1 Molecular Ecology (2000) 9, Comparison of genetic diversity of the invasive weed Blackwell Science, Ltd Rubus alceifolius Poir. (Rosaceae) in its native range and in areas of introduction, using amplified fragment length polymorphism (AFLP) markers L. AMSELLEM,* J. L. NOYER,* T. LE BOURGEOIS* and M. HOSSAERT-MCKEY *Centre de Coopération Internationale de Recherche Agronomique pour le Développement, Avenue du Val de Montferrand, BP 5035, Montpellier Cedex 1, France, Centre d Ecologie Fonctionnelle et Evolutive, CNRS 1919 route de Mende, Montpellier Cedex 5, France Abstract Theory predicts that colonization of new areas will be associated with population bottlenecks that reduce within-population genetic diversity and increase genetic differentiation among populations. This should be especially true for weedy plant species, which are often characterized by self-compatible breeding systems and vegetative propagation. To test this prediction, and to evaluate alternative scenarios for the history of introduction, the genetic diversity of Rubus alceifolius was studied with amplified fragment length polymorphism (AFLP) markers in its native range in southeast Asia and in several areas where this plant has been introduced and is now a serious weed (Indian Ocean islands, Australia). In its native range, R. alceifolius showed great genetic variability within populations and among geographically close populations (populations sampled ranging from northern Vietnam to Java). In Madagascar, genetic variability was somewhat lower than in its native range, but still considerable. Each population sampled in the other Indian Ocean islands (Mayotte, La Réunion, Mauritius) was characterized by a single different genotype of R. alceifolius for the markers studied, and closely related to individuals from Madagascar. Queensland populations also included only a single genotype, identical to that found in Mauritius. These results suggest that R. alceifolius was first introduced into Madagascar, perhaps on multiple occasions, and that Madagascan individuals were the immediate source of plants that colonized other areas of introduction. Successive nested founder events appear to have resulted in cumulative reduction in genetic diversity. Possible explanations for the monoclonality of R. alceifolius in many areas of introduction are discussed. Keywords: AFLP markers, biological invasions, genetic diversity, Indian Ocean islands, Rubus alceifolius, weeds Received 7 August 1999; revision received 2 November 1999; accepted 2 November 1999 Introduction Colonization often involves marked founder effects which reduce genetic variation and enhance genetic differentiation (Husband & Barrett 1991). This is particularly the case for weedy species introduced by man to new areas. The reproductive system of weedy plants involves predominantly selfing or apomixis (Husband & Barrett 1991). Introduced populations often descend from a few Correspondence: M. Hossaert-McKey. Fax: ; founders, and are completely isolated from source populations. These factors combine to reduce genetic variation in introduced populations. In contrast, in outbreeding species, relatively less reduction in genetic variation may occur if large population sizes are maintained following introduction (Brown & Marshall 1981). However, even initially outbreeding species may undergo reduction in genetic variation upon introduction, because breeding systems are known to evolve in introduced populations, in comparison with native ones (Barrett 1996). For example, selfcompatibility systems may be favoured in plants colonizing 2000 Blackwell Science Ltd
2 444 L. AMSELLEM ET AL. new areas during foundation events and population extension (Carr et al. 1986; McMullen 1987; Webb & Kelly 1993). Despite their potential interest for investigating colonization events, few comparisons have been made of the genetic diversity in introduced island populations and native mainland populations of the same plant species. The few existing comparisons between conspecific mainland and island populations (which concern species naturally found on islands, rather than introduced there) generally indicate that island populations contain less genetic variation than those from mainland areas (reviewed in Barrett (1996)). This lower genetic diversity in islands is assumed to be due to founder effects, inbreeding, and less selection exerted by predators. The genus Rubus is a good model for studying the effects on genetic variation of the introduction of species on islands. Rubus species have been shown to be often very invasive, particularly on islands such as Hawaii and island continents such as Australia (Howarth et al. 1997; Evans et al. 1998). The genus Rubus L. is usually hermaphroditic (Nybom 1986; Richards 1986). Thus, while outcrossing occurs (Antonius & Nybom 1995), plants are also potentially self-pollinating, and selfing is frequent (Nybom 1988). Rubus spp. propagate vegetatively very vigorously, enabling clonal spread of single individuals in a patch of habitat. Furthermore, apomixis (asexual reproduction through seed formation; Asker 1979) occurs in the two subgenera most important in horticulture: polyploid species of Rubus (blackberries and relatives) and, very rarely, Idaeobatus (raspberries) from polyploid species deviant of normally diploid species (Gustafsson 1943; Nybom 1988). Thus, propagules capable of longdistance dispersal between habitat patches can also be asexually produced. R. alceifolius Poir. (Rosaceae, subgenus Malachobatus Focke), is a simple-leafed southeast Asian bramble that has been a serious weed on La Réunion island since the early 20th century (Lavergne 1978; Quere 1990; Soulères 1990; Sigala & Lavergne 1996). Although the precise origin of this weed and the time of its introduction in this island are still unknown, records suggest an introduction sometime in the mid-19th century. R. alceifolius invades native vegetation and road edges from sea level to about 1600 m elevation. Introduced populations of this bramble also occur on other Indian Ocean islands (Madagascar, Mayotte, Mauritius) and in Australia (Queensland). To our knowledge, no records of its geographical origin(s) and the date(s) of introduction to the areas of introduction are known. According to E. Bruzzese (personal communication), its introduction to Queensland may have been very recent (about 1950). This species thus offers the possibility of studying multiple introduction events. This is interesting for two reasons. First, for some comparisons each introduction can be treated as independent, giving the replication necessary for answering certain questions: Does an introduction always result in lowered variation relative to the source population? Second, one introduced population may be derived from another. Do successive, nested introduction events result in repeated population bottlenecks and cumulative reduction of genetic variation? Both of these questions require knowledge or inference of the history of introduction and the relationships among native and introduced populations. Also, the development of a strategy of biological control of these invasive plants requires information on their geographical origin and on the genetic composition of native and introduced populations. For all these reasons, we began a study of genetic diversity, using AFLP (amplified fragment length polymorphism), of this bramble on La Réunion, where this plant is a serious weed, in order to determine the potential geographical origin(s) of introduced R. alceifolius. Our study compared genetic variation in southeast Asia where this species occurs naturally (Thailand, Laos, Vietnam, Sumatra and Java) and in localities within five areas where the plant has been introduced (La Réunion, Mayotte, Mauritius, Madagascar, Australia). We compared intra-area genetic diversity, and used the patterns of variation to make inferences about the reproductive biology of this weed in areas where it is native or introduced. We also analysed genetic diversity among areas in order to quantify genetic distances among populations, and thereby evaluate alternative scenarios of introduction of R. alceifolius. Our study was designed to obtain an overall view of world-wide genetic diversity of R. alceifolius in both its native range and areas of introduction, and to permit reconstruction of possible routes of transport and introduction of this weed. Materials and methods Bibliographic references about the introduction of Rubus alceifolius in La Réunion island Although different sources agree on the approximate time of introduction of R. alceifolius in La Réunion, about 1850, they disagree on its geographical origin. Some sources suggest that R. alceifolius was introduced from populations in India (from the Botanical Garden of Calcutta (Cordemoy 1895), where there are no current traces of this plant in herbaria and archives (L. K. Banerjee, personal communication)), Malaysia (Rivals 1960; Friedmann 1997), northern Vietnam ( Jolivet 1984), or Thailand (Figier & Soulères 1991). Another reference suggests that the plant may have first been introduced into Madagascar from Indonesia, and then introduced to La Réunion from Madagascar (Owadally 1980). Conversely, Vaughan (1937) suggests that the plant considered to be R. alceifolius in Madagascar
3 GENETIC DIVERSITY OF R. ALCEIFOLIUS 445 is in fact an endemic species: R. roridus (Lindl.), which is differentiated from R. alceifolius only by the aspects of the hairs on stipules and bracts. How the plants were introduced (by seeds or cuttings), and the number of initial founders, are both unknown. Plant material Samples were collected in numerous localities in both the native range and the areas of introduction of this plant (Table 1), and provided a good coverage of the known world distribution of R. alceifolius (Kalkmann 1993; Friedmann 1997). The native range was well sampled, with localities ranging from northern Vietnam to Java. However, one part of the native range was not represented in our sample. R. alceifolius is reported to be present in northeastern India (Arunachal Pradesh, Assam, Meghalaya and Manipur: L. K. Banerjee, personal communication). Unfortunately, it was not possible for us to obtain samples from these regions. We sampled large numbers of individuals and populations in Thailand, especially in Kaho Yaï National Park, where six populations of R. alceifolius and one population of another simpleleafed Rubus sp. were sampled. Sampling in this locality was extended in order to quantify variability within and between natural populations in the native range. Samples from other localities in the native range came from scattered individuals rather than a set of populations. Populations in five areas in the overall areas of introduction were also well sampled, especially La Réunion, the current focus of biological control efforts. In this island, sampling took place in as many populations as possible in both upland and lowland sites, in order to increase the resolution for this locality, and thus to minimize the risks of not observing a very localized polymorphism. Samples from localities other than Vietnam, Thailand and La Réunion were sent by several persons. In order to test the capacity of discrimination of the molecular marker used, various other species of Rubus more or less closely related to the focal species were added. We included Rubus spp. with compound leaves belonging to the subgenus Idaeobatus, and colleagues with simple leaves from the same subgenus as R. alceifolius Malachobatus (considered to be the more closely related to R. alceifolius). All these Rubus spp. were introduced in this study to compare taxonomic relationships based on morphology with those suggested by the molecular marker (see Table 1 for taxonomic details). Molecular techniques Total DNA was extracted, according to the protocol of Bousquet et al. (1990), from approximately 100 mg of foliar tissues dried with silica gel. The integrity of the DNA was estimated on agarose gels, and the quantity was determined using a fluorometer. AFLP (Zabeau & Vos 1993) is frequently used to study molecular polymorphism at the infraspecific level (Sharma et al. 1996; Roa et al. 1997; Escaravage et al. 1998; Russell et al. 1999). We used commercial AFLP kits (Life Technologies ) with the given pair of restriction enzymes (EcoRI and MseI), and made radioactively marked fragments with γ 33 P migrate on a urea polyacrylamide gel electrophoresis (PAGE). The result is displayed on an autoradiograph (Fig. 1). Four primer pairs were used in preliminary screenings. All of them revealed polymorphism and thus were retained for this study. The details of each primer pair are given in Table 2. For each analysis, the same three Réunionese specimens were used as controls. The purpose of this was not only to check the robustness and the reproducibility of the method, but also to calibrate each gel on known band levels, in order to avoid any shifting or confusion from one gel to another for a given primer pair. The data were coded binarily (1 = presence; 0 = absence) on each band level and for each considered individual. Only individuals with unambiguous signal intensities and banding patterns after two independent readings of AFLP profiles were retained. Methods of analysis The study was focused on two levels of comparison: among individuals and among areas. Resemblances/differences among individuals Similarity among individuals. The index of similarity of Sokal and Michener (or simple matching ) (Sokal & Sneath 1963) was calculated between all the individuals, using the software ntsys-pc (Rohlf 1993). Its value varies between 0 (for two individuals not having any common band) and 1 (for two identical individuals). This index allowed us to quantify intra-area genetic variability, and to draw a phenogram. Quantification and comparison of intra-area diversity. The matrix of similarities allowed us to quantify genetic variability within each area, by taking for each individual the mean of the index for all pairwise comparisons with other individuals of the locality concerned. In the areas of introduction, we first studied each area, and then we considered the group (Mauritius + Mayotte + Queensland) as an entity. This allowed us to estimate the genetic variability among these three areas, and to compare it with each separate area in the overall areas of introduction. A Duncan s multiple range test (SAS 1996, version 6.12, proc glm) was performed to detect statistical differences among means of similarity indices by areas. Although the assumption of independent variables is violated in the
4 446 Table L. 1 Details AMSELLEM of sample collection ET AL. localities for each area considered, and description of sample size by species and site of collection Locality Species Subgenus Site or region names Type of leaves Sample size Area of introduction La Reunion (RE) R. alceifolius 67 sites Rubus alceifolius R. apetalus Ideobatus Tevelave Compound 1 3 Rubus spp. R. fraxinifolius Ideobatus Plaine des palmistes Compound 1 R. rosifolius Ideobatus Cilaos Compound 1 Mayotte (MT) R. alceifolius Unknown 8 8 Rubus alceifolius Mauritius (MU) R. alceifolius Petrin 3 7 Rubus alceifolius R. alceifolius Plaine Paul 4 Madagascar (MD) R. alceifolius Ampasimandinika 1 19 Rubus alceifolius R. alceifolius Bengaly 1 R. alceifolius East Ranomafana 2 R. alceifolius Fanandrana 3 R. alceifolius Ivoloina 2 R. alceifolius Perinet (Andasibe) 7 R. alceifolius Tamatave 2 R. alceifolius Unknown 1 Queensland (AU) R. alceifolius Babinda 1 7 Rubus alceifolius R. alceifolius Banston Beach 1 R. alceifolius Millaa Millaa 3 R. alceifolius Tully 2 Native range Thailand (TH) Rubus sp. Unknown Doï Inthanon Simple 1 58 Rubus alceifolius Rubus sp. Unknown Doï Suthep Simple 4 13 Rubus spp. Pop. 1: R. alceifolius Kaho Yaï National Park 15 Pop. 2: R. alceifolius Kaho Yaï National Park 15 Pop. 3: R. alceifolius Kaho Yaï National Park 13 Pop. 4: R. alceifolius Kaho Yaï National Park 9 Pop. 5: R. alceifolius Kaho Yaï National Park 5 Pop. 6: R. alceifolius Kaho Yaï National Park 2 Pop. 7: Rubus sp. Unknown Kaho Yaï National Park Simple 3 R. blepharoneurus Malachobatus Doï Inthanon Simple 1 R. ellipticus Ideobatus Doï Inthanon Compound 1 R. rufus Malachobatus Pilok Simple 2 R. rugosus Malachobatus Doï Inthanon Simple 1 R. rugosus Malachobatus Kaho Yaï National Park Simple 1 Vietnam (VT) R. alceifolius Bavi 1 31 Rubus alceifolius R. alceifolius Cuc Phuong 8 9 Rubus spp. R. alceifolius Lang Son 5 R. alceifolius Tam Dao 8 R. alceifolius Yen Baï 8 R. clemens Malachobatus Tam Dao Simple 1 R. leucanthus Ideobatus Yen Baï Compound 1 R. moluccanus Malachobatus Lang Son Simple 1 R. moluccanus Malachobatus Tam Dao Simple 1 R. parkeri Malachobatus Yen Baï Simple 1 R. polyadenus Malachobatus Lang Son Simple 1 R. rosifolius Ideobatus Lang Son Compound 1 R. rosifolius Ideobatus Tam Dao Compound 1 Laos (LA) R. alceifolius Houay Leuk 1 1 Rubus alceifolius R. rugosus Malachobatus Houay Leuk Simple 2 2 Rubus sp. Java (JA) R. alceifolius Sumedang 4 4 Rubus alceifolius Sumatra (SU) Rubus sp. Unknown Sumut Deli Sibolanguit Simple 1 14 Rubus alceifolius Rubus sp. Unknown Sumut Samosir Toba Simple 1 3 Rubus spp. R. alceifolius Sipirok Tapanuli Utara 2 R. alceifolius Sumut Simalengue Aek Nauli 3 R. alceifolius Sumut Samosir Toba 9 R. moluccanus Malachobatus Baligue Sibooromborong Aek Nauli Simple 1 Total number of samples = 254 of which Rubus alceifolius = 224 and other Rubus spp. = 30.
5 GENETIC DIVERSITY OF R. ALCEIFOLIUS 447 Fig. 1 Example of a monomorphic (Réunionese individuals) and a polymorphic (Vietnamese individuals) gel. The gel photographs have been disposed in order to align the homologous band levels (with difference in the migration time, however). (Primer pair represented: E-AAC/M-CTT; locus numbered from 57 to 123). Table 2 Polymorphism revealed for each primer pair Primer pair Number of bands Number of polymorphic bands % of polymorphic bands E-AAC/M-CAA E-AAC/M-CAT E-AAC/M-CTA E-AAC/M-CTT Total Mean Standard deviation case of genetic distances among a finite set of individuals (Day & Quinn 1989; SAS 1989), this test was used because no more appropriate test appears to be available. Thus, our results were interpreted with caution. Phenogram among individuals. A phenogram was constructed from the matrix of similarity according to the neighbour-joining algorithm, using the software phylip (Felsenstein 1993). The statistical robustness of each node
6 448 L. AMSELLEM ET AL. was estimated by bootstrap with 2000 replicates as defined by Hedges (1992), and using a program kindly provided by Jean-Marie Cornuet (INRA Montpellier, Laboratory of Modelling and Evolutionary Biology). It is important to note the limits of interpretation of such a phenogram by individuals. The index used here only allows inspection of phenetic, not phylogenetic, relationships between individuals and populations, because of the dominance of the molecular marker used. Also, the phenogram presents only a unidimensional image of similarities among individuals. The Rubus spp. from the subgenera Idaeobatus and Malachobatus (other than R. alceifolius) used in construction of the phenogram were omitted from other analyses, in order to avoid background noise resulting in loss of precision and resolution of our results. Their sole purpose in this study was to confirm the reliability of the analysis, and to root the phenogram. Factorial correspondence analysis (FCA). In order to obtain a more synthetic picture of the organization of molecular variation among all the R. alceifolius sampled compared with the phenogram, a FCA was performed on the binary data matrix presence/absence (SAS 1996; proc corresp). The FCA identifies several independent axes (or dimensions) that account for the largest part of the variation. For each dimension, each individual has an ordinate representing its position compared with the other individuals considered. Thus, it was possible to represent in the form of scatter plots all the individuals. Often, the first three dimensions account for a large part of the variation, so that position of each individual can be visualized using these three dimensions in two separate graphs. Comparison among areas. We also quantified resemblances/ differences among areas. To examine biogeographical patterns, individuals within each area were grouped as an entity, and a phenogram of similarities among areas was constructed by the neighbour-joining algorithm, directly from the presence/absence matrix. Robustness of the nodes was evaluated by bootstrap (2000 replicates). Rubus spp. from the subgenus Malachobatus were omitted from this analysis. The phenogram was rooted with the Rubus spp. from the subgenus Idaeobatus. Results Resemblances/differences between individuals Intra-area variability. Mean values for the simple matching index, calculated within each locality, are given in Table 3. In all populations in areas of introduction, except in Madagascar, this index of similarity showed a low intraand inter-area variability. In these areas, this negligible level of variability corresponded to levels expected from a reasonable error rate in the experimental procedure, assuming no genetic variability at all and a single genotype. It is interesting to note that the group (Mayotte/ Mauritius/Queensland) in the phenogram is comparable in variability to Mayotte, Mauritius, or Queensland taken alone, suggesting that the plants of these three areas are in fact a single clone. The individualization of the Réunionese Rubus alceifolius in the phenogram, associated with the negligible intra-area genetic variability suggests that another clone is present on that island, compared with the group mentioned above. In Madagascar, the similarity index presented a higher level of intra-area variability. In areas within the native range, the index showed a higher degree of intra- and inter-area variability, except in Java, where only four individuals were sampled, and where nothing is known about their spatial distribution. This high similarity observed between the Javanese individuals may be an artefact due to small sample size, and perhaps to close geographical proximity of the few individuals sampled (several ramets of very few genets may have possibly been sampled in this locality). In Sumatra, the apparent low variability observed among the individuals is compensated by a rather large standard deviation, and may here again correspond to the analysis of several ramets of some few dissimilar genotypes. Phenogram of diversity by individual. On this phenogram (Fig. 2), the Rubus spp. from the subgenus Idaeobatus collected on La Réunion, in Thailand and in Vietnam are Table 3 Means and standard deviations (SD) of intra-area indices of similarity of simple matching. Different letters accompanying means in parentheses indicate significant differences (P < 0.05, Duncan s multiple range test) Area of introduction Native area RE MT MU AU MD MT/MU/AU TH VT LA JA SU Sample size Mean 0.99(a) 0.99(a) 0.98(a) 0.98(a) 0.83(c) 0.98(a) 0.77(d) 0.73(e) 0.99(a) 0.93(b) SD
7 GENETIC DIVERSITY OF R. ALCEIFOLIUS 449 Fig. 2 Phenogram constructed using the neighbour-joining method, according to the simple matching index, and rooted on the Rubus spp. with compound leaves. The numbers indicate bootstrap values (2000 replicates) for each node. The solid triangles are proportional to the sample size of the designed group. clustered together (R. rosifolius, R. leucanthus, R. apetalus, R. ellipticus, and R. fraxinifolius, representing a total of seven individuals). We considered this group as the outgroup, and used it to root the phenogram. After the outgroup, the Rubus spp. from the subgenus Malachobatus are clustered: R. rufus, R. blepharoneurus, R. parkeri, R. moluccanus, R. clemens, and R. polyadenus (representing a total of 23 individuals), followed by R. alceifolius from the various studied areas. Two major clusters of R. alceifolius appear in this phenogram. The first major cluster represents localities from areas of introduction. The node distinguishing this group is statistically very strong (bootstrap of 92%). R. alceifolius within areas of introduction is strongly differentiated from R. alceifolius within the native range in Asia. Moreover, each of the five areas of introduction is clustered apart from the others, here again with statistically robust nodes. From this phenogram, only one Madagascan individual is placed with plants from an area within the native range (Vietnam). R. alceifolius of Queensland is placed firmly in the middle of a cluster including all plants from Mauritius and Mayotte. These three areas form a group with a bootstrap value of 100%. This result is in accord with those comparing intra-area diversity. The second major group in the phenogram is represented by each area within the native range, which each forms a
8 450 L. AMSELLEM ET AL. cluster with low bootstrap values for the highest-order nodes. It is interesting to note that individuals from each Thai population are clustered together. FCA. The five first dimensions accounted for 49% of the total variance in the data matrix (Table 4). Only the first three dimensions (explaining 41.44% of the total explained variance) were retained, to establish spatial representations of the individuals. Representation in two dimensions are given in Fig. 3a (dimension 1/dimension 2) and Fig. 3b (dimension 1/ Table 4 Five principal dimensions of the factorial correspondence analysis (FCA) and their respective contributions to the total variance Dimensions % % % % % Total 49.03% Contributions to the original variance Fig. 3 Factorial correspondence analysis on the Rubus alceifolius individuals (n = 224). (a) Dimensions 1 and 2; (b) Dimensions 1 and 3.
9 GENETIC DIVERSITY OF R. ALCEIFOLIUS 451 Fig. 4 Phenogram of the various areas considered, constructed using the neighbour-joining method. Bootstrap values are indicated for each corresponding node (2000 replicates). dimension 3). They confirm our results obtained with the phenogram and the quantification of intra-area diversity. Globally, genetic diversity is much higher in the native range, where geographical areas are discriminated even at the scale of intra-area diversity. The areas of introduction are separated into two clusters: Madagascar showing higher genetic variability among individuals, and the other localities (representing 99 individuals) showing less variation and forming a single entity. Resemblances/differences among areas The phenogram obtained when individuals are grouped by area is given in Fig. 4. All the nodes of this phenogram are very robust (> 83%), except for the node connecting the localities of Mauritius with Queensland and Mayotte, certainly because of the same genotype existing in those three islands. The areas of introduction and the native range are very well separated, except for Vietnam, which is situated as a sister group of the rest of the native range. Discussion Our results clearly demonstrate that levels of genetic diversity are very different between native Asian Rubus alceifolius and introduced populations in the Indian Ocean islands and Queensland. In the native range, genetic variability among different areas and even between geographically close populations within an area is high. In contrast, except for Madagascar, in the areas of introduction, intra-area diversity is negligible, and there is very little genetic variation among areas, suggesting a clonal mode of propagation. In Madagascar, genetic variability is intermediate between the level of variability observed in other areas of introduction and that observed among localities in the native range. Genetic diversity of R. alceifolius in native and introduced ranges The strong contrast in genetic variability between the range of introduction and the native range is in accord with predictions that introduction of a small number of founders and their expansion in a new area will lead to a strong reduction in genetic variation (Husband & Barrett 1991). This reduction will be stronger in weedy plants that are predominantly selfing or apomictic than in outbreeding species (Brown & Marshall 1981). Genetic diversity within the complex (Mauritius/ Mayotte/Queensland) is not higher than that within each of these areas taken alone, suggesting that the plants of these three areas are a single clone. Moreover, the representation of the area of introduction in the phenogram, showing three major groups (Madagascan individuals forming three related clusters, La Reunion island, and Mayotte/Mauritius/Queensland, the two last groups being branched with a bootstrap of 100), tends to show that two clones, related to genotypes of some Madagascan plants, invaded other Indian Ocean islands: one established in Mayotte and Mauritius (and later in Queensland), the other in La Réunion. It is interesting to note that in Madagascar, where genetic diversity is relatively high, R. alceifolius is not especially invasive. In contrast, in La Réunion, Mauritius, Mayotte, and Queensland, where this bramble is a serious weed, genetic variation is much lower. There may be a causal connection between diversity and invasiveness, in the sense that a particular well-adapted genotype to a specific biotope may propagate through an area very quickly via asexual reproduction. Possible scenarios of introduction of R. alceifolius in the Indian Ocean islands The relatively high genetic variability of Madagascan individuals in the area of introduction leads to the suggestion that the plant may first have been introduced here. In light of our results, we can propose three possible scenarios of introduction of R. alceifolius. A first possible scenario is that this bramble was first introduced in Madagascar from southeast Asia, perhaps on a single occasion. Our results (Fig. 3a,b) show some genetic proximity among individuals from Madagascar and from Vietnam. This could indicate that Vietnam may be the most probable source of Madagascan Rubus, but data using codominant markers are necessary to confirm this result. Contemporary genetic diversity in Madagascar may be explained by a historically remote introduction of several individuals, with an increase in genetic diversity in the long period since introduction. We can imagine an introduction at the time of migratory flows of humans coming from Malaysia, to colonize this island (it has been suggested that sugarcane, Saccharum officinarum, was introduced in this way in Madagascar, about 400 ad (Dahl 1951). The second possibility is that there were several
10 452 L. AMSELLEM ET AL. independent introductions of R. alceifolius in Madagascar, with genetic diversity among introductions. In this context, it is perhaps significant that the Madagascan distribution of this bramble is on the eastern coast, always near Tamatave, a big trading port with a long history of commerce with Asia. In addition to climatic factors (the eastern coast is by far the most humid part of the island), the distribution of this bramble might reflect the possible points of introduction. A third possibility is the introduction of a few founders from southeast Asia, which reproduced sexually and generated genetically variable offspring. From Madagascan populations, two clones separately invaded other Indian Ocean islands, their propagules being carried either by birds or, perhaps more likely, by humans. Loss of genetic variability in the Indian Ocean islands Genetic diversity of R. alceifolius in Madagascar, while considerable, is still reduced in comparison with that observed in the native range. Introduction of R. alceifolius into the Indian Ocean islands from source populations in Madagascar would have led to a further population bottleneck and restriction of genetic diversity. Thus, the reduced genetic diversity in each of these populations is not surprising. Barrett & Richardson (1986) define the parameters most frequently met for invading plants, and their possible significance. First of all, an inevitable character is a process of colonization resulting from a strong reproductive potential by asexual means (= apomixis, vegetative propagation (cuttings, layering, suckering)). Then, a second significant character seems to be a high ploidy level (Chondrilla juncea, weed in Australia (Bergman 1952; Cuthbertson 1974), the water hyacinth, Eichhornia crassipes invading Asian rivers (Barrett 1982) and Spartina anglica (Thompson 1991)). High polyploidy levels allow the plant to diversify its genome (adaptive plasticity; Roose & Gottlieb 1976), and to reduce the cost of inbreeding. However, the monoclonality of populations on each of the Indian Ocean islands except Madagascar is difficult to explain. Even if each population originated from a single founder, recombination in sexual reproduction should produce some genetic diversity, even in selfed or inbred progeny. Our results suggest that reproduction of the plant in these islands may be entirely asexual. The fact that within La Réunion island monoclonality is observed at all spatial scales in extensive sampling further suggests that spreading may occur not only via clonal growth but also perhaps via apomictically produced seeds dispersed over longer distances. Several possible explanations may account for such a monoclonality within these localities. First, several genotypes may have been introduced. This species is considered to have therapeutic properties and has also been used as an ornamental plant or considered as a botanical curiosity. Intentional introduction of plants probably has a long history on La Réunion, and this plant may well have been introduced several times independently. Among the genotypes initially present, a particularly aggressive genotype in the Réunionese environment may have been fixed by some combination of competition and genetic drift. This genotype may have been one that possessed traits adapted for colonization, such as apomixis. A second possibility is that only a single genotype, already present in Madagascar, colonized La Réunion. In the unsaturated communities (Carlquist 1966) of this oceanic island, lacking natural enemies and many potential competitors of this plant, it may have spread with extraordinary speed. It is likely that R. alceifolius was introduced in La Réunion around 1850; 50 years later it was already considered as a serious weed. As already mentioned, R. alceifolius is less invasive in Madagascar than in other areas of introduction. An important question is whether the plant underwent changes in its biology, either in Madagascar or during the process of colonization, that made it more invasive, or whether its invasiveness in Mayotte, the Mascarenes, and Queensland is simply due to the different ecological context. A comparative survey of ploidy level and reproductive biology (for example, breeding system, occurrence of apomixis) in the native range and in different areas of introduction is necessary to address this question. It seems in any event probable that the immediate source of introduced populations in the other Indian Ocean islands was Madagascar. In each of the areas of introduction our sampling covered most of the zones where R. alceifolius is found, and we believe our sampling is representative of the real genetic diversity in each area. However, no genetic variation was detected in the insular localities and in Queensland, in strong contrast to Madagascar, where variability was intermediate in relation to that found in Asia. Moreover, the fact that Madagascar is placed as a sister group of all other areas of introduction supports the hypothesis that the plant was first introduced into Madagascar which then served as the source for subsequent introductions. Our results suggest that the probable source of Australian R. alceifolius was another of the Mascarenes islands, namely Mauritius. Perfect identity between R. alceifolius in these two areas, in regard to the hypervariable markers used, may have a historical explanation. Indeed, Queensland and Mauritius are two old English colonies. There were great migratory flows (labour and trading, sugar in particular) from Mauritius towards Australia (A. Kirk, personal communication). This bramble could have been introduced on purpose (as a plant with therapeutic properties) or by accident (seeds) in Australia. This scenario,
11 GENETIC DIVERSITY OF R. ALCEIFOLIUS 453 or the date (1950) proposed by E. Bruzzese (personal communication), could explain why the Queensland populations are not genetically differentiated from Mauritius populations. Confrontation of the marker used with systematics The results obtained with the molecular marker used in this study show strong congruence with the interspecific relationships suggested by taxonomy based on morphological characters. The separation between Rubus species from Malachobatus and those from the subgenus Idaeobatus is strongly supported by AFLP markers. Within the Rubus spp. from the subgenus Malachobatus, AFLP markers differentiated several species. Moreover, R. moluccanus and R. alceifolius, considered as sister species, are well separated by these markers. Moreover, it is interesting to notice that the Madagascan Rubus, considered to be R. roridus by Vaughan (1937), is placed firmly within the cluster of all introduced R. alceifolius. This supports the contention that R. roridus is synonymous with R. alceifolius, of which it may be an ecotype (Kalkmann 1993). Taken together, the results show the usefulness of AFLP markers for addressing questions at the level of conspecific populations, such as infraspecific biogeography, and relationships between closely related species. Conclusions and perspectives This study of genetic diversity has given insights on the probable origins of introduced Rubus alceifolius. Moreover, the genetic patterns obtained between the native range and the areas of introduction may suggest a notable change in biology of this species during its route towards the Indian Ocean, conferring on this weed a high degree of invasibility in the area of introduction. These changes could affect the reproductive biology of R. alceifolius, as implied by the presence of two clones in the Indian Ocean islands (except in Madagascar), which propagate mainly through asexual reproduction. A comparative study of reproductive biology, with a codominant molecular marker (microsatellites) will be performed from germinated seeds from introduced or native populations of this bramble, to test the relative frequencies of outcrossing, selfing, and apomixis in the native range and the areas of introduction. This future work will detect if there has been a switch towards a higher frequency of asexual reproduction in the area of introduction. In a study of genetic diversity with variable number tandem repeat (VNTR) DNA probes on four populations of R. moluccanus in the Philippines, Busemeyer et al. (1997) failed to detect evidence of apomixis, and showed that relatively high gene flow exists between three populations separated by distances ranging from 115 to 245 km. Moreover, significant differences can exist among populations. From these previous results, it would be interesting to compare the genetic status of R. alceifolius with its sister species in their native area. Similarly, cytogenetic studies of R. alceifolius in Asia and in the Indian Ocean islands will be undertaken to test the hypothesis of a change in ploidy level between these two geographical areas. A higher level of ploidy could help to explain the invasive character of R. alceifolius in the area of introduction. Nybom (1986) investigated 11 species of Rubus from the subgenus Malachobatus (R. alceifolius was not represented). All of them were polyploid (from octoploid; 2x = 28 94). Knowing that in the genus Rubus hybridization between closely related species is very frequent and gives viable and fecund offspring (Nybom 1986, 1988; Richards 1986), it is possible that introduced R. alceifolius are of hybrid and of higher ploidy origin, as are many introduced weeds in colonizing situations (Barrett & Husband 1990; Thompson 1991). On La Réunion island, there is a great need for limitation, and if possible reversal, of the spread of R. alceifolius through open areas, by biological control. Our study of genetic diversity on La Réunion, compared with other areas where R. alceifolius has been introduced, and within populations in its native range, will help us to evaluate the genetic status and the reproductive biology of this weed, and will provide data required prior to any eventual programme of biological control. Acknowledgements The authors wish to thank Yan Linhart (University of Colorado, Boulder, USA), John Thompson (CEFE, Montpellier, France), Patrice David (CEFE, Montpellier, France), Doyle McKey (CEFE, Montpellier, France), and two anonymous reviewers for their helpful comments on the manuscript. Pimpun Sommartya (National Biological Control Research Centre, Thailand), Gilles Chaix (CIRAD, Madagascar), Ambroise Dalecky (CEFE, Montpellier, France), and Roch Desmier de Chenon (CIRAD, Indonesia) are thanked for providing plant material. Nathalie Faure (INRA, Montpellier, France) is thanked for her able assistance in reading autoradiographies. Ange-Marie Risterucci (CIRAD, Montpellier, France) and Laurent Grivet (CIRAD, Montpellier, France) are thanked for their helpful advice on data analysis. Jean-Marie Cornuet is thanked for providing us with the bootstrap program. This work was financially supported by La Réunion district, through a contract with the CIRAD (Marché REG ). References Antonius K, Nybom H (1995) Discrimination between sexual recombination and apomixis/automixis in a Rubus plant breeding programme. Hereditas, 123, Asker S (1979) Progress in apomixis research. Hereditas, 91, Barrett SCH (1982) Genetic variation in weeds. In: (eds Charudattan R, Wlajker H) Biological Control of Weeds with Plant Pathogens, pp John Wiley, New York.
12 454 L. AMSELLEM ET AL. Barrett SCH (1996) The reproductive biology and genetics of island plants. Philosophical Transactions of the Royal Society of London, 351, Barrett SCH, Husband B (1990) The genetics of plant migration and colonisation. In: (eds Brown AHD, Clegg MT, Kahler AL, Weir BS) Plant Population Genetics, Breeding, and Genetic Resources, pp Sinauer, Sunderland. Barrett SCH, Richardson BJ (1986) Genetic attributes of invading species. In: (eds Groves RH, Burdon JJ) Ecology of Biological Invasions: an Australian Perspective, pp Australian Academy of Science, Canberra. Bergman BT (1952) Chromosome morphological studies in Chondrilla juncea and some remarks on microsporogenesis. Hereditas, 38, Bousquet J, Simon L, Lalonde M (1990) DNA amplification from vegetative and sexual tissues of trees using polymerase chain reaction. Canadian Journal of Forestry Research, 20, Brown AHD, Marshall DR (1981) Evolutionary changes accompanying colonisation in plants. In: (eds Scudder GEC, Reveal JL) Evolution Today, Proceedings of the Second International Congress of Systematic and Evolutionary Biology, pp Hunt Institute for Botanical Documentation, Carnegie-Mellon University, Pittsburgh. Busemeyer DT, Pelikan S, Kennedy RS, Rogstad SH (1997) Genetic diversity of Philippines Rubus moluccanus L. (Rosaceae) populations examined with VNTR DNA probes. Journal of Tropical Ecology, 14, Carlquist S (1966) The biota of long-distance dispersal. I. Principles of dispersal and evolution. Evolution, 41, Carr GD, Powell EA, Kyhos DW (1986) Self-incompatibility in the Hawaiian Madiinae (Compositae): an exception to Baker s Rule. Evolution, 40, Cordemoy EJ (1895) Flore de L île de la Réunion. F. Klincksieck, Paris. Cuthbertson EG (1974) Seed development in Chondrilla juncea L. Australian Journal of Botany, 22, Dahl OC (1951) Malgache et Maanjan: une comparaison linguistique, pp Egede-Instituttet i hovedkommisjon hos Arne Gimnes forlag, Oslo. Day RW, Quinn GP (1989) Comparisons of treatments after an analysis of variance in ecology. Ecological Monographs, 59, Escaravage N, Questiau S, Pornon A, Doche B, Taberlet P (1998) Clonal diversity in a Rhododendron ferrugineum L. (Ericaceae) population inferred from AFLP markers. Molecular Ecology, 7, Evans KJ, Symon DE, Roush RT (1998) Taxonomy and genotypes of the Rubus fruticosus L. aggregate in Australia. Plant Protection Quarterly, 13, Felsenstein J (1993) PHYLIP (Phylogeny Inference Package), Version 3.5c. Distributed by the author. Department of Genetics, University of Washington, Seattle. Figier J, Soulères O (1991) Le problème des exotiques envahissantes. Bois et Forêts des Tropiques, 229, Friedmann F (1997) Rosaceae. In: (eds Bosser J, Cadet T, Marais W, Guého J) Flore des Mascareignes. La Réunion, Maurice, Rodrigues, Vol. 81, pp. 1 11, ORSTOM, Paris. Gustafsson A (1943) The genesis of the European blackberry flora. Acta Universitatis Lund, 39, Hedges SB (1992) The number of replications needed for accurate estimation of the bootstrap P value in phylogenetic studies. Molecular Biology and Evolution, 9, Howarth DG, Gardner DE, Morden CW (1997) Phylogeny of Rubus subgenus Idaeobatus (Rosaceae) and its implications towards colonisation of the Hawaiian islands. Systematic Botany, 22, Husband BC, Barrett SCH (1991) Colonisation history and population genetic structure of Eichornia paniculata in Jamaica. Heredity, 66, Jolivet P (1984) Phaedon fulvescens Weise (Col. Chrysomelidae chrysomelinae), un auxilliaire possible dans le contrôle des Rubus aux tropiques. Bulletin de la Société Linnééne de Lyon, 7, Kalkmann C (1993) Rosaceae. In: Flora Malesiana, Series I, Spermatophyta: Flowering Plants, Vol. 11 (Foundation Flora Malesiana), pp Rijksherbarium, Leiden, the Netherlands. Lavergne R (1978) Les pestes végétales de l île de La Réunion. Info-Nature Ile de la Réunion, 16, McMullen CK (1987) Breeding systems of selected Galápagos Islands angiosperms. American Journal of Botany, 74, Nybom H (1986) Chromosome numbers and reproduction in Rubus subgen. Malachobatus. Plant Systematics and Evolution, 152, Nybom H (1988) Apomixis versus sexuality in blackberries (Rubus subgen. Rubus, Rosaceae). Plant Systematics and Evolution, 160, Owadally M (1980) Some forest pests and diseases in Mauritius. Revue agriculture sucrière, de l île Maurice, 59, Quere P (1990) Le problème des pestes végétales sur l île de La Réunion. Rapport de l Office National Des Forêts. ONF, France. Richards AJ (1986) Plant Breeding Systems. Allen G. Unwin, London. Rivals P (1960) Les espèces fruitières introduites à La Réunion (Notes historiques et biologiques). In: Travaux du laboratoire forestier de Toulouse, Vol. 1, Art. 3. Laboratoire forestier, Université de Toulouse. Roa AC, Maya MM, Duque MC, Tohme J, Allem AC, Bonierbale MC (1997) AFLP analysis of relationships among cassava and other Manihot species. Theoretical and Applied Genetics, 95, Rohlf FJ (1993) NTSYS-PC: Numerical Taxonomy and Multivariate Analysis System, Version Exeter Publishers, Setauket. Roose ML, Gottlieb LD (1976) Genetic and biochemical consequences of polyploidy in Tragopogon. Evolution, 30, Russell JR, Weber JC, Booth A, Powell W, Sotelo-Montes C, Dawson IK (1999) Genetic variation of Calycophyllum spruceanum in the Peruvian Amazon Basin, revealed by amplified fragment length polymorphism (AFLP) analysis. Molecular Ecology, 8, SAS (1989) SAS/STAT User s Guide, Version 6, 4th edn, Vol. 1. SAS Institute, Cary. SAS (1996) SAS/STAT User s Guide, Release SAS Institute, Cary. Sharma K, Knox MR, Ellis THN (1996) AFLP analysis of the diversity and phylogeny of Lens and comparison with RAPD analysis. Theoretical and Applied Genetics, 93, Sigala P, Lavergne C (1996) Les plantes exotiques envahissantes des forêts de l île de La Réunion: perspectives de lutte biologique. In: Ixème Symposium International Sur le Contrôle Biologique des Mauvaises Herbes (21 26 Janvier) Afrique du Sud. ONF, Paris. Sokal RR, Sneath PHA (1963) Principles of Numerical Taxonomy. Freeman, San Francisco. Soulères O (1990) Problème des exotiques envahissantes à La Réunion. ONF Bulletin Technique, 21,
13 GENETIC DIVERSITY OF R. ALCEIFOLIUS 455 Thompson JD (1991) The biology of an invasive plant. What makes Spartina anglica so successful? Bioscience, 41, Vaughan RE (1937) Catalogue of the Flowering Plants in the Herbarium Mauritius. Bulletin of the Mauritius Institute, 1, Webb CJ, Kelly D (1993) The reproductive biology of the New Zealand flora. Trends in Ecology and Evolution, 8, Zabeau M, Vos P (1993) Selective restriction fragment amplification: a general method for DNA fingerprinting. European Patent Application, publication no: EP This study is part of Laurent Amsellem s PhD thesis, conducted under the supervision of Martine Hossaert-McKey, on the comparison of genetic diversity and reproductive biology of Rubus alceifolius in its area of introduction and its native range. This study is being undertaken preparatory to initiation of a biological control programme in La Réunion island against this weed, with the support of local authorities.