Temperature drives diversification in a model adaptive radiation

Abstract

The warmer regions harbour more species, attributable to accelerated speciation and increased ecological opportunities for coexistence. While correlations between temperature and energy availability and habitat area have been suggested as major drivers of these biodiversity patterns, temperature can theoretically also have direct effects on the evolution of diversity. Here, we experimentally studied the evolution of diversity in a model adaptive radiation of the bacterium Pseudomonas fluorescens across a temperature gradient. Diversification increased at higher temperatures, driven by both faster generation of genetic variation and stronger diversifying selection. Specifically, low temperatures could limit the generation of diversity, suggested by the observation that supply of genetic variation through immigration increased diversity at low, but not high temperatures. The two major determinants of mutation supply, population size and mutation rate, both showed a positive temperature dependence. Stronger diversifying selection in warmer environments was suggested by promoted coexistence, and further explicitly inferred by the ability of evolved phenotypes to invade the ancestral type from rare. We discuss possible physiological and environmental mechanisms underlying the findings, most of which are likely to be general.

1. Introduction

Temperature is believed to be a major determinant of global patterns of biodiversity, most notably the latitudinal diversity gradient [1–9]. While these patterns may be driven to some extent by environmental variables that co-vary with temperature (including energy availability and habitat area), temperature itself can have causal effects on the rate of evolutionary diversification. First, higher temperatures may increase the rate at which genetic variation is generated. This will arise as a consequence of elevated mutation rates [10–13], shorter generation times [11,12] and larger population sizes; the latter because of the effects of temperature on the rate of energy use by organisms and thus primary productivity [11,14,15]. Second, diversifying selection could be enhanced by increased temperature. The physiological constraints imposed by low temperatures that reduce the potential advantage of many mutations will be relieved in warmer environments [11,16]; and the greater population sizes in high-temperature environments also increase the benefits of within-habitat niche differentiation [1,5,11,17–19]. Third, a decrease in the importance of drift in larger populations may permit a greater number of beneficial mutations to escape stochastic extinction and thus accelerate evolutionary processes [20]. Despite a large body of correlational studies exploring the different mechanisms, experimental studies are lacking.

Here, we experimentally address the effects of temperature on diversification in the model adaptive radiation of the bacterium Pseudomonas fluorescens. In spatially structured environments (static tubes containing growth medium), initially isogenic P. fluorescens populations can rapidly diversify into numerous phenotypes that show heritable differences in colony morphology and the spatial niches they occupy. The diversified populations are usually dominated by two categories of variant types: smooth morph (SM) types that resemble the ancestor and occupy the broth, and wrinkly spreaders (WS) that form a biofilm at the air–broth interface [21–23]. Diversity in this system is maintained through negative frequency-dependent selection [21,24,25] as well as stochastic persistence of phenotypes with nearly equal fitness [26,27].

2. Methods

(a). Diversification across a temperature gradient

The bacterium P. fluorescens SBW25 [21] has an upper temperature limit for positive growth of approximately 34°C and a lower limit of approximately 5°C under our experimental conditions. We studied the diversification of this bacterium across a temperature gradient from 9°C to 30°C. A single bacterial isolate was used to initialize 96 evolution lines. The cultures were grown in static microcosms (30 ml of tubes) each of which contained 5 ml of M9KB medium; and every microcosm was initially inoculated with 10⁶ isogenic bacterial cells. Twelve microcosms were grown at each of eight temperatures from 9°C to 30°C (with a 3°C interval); six replicate microcosms evolved in isolation (no immigration) and the other six with immigration. Specifically, 5 µl of culture from each ‘no immigration’ microcosm was transferred into 5 ml of fresh medium every 2 days. For the ‘with immigration’ microcosms, 4.5 µl of each culture was transferred into fresh medium, together with 0.5 µl of pooled culture from all the 96 microcosms (this immigration regime mimics a situation where there is no dispersal limitation). After six transfers of evolution (approx. 60 generations), dilutions of cultures were spread on M9KB agar plates and grown for 2 days at 28°C, and then scored for the densities of all morphologically distinguishable phenotypes. Across all the microcosms eight phenotypes were identified: large SM, small SM, tiny SM, large WS, small WS, round (middle-sized) WS, SM-like WS and wheel-like WS. Estimation of phenotypic richness and diversity of each microcosm was based on 100 randomly chosen colonies [22,23], and diversity was expressed as the complement of Simpson's index, where p_i is the frequency of the ith phenotype [28].

(b). Temperature dependence of mutation rate

The mutation rate of the ancestral strain was estimated using fluctuation tests [29]. At each of the eight temperatures, six replicate microcosms were inoculated with 10⁵ cells of ancestral bacteria and grown for 2 days. Final bacterial densities were determined by plating diluted cultures on non-selective agar plates (M9KB), and the number of rifampicin-resistant mutants was estimated by plating 200 µl of each culture on selective solid medium (M9KB supplemented with 50 mg l^–1 rifampicin). We examined the starting bacterial populations to ensure that there was no pre-existing resistant mutant. The MSS maximum-likelihood method was used to calculate mutation rates [30,31].

(c). Temperature dependence of selection strength

A pooled culture was assembled from the six ‘with immigration’ evolution lines from 21°C at the final transfer, from which six phenotypes were identified: large SM, small SM, large WS, small WS, round WS and SM-like WS. A single colony was isolated for each of the six phenotypes. Invasion assays were carried out to measure fitness when rare of those phenotypes relative to the ancestor at each of the eight temperatures. A modified variant P. fluorescens SBW25EeZY6KX [32] was used as the ‘ancestor’ in the invasion assays, which showed no detectable difference in fitness from the wild-type SBW25 strain. Each of the evolved phenotypes and the ancestral strain was grown overnight in shaken microcosms at each of the eight temperatures. These cultures were used to inoculate microcosms of invasion assays, with 5.0 µl of ancestor and 0.05 µl of evolved phenotype for each microcosm. The microcosms were grown statically for 2 days, with initial and final bacterial densities measured by plating dilutions on M9KB agar plates supplemented with X-gal, where the ‘ancestral’ strain (SBW25EeZY6KX) showed a blue colony colour, and colonies of the evolved phenotypes (SBW25) were yellow. The fitness of each evolved phenotype relative to the ancestor was calculated as the difference in the estimated Malthusian parameters, that is, a selection coefficient, S = (m_evolved – m_ancestor)/m_ancestor, where m = ln(N_f/N₀), with N₀ and N_f the relevant initial and final densities, respectively [33]. Each competition assay was repeated three times.

(d). Statistical analysis

Our study was a split-plot experiment (multiple replicate microcosms in each incubator, and one incubator for each temperature), therefore we used linear mixed-effect models to analyse the temperature responses of phenotypic richness, diversity, proportion of SM types, population size and phenotype fitness, where temperature was considered as a continuous variable and incubator ID a random factor. The general linear model was used for analysing the temperature–mutation rate relationship. Two-sample t-tests were used to examine the differences between ‘no immigration’ and ‘with immigration’ microcosms in diversity and population size at each temperature, and one-sample t-tests for the difference of fitness values from zero for every phenotype at each temperature (results of t-tests were briefly provided in the main text, and detailed in the electronic supplementary material). Richness and population size (density) data were log-transformed and proportional data arcsine-transformed before analysis.

3. Results and discussion

(a). Diversity generation was constrained at lower temperatures

Pseudomonas fluorescens populations that evolved in isolated microcosms showed an increase in phenotypic richness (figure 1a; F_1,6 = 42.86, p < 0.001) and diversity (the 1 − λ index; figure 1b; F_1,6 = 23.62, p = 0.003) with increasing temperature, and this corresponded with a decrease in the frequency of the SM types that occupied the ancestral niche (figure 1c; F_1,6 = 23.54, p = 0.003). This pattern may have arisen because of limited potential to generate diversity and/or weaker diversifying selection at lower temperatures. We first inferred the existence of a limitation of diversity generation by examining how the supply of genetic variation with gene flow affected diversification. To this end, we compared diversity in microcosms that evolved in isolation, and those receiving repeated immigration from a global pool of cultures: increased diversity resulting from immigration would indicate a limitation of within-population diversity generation. Immigration had no effect on diversity at the two highest temperatures, 27°C and 30°C (figure 1a–c; two-sample t-tests for phenotypic richness, 1 − λ diversity index, and SM proportion, p > 0.200), but resulted in greater diversity at each temperature within the range 9–24°C (difference between treatments statistically significant at 12–24°C, and almost significant at 9°C; see details in electronic supplementary material, table S1). This suggests the existence of a limitation of diversification at the intermediate and low temperatures.

Figure 1.

Phenotypic richness (a), the Simpson diversity 1 − λ (b), the proportion of SM types (c), and total bacterial population density (d) in P. fluorescens populations that evolved across a temperature gradient. Open symbols indicate microcosms evolving in isolation and filled symbols, microcosms with immigration. Data show mean ± s.e. (n = 6).

We next investigated factors that likely contributed to a lower potential for diversity generation at lower temperatures, notably population size and mutation rate [20]. There was a positive temperature–biomass relationship in both microcosms evolving in isolation (figure 1d; F_1,6 = 26.21, p = 0.002) and those with immigration (F_1,6 = 25.00, p = 0.003); and there was no significant difference in population size between microcosms with and without immigration at any temperature (p > 0.100; electronic supplementary material, table S1). Mutation rate also increased at higher temperatures (figure 2; F_1,6 = 8.89, p = 0.025). Note that the temperature influences on the ‘effective evolutionary time’ (number of generations per unit of absolute time), commonly observed in natural populations [11,12], did not apply in our experimental system. All our experimental microcosms evolved for approximately the same number of generations that was set by the serial dilution protocol (note that we cannot rule out the possibility that cell death and birth may take place during the stationary growth phase, and if that were the case, the populations in the high-temperature environments that spent more time in the stationary phase could have experienced a slightly larger number of generations; but cell death during the stationary phase is likely negligible compared with mortality resulting from the dilution). Moreover, our bacterial populations were of fairly large sizes (bottleneck population sizes greater than 2 × 10⁵ cells), therefore drift probably played a minor, if any, role in the evolutionary dynamics.

Figure 2.

The mutation rate of the ancestral strain (per cell per generation, for rifampicin resistance) as a function of temperature.

The positive temperature response of mutation rate may result from temperature effects on growth rate and metabolic activity that determine the chance of replication errors and oxidative DNA damage, respectively [12,13,34]. Alternatively, higher temperatures can lead to faster exponential growth and thus a longer period of stationary phase, where nutrient limitation can increase the mutation rate, for example, by reducing the availability of energy for the DNA repair function [35–38].

(b). The strength of diversifying selection increased with increasing temperature

In addition to the emergence of genetic diversity affecting the rate of adaptive diversification, processes that affect coexistence may also play an important role. On the one hand, diversifying selection may become stronger at higher temperatures where an alleviation of physiological constraints allows a larger number of beneficial mutations [11,16]. On the other hand, the temperature may increase niche availability through its effects on the intensity and complexity of biotic interactions resulting from increased productivity [1,5,11,17–19].

Microcosms with immigration (that alleviated mutation supply limitation) still showed a positive temperature–diversity relationship, suggesting greater diversifying selection at higher temperatures (figure 1a, F_1,6 = 46.23, p < 0.001; figure 1b, F_1,6 = 46.23, p < 0.001). Crucially, the increase in diversity with temperature coincided with a decrease of the frequency of the ancestor-like SM types (figure 1c; F_1,6 = 105.66, p < 0.001), consistent with a role of temperature-mediated changes in diversifying selection. We explicitly measured the strength of diversifying selection using six evolved phenotypes isolated from microcosms of a particular environment (21°C). A key signature of diversifying selection is that new phenotypes can invade the ancestral type from rare, while not necessarily competitively excluding the ancestor. We, therefore, measured the fitness when rare of the six evolved phenotypes relative to the ancestor across a temperature gradient, and found that the fitness of rare genotypes did increase with temperature (figure 3; phenotype, F_5,126 = 59.12, p < 0.001; temperature, F_1,6 = 195.37, p < 0.001; phenotype × temperature, F_5,126 = 5.36, p < 0.001). All the six evolved phenotypes could invade the ancestor from rare when cultured at relatively high temperatures, but had fitness equal to or lower than the ancestor at low temperatures (electronic supplementary material, table S2).

Figure 3.

The fitness of six evolved phenotypes as a function of temperature. Fitness shown here is from competition assays for the ability of each evolved phenotype to invade the ancestral type from rare. Data show mean ± s.e. (n = 3). Data points are plotted with jitter to avoid overlapping.

There are two possible explanations for an increase in the strength of diversifying selection with increasing temperature. First, in cold environments, all chemical reactions crucial for protoplasm functions are limited by temperature, hence there are likely fewer mutations, compared with warmer temperatures, that confer a fitness advantage [11,16]. Second, high temperatures may cause environmental changes that create opportunities for niche differentiation [1,5,11,17–19]. The physiological, as opposed to the environmental, mechanism is likely to have played an important role in the temperature dependence in the fitness of the two SM types in our experiment, simply because they occupied a similar ecological niche to the ancestor. Both SM genotypes had equivalent competitive ability with the ancestor at temperatures ≤ 15°C (one-sample t-test, p > 0.05; electronic supplementary material, table S2), but became more competitive in the warmer environments (p < 0.05). By contrast, the environmental mechanism likely had a relatively more important role in the temperature response of the four WS types that occupied the novel ecological niches of the air–liquid interface. Specifically, the greater biomass at higher temperatures likely led to more intense competition for the oxygen resource, allowing the costs of WS over-expression of polymer for biofilm formation to be partially offset by their ability to colonize the air–liquid interface where oxygen is abundant [22–24,39–41]. Consistent with this view, the fitness of WS was positive at higher temperatures (greater than or equal to 21°C, p < 0.05), but could become negative at extremely low temperatures (e.g. large WS and small WS at 9°C and 12°C; p < 0.05; electronic supplementary material, table S2). It is of course possible that the physiological and environmental mechanisms operated simultaneously, but it is challenging to unambiguously distinguish between them.

Our results suggest that the effect of temperature on diversification was governed by both the rate of diversity generation and the strength of diversifying selection. We used a variation partitioning approach to determine the relative importance of these mechanisms, based on which we tentatively suggest that mutation supply rate played a larger role (electronic supplementary material, figure S1). Furthermore, it is important to emphasize that the increased population sizes associated with higher temperatures may also have promoted neutral coexistence [42,43]. In an earlier study with our experimental system (the same culture conditions but at 28°C), approximately two-thirds of the diversified phenotypes showed dynamics that were not different from neutral when their abundances were artificially manipulated [27]. The increase in diversity with temperature we observed therefore almost certainly involved increase neutral, as well as selected, diversity. Finally, it is worth noting that high temperatures may actually reduce population sizes, and therefore diversification, under certain energy-limited conditions because of accelerated metabolic activity and reduced efficiency of energy use [14,15]; such a scenario deserves experimental investigation in future.

4. Conclusion

Here, we have shown experimentally that increasing temperature could promote adaptive diversification in a model microbial system. This was a consequence of multiple processes that resulted in both a net increase in the generation of diversity as well as stronger selection acting on the maintenance of diversity. Our results suggest that temperature should not be considered only as a proxy of energy in studies of biodiversity patterns. We expect the temperature effects on mutation supply rate and selection strength to be general, although the relative importance of these processes may vary across adaptive radiations in different taxa or different environments. These results may not only add new findings to the existing understanding of the causes of biodiversity patterns [22–24,44–48], but also have implications for understanding contemporary evolution [49–51].

Data accessibility

From the Dryad Digital Repository at: http://dx.doi.org/10.5061/dryad.t31q738 [52].

Authors' contributions

Q.G.Z. designed study and analysed data; Q.G.Z. and H.S.L. performed experiments; Q.G.Z. and A.B. wrote the paper.

Competing interests

We have no competing interests.

Funding

The study was funded by the National Natural Science Foundation of China (31725006 and 31670376), the 111 project (B13008) and the Fundamental Research Funds for the Central Universities of China (2017EYT20 and 2017STUD19); and A.B. was supported by the Royal Society.