Naturwissenschaften (2012) 99:883–892 DOI 10.1007/s00114-012-0970-9

ORIGINAL PAPER

Alkaloid concentration of the invasive plant species in relation to geographic origin and herbivory

Benjamin Hornoy & Anne Atlan & Michèle Tarayre & Sébastien Dugravot & Michael Wink

Received: 3 July 2012 /Revised: 10 September 2012 /Accepted: 13 September 2012 /Published online: 25 September 2012 # Springer-Verlag Berlin Heidelberg 2012

Abstract In the study of plant defense evolution, invasive genistae-tinctoriae were negatively correlated to concentration plant species can be very insightful because they are often of the QA lupanine. Quinolizidine alkaloid concentration was introduced without their enemies, and traits linked to defense very variable between individuals, both within and among can be released from selective pressures and evolve. Further, populations, but it was not different between native and invaded studying plant defense evolution in invasive species is impor- regions, suggesting that no evolution of decreased resistance tant for biological control and use of these species. In this study, occurred after gorse lost its enemies. Our study also suggests we investigated the evolution of the defensive chemicals qui- that QA concentrations are traits integrated into seed predation nolizidine alkaloids (QAs) in the invasive species gorse, Ulex avoidance strategies of gorse, with plants that mass-fruit in europaeus. Using a common garden experiment, our goals were spring but do not escape pod infestation in time being richer to characterize the role of QAs relative to specialist enemies of in QAs. gorse and to investigate if QA concentration evolved in invaded regions, where gorse was introduced without these enemies. Keywords Biological invasions . Defense syndrome . Our results showed that pod infestation rate by the seed predator Gorse . Plant– interaction . Quinolizidine alkaloids ulicis and infestation by the rust pathogen Uromyces

Communicated by: Sven Thatje Introduction Electronic supplementary material The online version of this article (doi:10.1007/s00114-012-0970-9) contains supplementary material, Plants have developed a large diversity of defense mechanisms which is available to authorized users. to cope with herbivores. Among these mechanisms, defensive B. Hornoy (*) : A. Atlan : M. Tarayre chemicals such as secondary metabolites are widespread and ECOBIO, Université de Rennes 1, can confer resistance to a wide spectrum of enemies, from Campus de Beaulieu, microorganisms to vertebrates. For example, alkaloids have 35042 Rennes Cedex, France been shown to be feeding deterrent and/or toxic to bacteria, e-mail: [email protected] fungi, plants, , nematodes, mollusks, and vertebrates S. Dugravot (Wink 1998). Many other traits may also be involved in resis- BIO3P, Université de Rennes 1, tance, such as physical defenses (e.g., thorns), escape in time Campus de Beaulieu, (e.g., fruiting phenology), and indirect defenses (e.g., volatiles 35042 Rennes Cedex, France attracting natural enemies of herbivores; Wink 1988; Agrawal M. Wink and Fishbein 2006). Plant defense strategies are thus better Institut für Pharmazie und Molekulare Biotechnologie, viewed as a suite of traits that act together against herbivory Universität Heidelberg, (i.e., a defense syndrome), rather than single traits (Agrawal and Im Neuenheimer Feld 364, 69120 Heidelberg, Germany Fishbein 2006; Carmona et al. 2011). The expression and evolution of these defensive traits are driven by environmental Present Address: factors such as resource availability and the composition of the B. Hornoy herbivore community. In particular, the reciprocal selection Institut de Biologie Intégrative et des Systèmes, Université Laval, 1030 Avenue de la Médecine, pressure between a plant species and an associated herbivore Québec, QC, Canada can lead to coevolution (Janzen 1980;Thompson2005), 884 Naturwissenschaften (2012) 99:883–892 especially when their relationship is very specific and the impact other reduce seed predation by a mass flowering resulting in of the two species on each other is strong (Gandon 2002). For predator satiation (Tarayre et al. 2007; Atlan et al. 2010). example, Berenbaum and Zangerl (1998) found a very close However, these flowering strategies do not explain all the match between populations of wild parsnip (Pastinaca sativa, variation observed in pod infestation rates (Atlan et al. Apiaceae) to produce defense compounds and the ability of the 2010). Other defenses must therefore be involved, and they parsnip webworm Depressaria pastinacella (Oecophoridae) to likely include quinolizidine alkaloids (QAs), which are char- metabolize them. The absence or loss of herbivores can thus acteristic of some tribes of the Fabaceae (Wink 1993; Wink et greatly affect the evolution of plant traits that are part of a al. 2010). Indeed, these compounds are known to have a defense strategy (e.g., Janzen 1975; Benkman 1995). defensive role against a wide array of herbivores, including An evolutionary decrease of costly defense traits is expected pathogens and phytophagous insects (reviewed in Wink 1992, in the absence of herbivores and pathogens (Blossey and 1998). In addition, QAs are produced in green parts of the Nötzold 1995; Müller-Schärer et al. 2004; Orians and Ward plant, are transported through the phloem, and accumulate in 2010). This situation is often observed in invasive plants be- epidermal cells and in fruits, thus being strategically posi- cause most of them have been introduced without their herbi- tioned to act against herbivores, among which phloem sap vores and pathogens, experiencing the so-called enemy release feeders, pathogens, and seed predators (Wink 1992;Winkand (Keane and Crawley 2002; Mitchell and Power 2003;Liuand Roberts 1998). Outside its native range, U. europaeus was Stiling 2006). The evolution of chemical defenses in invasive introduced without its herbivores and pathogens and is not plants depends on their cost and on the relative abundance of attacked by any local herbivore. Seed predators and other specialist and generalist enemies in the introduced range com- specialist herbivores from the native range were introduced pared to the native range (Müller-Schärer et al. 2004;Orians later for biological control in some invaded regions, such as in and Ward 2010). As a consequence, a decrease of defense traits New Zealand, where U. europaeus was present before 1835 has been observed in some but not all invasive species and where E. ulicis was introduced in 1931 and C. succedana (reviewed in Bossdorf et al. 2005, and Orians and Ward 2010). in 1992 (Hill and Gourlay 2002). In other regions (e.g., In this context, the evolution of plant defensive chemistry in Reunion Island), there is no biological control program, and invasive species is not only an important topic in the study of gorse plants are still free from their specialist enemies. evolution of defense and plant–herbivore interactions but also The main goal of the study was to investigate if QA concen- for biological control (Müller-Schärer et al. 2004)oruseof tration in gorse evolved after enemy release in invaded regions. these species (e.g., as fodder). Common gorse, Ulex europaeus To address this issue, we first described the pattern and variation (Fabaceae), is a long-lived spiny shrub that can reach more than of QAs in gorse and their ecological relevance relative to its 4 m height. It is native to Western Europe but has been intro- natural enemies. Specifically, the following questions were duced by Europeans, mainly since the beginning of the studied: (1) do gorse individuals richer in QAs show lower nineteenth century, in several regions of the world where it attack rates by their main natural enemies? (2) Does QA con- often became invasive, including New Zealand, Australia, centration differ between native and invaded regions? To an- South and North America, and the islands of Hawaii and swer these questions, we used a common garden experiment Reunion (Holm et al. 1997). In Europe, gorse is attacked by with individuals from native and introduced gorse populations several pathogens and herbivores. However, adult gorse plants and measured the QA concentration in their shoots, their life are protected from large grazers by their thorns and are only history traits, and the level of infestation by herbivores. attacked by small specialists, of which the most common and harmful are the specific Exapion ulicis (Coleoptera: Apionidae) and to a lesser extent the moth Cydia succedana Materials and methods (Lepidoptera: Tortricidae; Barat et al. 2007, 2008). Exapion ulicis spends its whole life cycle on gorse, feeding on gorse Experimental design shoots and flowers. Females lay their eggs inside young pods in spring, and larvae develop at the expense of seeds (Davies Studied gorse plants come from two native regions, Brittany 1928; Barat et al. 2007). Females of C. succedana lay eggs (France) and Scotland (UK), and from two invaded regions, outside the pods and the emerging larvae chew the pod wall to Reunion Island (Indian Ocean) and New Zealand (Pacific enter the pods and develop at the expense of the seed. Several Ocean). Population genetic analyses showed that gorse from other enemies attacking the vegetative parts of gorse were also these two invaded regions likely came from the northern described (e.g., aphids, spider mites, and fungi; Zwölfer 1962). part of Western Europe (France and UK; Hornoy 2011). In its native range, gorse resistance against seed predation Between 1999 and 2005, seeds were collected on an indi- was shown to be partly achieved through a genetically based vidual basis on up to 30 individuals per population, in three polymorphism of flowering strategies: some plants escape populations in each of these four regions. Seed dormancy attack in time by flowering before the infestation peak, while was broken by a cold treatment (4 months at 4 °C) and Naturwissenschaften (2012) 99:883–892 885 manual scarification, which resulted in a germination rate above was suspended in a known volume of methanol or in 100 μL 80 %. Furthermore, seedling mortality was very low (below of acetone containing an internal standard, respectively (see 5 %). Seed dormancy or seedling mortality could thus not have below). Given that pod wall samples resulted in very small substantially affected the results of the experiment. In October amounts of powder, extraction, dilution, and injection vol- 2006, one seed per mother plant was randomly chosen and the umes were reduced accordingly. resulting seedlings were grown for 1 year in the greenhouse. In November 2007, ten seedlings per population were randomly Alkaloid identification To identify QAs in our samples, some chosen and randomly transplanted in the common garden lo- representative shoot and pod wall alkaloid extracts were studied cated on the Campus of Rennes (Brittany, France), an area with by gas liquid chromatography–mass spectrometry (GLC-MS), several nearby gorse populations and prone to natural infesta- since this method is well suited to analyze complex mixtures of tion by herbivores and pathogens. The common garden thus QAs (Wink 1993). Instruments and conditions were the follow- contained N0120 gorse individuals (four regions × three pop- ing: Hewlett Packard, Series II 5890 Gas Chromatograph ulations × ten individuals). See Hornoy et al. (2011)forfull equipped with an OV-1 column (30 m×0.25 mm, bonded details on the setup of the common garden. 0.25 μm), injection volume 2 μL, split ratio 1:30, injector In gorse, the vegetative season occurs in late spring and 250 °C, and carrier gas helium. Mass spectrometer (Finnigan summer. The young shoots produced in September are thus soft MAT, SSQ 700) was in positive mode, scan time 0.5 s, first and very vulnerable to attacks by natural enemies. We therefore mass/lastmass40/450,electronenergy70eV,andionsource collected green shoots from every individual plant in the garden temperature 175 °C. The temperature program was 140 °C, in September 2009. Shoot samples were oven-dried for 4 days 10 min isothermal, 140–200 °C with 6 °C/min, 200–248 °C at 55 °C. Dried samples were ground to a very fine powder with4°C/min,248–300 °C with 12 °C/min, then 6 min iso- using a planetary ball mill (RS 100, Retsch). Finally, powder thermal. Mass spectral data and retention indexes were com- samples were stored in dry tubes (containing silica gel), in a pared with previous data (Wink 1993) to identify QAs dark cold room (ca 5 °C) until alkaloid extraction. To check if unequivocally. the alkaloid pattern of individuals was consistent between con- secutive years, we also collected shoots in September 2010 on Alkaloid quantification QAs were routinely quantified by 12 individuals (from populations of the four studied regions). capillary GLC using a GC (Varian 3400) equipped with an Finally, to check the correlation between QA concentration in OV-1 30 m × 0.25 mm column and the following conditions: shoots and in pods, we collected pods in June 2011 on nine injection volume 2 μL, injection temperature 250 °C, and individuals belonging to the four regions studied, and display- detector temperature 300 °C. Temperature program was the ing contrasted concentrations of QAs in shoots. Ten uninfested same as for the GLC-MS analysis (see above). Using this pods were collected per plant, and their walls were processed method, we were able to quantify three main alkaloids out of the same way as the shoot samples. the six QAs detected: lupanine, 5,6-dehydrolupanine, and ana- gyrine. Gas chromatography was performed both with tetraco- Alkaloid analysis sane(AldrichChemCo)asaninternalstandardineachsample and with lupanine as an external standard. The three other Alkaloid extraction Alkaloids were extracted following a detected QAs (cytisine, N-methylcytisine, and N-formyl- protocol outlined in Wink (1993). For each individual, 5 g cytisine) were not quantifiable on all the individuals of dried shoot powder (when available) was suspended in because of other compounds in gorse extracts having 30 mL 0.5 M HCl and was left under continuous stirring similar retention times. We nevertheless assayed their overnight at ambient temperature. The suspension was then concentration from the GLC-MS chromatograms on a centrifuged for 12 min at 3,000×g, and the supernatant (i.e., subsample of the individuals. the alkaloid extract) was collected and made alkaline (to pH All the chromatograms were analyzed with PeakSimple

12–14) with NH4OH. The alkaline extract was added to an 3D (SRI Instruments). Peak areas were used to quantify the extraction column filled with a matrix that adsorbs water individual alkaloids. Concentrations of individual QAs that (Varian Chem Tube Hydromatrix). After adsorption, the col- were quantified with the internal standard (tetracosane) and umn was progressively eluted with 90 mL methylene chloride. the external standard (lupanine) were nearly perfectly cor- The eluate was collected in a flask and the solvent was then related to each other (RSp>0.95 for the three assayed QAs, evaporated in a rotavapor at 45 °C. The crude alkaloid extract N0116, P<0.001). Consequently, we used the lupanine was dissolved in a small volume of methylene chloride (ca. standard for quantification. Because anagyrine concentra- 1.5 mL) and transferred to a sample vial. Finally, the solvent tion was an order of magnitude higher than the two other was evaporated from the vial by delivering a steady stream of alkaloids, we calculated a standardized “total” alkaloid con- nitrogen gas, and the vials were stored at −20 °C until analysis. centration (QAsum), which was the sum of the standardized Just before identification or quantification, the crude extract concentrations of the three assayed alkaloids. 886 Naturwissenschaften (2012) 99:883–892

Measurements of herbivory Statistical analysis

We measured herbivory by the weevil E. ulicis that feeds on Comparisons of alkaloid concentrations between popula- shoots and whose females also feed on pod walls before tions and regions were performed using a two-level nested laying eggs within pods, by the moth C. succedana that ANOVA in which populations (random effect) were nested feeds on pod walls to enter the pods and eat the seeds, by the within regions (fixed effect). The effect of plant traits and aphid Aphis ulicis (Homoptera: Aphididae) that feeds on QAs on pod infestation rates was assessed with GLMs on phloem sap, by the spider mite Tetranychus lintearius (Ac- rank-transformed variables. QA concentrations were rank- ari: Tetranychidae) that feeds on gorse shoots, and by the transformed before performing these analyses because data rust fungus Uromyces genistae-tinctoriae (Pucciniaceae; were not normally distributed (Conover and Iman 1981). Mireille Jourdan, personal communication) that affects Correlations involving alkaloid concentrations or plant gorse green shoots. Herbivory was measured from Septem- traits measured in the common garden were estimated using ber 2008 to July 2009. For each gorse individual in the Spearman’s rank order correlation coefficient. All the anal- common garden, we determined infestation by , yses were performed using the R software (R Development moths, aphids, spider mites, and rust. Pod infestation rate Core Team 2010). was estimated in late June 2009, when the highest propor- Due to some missing data in trait measurements (for tion of individuals was fruiting synchronously. We opened example if pod production was too low to estimate pod 30 ripe pods per individual (when available) and recorded infestation rate), sample sizes may vary slightly between the number of insects. A pod was considered as infested by different analyses, depending on the traits analyzed. They E. ulicis weevils if it contained weevils at any developmen- are reported in tables, figure captions, and when the results tal stage (from to adult) or weevil parasitoids. A pod of statistical tests are given. was considered as infested by C. succedana if it contained a moth larva or when its presence was revealed by droppings or holes. Infestation rates were estimated when at least ten Results ripe pods were available. Natural infestation by the rust fungus U. genistae-tinctoriae, the aphid A. ulicis, and the Phytochemistry spider mite T. lintearius was qualitatively recorded (i.e., presence or absence) for each individual during the time of GLC and GLC-MS revealed the presence of anagyrine, and the experiment. to a lesser extent cytisine and N-methylcytisine, as major QAs, and lupanine, 5,6-dehydrolupanine, and N-formylcy- Measurements of life history traits tisine as minor QAs (Table 1). Across our gorse shoots samples, we thus detected six QAs of two different types: For every individual, plant height was measured in October the lupanine type and the α-pyridone type (Table 1). These 2008, the date of flowering onset (the date at which the first six compounds were found in samples from the four studied flowers were fully open) was estimated during the flowering regions. We also detected ammodendrine, which is a piper- season lasting from September 2008 to May 2009, and pod idine alkaloid often found along with QAs. Single QA density (the number of pods produced per centimeter of concentrations varied between 0 and 541 μg/g dry weight, shoot, averaged over five random shoots) was determined with mean concentration across all individuals being be- in May 2009. See Hornoy et al. (2011) for full experimental tween 2.51 and 46.37 μg/g (Table 1). details.

Table 1 Concentration of the quinolizidine alkaloids detected Concentration (μg/g dry weight) in gorse shoots Alkaloid Type Number Range Mean±SD

Lupanine Lupanine 116 0–89 2.88±8.87 5,6-Dehydrolupanine Lupanine 116 0–19 3.01±3.25 Anagyrine α-Pyridone 116 0–541 46.37±91.32 Cytisine α-Pyridone 15 0–13 15.99±33.43 nd not determined N-formylcytisine α-Pyridone 15 0–12 2.51±3.78 aAmmodendrine is not a quino- N-methylcytisine α-Pyridone 14 0–68 12.21±17.29 lizidine alkaloid but it is often Ammodendrinea Piperidine nd nd found along with them Naturwissenschaften (2012) 99:883–892 887

Lupanine (µg/g) Variation of QA concentration

20 Native range Invaded range Concentration of the three assayed QAs were highly corre- 16 lated with each other (lupanine × 5,6-dehydrolupanine: 12 RSp00.69; lupanine × anagyrine: RSp00.61; 5,6-dehydrolu- 8 panine × anagyrine: RSp00.58; in all three cases N0116 and P<0.001). For the 12 individuals sampled both in 2009 and 4

2010, QA concentrations were correlated between years 0 BCC BCV BKE SBA SCR SST RLB RMA RPB ZAU ZCH ZWE (lupanine: RSp00.42; 5,6-dehydrolupanine: RSp00.53; ana- 0 gyrine: RSp 0.54), but due to lack of statistical power, these 5,6-dehydrolupanine (µg/g) high correlations were not significant. Consistently, the val- 8 ues of the standardized sum of these three QAs (QAsum) 6 were significantly correlated between 2009 and 2010 (RSp0 0.66, N012, P00.02). Furthermore, QA concentrations were similar between 2009 and 2010, with the slope of the 4 linear regression being close to 1, and the intercept being 2 close to 0 relative to the range of QAsum (linear regression: QAsum(2010)00.88+0.71 × QAsum(2009), N012). 0 BCC BCV BKE SBA SCR SST RLB RMA RPB ZAU ZCH ZWE Variation in QA concentration among individuals was very high (Table 1); the coefficient of variation being larger than Anagyrine (µg/g) 100 % for the six detected QAs, and even reaching 308 % for lupanine. At the among-population level, variation was large for 200 anagyrine and 5,6-dehydrolupanine (Fig. 1). For these two 160 QAs, the population effect was significant or nearly so, but no 120 significant difference existed between the four regions (Table 2). 80 For lupanine, variation occurred mainly among regions (Fig. 1) 40 but not between populations within regions (Table 2). QAsum 0 revealed the same pattern with a marginally significant region BCC BCV BKE SBA SCR SST RLB RMA RPB ZAU ZCH ZWE effect, but no population effect (Table 2). When region effects are present, the significant difference is mainly due to the QAStandardisedsum sum highest QA concentration in populations from Brittany com- 10 pared to the three other regions (Fig. 1). Our results thus do not 8 show any difference in the QA concentration of gorse shoots 6 between the two native regions and the two invaded regions. 4

Correlation between QA concentrations and herbivory level 2

0 Seed predators Maximum QA concentrations of pod walls in BCC BCV BKE SBA SCR SST RLB RMA RPB ZAU ZCH ZWE 2011 were higher than those observed in shoots in 2009 (lupanine, 0–947 μg/g; 5,6-dehydrolupanine, 0–779 μg/g; Fig. 1 Quinolizidine alkaloid concentration in shoots of gorse indi- anagyrine, 19–3,620 μg/g; N09 in all cases). However, the viduals grown in the common garden. Means are given with 1 SE. N0 variations in concentration of the lupanine-type QAs between 116. The first letter in each population ID refers to its region of origin: B for Brittany, S for Scotland, R for Reunion, and Z for New Zealand. shoots and pods were strongly and significantly correlated Second and third letters are population names (lupanine: RSp00.86, P00.003, N09; 5,6-dehydrolupanine: RSp00.78, P00.017, N09). In particular, individuals that concentrations as a proxy for variation in pod wall QA con- showed the highest concentrations in shoots also showed the centrations, in order to study the effect of QAs on pod infes- highest concentrations in pods (Fig. S1). The correlation, tation rates. although still positive and quite strong, was weaker and not Overall, pod infestation rate by seed predators was 25 % significant for anagyrine and thus for the standardized sum in spring 2009 and 43 % in spring 2010 and was correlated

(anagyrine: RSp00.53, P00.15, N09; QAsum(2009): RSp00.52, between years (RSp00.73, P<0.001, N0100). For pod in- P00.16, N09). The correlation between pod wall QA con- festation rate by weevils, the GLM with lowest AIC in- centration and shoot QA concentration, in particular for volved lupanine concentration, pod density, and plant lupanine-type QAs, allows us to use variation in shoot QA height as significant explanatory variables. Pod infestation 888 Naturwissenschaften (2012) 99:883–892

Table 2 Results of the nested ANOVA testing for differences in regression on the left side and on the right side of the quinolizidine alkaloid concentration in gorse shoots between popula- breakpoint. The breakpoint we found was a lupanine tions and regions concentration of 1.358 μg/g. Using the same model as Populations Regions above, we found that below this threshold concentration, lupanine was not correlated with infestation rate by Number d.f. F d.f. F weevils (slope0−0.24, P00.22). Above it, lupanine con- Lupanine 116 8 0.94 3 6.62* centration was negatively correlated to the infestation 0− 0 5,6-Dehydrolupanine 116 8 1.78# 3 1.76 rate by weevils (slope 0.92, P 0.01). Anagyrine 116 8 2.15* 3 2.28 # Parasites of vegetative parts In September 2008, 39 % of QAsum 116 8 1.67 3 3.38 gorse individuals were rust-infested. In May 2009, 59 % of # P<0.10; *P<0.05 individuals were infested by aphids. QA concentration in 2009 did not differ between plants infested by aphids and rate by moths was best explained by a model containing plants devoid of aphids, neither for single QAs nor for QAsum only pod density and plant height. (Mann–Whitney test, N059 and N057, respectively, U> These results confirmed the influence of plant height and 1,544 and P>0.05 in all cases). Plants that were resistant to pod density on pod infestation rate by weevils and moths rust infestation showed the highest QA concentrations (weevils: regression slopes were −0.29 and −0.21, P values (Fig. 3). This pattern led to a significant difference in lupanine were 0.004 and 0.03, respectively; moths: slopes were 0.16 concentration between rust-resistant individuals and rust- and 0.20, P values were 0.07 and 0.02, respectively). These infested individuals: median lupanine concentration was results also suggest that lupanine concentration significantly 1.32 μg/g (N071) and 0.68 μg/g (N045), respectively and negatively affected pod infestation rate by weevils (re- (Mann–Whitney test, U02,004.5, P00.02). There was no − 0 gression slope was 0.26, P 0.004). The two other QAs such difference with the other single QAs or QAsum (Mann– had no significant effect on infestation by weevils. None of Whitney test, U>1,871 and P>0.05 in all cases). No attack by the three assayed QAs did significantly impact pod infesta- the spider mite T. lintearius was observed during the time of tion by moths. this experiment. The correlation between lupanine concentration and pod infestation by weevils seemed to show a threshold effect: for Correlation between QA concentrations and life history traits lower lupanine concentrations, pod infestation showed no apparent trend relative to lupanine, whereas for higher lupa- The correlation between pod density and QA concentration nine concentrations, pod infestation strongly decreases with was positive and significant for the three QAs (lupanine: increasing lupanine concentrations (Fig. 2). We thus per- RSp00.27, P00.006; 5,6-dehydrolupanine: RSp00.23, P0 formed a piecewise regression between lupanine concentra- 0.02; anagyrine: RSp00.20, P00.04; N0104 in the three tion and pod infestation rate by weevils, following the method cases). Consistently, QAsum was correlated with pod density described in Crawley (2007). This analysis aims at first to (RSp00.24, N0104, P00.02). When the date of flowering determine a breakpoint (finding the breakpoint minimizing onset (which is linked to pod density; Atlan et al. 2010) was residual deviance in the GLM) and then perform the taken as a covariable, these correlations remained similar in

Fig. 2 Relationship between Pod infestation by weevils (%) lupanine concentration in 80 shoots of gorse individuals and pod infestation rate by E. ulicis 70 weevils (N0101). The dotted 60 line represents the threshold found with the piecewise 50 regression analysis (1.358 μg/g, see text). Lupanine 40 concentrations were log- transformed before plotting to 30 reduce the size of the x-axis 20

10

0 00110 100 lupanine (µg/g) Naturwissenschaften (2012) 99:883–892 889

Lupanine (µg/g) the year, observing similar QA concentration between years 100 suggests that QAs are not very inducible and is consistent with the fact that such toxins are thought to incur low costs (Stamp 2003; Müller-Schärer et al. 2004; see also Strauss et al. 2002). 30 Comparison of QA concentration between native 15 and invaded regions 10 In addition to the large variation observed among individuals within populations, we observed some significant differences between populations or regions, depending on the QA. An effect of the populations and regions of origin was also ob- served by Boschin et al. (2008) in a study of the QA concen- tration of Lupinus albus (Fabaceae), and by Macel et al. (2004) 0 in a study of pyrrolizidine alkaloid concentration of Senecio not infested infested jacobaea (Asteraceae). However, our results contrast with the results of studies that compared native and invaded regions of Fig. 3 Lupanine concentration in shoots of gorse individuals not infested invasive species. Indeed, we did not observe significant differ- or infested by the rust pathogen U. genistae-tinctoriae. N071 and N045, respectively. Boxes display the first and third quartile (lower and upper ences between native and invaded regions, while most studies limit of the box), and the median (thick line). Each upper or lower vertical found an increase in toxin concentrations in the invaded range: bar is 1.5 times the interquartile range, and open circles are data points pyrrolizidine alkaloids in the Asteraceae S. jacobaea, Senecio outlying off this range. Lupanine concentrations were log-transformed inaequidens,andSenecio pterophorus (Stastny et al. 2005; before plotting to reduce the size of the y-axis Cano et al. 2009); flavonoids in the tree Triadica sebifera (Wang et al. 2012); or the main glucosinolate in the Brassica- strength and significance. The correlation between the date ceae Lepidium draba (Müller and Martens 2005; but see of flowering onset or plant height and QA concentration was Cipollini et al. 2005). This change is expected when introduced not significant for any single QA nor for QAsum, even when populations escape specialists, but not generalist herbivores pod density was taken as a covariable (RSp<|0.13|, P>0.23 (Müller-Schärer et al. 2004). However, in gorse, both specialist in all cases). and generalist herbivores are lacking in the introduced range, and in this case, we expect a decrease in toxins (if their produc- tion is costly; Orians and Ward 2010). The absence of signifi- Discussion cant decrease in the two invaded regions (Reunion and New Zealand), especially when compared to Scotland, is thus con- In our gorse shoot and pod wall samples, we detected six of the sistent with the fact that no generalist enemies are present in the QAs known to be present in aerial parts of U. europaeus invaded range of gorse and that QA may not be very costly in (Maximo et al. 2006) and belonging to the lupanine type this species, although the latter remains to be investigated. In (lupanine and 5,6-dehydrolupanine) and the α-pyridone type addition, the pattern of variation of QA concentration between (anagyrine, cytisine, N-methylcytisine, and N-formylcytisine). populations and regions is consistent with the results obtained QA concentrations we found in gorse shoots ranged from for other gorse life history traits in previous studies. Indeed, 0.0005 to 0.5 mg/g dry weight, which is in the range of values large individual and population variations were observed for found in shoots/stems or leaves of several Fabaceae species, reproductive and growth traits as well as for pod infestation with QA concentrations ranging from 0.01 to 50 mg/g dry rates (Tarayre et al. 2007; Atlan et al. 2010), but their values weight (Wink and Witte 1991;Wink1992; Bermudez-Torres were independent of the native or invaded status of the region of et al. 2002;Maximoetal.2006;Bermudez-Torresetal.2009). origin (Hornoy et al. 2011). The quantitative analysis revealed a very large variation in QA concentration among individuals, even within populations Effects of QAs on gorse herbivores and pathogens (Fig. 1). The fact that gorse plants have been randomized in the common garden, and that QA concentration was correlated Effects of QAs on specialist seed predators Some of the QAs and similar between the 2 years, suggests that the observed we found in gorse (lupanine and cytisine) have been shown to variations have a genetic basis, which is consistent with other have feeding deterrent and/or toxic effects on insects such as studies on QAs (e.g., Carey and Wink 1994;WinkandCarey Coleoptera and Lepidoptera (reviewed in Wink 1992). Gorse 1994). Since the amount and nature of the attacks by phytoph- pod infestation by the weevil E. ulicis and the moth C. agous insects and pathogens were very different depending on succedana could thus be influenced by these QAs. In this 890 Naturwissenschaften (2012) 99:883–892 study, we found a negative correlation between lupanine con- By contrast, lupanine seemed to have a strong nega- centration and pod infestation by weevils, but not by moths. In tive effect on infestation by the rust U. genistae- our shoot samples, the maximum QA concentrations (between tinctoriae. This QA has already been shown to have 0.1 and 0.5 mg/g dry weight, i.e., ca. 0.01 and 0.05 mg/g fresh antifungal activities, since it inhibited conidia germina- weight) were close to QA concentrations that have been tion and further development of the powdery mildew shown to deter insect feeding (Wink 1992). This suggests that fungus Erysiphe graminis f.sp. hordei (Wippich and although QA concentration is relatively low in gorse shoots, Wink 1985). The fact that lupanine would contribute some individuals may have QA concentrations high enough to to the resistance against both weevils and rust is con- entail feeding deterrence on insects. In pod walls, the maxi- sistent with the observation that QAs are broad range mum QA concentrations were much higher than in shoots, bioregulators, being toxic to many species belonging to which is consistent with the fact that tissues closely related to very different taxonomic groups, from microorganisms fitness, such as fruits, may be more defended (e.g., Zangerl to vertebrates (Wink 1992, 1998). and Rutledge 1996). Altogether, these results suggest that deterrence of pod infestation by weevils may occur at two levels. On the one hand, plants whose shoots are rich enough Role of QAs in pod infestation avoidance strategies in lupanine could deter adult weevils and thus be less pod- in U. europaeus infested in spring. On the other hand, pod lupanine concen- tration could directly deter oviposition by females. While lupanine seems clearly involved in the pod infestation Interestingly, the nature of the QA was more important avoidance strategy of U. europaeus, pod infestation rate than its concentration. Indeed, lupanine that showed low varied widely, even for individuals with similar low lupa- concentrations had the strongest link with pod infestation, nine concentrations. This suggests that other traits are in- while anagyrine that showed the highest concentrations had volved in resistance against seed predators. Consistently, in no measurable effect. Other QAs, such as cytisine, which gorse, avoidance of pod infestation has been previously was detected but not quantified here, may also play a role, shown to involve flowering phenology through two differ- since it has been shown to have strong insecticidal effects in ent strategies: escape in time, for plants that start producing other plant species (Wink 1992). pods before the reproductive period of seed predators, and The absence of effect of 5,6-dehydrolupanine, anagyrine, predator satiation, for plants that produce fruits massively in and low concentrations of lupanine on pod infestation is con- spring, during the oviposition period of seed predators sistent with the fact that the studied phytophagous insects and (Tarayre et al. 2007; Atlan et al. 2010). In addition to this pathogens are specialists of gorse (Barat et al. 2007;Zwölfer polymorphism of flowering phenology, the present results 1962). Indeed, specialist enemies such as insects have been suggest that there is also a polymorphism in the strategy of shown to be able to cope with toxic secondary metabolites of QA production. Indeed, while for some plants lupanine their host, among which alkaloids, by eliminating toxins with seems to be used as an efficient defense against pod infes- the feces, by sequestering them, or after a mutation at a toxin tation and herbivory, other plants produce a basal level of target site (Wink 1992; Desprès et al. 2007). However, interest- lupanine that is too low to have any effect. Furthermore, ingly, even in the specialist E. ulicis, QA concentration seems to these two types of polymorphism seem related: the effect of negatively influence the feeding and/or oviposition choice of lupanine on pod infestation increases with pod density, females. Such a situation was already found by Wink et al. suggesting that the strategy of mass-fruiting also involves (1982) with the broom aphid Aphis cytisorum, a monophagous QA-based defense. This result supports the view that plant species specialized on broom, Cytisus scoparius (see also defense against herbivores involves a suite of traits that are Agrawal and Kurashige 2003). By contrast, C. succedana part of a defense syndrome and act together against herbi- larvae’s pod choice does not seem influenced by any QA vores (Agrawal and Fishbein 2006; Johnson et al. 2009; assayed here, suggesting that this species is better adapted to Carmona et al. 2011). gorse’s QAs than the weevils or that it is deterred by other QAs.

Effects of QAs on aphids and rust The QAs quantified here seemed to have no effect on infestation by the Acknowledgments This work was funded by the Groupement de Recherche “Ecologie Chimique.” The authors thank Louis Parize, aphid A. ulicis. Indeed, there was no difference in QA Astrid Backhaus, Frank Sporer, and Armelle Racapé for technical concentration between infested and not-infested gorse assistance and Véronique Martel and two anonymous reviewers for individuals. This could reflect the fact that aphids are helpful comments on a previous version of the manuscript. able to store dietary QAs, like in A. cytisorum, Aphis Ethical standards The experiments described in this manuscript genistae,andMacrosiphum albifrons (Wink et al. 1982; comply with the current laws of the countries in which they were Wink and Witte 1991). performed. Naturwissenschaften (2012) 99:883–892 891

Conflict of interest The authors declare that they have no conflict of biological control of gorse, Ulex europaeus L., in New Zealand. interest. Biol Control 25:173–186 Holm L, Doll J, Holm E, Pancho J, Herberger J (1997) World weeds: natural histories and distribution. Wiley, New York References Hornoy B (2011) Processus évolutifs contribuant au succès envahissant de l’ajonc d’Europe, Ulex europaeus. PhD thesis, University de Rennes 1 Agrawal AA, Fishbein M (2006) Plant defense syndromes. Ecology Hornoy B, Tarayre M, Hervé M, Gigord L, Atlan A (2011) 87:S132–S149 Invasive plants and enemy release: evolution of trait means Agrawal AA, Kurashige NS (2003) A role for isothiocyanates in plant and trait correlations in Ulex europaeus. PLoS One 6(10): resistance against the specialist herbivore Pieris rapae. J Chem e26275 Ecol 29:1403–1415 Janzen DH (1975) Behavior of Hymenaea courbaril when its predis- Atlan A, Barat M, Légionnet AS, Parize L, Tarayre M (2010) Genetic persal seed predator is absent. Science 189:145–147 variation in flowering phenology and avoidance of seed predation in Janzen DH (1980) When is it coevolution? Evolution 34(3):611–612 native populations of Ulex europaeus. J Evol Biol 23(2):362–371 Johnson MTJ, Agrawal AA, Maron JL, Salminen JP (2009) Heritabil- Barat M, Tarayre M, Atlan A (2007) Plant phenology and seed preda- ity, covariation and natural selection on 24 traits of common tion: interactions between gorses and weevils in Brittany (France). evening primrose (Oenothera biennis) from a field experiment. J Entomol Exp Appl 124:167–176 Evol Biol 22:1295–1307 Barat M, Tarayre M, Atlan A (2008) Genetic divergence and ecological Keane RM, Crawley MJ (2002) Exotic plant invasions and the enemy specialisation of seed weevils (Exapion spp.) on gorses (Ulex release hypothesis. Trends Ecol Evol 17(4):164–170 spp.). Ecol Entomol 33:328–336 Liu H, Stiling P (2006) Testing the enemy release hypothesis: a review Benkman CW (1995) The impact of tree squirrels (Tamiasciurus)on and meta-analysis. Biol Invasions 8:1535–1545 limber pine seed dispersal adaptations. Evolution 49(4):585–592 Macel M, Vrieling K, Klinkhamer PGL (2004) Variation in pyrrolizi- Berenbaum MR, Zangerl AR (1998) Chemical phenotype matching dine alkaloid patterns of Senecio jacobaea. Phytochemistry between a plant and its insect herbivore. PNAS 95:13743–13748 65:865–873 Bermudez-Torres K, Robledo Quintos N, Barrera Necha LL, Wink M Maximo P, Lourenço A, Tei A, Wink M (2006) Chemotaxonomy of (2002) Alkaloid profile of leaves and seeds of Lupinus hintonii Portuguese Ulex: quinolizidine alkaloids as taxonomical markers. C.P. Smith. Z Naturforsch 57:243–247 Phytochemistry 67:1943–1949 Bermudez-Torres K, Matinez Herrera J, Figueroa Brito R, Wink M, Mitchell CE, Power AG (2003) Release of invasive plants from fungal Legal L (2009) Activity of quinolizidine alkaloids from three and viral pathogens. Nature 421:625–627 Mexican Lupinus against the lepidopteran crop pest Spodoptera Müller C, Martens N (2005) Testing predictions of the ‘evolution of frugiperda. BioControl 54:459–466 increased competitive ability’ hypothesis for an invasive crucifer. Blossey B, Nötzold R (1995) Evolution of increased competitive ability in Evol Ecol 19:533–550 invasive nonindigenous plants: a hypothesis. J Ecol 83(5):887–889 Müller-Schärer H, Schaffner U, Steinger T (2004) Evolution in inva- Boschin G, Annicchiarico P, Resta D, D’Agostina A, Arnoldi A (2008) sive plants: implications for biological control. Trends Ecol Evol Quinolizidine alkaloids in seeds of lupin genotypes of different 19(8):417–422 origins. J Agric Food Chem 56:3657–3663 Orians CM, Ward D (2010) Evolution of plant defenses in nonindig- Bossdorf O, Auge H, Lafuma L, Rogers WE, Siemann E, Prati D enous environments. Annu Rev Entomol 55:439–459 (2005) Phenotypic and genetic differentiation between native Stamp N (2003) Out of the quagmire of plant defense hypotheses. Q and introduced plant populations. Oecology 144:1–11 Rev Biol 78:23–55 Cano L, Escarré J, Vrieling K, Sans FX (2009) Palatability to a Stastny M, Schaffner U, Elle E (2005) Do vigour of introduced pop- generalist herbivore, defence and growth of invasive and native ulations and escape from specialist herbivores contribute to inva- Senecio species: testing the evolution of increased competitive siveness? J Ecol 93:27–37 ability hypothesis. Oecol 159:95–106 Strauss SY, Rudgers JA, Lau JA, Irwin RE (2002) Direct and indirect Carey D, Wink M (1994) Elevational variation of quinolizidine alka- ecological costs of resistance to herbivory. Trends Ecol Evol loid contents in a lupine (Lupinus argenteus) of the rocky moun- 17:278–285 tains. J Chem Ecol 20:849–857 Tarayre M, Bowman G, Schermann-Légionnet A, Barat M, Atlan A Carmona D, Lajeunesse MJ, Johnson MTJ (2011) Plant traits that (2007) Flowering phenology of Ulex europaeus: ecological con- predict resistance to herbivores. Funct Ecol 25:358–367 sequences of variation within and among populations. Evol Ecol Cipollini D, Mbagwu J, Barto K, Hillstrom C, Enright S (2005) Expression 21:395–409 of constitutive and inducible chemical defenses in native and invasive Thompson JN (2005) The geographic mosaic of coevolution. The populations of Alliaria petiolata. J Chem Ecol 31:1255–1266 University of Chicago Press, Chicago Conover WJ, Iman RL (1981) Rank transformation as a bridge between Wang Y, Siemann E, Wheeler GS, Zhu L, Gu X, Ding J (2012) Genetic parametric and nonparametric statistics. Am Stat 35:124–129 variation in anti-herbivore chemical defences in an invasive plant. Crawley MJ (2007) The R book. Wiley, Chichester J Ecol 100:894–904 Davies WM (1928) The bionomics of Apion ulicis Forst. (gorse wee- Wink M (1988) Plant breeding: importance of plant secondary metab- vil) with special reference to its role in the control of Ulex euro- olites for protection against pathogens and herbivores. TAG paeus in New Zealand. Ann Appl Biol 15:263–286 75:225–233 Desprès L, David JP, Gallet C (2007) The evolutionary ecology of Wink M (1992) The role of quinolizidine alkaloids in plant insect insect resistance to plant chemicals. Trends Ecol Evol 22:298–307 interactions. In: Bernays EA (ed) Insect–plant interactions, vol Development Core Team R (2010) R: A language and environment for IV. CRC, Boca Raton, pp 133–169 statistical computing. R Foundation for Statistical Computing, Vienna Wink M (1993) Quinolizidine alkaloids. In: Waterman P (ed) Methods Gandon S (2002) Local adaptation and the geometry of host–parasite in plant biochemistry, vol 8. Academic, London, pp 197–239 coevolution. Ecol Lett 5:246–256 Wink M (1998) Chemical ecology of alkaloids. In: Roberts MF, Wink Hill RL, Gourlay AH (2002) Host-range testing, introduction, and M (eds) Alkaloids: biochemistry, ecology and medicinal applica- establishment of Cydia succedana (Lepidoptera: Tortricidae) for tions. Plenum, New York, pp 265–300 892 Naturwissenschaften (2012) 99:883–892

Wink M, Carey D (1994) Variability of quinolizidine alkaloid profiles Wink M, Botschen F, Gosmann C, Schäfer H, Waterman PG of Lupinus argenteus (Fabaceae) from North America. Biochem (2010) Chemotaxonomy seen from a phylogenetic perspective Sys Ecol 22:663–669 and evolution of secondary metabolism. Annu Plant Rev Wink M, Roberts M (1998) Compartmentation of alkaloid synthesis, 40:364–433 transport and storage. In: Roberts MF, Wink M (eds) Alkaloids: Wippich C, Wink M (1985) Biological properties of alkaloids: influ- biochemistry, ecology and medicinal applications. Plenum, New ence of quinolizidine alkaloids and gramine on the germination York, pp 239–262 and development of powdery mildew, Erysiphe graminis f. sp. Wink M, Witte L (1991) Storage of quinolizidine alkaloids in Macro- hordei. Experientia 41:1477–1479 siphum albifrons and Aphis genistae (Homoptera: Aphididae). Zangerl AR, Rutledge CE (1996) Probability of attack and patterns of Entomol Gen 15:237–254 constitutive and induced defense: a test of optimal defense theory. Wink M, Hartmann T, Witte L, Rheinheimer J (1982) Interrelationship Am Nat 147:599–608 between quinolizidine alkaloid producing legumes and infesting Zwölfer H (1962) Ulex europaeus project; European investiga- insects: exploitation of the alkaloid-containing phloem sap of tions for New Zealand. Unpublished report, No. 2. Com- Cytisus scoparius by the broom aphid, Aphis cytisorum. Z Natur- monwealth Institute of Biological Control, Delémont, forsch 37c:1081–1086 Switzerland