) at the start of the nutrient incubation experiments.If nitrate plus nitrite (no X ) concentration was below the detection limit (BDl), we assumed it was zero for the purpose of calculating the dissolved inorganic nitrogen (DIn) to soluble reactive phosphorus (srP) ratio.Mean and standard deviation for trace metals are estimated using Kaplan-Meier methods due to the presence of values below the detection limit (see detection limits in text; helsel 2005).when two or more samples were BDl, we did not calculate a mean and standard deviation.MB -Maumee Bay (lake erie); gBM -green Bay at the mouth of the fox river (lake Michigan); gBo -open waters of green Bay (lake Michigan), tP -total phosphorus, tn -total nitrogen.site Month DIn:srP tn:tP chl a (µg l -1 ) Vss (mg l -1 ) fe (µg l -1 ) ni (µg l -1 ) Mo (µg l -1 ) Zn (µg l -1 ) Microcystis Dna copies (millions l Trace metals are essential for many core metabolic functions in cyanobacteria and other phytoplankton taxa.Our experiments focused on indicators of biomass or abundance, but many other aspects of cyanobacterial physiology are potentially affected by metal availability.For example, the most commonly observed cyanotoxins in many North American waters are microcystins   (Loftin et al. 2016, Graham et al. 2020).Microcystins have an ancient origin, and their functional role is unclear (Rantala et al. 2004, Omidi et al. 2018).One hypothesis is that microcystins act to manage biologically important metals (Utkilen andGjølme 1995, Omidi et al. 2018).This hypothesis is supported by evidence that iron limitation triggers microcystin production (Sevilla et al. 2008), microcystin forms complexes with some trace metals (Klein et al. 2013), and experimental evidence that microcystins prevent the loss of photosynthetic function in low-Fe conditions by storing Fe (Alexova et al. 2011, Ceballos-Laita et al. 2017).Microcystin appears to form similar complexes with Mo, Cu, and Mn, but not Zn (Ceballos-Laita et al. 2017).If microcystin production is increased when metal demand increases, then even if trace metals are present at sufficient concentrations to avoid growth limitation, toxin production could be affected.During algal blooms in Green Bay and the nearshore waters of Lake Erie near Maumee Bay, labile inorganic forms of N and P often reach low water column concentrations (Larson et al. 2019, Rowland et al. 2020), presumably because the large phytoplankton biomass is rapidly removing these labile nutrient forms from the water column.In this scenario, metals would be required to make use of other N and P forms.This is occurring when a high concentration of phytoplankton biomass is present, and therefore demand for these metals would be high, and so we had anticipated this would be an ideal occasion to observe metal limitation.We did not observe evidence of metal limitation of phytoplankton growth, and therefore conclude metal supply was sufficient to support growth in these communities during the time our sampling occurred.Our study was very limited in spatial and temporal extent, and therefore it is difficult to extrapolate these results to other times and locations where blooms might occur.

Figure 1 .
Figure1.chlorophyll a (chl a) concentration (µg l −1 ) in replicate incubation chambers started with naturally occurring phytoplankton communities from Maumee Bay (lake erie, near toledo, ohio) and green Bay (lake Michigan, near green Bay, wisconsin).open symbols are individual samples and filled symbols are the mean.In a few treatments, one sample was lost during processing, in which case we did not plot the mean of the two remaining samples.Incubation chambers were treated with different amendments of nutrients or metals: c -control (no added nutrients or trace metals); n -nh 4 amended; P -Po 4 amended; t -fe, ni, Mo, and Zn amended.

Figure 2 .
Figure 2. Volatile suspended solid (Vss) concentration (mg l −1 ) in replicate incubation chambers started with naturally occurring phytoplankton communities from Maumee Bay (lake erie, near toledo, ohio) and green Bay (lake Michigan, near green Bay, wisconsin).open symbols are individual samples and filled symbols are the mean.In a few treatments, one sample was lost during processing, in which case we did not plot the mean of the two remaining samples.Incubation chambers were treated with different amendments of nutrients or metals: c -control (no added nutrients or trace metals); n -nh 4 amended; P -Po 4 amended; t -fe, ni, Mo, and Zn amended.

Table 4 .
results of generalized linear models relating nutrient or metal treatments to volatile suspended solids in experimental chambers.treatment effects (mg l −1 ; with 95% confidence intervals) are shown.treatment effects with 95% confidence intervals that do not overlap zero are in bold.sites are as follows: MB -Maumee Bay (lake erie); gBM -green Bay at the mouth of the fox river (lake Michigan); gBo -open waters of green Bay (lake Michigan).treatments: c -control (no added nutrients or trace metals); n -nh