Major Outbreaks in the Nineteenth Century Shaped Grape Phylloxera Contemporary Genetic Structure in Europe

Grape phylloxera is native to North America, where Vitis spp. acquired different mechanisms of resistance to leaf and root attack. Its appearance in European vineyards at the beginning of the 1860s, where the phylloxera-susceptible grapevine species V. vinifera L. is majorly cultivated, caused the devastation of a great number of vineyards, generating a deep crisis in the European wine production and trade industries. However, the origin and genetic structure of this pest across European vineyards still remain controversial and uncertain. Herein, we analysed the genetic structure of 1173 grape phylloxera individuals collected from 100 locations across eight European countries. Structure and phylogenetic analyses show that contemporary grape phylloxera populations in Europe are the result of at least two independent introductions from the native range that mirrors the historical records that also suggest two major outbreaks in Europe. The comparative analysis with samples from the native range trace back one of these two genetic groups to plants imported from the North East coast of North America, where the American species V. riparia and V. labrusca dominate. This study clarifies the level of genetic diversity of grape phylloxera in Europe and provides relevant information to resolve previous controversy about its origin.


Grape phylloxera genotyping and identification of multilocus genotypes (MLGs). A set of 1173
leaf-feeding and root-feeding grape phylloxera individuals collected from Europe (Austria, Croatia, Germany, Hungary, Italy, Romania, Serbia and Switzerland) and two other introduced regions (South Africa and Uruguay) from a diverse variety of hosts (V. vinifera L. cultivars, Vitis spp. interspecific rootstocks, Vitis spp. interspecific hybrid direct-producers and Vitis spp. interspecific resistant grape varieties) were genotyped with 7 SSR markers ( Table 1). A low number of missing data was obtained, with only 16 individuals presenting one missing locus, three individuals with two missing loci, and four individuals with three missing loci. Thus, the percentage of not-genotyped loci ranged from 0.0% to 1.45% (for PhyIII_30 and DVSSR4 loci, respectively). All the SSR markers were polymorphic among grape phylloxera samples, with a mean number of 9.1 alleles per locus. The number of alleles ranged from 5 (for DV8) to 12 (for PhyIII_55 and DVSSR4). The combination of this genetic information led to the identification of 774 unique MLGs (Supplementary Table S1). Whilst 610 grape phylloxera samples were identified only once (single MLGs), 164 MLGs were represented by 563 samples (repeated MLGs) ( Table 2). The most abundant MLG was MLG16 that consisted of 18 samples collected from the leaves of Vitis spp. interspecific rootstocks in Austria, followed by the MLG380, represented by 16 leaf phylloxera samples collected from diverse Vitis spp. interspecific rootstocks and Vitis spp. interspecific hybrid direct-producers grown in the municipalities of Stankovci (Croatia) and Quinten (in Switzerland), separated by ca. 620 km. MLG380 was not the only case in which the same MLG was collected from sites separated by large distances. MLG113 was found in Nuβdorf ob der Traisen (Austria) and in Rudolfingen (Switzerland), separated by ca. 330 km. Similarly, we found MLG475 in Köveskál (Hungary) and in Rafz (Switzerland), locations separated by ca. 690 km. On the other hand, we found seven MLGs (MLG113, MLG318, MLG329, MLG360, MLG463, MLG469 and MLG521) feeding in both leaves and roots. In addition, several examples of the same MLG in Vitis spp. interspecific hybrid direct-producers and Vitis spp. interspecific rootstocks (MLG113, MLG380, MLG394, MLF469 and MLF475), in Vitis spp. interspecific hybrid direct-producers and V. vinifera L. cultivars (MLG247 and MLG381), and in Vitis spp. interspecific rootstocks and V. vinifera L. cultivars (MLG360, MLG714 and MLG730) were found (Supplementary Table S1).
Population structure analysis of grape phylloxera in Europe. The genetic structure of the whole dataset was explored using STRUCTURE, PCoA and NJUw analyses. This study includes three variables for the samples: sampling site 20 , feeding form 22 and host plant 3 ). To avoid the impact of uneven sampling for these variables, STRUCTURE output was evaluated considering the MedMeaK, MaxMeaK, MedMedK and MaxMedK statistics to determine the most probable number of genetic groups. MedMedK, MaxMedK and MaxMeaK suggested K = 3 as the optimal level of structuring when using the sampling site (country) as correcting factor, whereas MedMeaK pointed out K = 2. On the other hand, the four metrics indicated K = 2 as the most probable number of genetic groups when using feeding form or host plant as co-factor ( Supplementary Fig. S1). At K = 2, and considering a coefficient of ancestry (Q) over 0.80 for group assignation, 302 MLGs (39.0%) were associated with one of the two genetic groups (denominated Eu1 and Eu2), whereas 472 MLGs (61.0%) were identified as "admixed" (Fig. 1A). PCoA results supported this level of structuring, as the two genetic groups identified by the previous analysis (Eu1 and Eu2) could be differentiated according to PCoA1 loadings (Fig. 1B). The neighbor-joining unweighted dendrogram built for the 295 MLGs associated to either Eu1 or Eu2 (seven were excluded because of the presence of missing data) clearly supported the presence of two clusters (Fig. 1C). One of the clusters included 158 MLGs (2.5% assigned to Eu1 and 97.5% to Eu2), and the other cluster consisted of 137 MLGs (100% assigned to Eu1). Altogether, K = 2 was considered as the most likely level of structuring in our dataset; only these results are discussed.
Genetic parameters indicated a similar level of genotypic diversity for Eu1 and Eu2 genetic groups, with values rather similar to those calculated for the whole dataset ( Table 2). The clonal origin of all the repeated MLGs found in Eu1 and Eu2 was confirmed through the significant P sex values (<0.01). The average number of alleles per locus is similar in Eu1 (6.14) and Eu2 (5.71). We detected 10 and 13 private alleles in Eu1 and Eu2, respectively, which were found across all the loci screened. A lower than expected level of heterozygosity was observed in Eu1 and Eu2, suggesting the effect of inbreeding forces on both populations. Lastly, F IS values were positive for most of the loci analysed in this work, but for PhyIII_36 in Eu1 and PhyIII_30 and DVSSR4 in Eu2.
Analysis of the factors affecting grape phylloxera population structure in Europe. The three-way ANOVA analysis revealed a major significant effect of sampling site (country) (F = 23.50; p ≤ 0.05) on grape phylloxera genetic structure, as well as a minor effect of host plant (F = 3.49; p ≤ 0.05). No significant effect for the feeding behaviour was observed (F = 1.09; p = 0.29), neither for the two-way or three-way interactions between the factors evaluated. Linear regression modelling results revealed a significant effect (p ≤ 0.05) of longitude and latitude of the sampling sites on grape phylloxera population structure. Attending to the standardised beta coefficients given by the model to each variable to compare their relative relevance on the genetic structure obtained, we observed a major effect of longitude (beta = 0.626) over latitude (beta = 0.171).

Comparative analysis of grape phylloxera MLGs in Europe, its native range and other introduced regions.
To gain insight into the origin of grape phylloxera in Eu1 and Eu2 genetic groups, the genetic information of the 302 MLGs assigned to these groups was combined with that of 319 MLGs from the native range and several introduced regions, obtained from Lund et al. 3 . STRUCTURE analysis and the ΔK criteria indicated K = 2 as the optimal level of structuring, although additional levels of structure were suggested at K = 3 and K = 5 ( Supplementary Fig. S2). At K = 2, a clear separation between the MLGs from the native range and Europe was obtained (Fig. 3). Considering a critical Q-value of 0.80 for group assignation, group 1 (NR) was formed by 212 MLGs, 211 from the native range of grape phylloxera in the current states of Tennessee (42 MLGs), Missouri (34), Virginia (26), North Carolina (17), New Mexico (17) and Arizona (15) (among others), and one MLG sampled in Germany (assigned to Eu1). Group 2 (Eu) clustered 376 MLGs, mainly from Europe (140 of Eu1 and 156 MLGs of Eu2). The remaining MLGs clustering in this group corresponded to samples isolated from the native range, including, among others, isolates from the states of New York (25 MLGs), Pennsylvania (8 Figure 1. Population structure of grape phylloxera in Europe. In A, population structure results obtained with STRUCTURE 39 considering 774 grape phylloxera MLGs and 7 SSR loci is shown. The optimal number of genetic groups (K = 2, Eu1 and Eu2) was set considering the method of Puechmaille 40 . Every MLG is shown as a vertical line, whose colour(s) indicates their estimated membership to Eu1 (red) or Eu2 (blue). Considering a critical ancestry coefficient of Q ≥ 0.80, 143 and 159 MLGs were assigned to Eu1 and Eu2, respectively (472 MLGs were considered as admixed). In B, a principal coordinate analysis (PCoA) obtained from a dissimilarity matrix calculated in DARwin from genetic data (7 SSRs) from 752 MLGs is shown. MLGs assigned to Eu1 and Eu2 are indicated as red and blue dots, respectively. Admixed MLGs are shown as yellow dots. The variance explained by the PCoA1 and PcoA2 is indicated (%). In C, the neighbor-joining unweighted dendrogram obtained in DARwin considering the MLGs grouped in Eu1 (red) and Eu2 (blue) according to STRUCTURE results is shown.  Table S2). The increase in K generated the separation of one of these two main genetic groups in several subgroups (Fig. 3). At K = 3, the NR genetic group had no major changes (it clustered 195 MLGs from the native range, and no MLGs from Europe or other introduced regions were found in this genetic group), but the Eu genetic group was divided into two genetic groups (Eu1' and Eu2'). This division matched with the genetic structuring previously observed when analysing the 774 MLGs from Europe. Eu1' grouped 188 MLGs, 122 of which were previously assigned to the Eu1 genetic group. It also clustered samples from the states of New York (22 MLGs), South Dakota (10), Pennsylvania (8), Massachusetts (7), Indiana (5), Minnesota (4) and Arkansas (4), as well as from the introduced regions of Uruguay (3 MLGs), Brazil (2) and Hungary (1). Eu2' grouped 160 MLGs, 151 of which were previously assigned to the Eu2 genetic group. The remaining MLGs corresponded to 1 MLG from Eu1, 2 from the Arizona state, 5 from Austria and 1 from Hungary (the latter corresponding to isolates from Lund et al. 3 Table S2). Eu1' and Eu2' genetic groups were well conserved at K = 5, but NR split into three genetic subgroups: NR1, NR2 and NR3 (Fig. 3), corresponding largely to major geographic regions of the grape phylloxera native range. NR1 clustered 62 MLGs, mainly from the states of Virginia (18 MLGs), North Carolina (15) and Arkansas (11). NR2 grouped 97 MLGs, mostly from Tennessee (36) and Missouri (26). Lastly, NR3 clustered 30 MLGs, most of them sampled in Arizona (21) and New Mexico (8) (Supplementary Table S2). Every country is represented as a circle, which size is proportional to the number of individuals sampled, and its colour represent the number of MLGs assigned to Eu1 (red), Eu2 (blue) or unassigned (admixed, yellow). In B, the percentage of grape phylloxera samples collected from leaves (green) and roots (brown) are shown for each genetic group (Eu1, Eu2). Samples from unknown origin are shown in grey. In C, the percentage of grape phylloxera samples collected from Vitis spp. interspecific hybrid direct-producers (dark red), Vitis spp. interspecific resistant grape varieties (light green), Vitis spp. interspecific rootstocks (yellow) and Vitis vinifera L. cultivars (violet) are shown for each genetic group (Eu1, Eu2). Samples collected of plants whose identity could not be confirmed are shown in grey. Map shown in A was created by using MapChart (https://mapchart.net).

) (Supplementary
The neighbor-joining unweighted dendrogram calculated for 575 MLGs (141 previously assigned to Eu1, 157 to Eu2, and 277 from the native range or the introduced regions analysed by Lund et al. 3 ) showed three distinct groups (Fig. 4), which matched with the distribution of the MLGs obtained by STRUCTURE at K = 3 (Fig. 3). Cluster I was mainly composed by MLGs previously assigned to Eu2' and some of the MLGs retrieved from Lund et al. 3

Discussion
Unravelling the origin of grape phylloxera in Europe is critical to resolve previous controversy about its origin. Likewise, it provides relevant information to set efficient viticultural tactics to deal with this pest in the upcoming decades, especially considering issues derived from climate change and new vineyard management practices. In this study, we analysed the genetic diversity and structure in 1173 grape phylloxera individuals through phylogenetic and population structure analyses. Samples were collected considering the latitudinal range comprised between the Rhine district in Germany and the Tuscany in Italy, and the longitudinal range between the municipalities of Nyon (western Switzerland) and Iași (eastern Romania), which represents a quantitative improvement compared to previous works 20, 21 .
Our results strongly indicate the presence of at least two independent grape phylloxera introductions from the native range into Europe, which derived into two genetic groups (Eu1 and Eu2). The structuring of grape phylloxera populations into two groups was already suggested by Forneck et al. 20 , who found a loose correlation with sampling latitude. Unfortunately, authors did not test the effect of longitude. Here, we also found a significant effect of latitude on the genetic structure, but longitude had a major effect. In this regard, Eu1 is mostly formed by grape phylloxera genotypes that dominate in western regions (Germany, Italy and Switzerland), whereas Eu2 genotypes dominate in eastern areas (Austria, Hungary, Croatia and Serbia). This genetic structure observed from nuclear genomic analyses is in accordance with previous mtDNA phylogenetic analyses that indicate a common origin for grape phylloxera in Germany and France, but independent to those from Hungary 23 . The effect of host plant on the selection of specific grape phylloxera clones has been already suggested 1,24 , especially in the native range 3,25 as a result of centuries of plant-insect coevolution 4 . Here, we observed a significant effect of host plant on the genetic structure, but it can be the result of the non-random distribution of the different plant hosts in the diverse regions analysed in this work (e.g.: no Vitis spp. interspecific rootstocks were sampled in Germany; no Vitis spp. interspecific hybrid direct-producers were sampled in Austria). In this line, a fine-scale exploration is needed to separate host plant-and geography-based associations. . STRUCTURE analysis to elucidate the origin of grape phylloxera in Europe. The population structure of 302 grape phylloxera MLGs from Europe (143 of Eu1 and 159 of Eu2) and 319 MLGs from the native range (NatR) and diverse introduced regions (InR, data from Lund et al. 3 ) was analysed considered 4 SSR loci. In each plot, every MLG is shown as a vertical line, whose colour(s) indicate ancestry proportions. The three levels of genetic stratification shown at K = 2 (A), K = 3 (B) and K = 5 (C) were determined according to the ΔK criteria proposed by Evanno et al. 46 . Considering a critical ancestry coefficient of Q ≥ 0.80, two genetic groups (Eu and NR, with 376 and 212 MLGs respectively) were obtained at K = 2. At K = 3, three genetic groups (Eu1' , Eu2' and NR, with 188, 160 and 195 MLGs, respectively) were observed. At K = 5, five genetic groups (Eu1' , Eu2' , NR1, NR2 and NR3, Eu1' and Eu2' , with 172, 157, 62, 97 and 30MLGs, respectively) were obtained. (2019) 9:17540 | https://doi.org/10.1038/s41598-019-54122-0 www.nature.com/scientificreports www.nature.com/scientificreports/ Considering the geographic distribution and historical reports, it seems plausible that genotypes of Eu1 derived from the grape phylloxera population introduced in the South of France in the 1860s 8 , which spread northwards and southwards to infest Swiss and Italian vineyards, respectively 10,14 . In addition, our results indicate that this introduction might have been responsible of the outbreaks reported in the west of Germany between 1874 and 1900 too 10 , as no additional stratification between Swiss, German and Italian samples was observed. On the other hand, grape phylloxera genotypes of Eu2 might have directly derived from the grape phylloxera outbreak detected in 1868 in the nurseries of the formerly known as the Imperial and Royal Viticulture Research Station at Klosterneuburg (Austria) 16 , where vine cuttings from different regions (including the USA and France) were received 26 . According to historical records, vine cuttings were distributed as a solution against oidium before grape phylloxera symptoms were evident, easing the dispersal of the insect across the former Austro-Hungarian Empire 10,26 . The independence on the origin of these two outbreaks was already evidenced in the 1870s, as no grape phylloxera injuries were detected when all the plant material received from France was screened at the Klosterneuburg nurseries 27 .
Our analysis confirmed previous findings that indicated a close relationship between European grape phylloxera populations and north-eastern American populations 3,19,23 . Nevertheless, we found this statement to be exclusively true for the Eu1 genetic group, which was likely founded by grape phylloxera populations originating from the North East coast of North America, where V. riparia and V. labrusca dominate 4 . How grape phylloxera was introduced into France is still controversial, although it is generally accepted it might have been introduced on roots of V. labrusca and/or leaf-galled V. riparia plants 4 , agreeing with our results. Some historical records suggest that the French outbreak probably arose from the cultivation of infested plants of the hybrid grape cultivar Isabella à grains noirs, introduced from Pennsylvania 28 . Some reports additionally indicate that first grape phylloxera populations could have been introduced to France via United Kingdom, where grape leaf galls were already observed in table grape vines by 1863 11 . The cultivation of Isabella à grains noirs vines (probably corresponding to the cultivar nowadays known as Isabella, V. labrusca × V. vinifera) was rather common in the South of France due to its pleasant fragrance and raspberry-like flavour 26,28 . Unfortunately, this trend favoured the rapid dissemination of the insect across neighbouring vineyards, and caused the decease of a great number of vines throughout west and central Europe. On the other hand, historical reports indicate that the strains initially introduced into Austria (founder of the Eu2 genetic group) could have been transported within a set of 20 American varieties imported to Klosterneuburg from New Jersey (USA), which were planted in the experimental vineyard to evaluate their fitness against oidium 29 . Nevertheless, we could not place the point of origin of this genetic group, as genotypes in Eu2 clustered only with samples from other introduced regions and not with those of the native range. In this line, it would be of interest to cross our data with a major number of grape phylloxera samples collected in the native range, as a wide insect genetic diversity exists as the result of plant-insect coevolution 4 .
Samples from the introduced regions of Uruguay and South Africa did not form a specific cluster, suggesting that (I) either both regions received infested plant material from Europe, or (II) the introduction came from the same source(s) as those that devastated European vineyards. Accordingly, it has been recently indicated that Uruguayan populations might have resulted from the unintended introduction of infested plant material from Europe 30 .
Our analysis revealed an important role of the first introductions that occurred in the nineteenth century in Europe. Considering historical records, it is likely that the introduced grape phylloxera populations rapidly spread across the continent by the unintended transportation of infested American vine cuttings to fight other disease-causing agents. This human-assisted dispersal counteracted the sessile nature of the insect for most of its life cycle 2 , which presents a low capability of colonizing new vines or vineyards 5 . The lack of natural resistance in V. vinifera L. and the absence of natural enemies in European vineyards favoured the fast colonization of the new environment. Genetic diversity parameters indicate that these initial populations were subjected to different selection pressures that lead to the selection of different alleles for local adaptation, generating a series of clones mainly by parthenogenesis, as suggested by significant P sex values. In fact, parthenogenesis is acknowledged as the preferable mechanism for grape phylloxera reproduction in different introduced regions 21,31-33 . Clonal reproduction is still capable of generating a high level of genetic variation per generation 34 . This reproductive mechanism likely facilitated the long-term persistence of early introduced populations until today, as this strategy favours aphid populations persistence in areas with temperate climate 35 , like European winemaking regions 36 . Unravelling the mechanisms that drove to the generation of site-and host-adapted grape phylloxera populations in Europe will aid to understand why some of the rootstocks generated more than a century ago are still useful to control grape phylloxera effects, as well as to plan effective measures of control (including the breeding of new more resistant rootstocks) to anticipate to new, more aggressive outbreaks.

Material and Methods
Grape phylloxera sampling. Leaf and soil-emerged grape phylloxera individuals were collected from V. vinifera L. cultivars (such as Cabernet Sauvignon and Chardonnay), Vitis spp. interspecific rootstocks (like Kober 5BB and Teleki 5C, from both abandoned and commercial vineyards), Vitis spp. interspecific hybrid direct-producers (like Léon Millot and Maréchal Foch), and Vitis spp. interspecific resistant grape varieties (also known as PIWIs, like Kosmopolita and Muscaris) in 100 locations from eight European countries (Austria, Croatia, Germany, Hungary, Italy, Romania, Serbia and Switzerland). It corresponded to a latitudinal gradient from 43.0°N to 50.0°N and a longitudinal gradient between 6.2°E and 27.6°E. Samples were collected from 2012 to 2019 ( Supplementary Fig. S3). In addition, some leaf-feeding grape phylloxera samples from Uruguay and South Africa were considered to acquire a representation of grape phylloxera from other introduced regions (Table 1). Grape phylloxera feeding form and host plant was annotated for each collected sample, and GPS coordinates (latitude and longitude) were recorded for each sampling site as decimal degrees. Supplementary Table S3 provides more information of the samples collected, and details on most of the samples analysed from Austria, Switzerland and Germany can be found elsewhere 18,19 . Leaf-feeding grape phylloxera individuals were obtained by collecting galled leaves from infested vines, which were stored at −20 °C until further analyses 19 . Root-feeding grape phylloxera individuals were sampled using emergence traps, as detailed in Powell et al. 37 . After insect collection, root-feeding grape phylloxera individuals were stored in ethanol at 2 °C for DNA extraction. In total, 1173 individuals (1121 from Europe, 44 from Uruguay and 8 from South Africa) have been analysed in this work.
DNA isolation and genotyping. Whole genomic DNA was extracted from grape phylloxera individuals as indicated in Forneck et al. 19 and stored at −20 °C for further use. A total of 1173 samples were genotyped with seven fluorescently labelled (6-FAM or HEX) simple sequence repeat (SSR) primers (Phy_III_55, Phy_III_30, Phy_III_36, Dvit6, DV4, DV8 and DVSSR4). DNA amplification, fragments selection and allele calling were performed as detailed previously 19,38 . In every set of samples, we included six control genotypes with known allele size to keep allele calling consistent between different runs. This analysis revealed the presence of 774 grape phylloxera MLGs.
Grape phylloxera MLGs from the american native range and introduced regions. In a previous work, Lund et al. 3 explored the population genetic diversity of grape phylloxera in its native range, using samples collected across the USA, as well as some samples obtained from Europe (Austria and Hungary) and South America (Argentina, Brazil, Peru and Uruguay). These samples were genotyped using 32 SSRs, including the Genetic analyses. Population structure analysis of grape phylloxera in Europe. The Bayesian clustering software STRUCTURE v.2.3.4 39 was used to infer the number of genetic groups and to assign individuals to the inferred populations, using our dataset of 774 MLGs based on 7 SSR markers. The existence of genetic structure was tested in a number of hypothetical genetic groups (K) from 1 to 15, using a cycle of 250.000 burn-in steps followed by 500.000 Markov Chain Monte Carlo iterative steps. Ten replications per K value were run to assess the consistency of the results, and each one was performed considering an admixture model with uncorrelated allele frequencies. Aware of the uneven sampling for the factors likely affecting grape phylloxera population structure (sampling site, feeding form, host plant), four estimators of the most likely K value (median of means (MedMeaK), maximum of means (MaxMeaK), median of medians (MedMedK) and maximum of medians (MaxMedK)) were determined using a membership coefficient threshold of 0.5, as proposed by Puechmaille 40 and implemented in STRUCTURE SELECTOR 41 . Thus, these four statistics were independently calculated using the different group classes obtained for sampling site (Austria, Croatia, Germany, Hungary, Italy, Romania, Serbia, South Africa, Switzerland or Uruguay), feeding form (leaf or root) or host plant (V. vinifera L. cultivars, Vitis spp. interspecific rootstocks, Vitis spp. interspecific hybrid direct-producers or Vitis spp. interspecific resistant grape varieties) as co-factors. The statistics of the three scenarios were then compared, and the most likely number of genetic groups was determined as the most repeated K value within the 12 calculations. Once the most probable K value was set, MLGs were assigned to a genetic group considering an ancestry coefficient Q ≥ 0.80; otherwise, a MLG was considered as "admixed". In parallel, a dissimilarity matrix was calculated between 752 grape phylloxera MLGs (22 were discarded because of the presence of missing data) using the DARwin software package v. 6.0.21 42 on the basis of 10.000 bootstrap steps, which was further used for a Principal Coordinate Analysis (PCoA). As indicated by Emanuelli et al. 43 , a Neighbor-Joining Unweighted (NJUw) dendrogram was constructed for the subset of MLGs clustered in the genetic groups previously determined to support the consistency of STRUCTURE and PCoA clustering methods. The dendrogram was constructed on the basis of a dissimilarity matrix with 10.000 bootstrap steps, and it was drawn considering 1.000 bootstrap replicates. MLGs with missing data were excluded from the analysis. The three outcomes (STRUCTURE, PCoA and NJUw) were used to determine the optimum level of genetic structure of grape phylloxera in Europe.
For each genetic group, we calculated the mean number of alleles per locus, the observed (H obs ) and expected (H e ) heterozygosities under Hardy-Weinberg expected equilibrium, and F IS values, which were independently calculated for each locus. The genotypic diversity index (R) was estimated as R = (MLG-1)/(N-1), and the clonal diversity index (Pd) as Pd = MLGs/N, MLG being the number of multilocus genotypes identified, and N the number of sampled individuals. All calculations were performed using MLGsim 2.0 44 with 10.000 simulation steps. In addition, the same software was used to calculate the P sex statistic to estimate the likelihood that grape phylloxera individuals with the same MLG are products of distinct sexual reproductive events or truly clonal. Thus, a multiple MLG was considered to be a true clone if P sex < 0.01 (i.e. low probability of being the result of sexual reproduction).
Analysis of the factors affecting grape phylloxera population structure in Europe. To evaluate the effect of sampling site (country), feeding form and host plant on the genetic structure of grape phylloxera in Europe, a three-way ANOVA was calculated focusing in the European MLGs with an ancestry coefficient Q ≥ 0.80 in any of the identified genetic groups (n = 302). For this analysis, three variables (country, feeding form and host plant) were used as independent variables, defining eight levels for country, two for feeding form and four for host plant, as detailed above. As dependent variable we used the Q values given by STRUCTURE to each MLG (Q-matrix). Factors were considered to have a significant effect on the Q-matrix at p ≤ 0.05. To deepen into the effect of geography on the genetic structure, we calculated a linear regression model using the Q-matrix as dependent variable and latitude and longitude values of the sampling sites as independent variables. Factors were considered significant if p ≤ 0.05. Analyses were performed using SPSS v.24.0 (IBM, Chicago, IL, USA).
Population structure analysis of grape phylloxera in Europe and its introduced region. A second population structure analysis was performed with STRUCTURE 39 to analyse the relationship between the 302 European MLGs clustered in the final genetic groups previously identified and the MLGs reported in Lund et al. 3 sampled in the native range and in diverse introduced regions. In total, 621 MLGs have been included in this analysis. To this aim, we used the procedure described above, but using the information available for four SSR loci (Phy_III_55, Phy_III_30, Phy_III_36 and Dvit6). The most likely number of genetic groups was assessed by plotting the ΔK value of the data over 10 runs, as implemented in STRUCTURE HARVESTER 45 , and defined in Evanno et al. 46 . MLGs were assigned to a genetic group if they have a coefficient of ancestry (Q) over 0.80; otherwise, they were considered as "admixed". A dissimilarity matrix was calculated considering genetic data on 4 SSRs for 575 MLGs as previously detailed, discarding 46 MLGs for the presence of missing data. The matrix was then used to construct an NJUw dendrogram with 1.000 bootstrap replicates to get insights into the relationship between the genetic groups identified in our dataset and those previously reported from grape phylloxera native range and different introduced regions 3 .