Spodoptera littoralis were obtained from an old laboratory strain of the Institut National de la Recherche Agronomique, Versailles, France, which had been maintained on a semi-artificial diet since the 2000s on a semi-artificial diet (Durand et al., 2010; Meslin et al., 2022). Knockout of the gene encoding kynurenine 3-monooxygenase was carried out by the CRISPR/Cas9 method as described below. KMO-ko larvae (homozygous for the knockout) and WT larvae (wild type) were grown in separate petri dishes containing white lima bean based artificial diet as described by (Spiteller et al., 2000). This food was mixed with the frass (frass: food = 1:5 w/w) from larvae of a different laboratory strain purchased from Syngenta Crop Protection Münchwilen AG (Münchwilen, Switzerland), because the previous experiments that studied the bacterial population of S. littoralis larvae were conducted on this laboratory strain (Tang et al., 2012; Teh et al., 2016). Hence, the microbial population of the Syngenta line was allowed to colonize the Versailles line (WT and KMO-ko), to maintain uniformity in the gut population among these two strains. The larvae were separately reared and maintained based on family numbers to prevent them from inbreeding.

Five larvae from second, third, fourth and fifth instar stages of the KMO-ko mutant and WT lines fed with artificial diet spiked with frass from Syngenta strain, were sampled and collected. The insects were frozen for 20 min before the dissection, after which they were surface sterilized in 70% ethanol for three times, followed by rinsing in sterile water. All the samples were flash frozen in liquid nitrogen and stored at −80°C until DNA extraction. Three technical replicates containing 0.25 g of food mixed with the frass was also sampled the same way.

8-Hydroxyquinoline-2-carboxylic acid (98%, ACROS Organics) was dissolved in distilled water with pH adjusted to 10 using a solution of 1 M sodium hydroxide for better solubility. This way, a gradient of concentration of 8-HQA was fed to batches of KMO larvae: 1, 3, 5, and 10 mm: 0.8–1 g diet, plus 100 μl of the solution containing the aforementioned final concentrations of 8-HQA, in order to approximate the concentration of 8-HQA naturally occurring in WT larvae. A control batch of the same (termed as 0 mm 8-HQA) was fed with diet spiked with 100 μl of the solvent. Five individuals from each concentration were sampled in the same way.

The CRISPR/Cas9 system was used to obtain a knockout line of S. littoralis that does not produce 8-HQA. In total, three target genes were knocked out in S. littoralis embryos in this study. Initially, two genes potentially coding for alanine-glyoxylate aminotransferase, AGT1 (GenBank Accession ON808443) and AGT2 (GenBank Accession ON808444), were identified based on a study on mosquitoes (Rossi et al., 2005) from an in-house S. littoralis transcriptome. Subsequently, kynurenine 3-monooxygenase (KMO, GenBank Accession ON808442) was identified from the transcriptome. The ZiFit Targeter version 4.2 website¹ was used to identify targets for the knockout of the three genes with the Cas9 endonuclease (target sequences in Table 1).

In the case of AGT1 and AGT2, sgRNA and Cas9 mRNA were synthesized in vitro and injected into freshly-laid eggs as described in Khan et al. (2017). Microcapillary injection needles with internal filament were pulled on a Sutter P-97 pipette puller and injection was accomplished with an Eppendorf Femtojet at 50 hPa resting pressure, 500 hPa injection pressure. The concentration of the injection solution was 100 ng/μl sgRNA and 250 ng/μl Cas9 mRNA.

For KMO, tracrRNA, crRNA and Alt-R^® S.p. Cas9 Nuclease were purchased from IDT (Integrated DNA Technologies, BVBA, Leuven, Belgium) and processed according to the manufacturer’s instructions. The injection solution contained 10 pmol tracrRNA + crRNA and 5 pmol Cas9 enzyme per 1 μl.

Single pair matings were set up in paper cups covered with gauze. Oviposition was monitored and egg masses on the gauze were collected and injected within 1 h after oviposition. The scales and top layers of the egg masses were removed using sticky tape. The eggs were injected with the different injection solutions loaded into a self-made glass needle. Glass needles were made by pulling a capillary (GB100TF-10, 0.78 × 1.00 × 100 mm; Science Products, Hofheim, Germany) on a P-97 micropipette puller (Sutter Instrument, Novato, California). Tips of the needles were cut with a razor blade. Injections of egg masses were done with an Eppendorf FemtoJet 5,247 at 50 hPa resting pressure, 500 hPa injection pressure. Injected egg masses were held in a humidified chamber at room temperature until hatching. Surviving larvae were screened for mutations and reared to adulthood.

For mutation screening, DNA was extracted from a single larval leg, following (Haenniger et al., 2020) with the modification that only 200 μl of the Chelex^® 100 Resin solution was used per sample. Each target region was amplified with PCRs with 5% of the forward primer substituted with a 5’IRD700 or 5’IRD800 labelled primer (Table 1). This way the amplicon could be detected in a Licor 4,300 DNA analyzer (LI-COR Biosciences, Bad Homburg v.d.H., Germany). 96 samples were run simultaneously and the visual output allowed for detection of insertions and deletions as small as 2 bp and up to a size of ~1,000 bp.

No AGT1 injected eggs survived to hatching, possibly due to accumulation of the toxic 3-hydroxykynurenine (Figure 1). For AGT2, one surviving larva which developed into a female showed 2 different mutations in the germline, producing two different mutant lines: one with a 2 bp insertion causing a frame shift in the gene, and one with a 3 bp insertion (Figure 2). Subsequent regurgitant measurements in these mutant lines showed 8-HQA levels similar to the WT larvae (data not shown). Thus, the AGT2 line was discontinued and KMO was used as the new target gene for knockout.

Injections with KMO tracrRNA+crRNA resulted in many different mutant genotypes (Figure 2) and a visible phenotype in the adults. KMO-ko mutant adults showed bright golden eyes instead of dark brown eyes of the wild types (Supplementary Figure 1). Additionally, their wing and body scale color was lighter. This phenotype was subsequently used to identify homozygous KMO-ko mutants, regardless of their specific induced mutation and the line was maintained with a mix of mutant KMO-ko alleles.

Ten fifth instar larvae from each of KMO-ko (fed with 0, 1, 3, 5, and 10 mm 8-HQA) and WT conditions were starved for 5 h and their regurgitant was collected in individual GC vials. To collect the regurgitant, each larva was pressed at the foregut area using a pair of forceps on a sterile petri dish, and the ejected fluid was collected using a pipette and weighed. The regurgitant samples were derivatized by means of tetrabutylammonium hydrogen sulfate (TBAS)-assisted anhydrous pentafluorobenzylation following a modified protocol published by Naritsin et al. (1995; Figure 3A). This method involves addition of 250 μl of 0.1 M TBAS in 1 M NH₄OH solution at pH 10.4 to the regurgitant. After drying in a SpeedVac for 3 h, 400 μl of fresh pentafluorobenzyl bromide (PFBB)/N,N-diisopropylethylamine in acetonitrile solution (1.5 ml N,N-diisopropylethylamine, 0.6 ml PFBB and 27.9 ml dry acetonitrile, for 75 reactions), were added and heated at 65°C for 15 min without mixing. After cooling down to room temperature, 100 μl decane and 1 ml of sulfuric acid (1 M) were added and vortexed for 10 min at 1000 rpm. After centrifuging in a SpeedVac at 1,400 rpm (215× g) for 10 min for phase separation the upper organic layer was transferred to GC-vials with 100 μl micro inserts.

Samples were analyzed by gas chromatography mass spectrometry (GC–MS) with an ITQ GC-ion-trap MS system (Thermo Scientific, Bremen, Germany) equipped with a fused silica capillary column ZB–5 (30 m × 0.25 mm × 0.25 μm with 10 m guard column, Zebron, Phenomenex, United States). One μl of the sample was injected using helium at 1 ml · min⁻¹ as the carrier gas with an injector temperature of 320°C running in splitless mode. Separation of the compounds was achieved under programmed temperature conditions (175°C for 1 min, then at a rate of 10°C min⁻¹ up to 320°C kept for 1 min). The MS was run in EI mode (70 eV) with a scan range of 35 to 800 amu, a transfer line temperature of 320°C, and an ion source temperature of 200°C. Data acquisition was performed using Xcalibur 3.1 (Thermo Fisher Scientific). The basepeak m/z 368 at retention time 12.90 (corresponding the fragment [M-PFT]⁺) was used to plot a calibration curve and calculate the levels of 8-HQA in the samples.

DNA was extracted from 2nd, 3rd, 4th and 5th instar larvae using the DNeasy Power Soil kit (Qiagen). Prior to the extraction of DNA, each larva was homogenized with a bead mill (TissueLyser LT, Qiagen) for 3 min at a frequency of 29 Hz. The homogenate went through a Proteinase K treatment (50 μg μL⁻¹ stock solution, Ambion) at 60°C overnight. The extraction was done as per manufacturer’s protocol. The resulting DNA was purified using the Zymo Research DNA Clean & Concentrator-5 kit (Orange City, California). After the extraction and purification, the DNA concentrations of all the samples were measured using a NanoDrop One Spectometer (Thermo Scientific) and stored at-20°C.

Each extracted sample DNA went through a quality control step which was a PCR amplification of the V4 region of the 16S rRNA gene. The PCR was conducted using forward primer 515F and reverse primer 806R (Table 2), The PCR was conducted using Thermo Fisher Scientific Taq DNA Polymerase (2 Units), 10 mm dNTP (0.2 mm), 50 mm MgCl2 (1.5 mm) and 0.5 μm of each forward and reverse primers. The PCR reaction cycle was an initial denaturation step at 94°C for 3 s, followed 35 cycles of 94°C for 45 s, 50°C for 30 s, and a final elongation step at 72°C for 90 s. Finally, an extended elongation step at 72°C for 10 min.

The DNA marker that was used to determine the band size was Gene Ruler 1 kb Plus DNA Ladder. The samples were run on 1.5% Agarose gel containing ~3–5 μl Midori green Advance DNA stain for visualization under UV illumination. The gel electrophoresis Bio-Rad Wide Mini-Sub^® Cell GT chamber was used to run the samples at voltage of 150 V and 120 mA current for 35–40 min. The gel was visualized using Gel Doc^™ XR+ System and band of ~390 kb size was obtained.

Each 10 μl PCR reaction consisted of 5 μl of SsoAdvanced Universal SYBR Green Supermix (Bio-Rad, Germany), 250 nM of forward and reverse primers: eIF4a to amplify a host gene from midgut tissue, and SRV3 to amplify 16S bacterial rDNA (Lee et al., 1996; Table 3), with the final volume made up with RT-PCR grade nuclease-free water (Thermo Fisher Scientific, Germany). The PCR reactions were conducted in the Bio-Rad CFX96 real-time PCR detection system. The reaction comprised the following steps: 98°C for 3 min, followed by 35 cycles of heating at 98°C for 15 s and 60°C for 30 s. The melt curve was plotted from 65 to 95°C in increments of 0.5°C. For each biological replicate, two technical replicates were measured. The Ct values were determined by log 10 fold-difference Ct = Ct (target gene)/Ct (reference gene), where the target gene is the 16S rDNA amplified by the SRV3 primers, and the reference gene is eukaryotic initiation factor 4a (eIF4a), a housekeeping gene of S. littoralis, used for normalization of the 16S rDNA copy numbers (Kang et al., 2019) relative to host DNA. The amplification efficiency of the two primer pairs was >95%, and was determined with serial dilutions of the amplified gene products of SRV3 and eIF4a primers, respectively.

Standard curves to calculate the copy numbers were plotted by amplifying 16S rDNA (using SRV3 primers) from DNA of E. mundtii and eIF4a from surface-sterilized legs of S. littoralis adults. The PCR was conducted using Thermo Fisher Scientific Taq DNA Polymerase (2 Units), 10 mm dNTP (0.2 mm), 50 mm MgCl2 (1.5 mm) and 0.5 μm of each forward and reverse primer. The PCR reaction cycle was an initial denaturation step at 94°C for 3 s, followed 35 cycles of 94°C for 45 s, 60°C for 30 s, and a final elongation step at 72°C for 90 s. Finally, an extended elongation step at 72°C for 10 min. The segments of length 203 and 289 bp, respectively, were cut with a scalpel after running the amplified product in 1.5% agarose gel as described in S2. The cut gel segments were eluted using QIAquick Gel Extraction Kit (Qiagen). The extracted DNA was serially diluted (1 ng to 10–8 ng) and was used for plotting the respective standard curves.

To study the effect of larval stage and genotype on the bacterial numbers, the qPCR data were analyzed with the help of a nested ANOVA model as follows: the dependence of the ratio of gene copy number (copy numbers of 16S rDNA/eIF4a) was analyzed according to the stage of the insect, genotype (KMO-ko/WT), and additionally, the Ct value of technical replicates nested within the stage of the insect, which is further nested within the genotype. The model checks for the dependence on stage and genotype independently as well as their interaction. Tukey’s post-hoc test was done thereafter. The statistics were performed with R (version 3.6.3).

Extracted DNA from 5 biological replicates of second, third, fourth and fifth instar larvae from KMO-ko and WT strains, and the KMO-ko strains fed with 0, 1, 3, 5, and 10 mm 8-HQA were sent for sequencing to MR DNA (Molecular Research LP), Texas. Sequencing of 2 × 250 bp paired-ends was performed on the IlluminaMiSeq platform using primer pair 515F (Parada et al., 2016), 806R (Apprill et al., 2015; Table 2). Each extracted DNA sample went through a quality control step which was a PCR amplification of the V4 region of the 16S rRNA gene prior to sending for sequencing.

The QIIME 2.0 platform was used to analyze the sequencing data. Adapters and primers of the sequences were removed using the Cutadapt plugin (Estaki et al., 2020), Martin (2011) followed by denoising the sequences using the DADA 2 plugin (Callahan et al., 2016; Estaki et al., 2020; truncation of both forward and reverse reads was set at 230 and chimeras were removed using the consensus method). Classification of taxon was done by training Naive Bayes classifier using Greengenes reference sequences (Greengenes Database Consortium, version 13_8, clustered at 99% similarity; DeSantis et al., 2006). Representative sequences from each OTU were generated and mapped against the Greengenes 16S database for taxon assignment. All samples were rarefied to the least number of reads (55,599, 280,187 pertaining to results 2 and 3, respectively). Taxa corresponding to mitochondria and chloroplast were filtered. R packages: Phyloseq and ggplot2 were used to visualize the analysis of the metagenomics data. Alpha diversity plots were based on the Shannon index (Gauthier and Derome, 2021) or the Faith PD (phylogenetic diversity) index for individual samples. Beta diversity plots depicted principal components analysis (PCoA) based on Bray-Curtis distances (Beals, 1984) or Canberra distances (Warton et al., 2012). The Adonis method for multivariate analysis of variance in the vegan package (Dixon, 2003) was used to check the variance among replicates of the different treatment groups. The Kruskal-Wallis test (p < 0.05) was carried out for each index of Alpha diversity, followed by the Wilcoxon test for pairwise comparison of KMO-ko and WT samples across developmental stages. R-packages-Phyloseq (McMurdie and Holmes, 2013), ggplot2 (Wickham, 2011), vegan (Dixon, 2003), ggpubr (Kassambara and Kassambara, 2022) were used for final visualization of data and statistics on R-studio using R version 3.6.3.

The raw genomic data have been deposited in the NCBI Short Read Archive (SRA). The BioProject ID is PRJNA727271. The S. littoralis genes have been deposited in GenBank (AGT1: GenBank Accession ON808443, AGT2: ON808444, KMO: ON808442). Additional data are in the Supplementary information.

In order to confirm the absence of 8-HQA in the larvae with mutants of the gene encoding kynurenine 3-monooxygenase, the concentration of 8-HQA was measured in the regurgitant of both WT and KMO-ko larvae. Figure 3B shows the levels of 8-HQA compared to the wild type (WT) counterparts (average 142 μg ml⁻¹of regurgitant). In KMO-ko individuals, 8-HQA could not be detected (Wilcoxon test p < 0.05; Supplementary Data Sheet 1).

The normalized 16S copy numbers (designating the bacterial abundance) in the KMO-ko are consistently higher than the corresponding WTs, along the stages of second, third and fourth instar (Figure 4A). The nested ANOVA (Table 3) showed significant results (Tukey’s test, p < 0.05) for 3rd, 4th and 5th instars.

The KMO-ko and WT larvae showed differences in their bacterial composition over different developmental stages. Not only do KMO-ko harbor higher numbers of bacteria (Figure 4A) than the WTs, their bacterial composition is dominated to an even greater extent by the class of Bacilli of the phylum Firmicutes (Figure 4B; Supplementary Figure 3). The KMO-ko of the Versailles strain was fed with frass from the Syngenta strain because the previous experiments on the bacterial population of S. littoralis larvae were conducted on this strain (Tang et al., 2012; Teh et al., 2016). The frass mixed with the food itself had a dominance of Bacilli and the rest consisted of Actinobacteria and Proteobacteria (Supplementary Figure 2). Only in the 2nd instar KMO-ko can one find more diversity than the KMO-ko of other stages. This condition in particular harbours 5 and 2% of Beta-and Gammaproteobacteria, respectively. Although the Firmicutes are still the most abundant phylum, Actinobacteria, and Proteobacteria (Supplementary Figure 3) show their presence in differing levels among the WTs of the four larval stages. The 2nd instar shows a 10 and 3% population of Beta-and Gammaproteobacteria respectively, whereas the 4th instar harbors 16% Actinobacteria and 9% Gammaproteobacteria.

Alpha diversity was measured on the basis of Shannon (Figure 4C), and Faith phylogenetic diversity indices (Supplementary Figure 4). The Shannon index measures species diversity in terms of richness of distribution, while the Faith phylogenetic diversity index is based on phylogeny (Faith, 1992). The Kruskal-Wallis test (p < 0.05) followed by the Wilcoxon test was performed for pair-wise comparisons and it showed a significant difference (p < 0.05) between the two genotypes of KMO-ko and WT in the 3rd and 4th instars prominently in case of Shannon index (Figure 4C; species richness), and 3rd instar when measured using Faith PD (Supplementary Figure 4).

Multivariate statistical analyses were employed to study the differences in bacterial composition across the genotypes (KMO-ko and WT) and individual stages (2nd, 3rd, 4th, 5th instars). Thus, beta diversity analysis was performed using Canberra distances (Warton et al., 2012) and Bray-Curtis (Beals, 1984; Figure 2D; Supplementary Figure 9), using Principal Coordinate analysis (PCoA) to analyze how the individual conditions cluster according to their taxonomic compositions. KMO-ko of all the four instars show clustering when the beta diversity is analysed using Canberra (Figure 4C) and Bray-Curtis distances (Supplementary Figure 5) on PCoA plots.

Analysis of variance (Adonis) using beta-diversity matrices-Bray-Curtis and Canberra distances gave significant results indicating that the bacterial composition was significantly different among the different conditions (p = 0.001; Figure 4D; Supplementary Figure 5). Overlapping clusters especially among stages of WT or KMO-ko is due to several shared common taxa. Considering all the KMO-ko and WT together, they form two separate clusters on a PCoA plot according to their Bray-Curtis (Supplementary Figure 5) and Canberra distances (Figure 4D) because of their different taxonomic composition (Adonis, p = 0.001).

KMO-ko larvae were fed with a range of concentrations of 8-HQA to study the effect on the microbial population. The WT condition was compared to KMO-ko on 0, 1, 3, 5 and 10 mm 8-HQA, added to their artificial diet. The 8-HQA recovered from the regurgitant was directly measured (Supplementary Figure 6; Supplementary Data Sheet 2). The concentration of 8-HQA from the KMO-ko larvae at 10 mm 8-HQA was closest to that of WT (Wilcoxon test, p > 0.5). In terms of bacterial content, these larvae were expected to behave more like WTs. Interestingly the bacterial numbers of 8-HQA-supplemented KMO-ko approached that of WTs (5th_WT; ANOVA, p < 0.05, Tukey’s test between 0 mm 8-HQA and WT p < 0.05; whereas p > 0.05 for other 8-HQA concentrations vs. WT; Figure 5A). However, the bacterial richness measured by the Shannon diversity index did not change in KMO-ko upon feeding with 8-HQA (Supplementary Figure 8, Wilcoxon test p < 0.05 for 8-HQA-fed KMO-ko vs. WT). They harboured 99% Enterococci (Figure 5B; Supplementary Figure 7), similar to the control KMO-ko (0 mm 8-HQA and 5th_KMO), as opposed to 64% Enterococci and 30% Betaproteobacteria in the WTs (5th_WT). The beta diversity analysis using Canberra distances show clustering of 5th instar WTs away from all the KMO-ko fed with 8-HQA, suggesting that an external administration of 8-HQA to the KMO-ko is not recapitulating the bacterial composition in KMO-ko as that of WTs (Supplementary Figure 9).