Release from foliar and floral fungal species does not explain the geographic spread of naturalized North American plants in Europe

Mark van Kleunen* and Markus Fischer

Institute of Plant Sciences, University of Bern, Altenbergrain 21, CH-3013 Bern, Switzerland

Summary 1. During the last centuries many alien species have established and spread in new regions, where some of them cause large ecological and economic problems. As one of the main explanations of the spread of alien species, the enemy-release hypothesis is widely accepted and frequently serves as justification for biological control. 2. We used a global - plant host distribution data set for 140 North American plant species naturalized in Europe to test whether alien plants are generally released from foliar and floral , whether they are mainly released from pathogens that are rare in the native range, and whether geographic spread of the North American plan t species in Europe is associated wi th release from fungal pathogens. 3. We show that the 140 North American plant species naturalized in Europe were released from 58% of their foliar and floral fungal pathogen species. However, when we also consider fungal pathogens of the native North American host range that in Europe so far have only been reported on other plant species, the estimated release is reduced to 10.3%. Moreover, in Europe North American plants have mainly escaped their rare, pathogens, of which the impact is restricted to few populations. Most importantly and directly opposing the enemy-release hypothesis, geographic spread of the alien plants in Europe was negatively associated with their release from fungal pathogens. 4. Synthesis. North American plants may have escaped particular fungal species that control them in their native range, but based on total loads of fungal species, release from foliar and floral fungal pathogens does not explain the geographic spread of North American plant species in Europe. To test whether enemy release is the major driver of plant invasiveness, we urgently require more studies comparing release of invasive and non-invasive alien species from enemies of different guilds, and studies that assess the actual impact of the enemies. Key-words: biological invasions, enemies, fungal pathogens, invasiveness, mildews, plant inva­ sions, rust fungi, smut fungi

The enemy-release hypothesis is now one of the most pro­ Introduction minent hypotheses in invasion biology (Hierro el al. 2005; During the last centuries many alien species have successfully lnderjit el al. 2005) and is frequently used as one of the main established and spread in new regions, where some of them justifications for classical biological control programs cause large ecological and economic problems (Nentwig (Mitchell & Power 2003; Hajek 2004). Although most of these 2007). An explanation for the success of these alien species programs have failed (Babendreier 2007), the few that have has been offered by the enemy-release hypothesis (Elton 1958; been successful have frequently been regarded as evidence Crawley 1987), which postulates that successful alien species supporting the enemy-release hypothesis (Crawley 1987; but have a fitness advantage over species native to the new see Keane & Crawley 2002). To date, the strongest support for region because alien species have been fully or partly released the enemy-release hypothesis comes from biogeographical from their natural enemies, including a nd studies that compare infestation or damage by enemies pathogens. between native and invasive populations of the same species (Colaulli el al. 2004; Liu & Stiling2006). Most of these studies, *Correspolldellce author. E-mail: [email protected] however, included only widespread invasive alien plant species 386

and no non-successful alien species. The latter are, howevet; non-noxio us and non-invasive a lien species were excluded also li ke ly to be released from natural enemies, and therefore, from the data set. Noxious weeds were, however, not released it remains largely unknown whether the establishment success from more fungal pathogens than non-noxious species and spread of ali en species is associated with release from (Mitchell & Power 2003), suggesting that fungal pathogen enemies. release might not be the major determinant of noxiousness There are two notable exceptions to the general lack of tests and invasiveness of a lien plants. Finally, even if alie n plants of the actual association between enemy release and invasive­ have escaped some of their fungal pathogens, it is li kely that ness or spread of alien plant species. Cappuccino and Carpenter they have mainly escaped the pathogens that are rare in their (2005) assessed damage by herbivores for nine invasive and native range. Although rare pathogens might be highly virulent nine non-invasive alien plant species in Ontario and found the in local populations, it is unclear to what extent the loss of rare invasive ones significantly less damaged than the non-invasive enemies wi ll result in an effective release from damage for species. These results are in line with the enemy-release their host (Colautti et al. 2004). For these reasons the ro le of hypothesis. However, the limited number of species studied release from fungal pathogens as a cause of plant invasiveness and the lacking assessment of plant damage by herbivores in might have been overestimated. the native range of the species precludes generalizations on Here we analyse fungus- host plant distribution data for release from herbivores. 140 North American plant species reported as established in In another inspiring study, Mitchell and Power (2003) Europe. T hi s allows us to test whether the results reported by tested the release from viral pathogens and from foli ar and M itchell and Power (2003) for European plants natura lized in floral fun gal pathogens of 473 E uropean plant species that North America are of a more general nature, and to take have establi shed (i .e. have become natura lized) in North potential biases into account. In particular, we accounted for America. By using a global fungus-host plant distribution potential bias due to differences in sampling effort between data base, these authors showed' that, on average, 84% fewer Europe and North America by adding to the European fungal fungal species infect a plant species in its naturalized North pathogens of a host, the ones of its native North American American range than in its native European range. Such a range that have been reported in Europe but so far only on large degree of release from pathogenic fungi is surprising other plant species. Furthermore, we accounted for sizeof the because spores of pathogenic fungi might easily be imported native di stribution of the host species, which is likely to be accidentally with plant material (Hulme et al. 2008). Moreover, positively associated with the number of pathogens (Mitchell spores of some fungal pathogens might cover huge - even & Power 2003), and for minimal residence time of the host transatl antic - distances by wind currents (Bowden et al. plant in Europe, which is likely to be positively associated 197 1; Nagarajan & Singh 1990). Mitchell and Power (2003) with geographic range size of the host plant (Pysek & Jarosik also showed that among the 45 declared noxious weeds (i.e. 2005) . We addressed the fo llowing specific questions: (i) Are primarily species with a negative impact on agriculture) and North American plant species released from their floral and among the 92 declared invasive plant species in their data set, fo liar fungal pathogen species in Europe? (ii) Are North the proportion of United States in which a species was American plant species mainly released from their rare declared noxious and the proportion of United States in which a pathogen species? (iii) Is geographic spread of the North species was declared invasive were positively associated with American plant species in Europe associated with their the release from fungal pathogens. As a consequence, this release from fungal pathogen species? study is the most widely cited empirical study in support of the enemy-release hypothesis, and has considerably contributed to general acceptance of enemy release as a key determinant Methods of the spread of alien plants. However, it is not known whether these results can be generali zed to target regions other than DAT A COLLECTION North America. In our analysis, we included all 140 plant species that are natura li zed T he results apparently supporting the enemy-release (i.e. have established wi ld populations) in Germany (i .e. Central hypothesis deserve a careful closer look (Colautti et al. 2004). Europe) according to the BioFlor data base (www.ufz.de/biolflor/ T he reported high rate of release of European plants from index .jsp; Kl otz el at. 2002) and are native to (i .e. originate from) fungal pathogens in North America might simply reflect an North America according to the USDA Plants data base (http:// overall higher sampling effort in Europe than in North America plants. usda.gov). We restricted o ur study to the species of the due to a higher density of mycologists and a longer hi sto ry of BIOLFLOR data base, because these species share broadly similar mycology in E urope. Moreover, because a li en plants are climatic preferences, and because for most of the exotic species in this data base the minimum resid ence time is known. Like Mitchell and relatively new to the invaded habitat, one cannot exclude the Power (2003), we used the USDA Fungus- Host Distribution data possibility that mycologists have under-sampled alien com­ base (http://nt.ars-grin .gov/fungaldatabaseslfungushost/fungushos1.cfm) pared to native species. T his would imply that a lien plant to assess the number of fo liar a nd floral fu nga l pathogens reported species might have accumulated more pathogens than we are o n each of the 140 plant species and th e number of records of each currently aware of. F urthermore, the positive associations of fungus- host plant combinatio n in North America and in Europe. noxiousness and invasiveness with release from funga l path­ The funga l pathogens included rust (Basidio I11 yco ta-rusts) and SI11 ut ogens (Mitchell & Power 2003) were o nly fo und when the (Basidiomycota -sI11 uts) fungi a nd powdery (Erysiphales), downy 387

(Peronosporales) and black (Meliolales) mildews. Because the USDA plant species between North America and Europe (i.e. the absolute Fungus- Host Distribution data base does not include data from two release from funga l pathogens). Additionally, to correct at least partly authoritative compilations of Central European (Gaumann 1959) for a potential under-sampling of the alien species, we analysed our and British (Wilson & Henderson 1966) rusts, we also added data for estim ate of minimum release from fungal pathogens in which we also rust pathogens observed on the 140 host plants from these two considered those fun gal pathogens of the native North A merican compilations (also see Mitchell & Power 2003). host range that in Europe so far have only been reported on other For each of the plant species, as a measure of geographic spread, plant species. To correct in both analyses for the number of fungal we assessed the number of European geographic regions in which pathogens recorded on a plant species in North America, we the species is established from the recently completed DAISI E data included that number in the model before the absolute release from base (www.europe-aliens.org/; Lambdon el al. 2008). The number of fungal pathogens. To assess whether the association between geo­ geographic regions is a crude estimate of the naturalized range graphical spread and pathogen release is robust, we performed size of a species, beca use it does not account for the local abundance additional analyses in which we corrected for several variables that of species. However, in our data set the number of Euro pean might affect geographical spread of the North American host plants geographic regions occupied by a species correlates positively in Europe. Because the number of fungal pathogens found on a species (Spearman's p = 0.693,11 = 140, P < 0.001) with the number of 10' in North America may depend on the species' range size there (Clay longitude x 6' latitude grid cells the species occupies in Germany 1995), and because plants with a large native range size might have (www.florawebde). This indicates that our measure of geographic been more frequently introduced to Europe (i.e. have a higher prop­ spread is robust. agule pressure; Pysek el al. 2004), we included the logarithm of the As a measure of native range size, we in cluded the number of number of continental United States and Canadian provinces and continental United States and Canadian provinces and territories in territories in which the host plant occurs as a covariate in the model. which the species occurs from the USDA Plants data base. Fo r 134 It has been suggested that only species that perform better in their of the 140 plant species, we obtained data on minimum residence natura li zed than in their native range should be considered invasive time in Europe from the BioFlor data base (Klotz el al. 2002), (Hufbauer & Torchin 2007). Therefore, an additional b enefi t of and supplemented data for species missing in BioFlor from the correcting spread in the natura li zed range for spread in the native Catalogue of alien plants of the Czech Republic (Pysek el al. range is that our analysis conforms to this definition of invasiveness. 2002), the NOBA NIS data base (www.nobanis.org/) and the DAISIE Furthermore, because the size of the natura li zed range may increase data base. with time since introduction and might depend on the taxonomic affinity of species, we included the logarithm of minimum residence time as a covariate and taxonomic subclass as a factor, respectively, ANALYSES in the model. Because data on the number of fungal pathogen species per plant To assess whether net effects of release were mainly due to the species in North America and Europe were not normally distributed, escape from North American fungal pathogens or the gain of new we compared them with the nonparametric Wilcoxon's-signed-rank European fungal pathogens, we also ran logistic regressions analysing test for paired data. To test whether the degree of release differed escape from North American pathogens and gain of new E uropean between the five groups of fungal pathogens, we used the nonpara­ pathogens separately. metric Friedman test for non-independent data . Both nonparametric tests were done with the statistical software SPSS, version 15 (SPSS Inc. 2006). Results To test whether North American plant species are more likely to On average, 58.0% fewer fungal pathogen species were be released from their rare funga l pathogen species than from their recorded for the 140 North American plant species in their common fungal pathogen species in Europe, we used logistic regression natura lized European range than in their native North Amer­ as implemented in the stati stical software GENSTAT, 9th edition ican range (Fig. I a; Wilcoxon's-signed-rank tes t: Z = - 6.634, (Payne el al. 2005). In this analysis, the binomial response variable 11 140, P < 0.001). The release of North American plants was the European presence (yes, no) of each fungal pathogen that = had been recorded on the host in North America, and the independen t was significantly dilTerent between the five groups of fungal 2 variable was the number of records of each of these fungus- host pathogen species (Friedman test: X 4 = 70.9 1, P < 0.00 I). plant combinations in North America. Because each host plant may North American plants were significantly released from smut have more than one associated fungal pathogen, we corrected for fungi (- 85.9%; Wi1coxon's-signed-rank test: Z = - 5.978, this non-i ndependence by in cluding ' host species' as a random n = 140, P < 0.001), rust fungi (- 72.0%; Z = -4.131, n = 140, factor in the analysis. P < 0.001) and downy mildews (- 75.4%; Z = -4.548, 11 = 140, To test whether the release from fungal pathogens is associated P < 0.001), but there was no signifi ca nt release from powdery with the geographic spread of the North American plants in Europe, mildews (+3.1%; Z=-O. 125, 11 = 140, P = 0.901 ; Fig. SI in we also used logistic regression . In this analysis, the response variable Supporting Information). Black mildews were only recorded wa s the number of European regions in which the species has estab­ in North America, but only on three of the 140 species, and as lished nat ura li zed popUlations. To analyse the number of regions as a consequence the rel ease from black mildews was not signi­ a bi nomial variable in the logistic regression, we set the bin omial total to 52, which equals the number of geographic regions covered ficant (Z = - 1.342, n = 140, P = 0.180; Fig. S I). by th e DAISIE data base for terrestrial plants (i.e. effect ively, for When we considered the estimate of minimum pathogen each plant species we had presence-absence data for 52 regions, and release rather than the reported release by adding to the the results apply to the proportion of European regions in which a European fungal pathogens of a host, the ones of its native plant species has established). The main independent va riable of N orth American range that have been reported in Europe, but interes t was the difference in the number of fun ga l pathogens on a so far only on other plant species, the average release from 388

(a) 1.5 25 • • I. • ea." "a. • ::l 20 e • w :0 1.0 .. w • .£ -o ,-c rt-! .. I ... iii ... 0 'Ciii • .c; 00 • 15 O.c; C :§c .- 0 '0, " 0 ~ (/J ~ c - :0 0.5 .2 10 'COl 0 ~ § • c • 'C ~15 .... t: « "u "0 c .... -. a. 0.0 i": .' Q) V> ..... a: "i!! a. ....• • ~. ". .. -0.5

Reported no. fungi on host in North America o 10 20 30 40 50 No. records of fungus on host in North America

Fig. 2. In Europe, North American plant species are mainly released (b) 25 from their rare funga l pathogen species. Each dot indicates the estimated likelihood of the presence of the fungus on its host in

Q) a. Europe, presented as an adj usted proportion value. These va lues e 20 deviate from the original binomial proportions 0 (not present) and I :0 w (present), because they have been adj usted for host identity. The line .£ Vi is the fitted relationship from the logistic regression. 0 .c; c 0 '0, c .2 F ig. S I). The release from black mildews did not change, 0 c but there was no significant release from downy mildews "iil '"E anymore (- 8.2%; Z= - 1.387, 11= 140, P=0. 166; Fig. SI). J!! 0 Moreover, there was even a significant gain in powdery 0.. mildews (+73.4%; Z = - 5.417,11 = 140, P < 0.001; Fig. SI). This indicates that despite a significant, but small , release of North American plant species from fungal pathogens in 25 Europe, enemy-release does not apply to a ll types of fungal Reported no. fungi on host in North America pathogens, and that there is potentially an enemy-gain of Fig. I. Release of North American plant species from fungal some types of fungal pathogens. pathogens in Europe. (a) In Europe. North American plant species North American plants naturalized in Europe were on appear to be strongly released from funga l pathogen species based on average much more released from their rare fungal pathogens, act ually recorded fungus- host plant combinations. (b) Release from pathogens appears much sma ller when in addition to the funga l that is, from pathogens with lower numbers of records on the pathogens recorded on a North American host in Europe, we added host in North America, than from their more common fungal those of its native range that have been reported in the naturalized pathogens (Fig. 2; quasi-Fl.474 = 4.35, P = 0.038). This might range, but so far on ly on other plant species. The size of the bubbles imply that the release from fungal pathogens might have little is proportional to the number of species wi th that combination of impact on the range dynamics of the host species. numbers of funga l pathogen species in North America and Europe. The bold lines are the fitted relationships from regression and are Geographic spread of the North American plant species, shown as illustration of fungal pathogen relea se. Release from fungal measured as the proportion of European regions in which pathogens is apparent as the fitted lines are below the bisecting lines they have establi shed naturalized populations, was signi­ indicating identical numbers of fun ga l pathogens in North America ficantly negatively associated with fungal pathogen release and Eu rope. (quasi-Fwi = 8.40, P = 0.004), after correction for the number of fungal pathogens recorded on a plant species in North America (quasi-FI.IJI = 3.74, P = 0.055). The negative fungal pathogens decreased to 10.3% (Fig. I b; Z = - 2.039, association between geographic spread a nd fungal pathogen 11= 140, P = 0.041). The release of North American plants release remained significant (Fig. 3a; quasi-Fl.I2o = 6.89, was still significantl y different between the five groups of P = 0.0 I 0) a fter correction for the positive effects of size of fungal pathogens (Friedman test: x; = 99.96, P < 0.001). native range (quasi-FI.I 2o = 23. 12, P < 0.00 I) a nd minimum These a na lyses yielded a small er, but still significant, release residence time (quasi-Fl.I2o = 20.07, P < 0.001) on size of from smut fungi (- 30.6°;(l; Z = - 4.565, 11 = 140, P < 0.001) the naturalized range and for plant taxonomic subclass and rust fungi (- 39. 1%; Z = - 2.004,11 = 140, P = 0.045; (quasi-F9,12o = 1. 88, P = 0.061). 389

Fig. 3. European geographic spread of North American plant species is negatively correlated with release from fungal pathogens. (a) Release based on actually recorded number of fungal species on host species in North America and Europe. (b) Similar to (a), but with those pathogen species added to the European fungal pathogens of a host, which have been recorded on the host in its native range and also have been recorded in the naturalized range, but so far only on other plant species. Proportions of European regions in which the plant species have established have been adjusted for taxonomic plant subclass, size of the native range, minimum residence time and number of fungal pathogen species recorded on the host in North America. The lines are the filted relationships from logistic regressions.

Net release from fungal pathogens is composed of the reported release of 84% fewer fungal pathogen species on escape from North American fungal pathogens and the European plant species in North America (Mitchell & Power accumulation of new European fungal pathogens. Geographic 2003). Possibly, this difference in the rate of release from funga l spread of the North American plant species in Europe was pathogens reflects a higher sampling intensity in Europe not significantly associated with escape from North American compared to North America rather than a true biological fungal pathogens (Fig. S2a; quasi-Fl.I2o = 1.84, P =0.177) , pattern. In an additional analysis for 122 species that have whereas it was positively associated with the gain of new established in both Europe and North America but originate European funga l pathogens (Fig. S2b; quasi-Fl.I 2o = 7.22, from other continents, we found that, on average, 54.2% more P = 0.008). This indicates that the negative association be­ fungal pathogen species were reported on each of these tween spread and fungal pathogen release of North American species in Europe than in North America. This suggests that plant species in Europe was mainly driven by the gain of new there is a sampling bias between North America and Europe. pathogens in Europe. When we correct the 84% release reported by Mitchell & When we considered the estimate of minimum enemy Power (2003) for this sampling bias, the actual release is re lease rather than the reported release by adding to the red uced to 75%. European fungal pathogens of a host, the ones of its native Despite the apparently higher sampli ng intensity in range that have been reported in the naturalized range, but so Europe, the observed release of North American plant species far only on other plant species, the negative association from fungal pathogen species in Europe could also be due to remained (Fig.3b; quasi-Fl.I 2o = 10.59, P = 0.001). Moreover, a general sampling bias against alien compared to native also when we restricted the analysis to the 69 North American plan t species. In other words, some of a host's fungal pathogen species that are categorized as natural-area invaders in species might be present in the naturalized range but might Europe (so-called agriophytes; K lotz et al. 2002), the negative simply not have been reported for the host species yet association between geographic spread and release from (Colauui et al. 2004). Indeed, when we additionally considered fungal pathogens remained significant (quasi-FI.56 = 10.91 , fungal pathogen species of the native, North American, host P = 0.002). These resu lts indicate that, based on total loads of range that in Europe so far have only been reported on other fungal species, re lease of North American plant species from plant species, the average release from fungal pathogens fungal pathogens in Europe does not explain geographic decreased to 10.3%. Even if some of these fungal pathogen spread of these plant species. species that occur in the hosts' native ranges do truly not occur on the hosts in Europe yet, it is likely that the hosts will encounter them in the future, and thus that enemy release will Discussion decay over time. The primary assumption of the enemy-release hypothesis is RELEASE FROM FUNGAL PATHOGENS that alien organisms are released from their enemies. However, On average, 58.0% fewer fungal pathogen species were recorded the rate of release might depend on the nature of the enemies, on the 140 plant species in their natura li zed, European, range in particular on their degree of host specialization (Mliller­ than in their native, North American, range. Although this Scharer el al. 2004). While there was a significant release from indicates a large release from fungal pathogens of North smut and rust fungi, there was no significant release from American plants in Europe, it is considerably smaller than the downy mildews, and even a potential gain in powdery mildews 390

(Fig. S I). It is notoriously difficult to classify organisms as guilds of enemies such as soil pathogens, nematodes, bacteria, specialists or generalists (Novotny et al. 2002), and particularly and foliar and floral fungal pathogens. so for pathogens that may be associated with several hosts (implying that they are generalists), but have a high host DEPENDENCE OF ESCAPE ON RARITY OF THE FUNGAL specificity for some of their life cycle stages (Callan & Carris PATHOGEN 2004). Nevertheless, the potential gain in powdery mildews is probably due to the more generalized nature of powdery Fungal pathogens that are rare in the native range of a host mildews compared to the other groups of pathogenic fungi in plant might be as virulent as or even more virulent than our analyses (Callan & Carris 2004). Our results for enemy widespread pathogens, but the impact of rare pathogens on release from different groups of fungal pathogens clearly show population dynamics is more localized and therefore less that despite an overall release from enemies, enemy-release likely to drive the range dynamics of the host species in the does not apply to all types of enemies. native range. In the naturalized range these rare pathogens In addition to foliar and floral fungal pathogens, plants are might, however, be as important as or even more important affected by many other guilds of enemies, such as soil pathogens, than widespread pathogens in driving the range dynamics nematodes, bacteria, viruses and herbivores that might control when the naturalized range is still small, while this is less likely plant range dynamics. Mitchell & Power (2003) showed that the case when the naturalized range is already large. There­ European plant species have also been released from viruses fore, it is unclear whether the release from rare pathogens will in North America but to a lesser extent than from fungal drive range dynamics of the alien plant species (Colautti e/ al. pathogens. Unfortunately, there are no large enemy-host 2004). We found that North American plant species naturalized plant distribution data sets available for most of the other in Europe were on average much more released from their rare guilds of enemies. Empirical studies on single or small groups pathogens, that is, from pathogens with lower numbers of of invasive plants frequently revealed plant release from these records in North America, than from their more common other guilds of enemies (herbivores and fungal pathogens: pathogens (Fig. 2). Possibly, to some degree this association Wolfe 2002; DeWalt et al. 2004; soil pathogens: Reinhart might simply reflect that pathogens that are rare in the native ef al. 2003; also see overviews in Colautti et al. 2004 and Liu range of a host are also bound to be rare in the non-native & Stiling 2006). However, it would be premature to generalize range, and as a consequence are less likely to be detected. from such a limited number of species, and because these However, the main biological consequence of their rarity in studies only include , they do not provide the native range is a lower chance to be introduced into the insight into whether the degree of invasiveness or spread is non-native range. Therefore, the detected much higher related to enemy release. release from rare than from common pathogen species Another approach to test the enemy-release hypothesis is to might imply that, overall, the release from fungal pathogens compare richness of enemies or damage by enemies between could have little impact on the range dynamics of the host invasive and native plant species. Carpenter and Cappuccino species. (2005) found that the average damage by herbivores was larger on 30 native species than on 39 exotic species in GEOGRAPHIC SPREAD OF PLANTS AND RELEASE Ontario. Van Grunsven et al. (2007) found that there was a FROM FUNGAL PATHOGENS smaller negative plant- soil feedback for three exotic species compared to three native species in the Netherlands. However, a While we found that, overall, North American plant species comparative study between 30 taxonomically paired native are released from fungal pathogen species in Europe, there and exotic plants in Ontario showed that escape generally was was large variation in the degree of release among the 140 inconsistent among different guilds of enemies (Agrawal plant species, and some species were associated with even et al. 2005). Therefore, it remains open whether, averaged more fungal pathogen species in Europe than they were in over all possible enemies, enemy-release is a major driver of North America (Fig. I). Therefore, a test of the enemy-release plant invasiveness. Another open question is whether effects hypothesis should not be restricted to only investigating of release from different enemies might interact. Such inter­ whether there is a release of alien plants from enemies but also actions are likely because infection with pathogens might whether actual invasiveness or geographic spread of the host make some plant species less (Rayamajhi et al. 2006) or more species is associated with such enemy release (Mitchell & (Ericson & Wennstrom 1997) attractive to herbivores and Power 2003; Cappuccino & Carpenter 2005). Directly opposing vice versa. the predictions of the enemy-release hypothesis, the proportion Although many studies have reported release of alien of European regions in which a North American plant species plants from enemies, we are aware of only one experimental has established naturalized populations - the only available study that tested whether invasiveness is driven by enemy measure of geographic spread of these plant species in their release. Cappuccino and Carpenter (2005) found that damage new range - was significantly negatively associated with fungal by herbivores was lower for nine invasive alien plant species pathogen release after correction for the number of fungal than for nine non-invasive ones in Ontario. This is a promising pathogens recorded on a plant species in North America. approach and future studies should assess experimentally This is also the opposite result to the positive association how invasiveness is related to the degree of release from other between invasiveness and proportional release from pathogens 391

reported for European plants naturalized in North America hardly any attention in studies on plant invasiveness (Colautli (Mitchell & Power 2003). In contrast to Mitchell and Power el al. 2004). The enemy-inversion hypothesis postulates that (2003), we corrected for taxonomic subclass, minimum re­ natural enemies could have a positive effect on their hosts in sidence time and native range size. However, even when we the naturalized range because of interactions with abiotic factors did not correct for these variables, the negative association that differ from the native range or through a restructuring of between geographical spread of the North American plants in multi-species interactions (Pearson et al. 2000; Pearson & Europe and their pathogen release was significant. Moreover, Callaway 2003; Colautti el al. 2004; Parker & Gilbert 2007). when we, like Mitchell and Power (2003) did, included in the For example, Pearson el al. (2000) found that two flower-head analysis proportional pathogen release as the ratio between inhabiting gall flies introduced to North America as biological­ num ber of pathogens in Europe and the num ber of pathogens control agents of the invasive Centaurea maculosa became the in North America, which biases the data set against plant main food resource for deer mice that thereby inadvertently species without pathogens in North America, the association dispersed seeds of C. maculosa. Another hypothesis, the between geographic spread and pathogen release remained enemy-of-my-enemy hypothesis, postulates that enemies significantly negative (results not shown). Furthermore, when co-introduced with an invasive organism may have a larger we, like Mitchell and Power (2003) did, restricted our data set impact on competitors in the introduced range than on the to naturalized species categorized as natural-area invaders, original host itself (Sabel is el al. 200 I; Tompkins et al. 2003; the negative association remained significant. Therefore, it is Colautti el al. 2004). For example, the parapox that was very unlikely that the contrasting results for European plant introduced into Europe with the grey squirrel from North species in North America and North American plant species America is reducing populations of the European red squirrel in Europe are a consequence of differences in methodology. to a larger extent than the populations of the grey squirrel The negative association between geographic spread of (Tompkins et al. 2003). Clearly, understanding and counter­ North American plants in Europe and pathogen release could acting spread of alien plants very urgently requires tests of the reflect that North American plants with a restricted naturalized general importance of these currently neglected alternative range in Europe are less likely to have encountered their hypotheses. potential pathogens yet. If this were the case, one would expect a negative association between size of the naturalized Conclusions range and the number of pathogens from the native range that have so far only been reported in Europe on other plant species. Our results show that North American plants accumulate not We, however, did not find such a negative association much fewer fungal pathogen species in Europe than in their (Spearman's p = 0.142, N = 140, P = 0.093; note that the native range. Moreover, we show that the North American trend of the non-significant correlation is even positive). plant species are mainly released from their rare pathogens. Therefore, also the strength of the negative association Most importantly, we show that geographic spread of the between geographic spread and pathogen release did not North American plant species in Europe is negatively instead decrease after adding to the European fungal pathogens of a of positively associated with the release from fungal patho­ host, the ones of its native range that have been reported in gens. One obvious limitation of our study is that we assessed Europe but so far only on other plant species (Fig. 3). Thus, fungal pathogen richness while we could not assess pathogen based on total loads of fungal species, release from foliar virulence. Although it is likely that a large number of fungal and floral fungal pathogens does not explain the geographic pathogens will include some highly virulent ones, this is not spread of North American plant species in Europe. necessarily the case. Future studies should therefore assess the One possible explanation for the negative association impact of fungal pathogens in the native and naturalized between geographic spread of North American plants in ranges of successful and unsuccessful alien plant species. Europe and their release from fungal pathogens is that it Furthermore, our fungus-focused study cannot preclude that reflects a trade-off between the high expansion capacity of the predictions of the widely accepted enemy-release hypothesis some species and their ability to defend themselves against hold for other guilds of plant enemies. Our results show, enemies. If species with a high capacity for spread are more however, that, based on total loads of fungal species, release vulnerable, they are likely to accumulate more enemies in the from floral and foliar fungal pathogens does not explain non-native range. This idea corresponds well with our finding geographic spread of North American plant species in that the negative association between geographic spread and Europe. To test whether enemy release is the major driver of release mainly reflects a positive association between spread spread of alien plant species, we urgently require more studies and accumulation of new European fimgal pathogens. At the comparing release of successful and non-successful alien same time this finding indicates that the accumulated new species from other guilds of enemies. fungal pathogens hardly influence the geographic spread of their new host species. Acknowledgements While the negative association between geographic spread of alien plant species and their release from pathogen species The authors thank Beatrice Senn-Irlet and Volker Kummer ror help with the rungal pathogen categorization. Erin McCray ror data files rrom the USDA contradicts the predictions of the enemy-release hypothesis, it Fungus-I-Iost-Distribution data base. and Wim van der Putten, Ragan Calla­ supports some of the opposing hypotheses that have received way and two anonymous referees for helpful comment s on an earlier version of 392

Ihis manuscrip!. The aUlh ors acknowl edge funding by Ih e Swiss Science Foun­ Nagarajan, S. & Singh, D. V. (1 990) Long-di slance dispersion of r USI palho­ dalion (ProjecI number 3100AO-11772211) and Ihe Swiss Nalional Cenlre of gens. AIII/llal Rel'iew ~r Phy topalhology, 28, 139- 153. Competence in Research - Pl ant Survival. Nenlwig, W (2007) Biologicalllll'asiol/s. Springe r-Verlag. Berlin. Novolny, V.. Bassel. Y. . Miller, S.E .. Weiblen, G. D., Bremer, B. , C ize k, L. & Drozd, P. (2002) Low hosl spec ifi cily of herbivo rous insecls in a Iropical References fore s!. Nalllre, 416, 841 - 844 . Parker, I.M . & Gilber!. G.S. (2007) Wh en Ihere is no escape: Ih e efTecls of Agrawal. A.A .. Kolanen. P.M., Milchell . CE., Power, A.G .. Godsoe, W & natural enemies on nati ve, invas ive. and noninvasive plants. , 88. Klironomos. 1. (2005) Enemy release? An ex perimenl wilh conge neri c 1210- 1224. planl pairs and dive rse above- and belowground enemi es. Ecology. 86. 2979- Payne, R.W , Harding, S.A .. Murray. D.A. , Soular, D. M ., Ba ird. D.B.. 2989. Welham, SJ .. Kane. A.F., Gilmour, A.R. , Thompson. R .. Websler. R. & Babendreier. D. (2007) Pros and cons of biological conlrol. Biologicallllvasiolls Tunnicl ifTe Wilson, G. (2005) The Gllide to Gellslla Release 9: S ia listics Part (ed. W Nenlwig), pp. 403- 41 8. Springer Verlag, Berlin . 2. VSN Inlernalional, Oxford . Bowden. 1. , Gregory, P.H. & Johnson, CG. (1 97 1) Possible wind lranspo r! of Pearson, D.E. & Callaway, R.M. (2003) IndireclefTeclsof hosl-specific biological cofTee leaf rusl across Ihe Allanlic Ocean. Natllre. 229. 500- 50 I. conlrol agenls. Trel/ds ill Ecology & Evolllliol/. 18, 456 - 46 1. Call an. B. E. & Carris, L.M. (2004) Fun gi on living planl subslrala, including Pearson, D.E., McKelvey, K.S. & Ru ggiero. L.F. (2000) Non-Iarget etrecls of fruils. Biodiversity 0/ Fllllgi (eds G. M. Muell er, G.F. . Bills & M.S. Fosler). an inlroduced biological conlrol agenl on deer mouse ecology. Oecologia . pp. 105- 126. El sevi er Academi c Press. Burlin glon. 122, 121- 128. Cappuccino. N. & Carpenler, D. (2005) Invasive exolic pla nls suffer less Pyse k, P. & Jarosik, V. (2005) Res idence lime delermines Ihe di slribuli on of herbivo ry Iha n non-invasive exolic plan Is. Biology Lelfers. 1. 435- alien planls. II/ vasil'e Plal/ts: Ecological alld Agriclllillral Aspecls (ed. S. 438. Inderjil), pp. 77- 96. Birkhiiuscr Verlag. Base l. Carpenler. D. & Cappuccin o. N. (2005) Herbivo ry, lime since inlroduclion and Pyse k, P., Sildlo, 1. & Manditk, B. (2002) Calalogue of alien planls of Ihe Czech Ih e invasive ness of exolic planls. JOIIl"llal 0/ Ecology. 93, 315- 32 1. RepUbl ic. Prestia. 74, 97- 186. Clay. K. (1995) Correlales of palhogen species richness in Ihe grass family. Pysek , P, Richardson. D.M . & Williamson. M. (2004) Pred icling an d explain ­ Calladiall JOIIl"llal 0/ BotallY, 73, S42- S49. in g plant invasions through analysis of source area flo ra s: some critical Colaulli, R.I., Ricciardi, A., Grigorovich. I. A. & Macisaac. H.1. (2004) Is inva­ considerations. Dil'ersily al/d Disiriblliiol/s, 10, 179- 187. sion success explained by Ihe enemy release hypolhesis? Ecology Lelfers. 7. Rayamajhi, M.B. , Van , T.K., Prall, P.D. & Cenler. T.D. (2006) Inleraclive asso­ 721 - 733. ciati on between Pliccinia psidii an d Oxyops viljostl, two introduced natu ra l Crawl ey , M.J. (1987) Whal makes a communilY invasible') Colollizatioll . enemies of M elalellCll qllillqllellerl'ia in Florida. Biological COlltrol, 37. 56- S lI ccessioll alld S tabililY (cds A.J. G ray. M.1. C rawl ey & P.J. Edwards), 67. pp. 429- 453. Blackwell Science, Oxford . Reinharl, K.O., Packer, A., van der PUllen, WH. & Clay, K . (2003) Planl-soil DeWall, S.1. , Denslow, 1.S. & Ickes, K. (2004) Nalural-enemy rel ease fa ci lilales biota interactions and spatial di stribution of black cherry in its native and habilal expansion of Ihe invasive Iro pi cal shrub Clidelllia hirta. Ecology. 85. invasive ranges. Ecology Lellers. 6. 1046- 1050. 47 1-483. Sabelis, M.W , Janssen, A. & Kanl , M.R. (2001)The enemy of my enemy is my Ellon, C S. (1958 ) The Ecology 0/ III \'asiolls by Allilllals alld Plallis. Melhuen. ally. Sciel/ce, 291 , 2104 - 2 105 . London. SPSS Inc. (2006) spss Base 15. 0jar Wil/dows User's Gllide. SPSS Inc., Chicago. Eri cson, L. & Wennslrom, A. (1 997) The efTecI of herbivory on Ihe inleracli on Tompkins, D.M .. While, A.R. & Bools. M. (2003) Ecological replacemenl of between th e cl onal plant Trielltalis ellrojJa ea an d its smut fun gus UI'OC'yslis nalive red squirrels by invasive greys drive n by disease. Ecology Lellers, 6, trielltalis. Oikos, 80, 107- 111. 189- 196. Gaumann, E.A. (1959) Die Roslpilze M illelellropas lIIil Besollderer Beriich ich­ va n G runsven, R.H.A., van der PUllen. WH .. Bezemer, T.M., Tamis, WL.M .. tigllllg del' S ch",eiz. Buchdruckerei Biichler. Bern. Berendse, F. & Veenendaal , E. M. (2007) Reduced pia Ill-soil fe edback of Hajek, A. (2004) Natllral Ellelllies - All IlIlrodllctioll 10 Biological COlllrol. pl ant species expanding their range as compared to natives. Journal of Ecology, Cambridge UniversilY Press, Cambrid ge. 95, 1050- 1057. H ierro:1. L. , Maron, 1. L. & Callaway, R. M. (2005) A biogeographical approach Wilson, M. & Henderson. D. M. (1966) British Rllsi FIII/gi. Cambridge Un iver­ to plant invasions: the import ance of studyin g exotics in their int roduced sity Press, Cambridge. and nalive range. JOIIl"llal 0/ Ecology. 93. 5- 15. Wolfe, L.M . (2002) Why alien in vaders succeed : support for Ihe escape­ Hufba uer. R.A. & Torchin, M.E. (2007) Inlegraling ecological and evolulion­ from-enemy hypolhesis. Americall Narllralist, 160, 705- 711. ary Iheory on biological invasions. Biologicallllvasiolls (cd. W Nenlwig), pp. 79- 96. Springe r-Verlag, Berlin . Hulme. P.E. , Bacher, S. , Keni s, M., Klolz, S., Kiihn, I. , Minchin. D. et al. (2008) Grasping al Ihe roules of bi ological invasions: a framewo rk for inlegraling palhways inlo po li cy. JOIIl"llal 0/ Applied Ecology. 45. 403- 4 14 . Supporting Information Inderj il, Cadolle, M. W & Colaull;, R.I. (2005) The ecology of biological in va­ sions: past, present an d futu re. In vasive Plants: Ecological and Agriculrural Additional Supporting Information may be found in the Aspects (cd. Inderjil), pp. 19- 43. Bi rkhiiuser Verlag. Basel. online version of this article: Kea ne, R.M. & Crawley, M.J. (2002) Exolic planl invasions and Ihe enemy release hypolhesis. Trellds ill Ecology & Evoilltioll , 17. 164- 170. Fig, S1. Release of North American plants from different K 10lZ, S.. Kiihn, I. & Durka. W. (2002) BIO I.FI.OR - eille Darellballk : 11 Biologisch­ Okologische M erkllllliell der Gejiis;pjlllllzell ill Delltsclillllld. Biindesaml fiir groups of fungi. Naturschutz, Bonn . Lambdon, P.W , Pyse k, P , Basnou, C. Hejda, M. Aranoul so u. M., Essi. F. Fig, S2, Spread of North American plant speci es in Europe el al. (2008) Alien fl ora of Eu rope: species diversil Y, lempora l Irends. in relation to escape from native fun gi and accumulation of geographical pallem s and research needs. Preslia, 80, 101 - 149. Li u, H. & Sliling, P. (2006) Tesling Ihe enemy release hypolhesis: a review and new fungi. mela-analysis. Biologicaill/I'asiol/s. 8, 1535- 1545. Milchell , CE. & Power. A.G. (2003) Re lease of invasive planls from fun gal and viral palhogens. Natllre, 421 .625- 627. Miiller-Scharer, H .. Schafrne r. U. & Slein ge r. T. (2004) Evolul ion in in vasive plant s: implicati ons for bi ologic,d control. Ii'ends ill Ecology & Evolwion, 19, 417-422.