The primers used to sequence the 18S rRNA gene captured a region varying in length from 506–530 bp. These sequences resulted in a poorly resolved tree, but still distinguished two distinct species in the October time block. In the April and June time blocks, all of the stock lines belonged to the same monophyletic group and were considered the same species (Fig. 1). Lines within this clade shared identical sequences, except for an insert at the beginning and end of the sequence in some lines. None of the reference sequences fall within this same group, however, indicating that this portion of this gene in this species has not previously been sequenced. In October, at least one of the stock lines belonged to this same monophyletic group, but two other stock lines do not. One of these lines (Oct1) is likely Colpoda steinii (Fig. 1) and differs from the previously unsequenced species by 32 bp. The other October line (Oct2) falls into its own clade (Fig. 1) and differs from the previously unsequenced species by 16 bp.
Figure 1. Bootstrap consensus tree using parsimony generated using nucleotide sequences from a region of the 18S rRNA gene. Samples in bold were sequenced in this study. Oct, Apr and Jun samples represent a portion of the lines from different time blocks; lines from Aug were collected in similar fashion, but not used in any experiment. Reference sequences were obtained from Genbank (Appendix B).
Download figure to PowerPoint
Prior to the April and June experiments, I recognized two distinct morphotypes among samples and used only a single morphotype thereafter. All stock lines in the previously unsequenced species have similar morphologies and are conspicuously larger than Colpoda steinii, with more obvious colouration and frequently visible vacuoles within the cell. Further, the gliding movement of the unknown species is smoother than the more jerky movement of Colpoda steinii. The Oct2 line is composed of both morphotypes, which may explain why it does not group distinctly with either group. Although several stock lines were unable to be recovered for sequencing, the unrecovered lines were morphologically similar to those of the unsequenced species. To account for potential species-level selection in the October time block, I analysed the trait values with and without this block. Results were qualitatively similar, particularly with respect to the selection treatment, and for simplicity, I present the anova results with only a single morphotype (April and June time blocks). The analysis including the October time block is available in Appendix C.
Evolutionary effects of competition
Selection environment had a significant effect on the evolution of Colpoda traits (manova: F6,6 = 52.1, P < 0.001) and affected the evolution of three of the six traits measured: population growth rate, cell size and cyst production. Measurement environment also had a significant effect on Colpoda traits measured (manovaF6,6 = 10.6, P = 0.006), largely because of cyst production and refuge use.
Population growth rate of Colpoda was significantly higher in populations that had evolved with competition than those that evolved in monoculture (F1,27 = 6.72, P = 0.015, Fig. 3a). Neither measurement environment (F1,1 = 0.36, P = 0.65, Fig. 3a) nor time block (F1,1 = 11.1, P = 0.19) had a significant effect on Colpoda population growth rate. The best-fit model included a nonsignificant interaction between time block and measurement environment (F1,27 = 2.49, P = 0.13). The initial population growth rate of Colpoda was lower than both selection environment treatments (r = 0.118 ± 0.009). Population growth rates of P. alpestris were more difficult to measure because populations had not peaked within 48 h. Fitting an exponential curve to the available data may overestimate population growth rate, but gave a population growth rate of 0.086 (± 0.006 SE) – considerably lower than the population growth rate of Colpoda (Fig. 3a).
Figure 3. Averages (±SE) of six traits of replicate Colpoda populations from two different selection environments (monoculture: white bars; interspecific competition: grey bars) measured in two different common garden conditions. Hatched bars represent the ancestral state of population traits (when available) before the selection experiment. In (a), the units are per capita growth rates per hour. In (d), positive values indicate greater density in the refuge environment; negative values indicate greater density in the water column.
Download figure to PowerPoint
Selection environment had a marginally significant effect on cell size (F1,1 = 52.0, P = 0.088). When statistical power was increased by including the October time block, this effect was significant (P = 0.009, Appendix C). Individuals that evolved with competition were smaller than those that evolved in monoculture (Fig. 3b). Although measurement environment (F1,1 = 36.2, P = 0.105) and time block (F1,1 = 7.05, P = 0.229) had no significant effect on cell size, they did have a large effect when the October data were included in the analysis (Appendix C). The best-fit model included all two-way interactions. The interaction between measurement and selection environment was one of magnitude and not direction (F1,25 = 69.2, P < 0.001). Cells selected in competition were always smaller than those in monoculture, and cells measured in competition were always larger than those measured in monoculture (Fig. 3b). Similarly, the interactions between selection environment and time block (F1,25 = 15.4, P < 0.001) and measurement environment and time block (F1,25 = 3.81, P = 0.062) did not affect the interpretation of the main effects. The initial cell size of Colpoda grown in monoculture was larger than either selection environment treatment (1.75 ± 0.1 × 10−3 mm). The average cell size of P. alpestris (mean cell area = 9.5 × 10−4 mm ± 4.6 × 10−5) was smaller than Colpoda (Fig. 3b).
Selection environment has a significant effect on cyst production (F1,28 = 5.67, P = 0.024), as populations of Colpoda that evolved with competition produced significantly fewer cysts than those that evolved in monoculture (Fig. 3c). Measurement environment had a similar effect on cyst production in that populations measured in monoculture produced more cysts than those measured in competition (F1,28 = 6.14, P = 0.020, Fig. 3c). Time block also had a significant effect on cyst production, with more cysts produced in June (F1,28 = 7.89, P = 0.009). The best-fit model did not include any interactions.
Selection environment had no significant effect on Colpoda refuge use (F1,28 = 0.65, P = 0.426, Fig. 3d). Measurement environment had a marginally significant effect on refuge use (F1,28 = 3.73, P = 0.064). Colpoda measured in monoculture used the refuge more heavily than in competition (Fig. 3d). Time block had a significant effect on refuge use (F1,28 = 13.3, P = 0.001), as Colpoda used the refuge more heavily in June than in April. The best-fit model included only these main effects. Conversely, P. alpestris used the refuge very little and were found in higher densities in the water column relative to the refuge (refuge use = −0.36 ± 0.12).
There was no effect of selection environment (F1,26 = 0.002, P = 0.970, Fig. 3e), measurement environment (F1,1 = 0.002, P = 0.972, Fig. 3e) or time block (F1,1 = 0.005, P = 0.955) on the peak population density. The best-fit model included two interactions (measurement × selection environment and block × measurement environment), but neither were significant (P > 0.60). Selection environment had no effect on the speed of Colpoda (Fig. 3f, F1,27 = 0.024, P = 0.878). Neither measurement environment (F1,1 = 4.55, P = 0.279, Fig. 3f) nor time block (F1,1 = 19.9, P = 0.140) had a significant effect on cell speed. The best-fit model included a nonsignificant block × measurement environment interaction (F1,27 = 0.014, P = 0.906). The initial speed of Colpoda prior to selection was slightly slower than either selection environment treatment (0.508 ± 0.060 mm s−1). The speed of P. alpestris (0.20 ± 0.007 mm s−1) was much slower than the average speed of Colpoda (Fig. 3f).
Most traits were significantly correlated with other traits, although the nature and strength of the correlation was dependent on the specific traits under consideration (Table 1). When data were subdivided into different measurement and selection environments, correlations were qualitatively similar to one another. As such, the values in Table 1 reflect pooled data with increased power. The one exception to this case was the correlation between cell area and refuge use. The overall correlation with pooled data was significantly negative, but this was largely because of a strong negative correlation of individuals from the monoculture selection environment grown in the monoculture measurement environment (r = −0.45, P < 0.05). In all other selection and measurement environments, the correlation between cell size and refuge use was nonsignificant (range of r = −0.04–0.16, P > 0.05).
Table 1. Average correlation coefficients among six traits after bootstrapping data pairs. Asterisks indicate values for which the 95% confidence interval does not include zero.
| ||Pop. growth rate||Peak abundance||Cell area||Swimming speed||Cyst production||Refuge use|
|Pop. growth rate||1.0||0.36*||−0.56*||−0.32*||0.25*||0.29*|
|Peak abundance|| ||1.0||−0.28*||−0.37*||0.29*||0.15|
|Cell area|| || ||1.0||0.36*||−0.17||−0.21*|
|Swimming speed|| || || ||1.0||−0.34*||−0.22*|
|Cyst production|| || || || ||1.0||0.18|
|refuge use|| || || || || ||1.0|