The interplay between movement, morphology and dispersal in Tetrahymena ciliates

Understanding how and why individual movement translates into dispersal between populations is a long-term goal in ecology. Movement is broadly defined as ‘any change in the spatial location of an individual’, whereas dispersal is more narrowly defined as a movement that may lead to gene flow. Because the former may create the condition for the latter, behavioural decisions that lead to dispersal may be detectable in underlying movement behaviour. In addition, dispersing individuals also have specific sets of morphological and behavioural traits that help them coping with the costs of movement and dispersal, and traits that mitigate costs should be under selection and evolve if they have a genetic basis. Here, we experimentally study the relationships between movement behaviour, morphology and dispersal across 44 genotypes of the actively dispersing unicellular, aquatic model organism Tetrahymena thermophila. We used two-patch populations to quantify individual movement trajectories, as well as activity, morphology and dispersal rate. First, we studied variation in movement behaviour among and within genotypes (i.e. between dispersers and residents) and tested whether this variation can be explained by morphology. Then, we addressed how much the dispersal rate is driven by differences in the underlying movement behaviour. Genotypes revealed clear differences in terms of movement speed and linearity. We also detected marked movement differences between resident and dispersing individuals, mediated by the genotype. Movement variation was partly explained by morphological properties such as cell size and shape, with larger cells consistently showing higher movement speed and higher linearity. Genetic differences in activity and movement were positively related to the observed dispersal and jointly explained 47% of the variation in dispersal rate. Our study shows that a detailed understanding of the interplay between morphology, movement and dispersal may have potential to improve dispersal predictions over broader spatio-temporal scales.


51
Individual movement is a universal feature of life with broad implications for the ecology and 52 evolution of species (Turchin 1998). As most environments are spatially structured, 53 understanding how individuals move across increasingly fragmented landscapes is of crucial   between dispersal and movement experimentally, because they allow tight control of the 112 genetic and environmental context and hence allow these to be disentangled.

113
Experimental approaches with microscopic organisms are a convenient way to measure 114 movement and dispersal simultaneously and hence allow us to study pattern and process at a  There is compelling evidence that dispersal in this organism is not solely a diffusive process, In this study, we investigate the relationships between small-scale individual movement (i.e.  Figure S1). To start the experiment, cells of a single genotype were pipetted into the "start" tube to obtain a density of 300000 cells/mL, an 170 intermediate cell density commonly observed under our culturing conditions. After mixing the 171 medium to distribute cells evenly in the start tube and 30 minutes of acclimation, the connecting 172 pipe was opened, and cells could freely disperse. At the end of the experiment after six hours, 173 the pipe was closed by a clamp and five independent samples were taken from both the start 174 and the target tubes of each dispersal system. Cells found in the "start" or "target" are 175 subsequently referred to "residents" or "dispersers", respectively, the two modalities possible  replicate as random effect nested in genotype but crossed with dispersal status. Genotype was 242 considered as a fixed effect, despite its common consideration as a random effect (e.g. Crawley 243 2007). This is because the set of genotypes cannot be considered as a random sample of the 244 genetic variation exhibited by the species in the wild (some genotypes were selected due to 245 previous results or based on their phenotypic characteristics, some others were created by 246 inbreeding in the laboratory). Dispersal status was crossed with replicate because the data for 247 the two statuses (disperser and resident, i.e. target and start tubes respectively) were paired for 248 each dispersal two-patch system. Speed and linearity (tau) were ln-transformed to improve 249 normality.

250
All cells belonging to the same genotype should have the same genetic make-up; however, 251 environmental differences encountered during the cell life cycle may lead to different 252 morphologies and cell states. Therefore, to answer our second question, we tested whether 253 differences in movement behaviour between residents and dispersers may be explained by 254 morphological differences such as cell size and shape. To see if there were differences between 255 residents and dispersers, we built ANCOVA models that related movement speed and linearity 256 to morphology properties (size and shape) across genotypes, accounting for differences due to 257 dispersal status. As some of the observed variation may be due to variation across replicates, 258 we investigated how within replicate differences in morphology affect differences in 259 movement. We used the Akaike Information criterion (AIC) to determine the most 260 parsimonious model, i.e. the simplest model (in terms of number of parameters) within 2 units 261 (deltaAIC < 2) of the best model (i.e. with the lowest AIC).

262
To address our third question about the power of movement behaviour to predict dispersal rate, 263 we assessed how much variation in dispersal rate was explained by genotype-specific activity, 264 movement speed, movement linearity and all predictors together. We used the R² of a multiple 265 regression and compared the three models with the Akaike Information criterion (AIC) to find 266 the best fitting model. For this analysis, movement metrics (activity, movement speed and 267 linearity) were averaged at the genotype level, i.e. over dispersers and residents. 270 Model selection across the four types of correlated velocity models revealed that the advective 271 correlated velocity model (ACVM) was the most common across genotypes, indicating the 272 genotypes show directed movement. The dispersal status did not change the overall pattern, 273 but genotypes showed variation in the relative frequencies of movement models (Figure 1).

274
Genotypes differed in activity (min. 39% to max. 70% of total cell population moving) and 275 movement parameters extracted from the correlated velocity models: movement speed (min.   was also affected by shape differences: more elongated disperser cells moved faster, whereas 291 the opposite was observed for residents (Tab. S2). We also found that larger cells moved 292 straighter. The slope of this relationship did not differ among dispersal status, however, 293 dispersers moved straighter on average (Tab. S3). The relationship between shape and linearity 294 again was dependent on the dispersal status: whereas higher elongation led to more linear 295 movement for dispersers, residents showed no pattern with higher elongation (Tab. S3). Within 296 genotypes, larger relative size of dispersers compared to residents led to higher relative 297 movement speed, whereas a larger relative elongation resulted in a decrease in relative speed 298 ( Figure S4, Tab. S4-S5).    and a dispersive type, showed co-variation between growth rate and dispersal ability (e.g. well 408 growing but poorly dispersing resident vs. poor growing and well dispersing establisher) and 409 if the ratio between genotypes in these parameters varied two-to ten-fold. Looking at the 410 variation of our genotypes (Figure 4), we see that the ratio in dispersal rate can be up to ten-    Table 1: Three-way ANOVA to assess the effect of genotype and the dispersal status (i.e. difference between dispersers and residents) on three 599 movement metrics: activity (proportion of moving cells), movement speed and linearity. Genotype and dispersal status were considered as 600 crossed and fixed effects, and replicate as random effect nested in genotype but crossed with dispersal status because data from the two status 601 were paired per replicate (i.e. the start and target tubes of one dispersal system). Arrows indicate the error term used to test for each effect, 602 according to the ANOVA model; "-" denote the factors that cannot be tested because the error has no degrees of freedom in this model.