Introduction history and population genetics of intracontinental scotch broom (Cytisus scoparius) invasion

Biological invasions at the intracontinental scale are poorly studied, and intracontinental invasions often remain cryptic. Here, we investigate the recent range expansion of scotch broom (Cytisus scoparius) into Norway and clarify whether the genetic patterns indicate natural spread or human introduction. Furthermore, we investigate whether plants were moved within the native range and how this influences invasion success. We also infer the level and structuring of genetic diversity within and between the putative native and introduced range.


| INTRODUC TI ON
Despite more than 50 years of research on the biological invasion of alien species across continents, studies at the intracontinental scale are rare, because it is difficult to prove whether a species is native or alien (Hulme et al., 2016;Webber & Scott, 2012). Such cases are termed cryptogenic species (Carlton, 1996) or interspecific cryptic invasions. Even when available records indicate a recent colonisation of a species, it should only be classified as alien if it was directly introduced by humans and did not expand its range by natural dispersal from its native range (Gilroy, Avery, & Lockwood, 2017). Throughout this publication, we use the term 'alien' for a species or genotype that occurs at a location as a direct result of human introduction (Hulme et al., 2016;Webber & Scott, 2012) and 'invasive' for an alien species (or genotype) that spreads beyond the location where it was introduced (Cristescu, 2015;Richardson et al., 2000;Wilson, Dormontt, Prentis, Lowe, & Richardson, 2009). When alien genotypes are introduced to a distant location but are within the species' native range, the invasion may be undetected or cryptic (Morais & Reichard, 2018;Saltonstall, 2002). Alien genotypes compete against native genotypes, and if they hybridise, they may genetically contaminate or swamp the native populations, which can lead to the loss of native gene pools and may cause phenotypic changes of the species (Morais & Reichard, 2018;Nielsen, Brandes, Kjaer, & Fjellheim, 2016). The admixture of geographically distant genotypes within the native range can also create a source for introductions into the novel range, which will become invasive (Ellstrand & Schierenbeck, 2000).
However, while it is generally accepted that the source origin of an introduced organism affects its ability to persist and invade (Henery et al., 2010), the possibility of genetic admixtures of different genotypes prior to their introduction has rarely been considered (Le Roux, Richardson, Wilson, & Ndlovu, 2013).
The introduction history of an alien species is crucial for its establishment and invasiveness (Colautti, Grigorovich, & MacIsaac, 2006;Estoup & Guillemaud, 2010). Of special importance are the quantitative aspects of propagule pressure (Lockwood, Cassey, & Blackburn, 2005;Simberloff, 2009) and the qualitative aspects of propagule origin. Propagule pressure quantifies the total number of introduced propagules, which is affected by the number of introduction events (propagule number) and by the number of individuals introduced in each introduction event (propagule size; Blackburn, Lockwood, & Cassey, 2015;Wittmann, Metzler, Gabriel, & Jeschke, 2014). High propagule pressure not only increases the probability of naturalisation by compensating for stochastic mortalities and increasing the chance of being introduced into favourable habitats, it also shapes the genetic structure of introduced populations (Lockwood et al., 2005). High propagule pressure has been observed in the majority of successful invasions at the intercontinental scale (Blackburn et al., 2015). As the likelihood of multiple introduction events increases over time, propagule pressure may explain the commonly observed lag phases of alien species (Bock et al., 2015;Simberloff, 2009). In addition, once an alien species has become established and reached maturity, its seeds may serve as an additional source of propagules, causing secondary spread in the invaded range. In addition to the number of introduction events, the origins of those introductions affect the invasive potential by increasing genetic diversity and the possibility of preadaptation to the novel environment (Buckley & Catford, 2016). Further, the genetic admixtures resulting from the crossing of plants from geographically distant origins produces novel genotypes that may have stronger invasive potential (Bock et al., 2015;Ellstrand & Schierenbeck, 2000;Morais & Reichard, 2018) and increased ability to adapt to the new environment. Propagules from different origins may have reached the new range by direct introduction, or they may have been moved to another location first where they intermixed with others before moving further into the new range (bridgehead invasion; van Boheemen et al., 2017;Lombaert et al., 2011). Thus, even a limited number of introduction events (of large propagule size) can result in diverse propagule origins (i.e. where they originally evolved) when propagules have already mixed in the source population of the introduction event.
The shrub scotch broom (Cytisus scoparius (L.) Link) is native across central Europe and the British Isles but has been classified as an invasive species elsewhere (e.g. in the United States, Chile, Australia, New Zealand (Kang, Buckley, & Lowe, 2007), Brazil (Cordero, Torchelsen, Overbeck, & Anand, 2016), South Africa (Mkhize, Mhlambi, & Nanni, 2013) and India (Srinivasan, Shenoy, & Gleeson, 2007). Since its first record in southern Norway in 1876 (Blytt, 1876), it has spread northwards ( Figure 1, see Appendix S1) and has also increased in abundance in threatened habitats, such as endangered coastal heathlands (Lindgaard & Henriksen, 2011). This species was included on the Norwegian Black List in 2007 but was removed in the 2012 revised list, considering that it might be native (Gederaas, Salvesen, & Viken, 2007;Gederaas, Moen, Skjelseth, & Lar sen, 2012). The ongoing range expansion might simply reflect a lag in dispersal to northern habitats that became suitable after the last glacial retreat. This could be due to the species' short distance seed dispersal (Bossard, 1991;Malo, 2004) and geographical barriers (as is proposed for European trees, e.g. Abies alba and Larix decidua in Svenning & Skov 2004). Alternatively, its recent range expansion can be a result of human introduction. In fact, scotch broom has been planted for soil improvement and stabilisation (Vesthassel, 1926) and as a garden ornamental, and the extended viability of its seeds (Magda, Gleizes, & Jarry, 2013) allows for accidental spread in soil (ballast soil of ships in the past, infrastructure K E Y W O R D S cryptic invasion, cryptogenic species, Cytisus scoparius, intracontinental plant invasion, introduction history, population genetics, propagule pressure, range expansion, scotch broom construction). However, natural expansion and introduction by humans are not mutually exclusive. Both native and introduced populations of scotch broom have been identified in Denmark (Nielsen et al., 2016;Rosenmeier, Kjaer, & Nielsen, 2013). With high certainty, a native gene pool exists in Denmark, where it was observed as early as 1648 (Paulli, 1648), and in 1958 this distinct phenotype was described (dwarfed growth form and increased cold hardiness; Böcher & Larsen, 1958; Rosenmeier et al., 2013). More recent studies have revealed that fast spreading, invasive plants in the Danish landscape are likely to constitute an introduced gene pool (Nielsen et al., 2016;Rosenmeier et al., 2013).
This study investigates the range expansion of the intracontinental invader scotch broom in its native distribution range in Europe, with a focus on its northernmost expansion front. More explicitly, we aim to (a) establish whether this range expansion is natural or is caused by human introductions, (b) determine whether the expansion is the result of one or multiple introductions (propagule pressure) and pinpoint the origin of these introductions (propagule origin) to establish the role of propagule pressure and origin in invasion success and (c) quantify the level and geographical structure of genetic diversity to investigate its possible role in invasion success.

| Sampling
We collected leaf samples from a total of 172 populations of scotch broom. We sampled both fresh leaves and leaves from herbarium specimens (see Appendix S2).  Rosenmeier et al. (2013). In addition, three individuals of scotch broom were collected in garden centres in eastern Norway (see Appendix 2).
Leaf samples of herbarium specimens were collected from 11 European herbaria (see Acknowledgements). For the herbarium specimens, a single individual represented the population from which the herbarium specimen originated. From the Norwegian herbarium samples, 27 were selected focusing on the four areas representing the species' oldest observations (Figures 1 and 4). They also covered the time span of the Norwegian collections (from 1869 to 1998). For the selection of 81 specimens from other European countries, the aim was to cover an equal number for each country of the oldest (1835-1890) and the most recently  collected specimens (see Appendix S2). As most specimens had no GPS coordinate record, they were estimated based on the specimen location description.

| DNA extraction
We extracted DNA from up to 10 mg dry tissue, or approximately 50 mg fresh tissue, following the DNeasy Mini Plant Kit (Qiagen) for most fresh samples. To obtain a higher yield, we prolonged the chemical cell lysis step (in buffer AP1, at 65°C) from 10 to 30 min.
For fresh samples, we implemented the final DNA elution only once, and for herbarium species the first eluate was passed through the filter a second time.

| Chloroplast sequencing
For the chloroplast genetic analysis, most populations were represented by one individual, but for 17 populations from Norway and the two populations from Denmark we sequenced five to seven individuals to test whether these populations originated from a single or from multiple maternal lines (see Appendix S2
First, the fresh samples and the herbarium samples were aligned separately. Unique polymorphisms (present in a single sample only) were verified by resequencing or were otherwise removed.
Mononucleotide repeats were excluded from the analysis. To allow the joint analysis of fresh samples and herbarium samples, all sequences were shortened to the herbarium length. Gaps were coded by simple index coding (SIC; Simmons & Ochoterena, 2000). In two cases of complex indels, they were each coded as two separate indels. ac.nz) and the median joining algorithm (Bandelt, Forster, & Rohl, 1999).
The spatial distribution of the haplotypes was visualised on maps using qGiS 2.14.20 (QGIS Development Team, 2018). To assess the relationship between the genetic distance and the geographical distance of all haplotype samples of scotch broom, we used a Mantel test with 999 permutations in Genalex 6.5 (Peakall & Smouse, 2006 We also assessed changes in the abundance of the haplotypes over time by plotting the cumulative count for each haplotype of all herbarium samples in r 3.5.0 (R Core Team, 2018).

| Nuclear SNPs
The fresh samples were analysed for single-nucleotide polymorphic (SNP) markers (see Appendix S2). This included up to 26 samples for each of the following populations: 20 populations from southern Norway, five from western Norway, two from Denmark, one from France and one from Germany. The seedlings grown from eight seed bank accessions were also included. However, for some of the seed bank populations, only a few individuals were available (see Appendix S2).

| Nuclear SNP marker development and genotyping
For the nuclear genetic analysis, SNP markers had been devel-

| Nuclear SNP data analysis
Genalex 6.5 (Peakall & Smouse, 2006  10,000 burn-in and MCMC repeats. This was run for K = 2 to K = 10 groups, repeated five times each, and the K with the strongest support was calculated in cluMPaK (Kopelman, Mayzel, Jakobsson, Rosenberg, & Mayrose, 2015) based on the Evanno methodology (Evanno, Regnaut, & Goudet, 2005). We also conducted Mantel tests for the nuclear markers with 999 permutations in Genalex.  and Austria; however, Austria was represented by only two samples.

| Spatial and temporal patterns of chloroplast haplotype diversity
The same general results were obtained after applying rarefaction to a sample size of five (Table 1). Nucleotide diversity (π) of the haplotypes grouped by country ranged from 0.00079 to 0.00739 and showed similar patterns as Hd, with a few exceptions ( Table 1).
The exact test resulted in some significant differentiation of haplotype frequencies between countries, most notably between the southwestern (France, Spain and Portugal) and the northeastern countries (see Appendix S5). Furthermore, Norway showed significant differences in haplotype frequencies from seven countries, but this was likely an artefact of the large sampling effort in Norway.
The haplotype network revealed two groups of haplotypes that were separated by six mutations (Figure 3) for which several individuals were analysed, three populations consisted of two haplotypes, and one had three haplotypes ( Figure 5).

| Genetic diversity and distribution of nuclear SNP markers
Most Norwegian populations had highly polymorphic SNP markers (see Appendix S9), ranging from 72% to 97%, except for population 3 (61%) and population 14 (67% The STRUCTURE analysis showed little genetic structure among the populations (Figure 6). Using Evanno's approach (Evanno et al., 2005), the best support was found for four clusters (see Appendix S8). Most noticeable were the distinct ancestry estimates for the populations 3, DN and 8Wa, which contained the chloroplast haplotype C (Figure 6, see Appendix S7). The samples of the remaining haplotypes were highly admixed, and no geographical patterns emerged.

| Northward expansion of scotch broom was driven by human introductions
Our study suggests that the recent range expansion of scotch broom was mainly driven by human introductions; therefore, the species is alien to Norway, and its cryptogenic status is resolved. In Norway, around the year 1900, at least three different haplotypes were  Figure 5). The northward expansion during the last century has been extensive ( Figure 1) and far exceeds the natural dispersal ability of up to 10 m every three years (Bossard, 1991;Malo, 2004).
The eight different haplotypes in this northern expansion range, combined with a lack of temporal and spatial patterns ( Figure 5, see  where the latter is a common glacial refugium. Another Norwegian haplotype (D) was not found in neighbouring countries but was present in most other countries ( Figure 2).
Nevertheless, we cannot completely exclude the possibility that some of the earliest records in Norway arose from a natural northwards expansion. Previous publications show that natural recolonisation reached as far north as Denmark by 1648 (Paulli, 1648) and maintained a specific phenotype there (Böcher & Larsen, 1958;Nielsen et al., 2016). These native Danish populations belong to chloroplast haplotype C. This haplotype is also present in one remote location in Norway, where it has been present since at least 1896 (see Appendix S2, Figure 5; Agder Museum of Natural History and Botanical Garden; GBIF.org, 2017). This population has not spread beyond its initial location and thus does not contribute to the recent invasion of scotch broom across Norway. This is in line with observations in Denmark, where the low-growing, putative native type is apparently not involved in the spread of the species across the Danish landscape.
Our findings suggest likely introduction routes as stowaways on ships in the past and as escapees from gardens and arboriculture (classification by Hulme et al., 2008). The three locations of the oldest introduction records, which also have the highest haplotype diversity, include an international trading port, a prestigious British-inspired ornamental garden, and a nursery that propagated scotch broom for soil stabilisation (see Kristiansand, Mandal and Kjørrefjord in Figure 5; Lundberg & Rydgren, 1994;Vesthassel, 1926). Another location with an old record contained only a single haplotype (see Telebukta in Figure 5), but this can be explained by the remote and isolated location, which is less frequented by humans. The samples from the Norwegian garden centres revealed the first and third most common chloroplast haplotypes observed in Norway and northern Europe (haplotypes A and C in Figure 2).
The one haplotype (A) that increased its relative abundance the most around the years 1960-1980 (Appendix S6) was also the most common haplotype among the garden cultivars ( Figure 2). This haplotype may have been preferred for propagation and trade by the horticulture industry. In fact, many invasive plants have been introduced as garden ornamentals (Milbau & Stout, 2008). Given that the three most common haplotypes are genetically quite similar and, at the same time, distant from the group of haplotypes that dominate the Iberian Peninsula (Figure 3), this indicates that geographical regions other than the southwestern part of Europe have been sourced for plant material to be used in the plant industry. For instance, is it known that Italian genotypes have been extensively used for soil improvement in Denmark (Rosenmeier et al., 2013).

| In the native range: Cryptic invasion by alien genotypes
In an attempt to determine whether human-caused spread occurred only in the species' recent expansion range or also within the spe-  -1910) showed no clear spatial pattern of chloroplast haplotypes and did not differ from present-day distribution (Figure 4). This indicates that human-caused dispersal was already occurring before the early 19th century.
We did, however, also observe some remnants of haplotype distribution patterns that we would have expected to see with the natural postglacial recolonization of scotch broom in its native European range. Most of the private haplotypes were found in former glacial refugial areas (Figures 2 and 3), especially in Spain and Portugal (seven out of 11). The haplotype diversity and genetic diversity calculated for each country also decreased from south-west to north-east (Table 1, Figure 2). This agrees with other studies observing genetic loss during the natural recolonization of Europe from southern refugia after the last glacial maximum (Petit et al., 2003).
It must be noted that old herbarium material may be prone to degradation and post-mortem DNA modifications (Staats et al., 2011).
We cannot completely exclude that this phenomenon occurred in some of the older herbarium specimens we used. However, only two of the private haplotypes from herbarium material were more than 50 years old, and the rate of post-mortem DNA damage modification was found to be very low (Staats et al., 2011

| Diverse introduction history and high propagule pressure
We suggest that frequent and diverse introductions were crucial for the invasiveness of scotch broom in its northern expansion front. In Norway, scotch broom became established approximately 150 years ago, but only in the last decades was it recognised as invasive (i.e. spreading and increasing in density; Figure 1, see Appendix S1). This is in line with Aikio, Duncan, and Hulme (2010) who calculated a relatively long lag phase of 97 years from introduction to rapid expansion of scotch broom in New Zealand.
Our results support that introductions have occurred repeatedly from a range of origins. This high propagule pressure (quantity of introductions) likely contributes to its invasion in the novel range (Blackburn et al., 2015;Colautti et al., 2006), similar to what has been suggested for many other invasive species (Jogesh et al., 2015;Kelager et al., 2013). Further supporting the high propagule pressure is the fact that the species is widely used as a garden ornamental and is commonly sold in garden centres in Norway.
The quality of the propagules (i.e. place of origin) also influences a plant's ability to establish itself and become invasive, as a plant should be suitable to the novel habitat or have a high genetic potential to adapt. We could not locate the geographical origins of the haplotypes because of the lack of spatial patterns across the native range (Figure 4). The possibility that haplotypes from different origins are already mixed across the native range (i.e. intraspecific cryptic invasion) increases the likelihood of introducing haplotypes of different origins, even from a small source range. A mixed ancestry of introduced populations has also been shown for Spartina alterniflora during intentional introduction China (Bernik, Li, & Blum, 2016).

| A uniform invasion front with high genetic diversity
We observed strong genetic admixture of the nuclear markers  Nearly all of the genetic variation was found within the populations, and much less was found among populations (Table 2), a pattern that has been observed in other invasive alien species (Ellstrand & Schierenbeck, 2000;Kelager et al., 2013). Multiple introduction events from different origins can also facilitate biological invasions when these diverse sources hybridise (Ellstrand & Schierenbeck, 2000;Rius & Darling, 2014), resulting in increased genetic diversity at the population and individual level (Colautti & Lau, 2015;Dlugosch & Parker, 2008;Fennell, Gallagher, Vintro, & Osborne, 2014) and possibly creating novel invasive genotypes (Bock et al., 2015).
There is also the possibility of gene flow between a genotype that is adapted to northern climates (i.e. native Danish populations) and introduced genotypes. In Demark, putative native and introduced genotypes readily cross, causing the genetically polluted native plants to change in phenotype (Nielsen et al., 2016). In our study, however, three populations that shared the haplotype of the Danish native population (3, DN, 8Wa) showed a distinct genetic structure and remarkably low nuclear genetic diversity.
It has been predicted that a low capacity to tolerate the strong winter conditions in northern temperate areas restricts the ability of invasive species to invade northern temperate areas (L. Rosef & E. Heegaard personal communication) . Surprisingly, the identified haplotype of the putative native Danish populations, which is adapted to a northern climate (Böcher & Larsen, 1958), appears to play a minor role in the northern expansion process.

| CON CLUS ION
This study represents one of the very few examples of intracontinental plant invasion, indicating the complex aspects of the introduction history. The rapid northern range expansion of scotch broom in the recent century was driven by human introductions, thus scotch broom can be classified as alien in Norway. Our results indicate the importance of a diverse introduction history for invasion success, with high propagule pressure and genetic admixture resulting in high genetic diversity. Further, we point out the value of extending the genetic analyses of biological invasions to a species' native range. We found indications that the different genotypes of scotch broom have been moved and combined within the native range already. This is likely to lead to intraspecific admixture and high genetic diversity in the source populations, and it might affect the establishment and invasion at the onset of introduction into a novel range.

ACK N OWLED G EM ENTS
The following herbaria and seed banks generously sent specimens