D. radicum and D. floralis samples derived from a laboratory culture at the German Centre for Integrative Biodiversity Research in Leipzig, Germany which was established 7 years ago. The D. radicum rearing started with collected pupae from a cabbage field in Brittany, France, kindly provided by Dr. Anne-Marie Cortesero (University of Rennes, France) and the D. floralis rearing with pupae from a laboratory culture which were kindly provided by Dr. Maria Björkman (Bioforsk–Norwegian Institute of Agricultural and Environmental Research, Norway). The colonies have been maintained under controlled environmental conditions as described in Sontowski et al. (2019). We collected eggs and larvae for all following experiments from these cultures and performed the experiments at constant temperature of 20°C, 16:8 h light:dark and relative humidity of 70 ± 5% in a Percival Reach-In chamber (CLF Plant climatic, Wertingen, Germany).

In the first experiment, we tested whether gut extracts from D. radicum and D. floralis larvae, which include also their gut microbiome, contain enzymes to catabolize GSLs. Larvae were fed on turnip and at the end of the 2^nd larval developmental stage (instar), the larvae were frozen at −20°C for 45 min, surface-sterilized (2 min in 0.2% bleach, neutralized by 1 min in potassium thiosulfate and rinsed three times with 70% ethanol) and had their guts dissected. We pooled 10 guts per replicate and manually homogenized them in 100 µl autoclaved ddH₂O using a pestle. To exclude effects of different protein concentrations in the extracted samples between the different species, we measured the protein concentration in a 1:500 dilution of the samples using the Micro BSA protein assay kit (Thermo Scientific, Rockford, IL, United States) according to the supplier’s recommendations. Protein concentration was measured on a Jasco V-630 spectrophotometer (Jasco, Groß-Umstadt, Germany), and determined from three replicates. Extracts prepared from both species contained the same range of protein concentrations. For this experiment, we used the following treatments: 1) gut extracts of D. radicum larvae or 2) D. floralis larvae incubated at room temperature. 3) gut extracts of D. radicum larvae or 4) D. floralis larvae heated for 7 min at 95°C to reduce the microbial activity and to denature proteins. Each treatment contained three biological replicates. To test whether GSLs were degraded by the larval gut extracts, we added 67 µl of 2PE-GSL solution (300 μg/ml in H₂O, 2-phenylethyl glucosinolate, also called gluconasturtiin; class: benzenic GSL, Phytoplan, Heidelberg, Germany, >97.0% purity) and 70 µl 4MSOB-GSL solution (300 μg/ml in H₂O, 4-(methylsulfinyl)butyl GSL, also called glucoraphanin; class: aliphatic GSL, Phytoplan, Heidelberg, Germany, >97.0% purity) to all samples and incubated them for 1 h at 25°C. In addition, we analyzed three replicates of 10 normal gut extracts treated as described previously, but without adding external GSLs to consider GSL residues from the food in the gut. All reactions were stopped by adding 85% methanol and boiling the samples for 5 min at 92°C. The GSLs were extracted and analyzed according to Grosser and van Dam (2017). Briefly, GSLs were desulphated and analyzed with reverse phase Ultra High Performance Liquid Chromatography (UHPLC) equipped with a photodiode array detector (PDA; Thermo Scientific Ultimate 3000 Series, Thermo Fisher Scientific, Waltham, MA, United States) at 229 nm. We injected 50 µl per sample. Desulphated GSLs were separated with a reverse-phase C₁₈ column (4.6 × 150 mm, 3 μm, Thermo Fisher Scientific, Schwerte, Germany) connected with a C₁₈ pre-column (10 × 4.6 mm, 5 μm particle size) using the parameters described in Grosser and van Dam (2017). After separation, the identification of desulphated GSLs was carried out based on retention time and UV spectra compared to commercially available reference standards (Phytoplan, Heidelberg, Germany). Desulphated GSLs were quantified using sinigrin as an external standard and response factors as described in Grosser and van Dam (2017). Data were processed using Thermo Scientific Chromeleon Chromatography Data System software vs 7.2 SR5 MUa (Thermo Fisher Scientific, Waltham, MA, United States).

To test whether ITCs were degraded by the larval gut extracts, which includes also their gut microbiome, we repeated the experiment above (GSL-breakdown experiment) regarding the preparation of larvae, treatments and replicates. Again, we pooled the extracted guts of 10 larvae and mashed them in 80 µl potassium phosphate buffer (0.1 M, pH 7). Before adding ITCs instead of GSLs, we added 10 µl ZnSO₄ (0.001 M) to each sample. Thereafter, 10 µl of 4MSOB-ITC (1 mg/ml in ethanol, MCE, Sollentuna, Sweden, purity >98%) or 10 µl of 2PE-ITC solutions (0.5 mg/ml in DMSO, Sigma-Aldrich, St. Louis, Missouri, United States, purity 99%) were added to the samples (n = 3 per condition group and ITC). To detect possible 4MSOB-ITC or 2PE-ITC from the food residues in the extracted guts, three additional replicates of larval guts from both species were tested without adding external ITCs. These samples were treated as described previously but instead of ITC, only the corresponding solvents were added (10 µl of 96% ethanol or 10 µl of 100% DMSO). All reactions were incubated for 1 h at 25°C. To stop the reactions, we added 10 µl of glacial acetic acid and vortexed quickly. All samples were subsequently centrifuged for 10 min at 10°C at 12.000 rpm (adapted from Schramm et al. (2012) and Jeschke et al. (2016)). The supernatant was transferred to a new vial and stored at −20°C until measurement. An aliquot of 1 µl was injected on an Agilent 1260 series HPLC system (Agilent Technologies, Boeblingen, Germany) coupled with an API5000 tandem mass spectrometer (Applied Biosciences, Darmstadt, Germany) using the HPLC parameters described in Chen et al. (2020). Separation of compounds was performed using a Zorbax Eclipse XDB-C-18 column (50 × 4.6 mm, 1.8 μm; Agilent) with chromatographic signals compared to authentic standards where available. 4MSOB-glutathione, 4MSOB-cysteine and 4MSOB-N-acetylcysteine were obtained from Santa Cruz Biotechnology (Dallas, TX, United States). 4MSOB-amine was obtained from Enamine (Monmouth Junction, NJ, United States). 4MSOB-cysteinylglycine was synthesized as described in Schramm et al. (2012). The cyclic 4MSOB-cysteine derivative (2-(4-(methylsulfinyl)butylamino)-4,5-dihydrothiazole-4-carboxylic acid, (Falk et al., 2014), was synthesized as described in Beran et al. (2018) (as “cyclic-Cys conjugate A”). 2PE-amine was obtained from Acros (Geel, Belgium), and raphanusamic acid was purchased from Sigma-Aldrich (St. Louis, Missouri, United States). MRM parameters for parent ion to product ion fragmentation were set as follows (includes parameters used in Gloss et al., 2014; Beran et al., 2018; Chen et al., 2020), and gave the corresponding retention times: m/z 178.11 →114 (CE, 13 V; DP, 56 V; RT 2.62 min) for 4MSOB-ITC; m/z 485.11 →179.1 (CE, 29 V; DP, 81 V; RT 2.04 min) for 4MSOB-glutathione; m/z 356.07 →136.1 (CE, 15 V; DP, 51 V; RT 1.93 min) for 4MSOB-cysteinylglycine; m/z 299.06 →136.1 (CE, 15 V; DP, 56 V; RT 1.90 min) for 4MSOB-cysteine; m/z 341.07 →178.1 (CE, 17 V; DP, 56 V; RT 2.22 min) for 4MSOB-N-acetylcysteine; m/z 265.1 →201 (CE, 25 V; DP, 56 V; RT 1.04 min) for 4MSOB-cyclic Cys; m/z 136 →72 (CE, 17 V; DP, 56 V; RT 0.45 min) for 4MSOB-amine; m/z 122 →105 (CE, 15 V; DP, 56 V; RT 1.21 min) for 2PE-amine, and m/z 164 →117.8 (CE, 17 V; DP, 61 V; RT 2.08 min) for raphanusamic acid.

To identify the genes expressed upon ITC exposure, and potentially involved in detoxification of ITC, total RNA was extracted from D. radicum and D. floralis larvae at the 2nd instar. The larvae were reared on a semi-artificial diet containing milk powder, yeast, freeze-dried turnip, agar (2:2:2:1) and 90% water. The diet of ITC exposed larvae was spiked with 2PE-ITC (0.35 µl of a 6 mmol/ml solution DMSO per g diet, Sigma-Aldrich, St. Louis, Missouri, United States, purity 99%). The larvae, 15 per treatment group and species, were reared on control or ITC diets for 7 days. Diets were refreshed every other day to ensure that the concentration of the ITCs and amines would not decrease too much (Muller et al., 2015). The experiment was set up in a climate cabinet (Percival Scientific, Perry, Iowa, United States) at constant conditions (see above). After 7 days, the remaining larvae were shock frozen at −80°C for 45 min. Then the larval surface was rinsed with autoclaved distilled water before the whole larva was manually crushed in lysis buffer from the ReliaPrep RNA Tissue Miniprep kit (Promega, Madison, United States). Total RNA was extracted from single larvae following the supplier’s recommended protocol. Qualitative and quantitative RNA analyses were performed using gel electrophoresis (1% agarose), a NanoPhotometer^® P330 (Implen, Munich, Germany) and a Qubit 2.0 (Invitrogen, Carlsbad, CA/United States, BR RNA kit). We pooled three samples and used three replicates per species and condition (with or without 2PE-ITC). Poly(A)-enriched strand-specific library preparation and RNA sequencing were performed by the Deep Sequencing group of Biotech TU Dresden, Germany on an Illumina NextSeq next-generation sequencer. In total, approximately 570 Mio read pairs having a length of 75 bp were generated.

The recently published reference genome of D. radicum (of iDiv_Dra_1.0, GenBank accession number GCA_021234595.1, (Sontowski et al., 2022) was used as a starting point for the expression analysis in both species. To obtain a reference sequence suitable to analyze D. floralis expression, we first used Illumina RNAseq reads (NCBI BioSamples: SAMN25131701 and SAMN25131702) of D. floralis transcriptomes to polish the existing D. radicum reference genome. Reads were first mapped on the iDiv_Dra_1.0 D. radicum genome using the Burrows-Wheeler Alignment tool with the BWA MEM algorithm version 0.7.17 (Li and Durbin 2009). The bam file was then passed to Pilon v1.23 (Walker et al., 2014) with default settings, in order to obtain an edited version of the genome more compatible with D. floralis (e.g. SNPs and indels with regard to the D. radicum genome have been corrected). The existing annotation of iDiv_Dra_1.0 was transferred to this new sequence using the tool liftoff in version 1.6.1 (Shumate and Salzberg 2021). In a second round, and in order to recover larger missing regions in the D. radicum reference, all D. floralis RNA-seq reads were mapped on the iDiv_Dra_1.0 reference using STAR v2.7.8a (Dobin et al., 2012). All unmapped reads were pulled together and assembled into contigs using Trinity v2.9.1 (Henschel et al., 2012). Those new contigs were annotated with coding regions and UTRs using TransDecoder (version 5.5.0, https://github.com/TransDecoder/TransDecoder). Annotations were filtered for a proper start and end of protein-coding transcripts by applying the GeMoMaAnnotationFilter (GAF, GeMoMa version 1.7.2, Keilwagen et al., 2016; Keilwagen et al., 2018) according to the workflow applied for the annotation of the genome of D. radicum (Sontowski et al., 2022). The resulting gene annotations were appended to the liftoff-derived gene annotation file of D. floralis (iDiv_DFl_1.0. gff) and accordingly, sequences of the new contigs containing these annotations were added to the set of pilon-derived genome sequences of D. floralis (iDiv_DFl_1.0. fasta).

Gene expression was assessed for both species using the nf-core/rnaseq pipeline in version 3.0 (Ewels et al., 2020), which is a standard nextflow pipeline for RNA-seq analysis, relying on STAR v2.6.1d (Dobin et al., 2012) for the alignment and Salmon 1.4.0 (Patro et al., 2017) for the quantification. Gene expression raw counts were used for differential gene expression analysis.

Differential gene expression analysis was performed using DESeq2 v1.3.0 (Love et al., 2014) and both species were treated separately. First, only genes with sufficient counts (e.g. at least 2 samples with more than 5 normalized counts) were retained. Differential expression analysis was then performed using a simple “∼ treatment” design, with 3 replicates for each condition (control and exposed to ITC). Differentially expressed genes were selected based on Benjamini–Hochberg (BH) false discovery rate (FDR) procedure; with a 5% significance threshold. Total number of differentially expressed genes in response to 2PE-ITC containing diet in D. radicum and D. floralis larvae were presented with a Venn diagram using jvenn (Bardou et al., 2014).

All genes were functionally annotated using PANNZER2 (Protein ANNotation with Z-scoRE, Törönen et al., 2018). Enrichment of GO terms of differentially expressed genes with regard to all expressed genes was tested using the topGO R package in version 2.44.0 (Aibar et al., 2015). Significantly enriched terms within the “Biological Process” and “Molecular Function” families were selected using a Kolmogorov-Smirnov test and a 5% significance threshold.

From the gene expression data, we performed a targeted screening of differentially expressed genes between the 2PE-ITC and the control group with an adjusted p-value <0.05. In the next step, we focused on genes associated with the mercapturic acid conjugation pathway (GSTs, γ-glutamyltransferases, dipeptidases, N-acetyltransferases) and general detoxification genes (cytochrome P450s, Danielson et al., 1997). From the mercapturic acid pathway, we selected one candidate gene, coding for enzymes from each enzyme class, that was significantly upregulated as a gene of interest. If more candidates fulfilled these expectations, we selected the most reliable candidate gene based on a high number of mapped RNAseq reads and a high log-fold change. Following this procedure, we selected the CYP6A1 gene as a gene of interest representing the cytochrome P450s, although it did not achieve the targeted p-value. To prove that these genes responded specifically to ITCs, we designed qPCR primers for the selected genes (Supplementary Table S1) using the online version of Prime3 v. 4.1.0 (Untergasser et al., 2012). In addition, two primer sets were designed for the housekeeping genes GAPDH and EF-1α (Supplementary Table S1). We verified that these genes were stable in their expression over different ITCs and concentrations. We set up an experiment, in which 2^nd instar of D. radicum and D. floralis were fed with a semi-artificial diet containing 0, 1 or 2 µmol 2PE-ITC (0.02, 0.2 or 0.4 µl of a 6 mmol/ml solution in DMSO per g diet) or 4MSOB-ITC (0.2, 2 or 4 µl of a 0.5 mmol/ml solution in ethanol per g diet). To the control diets, either 0.4 µl DMSO or 4 µl ethanol per g diet was added. After 6, 24 and 48 h, we collected 4-5 larvae from each diet condition and each fly species (in total 2 species x 6 conditions x 3 time points x 4-5 replicates per ITC = 150 samples in total). Due to low replicate numbers of D. floralis larvae on 2PE-ITC after 48 h, this time point was excluded from the analysis. Larvae were shortly rinsed with tap water to remove soil and diet particles, shock frozen at -80°C, and stored until further use. Single individuals were manually crushed in 500 µl TRIzol™ Reagent (Thermo Fisher Scientific, Schwerte, Germany) and total RNA was extracted following the supplier’s protocol. RNA quantity and quality were assessed using a NanoPhotometer^® P330 (Implen, Munich, Germany) and a gel electrophoreses (1% agarose). DNA was digested using DNAse I (Thermo Scientific, Waltham, MA, United States) according to the supplier’s instructions. The quality and quantity of the DNA-free RNA was assessed as described before. This RNA was translated into cDNA using RevertAid H Minus Reverse Transcriptase (Thermo Fisher Scientific, Waltham, MA, United States) according to the supplier’s instructions. Gene expression of candidate genes was examined using qPCR. The qPCR reactions were performed on a CFX384 Touch Real-Time PCR detection system (Bio-Rad, Feldkirchen, Germany) using 5 µl PerfeCTa SYBR Green Supermix (Quantabio, Beverly, MA, United States), 0.5 µl of 10 µM forward and reverse primer (Supplementary Table S1), 3.5 µl water and 1 µl cDNA. The qPCR program was as follows: initial incubation at 95°C for 5 min, followed by 40 cycles at 95°C for 30 s, 58°C for 30 s and 72°C for 30 s, followed by a 0.2°C increment melt curve from 60° to 95°C to confirm that there was a single amplified product representing high specificity. Individual samples were run in triplicate and a negative control with no DNA template was examined for each primer set on each plate. Housekeeping genes and target genes of the same sample were run on the same plate. To compare the expression levels after larval feeding at different ITC concentrations, we calculated the 2^−ΔΔCt values.

To test the effect of ITCs and their breakdown products on the performance of both fly species, we collected 500 eggs from each Delia species (D. radicum and D. floralis) and used 100 eggs per treatment. Eggs from each condition were placed in a plastic box (10 × 10 × 6 cm) filled with autoclaved and moistened sand (Gerhard Rösl GmbH, Jesewitz, Germany). The boxes were closed with a transparent lid and covered with Parafilm. The experiment was set up in a climate cabinet (Percival Scientific, Perry, Iowa, United States) at constant conditions (see section 2.1). After hatching, larvae of the control groups were fed with a semi-artificial diet (see section 2.4) and the ITC groups with the same diet spiked with 2PE-ITC (0.4 µl of a 6 mmol/ml solution in DMSO per g diet, Sigma-Aldrich, St. Louis, Missouri, United States, purity 99%), 4MSOB-ITC (4 µl of a 0.5 mmol/ml solution in ethanol per g of diet, MCE, Sollentuna, Sweden, 98% purity), 2PE-amine (24 µl of a 0.08 mmol/ml solution in DMSO per g diet, 99% purity) or 4MSOB-amine (3.8 µl of a 0.5 mmol/ml solution in ethanol per g diet, 95% purity). In total, we prepared 10 boxes (2 species x 5 conditions with 100 eggs each) including the following conditions: larvae fed on 1) diet without ITC as a control, 2) diet with 2PE-ITC, 3) diet with 4MSOB-ITC, 4) diet with 2PE-amine, and 5) diet with 4MSOB-amine. The diet was replaced with fresh diet every other day and the sand was moistened when needed. The number of larvae and pupae were recorded, and insects were weighed on the 11th, 18th and 25th day after starting the experiment for D. radicum and on the 14th, 21st and 31st day for D. floralis. Temporal differences between species are based on differences in developmental times. In addition, the number and sex of the emerged adults were determined.

Statistical analyses of data obtained from the GSL-breakdown, ITC-detoxification, performance experiments and qPCR analyses were run on R version 4.03 (R Core Team 2020). Homogeneity of variance across groups was tested using Levene’s test in the “car” package (Fox and Weisberg 2019). Normal distribution of the residuals was assessed visually and with the Shapiro-Wilk test. Data of the ITC breakdown experiment and qPCR data were normalized by log₂ transformation before analyses. Measured GSL values were statistically compared using a Student’s t-Test. The LC-MS/MS peak areas of ITCs, amines and conjugates retrieved in the ITC breakdown experiments were compared using a one-way ANOVA on each compound separately followed by a Tukey post-hoc test. For qPCR data, differences between groups were determined using a two-way ANOVA within each species and gene using time and ITC concentration as independent variables, in combination. Significant differences among conditions within time were identified using Tukey’s “Honest Significant Difference” method as post-hoc test. In the performance experiment, differences between treatments were compared using a log-rank test with the “survival” package 3.2-7 in R (Therneau and Lumley 2015). To test differences in the female-male distribution when larvae fed on different diets, we used a one-proportion z-test. The weight of the larvae and pupae were compared within single species and among conditions using a Mann-Whitney test with a Bonferroni correction of the resulting p-value.

Neither 2PE-GSL nor 4MSOB-GSL levels differed significantly between incubations with normal and heated larval gut extracts of D. radicum and D. floralis (Figure 1; Table 1; Supplementary Table S3). We did not detect GSLs in the negative controls containing only extracted guts from D. radicum or D. floralis, confirming that no residual GSLs were stored in the gut from the diet (Supplementary Table S3).

After incubation with 2PE-ITC, we detected a high proportion of 2PE-amine in the normal gut extracts of both D. radicum and D. floralis larvae (Figures 2A,B; Table 2; Supplementary Tables S3, S4). Very low levels of this breakdown product were detected in the heated gut extracts of both species. No 2PE-amines were detected in gut extracts without 2PE-ITC (Supplementary Table S3). 2PE-ITC itself does not produce a signal under LC-MS conditions utilized, and 2PE-ITC breakdown products of the mercapturic acid conjugation pathway (PE-glutathione, PE-cysteinylglycine, PE-cysteine, PE-N-acetylcysteine) could not be identified due to the lack of authentic standards for these compounds.

When incubated with 4MSOB-ITC, normal gut extracts of both species showed reduced levels of this substrate, whereas the ITC breakdown products 4MSOB-amine, 4MSOB-cysteine and cyclic 4MSOB-cysteine (2-(4-(methylsulfinyl)butylamino)-4,5-dihydrothiazole-4-carboxylic acid) were observed in these samples (Figure 2C, D; Table 2; Supplementary Tables S3, S4). In the heated gut samples of both species, 4MSOB-ITC stayed high and only very low levels of 4MSOB-amine and 4MSOB-cysteine were detected, probably as a result of non-enzymatic reactions. Other 4MSOB-conjugates commonly found as products of the mercapturic acid conjugation pathway (4MSOB-glutathione, 4MSOB-cysteinylglycine, 4MSOB-N-acetylcysteine), as well as the potential breakdown product of ITC-cysteine conjugates raphanusamic acid, were not detected. Combined these results indicate that Delia larval guts possess the enzymatic machinery necessary to break down dietary ITCs.

Using both sequence polishing and de novo assembly, the reference genome for D. radicum was edited to be used as a reference for D. floralis. Pilon detected and corrected 81,723 SNPs and indels after aligning 28.5 M of D. floralis Illumina RNAseq reads on the D. radicum reference genome. De novo assembly of the unmapped reads allowed us to reconstruct 36,573 transcripts clustered in 25,064 clusters. In total, 1,948 of those contigs were annotated to contain 2160 protein-coding open reading frames and were added to the edited genome of D. radicum to yield a comprehensive genome for D. floralis (https://doi.org/10.5281/zenodo.6044094).

RNAseq sequencing yielded on average 46.8 M read pairs for D. floralis (min: 41.0 M, median: 47.5 M, max: 52.0 M) and 48.2 M read pairs for D. radicum (min: 40.9M, median: 49.8M, max: 53.0 M) per sample. On average, 87.63% of the reads were correctly aligned for D. floralis (min: 86.23%, median: 87.68%, max: 88.93%) and 94.44% for D. radicum (min: 93.91%, median: 94.52%, max: 94.64%). After filtering of genes with low expression levels, 18,646 genes were considered for differential expression in D. floralis, and 26,351 for D. radicum.

We found 592 upregulated and 349 downregulated genes in D. radicum larvae when exposed to 2PE-ITC (Figure 3). The reverse was found in D. floralis larvae, where more genes were downregulated (504) than upregulated (291). Only four genes were up- and six downregulated in both species. More genes (in total 73) were upregulated in one species and downregulated in the other. Overall, genes of the same gene ontology (GO) term classes responded to ITCs in the larvae of both species, but in different directions (Figure 4, Supplementary Figure S1). While upregulated genes in D. radicum corresponding to 22 GO term classes, only six corresponding classes were observed to respond in D. floralis. Upregulated GO-classes and genes in D. radicum were mainly involved in glutathione, amino acid, peptide, sulfur compound and amine metabolic processes as well as sulfur compound binding (Figure 4; Supplementary Figure S1). The gene expression in D. floralis larvae comprised metamorphosis-related terms and whereas the glutathione, amino acid and sulfur compound metabolic processes were downregulated (Figure 4). Zooming in to the genes connected to the mercapturic acid conjugation pathway in D. radicum larvae revealed eight significantly upregulated genes encoding for GSTs, one for γ-glutamyltransferases, two for dipeptidases and five for N-acetyltransferases (Supplementary Table S1). We identified one significantly downregulated gene encoding for a GST and one for an N-acetyltransferase. In D. floralis larvae we found no significantly upregulated genes from any of these enzyme classes; actually, we found three downregulated GSTs and four N-acetyltransferases in this species.

The qPCR analyses showed that the selected candidate genes of the GST, γ-glutamyltransferase and cytochrome P450 (CYP6A1) family were upregulated in D. radicum larvae that were exposed to 2 µmol 2PE-ITC for 48 h (Figure 5A,B,E; Table 3; Supplementary Table S4). The expression of the selected genes for dipeptidase and N-acetyltransferase did not change upon ITC exposure (Figure 5C,D; Table 3; Supplementary Table S4). We found no differential expression of any selected detoxification genes in D. floralis feeding on 2PE-ITC containing diets (Supplementary Figure S2; Table 3; Supplementary Table S4). None of the selected candidate genes responded to 4MSOB-ITC in either species at any concentration or time point, (Supplementary Figures S3,S4, Table 3; Supplementary Table S4).

In order to determine the possible advantages of metabolizing 2PE- and 4MSOB-ITCs to Delia larvae, we fed ITCs and their major amine detoxification products to the insects in a semi-artificial diet. In general, consuming 4MSOB-ITC, 2PE-ITC, 4MSOB-amine or 2PE-amine all negatively affected the survival rate of D. radicum (Figure 6A; Table 4). Among these conditions, larvae performed markedly better when feeding on a diet containing the detoxification product 2PE-amine than on those with the toxin 2PE-ITC. A similar trend (p > 0.05) was detected in the 4MSOB-ITC and 4MSOB-amine treatments. Additionally, D. radicum larvae survived better when feeding on 4MSOB-ITC than on 2PE-ITC-containing diet; in fact, none of the larvae reached the second instar when fed on 2PE-ITC containing diets.

While larval weight gain was not affected by exposure to the different compounds, D. radicum pupae were slightly heavier when the larvae had been raised on diet containing 4MSOB-amine (Supplementary Figure S5). Additionally, the consumption of 4MSOB-amine and 2PE-amine affected the sex ratio of the emerging adults; diets with amines shifted the female-male ratio towards an increased proportion of females (Figure 6C; Supplementary Table S5).

D. floralis performed similarly to D. radicum when exposed to 2PE-ITC; none of the D. floralis larvae reached the 3rd instar when fed 2PE-ITC containing diet. We also found a higher survival of larvae fed on diets with amine compare to those fed on ITC (Figure 6B; Table 4). Surprisingly, the presence of 4MSOB-ITC or 4MSOB-amine resulted in a more than two-fold larger survival rate for D. floralis larvae relative to the larvae fed on control diet.

Although several of the growth effects of ITCs and amines were similar between both fly species, we observed species-specific effects on larval and pupal weights. While the larval weight of D. floralis decreased during the consumption of 4MSOB-amines, the pupal weight increased when consuming 4MSOB-ITCs (Supplementary Figure S5). Species-specific effects were also detected in the female-male ratio: contrary to D. radicum, no effect of these chemicals was detected on the sex ratio of emerging D. floralis (Figure 6D; Supplementary Table S5).