A specific innate immune response silences the virulence of Pseudomonas aeruginosa in a latent infection model in the Drosophila melanogaster host

Microbial pathogenicity often depends on the route of infection. For instance, P. aeruginosa or S. marcescens cause acute systemic infections when low numbers of bacteria are injected into D. melanogaster flies whereas flies succumb much slower to the continuous ingestion of these pathogens, even though both manage to escape from the gut compartment and reach the hemocoel. Here, we have developed a latent P. aeruginosa infection model by feeding flies on the bacteria for a short period. The bacteria stably colonize internal tissues yet hardly cause any damage since latently-infected flies live almost as long as noninfected control flies. The apparently dormant bacteria display particular characteristics in terms of bacterial colony morphology, composition of the outer cell wall, and motility. The virulence of these bacteria can however be reactivated upon wounding the host. We show that melanization but not the cellular or the systemic humoral response is the predominant host defense that establishes latency and may coerce the bacteria to a dormant state. In addition, the lasting activation of the melanization responses in latently-infected flies provides a degree of protection to the host against a secondary fungal infection. Latent infection by an ingested pathogen protects against a variety of homologous or heterologous systemic secondary infectious challenges, a situation previously described for the endosymbiotic Wolbachia bacteria, a guard against viral infections.

Temperature is definitely an important parameter and bacteria appear to be somewhat more virulent at 25°C.This point is now addressed in the Material and Methods section (lines 675-679) and in Fig. S6I.
As regards the influence of the composition of the infection solution, it does not seem to be a critical parameter that we have described in the context of continuous feeding on the bacterial solution (Limmer et al., PNAS, 2011).In preliminary experiments, we had tried LB or BHB medium to grow the bacteria and this did not make any difference (see Panel A below for LB [BHB used in our experiments]).As regards the sucrose concentration to the infection solution, we have tried two concentrations and did not observe any difference as regards the establishment of the latent infection.(see panel B below for 50 mM sucrose [100mM used in our experiments]).Of note, P. aeruginosa does not grow on sucrose solution alone.However, a latent infection was still established upon feeding the flies with PAO1 in sucrose alone, albeit likely with a mildly increased virulence, in the absence of any BHB medium (see panels C-D) below.-Figure 1D and other figures > CFUs given as Log2 are unconventional.One cannot easily deduce the burden unless e.g.translate 2e10 to ~1000 and 2e30 to ~10e9 CFUs.
True, but bacterial titer increases by a factor of two at each division cycle.Even though we have previously used a Log10 representation, we now prefer using a Log2 representation.This representation has also been used by other authors in the field, e.g., Duneau et al., eLife, 2019.
-Figire S1DB (now S1C)> "but from the outside of the gut".The given experiment does not prove that statement.
This issue has been already dealt with in the Nehme et al.PLoS Pathogens 2007 article, as cited in the manuscript.We further provide in Fig. S1B pictures documenting the presence of bacteria associated with visceral muscles.Finally, we also show that the gut lumen is essentially cleared of bacteria after a period of feeding on a sucrose solution or gentamicin.Hence, most bacterial colonies originate from the outer layer of the gut.We clarify the issue in the text (lines 154-158).
-Lines 146-7 > data are missing in support to the statement.
We have now added Fig. S1B to document that the gentamicin treatment does work, as actually does feeding on sucrose solution alone, as previously documented in Limmer et al., 2011 (Fig. S2B).Of note, we cannot exclude that a few bacteria remain, especially in the crop, but those would be at very low titer.Please, see also reply to Reviewer 2.
-Figure S1C > The effect of injury seems to be huge, and may account for much/most of the differences observed (including those between latent and active infection).This is further supported by Figure4A, injury may account for gut collapse and/or systemic stress.
It is well known that injury alone induces the systemic IMD pathway response 6 hours after injury but largely subsides by 24 hours.The point of Fig. S1C is that the level of induction reached during latent infection is very low as compared to that observed during a systemic infection, here obtained for reference with an Escherichia coli injection and to a lesser extent with a PBS injection.In our latent infection model, we do not perform any injury, except as noted by the reviewer in Fig. 4: the effects of an experimental injury are observed only while the bacteria are crossing the intestinal barrier and hardly any effect is observed when the injury is performed on day 10 (Fig. S4B).
The fluorescent flies were sorted out and their subsequent survival was monitored.As compared to nonfluorescent flies from the same batch, they died within two days of sorting them.
We thank the reviewer for pointing out this problem.The line is not the average but the median.
We have now added a precise description of the bars to all the figure legends.
-Lines 382-5 > what is the evidence of gut damage (or the absence of it)?How do the bacteria escape the gut?
The absence of major gut damages has been documented in Limmer et al, PNAS, 2011.How the bacteria escape the gut remains an open question (intracellular and/or paracellular route).
-Lines437-442 > The distinction between dormant P. aeruginosa in the fly tissues and persister cells (upon antibiotic treatment) cannot be justifies on the basis of relative bacterial numbers in the two systems.The extent of resistance to antibiotics though my serve that purpose.
In our latent infection model, most of the bacteria that have crossed the gut barrier become dormant and are associated with tissues, except at the beginning of the infection.In contrast, when a bacterial culture is treated with antibiotics, most of the bacteria are killed by the treatments and only a few ones persist, likely because of an inactive metabolism.Thus, the induction of dormancy in our latent infection model does not rely on the selection of a few metabolically-inactive bacteria able to withstand an immune response or an antibiotic treatment.

Significance
The study is a significant advance to our knowledge.
Notwithstanding further explanations, it provides a solid basis of understanding active versus dormant bacteria.It further establishes a mode of intestinal to systemic infection as a tool for further explorations.

Summary:
In this study, Chen and colleagues investigated a new latent infection model for Pseudomonas aeruginosa using Drosophila melanogaster as a host.First, the authors established a new model for latent Pseudomonas infection.The key feature of this model is the translocation of P. aeruginosa from the gut to the hemolymph and the colonization of fly tissues by the dormant bacteria.Bacteria that translocated from the gut appeared strikingly different in morphology and resistance to antibiotics compared to bacteria that were directly injected.Next, the authors suggest that melanization but not the Imd pathway or hemocytes are necessary to promote dormancy and colonization of fly tissues.Finally, flies with latent P. aeruginosa infection exhibit improved survival after secondary infections in a melanisation-dependent manner.The study reports an interesting model for latent infection, provides insights into the host factors promoting latency and describes some of the consequences of such latent infection for the host.However, some of the conclusions are not fully supported by the data and need further experimental evidence.
Major comments: 1.The latent infection model requires some clarifications.First, temperature.Could the authors explain why they used 18 {degree sign}C and could low temperature contribute to the establishment of dormancy?
As shown in Fig. S6I, the latent infection model is less compelling at 25°C in terms of survival curves, which may reflect an increased rate of spontaneous reactivation of the virulence, a phenomenon we have not studied at 25°C.In another manuscript in preparation (Lin et al.), we actually show that a small heat shock does contribute to reactivation of the bacteria, an issue that is outside of the scope of the present study.Please, see also reply to reviewer 1.
Second, the use of gentamycin.How does gentamicin affect PAO1 outside the gut?From Fig. 1C It looks like the cfus in the hemolymph diminished rapidly after gentamicin treatment (around day 3), suggesting the potential effect of the antibiotic.Once the bacteria have crossed the gut and entered the hemolymph, they could still be affected by feeding flies the antibiotic.Is there a possibility that gentamicin treatment is a stress factor that could trigger or facilitate the transition to dormancy?The authors could test this experimentally either by omitting the antibiotic and assessing dormancy or by feeding injected flies with gentamycin and scoring dormancy.
We had actually tested the issue about the potential role of gentamicin outside of the gut compartment.We have thus fed flies on different concentrations of gentamicin and monitored the survival of those flies to the injection of PAO1 bacteria (please, see Figure below).When flies were feeding on the highest concentration of gentamicin tested, 32 mg/mL, they were succumbing fast to the PAO1 challenge, but not as fast as nontreated positive control PAO1 injected flies.The use of lower concentrations (16, 8, and 4 mg/mL) led to a progressively stronger protection from PAO1 injection that inversely correlated with the gentamicin dose.We interpret the data with high gentamicin concentrations as an indication that gentamicin at such concentrations is likely directly toxic to the flies, an issue that could be experimentally tested but is not relevant to this study.Interestingly, lower doses led to a much-decreased protection from PAO1 (2mg/mL) to no protection at the dose we use to establish latent infection (100 µg/mL).Thus, these data show that gentamicin can pass the gut barrier when provided at high concentrations, down to 2 mg/mL.However, there is no proof of such a passage at the dose we use.In keeping with this latter possibility, we made a control experiment in which the gentamicin treatment step was replaced by simply feeding on the sucrose solution: the bacterial titer decreased in the hemolymph at the same rate as for gentamicin-treated flies (new Fig. S1F), demonstrating that ingested gentamicin does not contribute to the decreased titer.Rather, the likely depletion of the "source", that is PAO1 in the gut lumen, best accounts for this phenomenon.
We have now cited references which document a lack of permeability of the gut barrier to ingested gentamicin in vertebrate animals (lines 130-133).
As regards the possibility that gentamicin acts as a stress factor on bacteria, our data do not support this possibility, as a latent infection is established in the absence of gentamicin by just feeding the flies on a sucrose solution.We had previously reported that flies fed with P. aeruginosa for up to three days do not succumb within the next two weeks when they are fed on a sterile sucrose solution after having ingested the bacterial solution (Limmer et al., PNAS, 2011; Fig. 1C).Under the conditions of two days of PAO1 ingestion, we document in novel Fig. S1G that the carcass is equally well colonized under these conditions.

Figure: impact of gentamicin ingestion at diverse concentrations on the survival of injected PAO1 bacteria.
The ingested antibiotics can act on bacteria present in the hemocoel at concentrations over 2 mg/mL and not at that used in our experiments (100 µg/mL).
2. Does melanization really induce the dormant state of the bacteria?I am not sure the provided data fully support this claim.Addressing these questions might provide a stronger evidence: Fig. 2 A-F: What causes the morphological changes of the bacteria?Melanization or the passage through the gut?Do authors see the same changes in bacteria retrieved from PO-deficient mutant flies?Fig. 2G: Do the authors see the same resistance of PAO1 that colonized PO mutant flies to antibiotics?
In a novel Fig. S4, we now document comprehensively the physiological state of PAO1 bacteria fed to PO-deficient flies.We find that these bacteria are susceptible to antibiotics treatment as they can be rescued from death by the injection of antibiotics on day 3 (Fig. S4A-B).Second, they show a mixed phenotype in terms of colony morphology (Fig. S4C) and bacterial morphology and cell wall properties: even though most bacteria appeared to have kept a rounded morphology, they predominantly (about 75%) expressed the O5-LPS antigen.We interpret these data in terms of a slower transition to virulence than in a septic injury model.Melanization thus strongly contributes to the establishment of latency, even though it is likely that other factors contribute to the establishment of dormancy, but at best provide a minor contribution.
How do PO mutants behave after PAO1 injection?Are they similarly more susceptible?
PPO1-PPO2 mutants are not more susceptible to PAO1 injection than wt controls (new Fig. S3C).
3. Fig. 3F: PPO1 is believed to be the fast-acting PPO, whereas PPO2 is deployed later in infection.
This statement is based on experimental data gained on larvae, not adults.It is not really clear whether the about 10% adult hemocytes that express PPO2 actually contain crystals, in as much as the adult may be better oxygenated than larvae that grow in a hypoxic environment (description by the laboratory of Prof. Jiwon Shim of a role for PO in respiration at the latest EDRC meeting).This experiment is not needed given the explanation provided above.
This observation is difficult to account for and the explanations we can put forward at this stage are somewhat speculative.It appears that bacteria found in the tissues in PO-deficient flies have a morphology found in in vitro culture and within the gut, which does not correlate with virulence but also not with the avirulence state since they are LPS O5 positive.Given the shallow survival curves, we envision that there is a progressive release of bacteria from the tissue and then quick proliferation in the hemolymph in a few flies that would then die, but at a frequency too low to reliably ascertain in our hemolymph titer data, with a few flies displaying a high titer (Fig. 3D).By day5, the decreased titer in the carcass may reflect the progressive depletion of tissueassociated bacteria as they progressively become planktonic.

A)
Comparison of LB vs. BHB B) Establishment of latent infection with 50mM sucrose C) Establishment of a latent infection with a sucrose-only bacterial solution D) Colonization of host tissues by PAO1 ingested in a sucrose-only bacterial solution Minor issues: -Lines 579-581> How were the PAO1GFP/RFP constructed (details are needed) Done; please, see lines 641-643.
How does the Western blot look for PPO1?Will it show an early induction of melanization that could drive the change into the dormant state?We provide below a characterization of the PPO antibody we use by Western blot analysis.This antibody had originally been raised by the late Dr. Hans-Michael Müller against a PPO from mosquito cell lines, hence explaining its cross-reaction to both Drosophila PPO1 and PPO2 (Muller, H.M., Dimopoulos, G., Blass, C., and Kafatos, F.C. (1999)).A hemocyte-like cell line established from the malaria vector Anopheles gambiae expresses six prophenoloxidase genes.J BiolChem 274, 11727-11735.).It follows that at least one PO is partially cleaved at day 2 and that both are fully cleaved by day6 of the establishment of the latent infection (Fig.3F, Fig.S3F).characterization of the antibody raised against A. gambiae PPO Alternatively, the induction of melanization could also be measured with an L-DOPA test.