of freshwater mixotrophic plankton
28 may 2021, Kampen
By Merel Lanjouw
Supervisor Jolanda Verspagen
Examiner Susanne Wilken
Abstract
Mixotrophic plankton are organisms which utilize photo-autotrophy and phago-heterotrophy as nutrient and energy acquisition strategies. Although mixotrophy is the main trophic strategy, depending on trophic status of the environment and the physiology of the plankton, species can predominantly use one certain strategy. Some species are phototrophic and use phagotrophy under certain extreme conditions or to supplement nutrients. Similarly, mixotrophic plankton can also be primarily phagotrophic. Mixotrophic plankton are widespread in the aquatic systems, but much is still unknown about what drives freshwater mixotrophs though it is understood they fulfill an important part in the ecosystem. We investigated the effect of several lake variables on the abundance of mixotrophic genera. We looked at a survey performed on over a thousand lakes in the United States, which measured biological (e.g., plankton biovolume), chemical and physical stressors on lake water quality. The data on each genus, such as their morphology and the dominant nutrient acquisition strategy, were based on literature. A multiple stressor analysis was performed on the data using Rstudio. Our results show that indicators for eutrophic lakes, such as chlorophyll, phosphorus and prey (phytoplankton) concentrations are important predictors for primarily autotrophic genera such as Ceratium, Dinobryon, Cryptomonas and Peridinium. Furthermore, for some mixotrophs which rely on phagotrophy a positive correlation was found to organic nutrients and prey concentration such as Cryptomonas, Dinobryon, Peridinium and Plagioselmis. We have found that there is a relation between the trophic strategy of certain genera to the variables which predicts the abundance. Primarily autotrophic genera which naturally thrive in eutrophic lakes show a positive correlation with chlorophyll and phosphorus concentrations. Mixotrophs which rely on phagotrophy for essential nutrients show a positive correlation with prey concentrations. Genera which naturally occurs in shallow or stratified lakes show a negative correlation with lake depth.
KEY WORDS: Mixotrophy · Lake stressors · Ceratium · Cryptomonas · Dinobryon · Euglena · Peridinium · Plagioselmis · Trachelomonas · United States Lakes · Trophic strategy · multiple stressor analysis
Table of Contents
Abstract ... 1
Introduction... 4
Methodology ... 6
Study area and variable selection ... 6
Literature review: Mixotrophic genera ... 6
Data processing ... 7
Transformation, collinearity and variable importance ... 7
Model inferencing and linear regression model fitting ... 8
Results... 8
Functional traits of mixotrophic genera ... 8
Predictors ... 1 Variable interaction ... 1 Discussion ... 1 Ceratium ... 1 Cryptomonas ... 1 Dinobryon ... 2 Euglena ... 2 Peridinium ... 3 Plagioselmis ... 3 Trachelomonas ... 3 Relevance ... 4 Further research ... 4 Conclusion ... 4 References ... 5 Acknowledgements ... 10 Appendices ... 11
Appendix 1: Shapiro-Wilk and p values of the transformed data... 11
Appendix 2: histograms of untransformed, log transformed and bestNormalize transformed data ... 1
Appendix 3... 1
Ceratium ... 1
Cryptomonas ... 2
Euglena ... 7
Peridinium... 9
Plagioselmis ... 12
Introduction
Plankton are an essential part of worldwide ecosystems. Phytoplankton are primary producers, which allows them to capture CO2 and become the basis of extensive food chains (Dodds & Whiles, 2010). Zooplankton, on the other hand, are secondary producers and consume organic particles, bacteria or even other plankton, which is essential for the circularity of organic material in a system (Britannica, 2019). For years these two types of plankton were thought to be distinct. However, research has shown that some plankton can be both primary and secondary producers. Some phytoplankton were found to use osmosis (diffusion through the membrane) for ingesting certain organic particles, and some even possess mechanisms for phagotrophy (engulfs particle in membrane). On the other hand, zooplankton ingested cyanobacteria and retained the photosynthetic machinery to supply themselves with additional energy or nutrients (Mitra et al., 2016). Any species which can take up inorganic nutrients as well as organic nutrients is called mixotrophic. These nutrients often refer to carbon, though there are mixotrophs which use carbon from one trophic strategy and supply themselves with nutrients through multiple strategies. Mixotrophs can be seen in nature as parasitic or even carnivorous plants, but in aquatic systems they are much more widespread, and fulfill an important role in the nutrient cycle. Recent attention has focused on marine mixotrophs, which contribute to the carbon pump and of which some species are responsible for harmful algal blooms (Mitra et al., 2014). Less is known about freshwater mixotrophs. Freshwater systems are highly diverse and prone to fluctuations and extremes in temperature, pH, turbulence and solute concentrations. This wider range of variability, between and within lakes, results in large differences in plankton abundance and composition.
Mixotrophs can be categorized based on their nutrient/energy acquisition, the extent to which they rely on photo- or heterotrophy, but also on the presence of flagella. The nutrient acquisition of freshwater mixotrophs show a diverse range depending on the environment. Some mixotrophs are almost exclusively phototrophic but use phagotrophy when light is absent or extremely limiting and vice versa. When a mixotrophs needs a certain strategy for growth, this is called obligate. When the strategy is supplemental or only used under certain conditions, it is called facultative. Depending on which mechanism is obligate and which is facultative, certain mixotrophs will thrive in eutrophic systems while others thrive in oligotrophic systems. Saad et al. (2016) showed that mixotrophic plankton which primarily use heterotrophy thrive in oligotrophic lakes. Therefore, a higher abundance is expected at low nutrient concentrations (nitrogen, phosphorus and CO2) and high prey concentration. Similarly, Saad et al. showed that mixotrophic plankton which are obligate phototrophs thrive in eutrophic lakes, therefore the expectation is that plankton will show a higher abundance at higher nutrient concentrations.
Another distinction can be made on the presence of carbon concentrating mechanisms (CCM). CCM’s are beneficial for exclusive autotrophic species, these species acquire their nutrients from purely inorganic sources. The general inorganic source for carbon is CO2, but in combination with other phytoplankton or high nutrient concentrations, the CO2 concentration can dip below the uptake threshold (for growth). Some species have developed a mechanism which allows them to take up bicarbonate as well, allowing higher growth rates and a competitive advantage. Carbon concentrating mechanisms have been understudied in mixotrophs, only of some genera it is known whether they possess the CCM. Furthermore, since mixotrophs use organic carbon as well, the need for a CCM might be lower.
Because mixotrophy is a relatively new concept, much is still unknown about these plankton. This research looks at the environmental (e.g. nutrients, temperature, light) and biological (e.g. presence of prey) conditions which drive mixotrophic plankton abundance. A distinction is made between mixotrophic genera, based on the presence of flagella, carbon concentrating mechanisms and dominant trophic strategy. The aim of this research is summarized in the following research question:
Which (environmental) variables can predict the abundance of mixotrophic plankton genera with certain trophic strategies best?
Mixotrophs are defined here as plankton which can use photo-autotrophy and phago-heterotrophy. The decision to research on the genus level is supported by a research performed on the same dataset by Sodré et al (2020) which concludes that coarse (genus level) classification is applicable for inferring responses of environmental variables. Based on literature the following hypothesis is drawn up. Mixotrophs with flagella will likely thrive in more stratified lakes (Saad et al., 2016), therefore a negative correlation with depth is expected, since deeper and larger lakes are more likely to be turbulent. Certain genera will likely thrive under eutrophic conditions whereas other might favor oligotrophic conditions, this and the dominant trophic strategy will be expanded on in the results. Obligate phototrophs will show lower presence in low light and high DOC conditions (Jones, 2001).
The data used in this research was obtained in >1000 lakes in the United States, for locations see Figure 1 (National Lakes Assessment | US EPA, 2021). This program measured physical, biological and chemical stressors on lake water quality on multiple locations in each lake.
Figure 1: a map with all lake locations at which measurements were done (National Lakes Assessment | US EPA, 2020)
Methodology
Study area and variable selection
The United States Environmental Protection Agency has collected biological, chemical and environmental data in over a thousand lakes in the US (National Lakes Assessment | US EPA, 2021). The data has been published under the name National Lakes Assessment (NLA), and is available for downloading from their website. The datasets in Table 1 from lakes 2012 survey were used in this research. From these datasets the mixotrophs genera could be selected. The predictor pCO2, was calculated according to this
spreadsheet.
Literature review: Mixotrophic
genera
Over two hundred genera were counted in the dataset. The mixotrophic genera were found by selecting for genera which occurred > 150 observations, selected using lists of genera that are known to be mixotrophs (Vogt et al. 2017, Hansson et al., 2019). After finding the mixotrophic genera a literature research was performed on the trophic strategy of the genera as well as the presence of CCM and flagella. Scientific articles were found using the combination of terms, using the genera, if known the species, and either “CCM”, “carbon concentrating mechanism”, “flagella”, “trophic strategy”, “nutrient acquisition”, “mixotrophy” or “phagotrophy”. Furthermore, review articles were used per genera as a basis for information and a resource for more articles.
Table 1; Datasets from the EPA, with variables names, predictors and units. Phytoplankton density is used to calculate biovolume per cell. ANC, acid neutralizing capacity, was used to calculate pCO2. The depth and area
of the lake are used as an indicator of turbulence.
Dataset Predictor Unit
Chlorophyll-a Chlorophyll a ug/L Phytoplankton Count Biovolume um3/ml Density Cells /ml Profile pH - Temperature C Oxygen mg/L Conductivity uS/cm Secchi Secchi m
Water chemistry ANC ueq/L
Aluminum mg/L Magnesium mg/L Nitrite mg N/L Nitrate mg N/L Silica mg/L Sodium mg/L DOC mg/L Calcium mg/L Sulfate mg/L Chloride mg/L Nitrate_nitrite mg N/L Potassium mg/L Conductivity uS/cm (25c) Turbidity NTU Ammonia mg N/L Zooplankton Count Zooplankton biomass Nr of organisms / ml
Key variables Depth m
Area HA
Nitrogen mg/L
Data processing
Multiple regression analysis was performed with the abundance of each mixotroph genus as a response variable, and the environmental variables as predictor variables using the multiple stressor ‘cookbook’ by Feld et al. (2016) and Rstudio (2021). In addition to the predictor variables listed in Table 1, the phytoplankton prey volume was calculated because mixotrophs can feed on other phytoplankton. Notably, most mixotrophs feed on bacteria, but there are no data on bacterial abundances in the dataset. Some mixotrophs have specific selectivity for prey size but overall, prey which are too small or too big are not suitable for ingestion. The actual size is uncertain but for this research a range is chosen between 0.4 and 2 um (Gerea et al., 2019).
The following formula was used to calculate the limits from the chosen diameter as volume: Lower limit V =4 3π 0.2 3 = 0.0335 μm3 Higher limit V =4 3π 1 3 = 4.1888 μm3
Cell biovolume in the dataset also needed to be calculated since only total biovolume was available. The following formula was used in R studio to convert the biovolume and density to cell volume:
biovolume (cells μmml3)
density (mlnr) = cell volume ( μm3
cell)
The next step was making a table containing all predictors for each of the genera, using the unique code for each test location. This code allowed the combining of each variable based on location. The variables needed to be transformed in order to follow the normal distribution as much as possible.
Transformation, collinearity and variable importance
In order to use a linear regression model on data it is important that the data is distributed as closely to the normal distribution as possible. Skewed data will lead to incorrect estimate of the impact of a predictor and ultimately the wrong conclusion. Two methods were used to transform the data, in order to compare methods and obtain more reliable results. For the first method a loop was written, which tested all variables with transformation functions such as square root, log and boxcox. A Shapiro-Wilk test was performed to select the function which approached normality best as possible. The variable was transformed using this function and saved in a new data frame. The other method used the best.normalize package which suggests the optimal tool for each variable to obtain normal distribution (Peterson, 2019). The data was transformed using the predict function on the bestNormalize results.
Collinearity was checked with the VIF function from the usdm package, and collinear predictors were removed based on a threshold of 7 (Naimi et al., 2014). In order to find the hierarchy and interaction of/between the variables, the random forest tool was used (Ishwaran & Kogalur, 2021). The outcome and selected variables of the random forest tool were compared to a linear model with all the predictors in order to select the most relevant predictors for model inferencing. Variable importance was visualized using the ggRandomForests package (Ehrlinger, 2016).
Model inferencing and linear regression model fitting
Multi model inferencing was used to determine relevant predictors for the model (Barton, 2020). Two different kinds of rankings were used, the AIC which ranks the model best on the likelihood of being the best model for the data, and the BIC which ranks models based on the likelihood of being true. The BIC has a higher penalty for adding variables, meaning that it is more likely to supply models with fewer variables than AIC. The dredge function was run with different combinations of predictors, based on the strength of their relation to the response and their overall power to predict the model. The reason for running different combinations is due to the fact that the dredge function does not allow over 31 predictors/predictor combinations. Predictors with high p values or very low estimates were eliminated until the best model was left. The last step was adding relevant predictors to a linear model and analyzing the results.
Full reproducible code is available in the data repository.
Results
Functional traits of mixotrophic genera
The mixotrophs that were observed in > 150 lakes belonged to the genera: Ceratium, Cryptomonas, Dinobryon, Euglena, Peridinium, Plagioselmis and Trachelomonas (table 2). Notably for two genera the largest part of the observations consisted of one species. For Plagioselmis it was the P. nannoplanctica and for Ceratium the C. hirundinella. P. nannoplanctica is also the only freshwater Plagioselmis currently known in this genus (Wehr et al., 2015).
Figure 2: top left: Ceratium hirundinella (Fucikova, 2016), top middle: Trachelomonas ellipsoidalis (Plewka, 2018) , top right: Dinobryon divergens (NIES, 2017) , lower left: Plagioselmis nannoplanctica (Bruun, 2013) second lower left: Peridinium (Skjelbred, 2015), lower middle: Euglena acus (Oyadomari, 2010) lower right: Cryptomonas ovata (Plewka, 2011),
All selected genera use their flagella to navigate the water column, they ascend to catch sunlight during the day and descend during the night to capture nutrients (or prey) which are present in the lower water layers. This ability allows a competitive advantage to non-motile genera but results in a sensitivity to turbulence (Carty, 2014). Since there are no genera present without flagella there is no concise way to compare this sensitivity. Furthermore, turbulence is sometimes needed for cyst dispersion (Mertens et al., 2012). In table 2 the trophic strategy of each genus is described, this can be an indication for which predictors are relevant. Certain genera, such as Ceratium, Cryptomonas and Plagioselmis, are common in eutrophic lakes (Saad et al., 2016) and are therefore expected to positively correlate with increasing nutrient concentrations. Dinobryon, on the other hand, favors oligotrophic lakes (Saad et al., 2016) and is therefore expected to correlate negatively with increasing nutrient concentrations.
Table 2; functional traits and trophic strategy of mixotrophic genera
1 Rost et al., 2006 2 Smalley et al., 1999 3 Guiry & Guiry, 2021 4 Smalley and Coats, 2005 5 Carty, 2014
6 Reynolds et al., 2002
7 Kimura and Ishida 1986 8 Nygaard and Tobiesen 1993 9 Rothhaupt, 1996
10 Urabe et al. 1999 11 Tranvik et al., 1989 12 Urabe et al., 2000
13 Saad et al., 2016 14 Bird & Kalff, 1987 15 Maberly et al., 2009 16 Caron et al., 1993 17 Liu et al., 2016
18 Colman & Balkos, 2005
19 Britannica, 2020 20 Ogbonna et al., 2002 21 Wolken, 1967
Genera CCM Flagella Autotrophy vs heterotrophy Trophic strategy Genus
size Ceratium (hirundinella) 123 Unknown, present in marine C. lineatum 2 flagella of different lengths
Primarily autotrophic 4 Captures prey with pseudopod and retracts into theca 5. Competitive advantage in eutrophic lakes or lakes with low carbon concentrations 6. Forms blooms 5
20-200 um length
Cryptomonas 3 Unknown Two unequal
flagella.
Obligate phototrophy for carbon an energy and essential nutrients from phagotrophy 12
Phagotrophy only under low light condition and is only used for nutrients instead of energy and carbon 789 101112 Inorganic carbon is therefore essential. Sensitive to other phagotrophs 6. High abundance in eutrophic lakes 13
40 um
Dinobryon 314 No 15 Yes, two
flagella
Obligate phototrophy supplemental phagotrophy, high bacterivory 2714. 16 17
Tolerant to very low nutrient concentrations, sensitive to low CO2 concentration 6. Abundant in oligo and mesotrophic lakes with intermediate DOC content, nearly absent in eutrophic lakes 13
20 um length, 10 um width Euglena 318 Not in acidophilic
species E. mutabilis, present in acid tolerant E. gracilis from alkaline waters
Yes, two of unequal length. Uses eyespot to swim toward light 3. Primarily heterotrophic 19, high light intensities inhibit photoheterotrophic growth 20
Favors OM rich waters, can change cell chemistry, cell morphology depending on light or dark conditions 21 15-500 um length 8-18 um width
22 Berman-Frank et al., 2008 23 Raven et al., 2020 24 Porter, 1988 25 Rogers, 2013 26 Xu et al., 2016 27 Gerea et al., 2019 28 Saad et al., 2016
Peridinium3222324 Unknown, present in marine P. gatunense
Usually two flagella but also
nonmotile.
Primarily autotrophic Most species are capable of phagotrophy 25. P. umbonatum is known to cause harmful blooms 26. Found in stratified lakes, often mesotrophic 6
30-70 um Plagioselmis nannoplanctica (Rhodomonas minuta) 3
Unknown Yes, acute
tail.
Low contribution of phagotrophy to growth 27
High prey ingestion rates and grazing impact, 28. Highest proportion in meso/eutrophic lakes13
10 um diameter
Trachelomonas 3 Unknown Yes, two of unequal length
Found in acidic or neutral freshwater, with high concentrations of manganese and iron. Found in shallow waters usually mesotrophic 6
5-100um
Predictors
The transformation of the variables can be seen plotted as a histogram in Appendix 2 and as Shapiro-Wilk and p values in . The following variables showed high collinearity and were removed as predictors: nitrate/nitrite, nitrogen, conductivity, magnesium, sodium and secchi. The data transformed by the bestNormalize function approach normality significantly better. With some variable even obtaining near full normality. For this the following variables were removed: nitrate/nitrite, sodium, conductivity, secchi, nitrogen and magnesium.
Phytoplankton was used as prey variable, however, a relation between prey and chlorophyll is expected since both are higher in high light and nutrient environments. In Figure 3 prey biovolume is plotted against chlorophyll, here it can be seen that there is a relation present, though not very strong. Similarly, for zooplankton a relation with prey was tested. In Figure 4 zooplankton density is plotted against prey biovolume, again a weak relation can be seen.
Variable interaction
For all genera between 25% and 50% of the variation was explained by the variables, all models were significant with a p value of < 2.2e-16, though not all predictors were significant. Interaction terms, estimates, p values and standard error for each predictor and intercept can be found in Appendix , When using the linear model after data transformation an intercept of around zero should be expected, because the data is transformed in such a way that the data collects around zero. The probability cannot be zero, because it would mean there is no intercept, and therefore often no significant p values will be found. The main results are summarized in Table 3 and Table 4.
When power is mentioned, it refers to the random forest results which can be found in the data repository as well as the full models.
Chlorophyll, phosphorus and prey were the main significant predictors for Ceratium. Turbidity, contrarily, showed relatively high power, however, when looking at the full model turbidity had a moderate effect (-0.31) with a high standard error (0.19) and high p value (0.11). Chlorophyll, prey biovolume and temperature are important predictors for
Cryptomonas. In Figure 5 and Figure 6 a positive relation can be seen with chlorophyll and prey, however, a negative estimate (-0.06, p = 0.05) is found for the
interaction term between chlorophyll and prey on the biovolume of Cryptomonas. This shows, that even though
Figure 4: Zooplankton density plotted against prey biovolume, t = 4.43, df = 512, p-value = 1.15e-05, r = 0.19
Figure 5; Cryptomonas biovolume against chlorophyll concentrations t = 14.021, df = 436, p-value < 2.2e-16, r = 0.5574724
Figure 3: prey biovolume plotted against chlorophyll, t = 8.80, df = 512, p-value < 2.2e-16, r = 0.36
both positively interact with Cryptomonas abundance, the combination dampens the effect. This interaction is visualized in Figure 7. Turbidity, prey, zooplankton and temperature influence Dinobryon abundance. A strong relation can be seen between Dinobryon biovolume and prey biovolume. Even though zooplankton shows high power and a moderate effect on biovolume, it did not end up in the BIC model, when looking at the AIC model a relatively high p value (0.034) can be seen, which explain this difference. Chlorophyll, turbidity and depth are the main predictors for Euglena. Phosphorus can only be seen in the AIC model, though insignificantly. The chlorophyl*turbidity interaction enhances the effect on biovolume, and turbidity*phosphorus dampens the effect. Chlorophyll, phosphorus, zooplankton and DOC are the main predictors for Peridinium. Zooplankton showed high power as an predictor but was not included in the BIC model, likely due to the low estimate (0.09) and relatively high p value in the BIC model (0.058). Turbidity, phosphorus and prey are the main predictors for Plagioselmis. However, a negative relation with pH (0.03) can be found in the AIC model of the standard transformed data along with a negative relation with prey (-0.21). However, both estimates are insignificant (0.52 and 0.61). All other model show a positive significant relation with prey (0.27, p = 2.48e-10; bestNormalize 0.21, p = 7.21e-06 and 0.29, p = 4.58e-11). Chlorophyll, prey, depth and oxygen are the main predictors for Trachelomonas. Phosphorus is an additional predictor in the AIC model but is insignificant (p = 0.11).
Figure 6; Cryptomonas biovolume against prey biovolume concentrations t = 8.07, df = 436, p-value = 7.13e-15, r =0.36
Figure 7; Cryptomonas biovolume against chlorophyll prey interaction t = 2.58, df = 436, p-value = 0.010, r = 0.12
Table 3; Each genus with its most important predictor variable, using log and z transformation. Run in two models, the first one for the model as a result of BIC model inferencing, for which all variables were significant, and the second was a result of the AIC model inferencing with the highest R squared but often insignificant estimates. Values in cells are estimates for the transformed data (log and z), predictor variables in red had insignificant p values. Because most models also included interaction terms, the R squared in the table might differ while the variables are the same.
Genus Model
selection
Intercept Chl Turb Phosp Prey Depth DOC O2 PH Temp Zoo R squared
Ceratium BIC 0.09 0.44 0.21 0.23 0.45 AIC 0.09 0.44 0.21 0.23 0.45 Cryptomonas BIC 0.15 0.47 0.19 -0.08 0.33 AIC 0.18 0.40 0.03 0.05 0.20 -0.09 0.34 Dinobryon BIC 0.10 0.25 0.36 -0.17 0.24 AIC 0.14 0.19 0.36 -0.23 0.13 0.28 Euglena BIC -0.01 0.57 -0.17 0.46 AIC -0.06 0.50 0.17 -0.13 -0.14 0.49 Peridinium BIC 0.03 0.49 -0.16 0.14 0.24 AIC -0.01 0.47 -0.19 0.12 0.09 0.25 Plagioselmis BIC 0.05 0.15 0.14 0.27 0.23 AIC -0.16 0.13 0.19 -0.21 0.03 0.25 Trachelomonas BIC 0.14 0.36 0.14 -0.18 -0.12 0.33 AIC 0.14 0.30 0.09 0.14 -0.15 -0.12 0.33
Table 4; Each genus with its most important predictor variable, for data transformed by bestNormalize. Run in two models, the first one for the model as a result of BIC model inferencing, for which all variables were significant, and the second was a result of the AIC model inferencing with the highest R squared but often insignificant estimates. Predictor variables in red had insignificant p values. Cells in orange show that different predictor variables were selected compared to the log transformed data, cells in green show a higher R squared compared to the normal transformation and cells in blue a lower R squared. Because most models also included interaction terms, the R squared in the table might differ while the variables are the same.
Genus Model
selection
Intercept Chl Turb Phosp Prey Depth DOC O2 PH Temp Zoo R squared
Ceratium BIC 0.10 0.46 0.23 0.20 0.50 AIC 0.10 0.46 0.23 0.20 0.50 Cryptomonas BIC 0.15 0.49 0.19 -0.13 0.32 AIC 0.17 0.41 0.04 0.06 0.20 -0.14 0.35 Dinobryon BIC 0.15 0.17 0.36 -0.25 0.16 0.27 AIC 0.15 0.19 0.36 -0.25 0.15 0.29 Euglena BIC -0.01 0.48 0.14 -0.15 0.48 AIC -0.05 0.49 0.23 -0.11 -0.16 0.50 Peridinium BIC 0.03 0.38 0.12 0.23 AIC 0.00 0.44 -0.15 0.12 0.10 0.25 Plagioselmis BIC 0.04 0.15 0.15 0.29 0.24 AIC 0.01 0.18 0.10 0.21 0.02 0.27 Trachelomonas BIC 0.14 0.36 0.15 -0.18 -0.13 0.33 AIC 0.17 0.30 0.10 0.15 -0.16 -0.12 0.34
Discussion
Due to the nature of this data analysis, the results indicate a relation, but no causation can be proved. Laboratory research is needed to prove any of the relations found here. However, the result does provide an indication as to which variable influence the response and to which extent. Furthermore, when using a coarse classification, such as genera, differences between species will be overlooked. However, using coarse level classification, such as genera instead of species, in research will make results more comprehensible to people untrained in phycology (the science of algae) and aid policy makers (Sodré et al., 2020). Lastly, Kurk et al. (2010) showed that phytoplankton with flagella are often grazed on less by zooplankton and are better at nutrient foraging, this can be seen in the low correlation with zooplankton and nutrients.
Ceratium
The reason for the absence of other predictors but chlorophyll, phosphorus and prey possibly lie with the morphology of Ceratium, this species is known to migrate through the water column to benefit from the higher nutrient concentrations at the lower layers, and increasing light in the top layers (Meichtry de Zaburlín et al., 2016). Ceratium’s medium size and low density also helps to reduce sinking, which together with the flagella allows for optimal movement through the column (Kruk et al., 2010). If Ceratium can take up nutrients at levels, where the other genera fail, it might explain the correlation with prey biovolume and phosphorus but not with turbidity. In this case, the genera would likely be sensitive to turbulence. There are multiple reasons why turbulence is not a significant predictor for Ceratium. Since turbulence is not a variable present in the dataset, depth is used as an indicator. However, the relation between turbulence and depth is not always constant which makes it difficult to conclude the absence of this relationship. Furthermore, if Ceratium is mostly present in shallow lakes, the necessary number of observations for deeper lakes might just not be available.
Cryptomonas
Even though Cryptomonas is commonly thought of as an obligate phototroph, the genus still depends on phagotrophy for essential nutrients (Urabe et al., 2000). This research has shown that Cryptomonas also might use phytoplankton as prey, whereas most literature suggest bacterivory as main phagotrophic strategy (Nygaard & Tobiesen, 1993; Tranvik, et al., 1989). Preceding research found that Cryptomonas has indeed a high presence in phytoplankton turbid lakes (Saad, et al., 2016). Similarly to Ceratium, Cryptomonas is also of a medium size and relatively low density, which reduces sinking and enable optimal mobility over the water column (Kruk et al., 2010). High chlorophyll concentrations are an indicator of high phytoplankton presence, which again can indicate a direct relation to Cryptomonas or be an indirect indicator of prey concentrations (t = 8.8046, df = 512, p-value < 2.2e-16, r = 0.36 chlorophyll-prey). Other predictors were less significant, this might be a sign that Cryptomonas is less dependent on light intensity, and less sensitive to turbulence than the other genera used here, or that Cryptomonas is not present in large quantities in deeper lakes. Lastly, temperature shows a negative correlation with Cryptomonas abundance. Even though generally, the genus increases with increasing light and temperature. At low light (e.g. turbid
lakes) growth of Cryptomonas was found to be extremely limited at higher temperature (>20°C) and was sometimes even negative (Wirth et al., 2018).
Dinobryon
Chlorophyll was not taken up in this model, Dinobryon is often found in oligotrophic lakes and very rarely in lakes with high nutrient concentrations (Saad et al., 2016). Chlorophyll is an indicators of high phytoplankton biovolume and therefore meso/eutrophic lakes (Canfield et al., 2019). It is not unlikely that the chlorophyll concentrations are low and constant in lakes in which Dinobryon thrives, that no significant relation can be found (t = 3.6069, df = 295, p-value = 0.0003638, r = 0.21). Furthermore, turbidity showed to be relevant to Dinobryon, however, since Dinobryon is an obligate phototroph, a negative relation with turbidity is expected. The relation found here (estimate=0.11-0.25, p < 0.05) is positive. A possible explanation for this phenomenon
is the reduced competition with phytoplankton, which depends even more on light then Dinobryon does. This way less nutrient are taken up, and until a certain threshold, enough light is present to sustain growth. Temperature showed a negative relation with Dinobryon, which indicates that Dinobryon prefers colder waters. This corresponds to what is found in literature, Heinze et al. (2013) showed that Dinobryon was abundant at a temperature range of 9-18 degrees Celsius, at higher temperatures the abundances dropped. When looking at the interaction terms, a negative correlation with temp-turb (-0.15) is found, which can be explained by the fact that under low light conditions temperature does not benefit photosynthesis anymore (Heinze et al., 2012). Lastly, zooplankton was present as a predictor for Dinobryon with a positive correlation in three models. Dinobryon is not high-quality food for zooplankton (Vad et al., 2019) and the presence of flagella often make phytoplankton less suitable prey (Kurk et al., 2010), explaining the lack of a negative relation. However, Dinobryon and zooplankton can thrive under similar conditions: high nutrient and prey concentration, which accounts for the positive relation.
Euglena
Euglena is primarily heterotrophic, and high light intensities slow down growth (Ogbonna et al., 2002). This can be seen by the positive relation between Euglena abundance and turbidity. Euglena uses the eyespot to direct itself toward the light, therefore the negative relation with depth could be explained by the fact that lakes often become more turbulent with depth. Depth makes positioning more difficult for Euglena which is essential for moving through the water column. This
hypothesis is supported by the shape of Euglena, which is in practice and in proportion the longest and smallest genera used in this research (Figure 9), indicating that positioning for this plankton might be more essential than for the other genera.
Figure 8; Dinobryon biovolukme plotted against turbidity, t = 5.0543, df = 295, p-value = 7.592e-07, r = 0.2823037
Peridinium
Notably phosphorus appeared with a negative relation in the Peridinium model. Peridinium cysts are sensitive to eutrophication, which might explain the negative relation seen here (McCarthy et al., 2011) and Peridinium naturally occurs in mesotrophic lakes (Carty, 2014), this might explain the positive relation with chlorophyll and DOC but a negative relation with phosphorus, since an increase is expected from low to medium nutrient concentrations and a decrease from medium to high concentrations.
Plagioselmis
Plagioselmis depends heavily on phototrophy to grow, turbid lakes limit photosynthesis and therefore growth. The model, however, resulted in a positive correlation with turbidity. Since this species is often abundant is meso/eutrophic lakes, similarly to Ceratium and Cryptomonas this genus has found a way to find an advantage in turbidity. Plagioselmis shows a negative correlation with prey, in the large model with chlorophyll, pH, phosphorus and turbidity, but a positive correlation with prey in the smaller model with turbidity and phosphorus. For data transformed with bestNormalize they are both positive. When looking at the plot the negative relation can be understood, even though the data does show a very clear direction, which is less obvious in the estimates, which are 0.27, -0.21, 0.29 and 0.21. When looking at the plot (Figure 10), a positive relation seems more likely.
Trachelomonas
Chlorophyll, depth, oxygen and prey were included in both models, additionally phosphorus was included in the bigger model (not significant). Trachelomonas is often found in shallow mesotrophic lakes, this explains the positive correlation with chlorophyll and prey and negative correlation with depth. The negative correlation with oxygen is unexpected. Trachelomonas prefers high oxygen concentrations and abundance decreases when the water is deoxygenated (Grabowska & Wolowski, 2014). Deoxygenation takes places when lakes become turbid and not enough light infiltrates the column and not enough oxygen is produced to compensate for oxygen usage (Lee & Jones-Lee, 1983). The phytoplankton samples are taken from May to September, usually oxygen concentrations are lower at the end of the season because during algae blooms less light infiltrates the column. Trachelomonas occurs often during the bloom of Planktothrix agardhii which can cause de-oxygenation (Grabowska & Wolowski, 2014), this response of the Trachelomonas to the deoxygenation might be slightly delayed causing the relation to appear negative. However, in order to fully understand the relationship between Trachelomonas and oxygen concentrations, laboratory research is needed.
Figure 10; Plagioselmis biovolume plotted against prey biovolume as predictor. t = 10.04, df = 503, p-value < 2.2e-16, r = 0.4085775
Relevance
Mixotrophs are capable of taking up nutrients beneath the critical threshold needed for effective growth (Naselli-Flores & Barone, 2019). This ability makes mixotrophs the perfect candidate for processes surrounding wastewater utilization and biodiesel production, alongside the often higher growth rates compared to exclusive auto- or heterotrophs (Wang et al., 2014) (Zhu et al., 2017) (Zhang et al., 2021). The result of this research can be used to better understand and implement mixotrophs in these processes. Mixotrophs can be used, like phytoplankton, as extractive species to reduce eutrophication. Furthermore, plankton are often low in the trophic level and have short life cycles, this means they can be rapid responders to changes in the environment (OSPAR, 2018). A change, from a stable system to a system which is dominated by other species, is often an indicator of a change in nutrients, temperature, pH, salinity or turbidity. Therefore, possible inference can be made from plankton species abundance about the environmental state of the lake, resulting in pre-emptive recognition of conditional changes (Hart & Wragg, 2009). Lastly, mixotrophic plankton are represented in higher numbers in established oligotrophic and eutrophic systems, where they can stabilize the food system by reducing excessive harmful algal/bacterial blooms and dampen predator-prey relationships. Examples of reduction of harmful bacterial blooms by mixotrophy was investigated by Wilken et al. (2014), this research found that the toxic cyanobacteria Microcytis aeruginosa could be suppressed by the mixotroph Ochromonas, though most efficiently at low nitrogen concentrations.
Further research
Due to the low predictability of the models, further research is needed to truly understand the variables that drive mixotrophic plankton abundance. Is it merely due to the spatial distribution of these genera or do other factors such as climate, fluctuations in temperature or nutrient concentration play a role? Phytoplankton have different tolerations for O2, CO2 and light intensity per species, but also depending on the location and time of the year (Clegg, 2007), especially since phytoplankton are seasonal bloomers (Carty, 2014). In order to eliminate some of these options, subsequent research could study temporal data which allows the change over time to be considered. Furthermore, data on bacterial abundance could help understand the prey specificity better. In order to confirm whether species which correlate with phytoplankton as prey, ingest the phytoplankton, laboratory experiments are needed.
Conclusion
Mixotroph abundance is driven by different variables based on the specific trophic strategy of the genera. Genera which naturally thrive in eutrophic lakes show a positive correlation with chlorophyll and phosphorus concentrations. Mixotrophs which rely on phagotrophy for essential nutrients show a positive correlation with prey concentrations. Genera which naturally occurs in shallow or stratified lakes show a negative correlation with lakes depth. Some main findings compared to previous research are: Cryptomonas also use phytoplankton as prey, Plagioselmis abundance is influenced by prey concentrations and Dinobryon, Euglena and Plagioselmis thrive in turbid lakes. Reducing oxygen concentrations might have less negative impact on Trachelomonas then previously thought.
References
Barton, K (2020). MuMIn: Multi-Model Inference. R package version 1.43.17. https://CRAN.R-project.org/package=MuMIn
Berman‐Frank, I., Kaplan, A., Zohary, T., & Dubinsky, Z. (1995). Carbonic anhydrase activity in the bloom‐forming dinoflagellate Peridinium gatunense. Journal of Phycology, 31(6), 906-913.
Bird, D. F., & Kalff, J. (1987). Algal phagotrophy: regulating factors and importance relative to photosynthesis in Dinobryon (chrysophyceae) 1. Limnology and Oceanography, 32(2), 277-284.
Brierley, A. S. (2017). Plankton. Current Biology, 27(11), R478-R483.
Britannica, T. Editors of Encyclopaedia (2019, November 12). Protist. Encyclopedia Britannica.
https://www.britannica.com/science/protist
Britannica, T. Editors of Encyclopaedia (2020, May 5). Euglena. Encyclopedia Britannica.
https://www.britannica.com/science/Euglena
Bruun, K. (2013, 17 September). Rhodomonas minuta Skuja [Photo]. AlgaeBase.
https://www.algaebase.org/search/species/detail/?species_id=35730
Canfield Jr, D. E., Bachmann, R. W., Hoyer, M. V., Johansson, L. S., Søndergaard, M., & Jeppesen, E. (2019). To measure chlorophyll or phytoplankton biovolume: an aquatic conundrum with implications for the management of lakes. Lake and Reservoir Management, 35(2), 181-192.
Caron, D. A., Sanders, R. W., Lim, E. L., Marrasé, C., Amaral, L. A., Whitney, S., ... & Porters, K. G. (1993). Light-dependent phagotrophy in the freshwater mixotrophic chrysophyte Dinobryon cylindricum. Microbial ecology, 25(1), 93-111.
Carty, S. (2014). Freshwater Dinoflagellates of North America. Cornell University Press.
Clegg, M. R., Maberly, S. C., & Jones, R. I. (2007). Behavioral response as a predictor of seasonal depth distribution and vertical niche separation in freshwater phytoplanktonic flagellates. Limnology and Oceanography, 52(1), 441-455.
Colman, B., & Balkos, K. D. (2005). Mechanisms of inorganic carbon acquisition in two Euglena species. Canadian Journal of Botany, 83(7), 865-871.
Dodds, W. K., & Whiles, M. R. (2010). Freshwater Ecology (2de edition). Amsterdam University Press.
https://doi.org/10.1016/C2009-0-01718-8
2.0.1. https://CRAN.R-project.org/package=ggRandomForests
Fucikova, K. (2016, 24 april). Ceratium hirundinella [Photo]. iNaturalist.
https://www.inaturalist.org/observations/3022141
Gerea, M., Queimaliños, C., & Unrein, F. (2019). Grazing impact and prey selectivity of picoplanktonic cells by mixotrophic flagellates in oligotrophic lakes. Hydrobiologia, 831(1), 5-21.
Grabowska, M., & Wołowski, K. (2014). Development of Trachelomonas species (Euglenophyta) during blooming of Planktothrix agardhii (Cyanoprokaryota). In Annales de Limnologie-International Journal of Limnology (Vol. 50, No. 1, pp. 49-57). EDP Sciences.
Hansson, T. H., Grossart, H. P., del Giorgio, P. A., St‐Gelais, N. F., & Beisner, B. E. (2019). Environmental drivers of mixotrophs in boreal lakes. Limnology and Oceanography, 64(4), 1688-1705.
Hart, R. C., & Wragg, P. D. (2009). Recent blooms of the dinoflagellate Ceratium in Albert Falls Dam (KZN): History, causes, spatial features and impacts on a reservoir ecosystem and its zooplankton. Water SA, 35(4).
Heinze, A. W., Truesdale, C. L., DeVaul, S. B., Swinden, J. & Sanders, R. W. 2013. Role of temperature in growth, feeding, and vertical distribution of the mixotrophic chrysophyte Dinobryon. Aquat. Microb. Ecol. 71:155-63.
Ishwaran H. and Kogalur U.B. (2021). Fast Unified Random Forests for Survival, Regression, and Classification (RF-SRC), R package version 2.11.0.
Jones, R. I. (2001). Mixotrophy in planktonic protists: an overview. Freshwater biology, 45(2), 219-226. Kimura, B. & Ishida, Y. (1986). Possible phagotrophic feeding of bacteria in a freshwater red tide Chrysophyceae Uroglena americana. 日本水産学会誌, 52(4), 697-701.
Kruk, C., Huszar, V. L., Peeters, E. T., Bonilla, S., Costa, L., Lürling, M., ... & Scheffer, M. (2010). A morphological classification capturing functional variation in phytoplankton. Freshwater biology, 55(3), 614-627.
Lee, G. F., & Jones-Lee, A. (1983). Mechanisms of the deoxygenation of the hypolimnia of lakes. In in Proc. Water Pollut. Control Fed. Symposium on Sediment Oxygen Demand.–1983 b: Impact of Eutrophication on Fisheries.—Prepared for Amer. Fish. Soc. Submitted for publication.
Liu, Z., Campbell, V., Heidelberg, K. B., & Caron, D. A. (2016). Gene expression characterizes different nutritional strategies among three mixotrophic protists. FEMS microbiology ecology, 92(7), fiw106. Maberly, S. C., Ball, L. A., Raven, J. A., & Sültemeyer, D. (2009). INORGANIC CARBON ACQUISITION BY CHRYSOPHYTES 1. Journal of phycology, 45(5), 1052-1061.
McCarthy, F. M., Mertens, K. N., Ellegaard, M., Sherman, K., Pospelova, V., Ribeiro, S., ... & Vercauteren, D. (2011). Resting cysts of freshwater dinoflagellates in southeastern Georgian Bay (Lake Huron) as proxies of cultural eutrophication. Review of Palaeobotany and Palynology, 166(1-2), 46-62.
M.D. Guiry in Guiry, M.D. & Guiry, G.M. 2021. AlgaeBase. World-wide electronic publication, National University of Ireland, Galway. http://www.algaebase.org; searched on 26 March 2021. https://www.algaebase.org/search/genus/detail/?genus_id=43626&sk=0
M.D. Guiry in Guiry, M.D. & Guiry, G.M. 2021. AlgaeBase. World-wide electronic publication, National University of Ireland, Galway. http://www.algaebase.org; searched on 04 May 2021. https://www.algaebase.org/search/genus/detail/?genus_id=43657&
M.D. Guiry in Guiry, M.D. & Guiry, G.M. 2021. AlgaeBase. World-wide electronic publication, National University of Ireland, Galway. http://www.algaebase.org; searched on 04 May 2021. https://www.algaebase.org/search/genus/detail/?genus_id=43651&sk=0
M.D. Guiry in Guiry, M.D. & Guiry, G.M. 2021. AlgaeBase. World-wide electronic publication, National University of Ireland, Galway. http://www.algaebase.org; searched on 04 May 2021.
Meichtry de Zaburlín, N., Vogler, R. E., Molina, M. J., & Llano, V. M. (2016). Potential distribution of the invasive freshwater dinoflagellate Ceratium furcoides (Levander) Langhans (Dinophyta) in South America. Journal of Phycology, 52(2), 200-208.
Mertens, K. N., Rengefors, K., Moestrup, Ø., & Ellegaard, M. (2012). A review of recent freshwater dinoflagellate cysts: taxonomy, phylogeny, ecology and palaeocology. Phycologia, 51(6), 612-619. Mitra, A., Flynn, K. J., Burkholder, J. M., Berge, T., Calbet, A., Raven, J. A., ... & Zubkov, M. V. (2014). The role of mixotrophic protists in the biological carbon pump. Biogeosciences, 11(4), 995-1005.
Mitra, A., Flynn, K. J., Tillmann, U., Raven, J. A., Caron, D., Stoecker, D. K., ... & Lundgren, V. (2016). Defining planktonic protist functional groups on mechanisms for energy and nutrient acquisition: incorporation of diverse mixotrophic strategies. Protist, 167(2), 106-120.
Moresco, G. A., Bortolini, J. C., Rodrigues, L. C., Jati, S., & Machado Velho, L. F. (2020). A functional deconstructive approach to mixotrophic phytoplankton responds better to local, regional and biogeographic predictors than species. Austral Ecology, 45(2), 249-263.
Morgan, K. C., & Kalff, J. (1979). Effect of light and temperature interactions on growth of Cryptomonas erosa (Cryptophyceae) 1. Journal of Phycology, 15(2), 127-134.
Naimi B, Hamm Na, Groen TA, Skidmore AK, Toxopeus AG (2014). “Where is positional uncertainty a problem for species distribution modelling.” _Ecography_, *37*, 191-203. doi: 10.1111/j.1600-0587.2013.00205.x (URL: https://doi.org/10.1111/j.1600-0587.2013.00205.x)
Naselli-Flores, L., & Barone, R. (2019). Mixotrophic phytoplankton dynamics in a shallow Mediterranean water body: how to make a virtue out of necessity. Hydrobiologia, 831(1), 33-41.
NIES. (2017). Dinobryon divergens 284 [Photo]. Algae Resource Database. https://shigen.nig.ac.jp/algae/imageListQueryAction.do;jsessionid=C617203EE9F753C1599ED974CEA3C8C1?pho toType=1&className=Chrysophyceae
Nygaard, K., & Tobiesen, A. (1993). Bacterivory in algae: a survival strategy during nutrient limitation. Limnology and Oceanography, 38(2), 273-279.
Ogbonna, J., Ichige, E., & Tanaka, H. (2002). Interactions between photoautotrophic and heterotrophic metabolism in photoheterotrophic cultures of Euglena gracilis. Applied microbiology and biotechnology, 58(4), 532-538.
OSPAR. (2018, 16 May). Changes in Phytoplankton Biomass and Zooplankton Abundance.
https://oap.ospar.org/en/ospar-assessments/intermediate-assessment-2017/biodiversity-status/habitats/plankton-biomass/
Oyadomari, J. (2010, 27 mei). Euglena acus [Photo]. Keweenaw Algae.
https://www.keweenawalgae.mtu.edu/gallery_pages/euglenoids.htm
Plewka, M. (2011, 14 April). Ovaler Schlundflagellat (Cryptomonas ovata) [Photo].
http://www.limnolife.de/Science/Atlas/Kennkarten%20Algen/AndereAlgen/Source/Cryptomonas%20ova ta.html.
Plewka, M. (2018). Trachelomonas ellipsoidalis [Photo].
https://www.plingfactory.de/Science/Atlas/Kennkarten%20Algen/01_e-algae/Euglenophyceae/e-source/Trachelomonas%20ellipsoidalis.html
Porter, K.G. Phagotrophic phytoflagellates in microbial food webs. Hydrobiologia 159, 89–97 (1988). https://doi.org/10.1007/BF00007370
Raven, J. A. (1995). Photosynthetic and non-photosynthetic roles of carbonic anhydrase in algae and cyanobacteria. Phycologia, 34(2), 93-101.
Raven, J. A., Suggett, D. J., & Giordano, M. (2020). Inorganic carbon concentrating mechanisms in free‐ living and symbiotic dinoflagellates and chromerids. Journal of phycology, 56(6), 1377-1397.
Reynolds, C. S., Huszar, V., Kruk, C., Naselli-Flores, L., & Melo, S. (2002). Towards a functional classification of the freshwater phytoplankton. Journal of plankton research, 24(5), 417-428.
Rogers, K. (2013, 19 July). Peridinium | Characteristics, Classification, & Facts. Encyclopedia Britannica.
Rost, B., RICHTER, K. U., Riebesell, U. L. F., & Hansen, P. J. (2006). Inorganic carbon acquisition in red tide dinoflagellates. Plant, Cell & Environment, 29(5), 810-822.
Rothhaupt, K. O. (1996). Utilization of substitutable carbon and phosphorus sources by the mixotrophic chrysophyte Ochromonas sp. Ecology, 77(3), 706-715.
Ryan A. Peterson (2019). Ordered quantile normalization: a semiparametric transformation built for the cross-validation era. Journal of Applied Statistics, 1-16.
RStudio Team. (2021). RStudio: Integrated Development Environment for R. Boston, MA. Retrieved from http://www.rstudio.com/
Saad, J. F., Unrein, F., Tribelli, P. M., López, N., & Izaguirre, I. (2016). Influence of lake trophic conditions on the dominant mixotrophic algal assemblages. Journal of Plankton Research, 38(4), 818-829.
Smalley, G. W., Coats, D. W., & Stoecker, D. K. (2003). Feeding in the mixotrophic dinoflagellate Ceratium furca is influenced by intracellular nutrient concentrations. Marine Ecology Progress Series, 262, 137-151.
Smalley, G. W., & Coats, D. W. (2005). Ecology of the red‐tide dinoflagellate Ceratium furca: distribution, mixotrophy, and grazing impact on ciliate populations of Chesapeake Bay. Journal of Eukaryotic Microbiology, 49(1), 63-73.
Tranvik, L. J., Porter, K. G., & Sieburth, J. M. (1989). Occurrence of bacterivory in Cryptomonas, a common freshwater phytoplankter. Oecologia, 78(4), 473-476.
Urabe, J., Gurung, T. B., Yoshida, T., Sekino, T., Nakanishi, M., Maruo, M., & Nakayama, E. (2000). Diel changes in phagotrophy by Cryptomonas in Lake Biwa. Limnology and oceanography, 45(7), 1558-1563. US EPA. 2021. National Lakes Assessment | US EPA. [online] Available at: <https://www.epa.gov/national-aquatic-resource-surveys/nla> [Accessed 22 April 2021].
US EPA. (2020). National Lakes Assessment [Photo]. US EPA. https://www.epa.gov/national-aquatic-resource-surveys/nla
U.S. Environmental Protection Agency. [2016]. National Aquatic Resource Surveys. [National Lakes Assessment 2012] (data and metadata files). Available from U.S. EPA web page: https://www.epa.gov/national-aquatic-resource-surveys/data-national-aquatic-resource-surveys. Date accessed: 2021-04-22.
Vogt, R. J., St‐Gelais, N. F., Bogard, M. J., Beisner, B. E., & Del Giorgio, P. A. (2017). Surface water CO 2 concentration influences phytoplankton production but not community composition across boreal lakes. Ecology letters, 20(11), 1395-1404.
Wang, J., Yang, H., & Wang, F. (2014). Mixotrophic cultivation of microalgae for biodiesel production: status and prospects. Applied biochemistry and biotechnology, 172(7), 3307-3329.
Wehr, J. D., Sheath, R. G., & Kociolek, J. P. (Eds.). (2015). Freshwater algae of North America: ecology and classification. Elsevier.
Wirth, C., Limberger, R., & Weisse, T. (2019). Temperature× light interaction and tolerance of high water temperature in the planktonic freshwater flagellates Cryptomonas (Cryptophyceae) and Dinobryon (Chrysophyceae). Journal of phycology, 55(2), 404-414.
Wilken, S., Verspagen, J. M., Naus-Wiezer, S., Van Donk, E., & Huisman, J. (2014). Biological control of toxic cyanobacteria by mixotrophic predators: an experimental test of intraguild predation theory. Ecological applications, 24(5), 1235-1249.
Wolken, J. J. (1967). Euglena: an experimental organism for biochemical and biophysical studies. Springer Science & Business Media.
Xu, J., Wu, X., Yang, Y., Xu, S., Kang, Y., Fu, X., ... & Wu, Z. (2016). Changes in growth, photosynthesis and chlorophyll fluorescence in the freshwater dinoflagellate Peridinium umbonatum (Peridiniales, Pyrrophyta) in response to different temperatures. Phycologia, 55(4), 469-477.
Zhang, Z., Sun, D., Cheng, K. W., & Chen, F. (2021). Investigation of carbon and energy metabolic mechanism of mixotrophy in Chromochloris zofingiensis. Biotechnology for biofuels, 14(1), 1-16.
Zhu, L., Nugroho, Y. K., Shakeel, S. R., Li, Z., Martinkauppi, B., & Hiltunen, E. (2017). Using microalgae to produce liquid transportation biodiesel: what is next?. Renewable and Sustainable Energy Reviews, 78, 391-400.
Acknowledgements
I thank Jolanda Verspagen, who has acted as my supervisor and whose feedback sessions were always useful and Susanne Wilken, for acting as examiner. I also want to thank the United States Environmental Protection Agency for supplying their data to be used by all.
Appendices
Appendix 1: Shapiro-Wilk and p values of the transformed data
Table x Shapiro-Wilk and p value for standard transformed data
variable W p_value SECCHI 0.987165 1.16E-05 CHLX_RESULT 0.993561 0.005379 ALUMINUM_RESULT 0.661607 1.52E-34 MAGNESIUM_RESULT 0.994237 0.011304 NITRITE_N_RESULT 0.423034 7.62E-42 NITRATE_N_RESULT 0.640182 2.39E-35 SILICA_RESULT 0.954857 1.5E-13 SODIUM_RESULT 0.972681 6.35E-10 DOC_RESULT 0.99079 0.00031 CALCIUM_RESULT 0.966073 2.07E-11 SULFATE_RESULT 0.985828 3.87E-06 CHLORIDE_RESULT 0.981826 1.91E-07 NITRATE_NITRITE_N_RESUL T 0.640435 2.44E-35 POTASSIUM_RESULT 0.98605 4.63E-06
COND_RESULT 0.988321 3.16E-05 TURB_RESULT 0.971832 3.98E-10 AMMONIA_N_RESULT 0.707082 1.07E-32 TEMPERATURE 0.966352 2.37E-11 OXYGEN 0.985477 2.92E-06 PH 0.986675 7.73E-06 INDEX_SITE_DEPTH 0.976651 6.41E-09 AREA_HA 0.944911 3.63E-15 NTL_RESULT 0.99458 0.016553 PTL_RESULT 0.978291 1.78E-08 pCO2 0.97986 4.96E-08 ZOO_DENSITY 0.984446 1.31E-06 PREY_BIOVOLUME 0.621058 4.94E-36
Table x Shapiro-Wilk and p value for best normalized transformed data
variable W p_value
SECCHI 0.997501 0.403241
ALUMINUM_RESULT 0.858843 3.67E-24 MAGNESIUM_RESULT 0.999881 1 NITRITE_N_RESULT 0.205432 1.43E-46 NITRATE_N_RESULT 0.905064 4.29E-20 SILICA_RESULT 0.987102 1.1E-05 SODIUM_RESULT 0.998266 0.740776 DOC_RESULT 0.995037 0.027675 CALCIUM_RESULT 0.999735 1 SULFATE_RESULT 0.999899 1 CHLORIDE_RESULT 0.999913 1 NITRATE_NITRITE_N_RESULT 0.906497 6.03E-20 POTASSIUM_RESULT 0.985728 3.57E-06 COND_RESULT 0.99991 1 TURB_RESULT 0.998987 0.97473 AMMONIA_N_RESULT 0.999247 0.996285 TEMPERATURE 0.999747 1 OXYGEN 0.985477 2.92E-06
PH 0.996652 0.169321 INDEX_SITE_DEPTH 0.990642 0.000269 AREA_HA 0.992759 0.002281 NTL_RESULT 0.99458 0.016553 PTL_RESULT 0.997293 0.330335 pCO2 0.999925 1 ZOO_DENSITY 0.985738 3.6E-06 PREY_BIOVOLUME 0.941506 1.13E-15
Appendix 3
Ceratium
The most important predictors for Ceratium were chlorophyll, turbidity, DOC and phosphorus. Chlorophyll and turbidity also show high power in predicting the biovolume of Ceratium, whereas DOC and especially phosphorus do not. Prey was added to the model, after showing a significant relation in the full model. By model inferencing two of the best models were selected.
Model choice 1: Highest R squared
AIC weight 0.09 BIC weight 0.402 Multiple R squared 0.447 Adjusted R squared 0.4369 P value < 2.2e-16 Notes: Coefficients: Model 1:
Estimate Std. Error t value Pr(>|t|) (Intercept) 0.08561 0.06056 1.414 0.159339 CHLX_RESULT 0.44104 0.09119 4.837 3.02e-06 *** PREY_BIOVOLUME 0.22924 0.06292 3.643 0.000361 *** PTL_RESULT 0.21114 0.08902 2.372 0.018867 *
The same method was applied to the data transformed by the bestNormalize function.
Model choice 1: Highest R squared
AIC weight 0.04
BIC weight 0.42
Adjusted R squared 0.4929
P value < 2.2e-16
Notes:
Model 1
Estimate Std. Error t value Pr(>|t|) (Intercept) 0.10012 0.05567 1.799 0.073931 . CHLX_RESULT 0.45889 0.08348 5.497 1.45e-07 *** PREY_BIOVOLUME 0.19979 0.05783 3.455 0.000701 *** PTL_RESULT 0.22831 0.08172 2.794 0.005826 **
Cryptomonas
The most important predictors for Cryptomonas were chlorophyll, prey biovolume, turbidity, temperature and phosphorus. Chlorophyll, prey biovolume, phosphorus and turbidity also show high power in predicting the biovolume of Ceratium.
By model inferencing two of the best models were selected.
Model choice 1: Highest R squared 2: Highest weight (BIC)
AIC weight 0.065 0.025
BIC weight 0.00 0.415
Multiple R squared 0.341 0.3295
Adjusted R squared 0.3332 0.3266
P value < 2.2e-16 < 2.2e-16
Notes: Phosphorus and turbidity are not
significant
Coefficients: Model 1:
Estimate Std. Error t value Pr(>|t|) (Intercept) 0.17617 0.03648 4.830 1.6e-6 *** CHLX_RESULT 0.40288 0.05509 7.313 < 2.2e-16*** PREY_BIOVOLUME 0.20345 0.03288 6.188 1.0e-9 *** PTL_RESULT 0.05328 0.04622 1.153 0.24942 TEMPERATURE -0.08979 0.03271 -2.745 0.00620 ** TURB_RESULT 0.02515 0.05165 0.487 0.62644
CHLX_RESULT:PREY_BIOVOLUME -0.06037 0.03065 -1.969 0.04931 * CHLX_RESULT:PTL_RESULT 0.10983 0.04390 2.502 0.01259 * PTL_RESULT:TURB_RESULT -0.10755 0.03809 -2.823 0.00489 ** Model 2:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.15315 0.02974 5.149 3.4e-7 ***
CHLX_RESULT 0.46723 0.03490 13.387 < 2.2e-16***
PREY_BIOVOLUME 0.19247 0.03261 5.903 5.6e-9 ***
TEMPERATURE -0.08418 0.03238 -2.600 0.00953 ** The same method was applied to the data transformed by the bestNormalize function. For this data the most important predictors were chlorophyll, turbidity, phosphorus, temperature and prey biovolume.
Model choice 1: Highest R squared 2: Highest weight (BIC)
AIC weight 0.09 0.009
BIC weight 0.00 0.724
Multiple R squared 0.3519 0.3382
Adjusted R squared 0.3444 0.3353
P value < 2.2e-16 < 2.2e-16
Notes: Phosphorus and
turbidity are not significant
Model 1
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.17200 0.03600 4.778 2.17e-06 *** CHLX_RESULT 0.41236 0.05431 7.592 1.03e-13 *** PREY_BIOVOLUME 0.20422 0.03274 6.238 7.73e-10 *** PTL_RESULT 0.05685 0.04524 1.256 0.20938 TEMPERATURE -0.14060 0.03293 -4.270 2.23e-05 *** TURB_RESULT 0.03804 0.05429 0.701 0.48366 CHLX_RESULT:PREY_BIOVOLUME -0.06267 0.02989 -2.096 0.03641 * CHLX_RESULT:PTL_RESULT 0.12826 0.04491 2.856 0.00442 ** PTL_RESULT:TURB_RESULT -0.12509 0.04110 -3.044 0.00243 ** Model 2
(Intercept) 0.14641 0.02949 4.965 8.64e-07 ***
CHLX_RESULT 0.48721 0.03488 13.968 < 2e-16 ***
PREY_BIOVOLUME 0.19492 0.03227 6.039 2.52e-09 ***
TEMPERATURE -0.13013 0.03240 -4.017 6.55e-05 ***
Dinobryon
The most important predictors for Dinobryon were chlorophyll, zooplankton density and prey biovolume. And to a lesser extent turbidity, phosphorus and lake depth. Prey biovolume and zooplankton also show high power in predicting the biovolume of Ceratium, chlorophyll, turbidity and depth show less power and phosphorus very low power. Depth was removed as a predictor in multiple model inferencing due to its insignificance
By model inferencing two of the best models were selected.
Model choice 1: Highest R squared 2: Highest weight
Weight 0.28 0.28
AIC/BIC AIC BIC
Multiple R squared 0.2782 0.2372
Adjusted R squared 0.2607 0.2294
P value < 2.2e-16 < 2.2e-16
Notes: Turb*zoo is not significant
Coefficients: Model 1:
Estimate Std. Error t value (Intercept) 0.13664 0.05357 2.550 PREY_BIOVOLUME 0.35979 0.05090 7.068 TEMPERATURE -0.23026 0.05532 -4.162 TURB_RESULT 0.19374 0.06294 3.078 ZOO_DENSITY 0.12640 0.05937 2.129 PREY_BIOVOLUME:TURB_RESULT -0.09112 0.04563 -1.997 TEMPERATURE:TURB_RESULT -0.15317 0.05279 -2.902 TURB_RESULT:ZOO_DENSITY 0.09257 0.04855 1.907 Pr(>|t|) (Intercept) 0.01128 * PREY_BIOVOLUME 1.18e-11 *** TEMPERATURE 4.16e-05 *** TURB_RESULT 0.00228 **
ZOO_DENSITY 0.03411 * PREY_BIOVOLUME:TURB_RESULT 0.04677 * TEMPERATURE:TURB_RESULT 0.00400 ** TURB_RESULT:ZOO_DENSITY 0.05755 .
Model 2:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.10160 0.04972 2.043 0.04190 *
PREY_BIOVOLUME 0.35921 0.05167 6.952 2.35e-11 ***
TEMPERATURE -0.17418 0.05385 -3.235 0.00136 **
The same method was applied to the data transformed by the bestNormalize function.
Model choice 1: Highest R squared 2: Highest weight (BIC)
Weight 0.25 0.46
AIC/BIC AIC BIC
Multiple R squared 0.2883 0.2733
Adjusted R squared 0.2711 0.2608
P value < 2.2e-16 < 2.2e-16
Notes: Turb*zoo is not
significant Model 1
Estimate Std. Error t value
(Intercept) 0.14861 0.05502 2.701 PREY_BIOVOLUME 0.36480 0.05110 7.138 TEMPERATURE -0.25100 0.05537 -4.533 TURB_RESULT 0.18985 0.06111 3.107 ZOO_DENSITY 0.14628 0.05903 2.478 PREY_BIOVOLUME:TURB_RESULT -0.09821 0.04589 -2.140 TEMPERATURE:TURB_RESULT -0.15634 0.05263 -2.971 TURB_RESULT:ZOO_DENSITY 0.08828 0.05177 1.705 Pr(>|t|) (Intercept) 0.00732 ** PREY_BIOVOLUME 7.61e-12 *** TEMPERATURE 8.52e-06 *** TURB_RESULT 0.00208 ** ZOO_DENSITY 0.01378 * PREY_BIOVOLUME:TURB_RESULT 0.03319 * TEMPERATURE:TURB_RESULT 0.00322 ** TURB_RESULT:ZOO_DENSITY 0.08922 . Model 2
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.14975 0.05227 2.865 0.00447 ** PREY_BIOVOLUME 0.35505 0.05090 6.976 2.04e-11 *** TEMPERATURE -0.24868 0.05540 -4.489 1.03e-05 *** TURB_RESULT 0.17442 0.06111 2.854 0.00463 ** ZOO_DENSITY 0.15862 0.05889 2.693 0.00748 ** TEMPERATURE:TURB_RESULT -0.11738 0.04695 -2.500 0.01296 *
Euglena
The most important predictors for Euglena were chlorophyll, depth, turbidity and phosphorus, which also show the largest power in predicting.
By model inferencing two of the best models were selected.
Model choice 1: Highest R squared 2: Highest weight 3. Second higest weight and R squared
Weight 0.17 0.60 0.37
AIC/BIC AIC BIC AIC
Multiple R squared 0.4949 0.4632 0.4856
Adjusted R squared 0.481 0.4595 0.4768
P value < 2.2e-16 < 2.2e-16 < 2.2e-16
Notes: Phosphorus, phos*turb
and chlx*phosp*turb were not significant
Higher r squared still significant predictors Coefficients: Model 1: Estimate Std. Error (Intercept) -0.05785 0.04990 CHLX_RESULT 0.49566 0.06671 INDEX_SITE_DEPTH -0.14429 0.04794 PTL_RESULT -0.12598 0.07414 TURB_RESULT 0.17284 0.07448 CHLX_RESULT:PTL_RESULT -0.01203 0.05966 CHLX_RESULT:TURB_RESULT 0.17790 0.05393 PTL_RESULT:TURB_RESULT -0.11322 0.05800 CHLX_RESULT:PTL_RESULT:TURB_RESULT 0.06118 0.03118 t value Pr(>|t|) (Intercept) -1.159 0.24724 CHLX_RESULT 7.430 1.21e-12 *** INDEX_SITE_DEPTH -3.010 0.00284 ** PTL_RESULT -1.699 0.09037 . TURB_RESULT 2.321 0.02099 * CHLX_RESULT:PTL_RESULT -0.202 0.84028 CHLX_RESULT:TURB_RESULT 3.299 0.00109 ** PTL_RESULT:TURB_RESULT -1.952 0.05189 .
CHLX_RESULT:PTL_RESULT:TURB_RESULT 1.962 0.05073 . Model 2:
Estimate Std. Error t value Pr(>|t|) (Intercept) -0.008726 0.041234 -0.212 0.832546 CHLX_RESULT 0.573326 0.045236 12.674 < 2e-16
INDEX_SITE_DEPTH -0.170026 0.045236 -3.759 0.000206
Model 3:
Estimate Std. Error t value (Intercept) -0.03877 0.04893 -0.792 CHLX_RESULT 0.48977 0.06063 8.078 INDEX_SITE_DEPTH -0.12766 0.04661 -2.739 TURB_RESULT 0.17632 0.06589 2.676 CHLX_RESULT:TURB_RESULT 0.15299 0.04808 3.182 TURB_RESULT:PTL_RESULT -0.11791 0.04937 -2.388 Pr(>|t|) (Intercept) 0.42880 CHLX_RESULT 1.72e-14 *** INDEX_SITE_DEPTH 0.00655 ** TURB_RESULT 0.00787 ** CHLX_RESULT:TURB_RESULT 0.00162 ** TURB_RESULT:PTL_RESULT 0.01756 *
The same method was applied to the data transformed by the bestNormalize function.
Model choice 1: Highest R squared 2: Highest weight (BIC) 3. Second highest weight and R squared
Weight 0.29 0.65 0.35
AIC/BIC AIC BIC BIC
Multiple R squared 0.4952 0.4688 0.4767
Adjusted R squared 0.4849 0.4652 0.4714
P value < 2.2e-16 < 2.2e-16 < 2.2e-16
Notes: Phosphorus is not
significant
More variables than 2, still significant
