Predicting the impacts of environmental change on marine organisms, food webs, and biogeochemical cycles presently relies almost exclusively on short-term physiological studies, while the possibility of adaptive evolution is often ignored. Here, we assess adaptive evolution in the coccolithophore Emiliania huxleyi, a well-established model species in biological oceanography, in response to ocean acidification. We previously demonstrated that this globally important marine phytoplankton species adapts within 500 generations to elevated CO2. After 750 and 1000 generations, no further fitness increase occurred, and we observed phenotypic convergence between replicate populations. We then exposed adapted populations to two novel environments to investigate whether or not the underlying basis for high CO2-adaptation involves functional genetic divergence, assuming that different novel mutations become apparent via divergent pleiotropic effects. The novel environment “high light” did not reveal such genetic divergence whereas growth in a low-salinity environment revealed strong pleiotropic effects in high CO2 adapted populations, indicating divergent genetic bases for adaptation to high CO2. This suggests that pleiotropy plays an important role in adaptation of natural E. huxleyi populations to ocean acidification. Our study highlights the potential mutual benefits for oceanography and evolutionary biology of using ecologically important marine phytoplankton for microbial evolution experiments.

Experimental evolution has revolutionized evolutionary biology, providing the unique opportunity to investigate evolutionary processes in real time. Model systems for experimental evolution studies are typically unicellular microbes that are easy to cultivate, reproduce quickly, and provide a range of genomic and recombinant technologies, so that organisms such as Escherichia coli, Pseudomonas, and Saccharomyces are the microbes of choice for most experimental evolution (Elena and Lenski 2003; Buckling et al. 2009). Although experimental evolution has made inroads as an applied science for medically relevant organisms such as Pseudomonas (e.g., Perron et al. 2006), most research is still done using organisms where we know little about their ecology outside the lab. In contrast, a model system with a clear ecological role that is also studied in the context of biogeochemical consequences of global climate change provides the opportunity to use microbial experimental evolution to address contemporary questions of pressing ecological and socioeconomic relevance.

Potential model systems with well-studied ecology include marine phytoplankton; prokaryotic and eukaryotic microbes that are responsible for about 50% of the global primary production (Field et al. 1998). One question of immediate concern is how future marine phytoplankton communities are likely to respond (including the possibility of adaptive evolution) to a warmer, more stratified, and more acidic ocean of the coming decades (Riebesell et al. 2011; Collins 2012; Reusch and Boyd 2013). The coccolithophore Emiliania huxleyi (Fig. 1) is a prominent model organism in biological oceanography. This microscopic alga thrives under a wide range of environmental conditions, with a ubiquitous distribution, except for polar waters, and has the ability to form extensive blooms (Paasche 2002) that are large enough to make them visible from space (Fig. 2). These characteristics suggest an important role of E. huxleyi for marine primary production and carbon cycling (Westbroek 1989). Besides photosynthetic carbon fixation, coccolithophores possess a second biogeochemically relevant trait—they produce minute calcite scales that can accelerate the export of organic matter from the surface into the deep ocean and thereby also affect the oceanic carbon cycle (Armstrong et al. 2002). Increasing atmospheric concentrations of fossil fuel derived CO2 changes ocean chemistry; it leads to a lowering in seawater pH termed ocean acidification (Caldeira and Wickett 2003). This in turn, may negatively affect growth and calcite production in coccolithophores, which has reinforced the attention given to E. huxleyi (Riebesell et al. 2009; Riebesell et al. 2011).

Figure 1.

Scanning electron micrograph of an Emiliania huxleyi cell covered with calcite scales. Image provided by L. T. Bach, GEOMAR, credits to the Institute for Geosciences, University of Kiel.

Figure 2.

Satellite image of an Emiliania huxleyi bloom off the Norwegian coast. The light blue color, caused by increased light reflection due to the calcite platelets produced by the cells, marks the extension of the bloom. Image provided by the SeaWiFS Project, NASA/Goddard Space Flight Center and GeoEye.

While short-term physiological responses to changes in various environmental parameters are reasonably well studied in E. huxleyi (Paasche 2002; Zondervan 2007), long-term experiments that investigate evolutionary responses are needed to predict future responses of marine phytoplankton to global change (Riebesell 2000; Collins 2012; Lohbeck et al. 2012). This is where experimental evolution and biological oceanography meet to open a new and promising interdisciplinary research area with the potential to benefit both fields (Reusch and Boyd 2013).

To investigate the evolutionary response of E. huxleyi to elevated CO2, we carried out an experimental evolution experiment with freshly isolated genotypes from a coastal, temperate site near Bergen, Norway (Lohbeck et al. 2012). After the first 500 generations of asexual growth, we found 3.3% increase in fitness and 51% restoration in calcification rate in high CO2 adapted E. huxleyi populations relative to ambient CO2 selected control populations when tested under high CO2 conditions. Based on laboratory experiments, adaptive evolution in this key phytoplankton species appears possible and swift enough to keep pace with global climate change scenarios, even in the absence of sex, migration, or other mechanisms that may speed up adaptation in natural populations.

Interestingly, replicate selection lines revealed strikingly similar phenotypes in terms of direct and correlated responses to selection under CO2 enrichment. All five replicate populations independently revealed similar fitness increases relative to ambient CO2 selected control populations, with within-group variances being similar between adapted and control populations. There are two possible explanations for such phenotypic convergence. First, the same mutations may have occurred and increased in frequency throughout all five adapting replicates (Wichman et al. 1999; Woods et al. 2006). Alternatively, different mutations characterize phenotypically convergent replicate populations (Travisano et al. 1995; Ostrowski et al. 2008). If so, the pleiotropic effects of these new allelic variants are likely to be different in another, novel environment. Phenotypic divergence then can be taken as indirect evidence for a different genetic basis of CO2-adapted phenotypes (Travisano et al. 1995; MacLean and Bell 2003; Ostrowski et al. 2008). Such functional genetic divergence can constrain evolution in the face of additional stressors and therefore be a key variable when assessing phenotypic responses to a multitude of complex selective forces natural phytoplankton populations will encounter in the future ocean (Boyd 2011).

To distinguish among these alternatives, we used two “challenge” assays to resolve to what extent the populations founded from a single genotype had diverged during adaptation to the high CO2 environment. If the replicate populations have not diverged genetically, they would have the same phenotypes in the novel environment. If, on the other hand, they have diverged genetically, it is expected that their phenotypes in some novel environments differ, either due to divergent pleiotropic effects of adaptive mutations fixed during selection at high CO2, or to replicate-specific genetic variation that is neutral at high CO2 but that affects fitness in a novel environment. We used high-light and low-salinity environments that were not experienced by the populations during the high CO2 selection experiment, to test for functional genetic divergence in high CO2 adapted populations that may affect responses to other environmental changes.

Materials and Methods


Clonal cultures were founded by single cell isolation from a natural E. huxleyi bloom in Raunefjorden, Norway in May 2009 and verified to originate from a single diploid cell by microsatellite genotyping as described in Lohbeck et al. (2012). We used nonaxenic cultures, which is a common approach for E. huxleyi culture experiments, to allow for long-term batch culturing over many years. To exclude the presence of a significant bacteria fraction, we regularly performed automated cell counts and microscopy throughout the experiments.

Artificial seawater (ASW) was produced as in Kester et al. (1967) with the exception of 2400 μmol bicarbonate kg−1 ASW addition (Merck, Germany), resulting in a total alkalinity (TA) of 2444 μmol kg−1 at a salinity of 35 (ASW35). Low-salinity ASW (ASW15) was produced by adding 1.334 kg MilliQ water kg−1 ASW35 and subsequent addition of 1314 μmol bicarbonate kg−1 low salinity ASW, resulting in ASW with a salinity of 15 (ASW15) and a TA of 2367 μmol kg−1 ASW15. All ASW media were 0.2-μm sterile filtered (Whatman Polycap 75AS, GE Healthcare, UK) and supplemented with 64 μmol kg−1 nitrate, 4 μmol kg−1 phosphate (nutrient ratio after Redfield 1958), trace metals and vitamins according to f/8 adapted from Guillard and Ryther (1962), 10 nmol kg−1 selenium after Danbara and Shiraiwa (1999), and 2 mL kg−1 sterile filtered North Sea water to exclude any limitations by micronutrients. Initial nutrient concentrations were sufficient to ensure that under all final cell densities observed in the experiments an adequate remaining nutrient concentration prevents any nutrient limitation effects.

Prior to inoculation, ASW media were aerated for 24 h using a controlled CO2 gas mixing system at the treatment levels of 400 and 2200 μatm pCO2. Evaporation was minimized by prefixed gas wash bottles filled with MilliQ water to saturate gases with humidity. After CO2 manipulations, ASW media were carefully pumped into the respective culture flasks using a sterile silicon hose to ensure minimal gas exchange. For each treatment, two extra flasks were prepared for dissolved inorganic carbon (DIC) and TA measurements.

Experimental cultures in the CO2 selection and assay experiments were grown under the same conditions as given here for the challenge experiments but under continuous rotation. Details are described in Lohbeck et al. (2012). In the challenge experiments, cultures were grown in 250-mL Duran square flasks (Schott, Germany) filled with 310 mL ASW medium leaving a minimum headspace. Cultures were incubated in a RUMED Light Thermostat 1301 (Rubarth Apparate GmbH, Germany) at 15°C and a photon flux density of 155 ± 2 μmol photons m−2 s−1 under a 16:8 light:dark cycle. All culture flasks were manually rotated 10 times, twice a day (10 and 17 h).


Replicate selection lines were propagated for 750 and 1000 generations under ambient (400 μatm) and high (2200 μatm) pCO2 and assayed under reciprocal CO2 conditions using the same protocol as in Lohbeck et al. (2012). To account for technical variation in the challenge experiments, triplicate subreplicates of each individual replicate population were inoculated into an acclimation batch cycle prior to an assay batch cycle in their respective control and challenge environments. We applied one full acclimation cycle (approximately six to eight cell divisions) prior to the assay cycle to account for nongenetic effects after transfer into a new environment. Growth in novel environments (high light, low salinity) was used to detect functional genetic divergence. Populations were grown in the CO2 environment that they had been selected at for either 750 or 1000 generations. Populations selected at high CO2 are referred to as “high CO2 populations” and those selected at ambient CO2 are referred to as “ambient CO2 populations.” Each population was then subjected to a novel environment that is not obviously connected to CO2 levels.

Cell densities were measured in triplicate using a Z2 Particle and Size Analyzer (Beckman Coulter, Brea, CA). Exponential growth rates (μ) were calculated from cell densities where μ = (ln N1 – ln N0)/d, and N1 and N0 are cell concentrations at the beginning and end of a batch cycle and d is the duration of the batch cycle in days. Experimental cultures were started with a population size of 100,000 cells per flask (323 cells/mL). In both challenge experiments, control populations from ambient and high CO2 selection lines were grown in five-day batch cycles (155 μmol photons m−2 s−1, salinity 35). High light challenged populations (800 μmol photons m−2 s−1, salinity 35) were grown in four-day batch cycles and salinity challenged populations in six-day (400 μatm pCO2, salinity 15) and 12-day (2200 μatm pCO2, salinity 15) batch cycles, respectively. Throughout the experiments, all populations grew exponentially and never reached the stationary phase.


Carbonate chemistry was determined by DIC and TA measurements. DIC samples were taken for all treatment groups prior to inoculation and measured with an AIRICA system (Marianda, Germany). TA was measured from all ASW batches by open-cell acidimetric titration using a Basic Titrino 794 (Metrohm, Switzerland). CO2 partial pressure in the culture media was calculated from DIC and TA using the software CO2SYS (Lewis and Wallace 1998) with solubility constants after Roy et al. (1993). Average culture pCO2 values of 351 ± 12 and 2065 ± 103 μatm (ambient and high CO2 treatments) were calculated from DIC and TA measured prior to inoculation and the draw-down estimate from final cell numbers following the approach of Bach et al. (2011). Under these conditions, the batch culture environment does not constitute a constant but a slightly decreasing CO2 environment due to DIC drawdown during population growth that accounted for a maximum DIC drawdown of 7.6%.

We were not able to measure DIC samples taken from the high-light challenge experiment, but the average culture CO2 values were almost certainly within the range set up by the CO2 gas system for two reasons. First, we have more than 1000 generations of culture work experience using this CO2 gas mixing system and so far the expected CO2 levels have always been confirmed by DIC and TA measurements. Second, growth rate is a sensitive indicator for CO2 levels in this culture media. We prepared control and challenged media from the same aerated ASW batch and found growth rates of control populations to be absolutely in line with expectations for ambient CO2 culture media. As such, the average culture CO2 values from all experiments were unlikely to differ significantly from the desired treatment levels of 400 and 2200 μatm pCO2 that are given in the figures.


To test for heterogeneity of variances among replicates within the challenge experiment, we first performed an overall nested analysis of variance with replicate population nested into the respective treatment combinations. In the challenge experiments, the treatment structure was a factorial composed of the factor combination high versus low CO2 selection × normal/high light challenge, or normal/low salinity challenge, respectively. Only if the replicate lines were heterogeneous at α < 0.05 did we proceed with pairwise comparisons of the variances within the same selection treatment (ambient or high CO2) among control and challenge. To this end, we used a direct F-test comparing the within-treatment variances in numerator (larger variance) and denominator (smaller variance). The variances were formulated using the five subreplicate means per treatment combination. Hence, numerator and denominator have 4 degrees of freedom each. The hypothesis being tested was that only high CO2 adapted replicate populations would diverge in their variance under challenge conditions.



Here, we show the results of the second year of selection of E. huxleyi populations under high CO2 conditions simulating future ocean acidification. Although there was a direct response to selection at high CO2 during the first year of the experiment (0–500 generations), no further increases in fitness relative to ambient CO2 selected populations were detected in the second year (after 750 and 1000 generations) (Figs. 3, 4). Direct responses were measured as the growth rate of the high CO2 adapted populations relative to the ambient CO2 adapted populations in a high CO2 assay environment. In contrast, correlated responses, measured as growth rate of high CO2 adapted populations assayed in the ancestral ambient CO2 environment, were found to have increased by 4% relative to ambient CO2 selected populations after 500 generations but showed a decrease by 3% relative to ambient CO2 selected populations after 750 and 1000 generations, respectively (Figs. 3, 4).

Figure 3.

Mean exponential growth rate (±1 SD) of replicate high CO2 adapted (gray background) and ambient CO2 adapted (white background) Emiliania huxleyi populations (N = 5) tested in the respective assay environment after 500 (blank), 750 (dotted), and 1000 (striped) generations.

Figure 4.

Mean fitness in direct and correlated response of high CO2 selected populations relative to ambient CO2 selected populations (N = 5) at the start of the experiment and after 500, 750, and 1000 generations. Fitness ratios are meant to illustrate the time course of direct and correlated response based on fitness data shown in detail in Figure 3. No error bars were calculated for the depicted fitness ratios because the evaluation of statistically significant effects was based on the treatment means, not the ratios. Direct response is the fitness of high CO2 selected populations relative to the fitness of ambient CO2 selected populations in the environment they were selected for (high CO2). Correlated response is the fitness of high CO2 selected populations relative to ambient CO2 selected populations in any other environment they were not selected for (e.g., ambient CO2).

Consistent with earlier assessments from reciprocal transplants after 500 generations (Lohbeck et al. 2012), we found low within-treatment variances in exponential growth rate among replicate selection lines also after 750 and 1000 generations, indicating phenotypic convergence in all selection lines under high CO2 selection (Fig. 3).


Populations transferred into a novel high-light environment showed an increased mean fitness by 3% under ambient and 7% under high CO2, respectively (Fig. 5A). There was no significant within-treatment heterogeneity among the four treatment combinations ambient versus high CO2 selection × challenge versus control (nested ANOVA, F3,16 = 1.865, P = 0.056). Because the P-value was close to statistical significance, we also compared the among treatment variances pairwise across selection environments. We found no significant difference of within-treatment variance between the challenge and control treatments in either the ambient or the high CO2 selected populations (all P > 0.4). Here, the high-light challenge did not detect any functional genetic divergence.

Figure 5.

Mean exponential growth rate (N = 3) of all individual replicate populations (N = 5), and their overall mean (horizontal line ±1 SD) selected under ambient CO2 (circles) and high CO2 (triangles) when tested under (A) control light (open symbols) and high-light challenge (closed symbols) and (B) control salinity (open symbols) and low-salinity challenge (closed symbols) assay conditions. Note that SD values (horizontal thick line) are so small that they are collapsed with the overall mean (thin horizontal line).

In contrast, the low-salinity challenge revealed that the high CO2 selected lines had diverged genetically. Selection lines transferred into a novel low-salinity environment showed a consistent decrease in fitness by an average of 24% in the ambient CO2 treatment and of 67% in the high CO2 treatment (Fig. 5B). The combination of high CO2 and low salinity showed an additive effect on mean fitness. When grown under lower salinity, a significant within-treatment variance indicated that phenotypes were no longer convergent (nested ANOVA, F3,16 = 19.94, P < 0.0001). We identified which within-treatment/among replicate variance produced this heterogeneity. The within-treatment variance of populations selected at ambient CO2 levels was not significantly different among challenged and control treatments (direct F-test to compare variances, F4,4 = 1.41, P > 0.1), but the high CO2 adapted populations had a much greater variance in the challenge treatment than in their respective control treatments (direct F-test to compare variances, F3,4 = 249, P << 0.0001). In line with this, under low-salinity challenge, all three subreplicates from one high CO2 selected population failed to reach sufficient cell numbers for cell counting and inoculation within the acclimation cycle of 12 days. Growth rates of zero (populations failed to grow within the 12-day batch cycle) were included in the graph (Fig. 5B), but were omitted from the statistical analysis. Because it neglects the most extreme phenotype seen in the low-salinity environment, the homogeneity of the fitness response is overestimated in the salinity challenge × CO2 adaptation treatment, making our statistical test for detecting functional genetic divergence conservative.


Replicate populations of the globally important marine phytoplankton species E. huxleyi were grown for 1000 generations under ambient and elevated CO2 levels, representing a contemporary control and a future ocean acidification scenario. Reciprocal transplants revealed adaptation to ocean acidification in populations selected under high CO2. Over the experiment, all replicate populations in a treatment group showed similar fitness responses, indicating phenotypic convergence, although it was not known whether this reflected underlying genetic convergence. Using an experimental approach that utilizes the hypothetical pleiotropic role of novel, fitness conferring mutations (Travisano et al. 1995; MacLean and Bell 2003; Ostrowski et al. 2008), we demonstrated functional genetic divergence underlying adaptation to high CO2 in single clone derived E. huxleyi populations. Although no functional genetic divergence was detected in a novel high-light environment, we found strikingly divergent phenotypes only among the populations adapted to high CO2, but not ambient CO2 levels, when challenged with low salinity.


The dynamics of adaptation to high CO2 in E. huxleyi populations using de novo variation happened in two phases during our selection experiment. Over the first 500 generations, we observed rapid adaptation characterized by a marked increase in fitness in high CO2 selected populations, after which fitness remained constant for the following 500 generations. This adaptive pattern suggests that high CO2 exerts strong selective pressure on E. huxleyi populations, and that initially beneficial mutations (either few mutations of large effect or many mutations of small effect) were common and rose to high frequencies within the populations quite rapidly (≤500 generations). Because there was no detectible rise in fitness in the second half of the experiments, one could conclude that further beneficial mutations are rare or even absent on this genetic background. If further rare beneficial mutations that cause fitness increases relative to the ancestor exist, then we expect those mutations to eventually occur in a subset of the high CO2 selected populations (Blount et al. 2008). Because it is very unlikely that rare mutations would simultaneously occur in all five replicate populations, this would lead to divergent direct responses to selection.

Although no obvious change in the direct response to selection occurred between generations 500 and 1000, we observed changes in the correlated response of high CO2 selected populations. Because these changes occurred in all replicate populations and were relatively rapid, and the variance in fitness within populations did not increase (direct F-test of variance comparison to compare variances, P > 0.5), they are consistent with continued adaptive evolution rather than a relaxation of selection pressure. The pattern seen here may indicate temporary stasis, slow-paced fitness increase not detectable on the short time scales investigated, or nontransitive increases in fitness, despite continued genomic adaptation. A similar pattern has also been reported by Lenski and colleagues (Lenski and Travisano 1991; Barrick et al. 2009) during the first few thousand generations of their E. coli selection experiment.


Phenotype convergence was observed during 1000 generations CO2 selection in the diploid marine phytoplankton species E. huxleyi. Convergent phenotypes may either result from the same sets of mutations being independently acquired in all replicate populations or from different mutations that produce a similar phenotype (Lenski et al. 1994; Travisano et al. 1995). If more than one beneficial mutation exists on this genetic background, and there is no hypermutable region or genetic switch involved, then it is unlikely that the same mutations occur twice, let alone in all five replicate populations (Woods et al. 2006). In this case, the expectation is that different evolutionary trajectories have resulted in similar phenotypic adaptations to high CO2, but divergent genetic bases for adaptation, in replicate populations. If this is true, pleiotropy has the potential to constrain adaptive evolution in natural E. huxleyi populations in the face of the complex environmental changes associated with an acidifying ocean.

To further investigate the underlying basis for phenotypic convergence, we exposed replicate populations already adapted to ambient or high CO2 to two novel environments. In a novel environment where fitness relies on sets of genes not under selection in the high CO2 environment, conditionally neutral mutations and pleiotropic interactions can be detected by an increased variance in fitness among replicate populations (Travisano et al. 1995; Ostrowski et al. 2008). However, the coincidence of a conditionally neutral mutation and a beneficial mutation shortly thereafter in the same cell lineage appears unlikely.

The two novel environments used here produced different outcomes. Exposure to a novel high-light environment did not reveal increased variance in fitness among replicate populations, indicating that no functional genetic divergence had occurred in genes involved in responses to non-stressful high-light levels that are within the acclimation capacity of genotypes of the study populations (Paasche 2002). In contrast, exposure to a stressful environment resulted in significant increase in variance in fitness among high CO2 selected populations but not among ambient CO2 selected populations. This indicates that different mutations have become fixed or frequent in the populations that have adapted to high CO2, and that several different genetic solutions exist to the problem of adaptation to ocean acidification in E. huxleyi populations (Travisano et al. 1995; Ostrowski et al. 2008).

We used a stressful environment that is probably at the far end of natural environmental conditions experienced by coastal E. huxleyi genotypes (Paasche et al. 1996). However, more moderate salinity changes are likely to occur regularly in the coastal environment and will be important to adaptation of coastal E. huxleyi populations to future ocean changes. The marked fitness decrease in populations exposed to low salinity and high CO2, relative to the fitness decrease under low salinity and ambient CO2, illustrates the importance of additive interactions of these two environmental factors. Complex changes of multiple environmental parameters can exert a very different selective force than anticipated from single factor selection experiments and, in addition to our results on pleiotropic effects, potentially play an important role in the adaptive evolution of natural E. huxleyi populations to ocean acidification. In particular, the range of adaptive responses to elevated CO2 may be constrained by having to maintain the ability to deal with fluctuations in salinity. This may be especially important for calcifying phytoplankton populations that live in or close to estuaries with widely fluctuating salinity levels (Paasche et al. 1996).

That we could not find divergent pleiotropic effects in the high light but in the low-salinity environment may have several explanations. First, the two challenge experiments were performed about 250 generations apart from each other. Although we cannot rule out an effect of time, this explanation appears unlikely because we had already found a marked adaptive response in all replicates at the 500 generations mark. The different outcome of the chosen novel environments to produce divergent phenotypes may also be attributed to qualitative differences between the environments. The low-salinity environment may fall outside the boundaries for an environmental parameter for which the population has acquired adaptive plasticity in the form of acclimation whereas the high-light environment does not (Alpert and Simms 2002; Paasche 2002). In consequence, the low-salinity environment is stressful, sensu decreasing Darwinian fitness, whereas the benign high-light environment actually increases fitness in our experiment. Depending on the cause of the pleiotropy involved, it may only become apparent in a stressful environment. We would expect this to be the case for trade-offs that are the result of energy allocation, for example. Alternatively, the inability to detect divergent pleiotropy may reflect the absence of such pleiotropic mutations in genes relevant to the high-light environment. The same genes may be under selection in the ambient and high-light environments. This assumption would further suggest a very limited linkage of genes relevant to CO2 and light adaptation in E. huxleyi. The reciprocal assay environments used to test for adaptation in the CO2 selection experiment may also be considered as novel environments. That we did not find increased variance when transplanting high CO2 selected populations into ambient CO2 may indicate, similar to the latter explanation brought for high light, that the same genes are relevant to fitness under high and ambient CO2 and therefore adaptation to one CO2 environment results in a homogeneous response in a certain range of other CO2 environments.


One objective of our selection experiment was to incorporate an established model system from biological oceanography, the coccolithophore E. huxleyi, into microbial experimental evolution. We used a modified microbial selection experiment spanning 1000 generations of selection to investigate adaptive evolution in this globally important phytoplankton species to ongoing ocean acidification.

Interdisciplinary studies, assessing adaptive evolution in marine phytoplankton species to climate change scenarios, should be more widely used to fill the data gap commonly masked by extrapolating short-term physiological studies to predict phytoplankton responses in a future ocean (Riebesell et al. 2011; Collins 2012; Reusch and Boyd 2013). New model systems also hold the opportunity for evolutionary biology to test for generality of concepts beyond the limited group of classical model organisms. In particular, diploid eukaryotes from the marine phytoplankton provide important study systems for evolutionary biology because they represent deeply divergent polyphyletic groups (Falkowski et al. 2004) that can be used to study fundamental evolutionary concepts. For example, the bulk of comparative studies on the consequences of diploidy versus haploidy for the rate of adaptation come from yeast (e.g., Zeyl et al. 2003) while phylogenetically deeply divergent eukaryotic (diploid and haploid) and prokaryotic (haploid) phytoplankton species may allow for more systematic and inclusive tests of the role of ploidy for adaptation. Finally, there are also obvious drawbacks of using marine algae in experimental evolution under climate change scenarios. Most challenging are probably the sophisticated and labor-intensive culturing work and seawater carbonate chemistry manipulations and measurements that by far exceed space and time requirements of classical model species such as E. coli or yeast. Nonetheless, considering the need for reliable estimates of future ocean responses to ongoing climate change (Bell and Collins 2008; Riebesell et al. 2011; Boyd et al. 2010), the gain in value from an ecological and socioeconomical perspective makes such applied evolution using marine microbes worthwhile.


We thank J. Meyer, R. Klapper, L. Miersch, and K. Beining for laboratory assistance. This project was funded by the German Federal Ministry of Education and Research (program “BIOACID”) and the Petersen foundation fellowship. The authors have no conflict of interest to declare.