P. armoraciae Rearing
The laboratory culture of P. armoraciae was established in 2012 using beetles collected from horseradish plants in Laasdorf, Thuringia, Germany. Adult P. armoraciae beetles were reared on three- to four-week old potted Brassica juncea cv. Bau Sin plants (Known-You Seed Co. Ltd., Kaohsiung, China) in a controlled environment chamber at 24 °C, 60% relative humidity, and a 14:10 h light:dark cycle. After one week, plants with eggs were transferred to a separate cage for larval development. Three weeks later, any remaining plant material was removed, and the soil containing pupae was kept in plastic containers (9 L volume, Lock&Lock, Seoul, South Korea) until adults emerged. Field-collected beetles were added to the colony every year to prevent an inbreeding depression. The experiments described here were carried out between 2014 and 2017.
Localization of Sequestered GLS in P. armoraciae Beetles
To determine where sequestered GLS are stored in P. armoraciae beetles, we dissected adults collected from the laboratory colony. First, we collected hemolymph by inserting a thin glass capillary into the hemocoel between thorax and abdomen, and then separately collected head, legs, elytra, hindwings, thorax, integument, gut, fat body, and reproductive organs. Hemolymph and tissues of five males and five females were pooled in 500 μL of 80% (v/v) methanol on ice, and samples were stored at −20 °C until GLS analysis. Dissected tissues were homogenized using metal beads (2.4 mm diameter, Askubal, Korntal-Münchingen, Germany) using a TissueLyser II (QIAGEN, Hilden, Germany). After adding 50 nmol of 4-hydroxybenzyl GLS (sinalbin) as an internal standard to each sample, GLS were extracted, analyzed by HPLC-UV at 229 nm and quantified as previously described in Beran et al. (2014). The distribution of GLS in hemolymph and different beetle tissues was expressed relative to the total GLS amount detected in all samples, which was set to 100%.
Design of the GLS Sequestration Experiment with P. armoraciae and A. thaliana
To determine how ingested GLS affect sequestration and excretion in P. armoraciae, we performed a feeding experiment with newly emerged beetles and three different A. thaliana genotypes that differ about four-fold in GLS contents and in their GLS composition. We used the Col-0 wild type of Arabidopsis that mainly produces 4-methylsulfinylbutyl (4MSOB) GLS and other methylsulfinylalkyl GLS as well as minor amounts of indolic GLS, and two double knock-out mutants, myb28myb29 (myb) and cyp79b2cyp79b3 (cyp) in the Col-0 background, which are devoid of aliphatic and indolic GLS, respectively (Sønderby et al. 2007; Zhao et al. 2002). Arabidopsis plants were cultivated in a controlled environment chamber at 21 °C, 55% relative humidity and a 10:14 h light:dark cycle.
To compare GLS in P. armoraciae beetles before and after feeding on Arabidopsis, we randomly assigned newly emerged male and female beetles to the following four treatments, newly emerged (control before feeding), fed on wild type, fed on myb, and fed on cyp, each with ten replicates. Control beetles were immediately collected in groups of five beetles, weighed, frozen in liquid nitrogen, and stored at −20 °C for GLS analysis.
For feeding, we placed five beetles together with one detached leaf from a six- to seven-week old Arabidopsis plant into a Petri dish (60 mm diameter, Greiner Bio-One, Frickenhausen, Germany). The leaf petiole was inserted into a 0.2 mL reaction tube containing 0.1 mL ultrapure water to prevent wilting. Adults were provided with a new leaf taken from an undamaged Arabidopsis plant every day until day five, and were then starved for one additional day before sampling as described for the newly emerged adults.
To calculate how much plant tissue the beetles had ingested, we weighed each leaf before and after feeding. Because we noticed that the fresh weight of detached leaves increased by 23 ± 3% for Arabidopsis wild type, 22 ± 3% for myb mutant, and 15 ± 2% for cyp mutant, under our conditions (mean ± SD, N = 8–9), we corrected the initial leaf weight before calculating the amount of fed plant tissue. Fed leaves were frozen in liquid nitrogen and stored at −20 °C until they were freeze-dried for later GLS analysis. Feces were collected every day using 100 μL of ultrapure water per Petri dish. Each aqueous feces sample was mixed with 300 μL pure methanol (purity ≥99.9%, Carl Roth GmbH & Co. KG, Karlsruhe, Germany) and stored at −20 °C until GLS extraction.
P. armoraciae beetles were homogenized in 1 mL 80% (v/v) methanol containing 50 nmol sinalbin using a plastic pestle. Freeze-dried Arabidopsis leaves were homogenized using metal beads to a fine plant powder that was extracted with 1 mL 80% (v/v) methanol containing 50 nmol sinalbin. Feces samples were homogenized with metal beads and the volume was adjusted to 1 mL using 80% (v/v) methanol containing the internal standard sinalbin. The GLS extraction, analysis and quantification was done as described in Beran et al. (2014).
Question 1: How are the GLS levels and composition in adult P. armoraciae beetles affected by ingested GLS? The GLS compositions in the three Arabidopsis lines used in our sequestration experiment are shown in Table S1. We compared the individual and total GLS amounts (in nmol per beetle) and concentrations (in nmol per mg beetle) in newly emerged beetles with those in fed beetles by different statistical methods depending on the variance homogeneity and the normality of residuals. Comparisons by Student’s t test, Mann-Whitney U test, and analysis of variance (ANOVA) were carried out in Sigma Plot 11.0 (Systat Software, Inc., Erkrath, Germany). Analyses using the method of generalized least squares were done in R 3.5.1 (nlme package, Pinheiro et al. 2019; R Core Team 2018). If necessary, data were transformed prior to analysis. For data analyzed with the generalized least squares method, the varIdent variance structure was applied, allowing each group to have a different variance. The P value was obtained by removing the explanatory variable and comparing both models using a likelihood ratio test (Zuur et al. 2009). Factor level reductions were used to reveal significant differences between groups (Crawley 2013). Details of statistical methods are provided in Supplementary Tables S2 and S3.
Since the amounts of previously sequestered allyl GLS were lower in fed beetles than in newly emerged beetles, we examined the influence of the total ingested GLS amount and the food plant on the allyl GLS levels in beetles and feces, respectively, by analysis of covariance (ANCOVA) in R (R Core Team 2018). The total ingested GLS amount was calculated based on the ingested amount of leaf tissue and the corresponding GLS concentration in each fed leaf. Although GLS are unevenly distributed in Arabidopsis rosette leaves (Shroff et al. 2008), adult feeding damage was randomly distributed across leaves in our experiment (Fig. S1). In both analyses, the ingested GLS amount per beetle was log-transformed. Allyl GLS amounts per beetle were log-transformed, and excreted allyl GLS amounts per beetle were square-root transformed in order to achieve homogeneity and normality of the residuals. Factor level reductions were used to reveal significant differences between groups (Crawley 2013). To elucidate which proportion of the lost allyl GLS was excreted, we expressed the allyl GLS amount detected in the feces relative to the lost allyl GLS amount in adults, which was set to 100%.
Question 2: Are ingested GLS from Arabidopsis wild type leaves selectively sequestered and metabolized in P. armoraciae adults? To determine whether P. armoraciae accumulated individual GLS from Arabidopsis wild type leaves selectively, we expressed the concentration of each GLS in adults relative to the average concentration in feeding-damaged leaves (set to 1). The relative (fold) accumulation of different GLS in P. armoraciae was compared using the generalized least squares method (nlme package, Pinheiro et al. 2019). Data were square-root-transformed prior to analysis.
The analysis of the relative accumulation of individual GLS from Arabidopsis wild type leaves in P. armoraciae revealed a disproportionately high accumulation of 4-methylthiobutyl (4MTB) GLS in beetles. In addition, these beetles contained significantly higher amounts of 3-butenyl (3But) GLS, although this GLS was not present in their food plant. To determine whether P. armoraciae converts 4MSOB GLS, the major aliphatic GLS in Arabidopsis wild type leaves, to 4MTB GLS and 3But GLS, we fed newly emerged adults with an aqueous solution containing 10 nmol 4MSOB GLS (purchased from Phytoplan, Heidelberg, Germany). We placed newly emerged beetles in a Petri dish with a 0.2 μl droplet containing the GLS, or pure water as a control, and observed each beetle until it had finished drinking the droplet under a microscope. To allow adults to metabolize the ingested 4MSOB GLS, they were fed for three days with detached B. juncea leaves, which do not contain 4MSOB GLS and 4MTB GLS (Beran et al. 2014). Afterwards, adults were frozen in liquid nitrogen and stored at −20 °C until GLS extraction. We analyzed four and six replicates for the control and 4MSOB GLS treatments, respectively, each consisting of five adults. Because the 4MSOB GLS solution fed to beetles contained a small amount of 4MTB GLS as contaminant, we compared the ingested 4MTB GLS amount (present in the fed 4MSOB GLS solution) with the amount detected in fed beetles by Student’s t test. Because 3But GLS was detected in control and fed beetles, we compared the levels in both groups by Student’s t test. In addition to 4MSOB GLS, Arabidopsis contains also other methylsulfinylalkyl GLS, including 7-methylsulfinylheptyl GLS and 8-methylsulfinyloctyl GLS, which might be converted to the corresponding methylthioalkyl GLS. Since the chromatographic conditions used for the GLS analysis in beetles and feces did not allow the detection of 7-methylthioheptyl- and 8-methylthiooctyl GLS, we analyzed several samples by HPLC-UV at 229 nm using a modified solvent gradient of 0.2% formic acid (solvent A) and acetonitrile (solvent B): 1.5% (v/v) B (1 min), 1.5–5% (v/v) B (5 min), 5–7% (v/v) B (2 min), 7–21% (v/v) B (10 min), 21–29% (v/v) B (5 min), 29–43% (v/v) B (7 min), 43–100% (v/v) B (0.5 min), 100% (v/v) B (2.5 min), 100 to 1.5% (v/v) B (0.1 min), and 1.5% (v/v) B (4.9 min). In addition, we determined the presence of methylthiolalkyl GLS in beetle and feces samples by liquid chromatography coupled with mass spectrometry. Chromatographic analyses (as described above) were carried out on 1100 series equipment (Agilent Technologies, Waldbronn, Germany) coupled to an Esquire 6000 ESI-Ion Trap mass spectrometer (Bruker Daltonics, Bremen, Germany) operated in positive ionization mode in the range of m/z 60–1000, with a skimmer voltage of 52.8 V, capillary exit voltage of 117.3 V, capillary voltage of 3000 V, nebulizer pressure of 35 psi, drying gas of 11 L/min, and gas temperature of 330 °C. Elution was accomplished at a flow rate of 1 mL/min at 25 °C under chromatographic conditions as described above. Flow coming from the column was diverted at a ratio of 4:1 before reaching the electrospray ionization (ESI) unit. We detected 3-methylthiopropyl-, 7-methylthioheptyl- and 8-methylthiooctyl GLS as desulfo-GLS in samples by comparing the retention times, UV spectra, mass spectra and in-source fragmentation patterns to those of isolated standards (Brown et al. 2003). The presence of 5-methylthiopentyl GLS was analyzed according to its UV spectrum, mass spectrum, and in-source fragmentation pattern.
Question 3: Do P. armoraciae adults excrete GLS selectively? The amounts of individual and total GLS detected in feces were compared by different statistical methods as described above. Details of statistical analyses are provided in Table S4. To determine whether beetles excreted previously sequestered allyl GLS and ingested GLS selectively, we compared the amounts of allyl GLS, 4MSOB GLS, and indol-3-ylmethyl (I3M) GLS excreted by adults fed on Arabidopsis wild type leaves using a linear mixed effects model. The lme function (Pinheiro et al. 2019) was applied to account for the different beetle groups. GLS and day were treated as fixed effects, and beetle groups as random effect. The GLS amount was log-transformed prior to analysis. P values and significant differences between groups were obtained as described above.
Question 4: Does the metabolic fate of ingested GLS in P. armoraciae depend on GLS type, total ingested GLS amount, and the GLS composition in the food plant? To analyze the metabolic fate of ingested aliphatic and indolic GLS, we calculated the percentage of sequestered and excreted aliphatic and indolic GLS relative to the total amount of ingested aliphatic and indolic GLS, respectively (set to 100%). Because a low background of indolic GLS was present in newly emerged adults, we subtracted the average amount of each indolic GLS detected in newly emerged beetles from the corresponding GLS amounts detected after feeding on Arabidopsis. To analyze whether the total ingested GLS amount (covariable), the GLS type (aliphatic or indolic GLS as explanatory variable) or the GLS composition (Arabidopsis line as explanatory variable) affect the metabolic fate of ingested GLS, we performed ANCOVA or analyzed the data using the method of generalized least squares, or linear mixed effects models (nlme package, Pinheiro et al. 2019) with GLS ingestion and GLS type as fixed effects and beetle groups feeding on a certain plant as random intercept. ANCOVA analyses were conducted with type II variance partitioning of the car library (Fox and Weisberg 2011) to adjust each effect for other effects (Kabacoff 2011). If necessary, data were transformed prior to analysis. To determine the appropriate variance structure for the generalized least squares analyses, models fitted with different variance structures were compared based on the Akaike information criterion (AIC) (Zuur et al. 2009). P values were obtained as described above. We used the total ingested GLS amount instead of the ingested amounts of aliphatic and indolic GLS in our analyses, respectively, because a Spearman’s rank correlation coefficient analysis showed a strong positive correlation between the total ingested GLS amount, the ingested aliphatic GLS amount and the ingested indolic GLS amount, respectively (rho ≥ 0.770, P ≤ 0.014; Fig. S2). Details of statistical analyses are given in Table 2.