Mycobacterium abscessus resists the innate cellular response by surviving cell lysis of infected phagocytes

Mycobacterium abscessus is the most pathogenic species among the predominantly saprophytic fast-growing mycobacteria. This opportunistic human pathogen causes severe infections that are difficult to eradicate. Its ability to survive within the host was described mainly with the rough (R) form of M. abscessus, which is lethal in several animal models. This R form is not present at the very beginning of the disease but appears during the progression and the exacerbation of the mycobacterial infection, by transition from a smooth (S) form. However, we do not know how the S form of M. abscessus colonizes and infects the host to then multiply and cause the disease. In this work, we were able to show the hypersensitivity of fruit flies, Drosophila melanogaster, to intrathoracic infections by the S and R forms of M. abscessus. This allowed us to unravel how the S form resists the innate immune response developed by the fly, both the antimicrobial peptides- and cellular-dependent immune responses. We demonstrate that intracellular M. abscessus was not killed within the infected phagocytic cells, by resisting lysis and caspase-dependent apoptotic cell death of Drosophila infected phagocytes. In mice, in a similar manner, intra-macrophage M. abscessus was not killed when M. abscessus-infected macrophages were lysed by autologous natural killer cells. These results demonstrate the propensity of the S form of M. abscessus to resist the host’s innate responses to colonize and multiply within the host.


Introduction
Drosophila is a well-established organism model for studying pathophysiology of bacterial infections, such as those with Listeria monocytogenes and Staphylococcus aureus [22,23]. Its genetic tractability makes it one of the best models to combine functional genetics with immunity. Indeed, to note, Toll-like receptors were discovered in Drosophila [24]. In the context of mycobacterial infections, Drosophila has mainly been used to model tuberculosis, with M. marinum infection [25]. As M. tuberculosis in humans, M. marinum causes a wasting in Drosophila [26], associated with a metabolic switch [27]. Drosophila has allowed to highlight the crucial role of the STAT-ATG2 pathway in the control of mycobacterial infections by macrophages [28]. Few studies have been conducted in Drosophila with M. abscessus, mainly to test the efficacy of some drug combinations against M. abscessus [21].
The availability of a model of susceptibility to infection by the S form has allowed us to study the propensity of M. abscessus to resist the protective innate responses of the infected host. Indeed, we set up a modified Drosophila infection model by administering S M. abscessus intrathoracically with full control of the injected inoculum. After systemic injection, M. abscessus was rapidly internalized by phagocytes, allowing it to avoid the antimicrobial peptide response, and with intracellular growth before spreading into the circulation. Our results highlight that M. abscessus resists the lysis and caspase-dependent apoptotic cell death of Drosophila infected phagocytes, which was further confirmed in mice, with murine M. abscessusinfected macrophages lysed by autologous natural killer cells.

M. abscessus is more pathogenic for Drosophila compared to some other RGM and SGM
Thoracic nano-injections were performed with different doses of S M. abscessus. All flies injected with 1,000 colony-forming-units (CFU) of S M. abscessus died after 6 days post-infection (p.i.). A nearly complete absence of fly death was observed when S M. abscessus heat-or PFA-killed were injected (Fig 1A). At the same time point (6 days p.i.), 50% of lethality was observed for the lowest dose (10 CFU) (Fig 1B), compared to 87% of flies infected with 100 CFU, showing that death of infected flies was concentration dependent.
We next infected flies with 1,000 CFU of S and R M. abscessus. 50% of R M. abscessus infected-flies died on day 3 p.i. whereas the median survival of S M. abscessus infected flies was delayed to day 5 p.i. (S1 Fig), suggesting that R M. abscessus is more virulent than S M. abscessus in Drosophila.
Comparison of fly survival after infection with other mycobacteria at the 10 CFU dose confirmed the acute virulence of S M. abscessus in Drosophila (Fig 1C), with delayed death when flies were infected by M. marinum, a strict pathogenic SGM for humans, or absence of death when flies were infected by M. smegmatis, a saprophytic RGM (Fig 1C).
We also compared the survival of flies infected with subspecies of M. abscessus and with M. chelonae (Fig 1D). M. chelonae was the most virulent, killing all infected flies on day 10 p.i. whereas M. abscessus subsp bolletii was the least "pathogenic", killing only half of the population at the same time point. M. abscessus subsp massiliense behaved in the same way as S M. abscessus ( Fig 1D).
Finally, we infected flies with mutant strains of M. abscessus that we have isolated or generated and validated in our previous works for their attenuation in intracellular growth in cellular and/or zebrafish models [9,29,30]. Thus, the mmpL8_ MAB mutant is characterized by an impaired adhesion to macrophages, a decreased intracellular viability, a delay in making cytosol/phagosome contact and an attenuated virulence in zebrafish [29]. The MAB_4532c is also strongly impaired in intracellular viability and is unable to induce phagosomal membrane damage and to prevent reactive oxygen species (ROS) production by macrophages [30]. Injection of each of these mutants caused a lower mortality as compared to wild-type M. abscessus infection (Fig 1E and 1F). We also infected flies with three M. massiliense 43S mutants strains, known to have a transposon (Tn) insertion in a gene of the ESX-4 locus and to be impaired in

PLOS PATHOGENS
Mycobacterium abscessus infection and cellular response intracellular growth [9]. Two mutants correspond to Tn insertions in eccC and eccE genes, encoding for two ESX-4 structural proteins, and the third mutant correspond to a Tn insertion in the gene encoding for the ESX-secreted protein espI. All the 3 Tn mutants were less virulent than the control (S1B Fig).
Altogether, these results confirmed the sensitivity of Drosophila to S M. abscessus infection, which is related to bacterial virulence, as demonstrated by the behavior of M. abscessus mutants.
This has enabled us to investigate how S M. abscessus resists the host innate response, allowing it to colonize and ultimately trigger an infection which kills the flies.

M. abscessus infection induces an inefficient antimicrobial peptide-based response in Drosophila
Drosophila innate immune response relies on both acellular and cellular responses. The acellular response, named humoral by drosophilists, is mainly based on the production of antimicrobial peptides (AMPs) and mediated by two conserved NFkB pathways, Toll and immune deficiency (Imd). The cellular response relies on immune blood cells, called hemocytes, among which more than 90% are macrophages (plasmatocytes) [31].
First, we studied whether S M. abscessus infection induced the Imd-and/or Toll-regulated Drosophila humoral responses, by quantifying the transcription of the main AMPsencoding genes, during the course of infection. Injection of 10 CFU of S M. abscessus resulted in increased levels of the Imd-regulated Attacin-A and Diptericin transcripts, and of the Toll-regulated Metchnikowin, as compared to controls (Fig 2A). This induction peaked on day 3 p.i. Injection of 1,000 CFU of S M. abscessus resulted in higher transcript levels for almost all Imd-and Toll-regulated genes ( Fig 2B). As a control, the wound resulting from nano-injection of water did not induce any expression of AMP-encoding genes on days 0 and 3 (S2A Fig).
We then infected AMP deficient flies generated by CRISPR/Cas9 gene editing technology [32]. These flies were defective in genes encoding AMP regulated by either the Imd pathway (group B:AttC Mi ,Dro-AttA-B SK2 ,DptA-B Ski ;AttD Ski ), or the Toll pathway (group C: Mtk R1 ;Drs R1 and Bomanins: Bom Δ55C ) or both (group A: Def SK3 ). These mutant flies were no more sensitive than wild-type controls to water injection (S2B Fig). Mutant flies for Relish (Rel E20 ) and spatzle (spz rm7 ) genes were used as control for the Imd and the Toll pathway respectively. Mutant flies were infected with low and high doses of S M. abscessus (10 and 1,000 CFU). We did not observe differences in terms of fly survival regardless of the AMP pathway impacted, with an equivalent mortality for all mutated flies (Fig 2C and 2D).
Similar experiments with the same altered flies, but this time performed with the attenuated M. abscessus mutants, also showed no difference in fly survival, regardless of the AMP pathway impacted (Fig 2E and 2F). Comparatively, the same AMPs mutant flies died when infected with 10 CFU of B. cepacia, a Gram-negative bacterium (S2C Fig).
Taken together these results show that S M. abscessus infection induces expression of AMPs-encoding genes. We also show that the absence of AMPs does not modify the mortality of the flies, as opposed to what was observed with B. cepacia, suggesting that they do not play a major role in the resistance to infection by S M. abscessus.

The cellular response of Drosophila to S M. abscessus infection is critical for fly survival
The cellular response relies on immune blood cells, called hemocytes. Plasmatocytes represent the majority of total hemocyte population in flies, and thus the cellular part of the protective innate response. To assess whether M. abscessus was internalized by phagocytic plasmatocytes during the course of infection, we used 500 CFU of Td-Tomato fluorescent S M. abscessus to infect reporter flies with GFP-producing plasmatocytes (hml>GFP) [33,34]. Red fluorescent mycobacteria were observed inside GFP-producing plasmatocytes as early as 30 min. p.i. (Fig  3A), and up to 24 hours p.i. Increased red fluorescence, observed up to 4 days p.i. (Fig 3A), indicates that M. abscessus survives and potentially grows inside the GFP-producing plasmatocytes. Moreover, by looking at the whole fly, up to 5 days p.i., we observe that S M. abscessus represent the standard deviations. Data were analyzed using two-way analysis of variance (ANOVA) (*p<0.05; **p<0.005; ***p<0.0005) in (A-B). Survival was analyzed on 120 flies per genotype in (C) and for 60 flies per genotype in (D-F). Data were analyzed using the log-rank test (*p = 0.0475).
https://doi.org/10.1371/journal.ppat.1011257.g002  The progressive effect of the S M. abscessus infection, with a potential dissemination only at the 4th -5th day, might indicate a protective role of plasmatocytes in the control of the infection, at least at its beginning. To test this hypothesis, we treated flies with clodronate containing liposomes. This treatment, depleting phagocytic plasmatocytes [35] (S4 Fig), led to a very high mortality rate, with all clodronate pre-injected flies dying on day 6 p.i. as compared to control flies (Fig 3B), and was associated with a significant increase in S M. abscessus growth (Fig 3C).
To confirm the critical role of phagocytic plasmatocytes on the infection control, we used the UAS/GAL4 system to perform genetic depletion of these immune cells, as previously described [36,37]. Indeed, the UAS/GAL4 system allows the spatial and temporal control of transgene expression. It is based on the use of the yeast transcription factor, Gal4. The binding of this transcriptional activator on a minimal regulatory sequence, called UAS (Upstream Activating Sequence) drives the expression of a sequence located downstream of this UAS sequence [38]. Here, we used transgenic UAS-debcl flies, allowing a Gal4-dependent expression of a proapoptotic gene, in order to kill the cellular populations of interest. To drive its expression in plasmatocytes, we crossed UAS-debcl flies with transgenic Hemese (He)-GAL4 or croquemort (crq)-GAL4 plasmatocytes driver lines. He and crq correspond to transcriptional enhancer controlling expression of plasmatocytes markers and so the expression of the GAL4 gene in these driver lines. Therefore, flies carrying both transgenes (driver and UAS-debcl) express debcl pro-apoptotic gene in their plasmatocytes.
Flies expressing debcl receiving a water injection remained alive throughout the experiments ( Fig 4A) and was confirmed by the increased bacterial load on day 3 p.i. for the crq>debcl fly genotype ( Fig 4B).
Taken together these results demonstrate that phagocytic plasmatocytes rapidly internalized S. M. abscessus. Their absence makes the flies hyper-susceptible to infection, indicating that these cells play a protective role against S M. abscessus, at least during the first days of infection.

S M. abscessus infection is favored by a deleterious Drosophila immune cell population
Surprisingly, depletion of Hemese (He) expressing cells conferred an opposite phenotype to crq>debcl flies, with increased resistance of the He>debcl flies to S M. abscessus infection and a better control of mycobacterial growth (Fig 4C and 4D).
Several recent works [39][40][41][42] tend to indicate that the population of plasmatocytes is more diverse than expected. At least six major populations have recently been defined within larval plasmatocytes [43]. Based on our two opposite phenotypes, we investigated which gene markers were differentially expressed within these different sub-populations. Hemese, Tep4 and ance genes have been reported to be more highly expressed than crq gene in a new class of plasmatocytes called thanacytes (as we will refer to them throughout the text for a better understanding) at the larval stage [39]. Interestingly, this population might correspond to the secretory Plasmatocytes, very recently described at the pupal stage, strongly suggesting a persistence of thanacytes in adult flies [44]. We thus confirmed the presence of these cells in adult flies by confocal microscopy by observing Tep4>GFP and ance>GFP positive hemocytes (S6A Fig).
By using the same UAS/GAL4 approach, we looked more specifically at whether depleting thanacytes, by driving UAS-debcl by ance-GAL4, resulted in a resistance to infection by S M. abscessus, as observed when we depleted for He-positive plasmatocytes. Firstly, ance>debcl

PLOS PATHOGENS
Mycobacterium abscessus infection and cellular response flies showed an increased survival as compared to controls and CFU counts were not significantly different (Fig 5A and 5B). Similar results were observed when using the TARGET system [45], allowing debcl expression only at the adult stage (S6B Fig). In this system, Gal4 activity is inhibited by Gal80 and this inhibition is conditionally restricted to the period starting from the development to the beginning of the adult stage. Finally, no significant difference was observed in AMPs encoding gene expression level between M. abscessus-infected ance> and M. abscessus-infected ance>debcl flies (S6C Fig), suggesting that the increased survival of ance>debcl flies was not due to an increased AMPs expression. Moreover, there was no difference in survival between control and ance>debcl flies infected with the avirulent RGM M. smegmatis (S6D Fig).
Taking all of these results together, we confirm the existence of a cell population, potentially the thanacytes, as defined by the drivers used, which, when depleted, protects the flies from S M. abscessus infection. Since this improved protection cannot be due to AMPs produced during infection, because of the lack of preferential induction of AMPs in the mutated fly compared to the wild type, we wondered what could be their function that makes them deleterious in S M. abscessus infected flies.
Thanacytes are defined by a high expression of two serine protease encoding genes, CG30088 and CG30090, which have been proposed to be respectively homologous to granzyme B and granzyme H encoding genes, expressed by mammalian Natural Killer (NK) cells and cytotoxic T CD8 + lymphocytes [39]. We thus hypothesized that thanacytes, through the expression of CG30088 and CG30090, could lyse infected phagocytic cells without killing the intracellular M. abscessus, and then, promote M. abscessus spreading.
We individually depleted CG30088 and CG30090 transcripts in thanacytes by RNA interference (RNAi), by crossing UAS-RNAi lines with thanacytes-GAL4 drivers. We obsreved an increased survival of ance>CG30088-RNAi and ance>CG30090-RNAi flies as compared to wild-type flies, with 85% and 80% flies still alive on 7-day p.i. respectively (Fig 5C). Similar results were obtained when the transcripts were only depleted in adult thanacytes (S6B Fig). An increased survival was also observed for Tep4>CG30090-RNAi compared to control flies ( Fig 5D). As a control, we validated the efficiency of these RNAi lines by qRT-PCR. Interestingly, we observed an increase of the quantity of both transcripts upon infection and that both RNAi significantly reduce these increases (S6E and S6F Fig). Collectively, these results, although indirect because they rely on the level of reduction of the cellular serine protease gene expression, still indicate that the thanacytes, through the production of CG30088 and CG30090, are deleterious for Drosophila survival during S M. abscessus infection.
We next assessed whether depletions of thanacytes' products CG30088 and CG30090 could also confer a protection to flies infected with other bacteria. We observed this protective for infection with a strict pathogenic and slow-growing non-tuberculous mycobacterium, M. marinum, as compared to another fast-growing mycobacterium M. chelonae and an extracellular Gram-negative bacterium B. cepacia (S7A- S7C Fig). These results suggest that S M. abscessus and M. marinum share a virulence trait linked to a an intrinsic resistance to Drosophila serineprotease response, absent in less or non-pathogenic mycobacteria such as M. chelonae and M. smegmatis.

M. abscessus infection leads to a caspase-dependent apoptosis of fly infected phagocytes
We have produced two essential results leading to the same resistance phenotype when thanacytes are depleted or CG30090/CG30088 transcripts are depleted. Our working hypothesis was that thanacytes, through the expression of CG30088 and CG30090, could lyse infected phagocytic cells without killing the intracellular M. abscessus, and then, promote M. abscessus spreading. We then tested whether plasmatocytes infected by S. M. abscessus could be lysed.
To do so, we measured the transcript levels of crq and nine additional genes (Mmp2, NimC2, CAH7, Robo2, Mbc, NimB4, NimB5, Nplp2, eater) described as highly expressed in plasmatocytes with phagocytic activity [39,42]. Their abundance would indirectly reflect phagocyte numbers in ance> and ance>CG30090-RNAi flies. On day 4 p.i., relative expression levels of the ten genes were higher in infected ance>CG30090-RNAi flies compared to infected ance> (Fig 6A), relatively to uninfected flies of the two genotypes. These results suggest that the increased expression of these markers during infection, upon serine protease depletion, would be related to a larger phagocytic population, maintained during this infection, unlike the control flies where this expression is reduced. The reduction in the infected phagocytic population might be through caspase-dependent apoptosis [46], with regards to the property conferred to thanacytes via the serine-protease activity [39].

Mycobacterium abscessus infection and cellular response
We thus collected infected or uninfected ance> and ance>CG30090-RNAi flies' hemocytes, extracting their proteins and performing a quantification of caspase activity using a synthetic substrate containing a caspase-3 cleavage site conjugated to a fluorochrome as described in [47]. Protein extracts of larval wing disc of vg>Rbf genotype, known to be apoptotic [48], were used as a positive control. A significant increase of caspase activity was observed in M. abscessus-infected ance> extracts compared to non-infected ones (Fig 6B). Interestingly, this caspase activity significantly decreased when the CG30090 serine protease transcripts were depleted (Figs 6B and S8), showing that the observed caspase activation of infected plasmatocytes is dependent on the products of CG30090.
We then inhibited apoptosis in phagocytes by inhibiting caspase activity. We hypothesized that this should increase Drosophila survival and phenocopy thanacytes and CG30090 transcripts depletions. To test this, we inhibited caspase activation in the adult crq-expressing phagocytic plasmatocytes, by expressing in the latter a transgene encoding the caspase inhibitor baculovirus protein p35. Expression of p35 in adult phagocytic plasmatocytes significantly increased fly survival compared to control (Fig 6C).
With all of these results taken together, we propose that M. abscessus infection leads to caspase-dependent apoptosis of the infected phagocytes, leading to the progressive depletion of these latter by thanacytes. This might explain the resistance phenotype of thanacytes or serineproteases depleted flies, by the maintenance of the phagocytic cell reservoir, the best able to control S M. abscessus infection. We now have to confirm whether this peculiar trait was also observed in a mammalian host.

Intracellular M. abscessus resists lysis of murine macrophages by autologous NK cells
The observed behavior of S M. abscessus or M. marinum, as compared to M. smegmatis, in the fly, led us to evaluate whether, like another slow-growing pathogenic mycobacterium, M. tuberculosis, we might find the same phenotype of resistance to NK lysis described for M. tuberculosis [49]. In fact, the CD8 + or NK cytotoxic response in human tuberculosis has been shown to be involved in controlling M. tuberculosis [49]. This response follows two pathways in humans, the granzyme and the perforin-granulysin pathways. M. tuberculosis, a strict pathogen of humans, is resistant to the granzyme-mediated CD8 + and NK cytotoxic response [50]. The observation that the opportunistic S M. abscessus behaves in a similar way in Drosophila obliges us to test this hypothesis of S M. abscessus resistance to the murine NK response.
To assess whether S M. abscessus might resist to the lysis of infected macrophages induced by NK cells, we performed co-cultures of purified mouse primary NK cells and autologous macrophages. These latter were infected with M. abscessus at a MOI of 1:10; then, NK cells were added or not 4 hours p.i. On day 2 p.i., macrophages survival was not decreased by S M. abscessus infection and even seem to be improved (Fig 7A upper panel and 7B). NK cells addition to non-infected macrophages decreased their survival (Fig 7A lower panel and 7B). This decrease was enhanced when macrophages were infected with S M. abscessus (Fig 7A lower  panel and 7B), reinforcing the critical role of NK cells during infection. Importantly, the bacterial load of intracellular S M. abscessus was similar both in the presence or absence of NK cells despite the drastic decreased survival of infected macrophages (Fig 7C), corroborating our results in Drosophila. Overall, these results show that NK cells can kill S M. abscessus-infected macrophages, whereas the intracellular mycobacteria are resistant to this lysis.

Discussion
Over the last decades, in part due to the relative conservation of molecular and genetic pathways of innate immunity with mammals, D. melanogaster has emerged as a good model among the non-mammalian hosts for studying interactions between host and intracellular pathogens [51,52]. Thus, RNA interference (RNAi) screens, performed on S2 cells (embryonic derived macrophage-like cells), have allowed the dissection of host factors involved in Legionella pneumophila and Listeria monocytogenes invasion and intracellular replication [53][54][55] and those required for the entry and survival of M. fortuitum [56] and M. smegmatis [57]. Nevertheless, M. marinum, a strict pathogenic mycobacterium, remains the most studied species in vivo in Drosophila [58]. In adult flies, this bacterium proliferates within phagocytic plasmatocytes then spreads systemically leading to death [25].
Despite being a RGM, S M. abscessus survived and proliferated within phagocytic plasmatocytes, similarly to the strict pathogenic SGM M. marinum. The ability of M. abscessus to survive in protozoa including amoeba, even after encystment [59], provides it with advantages [60] that might reflect defense mechanisms acquired by the bacterium in contact with these

PLOS PATHOGENS
Mycobacterium abscessus infection and cellular response predators, allowing it to survive intracellularly. Moreover, mutant strains unable to proliferate in phagocytic cells were less virulent in Drosophila, validating the results obtained in environmental phagocytes (amoebae) and mammalian macrophages [9,30], in which an intracellular growth deficit affects the survival of the M. abscessus within its host [9,29,30].
D. melanogaster has the undeniable additional advantage of allowing the analysis of the host's innate immune response [61]. Unlike infections with M. marinum or M. smegmatis in which no AMP production was reported [25], two studies have shown that Drosophila infection with S M. abscessus induces the expression of some AMPs, a Toll related AMP-encoding gene [20,62]. We have confirmed and extended this observation by quantifying the expression of the main AMP-encoding gene transcripts, resulting in the induction of almost all Toll but also Imd-related AMP-encoding genes. Strikingly, when we tested whether the induced AMPs were necessary to control S M. abscessus, we found that their absence, using Drosophila strains lacking AMPs, did not increase fly susceptibility to the infection. Non-mutually exclusive interpretations of these results are either that the cellular response prevails over humoral responses to control S M. abscessus infection or that the intracellular survival of mycobacteria after internalization by phagocytic cells protects them from the AMPs.
We reveal that M. abscessus was found in Drosophila macrophages (plasmatocytes), where it appears to multiply, although we do not exclude the possibility that plasmatocytes re-internalize extracellular bacteria. M. abscessus knock-out mutants, impacted in their intracellular survival, are also affected in their ability to grow in Drosophila, leading to a parallel between phagocytosis resistance in humans and in Drosophila. The importance of phagocytic plasmatocytes is demonstrated by the increased susceptibility of flies observed after either clodronate-mediated depletion of phagocytic cells or genetically mediated depletion of crq-expressing plasmatocytes. This population can be considered as the main immune cell type controlling the infection during the first days post-infection but also as an intracellular reservoir for S M. abscessus.
We have highlighted another sub-population of plasmatocytes, expressing He, ance, Tep4, CG30088 and CG30090, that is detrimental for fly survival. Due to this expression profile, they are presumably the subtype recently identified by single cell sequencing of Drosophila larval hemocytes called either thanacytes [39] or PL-Pcd/PL-AMP [42,43], which do not exhibit this capacity for phagocytosis [43] and even more recently described in pupae as Seceretory-PL [44]. Interestingly, Drosophila are protected from S M. abscessus infection when He-or anceexpressing cells were depleted by expressing the pro-apoptotic debcl gene. A similar protection was observed when CG30088 and CG30090 transcripts were depleted by RNAi in either anceor Tep4-expressing cells. Similar phenotypes were observed with M. marinum infection. However, no protection was observed in these flies when infected by M. chelonae, a closely related species to M. abscessus and B. cepacia, an extracellular bacterium. This highlights the behavior of M. abscessus, similar to strict pathogenic mycobacteria towards the host innate cellular response.
Thanacytes, through CG30090 production, induce death of phagocytes through a caspasedependent apoptosis. However, this cellular response was not effective against the mycobacterium. We confirm this resistance of M. abscessus to phagocytes cytotoxic lysis in mice. We do not provide a complete mechanism here, but in view of the conserved role of granzymes, our results suggest that the NK-dependent killing of M. abscessus-infected macrophages may pass through induction of caspase-dependent apoptosis after formation of pores in the cell membrane. This mechanism is consistent with what is observed with M. tuberculosis infection of dendritic cells. Indeed, NK cells lysed M. tuberculosis-infected phagocytes. Likewise, they also lysed non-infected activated phagocytes (called bystander cells) depleting the mycobacterial cellular reservoir and thus counterbalancing the inflammatory response at the expense of the host [49,50,63].
Our results support the notion that in both Drosophila and murine primary cells, intracellular M. abscessus resists the host innate cellular response by resisting macrophages' death. This observation is consistent with pioneering reports on the understanding of T cell cytotoxic responses against M. tuberculosis. Indeed, human NK cells (defined at the time as double negative lymphocyte for CD4 and CD8) killed M. tuberculosis-infected macrophages without affecting the viability of the intracellular mycobacteria. This lysis of macrophages by NK cells favors tolerance to the infection, by depleting the mycobacterial cellular reservoir and therefore, reducing the inflammatory response [49]. Likewise, CD8 T cells lysed infected macrophages, but also killed intracellular mycobacteria during this process contrarily to NK cells [49,50]. This M. tuberculosis destruction during macrophage cytotoxic lysis by CD8 T cells constitutes a mechanism of bacterial control and thus, a protection [49,50,64]. These opposing effects of macrophage lysis by NKs and CD8s on mycobacterial viability promote a balance between the inflammatory response and tolerance to the infection [63,65,66].
To summarize, Drosophila infection with S M. abscessus has allowed us to demonstrate how this mycobacterium escapes the host innate response. Similar to strict human and animal pathogenic SGM, such as M. tuberculosis or M. marinum, S M. abscessus is internalized by phagocytes, within which it survives and seems to replicate. These host cells might constitute a shield from the AMPs response to which M. abscessus is resistant. Nevertheless, both in Drosophila and mice, the resistance of M. abscessus to the killing of the phagocytes might constitute a mode for mycobacterial spreading and, at least, it might result in a severe depletion of the main cell reservoir able to control M. abscessus replication. This propensity of M. abscessus to resist the host innate immune response, typical of strict pathogenic SGM, might partially explain its superior pathogenicity among RGM that are predominantly saprophytic.

Materials and methods
Key materials used in this study are listed in the Table 1.

Experimental details
Bacterial strains and cultures. All mycobacterial strains were grown at 37˚C, except M. marinum (28˚C), and M. chelonae (30˚C), in Middelbrook 7H9 medium (Sigma-Aldrich, Saint-Louis, USA) supplemented with 1% glucose and glycerol 0.2% under aerobic condition until an OD 600 between 0.6 and 0.8. B. cepacia was cultured in standard Luria-Bertani (LB) medium. Bacterial cultures were then centrifuged to obtain concentrated aliquots, which were frozen at -80˚C in 10% glycerol.
Drosophila maintenance, crosses and infection. The flies were raised on a standard corn agar medium at 25˚C. Crosses were performed at 25˚C. An exception is for the TARGET experiments, for which 18˚C was used until pupal eclosion and a shift in adults at 29˚C. The UAS-GAL4 system [38] was used to express transgenes. For infections, frozen bacterial aliquots were thawed on ice and homogenized using a 30-gauge insulin needle (Becton-Dickinson, France) to avoid clumps. Serial 10-fold-dilutions were done and 30 μL of each dilution was spread on a blood agar plate for mycobacteria (COS, bioMérieux, France) or on a classic LB agar plate for B. cepacia. Plates were then stored at 28˚C or 37˚C for 2 or 3-5 days depending on the bacteria and colony forming unit (CFU) counts were determined.
The bacterial inoculum was diluted with water to obtain a suitable concentration. 5-7 days old virgin female flies were anesthetized with CO 2 (Inject-Matic, Switzerland), and were infected with 50 nL of the suspension containing 10, 100 or 1,000 bacteria by injection into the sternopleural suture. Infections were performed using a Nanoject III (Drummond Scientific Company, USA) nano-injector charged with a calibrated pulled glass needle made with a DrosDel iso w 1118 Dr B. Lemaitre [32] iso Bom Δ55C Dr B. Lemaitre [32] (Continued )

PLOS PATHOGENS
Mycobacterium abscessus infection and cellular response  To validate plasmatocyte depletion by clodrosome injection, water-and clodrosomeinjected crq> ds-Red flies were mounted in Washable Clear Glue (Elmer's) between a slide and a coverslip 24h after injection. Whole flies were imaged using an IX-83 microscope (Olympus) with a 10x objective. Acquisitions were performed using CellSens software and images were reconstituted in 3D using IMARIS software (Bitplane).
Plasmatocyte isolation and observation were performed on 5-7 days old female flies of hml-Gal4; UAS-eGFP genotype that were individually infected with 500 CFU of M. abscessus Td-Tomato. 6-8 flies were dissected as described previously [68]. Circulating and sessile hemocytes were collected in 100 μL of PBS containing 1% Hoechst (Invitrogen Hoechst 33342, USA) at 30 minutes, 24h, 48h, 72h and 96h p.i. 10 μL of the solution were placed between a glass slide and a coverslip and observed under a Leica SP8 X laser scanning confocal microscope. The acquired images were treated with ImageJ software (NIH).
qRT-PCR. Total RNA was extracted from 20 female flies per condition using TRIzol reagent (TRI reagent, ThermoFisher, Waltham, USA), chloroform, and isopropanol. Genomic DNA was removed from the extracted RNA using a Turbo DNA-free kit (Invitrogen AM1907, Invitrogen, USA), and cDNA was generated using Superscript III (Invitrogen 18080051, Invitrogen, CA, USA), following the manufacturer's instructions. qPCR was performed using Maxima SYBR Green Master Mix (ThermoFisher K0221, ThermoFisher, USA), 100 ng of cDNA as a template and 10 μM of target gene-specific primers. Primers used are listed in

PLOS PATHOGENS
Mycobacterium abscessus infection and cellular response Table 1. RpL32 transcript levels were used for normalization and the ΔΔct method was used for relative expression. Caspase activity assay. For each experimental condition, 30-50 female flies were dissected and hemocytes were collected as in [68] in 100 μl of chilled caspase assay lysis buffer (HEPES 50 mM (pH 7.5), NaCl 100 mM, EDTA 1mM, CHAPS 0.1%, sucrose 10%, DTT 5mM, Triton 0.5mM (X-100), glycerol 4%, Protease inhibitor cocktail 1x (cOmplete, Roche)). Hemocytes were collected in 1.5 mL microcentrifuge tubes. Proteins were extracted as previously described [47]. Briefly, hemocytes were homogenized with a handled pestle (5 strokes), lysed by freezing in liquid nitrogen, and rapidly thawed at room temperature 3 times. The lysates were centrifuged at 16,000 x g for 20 min. at 4˚C and the supernatant was transferred to a new tube. Protein concentrations were determined using the bicinchoninic acid method (Pierce BCA protein assay kit, ThermoFisher, Waltham, USA) following the manufacturer's instructions. The caspase activity assay was performed at 37˚C in a 96-well plate using 20 μg of protein per condition in 100 μM of the caspase-3 substrate Ac-DEVD-AFC (ALX-260-031, VWR, Radnor, USA) in a total volume of 100 μL, according to the manufacturer's instructions. Fluorescence was quantified over time using a spectrophotometer (Tecan Infinite M200, Life Sciences) with excitation at 385 nm and emission at 460 nm.

Murine immune cells isolation, infection and analysis.
Spleens from female C57Bl6 mice were collected and cell suspensions were prepared after a 20 min. treatment with Collagenase D (2 mg/mL, Roche Diagnostics, Switzerland) at 37˚C to release macrophages. After red blood cell lysis using Ammonium-Chloride-Potassium lysis buffer, CD11b + cells were isolated by positive magnetic selection, using anti-CD11b microbeads and an AutoMacs Pro Separator according to the manufacturer's instructions (Miltenyi Biotec, USA). Negative fractions were used to isolate NK cells after staining with FITC-conjugated anti-mouse NK1.1 (clone NKRP1A, BD Biosciences) and PE-conjugated anti-mouse CD3 (clone 145-2C11, BioLegend, USA) (15 min. at 4˚C) to exclude NKT cells. Pure NK cells were isolated using a BD Aria III cell sorter (BD-Biosciences, USA).
Biostatistical analysis. All data were analyzed using GraphPad Prism 9.0.0 (GraphPad Software Inc., USA). The log-rank (Mantel-Cox) test for Kaplan-Meier survival curves was used to evaluate the significance of survival statistics. Quantification of CFU and AMP transcript levels was compared by two-way ANOVA and caspase activity by one-way ANOVA. Comparisons of phagocytic plasmatocyte transcript levels were performed using a multiple Student's t-test. Statistical significance was set to 0.05.