B. thuringiensis sv. morrisoni biovar tenebrionis isolate 4AA1 (Btt) (Bacillus Genetic Stock Center, Ohio State University, Columbus, OH, USA) was grown in HCT medium (5 g tryptone, 2 g Bacto casamino acids, 6.8 g KH₂PO₄, 0.1 g MgSO₄, 0.002 g MnSO₄, 0.014 g ZnSO₄, 0.15 g CaCl₂, and 0.022 g ammonium ferric citrate in 1 L dH₂O) (29) for 96 h at 30°C, 200 rpm. The bacterial cultures were then washed twice with 20 ml phosphate-buffered saline (PBS), pH 7.4, with a centrifugation step at 15,000 rpm at 4°C and a final dilution in 20 ml PBS. The presence of spores and toxin crystals was verified by light microscopy observation. Bradford protein analysis (30) using the Bio-Rad microprotein assay was performed to quantify the protein content of the Btt toxin crystal, and 10% SDS-PAGE was used to qualify their nature (relative part of the Cry3Aa 65-kDa toxin). The results were used to calculate the toxin content (in micrograms per microliter) of the Btt culture.

The infection suspension was exposed to high-temperature treatment (80°C) for 10 min to deactivate vegetative cells in order to provide the insects with Btt spores only. The spore load was then determined by CFU counting on Luria–Bertani agar (LBA) after an incubation period of 24 h at 30°C. The ratio of the toxin/spore suspension in the feed during the assay was expressed in micrograms toxin per milligram feed, and the volume of suspension added in the feed was 100 µl/100 mg feed. The concentration of Btt toxin used for the infection was prepared by concentrating the infection suspension in order to obtain a final load of 2.5 μg toxin/mg feed and 2 × 10⁷ spores CFU/mg feed.

The M. brunneum isolate KVL 12-30 (Department of Plant and Environmental Sciences, University of Copenhagen) was grown on Petri dishes (9 cm diameter, triple vented) containing Sabouraud dextrose agar (SDA; 65 g/L) at 23°C in complete darkness for 21 days. The conidia were harvested with a Drigalski spatula by pouring 10 ml of Tween 20 (0.05%) into the Petri dishes. Two washing steps with 15 ml of Tween 20 (0.05%) with centrifugation (3,000 rpm for 3 min) to discard the supernatant each time were followed by a dilution step with water to obtain a 1,000 times diluted suspension. The conidia were counted under a light microscope at ×400 magnification after pouring 20 μl of the suspension into a 0.2-mm Fuchs-Rosenthal hemocytometer. The conidia were counted in four squares (one square containing 16 cells) diagonally starting on the lower left corner. The concentration (conidia per milliliter) was then calculated by multiplying the average number of conidia per square with 5,000,000/ml (1,000 times dilution/0.0002 ml volume). The spore suspensions were then prepared from the stock suspension using the following formula: C ₁ * V ₁ = C ₂ * V ₂ (where C is the concentration and V is the volume). All spore suspensions were prepared and used for inoculation on the same day. A germination test was performed to ensure the ability of the fungal conidia to germinate. On the day of the assay, 100 μl of the infection suspension was plated on the SDA medium and incubated for 24 h at 28°C at 65% RH in the dark. Conidial germination was recorded by calculating the percentage of conidia per Petri dish in 3 Petri dishes. Values of conidial germination above 99% were accepted, which indicate that the fungi were able to germinate in the conditions of infection. The M. brunneum conidial suspension used for the infection was 30,000 conidia/mg diet.

Larvae of T. molitor with a body mass of 20 mg from the 10 different treatments were selected for the infection experiments. Groups of 20 larvae were first placed without feed in sterile plastic cups (60 ml; Qualibact^® resistant 95 kPa, Labelians, France) to obtain a density of 0.07 g/cm² for 24 h ( Figure 2 ) in four replicates for each infection treatment. An amount of 100 mg of diet, composed of a mixture of WB (90 mg) and WE (10 mg) or WB alone, was poured into each cup. A volume of 100 µl of the Btt suspension or the M. brunneum suspension or 50 µl Btt + 50 µl M. brunneum suspension or 100 µl sterile water as the negative control treatment, giving 1 μl/mg feed, was poured homogenously in the feed for inoculation. A wet paper was inserted into the cups to maintain the humidity. Subsequently, the cups were covered with tissue paper and incubated at 28°C for 72 h. After 72 h, the dead insects, frass, and remaining feed were mechanically separated by sieving (Ø = 200 mm; pore sizes, 2 mm, 1.4 mm, and 500 μm) (Retsch, Eragny sur Oise, France). The larvae (about 400 mg larval biomass) were then transferred into four new sterile cups per infection treatment containing the WB diet (30% of the larval biomass) and 1% water agar (about 60% of the larval biomass). The number of surviving larvae and the weight of all larvae in the cup were subsequently recorded every 2 days for 11 days ( Figure 2 ).

Statistical analysis of the survival, relative growth, and the individual mean mass (IMM) of the larvae was performed using R 4.2.2 (31) on three replicates. Each of these variables was modeled as a function of the probiotic, diet, pathogen, and their interactions.

A generalized linear model with a binomial error distribution was used to model the number of surviving larvae. Given that one of the categories presented complete separation (i.e., all larvae survived), a bias reduction method (brglmFit) from the R package brglm2 (32) was used. Wald tests were then used to assess the statistical significance of the factor effects using the lmtest package. Thereafter, estimated marginal means and contrasts (emmeans package) (33) were used to compare differences between the different treatments and the control category (no pathogen and no probiotics) within a given diet and pathogen inoculation category. The p-values were adjusted using an approximation of the Dunnett’s method. Similarly, contrasts were used to estimate the differences in survival between diets with the same probiotic and pathogen status.

The IMM was calculated by dividing the biomass of the alive larvae at the end of the assay by the total number of larvae, and relative growth was determined by dividing the difference between the final and the initial weight of the population (all larvae per cup) by the initial weight following the methods previously used by Dupriez et al. (34). Analysis of variance (ANOVA) was used to test the statistical significance of the factors previously mentioned on larval relative growth and IMM. Shapiro-Wilk’s and Bartlett’s tests were run on the residuals of the ANOVA model to test for normality and homogeneity of the variances among factor groups. No departure from these conditions was found (α = 0.05). Thereafter, estimated marginal means and contrasts were used for survival. The p-values for contrasts among factors were adjusted with Tukey’s and Dunett’s methods for relative growth and IMM, respectively.

Larvae were sampled from each treatment when an individual body mass of 20 mg was reached (prior to pathogen exposure, day 0), 72 h after pathogen infection treatments (day 3), and 11 days after pathogen removal (day 14). DNA extraction was performed on 4 replicates composed of 3 individuals each (n = 12 larva per treatment per day) ( Figure 2 ). The larvae were surface sterilized with 3 washing steps consisting of successive immersions in diethyl pyrocarbonate (DEPC) water, EtOH 70%, and DEPC water for 30 s. Three larvae were pooled and then ground with liquid nitrogen. DNA extraction was performed with the DNeasy® PowerSoil® Pro Kit (Qiagen®, Hilden, Germany). The quality of the genetic material was determined by NanoDrop and Qubit analysis. Samples were stored at −80°C.

The V3–V4 hypervariable regions of the ribosomal RNA subunit 16S rDNA gene were amplified from the DNA extracts during the first PCR step with the universal primers PCR1F_343 and PCR1_R784 ( Table 2 ), which are fusion primers as described by Nadkarni et al. (35). This PCR was performed using 2 U of a DNA-free Taq DNA polymerase and 1× Taq DNA polymerase buffer (MTP Taq DNA Polymerase; Sigma-Aldrich, St. Louis, MO, USA). The buffer was completed with 10 nmol of a dNTP mixture (Sigma-Aldrich, St. Louis, MO, USA), 15 nmol of each primer (Eurofins, Luxembourg), and nuclease-free water (Qiagen, Hilden, Germany) in a final volume of 50 μl.

The PCR reaction was carried out in a T100 Thermal cycler (Bio-Rad, Hercules, CA, USA) as follows: an initial denaturation step (94°C for 10 min) was followed by 30 cycles of amplification (94°C for 1 min, 68°C for 1 min, and 72°C for 1 min) and a final elongation step at 72°C for 10 min. Amplicons were then purified using CleanPCR magnetic beads (Clean NA, GC Biotech B.V., Waddinxveen, the Netherlands) in a 96-well format. The concentration of the purified amplicons was controlled using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA), and a subset of amplicon sizes was controlled on a Fragment Analyzer (AATI, Hialeah, FL, USA) with the reagent kit ADNdb 910 (35–1,500 bp).

Sample multiplexing was performed on the Abridge platform (INRAE, Jouy en Josas, France) by adding tailor-made 6-bp unique indices during the second PCR step at the same time as the second part of the P5/P7 adapters to obtain primer PCR2_P7F and reverse primer PCR2_P7R ( Table 2 ). This second PCR step was performed on 50–200 ng of purified amplicons from the first PCR using 2.5 U of a DNA-free Taq DNA polymerase and 1× Taq DNA polymerase buffer. The buffer was completed with 10 nmol of a dNTP mixture (Sigma-Aldrich, St. Louis, MO, USA), 25 nmol of each primer (Eurofins, Luxembourg), and nuclease-free water (Qiagen, Hilden, Germany) up to a final volume of 50 μl. The PCR reaction was carried out on a T100 thermal cycler with an initial denaturation step (94°C for 10 min), 12 cycles of amplification (94°C for 1 min, 65°C for 1 min, and 72°C for 1 min), and a final elongation step at 72°C for 10 min. Amplicons were purified as described for the first PCR reaction. The concentration of the purified amplicons was measured using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA), and the quality of a subset of amplicons (12 samples per sequencing run) was controlled on a Fragment Analyzer (AATI, Hialeah, FL, USA) with the reagent kit ADNdb 910 (35–1,500 bp). Controls were carried out to ensure that the high number of PCR cycles (35 cycles for PCR1 + 12 cycles for PCR2) did not create significant amounts of PCR chimeras or other artifacts. The region of the 16S rDNA gene to be sequenced has a length of 467 bp, for a total amplicon length of 522 bp after PCR1 and of 588 bp after PCR2 (using the 16S rDNA gene of Escherichia coli as a reference).

Negative controls were included to assess the technical background using nuclease-free water (Qiagen, Hilden, Germany) in place of the extracted DNA during the library preparation.

All libraries were pooled with equal amounts to generate equivalent numbers of raw reads for each library. The DNA concentration of the pool (no dilution or diluted 10× and 25× in EB + Tween 0.5% buffer) was quantified on a Qubit fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). The pool, at a final concentration between 5 and 20 nM, was used for sequencing.

The pool was denatured (0.1 N NaOH) and diluted to 7 pM. PhiX Control v3 (Illumina, San Diego, CA, USA) was added to the pool at 15% of the final concentration as described in the Illumina procedure. A total of 600 μl of this pool and the PhiX mixture were loaded onto the Illumina MiSeq cartridge according to the manufacturer’s instructions using MiSeq Reagent Kit v3 (2 × 300 bp paired-end reads, 15 Gb output). FastQ files were generated at the end of the run (MiSeq Reporter software, Illumina, San Diego, CA, USA) for quality control. The quality of the run was checked internally using PhiX Control, and then each paired-end sequence was assigned to its sample using the multiplexing index.

The resulting sequences were received in the format of de-multiplexed FASTQ file and analyzed using FROG (version 4.0.1) (36). The sequences obtained from the 267 samples were analyzed according to the following workflow: pre-processing (maximum read size, 250; maximum rate of mismatch in the overlap region, 0.1; Vsearch software for merging paired-end reads, 150–450 limits for amplicon sizes; primers, 5′-ACGGRAGGCAGCAG and 3′-AGGATTAGATACCCTGGTA), swarm clustering (distance = 1), chimera removal, and OTU filtering (minimum OTU abundance as proportion or count = 0.00005, as recommended by Bokulich et al., (37), to retain OTUs with at least 0.005% of all sequences; PhiX database for contaminants). Affiliation by blast (version 2.10) with the reference DATABASE SILVA 138.1 (38). The indices Chao1, Shannon, and Simpson were used to determine the alpha diversity. Beta diversity was determined with the weighted UniFrac (39) method and principal coordinate analysis (PCoA). A permutational multivariate ANOVA using distance matrices (vegan::adonis) and Wilcoxon’s test corrected with the Benjamini–Hochberg false discovery rate method were performed to determine whether the paired comparisons of the treatments were significant.

Larval survival was recorded on day 14, 11 days after the larvae were moved from the pathogen-inoculated feed, which corresponded to 14 days after contact with the probiotic feed (see Figure 2 ). A model that included the second- and third-order interactions improved the model fit over a model that included only the main effects (χ ² = 50.148, df = 31, p = 0.01619). The results showed significant differences in the survival of T. molitor larvae reared on WB with provision of P. pentosaceus KVL B19-01, live (Pp) and deactivated (Pp_t), with 92.6% and 94%, respectively, in the co-infection treatment (met_btt) compared with the larvae reared on the WB control diet and exposed to the same two pathogens, where survival was only 74.6% (Pp: p = 0.0415; Pp_t: p = 0.0227) ( Figure 3 ). No effect of the provision of Lb. plantarum was found on survival in either infected or uninfected larvae provided with WB or WE (p > 0.05). Individuals reared on the WE diet were less susceptible to pathogen infection, as shown by the higher survival probabilities 14 days post-pathogen exposure ( Figure 3 ).

The relative growth (weight gain between days 3 and 14) in the control compared to the other treatments during the assay was significantly influenced by the presence of the pathogens and not by the diet composition or by probiotic addition (ANOVA, type III test: pathogens, p = 0.003; diet, p = 0.117; probiotic, p = 0.137) ( Figure 4 ). Significant differences were observed between the met_btt co-infection and the control (p = 0.003) and between met_btt and Btt (p = 0.033). Similarly, significant differences were observed between the met_btt co-infection and met infection (p = 0.045). No significant differences were detected in single infections between the control and Btt infection (p = 0.84) or the control and met infection (p = 0.78) or between the Btt and met infections (p = 0.99).

The IMM was impacted by the interaction of probiotics, diet, and pathogen exposure (F _12,80 = 1.81, p = 0.06). Larvae fed the WB diet supplemented with deactivated Lb. plantarum (lacto_yes) and infected with M. brunneum (met) and those fed the WB diet supplemented with live Lb. plantarum (lacto_no) and infected with both pathogens (met_btt) presented higher IMM than the WB with pathogen “control” (p = 0.05 and p = 0.038, respectively), indicating that the addition of the probiotic reduced the negative impact of the pathogens on larval growth ( Figure 5 ).

The effects of probiotic treatment on the composition of the bacterial community were investigated in the two diets (WB and WE) from individuals exposed to different treatments [probiotics and pathogen(s)]. The aim was to identify the level at which the diets and the presence of probiotics and pathogens modify the bacterial composition and the OTU abundance over time after the removal of the treatments. A key interest was also to analyze the extent to which the probiotics or the pathogen could persist in the larvae. The analyses were performed on larvae from day 0 (just after removal from the diet with probiotics), from day 3 (just after removal from the diet with pathogens), and on day 14, corresponding to 11 days fed with WB or WE only. All data were compared with those of the control larvae fed only with WB or WE throughout the whole period from egg to the 14-day assay ( Figures 1 , 2 ).

The 16S rDNA Illumina sequencing analysis resulted in a total of 12,890,914 Nb sequences. Pre-processing allowed selecting a number of pre-paired sequences of 1,849,872 Nb (14.35%). After chimera removal (27.5%), the OTU filter processing retained 9,841,546 (92.7%) sequences.

Taxonomy was assigned with the SILVA 138.3 database (38), which determined the affiliation of 100% sequences with 0.005% sequence abundance and an identity of acceptance of 97%. A total of 390 OTUs were obtained from 275 samples and were classified into 6 phyla, 8 classes, 26 orders, 41 families, 85 genera, and 124 species ( Supplementary Table S1 ).

On day 0, the microbial composition was mainly represented by Enterococcus [Relative abundance (RA%): WE control = 39%, WE_Lp = 6%, WE_Pp = 23%, WE_Lbt = 0%, and WE_Ppt = 6%], Lactococcus (%RA: WE control = 22%, WE_Lp = 55%, WE_Pp = 9%, WE_Lbt = 86%, and WE_Ppt = 2%), and Listeria (%RA: WE Ppt = 36%) in larvae reared on the WE diet. T. molitor fed the WB diet presented prevalence of the genera Lactococcus (%RA: WB control = 41%, WB_Lp = 12%, WB_Pp = 14%, WB_Lbt = 7%, and WB_Ppt = 14%) and Staphylococcus (%RA: WB control = 53%, WB_Lp = 21%, WB_Pp = 29%, WB_Lbt = 26%, and WB_Ppt = 1%). The family Enterobacteriaceae was also widely present. In all the samples, P. pentosaceus was detected from day 0 throughout the assay in all conditions at the species level, while Lb. plantarum was not detectable at all.

On day 0, the impact of probiotic provision was significantly different for T. molitor individuals fed the WB- and WE-based diets. In larvae reared on the WE diet, the alpha indices Chao1 (p > 0.05), Shannon (p = 0.00056), Simpson (p < 0.0001), and InvSimpson (p = 0.009) showed significant differences in the species richness of the larvae exposed to the different probiotic treatments. Paired comparisons highlighted differences between the control (no) and Lb. plantarum deactivated (Shannon: p = 0.008; Simpson: p = 0.001; InvSimpson: p = 0.016) and in both P. pentosaceus live and deactivated (Shannon: p = 0.013; Simpson: p < 0.0001). In larvae reared on the WB diet, the alpha indices for microbial community composition were not significantly different as a result of probiotic treatment. These findings suggest that the addition of egg white protein (WE diet) had a direct or an indirect influence on the probiotics and the composition of the commensal bacterial microbiota.

Overall, when the microbial communities were compared after 14 days of probiotic removal, the alpha values for OTU composition were not significantly different, suggesting little diversity between the larvae exposed to the different treatments. However, variation was found within each treatment over time. Larvae fed with WE presented significant differences in the alpha values in pairwise comparisons between day 0 and day 14 (Chao1: p = 0.002; Shannon: p < 0.0001; Simpson: p < 0.001) and between day 3 and day 14 (Chao1: p < 0.001; Shannon: p = 0.0001). Moreover, the WB samples presented significant differences in the alpha values in pairwise comparisons between day 0 and day 14 (Chao1: p < 0.0001; Shannon: p < 0.0001; Simpson: p < 0.001) and between day 3 and day 14 (Chao1: p < 0.00001; Shannon: p < 0.00001; Simpson: p = 0.023). These results are illustrated in the PCoA of the WB and WE diets ( Figure 6 ). There is clear clustering of the data. The day 14 clusters were different from those of day 0 and day 3, in which more overlap was observed for both WB and WE. Interestingly, when comparing WB and WE, the clusters were differently positioned in the PCoA.

Comparisons between the live and deactivated probiotic treatments on day 14 showed differences in the relative abundances of the microbial community at the genus level. Larvae fed with live P. pentosaceus (WB-Pp) presented higher abundances of species in the genera Enterococcus and Sphingobacterium compared to those provided with the deactivated strain (WB_Ppt). The same live strain inoculated in the WE diet resulted in higher abundances of Enterococcus, Acinetobacter, Enterobacter, Erwinia, and Pantoea.

Lb. plantarum WJB had no major impact on the microbial community composition in WE-fed larvae compared with the control. On the other hand, in WB supplemented with the live bacteria, the abundance of species in the genera Sphingobacterium, Chryseobacterium, and Staphylococcus was higher, while a lower abundance of Acinetobacter was observed.

Comparison between the probiotic treatments and the day of sampling was made using multivariate ANOVA. The two diets showed different results. A significant interaction effect on microbial composition occurred between probiotic treatment and day for larvae fed the WE diet on day 14 (R ² = 0.222, df = 8, F = 2.66, p = 0.001). For the WB diet, this interaction was not significant (R ² = 0.089, df = 6, F = 1.5873, p = 0.114).

Further analyses were performed on the microbiota composition of the OTUs by applying Wilcoxon’s test corrected with the Benjamini–Hochberg false discovery rate method to compare the probiotic treatments and the day of sampling. Probiotic provision did not result in statistical differences in the OTU composition compared with the control treatments (p > 0.05). The microbiota composition of larvae reared on the same probiotic treatments was mainly affected by the time, with differences between day 0 and day 14 being less evident than those between day 3 and day 14 ( Supplementary Figure S2 ).

The WB- and WE-based diets did not influence the microbial community composition on day 0 in uninfected larvae (multivariate ANOVA: p > 0.05). The family Enterobacteriaceae (RA% = 24%) and the genera Staphylococcus (%RA = 10%–50%), Chryseobacterium and Sphingobacterium (%RA = 5%–16%), Lactococcus (%RA = 3%–12%), Pseudomonas (%RA = 8%), and Weisella (%RA = 3%–7%) were the most represented in larvae from all control treatments. Among the diet treatments, a difference in the abundance of the OTUs belonging to the genus Pediococcus on day 14 was higher, but not significant, in WE-fed larvae compared to WB-fed larvae ( Supplementary Figure S1 ).

T. molitor larvae were exposed to pathogens from day 0 to day 3 (72 h), placed in sterile cups, and then provided with non-contaminated feed for 11 days. Overall, the addition of probiotics resulted in significant differences in the larval microbial community abundance on both days 3 and 14 in the case of infection. The probiotic supplementation in WB and WE resulted in significant effects on the microbial community composition on day 3 (R ² = 0.2286, F = 5.74, df = 4, p = 0.0001; R ² = 0.2394, F = 6.7579, df = 4, p = 0.0001) and on day 14 (R ² = 0.1735, F = 4.0275, df = 4, p < 0.00001; R ² = 0.1673, F = 4.831, df = 4, p < 0.00001). When considering the effect of probiotic species on pathogen-infected larvae in the two diet treatments on day 3, the analysis showed no significant effects in WB-fed larvae (p > 0.05), but significant differences in WE-reared larvae (R ² = 0.0739, F = 2.9021, df = 12, p = 0.0001). Indeed, the microbial community of larvae exposed to single infection with either Btt or met or to co-infection presented different microbial community compositions on day 3 compared with uninfected individuals in WE (Chao1: control-Btt, p < 0.0001; control-Btt/met, p = 5.468e−7; control-met, p = 0.00031). In WB diet treatments, the effect of pathogens on the microbiota composition was less pronounced than that in WE, as examined based on the p-values (Chao1: control-Btt, p = 0.00190; control-Btt/met, p = 0.02439; Shannon: control-met, p = 0.02544; InvSimpson: control-met, p = 0.01369) ( Figure 7 ).

After 3 days, in all the control larvae exposed to M. brunneum, Lactococcus (%RA = 46%) was the most abundant genus, followed by Weissella and Staphylococcus and the family Enterobacteriaceae. The controls presented similar microbial composition, with the Chryseobacterium and Sphingobacterium genera detected with RA ranging from 20% to 53% in infected larvae on day 14. Individuals reared on probiotic treatments presented lower RA of the same genus (%RA < 10%). Bacillus species had RA of 3% in infected individuals on day 3, with lowered values in all probiotic-treated larvae (<2%) and were no longer detected on day 14. Bacillus species were completely absent in samples fed the WE or WB diet supplemented with deactivated probiotic species.

Interactions between probiotic treatments and infections were evident at the end of the experiment (R ² = 0.3478, F = 3.3468, df = 12, p < 0.0001; R ² = 0.2941, F = 2.4474, df = 12, p < 0.0001). Indeed, the effects of the treatments on OTU abundance were still detectable in the control larvae and in those infected with Btt (Chao1: p = 3.978E−06; Shannon: p = 1.446E−06; Simpson: p = 0.002475; InvSimpson: p < 0.0001). On day 14, WE-fed larvae infected with Btt, met, or with both pathogens presented significant differences in microbial community composition (met-Btt/met: Chao1, p < 0.0001; Shannon, p = 0.00997; met-Btt: Chao1, p = 0.00814; Shannon, p = 0.00577; InvSimpson, p = 0.00094).

The same trend was not observed in WB-fed larvae, for which the infection treatments did not show any significant impact on the microbial composition on day 14 (p > 0.05).