D-alanine esterification of teichoic acids contributes to Lactobacillus plantarum mediated Drosophila growth promotion upon chronic undernutrition

Summary

The microbial environment influence animal physiology. However, the underlying molecular mechanisms of such functional interactions are largely undefined. Previously, we showed that upon chronic undernutrition, strains of Lactobacillus plantarum, a major commensal partner of Drosophila, promote host juvenile growth and maturation partly via enhanced expression of intestinal peptidases. By screening a transposon insertion library of Lactobacillus plantarum in gnotobiotic Drosophila larvae, we identify a bacterial cell wall modifying machinery encoded by the pbpX2-dlt operon that is critical to enhance host digestive capabilities and promote animal growth and maturation. Deletion of this operon leads to bacterial cell wall alteration with a complete loss of teichoic acids D-alanylation. We show that L. plantarum cell walls bearing D-alanylated teichoic acids are directly sensed by Drosophila enterocytes to ensure optimal intestinal peptidase expression and activity, juvenile growth and maturation upon chronic undernutrition. We thus conclude that besides peptidoglycan, teichoic acids modifications participate in the host-commensal bacteria molecular dialogue occurring in the intestine.

Introduction

Metazoans establish complex interactions with their resident microorganisms for mutual benefits^1–3. When in homeostasis, these interactions contribute to different aspects of host physiology^3–5. In the gut, microbial communities enhance digestive efficiency by providing enzymatic functions that help their hosts optimize extraction of dietary energy and nutrients. In addition, the gut microbiota promotes proper immune system development, local immune homeostasis and limits pathogen colonization^6–8. Despite the renewed interest in understanding the functional impact of gut microbiota on host physiology, a clear view of the molecular dialogue engaged upon host/microbiota interaction remains elusive^9,10. Therefore the use of simple animal models, such as Drosophila, may help unravel the evolutionarily conserved mechanisms underlying the impact of intestinal bacteria in their host physiology, since it combines genetic and experimental tractability with a cultivable microbiota and the ease to generate germ-free animals^11,12.

Drosophila gut microbiota is composed of simple and aerotolerant bacterial communities (mostly Acetobacteraceae and Lactobacillaceae families) with five prevalent species: Acetobacter pomorum, Acetobacter tropicalis, Lactobacillus brevis, Lactobacillus plantarum and Lactobacillus fructivorans^13,14. The genus Lactobacillus gathers bacteria with high phylogenetic and functional diversity^15,16. They have been largely used as model lactic acid bacteria and are recognized as potential health beneficial microorganisms in the human gastrointestinal tract^17,18. As a prevalent member of Drosophila microbiota, L. plantarum is involved in several aspects of host physiology such as social behaviour^11,19, protection against infection^3,20, gut epithelial homeostasis^21,22, nutrition^23–25 and post-embryonic development^3,4,23,26. We previously reported that, upon chronic undernutrition, certain strains of L. plantarum (such as L. plantarum^WJL, Lp^WJL) fully recapitulate the beneficial effect of a more complex microbiota by promoting Drosophila juvenile growth and maturation rate^4,23. Lp^WJL exerts its beneficial effect on larval growth through the host nutrient sensing system that relies on the tissue specific activity of the TOR kinase, which subsequently modulates hormonal signals controlling growth and maturation^4,27. Importantly, using conventional and gnotobiotic mice we recently demonstrated that the intestinal microbiota and some strains of L. plantarum also influence linear growth in mammals^3,11. These results suggest that the still unknown molecular mechanisms underlying microbiota-mediated juvenile growth promotion are likely conserved during evolution. Recently, we showed that L. plantarum influences juvenile growth at least partly through the increased expression of a set of specific host digestive enzymes in the intestine^1,3. We have shown that Lp^WJL promotes the expression of peptidases, such as Jon66Ci and Jon66Cii, in the enterocytes in both PGRP-LE/Imd/Relish-dependent and independent manner. The resulting enhanced peptidase activity in the midgut increases the digestion of dietary proteins into dipeptides and amino acids as well as their uptake. Circulating dipeptides and amino acids are sensed in endocrine tissues by the TOR kinase pathway, which promotes the production of Drosophila growth factors, the insulin-like peptides (dILPs) and a precocious pic of production of the molting steroid hormone Ecdysone^3,4.

Here, we aim to identify the bacterial genetic determinants involved in L. plantarum-mediated juvenile growth promotion. Through the generation of a random transposon-mediated insertion library in the growth promoting strain L. plantarum^NC8 (Lp^NC8), we identified a set of transposon insertions altering host growth promotion. Among these insertions, we further characterized the insertion in the pbpX2-dlt operon with predicted functions in cell wall biogenesis and remodelling. Deletion of this operon alters the bacterial cell wall due to complete loss of teichoic acids D-alanylation, and significantly impairs Drosophila larval growth, maturation and intestinal peptidases expression and activity. Our analysis points to the existence of additional host sensing and signalling mechanisms, besides PGRP-LE/Imd/Relish, involved in the sensing of the cell wall features defined by the gene products of pbpX2-dlt operon. Taken together our results demonstrate that D-alanine esterification of teichoic acids directly contributes to L. plantarum mediated Drosophila growth promotion upon chronic undernutrition.

Results

Generation of a L. plantarum^NC8 random-mutagenesis library

To identify L. plantarum genes required to sustain Drosophila juvenile growth and maturation, we adopted a classical unbiased forward genetic approach via transposon-mediated mutagenesis of the bacterial chromosome coupled to a phenotypical screening in mono-colonized animals. The previously characterized strain Lp^WJL has low transformation efficiency (Matos R., unpublished results) and carries plasmids^6,8,28,29 that preclude random transposition in the bacterial chromosome. We therefore chose a L. plantarum strain with high transformation efficiency, no native plasmids and capable to promote host growth in mono-association experiments to the same extent as Lp^WJL. This strain, designated as Lp^NC8, is suitable for transposon mutagenesis library construction^3,9. Upon mono-colonization both strains strongly support Drosophila’s larval linear growth (Fig. 1a) and maturation (i.e entry to metamorphosis; Fig. 1b) under chronic undernutrition when compared to the moderate growth promoting strain L. plantarum^NIZO2877 (Lp^{NIZO2877 11,30} or the germ-free (GF) condition (Fig. 1a-b). Thus, we mutagenized the Lp^NC8 chromosome using the P_junc-TpaseIS₁₂₂₃ transposon mutagenesis system previously described by^3,13. This system was developed for lactic acid bacteria and has been successfully applied to Lactobacillus casei^3,15 and Lactobacillus pentosus^9,17. It couples a thermo-sensitive plasmid (pVI129) expressing transiently the IS₁₂₂₃ transposase and a suicide plasmid (pVI110) encoding the IS₁₂₂₃ transposon, which together lead to random insertion of IS₁₂₂₃ sequences into the bacterial chromosome. Strain NC8pVI129 (Supplementary Table 1) was transformed with pVI110 and 2091 colonies were randomly selected and stocked as individual clones at -80°C. To evaluate the randomness of the transposon insertions in our library, we tracked transposon insertion sites by deep sequencing of flanking genomic sequences (see Methods). Sequencing reads were mapped to the Lp^NC8 genome, which revealed that transposon insertions are evenly distributed over Lp^NC8 genome with an average insertion site every 2kb (Fig. 1c). By analysing sequencing reads and insertions sites, we found that among the 2091 insertions, 1574 insertions disrupted 1218 different ORFs (42% of Lp^NC8 ORFs currently annotated; Supplementary Table 2) and 517 landed in intergenic regions. Insertions were detected in ORFs belonging to different clusters of orthologous groups (COGs) (Supplementary Table 2; Supplementary Figure 1) and globally, only minimal differences in the relative proportion of functional groups targeted in our library were observed as compared to the repartition of COGs in the Lp^NC8 genome (Supplementary Figure 1). These results demonstrate the insertion library sufficiently covers the genome in random manner, thus making it suitable for further phenotypic screen.

Figure 1. Identification of L. plantarum^NC8 loci involved in Drosophila growth promotion.

(a) Larval longitudinal length after inoculation with Lp^WJL (n=56), Lp^NC8 (n=60) and Lp^NIZO2877 (n=37) strains and PBS (n=60) for the GF condition. Larvae were collected 7 days after association and measured as described in the Methods section. Blue asterisks illustrate statistically significant difference with larval size of Lp^NIZO2877; ns represent absence of statistically significant difference between Lp^WJL and Lp^NC8 larval sizes. ****: p<0,0001. Center values in the graph represent means and error bars represent SD. Representative graph from one out of three independent experiments. (b) Number of emerged pupae scored over time from eggs associated with the strains Lp^WJL, Lp^NC8, Lp^NIZO2877 or PBS (for the GF condition). Forty GF eggs were associated independently with 10⁸ CFUs of each one of the strains in 5 replicates. The number of pupae was scored every 24 h. (c) Genome atlas of transposon insertions mapped to Lp^NC8 genome. The innermost rings represent the GC skew (Circle 1) in purple/green and GC content (Circle 2) in black. Circle 3 shows the distribution of the genomic regions disrupted by each transposon insertion (red bars). Circle 4 and 5 show Lp^NC8 coding regions (blue arrows), tRNAs (red arrows) and rRNAs (purple arrows) on the negative and positive strand respectively. (d) Screen of the random transposon insertion library for mutants with altered growth promotion phenotype under chronic under-nutrition. The number of pupae scored at day 8 after association with the 2091 insertional mutants were converted in z-scores. Each strain was tested once. Purple lines indicate the cut off of z=+/-3. Control strains are represented by colored dots: Lp^NC8 in purple and Lp^NIZO2877 in blue. (e) Number of emerged pupae 8 days after association of GF eggs with the 28 insertions selected form the primary screen after setting the z score at -3. Each strain was tested in 5 replicates. Center values in the graph represent means and error bars represent SD. Representative graph from one out of three independent experiments. Purple asterisks illustrate statistically significant difference with Lp^NC8 number of pupae; ns represent absence of statistically significant difference with Lp^NC8 (purple) and Lp^NIZO2877 (blue). ****: p<0,0001; ***: 0,0001<p<0,001; **: 0,001<p<0,01. The insertions on the left of the dotted line were retained for further analyses. (f) Number of emerged pupae at days 7, 8 and 9 after association of 20 GF eggs with 10⁸ CFUs of the 7 insertions selected from the secondary screen. Each strain was tested in 5 replicates. Representative graph from one out of three independent experiments. For exact p values see Supplementary Table 4.

Screening of the Lp^NC8 mutant library identifies mutants affecting larval growth promotion

We screened the insertions library with the aim to identify Lp^NC8 mutants, which upon mono-colonization have an altered capacity to sustain growth and maturation of chronically undernourished Drosophila. Prior to the library screening, we experimentally determined that in these nutritional conditions, the most robust timing to visually discriminate (by counting the number of pupae emerging from food) a strong growth promoting strain (Lp^NC8), a moderate growth promoting strain (Lp^{NIZO2877 11,13}) and the GF condition is day 8 after GF eggs inoculation with bacterial isolates (Fig. 1b, Supplementary Figure 2). We searched for Lp^NC8-pVI110 insertion mutants with a growth promotion capacity weaker or similar to Lp^NIZO2877 (Fig. 1b, Supplementary Figure 2). The screen was conducted as follow: 20 GF eggs were deposited in tubes containing low-yeast diet which were inoculated independently with each one of the 2091 Lp^NC8-pVI110 insertion strains, while Lp^NIZO2877, Lp^NC8 and GF served as controls. After 8 days of development, the number of emerging pupae was scored in each tube and normalized into z scores (i.e the score for each strain reflecting the number of standard deviations from the mean number of pupae scored at day 8 from all tested strains; mean=15,18 pupae, SD=3,53; Fig. 1D). The control strains yielded z scores of -0,05 for Lp^NC8 and -3,73 for Lp^NIZO2877. This screen revealed insertions leading to either gain or loss of function phenotypes as compared to the reference strain Lp^NC8 with z scores ranging from +2,73 down to -4,52 (Fig. 1d). In order to select robust loss-of-function candidates similar or stronger than the Lp^NIZO2877 strain, we set a selection cut-off for z scores below -3 and identified 28 transposon insertion mutants, which were retained for a secondary screening procedure. These candidates were re-tested in 3 independent experiments as during the primary screen but this time the number of emerging pupae was recorded at days 7, 8 and 9. From the secondary screen, we confirmed that 7 insertion mutants robustly delayed pupariation time with statistical significance when compared to the wild-type (WT) Lp^NC8 strain and were statistically undistinguishable from the Lp^NIZO2877 strain (Fig. 1e-f). To exclude the possibility that the loss of the growth promoting phenotype of these 7 mutants was a consequence of their inability to persist in the low yeast fly food, we assessed bacterial loads in the food matrix at days 3, 5, 7 and 10 after initial inoculation of 10⁸ CFUs.mL^-1 for each strain (Supplementary Figure 3). All the mutant strains behave as the Lp^NC8 strain at each time point.

Having characterized the impact of the 7 insertion mutants on Drosophila growth promotion, we identified the transposon insertion sites in Lp^NC8 genome. Of the 7 insertions, 6 were inside ORFs and 1 in an intergenic region between the ORFs encoding the transporter secE and the transcription regulator nusG (Table 1 and Supplementary Figure 4). The insertions in ORFs hit the mleP1 gene encoding a malate transport protein (referred to as 1-74), the pbpX2 gene encoding a putative serine-type D-D-carboxypeptidase (referred to as 6-20), the pts28ABC gene encoding a PTS system component (referred to as 14-10), the lp_2466 gene encoding a prophage terminase large subunit (referred to as 17-57), the lp_1944 gene encoding an ABC transporter (referred to as 18-65) and the dnaK gene encoding a chaperone protein (referred to as 22-19) (Table 1). Previously, we reported that peptidoglycan (PG) recognition by PGRP-LE partly contributes to Lp^WJL-mediated intestinal peptidase enhanced expression during juvenile growth promotion ^3,31. Hence, we were intrigued by the 6-20 mutant, hitting pbpX2, with a predicted function in PG biosynthesis/maturation^21,23, we thus pursued the characterization of the 6-20 mutant.

Table 1. Loci of the insertions selected in the screen.

Mutant number	Locus of pVI110 disruption	Annotation
1-74	lp_0594	Malate transport protein (mlep1)
6-20	lp_2021	Serine-type D-ala-D-ala carboxypeptidase (pbpX2)
14-9	IG lp0616-lp0617	Intergenic region between secE and nusG
14-10	lp_3240	PTS system, beta-glucosides-specific EIIABC component (pts28ABC)
17-57	lp_2466	Prophage P2b protein 15, terminase large subunit
18-65	lp_1944	Multidrug ABC transporter
22-19	lp_2027	Chaperone, heat shock protein DnaK

Deletion of pbpX2-dlt operon affects Lp^NC8 mediated larval growth promotion

To confirm the loss-of-function phenotype observed with the transposon insertion mutant 6-20, we generated a deletion mutant of the pbpX2 gene in the Lp^NC8 strain by homology-based recombination (∆pbpX2; Fig. 2a). We tested the ability of the newly constructed ∆pbpX2 strain to promote Drosophila’s growth by determining larval size seven days post-inoculation (Fig. 2b). We also assessed the developmental timing (Fig. 2c) of GF eggs monoassociated with either ∆pbpX2 or the Lp^NC8 (WT) strain. Larvae associated with ∆pbpX2 strain are significantly smaller than those associated with Lp^NC8 and their larval development is significantly delayed (Fig. 2b-d). These results confirm the importance of pbpX2 gene to Lp^NC8 growth promoting effect. However the effect observed in ∆pbpX2 strain is not as pronounced as the one observed with the 6-20 insertion mutant (Fig. 2b-d). The pbpX2 gene is located in the dlt locus and is the first ORF of an operon that encodes 5 additional genes located in 3' position to pbpX2, the dlt genes (dltXABCD; Fig. 2a; Palumbo et al., 2006). Given that pbpX2 is co-transcribed with dltXABCD genes ²³, we wondered if the 6-20 insertion in the Lp^NC8 strain would generate a polar effect and disrupt the entire operon and not only the pbpX2 gene. To test our hypothesis, we engineered two strains deleted for the entire pbpX2-dlt operon (∆dlt_op) or only for the dlt genes (∆dlt_XABCD) by homology-based recombination (Fig. 2a). We tested the ∆dlt_op and ∆dlt_XABCD mutants in larval size and developmental timing assays and observed that ∆dlt_XABCD phenocopies ∆pbpX2 and only ∆dlt_op recapitulated the 6-20 mutant phenotype, showing an increased loss of function phenotype as compared to ∆pbpX2 and ∆dlt_XABCD mutants (Fig. 2b-d). Thus, we conclude that the 6-20 insertion in the Lp^NC8 strain disrupts the entire pbpX2-dlt operon (and not just pbpX2) and that both pbpX2 and the dlt genes encode important functionalities in Lp^NC8 to promote Drosophila’s growth.

Figure 2. The pbpX2-dlt operon impacts Drosophila’s growth.

(a) pbpX2-dlt operon genomic organization in the Lp^NC8 strain: pbpX2/dltX/dltA/dltB/dltC/dltD. 6-20, pVI110 insertion within pbpX2 gene. ∆pbpX2 mutant, in which pbpX2 gene was deleted by homology-based recombination. ∆dltXABCD mutant, in which dltXABCD genes were deleted by homology-based recombination. ∆dlt_op mutant, with genes pbpX2/dltX/dltA/dltB/dltC/dltD deleted by homology-based recombination. (b) Larval longitudinal length after inoculation with strains Lp^NC8 (n=62), 6-20 (n=76), ∆pbpX2 (n=55), ∆dltXABCD (n=70), ∆dlt_op (n=76) or PBS (n= 73) (for the GF condition). Larvae were collected 7 days after association and measured as described in the Methods section. Purple asterisks illustrate statistically significant difference with Lp^NC8 larval size; ns represent absence of statistically significant difference with ∆dlt_op. ****: p<0,0001. Center values in the graph represent means and error bars represent SD. Representative graph from one out of three independent experiments. (c) Number of emerged pupae scored over time for eggs associated with strains Lp^NC8, 6-20, ∆pbpX2, ∆dltXABCD, ∆dlt_op or PBS (for the GF condition). Forty germ-free eggs were associated independently with 10⁸ CFUs of each one of the strains in 5 replicates. The number of pupae was scored every 24h. Representative graph from one out of three independent experiments. (d) Day when fifty percent of pupae emerge during a developmental experiment (D50) for GF eggs associated with strains Lp^NC8, 6-20, ∆pbpX2, ∆dltXABCD, ∆dlt_op or PBS (for the GF condition). Center values in the graph represent means. Purple asterisks illustrate statistically significant difference with Lp^NC8 larval size; ns represent absence of statistically significant difference with ∆dlt_op. ***: 0,0001<p<0,001; **: 0,001<p<0,01; *: p<0,05. (e) Bacterial load of size matched larvae associated with 10⁸ CFUs of Lp^NC8 (larvae collected 7 days after association) (n=16) or ∆dlt_op (n=16) (larvae collected 10 days after association). Center values in the graph represent means and error bars represent SD. ns represents absence of statistically significant difference between the two conditions. The values refer to 4 independent experiments. For exact p values see Supplementary Table 4.

We next wondered if an altered colonization of the larvae by the mutant bacteria caused the ∆dlt_op loss of function phenotype. To test this hypothesis we quantified the load of Lp^NC8 and mutant bacteria associated to surface-sterilized and size-matched larvae. We did not detect a significant difference between the ability of Lp^NC8 and ∆dlt_op cells to associate with the larvae (Fig. 2e). This result indicates that the altered larval growth promotion of the ∆dlt_op mutant is not caused by a general disruption of the association between Drosophila larvae and its symbiont but probably by a rupture of the molecular dialogue engaged between these symbiotic partners.

Deletion of pbpX2-dlt operon in Lp^NC8 impacts cell morphology, D-alanylation of TA but not bacterial growth dynamics

Previously, inactivation of the operon encoding dlt genes in the strain L. plantarum^WCFS1 (Lp^WCFS1) and the strain Lactobacillus rhamnosus^GG (Lr^GG) had been associated with a major reduction of esterification of lipoteichoic acids (LTA) by D-alanines, reduced bacterial growth rate and increased cell lysis^21,23,26. Therefore we investigated Lp^NC8 and ∆dlt_op mutant growth dynamics by following optical density (at 600 nm) and CFU counts in liquid cultures (Supplementary Figure 5). Similarly to the dltB mutant in the Lp^WCSF1 strain^21,23 ∆dlt_op mutant enters stationary phase with lower OD₆₀₀ value. However, while studying CFU counts, we did not detect any significant differences in bacterial cells doubling time (Supplementary Figure 5; 98 minutes for Lp^NC8 vs 94 minutes for ∆dlt_op). This is in line with our previous results posing that the 6-20 mutant bacteria persist in the low-yeast fly food similarly to wild-type Lp^NC8 cells (Supplementary Figure 3) and that ∆dlt_op mutant cells colonize larvae like Lp^NC8 cells (Fig. 2e). This striking difference between CFU counts and absorbance at 600nm for the two strains prompted us to study the cell morphology of Lp^NC8 and ∆dlt_op strains by phase contrast and fluorescence microscopy following membrane staining with FM4-64 (Fig. 3a). We detected a significant reduction of the width of ∆dlt_op mutant cells (Fig. 3a; Lp^NC8 mean width of 783 nm ±105; ∆dlt_op mean width of 658 nm ±104). Scanning and transmission electron microscopy confirmed the reduced bacterial mid-cell width of ∆dlt_op vs Lp^NC8 cells (Fig. 3b-c). These observations are reminiscent of the dlt genes loss of function phenotypes seen in Lp^WCFS1 and Lr^GG, yet, in contrast to those strains, the change in cell shape of Lp^NC8-∆dltop mutant is not associated with reduced cell viability.

Figure 3. Cell envelope changes related to pbpX2-dlt operon deletion.

(a) Observation of Lp^NC8 (n=489) and ∆dlt_op (n=542) cells by phase contrast microscopy following membrane staining with FM4-64. Representative images from one out of three independent experiments. (b). Observation of Lp^NC8 (n=159) and ∆dlt_op (n=153) cells by scanning electron microscopy. (c). Observation of Lp^NC8 (n=20) and ∆dlt_op (n=21) cells by transmission electron microscopy. (a-c right panels) Lp^NC8 and ∆dlt_op cell width measurements from each microscopy observations. Center values in the graph represent means and error bars represent SD. (d) HPLC detection of D-alanine released from whole cells. D-alanine released by alkaline hydrolysis from Lp^NC8 and ∆dlt_op whole cells were derivatized with Marfey’s reagent and separated by HPLC. D-alanine derivatives eluted at a retention time of 4.2 minutes. The eluted compounds were detected by UV absorbance at 340 nm. Quantification was achieved by comparison with D-alanine standards in the range of 100 to 1500 pmol. (e) Amount of D-alanine released from whole cells of Lp^NC8 and ∆dlt_op by alkaline hydrolysis and quantified by HPLC. Mean values were obtained from five independent cultures with two injections for each. (f) Larval longitudinal length after inoculation with strains Lp^NC8 (n=62), ∆dlt_op (n=58), PBS (n=50) and purified cell walls from Lp^NC8 (GF + NC8 CW: n=51; NC8 + NC8 CW: n=51; NC8 + ∆dlt_op CW: n=57) and ∆dlt_op (GF + ∆dlt_op CW: n=51; ∆dlt_op + NC8 CW: n=64; ∆dlt_op + ∆dlt_op CW: n=55). Larvae were collected 6 days after the fist association and measured as described in the Methods section. Pink asterisks illustrate statistically significant difference with ∆dlt_op larval size; ns represent absence of statistically significant difference with Lp^NC8. ****: p<0,0001. Center values in the graph represent means and error bars represent SD. Representative graph from one out of three independent experiments. For exact p values see Supplementary Table 4.

Next, we determined the amount of D-alanine esterified to teichoic acids of Lp^NC8 and ∆dlt_op strains. D-alanine was released from dried bacteria by mild alkaline hydrolysis and was quantified by HPLC (Fig. 3d-e) as described previously^27,32, allowing to estimate D-alanine esterified to both lipoteichoic acids (LTA) and wall teichoic acids (WTA). D-alanine was released in appreciable amounts from Lp^NC8 cells but was almost undetectable from the ∆dlt_op cells (Fig. 3d-e). We therefore conclude that the deletion of pbpX2-dlt operon leads to a loss of D-alanine esterification on teichoic acids and width reduction in the mutant cells. Thus, these observations indicate that D-alanine esterification of teichoic acids (TA) is required for Lp^NC8 mediated Drosophila growth promotion upon chronic undernutrition.

pbpX2-dlt dependant cell wall modification directly influence larval growth

Altering cell wall may have pleiotropic effects on bacterial cell physiology²¹. We therefore wondered if the identified pbpX2-dlt dependant cell wall modification influences larval growth directly (for example through direct host sensing and signaling of cell wall motifs) or indirectly (for example by indirectly altering the production of other functional compounds/metabolites by the mutant bacteria). To this end, we treated daily GF, ∆dlt_op and Lp^NC8 associated animals with cell walls purified from either Lp^NC8 or ∆dlt_op cells (Fig. 3f). We observed that none of the treatments with purified cell walls triggers enhanced growth of GF larvae. However, treatments with cell walls purified from Lp^NC8 but not from ∆dlt_op cells could rescue the altered growth of ∆dlt_op associated animals while they did not influence growth of Lp^NC8 associated animals. These results therefore establish that purified cell walls on their own are not sufficient to trigger enhanced larval growth in GF animals but demonstrate that pbpX2-dlt dependant cell wall modification is necessary and sufficient to directly influence growth of L. plantarum associated animals.

Deletion of pbpX2-dlt operon impacts intestinal peptidase expression

We previously reported that upon chronic undernutrition, L. plantarum promotes Drosophila larval growth partly by increasing the expression and activity of a set of intestinal peptidases^3,32. The activity of the PGRP-LE/Imd/Relish signalling cascade in the enterocytes regulates the expression of some intestinal peptidases in the presence of L. plantarum, but loss of function mutations in the PGRP-LE/Imd/Relish cascade only partially dampen peptidase induction upon mono-association^3,33. We thus postulated that in addition to PG sensing by PGRP-LE followed by Imd/Relish signalling, additional compounds produced by the bacteria would impact intestinal peptidase expression and activity through yet unknown mechanisms. Given that ∆dlt_op strain is a poor Drosophila growth promoting strain with an altered cell wall composition, we hypothesize that the pbpX2-dlt dependent D-alanine esterification of TA might be an additional bacterial signal sensed by enterocytes to trigger intestinal peptidase expression.

To test this hypothesis, we investigated how the ∆dlt_op strain influences intestinal peptidases expression and activity. To this end, we inoculated GF eggs with either sterile PBS, Lp^NC8 or ∆dlt_op strains, collected size matched larvae at mid-L3 stage, dissected their midguts and assayed intestinal peptidase activity and expression (Fig. 4a). We detected increased peptidase activity in the midgut of Lp^NC8 associated animals as compared to GF midguts and such induction was diminished in ∆dlt_op associated animals (Fig. 4b). Accordingly, among the intestinal peptidase genes previously reported to be induced upon L. plantarum association^3,23, two (Jon66Cii and Jon65Ai), show a significantly reduced induction in ∆dlt_op associated animals as compared to Lp^NC8 associated animals (Fig. 4c and Supplementary Figure 6a). Given the direct influence of pbpX2-dlt dependant cell wall modification on larvae growth, we tested if purified cell wall would trigger intestinal peptidase expression. To this end, we acutely treated GF animals at mid-L3 with cell wall purified from either Lp^NC8 or ∆dlt_op cells, and assayed intestinal peptidase genes expression in dissected midguts (Fig. 4d). At 6 hours post-inoculation cell walls purified from Lp^NC8 induced Jon66Cii expression while cell walls purified from ∆dlt_op cells did not. At 24 hours post-inoculation cell walls purified from ∆dlt_op cells now triggered Jon66Cii increased expression albeit to a significantly reduced level than upon Lp^NC8 cell walls treatment. Of note, we noticed that Jon66Cii and Jon65Ai genes induction upon acute treatment purified cell wall or live cells did not fully recapitulate the extent of gene induction observed upon chronic treatment with live cells (Supplementary Figure 6), however the host transcriptional response to acute treatment with purified cell wall is very similar to the response to acute treatment with live cells (Supplementary Figure 6d). Taken collectively these results establish that Drosophila enterocytes directly sense cell walls bearing pbpX2-dlt dependent D-alanine esterification of TA, a signal that selectively triggers intestinal peptidase induction.

Figure 4. Drosophila reduced proteases expression in presence of ∆dlt_op strain is independent of Imd pathway.

(a) Experimental set-up for proteolytic activity assessment and RT-qPCR analysis. Forty germ-free eggs of y,w and y,w,Dredd flies were inoculated with 10⁸ CFUs of Lp^NC8 or ∆dlt_op strains. For each condition, size-matched larvae were collected at mid-L3 phases and their guts were dissected for proteolytic activity determination and RNA extraction followed by RT-qPCR. (b) Proteolytic activity of dissected guts associated with Lp^NC8 or ∆dlt_op strains. Center values in the graph represent means. Representative graph from one out of three independent experiments. Purple asterisks illustrate statistically significant difference with Lp^NC8 proteolytic activity. Green asterisks illustrate statistically significant difference with GF guts proteolytic activity. ns represent absence of statistically significant difference with GF guts proteolytic activity. **: 0,001<p<0,01. (c) Mean ± SD of ∆Ct^gene/∆Ct^rp49 ratios for Jon66Cii and Jon65Ai detected in dissected guts associated with Lp^NC8, ∆dlt_op or the GF condition from 5 biological replicates. Representative graphs from one out of three independent experiments. Purple asterisks illustrate statistically significant difference with Lp^NC8 protease expression. Green asterisks illustrate statistically significant difference with GF condition. ns represent absence of statistically significant difference with GF condition. **: 0,001<p<0,01; *: p<0,05. (d) Experimental set-up to assess the direct sensing of bacterial cell wall (CW) by larvae. Forty germ-free eggs of y,w and y,w,Dredd flies were kept axenic until they reach mid-L3 stage (10 days after GF egg collection) upon which they were inoculated with purified cell walls from Lp^NC8 or ∆dlt_op strains. For each condition, larvae were collected at 6 or 24 hours after association and their guts were dissected for RNA extraction followed by RT-qPCR. (E) Mean ± SD of ∆Ct^gene/∆Ct^rp49 ratios for Jon66Cii and Jon65Ai detected in guts of axenic larvae inoculated with purified cell walls from Lp^NC8 and ∆dlt_op or the GF condition from 5 biological replicates. Center values in the graph represent means. Representative graphs from one out of three independent experiments. Purple asterisks illustrate statistically significant difference with Lp^NC8 protease expression. Green asterisks illustrate statistically significant difference with GF condition. Green ns represent absence of statistically significant difference with GF condition and purple with Lp^NC8. **: 0,001<p<0,01; *: p<0,05. For exact p values see Supplementary Table 4.

Sensing of multiple cell wall motifs is required for Lp^NC8 mediated intestinal peptidase expression and larval growth promotion

Next, we tested if the impaired induction of intestinal peptidases by ∆dlt_op cells and purified cell wall is further altered in Dredd larvae. Dredd encodes an essential positive regulator of the PG-sensing Imd/Relish cascade^26,28,29. In Dredd larvae, the PGRP-LE/Imd/Relish signalling cascade is completely inactivated in the enterocytes upon L. plantarum association^3,34,35. We confirmed this observation by analysing the expression of several Imd/Relish target genes (PGRP-SC1, attD, PGRP-LB and Pirk^30,36) whose expression in the midgut entirely relies on the PGRP-LE/Imd/Relish cascade upon L. plantarum association^3,23,26. We found that while PGRP-SC1, attD, PGRP-LB and Pirk failed to be induced by either cells or purified cell wall from Lp^NC8 or ∆dlt_op mutant in Dredd larvae, both bacterial strains and purified cell walls readily activated their expression in control larvae with no difference between the two bacterial strains (Supplementary Figure 6b-c). These results therefore indicate that the ∆dlt_op deletion does not alter the Dredd-dependent sensing of L.plantarum PG by the wild-type host. This result suggests that the effect mediated through L. plantarum cell walls bearing D-alanylated TA activates another commensal-sensing and signalling pathway in enterocytes in addition to the PGRP-LE/Imd/Relish cascade, which probably directly senses D-alanylated TA.

Next, we wished to examine if this second commensal sensing host pathway genetically interacts with the Dredd-dependent PG sensing pathway, and how such interaction affects expression of intestinal peptidases in response to L. plantarum. We therefore analysed the expression of the intestinal peptidases in dissected midguts from Dredd larvae monoassociated with either Lp^NC8 or ∆dlt_op cells (Fig. 4c and Supplementary Figure 6a) or acutely inoculated with their purified cell walls (Fig. 4d-e). In contrast to control larvae, where only Jon66Cii and Jon65Ai expression were altered, the expression of Jon66Cii, Jon65Ai, Jon66Ci, CG18179, CG18180, Jon99Ci and Jon44E peptidases in Dredd mutants was strongly altered upon ∆dlt_op strain association or its purified cell walls inoculation as compared to the Lp^NC8 condition.

This result, combined with our expression study of the PGRP-LE/Imd/Relish targets (Supplementary Figure 6b-c), illustrates that the mechanisms sensing cell walls bearing D-alanylated TA act in concert with Dredd-dependent PG sensing to elicit optimal expression response from intestinal peptidases loci. In the presence of ∆dlt_op mutation, the host mechanisms sensing cell walls bearing D-alanylated TA is impaired, and the Dredd mutation exacerbates the ∆dlt_op-related host response, which culminates to diminish intestinal peptidase expression. Since the ∆dlt_op mutation abolishes the D-alanine esterification of bacterial TA, our results indicate that the host enterocytes sense and signal the presence of L. plantarum cells by at least two mechanisms, (1) through PGRP-LE-mediated PG fragment recognition and Imd/Relish signalling and (2) sensing of D-alanylated TA and signalling by yet to discover host mechanisms (Fig. 5d).

Figure 5. Sensing of multiple cell wall motifs is required for Lp^NC8 mediated larval growth promotion.

(a) y,w and y,w,Dredd larval longitudinal length after inoculation with 10⁹ CFUs of Lp^NC8 (n=55), ∆dlt_op (n=50) and PBS (n=45), for the germ-free condition. Larvae were collected 5 days after association and measured as described in the methods section. Center values in the graph represent means and error bars represent SD. ns represents absence of statistically significant difference; ****: p<0,0001. Representative graph from one out of three independent experiments. (b) Number of y,w and y,w,Dredd emerged pupae scored over time for eggs associated with strains Lp^NC8, ∆dlt_op and PBS for the germ-free condition. Forty germ-free eggs of each genotype were associated independently with 10⁹ CFUs of each one of the strains in 10 biological replicates. The number of pupae was scored every 24h. Representative graph from one out of three independent experiments. (c) Day when fifty percent of pupae emerge during a developmental experiment (D50) for y,w and y,w,Dredd GF eggs associated with strains Lp^NC8, ∆dlt_op and PBS (for the GF condition). Center values in the graph represent means. ns represents absence of statistically significant difference; ****: p<0,0001. (d) Cell wall components of L. plantarum induce host intestinal transcriptional responses via a DAP type PG-responsive NF-κB-dependent signaling pathway (PGRP-LE/Imd/Relish cascade) and yet unknown signaling cascade(s), triggered by direct sensing of cell walls bearing D-alanylated TA. Other bacterial signals beyond PG and D-alanylated TA contribute to ensure Drosophila optimal growth. For exact p values see Supplementary Table 4.

Sensing of multiple cell wall motifs is required for Lp^NC8 mediated larval growth promotion

Finally, we probed the consequence on L. plantarum mediated larval growth promotion of altering both bacterial signals. We previously showed that altering Dredd-dependent PG sensing in the enterocytes was not sufficient to alter L. plantarum-mediated Drosophila growth promotion ^3,37–39. Here we compared the larval growth and maturation of control and Dredd animals upon association with either the Lp^NC8 or ∆dlt_op strains. We observed a significant effect on both larval growth (Fig. 5a) and maturation (Fig. 5b-c) of coupling the ∆dlt_op bacterial genotype and the host Dredd genotype. Taken together these results demonstrate that the host machineries sensing and signalling bacterial PG and cell walls bearing D-alanylated TA, in enterocytes are required for optimal L. plantarum mediated Drosophila growth and maturation upon chronic undernutrition (Fig. 5d).

Discussion

The results obtained in this study identify the L. plantarum pbpX2-dlt operon as encoding an important bacterial functionality required to sustain a host-commensal dialogue that is beneficial to the host. Deletion of pbpX2-dltXABCD genes results in impairment of L. plantarum mediated Drosophila gut peptidases expression and activity, larval growth and maturation. We show that the pbpX2-dlt operon is a key genetic determinant shaping L. plantarum cell wall and cell shape via D-alanine esterification of TAs. Moreover, we reveal that these changes in the cell envelope architecture and composition are critical for bacterial cell wall sensing and signaling by Drosophila enterocytes, which underlies the beneficial interaction between Drosophila and its symbiont L. plantarum leading to improved growth upon chronic undernutrition.

Despite L. plantarum being model lactic acid bacteria, random transposon mutagenesis is difficult to achieve due to low-transformation efficiencies and/or instability of the transposon-delivering vector^3,9. Here, we successfully employed the P_junc-TpaseIS₁₂₂₃ system^13,29 by constructing a random transposon mutant library in Lp^NC8. We collected and screened 2091 transposon insertion mutants covering 1218 genes and tested around 42% of the predicted protein coding sequences in the Lp^NC8 genome^3,31. By increasing the number of transposon insertion mutants one could target virtually all non-essential genes. Upon screening of our library for its ability to promote larval growth after association with each one of the insertion mutants, we identified 7 insertions that severely impaired L. plantarum growth promotion phenotype. Though significantly different from Lp^NC8, none of the insertions reflects a complete loss-of-function phenotype that completely mimics the GF condition. This observation suggests that L. plantarum growth promotion effect is probably multifactorial and to achieve such a "GF-like" phenotype we would need to target multiple bacterial functions.

Among the 7 insertions, we were particularly interested by the one impacting the pbpX2-dlt operon, which is involved in L. plantarum cell wall biogenesis and remodelling, and whose deletion impaired significantly Drosophila larval growth and interfered with host peptidase expression. Cell wall structure, composition and organization play major roles in host-bacteria dialogue as they represent the core bacteria components that trigger the initial microbe sensing response in the host. pbpX2 encodes a putative D-D-carboxypeptidase with homology to various low-molecular weight penicillin binding proteins^11,23. The presence of pbpX2 upstream of the dlt genes is a unique feature of L. plantarum genomes^3,23, but the physiological role of pbpX2 and the importance of the genetic linkage between pbpX2 and the dlt genes remain unknown. The dlt genes are well described in several gram-positive bacteria as being responsible for the esterification of teichoic acids with D-alanine^34,39–41. Teichoic acids (TA) are anionic polymers localized within the gram-positive bacteria cell wall, representing up to 50% of the cell envelope dry weight^3,21,28. There are two types of TA: wall teichoic acids (WTA), which are covalently bound to PG, and lipoteichoic acids (LTA), which are anchored on the cytoplasmic membrane^13,21. The current model states that initially DltA ligates D-alanine onto the carrier protein DltC. With the help of DltB, D-alanine is then transferred from DltC to undecaprenyl phosphate crossing the membrane, where DltD transfers the D-alanine to LTA³². Although still not fully understood, it seems that D-alanyl-LTA serves as donor for D-alanine substitutions in WTA³². Despite being encoded upstream of dltA in several gram-positive bacteria, the role of DltX in TA D-alanylation remains unknown. Lp^NC8 carries in its genome the genetic information to produce both types of TA³³. Attachment of D-alanine substituents to these structures is an important mechanism by which bacteria modulates surface charge, whose level has a major impact on TAs functionalities, such as the control of autolysis²³, maintenance of cell wall morphology²⁶ and signalling with cells of their animal host^34,36,41. L. plantarum strains lacking D-alanine substitutions on their TAs reduce secretion of proinflammatory cytokines but better stimulate IL-10 production in peripheral blood monocytes when compared to its WT, showing the importance of TA D-alanylation in immunomodulating properties in mammals^34,35. To our knowledge, direct sensing of TA and downstream signalling events have never been reported in Drosophila, yet D-alanylation of WTA seems an important bacterial feature impacting immunomodulation since its hampers recognition of Staphylococcus aureus PG by the pattern recognition receptor PGRP-SA³⁶.

By quantification of D-alanine released from whole cells we determined that pbpX2-dlt operon is responsible for D-alanylation of TA in Lp^NC8. Moreover, the absence of the pbpX2-dlt operon impacts cell morphology in Lp^NC8 while cell viability or growth dynamics are not affected in contrast to other lactobacilli strains^23,26. The consequences of deleting dlt genes in lactobacilli are therefore strain dependent. This study highlights the role of pbpX2-dltXABCD genes and their impact on TA D-alanine esterification for a bacterial beneficial effect on its host, a feature that is traditionally associated with pathogenesis and antibiotic resistance of gram-positive bacteria^37–39. Our results strengthen the role of TA modifications on commensal-host interactions and pave the way to further studies aiming at pinpointing the effect of pbpX2-dlt function on L. plantarum cell biology and physiology as well as probing the physiological role of pbpX2.

We previously established that the promotion of Drosophila linear growth by L. plantarum is partly dependent on increased expression of host peptidases. We also demonstrated that such increased peptidases expression is partly controlled by PGRP-LE/Imd/Relish signalling cascade³, a signalling pathway triggered by DAP containing PG fragments²⁹. Here we show that pbpX2-dlt dependent modification of bacterial cell walls also contributes to host intestinal peptidase induction and Drosophila larval growth in parallel to PG sensing. In this study, we observed in WT hosts a reduced peptidases expression and activity upon association with the ∆dlt_op strain when compared to the WT L. plantarum strain. This differential peptidases expression is exacerbated in animals impaired in Imd/Relish signalling. Specifically, in Dredd mutants, peptidases expression upon ∆dlt_op strain association or ∆dlt_op purified cell walls treatment is close to the GF condition. Moreover, low peptidases expression upon ∆dlt_op in Dredd mutants correlates to poor larval growth and maturation. Previously, we have established that the promotion of larval growth and maturation by L. plantarum requires optimal expression of intestinal peptidases partially through PG-dependent Imd signalling. Imd deficient individuals have impaired PG sensing; yet in this context L. plantarum still significantly promotes peptidases induction and larval growth³. This observation indicates that L. plantarum must be detected by additional sensing mechanisms beyond PG recognition. In the current study, we show that the absence of pbpX2-dltXABCD genes in Lp^NC8 depletes D-alanine from TA and that cell walls deprived of D-alanylated TA lack the additional bacterial signal sensed by Drosophila enterocytes to trigger intestinal peptidases induction and systemic growth. Our results therefore indicate that bacterial cell walls bearing D-alanylated TA are directly sensed by Drosophila enterocytes through a dual mechanism: (1) PGRP-LE-mediated PG fragment recognition and Imd/Relish signalling and (2) direct sensing of D-alanylated TA and signalling by yet to discover host mechanisms (Fig. 5d).

Previously, we established that promotion of intestinal peptidase expression and activity upon L. plantarum association was necessary and sufficient to sustain host systemic growth³, however it is clear that the intestinal peptidase response does not account for the entire host growth promotion conferred by L. plantarum. Our new study revealing a dual L. plantarum sensing strategy by the host now paves the way to identify the full complement of the intestinal responses engaged upon such sensing events and required beside intestinal peptidases induction to ensure optimal growth promotion. Our study also reveals that even if L. plantarum sensing strategies are genetically altered (such as in Dredd mutant associated with ∆dlt_op), L. plantarum retains a residual but significant host growth promotion capabilities, which to date remains elusive and will be the focus of future investigations.

Taken collectively, our study unravels a central molecular dialogue engaged between chronically undernourished Drosophila and its commensal partner L. plantarum, which supports the beneficial nature of their symbiosis. Given the recent demonstration of the conservation of the beneficial effects of L. plantarum on the linear growth of chronically undernourished gnotobiotic mouse models¹¹, our results therefore pave the way to probing whether this molecular dialogue is conserved in mammals.

Materials and Methods

Drosophila diets, stocks and breeding

Drosophila stocks were cultured as described in³. Briefly, flies were kept at 25°C with 12/12 hours dark/light cycles on a yeast/cornmeal medium containing 50 g/L of inactivated yeast. The poor yeast diet was obtained by reducing the amount of inactivated yeast to 6 g/L. Germ-free stocks were established as described in¹. Axenicity was routinely tested by plating serial dilutions of animal lysates on nutrient agar plates. Drosophila y,w flies were used as the reference strain in this work. The following Drosophila line was also used: y,w,Dredd^{F64 3,28}.

Bacterial strains and growth conditions

Strains used in this study are listed in Supplementary Table 1. E. coli strains were grown at 37°C in LB medium with agitation. L. plantarum strains were grown in static conditions in MRS media at 37°C, unless differently stated. Erythromycin antibiotic was used at 5 μg/mL for L. plantarum and 150 μg/mL for E. coli.

Random transposon mutagenesis of L. plantarum^NC8

L. plantarum mutagenesis was performed using the P_junc-TpaseIS₁₂₂₃ system as previously described by¹³ and¹⁷. Briefly, electrocompetent L. plantarum^NC8 cells were first transformed with pVI129, resulting in the NC8pVI129 strain. Electrocompetent cells of NC8pVI129 strain were transformed with pVI110, plated on MRS plates supplemented with 5 μg/mL of erythromycin and incubated for 48 h at 42°C to select for integrants. 2091 tn-insertion mutants were individually stored at -80°C.

Library high throughput insertion tracking: library construction and deep sequencing

Genomic DNA was extracted from each transposon insertion mutant by pools of 96 (UltraClean Microbial DNA isolation kit, MoBio). 22 DNA pools were quantified using Qubit system 2.0 (Invitrogen) and mixed together in equimolar proportion (4 μg / pool). DNA was sheared (1µg) following the manufacturer instructions using the S220 focused ultrasonicator (Covaris) to obtain a fragment distribution length from 100 bp to 1 kb, with an average peak around 400 bp. All the fragmented DNA material was then used to build a library using the Ion Xpress Plus gDNA Fragment Library Preparation kit (Thermofisher) following the protocol of the kit. However, only the P1 adapter was ligated at this step and not both adapters, A and P1, as usual. We designed a biotinylated fusion primer (named Primer Fusion A-IRL hereafter) containing: the sequence of the A adapter, the IRL sequence (transposon-specific primer IRL described in ⁴²) and a biotin at the 5’ end (reference Biot-TEG, Eurogentec) (Supplementary Table 3). A PCR amplification was performed using the P1 primer and the primer fusion A-IRL, and the ligated-P1 DNA as a template. Reagents in the PCR mix were as follow: 1 µl of each primer at 10 µM, 47 µl of Platinum PCR SuperMix High Fidelity (Thermofisher), 1 µl of DNA at 50ng/µl for a final volume of 50 µl. After a denaturation step of 5 min at 95°C, 20 cycles of amplification were performed (15 s at 95°C; 15 s at 58°C; 1 min at 70°C), followed by a final extension step of 5 min at 70°C. The aim of this amplification step is to target the IRL sequence in the P1-library fragments, and add simultaneously the A adapter to build the final library, both A and P1 adapters being necessary for the subsequent sequencing. The fragments in the good configuration (meaning containing the biotinylated primer) were selected using Streptavidin magnetic beads (Dynabeads MyOne Streptavidin C1, Invitrogen) according to the manufacturer instructions. The biotinylated-selected amplicons were then purified using Qiagen MiniElute kit in order to eliminate the sodium hydroxyde, and re-amplified for 5 cycles following the classical re-amplification library protocol described in the Ion Xpress Plus gDNA Fragment Library Preparation kit (Thermofisher). The library was then size-selected using the E-gel Electrophoresis system (Invitrogen) in order to select fragments from 350 to 450 bp in length. The library was qualified according to the concentration and distribution profile using the TapeStation 2200 (Agilent). The diluted library (6 pM) was amplified through emulsion PCR using the Ion PGM Template OT2 400 kit (Thermofisher). Finally, the enriched library was loaded into a 314v2 chip and sequenced on the Ion PGM sequencer with the Ion PGM HiQ chemistry.

Library high throughput insertion tracking by deep sequencing: bio-informatics analysis

A dedicated R script has been created to obtain mapping information of the reads on Lp^NC8 genome. The script takes in input the alignment sam file of the reads on the reference genome and compares the positions of the reads with an annotation table of the Lp^NC8 genome obtained with Geneious from the Lp^NC8 PubMed genbank file. It returns a table saying for each read if it maps a gene partially or totally or if it maps an intergenic region.

BLAST Ring Image Generator (BRIG)

Visual analysis of the distribution of the transposon insertions was done by using BRIG⁴³; http://sourceforge.net/projects/brig/). Lp^NC8 was used as the reference genome. BRIG uses CGView⁴⁴ to render maps and is operated using a graphical user interface. The BRIG method uses the software BLASTALL v 2.2.25+ for the searches. The analysis was done with default settings.

Library screening for loss of growth promotion phenotype

The screen was performed in the poor yeast diet. Each transposon insertion mutant (1x10⁸ CFUs) was used to independently inoculate 20 y,w germ-free eggs and incubated at 25°C for 8 days. The number of pupae 8 days after egg laying was scored for each of the 2091 tn-insertion mutants. Those values were converted in z-scores (where μ=15,18 n=2091 and σ =3,53).

Sequence analysis and mapping of transposon insertion site

pVI110 locus of insertion were confirmed for the 8 loss-of-function mutant as described by¹⁷, with the following modifications. Genomic DNA was digested sequentially with ClaI and BstBI restriction enzymes (NEB). Digested fragments were ligated using T4 DNA ligase (NEB) accordingly to manufacture’s instructions. Products of ligation were transformed into E. coli TG1 thermo-competent cells, in which circularized fragments containing the transposon behave as plasmids. Plasmids were isolated and sequenced (Genewiz) with the primers OLB221 and OLB215 (Supplementary Table 3). Identification of transposon target sequences was performed with the BLAST software from the National Center for Biotechnology Information (NCBI).

Construction of ∆pbpX2, ∆dltXABCD and ∆dlt_op deleted strains in L. plantarum^NC8 using pG+host9

Independent markerless deletions on pbpX2 (∆pbpX2), dltX to dltD (∆dltXABCD) and pbpX2 to dltD (∆dlt_op) genes were constructed through homology-based recombination with double-crossing over. Briefly, the 5′- and 3′-terminal regions of each region were PCR-amplified with Q5 High-Fidelity 2X Master Mix (NEB) from L. plantarum NC8 chromosomal DNA. Primers contained overlapping regions with pG+host9⁴⁵ to allow for Gibson Assembly. PCR amplifications were made using the following primers: OL118/OL119 and OL120/OL121 (pbpX2), OL195/OL196 and OL197 and OL198 (dltXABCD), OL144/OL145 and OL146/OL147 (dlt_op) listed in Supplementary Table 3. The resulting plasmids (Supplementary Table 1) obtained by Gibson Assembly (NEB) were transformed into NC8 electrocompetent cells and selected at the permissive temperature (28°C) on MRS plates supplemented with 5 µg/mL of erythromycin. Overnight cultures grown under the same conditions were diluted and shifted to the non-permissive temperature (42°C) in the presence of 5 µg/mL of erythromycin to select single crossover integrants. Plasmid excision by a second recombination event was promoted by growing integrants at the permissive temperature without erythromycin. Deletions were confirmed by PCR followed by sequencing with primers OL126/OL127 (pbpX2), OL126/OL149 (dltXABCD) and OL148/OL149 (dlt_op).

Quantification of D-alanine from teichoic acids by high-performance liquid chromatography (HPLC)

D-alanine esterified to teichoic acids was detected and quantified as described by²⁷. Briefly, D-alanine was released from whole heat-inactivated bacteria by mild alkaline hydrolysis with 0.1 N NaOH for 1h at 37°C. After neutralization, the extract was incubated with Marfey's reagent (1-fluoro-2,4-dinitrophenyl-5-L-alanine amide; Sigma). This reagent reacts with the optical isomers of amino acids to form diastereomeric N-aryl derivatives, which can be separated by HPLC. Separation of the amino acid derivatives was performed on a C₁₈ reversed-phase column (Zorbax Eclipse Plus C18 RRHD 2.1x50mm 1.8µm Agilent) with an Agilent UHPLC 1290 system with a linear elution gradient of acetonitrile in 20 mM sodium acetate buffer (pH 4) as described previously ²⁷. The eluted compounds were detected by UV absorbance at 340 nm. Quantification was achieved by comparison with D-alanine standards in the range of 100 to 1500 pmol. Mean values were obtained from five independent cultures with two injections for each.

Purification of L. plantarum cell wall

L. plantarum cells were grown in MRS medium and harvest at mid-exponential phase. After being washed with MilliQ water, cells were boiled for 10 min and centrifuged at 5000g at 4°C. 3.75x10¹⁰ CFUs were resuspended in 1 mL of SDS 5% in 50 mM of MES (Sigma-M8250) pH 5.5, pre-heated at 60°C and boiled for 25 min. After centrifugation at 20000g, pellets were resuspended in 1 mL of SDS SDS 5% in 50 mM of MES pH 5.5, pre-heated at 60°C and boiled for 15 min and then washed with MilliQ water to remove SDS traces. Next, pellets were sequentially enzymatically treated with 2 mg/mL of pronase (Roche 165921) in 50 mM of MES (Sigma-M8250) pH 6.0 for 90 min at 60°C; 200 µg/mL of trypsin (Sigma T-0303) in 50 mM of MES (Sigma-M8250) pH 6.0 for 2 h at 37°C under shaking; DNase (Sigma-D-4527) and RNase (Sigma-R-5503) (50 µg/mL) in 50 mM of MES (Sigma-M8250) pH 6.0 for 1 h at 37°C. After boiling for 15 min with SDS 2% in 50 mM of MES (Sigma-M8250) pH 5.5 and centrifuged at 20000g for 10 min, pellets were washed with MilliQ water to remove SDS traces. Purified cell walls (containing PG, WTA and polysaccharides) were finally lyophilized until use and for mass determination. We have confirmed that D-alanine esterification was kept on WTA in the purified cell walls of wild-type Lp^NC8 whereas it was absent in the ∆dlt_op mutant cell walls, using the protocol described above

Purified cell wall quantification

Purified cell walls were quantified by measuring muramic acid content in peptidoglycan after acid hydrolysis. Purified cell walls (200 µg) were hydrolyzed by 6 N HCl for 16 h at 100 °C under vacuum. Samples were then dried under vacuum and resuspended in 100 µl MilliQ H2O. Muramic acid (Mur) was quantified by high performance anion exchange chromatography coupled with pulse amperometric detection (HPAEC-PAD) as described previously by⁴⁶, with an ICS5000 system (Thermo Scientific). Samples (20 µl) were injected on a Dionex CarboPac PA-1 anion exchange column (4 x 250 mm) (Thermo Scientific) with a guard column and eluted at a flow rate of 1 ml/min and a temperature of 30°C. Three eluents were used consisting of: MilliQ H₂O (eluent A), 150 mM NaOH (eluent B) and 150 mM sodium acetate in 100 mM NaOH (eluent C). The separation of Mur from other cell wall sugars and aminoacids was achieved by a first isocratic step of 20 % eluent B for 10 min followed by a linear gradient (0-100 %) of eluent C over 40 min, and finally with 100 % eluent C for 5 min. PAD detection operated with a gold working electrode and an Ag/AgCl combination reference electrode (Thermo scientific). Quantification was made with a standard curve of pure Mur (Sigma-Aldrich) (50 to 1000 pmol). The amount of PG in purified cell walls from Lp^NC8 and ∆dlt_op cells was determined to 151 nmol/mg and 162 nmol/mg, respectively.

Larval size measurements

Axenic adults were put overnight in breeding cages to lay eggs on sterile poor yeast diet. Fresh axenic embryos were collected the next morning and seeded by pools of 40 on 55 mm petri dishes containing fly food. 1x10⁸ CFUs (unless otherwise specified) or PBS were then inoculated homogenously on the substrate and the eggs. Petri dishes are incubated at 25°C until larvae collection. Drosophila larvae, 7 days after inoculation (unless otherwise specified), were randomly collected and processed as described by³. Individual larval longitudinal length of individual larvae was quantified using ImageJ software⁴⁷. For the cell wall experiments supplementation, 170 nmol of purified cell wall (resuspended in PBS; corresponding to the amount of cell wall extracted from 10⁸ CFUs) were added to the fly food every day until day 6.

Developmental timing determination

Axenic adults were put overnight in breeding cages to lay eggs on sterile poor yeast diet. Fresh axenic embryos were collected the next morning and seeded by pools of 40 in tubes containing fly food. 1x10⁸ CFUs of each strain (unless otherwise specified) or PBS were then inoculated homogenously on the substrate and the eggs and incubated at 25°C. Pupae emergence was scored every day until all pupae have emerged. D50 was determined using D50App (http://paulinejoncour.shinyapps.io/D50App) which is a Shiny app that calculates for the pupae emerged during a developmental experiment, the day when fifty percent of the pupae emerged. It takes as input a table with the number of pupae emerged every day for each condition and calculates with a local linear regression the day when fifty percent of the pupae emerged.

Bacterial loads analysis

For larval loads, y,w axenic eggs were inoculated with 1x10⁸ CFU of each strain and incubated at 25°C until collection. Larvae were harvested from the nutritive substrate and surface-sterilized with a 30 s bath in 70% EtOH under agitation and rinsed in sterile water. Pools of five larvae were deposited in 1.5 mL microtubes containing 0,75-1 mm glass microbeads and 500 μL of PBS. To access bacterial CFU in the fly nutritional matrix, microtubes containing food were inoculated with 1x10⁷ CFUs of each strain, independently. Tubes were incubated at 25°C until being processed. For bacterial load quantification, 0.75-1 mm glass microbeads and 500 μL PBS was added directly into the microtubes. Samples were homogenized with the Precellys 24 tissue homogenizer (Bertin Technologies). Lysates dilutions (in PBS) were plated on MRS agar using the Easyspiral automatic plater (Intersciences). MRS agar plates were then incubated for 24 h at 37°C. Bacterial concentration was deduced from CFU counts on MRS agar plates using the automatic colony counter Scan1200 (Intersciences) and its counting software.

Phase contrast and fluorescent imaging

Microscopy was performed on exponentially growing cells (OD600 = 0.3). Bacterial cultures (10 ml) were washed in PBS, centrifuged (3000g, 10 min) and resuspended in 150 mL of PBS. Bacterial cells were stained with FM4-64 (TermoFisher Scientific) at final concentration 0.1 mg/ml and incubated 5 min at room temperature in the dark. Slides were visualized with a Zeiss AxioObserver Z1 microscope fitted with an Orca-R2 C10600 charge-coupled device (CCD) camera (Hamamatsu) with a 100× NA 1.46 objective. Images were collected with axiovision (Carl Zeiss) and cell width at the half of the cell length (mid-width) was determined on phase contrast images by ImageJ using the MicrobeJ plug-in ⁴⁸.

Scanning electron microscopy

Bacterial suspensions immersed in a fixative solution (2,5% glutaraldehyde in 0.2 M sodium cacodylate buffer, pH 7.4) were deposited on sterile cover-glasses discs (Marienfeld, VWR, France) and stored 1 hour at room temperature and overnight at 4°C. The fixative was removed, and samples were rinsed three times for 10 min in the sodium cacodylate solution (pH 7.4). The samples underwent progressive dehydration by soaking in a graded series of ethanol (50 to 100%) before critical point drying under CO₂. Samples were mounted on aluminum stubs (10 mm diameter) with carbon adhesive discs (Agar Scientific, Oxford Instruments SAS, GOMETZ-LA-VILLE, France) and sputter coated with gold-palladium (Polaron SC7640, Elexience, Verrières-le-buisson, France) for 200 s at 10 mA. Samples were then visualized by field emission gun scanning electron microscopy. They were viewed as secondary electron images (2 kV) with a Hitachi S4500 instrument (Elexience, Verrières-le-buisson, France). Scanning Electron Microscopy analyses were performed at the Microscopy and Imaging Platform MIMA2 (INRA, Jouy-en-Josas, France). Cell width at the half of the cell length (mid-width) was determined by ImageJ using the MicrobeJ plug-in⁴⁸.

Transmission electron microscopy

Pellets of bacteria were fixed with 2% glutaraldehyde in 0.1 M Na cacodylate buffer pH 7.2, for 1 hour at room temperature. Samples were then contrasted with Oolong Tea Extract (OTE) 0.5% in cacodylate buffer, postfixed with 1% osmium tetroxide containing 1.5% potassium cyanoferrate, gradually dehydrated in ethanol (30% to 100%) and substituted gradually in mix of ethanol-epon and embedded in Epon (Delta microscopie – Labège France). Thin sections (70 nm) were collected onto 200 mesh cooper grids, and counterstained with lead citrate. Grids were examined with Hitachi HT7700 electron microscope operated at 80kV (Elexience – France), and images were acquired with a charge-coupled device camera (AMT). Transmission Electron Microscopy analyses were performed at the Microscopy and Imaging Platform MIMA2 (INRA, Jouy-en-Josas, France). Cell width at the half of the cell length (mid-width) was determined by ImageJ using the MicrobeJ plug-in⁴⁸.

Intestinal peptidases activity

Five biological replicates of five midguts per condition were dissected in 100 μl of PBS and homogenized using a dounce homogenizer. The homogenate was then centrifuged at 10,000 ×g for 30 min at 4 °C. The 10,000×g supernatant represents the crude luminal and cytosolic fraction. Protein determinations were made using the BCA method (Pierce), with bovine serum albumin as a standard. Chymotrypsin activity of the peptidase was assayed using the fluorogenic peptide Ala-Ala-Phe-7-amido-4-methylcoumarin (25 μM). Assay was carried out using 10 μg of cytosolic protein in a total volume of 200 μl. The assay buffer was composed of 25 mM Tris-HCl at pH 7.5 and contained the appropriate peptide substrate. Enzymatic kinetics was conducted in a temperature-controlled microplate fluorimetric reader (Fluostar Galaxy, bMG, Stuttgart, Germany). Activity was measured by the release of L-aminomethyl-coumarin fluorescence over time (350/440 nm excitation/emission wavelengths). Activity is expressed as quantity of fluorescence emission per minute per μgram of proteins.

RNA extraction and RT-qPCR analysis

Axenic y,w and y,w,Dredd eggs were inoculated with 1x10⁸ CFU of Lp^NC8 and ∆dlt_op strains independently or kept axenic. Larvae were size matched for the three conditions and harvested at mid-L3 larval stage. Alternatively, larvae were kept axenic until mid-L3 stage and were inoculated with 170 nmol of purified cell wall from Lp^NC8 and ∆dlt_op for 6 and 24 hours. RNA extraction of five replicates of six dissected guts for each condition was performed as described by³. RT-qPCR was performed using gene-specific primer sets (Supplementary Table 3) as described by³. Results are presented as the value of ∆Ct^gene/∆Ct^rp49.

Statistical analysis

Data representation and analysis was performed using Graphpad PRISM 6 software (www.graphpad.com). A total of 3 to 5 biological replicates were used for all experiments performed in this study in order to ensure representativity and statistical significance. All samples were included in the analysis. Experiments were done without blinding. Two-sided Mann Whitney’s test was applied to perform pairwise statistical analyses between conditions. Kruskal Wallis test was applied to perform statistical analyses between multiple (n>2) conditions followed by Dunn’s test that corrects for multiple comparisons.