Three previously well-characterized MDR-Ec strains were used for Z. morio larvae colonization experiments: Ec-4901.28 (ST131, CTX-M-15 ESBL producer) (Seiffert et al., 2013; Bernasconi et al., 2020), Ec-042 (ST410, OXA-181 carbapenemase producer) and Ec-050 (ST167, NDM-5 carbapenemase producer) (Endimiani et al., 2020). The susceptibility of these strains to the INTESTI bacteriophage cocktail (10 mL x 5 ampules, lot no. M2-1201) was tested using the double-layer agar method (DLA) (Clockie and Kropinski, 2009): Ec-4901.28 was susceptible (++; opaque lysis: turbidity throughout the cleared zone), whereas Ec-042 and Ec-050 were resistant (R; no clearing). The overall characteristics of the three MDR-Ec strains - including antibiotic resistance profiles, antimicrobial resistance genes (ARGs) and virulence factors – are summarized in Table 1.

Zophobas morio larvae (box of 40 g, ~60–80 larvae; BUGS-International GmbH) were purchased on different occasions from a Swiss pet store (QUALIPET) and kept in polypropylene containers (33x8x11 cm) at room temperature (23 ± 1°C) with regular day and night light conditions. Larvae were fed daily with slices of fresh pears (rinsed three times with distilled water and then surface disinfected with 70% ethanol) and dry food for cats (~20 g and ~ 3 g, respectively / container) mixed in a substrate of oat flakes (height, ~1 cm). This diet is fully compatible with the physiology of Z. morio larvae (Rumbos and Athanassiou, 2021). Moreover, it is commonly used by professionals to bread larvae used for feeding reptiles in captivity (data not shown). Adult larvae (700 ± 50 mg and 45 ± 5 mm in length) were used for all experiments after 2–3 days of acclimatization (Figure 1A).

Notably, direct culture-based analyses on MacConkey II and Columbia sheep blood agar plates (Becton-Dickinson) revealed that disinfected pears were contaminated with Bacillus cereus [~3 × 10¹ colony forming units (CFU)/mL], whereas dry food contained Moraxella osloensis and Micrococcus luteus (~2–5 × 10² CFU/mL); the oat substrate was constantly negative. Moreover, the three types of food were regularly screened for the presence of extended-spectrum cephalosporin-resistant (ESC-R)-Ent using an overnight broth enrichment and selective agar plates as previously done (Endimiani et al., 2020; Campos-Madueno et al., 2023a). As a result, no food used throughout experiments resulted contaminated with ESC-R-Ent (data not shown).

To induce gut colonization with the MDR-Ec, larvae (at least n = 48) were fed for 1 week [from day 0 (T0) to day 7 (T7)] with food contaminated with the specific strain (Figures 1B,F). In particular, colonies grown on a MacConkey II plate were incubated overnight with 10 mL Luria-Bertani (LB) broth (Thermo Fisher Scientific) (Figure 1C). After that, the overall broth was poured into a sterile Petri dish containing slices of pears and dry food, and then further incubated for 1 h at 36 ± 1°C before administration to larvae (Figures 1B,D). Notably, bacterial concentration in the overnight LB broth was ~10⁹ CFU/mL for each of the three tested MDR-Ec strains (colony count performed; data not shown).

After T7, larvae were transferred to a new clean cage with a fresh oat substrate where they received non-contaminated food for the next 21 days. For the overall 28 days, the administered food (from T0 to T7 contaminated and from T7 to T28 non-contaminated) was removed every day using sterile tweezers and replaced with a fresh one (Figure 1F).

Colonization experiments were replicated at least 4 times for each of the three MDR-Ec strains tested (Supplementary Table S1). Finally, 4 experiments where larvae received only non-contaminated food from T0 to T28 were also performed as negative controls (Neg-Cs; Figure 1F).

The INTESTI cocktail was used to test its potential to decolonize larvae from the MDR-Ec strains. This cocktail is a sterile-filtrate phage lysate [~10^5–6 plaque forming unit (PFU)/mL] of E. coli, Shigella spp., Salmonella spp., Proteus vulgaris/mirabilis, Pseudomonas aeruginosa, Staphylococcus spp., and Enterococcus spp. strains. It was fully characterized with a metagenomic approach (Zschach et al., 2015).

Larvae receiving bacteriophages followed the same protocol as the colonization experiments (see above). However, a sub-group of larvae (at least n = 16) received 10 μL of INTESTI cocktail administered per os by force-feeding on both T7 and T10 using a blunt 26 s-gauge needle connected to a 250 μL Gastight syringe (Hamilton) (Figures 1E,F). For each MDR-Ec tested, force-feeding with bacteriophages was replicated in 3 experiments.

As commonly done (e.g., Zhang et al., 2019), colonized larvae were also subjected to 2 further force-feeding experiments in which on both T7 and T10 they received 10 μL of sterile 1X Dulbecco phosphate-buffered saline (1X dPBS; Biochrom GmbH) used as control (Figure 1F; Supplementary Table S1). Notably, larvae were inspected for the first 30 min after any force-feeding and, in case of vomiting, they were discharged and replaced by new ones.

For each experiment, 4 random larvae were simultaneously sampled at 8 specific time points: T0 (before administering food contaminated with the MDR-Ec), T2, T4, T7 and T10 (both before the corresponding force-feeding with bacteriophages or 1X dPBS, if any), T14, T21, and T28. For simplicity, samples taken during experiments with the administration of bacteriophages will be hereafter referred to as T10+, T14+, T21+, and T28+ (Figure 1F).

The 4 larvae taken at each time point were placed inside a 50 mL polypropylene tube (TPP Techno Plastic Products AG) and euthanized by placing them at −20°C for 1 h. After that, their exoskeleton was disinfected by adding 20 mL of 70% ethanol for 3 h at 4°C. The Precellys Evolution Touch tissue homogenizer apparatus (Bertin Technologies) was then implemented to homogenize the 4 disinfected larvae. In particular, two 7 mL high-impact bead beating tubes containing 1.4 mm ceramic beads (Labgene Scientific SA) plus 2 mL of sterile 1X dPBS were loaded with 2 larvae each. Tubes were processed as follows: 20 s/5500 rpm, break 30 s, and 20 s/5500 rpm. Then, the 2 homogenized liquid samples were pooled into a single 15 mL conical tube (Sarstedt AG & Co.) for a final volume of ~3–4 mL.

Final samples were properly diluted in sterile 1X dPBS and aliquots of 100 μL were plated on selective ChromID® ESBL agar plates (bioMérieux), followed by an overnight incubation at 36 ± 1°C in ambient air to detect ESC-R Gram-negatives. After bacterial species identification (see below), colony count (CFU/mL) for the specific MDR-Ec strain (ESC-R due to the production of ESBLs and/or carbapenemases) used to colonize the larvae was performed for all the 8 time points. Notably, colony count results were expressed as CFU/mL consistently with similar studies (e.g., Cools et al., 2019). All homogenized samples were stored at −80°C in 2 mL cryogenic Cryo.s tubes (Greiner Bio-One) supplemented with 20% glycerol.

Colonies grown on ChromID® ESBL agar plates (i.e., ESC-R) were identified at the species level using the matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS; Brucker).

Six ESC-R-Ec strains obtained from different experiments at T28/T28+ (or T21/T21+ if no bacterial growth at T28/T28+) also underwent antimicrobial susceptibility tests (ASTs), phenotypic assay for the INTESTI bacteriophage cocktail, whole-genome sequencing (WGS) and single nucleotide variant (SNV) analyses. All results were compared for consistency with the original MDR-Ec strains (Table 1) administered from T0 to T7 with the contaminated food.

ASTs were performed using the MIC microdilution Sensititre GNX2F panels (Thermo Fisher Scientific) and interpreted according to the 2023 European Committee on Antimicrobial Susceptibility Testing (EUCAST) breakpoints for Ent (v13.0).¹ Susceptibility to bacteriophages was evaluated using the DLA method (Clockie and Kropinski, 2009).

WGS was achieved using both short- and long-read sequencing. Short-read sequencing was conducted with the Illumina NovaSeq 6,000 platform applying the NEBNext® Ultra™ II DNA library prep kit (2 × 150 bp paired-end reads) by Eurofins Genomics (Ebersberg, Germany), while for long-reads the MinION sequencer (Oxford Nanopore Technologies) was implemented using a rapid barcoding library prep (SQK-RBK004) and FLO-MIN 106D R9.4.1 flow cells. Sequencing adaptors were removed from both short- and long-reads using Trimmomatic v0.36 and Porechop v0.2.3, respectively. Complete and circular genomes were generated using Unicycler v0.4.8² using the hybrid pipeline as previously performed (Campos-Madueno et al., 2020, 2021, 2022). Results were interpreted using the tools of the Center for Genomic Epidemiology,³ such as ResFinder v4.1, VirulenceFinder v2.0, PlasmidFinder v2.1, and MLST (all with default parameters).

For the SNV analyses, individual chromosomal and plasmid sequences from every strain were used to generate core-genome alignments using Parsnp v1.7.4⁴ as previously described (Bernasconi et al., 2020; Campos-Madueno et al., 2022; Moser et al., 2022). Each chromosomal and plasmid sequences of the original MDR-Ec strains were used as a reference to the recovered ESC-R-Ec for core-genome alignment. The recombination filtration (−x parameter) was used and the rest of the parameters were set as default. Alignment coverage of the core-genome was determined automatically with Parsnp. SNV sequences were extracted from the Variant Call Format (VCF) output using Harvest-Tools v1.2.⁵ Only high quality SNVs (PASS) were considered.

The DLA method was also implemented for bacteriophage enumeration (titration) in the homogenized larvae tissues as previously done (Bernasconi et al., 2020). In particular, samples from larvae treated with bacteriophages (250 μL of tissues diluted 1:1 in 1X dPBS) were filtrated using a 0.22 μm pore size PES sterile syringe (Carl Roth Gmbh). The titration was performed for the following time points; T7 (null control), T10+, T14+, T21+, and T28 + .

Briefly, 1.5% Brain Heart Infusion (BHI; Becton-Dickinson) agar was prepared and distributed as a first layer in a sterile Petri dish. Then, 100 μL of the filtrated tissues were added to a 15 mL tube containing 1 mL of BHI broth and incubated for 5 min at room temperature. Subsequently, 100 μL of the previously characterized phage susceptible E. coli strain 56-M3-Ec (++++; confluent lysis: complete clearing) grown overnight in BHI broth (Bernasconi et al., 2017) plus 4 mL of 0.6% of BHI agar were added to the 15 mL tube and then distributed on top of the first agar layer. Upon an overnight incubation at 36 ± 1°C, confluent lysis plaques on the agar plates were counted for the viral titer (PFU/mL).

For each MDR-Ec strain tested, microbiota was analyzed in three different experiments (named A, B, C for simplicity; Supplementary Table S1) where, at T7, larvae were split in two groups: non-treated larvae (samples analyzed: T0, T7, T14, and T28) and larvae receiving bacteriophages (samples analyzed: T14+ and T28+). Finally, microbiota was also analyzed at T0, T7, T14, and T28 for larvae not receiving neither contaminated food nor bacteriophages (i.e., the 4 Neg-Cs; see above).

DNA from the homogenized samples was extracted using the QIAamp PowerFecal Pro DNA Kit (Qiagen). Total gDNA was purified using CleanNA CleanNGS purification beads (Labgene) and resuspended in 10 mM Tris–HCl buffer at pH 8.0 (Sigma-Aldrich). DNA quantification and purity were determined using the NanoDrop™ One/OneC Microvolume UV–Vis spectrophotometer (Thermo Fisher Scientific) and Qubit™ 3.0 Fluorometer (Invitrogen).

Larvae microbiota characterization was achieved through 16S rRNA amplicon sequencing. gDNA extracts were sent for Illumina sequencing to Microsynth AG.⁶ For library preparation, the 16S rRNA V4 region was subjected to Nextera two-steps PCR amplification using the 515F Parada primer (GTGYCAGCMGCCGCGGTAA), and 806R Apprill primer (GGACTACNVGGGTWTCTAAT) (Walters et al., 2016). Sequencing data were generated as adaptor trimmed, demultiplexed and quality checked raw reads in fastq format. Notably, sequence counts of all samples were above 23,000 reads. Identification of the Amplicon Sequence Variants (ASVs) was generated using DADA2 v1.26.0⁷ R package. Taxonomy was assigned using the SILVA-based (v138.1) bacterial reference alignment. Microbial community analysis was performed using the phyloseq v1.38.0⁸ package for R v4.4.2 (McMurdie and Holmes, 2013). The abundance of bacterial taxonomic composition was obtained at the genus level using the DADA2 pipeline and the mirlyn v1.4.0 R package (Moor et al., 2021).

Diversity indexes were determined based on the entire ASV data obtained. The output generated by the DADA2 pipeline was imported as a phyloseq object in R for diversity analyses (McMurdie and Holmes, 2013).

The alpha diversity was analyzed obtaining both richness and the Shannon diversity index (SDI) (Kim et al., 2017). Before conducting within samples comparison, the phyloseq object was normalized using the TMM method (Trimmed Mean of M-value) from the edgeR v3.36.0 R package. Beta diversity was then estimated using the Bray–Curtis dissimilarity index (BCDI) using the vegan package v2.6.4⁹ and the phyloseq package, while the compositional differences between the samples were tested using the permutational analysis of variance (PERMANOVA).

All statistical analyses were performed using GraphPad Prism version 9.4.0 for Windows (GraphPad Software). CFU/mL count data were analyzed starting from T10/T10+ to assess differences in bacterial growth rates between the three groups (i.e., larvae not receiving bacteriophages, larvae receiving bacteriophages, and larvae receiving 1X dPBS) using the two-tailed F test of variance.

Microbiota beta diversity of BCDI mean values for (i) Neg-C experiments #1, #2, #3 and #4, and (ii) experiments A, B and C for the three tested MDR-Ec strains, were, respectively, statistically compared using two-way ANOVA, followed by Tukey post hoc test. A p-value <0.05 was considered to be statistically significant.

All 16S rRNA gene sequenced samples have been deposited in GenBank under BioProject accession number PRJNA992250. The re-assembled 4901.28_2 strain is deposited under PRJNA551948 BioProject, BioSample ID SAMN38456712. Whole-genome sequences of the 6 MDR-Ec strains are deposited under BioProject accession number PRJNA1045999.

A summary (mean) of the growth curve experiments for larvae fed with the three MDR-Ec strains is shown in Figure 2 (left panel). Notably, during all of these colonization experiments none of the Z. morio larvae died.

After administering the food contaminated with the ST131 Ec-4901.28 strain, Z. morio larvae became rapidly colonized with a high bacterial load (T7 = 6.95×10⁶ CFU/mL). The removal of contaminated food on T7 induced a drop in bacterial count (T10 = 6.91 × 10⁴ CFU/mL), but this value remained constant in the larvae until the end of experiments (T28 = 5.07 × 10⁴ CFU/mL). Notably, for this specific MDR-Ec strain, experiment #6 was extended to 35 days, with the bacterial count still being 1.19×10⁴ CFU/mL (Supplementary Table S1).

The ST410 Ec-042 strain, showed a similar capacity to colonize Z. morio larvae (T7 = 9.22×10⁶ CFU/mL) as Ec-4901.28, although at the end of experiments bacterial load was slightly lower (T28 = 7.60×10³ CFU/mL). In contrast, the ST167 Ec-050 strain displayed a different colonization behavior. Specifically, feeding with contaminated food generated a colonization load of 1.71 × 10⁷ CFU/mL on T7, but then, bacterial count in larvae rapidly declined to 0 CFU/mL on T28.

Colony counts for homogenized larvae that underwent force-feeding with the INTESTI bacteriophage cocktail, together with the corresponding viral titers, are summarized in Figure 2 (left panel). Statistical analysis of the overall growth curves is also shown in Figure 2 (right panel). Notably, ~15% of force-fed larvae were discharged because of vomiting, while none of those remaining in the experiments died by T28+.

For the INTESTI-susceptible Ec-4901.28 strain, treatment with bacteriophages induced a decrease in the MDR-Ec load that ended up with CFU/mL count lower than the limit of detection (LOD) on T28+. This final effect was consistent with a high viral titer recorded at T14+ (1.22×10⁴ PFU/mL); such titer decreased to 0 on T28+, in parallel to the disappearance of Ec-4901.28. Notably, treatment with 1X dPBS (control) did not have any effect (T28 = 1.03×10⁵ CFU/mL), as shown by the resultant bacterial growth curve that was similar to the one obtained for larvae not receiving the bacteriophage cocktail. Statistical analysis showed that the Ec-4901.28 growth curves for larvae receiving and not receiving bacteriophages displayed a significant difference to each other from T14/T14+ (p = 0.042) to T28/T28+ (p < 0.001). In contrast, at T28, growth curves for non-treated larvae and those receiving 1X dPBS were not statistically different.

Bacteriophages administration decolonized Z. morio larvae from the INTESTI-resistant Ec-042 on T21+ (CFU/mL count <LOD). This effect was in line with the viral titer curve at T14+ (2.10×10² PFU/mL) and T21+ (0 PFU/mL). However, after T21+, Ec-042 started to regrow and larvae returned colonized with this pathogen (T28 + =6.33 × 10⁴ CFU/mL). Analysis of variance indicated that the Ec-042 growth curves of larvae treated and non-treated with bacteriophages were significantly different from T14/T14+ to T28/T28+ (p < 0.001). Of note, the 1X dPBS treatment curve resembled the curve referring to the bacteriophage’s treatment more than to the one obtained for untreated larvae.

With regard to the INTESTI-resistant Ec-050 strain, the double dose of bacteriophages did not have an effect on the bacterial load of larvae (T7 = 1.71×10⁷ CFU/mL to T14+ =2.07×10⁴ CFU/mL), although the viral titer reached its peak on T14+ (5.87×10² PFU/mL). In contrast, administration of 1X dPBS led to a decrease of the CFU/mL count of Ec-050 under the LOD within T21. However, additional growth curve experiments indicated that 1X dPBS had no effect on the in vitro growth of Ec-050 (Supplementary Table S2, including methods implemented). Finally, statistical analyses showed that the three bacterial growth curves for this MDR-Ec strain were significantly different to each other from T14/T14+ to T28/T28+ (e.g., T28/T28+: all p < 0.001).

The six MDR-Ec strains that underwent full characterization were obtained as follows: experiment #9 (T28) and #11 (T21+) for Ec-4901.28, experiment #2 (T28) and #6 (T28+) for Ec-042, and experiment #2 (T21) and #7 (T28+) for Ec-050.

Antibiotic/bacteriophage susceptibility profiles and genetic backgrounds (e.g., ARGs, plasmids, ST) of the above six strains were consistent with the corresponding original MDR-Ec administered with the contaminated food (Table 1). Moreover, the chromosomal SNVs recorded for Ec-4901.28, Ec-042 and Ec-050 at the end of in vivo experiments were: T28= 14 and T21+ =29 SNVs; T28 = 5 and T28+ =26 SNVs; and T21= 13 and T28+ =16 SNVs, respectively (Supplementary Table S3).

Relative abundance of bacterial genera and SDI of larvae’s microbiota at T0 are described in Figure 3A and Supplementary File S1.

Microbiota patterns for the 12 T0 samples (including the 4 Neg-Cs) appeared heterogeneous, with SDIs ranging between 1.1 and 2.7. Furthermore, comparison to each other (beta diversity) demonstrated BCDIs ≥30% and up to 73% (Figure 3B). Nevertheless, these different initial microbiota patterns had a frequent presence of the following bacterial genera (relative abundance, range): Lactococcus (7–88%), Enterococcus (2–36%), Spiroplasma (1–22%), Hafnia-Obesumbacterium (0–31%), Pediococcus (0–22%), Cronobacter (0–13%), Latilactobacillus (0–11%), Klebsiella (0–7%), and Citrobacter (0–5%) spp.

The larvae microbiota dynamics during the 28 days in the 4 Neg-Cs are depicted in Figure 4, Supplementary Figure S1, and Supplementary File S1.

Considering the relative abundance, two genera were constantly present from T0 to T28: Lactococcus (rates up to 88 and 46%, respectively) and Spiroplasma (rates up to 21.5 and 33.4%, respectively). Moreover, the SDI ranges for Neg-Cs #1, #2, #3 and #4 were 1.1–2.0, 1.6–2.5, 1.5–3.0 and 1.9–2.2, respectively (Figure 4A). Although the BCDIs between samples of the same experiment were ≥ 20% and up to 66%, statistical comparison showed that they were not significantly different (Figure 4B).

The dynamics of relative genus abundance and SDI recorded in three different experiments (A, B, and C) performed with or without administering bacteriophages are depicted in Figure 5 (left panel), Supplementary Figure S2, and Supplementary File S1.

Overall, several general aspects can be noted. First, the Escherichia-Shigella genera appeared in larvae only at T7 (relative abundance range: 1–20%), but it mostly disappeared (<1% abundance) at T14/T14+, regardless of the administration of bacteriophages. Second, Enterococcus spp. tended to decrease from the beginning of colonization experiments and, irrespective of the use of INTESTI cocktail, its relative abundance at T28/T28+ was usually lower than at the corresponding T0. Third, when Citrobacter spp. was part of the microbiota at T0, the use of bacteriophages was related to its elimination at T28+; this phenomenon was only partially true when Klebsiella spp. was naturally present in the microbiota of larvae. Last, in most of the experiments, community diversity (SDI) of individual samples at T28/T28+ was slightly lower than the initial microbial diversity at T0; although a statistical difference between the BCDIs was not recorded (Figure 5, right panel).

Notably, in the experiments where the INTESTI-susceptible strain Ec-4901.28 was used, the bacterial diversity after the administration of bacteriophages was always lower than when no treatment was implemented (e.g., in experiment A: SDI = 1.9 at T28+, whereas SDI = 2.2 at T28); however, no statistical differences between the BCDIs were observed (Figure 5, right panel).

To establish our new model, larvae were first tested with three clinically and epidemiologically relevant MDR-Ec strains that nowadays are frequently responsible for gut colonization in humans and animals (Table 1). Consequently, larvae were rapidly colonized with a high bacterial load of > 10⁶ CFU/mL at T2 and > 10⁷ CFU/mL at T7. Nevertheless, each specific strain had a different colonization behavior after the cessation of contaminated food administration (Figure 2).

The first strain tested (Ec-4901.28) was a CTX-M-15 producer of ST131 lineage, being globally the most dominant extra-intestinal pathogenic E. coli strain often responsible for community- and hospital-acquired urinary-tract infections and/or bacteremia (Seiffert et al., 2013; Peirano and Pitout, 2019; Pitout and Finn, 2020). The strain was able to persist within the larvae with a stable bacterial load from T10 to T28 (~10⁵ CFU/mL; Figure 2). This phenomenon was also observed in our previous work where we used a bioreactor to simulate the human gut (Bernasconi et al., 2020). Overall, these results can be explained by the molecular features of ST131 that favor its fitness and long-term intestinal colonization capacities (Pitout and Finn, 2020). In fact, Ec-4901.28 carries a higher number of virulence factors (VFs) that aid the adherence/invasion of host intestinal epithelial cells, siderophores to scavenge iron, bacteriocins and corresponding immunity proteins that confer an advantage over sensitive bacterial populations (Table 1).

The second MDR-Ec strain used (Ec-042) was an OXA-181 producer belonging to ST410, a rapidly emerging pandemic clone particularly able to colonize the intestinal tract of humans and animals (Nigg et al., 2019; Moser et al., 2021; Pitout et al., 2024). As recorded for Ec-042 (Table 1), the ST410 lineage possesses patterns of VFs (e.g., fimbriae, bacteriocins) very similar to those of ST131 (Roer et al., 2018; Feng et al., 2019; Pitout et al., 2024). Therefore, it is not surprising that Ec-042 also rapidly colonized Z. morio larvae and persisted with a high bacterial load until T28 (~10⁴ CFU/mL; Figure 2).

In contrast to the above two MDR-Ec, Ec-050 expressed the weakest colonization effect on larvae as no colonies were present after T21 (Figure 2). This strain produces the NDM-5 carbapenemase and belongs to another worldwide emerging high-risk clone (ST167) that spreads in human and non-human settings (Endimiani et al., 2020; Schmidt et al., 2020; Linkevicius et al., 2023). As for Ec-4901.28 and Ec-042, Ec-050 carried numerous genes encoding VFs (Table 1). Therefore, we are unable to clearly explain its inability to persistently colonize larvae. Nevertheless, based on the findings of larvae receiving bacteriophages, a hypothesis can be raised (see below).

As a proof of concept to test new strategies to decolonize intestinal carriers of MDR-Ec (Campos-Madueno et al., 2023b), our colonized larvae were challenged with two doses of INTESTI bacteriophage cocktail at T7 and T10 (Figure 2).

Numerous studies have shown that bacteriophages are highly active in vivo against infections due to MDR-Ec. In contrast, data regarding the effect of bacteriophages against the intestinal colonization due to MDR-Ent in in vivo models are scarce (Lin et al., 2017; Campos-Madueno et al., 2023b). Using four lytic phages, Javaudin et al. were unable to decolonize mice carrying ESBL- or OXA-48-producing E.coli strains (Javaudin et al., 2021). Fang et al. showed that administration of two lytic phages to mice colonized with a carbapenem-resistant K. pneumoniae strain generated phage-resistant mutants (Fang et al., 2022). In the study of Mirza et al., administration of two bacteriophages decreased the CFU count of a carbapenemase-producing K. pneumoniae colonizing the gut of G. mellonella, but the difference was not statistically significant (Mirza et al., 2024).

In our study, larvae were decolonized from the INTESTI-susceptible strain Ec-4901.28 within T28+ (p < 0.001), and no phage-resistant mutants were detected (Figure 2). This latter phenomenon was unexpected, since in our past bioreactor experiments with Ec-4901.28 mutants were sometimes isolated, though their mechanism of resistance was not elucidated by WGS analyses (Bernasconi et al., 2020). We therefore speculate that the combination of the multiple lytic phages into the INTESTI cocktail, together with the host immune response, reduced the chances of developing resistance against bacteriophages (Lin et al., 2017).

On the other hand, administration of the INTESTI cocktail was ineffective for Z. morio larvae colonized with phage-resistant strains (Figure 2). In the case of Ec-042, there was initially a difficult to explain decrease of the pathogen load (less than the LOD) followed by a rapid regrow (~10⁵ CFU/mL at T28+). More interestingly, Ec-050 was not at all affected by the bacteriophages activity and maintained a load significantly higher than during experiments with larvae not receiving treatment (T28 vs. T28+: p < 0.001). To explain this interesting data, we hypothesize that there are specific bacterial competitors of Ec-050 in the intestinal tract of Z. morio larvae that protect against its colonization (a phenomenon known as “colonization resistance”) (Caballero-Flores et al., 2023). Since the INTESTI cocktail contains bacteriophages inhibiting multiple species (Zschach et al., 2015) - including the hypothetical competitors - Ec-050 could persistently colonize larvae (~10^4–5 CFU/mL at T28+).

During the experiments with Ec-050, we also noted a strong decolonization activity generated by the administration of 1X dPBS (control; Figure 2). This surprising phenomenon did not find an explanation after verifying that Ec-050 - as for Ec-4901.28 and Ec-042 - is not inhibited by 1X dPBS in vitro (Supplementary Table S2). Therefore, we speculate that the use of 1X dPBS might: (i) stimulate some species part of the natural larvae microbiota to produce inhibitors (e.g., bacteriocins) against Ec-050 (Wang et al., 2024); (ii) affect the characteristics of the gut (e.g., expression of glycoproteins in the epithelium) in an unfavorable way for the Ec-050 colonization (Foley et al., 2021); and/or (iii) improve the host immune response against Ec-050 (Lange et al., 2019; von Bredow et al., 2023).

Recently, we have shown that Ec-042 and Ec-050 were genetically highly stable after 20 propagation steps on selective agar plates (0 and 0–9 SNVs, respectively) (Moser et al., 2022).

In the present work (Supplementary Table S3), we demonstrated that the three hyperepidemic strains tested behave in a similar way during the in vivo experiments with larvae (e.g., 14, 5, and 13 chromosomal SNVs for Ec-4901.28, Ec-042, and Ec-050, respectively), though administration of bacteriophages seems to slightly increase the chance of mutations (e.g., 29, 26, and 16 chromosomal SNVs, respectively). This is consistent with the notion that bacteriophages stress may select for mutants that frequently involve surface receptors (Chevallereau et al., 2022). Overall, our data indicate that the three MDR-Ec strains used in our experiments possess stable phenotypic and molecular features that make them very useful for any kind of reproducible in vivo experiment.

Zophobas morio larvae acquired from the pet shop possessed a very diversified gut microbiota, with Lactococcus, Enterococcus and Spiroplasma as dominant and constantly present bacterial genera (Figure 3; Supplementary File S1). This variety may possibly be due to the type of food administered to larvae during their industrial breeding, although the origin and type of these sources are not stated by the provider.¹⁰ In this context, some authors have reported that restaurant/household/gardening waste, slaughterhouse products and animal manure might be used to breed Z. morio larvae (Harsanyi et al., 2020). As a consequence, a high load of human/animal derived bacteria (e.g., Enterococcus spp. and Ent) could be ingested by the larvae.

We emphasize that the natural richness of bacterial species found in Z. morio larvae has no correspondence to other invertebrates (e.g., G. mellonella larvae) or genetically modified vertebrates (e.g., gnotobiotic mice) that are frequently used for laboratory experiments (Allonsius et al., 2019; Darnaud et al., 2021). Moreover, the Z. morio microbiota recorded at different T0 samples (Figure 3) has a certain stability over time (Figure 4).

When colonization with the MDR-Ec strains was induced, larvae microbiota gradually showed a non-significant reduction of the bacterial diversity (SDI) during the 28-day experiments (Figure 5). Since the presence of the colonizing MDR-Ec was recorded only at T7 (i.e., Escherichia-Shigella genus), we speculate that this slight diversity reduction was mainly due to the fixed diet that we administered to larvae rather than the direct effect of the colonizer per se (Mason et al., 2020). As anticipated above (see methods), oats, pears and dry food were not contaminated with Enterococcus spp. and Ent, justifying the reduction of these bacterial species in the gut of Z. morio larvae.

As previously observed in humans and animal models (Hong et al., 2016; Febvre et al., 2019), the use of bacteriophages generated in larvae a non-significant reduction in bacterial diversity as for the experiments without treatment; this was independent on the specific MDR-Ec strain tested (Figure 5). Moreover, though the microbiota of larvae receiving or not receiving bacteriophages showed to be quite dissimilar (e.g., mean of BCDI at T28+ vs. T28 of 37–52%), such patterns were not significantly different. Nevertheless, administration of bacteriophages proved to neutralize Citrobacter spp. that by definition should be susceptible to the INTESTI cocktail (Zschach et al., 2015). In contrast, Klebsiella spp. was not affected due to the lack of lytic phages into the cocktail and some residual Enterococcus spp. persisted at T28+ probably due to their phage-resistant profile. Overall, our results confirm that bacteriophages are highly host-specific and have limited effect on the untargeted natural gut bacteria when implemented (Hong et al., 2016; Febvre et al., 2019).