The larvae of Hermetia illucens L. (black soldier fly, BSFL) used in this study were bred at the German Institute of Food Technologies (DIL; Quakenbrück, Germany). The breeding system was installed in a greenhouse at 27 °C with 60–70% relative humidity and 14 h:10 h (light: dark) time ratio. BSF were kept in cages of 1.8 m³ with a 1.5 mm mesh size. Commercial laying hen meal (GS Agri, Schneiderkrug, Germany) with addition of 5% (w/w) sunflower oil was used as a feed substrate for the first 6-day growth phase of larvae (Khan et al., 2025).

The following agro-industrial by-products were used as feed substrates: Carrot pomace (AHN GmbH, Warmsen, Germany), brewer’s yeast as well as spent grains (Leiber, Bramsche, Germany), and rape press cake (Brökelmann + Co – Oelmühle GmbH + Co, Hamm, Germany). Commercial laying hen meal (GS Agri, Schneiderkrug, Germany) supplemented with 5% (w/w) sunflower oil was used as control feed. All substrates had no legal limitations to be used as substrates for animal feed in farms, either as a single component or in blends (EC, 2001). Each substrate’s nutritional composition, moisture content, and contents of starch, monosaccharides, disaccharides, and ash were determined according to German standard methods for feed described in §64 LFGB, L00.00-19/3 2004-07 (Bundesamt für Verbraucherschutz und Lebensmittelsicherheit, 2004). All analyses were performed in duplicates. The amount of digestible carbohydrates was calculated as the sum of the contents of starch, monosaccharides, and disaccharides (Dossey et al., 2016). All substrates were packed, kept in vacuum-sealed plastic bags, and stored at 2 °C for a maximum of 4 months. Before use, single substrates were ground to a particle size <1 mm using a lab mill with a 1 mm sieve (GM 200, Retsch, Haan, Germany) (Kieβling et al., 2023).

Six days after hatching, approximately 2,000 larvae of an average weight of 3.3 ± 0.4 mg/larva were placed into each rearing box (32 cm × 25 cm × 10 cm). Each rearing box represented one replicate; thus, approximately 2,000 larvae were used per replicate. The experiment consisted of five groups in total: one control feed group and four experimental groups corresponding to the four substrates. Each group was set up in three replicates in individual rearing boxes. Water was added to the dry substrates to obtain a wet substrate containing 67% moisture. In every box, 2.3 kg (wet weight) of freshly prepared substrate was filled in. The amount of wet substrate was given in surplus and was not exchanged during the rearing period. The BSFL were kept under the same conditions as described for colony maintenance. The feeding period was 11 days. For microbiological analysis, approximately 15 g of larvae were collected manually from each rearing box and placed in sterile stomacher bags. The adhering substrate was removed from the larval surface through manual stripping with a tweezer. Then the bags were sealed and stored at 8 °C till analysis. The remaining larvae per rearing box were used for weight determination and chemical analysis; larvae were frozen at −20 °C (Ma et al., 2018; Kieβling et al., 2023).

For nutritional values of BSFL, the nutritional composition, moisture content, and contents of starch, monosaccharides, and disaccharides were determined according to the German standard methods for feed described in §64 LFGB, L00.00-19/3 2004-07 (Bundesamt für Verbraucherschutz und Lebensmittelsicherheit, 2004). Each analysis was carried out in duplicate. The total amounts of starch, monosaccharides, and disaccharides were summed up to calculate the amount of total digestible carbohydrates. For the determination of fatty acids of the larvae, proteins and carbohydrates were separated from the lipid fraction. For preparation of fatty acid methyl esters, samples were heated directly with a methanolic sulfuric acid solution after freeze-drying in accordance with ISO 12966-2. In the second processing step (ISO 12966-4), the fatty acid methyl esters were separated by capillary gas chromatography on a highly polar stationary phase depending on their chain length, the degree of saturation, and the geometry and positions of the double bonds [gas chromatography with flame ionisation detector (GC-FID)]. The equipment used consisted of a gas chromatograph GC-2010 plus FID (Shimadzu Europa mbH, Duisburg, Germany) with the separation column Agilent HP 88, 100 m, film thickness 0.20 m, and ID 0.25 mm.

The collected larval samples (three replicates per treatment; see 2.3) were homogenised in their sample bags by crushing with a rolling pin. From the homogenised larvae or the previously collected substrate samples (one sample per substrate; see 2.3), 10 g were aseptically transferred to stomacher bags (INTERSCIENCE) and diluted with 90 mL of sterile maximum recovery diluent (Oxoid). Samples were homogenised in a stomacher (BagMixer 400; INTERSCIENCE) for 2 min (first dilution). From these first dilutions, 5 mL were frozen and stored at −20 °C until needed for 16S rRNA gene sequencing. The remaining first dilutions were immediately used to determine the microbial counts.

Decimal dilutions were prepared, and aliquots of 0.1 mL of the appropriate dilutions were spread and incubated in duplicates using the following parameters: Plate count agar (PCA, Oxoid) incubated at 30 °C for 3 days to count mesophilic aerobic bacteria; Yeast extract glucose chloramphenicol agar (YGC; Oxoid) incubated at 25 °C for 4 days to count yeasts and molds; Violet red bile dextrose agar (VRBD; Oxoid) incubated anaerobically at 30 °C for 2 days in an anaerobic jar (AnaeroGen, Oxoid) to count Enterobacteriaceae. After appropriate incubation, the colony-forming units (CFU) were counted.

DNA was isolated using the DNeasy PowerLyzer PowerSoil kit (Qiagen) according to the manufacturer with some modifications. Briefly, 1.8 mL homogenate were pelleted by centrifugation at 10,000 g for 2 min. The pellet was resuspended in 750 μL of PowerBead solution and transferred to a PowerBead tube. Sixty μl of solution C1 were added, and the sample was homogenised in a Bead Ruptor 24 Elite (OMNI International) at 6 m/s for 4 min. For insect samples, an additional Proteinase K treatment was introduced after bead milling. Hundred μg Proteinase K (A&A BIOTECHNOLOGY) were added to the homogenate and incubated overnight at 56 °C and 600 rpm (Rubin et al., 2014). For both feed and insect samples, the DNA isolations were completed using the entire PowerSoil protocol with a final elution in 50 μL of solution C6. DNA quality was examined with a NanoDrop™ One microvolume UV-Vis spectrophotometer (Thermo Scientific), and the concentration was measured with a Qubit 4 fluorometer (Invitrogen) using the 1× dsDNA HS kit (Invitrogen) according to the manufacturer.

PCRs of isolated DNA samples were performed using custom-made barcoded primer pairs in unique combinations targeting the full-length 16S rRNA gene (Supplementary Table 1). The exact PCR composition was as follows: 0.5 μL of 10 μM primer each, 25 μL of 2× KAPA HiFi HotStart ReadyMix (Roche), 25 ng of template DNA and filled up to a final volume of 50 μL with PCR-grade water. In the case of carrot pomace and spent grains, 24 μL of the template DNA were used due to low DNA concentration. PCR cycling parameters were: 1 min at 95 °C, and 35 cycles with 95 °C for 20 s, 55 °C for 30 s and 72 °C for 2 min with a final extension at 72 °C for 5 min. The PCR products were purified with an Ampure XP bead clean-up (BECKMAN COULTER Life Sciences) according to the manufacturer using 30 μL beads. The purified PCR products were then used for the library preparation for the 16 s rRNA gene sequencing according to the Ligation sequencing amplicons protocol (SQK-LSK109; Oxford Nanopore), and sequencing was performed on a MinION sequencing platform according to the manufacturer (Oxford Nanopore).

SigmaPlot statistical software (Systat Software, Windows version 13.0, SigmaPlot, San José, CA, USA) was used to investigate the statistical significance of changes in mean larval live weight, ANOVA (p < 0.05) followed by Tukey HSD tests. To evaluate variability, nutritional and fatty acid compositions of the samples were statistically analysed by calculating means and standard deviation with Microsoft Excel. For microbial counts, mean values and standard deviations were calculated and logarithmised (log10) in Microsoft Excel. After sequencing, bases were called in super high accuracy mode using Guppy v6.4.8 (Oxford Nanopore), and demultiplexing was done using Minibar v0.25 (Krehenwinkel et al., 2019) with a barcode edit distance value of 11 and barcode trimming. Taxonomic identification was done using Emu v3.4.4 (Curry et al., 2022) with the prebuilt database (version from 18th August 2022) expanded by plant species likely occurring in the substrates (Supplementary file: List S1). The 16S sequences of the plants were taken from PhytoRef and added manually to the existing database (Decelle et al., 2015). Further data analysis was performed in R v4.2.2 (R Core Team, 2022), where first plant hits and unassigned hits were filtered out. The highest portion of plant and unassigned sequences was filtered out for the rape press cake feed substrate, with mean values at 21.5 and 38.2%, respectively. For the other feed substrates, portions of plant and unassigned sequences being filtered out ranged from 0 to 20.7% and from 1.2 to 34.4%. The larval data set contained no plant sequences, and portions of unassigned sequences ranging from 0.8 to 12.7% were filtered out. Further Shannon indices, the PCoA map based on Bray–Curtis dissimilarity approach and the heatmap were calculated using phyloseq v1.42.0 (McMurdie and Holmes, 2013). Additionally, graphs were produced with ggplot2 v3.4.1 (Villanueva and Chen, 2019).

Chemical composition of the agro-industrial by-products carrot pomace, brewer’s yeast, spent grains, and rape press cake used as feed substrates, as well as of the control feed, are listed in Table 1. Substantial differences across all chemical parameters were detected. Based on dry matter, crude protein, crude lipid, digestible carbohydrate, and crude fibre content of the feed substrates varied widely between 7.9 to 47.7%, 2.3 to 19.0%, 4.1 to 48.8%, and 0.5 to 28.4%, respectively. Crude protein was found to be highest in brewer’s yeast (47.7%), whereas crude lipid was highest in rape press cake (19.0%). Remarkably, carrot pomace showed the lowest contents of crude protein (7.9%) and crude lipid (2.3%), however, the highest content in crude fibres (28.4%). Regarding the fatty acid composition, the concentration of saturated, monounsaturated, and polyunsaturated fatty acids varied from 8.2 to 63.0%, 5.5 to 58.6%, and 8.2 to 66.5%, respectively (Table 1).

BSFL grew on all feed substrates, resulting in average individual larval biomasses of 210.5 mg, 184.3 mg, 150.2 mg, 42.2 mg, and 36.5 mg for the control feed, brewer’s yeast, rape press cake, spent grains and carrot pomace, respectively (Supplementary Table 2). As shown in Table 2, the chemical compositions of BSFL biomasses were strongly impaired by the type of feed substrate. Differences were highest in the crude lipid contents (approximately up to 6-fold), ranging from 5.8% for carrot pomace up to 30.3% for rape press cake. For the crude protein content, smaller differences were found (approximately up to 1.4-fold), being highest in the BSFL fed with brewer’s yeast (61.1%). A certain difference was also observed for the ash content, being lowest in the BSFL fed with the brewer’s yeast (5.6%) and highest using the carrot pomace as substrate (14.7%). The pH values of the BSFL biomass were 7.9, 6.6, 7.8, 6.9, and 7.1 when grown on the carrot pomace, brewer’s yeast, spent grains, rape press cake, and control feed, respectively. The data did not reveal a consistent or direct relationship between substrate chemical composition and nutrient profile of the BSFL (Table 2).

Determination of the compositions of saturated, monounsaturated, and polyunsaturated fatty acids of BSFL biomasses (Tables 2 and 3) showed an impact of the feed substrate (Table 1). Saturated fatty acid (SFA) contents were highest in the BSFL grown on brewer’s yeast (18.70%) and low when BSFL were fed with carrot pomace (3.46%) and spent grains (4.47%). In contrast, monounsaturated fatty acid (MUFA) contents were high in BSFL grown on rape press cake (13.70%), compared to the use of carrot pomace (1.24%) or spent grains as feed substrate. Similarly, polyunsaturated fatty acid (PUFA) contents were found to be highest in BSFL grown on rape press cake (7.03%) and lowest when grown on the brewer’s yeast. As shown in Table 3, lauric acid (C12:0) was by far the most abundant SFA in BSFL, ranging in content from 52.16% when grown in brewer’s yeast compared to rape press cake (18.37%), followed by palmitic acid (C16:0), ranging from 7.07 to 17.03%. Regarding the MUFAs, oleic acid (C18:1) was found in high concentrations, ranging from 11.80 to 40.04%, when the BSFL were grown on brewer’s yeast and rape press cake, respectively. Linoleic acid (C-18:2) was most abundant among the PUFAs and present in high concentrations in BSFL grown on carrot pomace, brewer’s yeast and rape press cake (Tables 2, 3). Moreover, the concentration of alpha-linolenic acid (18:3, ω-3) in BSFL also depended on the feed substrate and was found to be highest in BSFL grown on rape press cake (6.53%).

Microbial counts of the feed substrates were found to be generally low (<approximately 5.0 log₁₀ CFU/g), except for the count of mesophilic aerobic bacteria for spent grains (7.33 log₁₀ CFU/g, Table 4). In all feed substrates, counts of Enterobacteriaceae and yeasts were below the limit of quantification, except for the control feed (4.48 and 2.00 log₁₀ CFU/g, respectively). In contrast, after rearing, BSFL showed much higher microbial counts than the respective feed substrates (Table 4). Mesophilic aerobic bacterial counts ranged from 8.39 to 9.33 log₁₀ CFU/g, and counts of Enterobacteriaceae were found to be higher than 6.0 log₁₀ CFU/g. With carrot pomace, the BSFL biomass showed high counts of yeast and molds (>7.0 log₁₀ CFU/g). Again, no correlation was found between the microbial loads of the substrates and those of the larvae, independent of the microbial groups.

Alpha diversity at genus level was measured using the Shannon diversity index (Figure 1). BSFL reared on brewer’s yeast, spent grains and rape press cake had diversity levels in ascending order which are similar to the levels of the corresponding feed substrates (Figure 1). In contrast, BSFL reared on carrot pomace and control feed had intermediate diversity levels, while the feed substrate carrot pomace showed relatively low diversity, comparable to brewer’s yeast, and the control feed even higher diversity like the outlier of rape press cake. The findings were also reflected by the bacterial abundances at the genus level. The control and rape press cake feed harboured numerous different genera (>20) in comparable abundances (Figure 2), while the corresponding BSFL mainly contained Enterococcus (30–45%), Staphylococcus (14–29%), and Corynebacterium (10–16%) and Bacillus (21–31%), Pseudogracilibacillus (5–12%), Atopostipes (10%), and Brevibacterium (5–9%), respectively (Figure 3). In contrast, the carrot pomace feed (Figure 2) is mostly dominated by Weissella (70–73%,), while the corresponding BSFL (Figure 3) harboured a more diverse microbiota consisting among others of Paenibacillus (4–33%), Enterococcus (11–29%), Klebsiella (9–22%), and Sphingobacterium (3–13%). The other feed substrates were mainly dominated by one up to a couple bacterial genera, namely Lactobacillus (68–82%) in brewer’s yeast or Bacillus (44–49%), Paenibacillus (11%), Aneurinibacillus (11%), Thermobacillus (10–12%), and Brevibacillus (6%) in spent grains (Figure 2). The first three of these genera were also found in high relative abundances in the corresponding larvae (Figure 3), suggesting the presence of similar taxa in both feed substrates and larvae, which may reflect potential transfer and/or selective enrichment. In brewer’s yeast-fed larvae, Lactobacillus accounted for 10–38%, while in spent grains-fed larvae, Bacillus (38–42%) and Paenibacillus (15–19%) were dominant. Furthermore, the genus Enterococcus was consistently detected at high relative abundances in larvae regardless of the feed substrate (11–54%).

Further evidence for shared bacterial taxa between feed substrates and corresponding larvae in the case of brewer’s yeast and spent grains is provided by the heat map (Figure 4), which shows the relative abundance of the 50 most abundant species. For brewer’s yeast, Levilactobacillus brevis was detected in both the feed substrate and larvae, suggesting potential transfer or selective enrichment. This pattern was more pronounced for spent grains, where numerous Bacillus, Paenibacillus, Thermobacillus, Aneurinibacillus, and Brevibacillus species were present in both the feed substrate and corresponding larvae. Notably, around 50% of those bacterial species are also found in the brewer’s yeast feed substrate, though in lower abundances, but not in the corresponding larvae (Figure 4). To a lesser extent, Bacillus subtilis might also be transferred from the feed substrate, rape press cake and control feed to the corresponding larvae. Contrarily, it was also found in spent grains but not in the corresponding larvae and in brewer’s yeast-fed larvae but not in the substrate. Besides these findings, it becomes apparent that the microbiota of different reared larvae is much more similar to each other than to the corresponding feed substrates (Figure 4). This is supported by the presence of Actinomyces pacaensis, Klebsiella pneumoniae, Morganella morganii, and several Enterococcus species in the microbiota of all larvae (Figure 4). This bacterial microbiota conformity of the different larvae treatments is further supported by their beta diversity, which was measured using the Bray-Curtis method at the genus level (Supplementary Figure 1). Besides the feed substrate, spent grains and the corresponding larvae, all other feed substrates and likewise all other larvae were grouped, indicating that their microbiota are more similar to each other than to the corresponding larvae or feed substrate (Supplementary Figure 1).