Results
The haplotype network constructed using only the North American specimens (1181 bp sequence) included 20 haplotypes encompassing 53 polymorphic sites and did not indicate the presence of regional population structure (Figure S1). When we included samples from all continents (928 bp sequence; Figure 2), we identified 64 haplotypes encompassing a total of 46 polymorphic sites (Table S5). Only one of these haplotypes was shared between two introduced regions (H_25; North American and Australia, Figure 2, Figure 3). Although haplotypes from North America and Australia were phylogenetically similar, South African haplotypes were completely separated from the other invasive haplotypes, by a minimum of eight mutations. Four native range haplotypes were found in North America (H_11, H_25, H_29, H_31). Overall, native range haplotypes were well distributed across the four-continent network but not all basal haplotypes (e.g. H_51) were represented in the native range samples. Table S5 indicates polymorphic positions across haplotypes from all four populations and integrates naming conventions from the present study, and those of Australian and South African populations.
            FST values from pairwise comparisons between the native range and the three introduced populations ranged between 0.060 (North America) and 0.174 (Australia) and were all highly significant (Table S2). When the North American population was separated into three separate regions (Eastern, Central and Western sampling sites), the FST values for all comparisons were low (FST ≤ 0.044) and only significantly different for the Central vs. Western US comparison (Table S3).
Haplotype diversity and richness was highest in the native range, followed by the North American population (Table 2). Tajima’s D values were nonsignificant in all four populations. Fu’s Fs values were negative in North America and the native range, but only significant in the latter (Table 2). The mismatch distribution model for sudden (demographic) expansion was significantly different than empirical data from the South African population (SSD=0.081, P=0.04) but not from the Australian (SSD=0.107, P=0.07) nor North American populations (SSD=0.15, P=0.10) (Figure 4). The mismatch distribution model for spatial expansion was not different to that of empirical data from any of the three invasive populations (South Africa: SSD=0.075, P=0.06; Australia: SSD=0.057; P=0.26; North America: SSD=0.009, P=0.56; Figure 4).
 
Discussion
Starlings are a highly successful invasive species occupying a wide breadth of environments across the world, resulting from introductions of varying age and intensity. This system enables a unique opportunity to study molecular evolution and adaptation. Here we use mitochondrial sequence data to compare the population genetic structure and diversity of the three best-studied starling invasions: North America, Australia and South Africa. Overall, our findings and those from data of other studies included here suggest that low genetic diversity is not an obstacle for this species’ rapid expansion and establishment in new environments (Dlugosch & Parker 2008; Rollins et al. 2013).
As expected, the invasive populations had lower genetic diversity than the population in the native range, likely caused by genetic bottlenecks at introduction. The highest haplotype richness (which accounts for differences in sample size) was found in the UK (R=30.0); although only 45 individuals were sampled, we identified 30 haplotypes in this population. Surprisingly, despite higher propagule pressure in Australia as compared to that of North America or South Africa, Australia harbored the lowest haplotype richness (R=7.7). The North American population, which was intermediate in terms of propagule pressure has retained the most genetic diversity (R=14.7). Given the timescales involved, this is unlikely to be caused by novel mutations arising in North America (but see Rollins et al. 2016). However, it could be caused by differences in genetic diversity of founders or by higher levels of differential survival between haplotypes in Australian or South African starlings as compared to those from North America. It may be that some haplotypes have been lost in the native range since founders were collected. Differences in population expansion rates in novel environments also could be responsible for the differences in genetic diversity we found, with faster expansion resulting in higher haplotype diversity and lower nucleotide diversity (Halliburton 2004).
The haplotype network including all populations (Figure 2) revealed some interesting relationships among haplotypes. South African starlings are phylogenetically distinct from those of North America and Australia, suggesting that the founders for this population may have been sourced from a different region of the UK. North America and Australian starlings are phylogenetically similar (intermixed in the network), but only shared a single haplotype (H_25), suggesting that the founders for these populations may have been sourced from the same region of the UK, but were likely to have been genetically distinct. As expected, UK samples were well-distributed across the network, but many of the invasive haplotypes were not found in UK samples, highlighting the paucity of information that exists about starlings in their native range and making it difficult to further interpret sources of founding populations. For this reason, and because European starling populations are in decline in their native range (Heldbjerg et al.,2019), it may be important to further characterize this population.
Previous studies have investigated population structure within introduced populations of starlings. Within Australia, genetically distinct groups of starlings have been characterized using nuclear and mitochondrial markers (Rollins et al. 2009, 2011) and evidence of local adaptation to the Australian environment has been described (Cardilini et al. 2016, 2020). However, in South Africa, no evidence of population structure was found (Berthouly-Salazar et al. 2013). The regional analysis conducted within North America in the present study also found little evidence of population structure in this invasive population. We did see a slight (FST = 0.04) albeit statistically significant difference between Central and Western samples but this may be due to the low sample size from the Central US (N=20). Overall, our findings are consistent with an earlier investigation of this population, which utilized allozyme data (Cabe, 1998), and a recent study using genome-wide SNPs (Hofmeister et al., 2019). However, the latter indicated that there are genotypes associated with specific environmental features such as precipitation and/or temperature. This may imply that over time, population structure could develop in this invasive population, despite apparent high levels of dispersal. Interestingly, migration rates between Central and Western sites differ (Hoffmeister et al. 2019) and banding data in North America have shown that the starlings are found to migrate in unpredictable ways, not always in the North and South direction, but also in the East and West directions (Brewer, 2000). Therefore, the genetic pattern we found may be due to the high dispersal rates and these unpredictable and latitudinal migration patterns.
            When we investigated genetic differentiation across continents, we found that invasive populations were genetically divergent (FST ranged from 0.17-0.26, all highly significant) and all significantly different from populations in the native range (FST ranged from 0.06-0.17). North America was most similar to the UK and Australia was least similar. These differences are likely caused by a combination of discrete introduction sources and founder effects. However, this could also be due to differences in timing of introductions; the Australian introduction occurred earlier than the others (mid-19th century) so it is possible that these differences reflect shifts that occurred in the native range in the latter half of the 19th century.
Not surprisingly, we found genetic evidence of spatial expansion in all three invasive populations. While there was genetic support for demographic expansion in both North America and Australia, the mismatch analysis of South African data did not support the sudden (demographic) expansion model (Figure 4). This may mean that the South African starling population may still be in the “lag phase”, which typically occurs following introduction (Sakai et al. 2001). Neither Tajima’s D nor Fu’s Fs values supported the presence of population expansion in any of the invasive populations. However, Fu’s Fs was significantly negative in the native range, which suggests that this population may either be undergoing expansion or that it has an excess of recent mutations (Fu 1997). Given observations of population decline in the native range (described above), this might be a signal of directional selection, which could be a response to novel environmental stressors resulting from land use changes in the UK (Heldbjerg et al.,2019).
            It is also interesting to consider that differences in the environments of each of the three invasive ranges studies here may have influenced population expansion rates. The United Kingdom and surrounding parts of Europe (native range) are largely classified as temperate with a hot or warm summer (Beck et al., 2018). Temperate areas similar to the native range are the regions where most starling invasive range expansion has occurred. The starling population in North America is about the same latitude as that of the native range between 40° – 55°N, whereas the invasive populations in Australia and South Africa occur at about 30° – 35°S (Sulliven, 2009). In Australia and South Africa, starlings have not expanded to cover the same area that they have in a comparable amount of time in North America. In North America starlings spread from New York to Alaska from 1890-1970, which represents 80 years and a rate of 90km/year (Bitton and Graham 2014). In Australia, starlings rapidly expanded their range into south-eastern Australia and were in Western Australia by the 1970’s. However, starlings have not colonized the arid center (Higgins et al., 2006) of the continent, where the highest temperatures and lowest rainfall occur (Jones et al, 2009). In South Africa, starlings spread primarily eastward from Cape Town, and are now reported only as far north as Kruger National Park (Berthouly-Salazar 2013; Sulliven, 2009). Similar to Australia, large areas of South Africa are classified as an arid, hot, dessert with surrounding areas classified as arid, hot, steppe (Beck et al. 2019). These environmental differences pose an explanation for the rapid and continued expansion of European starlings in North America and associated mitigation of loss of genetic diversity, and suggests that the success of starlings in South Africa and Australia may have depended upon adaptation to novel climatic conditions.
There is still much to learn about the population dynamics and genetic structure of European starling invasions world-wide, and about the native range genetics of this species. The mitochondrial data set we have extended here is a useful tool to grow our knowledge of this species and, more generally, of invasion genetics. Despite our knowledge gaps, starlings provide an intriguing framework to study invasions of different ages and geographic extent (e.g. South America (small) vs. Australia (large), of similar or contrasting genetic backgrounds (e.g. North America vs. Australia (phylogenetically similar), North America vs. South Africa, (phylogenetically different) and across different environments (e.g. North America (temperate) vs. Fiji (tropical)). Together with the recent development of genomic resources (transcriptome: Richardson et al. 2018; genome (GCF_001447265.1: Nucleotide [Internet])), the features of this species make it ideal for advancing our knowledge of evolution in introduced ranges. Especially with continued global climate change, closely monitoring invasive species and understanding their outsized adaptive flexibility will be increasingly important to our ability to manage invasions and to understand how species adapt to a changing world.