• All data and metadata to be archived at the U.S. National Science Foundation Biological and Chemical Oceanography Data Management Office (BCO-DMO,


Increasing pCO2 (partial pressure of CO2) in an “acidified” ocean will affect phytoplankton community structure, but manipulation experiments with assemblages briefly acclimated to simulated future conditions may not accurately predict the long-term evolutionary shifts that could affect inter-specific competitive success. We assessed community structure changes in a natural mixed dinoflagellate bloom incubated at three pCO2 levels (230, 433, and 765 ppm) in a short-term experiment (2 weeks). The four dominant species were then isolated from each treatment into clonal cultures, and maintained at all three pCO2 levels for approximately 1 year. Periodically (4, 8, and 12 months), these pCO2-conditioned clones were recombined into artificial communities, and allowed to compete at their conditioning pCO2 level or at higher and lower levels. The dominant species in these artificial communities of CO2-conditioned clones differed from those in the original short-term experiment, but individual species relative abundance trends across pCO2 treatments were often similar. Specific growth rates showed no strong evidence for fitness increases attributable to conditioning pCO2 level. Although pCO2 significantly structured our experimental communities, conditioning time and biotic interactions like mixotrophy also had major roles in determining competitive outcomes. New methods of carrying out extended mixed species experiments are needed to accurately predict future long-term phytoplankton community responses to changing pCO2.

Future marine phytoplankton assemblages will be influenced by environmental changes that include transfer of fossil fuel-derived CO2 from the atmosphere to the surface ocean. Because of the tremendous physiological variability found among microalgal taxa, it is likely that differential responses to consequent ocean acidification will provide a selective advantages or disadvantages, resulting in “winners” and “losers” (Rost et al. 2008; Hutchins et al. 2009; Boyd et al. 2010). These two outcomes may be expressed either in terms of increased or decreased reproductive fitness (e.g., growth rates), or in terms of competitive success (relative abundance or biomass in mixed communities).

Experiments using natural communities have manipulated the partial pressure of CO2 (pCO2) in seawater to characterize algal physiological responses that may help us predict future phytoplankton community structure (Riebesell 2004; Kim et al. 2006; Hare et al. 2007; Feng et al. 2009, 2010). In contrast to these ocean acidification simulation experiments that typically last at most a few weeks, although, phytoplankton populations in nature will encounter gradually changing conditions over time scales of years to centuries. The long-term evolutionary responses of phytoplankton groups whose competitive interactions ultimately determine algal community structure have only begun to be examined, so we can only speculate at present on the composition of these future “greenhouse” phytoplankton assemblages.

Aside from assemblage-level effects, there is also a limited amount of information available on potential long-term evolutionary responses of single species of phytoplankton to high pCO2 (Collins and Bell 2004, 2006; Collins 2010; Müller et al. 2010; Crawfurd et al. 2011; Lohbeck et al. 2012). After 1000 generations of growth at high CO2, a freshwater green alga revealed no adaptive evolutionary response of specific growth rate (Collins and Bell 2004). Müller et al. (2010) found little difference between short- and long-term effects of high CO2 on two species of coccolithophores. The marine diatom Thalassiosira pseudonana revealed no evidence of genetic adaptation to high CO2 over 3 months (Crawfurd et al. 2011), but another recent study documented increases in growth rates and calcification in coccolithophore cultures adapted to high CO2 conditions (Lohbeck et al. 2012).

Dinoflagellates within the class Dinophyceae are important members of ocean ecosystems, especially in the coastal zone. These microscopic protists are typically motile with the aid of their two flagella, and many exhibit extensive diel vertical migrations within the water column. They are evolutionarily ancient organisms that exhibit diverse morphological forms, life history strategies, and modes of nutrition ranging from autotrophy to mixotrophy to heterotrophy (Smayda 1997, 2002). In particular, dinoflagellates have unique chromosomal attributes, including size, structure, and composition (Rizzo 2003). Some members of this class have environmental, economic and human health significance as they are capable of forming dense “blooms,” as well as manufacturing potent neurotoxins that can be accumulated through trophic transfer in marine food webs (Fu et al. in press).

To address some of the unknowns regarding the long-term assemblage-level responses to changing CO2 conditions, we used a novel experimental design that compared the outcome of competition in short-term (2-week) natural dinoflagellate community pCO2 manipulations with the results of competition between the same species in artificial communities after conditioning to the same pCO2 treatments in clonal cultures for 1 year. An objective of this work was to determine if short-term incubation experiments are reasonable proxies for predicting the effects of long-term processes on community structure, and thus address the question: Is the outcome of multispecies competitive interactions the same under short-term and long-term selection by pCO2?

Materials and Methods


A flow chart of the experimental design is shown in Figure 1, including collection and incubation of the natural bloom in a short-term (2 weeks) pCO2 experiment, isolation of cells of all four species from each pCO2 treatment, conditioning of the isolates at the pCO2 from which they were isolated for 1 year, and recombining isolates into artificial communities to compete for 2-week periods following 4, 8, and 12 months of conditioning. Growth rates of each species were assessed during conditioning in unialgal cultures at the 8-month timepoint, as well as in mixed communities during the initial bloom experiment and the 12-month artificial community experiment. “Switch” competition experiments in which clones conditioned at each pCO2 were competed in artificial communities at the other two pCO2 levels were also performed after 12 months.

Figure 1.

Flow chart showing the sequence of the major components of the year-long experiment.


A mixed natural dinoflagellate bloom dominated by Lingulodinium polyedrum, Prorocentrum micans, Alexandrium sp., and Gonyaulax sp. at a total cell density of approximately 700 cells per mL was collected off Venice Beach, California, in September, 2009. This large regional bloom extended throughout the Southern California Bight region. Samples were collected near shore for both the initial incubations and all experimental dilution water used throughout the 12-month experiment.

The experiment was incubated in the laboratory at 18°C under 90 photons m−2 s−1 of cool white fluorescent illumination on a 14-h light : 10-h dark cycle. Triplicate sterilized 1 L polycarbonate bottles were gently bubbled (60 bubbles min−1) using commercially prepared air/CO2 mixtures (230, 433, and 765 ppm, Praxair Gas). Preliminary experiments verified that growth rates of cultures bubbled at this rate were not significantly different from those of unbubbled cultures (data not shown), and these methods have been employed for other CO2 experiments (Fu et al. 2007; Hutchins et al. 2007), including dinoflagellate studies (Fu et al. 2008, 2010). Filtered seawater was amended with L1/20 nutrient, vitamin, and trace metal concentrations (Guillard and Hargraves 1993), except NH4Cl+ was substituted for NaNO3 and silicate was omitted. Nutrient concentrations at the Redfield ratio (by atoms) of 16 N : 1 P (Redfield 1958) were added initially to the incubation bottles, and replenished once at the 1 week dilution.

The CO2-amended treatments were maintained in active growth using semicontinuous culture methods (Tatters et al. 2012). Each bottle was diluted to the original time-zero in vivo chlorophyll a fluorescence value after 1 week with nutrient-amended filtered seawater. Aliquots were removed initially, and after 1 and 2 weeks for examination of carbonate buffer system parameters and community structure using microscopic cell counts. Samples for cell counts were obtained at the 1 week timepoint (after dilution), and after the 2 week incubation, to calculate acclimated growth rates (1–2 week rates) and final abundances of all species.


Three individual cells representing the four dominant genera were isolated from each incubation bottle at the end of the 2-week incubation of the natural community, and maintained in long-term culture (52 weeks) at the pCO2 from which they were obtained under conditions of temperature, light, nutrients, CO2 bubbling, etc., identical to the 2-week natural community experiment. Cultures were maintained in exponential phase using autoclave-sterilized enriched seawater growth medium with semicontinuous weekly dilutions based on specific growth rates within each bottle, calculated as in Tatters et al. (2012). The approximate number of generations during this time period was: L. polyedrum (48–62), P. micans (58–71), Alexandrium sp. (34–38), and Gonyaulax sp. (75–126).


The conditioned clonal cultures were recombined into artificial communities after 4, 8, and 12 months in the same relative proportions and cell densities as the original natural bloom assemblage. The 8- and 12-month experiments used triplicate communities of all four species, but because of logistical limitations, the 4-month experiment used only L. polyedrum, P. micans, and Alexandrium sp. in duplicate communities. Because Gonyaulax sp. was not included in this preliminary 4-month experiment and replication was different, it is not fully comparable to the other experiments. The dinoflagellates in the artificial community trials were allowed to compete under identical experimental conditions of light, temperature, nutrient availability, and pCO2 for the same time period and diluted exactly as in the original natural bloom incubation. Samples were collected for cell counts and carbonate system parameters (Table 1) in all experiments.

Table 1. Measured seawater carbonate buffer system values, pH, and total dissolved inorganic carbon (DIC) in the initial 2 week incubation with the natural bloom sample, in the artificial community competition experiments at 4, 8, and 12 months, and in the single species cultures during long-term conditioning at the 8-month timepoint. Also shown are pCO2 values calculated from the two measured parameters. For clarity, in the text pCO2 values of 230 to 336 are referred to as “low,” values of 433 to 506 are referred to as “medium,” and values of 709 to 792 are referred to as “high.”
2 weeks   
8 months   
12 months   
8 months, cultures   
 L. polyedrum   
P. micans   
Alexandrium sp.   
Gonyaulax sp.   


A triplicated CO2 switch experiment at the 12-month timepointconsisted of “switching” low CO2-conditioned cell lines to medium and high CO2 (low→medium and low→high), medium CO2-conditioned cell lines to low and high CO2 (medium→low and medium→high), and high CO2-conditioned clones to low and medium CO2 (high→low and high→medium). The same experimental bottles employed in the 12-month artificial community experiments described above were used to provide appropriate controls of low→low, medium→medium, and high→high. Other than the switched pCO2 treatments, protocols for these trials were exactly the same as in the other artificial community experiments.


Cell counts and growth rates

Growth rates were measured for each species in mixed communities during the second week of the initial natural community incubation, and during the 12-month and switch artificial community experiments. Because of logistical limitations and the fact that dinoflagellates cannot be cryopreserved for subsequent growth rate measurements, growth rates were determined immediately for all isolates in unialgal culture only at the 8-month timepoint. These values are representative of their long-term steady-state exponential growth rates throughout the conditioning period. Final cell abundances of each species were measured in every natural or artificial community competition experiment. Algal cells were preserved in acidified Lugol's solution and enumerated using an Accu-Scope 3032 inverted microscope using the Utermöhl method (Utermöhl 1931).

Carbonate buffer system

Dissolved inorganic carbon analysis used a CM5230 CO2 coulometer (UIC; King et al. 2011). pH was determined on freshly collected samples using a calibrated Orion 5-star plus pH meter using an NBS buffer system with three-point calibration. Experimental pCO2 was calculated using CO2SYS software as in Tatters et al. (2012) (Table 1). Because of unavoidable minor variability in calculated pCO2 levels between experiments (largely from differences in batches of commercial gas mixtures), for clarity, throughout the text pCO2 values of 230 to 336 ppm are referred to as “low,” values of 433 to 506 ppm are referred to as “medium,” and values of 709 to 792 ppm are referred to as “high.”

Multivariate statistical methods

Multivariate analyses used the PRIMER v6 statistics package (Clarke and Warwick, 2001) with the permutational analysis of variance/multivariate analysis of variance (PERMANOVA) add-on. Final cell abundances for each species from replicate bottles were square-root transformed prior to community structure comparisons based on Bray–Curtis similarities and log-transformed species growth rates compared based on Euclidean distance measures. ANOSIM permutation tests were used to test the impact of differing pCO2 competition levels on overall community structure and on relative abundance of the four species at the end of all incubations, and on their growth rates in the initial natural community, 12-month and switch experiments. These tests resulted in R values and significance levels where R = 0 implies no difference among groups, and R = 1 suggests that group separation is so large that all dissimilarities among groups are larger than any dissimilarity within them (Clarke and Warwick 2001).

Data from the natural community experiment and from competition trials after 8 and 12 months were further combined for PERMANOVA analyses to test for significant differences among and within predefined groups in response to both the pCO2 competition levels and the differing periods of conditioning to these pCO2 concentrations. PERMANOVA resulted in Pseudo-F and significance levels, where Pseudo-F = 1 implies a large overlap among sample groups being compared, whereas Pseudo-F > 1 indicates little or no overlap between them (Anderson et al. 2008). PERMANOVA also allowed us to test for interaction between the factors pCO2 competition level and conditioning period in forcing overall community structure.

For the 12-month switch competition trials, we used final cell abundances from all treatments to examine the comparative effects of differing pCO2 competition levels and differing pCO2-conditioning levels. Using a two-way crossed design for the ANOSIM routine, we tested the average effect on overall community structure and on the four individual species separately of pCO2 levels during competition removing differences in conditioning pCO2, and the average effect of conditioning pCO2 removing differences in competition pCO2 (Clarke and Warwick 2001).



Community structure

The natural bloom composition at the time of collection was dominated by L. polyedrum (82.5%), followed by P. micans (9.0%), Gonyaulax sp. (4.8%), and Alexandrium sp. (3.7%) (Fig. 2A). Other phytoplankton species were also present but they comprised < 1% of total cell abundance within the community, therefore we considered only the four dominant species. Each of the pCO2 treatments yielded a different dinoflagellate assemblage at the end of the 2-week incubation, indicating that community structure was strongly altered by pCO2 (average R = 0.97 at P = 0.004, one-way ANOSIM; Table 2). Treatment-specific trends were consistent within the triplicate experimental flasks, as evidenced by the clustering of Bray–Curtis similarities for each group of replicates illustrated in a nonparametric, multidimensional plot (MDS; Fig. 2A).

Figure 2.

Relative abundance community composition graphs (A) from the original bloom sample, (B) at the end of the corresponding 2-week natural community pCO2 incubation experiment, and at the end of the pCO2-conditioned artificial community competition experiments at (C) 4 months, (D) 8 months, and (E) 12 months.

Table 2. Results of one-way ANOSIM comparing overall algal community structure across three different pCO2 levels (low, ambient, and high) for each experiment after 0, 4, 8, and 12 months of conditioning of the clonal isolates at the three different pCO2 concentrations. All global tests were run at 280 permutations except for the 4 months dataset that was limited to 15 permutations. P = significance level; ns = not significant at P < 0.05
  1. 1R values typically range from 0 to 1, where R equal to 0 indicates community structures are not dissimilar among sample groups (H0 hypothesis), whereas R equal to 1 indicates strong dissimilarities among sample groups.

Community structure
 0 months0.970.004
 4 months0.67ns
 8 months0.890.004
 12 months0.600.004

Final relative abundance

After the original community incubation the final relative abundance of the dominant species L. polyedrum was highest in the low CO2 treatment (90%) and declined progressively with rising pCO2 at medium pCO2 (73%) and high pCO2 (67%). In contrast, relative abundance increased progressively with increasing pCO2 for Gonyaulax sp. (2%, 3%, and 8%). The relative abundances of the other two species were also highest at one or both of the two most elevated pCO2 levels, although stepwise trends were less well defined (P. micans 2%, 17%, and 13%; Alexandrium sp. 6%, 5%, and 10%; Fig. 2B).

1- to 2-week growth rates

Growth rates of L. polyedrum in the natural community incubation declined moderately (global R = 0.52, P = 0.029; ANOSIM) as pCO2 level increased (Fig. 4A). pCO2 competition level had the strongest effect on growth rates of P. micans (global R = 0.68, P = 0.007), resulting in higher rates at high pCO2 (Fig. 4A). In contrast, there was no significant effect of the three pCO2 levels on the growth rates of either Alexandrium sp. (global R = 0.39, P > 0.05) or Gonyaulax sp. (global R = 0.04, P > 0.05).

Figure 3.

Cluster-diagram showing Bray–Curtis similarity for the dinoflagellate assemblages analyzed (A) after the initial 2-week natural community pCO2 incubation, (B) after 4 months of conditioning, (C) after 8 months of conditioning, and (D) after 12 months of conditioning for each of the three pCO2 treatments (low, ambient, and high).

Figure 4.

Specific growth rates of the four species in the three pCO2 treatments, (A) during the second week of the initial natural community incubation experiment, (B) in the clonal isolates after 8 months of conditioning in unialgal cultures, and (C) during the second week of the 12-month artificial community recombination experiment. Error bars represent the standard deviations of triplicate bottles.


Growth rates of each clone were assessed in triplicate after 8 months of conditioning under their respective pCO2 conditions. A striking observation was that growth rates of the dinoflagellates at all pCO2 levels were in general much higher in the original natural community incubation (Fig. 4A) than in unialgal cultures (Fig. 4B), despite identical growth conditions. For L. polyedrum, clonal culture growth rates were 50% (low pCO2), 32% (medium), and 27% (high) lower than in the mixed natural community incubation, whereas for P. micans these rates decreased by 65% (low), 53% (medium), and 82% (high). Alexandrium sp. also showed large decreases when brought into clonal culture, with growth rates that were 64% (low), 38% (medium), and 72% (high) lower compared to the original experiment. Gonyaulax sp. growth rates were least affected by unialgal culture conditions, with rates that were 33% lower (low), 44% higher (medium), and 14% lower (high) than in the natural community experiment (Fig. 4B).

The growth rates of some of the dinoflagellates in unialgal culture were affected by pCO2, despite being lower overall than in the mixed natural community. Specific growth rates for L. polyedrum were 0.12 day−1 (low), 0.11 day−1 (medium), and 0.09 day−1 (high; Fig. 4B). The 22% growth rate decrease from low to high pCO2 was significant (P < 0.05). Growth rates of P. micans were 0.11 day−1 (low), 0.13 day−1 (medium), and 0.14 day−1 (high; Fig. 4B), however the 15% increase from low to high pCO2 was not significant (P > 0.05). Clones of Gonyaulax sp. grew the fastest at all CO2 concentrations, ranging from 0.17 day−1 (low) to 0.22 day−1 (medium), and 0.25 day−1 (high; Fig. 4B). Growth rates were thus 32% higher at high versus low pCO2 (P < 0.001). Growth of Alexandrium sp. was the slowest of the four dinoflagellates at all CO2 concentrations, at 0.06 day−1 (low), 0.07 day−1 (medium), and 0.07 day−1 (high; Fig. 4B), and was not significantly different between any of the treatments (P > 0.05). Analysis of the growth rates of all species combined (i.e., independent of species) suggested a strong overall effect of pCO2 on growth rate (R = 0.704, P = 0.004).


Final community structure

Community structure varied significantly in response to different pCO2 levels in two out of the three artificial competition trials (Table 2). A trend of a steady decline of R-values with increasing conditioning period was revealed as illustrated in side by side MDS plots for each of the experiments, from the natural community experiment (Fig 3A) to the 4-month (Fig. 3B), 8-month (Fig. 3C), and 12-month (Fig. 3D) artificial community competition trials. No conditioning resulted in the highest dissimilarities at the end of the initial natural community incubation, whereas increasing conditioning periods resulted in still significantly different, but less dissimilar, assemblages after 8 months (average R = 0.89 at P = 0.004) and 12 months (average R = 0.6 at P = 0.004). Our 4-month competition experiment with two replicates and a lower number of competing species (3) resulted in a limited number of permutation runs for the ANOSIM test (15) and was the only trial that did not yield a significance level (at P < 0.05) when overall community structure was compared. PERMANOVA (Type 1, sequential at 9999 permutations) analyses of three out of the four competition experiments (4-month trial excluded because of differences in design) indicated that both factors, pCO2 level and conditioning time, interacted in forcing overall community structure (Pseudo-F = 4.9 at P = 0.0001). PERMANOVA test results further suggested the effect of different conditioning periods (Pseudo-F = 161.8 at P = 0.0001) was slightly stronger than that of differing pCO2 competition levels (Pseudo-F = 17.9 at P = 0.0001).

Final relative abundance

The strongest effect of changed pCO2 level was observed for the final abundance of L. polyedrum and P. micans (global R = 0.54 and 0.55, respectively; P = 0.0001; ANOSIM), with lesser but still significant effects for Alexandrium sp. and Gonyaulax sp. (global R = 0.32 and 0.24 and P = 0.004 and 0.033, respectively; Table 3A). Generally, conditioning time appeared to have a strong effect on the outcome of the competition trials for all four species with a relatively lower global R value of 0.44 (P = 0.0001) for L. polyedrum compared to 0.73 to 0.82 (P = 0.0001; ANOSIM) for the remaining three species (Table 3A).

Table 3. Statistical results from two-way ANOSIM examining the effect of (A) differing competing pCO2 level (low, ambient, and high) and preconditioning time (0, 8, and 12 months) on relative abundance changes for each of the four dinoflagellate species; and (B) the effect of pCO2 competition level versus pCO2 conditioning levels prior to the 12-month switch experiment. All global tests were run at 9999 permutations. P = significance level; ns = not significant at P < 0.05
 Competing pCO2Conditioning Time
L. polyedrum0.540.00010.440.0001
Alexandrium sp.0.320.0040.730.0001
Gonyaulax sp.0.240.0330.810.0001
P. micans0.550.00010.820.0001
 Conditioning pCO2 LevelCompeting pCO2
L. polyedrum0.490.00030.14ns
Alexandrium sp.0.21ns0.12ns
Gonyaulax sp.0.16ns0.07ns
P. micans0.610.00020.560.0002

L. polyedrum comprised a significantly larger percentage of the community at low versus medium or high pCO2 levels in all three artificial community experiments (Fig. 2C–E). However, L. polyedrum became a significantly less successful competitor at all pCO2 levels with progressively increasing conditioning time, from 0 to 4 to 8 and 12 months (Fig. 2B– E).

P. micans was an increasingly successful competitor relative to the other dinoflagellates with progressively longer conditioning periods (Fig. 2). The differing competition pCO2 treatments affected P. micans relative abundance within each of the incubations, however, trends differed at the 12-month timepoint (highest abundance at low pCO2) from those in the initial, 4- and 8-month trials (higher abundance at higher pCO2). P. micans always increased in abundance at least twofold in each treatment for all recombination trials. It was clearly the winner at high CO2 after 4 months (Fig. 2C), and the dominant competitor at all CO2 concentrations after 8 months of conditioning (Fig. 2D). At the 12-month timepoint, P. micans made up > 50% of the community at low and medium pCO2, and closely approached that relative abundance at high pCO2 (Fig. 2E).

Gonyaulax sp. was always a winner in each artificial community trial (Fig. 2D, E), and similar to P. micans, its competitive success was increased by longer conditioning time (Table 3A). Gonyaulax sp. thus competed better in both the 8- and 12-month trials than in the original short-term natural community experiment. Its positive relative abundance trend with pCO2 was conserved throughout the year-long experiment, with the highest final abundance always observed at high pCO2 (Fig. 2).

Alexandrium sp. increased in relative abundance during each recombination trial (Fig. 2) and, similar to P. micans and Gonyaulax sp., conditioning time positively affected its competitive ability. The clones conditioned to medium and high pCO2 generally competed the best as reflected in the specific growth rate (Fig. 4A), but the significance level of this pCO2 effect on competitive success was lower than for either L. polyedrum or P. micans (R = 0.32, P = 0.004, Table 3A). Alexandrium sp. competed more successfully after 4 months at high pCO2 (Fig. 2C), similar to the original short-term incubation (Fig. 2B), and after 8 months it was a strong competitor under all pCO2 concentrations (Fig. 2D). This species achieved a higher relative abundance after 12 months at medium CO2, followed closely by high CO2 (Fig. 2E).

1- to 2-week growth rates

When the individually conditioned clones were recombined in the 12-month artificial community experiment, in most cases their growth rates in mixed culture were substantially higher than in unialgal culture (Fig. 4C). The exception to this was L. polyedrum, which exhibited a range of growth rates that compared to the unialgal culture rates were 73% lower (low), 78% higher (medium), and 53% lower (high). In contrast, P. micans growth rates in coculture with the other species increased by 339% (low), 287% (medium), and 257% (high). Gonyaulax sp. growth rates exhibited a mixed response to coculturing, ranging from 30% lower (low) to 30% higher (medium) and 86% higher (high). Alexandrium sp. growth rates increased most dramatically in the mixed communities, with increases of up to 650% across all pCO2 levels (compare Fig. 4B with Fig. 4C).

Effects of pCO2 on the growth rates of L. polyedrum were not significant in the 12-month artificial community experiment (global R = 0.08, P > 0.05; ANOSIM; Fig. 4C), unlike the natural community experiment (Fig. 4A) but similar to the 8-month unialgal culture measurements (Fig. 4B). pCO2 competition level significantly affected the growth rates of P. micans (global R = 0.49, P = 0.021), with values that were inversely related to pCO2 (Fig. 4C). This contrasted with the positive relationship in the original experiment (Fig. 4A), as well as with the lack of significant effect of pCO2 in the unialgal cultures (Fig. 4B). In the 12-month artificial community experiment, increasing pCO2 positively affected the growth rates of Gonyaulax sp. (global R = 0.49, P = 0.043), similar to the 8-month unialgal cultures (Fig. 4B), but unlike the original incubation experiment where its growth rates did not respond significantly to pCO2 (Fig. 4A). At 12 months, the effects of pCO2 competition level on Alexandrium sp. growth rates were not significant (global R = 0.24, P > 0.05), as was the case in both the natural community experiment and the unialgal cultures.


Final community structure

The pCO2 switch experiments at the 12-month timepoint (Fig. 5) showed that both pCO2 level during competition (global R = 0.60 at P = 0.0001) as well as pCO2 concentration during the conditioning period (global R = 0.78 at P = 0.0001) were forcing factors on community structure (two-way ANOSIM test). CO2 concentrations during conditioning thus appeared to have a slightly stronger effect on community structure than the pCO2 levels the algae were exposed to during the 2-week competition.

Figure 5.

Relative abundance composition graphs after 2-week pCO2 switch or reversion artificial community incubation experiments carried out at the 12-month timepoint, in which the isolates were allowed to compete at the two pCO2 levels other than the one that they had been isolated from and conditioned to. Controls in which the competition experiment was conducted at the same pCO2 to which the isolates were conditioned are the same data shown in Figure 2E, and are included again here for direct visual comparison with the switch treatment outcomes.

Final relative abundance

Competition pCO2 and conditioning pCO2 levels had varying effects on final relative abundances during the switch experiment. There was a significant effect of prior pCO2 conditioning on the dominance of L. polyedrum (global R = 0.49, P = 0.0003; Table 3B), but this was primarily a function of the cultures conditioned at low pCO2 (pairwise ANOSIM test results not shown). L. polyedrum competed best in all treatments following conditioning at low pCO2, and the clones conditioned at medium or high pCO2 were relatively poor competitors in the pCO2 switch experiments (Fig. 5D–F). Although its apparent relative abundance was somewhat lower in the low→low treatment than in the low→medium (Fig. 5B) and low→high (Fig 5C) treatments, its overall competitive success was not significantly different across the three pCO2 levels in any of the switch experiments (global R = 0.14, P > 0.05; Table 3B).

The influence of prior conditioning pCO2 on P. micans dominance was a significant influence at all three pCO2 levels (global R = 0.61, P = 0.0002; Table 3B), and this was greater than for the primarily low pCO2-conditioned treatment effect on L. polyedrum (pairwise ANOSIM test results not shown). Also unlike L. polyedrum, pCO2 competition level during the artificial community incubations significantly influenced the relative dominance of P. micans (global R = 0.56, P = 0.0002; Table 3B). P. micans was still the most abundant species at low→medium (Fig 5B) and low→high (Fig 5C), but it did not dominate the community to the same degree as in the low→low treatment (Fig. 5A). For clones conditioned to medium pCO2 levels, P. micans again was the dominant species after the incubation with the highest relative abundance attained at medium→low (Fig 5D). The medium→medium (Fig. 5E) and medium→high (Fig. 5F) pCO2 treatments yielded approximately 50% P. micans, whereas its relative abundance was higher in the high→low (Fig. 5G) and high→medium (Fig. 5H) treatments than in the high→high (Fig. 5I).

Alexandrium sp. was a good competitor in all switch communities (Fig. 5), but statistical comparisons showed no significant effects of either prior pCO2 conditioning level or the pCO2 treatments during the competition experiments (global R = 0.21 and 0.12, respectively; P > 0.05; Table 3B). Relative abundance trends for Gonyaulax sp. indicated it was in general a less successful competitor in these switch experiments (Fig. 5), but like Alexandrium sp. neither conditioning history nor experimental pCO2 levels (global R = 0.16 and 0.07, respectively; P > 0.05; Table 3B) influenced dominance trends.

1- to 2-week growth rates

In the 12-month switch experiment, growth rates were significantly affected by conditioning pCO2 for all species (L. polyedrum global R = 0.47, P = 0.002; P. micans global R = 0.56, P = 0.005; Alexandrium sp. global R = 0.42, P = 0.036; Gonyaulax sp. global R = 0.30, P = 0.037; ANOSIM). In contrast, competition pCO2 levels during the 2-week competition experiment only had a significant effect on growth rates of L. polyedrum (global R = 0.48, P = 0.014).

For L. polyedrum, there was a significant increase in growth rates for the cultures conditioned at low pCO2 (LL) when grown in the communities incubated at medium (LM) or high (LH) levels (p < 0.05). There were no significant differences in L. polyedrum growth rates between any of three treatments for the medium or high pCO2-conditioned cultures (Fig. 6A, P > 0.05). The same trends were observed for Gonyaulax sp. (Fig. 6D), but there were no clearly defined trends in growth rate across the three competition pCO2 levels for P. micans (Fig. 6B) or Alexandrium sp. (Fig. 6C). In general, for each of the four species, pairwise comparisons from the switch experiment did not provide evidence that extended conditioning resulted in faster growth rates during mixed culture competition at the conditioning pCO2 versus the other two pCO2 treatments.

Figure 6.

Specific growth rates for all four species during the second week of the 12-month switch experiments, including cultures conditioned at low pCO2 and competed at low (LL), medium (LM), and high (LH) pCO2; cultures conditioned at medium pCO2 and competed at low (ML), medium (MM), and high (MH) pCO2; and cultures conditioned at high pCO2 and competed at low (HL), medium (HM), and high (HH) pCO2. (A) Lingulodinium polyedrum, (B) Prorocentrum micans, (C) Alexandrium sp., and (D) Gonyaulax sp. Error bars represent the standard deviations of triplicate bottles.


We initially examined competitive dynamics in a 2-week CO2 incubation experiment using a natural community composed of four dominant dinoflagellate species. We then isolated clonal cultures of each species from these presumably genetically and phenotypically diverse populations, with the implicit assumption that by the end of the natural community experiment, the best-adapted variants of each species would have become dominant in each pCO2 treatment. These clonal isolates were then conditioned separately for 1 year to the abiotic factor CO2, without the adaptive evolutionary pressure of interactions between species. Periodically, these isolates were reassembled in artificial communities analogous to the original natural assemblage, to assess potential evolutionary influences of CO2 conditioning on their competitive success at each pCO2 level.

We employed artificial communities because enclosure artifacts can confound interpretation of long-term enclosed natural plankton assemblage experiments (Caron and Countway 2009). One natural plankton community mesocosm experiment spanning several years suggested that it is virtually impossible to predict long-term trends in species abundances, because of progressively greater chaotic behavior over time (Benincà et al. 2008). Our novel approach of periodically recombining and competing individually conditioned isolates was intended to minimize some of the stochasticity inherent in long-term conditioning of the natural dinoflagellate assemblage, by providing greater predictability in a simplified format. However, clearly our artificial communities differ in many ways from the original natural community, despite being composed of the same four dominant species grown under the same set of environmental conditions. To better distinguish between the potential effects of this winnowing of a set of diverse populations down to a handful of clonal culture lines, and the selective effects of pCO2 conditioning time, future studies using similar experimental designs could be improved by conducting an additional set of artificial community experiments directly after isolating the cultures.

The results of our 4-, 8-, and 12-month acclimated artificial community competition experiments had trends that were in some respects similar to, and other ways different from the outcome of the original mixed bloom incubations. Although the “winners” of the artificial community competitions in terms of competitive dominance were not the same as in the natural community experiment (i.e., P. micans instead of L. polyedrum), relative abundance trends of individual species (i.e., L. polyedrum) with pCO2 were often parallel to those in the original incubation. These observations provide some support for the trends in the responses of species abundance to pCO2 changes in short-term natural assemblage experiments, while emphasizing that such brief experiments have limited predictive power for long-term community structure outcomes.

There are many possible reasons for the differences in final relative abundance between the original short-term natural community and the long-term conditioned artificial community experiments. The assumption that the outcome of competition between our sympatric dinoflagellate species hinges only on a single factor (i.e., pCO2) may be unwarranted, as during the conditioning period the clones also may become acclimatized to and/or selected by other laboratory culture conditions. Statistical analysis suggested conditioning time was slightly more influential than pCO2 levels in structuring the artificial communities, as indicated by the emergence of the “lab weed” P. micans, which competed progressively better over time and eventually dominated in all treatments. In fact, length of time in culture became progressively more influential than pCO2 on overall community structure, suggesting the possibility that if we had extended the experiments long enough, the pCO2 treatments might have eventually become irrelevant to the species composition of the assemblage. It is obvious that like short-term manipulations, long-term culture studies have artifacts of their own, and caution should be used when deriving information from older isolates because of accumulated culture selection artifacts. In fact, many phytoplankton clones used as model organisms for physiological, genetic, and evolutionary studies have been in culture for decades, suggesting that fresh isolates could increase the environmental relevance of laboratory studies.

A striking and unexpected result was that growth rates of several species in unialgal culture were significantly lower than in the natural and artificial community experiments. These large differences were not because of abiotic factors, as light, nutrients, temperature, and seawater chemistry were kept constant throughout all of the experiments. Thus, this effect must have been because of biotic factors directly associated with the presence or absence of the other species. In fact, these and other dinoflagellate species often occur in mixed-species blooms worldwide. (Allen 1941; Marasovic et al., 1995; Tahri-Joutei et al., 2000; Amorim et al., 2001). Our experimental results suggest that these multiple species associations are probably not simply because of coincidental similarities of environmental growth preferences, but could be a result of specific interactions between species.

Several possibilities may have contributed to differences in the growth rates of the dinoflagellates in unialgal versus mixed cultures. Many dinoflagellates exhibit facultative mixotrophic (phagotrophic) behavior (Burkholder et al. 2008; Caron 2000), but how the relative degree of autotrophy versus heterotrophy in these organisms may be affected by future rising pCO2 is unknown (Caron and Hutchins in press). All four of our species are potentially mixotrophic, and in fact L. polyedrum, Alexandrium sp., and Gonyaulax sp. have all been demonstrated to ingest Prorocentrum (Jeong et al. 2005). Growth rates of Alexandrium spp. (Jacobson and Anderson 1996) and other dinoflagellates (Jeong et al. 1999; Adolf et al. 2006) increase significantly in the presence of suitable prey. Supporting the case for potential reliance on mixotrophy by Alexandrium our experiments, growth rates were very slow in unialgal culture but increased dramatically in coculture, and throughout the 12-month experiment, conditioning and competition pCO2 levels were not a significant influence on its dominance or growth rates. P. micans also responded with increased growth rates in the 12-month artificial community experiment, even though both conditioning and competition pCO2 were also significant influences on its dominance and growth. In contrast, L. polyedrum and Gonyaulax sp. exhibited much less stimulation or even a reduction of growth rates in some mixed culture treatments, perhaps indicating mortality losses to grazing. If mixotrophy was a significant influence on community structure, as seems likely, it appears that Alexandrium sp. and P. micans may have taken on the role of grazers at the expense of the other two species.

Other, less understood interspecies interactions such as allelopathy (Granéli and Hansen 2006) may have also influenced growth in our mixed cultures. For instance, Alexandrium fundyense can inhibit the growth of competing phytoplankton species through the production of allelochemicals (Hattenrath-Lehmann and Gobler 2011). Other variables, such as growth factors in the form of vitamins (Tang et al. 2010) and co-occurring bacteria may also have played a role in the algal population dynamics in our experiments. Whether the competition between the dinoflagellates was influenced by these types of interactions is unknown, but the common occurrence of mixed species blooms of dinoflagellates and other phytoplankton groups such as raphidophytes and diatoms (Zhang et al. 2006; Schnetzer et al. 2007) suggest that interspecific facilitation or inhibition interactions could be a widespread phenomenon. Despite these necessary qualifications about factors other than pCO2 that may have influenced our experimental results, outcomes of the four community trials consistently revealed a significant restructuring of our dinoflagellate communities based on pCO2. This suggests the possibility that ocean acidification may play an influential role in bloom dynamics of these organisms in the future ocean, but in combination with biotic factors including mixotrophy, alleopathy, and facilitation.

We used final relative abundance as an indicator of competitive success, whereas growth rates were assumed to be indicative of any fitness changes under our long-term pCO2 conditioning treatments. However, higher growth rates did not always correspond to competitive dominance at the end of the experiment. This is partly because initial abundances of each species were different by design (reproducing their initial abundances in the original community), and partly because we focused on growth rates during the second week of each competition experiment. We avoided drawing conclusions based on initial (first week) growth rates because we assume that the cultures may not have been fully acclimated directly after being placed in the mixed culture communities. Because the final abundance of each species integrated the relative growth rates in both weeks, as well as initial abundances, the second week growth rates presented may not have been uniformly predictive of final relative abundance trends.

The responses of the four species in the switch experiments conducted at the 12-month timepoint did not provide unambiguous evidence that either adaptation or acclimation during the conditioning period at a particular pCO2 subsequently provided a strong competitive advantage in a mixed community growing at that pCO2. Neither Alexandrium sp. nor Gonyaulax sp. showed statistical evidence that the conditioning pCO2 level affected their competitive success in any of the switch pCO2 treatments. L. polyedrum showed a highly significant effect of conditioning pCO2 on its competitive dominance, but only among the low pCO2-conditioned clones. Furthermore, there was no evidence that these L. polyedrum clones had adapted in a manner that favored their competitive success at the particular conditioned pCO2 (low→low treatments), as the net effect of long-term conditioning at low pCO2 was to increase the competitive success of L. polyedrum at all pCO2 levels tested (Fig. 6A–C).

For P. micans there was strong statistical support for an effect of all three conditioning CO2 concentrations on its relative dominance in the switch experiments. Because all conditioning treatments increased its competitive abilities, though, we cannot rigorously attribute its success to selection by pCO2 alone. P. micans could have also been favored by selection by one or more of other shared environmental conditions common to all the long-term cultures. However, low pCO2-conditioned clones of P. micans did compete significantly better in the low→low treatments than in the low→medium and low→high treatments (Fig. 6A–C), suggesting that in this single case, conditioning at low pCO2 may have conferred a competitive advantage in communities growing at this CO2 level.

Dinoflagellates are a ubiquitous protistan functional group in marine ecosystems, and their multispecies blooms are excellent systems in which to experimentally test community competitive interaction dynamics in response to changing environmental factors such as pCO2. However, these organisms are not typically amenable to cryopreservation, have extraordinarily large genomes (Hackett et al. 2004), and in general have slow growth rates compared to many microorganisms and thus present some inherent difficulties as experimental evolution model organisms. Our experiments were not intended to rigorously distinguish physiological acclimation from genetic adaptation, and it is questionable to what degree selection could have led to significant adaptive responses within the limited number of generations encompassed by the relatively slow-growing clonal cultures. In contrast to studies with much faster growing organisms (Collins and Bell 2004; Lohbeck et al. 2012), our dinoflagellates only completed approximately 35 to 120 generations over the 1-year period. In addition to low intrinsic growth rates, and the reduced growth in unialgal cultures relative to mixed communities discussed above, slow growth in our cultures may have also been partially because of our attempts to keep our experiments environmentally relevant by using relatively modest nutrient concentrations.

Contemporary populations of algae are adapted to CO2 concentrations roughly equivalent to our “medium” treatment. Nevertheless, they still may be exposed to large diurnal and seasonal ranges in dissolved inorganic carbon and pH levels. in habitats such as productive estuaries (Hinga 1992; Hansen et al. 2007) and coastal upwelling regimes (Feely et al. 2008). The seawater carbonate buffer system can vary dramatically during blooms, especially in semi-enclosed embayments. The low pCO2 treatment in our study can be viewed as simulating a preindustrial value, but also serves as a good model for contemporary CO2-depleted bloom water. Indeed, experiments suggest there are dinoflagellate taxa that thrive at low pCO2 (Hinga 1992). Particularly notable in our study was the observation that the relative abundance of L. polyedrum always increased as pCO2 decreased, suggesting that this organism may be better adapted to growth at lower CO2/higher pH than many co-occurring species. Because CO2 depletion commonly occurs in dense blooms, it may be no coincidence that L. polyedrum is usually the dominant dinoflagellate in recent recurring in the Southern California coastal region (Gregorio and Pieper 2000).

Shifts away from L. polyedrum at high CO2 and toward greater dominance by species such as Alexandrium and Gonyaulax spp. that can produce virulent phycotoxins (Rhodes et al. 2006; Schantz et al. 1966) may have implications for the future environmental impacts of Southern California coastal blooms. In addition to increased abundance, some harmful bloom species such as the dinoflagellate Karlodinium veneficum and the diatom Pseudo-nitzschia spp. produce much more toxin when physiologically acclimated to high pCO2 (Fu et al. 2010; Sun et al. 2011; Tatters et al. 2012). How toxin production rates will respond to selection by prolonged growth in acidified seawater is, however, presently unknown.

Our experiments examined only pCO2, and so cannot predict the full spectrum of community responses to interactive effects from multiple global change variables (Boyd et al. 2010). Experiments incorporating not only higher pCO2 but also projected concurrent warming, nutrient shifts, and irradiance changes need to be conducted with different assemblages (Rost et al. 2008; Boyd et al. 2010). Our experiments suggest a need for such multivariate experiments to encompass longer time scales. Despite the escalating complexity such experiments entail, elucidation of how multiple interactive variables will modulate algal acclimatization and adaptation is crucial for a better understanding of their ecology in a future greenhouse ocean. Investigations into how the phenotypic plasticity and genetic diversity of algal populations are involved in long-term responses to a complex changing environment are a critical component of this effort. Our research adds to the very sparse data on long-term phytoplankton community structure responses under global change regimes. To our knowledge our study is the first to take an experimental approach; other efforts to examine long-term phytoplankton community responses have considered mainly warming, and present observational (Cloern et al. 2005; Paerl and Huisman 2009; Hinder et al. 2012) or modeling (Moore et al. 2008) datasets. Future experimental efforts distinguishing physiological acclimation from genetic adaptation are needed to provide a foundation for a more in-depth understanding of these processes in the future.


This study was supported by funding from NSF OCE-0962309 and USC Sea Grant.