Ploidy Distribution of the Harmful Bloom Forming Macroalgae Ulva spp. in Narragansett Bay, Rhode Island, USA, Using Flow Cytometry Methods

Macroalgal blooms occur worldwide and have the potential to cause severe ecological and economic damage. Narragansett Bay, RI is a eutrophic system that experiences summer macroalgal blooms composed mostly of Ulva compressa and Ulva rigida, which have biphasic life cycles with separate haploid and diploid phases. In this study, we used flow cytometry to assess ploidy levels of U. compressa and U. rigida populations from five sites in Narragansett Bay, RI, USA, to assess the relative contribution of both phases to bloom formation. Both haploid gametophytes and diploid sporophytes were present for both species. Sites ranged from a relative overabundance of gametophytes to a relative overabundance of sporophytes, compared to the null model prediction of √2 gametophytes: 1 sporophyte. We found significant differences in cell area between ploidy levels for each species, with sporophyte cells significantly larger than gametophyte cells in U. compressa and U. rigida. We found no differences in relative growth rate between ploidy levels for each species. Our results indicate the presence of both phases of each of the two dominant bloom forming species throughout the bloom season, and represent one of the first studies of in situ Ulva life cycle dynamics.


Introduction
Macroalgal blooms typically consist of large accumulations of ephemeral macroalgal biomass. These blooms occur worldwide, often in shallow areas with relatively low water mixing that are affected by coastal eutrophication, and they have the potential to cause severe ecological and economic damage [1][2][3]. The largest documented bloom on record occurred four weeks before the 2008 Beijing Olympics, with a bloom of an estimated 20 million tons of Ulva prolifera in system of Narragansett Bay, Rhode Island [41,42]. Macroalgal densities (comprised mostly of Ulva) peak in the summertime and vary significantly across sites, seasons, and years [43][44][45].
Our research focuses on four central questions regarding the life cycles and biology of U. compressa and U. rigida. Firstly, what is the relative abundance of sporophytes and gametophytes of both species? Secondly, how do these relative abundances correlate with physical and biological factors? Thirdly, do the phases have different growth rates, and lastly, do the phases have cells of different sizes? We interpret our data in the context of macroalgal bloom dynamics and the impacts of environmental variables in structuring bloom formation.

Collection of Ulva
We collected Ulva spp. monthly from June to October 2013 at several publically accessible bloom-forming sites in Narragansett Bay, RI, including Chepiwanoxet, Sandy Point, Oakland Isomorphic biphasic life cycle of Ulva. Ulva cycles between two morphologically similar multicellular adult phases, a haploid gametophyte and a diploid sporophyte. Diploid sporophytes produce haploid zoospores that develop into gametophytes. Haploid gametophytes produce haploid gametes. When a "+" and "-" gamete fuse a zygote is formed, which develops into a diploid sporophyte.
Beach, Oakland Beach Cove, and Warwick City Park. We chose these sites to represent a range of typical Ulva spp. bloom intensity, with Oakland Beach Cove and Warwick City Park as high bloom sites, while Chepiwanoxet, Sandy Point and Oakland Beach as low bloom sites (Thornber, unpublished data). At each site, on each sampling date, we haphazardly collected individuals by hand from the shallow subtidal zone, put them in a plastic bag, and brought them back to the lab. We selected a minimum of 16 individuals and maximum of 40 individuals on each sampling date. Later, we identified U. compressa and U. rigida to the species level by microscopic examination and only used individuals with clear cellular characteristics based on the current molecular analyses of Ulva in Narragansett Bay [45]. A recent study by Mao et al. 2014 discovered the presence of U. laetevirens in Long Island Sound [46]. Since there are morphological similarities between U. rigida and U. laetevirens, we recognize the potential for species misidentification, however slight, in our study. Overall, we collected and analyzed 282 total Ulva individuals: 150 U. compressa and 132 U. rigida (S1 Table). Both species were collected at all sites, with a minimum of 10 individuals of each species at each site. Due to the nature of sampling and length of time necessary for preparing flow cytometry samples (which limited our ability to collect larger sample sizes), we present and analyze our data here in terms of the overall relative abundance of each Ulva species during the peak bloom-forming season at each site. However, we use collection date and month as covariates in building our logistic regression models for predicting the relative abundance of each phase (see Statistical Analysis section). We used sea surface temperature and sea surface salinity data for Greenwich Bay (Site F5) collected daily by the Rhode Island Department of Environmental Management Bay Assessment and Response Team (http://www.narrbay.org/d_projects/buoy/buoydata.htm; S2 Table).
We also determined Ulva biomass data from monthly subtidal surveys of the same sites, following the protocol in Guidone [44]. Briefly, at each site, we collected all algae in each of 0.16 m 2 subtidal quadrats placed 1 m apart along a transect line. All plots were < 2 m deep at mean lower low water (S1 Table).
Prior to thallus destruction for flow cytometry, we took a microscopic photograph at 400X of each individual that was analyzed for ploidy content. Using ImageJ (www.nih.gov), we created an overlying grid on each microscopic photograph, and measured the area of the exposed surface of the first ten cells that were at grid intersection points to assess cell size differences between phases (S3 Table). We examined the upper cell layer, as U. compressa and U. rigida are each two cells thick.

Flow Cytometry and Ploidy Analysis
We used flow cytometry to determine the relative abundance of gametophytes and sporophytes in U. rigida and U. compressa. Based on the C-values (haploid genome sizes) of U. compressa 0.13 pg [27] and U. rigida 0.16 pg [27], we used the freshwater unicellular alga Chlamydomonas reinhardtii as an external flow cytometry control, with a C-value of 0.12 pg [47]. We specifically selected the cell wall-deficient mutant CC-400 cw15 mt+ as our control (University of Minnesota Chlamydomonas Center, chlamycollection.org). The cell wall-deficient mutant was selected to easily rupture the cells and allow the PI/RNase Staining Buffer to reach the nucleus.
We used an enzyme solution developed specifically for efficient production of Ulva protoplasts [48], along with a modified version of the LB01 nuclear isolation buffer. Instead of the standard 0.1% v/v concentration for Triton X-100, we modified the buffer to contain a 1% v/v concentration to ensure the nuclei were cleanly isolated (15mM Tris, 2MM EDTA, 0.5mM Spermine tetrahydrochloride, 80mM KCl, 20mM NaCl, 1% vol/vol Tritron x-100, 15mM β-mercaptoethanol) [26].
We were concerned with successful protoplast isolation and not with the exact number of protoplasts obtained, so we chose a qualitative method for isolating protoplasts [48]. We weighed all Ulva samples to 0.50 g wet weight, rinsed with them raw seawater to remove debris and epiphytes, and then thoroughly scrubbed them manually in 20 μm filtered seawater to remove smaller particles. Ulva samples were chopped with a razor blade in a large (85 mm x 25 mm) plastic Petri dish for one minute, and then the tissue was transferred into a small (55 mm x 15 mm) Petri dish that contained 5 mL of enzyme solution [48].
Protoplasts were released by placing samples on a shaker at 50 rpm in the dark for two hours at room temperature (~21°C), then filtered with a 30μm nylon mesh into a 5mL polypropylene tube and spun for five minutes at 120 x g at 4°C. A total of 2mL of supernatant was then removed and replaced with 2mL of sterile filtered seawater. Centrifugation with subsequent replacement of fluid was repeated twice, and after the last round of centrifugation, all supernatant was removed and replaced with 1mL of sterile filtered seawater. We observed successful protoplast isolation via microscopic examination at 400X. In preparation for the flow cytometer samples were spun for five minutes at 120 g at 4°C, the supernatant was removed, and samples were kept refrigerated or on ice.
To liberate the nuclei, we added 1 mL of modified LB01 nuclear buffer kept on ice to each sample, vortexed and tapped the tube occasionally for eight minutes, and then added 0.5mL of PI/RNase Staining Buffer (BD Science). After five minutes the samples were run on a BD Influx flow cytometer at the RI EPSCoR Marine Life Sciences Facility on the University of Rhode Island's Narragansett Bay Campus. This machine was optimized for marine applications and is equipped with three lasers (355 nm, 488 nm, and 561 nm). We used a green (532 nm) or a blue (488 nm) laser and quantified fluorescence at 610 nm (20 nm bandwidth) on a linear scale. Since sporophytes have twice the amount of genetic material as gametophytes, sporophytes have twice the amount of fluorescence as gametophytes (Fig 2). To measure the spread of the distribution of the data we used the coefficient of variation (CV), which is the standard deviation expressed as a percentage of the population mean. The CV was calculated from replicate counts of the same prep from one thallus; our CV values ranged from 3-8%. This range is due to the small genome size and the predilection of PI to bind to remaining cell wall polysaccharides from the extraction of Ulva protoplasts, which makes obtaining CV values less than 3% challenging [49,50].

Growth Experiments
We assessed growth rates of gametophytes and sporophytes of U. rigida and U. compressa in outdoor flow-through ambient temperature seawater tanks on the University of Rhode Island's Narragansett Bay campus. We collected healthy Ulva individuals from the shallow subtidal zone in Greenwich Bay in the summer of 2013. In total, we used 90 U. compressa individuals (62 sporophyte and 28 gametophyte) and 61 U. rigida individuals (38 sporophyte and 23 gametophyte) for this analysis. We conducted growth experiments in June, July, and August to assess differences in growth over the peak bloom-forming months (S4 Table).
In the lab, we determined the species identity of each specimen via microscopic examination. We then spun individuals 20 times in a salad spinner prior to separating 1.0 g from the thallus. We placed one 1.0g Ulva individual in each 2.5 L bucket with mesh sides; after 14 days, all growth experiments concluded and the Ulva was re-weighed. For each month, we had a sample size of at least five (up to a maximum of 36) individuals of each phase of each species, except for U. rigida sporophytes in August, when we only had three individuals. All Ulva were spun 20 times in a salad spinner prior to each weighing on a digital scale to ensure consistent mass, and all individuals were analyzed using flow cytometry for ploidy content (see above).

Statistical Analyses
To assess ploidy ratios in field populations of U. compressa and U. rigida, we used a χ 2 analysis to determine if the relative abundances of each species, at each site, were significantly different from the null model hypothesis. We then assessed the relationship of several variables (site, species, salinity, temperature, month of collection, date of collection, total Ulva biomass, total algal biomass, total Ulva biomass) to the ploidy ratio, using a logistic regression model with a binomial response variable (gametophyte vs. sporophyte). We selected the model with the highest AIC as it best explained the distribution of gametophytes and sporophytes in Greenwich Bay (S1 Text).
The AIC measures the relative quality of a statistical model, taking into consideration the number of parameters and the information lost with the model. Model coefficient estimate values predict the odds ratio of gametophytes and sporophytes in the population. The model has a binomial response variable with sporophytes chosen as success and gametophytes as failure. Therefore, negative estimate values are associated with higher proportions of gametophytes while positive estimate values are associated with higher proportions of sporophytes.
Based on the results for the logistic regression model described above, we then selected the three significant continuous variables (salinity, salinity two weeks prior to specimen collection, and total Ulva biomass) and analyzed each individually in separate models for representation in graphical models. Data analyses were conducted in R [51,52] and JMP (JMP 1 , Version 10. SAS Institute Inc., Cary, NC, 1989-2013).
Relative growth data were analyzed with a two way fixed factor ANOVA to measure differences across ploidy levels and months. Cell sizes were compared between gametophytes and sporophytes for each species using t-tests with unequal variances in JMP. All data were checked for statistical test assumptions and transformed where appropriate prior to analysis.

Ethics Statement
All research was conducted on public beaches in Rhode Island. No specific permits were obtained for this research, as the Rhode Island state constitution guarantees its citizens the right collect seaweed from public beaches [53]. The study did not involve any endangered or protected species or any protected locations.

Ploidy
We found both gametophytes and sporophytes of each species present at each of the sampling location sites (S1 Table). There were significant differences among the relative ploidy levels at each site (Fig 3), compared to the null model prediction of p 2 gametophytes to 1 sporophyte (χ 2 likelihood test, Table 1). U. compressa in Oakland Beach Cove (OBC) and Sandy Point (SP) differed from this null prediction with a relative overabundance of sporophytes. U. rigida in Warwick City Park (WCP) and Sandy Point (SP) differed from the null prediction with a relative overabundance of sporophytes in WCP and dominance of gametophytes in SP. Based on AIC values, the strongest predictive model for ploidy relative abundance included the variables species, site, salinity at time of sampling, and total Ulva biomass ( Table 2; S2  Table; S1 Text) and not temperature, month of sampling, date of sampling, or total algal biomass. While salinity measurements with a time lag of two weeks prior were significant, they were not included in the model with the strongest AIC.
When we analyzed the significant continuous variables individually for their correlation to ploidy ratios, we found that the relative abundance of sporophytes was positively correlated with higher Ulva biomass at the time of collection (Fig 4A; χ 2 3 = 16.10, p<0.01). We found increasing proportions of Ulva sporophytes at higher salinities at the date of sampling for both species (Fig 4B;

Ploidy Distribution
Our data indicate that both phases are present for both U. compressa and U. rigida throughout the peak bloom-forming season, and that relative phase abundance is correlated with both abiotic and biotic factors. We found a high variability among sites in ploidy ratio among sites, with some sites matching the null model prediction of relative abundance, while others exhibited a significant overabundance of gametophytes or sporophytes. These deviations could be due to ecological differences among phases, environmental differences among sites, and/or temporal differences in life cycle dynamics among sites. Sandy Point, which differed from the null hypothesis for both species, is a more exposed site and experiences more water mixing than the other sites [54]. However, as U. compressa had an overabundance of sporophytes and U. rigida had an overabundance of gametophytes at this site, the relative impacts of environmental factors are challenging to assess and may represent specific environmental factors unique to each species. Warwick City Park and Oakland Beach Cove, which differed from the Population Dynamics of Bloom-Forming Ulva null hypothesis in U. compressa and U. rigida respectively, are more sheltered sites and experience less water mixing [54]. We found a significant correlation of physical and biological factors on the relative abundance of gametophytes and sporophytes in our study system ( Table 2, Fig 4). In this study system, low salinities are typically a result of increased freshwater flow from rivers caused by storms. In Narragansett Bay, increased flow in rivers yields higher concentrations of dissolved inorganic nitrogen and phosphorus [55]. Therefore, although nutrient data are not available for our sampling period, low salinities can be used as a proxy for increased nutrients. Lower salinities from the date of sample collection were correlated with higher relative levels of gametophytes, while lower salinities from two weeks prior to specimen collection were correlated with more sporophytes (Fig 4). This shift in ploidy ratios may be due to several factors, such as salinity tolerance, positive response to nutrient availability from one phase over the other, or a shift to asexual reproduction [56]. While it is unlikely that a reproductive event would result in the presence of new adults after only two weeks [57], lower salinities may trigger more rapid growth of one phase from a microscopic to a macroscopic size [35]. Due to the biphasic life cycle, increased nutrients may either impact mortality and/or fecundity rates of either phase [40,58], with differential effects on the relative balance of phases. In addition, vegetative fragmentation of mature blades, germination of unfused gametes, and/or asexual production of diploid spores by sporophytes may impact the ploidy ratio [18].
We also found a positive correlation between the relative abundance of sporophytes for total Ulva biomass for both species. This may be a byproduct of the positive correlation of temperature with bloom abundance [59,60] and growth rates [44], although we found no impact of temperature on the relative abundance of gametophytes and sporophytes in this study.
Previous studies have found a seasonal dominance of one ploidy phase [30] or a long term (11-20 month) non-seasonal cyclic dominance [19], or no seasonal trend [61]. As our sampling was limited to the bloom forming season, a cycling trend in ploidy for U. compressa and/ or U. rigida could exist. However, due to the scarcity of Ulva specimens during non bloom forming periods [43], this would be challenging to assess.

Growth and Cell Area
We did not find any significant differences in growth rates of adult gametophytes and sporophytes of either species, but this does not preclude the possibility of differences at the germling stage [35]. In addition, growth rates can vary based on nutrient levels [62]; as nutrient levels shift in Narragansett Bay over seasonal cycles [63,64], differences in Ulva growth rates between phases may emerge.
Based on our cell area data, future studies of U. compressa and U. rigida life cycle dynamics may be much more rapid. Individuals can be predicted as gametophytes or sporophytes based on their cell area, with a subset confirmed using ploidy analysis. This would increase the ability to have larger sample sizes and more rapid assessment.
Differences in U. compressa and U. rigida cell areas between phases may impact the surface area to volume ratio, allowing for faster uptake of nutrients in smaller cells [40]. This is especially relevant in single-celled spores, gametes, and small juveniles, and may impact Ulva individuals in their early growth stages. U. rigida zoospores are 9-15 μm x 5-10 μm while gametes are 7-11 μm x 4-6 μm [65]. Since gametes are smaller than zoospores, they may have a survival differential in their ability for nutrient uptake and storage capacity. There may also be other ecological differences between either phases across their lifespan, such as susceptibility to herbivores, light tolerance, salinity tolerances, and temperature optima [23,29,34,66], that may explain differences in ploidy ratios.

Flow Cytometry Method
We designed our flow cytometry ploidy analysis methods from similar analyses in higher plants [26,49,67,68], which has been successful for other macroalgal studies [27]. We first attempted chopping Ulva tissues with a razor blade in the presence of a nuclear isolation buffer to obtain isolated nuclei (essentially removing our protoplast isolation step). This method, which is successful in higher plants for flow cytometric analysis [69], was unsuccessful for Ulva. The number of nuclei obtained was small and contaminated with other materials, likely organelle genomes and bacteria [70]. In addition, Ulva has high concentrations of anionic polysaccharides in its cell walls [71] which can interfere with obtaining a sufficient number of nuclei by binding to the positively charged nuleus, inhibiting the propidium iodide from attaching. Given these constraints, protoplast isolation was necessary to obtain sufficient numbers of nuclei for flow cytometry analyses [50], which is successful yet time consuming [48], thus limiting our abilities to obtain larger sample sizes.
Supporting Information S1