This thesis describes the global molecular population structure of two shorebird species, in particular of the dunlin, Calidris alpina, by means of comparative sequence analysis of the most variable part of the mitochondrial DNA (mtDNA) genome. There are several reasons why mtDNA is the molecule of choice to probe the recent evolutionary history of a species. Most importantly, mtDNA accumulates substitutions at a high average rate that permits the tracing of genealogies within the time frame of speciation. The population structure of shorebirds, like that of arctic- breeding waterfowl (Ploeger, 1968), must have been influenced dramatically by the Pleistocene glaciations (mainly during the last one million years). The fastest evolving part of the mtDNA genome, the non-coding control region, offers sufficient genetic resolution to reveal differentiation of such recent origin. The typical mode of maternal inheritance, the absence of recombination, and the presumed neutrality of substitutions, are characteristics that add to the suitability of mtDNA for the construction of robust phylogenies ( Chapter 1 ).Cloning and sequencing of the control region of a turnstone ( Arenaria interpres ) facilitated subsequent amplification and direct sequencing of the homologous region in other turnstones, and dunlins as well. Comparison of this approximately 1200 basepairs (bp) region for several turnstones, dunlins and a chicken ( Gallusdomesticus ) revealed the presence of differentially evolving sequence blocks within the control region. Both shorebird species contain an AC repetitive sequence at the 3' end of the light strand, varying in size (around 100 bp) and composition between individuals. Sequence identity is highest in the central part of the control region, similar to the conservation of this part in other vertebrate species. Most single nucleotide substitutions, as well as insertions and deletions, are restricted to two segments, notably at the beginning and near the end of the control region. Overall, the organization of the avian control region resembles its human counterpart. Sequence comparison of the larger variable segment at the beginning of the control region (CR I) for worldwide samples of 25 tumstones and 25 dunlins demonstrated the utility of this region for the detection of intraspecific differentiation. The turnstone reveals few differences worldwide and identity of clones from distant regions, whereas the dunlin reveals divergent clusters of genotypes that are geographically restricted. It is concluded that the turnstone has been confined historically to one Pleistocene refugium, from which it has dispersed around the world to establish its current biogeography. The dunlin, on the other hand, became divided into several isolated populations during the Pleistocene and has retained a significant amount of intraspecific genetic diversity until the present ( Chapter 2 ).This remarkable difference in population genetic structure between the two shorebird species may be explained by their differing ecologies. The turnstone is a high arctic breeder, and depends mainly on cold tundra habitat, whereas the dunlin breeds mostly in the lower arctic and even in temperate zones. Cold tundra habitat may have disappeared almost entirely during the last interglacial (Eemian: around 125,000 years ago) that was characterized by high temperature peaks (Anklin et al., 1993). A very similar lack of global mtDNA differentiation has been observed in the knot ( Calidris canutus ) , another shorebird that is a typical breeder of the high arctic tundra (A.J. Baker and T. Piersma, personal communication).Representative samples of dunlin populations from four major regions in the world were analysed for 910 bases of mtDNA sequence from the control region and the cytochrome b gene. The regions comprised the Pacific coast of North America, the Atlantic coast of North America, the Atlantic coast of Europe and arctic central Siberia. Sequence comparison of the three amplified DNA fragments showed that most substitutions are located in the CR I fragment, and substantially less in another control region segment (CR II) and part of the coding sequence of the cytochrome b gene. The 50 substitutions that were found together defined 35 different genotypes. A genealogical tree relating these genotypes revealed five major clusters. Each cluster has high geographic specificity. The cluster containing the most divergent sequences is present along the Atlantic coast of North America and represents the dunlin population breeding in arctic central Canada. Two clusters of genotypes are located principally in western Europe and central Siberia. Evidence for a low level of gene flow between these latter two populations was provided by three individuals whose genotypes suggested they were immigrants. Two other clusters are found along the Pacific coast of North America. Whereas dunlins from southern Alaska assorted to one cluster, dunlins from the southerly wintering population revealed genotypes of both clusters.The genetic divergence of these major mtDNA lineages can be dated to the late Pleistocene based on a molecular clock for the control region of birds. Genotypic diversity within the population samples is extensive and the calculated long term effective (female) population sizes argue against strong historical bottlenecks. Overall, there is a negative correlation of mtDNA variation and previously defined morphometric variation in dunlins. This discordance is induced largely by the morphometrical similarity of the genetically most divergent populations from both North American coasts.A plausible scenario for the genetic divergence of the major dunlin lineages is the ancestral fragmentation of populations over tundra refugia, that were created by the extensive glaciations of the northern hemisphere during the Pleistocene. Prolonged isolation of populations of reduced size increased the effect of genetic drift and this may have led to the observed mtDNA monophyly. The different lineages continuously diverged by the process of mutation. This ancient subdivision has been retained after retreat of the icesheets, most likely as a result of the strong site-fidelity of dunlins to their breeding ground. Dunlin populations could thus not become homogenized genetically because gene flow is not extensive enough between them ( Chapter 3 ).The generally observed lack of genetic population differentiation in birds, in contrast to other vertebrate groups, has been interpreted as a sign of panmixia, caused by the high dispersive capabilities of birds (Cooke and Buckley, 1987). This conclusion is mainly based on the analysis of allozyme data, but more recently also on the analysis of mtDNA restriction polymorphism (Ball et al. 1988). However, allozymes are relatively conserved genetic markers, and thus do not provide resolution at shorter time scales of evolution. The dunlin is not exceptional in its degree of natal philopatry. Rather, the findings in dunlin indicate that population structure in this species is of recent evolutionary origin, that could be detected by virtue of the high rate of nucleotide substitution in the selected mtDNA sequences. In addition, the global coverage of this study is beyond the geographical scope of most avian studies, and thus had a better perspective for detecting major phylogenetic splits within a species.To elucidate the geographical distribution of mtDNA lineages over the circumpolar breeding range of the dunlin (intraspecific phylogeography: Avise et al., 1987), many additional samples from interspersed populations were analysed for both control region segments. No additional major lineages turned up among 155 breeding dunlins, but one lineage previously found among wintering dunlins in western North America could be located to the eastern Siberian breeding ground. Samples from breeding birds in Greenland, Iceland, the Baltic, southern Norway, northern Norway, and western Siberia revealed genotypes that cluster together in the major European lineage. The central Siberian lineage was found in northern Russia from the Lena river delta in the east, across the Taymyr peninsula in the middle, to the Yamal Peninsula in the west. A few of these 'central Siberian' genotypes were retrieved from dunlins breeding in Norway and eastern Siberia, indicating a restricted amount of gene flow between these populations. A zone of geographical overlap between the European and the central Siberian phylogeographic groups is present at the Yamal peninsula, where equal numbers of dunlins assorted to these respective major lineages. Dunlins captured in northern, western and southern Alaska all belonged to the same mtDNA lineage and thus constitute one genetic population.A large fraction of the total mtDNA variance in dunlins is distributed between the five major phylogeographic regions (76%). Extensive diversity also exists, however, among the individuals of a local population. This is induced by the high rate of substitution in CR I and renders the traditional population genetic correlation measure GST less applicable. Time estimates for the corrected sequence divergence of each phylogeographic group on the basis of a molecular clock indicate a repeated fragmentation of populations, and coincide well with the onset of glacial periods. The ancestral population in central Canada may have been separated from all other dunlin populations for over 200,000 years.Phylogeographic groups can be correlated to the global geography of morphometrically defined subspecies in the dunlin. Whereas several disputed subspecies gain support from the genetic data (i.e. C . a. hudsonia in central Canada and C . a. centralis in central Siberia), other subspecies merge into the same phylogeographic group. No major phylogenetic divisions are apparent among the morphometrically dissimilar populations in north-eastern Greenland, Iceland, the Baltic Sea, and Norway (recognized until now as three to four different subspecies). Gauged by the depth of the other phylogenetic splits in dunlins, they can jointly be referred to as C . a. alpina, Similarly, the dunlins from northern and southern Alaska can be merged under C. a. pacifica.Detailed comparison of populations in Europe reveals a developing geographic specificity of slightly divergent genotypes of the European genetic cluster. Intermediate genotypic correlation measures between locales are supported by measures of restricted gene flow, particularly for the Icelandic and Baltic populations. The genetic differences between European populations have likely evolved after retreat of the ice sheets, approximately 10,000 years ago. Post-Pleistocene colonization of newly exposed breeding grounds combined with the habit of strong site-fidelity can explain the population differentiation within Europe ( Chapter 4 ).It is thus revealed how morphology lacks an evolutionary perspective in the determination of intraspecific taxonomy. For the dunlin, a parallel morphological evolution of genetically divergent populations, as well as the opposing process of morphological divergence of evolutionary closely related populations, is observed. Morphometric characters employed in intraspecific avian taxonomy are suffering from homoplasy, either as a result of character plasticity and environmental induction (James, 1983), or because of very high mutation rates and strong directive selection acting on phenotypes (Turelli et al., 1988). Because morphometrically different dunlin populations are often mixed outside the short breeding season, environmental induction of morphology seems unlikely, although this possibility remains to be investigated. Although the concept of a molecular clock is debatable, general agreement exists as to the neutrality of most nucleotide substitutions in DNA and the cumulative character of the mutation process. On the basis of statistically reliable amounts of substitution, the phylogenetic branching order of intraspecific lineages can therefore be inferred with precision. This applies even more so to the non-coding mitochondrial control region. Although the oldest split in the dunlin mtDNA phylogeny is dated at approximately 200,000 years ago, the species itself is probably much older, in the range of a million years (Baker, 1992). This time discrepancy could imply that many populations have been transient in the intraspecific history of the dunlin. Only populations that radiated during the later part of the Pleistocene have survived until the present. The observed genetic differentiation within Europe thus represents the shallow branch tips in the phylogenetic tree of dunlins. The mtDNA assays suggest that measures to protect declining breeding populations in Europe, like the dunlins breeding around the Baltic Sea, cannot be argued for on the basis of a subspecific status of these populations. Rather, subspecies should be reserved for groups that represent a major source of intraspecific genetic diversity.Limited numbers of migratory and wintering dunlins from around the world were sequenced for both control region segments to trace lineages away from the breeding grounds. The mtDNA lineages detected in these birds were identical to those already known from the breeding grounds. Mixtures of major mtDNA lineages are present in different regions of the world. Samples of dunlins from both sides of the Pacific Ocean comprised two lineages that were found separately on the breeding grounds in eastern Siberia and Alaska. The two lineages identified in population samples from the western Palearctic (western Europe and western Asia) correspond to those present on the breeding grounds in Europe and central Siberia. Overall, it appears that dunlin populations breeding in different circumpolar regions occupy overlapping areas on migration and in winter through much of their southern range. Dunlins wintering along the North American west coast can be assigned to the Alaskan as well as to the eastern Siberian breeding grounds. In parallel, it is likely that dunlins migrating along the eastern Pacific coast of Asia originate from northern Siberia as well as from Alaska. Because the Alaskan genotype found in some eastern Asian dunlins occurs in high frequency only in birds breeding in northern Alaska, it appears that the northern and southern Alaskan populations migrate in different directions. The allocation of individual dunlins to their breeding population on the basis of their mtDNA genotype, can only be certain for those lineages that are geographically separated on the breeding grounds. Because of the limited gene flow between the European and central Siberian breeding populations, uncertainty exists in the population assignment of western Palearctic dunlins. Additional characters such as body mass and time of passage during spring migration or the presence of a particular moult pattern during fall migration can be instructive for the discrimination of dunlins of Siberian origin at European staging posts. These characters seem to be correlated with the possession of a central Siberian genotype by individual dunlins. Larger sample sizes remain to be tested, however, to obtain a better estimate of the diagnostic value of each of these methods ( Chapter 5 ).It is not clear what underlies the different genetic compositions of dunlin populations at the breeding grounds versus the wintering regions. Dunlins probably have also migrated during the Pleistocene, under the influence of seasonal temperature fluctuations. Their wintering quarters may have been fragmented by extensive glaciations, just as the breeding grounds were. Sharing of wintering grounds would likely have opened a route for exchange of individuals between the different populations. Such gene flow would have hampered the process of stochastic lineage sorting under the influence of genetic drift. The five major flyways that are recognized for dunlins around the world today, may still partially reflect the separate ranges that were occupied by populations throughout the last glacial period. Only the population migrating along the Atlantic coast of North America has remained geographically fully separate. What causes the mixing of the other populations outside their breeding range? The question might simply be reversed. Why does the subdivided population structure still exist over the northern breeding range? This can be explained by the imprinting of natal site-fidelity in the juvenile dunlin. Juvenile dunlins leave the breeding grounds indepently of the adult birds in a rough general direction, that may also be imprinted. Their exact direction of southward migration, however, is likely a learned behaviour and this is more prone to error or change (Rösner, 1990).This thesis demonstrates the utility of mtDNA in elucidating the population genetic structure of a bird species. By sequencing the most variable part of the mtDNA genome the major gene pools within a species can be detected together with their phylogenetic relationships. On this basis important insights into the evolutionary history and also the life history of the dunlin Calidris alpina were gained. This method should prove highly valuable not only in the detection and preservation of genetic diversity in dunlins but also in other (endangered) animal species.