The aim of this part was to compare the phenotype of individuals between populations in order to assess if a phenotypic differentiation has occurred in some introduced populations. As explained in the introduction, the phenotypic traits chosen were some morphological characters that were expected to vary across environments and that could be measured on both live individuals and museum specimens. We used two complementary methods to describe the morphology of individuals: classical measurements and geometric morphometrics. The advantage of classical measurements is that they are obtained rapidly and that they can be taken on every parts of the body. However, some studies show that museum specimens tend to shrink with time and thus there is a bias in the lengths measured (reviewed in Engelmoer et al. 1983; Eastham et al. 2000). In addition, the different parts of the body do not shrink with the same amplitude. Some correction factors exist in the literature but they are very dependent on the species, on the trait considered and on the age of the specimens. As there are no correction factors available for Psittacidae, we decided not to use any corrections but rather to bear in mind that there was a possible bias in our data because some individuals were measured alive while some others were measured as skins. Using museum specimens was our only way to obtain data from the native ranges of our two model species. Indeed, organizing field missions to Asia and Africa would have been too costly and time-consuming in the framework of a PhD thesis. In order to reduce the bias caused by the comparison of live individuals and museum specimens, we decided to also use geometric morphometrics to describe the morphology of our specimens. With this approach, it is the conformation of the body that is studied and not lengths. We hypothesized that this measure would be less sensitive to specimen shrinkage. Furthermore, we decided to work on the beak of the individuals as some authors suggest that it shrinks less than other parts of the body such as wings and tail (e.g. Engelmoer et al. 1983). However, this approach had its drawbacks as the digitization process is very long and it allowed us to work only on a small part of the body. Indeed, it would have been very difficult to work on other parts such as wings for example as this technique requires that the organ studied is photographed in standardized conditions for every specimen. It would have been impossible to study the wings of museum specimens as they are not always spread in the same way from one specimen to the other. This is why we used both classical measurements and geometric morphometrics to describe the morphology of individuals.
Classical morphological measurements were transformed into log-shape ratios in order to control for the size effect on the body parts measured (Mosimann & James 1979). Following this method, the overall size of each individual was defined as the mean of the log-transformed measurements. Each measurement was then standardized by subtracting the overall size of the individual to the log-transformed measured value.
A Generalized Procrustes superimposition (Rohlf & Slice 1990) of the points digitized for each individual was performed using TPSRELW (Rohlf 2010b). With this method the set of landmarks digitized for each individuals are transformed in order to minimize differences between individuals. This is done by adjusting their position, rotation and scale while conserving the shape they define (Adams et al. 2004). The size information is thus removed. Semi-landmarks are also slid along the curves they describe to match as well as possible the positions of the corresponding points in a reference specimen randomly chosen (Adams et al. 2004). The coordinates obtained after this step were those used for the analyses of beak shape. The size of the beak of individuals was defined as the log-transformed centroid size (square root of the sum of the square distances between the landmarks and their centroid).
Choice of sequences and amplification protocols
The sequences chosen for the analyses had to be variable enough to enable the discrimination of the different populations. For nuclear DNA, genes are generally too conserved at this scale and could not be used. Unfortunately, little genetic data is available on my two model species and we could not find introns variable enough to be used. I amplified the flanking regions of the microsatellite loci used in the population genetics study but they were also too conserved in my samples. The phylogeographic analyses are therefore based on mitochondrial sequences only. For both species, genes for which there were sequences deposited on Genbank were chosen to complete the sampling. The cytochrome oxidase subunit I (COI) and the NADH dehydrogenase II (ND2) were used for the Red-whiskered bulbuls, and the cytochrome b (Cytb) for the Ring-necked parakeets.
DNA extraction and amplifications were done using classical protocols except for museum specimens for which DNA was too fragmented to amplify genes in one piece. In these cases, genes were amplified in small overlapping fragments of 200-300 bp.
Choice of neutral genetic markers
In order to study the demographic history of a population, neutral genetic markers are needed (i.e. loci that are not under selection and that can thus track other evolutionary forces). We chose to use microsatellite loci as they are generally highly polymorphic. This was indeed necessary for the study of recently introduced populations. The regions flanking the microsatellites have also the advantage of being usually conserved across species. Thanks to this property, we were able to use primers developed in species closely related to our models. Finally, microsatellite loci can be multiplexed which reduces the cost of amplification and genotyping.
For both species, I tried to amplify an important number of microsatellite loci found in the literature on a small subsample of individuals from different populations. Only the loci that amplified well and that were polymorphic were kept. In the end I retained ten polymorphic microsatellite loci for the Red-whiskered bulbul and 18 for the Ring-necked parakeet. These loci were then amplified for all individuals in several multiplex and tagged with fluorescent forward primers. Genotyping was done at the lab on an Applied Biosystems 3130XL DNA sequencer. Genotypes were scored with GeneMapper 4.0 (Applied Biosystems) and checked manually.
Founder effect as the cause of the observed morphological differentiation
The genetic structure of the Red-whiskered bulbul on Reunion, coincides with the morphological differentiation observed by Amiot et al. . However, within each population (and thus morphotype) there was no isolation by distance on Reunion. Moreover, the comparison of the genetic structure on Reunion with that of Mauritius and Oahu showed that isolation and drift alone cannot explain the pattern we observe on Reunion. Indeed, there was no evidence for genetic structure on Mauritius and Oahu although similar natural barrier are present. Thanks to the comparison of invasion scenarios, we inferred that the populations of Red-whiskered bulbul on Reunion were founded by two introduction events. We conclude that a different origin, rather than local adaptation, is probably the cause of the genetic and morphological differences observed between the two coasts of Reunion. This is supported by the findings of Roussel et al.  who showed with an isotopic analysis that they were no differences in diet or feeding behaviour between bulbuls of the two coasts of Reunion. In that case, several non-exclusive mechanisms can explain the genetic and morphological differences between the two populations: 1) because of a sampling effect, the individuals which founded the two populations already had different genetic backgrounds and morphologies at the moment of the introduction, 2) drift and/or selection created the observed differences in-situ after the introduction.
Table of contents :
I.1.1. Why study rapid evolution?
I.1.2. Invasive species and rapid adaptation
I.2. What has been done on rapid adaptation, what remains to be done?
I.3. Question and methodological choices
I.4. Presentation of our two case studies
I.4.1. The Red-whiskered bulbul (Pycnonotus jocosus)
I.4.2. The Ring-necked Parakeet (Psittacula krameri)
II.1.1. Development of capture protocols
II.2. Morphological comparisons
II.2.1. Data acquisition
II.2.2. Data standardization
II.3.1. Choice of sequences and amplification protocols
II.4. Population genetics
II.4.1. Choice of neutral genetic markers
III.1. Synthesis of the results
III.2. Organisation of the results in the manuscripts
IV.1. Advances made
IV.1.1. Beware of correlations between phenotypic and environmental changes
IV.1.2. What about rapid adaptation and phenotypic plasticity?
IV.2.1. Improving sampling of introduced populations
IV.2.2. Refining the phylogeographic studies
IV.2.3. Studying other phenotypic traits and their heritability
IV.2.4. Describing the different environments studied
IV.2.5. Using complementary approaches
IV.2.6. Perspectives in the study of rapid adaptation in general