Population structure and diversity of an invasive pine needle pathogen reflects anthropogenic activity

Dothistroma septosporum is a haploid fungal pathogen that causes a serious needle blight disease of pines, particularly as an invasive alien species on Pinus radiata in the Southern Hemisphere. During the course of the last two decades, the pathogen has also incited unexpected epidemics on native and non-native pine hosts in the Northern Hemisphere. Although the biology and ecology of the pathogen has been well documented, there is a distinct lack of knowledge regarding its movement or genetic diversity in many of the countries where it is found. In this study we determined the global population diversity and structure of 458 isolates of D. septosporum from 14 countries on six continents using microsatellite markers. Populations of the pathogen in the Northern Hemisphere, where pines are native, displayed high genetic diversities and included both mating types. Most of the populations from Europe showed evidence for random mating, little population differentiation and gene flow between countries. Populations in North America (USA) and Asia (Bhutan) were genetically distinct but migration between these continents and Europe was evident. In the Southern Hemisphere, the population structure and diversity of D. septosporum reflected the anthropogenic history of the introduction and establishment of plantation forestry, particularly with Pinus radiata. Three introductory lineages in the Southern Hemisphere were observed. Countries in Africa, that have had the longest history of pine introductions, displayed the greatest diversity in the pathogen population, indicating multiple introductions. More recent introductions have occurred separately in South America and Australasia where the pathogen population is currently reproducing clonally due to the presence of only one mating type.


Introduction
Biological invaders are plants, animals, invertebrates, or microorganisms that have become established in a new area and that threaten, or have a detrimental effect on the biodiversity and ecology of the new environment (Pimentel et al. 2000;Sakai et al. 2001;Allendorf and Lundquist 2003;Anderson et al. 2004). It is well recognized that biological invasions by fungal plant pathogens have had huge impacts on natural and managed forest ecosystems as well as plantation forests (Elton 1958;Desprez-Loustau et al. 2007). Two well-known examples of devastation caused by introduced pathogens on native trees in natural forests include Chestnut blight caused by Cryphonectria parasitica in North America and Europe (Anagnostakis 1987;Heininger and Rigling 1994) and Dutch elm disease caused by Ophiostoma ulmi and O. novo-ulmi in Europe and North America (Brasier 1991). Human-created environments such as plantation forests, consisting of non-native trees, have also been severely damaged by diseases, particularly in the tropics and Southern Hemisphere (Wingfield et al. 2001;Wingfield 2003).
The ascomycete fungus, Dothistroma septosporum (teleomorph: Mycosphaerella pini), that causes Dothistroma needle blight (DNB), is by far the most important invasive pathogen of non-native pine species (Gibson 1972;Ivory 1987;Bradshaw 2004; Barnes et al. 2008a). This disease results in successive needle defoliation, a reduction in stem diameter and height growth and, in severe cases, tree death (Gibson et al. 1964). DNB has been reported from more than 63 countries worldwide, infecting over 82 different species of pine and occasionally other conifer species in their native and non-native ranges (Bedn a rov a et al. 2006;Watt et al. 2009).
One of the main factors contributing to the increase in biological invasions by fungal plant pathogens is the expansion of global travel and trade. This has promoted the anthropogenic introduction and spread of these organisms within and between countries, mainly via infected plant material (Richard and Lonsdale 2001;Rossman 2001;Wingfield et al. 2001). In the Northern Hemisphere, for example, planting stock of Pinus nigra and Pinus mugo infected with D. septosporum was intercepted when transported into the Czech Republic via Hungary (Jankovsk y et al. 2004). The intercontinental spread of D. septosporum was most likely due to the increase in air traffic and the establishment of non-native commercial pine plantations in the Southern Hemisphere, especially after World War II (Gibson 1972). In New Zealand, the pathogen is speculated to have been introduced by forestry officials who travelled to East Africa in 1957 to observe DNB (Hirst et al. 1999). The pathogen was discovered 5 years later causing disease in the central North Island forests of New Zealand (Gilmour 1967).
Introductions into new environments can also be due to natural events such as long distance dispersal via windblown spores (Brown and Hovmøller 2002;Stukenbrock et al. 2006). Natural long distance dispersal of D. septosporum into Australia from New Zealand via spores in mist clouds blown over the Tasman Sea, has been suggested by Edwards and Walker (1978). This view is supported by the fact that strict quarantine regulations are implemented in Australia, which would make it unlikely that an introduction via plant material had occurred (Edwards and Walker 1978;Bradshaw 2004).
A characteristic of a successful invasive species, after introduction, lies in its ability to become established in a new environment and then to spread to new areas (Sakai et al. 2001). Short distance spread of the spores of D. septosporum, either as asexual conidia or sexual ascospores is very effective (Gibson 1972;Bradshaw 2004). The ability of the pathogen to expand its range can thus be observed from the chronology of the first records from countries both in the Northern and the Southern Hemisphere.
In the Southern Hemisphere, the first report of D. septosporum was from Zimbabwe in 1940, where young Pinus radiata trees were severely damaged (Table  S1, Barnes 1970;Gibson 1972). In 1957, the pathogen was found in Tanzania and within 7 years, the associated disease had spread to all major P. radiata plantations in Kenya, Malawi, and Uganda (Gibson 1972). Similarly, in Chile and New Zealand, where 92% of the world's P. radiata is grown (Toro and Gessel 1999;Rogers 2002), D. septosporum has caused disease epidemics since 1957 (Gibson 1972) and 1963 (Gilmour 1967), respectively. In Australia, the pathogen was found much later, in 1975 (Edwards and Walker 1978).
During the past 50 years, D. septosporum has successfully invaded and become established in many European countries. During this time, the geographic range of the pathogen has expanded and serious disease epidemics have emerged. In Serbia, for example, the pathogen has been known since 1955 (Krsti c 1958) and by 1988, severe epidemics on both native and exotic hosts had occurred (Karadzi c 1988). From there, the pathogen spread northwards, and by 1969 it had entered the southern part of Hungary (Karadzi c 1989). By 1995, the DNB fungus had spread to virtually all P. nigra monocultures in Hungary (Szab o 1997), most of which had been established during a pine afforestation program in the 1960s ( AESZ 2002). Approximately 2 years later, the disease was recorded in the Southern tip of Slovakia, close to the border of Hungary (Barnes et al. 2011). Today, it is found throughout Slovakia on both native and non-native pine species (Z ubrik et al. 2006).
Given its importance, it is surprising how little is known regarding the origin of D. septosporum. In this regard, there are two hypotheses (Evans 1984;Ivory 1994). Based on its presence on native pines in the high elevation, minimally disturbed cloud forests of Central America, and in the absence of conspicuous epidemics, one view is that D. septosporum is native to that region (Evans 1984). The fact that the pathogen has also been found on indigenous pine trees in remote forests in the Himalayas prompted Ivory (1994) to suggest that it might also be native to these areas. Given the fact that two fungal species, D. pini and D. septosporum cause DNB (Barnes et al. 2004) and that these species are almost impossible to discriminate from each other without DNA-based identification techniques (Ioos et al. 2010;Barnes et al. 2011), reports such as those listed above, and hypotheses regarding the origin of these pathogens, are speculative at best. Recently, the presence of D. septosporum on native blue pine (P. wallichiana) trees in the Himalayas has, however, been confirmed (Barnes et al. 2008a) and this area of origin seems probable.
Although the biology and ecology of D. septosporum has been well documented, little is known regarding the global structure and population genetics of the pathogen. The aim of this study was to determine the diversity and genetic structure of D. septosporum populations from a broad collection of isolates from 14 countries across six continents using 12 polymorphic microsatellite markers.
More specifically, the aim was to determine whether the patterns of genetic diversity and structure of D. septosporum populations reflect the movement of its hosts from the Northern Hemisphere to the Southern Hemisphere or whether epidemic populations on non-native pines in both the Northern and Southern Hemisphere reflect recent introduction events. In addition, the degree of genetic differentiation and variation that exists between populations and groups of populations was determined and gene flow between geographic locations was estimated. For the possibility of sexual recombination to occur, both mating types need to be present in a population. Thus, the frequency and distribution of mating types in the D. septosporum populations was also determined.

Materials and Methods
Sampling, fungal isolations, and DNA extraction Isolates of D. septosporum were obtained from a variety of Pinus spp. representing 14 different countries (referred to as populations) from six regions (=continents). These included Africa (South Africa and Kenya), Europe (Austria, Czech Republic, Hungary, Poland, Romania, Slovakia), Asia (Bhutan), North America (U.S.A.), South America (Chile and Ecuador), and Australasia (Australia and New Zealand) ( Table 1). The main pine species from which collections were made included P. nigra in the Northern Hemisphere and P. radiata in the Southern Hemisphere. The sampling strategy from plantation forests included collecting a handful of diseased needles from every second tree along two or more transects. Samples from native pine species, or from hosts not growing in plantations were opportunistically collected in areas where infected trees could be located. Samples were collected at one or several locations in a country (Table 1).
Isolations were made from a single conidiomata on a needle, per tree, as described by Barnes et al. (2004). Cultures were grown on 2% malt extract agar (MEA, Biolab, Midrand, Johannesburg) at 20°C. All cultures are maintained in the culture collection (CMW) of the Forestry and Agricultural Biotechnology Institute (FABI), University of Pretoria, South Africa. Genomic DNA was extracted from each isolate using freeze-dried, ground mycelium, from 2to 3-month-old cultures with the aid of the DNeasy Plant Mini Kit (Qiagen, Hilden, Germany).
Polymerase chain reactions (PCR) were carried out in a total volume of 12.5 ll and contained 5-10 ng DNA template, 0.06 U FastStart Taq DNA Polymerase (5 U/lL) (Roche Diagnostics GmnH, Mannheim, Germany), 19 PCR buffer containing 2 mmol/L MgCl 2 , 0.25 mmol/L of each dNTP, 100 nmol/L of the forward and reverse primers, one of which was fluorescently labelled and 1 mmol/L additional MgCl 2 . The PCR conditions consisted of an initial denaturation step of 10 min followed by 10 cycles of 94°C for 30 s, specified annealing temperature for 45 s (as described in Barnes et al. 2008b) and 72°C for 1 min. A further 30 cycles were carried out using the same conditions as those described above except that a 0.5 s increment was added to the elongation time. A final elongation period of 30 min at 60°C was added to avoid the +A effect (Clark 1988;Magnuson et al. 1996) during genescan analyses. PCR amplicons were separated, with a 100 bp marker, on 2% low electroendosmosis agarose gels (Roche Diagnostics) stained with ethidium bromide and visually analyzed under UV light (Vilber Lourmat, Omni-Science, Cedex, France). PCR amplicons were not purified.
To facilitate multiplexing, PCR amplicons (for each individual) were combined according to the approximate size of the amplicons and type of fluorescent label attached to the primer. Samples were subjected to electrophoresis on an ABI 3100 sequencer. Allele assignments were determined using ABI-Prism â GENEMAPPER TM software version 3.0 (Applied Biosystems, Foster City, CA). Multilocus genotypes were obtained by combining the alleles present at all twelve loci for each isolate. All isolates having the same multilocus haplotype in a population were considered as clones.

Genetic diversity
Allele frequencies were estimated for each SSR locus using the program POPGENE version 1.32 (http://www.ualberta. ca/~fyeh) (Yeh et al. 1999). The total number of alleles, unique alleles and gene diversity (Nei 1973), were computed for each population and region across all 12 loci.
In addition to calculating gene diversity, estimates of allelic richness were computed for each population and region in the program FSTAT for windows, version 2.9.3.2   (http://www2.unil.ch/popgen/softwares/fstat.htm) (Goudet 2001). Unequal sample sizes were standardized, by rarefaction, to a uniform sample size of the smallest population and region (USA/North America, N = 10) as described by El Mousadik and Petit (1996). Romania was excluded from this analysis due to low sample size (N = 4). The clonal fraction was calculated for each population by dividing the number of genotypes observed in the population by the total population size and subtracting this from one. Measures of genotypic diversity (D) were quantified in MULTILOCUS version 1.3 (http://www.agapow.net/software/ multilocus) (Agapow and Burt 2001) as (n / n À 1) (1 À Σ i pi 2 ) where pi is the frequency of the ith genotype and n is the number of individuals sampled. Here, the multilocus genotype of all possible pairs of individuals is compared and the proportion of pairs that are different is calculated. Completely clonal populations would score a value of 0 while those where all individuals have different multilocus genotypes would score 1.
The hierarchical partitioning of molecular variation within and among populations and among regions was assessed with an AMOVA -test implemented in GENALEX version 6.1 (Peakall and Smouse 2006) using the complete dataset. The significance was tested by 1000 permutations of the dataset. Only regions with more than two populations were included in the analyses. The null hypothesis of no genetic difference between populations was rejected at P < 0.05.

Population structure
The genetic distance between all isolates was calculated using Nei's standard distance D A (Nei et al. 1983) in POPULATIONS version 1.2.32 (http://bioinformatics.org/ project/?group_id=84). The distance matrix generated was used to construct a Neighbor-joining (NJ) tree in MEGA version 5 (Tamura et al. 2011).
The program STRUCTURE version 2.2 (Pritchard et al. 2000;Falush et al. 2003) was used to determine the optimal number of populations (K) and to assign individuals to these distinct populations based on their genotype data without any prior indication of geographic location or collection sites. The program uses a Bayesian Monte Carlo Markov Chain (MCMC) clustering algorithm and the simulations assumed a model of mixed ancestry and correlation of allele frequencies within clusters. Individuals were assigned to clusters to minimize Hardy-Weinberg disequilibrium and linkage disequilibrium between the loci within each cluster. The MCMC scheme was run for 100,000 iterations after a burn-in period of 10,000. Twenty simulations were performed for K ranging from 1 to 14 to verify the convergence of the Log likelihood values for each value of K. Delta K (D K), a statistic based on the rate of change in the log probability of data with respect to the number of clusters, was used to interpret the real number of clusters (Evanno et al. 2005). After the optimal K was determined, a final parameter of 1 million MCMC replicates and a burn-in period of 100,000 was run for the assignment of individuals into K populations.

Multidimensional scaling using principal components
Principal Component Analyses (PCA) and their applications to control for population stratification were performed using the methods described by Patterson et al. (2006). This analysis involves three steps: (1) apply PCA to microsatellite data; (2) determine the significant principal components; and (3) identify the optimal number of clusters defined as the subpopulations after correcting for population structure. Each microsatellite allele was converted to a binary allele such that for each microsatellite locus, there were as many columns generated as there were alleles at that specific locus (Patterson et al. 2006). PCA was performed on normalized data to improve resolution of underlying population structure recovered using STRUC-TURE. Principal components were normalized to unit length and the number of significant axes of variation was determined using the Tracy-Widom statistic (Tracy and Widom 1994). The optimal number of distinct clusters was based on the Gap Statistic, which compares within cluster variance with that expected under a reference null distribution based on PCA (Tibshirani et al. 2001). The clustering was performed without any a priori knowledge of population units and was, therefore, an unbiased estimate of underlying population structure. The association of statistically significant clusters with geographic locality and host species was displayed graphically using scatter plots (3D) for consecutive combinations of three significant principal components, generated using the SCATTERPLOT3D (

Multilocus inference of migration
An MPI version of MIGRATE-N Version 3.4.4 (Beerli 2006) implemented in Mobyle SNAP Workbench was used to calculate the mutation-scaled population sizes (h) and the mutation-scaled immigration rates (M) between populations. M is defined as m/l, where m is the proportion of migrants in the population in each generation and l the mutation rate per generation per microsatellite locus. M therefore measures the importance of immigration versus mutation in introducing new variants into a population. The number of alleles per locus was used as a measure of mutation rate (l) for the microsatellite data. Theta (h), is the product of the inheritance scaler (=2 for haploids), the effective population size (N e ), and the mutation rate (l) per generation (h = 2N e l). MIGRATE uses a Bayesian approach based on the coalescent theory to estimate the population parameters. Microsatellite input data were converted to repeat numbers and a Brownian motion approximation to the stepwise mutation model was used. Starting values of h and M were generated from F ST assuming an inheritance scaler of 0.5 (for haploids). For the search strategy, the mutation rate was set to constant for all loci, the posterior distributions were generated using slice sampling and a static heating scheme with four chains of temperatures of 1,000,000, 3, 1.5, and 1 was used. Analyses were performed using one long chain, 100,000 steps recorded every 100 steps and a burn-in of 50,000. Three independent MIGRATE runs using different starting random seed numbers were performed on two datasets to ensure convergence of parameter estimates. The first dataset was used to compare migration between the six continents and the second between the 14 individual countries.

Mating type distribution
The mating types MAT1-1 and MAT1-2 were assayed using a set of primers developed by Groenewald et al. (2007), to amplify the mating-type idiomorphs of D. septosporum. Degenerate primers (Dot Mat1r 5 0 -TTGCCTGACCGGC TGCTGGTG-3 0 and Dot Mat2r 5 0 -CTGGTCGTGAAGTC CATCGTC-3 0 ) and species-specific primers (D. septo Mat2f 5 0 -GTGAGTGAACGCCGCACATGG-3 0 and D. septo Mat1f 5 0 -CGCAGTAAGTGATGCCCTGAC-3 0 ) were multiplexed in a single reaction. PCR reactions were carried out in a total volume of 12.5 lL and consisted of: 10-20 ng DNA, 19 PCR buffer containing 2 mmol/L MgCl 2 , 0.5 mmol/L MgCl 2 , 0.12 mmol/L of each dNTP, 200 nmol/L of each of the four primers and 0.032 U FastStart Taq DNA Polymerase (5 U/lL) (Roche Diagnostics). Cycling conditions consisted of an initial denaturation of 5 min at 94°C followed by 40 cycles of 94°C for 20 s, 65°C for 20 s, 72°C for 40 s and a final elongation of 7 min at 72°C. The presence of the MAT1-1 idiomorph was indicated by an amplicon size of 823 bp and these isolates were designated MAT 1. Presence of the MAT1-2 idiomorph was determined by the amplification of a fragment of 480 bp and these isolates were designated MAT 2. An exact binomial test, using two-tailed P-values (http://udel.edu/~mcdonald/ statexactbin.html) was used to determine whether populations deviated from the null hypothesis of a 1:1 ratio of mating types.

Haplotype identification
A total of 471 isolates were successfully recovered from the infected pine needles. All isolates screened with the internal diagnostic marker, Doth_A, produced an allele size of 124 bp, consistent with D. septosporum. Thirteen isolates from Hungary that produced an allele size of 109 bp were identified as Dothistroma pini (Barnes et al. 2011) and were excluded from further analyses (Table 1). Ultimately, 458 isolates were used in the study and after clone-correction 234 haplotypes were obtained across the populations (Table 2).

Genetic diversity
From 12 microsatellite loci, a total of 130 alleles were inferred ranging from two alleles at Locus_DCB2 to 24 alleles at Locus_L. The allele frequencies at each locus were recorded for all populations in Table S2 (Supporting information). Other indices of variation are reported in Table 2 for each population (country) and region (continent). Isolates from Austria had the highest percentage of unique alleles at 7.7% (based on a total of 130 in the entire data set) followed by the United States (6.2%). Only the Austrian and Hungarian populations showed 100% polymorphism for the 12 loci and contained the highest allelic richness (4.59 and 3.89, respectively). In the Southern Hemisphere, although two unique alleles were present in the Kenyan population, those from South Africa, Chile, Ecuador, and New Zealand contained no unique alleles (Table 2). Gene diversity per country ranged from a high value of 0.55 in Austria to a low value of 0.01 in New Zealand. Similarly, allelic richness ranged from 4.59 in Austria to 1.05 for New Zealand.
At the continental scale, the European population had the greatest number of alleles (109), percentage of unique alleles (36.2%), gene diversity (0.57) and allelic richness (4.79). The North American, Asian, and African populations contained moderate levels of diversity, with values of 6.2/0.25/2.33 and 1.5/0.33/2.97 and 1.5/0.38/2.47, for the percentage of unique alleles, gene diversity and allelic richness, respectively. The South American and Australasian populations had the lowest levels of diversity with Overall, the clonal fraction in the Northern Hemisphere populations was low (mean 17%), compared to those in the Southern Hemisphere (mean 65%) ( Table 3). The collection of isolates from the Czech Republic and Bhutan contained no duplicate haplotypes. The highest clonal fraction was observed for the New Zealand population at 91%. Genotypic diversities were higher for collections from Northern Hemisphere countries, ranging from 0.83 to 1 ( Table 2). The genotypic diversity for isolates from the Southern Hemisphere ranged from 0.09 (New Zealand) to 0.96 (Kenya). All but four multilocus haplotypes were unique to countries. The exceptions included those present in isolates from both Chile and Ecuador and one that was present in both New Zealand and Australian isolates.
AMOVA analyses showed that at a global scale, with the continents representing regions, most of the variation was distributed within populations (61%), but a significant proportion of the variation (32%) was also attributed to differences among regions (Table 4). Eight percent of the variation was partitioned among populations within regions. AMOVA analyses for individual regions showed that for both the South American and Australasian collections, a high percentage of the variation was within populations (99% and 96%, respectively), with no genetic differences observed among these populations (Table 4). A large proportion of the variation for the African collections was observed between Kenya and South Africa (33% ,  Table 4). Within Europe, although variation among populations was low (Φ-PT = 0.088, P = 0.001), it was still significant for population structure.
Population structure STRUCTURE analyses revealed the highest posterior probability for five clusters/populations (Fig. 1). Clusters consisted mainly of isolates from the same geographic areas of distribution. The Southern Hemisphere isolates were partitioned into two clusters. Cluster 1 comprised the African isolates from South Africa and Kenya while Cluster 2 comprised the rest of the Southern Hemisphere isolates from Chile, Ecuador, Australia, and New Zealand. For the Northern Hemisphere, three additional clusters were observed. The isolates from the United States mainly fell into Cluster 3 and those from Bhutan into Cluster 4. The European isolates contained mixed assignments, mainly into Cluster 5 (Romania, Hungary, Slovakia, Czech, Poland, and Austria), but with individuals from these countries also being assigned to Cluster 3 (mainly Poland) and a few into Cluster 4 (Romania, Hungary, Slovakia, Czech, and Austria).
The NJ tree based on Nei's genetic distance D A showed similar clustering of isolates into distinct clades based broadly on the country and continent from where they were collected (Fig. 2). Lineages have been coloured in accordance with the geographic clusters identified in STRUCTURE ( Fig. 1 and 5). Although STRUCTURE grouped the isolates from South America and Australasia into one cluster, distinct lineages can be seen here for the South American (yellow lineage) and Australasian (orange lineage) isolates (Fig. 2). In four cases, isolates from Ecuador and Chile share identical haplotypes (indicated as black dots) while isolates from Australia and New Zealand share one identical haplotype. Similarly, most of the isolates from South Africa cluster together, although geneti-cally distinct from Kenya (both indicated in pink). The isolates from the United States form their own unique, distinct lineage (blue), as did most of the isolates from Bhutan (indicated in red). The isolates from Europe, however, showed many divergent branches and formed several subgroups (indicated in green). Isolates from Austria, Romania, Hungary, the Czech Republic, and Poland  The null hypothesis of no population structure is rejected at P < 0.01.  were randomly distributed throughout the tree indicating a high level of migration.

Multidimensional scaling using principal components
Principal component and Tracy-Widom analyses revealed five significant axes of variation and five distinct clusters based on the gap statistic similar to the number of clusters inferred with STRUCTURE. Plots of the first three (PC1, 2, 3) and last three (PC3, 4, 5) significant axes are presented for cluster/geographic location ( Fig. 3A and B), cluster/Pinus species ( Fig. 3C and D), and a combination of geographic location/Pinus species (Fig. 3E and F). Inspection of these plots revealed P. radiata as harboring variation at the extremes of two axes of variation, PC3 and PC5, outlined using ovals in Fig. 3D. When only the isolates from P. radiata were examined (Fig. 4A), a total of eight clusters based on three significant axes of variation could be distinguished. The South African isolates fell into four of these distinct clusters as distinguished by PC3 (Fig. 4A). Approximately 20% of the variation separated the South African clusters from the New Zealand/Australian and Chile clusters. The latter two clusters were further separated by variation in PC2 and PC3 (Fig. 4A). When the first two significant axes were examined alone, then three distinct clusters were observed in the Southern Hemisphere; an African cluster with South African and Kenyan isolates (cluster 2), a South American cluster with isolates from Chile (cluster 1) and an Australasian cluster with isolates from New Zealand and Australia (cluster 0) (Fig. 4B).

Multilocus inference of migration
A total of 36 migration parameters were tested at a continental level (6 9 6 matrix for Africa, South America, Australasia, Europe, Asia, and North America) and 196 parameters at a global scale (14 countries) in three independent runs. While there was adequate chain mixing, independent runs did not converge in estimates of the magnitude of migration among countries, or continents, for any of the runs. While absolute values did not match, in many cases similar migration directionality was observed between runs. Results of the migration parameters for the comparisons between continents for a single MIGRATE run are shown in Fig. 5. Migration estimates indicated that the largest donors of migrants at a global scale came from Europe and Africa. Austria and Hungary were the largest contributors for Europe and South Africa emerged as the largest donor for Africa (Fig. 5, Table 5). Europe appeared to represent the largest sink population for migrants and these were especially from the Southern Hemisphere. In the Southern Hemisphere, the most significant number of migrants was observed going from Chile into Ecuador. The sink for donors in Europe and South America was identified as Australia while New Zealand received migrants from Australia. South Africa was the largest donor of migrants to Kenya.

Mating type distribution
Both mating types were identified in all of the Northern Hemisphere populations. Isolates from Poland and the United States deviated from a 1:1 ratio of random mating with the MAT1-2 idiomorph being significantly more common in Poland while the MAT1-1 was more common in the United States (Table 3). In the Southern Hemisphere, all the isolates from Chile, Ecuador, Australia, and New Zealand were MAT 2. Although both mating types were found in Africa, only the population from Kenya showed evidence of a randomly mating population.

Discussion
Clustering methods, based on multilocus microsatellite markers and applied to a broad collection of D. septosporum isolates, partitioned the isolates roughly corresponding to the geographical areas where they had been collected. These clusters also showed that the patterns of genetic diversity, population structure, and distribution of mating types in D. septosporum reflect anthropogenic activity relating to its distribution on non-native pines in the Southern Hemisphere. They also provide interesting insights into the establishment, reproduction and range expansion of D. septosporum on non-native and native trees in the Northern Hemisphere, particularly in Europe.
The European populations of D. septosporum were generally characterized by high levels of genetic diversity, high levels of migration between countries and the equal presence of both mating types. The collection of isolates from Austria, Hungary, Poland, the Czech Republic, Slovakia, and Romania constituted a single, large  Fig. 1 and 5). Lineages in green represent isolates from Europe (A: Austria, C: Czech, H: Hungary, P: Poland, R: Romania, S: Slovakia); blue from the United States; pink from Africa (Kenya and RSA:South Africa); red from Bhutan; yellow from South America (Ch: Chile, E: Ecuador) and those in orange represent isolates from Australasia (AL: Australia, NZ: New Zealand). Black dots indicate instances where a haplotype is shared between different countries as in the case of Chile and Ecuador and between Australia and New Zealand.  Species P radiata P muricata P nigra P ponderosa P contorta P sylvestris P mugo P cembra P peuce P wallichiana

Species
P radiata P muricata P nigra P ponderosa P contorta P sylvestris P mugo P cembra P peuce P wallichiana population, where most of the genetic variation (91%) is due to the large numbers of unique alleles present in the individual country collections. These populations, mainly on planted P. nigra, were genetically very similar, despite their occurrence in different countries. This can be explained by the fact that these countries border on each other and thus have no barriers to gene flow. Migration or range expansion of D. septosporum during the last 50 years in Europe, due to effective dispersal of asexual spores (conidia) and extensive movement of pine material (Tom sovsk y et al. 2012), has evidently resulted in the homogenization of allele frequencies in isolates from this region. This has apparently resulted in an even distribution of genetic structure of the pathogen in these countries. Such range expansion for an apparently native pathogen is not unusual and is well known for many other plant pathogens (Banke and McDonald 2005;Stukenbrock et al. 2006;Gladieux et al. 2008). The fact that both mating types were found in more or less equal proportions in the populations of D. septosporum from Europe, and the high levels of genetic diversity observed, suggests that recombination between individuals is occurring. This interpretation is further supported by the fact that in most cases, the populations consisted of unique haplotypes and by the fact that the sexual state of the fungus has been recorded in several European countries (Kowalski and Jankowiak 1998;Bradshaw 2004) including those considered in this study. It is apparent that regular sexual recombination as well as population growth and expansion over time has gradually dissipated most of the evidence of introduced populations in Europe, which would be revealed from genetic bottlenecks and multilocus linkage disequilibrium (Dlugosch and Parker 2008). The high genetic diversity of D. septosporum in Europe could be due to multiple introductions over time. However, based on the genetic evidence presented in this study, and others (Drenkhan et al. 2013;Kraj and Kowal-ski 2013), it is more likely that the pathogen is native to some areas of Europe although it has been, and still is, considered an alien invasive species and quarantine pest on this continent. The recent emergence of serious disease outbreaks and the observed range expansion of the pathogen might then be due to a combination of factors including climate change (Woods et al. 2005), increased availability of susceptible hosts arising from larger areas of planted forests and consequently, increased inoculum loads. A "spill-over" effect of the native pathogen onto newly available, susceptible hosts could then account for the high genetic diversity observed. A similar phenomenon has been observed in British Columbia (BC) where records of tree ring chronology suggest that D. septospo- Table 5. Population estimates of theta (h) and mutation-scaled immigration rates (M = m/l) between pairs of populations of Dothistroma  The directional migration routes between continents, as determined by MIGRATE, are shown with arrows. M represents the mutation-scaled effective immigration rate between continents while the number of immigrants per generation (hM) is indicated in parenthesis. Although Europe is the biggest donor of immigrants on a global scale, it also acts as the largest sink population. On a country level, the largest evidence of directional migration was from Chile into Ecuador (see Table 5).
rum has been present in this Canadian province since the 1800's (Welsh et al. 2009). The recent severe epidemics in BC appear to be linked to an enormous increase of planting P. contorta var. latifolia in monocultures on sites supporting naturally mixed species as well as an increase in precipitation and temperature during the last decades, greatly favoring infections (Woods et al. 2005). The successful establishment and population build-up of D. septosporum in BC has evidently been facilitated by the high genetic diversity of the pathogen in native and nonnative P. contorta var. latifolia stands (Dale et al. 2011) and the presence of both mating types (Groenewald et al. 2007). Patterns of genetic diversity and the clustering of D. septosporum isolates from the Southern Hemisphere countries revealed the presence of three distinct evolutionary lineages that reflect the anthropogenic movement of pines into Africa, South America, and Australasia (Table S1). In the African cluster, the high migration rates from Europe, the high levels of genetic variability in the Kenyan and South African populations, and the presence of both mating types suggests that D. septosporum has most likely been introduced into these areas more than once. This is consistent with the long history of pine being moved into these countries (Poynton 1977;Richardson et al. 2007) that would have provided many pathways and opportunities for introduction of the pathogen (Lundquist 1987;Diekmann et al. 2002). A similar concordance between levels of genetic diversity and numerous putative introduction events has been found for the pine shoot and canker pathogen Diplodia sapinea on P. radiata in the Southern Hemisphere (Burgess et al. 2001), Plasmopara viticola populations in Europe (Gobbin et al. 2006) and the European race of Gremmeniella albietina var. abietina in North America (Hamelin et al. 1998).
Founder effects were evident in D. septosporum populations from Ecuador, Chile, New Zealand and Australia. The low levels of diversity and clonal structure that were observed in isolates from these countries suggest that these populations have lost alleles, which is consistent with their anthropogenic movement to new environments. This is common for introduced pathogens, for example, the pitch canker pathogen Fusarium circinatum (Wikler and Gordon 2000) and the chestnut blight fungus C. parasitica in south-eastern Europe (Milgroom et al. 2008). Clonality and shifts in the reproductive modes of populations are characteristic features for pathogens that have been moved from their areas of origin (Taylor et al. 1999). Both of these characteristics were apparent in the populations of D. septosporum from South America and Australasia in this study. In both regions, only the MAT1-2 idiomorph was observed, which confirmed, using a larger sample size, the results of Groenewald et al. (2007).
Pathogen populations that have passed through bottlenecks resulting in lower levels of genetic diversity are thought to be at a disadvantage because of lost alleles that could potentially confer adaptive abilities to survive in their new environments (Sakai et al. 2001;Allendorf and Lundquist 2003). However, the low levels of genetic diversity and the presence of a small number of clonal lineages in the South American and Australasian clusters detected in this study showed that this has clearly not been a limiting factor for the successful establishment, spread and the occurrence of devastating D. septosporum epidemics in these Southern Hemisphere regions. In these cases, a small number of virulent genotypes have most likely thrived on the highly susceptible P. radiata under environmental conditions conducive for infection and spread exclusively via asexually produced spores. This would be similar to other introduced plant pathogens including, for example, Phytophothora infestans (Goodwin et al. 1994), D. sapinea (Burgess et al. 2001), and Ceratocystis platani (Engelbrecht et al. 2004). Direct movement of D. septosporum between Chile and Ecuador was strongly supported by the genetic evidence in this study. Both countries shared common genotypes, all the alleles were identical and there was a lack of population structure (1% among-population variation) between these countries. DNB has been known in Chile since 1957 (Gibson 1972), but was reported from Ecuador only in 1983 (Evans and Oleas 1983). Given the genetic similarity of the D. septosporum populations in these countries and the strong evidence of directional migration observed using MIGRATE, it is likely that the pathogen was accidentally introduced from Chile into Ecuador. This is an excellent example of human-mediated movement because Chile is known to have shared P. radiata germplasm with Ecuador (F. Montenegro, pers. communication).
Results of this study clearly show that the D. septosporum population in New Zealand is clonal. This is consistent with previous investigations (Hirst et al. 1999;Groenewald et al. 2007) in which only one haplotype and a single mating type was found in samples covering a relatively large area. In the study by Hirst et al. (1999), collections from the 1960's were identical to those made 30 years later. It is evident that the same clone continues to persist almost 50 years later. New Zealand is well known to have maintained very strict quarantine regulations for many years (http://www.maf.govt.nz/biosecurity-animal-welfare) and evidently, additional introductions of D. septosporum have not occurred.
It has been speculated that D. septosporum moved from New Zealand to Australia naturally across the Tasman Sea (Edwards and Walker 1978;Matheson 1985). MIGRATE analyses, however, suggest that the direction of migrants was most likely from Australia into New Zealand. The single haplotype present in New Zealand was also present in two locations sampled in Australia. However, the Australian samples included additional genotypes and unique alleles not found in New Zealand. This suggests that, despite rigorous quarantine regulations, there have been introductions from other sources into Australia, most likely from Europe and Africa. These would most likely have been human-mediated, due to the geographically isolated posi-tion of the continent, and they would most probably have occurred through the importation of pine germplasm.
The number of isolates from Kenya available for this study was limited. However, there were no shared genotypes between isolates from New Zealand and Kenya and only a small proportion (22%) of the alleles were common to the fungal populations from these areas. It was thus not possible to validate the hypothesis of Hirst et al. (1999) that D. septosporum in New Zealand or Australia originated in Kenya. Yet there is some evidence of low levels of migration from the African continent into Australasia.
The worldwide distribution of D. septosporum is an important example of human mediated, but unintentional movement of a plant pathogen. This movement has facilitated the establishment of D. septosporum in many areas of the world where pines are planted outside their natural ranges. Although available cultures from countries considered in this study varied in number, intriguing patterns of local range expansion and global spread emerged. There remains substantial opportunity to expand the knowledge gained in this study with larger numbers of isolates from areas that could not be sampled or where sampling was not optimal. In particular, it would be most interesting to consider the population genetics of isolates from North and Central America, which could provide intriguing insights into the possible area of origin of D. septosporum.
In the future, optimal climatic conditions, influenced dramatically by climate change, and the planting of susceptible trees in monocultures are likely to increase the potential for D. septosporum to infect new hosts, and to become established in new areas (Woods et al. 2005;Watt et al. 2009). This is apparently already occurring in areas such as British Columbia, Estonia, Finland, and Sweden (Woods et al. 2005;Brown and Webber 2008;Barnes et al. 2011;Drenkhan et al. 2013).
Dothistroma septosporum provides an outstanding model organism to study biological invasions because it has become established both in the Northern and Southern Hemisphere, where it can be viewed alternatively as a native pathogen or an alien invasive, influenced by planting practices, other human activities and climate change. Whereas populations of the pathogen in the Northern Hemisphere are genetically heterogenous and sexually recombining, those in the Southern Hemisphere, with the exception of Africa, have low levels of diversity and contain a single mating type gene. It will thus be important to monitor and prevent the further spread of different, possibly more virulent genotypes and opposite mating type genes, into these predominantly monoculture plantations of susceptible hosts in the Southern Hemisphere.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Table S1. Dates for when pines (and more specifically, P. radiata) were introduced into some Southern Hemisphere countries. Dates are also recorded for when the susceptible species P. radiata was extensively grown in plantations and when the first reports and epidemics of Dothistroma needle blight (DNB) occurred. Table S2. Allele frequencies of Dothistroma septosporum isolates from 14 different countries in six regions based on 12 microsatellite markers.