Phytoplankton species traits have been used to successfully predict the outcome of competition, but these traits are notoriously laborious to measure. If these traits display a phylogenetic signal, phylogenetic distance (PD) can be used as a proxy for trait variation. We provide the first investigation of the degree of phylogenetic signal in traits related to competition in freshwater green phytoplankton. We measured 17 traits related to competition and tested whether they displayed a phylogenetic signal across a molecular phylogeny of 59 species of green algae. We also assessed the fit of five models of trait evolution to trait variation across the phylogeny. There was no significant phylogenetic signal for 13 out of 17 ecological traits. For 7 traits, a non-phylogenetic model provided the best fit. For another 7 traits, a phylogenetic model was selected, but parameter values indicated that trait variation evolved recently, diminishing the importance of common ancestry. This study suggests that traits related to competition in freshwater green algae are not generally well-predicted by patterns of common ancestry. We discuss the mechanisms by which the link between phylogenetic distance and phenotypic differentiation may be broken.
Citation: Narwani A, Alexandrou MA, Herrin J, Vouaux A, Zhou C, Oakley TH, et al. (2015) Common Ancestry Is a Poor Predictor of Competitive Traits in Freshwater Green Algae. PLoS ONE 10(9): e0137085. https://doi.org/10.1371/journal.pone.0137085
Editor: Chih-hao Hsieh, National Taiwan University, TAIWAN
Received: February 17, 2015; Accepted: August 13, 2015; Published: September 8, 2015
Copyright: © 2015 Narwani et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Data Availability: All trait data are within the paper and its Supporting Information files. The phylogeny and all associated sequence data, results and tools have already been published and made available online as a part of that earlier publication (Alexandrou et al. 2015). All phylogenetic datasets and orthologs are available on Dryad (doi:10.5061/dryad.c574h). Raw Illumina files can be accessed using the NCBI SRA archive under BioProject Accession PRJNA237822. The bioinformatics tools can be accessed at http://galaxy-dev.cnsi.ucsb.edu/osiris, downloaded from https://bitbucket.org/osiris_phylogenetics/osiris_phylogenetics, and read about at http://osiris-phylogenetics.blogspot.com.
Funding: This study was supported by the US National Science Foundation’s DIMENSIONS of Biodiversity program grant to BJC and THO (DEB-1046121) (http://www.nsf.gov/funding/pgm_summ.jsp?pims_id=503446). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Phylogenetics has recently been integrated with community ecology to ask whether common ancestry between species, measured as their phylogenetic distance, can help explain phenomena ranging from community assembly [1–4], species invasions , priority effects , and biodiversity-ecosystem functioning relationships [7–10]. The desire to incorporate phylogenetics into community ecology has been based on many factors. First, the increased accessibility of genetic sequence data has made the construction of molecular phylogenies more feasible. Second, the availability of molecular phylogenies has made it possible to rigorously test some old and popular hypotheses, many originated by Darwin, about how species' relatedness should impact the strength of species interactions [3, 4, 11, 12]. Third, the possibility that phylogenetic relationships among species could be used to predict processes and patterns in community ecology would mean that phylogenetic information could be useful in making conservation and management decisions aimed at the protection of biodiversity, ecosystem functioning and ecosystem services [9, 10, 13, 14].
All investigations of the importance of phylogenetic distance as a predictor of ecological interactions and community assembly ultimately rely on the assumption that the traits determining species' ecologies display a phylogenetic signal . When ecological traits do display a phylogenetic signal, patterns of common ancestry can be used as reasonable proxies for species' trait variability. The expectation that closely related species should be more phenotypically similar than distantly related species is intuitive, and it is also a pattern that is predicted by certain null models of trait evolution. Under such models, species accumulate greater phenotypic differences from one another, the longer the time they have had to diverge from one another . However, a number of evolutionary processes can reduce or eliminate phylogenetic signal [17, 18], and recent reviews have shown that the presence of a phylogenetic signal in traits may not be as common as is widely presumed in studies of community phylogenetics [15, 18–21]. So, while phylogenetic approaches may offer great promise in community ecology, it is imperative that we first test the fundamental assumption that traits display a phylogenetic signal in individual systems in which we hope to use this assumption.
In this paper we investigate whether traits that are thought to predict the outcome of competition in freshwater algae display a phylogenetic signal. Algae have long been used as a model system in which to study community assembly and the ecological processes governing biodiversity [22–29]. This interest is partly due to the fact that one can find a great diversity of algal species coexisting in relatively homogenous environments, despite the fact that they compete for a relatively small number of limiting resources (Hutchinson’s famous “paradox of the plankton” ). Previous work, however, indicates that phylogenetic distance among freshwater algae is not a good predictor of the strength of species interactions , the probability two species will coexist , or how polycultures influence ecosystem functions like biomass production . These previous studies showed that the phylogenetic distances among species pairs do not predict the outcome of community assembly and species interactions for this group of algae. Such prior results suggest that traits related to competition and coexistence may not show a phylogenetic signal, but this assumption has never been tested. This is the first direct investigation of how species’ competitive abilities for inorganic resources are distributed across the phylogeny.
Here, we measured species' abilities to compete for limiting resources to test whether an absence of phylogenetic signal of traits related to competition may explain previous findings that phylogenetic distance is a poor predictor of algal ecology. Specifically, we measured species’ minimum resource requirements for light, nitrogen and phosphorus. We also measured species’ elemental content and stoichiometry for carbon, nitrogen and phosphorus because these elemental ratios reflect species’ consumption vectors—or the ratios at which species remove these elements from the environment. Algae are known to compete for these inorganic resources , and the abilities of species to sequester and survive on limiting quantities of these resources have been used to correctly predict the outcome of competition for limiting resources in other algae and plants [25, 26, 33] using resource ratio theory . Lastly, we measured two morphological traits because these traits have been hypothesized to be "master traits" that inherently control other traits related to competition due to allometric scaling relationships [35, 36], and because size confers a defense against grazing .
We used a variety of approaches to test whether phylogenetic distance is related to trait variation among species. First, we estimated the strength and significance of the phylogenetic signal of each trait across the phylogeny. Second, we tested whether phylogenetic distance (PD), measured as the sum of the branch lengths between species pairs [7, 9], was a good predictor of their trait differentiation. Third, we asked whether the data were more consistent with a null, Brownian Motion, model of evolution with other alternative phylogenetic models of trait evolution, or with a non-phylogenetic model of trait-variation. Model fitting cannot be used to infer the mechanisms of evolution (e.g. ecological speciation or adaptive radiation). However, it can be used to support or reject hypotheses about how evolution impacts trait variation over time . Overall, we found algal traits related to competition showed less phylogenetic signal than expected based on a Brownian Motion null model of evolution. Phylogenetic distance was not a good predictor of trait variation, and model comparisons suggest that traits tended to be either randomly distributed with respect to the tree, or consistent with accelerating evolution towards the tips of the tree.
While these results are the first demonstration that a lack of phylogenetic signal may explain previous findings showing that phylogenetic distance among species does not predict species interactions, coexistence, co-occurrence or ecosystem functioning, we acknowledge that phylogenetic distance among species on a tree is not equivalent to the phylogenetic distance of species in a community. Further work must be done to demonstrate that these traits explain community assembly in order to confirm that this lack of phylogenetic signal explains previous findings at the community level.
Materials and Methods
Species selection and resource requirement experiments
We sought to measure traits for 51 of the 59 species on the phylogeny (Table A in S1 File). We selected these species because they were either a) among the 50 most frequently occurring species in lakes across North America, as described in the US EPA's National Lakes Assessment survey for 2007, or b) they occurred in our study lakes at the University of Notre Dame’s Environmental Research Center in the Upper Peninsula of Michigan . This resulted in a list of 95 species which was then further restricted by our ability to c) obtain a culture of each species from culture collections, and d) to culture the species on a single common freshwater medium (COMBO ). These four conditions determined the species in our species pool.
We conducted resource-limitation experiments to determine how algal populations respond to gradients in resource availability, and to estimate species' minimum resource requirements (R*). We estimated species growth rates as a function of resource availability and used a Monod function to estimate the resource level at which the species could not grow (i.e. the zero net growth isocline, ). To do so, we performed three separate experiments in 48-well Falcon Tissue Culture Plates to describe the response of each species’ growth rate to increasing availabilities of nitrate, phosphate and light. Species were inoculated into sterile freshwater COMBO medium  and received 10% media exchanges every other day for 14 days. Well-plates were placed on a shaker table and rotated at 100 rpm and were illuminated on a 18L:6D photoperiod in an environmental chamber that was maintained at 20°C. We measured in situ fluorescence daily using a Biotek Synergy H1 Hybrid Reader, at an excitation wavelength of 435 nm and an emission wavelength of 685 nm to estimate in situ chlorophyll-a . We used chlorophyll-a fluorescence as a proxy for algal biomass , and estimated species growth rates as a function of resource availability by using a Monod function to estimate the resource level at which the species could not grow (i.e. the zero net growth isocline, ).
We estimated growth rates (r) by fitting data on fluorescence values to the exponential time-series: (1) where Ft is the fluorescence on day t, and F0 is the fluorescence on day zero. We did not include time series that showed zero or negative growth in the analysis. Not all time series showed exponential growth dynamics for the entire time series. In many cases species displayed density-dependent growth by the end of the two weeks. In order to obtain the best possible estimate of the exponential portion of growth, we eliminated time points from the end of the time series until the variance explained by the exponential growth curve was maximized—as estimated by the R2 value of the curve. The population growth rates (r) were then compiled across the treatment levels for each species, with each species by resource concentration treatment being replicated twice.
We estimated each species' minimum light requirement and light-use efficiency by subjecting well-plates to 10 different levels of light and measuring each species’ population growth rate. The well-plates were illuminated under AgroBrite T5 fluorescent bulbs. We used neutral density filters (Solar Graphics Inc.) to reduce the total amount of irradiance that the algae receive in the well-plates, without changing the distribution of wavelengths. Light intensities under the filters were measured using a Biospherical Instruments QSL-2100 Radiometer, and were: 0, 1, 7.5, 10, 32.5, 48, 113.5, 133, 162.5, 198.5, 210.5, 237, 248.5, 311.5 μEinstein m-2 s-2.
In order to estimate algal growth requirements for nitrogen, we exposed each species to six increasing molar concentrations of nitrate in the form of NaNO3 (0, 0.2, 2, 20, 100, and 200 μmol L-1). For comparison, standard COMBO contains 1000 μmol L-1 of NaNO3, and the minimum and maximum molar concentrations of total nitrogen in the EPA National Lakes Assessment Survey of 1,157 lakes in 2007 were 0.07 µmol L-1 and 1,863 mol L-1, respectively. Nitrogen was provided solely in the form of NaNO3. All other nutrient concentrations were kept constant. These concentrations of nitrate were chosen based on a previously published resource limitation experiments on freshwater algae , as well as pilot studies in which we estimated minimum resource requirements on a small subset of our selected taxa across a wider gradient of resource availabilities.
Similarly, in order to estimate algal requirements for phosphorus, we employed six different molar concentrations of phosphate, in the form of K2HPO4 (0, 1, 2, 4, 8, and 16 μmol PO4 L-1). Standard COMBO contains 50 μmol L-1 of K2HPO4, and the minimum and maximum molar concentrations of total phosphorus in the EPA National Lakes Assessment Survey were 0.03 μmol L-1 and 157 μmol L-1, respectively. In phosphate removal treatments, we held the availability of potassium constant by replacing the same molar concentration of potassium in the form of KCl. All other nutrient concentrations were maintained as in standard COMBO medium. We soaked the well-plates and pipette tips used in the phosphorus experiment in a 10% HCl acid bath for 24-hrs to eliminate the possibility of phosphate contamination, after which they were rinsed in DI water. For both the nitrate and phosphate experiments the algae were exposed to 100 μEinstein m-2 s-2 of irradiance from Cool-White fluorescent bulbs. Well-plates were randomly assigned a location on the shaker table.
Prior to their inoculation into well-plates, species were grown in batch culture in COMBO medium. In order to remove existing nitrogen and phosphorus from the batch medium for the nitrogen and phosphorus experiments respectively, each culture was centrifuged at 4000 RPM for 15 minutes and rinsed with nitrogen- or phosphorus-free media three times. The algal pellet was then re-suspended in sterile COMBO medium. The algae were kept in nitrogen-free or phosphorus-free medium for the nitrogen and phosphorus experiments, respectively for 72 hours prior to the start of the growth measurements. This nutrient 'starvation-period' was employed to ensure that species would use up any remaining dissolved or stored nutrient resources in their cells before being exposed to the experimental nutrient treatments—i.e. to prevent nutrient contamination. Re-suspended cells were then counted under a compound microscope and either concentrated or diluted to a concentration of 6,000 cells mL-1. Each experimental well contained 900 μL of media and was inoculated with 100 μL of algae to achieve an initial density of 600 cells mL-1. The well-plate lids were sealed with Parafilm to prevent evaporative losses.
We used non-linear least squares regression in R (R Development Core Team 2011) to estimate the parameters α and μmax of the Monod equation: (2) where μ(R) is the population growth rate at a given resource concentration (R), α is the initial rate of response of the population growth rate to increases in the resource, μmax indicates the maximal population growth rate in the condition of unlimited resources, and D is the dilution rate (0.05). The Monod equation is a saturating function that describes how the per capita population growth rate of a species increases with increasing resource availability. For the light experiment, we also fitted a second model which identifies an optimum light level (Ropt), at which the population growth rate is maximized and above and below which the population growth rate declines : (3)
We used the AIC of the two model fits to select the best model. The Monod model was selected for all but three species (Scenedsmus acuminatus, Selenastrum minutum and Spondylosium planum). For these three species the light optimum model (3 above) provided the better fit. We then solved the parameterized growth functions in order to determine the level of each resource at which growth rates went to zero (N*, P* and I*) using Mathematica v. 9.0.1 (Wolfram).
Stoichiometry & Cell measurements.
We estimated species' stoichiometric elemental ratios (e.g. N:P) because they reflect the homeostatic equilibrium of elements within each species' tissue—or the balance between what a species takes in and what it excretes [43, 44]. Stoichiometric ratios thus reflect a species' relative use of one element over another in carrying out its cellular functions, and they have been used as an estimate of species' resource consumption "vectors"—or the ratio at which a species is removing two resources from the environment . These vectors, along with minimum resource requirements for two species, are the two pieces of information needed to predict the outcome of competition for two resources . Furthermore, because certain elements are more important for cell growth (e.g. nitrogen) or cell division (e.g. phosphorus), stoichiometric ratios can also reflect variability in life history strategy among species, i.e. growing larger cells that divide slowly versus splitting into small cells that divide rapidly [43, 44]. Algal stoichiometry is known to display some plasticity (i.e. variation in response to environmental conditions) [36, 45], but intraspecific variation is generally lower that interspecific variation, and so species do generally have unique stoichiometric signatures . To estimate the C, N, and P stoichiometry of the focal species, we centrifuged 300 mLs of batch culture to achieve a pellet of concentrated biomass of each species and poured off the supernatant. The pelleted biomass was dried in a drying oven at 60°C for 5 days. The dried algal biomass samples were then massed and sent to the University of California Davis Analytical Lab for determinations of nitrogen, phosphorus, and carbon content, and stoichiometry.
We estimated cell size and one aspect of morphology (long axis length), because size and morphology are considered to be "master traits" that are linked to competitive abilities and resistance to grazing for phytoplankton . We measured cell dimensions for ten cells of each species using a camera attached to a compound light microscope to take photos and made size-calibrated measurements of each cell using Olympus' CellSens imaging software (Olympus Corporation of the Americas). Each species' cell shape was grouped according to standard shape-classifications a priori and cell dimensions were used to calculate the cellular biovolume of each cell (μm3) . All trait data and meta-data are available online (S1 and S2 Tables respectively).
In order to test whether ecologically relevant traits exhibit significant phylogenetic signal and to fit models of trait evolution, we relied on a recently published phylogenomic dataset for freshwater green algae . The phylogeny, full methods and results for the phylogenomic analyses, as well as supplementary trees and raw data are available or referenced in . All phylogenetic datasets and orthologs are available on Dryad (doi:10.5061/dryad.c574h). Raw Illumina files can be accessed using the NCBI SRA archive under BioProject Accession PRJNA237822. The bioinformatics tools can be accessed at http://galaxy-dev.cnsi.ucsb.edu/osiris, downloaded from https://bitbucket.org/osiris_phylogenetics/osiris_phylogenetics, and read about at http://osiris-phylogenetics.blogspot.com. To our knowledge, this is the most data-rich molecular phylogeny of freshwater green algae published to date.
For each trait, we chose to calculate three commonly used metrics of phylogenetic signal: Blomberg's K, Pagel's λ, and Moran's I using the Picante, Geiger and Adephylo packages in R respectively [50–52]. P-values for the significance of K and I were based on 999 randomizations each, whereas the significance test for λ was based on a likelihood ratio test. K and λ are considered to reflect an effect of shared ancestry on trait similarity, while I is considered to be a measure of autocorrelation of traits across the phylogeny. For K and λ, values close to zero indicate no phylogenetic signal, while values of one correspond to a Brownian motion expectation. Unlike λ, K can achieve values > 1, indicating that species are more similar than would be expected based on Brownian motion.
We tested for significant relationships between the pairwise phylogenetic distance among species (PD) and trait differentiation among species using Mantel tests, using the ‘ape’ package in R. PD was calculated using the ‘cophenetic’ function, and the trait distances were calculated as Euclidean distances using the ‘dist’ function. We considered trait differentiation and PD among species pairs to be significantly correlated at P≤0.05.
Most standard tests of phylogenetic signal assume that traits have evolved according to a Brownian Motion model [50, 51]. As a result, it is important to test whether or not a Brownian Motion model is indeed a good fit to the data, or if an alternative model provides a better fit. If alternative models provide a better fit, then metrics of phylogenetic signal based on a Brownian Motion model are not meaningful [17, 53]. Therefore, while some of the measured traits did display a phylogenetic signal, it was important to determine whether the Brownian Motion model of evolution was indeed a good explanation variation in these traits. Additionally, because many of our traits did not display a significant signal, we asked whether alternative models provided a better explanation of trait variation. For each trait, we tested nine models of continuous trait evolution using the Geiger package in R .
Five of the nine potential models of evolution in the Geiger package were retained for further analyses, either because they were the best fit model for one or more traits, or because they are useful for comparison with prior studies. The Brownian Motion (BM) model represents the evolution of trait as a random walk, and the model estimates a single rate parameter. The Ornstein-Uhlenbeck (OU) model aims to mimic the effect of stabilizing selection towards a single optimum and it therefore fits a model that describes the evolution of constrained trait variation. It incorporates a random trait walk, but includes two additional parameters, θ, which indicates the optimum trait value, and α, which represents the strength of the pull on trait values towards the optimum . As the current trait value at any point in time deviates further from the optimum, the impact of α becomes proportionately larger. As a result, OU models will generally lead to constrained trait variation among species compared to a BM model (Fig 1B). The White Noise (WN) model is a non-phylogenetic model that assumes that trait data come from a random normal distribution, and that species have no significant trait covariance (hence not shown in Fig 1). The δ tree-transformation model transforms the phylogeny to reflect the relative importance of recent divergence versus shared ancestry in determining trait variability. The δ parameter represents the influence of accelerating or decelerating evolution; values < 1 indicate that stem branches contributed disproportionately to trait evolution, while values > 1 indicate that the tip branches contribute more heavily to trait variation . For example, if distantly related species differ more than expected based on the phylogeny relative to recently diverged species, then it will transform the tree to shorten tip branches and lengthen stem branches, thereby lessening the impact of recent divergence on traits (δ < 1, Fig 1C). Conversely, if closely related species are more divergent than expected based on phylogeny, then it will transform the tree so as to lengthen the tip branches and exaggerate the impact of recent divergence (δ > 1, Fig 1D). The λ models also transform the tree so as to reflect the trait variability. However, λ only varies from 0 to 1, where λ = 1 retains the actual phylogeny and the traits are distributed across the tree as they would be for a BM model of evolution, and λ = 0 is a star phylogeny, where traits can be randomly dispersed across the phylogeny because common ancestry has no impact on trait variability (Fig 1E) . To determine the relative likelihood of the five retained models, we calculated each model's Akaike's Information Criterion (AIC) and Akaike weight. All analyses were conducted using the published phylogram of freshwater green algae . When no model achieved an AIC weight of 0.5 or more, we concluded that no model performed substantially better than the others (no choice).
We used the unsmoothed phylogenetic tree from Alexandrou et al. 2015  for these simulations. The branches lengths of the tree were visually “smoothed” for the figures by extending a horizontal line between the tips of the phylogeny and the maximum tip value. Values are based on simulations and therefore have no units or biological meaning. The scale was held constant for Fig 1a and 1b to allow a comparison of the impact of the OU model on trait variation relative to the BM model. Scales for other models were allowed to vary. The function that was used to generate the simulated trait data is an update of the picante package for R called evolve.trait.R and is available on the R Forge website (R-Forge@R-project.org).
Among the 17 traits that were measured, only three traits showed significant phylogenetic signal (measured as Blomberg's K): biovolume, long axis length and μmax light (Table 1). There was also a marginally non-significant phylogenetic signal for %C (although K = 0.401 is lower than K = 1 expected for a Brownian Motion random walk). There was also some indication of a signal for μmax phosphate (K = 0.857, p = 0.098), although it was not significant. There was no significant phylogenetic signal for traits related to resource requirements (R*s), responses to resource availability (αs), or stoichiometry (Table 1).
K = Blomberg's K (p-value based on randomization), Pagel's λ (p-value based on log-likelihood ratio test with λ set to zero), Moran's I (p-value based on randomization and one-way hypothesis test). Biovolume and long axis length are measures of algal size and morphology respectively; α light, α nitrate and α phosphate are the initial rates of response of species' growth rates to increases in resource availability; μmax light, μmax nitrate and μmax phosphate are the maximum per capita population growth rates of each species when that particular resource is unlimited; I*, N* and P* indicate concentrations of the resource (molar for N and P, μEinstein for I) at which the population growth rate equals zero; %C, %N, % P indicate the mass percent of carbon, nitrogen and phosphorus; C:N, C:P and N:P indicate the molar ratios of carbon (C), nitrogen (N) and phosphorus (P) of each species. N is the number of species.
The Mantel tests confirmed that the phylogeny provides some predictive value for trait variation for log biovolume and log long axis length (Table 2). There was a significant correlation between PD and the difference across species pairs in terms of their logged biovolumes and their logged long axis lengths. There was also a marginally non-significant correlation between PD and trait distance for the N:P and C:P molar ratios (Table 2). However, there was no significant relationship between PD and trait distance for any other traits related to competition. Altogether, this means that PD was a poor predictor of trait variation for 13 of the 17 traits that we measured.
The analysis was performed by comparing trait distance and PD matrices for each trait using the mantel.test function in the ‘ape’ package in R.
The model selections indicated that the WN model was the most commonly supported model. The WN model provided the best fit to the data for 7 out of 17 of the traits, including μmax light, I* (Fig 2A), α phosphate, μmax nitrate, % C (Fig 2B), % N, and the C:N molar ratio (Table 3). Phylogenetic models of trait evolution (e.g. δ and λ) were also supported for 7 traits. The δ model received the greatest amount of support for species' biovolume (Fig 2C), long axis length, α nitrate and N* (Table 3). For all of these traits, the δ parameter was > 3, suggesting that evolution of these traits occurred predominantly in the tips of the tree. This means that species are more different from one another than expected based on the extent of their shared ancestry and a Brownian Motion model of evolution. The λ model received the greatest support for %P (Fig 2D), C:P molar ratio and N:P molar ratio, although the WN model also received low to moderate support for all three traits. The λ parameter was close zero for these traits, indicating that a tree-transformation to a star phylogeny provided the best fit to the data (i.e. evolutionary relatedness was not correlated with trait similarities among species). For three traits, no model received strong support (AIC weights were all <0.50), namely, α light, μmax phosphate and P*. For the two phosphorus traits, μmax phosphate and P*, this lack of model support may have been due to the limited sample size and the resulting low statistical power for these traits (N = 7 species), which occurred because many species failed to grow under the experimental conditions after the imposed 72-hour nutrient starvation.
a) Distribution of I* across the phylogeny. For this trait a WN model was supported, indicating a random distribution of traits across the phylogeny; b) Distribution of % C across the phylogeny, a WN model was supported, again indicating a random trait distribution; c) Distribution of log-transformed algal biovolume across the phylogeny, a δ model was supported with a δ value of 7.971, suggesting a very weak influence of ancestral branches on trait variation; and d) Distribution of % P across the phylogeny, a λ model was supported with a λ value of 0.171, indicating that a tree-transformation to a nearly star-like phylogeny was the best fit. The phylogeny in this figure is the phylogeny published in Alexandrou et al. 2015 . We have smoothed the branches of the phylogeny only for purposes of the illustration.
BM = Brownian Motion model, OU = Ornstein-Uhlenbeck model, WN = White Noise model. Models were fitted to the trait data using the Geiger package and the fitcontinuous function in R. Algal traits are the same as in Table 1. See Fig 2 for traitgrams. The α parameter of the OU model, the δ parameter of the δ model and the λ parameter of the λ model are all also shown to allow interpretation of the model selections. In cases where a δ Model or a λ model received the greatest support from AIC weights, we tested whether the estimated parameter value was significantly different from a BM expectation by simulating 1,000 random walks of evolution. For each simulated walk, we then estimated δ and λ, creating sampling distributions of the parameters to determine the estimated parameter value from our data was outside of the 95% confidence interval for a random walk. Parameter values that are in bold are outside of the 95% confidence interval for a random walk of evolution (α CI = 0.00–3.16 (one-tailed), δ CI = 0.52–4.00 (two-tailed), λ = 0.87–1.00 (one-tailed)).
Despite the fact that Blomberg's K and Pagel's λ values indicate that there was significant phylogenetic signal in μmax light, %C, biovolume and long-axis length (Table 1), both the K and λ values tended to be less than one, indicating that the strength of the signal was weaker than would be expected based on a BM model of evolution. Furthermore, the model selections identified the WN model to be the best model for μmax light and %C and the δ model as the best model for both biovolume and long axis length, each with δ parameters of >4, suggesting that common ancestry does not heavily constrain variation in cell size and morphology among species. For biovolume and long axis length, the support for the δ model was strong (~1), and there was no support for the BM model. Blomberg's K values can be misleading when Brownian Motion is not a good underlying model of evolution [17, 53], and as a result, we do not interpret the significant K statistics as robust evidence of phylogenetic signal in these traits. However, the Mantel tests confirmed that there is a significant positive correlation between PD and trait distance among pairs of species for log biovolume and log long axis length, and so, we conclude that while phylogenetic distance does provide some information about differences in biovolume and long axis length among species, the correlation is weaker than expected based on a neutral model of trait evolution.
The evolutionary model selections and parameter values generally indicate that species' phylogenetic placements on the tree are not strongly related with their trait values. In some cases a non-phylogenetic model (WN) was selected, while for other traits, phylogenetic models (δ and λ) provided the best fit. The WN model indicates that traits were randomly distributed across the tree, while the model parameters for the δ and λ models indicated that the stem branches, or shared ancestry, had little influence on trait variation. These various model selections therefore lead to the similar interpretations. Using simulation, we show that traits that are randomly distributed across the tree, which are expected to return a WN model as the best fit model, also sometimes return δ and λ models (with parameter values reducing the influence of stem branches) as the best fit model (see Supporting Information). For traits that are randomly distributed across the phylogeny, which model is selected may have more to do with the statistical trait distribution than the association of the traits with the phylogeny. Regardless of the identity of the model selected, for 14 model selections, the biological interpretation of the model and parameter estimates selected is essentially very similar: shared ancestry has little ability to predict trait similarity.
Overall, our data do not lend much support to the hypothesis that ecological traits related to competition in freshwater green algae (resource requirements, resource assimilation, stoichiometry) are related to the phylogenetic placements of species on the tree of life. We found that for 13 of 17 traits, there was no significant phylogenetic signal. We also used a model-fitting approach to test whether the data supported a Brownian Motion model of evolution, a model of constrained evolution (OU), models of accelerating or decelerating evolution (δ and λ), or a White Noise (WN) model of essentially random trait distributions. The AIC weights indicated that a White Noise (WN), non-phylogenetic, model of evolution was most frequently supported. Even when the Blomberg’s K values indicated a significant phylogenetic signal, non-phylogenetic models were selected, or the best phylogenetic models indicated that traits have diverged since the most recent speciation events. Specifically, in the instances where phylogenetic models were supported (i.e. δ and λ), the estimated parameter values indicate that the phylogenetic signal was weak (K<1), evolution was accelerated in the tips of the phylogeny (δ ≥ 3), or that a star phylogeny transformation was the best fit (λ close to zero). Taken collectively, these model fits and parameter estimates suggest that ecological traits were not closely related to patterns of common ancestry (low signal), and that species' trait values often varied randomly across the tree (frequent selection of white noise as the best model). While other studies have shown that phylogenetic signal may not occur for all traits [19, 56–60], these results are novel because, to our knowledge, this is the first test of signal in traits related to resource competition in freshwater green algae. These results contrast with prior studies indicating that body size variation in diatoms evolved in basal lineages , and that variation in carbon uptake ability among groups of algae and other microbes occurs across larger taxonomic groups [62–64]. Similar to our findings however, research on prokaryotic microorganisms has also shown that variation in resource use, specifically in nitrate and phosphate uptake affinities, does not display a phylogenetic signal due to recent divergence [62, 65, 66].
The lack of phylogenetic signal in most of the ecological traits suggests that evolutionary relatedness does not generally reflect biological variation that is thought to underlie competitive abilities and resource use differentiation among freshwater green algae. While this contrasts with the expectation that more closely related species should be more similar in their traits than distantly related species, the absence of phylogenetic signal in species' traits is not uncommon in other taxa [19, 50]. Furthermore, these results may help to explain previous findings which demonstrated that the phylogenetic distance among pairs of freshwater green algal species is not a good predictor of the size of niche or fitness differences between them , the strength of their species interactions , or the level of ecosystem functioning they produce . These results may also provide an explanation for the observed lack of phylogenetic structure in 99% of the >1,000 natural communities of freshwater algae that have been surveyed in North American lakes . Specifically, if competition for resources is important in structuring natural communities , but competitive traits do not display a phylogenetic signal, then we should not expect natural communities to display phylogenetic structure .
Biological reasons for the absence of phylogenetic signal
There are a number of potential biological explanations for the lack of a relationship between species' placements on a molecular phylogeny and the trait values, as overwhelmingly observed in our dataset. While we cannot assess these directly with the data collected in this study, they are worth discussing in order to provide guidance for future work. First, it is possible that trait plasticity (variation of trait values within a species that occurs in response to variability in the environment) is obscuring phylogenetic signal. Some of the traits that we measured, e.g. light utilization traits, are known to be plastic [36, 45, 68]. When variation among individuals in a species is sufficiently large to approach variation between species, any correlation between species traits and their evolutionary relatedness can be masked. However, for many of our traits (e.g. stoichiometry and nutrient utilization traits), previous work has shown that interspecific trait variation is greater than intraspecific trait variation despite plasticity, such that species have unique trait signatures [36, 46]. As long as interspecific trait variation is greater than intraspecific trait variation, plasticity should not limit the detection of phylogenetic signal.
There is also evidence that algal traits can evolve rapidly, i.e. so quickly that they can affect the impact of ecological interactions as they are occurring [69–72]. Although this has only been demonstrated for traits related to herbivore defense, it is possible that the lack of phylogenetic signal that we found in competitive traits may be due to high trait lability. The individual functional genes controlling those traits may also be evolving at much faster rates than the rest of the genome. The phylogeny used in this study is based on many genes with the intention that using as many genes as possible will produce a phylogeny that best reflects the species tree. However, the genome is inevitably composed of genes that evolve at different rates, and individual gene trees may support alternative topologies . If the genes responsible for a particular ecological function are evolving at a different rate than the majority of the genome, or with a different pattern of descent, then ecological function would be decoupled from overall genetic sequence divergence, and therefore phylogenetic distance.
Lastly, we know very little about the relationship between genotype and phenotype evolution for algal species. For example, while many nucleotide substitutions are functionally synonymous, some do result in changes of large effect, e.g. . The functionality of any particular gene may also change independently of its sequence divergence via a number of mechanisms including changes in gene expression, methylation, RNA editing, or post-translational modification, among others. Thus, very different sequences can yield similar functions but large estimates of PD, while similar sequences with only a few unique base pairs may yield small estimates of PD but have vastly different functionality. Additionally, the traits we measured may be determined by multiple genes that interact to produce phenotypes in a non-additive or non-linear fashion (epistasis) [75, 76]. Future work should focus on better characterizing and incorporating these nonlinear impacts of genetic variation on phenotype into models of trait evolution. Gene knock-out or knock-down experiments could be used to identify and describe the way in which genes, sequence variability, and gene interactions control traits related to competition.
There are some important limitations to this study. First, while we are confident that the traits we measured do reflect species' abilities to acquire and survive on limiting resources, we did not directly assess the ability of these traits to predict the outcome of resource competition for these species. In order to confirm that the absence of a phylogenetic signal in these traits indeed explains our earlier findings that phylogenetic distances among freshwater green algae do not predict the strength of competition, or more generally, species interactions, community structure and ecosystem functioning, we would need to test whether these traits correctly predict these aspects of community ecology. Most studies of phylogenetic signal however, measure physiological, morphological and life history traits that are thought to be only indirectly linked to resource acquisition and the outcome of competition (e.g. [77, 78]). Others measure some aspect of resource use, such as diet or habitat preferences . Such traits may be the best possible, though indirect, estimates of resource requirements and competitive abilities for many taxa. The traits that we measured, however, are direct estimates of resource requirements and competitive abilities for inorganic resources, and have previously been shown to predict the outcome of competition for phytoplankton [25, 26, 79–81].
Secondly, some have argued that because phylogenetic signal in trait variation must occur at some phylogenetic scale , an absence of phylogenetic signal suggests that one has not studied the right phylogenetic scale [19, 82, 83]. For example, phylogenetic signal in the traits that we studied here may occur at smaller (e.g. among sister taxa, ) or larger phylogenetic scales (e.g. when comparing groups such as diatoms, coccolithophorids, dinoflagellates and green algae, ). Our goal here, however, was not to find the phylogenetic scale at which these traits display a phylogenetic signal. Rather our goal was to test whether common ancestry can be used to describe the variation in traits that reflect resource competition in freshwater green algae per se, because we had previously shown that these algae display great variation in the outcome of competition [3, 10, 31]. It remains possible however that these traits do display phylogenetic signal at other phylogenetic scales.
Finally, our species selections and methods may have impacted our conclusions because the species that grew in these experiments are only a subset of those found in nature. Specifically, we only included cultivable species that are common in lakes across North America and we excluded species whose traits we could not measure. Efforts to measure traits in situ and sample a greater species pool would help to test the generality of our findings.
In conclusion, our study demonstrates that phylogeny is not a good predictor of important ecological characters in freshwater green algae, and that maximizing phylogenetic distance among species is no more likely to capture diversity of some traits than randomly selecting species. While this is the first time this has been shown for competitive traits of freshwater green algae, a lack of phylogenetic signal has also been described in other traits and other taxa [19, 21, 56–60]. This suggests that tests of phylogenetic signal and evolutionary model fits of ecologically important characters should be regularly completed before using phylogenetic information to estimate phenotypic or functional biodiversity, to infer processes of community assembly, or to gauge ecosystem functioning.
S1 Table. Original trait data file.
S2 Table. Metadata file for trait data.
We would like to thank C. Hampton-Miller for thoughtful discussions and for providing useful R scripts that helped with the evolutionary model fitting. This study was supported by the US National Science Foundation’s DIMENSIONS of Biodiversity program in a grant to BJC (DEB-1046121) and THO (DEB-1046121).
Conceived and designed the experiments: AN JH. Performed the experiments: JH AV CZ. Analyzed the data: AN JH AV CZ. Contributed reagents/materials/analysis tools: THO BJC. Wrote the paper: AN MAA AV CZ THO BJC.
- 1. Cahill JF, Kembel SW, Lamb EG, Keddy PA. Does phylogenetic relatedness influence the strength of competition among vascular plants? Perspectives in Plant Ecology Evolution and Systematics. 2008;10(1):41–50. pmid:WOS:000254229400003.
- 2. Cadotte MW, Strauss SY. Phylogenetic Patterns of Colonization and Extinction in Experimentally Assembled Plant Communities. PLoS One. 2011;6(5). doi: e19363 pmid:WOS:000290305600022.
- 3. Narwani A, Alexandrou MA, Oakley TH, Carroll IT, Cardinale BJ. Experimental evidence that evolutionary relatedness does not affect the ecological mechanisms of coexistence in freshwater green algae. Ecol Lett. 2013;16(11):1373–81. pmid:WOS:000325976500006.
- 4. Godoy O, Kraft NJB, Levine JM. Phylogenetic relatedness and the determinants of competitive outcomes. Ecol Lett. 2014;17(7):836–44. pmid:WOS:000337534600008.
- 5. Strauss SY, Webb CO, Salamin N. Exotic taxa less related to native species are more invasive. Proc Natl Acad Sci U S A. 2006;103(15):5841–5. pmid:WOS:000236896200038.
- 6. Tan JQ, Pu ZC, Ryberg WA, Jiang L. Species phylogenetic relatedness, priority effects, and ecosystem functioning. Ecology. 2012;93(5):1164–72. pmid:WOS:000304368100021.
- 7. Cadotte MW, Cardinale BJ, Oakley TH. Evolutionary history and the effect of biodiversity on plant productivity. Proc Natl Acad Sci U S A. 2008;105(44):17012–7. Epub 2008/10/31. pmid:18971334; PubMed Central PMCID: PMC2579369.
- 8. Cadotte MW, Cavender-Bares J, Tilman D, Oakley TH. Using phylogenetic, functional and trait diversity to understand patterns of plant community productivity. PLoS One. 2009;4(5). doi: e5695 pmid:WOS:000266415000008.
- 9. Flynn DFB, Mirotchnick N, Jain M, Palmer MI, Naeem S. Functional and phylogenetic diversity as predictors of biodiversity-ecosystem-function relationships. Ecology. 2011;92(8):1573–81. pmid:WOS:000293459100004.
- 10. Fritschie KJ, Cardinale BJ, Alexandrou MA, Oakley TH. Evolutionary history and the strength of species interactions: testing the phylogenetic limiting similarity hypothesis. Ecology. 2014;95(5):1407–17. pmid:WOS:000336740500029.
- 11. Maherali H, Klironomos JN. Influence of Phylogeny on fungal community assembly and ecosystem functioning. Science. 2007;316(5832):1746–8. pmid:WOS:000247400500047.
- 12. Violle C, Nemergut DR, Pu ZC, Jiang L. Phylogenetic limiting similarity and competitive exclusion. Ecol Lett. 2011;14(8):782–7. pmid:WOS:000292864400008.
- 13. Faith DP. Conservation evaluation and phylogenetic diversity. Biological Conservation. 1992;61(1):1–10. pmid:WOS:A1992HX12300001.
- 14. Faith DP. Phylogenetic diversity, functional trait diversity and extinction: avoiding tipping points and worst-case losses. Philos Trans R Soc B-Biol Sci. 2015;370(1662). pmid:WOS:000347312400010.
- 15. Webb CO, Ackerly DD, McPeek MA, Donoghue MJ. Phylogenies and community ecology. Annual Review of Ecology and Systematics. 2002;33:475–505. pmid:WOS:000180007000018.
- 16. Blomberg SP, Garland T. Tempo and mode in evolution: phylogenetic inertia, adaptation and comparative methods. Journal of Evolutionary Biology. 2002;15(6):899–910. pmid:WOS:000178853600002.
- 17. Revell LJ, Harmon LJ, Collar DC. Phylogenetic Signal, Evolutionary Process, and Rate. Systematic Biology. 2008;57(4):591–601. pmid:WOS:000263702800005.
- 18. Kelly S, Grenyer R, Scotland RW. Phylogenetic trees do not reliably predict feature diversity. Diversity and Distributions. 2014;20(5):600–12.
- 19. Losos JB. Phylogenetic niche conservatism, phylogenetic signal and the relationship between phylogenetic relatedness and ecological similarity among species. Ecol Lett. 2008;11(10):995–1003. pmid:WOS:000259147800001.
- 20. Mouquet N, Devictor V, Meynard CN, Munoz F, Bersier L- F, Chave J, et al. Ecophylogenetics: advances and perspectives. Biological Reviews. 2012;87(4):769–85. pmid:WOS:000309416900001.
- 21. Gerhold P, Cahill JF, Winter M, Bartish IV, Prinzing A. Phylogenetic patterns are not proxies of community assembly mechanisms (they are far better). Functional Ecology. 2015;29(5):600–14.
- 22. Grover JP. Assembly rules for communities of nutrient-limited plants and specialist herbivores. Am Nat. 1994;143(2):258–82. pmid:ISI:A1994MX57600003.
- 23. Sommer U. An experimental test of the intermediate disturbance hypothesis using cultures of marine phytoplankton. Limnology and Oceanography. 1995;40(7):1271–7. pmid:ISI:A1995TK46900010.
- 24. Sommer U. The paradox of the plankton: Fluctuations of phosphorus availability maintain diversity of phytoplankton in flow-through cultures. Limnology and Oceanography. 1984;29(3):633–6. pmid:ISI:A1984SX74500016.
- 25. Tilman D. Resource compeition between planktonic algae: An experimental and theoretical approach. Ecology. 1977;58(2):338–48. pmid:ISI:A1977DN32600009.
- 26. Tilman D. Tests of resource competition theory using 4 species of Lake Michigan algae. Ecology. 1981;62(3):802–15. pmid:ISI:A1981LW72000034.
- 27. Huisman J, Weissing FJ. Biodiversity of plankton by species oscillations and chaos. Nature. 1999;402(6760):407–10. pmid:ISI:000083913600055.
- 28. Huisman J, Weissing FJ. Fundamental unpredictability in multispecies competition. Am Nat. 2001;157(5):488–94. pmid:ISI:000168023500002.
- 29. Huisman J, Johansson AM, Folmer EO, Weissing FJ. Towards a solution of the plankton paradox: the importance of physiology and life history. Ecol Lett. 2001;4(5):408–11. pmid:ISI:000172198700005.
- 30. Hutchinson GE. The paradox of the plankton. Am Nat. 1961;95(882):137–45. pmid:ISI:A1961WJ60200002.
- 31. Venail PA, Narwani A, Fritschie K, Alexandrou MA, Oakley TH, Cardinale BJ. The influence of phylogenetic relatedness on species interactions among freshwater green algae in a mesocosm experiment. Journal of Ecology. 2014;102(5):1288–99.
- 32. Reynolds CS. Ecology of phytoplankton. Cambridge: Cambridge University Press; 2006.
- 33. Dybzinski R, Tilman D. Resource use patterns predict long-term outcomes of plant competition for nutrients and light. Am Nat. 2007;170(3):305–18. pmid:WOS:000248964000002.
- 34. Tilman D. Resource Competition and Community Structure. Princeton, New Jersey: Princeton University Press; 1982.
- 35. Edwards KF, Thomas MK, Klausmeier CA, Litchman E. Allometric scaling and taxonomic variation in nutrient utilization traits and maximum growth rate of phytoplankton. Limnology and Oceanography. 2012;57(2):554–66. pmid:WOS:000302613200014.
- 36. Litchman E, Klausmeier CA. Trait-Based Community Ecology of Phytoplankton. Annual Review of Ecology Evolution and Systematics. 2008;39:615–39. pmid:ISI:000261725500029.
- 37. Smith KL, Harmon LJ, Shoo LP, Melville J. Evidence of constrained phenotypic evolution in a cryptic species complex of agamid lizards. Evolution. 2011;65(4):976–92. pmid:WOS:000289158300006.
- 38. Cottingham KL, Carpenter SR. Population, community, and ecosystem variates as ecological indicators: Phytoplankton responses to whole-lake enrichment. Ecological Applications. 1998;8(2):508–30. pmid:WOS:000073447900030.
- 39. Kilham SS, Kreeger DA, Lynn SG, Goulden CE, Herrera L. COMBO: a defined freshwater culture medium for algae and zooplankton. Hydrobiologia. 1998;377:147–59. pmid:ISI:000076947500013.
- 40. Seppala J, Balode M. The use of spectral fluorescence methods to detect changes in the phytoplankton community. Hydrobiologia. 1997;363(0):207–17. pmid:BCI:BCI199800359976.
- 41. Clesceri LS, Greenberg AE, Eaton AD, editors. Standard methods for the examination of water and wastewater. 20 ed. Washington D.C., USA: American Public Health Association; 1998.
- 42. Schwaderer AS, Yoshiyama K, Pinto PT, Swenson NG, Klausmeier CA, Litchman E. Eco-evolutionary differences in light utilization traits and distributions of freshwater phytoplankton. Limnology and Oceanography. 2011;56(2):589–98. pmid:WOS:000290677800015.
- 43. Elser JJ, Sterner RW, Gorokhova E, Fagan WF, Markow TA, Cotner JB, et al. Biological stoichiometry from genes to ecosystems. Ecol Lett. 2000;3(6):540–50. pmid:WOS:000165516300013.
- 44. Loladze I, Elser JJ. The origins of the Redfield nitrogen-to-phosphorus ratio are in a homoeostatic protein-to-rRNA ratio. Ecol Lett. 2011;14(3):244–50. Epub 2011/01/20. pmid:21244593.
- 45. Sterner RW, Elser JJ. Ecological stoichiometry: The biology of elements from molecules to the biosphere2002.
- 46. Klausmeier CA, Litchman E, Daufresne T, Levin SA. Phytoplankton stoichiometry. Ecological Research. 2008;23(3):479–85.
- 47. Litchman E, Pinto PD, Klausmeier CA, Thomas MK, Yoshiyama K. Linking traits to species diversity and community structure in phytoplankton. Hydrobiologia. 2010;653(1):15–28. pmid:ISI:000280092100003.
- 48. Hillebrand H, Durselen CD, Kirschtel D, Pollingher U, Zohary T. Biovolume calculation for pelagic and benthic microalgae. Journal of Phycology. 1999;35(2):403–24. pmid:ISI:000080222900021.
- 49. Alexandrou MA, Cardinale BJ, Hall JD, Delwiche CF, Fritschie K, Narwani A, et al. Evolutionary relatedness does not predict competition and co-occurrence in natural or experimental communities of green algae. Proceedings of the Royal Society B-Biological Sciences. 2015;282(1799). pmid:WOS:000345624600006.
- 50. Blomberg SP, Garland T, Ives AR. Testing for phylogenetic signal in comparative data: Behavioral traits are more labile. Evolution. 2003;57(4):717–45. pmid:WOS:000182815100003.
- 51. Pagel M. Inferring the historical patterns of biological evolution. Nature. 1999;401(6756):877–84. pmid:WOS:000083464700048.
- 52. Munkemuller T, Lavergne S, Bzeznik B, Dray S, Jombart T, Schiffers K, et al. How to measure and test phylogenetic signal. Methods in Ecology and Evolution. 2012;3(4):743–56. pmid:WOS:000306961700014.
- 53. Cooper N, Jetz W, Freckleton RP. Phylogenetic comparative approaches for studying niche conservatism. Journal of Evolutionary Biology. 2010;23(12):2529–39. pmid:WOS:000284277700002.
- 54. Harmon LJ, Weir JT, Brock CD, Glor RE, Challenger W. GEIGER: investigating evolutionary radiations. Bioinformatics. 2008;24(1):129–31. pmid:WOS:000251865000018.
- 55. Butler MA, King AA. Phylogenetic comparative analysis: A modeling approach for adaptive evolution. Am Nat. 2004;164(6):683–95. pmid:WOS:000225968000003.
- 56. Ndiribe C, Pellissier L, Dubuis A, Vittoz P, Salamin N, Guisan A. Plant functional and phylogenetic turnover correlate with climate and land use in the Western Swiss Alps. Journal of Plant Ecology. 2014;7(5):439–50. pmid:WOS:000343327200003.
- 57. Roncal J, Henderson A, Borchsenius F, Sodre Cardoso SR, Balslev H. Can phylogenetic signal, character displacement, or random phenotypic drift explain the morphological variation in the genus Geonoma (Arecaceae)? Biological Journal of the Linnean Society. 2012;106(3):528–39. pmid:WOS:000305182200006.
- 58. Perronne R, Mauchamp L, Mouly A, Gillet F. Contrasted taxonomic, phylogenetic and functional diversity patterns in semi-natural permanent grasslands along an altitudinal gradient. Plant Ecology and Evolution. 2014;147(2):165–75. pmid:WOS:000338914400003.
- 59. Agrawal AA, Fishbein M, Halitschke R, Hastings AP, Rabosky DL, Rasmann S. Evidence for adaptive radiation from a phylogenetic study of plant defenses. Proc Natl Acad Sci U S A. 2009;106(43):18067–72. pmid:WOS:000271222500008.
- 60. Fritz SA, Purvis A. Phylogenetic diversity does not capture body size variation at risk in the world's mammals. Proceedings of the Royal Society B-Biological Sciences. 2010;277(1693):2435–41. pmid:WOS:000279784000003.
- 61. Nakov T. Using phylogeny to model cell size evolution in marine and freshwater diatoms. Limnology and oceanography. 2014;59(1):79.
- 62. Martiny AC, Treseder K, Pusch G. Phylogenetic conservatism of functional traits in microorganisms. ISME J. 2013;7(4):830–8. http://www.nature.com/ismej/journal/v7/n4/suppinfo/ismej2012160s1.html. pmid:23235290
- 63. Young JN, Rickaby REM, Kapralov MV, Filatov DA. Adaptive signals in algal Rubisco reveal a history of ancient atmospheric carbon dioxide. Philosophical Transactions of the Royal Society B: Biological Sciences. 2012;367(1588):483–92.
- 64. Low-Décarie E, Fussmann GF, Bell G. The effect of elevated CO2 on growth and competition in experimental phytoplankton communities. Global Change Biology. 2011;17(8):2525–35.
- 65. Martiny AC, Huang Y, Li W. Occurrence of phosphate acquisition genes in Prochlorococcus cells from different ocean regions. Environmental Microbiology. 2009;11(6):1340–7. pmid:40642428.
- 66. Martiny AC, Kathuria S, Berube PM. Widespread metabolic potential for nitrite and nitrate assimilation among Prochlorococcus ecotypes. Proceedings of the National Academy of Sciences. 2009;106(26):10787–92.
- 67. Edwards KF, Litchman E, Klausmeier CA. Functional traits explain phytoplankton responses to environmental gradients across lakes of the United States. Ecology. 2013;94(7):1626–35. pmid:WOS:000321618400020.
- 68. Rhee GY. Effects of N-P atomic ratios and nitrate limitation on algal growth, cell composition, and nitrate uptake. Limnology and Oceanography. 1978;23(1):10–25. pmid:WOS:A1978EW90700002
- 69. Yoshida T, Ellner SP, Jones LE, Bohannan BJ, Lenski RE, Hairston NG Jr. Cryptic population dynamics: rapid evolution masks trophic interactions. PLoS Biol. 2007;5(9):1868–79. Epub 2007/09/07. pmid:17803356; PubMed Central PMCID: PMC1964773.
- 70. Yoshida T, Jones LE, Ellner SP, Fussmann GF, Hairston NG. Rapid evolution drives ecological dynamics in a predator-prey system. Nature. 2003;424(6946):303–6. pmid:ISI:000184183900039.
- 71. Becks L, Ellner SP, Jones LE, Hairston NG. The functional genomics of an eco-evolutionary feedback loop: linking gene expression, trait evolution, and community dynamics. Ecol Lett. 2012;15(5):492–501. pmid:WOS:000302288900012.
- 72. Jones LE, Becks L, Ellner SP, Hairston NG Jr., Yoshida T, Fussmann GF. Rapid contemporary evolution and clonal food web dynamics. Philos Trans R Soc Lond B Biol Sci. 2009;364(1523):1579–91. Epub 2009/05/06. pmid:19414472; PubMed Central PMCID: PMC2690500.
- 73. Huynen MA, Bork P. Measuring genome evolution. Proc Natl Acad Sci U S A. 1998;95(11):5849–56. Epub 1998/05/30. pmid:9600883; PubMed Central PMCID: PMCPMC34486.
- 74. Elena SF, Lenski RE. Evolution experiments with microorganisms: The dynamics and genetic bases of adaptation. Nat Rev Genet. 2003;4(6):457–69. pmid:WOS:000183202600015.
- 75. Phillips PC. Epistasis—the essential role of gene interactions in the structure and evolution of genetic systems. Nat Rev Genet. 2008;9(11):855–67. pmid:WOS:000260162500010.
- 76. Carlborg O, Haley CS. Epistasis: too often neglected in complex trait studies? Nat Rev Genet. 2004;5(8):618–U4. pmid:WOS:000223039200016.
- 77. Hertz PE, Arima Y, Harrison A, Huey RB, Losos JB, Glor RE. Asynchronous evolution of physiology and morphology in Anolis lizards. Evolution. 2013;67(7):2101–13. pmid:WOS:000321184500022.
- 78. Savage JA, Cavender-Bares J. Habitat specialization and the role of trait lability in structuring diverse willow (genus Salix) communities. Ecology. 2012;93(8):S138–S50. pmid:WOS:000307302400012.
- 79. Grover JP. Phosphorus-dependent growth-kinetics of 11 species of fresh-water algae. Limnology and Oceanography. 1989;34(2):341–8. pmid:WOS:A1989U538800007.
- 80. Grover JP. Effects of Si:P supply ratio, supply variability, and selective grazing in the plankton—an experiment with a natural algal and protistan assemblage. Limnology and Oceanography. 1989;34(2):349–67. pmid:WOS:A1989U538800008.
- 81. Miller T, E., Burns JH, Munguia P, Walters EL, Kneitel JM, Richards PM, et al. A Critical Review of Twenty Years’ Use of the Resource-Ratio Theory. Am Nat. 2005;165(4):439–48. pmid:15791536
- 82. Cavender-Bares J, Kozak KH, Fine PVA, Kembel SW. The merging of community ecology and phylogenetic biology. Ecol Lett. 2009;12(7):693–715. pmid:WOS:000266699000013.
- 83. Cavender-Bares J, Keen A, Miles B. Phylogenetic structure of floridian plant communities depends on taxonomic and spatial scale. Ecology. 2006;87(7):S109–S22. pmid:WOS:000239606900011.
- 84. Litchman E, Klausmeier CA, Schofield OM, Falkowski PG. The role of functional traits and trade-offs in structuring phytoplankton communities: scaling from cellular to ecosystem level. Ecol Lett. 2007;10(12):1170–81. pmid:ISI:000250700800007.