Quorum‐sensing regulator RhlR but not its autoinducer RhlI enables Pseudomonas to evade opsonization

Abstract

When Drosophila melanogaster feeds on Pseudomonas aeruginosa, some bacteria cross the intestinal barrier and eventually proliferate in the hemocoel. This process is limited by hemocytes through phagocytosis. P. aeruginosa requires the quorum‐sensing regulator RhlR to elude the cellular immune response of the fly. RhlI synthesizes the autoinducer signal that activates RhlR. Here, we show that rhlI mutants are unexpectedly more virulent than rhlR mutants, both in fly and in nematode intestinal infection models, suggesting that RhlR has RhlI‐independent functions. We also report that RhlR protects P. aeruginosa from opsonization mediated by the Drosophila thioester‐containing protein 4 (Tep4). RhlR mutant bacteria show higher levels of Tep4‐mediated opsonization, as compared to rhlI mutants, which prevents lethal bacteremia in the Drosophila hemocoel. In contrast, in a septic model of infection, in which bacteria are introduced directly into the hemocoel, Tep4 mutant flies are more resistant to wild‐type P. aeruginosa, but not to the rhlR mutant. Thus, depending on the infection route, the Tep4 opsonin can either be protective or detrimental to host defense.

Introduction

Opportunistic pathogens such as Pseudomonas aeruginosa or Serratia marcescens are able to infect a large range of hosts, from plants to humans 1, 2. They sense their environment and adapt their gene expression programs accordingly. Quorum‐sensing (QS) systems allow pathogenic microorganisms to implement collective processes only when the pathogen number is locally high enough to ensure the efficiency of these processes, for instance, the expression of virulence factors or biofilm formation 3, 4, 5. P. aeruginosa illustrates the complexity of QS signaling as its genome codes for at least three distinct intertwined QS systems that work in a hierarchical order (LasI/LasR > RhlI/RhlR > Quinolone Signal) 3, 6, 7, 8, 9. The first two QS systems belong to the LuxI/LuxR category in which the LuxI enzyme synthesizes a homoserine lactone (HSL), respectively, 3‐oxo‐C12‐HSL or C4‐HSL, that activates its cognate receptor when a threshold concentration is reached 10, 11, 12, 13.

Genetic model organisms such as the nematode Caenorhabditis elegans or the fruit fly Drosophila melanogaster have yielded unexpected insights into many key biological principles, including host–pathogen relationships 14, 15, 16, 17, 18, 19, 20, 21, 22. Bacteria ingested either by worms or by flies are exposed in the gut lumen to antimicrobial peptides (AMPs) and reactive oxygen species produced by intestinal epithelial cells 23, 24. Whereas P. aeruginosa PA14 or S. marcescens remain mostly confined to the nematode gut lumen until late in the infection, a few bacteria cross the Drosophila intestinal barrier within 2 h. The bacteria do not initially proliferate and remain in low numbers in the insect hemocoel where they are controlled by phagocytosis 25, 26, 27. Nevertheless, P. aeruginosa starts proliferating after a few days and systemic bacteremia ensues that ultimately kills the fly despite the induction of the systemic immune response that is mediated by the Toll and immune deficiency (IMD) pathways 28, 29.

The exact conditions of an intestinal infection modulate the outcome of infection and patterns of virulence. Thus, depending on the osmotic pressure in agar plates, nematodes can either succumb to phenazine‐mediated “fast killing” or to a multifactorial “slow killing” 21. Likewise, flies succumb to or survive ingested P. aeruginosa depending on the nature of the feeding solution 25. The infection route also determines the outcome of an infection as the bacteria experience distinct environments within the host. Indeed, whereas intestinal infections with either S. marcescens or P. aeruginosa kill flies slowly, the introduction of as few as 10 bacterial cells through a septic injury directly in the hemocoel kills flies within 24 and 36 h, respectively 25, 26, 27. Bacteria injected directly into the hemocoel immediately proliferate exponentially despite the release in the hemolymph of AMPs synthesized in the fat body 28, 29.

Flies immunodeficient either for the cellular immune response or for the Toll or IMD pathways succumb sooner to an intestinal infection than wild type 25. Previously, we found that among the P. aeruginosa PA14 QS systems, only RhlR is required for virulence of ingested bacteria that have escaped into the hemocoel, as it allows them to elude the cellular immune response by an unknown mechanism 25. This infection model thus underscores the importance of the cellular arm of host defense against Gram‐negative bacteria, which is overshadowed by the IMD systemic immune response in septic injury models 30, 31, 32. One of the major functions of cellular immunity in adult flies is the phagocytosis of microbes present within the hemocoel. In vertebrates, phagocytosis is enhanced by the prior complement‐mediated opsonization of microbes. Analysis of the Drosophila genome revealed the existence of complement family/α2‐macroglobulin homologs known as thioester‐containing proteins (TEPs). Most TEPs encode a GCGEQ motif in which an unstable bond between cysteine and glutamic acid has the potential to bind to hydroxyl or amine groups present on biological surfaces 33, 34. TEPs are found in many protostome groups and have been most intensively studied in mosquitoes and Drosophila 33, 35, 36, 37, 38, 39, 40, 41, 42. In the mosquito Anopheles gambiae, Tep1 plays a key role against infection by Plasmodium malarial parasites 35, 43.

The Drosophila genome encodes five functional Tep loci and a pseudogene (Tep5) 33. The expression patterns of Tep1‐4 have been documented by in situ hybridization 44. The exact function of each Tep in host defense remains poorly understood despite several studies and their potential role as opsonins has not been definitely determined 33, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54. Although an initial study failed to reveal a role of Tep1, Tep2, Tep3, or Tep4 in several infection models, even when Tep2‐Tep3‐Tep4 triple mutants were assayed 44, a quadruple Tep1‐Tep2‐Tep3‐Tep4 mutant was reported to be sensitive to some fungal and Gram‐positive bacterial infections, likely because Toll pathway activation is impaired in this quadruple mutant 55. In addition to this apparent redundancy in function, different Tep genes may also have pathogen‐specific functions: Whereas Tep2 mutant or Tep4 mutant was reported to be more sensitive to infections with Porphyromonas gingivalis 56, they unexpectedly were more resistant to challenge with Photorhabdus species 49, 57.

Besides the systemic and local immune responses that lead to the expression of AMPs, another humoral immune response in flies depends on the phenoloxidase activation cascade that leads to melanization at the wound site as well as to the production of reactive oxygen species 58, 59. This phenoloxidase activation cascade plays an important role in combating some fungal and Gram‐positive bacterial pathogens 60, 61. The proteolytic activation cascade that ultimately leads to the cleavage of the prophenoloxidase (PPO) into active PO is initiated by the binding of pattern recognition receptors to microbes 62.

Here, we report the unexpected observation that the P. aeruginosa rhlI and rhlR null mutant phenotypes are distinct in the Drosophila intestinal infection model and also upon ingestion of P. aeruginosa by C. elegans. We document the function of RhlR in escaping Tep4‐mediated opsonization and propose a model that gives an integrated view of Tep4 function in host defense against bacterial infections.

Results

ΔrhlI is more virulent than ΔrhlR in the Drosophila intestinal infection model

Ingested rhlI was less virulent than wild‐type P. aeruginosa PA14 but more virulent than rhlR (Fig 1A and B). Importantly, rhlI proliferated in the hemolymph, although it grew less rapidly than wild‐type PA14, whereas rhlR was cleared from the hemolymph (Fig 1C and D). Consistent with this latter result, rhlI but not rhlR triggered a systemic immune response, as monitored by measuring the expression of the Diptericin gene (Appendix Fig S1E and F). Similar results were obtained with independent rhlR and rhlI in‐frame deletion mutants constructed by another laboratory (Appendix Fig S1) 63. We conclude that RhlR and RhlI have differing functions in the Drosophila intestinal infection model.

Figure 1. RhlI displays a distinct phenotype from that of rhlR in the Drosophila intestinal infection model.

Wild‐type flies were fed on wild‐type (WT) PA14 bacteria, rhlR (ΔrhlR), or rhlI (ΔrhlI).

• A
  Representative survival curves of infected and non‐infected (NI) flies. Flies died faster from the infection with PA14 WT than with rhlR. Flies infected with rhlI exhibited an intermediate survival phenotype. One representative experiment is shown. Statistical analysis of the data is shown in (B).
• B
  Pooled LT₅₀ data of wild‐type flies (w ^A5001) following intestinal infections with PA14 WT, rhlR, or rhlI. LT₅₀ of flies after infection with PA14 WT was significantly lower than with rhlR or rhlI (***P < 0.0001 and *P = 0.0009, respectively). Flies were significantly more susceptible to infection with rhlI than with rhlR. The LT₅₀ data from seven survival experiments are displayed (biological duplicates are also shown as there was as much variability between experiments as within experiments). ***P < 0.001.
• C, D
  Bacterial titer of the hemolymph collected from flies that had ingested wild‐type or mutant PA14 as indicated, three (C) or five (D) days after the ingestion of PA14 WT or ingestion of mutants as indicated. In this series of experiments, flies infected with PA14 WT had started to succumb by day 5 and were therefore not analyzed. ***P < 0.001.
• E
  Survival curves of wild‐type and latex bead‐injected flies after intestinal infection with PA14 bacteria. In latex bead‐injected flies, rhlI regained virulence. Note, however, that the shift in virulence was of the same amplitude as that observed for PA14 WT and contrasts with the large shift observed with rhlR.
• F
  Pooled LT₅₀ data of latex bead‐injected flies (w‐LXB) survival experiments. w‐LxB flies died significantly slower after rhlR infection than with PA14 WT (*P = 0.015). A slight decrease in virulence was observed between PA14 WT and rhlI (*P = 0.02). No difference in virulence was detected between rhlR and rhlI. Data represent the LT₅₀s from five experiments (biological duplicates are also shown as there was as much variability between experiments as within experiments).
• G
  Differences of LT₅₀s between WT flies pre‐injected with PBS (WT, filled circles; these data are also shown in B) and flies pre‐injected with latex beads (LXB, open circles; these data are also shown in F) after intestinal infection with PA14 WT, rhlR, or rhlI, bars indicate medians. Δ represents the difference between PBS and LXB‐injected flies (LT₅₀[wt‐wt^LXB]), and this difference was highly significant for all bacterial genotypes (P = 0.0032 for WT PA14 and rhlI, P < 0.0001 for rhlR; not shown on the graph for simplicity). Data represent the LT₅₀s from five experiments (biological duplicates are also shown as there was as much variability between experiments as within experiments). *P < 0.05.

Data information: (B–D, F, and G) Bars represent medians, the upper and lower limits of boxes indicate, respectively, the first and third quartiles and whiskers encompass data points within 1.5 times the interquartile range. Data were analyzed using linear models and reported P‐values are relative to the post‐hoc pairwise comparisons between the groups of interest.Source data are available online for this figure.

rhlR and rhlI killed flies in which phagocytosis had been impaired by the prior injection of latex beads (LXB flies) much faster than PBS‐injected control flies, at approximately the same rate, but more slowly than wild‐type PA14 (Fig 1E and F). Does this reflect a regained virulence of the mutants in this context 25 or simply that these immunodeficient flies are more sensitive to P. aeruginosa intestinal infections? To answer this question, it is informative to measure the difference in LT₅₀ (the time for 50% lethality) values of control versus latex bead‐injected flies (Δ = LT₅₀[wt‐wt^LXB]) for each mutant and to compare it to the Δ measured for wild‐type PA14: If it is equal in magnitude, then the effect is likely non‐specific; if it is higher, it would suggest that the cellular immune response is targeted by the wild‐type version of the mutated bacterial gene. rhlR did recover virulence with a Δ of 4.7 days, compared to 2.4 days for wild‐type PA14 (Fig 1G), which corresponds to the level of recovered virulence reported earlier 25. With a value of 3.5 days, rhlI displayed an intermediate Δ, which was significantly distinct from that measured for ΔrhlR but not that of wild‐type PA14 (Fig 1G). Thus, we conclude that PA14 RhlR plays a predominant role in counteracting phagocytosis.

rhlI and wild‐type PA14, but not rhlR, strains colonize the C. elegans digestive tract

We tested whether the differences that we observed in the virulence of rhlR and rhlI mutants in the Drosophila intestinal infection assay were reflected in the C. elegans‐P. aeruginosa nematode “slow‐killing” survival assay 21, 64. Indeed, two independently constructed in‐frame rhlR deletion mutants in the PA14 background were markedly less virulent than two independently constructed rhlI deletion mutants in their ability to kill C. elegans (Fig EV1A). We next used a quantitative assay (Fig EV1B–E, see Materials and Methods) that monitors the accumulation of live bacterial cells in the intestine of the nematodes using PA14, rhlR, and rhlI bacteria expressing GFP to monitor the level of intestinal colonization. Live wild‐type P. aeruginosa PA14 cells start accumulating in the intestine 24–48 h post‐infection. Two independent rhlR mutants in the PA14 background colonized the C. elegans intestine at significantly lower levels than two independent rhlI mutants. Indeed, in this colonization assay, rhlI was indistinguishable from wild‐type PA14. An alternative explanation for these results is that C. elegans preferentially feeds on the rhlI compared to rhlR and simply overwhelms the immune response with a larger number of ingested cells. However, this possibility was ruled out by monitoring the pumping (feeding) rate of C. elegans feeding on wild‐type, rhlR, and rhlI. C. elegans pumped at the same rate on all three strains (Appendix Fig S2). These data suggest that RhlR but not RhlI is a predominant determinant for PA14 virulence in both fly and nematode intestinal infections and that the primary effect in C. elegans appears to be on colonization.

Figure EV1. rhlR mutants are impaired in their ability to kill C. elegans and do not colonize the C. elegans intestine.

• A
  An automated C. elegans lifespan machine was used to monitor the survival of worms in a P. aeruginosa‐mediated killing “slow‐killing” assay 64. Approximately 200 wild‐type C. elegans nematodes were fed WT PA14 (blue), rhlI (ΔrhlI: green), or rhlR (ΔrhlR: red) (constructed in the Ausubel or Bassler laboratories, hence two curves for each genotype); P < 0.001 (log‐rank test) for PA14 versus rhlI, PA14 versus rhlR, or rhlI versus rhlR. The experiment was repeated twice with similar results.
• B–E
  C. elegans wild‐type N2 animals were fed WT PA14, rhlI, or rhlR (constructed in the Ausubel (B and D) or Bassler (C and E) laboratories expressing GFP). At 48 h post‐infection, 13–18 worms infected with WT PA14, rhlI, or rhlR were imaged in the green fluorescent channel. (B, C) Representative images are shown. Scale bar is 200 μm. (D, E) Images were quantified using ImageJ. There was a significant difference in the levels of live bacteria between WT PA14 and rhlR (***P = 0.0003, D and E), and there were significant differences between the rhlR and rhlI mutants (***P = 8 × 10⁻⁷ (D) and 0.0005 (E)). The data fitted a negative binomial distribution and were analyzed using generalized linear models by one‐way ANOVA. The differences between rhlI and WT PA14 were not significant. The experiments were repeated at least two times with similar results. Bars represent medians, the upper and lower limits of boxes indicate, respectively, the first and third quartiles and whiskers encompass data points within 1.5 times the interquartile range.

Source data are available online for this figure.

Phagocytosis protects Drosophila against invasion of its hemocoel by wild‐type PA14 during the early phase of the infection

In our Drosophila intestinal infection models, flies constantly feed on the pathogen present on a filter. Using bacteria expressing different colored fluorescent proteins, we previously showed that S. marcescens continuously crosses the intestinal barrier during the infection 26. Figure 2A and B shows that when flies that had been feeding on dsRed‐labeled PA14 bacteria for 4 days were switched to a filter laced with GFP‐labeled PA14, the green bacteria progressively replaced the red bacteria both in the gut and in hemocoel compartments. We conclude that P. aeruginosa, like S. marcescens, continuously crosses the intestinal barrier during the infection.

Figure 2. Phagocytosis is required during the early stage of the infection in Drosophila .

• A, B
  Wild‐type Drosophila were orally infected with wild‐type PA14 expressing dsRed (PA14‐dsRed). After 4 days, infected flies were transferred to tubes containing wild‐type PA14 expressing GFP (PA14‐GFP). At day 5 of the infection (1 day after transferring flies to PA14‐GFP), most PA14 bacteria in the Drosophila gut expressed dsRed and only 10% expressed GFP (A). However, at day 6 (2 days after the transfer of flies to PA14‐GFP), only GFP‐positive bacteria were detected in the gut. Results were similar for bacteria retrieved from the hemolymph compartment (B). Each dot corresponds to a sample of 10 flies.
• C–E
  LT₅₀s from survival experiments of WT flies (w) after intestinal infection with PA14 WT (C), rhlR (D), or rhlI (E) mutants and injection of either latex beads (LXB, open circles) or PBS (filled circles) at different time points of the infection. Latex beads or PBS was injected either 1 day before the infection started (−1 day) or 4 h (+4 h), 1 day (+1 day), 4 days (+4 days), or 6 days (+6 days) after the infection started. Black filled circles correspond to the survival of infected, uninjected flies. (C) LT₅₀s of w ^A5001‐LxB were significantly lower than w ^A5001 only at −1 day (**P = 0.003) and +4 h (*P = 0.013). (D) LT₅₀s of w‐LXB flies were significantly lower than w at most times during the infection (−1 day: ***P = 5 × 10⁻⁵, +4 h: ***P = 3 × 10⁻⁶, and +4 day: **P = 0.01). (E) A similar phenotype was observed with flies infected with rhlI (−1 day: **P = 0.002, +4 h: **P = 0.006, +1 day: ***P = 8 × 10⁻⁶, and +4 days: *P = 0.03). Note, however, that for injections of latex beads at day 4, the difference was reduced, as compared to earlier time points of injection of latex beads. The cumulative LT₅₀ data from at least three experiments (only two experiments for rhlI) are shown, except for day 6.
• F
  Δ: Differences of LT₅₀s between WT flies injected with PBS (WT) and flies injected with latex beads (WT^LXB) after intestinal infection with PA14 WT, rhlR, or rhlI in at least two experiments. *P < 0.05; ***P < 0.001.
• G
  Guts of transgenic flies with GFP‐labeled hemocytes were dissected in a manner that preserves the association of hemocytes with the digestive tract and were examined by fluorescence confocal microscopy. Green: GFP; blue: DAPI staining of nuclei. Scale bars: 100 μm.
• H
  Analysis of hemocytes recruited to the fly intestine upon infection with either wild‐type PA14, rhlR, or rhlI. All three strains elicited recruitment of hemocytes to the gut (4 h after the beginning of the infection, for each bacteria ***P < 0.0001); at 3 days after the beginning of the infection, there were fewer hemocytes recruited after infection with rhlI (**P = 0.0025 between rhlR and rhlI). Data represent three pooled experiments. *P < 0.05.

Data information: (C–F and H) Bars represent medians, the upper and lower limits of boxes indicate, respectively, the first and third quartiles and whiskers encompass data points within 1.5 times the interquartile range. Data were analyzed using linear models and reported P‐values are relative to the post‐hoc pairwise comparisons between the groups of interest.Source data are available online for this figure.

Next, we asked at what time periods during an infection is phagocytosis important in preventing PA14 growth in the hemolymph. To this end, we saturated the phagocytic apparatus of hemocytes by injecting latex beads into flies at different time points during infection. As expected, blocking phagocytosis 1 day prior to the infection led to an earlier demise of the PA14‐infected flies compared to PBS‐injected control flies. Similar results were found when latex beads were injected 4 h or 1 day after infection, although in the latter case, the difference was not significant (its value was nevertheless similar to that obtained by injection 1 day prior to infection; Fig 2C). The injection of latex beads 4 or 6 days after the beginning of the ingestion of wild‐type PA14 did not affect the survival rate of flies. That is, the LT₅₀ values were similar to those of control (PBS‐injected) flies, consistent with the conclusion that starting about 4 days after infection the cellular immune response no longer plays a major role in limiting a wild‐type PA14 infection.

In contrast to wild‐type PA14, rhlR was kept in check by phagocytosis at least up to day four and to some extent up to 6 days after infection (Fig 2D and F). Phagocytosis was efficient against rhlI for approximately 4 days, which displayed again an intermediate phenotype (Fig 2E and F). These data suggest that hemocytes constantly keep in check bacteria that have crossed the intestinal barrier, but that as a consequence of RhlR function, wild‐type PA14 ultimately escapes this immune surveillance.

A recent study has reported that hemocytes are recruited to the gut after the ingestion of bacteria 65. We confirmed this finding in the case of P. aeruginosa oral infection, with a significant recruitment observed at 4 h after the beginning of the infection with either wild‐type PA14, rhlR, or rhlI mutants (Fig 2G and H, Appendix Fig S3). While hemocytes remained associated with the midgut for at least 3 days after the beginning of the ingestion of wild‐type PA14 or rhlR bacteria, they were not in the case of rhlI bacteria (Fig 2H). While some ingested bacteria could be detected in the hemocytes recruited to the gut, this phenomenon was not reproducible enough to allow reliable quantification of their uptake and thus to assess whether this recruitment of hemocytes to the gut may be involved in controlling bacteria that have crossed the intestinal barrier.

Drosophila Tep4 is required for host defense against ingested PA14

Given the importance of phagocytosis in controlling bacteria that have escaped from the digestive tract, we asked whether Teps might not be involved in this process as opsonins. We tested mutations affecting the Tep2, Tep3, and/or Tep4 genes 44 in the intestinal infection assay. Only Tep4 and a triple Tep2‐Tep3‐Tep4 mutant displayed an increased susceptibility to PA14 ingestion (Fig 3A and D). Of note, uninfected Tep3 mutants fed on a sucrose solution displayed an enhanced fitness when compared to wild‐type or other Tep mutant lines, which may contribute to its apparent decreased susceptibility to ingested PA14 (Appendix Fig S4C, Fig 3A and B). In the case of Tep1, since no mutants were available, we tested an RNAi transgene expressed either in hemocytes or in the fat body; however, we did not observe any change in the virulence of ingested PA14 in the Tep1 knockdown (Appendix Fig S4A and B). We conclude from these data that Tep4, but not other thioester‐containing proteins, is required for host defense against ingested PA14.

Figure 3. RhlR circumvents Tep4‐mediated host defense.

• A, B
  Drosophila wild‐type flies (w ^A5001), single Tep3 or Tep4 mutants, the double Tep2,3 mutant, and the triple Tep2,3,4 mutant were orally infected with PA14 WT (A) or rhlR (B) in parallel survival experiments. (A) Tep4 and Tep2,3,4 mutant flies were significantly more susceptible to PA14 infection compared to w ^A5001. No difference in survival was detected between the Tep2,3 mutant and w ^A5001. Surprisingly, Tep3 mutants seemed to be more resistant to infection. (B) A strong enhancement of rhlR virulence was observed with the Tep4 and Tep2,3,4 mutants compared to w ^A5001 flies. Tep2,3 and w ^A5001 exhibited nearly the same rate of death when challenged with rhlR. The Tep3 mutant seemed again to be more resistant to the infection. One representative experiment out of three (each with biological triplicates, except for uninfected flies) is shown.
• C
  The survival of Tep4 flies infected with PA14 WT, rhlR, or rhlI was examined. One representative experiment out of three (each with biological triplicates) is shown.
• D
  Quantification of the experiments shown in (C), which had been performed in parallel with w ^A5001 flies; filled circles: wild‐type flies, open circles: Tep4. The triplicates were analyzed as independent experiments as there was as much variability between experiments as within experiments. All differences between wild‐type flies and the Tep4 mutant (Δ) were highly significant (***P < 0.0001). Bars represent medians, the upper and lower limits of boxes indicate, respectively, the first and third quartiles and whiskers encompass data points within 1.5 times the interquartile range. Data were analyzed using linear models and reported P‐values are relative to the post‐hoc pairwise comparisons between the groups of interest. **P < 0.01; ***P < 0.001.

Source data are available online for this figure.

Next, we found that rhlR became almost as virulent as wild‐type PA14 only when ingested by Tep4 or Tep2‐Tep3‐Tep4 mutants (Fig 3A, B and D). rhlR recovered virulence to a much larger extent than rhlI when ingested by Tep4 flies (Fig 3C; compare the Δ for rhlR and rhlI, Fig 3D). Interestingly, the injection of latex beads in Tep4 flies only modestly increased the virulence of ΔrhlR bacteria when compared to PBS‐injected Tep4 flies (Appendix Fig S4D and E), suggesting that phagocytosis of rhlR is already severely affected in the Tep4 mutant. Hence, the behavior of the rhlR mutant is similar in eater, which encodes a putative phagocytic receptor 27, and Tep4 mutant flies 25, thereby raising the possibility that both fly genes are involved in the same process. These data suggest that a RhlR‐dependent process bypasses the Tep4‐mediated control of P. aeruginosa present in the hemocoel.

Phagocytosis of PA14 bacteria is impaired in Tep4 mutant hemocytes

To quantitatively monitor the uptake of PA14, we used assays that relied on larval hemocytes. First, we injected heat‐killed, pHrodo^®‐labeled bacteria in wild‐type or Tep4 third‐instar larvae. After 45 min, the larvae were bled and a phagocytic index was established. When analyzed globally by ANOVA, the data showed that wild‐type hemocytes ingested significantly more bacteria than Tep4 hemocytes (P = 0.004), but when comparisons were made according to the genotype of the phagocytosed bacteria, a significant difference was detected for wild‐type PA14 but not for rhlR (Fig 4A). Similarly, a global analysis revealed significant differences between heat‐killed wild‐type PA14 and rhlR uptake by hemocytes (P = 0.01), yet a difference between these two bacterial genotypes was significant only for Tep4 hemocytes (Fig 4A). This was not necessarily unexpected as heat‐killing likely inactivates rhlR‐dependent virulence factors and might also alter the surface of bacteria.

Figure 4. Tep4 is required for phagocytosis of rhlR mutant bacteria.

1. Heat‐killed pHrodo^®‐labeled bacteria of the indicated genotype were injected into either wild‐type or Tep4 third‐instar larvae and incubated for 45 min. The hemocytes were then retrieved. Bacteria present in phagosomes were fluorescent and used to measure the phagocytic index. rhlR bacteria injected into Tep4 (open circles) were significantly more phagocytosed than wild‐type bacteria (filled circles) (P = 0.02); PA14 WT were more readily phagocytosed by wild type than by Tep4 hemocytes (*P = 0.01).
2. The experiment is similar to that shown in (A), except that live bacteria were used and a differential antibody staining procedure was performed to reveal phagocytosed bacteria: Phagocytosed bacteria were stained only after permeabilization and fluoresced green whereas non‐ingested bacteria were stained by both secondary antibodies, red, and green, which were, respectively, used prior and after the permeabilization step (see Materials and Methods section). PA14 WT, rhlR, or rhlI were injected in wild‐type or Tep4 larvae. Again, PA14 WT was more readily phagocytosed by wild‐type than by Tep4 larval hemocytes (***P = 0.0006). rhlR were engulfed more efficiently than PA14 WT bacteria by Tep4 hemocytes (*P = 0.01).

Data information: Bars represent medians, the upper and lower limits of boxes indicate, respectively, the first and third quartiles and whiskers encompass data points within 1.5 times the interquartile range. Data were analyzed using generalized linear models assuming a negative binomial distribution and reported P‐values are relative to the post‐hoc pairwise comparisons between the groups of interest.Source data are available online for this figure.

We therefore modified the assay by using live bacteria 66 and an antibody we had raised against PA14 to differentially immuno‐stain the bacteria, prior to and after permeabilization of the fixed cells. As before, Tep4 hemocytes exhibited a decreased uptake of bacteria compared to wild‐type hemocytes when analyzed globally (P = 0.0002); however, in the case of the comparisons by bacterial genotypes, only the difference for wild‐type PA14 was significant (Fig 4B). When analyzing the data globally according to bacterial genotypes, a significant difference was found between wild‐type PA14 and rhlR (P = 0.0005) but not between wild‐type PA14 and rhlI nor between rhlR and ΔrhlI. When comparing wild‐type PA14 and rhlR specifically per fly genotype, the difference between these two strains was significant only in the Tep4 mutants. Upon repeating this experiment in wild‐type flies using bacterial strains obtained from another laboratory, we observed that rhlR bacteria were significantly better phagocytosed than either wild‐type PA14 or rhlI (Appendix Fig S1G). In summary, both assays revealed that wild‐type flies ingest wild‐type PA14 better than Tep4 mutant flies and that rhlR appears to be better phagocytosed than wild‐type PA14 in Tep4 mutant flies. Both phagocytosis assays, however, failed to discriminate between rhlR and rhlI, possibly because they are not sensitive enough.

Tep4 opsonizes rhlR better than rhlI or wild‐type bacteria

We next designed an experiment to assess whether Tep4 functions as an opsonin, that is, that it is deposited on the surface of bacteria to facilitate its detection and subsequent ingestion by hemocytes (Fig 5A). Wild‐type PA14 was poorly phagocytosed in this assay (medians of phagocytic index lower than 10), whether the bacteria were initially incubated in the hemolymph of Tep4‐containing wild‐type larvae or Tep4 larvae and then secondarily injected into wild‐type (Fig 5B) or Tep4 (Fig 5C) larvae. When these experiments were performed using rhlR bacteria that were injected into Tep4 recipient larvae (Fig 5C), it made a major difference whether these mutant bacteria had initially been pre‐exposed to wild‐type or Tep4 hemolymph. Tep4‐dependent opsonization led to a massive uptake of bacteria (median phagocytic value of 160), whereas non‐opsonized bacteria (pre‐incubated with Tep4 mutant hemolymph) were hardly ingested (median phagocytic value of 10) when the treated bacteria were injected into larvae. As expected, initially non‐opsonized bacteria that were then injected in wild‐type recipients (Fig 5B) were much better phagocytosed (median phagocytic value of 210), presumably because Tep4 was circulating in the hemolymph of wild‐type recipient larvae. They were nevertheless ingested less efficiently than opsonized bacteria (median phagocytic value of 380; Fig 5B). Finally, rhlI bacteria were also opsonized in a Tep4‐dependent manner, but significantly less than rhlR bacteria (Fig 5B: P = 0.0005 for opsonized bacteria, P = 0.0008 for non‐opsonized bacteria; Fig 5C: P < 0.0001 for opsonized bacteria). Again, they displayed a phenotype that was intermediate to that of wild‐type PA14 on the one hand, and rhlR on the other, although it was quantitatively closer to that of rhlR than that of wild‐type PA14. Of note, rhlR and rhlI bacteria behaved similarly in this assay when Tep4 was entirely absent (non‐opsonized bacteria in Tep4 recipient larvae (Fig 5C)). We conclude that RhlR plays a major role in eluding Tep4‐mediated opsonization of wild‐type PA14. Unexpectedly, RhlI also significantly contributes to this process, even though it does not play a predominant role in virulence.

Figure 5. Tep4 is an opsonin that preferentially detects rhlR over rhlI mutant bacteria.

• A
  Scheme of the experimental procedure. Live bacteria were incubated with either wild‐type or Tep4 hemolymph and were thereafter injected into naive larvae, which were either wild‐type or Tep4. The phagocytosis index was then measured as in Fig 4B.
• B, C
  Bacteria pre‐incubated with wild‐type or Tep4 hemolymph are represented in pairs, respectively, with filled (left) and open (right) circles. PA14 WT, rhlR, or rhlI bacteria pre‐incubated with either wild‐type (wt) or Tep4 hemolymph were injected into wild‐type larvae (B) or into Tep4 larvae (C). Data were analyzed using generalized linear models assuming a negative binomial distribution and statistically homogeneous groups (indicated by letters) were defined based on the post‐hoc pairwise comparisons between the different groups. The difference between each group is highly significant (P < 0.001). Bars represent medians, the upper and lower limits of boxes indicate, respectively, the first and third quartiles and whiskers encompass data points within 1.5 times the interquartile range.

Source data are available online for this figure.

RhlR may function cell autonomously

RhlR either functions at the level of the bacterial cell in which it is expressed, for example, by altering the structure of the cell wall to prevent detection by Tep4 or, alternatively, it might control the expression of secreted factors such as proteases that would inactivate Tep4 in a non‐cell‐autonomous process. To discriminate between these two possibilities, we performed the opsonization assay described above on a 1:1 mix of wild‐type PA14 and rhlR and compared their ingestion profiles to that of either strain alone subjected to the assay. Figure 6A shows that when mixed, wild‐type PA14 was unable to rescue the Tep4‐sensitivity phenotype of rhlR bacteria. A comparison of the profiles of phagocytosed bacteria (Fig 6B) supports the conclusion that the ingested bacteria in the mix are predominantly rhlR and not wild‐type PA14. The cell‐autonomous function of RhlR is in keeping with the observation that hemocytes remain able to phagocytose Escherichia coli even late during the infection 25.

Figure 6. RhlR appears to function cell autonomously in eluding opsonization.

• A, B
  PA14 WT, rhlR, or a 1:1 mix thereof was submitted to the opsonization assay using wild‐type larvae as donors and recipients. Panel (A) displays the measured phagocytic indexes: PA14 WT is hardly ingested by larval hemocytes, whereas rhlR and the mixture of wild‐type and rhlR bacteria display a similar distribution, as shown in (B), where the cumulative distribution of phagocytic indices is shown. The purple curve (PA14_WT_ΔrhlR_join) shows the distribution that would be obtained by pooling the data from rhlR and PA14 WT alone in this opsonization assay. The observed distribution of the phagocytic indices in the mix is clearly closer to that of rhlR (two‐sample Kolmogorov‐Smirnov test, P = 0.98), suggesting that the ingested bacteria are mostly of the rhlR genotype. Bars represent medians, the upper and lower limits of boxes indicate, respectively, the first and third quartiles and whiskers encompass data points within 1.5 times the interquartile range.

Data information: (A) Data were analyzed using generalized linear models assuming a negative binomial distribution and reported P‐values are relative to the post‐hoc pairwise comparisons between the different groups. *P < 0.05.Source data are available online for this figure.

Tep4 plays an adverse role in a PA14 systemic infection model in Drosophila

A recent study has reported that Tep4 mutants are more resistant than wild‐type flies to a systemic infection with the entomopathogenic bacterium Photorhabdus luminescens in a septic injury model 57. By injecting several doses of PA14, from 10 to 1,000 CFUs, directly into the thorax of flies, we also found that Tep4 mutants survived better than wild‐type flies in this systemic infection model (Fig 7A). This difference in survival between Tep4 and wild‐type flies was no longer observed when rhlR was injected, thereby establishing again a relationship of altered virulence of these bacteria in a Drosophila Tep4 mutant background (Fig 7B). Using the steady‐state expression of the antibacterial peptide gene Diptericin as a read‐out of the activation of the IMD pathway that regulates the systemic immune response against Gram‐negative bacteria, we found no consistent difference of expression between wild‐type and Tep4 (Fig 7C). As a higher level of induction of the IMD pathway is unlikely to account for the increased resistance of Tep4 mutants against PA14 infection, we tested whether the phenoloxidase cascade was more efficiently activated in this mutant background, as previously reported 57. Indeed, we found that prophenoloxidase was more extensively cleaved in Tep4 than in wild‐type flies (Fig 7D). These data suggest that injected wild‐type PA14 bacteria elude detection by the factors that trigger the phenoloxidase cascade and that Tep4 plays a role in this process of evasion from the melanization response.

Figure 7. Tep4 mutants are more resistant to a systemic infection with PA14.

• A, B
  The survival of wild‐type and Tep4 flies was examined, after injection of PBS as a non‐infected control (NI), or 10, 100, 1,000 CFUs of PA14 WT (A) or rhlR (B). One representative experiment out of 10 (A) or out of three (B) is shown. **P < 0.01; ***P < 0.001. A log‐rank test was used.
• C
  Diptericin expression was measured by RT–qPCR in wild‐type and Tep4 flies at 2, 8, or 24 h after injection of PBS (NI), or of 10, 100, 1,000 CFUs of PA14 WT bacteria. Each experiment was performed independently three times, and a representative experiment is shown. The Diptericin expression shown is relative to RpL32 expression and normalized to the non‐infected wild‐type control. Error bars represent the standard deviation. *P < 0.05; **P < 0.01 (unpaired t‐test).
• D
  The cleavage of the prophenoloxidase was analyzed 2 h after injection of 1,000 CFUs of PA14 WT bacteria or PBS, by Western blotting using a pan‐phenoloxidase antibody. The intensity of the prophenoloxidase (PPO) and of the cleaved active phenoloxidase (activated PO) bands was measured, and the ratio of the measurements is shown below the blot. This experiment was performed twice with a similar result.

Source data are available online for this figure.

Discussion

In this study, we analyzed the interactions of P. aeruginosa with Drosophila from the dual perspective of both pathogen and host. Our data lead us to propose a model in which the P. aeruginosa quorum‐sensing regulator RhlR plays a pivotal role in virulence by diminishing the ability of the cellular immune arm of the host defense response to detect P. aeruginosa once the bacteria have reached the internal body cavity of the insect in an intestinal infection model. Surprisingly, the function of RhlR in eluding opsonization by Tep4 is at least partially independent of RhlI, the enzyme that synthesizes C4‐HSL, the quorum‐sensing signaling molecule that corresponds to RhlR. We propose that this novel C4‐HSL‐independent function of RhlR is to modify the bacterial surface, thereby altering the detection of P. aeruginosa by the immune system through Tep4. Whereas Tep4 has a protective function in this intestinal infection model, it plays a detrimental role in Drosophila host defense in a septic injury model, thereby underscoring that the efficiency of specific host defenses depends on the infection route and virulence strategy of the pathogen.

A rhlI‐independent function of rhlR

Null rhlI and rhlR mutants display distinct virulence phenotypes as monitored in survival experiments in both the Drosophila and the Caenorhabditis intestinal infection models. The virulence of rhlI bacteria is moderately impaired whereas that of rhlR is severely affected. In the nematode worm, the strongly attenuated virulence of rhlR likely reflects its decreased ability to colonize the intestinal tract. In the fly, systemic bacteremia is thought to cause the demise of the host: Bacteremia in the hemolymph is delayed in the case of rhlI infection but does not occur for rhlR, thereby highlighting a major phenotypic difference between the two mutants. Thus, rhlR appears to have a rhlI‐independent function in vivo. In support of this conclusion, a recent study by Mukherjee et al 67 showed that rhlI and rhlR mutants exhibit different colony biofilm morphologies, that rhlI mutants are more virulent than rhlR mutants in a mouse infection model, and that there are three categories of rhlR‐dependent regulons that require rhlI function to varying degrees. The Mukherjee et al 67 study also established that RhlR activity in vitro depends on a second ligand, the nature of which remains elusive at this stage. It is unknown whether a ligand distinct from C4‐HSL is also required for virulence in flies, for instance, to stabilize the folding of the hydrophobic core of the RhlR regulators, a role thought to be assumed by HSL for LuxR regulators 3. Of note, RhlR does not appear to be activated by 3‐oxo‐C12‐HSL, the LasR ligand 67. Furthermore, lasR and lasI mutant bacteria display only a modestly decreased virulence phenotype in the Drosophila intestinal infection model (Appendix Fig S5). Because phagocytosis and Tep4‐mediated detection represent an efficient host defense at the very beginning of the infection when only few bacteria can be retrieved from the hemolymph, it seems unlikely that an alternative RhlR ligand is expressed in a quorum‐sensing‐dependent manner but rather as the result of the specific conditions the bacteria encounter in the host hemolymph or when going through its digestive tract.

Because rhlR function is required at a stage in which only a few bacteria can be retrieved, it is technically difficult to measure gene expression in this context and specific reporter transgenes will be needed to assess whether rhlR‐dependent‐rhlI‐independent genes are expressed. Nevertheless, because of the similarities observed between the fly intestinal infection model and the nematode slow‐killing infection paradigm observed in this study, we used the Drosophila intestinal infection model to screen more than 300 PA14 mutant lines that had been identified as less virulent in C. elegans 68. This approach, however, failed to identify mutants, or combinations thereof, with as strong a phenotype as rhlR (Haller et al, in preparation). Thus, this approach appears to have failed to identify any gene(s) encoding the enzyme synthesizing a putative second RhlR ligand.

RhlR counteracts the cellular host defense by eluding detection by Tep4

Our phagocytosis and opsonization data are consistent with a model in which RhlR controls the expression of gene products that mask the site being recognized by Tep4 or a Tep4‐associated protein, presumably on the bacterial cell wall. Alternatively, RhlR may actively inhibit the uptake of opsonized bacteria, for instance by impeding the cytoskeletal rearrangements in hemocytes that are necessary for phagocytosis, but this process would not involve factors secreted in the hemolymph.

Insect thioester‐containing proteins belong to the complement family and have been shown to be involved in the opsonization of bacteria in mosquitoes. The deletion of four Drosophila Teps did not render the flies susceptible to infection with non‐pathogenic Gram‐negative bacteria 55. Tep2 has been reported to be required for the uptake of E. coli, a Gram‐negative bacterium, by cultured S2 cells 48, a finding confirmed in vivo 49. In contrast, we find no involvement of Tep2 in our in vivo intestinal infection model with P. aeruginosa, at least in terms of virulence, but do detect a requirement for Tep4 in phagocytosis and opsonization assays. Given that the structure of mosquito thioester‐containing protein 1 is similar to that of complement family C3 34, a well‐described opsonin, our data are compatible with a model of direct opsonization of bacteria by Tep4. Why Tep4 and not other Teps mediate this effect will require further investigations and a better understanding of the exact mechanism of action of Drosophila Teps.

It is striking that the opsonization phenotype of rhlI is quantitatively closer to that of rhlR than that of wild‐type PA14, which appears to be poorly opsonized. In contrast, both wild‐type PA14 and rhlI ultimately cause a systemic bacteremia while rhlR is cleared from the hemolymph. Thus, a relatively minor difference in opsonization efficiency translates to a major effect with respect to the establishment of a systemic infection. This observation underscores the critical importance of opsonization levels in controlling pathogenic bacteria that escape from the digestive tract. Alternatively, the difference of virulence observed between rhlI and rhlR would not be linked to opsonization.

The opposite roles of Tep4 in host defense against the same pathogen according to the infection route

The finding that Tep4 plays a protective function in the intestinal infection model whereas it is detrimental in the case of a direct systemic infection is paradoxical. This may actually represent two faces of the same phenomenon. P. aeruginosa may have developed a stealth strategy to avoid detection by the immune system and thus may actively hide any features that might reveal its presence. We propose here that RhlR plays a critical role in a program that allows P. aeruginosa to function in a stealth mode that does not activate an immune response. As a result of RhlR action, only a few sites would be available on the surface of the wild‐type bacteria for Tep4 direct or indirect binding. There, Tep4 would mediate opsonization and then subsequent phagocytosis of the bacteria. Our data in the systemic infection model are compatible with the possibility that Tep4 competes for these sites with pattern recognition receptors (PRRs) that activate the phenoloxidase cascade since PO activation occurs efficiently only in the Tep4 mutant background. It is likely that small peptidoglycan (PGN) fragments released by proliferating bacteria represent a major trigger of the IMD pathway in addition to large PGN fragments directly sensed by PGRP‐LC, thereby accounting for the apparent normal expression of Diptericin when flies are challenged with injected PA14. In contrast, we have previously established that some fungi and Gram‐positive bacteria trigger the phenoloxidase cascade through defined PRRs 62. The situation is less clear as regards Gram‐negative bacteria. While the original characterization of PGRP‐LE suggested that it triggers the phenoloxidase activation cascade 69 and acts non‐cell autonomously 70, subsequent studies documented a role for PGRP‐LE as an intracellular sensor for PGN fragments 28, 71, 72. Thus, further work will be required to identify whether Tep4 actually competes with PRRs in the detection of pathogens. A major challenge will be to establish how RhlR influences the surface properties of PA14 or alternatively inhibits the uptake of opsonized bacteria.

Our results with Tep4 further suggest that the cellular immune response is a relevant defense when a few bacteria enter the hemocoel after escaping from the digestive tract in the intestinal infection model, but not in the septic injury model. Conversely, melanization mediated by activated phenoloxidase appears to be somewhat effective after injection of PA14. Thus, infection by different pathogens challenges the immune system in unique ways that depend not only on the pathogen but also on its infection route. Thus, depending on the pathogen and mode of infection, some host defenses may be important whereas others might be bypassed or neutralized or worse, become detrimental by attenuating the action of the appropriate defense responses.

Finally, rhlR bacteria exhibit reduced dissemination capacities in a rodent lung infection model when compared to rhlI or wild‐type PA14 67. By analogy to our findings in the Drosophila intestinal infection model, it will be interesting to determine whether the complement system restricts the systemic escape of rhlR from the mouse lung into the periphery.

Materials and Methods

Many of the methods employed in this study were described in detail in Haller et al 73.

Drosophila stocks and culture

The following strains were used: w ^A5001 as wild type 74, and the mutants Tep3 (d03976) 74, Tep4 (EY04656) 75, Tep2,3 double mutant, Tep2,3,4 triple mutant, and UAS‐Tep1 RNAi (ML2D) were described before 44. Of note, the Tep4 mutant is in a yw background; we have tested this mutant background in survival experiments and it did not differ significantly from the w ^A5001 background (Appendix Fig S6). The Tep4 and the compound Tep2,3,4 triple mutants are in distinct genetic backgrounds since the latter was generated by recombination between a Tep4 mutant in a yw background and a Tep2‐3 chromosome in a w ^A5001 background 44. Transgenic flies allowing to label hemocytes with GFP (hml::Gal4, UAS::GFP) were obtained by the Bloomington Drosophila Stock Center (BL70140). c564 Gal Gal80 ^ts and hml Gal4 Gal80 ^ts fly lines were used as drivers for UAS‐Tep1 RNAi 76. These crosses were done at 18°C. After hatching, F1 flies were transferred to 29°C for 6 days to allow a strong expression of the Tep1 RNAi construct. To block the phagocytosis ability of Drosophila, latex beads were injected in these flies 1 day before infection or upon the infection (Fig 2) as described before 73, except that flies were allowed to recover for 4 h after injection prior to being exposed to the bacteria.

Experiments were performed on 3‐ to 10‐day‐old female flies devoid of known pathogens, including Nora virus and microsporidia. Sample size was chosen according to the type of experiments and experiments repeated as needed.

Bacteria strains and growth conditions

Construction of the PA14 ΔrhlR mutant was described in Limmer et al 25. The PA14 ΔrhlI in‐frame deletion mutant was generated by replacing 2.018 kb of wild‐type PA14 sequence with a 1.502‐kb PCR‐amplified fragment that contained a 0.516‐kb deletion in the rhlI ORF. The PCR‐amplified fragment containing the deleted gene was subcloned into the XbaI site of pEX18Ap 77, generating plasmid pEX18rhlIΔ9. The resulting construct was used to introduce the rhlI‐deleted gene into the wild‐type PA14 genome by homologous recombination, resulting in the ΔrhlI 9‐9 mutant. Colonies of putative mutants obtained after homologous recombination were screened by PCR using appropriate flanking primers and were further confirmed by sequencing the corresponding PCR products.

P. aeruginosa strain PA14 wild‐type 78, ΔrhlR (rhlR), ΔrhlI (rhlI), ΔlasR, and ΔlasI mutants 25, 63 were grown in brain–heart infusion broth (BHB), overnight, at 37°C with agitation.

Drosophila intestinal infections

Infections were performed as described previously with PA14 73. At least 20 flies were assessed per infection tube. Infected and control flies were kept at 25°C. For survival assays, the number of surviving flies was computed daily.

C. elegans killing and intestinal accumulation assays

The “slow killing” of C. elegans by P. aeruginosa PA14 wild‐type and PA14 rhlR and rhlI mutants was monitored using automated image analysis as previously described 64, 79. To monitor the accumulation of P. aeruginosa in the C. elegans intestine, wild‐type (N2) animals were used for all experiments. Worms were reared on non‐pathogenic E. coli OP50 on nematode growth media at 25°C. Synchronized L4 worms (fourth larval stage) were transferred to slow kill (SK) nematode growth media agar plates containing a lawn of P. aeruginosa PA14::GFP. Post‐infection at 24 and 48 h, approximately 20 worms were picked onto a 2% agar pad that contain the paralyzing agent levamisole (1 mM). The worms were imaged in the GFP channel using a Zeiss Apotome microscope with the same exposure time for all the worms on wild‐type PA14 and the rhlR and rhlI mutants. The images were processed post‐acquisition using ImageJ software, and the area and fluorescence intensity were measured. The relative fluorescence intensity is plotted.

Analysis of hemocyte recruitment at the gut upon intestinal infection with PA14 or mutant bacteria

Female transgenic flies hml::Gal4, UAS::GFP, 3–5 days old were kept at 29°C for 3 days to induce the expression of the GFP. Flies were infected with wild‐type PA14, rhlR, or rhlI as described above. Non‐infected flies (kept on sucrose only) were used as control. At determined time points, flies were dissected and the gut was removed for analysis. In order to maintain the attachment of hemocytes to gut, flies were directly dissected in 4% PFA and processed as described before 65. Guts were analyzed using a ZEISS LSM700 confocal, and hemocytes attached to the gut were manually counted.

In vivo phagocytosis assay of pHrodo®‐labeled heat‐killed bacteria

The pHrodo^® dye becomes fluorescent when placed in an acidic environment such as that encountered in phagosomes. Wild‐type PA14 and rhlR overnight cultures in BHB at 37°C were washed in PBS and heat‐killed at 65°C for 60 min. Killed bacteria were then washed twice in PBS and labeled with pHrodo^® following the manufacturer's instructions (pHrodo^® Red succinimidyl ester, #P36600, Molecular Probes). Solutions at 5 × 10¹⁰ bacteria/ml were further aliquoted and kept frozen at −20°C before use. A5001 or Tep4 third‐instar wandering larvae (n = 23–35) were injected with 27.6 nl of the pHrodo‐labeled bacteria solutions, using a Nanoject apparatus (Drummond). After 40 min of incubation, one larva was bled in each well of an 8‐well pattern microscopy slide that contained a 0.5 μg/ml DAPI solution in PBS. The cells were left to settle to the bottom of the well for 30 min in a humid chamber. The slides were mounted in PBS and immediately analyzed using a Zeiss Axioskope 2 fluorescent microscope. 10–80 cells per larva were analyzed: The number of red‐fluorescent pHrodo‐labeled bacteria was counted for each DAPI‐positive hemocyte, and the phagocytosis index was calculated (percentage of phagocytes containing at least 1 bacterium) × (mean number of bacteria per phagocyte).

In vivo phagocytosis assay of live bacteria

Overnight cultures of wild‐type PA14, rhlR, and RhlI were concentrated to OD₆₀₀ = 20 in PBS. A5001 or Tep4 third‐instar wandering larvae (n = 20–38 for A5001; n = 4–8 for Tep4) were injected with 27.6 nl of the live bacteria solutions using a Nanoject apparatus (Drummond). After 40 min of incubation, one larva was bled in each well of an 8‐well pattern microscopy slide that contained PBS. The cells were left to settle to the bottom of the well for 30 min and then were fixed in 1% paraformaldehyde for 10 min, in a humid chamber. The fixed cells were washed twice in PBS, and after a 30‐min incubation in a PBS solution with 2% BSA, the samples were stained with a 1/100 to 1/500 diluted rabbit antiserum against PA14 in a PBS solution with 2% BSA overnight at 4°C or for 2 h at room temperature (the rabbit antiserum was raised against UV‐killed wild‐type PA14 by Covalab and also detects rhlI and rhlR bacteria). The cells were incubated with a Cy3‐labeled goat anti‐rabbit secondary antibody (Invitrogen) in a PBS solution with 2% BSA for 2 h at room temperature. After a 20‐min permeabilization step in a PBS solution with 0.1% Triton X‐100 and 2% BSA, a second round of staining with the diluted rabbit antiserum against PA14 in a PBS solution with 0.1% Triton X‐100 and 2% BSA was performed overnight at 4°C. The samples were then incubated with a FITC‐labeled goat anti‐rabbit secondary antibody (Invitrogen) in a PBS solution with 0.1% Triton X‐100 and 2% BSA for 2 h at room temperature. The slides were mounted in Vectashield with DAPI (Vector Laboratories) and analyzed using a Zeiss Axioskope 2 fluorescent microscope. 40–50 cells per larva were analyzed: The number of green fluorescent bacteria that were not also red (phagocytosed bacteria) was determined for each DAPI‐positive hemocyte, and the phagocytosis index was calculated (percentage of phagocytes containing at least 1 only‐green bacterium) × (mean number of only‐green bacteria per phagocyte).

Opsonization assay of live bacteria

Overnight cultures of PA14 wild‐type, rhlR, and RhlI mutants were concentrated to OD₆₀₀ = 10 in PBS. Twenty third‐instar larvae were bled in 150 μl of bacteria, resuspended in PBS at OD₆₀₀ = 10, and incubated at room temperature for 30–45 min to allow a hypothetical Tep4‐dependent opsonization to occur. Samples were centrifuged at 500 g for 15 min, and the pellet (containing larval debris) was removed. A second centrifugation was performed at 3,500 g for 15 min to retrieve bacteria in the pellet, which were resuspended in 10 μl PBS. These bacteria were then injected into either naive wild‐type or Tep4 mutant larvae prior to bleeding these injected larvae to be able to count the ingested bacteria as described above. In some experiments, a 1:1(v/v) solution from the PA14 wild‐type and rhlR solutions was prepared. A5001 or Tep4 third‐instar larvae were injected with 32.2 nl of the live bacteria solutions using a Nanoject apparatus (Drummond). After 60–90 min of incubation, one larva was bled in each well of an 8‐well pattern microscopy slide that contained PBS and stained as described in the previous section.

Phenoloxidase cleavage assay

The procedure was performed as described 80, except that hemolymph loads were not adjusted by measuring the protein content of the extracted hemolymph. The antibody was generated by Dr. H. M. Müller against Anopheles phenol oxidases 81. The ratio of cleaved to non‐cleaved form was determined by densitometry scanning.

Statistical analysis

Except when indicated, data were analyzed in a (generalized) linear modeling framework performed using R (version 3.4.2). Global differences in the mean response (e.g., LT50, bacterial counts or phagocytosis index) between the different experimental conditions (bacteria and fly genotypes, latex bead injected, and other experimental treatments) and their combinations were tested by analysis of variance/deviance (Anova function in the “car” package). Pairwise comparisons between specific conditions and “differences of differences” (e.g., comparison of the difference in LT50 between wild‐type and RhlR bacteria in WT and phagocytosis‐impaired flies) were done using the lsmeans and contrast function of the “lsmeans” package. Normally (Gaussian) distributed data (e.g., LT₅₀s, qPCR datasets) were tested using linear models (lm function in “base” R). Count data (i.e., phagocytosis index and bacterial counts) were analyzed using generalized linear models (GLM) assuming a negative binomial distribution (glm.nb function in the “MASS” package). This distribution is an extension of the Poisson distribution (commonly used to model count data), which allows inequality between means and variances. Normality of the residuals was assessed visually by quantile plots and tested by the Shapiro–Wilk test. Reported P‐values for tests involving multiple comparisons were adjusted using a sequential Bonferroni correction. Survival curve statistical analysis (Fig 7) was performed using the log‐rank (Mantel–Cox) as implemented in Graphpad Prism version 5 (Graphpad software Inc., San Diego, CA). Details are included in the legend of each figure. *P < 0.05; **P < 0.01; ***P < 0.001.

Author contributions

SH, AF, SL, and DF conceived the Drosophila experiments and analyzed the data. SH performed the majority of these experiments, with later work performed by AF and SL; SS performed the precursor experiments that led to this work. Work on the septic injury model was performed by JC, with some help from ZL. ΔlasR, ΔlasI, ΔrhlR, and ΔrhlI mutants were constructed by ED and SY, except when indicated otherwise. AH and FMA conceived the C. elegans experiments and analyzed the data, which were obtained by AH. NM performed the statistical analysis. SH, DF, and FMA wrote the manuscript, with inputs from other co‐authors.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Expanded View

Review Process File

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 6

Source Data for Figure 7

EMBO Reports (2018) 19: e44880