J Chem Ecol (2009) 35:10861095 DOI 10.

1007/s10886-009-9690-9

Feeding Deterrence and Detrimental Effects of Pyrrolizidine Alkaloids Fed to Honey Bees (Apis mellifera)
Annika Reinhard & Martina Janke & Werner von der Ohe & Michael Kempf & Claudine Theuring & Thomas Hartmann & Peter Schreier & Till Beuerle

Received: 3 June 2009 / Revised: 25 August 2009 / Accepted: 9 September 2009 / Published online: 24 September 2009 # Springer Science + Business Media, LLC 2009

Abstract Recent studies have shown the occurrence of plant derived pyrrolizidine alkaloids (PAs) in retail honeys and pollen loads, but little is known about how these compounds influence the fitness of foraging honey bees. In feeding experiments, we tested a mix of tertiary PAs and the corresponding N-oxides from Senecio vernalis, pure monocrotaline, and 1,2-dihydromonocrotaline in 50% (w/w) sucrose solutions. The bees were analyzed chemically to correlate the observed effects to the ingested amount of PAs. PA-N-oxides were deterrent at concentrations >0.2%. 1,2-Unsaturated tertiary PAs were toxic at high concentrations. The observed PAs mortality could be linked directly to the presence of the 1,2-double bond, a well established essential feature of PA cytotoxicity. In contrast, feeding experiments with 1,2-dihydromonocrotaline revealed no toxic effects. Levels of less than 50 g 1,2unsaturated tertiary PAs per individual adult bee were tolerated without negative effects. PA-N-oxides fed to bees were reduced partially to the corresponding tertiary PAs. Unlike some specialized insects, bees are not able to actively detoxify PAs through N-oxidation. To gain insight into how PAs are transmitted among bees, we tested for
A. Reinhard : M. Kempf : P. Schreier Lehrstuhl fr Lebensmittelchemie, Universitt Wrzburg, Wrzburg, Germany M. Janke : W. von der Ohe Niederschsisches Landesamt fr Verbraucherschutz und Lebensmittelsicherheit, Institut fr Bienenkunde, Celle, Germany A. Reinhard : C. Theuring : T. Hartmann : T. Beuerle (*) Institut fr Pharmazeutische Biologie, Technische Universitt Braunschweig, Braunschweig, Germany e-mail: t.beuerle@tu-bs.de

horizontal PA transfer (trophallaxis). Under laboratory conditions, up to 15% of an ingested PA diet was exchanged from bee to bee, disclosing a possible route for incorporation into the honey comb. In the absence of alternative nectar and pollen sources, PA-containing plants might exhibit a threat to vulnerable bee larvae, and this might affect the overall colony fitness. Keywords Pyrrolizidine alkaloids . Insects . Apis mellifera . Honey bee . Toxicity . Deterrence

Introduction Pyrrolizidine alkaloids (PAs) are plant defense compounds that are found regularly in four plant families within the angiosperms, i.e., the Asteraceae (tribes Senecioneae and Eupatorieae), the Boraginaceae, the Apocynaceae, and the genus Crotalaria within the Fabaceae. There are some 350 different structures known (Hartmann and Witte 1995). 1,2Unsaturated PAs are toxic and/or deterrent to insect herbivores (van Dam et al. 1995; Macel et al. 2005; Narberhaus et al. 2005). Because of this, PA-plants usually are avoided by generalist herbivores, but there are numerous examples of specialist insects of taxa that have developed specific mechanisms to deal with these potentially hazardous compounds. These adaptations range from tolerance mediated by fast excretion rates (Lindigkeit et al. 1997; Sadasivam and Thayumanavan 2003), to sequestration, safe storage, and integration into the insect-specific defense. The latter often includes sophisticated biochemical and behavioral adaptations that in leptidopterans guarantee optimal protection of the eggs, the most vulnerable stage in insect metamorphosis, from predators and parasitoids (for review see Eisner et al. 2002; Hartmann and Ober 2008).

J Chem Ecol (2009) 35:10861095

1087

The frequent occurrence of PAs in honey (RIVM 2007; Kempf et al. 2008), pollen, and pollen loads of bees (Boppr et al. 2005, 2008) shows that bees are confronted with PAs during natural foraging conditions. Honey foraged from PA-containing plants may contain up to 3.9 g PAs per gram of honey (Deinzer et al. 1977; Betteridge et al. 2005). It seems likely that the nectar of PA-plants contains PAs as well, but there is no direct confirmation. On the other hand, there are data on PAs in pollen. These suggest that pollen of PA-plants as well as pollen loads of foraging bees contain high PA concentrations. Up to 14 mg/g for Echium vulgare and 0.8 mg/g for Senecio jacobaea have been reported (Boppr et al. 2005, 2008). Since floral nectar always is contaminated with plant pollen (which may contain high amounts of PAs) it is difficult to determine the original nectar PA load. Apart from the ongoing discussion on the ecological role of toxic nectar in plant-pollinator relationships (reviewed by Adler 2000; Raguso 2008), there are numerous examples where plant alkaloids have been shown to be constituents of floral nectars (Baker 1977; Detzel and Wink 1993; Adler 2000; Nicolson and Thornburg 2007). Thus far, deterrence and mortality effects of monocrotaline on butterfly species have been investigated (Masters 1991; Landolt and Lenczewski 1993), but only the effects of a few PAs have been tested against honey bees (Detzel and Wink 1993; Tan et al. 2007). While some work on pollinator toxicity has been performed in the course of pesticide approval, there is only limited information on the mode of action for toxins from natural sources like nectar and pollen. The results of a recent pollen analysis of contaminated honeys (Kempf et al. 2008) indicate strongly that some PA-plants, like Echium vulgare are attractive bee plants. To date, there is only one report on the effect of PAs on the fitness and behavior of honey bees: Detzel and Wink (1993) have reported a LD50 of 0.1% for the PA heliotrine under no-choice conditions. The focus of this study was to investigate the effect of PAs in the bee diet on the behavior and health of bees. We were interested particularly to learn how bees handle and metabolize these hazardous plant constituents in order to escape their toxicity. Since honey bees are social insects that live in colonies of up to 80,000 individuals, we chose laboratory experiments with small caged bee groups. These tests were designed in accordance to the OECD guidelines for the testing of chemicals (OECD 1998: Honeybees, Acute Toxicity Test). Since only 1,2-unsaturated tertiary PAs but not their N-oxides are converted into toxic intermediates through bioactivation by cytochrome P450 enzymes, 1,2-unsaturated tertiary PAs are regarded as pro-toxic, while the respective

PA-N-oxides (PA-Noxs) are non-toxic per se. However, the N-oxides are reduced easily to the pro-toxic free bases in the presence of weak reducing agents. Generally, this happens in the intestines of any herbivore that feeds on a PA-plant. This is why PA-containing plants are essentially toxic although they often store PAs exclusively as N-oxides in their tissues (Hartmann and Witte 1995) (Fig. 1). 1,2-Unsaturated PA toxification by hepatic cytochromeP450 enzymes is well-studied in vertebrates, and acute, chronic, and genotoxic effects are documented (Culvenor et al. 1976; Mattocks 1986; Stegelmeier et al. 1999; Fu et al. 2004). The bioactivation is initiated by hydroxylation of the necine base followed by a spontaneous dehydratization to form pyrrolic ester structures that readily react with cellular nucleophiles (Fu et al. 2004). The structural prerequisite here is the concurrent presence of the 1,2-double bond and an allylic ester functionality. Similar mechanisms can be assumed for mutagenic and toxic effects observed in insects (Frei et al. 1992; Narberhaus et al. 2005; Hartmann and Ober 2008). Hence, the tests should shed light on the question whether there are different impacts on bees for tertiary PAs and PA-Noxs, respectively. These studies were complemented by feeding monocrotaline and 1,2-dihydromonocrotaline. These two compounds are structurally closely related, but the latter does not bear the generally required cytotoxic feature of the 1,2-double bond. By studying these compounds, it should be possible to link an observed mortality directly to the toxic principle of the 1,2-double bond. We also tested known steps of PA metabolism in insects to see whether bees have evolved mechanisms to cope with

Fig. 1 Examples of pyrrolizidine alkaloid (PA) structures and reactions

1088

J Chem Ecol (2009) 35:10861095

natural occurring PAs. In these studies, we intended to monitor elevated excretion of ingested PAs, detoxification of ingested tertiary PAs via N-oxidation, and toxification via reduction of 1,2-unsaturated PA-Noxs into 1,2-unsaturated tertiary PAs. Finally, we addressed the question to what extent trophallaxis is influenced by PA-food uptake. Inside bee colonies, a pronounced horizontal nectar transfer from bee to bee can be observed. Furthermore, foraging bees that return to the hive with poor-quality food experience difficulties in finding unloader bees that are in charge of unloading and storing the nectar in the hive. This in turn results in lower recruiting success from dance attendees to head for this poor food source. Overall, high PA content (in our sense i.e., poor quality) could trigger a lower PA intake into the hive if there were deterrence effects because of high PA concentrations.

Hydrogenation of the 1,2-double bond was performed as recently described by Marn-Loaiza et al. (2008). Further purification was achieved by column chromatography on silica gel 60 (mesh 230400) using the solvent system CH2Cl2/MeOH/NH4OHconc (83:15:2/ v:v:v). The purity of the collected fractions was checked by capillary GC and GC-MS analysis. Fractions that contained only 1,2-dihydromonocrotaline were combined. Reaction products of several reactions were combined yielding about 100 mg (average yield per reaction: 40%). Purity and identity were confirmed by GC-MS. Feeding Experiments Feeding experiments were conducted in the summer seasons 2007 and 2008 with two different bee colonies. Because of the relative early onset of fall conditions in summer 2007, it was not possible to finish all tests in time. To assure as much as possible the highest possible degree of reproducibility in the data set, we here refer to results obtained in 2008. Where possible, we complemented and confirmed the results with data from 2007. The experiments were conducted with minor modifications to the recommended OECD Guidelines for the testing of chemicals (1998). The OECD guidelines recommend one test series with 3 cages that contain ten bees for each concentration (30 bees per data point). In order to generate a more solid dataset, we conducted 3 independent test series for each compound and each concentration (equals 90 individual bees per data point). The particular test compounds (PA-mix, PA-Nox-mix, monocrotaline, and 1,2-dihydromonocrotaline) were dissolved in a 50% (w/w) sucrose solution at concentrations of 0.02, 0.2, and 2.0% (w/w). In the 2007 experiments, in addition to these concentrations, a PA level of 0.002% was tested, but omitted for subsequent experiments because it did not display any effects. In a preliminary experiment, PA-containing diets were offered to bees, and their effects on the feeding behavior and mortality were studied. The PA diets consisted of a purified mixture of 1,2-unsaturated PA isolated from a natural source (Senecio vernalis). Two diets were offered: A mixture of the tertiary PAs (free bases) and a mixture of the respective N-oxides (Nox). A 0.2% PA content is what one can expect for flower heads of PA-plants. Furthermore, the highest reported value for the PA content in pollen is 1.4% (Boppr et al. 2005). For floral nectar, unfortunately, no data are available. The tested concentrations ranged from 0.02 to 2%, and covered the maximum PA level that an individual honey bee might encounter in nature. The second test series (monocrotaline/1,2-dihydromonocrotaline) was conducted to examine two questions: (1) Are observed effects of the PA-mixes really caused by the PAs and not by contaminations that possibly were not removed completely during PA-mix purification? (2) Since

Methods and Material Plant Material Plants of Senecio vernalis were obtained in spring 2007 from a wild population in the vicinity of Braunschweig. The inflorescences were collected, lyophilized, powdered, and stored dry in the dark at room temperature until alkaloid extraction. Bees Summer bees (Apis mellifera) collected from a honeycomb from the edge of the colony were used for the feeding experiments. Immediately after the tests, bees where anaesthetized by gaseous CO2, stored at 18C, and lyophilized just before analysis. Preparation of the PA-mix Lyophilized plant material (Senecio vernalis, flowering specimens) was extracted according to Hartmann and Toppel (1987). The purified mixture of total plant PAs was analyzed by GC-FID/NPD and GC-MS. It contained senecionine (81%), seneciphylline (15%), integerrimine (1.5%), retrorsine (0.9%), and senecivernine (0.4%). No other components were detected by this method. This purified mixture of tertiary PA is referred to as PA-mix. Chemical N-oxidation of Tertiary Alkaloids The N-oxidation of the PA-mix from S. vernalis (see above) was performed as reported by Cymerman Craig and Purushothaman (1970). All alkaloids of the PA-mix were converted chemically to the corresponding PA-Nox (Fig. 1, A) with essentially the same quantitative composition as the PA-mix. This mixture is referred to as PA-Nox-mix. Preparation of 1,2-Dihydromonocrotaline Monocrotaline (99.0%) was obtained from Fluka (Steinheim, Germany).

J Chem Ecol (2009) 35:10861095

1089

the 1,2-double bond is a prerequisite for PA toxicity, what is the effect of the structurally related 1,2-saturated derivative? Monocrotaline was chosen because it is readily available and 1,2-dihydromonocrotaline because it can be obtained with high purity through chemical synthesis. Monocrotaline and 1,2-dihydromonocrotaline were not used to test for differences between 11- and 12-membered PA ring systems. All solutions were offered to starved bees for 2 h. Starvation status was checked approximately 2 h after the bees were separated from the colony by dissecting 10 randomly picked bees and checking the content of the honey stomach. If all honey stomachs were confirmed empty, the tests were started. After 2 h, the food was changed to a pure 50% sucrose solution ad libitum for an additional 46 h. The amount of the consumed PA solution was determined as the loss of weight of the feeding dish after the first 2 h feeding period. Together with each PA test series, two control groups were conducted. One control group (required for the OECD protocol) obtained pure sucrose solution during the first 2 h, while the second did not get any food during the first 2 h. Since all bees were already starving before the tests started, this adds additional stress (starvation) that might also cause mortality. A high mortality in the first 2 h, however, would be reflected in lower food consumption, which then would be misinterpreted as deterrence. The additional control was introduced to recognize deterrence and to distinguish whether mortality in the PA feeding test is a consequence of additional starvation stress induced by strong deterrence or the result of the PA content of the diet. All cages were kept in an incubator at 252C, in darkness, and at a relative humidity of 5070%. Dead bees of each cage were collected, and the numbers of dead bees were counted after 2, 4, 6, 24, 30, and 48 h. The dead individuals of each time point were stored separately, so subsequent chemical analyses could be done in consideration of the time of death. Alkaloid Extraction from Bees Lyophilized bees were ground with stainless steel balls (1 to 4 mm diam) in a paint shaker (Merris Minimax MK4, Glattbrugg, Switzerland) in two cycles of 90 s each. Extraction of the bees was conducted according to Hartmann and Toppel (1987). Briefly, homogenates were extracted twice with 500 L 1 M HCl and left to stand for 10 min. Subsequently, extracts were centrifuged (10 min at 13,000 rpm) and subjected to one of the following work-up methods: Method A: This was used for monocrotaline and 1,2dihydromonocrotaline feeding solutions and the horizontal transfer experiment. The extract was made basic with NH4OH (25%) and applied to an Extrelut NT20 column (Merck,

Darmstadt, Germany) (1.4 ml/g Extrelut). The PAs were eluted with CH 2 Cl 2 (6 ml/ g Extrelut). The organic solvent was evaporated, and the residue was dissolved in 50 l MeOH. The amount of monocrotaline, 1,2-dihydromonocrotaline or tertiary PA-mix per bee was quantified by using heliotrine as an external standard (concentration 1 mg/ml). Method B: This was used for analysis of extracts that contained exclusively PA-mix. Immediately after grinding of the bees, 50 l of a monocrotaline solution (1 mg/ml in MeOH) were added as internal standard. Monocrotaline was used instead of heliotrine for quantification, because of chromtographically insufficient peak separation for peaks of interest. The procedure was subsequently processed as described for method A. Method C: This was used for the determination of tertiary PAs and PA-Noxs. To evaluate the ratio of tertiary PAs vs. PA-Noxs, essentially method B was applied, but the supernatant obtained after centrifugation was divided into halves. One half was treated as described in method B to account for the amount of tertiary PAs. To the second half, an excess of Zn dust was added to reduce the PA-Noxs into the corresponding tertiary PAs (see also Fig. 1, A). After stirring at room temperature for 3 h, the solution was made basic, and the extraction was finished as described in method A. The result accounted for the sum of tertiary PAs and PA-Noxs (total PAs). The content of PA-Noxs was calculated by subtraction of total PAs by tertiary PAs. The total PA uptake was calculated on the basis of food consumption during the 2 h feeding period, and correlated to the recovered PA amount. Analytical Procedures GC-FID/NPD analysis: An Agilent 6890N Network GC System (Agilent Technologies, Wilmington, DE, USA) was equipped with a ZB-1 capillary column (30 m 0.32 mm 0.25 m fth, Phenomenex, Aschaffenburg, Germany). Conditions: injector 250C; temperature program 100C (3 min)6C min310C (3 min); split ratio 1:10; injection volume 1 l; carrier gas He (1.2 ml/min). The eluting compounds were detected simultaneously by using a fused-silica Y-splitter and an FID and an NPD detector. GC-MS analysis: An Agilent 6890N Network GC System was coupled with an Agilent mass spectrometer 5975B inert EI/CI MSD. For GC separation, a DB-1 fused-

1090

J Chem Ecol (2009) 35:10861095

silica column (30 m0.32 mm0.25 m fth, J&W Scientific, Waldbronn, Germany) was used. Conditions were the same as above. The quantitative analysis of the PAs was performed by capillary GC-FID, the identification of the individual PA structure was confirmed by GC-MS in comparison to spectra and retention indices (RIs) recorded from authentic reference compounds (Witte et al. 1993). The retention index (RI) was calculated by a set of hydrocarbons (even numbered from C10 to C28) by linear interpolation. Horizontal PA Transfer (Trophallaxis) The experiment of bee-to-bee transfer of PAs was performed in the summer season 2007. Recent studies have shown that PAs frequently are found in retail honey and in pollen loads collected by bees (RIVM 2007; Boppr et al. 2008; Kempf et al. 2008). We wanted to test whether the uptake of PA-food influences the exchange of PA-contaminated food among bees. These experiments were planned as a simulation of the natural occurring plant-to-bee and bee-to-bee transfer, and the final PA-deposition in the honey comb. Two cages with 10 bees each were supplied with a diet containing 0.2% or 2% PA-mix in 50% sucrose solution for 1 h. After 1 h the PA diet was removed and an additional ten bees were released into the cages. These bees were marked with yellow color spots and had no food access beforehand. After 1 h of contact (without any food supply) the experiment was stopped, and the two groups (marked/ not marked) were analyzed separately for PAs. The experiment was repeated three times. PA extraction and PA content was determined as described above. All bees of the experiment were analyzed per group (N=10 individuals per group) and thus the values represent the mean PA content per group. Statistical Methods Means of food consumptions and mortalities in the feeding experiments with PA-mix, PA-Nox-mix (Fig. 2), monocrotaline, 1,2-dihydromonocrotaline (Fig. 4), and the total tertiary PA amounts in the different bee-groups of the feeding experiment with 2%-PA-mix (Fig. 5) were tested for significant differences by One-way ANOVA followed by a Student-Newman-Keuls test for all pairwise comparisons. These tests were performed by using SigmaStat 3.1 (Systat Software GmbH, Erkrath, Germany).

Fig. 2 Comparison of food consumption and mortality (means SD) in experiments (N = 9) with tertiary PA-mix and PA-Nox-mix. The food consumption of 2% PA-Nox-mix was significantly different (P<0.01) to all other concentrations. The same applies for the mortality observed for 2% tertiary PA-mix (significant difference (P<0.01) to all other concentrations)

content, Fig. 2), indicating substantial deterrent effects. Concerning mortality, only the 2% PA-mix (containing 1,2-unsaturated tertiary PA) showed a marked increase (Fig. 2), while the PA-Nox-mix did not cause any mortality. In parallel to the regular control groups (0% PA) that had access to sucrose solution throughout the test (48 h), a second control group was established. The time-depending mortality of the starving control group was compared to the 2% PA-mix group and the 0% PA control group (Fig. 3). Figure 3 shows that there can be a high mortality of up to 45% caused obviously by starvation within the first 2 h. But as soon as the remaining individuals had access to regular food (after 2 h), they recovered immediately, and fatal casualties were no longer observed. On the other hand, the increase of mortality caused by the 2% PA-mix clearly showed a different time course. Up to 6 h the mortality rate increased slowly then escalated and kept on ascending and reached mortality rates up to 70% after 48 h (Fig. 3). As expected, the 0% PA control group showed almost no mortality. All individual test series exhibited the same tendencies. Because stress related mortality induced by starvation could be excluded, the observed mortality rate in PA feeding experiments must be linked directly to PA toxicity and not to PA deterrence and rejection of PA food. Monocrotaline and 1,2-Dihydromonocrotaline Feeding Experiments To expand the results for the PA-mix experiments, a second test series was conducted. Instead of the PA-mix, pure monocrotaline and its 1,2-dihydro derivative (Fig. 1, B) were applied. The feeding protocol was exactly

Results PA-mix and PA-Nox-mix Feeding Experiments The feeding of the PA-Nox-mix showed a significantly lower food consumption at the highest concentration tested (2% PA

J Chem Ecol (2009) 35:10861095

1091

as obtained for the PA-mix (Fig. 3). The mortality profile for these tested compounds established that the 1,2-double bond of the PA nucleus is an essential feature of PA toxicity in honey bees. Chemical Analysis of PA-Treated Bees Qualitative and quantitative PA analysis was conducted for distinct groups of bees. The observed effects of the PA feeding should be correlated to the total PA content (PA and PA-Nox) found in individuals of different experiments. The bees of the respective group were picked randomly and analyzed individually. We distinguished early deceased (2 to 30 h), late deceased (>30 h), and surviving bees. The results are summarized in Fig. 5. Bees of the first group showed on average 175 g PA/bee. The second group exhibited an average PA content of 250 g/bee. The surviving individuals (48 h) had the lowest PA load (in average 50 g PA/bee). Because of the selected experimental conditions (sucrose feeding, darkness, humidity, and temperature) bees did not defecate during the assays (48 h). Thus, the PA levels of bees after the treatments were not altered by excretion but only by metabolism. The comparison of the PA-profile of the PA-mix offered with the diet and the profile of PAs found in bees after 48 h (deceased and surviving bees) demonstrated no significant alteration (Fig. 6). This suggests no discrimination in the uptake of individual PAs from the mixture, and identical metabolic stability of the compounds stored in the bees bodies. To check for an overall recovery of administered PAs, all bees of one cage were analyzed quantitatively. About 61% of the consumed PAs could be extracted and analyzed by GC-FID/NPD and GC-MS methods. For a single bee, analyzed 46 h after feeding on the PA diet, a maximal load of 200 g was calculated (Fig. 5, C). Considering the average body weight of 100 mg, this amount of 0.2% PAs testifies to a relatively high tolerance towards PAs under these conditions. However, there is an enormous variation of almost one order of magnitude between the PA loads of the individual bees within one experimental set (Fig. 5, B). Conversion of PA-Noxs to Tertiary PAs In most PAcontaining plants, bees are facing the non-toxic PA-Noxs, which are the dominating plant constituents. As already mentioned, these N-oxides are easily reduced in the intestines of herbivores, yielding the pro-toxic tertiary PAs. Obviously, bees were not harmed by PA-Nox feeding (see Fig. 2). Hence, we wanted to check whether bees have developed mechanisms to maintain the PAs in the non-toxic N-oxide state. Individual bees of a group that had access to a PA-Nox-mix (2%) for 2 h and subsequently 46 h of regular food were analyzed for co-occurrence of both PAforms (Method C). The result indicated that some reduction

Fig. 3 Comparison of bee mortality in experiments with 2% PA-mix diet (offered for 2 h) (half filled symbols) and the control experiment starvation for 2 h (empty symbols) respective 0% PA diet (filled symbols). The curves a to c represent three individual test series (N=10). The time point of food replacement into 50% sucrose solution ad libitum is marked (2 h)

the same as that used for the PA-mix experiments. No significant deterrence was observed for these compounds (Fig. 4). Monocrotaline caused mortality effects at concentrations of 2%. The curve progression of monocrotaline resembled the result obtained for the PA-mix. On the other hand, increasing concentrations of 1,2-dihydromonocrotaline did not cause a marked increase in mortality (Fig. 4). The analysis of the control groups showed the same pattern

Fig. 4 Comparison of food consumption and mortality (means SD) in experiments (N=9) with monocrotaline and 1,2-dihydromonocrotaline. The mortality of 2% monocrotaline was significantly different (P<0.01) from all other concentrations

1092 Fig. 5 Amount of pyrrolizidine alkaloid (PA) per individual bee (means SD) after feeding of 2% PA-mix divided into different groups: Early deceased bees (between 2 and 30 h) (A); end point deceased bees (between 30 and 48 h) (B); surviving bees (48 h) (C). Total tertiary PA amounts in bees of group C were significantly different (P<0.05) from total tertiary PA amounts found in group A and B

J Chem Ecol (2009) 35:10861095

occurs. An average of 69% of the administered PA-Noxs was converted into tertiary PAs (Fig. 7). Thus bees have no specific mechanism to maintain the N-oxides in the harmless form. The maximum level of tertiary PAs did not exceed 50 g/bee (Fig. 7, bee #8). The overall amount of PAs seems rather low, but this is most likely caused by the strong deterrence effect of the PA-Nox-mix (see Fig. 2). Conversion of Tertiary PAs into PA-Noxs As mentioned above, some specialized insects are able to convert potentially hazardous tertiary PAs into PA-Noxs for safe storage or fast excretion. We analyzed bees that had access to the tertiary PA-mix (2%) for 2 h and subsequently 46 h of regular food for the occurrence of possible PA-Nox by using Method C. In all analyzed bees, no such conversion was detectable.

Horizontal Bee-to-bee PA Transfer The bees without direct PA contact showed approximately 4% (2% PA diet) and 15% (0.2% PA diet) of the PA load that was found in bees that had direct PA contact. Thus, at least under laboratory conditions, we demonstrated that horizontal transfer (trophallaxis) of PA contaminated food is possible. In addition, if the content of the honey stomach contained high PA concentrations, the observed exchange to other bees was reduced.

Discussion Recent studies have documented PA contamination of honey, and have raised concerns of potential health risks

Fig. 6 Comparison of GC-NPD chromatograms (A) of the components (B) of the PA-mix (A, upper chromatogram) and a PA profile found in bees after feeding 2% PA-mix diet and 48 h test duration (A, lower chromatogram)

J Chem Ecol (2009) 35:10861095

1093

Fig. 7 Total amounts of PA-Nox and tertiary PA found in individual surviving bees after feeding 2% PA-Nox-mix

for consumers (Edgar et al. 2002; Kempf et al. 2008). The food supply of honey bees depends to a great extent on the acquisition of nectar and pollen from flowers. It is still an open question whether floral nectar of PA-plants naturally contains PAs, since PA contamination found in honey also could be caused by pollen. Any honey that was found to contain PAs was also found to be contaminated with pollen of PA-plants (Kempf et al. 2008). The genuine occurrence of PAs in floral nectars has been suggested frequently but never proved unequivocally (Detzel and Wink 1993; Gaffal and Gammal 2003; Singaravelan et al. 2005; Gegear et al. 2007). It seems that the alkaloid content of nectar generally is lower than in pollen (Detzel and Wink 1993). Since up to 3% of flowering plants contain PAs (Culvenor 1980), there is a high probability that honey bees are exposed to these potentially toxic compounds. So far, no data have been available to explain how bees accomplish this challenge. Toxicity and Deterrence The first approach was to expose bees to a naturally occurring PA mixture. The PA-mix obtained from S. vernalis contained PAs of the senecioninetype, representing one of the most prominent and toxic classes of macrocyclic PAs (Hartmann and Witte 1995; Fu et al. 2004). The results of the feeding experiments with the 2% PA-mix showed high mortalities compared to the PA-free control. Monocrotaline showed a similar toxicity profile (increased mortality at >30 h). Thus far, only the influence of heliotrine, a representative of PA monoesters of the lycopsamine type, has been tested on honey bees, and a LD50 of 0.1% was determined (Detzel and Wink 1993). Since the test conditions were quite different (this study: 2 h PA feeding pulse; Detzel and Wink: continuous feeding of a PA diet for 48 h), it seems understandable why we did not observe such high mortality rates in our experiments at PA concentrations of 0.2%. In

fact, both experiments demonstrated that adult worker bees could safely handle quite high PA amounts in their diets. During the feeding period of 48 h, the bees did not defecate, and thus were unable to reduce the ingested PA load by excretion. The observed PA concentrations among one distinct group of bees varied by almost one order of magnitude (Fig. 5, B) and demonstrated impressively the dimension of biological variation even under idealized laboratory conditions and among closely related individuals. Strikingly, in the feeding experiment with the 2% PA diet, the average PA load of the surviving bees was markedly lower (50 g/bee) as compared to the two groups of deceased bees (175 and 250 g/bee, respectively; Fig. 5). The real PA load might be even higher, since our quantification measure only covered soluble and extractable PAs but not PAs bound to cellular structures or polar degradation or detoxification products. As a consequence, the results of the 2%-test might be interpreted as the dose makes the poison, and bees that only cautiously sipped the highly contaminated diet were able to survive. Since we did not observe any noticeable effects in repellency or mortality for the experiment with the 0.2% PA diet, it seems that bees can deal safely with PA concentrations found in their environment. A similar result was found for the alkaloids nicotine/anabasine and caffeine, which did not deter bees when administered in natural concentrations. Nicotine and caffeine were even found to stimulate feeding (Singaravelan et al. 2005). The design of the experiments allowed us to distinguish between mortalities caused by potential stress through starvation, possibly caused by deterrent effects of the food (Fig. 3), and mortality caused by the ingestion of pro-toxic tertiary 1,2- unsaturated PA (Figs. 2 and 4). The results indicate that the toxicity observed at high concentrations (2% PA diet) of the tertiary PA-mix or monocrotaline must be caused by the ingested PAs. PA-Noxs as well as tertiary PAs that lack the 1,2-double bond (i.e., 1,2-dihydromonocrotaline) were non-toxic. We, therefore, suggest that the toxic effects observed with 1,2-unsaturated tertiary PAs are caused by cytochrome P450-mediated bioactivation (see Introduction). The PA-Nox-mix showed a significant deterrence effect at a concentration of 2%. As mentioned, most potential nectar plants of bees (species of the Asteraceae and Boraginaceae) contain PAs mainly as N-oxides (Hartmann and Witte 1995; Boppr et al. 2005, 2008). The repellent effect may represent an essential adaptation of bees to recognize plant PAs and consequently to provoke individuals to prevent the ingestion of hazardous amounts. Metabolism The conversion of PA-Noxs to tertiary PAs is usually regarded as an unspecific reduction (see Introduction).

1094

J Chem Ecol (2009) 35:10861095

Therefore, we analyzed how bees metabolize PA-Noxs. In all instances, the PA-Noxs were at least partially converted to the corresponding tertiary PAs in the intestines. The extent of conversion varied strongly, but the overall dose of toxic tertiary PAs never exceeded 50 g tertiary PA/bee (Fig. 7). This was in accordance with the average content observed for the surviving bees in the feeding experiment with 2% PA diet (Fig. 5, C). We also examined the ability of bees to convert pro-toxic tertiary PAs into non-toxic PA-Noxs. This reaction is a well-known detoxification mechanism that has been demonstrated for several specialized insects that sequester PAs from their food plants and utilize them for their own benefit (Hartmann and Ober 2008). We never detected any significant N-oxidation of tertiary PAs in bees (data not shown). Thus, detoxification of PAs by N-oxidation can be excluded as a mechanism of detoxification in bees. This is corroborated by the fatalities observed with bees at high 1,2-unsaturated tertiary PA concentrations (Figs. 2 and 4). Trophallaxis Experiment Our experiments of the horizontal PA transfer among bees revealed that bees that ingested high amounts of PAs (feeding on 2% PA diet) donated only 4% of their load to other bees, while bees that ingested lower amounts of PAs (0.2% PA diet) transmitted more than 15%. Obviously, the receiving bees are able to recognize a possibly hazardous load brought in by the foraging workers. Speaking for the colony, this deterrent effect would be beneficial, since foraging bees with high PA loads are less likely to be unloaded in the hive and, therefore, are less likely to recruit new workers for their potentially dangerous nectar source (Gould and Gould 1995; Seeley 1995). Possibly, the same sensory system that elicits feeding deterrence in the field is also involved in the control of trophallaxis. In summary, our results suggest that adult worker bees are not deterred by PA concentrations that reflect natural conditions. Beyond that, bees apparently tolerate natural PA contamination of their food without adverse effects. Important questions and consequences arise, however, when we consider the observed PA tolerance of individuals in respect to PA transmission by trophallaxis within the bee community. If bees do not discriminate PA-containing food per se, PAs would be a natural component of honey and could only be avoided by limiting the access of bees to PAplants. Are bees able to protect their offspring from contact with PAs? The well-documented mutagenic effects of PAs would be even more harmful during early developmental stages of the larvae. Are there any chronic effects of low PA concentrations on the fitness of the colony? Further studies are in progress to address these questions. We expect that the answers will provide useful information as to what extent bees are adapted to handle naturally occurring toxins

such as PAs and thus prevent detrimental effects on their colonies. Perhaps this will improve our understanding of the phenomenon of unexplained mortality of bees, sometimes observed by beekeepers.

Acknowledgements Ms von der Ohe, Ms Schnberger, and Ms Warner are thanked for technical assistance with the handling of bees. This work was supported by the Deutsche Forschungsgemeinschaft (DFG), Bonn, to P. S. project no. SCHR 211/22-1 and SCHR 211/23-1 and to T. B. project no. BE 3200/1-1 and BE 3200/3-1.

References
ADLER, L. S. 2000. The ecological significance of toxic nectar. Oikos 91:409420. BAKER, H. G. 1977. Non-sugar chemical constituents of nectar. Apidologie 8:349356. BETTERIDGE, K., CAO, Y., and COLEGATE, S. M. 2005. Improved method for extraction and LC-MS analysis of pyrrolizidine alkaloids and their N-oxides in honey: Application to Echium vulgare honeys. J. Agric. Food Chem. 53:18941902. BOPPR, M., COLEGATE, S. M., and EDGAR, J. A. 2005. Pyrrolizidine alkaloids of Echium vulgare honey found in pure pollen. J. Agric. Food Chem. 53:594600. BOPPR, M., COLEGATE, S. M., EDGAR, J. A., and FISCHER, O. W. 2008. Hepatotoxic Pyrrolizidine Alkaloids in Pollen and DryingRelated Implications for Commercial Processing of Bee Pollen. J. Agric. Food Chem. 56:56625672. CULVENOR, C. C. J. 1980. Alkaloids and human disease, pp. 124141, in R. L. Smith and E.A. Bababunmi (eds.). Toxicology in the Tropics. Taylor and Francis, London. CULVENOR, C. C., EDGAR, J. A., JAGO, M. V., QUTTERIDGE, A., PETERSON, J. E., and SMITH, L. W. 1976. Hepato- and pneumotoxicity of pyrrolizidine alkaloids and derivatives in relation to molecular structure. Chemico-biological Interactions 12:299324. CYMERMAN CRAIG. J., and PURUSHOTHAMAN, K. K. 1970. Improved preparation of tertiary amine N-oxides. J. Org. Chem. 35:1721 1722. DEINZER, M. L., THOMSON, P. A., BURGETT, D. M., and ISAACSON D. L. 1977. Pyrrolizidine alkaloids: Their occurrence in honey from tansy ragwort (Senecio jacobaea L.). Science 195:497499. DETZEL, A., and WINK, M. 1993. Attraction, deterrence or intoxication of bees (Apis mellifera) by plant alelochemicals. Chemoecology 4:818. EDGAR, J. A., ROEDER, E., and MOLYNEUX, R. J. 2002. Honey from plants containing pyrrolizidine alkaloids: a potential threat to health. J. Agric. Food Chem. 50:27192730. EISNER, T., ROSSINI, C., GONZALEZ, A., IVENGAR, V. K., SIEGLER, M. V. S., and SMEDLEY, S. R. 2002. Paternal investment in egg defence, pp. 91116, in M. Hilker, and T. Meiners (eds.). Chemoecology of Insect Eggs and Egg Deposition vol. Blackwell, Oxford. FREI, H., LUTHY, J., BRAUCHLI, J., ZWEIFEL, U., WURGLER, F. E., and SCHLATTER, C. 1992. Structure/activity relationships of the genotoxic potencies of sixteen pyrrolizidine alkaloids assayed for the induction of somatic mutation and recombination in wing cells of Drosophila melanogaster. Chem. Biol. Interact 83:1-22. FU P. P., XIA Q., LIN G., and CHOU M. W. 2004. Pyrrolizidine alkaloidsgenotoxicity, metabolism enzymes, metabolic activation, and mechanisms. Drug Metab. Rev. 36:155.

J Chem Ecol (2009) 35:10861095 GAFFAL, P. K., and GAMMAL, E. 2003. Nektar und Honiganalyse als Prognose fr den Phloemtransport von natrlichen Pflanzengiften? Drogenreport 16:917. GEGEAR, R. J., MANSON, J. S., and THOMSON, J. D. 2007. Ecological context influences pollinator deterrence by alkaloids in floral nectar. Ecology Letters 10:375382. GOULD, J. L., and GOULD, C. G. 1995. The Honey Bee. Scientific American Library, New York. HARTMANN, T., and OBER, D. 2008. Defense by pyrrolizidine alkaloids: developed by plants and recruited by insects, pp. 213231, in A. Schaller (ed.). Induced Plant Resistance to Herbivory. Springer. HARTMANN, T., and TOPPEL, G. 1987. Senecionine N-oxide, the primary product of pyrrolizidine alkaloid biosynthesis in root cultures of Senecio vulgaris. Phytochemistry 26:16391643. HARTMANN, T., and WITTE, L. 1995. Chemistry, biology and chemoecology of the pyrrolizidine alkaloids, pp. 155233, in S. W. Pelletier (ed.). Alkaloids: Chemical and Biological Perspectives, vol. 9. Pergamon, Oxford. KEMPF, M., BEUERLE, T., BHRINGER, M., DENNER, M., TROST, D., VON DER OHE, K., BHAVANAM, V. B., and SCHREIER, P. 2008. Pyrrolizidine alkaloids in honey: risk analysis by gas chromatography-mass spectrometry. Mol. Nutr. Food Res. 52:11931200. LANDOLT, P. J., and LENCZEWSKI, B. 1993. Lack of evidence for the toxic nectar hypothesis: a plant alkaloid did not deter nectar feeding by Lepidoptera. Fla. Entomol. 76:556566. LINDIGKEIT, R., BILLER, A., BUCH, M., SCHIEBEL, H. M., BOPPR, M., and HARTMANN, T. 1997. The two faces of pyrrolizidine alkaloids: The role of the tertiary amine and its N-oxide in chemical defense of insects with acquired plant alkaloids. Eur. J. Biochem. 245:626636. MACEL, M., BRUINSMA, M., DIJSKTRA, S. M., OOIJENDIJK, T., NIEMEYER, H. M., and KLINKHAMER, P. G. L. 2005. Differences in effects of pyrrolizidine alkaloids on five generalist insect herbivore species. J. Chem. Ecol. 31:14931508. MARN-LOAIZA, J. C., ERNST, L., BEUERLE, T., THEURING, C., CSPEDES, C. L., and HARTMANN, T. 2008. Pyrrolizidine alkaloids of the endemic Mexican genus Pittocaulon and assignment of stereoisomeric 1,2-saturated necine bases. Phytochemistry 69:154167. MASTERS, A. R. 1991. Dual role of pyrrolizidine alkaloids in nectar. J. Chem. Ecol. 17:195205.

1095 MATTOCKS, A. R. (ed) 1986. Chemistry and Toxicology of Pyrrolizidine Alkaloids. Academic, London. NARBERHAUS, I., ZINTGRAF, V., and DOBLER, S. 2005. Pyrrolizidine alkaloids on three trophic levelsevidence for toxic and deterrent effects on phytophages and predators. Chemoecology 15:121125. NICOLSON, S. W., and THORNBURG, R. W. 2007. Nectar chemistry, pp. 215264, in S. W. Nicolson, M. Nepi, and E. Pacini (eds.). Nectaries and Nectar. Springer, Dordrecht. OECD/OCDE 1998. OECD guidelines for the testing of chemicals, Honeybees, Acute Toxicity Test. RAGUSO, R. A. 2008. Start making scents: the challenge of integrating chemistry into pollination ecology. Ent. Exper. Appl. 128:196 207. RIVM Rijksinstituut voor Volksgezondheid en Milieu. 2007. RIKILT Institute of Food Safety, Risicobeoordeling inzake de Aanwezigheid van Pyrrolzidine Alkaloiden in Honing, http://www.vwa.nl/cdlpub/ servlet/CDLServlet?p_file_id=22703 (accessed on May 20th, 2009). Wageningen, Netherlands. SADASIVAM, S., and THAYUMANAVAN, B. 2003. Molecular Host Plant Resistance to Pests. CRC, New York, pp 207209. SEELEY, T. D. 1995. The Wisdom of the Hive: The Social Physiology of Honey Bee Colonies. Harvard University Press, Cambridge. SINGARAVELAN, N., NEEMAN, G., INBAR, M., and IZHAKI, I. 2005. Feeding responses of free-flying honeybees to secondary compounds mimicking floral nectars. J. Chem. Ecol. 31:27912804. STEGELMEIER, B. L., EDGAR, J. A., COLEGATE, S. M., GARDNER, D. R., SCHOCH, T. K., COULOMBE, R. A., and MOLYNEUX, R. J. 1999. Pyrrolizidine alkaloid plants, metabolism and toxicity. J. Nat. Tox. 8:95-116. TAN, K., GUO, Y. H., NICOLSON, S. W., RADLOFF, S. E., SONG, Q. S., and HEPBURN, H. R. 2007. Honeybee (Apis cerana) Foraging Responses to the Toxic Honey of Tripterygium hypoglaucum (Celastraceae): Changing Threshold of Nectar Acceptability. J. Chem. Ecol. 33:22092217. VAN DAM, N. M., VUISTER, L. W. M., BERGSHOEFF, C., DE VOS, H., and VAN DER M EIJDEN, E. 1995. The raison detre of pyrrolizidine alkaloids in Cynoglossum officinale: deterrent effects against generalist herbivores. J. Chem. Ecol. 21:507-523. WITTE, L., RUBIOLO, P., BICCHI, C., and HARTMANN, T. 1993. Comparative analysis of pyrrolizidine alkaloids from natural sources by gas chromatography- mass spectrometry. Phytochemistry 32:187196.