Genetic variation in dispersal and life history strategies
Our study provided strong evidence for genetic variation in dispersal and life history strategies in Tetrahymena thermophila protozoans. Strains varied considerably and significantly in nearly all aspects of life history (growth performance, survival, cell morphology changes, and degree of phenotypic plasticity of cell shape) under different environmental conditions, and showed significant differences in the associations of these traits under our experimental conditions (axenic food resource or food free medium). Life history trait differences were in many cases significantly related with dispersal strategies, which also varied significantly between strains. Among-strain differences in life-history traits and trait associations were not just due to variation in strain quality, although some T. thermophila strains did perform better than others as regarded a series of core fitness traits (survival, dispersal, growth performance, and colonization ability). This is because strains also differed with respect to various trait associations that cannot as such be considered to represent differential quality but rather different strategies. Examples are the observed among-strain differences in growth rates r versus final population density in presence of nutrients (PC2G), cell shape versus size in presence of nutrients (PC1G), and the opposition between a high degree of phenotypic plasticity of cell shape (high variance in cell elongation) with some cells turning into dispersal morphs versus a more durable and greater, but less plastic, mean cell elongation under starvation (PC2S). The T. thermophila ten strains hence differed also with respect to the patterns of their investment along different trait axes. Hence, our study adds to the limited evidence (e.g. [24,38,39,41-47,58]) for genetic variance in dispersal syndromes within a species. Also, the among-strain variance in the degree of plasticity of cell shape found in T. thermophila is similar to recent findings of genetic variance in plasticity for core life-history traits in other animals, e.g. Caenorhabditis elegans nematodes  and in dispersal related morphological traits in pea aphids (a winged-non-winged polyphenism) .
The life-history trait association differences found among T. thermophila strains should be attributed to among-strain differences in genes of the macronucleus (somatic nucleus characteristic for ciliated protozoans; reviews by [75,76]), not directly the micronucleus (germline nucleus, not transcribed ). This is because macronucleus genes represent a modified, often variably amplified and rearranged subset of micronucleus genes [75,76,78]. The genetic trait variance expressed among clonal lines is therefore not necessarily the same that would be found in crossing experiments (sexual reproduction; involving micronuclei). Another curiosity of ciliates, the amitotic divisions of macronuclei making daughter cells from a clonal division potentially have slightly different alleles or amounts of alleles (, reviews by [75,76]) did not invalidate our study. Such differences among cells within a clone line would only make our tests for genetic differences among strains more conservative. Equally, non-genetic 'maternal' effects are not likely to have contributed importantly to among-strain differences and so confounded our study, because all strains were maintained under the same conditions for some 700 generations before the start of experiments (> one year). Finally, the observed strain differences in morphological responses to starvation were very unlikely to be artefactual. We used standard techniques (centrifugation, decanting of super-natant and addition of water) to eliminate nutrients from the medium [e.g. [62,69]], and any damage during centrifugation would rather lead to ciliae loss and so to a decreased, not a greatly enhanced swimming speed. Cells, moreover, started swimming within seconds after centrifugations indicating perfectly preserved cell integrity, verified also under the microscope (E.J.F and N.S. personal observations). Also, all strains were subjected to precisely the same manipulations (same centrifugal forces and length of centrifugation).
Covariation of dispersal strategies with life history traits
Dispersal strategies, growth performance, and survival
The association of T. thermophila dispersal rates in presence of nutrients with other life history traits fits partly with the classic 'colonizer syndrome'  envisioned typical for species living in variable patches in metapopulations (see also [21,23,31]). Strains with great short-distance dispersal rates in presence of nutrients had a better growth performance, and high colonization abilities as predicted ([30,31,79] but see ), but contrary to classical predictions they also had a better mean strain survival rate than did less dispersive strains. An actual negative co-variance (trade-off) between two life-history traits could be masked by differences among strains in overall condition, because some strains would have more resources to invest both in growth and in survival [28,29]. Our experiments were not set up to explicitly explore this issue. All our strains were, however, kept in the same conditions of nutrient availability in the growth experiments, and so should have had the same amount of resources available. Also we note that our findings could be explained by some alternative dispersal evolution models that do predict positive associations of survival with dispersal rates and reproductive effort for ruderal species, the precise expectations depending on landscape characteristics (productivity, demographics, spatio-temporal structure , review by ), that are only little known for T. thermophila.
The opposite end of the spectrum, namely T. thermophila strains dispersing little, growing less, and colonizing less well as single cells in presence of food, and surviving food-stress poorly, represented relative philopatry as long as the environment remained good, and likely long-distance dispersal (via fast-swimming dispersal morphs) when the environment turned bad. Such a long-distance colonizer strategy could allow coexisting with better competitor strains (see above). Alternatively, 'philopatric' strains may come from habitats with less local spatio-temporal variability (see e.g. [23,32]) but more catastrophic patch degradations (favouring the production of the dispersal morph). Finally, the variation among strains in dispersal strategies and life history could represent non-adaptive variance around an adaptive mean. In all cases, relative philopatry in presence of food did not signify an adaptation to a more intensive exploitation of already colonized patches (see e.g. [81,83,84,86]), because the growth performance of philopatric strains was inferior and also philopatry was not related significantly with r versus K strategies. Nevertheless, the lower dispersal rate of philopatric strains suggests that they freely tolerate higher densities without dispersing massively. Density-dependent dispersal strategies are indeed predicted by several theoretical models and have been reported in a number of species (e.g. [4,85-88]).
Potential density-dependence of dispersal strategies is particularly interesting given the primitive form of cooperation exhibited by T. thermophila. Cells secrete certain chemical compounds ('growth substances') that increase the survival of other cells [72,73] and also actively form aggregations , which should enhance the effective concentration of these substances. Those T. thermophila strains that disperse little even when at high density and are less good at colonizing as single cells, could be strains adapted to and dependent on greater cooperation than the more dispersive strains. Consistent with this, strains producing many dispersal morphs (generally philopatric strains) have smaller cell sizes (at carrying capacity; Figure 4E), and so each cell likely has fewer resources than is the case in more dispersive strains. The inferior rates of survival under starvation for philopatric strains corroborate this idea. If our hypothesis is correct, the philopatric strains should show a stronger degree of aggregation and only disperse much when densities are very high and benefits of 'cooperation' outweighed by costs of conspecific competition. We are currently testing this.
Morphological differences between dispersers and non-dispersers were found in our study, as in a series of other species ([41,58,79], see also [10,11]). Morph diversity was, however, greater than suggested by previous morphological studies on T. thermophila . Dispersing cells were more elongated than non-dispersers in our two-patch dispersal experiment, as expected, and strains that elongated much in presence of nutrients also showed greater mean cell elongation under starvation (but a lower variance and hence lower plasticity). Elongated disperser cells, however, differed in shape from the more elongated cells produced under starvation, and strains dispersing much in presence of nutrients surprisingly produced only few of the fast-swimming dispersal morphs described by Nelsen and Debault . Importantly, the differential production of dispersive morph types by different strains was linked to variation in core life history trait associations (growth, survival, colonization, short-distance dispersal), with strains performing poorly under normal conditions producing more fast-swimming dispersal morphs under starvation.
That the fast-swimming very elongated dispersal morphs likely constitute a commitment to long-distance dispersal in degraded environments was supported by our study. Firstly, our dispersal experiment showed that normal cell morphs or partially elongated cells were very capable of short-distance dispersal [69,70]. Secondly, our starvation experiment confirmed that considerable time is required for a cell to transform itself into the very elongated dispersal morph (4–8 hours in this study, which corresponds to half a generation; see temperature dependent data in [69,70]). Hence transformation, likely also energetically costly, should only occur when cells need to surpass the mobility capacities of normal or partially elongated morphs. Strains producing many fast-swimming morphs should therefore benefit from a dispersal-distance advantage, which may help explain the coexistence of low performance strains with strains doing well (see-colonization-competition models for coexistence , see also  and references in these).
Other selective factors could also have contributed to the observed variation among strains in dispersal distance strategies. For example different strains may come from habitats that vary with respect to the selective pressures of kin competition versus spatio-temporal variation in habitat quality, distributions of resources and kin in space, or the shapes of distance-dependent dispersal cost functions (e.g. [17,21,81,82,86,87,89]; see also ). Finally, we cannot reject that strain variation in life history traits could represent non-adaptive variation around an adaptive mean strategy. Intimate knowledge on the habitats of origin of different strains or, better still, experimental evolution studies will be required to test this.