AFLP-Based Genetic Structure of Lithuanian Populations of Small Balsam ( Impatiens parviﬂora DC.) in Relation to Habitat Characteristics

: Currently, there is an increasing focus on understanding the interactions between genetic features of the invader and environmental factors that ensure the success of invasion. The objective of our study was to evaluate the genetic diversity of Lithuanian populations of highly invasive small balsam ( Impatiens parviﬂora ) by ampliﬁed fragment length polymorphism (AFLP) markers and to relate molecular data to biotope features deﬁned by employing neighboring species of herbaceous plants. Low polymorphism of I. parviﬂora populations was observed at AFLP loci. Hierarchical analysis of molecular variance did not reveal differentiation of populations depending on biotope, geography, or road types. Bayesian analyses of AFLP data demonstrated many genetic clusters. Our results suggest multiple introductions of I. parviﬂora into Lithuania. The polymorphism of AFLP loci of populations signiﬁcantly correlated with the total coverage by herbaceous plants in the sites. Deﬁned by principal component analysis, the variability of study sites was most related to the coverage of herbaceous plants and least related to the molecular features of I. parviﬂora populations. The sites with I. parviﬂora were classiﬁed into agricultural scrubland, riparian forest, and urban forest biotopes. Of them, urban forest was distinguished by the highest coverage of I. parviﬂora and the lowest Ellenberg indicatory values for light, soil acidity, and richness in nutrients.


Introduction
Large areas of forests have been destroyed in favor of highways, urban territories, and agricultural areas. This has provided an increased commodities, food, and goods but has caused an irreversible loss of native habitats and decrease in species number [1]. In many cases, human interventions into forests have been accompanied by species invasions, which certainly hamper the effort to protect and restore nature [2]. Alien plant invasion affects the diversity and stability of the local community, and as a result, the functions of ecosystems and the services they provide are changing [3]. In Europe, 5789 alien plant taxa have been reported [4]. The rate of release of invasive alien species has increased in recent years. The European Union (EU) Biodiversity Strategy for 2030 [5] has set the challenge of preventing the introduction of new alien species into the EU territory and of controlling invasive aliens that have already become widespread. Similar guidelines have been provided by the EU The amplified fragment length polymorphism method has been employed in an assessment of broad range of invasive taxa, including Rubus alceifolius [44], Heracleum sosnowskyi [45], Lythrum salicaria [46,47], Phalaris arundinacea [48], Veronica hederifolia [49], Reynoutria japonica [50], Mikania micrantha [51], and Solidago canadensis [52]. These markers have been applied for some ornamental Impatiens species [53,54], including also examination of natural populations of I. noli-tangere [55] and I. capensis [56] and natural and exotic populations of I. capensis [57]. To the best of our knowledge, genetic diversity of I. parviflora, has been only evaluated by AFLP markers in one study of two Polish populations [58].
Numerous studies have focused on I. parviflora interactions with neighboring herbaceous species [15,16,19,21,22,31,[59][60][61]. Phytocenological data were mostly associated with either environmental (edaphic and climatic) properties of the sites with I. parviflora or with geographical, morphological, and physiological properties of I. parviflora populations, but the relationship with molecular features has been poorly investigated. As an extension of our pilot assessment of RAPD and ISSR loci-based diversity of I. parviflora populations [35,38], there was need to confirm and extend our findings by more advanced and precise method. Hereby, the present study is aimed at evaluation of Lithuanian populations of I. parviflora using AFLP markers and linking molecular parameters to the features of the biotic and abiotic environment.

Study Area
Small balsam (Impatiens parviflora DC.) individuals were sampled from 21 Lithuanian populations in south-eastern, central, and north-western regions ( Figure 1). . Information about the geography and climate of the sampling sites was provided in the previous study [35] and the other biotope features were thoroughly described previously as well [38]. In total, over 100 individuals of I. parviflora were sampled at the sites.

Molecular Analysis
Plant leaf DNA was isolated from 105 individuals (5 individuals in each of 21 populations) by DNA purification kit #KO512 (Thermo Fisher Scientific, Vilnius, Lithuania), as documented earlier [62].
Fragment analysis was performed on an ABI Prism 3130xl Genetic Sequencer (Applied Biosystems, Foster, CA, USA) using GeneScan 500 LIZ size standard. Manual binning and automatic scoring were performed in the range of 50-500 bp using GeneMapper version 3.7 (Applied Biosystems, Foster, CA, USA). Ten replicate samples were used for analysis and only highly reproducible loci were kept.

Abiotic and Biotic Features of the Sites
To link molecular and environmental data of populations, several features of the sampling sites were recorded. Sampling sites were attributed to three types of biotopes: agrarian scrubland, urban forest, and riparian forest [38]. Due to climatic differences, sampling sites were subdivided into three parts: south-eastern, central, and north-western Lithuania. In addition, traffic intensity/road vicinity was considered, distinguishing location along a blacktop road with intensive traffic, along a blacktop road of the city with low intensity traffic, and along a road without blacktop in forest [38]. Respectively, for hierarchical analyses of molecular variance (AMOVA), populations were grouped according to each parameter (biotope, geography, and road) into three clusters.
At each site, herbaceous plant species growing along with I. parviflora populations were assessed in 100 m 2 plots. Species names were given according to World Flora Online. Evaluations were done using two sets of the species number at sites: one set (named as NSp-138) included all herbaceous species (in total, 138 species were registered), the second set (named as NSp-76) excluded species that were found at single sites only (76 species). The coverage of each species at sites was calculated using Braun Blanquet methodology [64]. Total coverage by herbaceous plant species (covT-138 and covT-76), coverage by herbaceous plant species without I. parviflora (covHerb-138 and covHerb-76), and coverage by I. parviflora (covIP) were estimated. For each site, abiotic characteristics (light, temperature, climate continentality (hereinafter referred to as continentality), soil moisture, soil pH, and soil nutrients) were quantified by Ellenberg indicatory plant values (EIV) [13] averaged in proportions to each species coverage (weighted average method-WEIV) [8,65] or averaged without coverage data (unweighted average method-UEIV) [15].

Data Analysis
Standard parameters of genetic diversity, principal coordinate analyses (PCoA) for populations and individuals, and hierarchical AMOVA were obtained using GenAlEx v. 6.5. The unweighted pair group method with arithmetic mean (UPGMA)-based dendrogram was prepared by PopGene v. 1.32, similarly to that done for the data obtained by the other dominant multi-locus markers [35,62].
For identification of genetic clustering of I. parviflora populations, Bayesian analysis was performed with Structure v.2.3.1 [66], for significance of clustering patterns using Evanno et al. [67] methodology. The a priori number of clusters selected was set to K = 1-21, the maximum expected number of clusters corresponded to the number of analyzed populations. A total of 20 runs were carried out for each K and the rate of change in the log probability of the data between successive likelihood values (∆ K) was estimated using a 10 5 steps burn-in period followed by 10 6 iterations of Markov chain Monte Carlo.
Site abiotic characteristics were quantified by EIV of herbaceous species, using program STATISTICA v. 7.0. Plant species variation and the number of species at sites were evaluated by two-way cluster analysis, program PC-ORD v. 6.0 [68].

Analysis of Genetic Diversity
DNA fragments generated by AFLP primers ranged from 51 to 496 bp ( Table 1). The lowest number of AFLP fragments (13) was generated by using EcoRI-AGG-PET/MseI-CAC primer pairs and the highest number (40)-by EcoRI-AAC-FAM/MseI-CTG, the mean value per marker being 28. A total of 223 fragments of DNA were registered using 8 AFLP primer pairs, 167 of them were polymorphic (74.9%). Percentage of polymorphic loci (PLP) per population ranged from 11.2 to 34.1 (mean value, 20.1), with the lowest values for the population Sve in the south-eastern region and the highest values for the population KAS in the central region ( Table 2). The highest percentage of polymorphic loci was characteristic to I. parviflora growing in the central part of Lithuania (24.0%), slightly lower polymorphism was observed in south-eastern populations (21.1%), and the lowest polymorphism was observed in north-western populations (14.9%). The mean values of Nei's gene diversity and Shannon's index per populations were 0.092 and 0.130, respectively. Minimum and maximum values of these parameters were characteristic for the same populations as in the case of PLP. Among the populations, Nei's gene diversity ranged from 0.051 to 0.162 and Shannon's index ranged from 0.072 to 0.227, the most extreme values of these parameters differed 3.2 times.  (Nei, 1978), I-Shannon's information index.
Nei's [69] genetic distances (GD) at AFLP loci ranged from 0.021 to 0.124 (data are not provided in the tables). The most genetically distant pairs of the populations were in geographically neighboring areas Pre and Jon (GD = 0.124), Pla and Jon (GD = 0.122), or Vve and Jon (GD = 0.121). The shortest genetic distances were between the populations from neighboring geographic areas: Nid and KVa (GD = 0.021), DRa and KVa (GD = 0.031), Juo and Kva (0.033).
Using UPGMA for clustering of populations according to Nei's distances [69] at AFLP loci, 11 descending order clades were revealed ( Figure 2). The genetically closest clades contained populations of distinct geographical region: Pal, Zag (north-western), Sve (south-eastern), and KMa (central), another example was populations DRa (south-eastern), KVa (central), and Nid (west). Hereby, a stronger link between genetic and geographic distances between populations was not revealed. According to the PCoA of AFLP data, coordinate 1 accounted for 14.5%, coordinate 2 for 11.8%, and coordinate 3 for 9.2% of the total genetic variance of populations ( Figure 3). Correspondingly, the importance of the first two coordinates accounted for 26.3% and the importance of all three coordinates for 35.5% of the total genetic variance of I. parviflora populations ( Figure 3A-C). Relations among populations from distinct geographical regions were very close. None of the populations of some geographical part of Lithuania (northwestern, central, or south-eastern) formed a separate genetic cluster. In the PCoA plot, the north-western and south-eastern populations overlapped the most ( Figure 3C), while the north-western and central populations overlapped the least. The most genetically related populations were in north-western Lithuania and the most scattered populations were in the central Lithuania.
The findings for population PCoA were also true for individual PCoA. Individuals from the north-western and south-eastern populations overlapped the most ( Figure 3E,F), and individuals from the north-western and central part were genetically less similar ( Figure 3D).
The two-level AMOVA (i.e., within and among populations) partitioned 91% of the molecular variance within populations and 9% (Φ PT = 0.087; p = 0.001) among populations (Table 3A). Analysis of AFLP loci by three-level AMOVA did not reveal significant differences among population groups based on biotope (Table 3B), geographical zones (Table 3C), or neighboring road type (Table 3D). Molecular variance among populations within groups ranged from 8% to 9%, and molecular variance among individuals within populations was 91% in all cases (Table 3B- Figure 1. Principal coordinate analysis of AFLP data for individuals showed that coordinate 1 accounted for 8.5%, coordinate 2 for 7.5%, and coordinate 3 for 5.0% of the total variation. The importance of the first two coordinates comprised 16.0% and importance of the first three coordinates comprised 20.9% of the total genetic variance of I. parviflora individuals ( Figure 3D-F). Table 3. Analysis of AFLP loci variance for small balsam (Impatiens parviflora) populations. A. Twolevel analysis. B-D. Three-level analyses, differentiation of populations in relation to biotope (B), geographical zone (C), and road type (D). Amplified fragment length polymorphism marker-based Bayesian analysis revealed that the highest ∆K indicates K = 17, and the second and third highest ∆K suggest 20 and 14 clusters, respectively ( Figure 4).

Herbaceous Species Composition
A very scattered view was obtained classifying all herbaceous plant species (138) according to their presence or absence at sites ( Figure S1). Two-way cluster analysis grouped the species into a dendrogram consisting of clades of 20 order ( Figure S1, right side cladogram) and sites with I. parviflora were grouped into a dendrogram consisting of clades of 14 order ( Figure S1, top cladogram). We identified 62 species growing in single sites only. The Zag site contained the largest number of unique species (14), seven unique species were growing at VZi, Nid, and ATr sites each. At some sites, for example, Sve, Dra, KAS, and KZa, no unique species were found.
In some other studies, environmental conditions were assessed by excluding lowfrequency species [15]. After we eliminated species that grew at single sites only, the initial number of 138 herbs decreased to 76 ( Figures S1 and 5). The initial number of herbaceous species at one site ranged from 13 to 32 [38], and excluding unique species, the range narrowed to 10-25. Only a few of the removed species were aliens. Removal of unique species did not cause more significant changes neither in total coverage of herbaceous plant species per site (from 31.6-97% to 27.1-96.5%), nor in coverage of herbaceous plant species without the input of I. parviflora (from 12.2-81.9% to 10.8-79.3%, Figure S1. A total of 76 herbaceous plant species growing along with I. parviflora were grouped by two-way cluster analysis into species dendrogram encompassing clades of 17 order ( Figure 5, right side cladogram) and into site dendrogram encompassing clades of 10 order ( Figure 5, top cladogram).
Shift from NSp-138 to NSp-76 did not cause significant differences in coverage by all herbaceous plant species (covT-138 and covT-76) or in coverage by all herbaceous species but I. parviflora (covHerb-138 and covHerb-76) ( Figure 6). Coverage of I. parviflora ranged from 5% at Zag site to 70% at KAS and Juo sites.
There was a significant difference (p < 0.01) between mean values for light (L-WEIV, 5.44 and L-UEIV, 5.79). For all remaining parameters of environment, there were no significant differences between UEIW and WEIV.

Relations between Molecular and Environmental Variables
To evaluate the links between molecular and ecological features of sites with I. parviflora, Spearman rank correlation analysis was applied (Figure 8). No correlations were found between polymorphism of AFLP and ISSR or RAPD loci.
Coverage by I. parviflora correlated negatively with (1)

Molecular Features of Impatiens parviflora
Molecular diversity is a substantial part of species diversity which is assumed as opportunity of organisms to survive over fluctuating environmental conditions over a long historical span. Genetic polymorphism of a population is often discussed in relation to invader success [70]. Polymorphism of populations within invasive and native areas has been compared for many alien species [37,44,52,71], including I. glandulifera which is congeneric to the species assessed in the present study [72,73]. It would be valuable to note that to date, any studies of I. parviflora have been performed within invasive areas only [4,[15][16][17][18][19][20][21][22][23]. Furthermore, until now, information about the molecular polymorphism of I. parviflora populations has been limited to Poland [58] and our studies in Lithuania [35,38], despite the importance of this invasive species in forests and other biotopes at the European level. We found very low (mean value, 20.1%) polymorphism of 21 Lithuanian populations at 8 AFLP loci (Table 2). These results are similar to the findings of our previous studies which employed other DNA markers such as ISSR and RAPD and reported very low polymorphism values (16.5% and 21.0%, respectively). Comparison of population groups attributed to various biotopes revealed that the least polymorphic populations were not the same for different marker systems (urban forest at AFLP and ISSR loci, riparian forest at RAPD loci with the mean polymorphisms of 18.9%, 15.6%, and 17.1%, respectively), although the differences for means per population group were not significant. Hereby, the use of 3 different DNA marker systems (presumably related to different DNA sequences) resulted in the same low genetic diversity of I. parviflora populations in Lithuania. Our findings are similar to a Polish study which demonstrated exceptionally low polymorphism (6%) at four AFLP loci (two times less than in our study) for two populations [58]. Examination of a similar number (20 versus 21) of Lithuanian I. glandulifera populations at RAPD and SSR loci [62,74] showed much higher polymorphism compared to I. parviflora. Genetically, less polymorphic populations of I. parviflora were/are more successful invaders in Lithuania than I. glandulifera [4]. This means that Baltic territory invaders of Balsaminaceae family are in agreement with the assumption of the genetic paradox of invasive species [70,72].
Comparison of agricultural crop genotypes using both dominant and co-dominant marker systems showed that AFLP markers are less polymorphic than microsatellites [75][76][77]. Despite big attention being paid to I. parviflora, until now, microsatellite markers specific for that species have not been created. Our attempts to use the microsatellites of congeneric species I. glandulifera [78,79] for Lithuanian populations of I. parviflora were not successful [38]. In addition, not all known cases of using microsatellite markers were more helpful compared to AFLP. Microsatellite and AFLP loci analyses of populations of three Draba species revealed the AFLPs as more phylogeographic structuring than the microsatellites [80].
Geographic factors may have been important in determining the small extent of variability explained by PCoA in our study of I. parviflora (20.9%; Figure 3). In the former evaluation of I. parviflora from Lithuania and Czech Republic the first three principal coordinates explained 71.5% and 84.3% of population variance at RAPD and ISSR loci, respectively [81]. The most distinct populations in PCoA plots at AFLP, ISSR, and RAPD loci were located in different geographic zones of Lithuania (Jon, Dra, and Pre, respectively), indirectly indicating that different marker systems are related to distinct sequences and there were no special natural barriers for gene flow. Some anthropogenic factors such as manor activities might have importance for the time and source of the seeds for arrival of the foreign propagules. In addition, the extent of genetic variability might be species-/genera-specific. Higher explanation of variability by PCoA (44.8% for the first two axis) was demonstrated by a similar scope Lithuanian study of populations of Bunias orientalis at ISSR loci [82]. Similar to our results of I. parviflora, the genetic diversity at microsatellite loci in invasive populations of I. glandulifera was unusually low compared to the other invasive species [72]. In contrast, the application of four AFLP primer pairs populations of I. capensis Meerb in southern New England revealed significant differentiation (Φ PT = 0.32) between the five regions in the hierarchical model and the Mantel test showed significant correlation between geographic and genetic distances [57]. Molecular parameters of populations may differ between invasive and natural areas of distribution as it was documented for microsatellite loci of I. glandulifera: genetic diversity within invasive range was lower [72].
Principal coordinate analysis together with the dendrogram of population relationships based on the Nei's [69] genetic distances (Figure 2) did not reveal a stronger link between genetic and geographic distances of populations. In support of this, PCoA-AFLP data showed overlapping of south-eastern and north-western populations (Figure 3). These facts could not be explained either by natural distribution within secondary distribution range of the species, or the intense transportation of the unimpressive plant. Furthermore, according to the size of the seeds, I. parviflora fits very well for transportation by vehicle tires [20]. Thus, our data provide evidence of unintentional human-mediated spread as a prevailing type of dissemination within the Lithuanian territory.
Partitioning of AFLP diversity showed that only 9% of variability was explained by differences among populations (Table 3). Our data are in agreement with the hypothesis of the genetic drift role in differentiation of I. glandulifera Royle populations in Europe [72]. Evaluation of native Lithuanian populations of Lythrum salicaria L. [47], Nuphar lutea Smith [83], and Juniperus communis L. [84] also revealed higher diversity within populations compared to interpopulation variability. Data about genetic variation of Impatiens noli-tangere between regions in the UK was very contradictive [55]. No AFLP-based evidence for geographical structuring within region or continent was found among 53 populations of annual Arabidopsis thaliana (L.) Heynh. sampled in North America [85].
Despite the relatively small Lithuanian territory, south-eastern, central, and northwestern regions are distinct in terms of meteorological variables (Lithuanian Hydrometeorological Service under the Ministry of Environment) [86]. Molecular investigations revealed geography-related significant differentiation of Lythrum salicaria L. [47], Phalaris arundinacea L. [87], and Juniperus communis L. [84] populations in Lithuania. Our assessment of I. parviflora in different geographic regions of the country did not show significant differentiation at AFLP loci.
Levels of invasion of aliens increase with increasing proximity to roads [88]. We sampled I. parviflora populations close to roads or footpaths. Pedestrians, bicycles, and cars may be unequally important in transporting seeds [20].
The importance of roads for unintentional human-mediated dispersion of I. parviflora seed was also discussed in many other studies [15,22,88]. We used hierarchical AMOVA to find out whether different types of roads were of distinct importance for the genetic differences between populations; however, our results did not show road type-related differentiation of populations. Due to the unintentional transfer of seeds by various means of transport (trucks, cars, bicycles, and travelers' shoes), new populations may emerge either close or far from initial locations.
In the present study, Bayesian analysis revealed the existence of 17 genetic clusters at AFLP loci in I. parviflora populations (Figure 4) and the second suggestion for the number of clusters (20) was close to the total number of populations. It corresponded to the results of AMOVA analysis (Table 3), which indicated very high variability between individuals inside populations (91%). Very similar data were obtained in the previous assessments of the same populations by other molecular markers: the presence of 11 genetic clusters at ISSR loci and 13 genetic clusters at RAPD loci [38]. Even larger number of genetic clusters were reported for Mikania micrantha, i.e., 28 genetic clusters for 28 populations across its introduced range [51].
Lithuania is surrounded by several countries with historically different socio-economic background and cultural traditions. Therefore, it is possible that I. parviflora might have crossed Lithuanian borders at several different places at different times and through distinct pathways. Based on AFLP data, we came to the same conclusion about possible multiple introductions of I. parviflora into the territory of Lithuania as in our previous study based on other dominant markers [38]. This also supports the fact about multiple introductions of many other alien species within the invasive range of distribution [37,72]. Widespread invasive species such as I. parviflora [4] may have remained low variable also after repeated introductions as has been suggested for congeneric species I. glandulifera [72].

Peculiarities of Herbaceous Plant Assemblages
Three biotopes of Lithuanian I. parviflora (i.e., agricultural scrubland, riparian forest, and urban forest) corresponded to the habitats of I. parviflora analyzed in more southern Europe such as Austria's unmanaged riparian forests [89], parks and urban forests [90], areas of anthropogenic ruderals, and garden allotments in Poland [22]. In our study, I. parviflora coverage varied considerably among the sites, i.e., it ranged from 5 to 70%. When compared to agricultural scrubland or riparian forest biotopes, the urban forest was distinguished by the highest coverage of I. parviflora (Table 4).
In the former study, we analyzed all herbaceous plant species in the sites with I. parviflora; in addition, species lists of each site have not been analyzed and EIV data were calculated using all recorded species (138). Until now, there were some methodical uncertainties left concerning the number of species required for the assessment. In the study of I. parviflora in Slovakian forests, low-frequency herbaceous plant species were removed from the list [15]. Following the same approach, we tested two sets of species: with and without unique (growing in single sites only) herbaceous species and all data provided in the current assessment concerns evaluations of reduced number of the species (76) in comparison to the earlier assessed number (138) (Figures 5-7). A diminished number of the species (Figure S1, Figures 5 and 6) appeared to be very convenient evaluating species relatedness to sites. In phytocoenological studies, connections between the species composition and soil properties have been evaluated [15][16][17][18][19][20][21][22]. According to our study, I. parviflora populations did not show the greatest genetic similarity in the areas where the species composition was the most similar (Pal, Zag, Sve, and KMa, or DRa, KVa and Nid) (Figures 2 and 5).
The most frequent neighbors of I. parviflora were native perennials (mainly hemicryptophytes) and biennial Alliaria petiolata. Our results suggested that annual I. parviflora may survive using temporal and spatial gaps of local perennials. This assumption is supported by data showing that I. parviflora suppressed small early flowering heliophilous species in the beginning of the season due to competition for light [61]. Another opportunity for the alien annual to survive among local perennials might be good chemical compatibility. Essential oils of I. parviflora did not suppress root elongation, but they were toxic for some plant species at the germination stage [91]. In our study, the most frequent co-occurring species (found at more than 40% of sites with I. parviflora) were Aegopodium podagraria, Alliaria petiolata, Anthriscus sylvestris, Chelidonium majus, Galium aparine, Geranium robertianum, Geum urbanum, Glechoma hederacea, Stellaria media, Rubus caesius, Urtica dioica, and Veronica chamaedrys ( Figure 5).
Impatiens parviflora is an invasive species whose environment has been extensively assessed using EIV [38]. Our analysis of EIVs revealed a wide range of ecological tolerance of Lithuanian I. parviflora with respect to nutrients, soil reaction, light availability, and moisture within biotopes; this is in agreement with the findings in regions of more southern Europe latitudes [15,16,22].
There are different opinions concerning the use of EIV to characterize site conditions in the absence of direct chemical and physical measurements: some authors used WEIV [8,16,65], while others employed UEIV [15,92]. In the former study, we used WEIV only and the present assessment analysis two cases: WEIV and UEIV. We used both approaches and did not reveal significant differences between all EIV parameters but light (light availability was significantly higher using UEIV compared to WEIV) (Figure 7). Correlations between respective UEIV and WEIV were significant and very high ( Figure 8). All herbaceous plant (except Anthriscus sylvestris) species growing frequently (> 40% cases) next to I. parviflora had EIV values for light ≤ 6 which met the criteria for the forest plants [93]. EIVs of the recorded species also indicated high levels of soil nutrients (7-9) at sites. Investigations of nitrogen concentrations among populations of riparian herbaceous species of Lithuania revealed the highest leaf nitrogen concentrations (determined chemically) for another annual invader Echinocystis lobata (Michx.) Torr. & A. Gray [94]. In our study, median WEIV and UEIV values for temperature were 5.65 and 5.60, respectively, and corresponded with the data on I. parviflora along oak forests at southern latitudes (Austria, Germany, Poland, Czech Republic, and Slovakia) [16]. Since the time when EIVs were defined [13], climate change has significantly shifted the timing of major phenological events, such as spring advancement and autumn postponement [95], and might have caused changes in northern latitude species composition towards plants with higher optimum temperature for growth.
When compared to agricultural scrubland or riparian forest biotopes, the urban forest in our study was distinguished by lowest EIVs for light, soil acidity, and richness in soil nutrients (Table 4).

Importance of Genetical and Ecological Variables at Sites with Impatiens parviflora
Success of invasion of alien species depends on complex admixture of both plant and environment traits. However, before our former and present studies, the relations between genetic data and assemblages of herbaceous species in sites with I. parviflora had never been traced for the species. When compared to ISSRs or RAPDs [38], the AFLP marker system appeared to be the most important searching for relationships between genetic diversity of I. parviflora and biotic and abiotic features of environment. The present study revealed that AFLP polymorphism of I. parviflora populations correlated significantly with the total coverage of herbaceous plant species (Figure 8), and in the former assessment of the same sites, it was documented that I. parviflora coverage data significantly correlated with the extent of ISSR-based polymorphism [38]; however, principal component analysis ( Figure 9) revealed lower importance of DNA marker systems for variability of sites with I. parviflora when compared to phytocoenological and indicatory data of herbaceous plant species. In accordance with the first two principal components, the highest variability of Lithuanian sites with I. parviflora was caused by coverage of herbaceous plant species excluding coverage by I. parviflora. In many studies of alien species, coverage by an invader has been interpreted as a measure of successful invasion [15,16,22].
In most studies of invasive populations, molecular data have been linked to geography [37,72]. Only a few investigations aimed to relate genetic features to biotic and abiotic environment [88,96]. Relationships between genetic differentiation and various environmental variables such as temperature, humidity, and nitrogen were documented for Phalaris arundinacea within invasive distribution range [96]. Our studies of I. parviflora represent the first attempts to relate genetic features of populations to the biotic features of their environment.
Combining data from several DNA marker systems might assist in better detection of differences between populations. However, the lack of differentiation effects in the present study could be due to noncoding regions to which molecular markers are mainly related and may not correlate with adaptive characters [37,97], which are well-expressed for I. parviflora, known as species of wide ecological amplitude [15][16][17]19,21,22].
Our methodical assessments concerning the comparisons of number of herbaceous plant species at sites and EIV application approaches might be the source of information for the development of cost-effective methods in future research. Impatiens parviflora is one of the invasive species for which management priority should be set at both European and national levels. The results of this study might be useful for forest owners implementing programs for development of a sustainable forest.
Other methods of molecular research should be used for the future assessments of I. parviflora. The further molecular investigations of I. parviflora populations should extend longitudinal and latitudinal areas, additionally encompassing the native range of distribution of the species.