Cytosolic Access of Mycobacterium tuberculosis: Critical Impact of Phagosomal Acidification Control and Demonstration of Occurrence In Vivo

Mycobacterium tuberculosis (Mtb) uses efficient strategies to evade the eradication by professional phagocytes, involving—as recently confirmed—escape from phagosomal confinement. While Mtb determinants, such as the ESX-1 type VII secretion system, that contribute to this phenomenon are known, the host cell factors governing this important biological process are yet unexplored. Using a newly developed flow-cytometric approach for Mtb, we show that macrophages expressing the phagosomal bivalent cation transporter Nramp-1, are much less susceptible to phagosomal rupture. Together with results from the use of the phagosome acidification inhibitor bafilomycin, we demonstrate that restriction of phagosomal acidification is a prerequisite for mycobacterial phagosomal rupture and cytosolic contact. Using different in vivo approaches including an enrichment and screen for tracking rare infected phagocytes carrying the CD45.1 hematopoietic allelic marker, we here provide first and unique evidence of M. tuberculosis-mediated phagosomal rupture in mouse spleen and lungs and in numerous phagocyte types. Our results, linking the ability of restriction of phagosome acidification to cytosolic access, provide an important conceptual advance for our knowledge on host processes targeted by Mtb evasion strategies.


Introduction
The pathogenic potential of Mycobacterium tuberculosis (Mtb), the etiologic agent of human tuberculosis (TB), depends largely on the type VII secretion system ESX-1 [1,2], which is responsible for the secretion of the 6-kDa Early Secreted Antigenic Target (ESAT-6), its protein partner, the 10-kDa Culture Filtrate Protein (CFP-10), and several ESX-1 associated proteins (Esps) [3,4]. ESX-1 secretion is evolutionary conserved in most members of the M. tuberculosis complex [5], and the more distantly related tubercle bacilli of the Mycobacterium canettii clade [6,7], as well as in some non-tuberculous mycobacteria such as Mycobacterium marinum [8]. This secretion system governs numerous aspects of interaction between pathogenic mycobacteria and the host cell [1,2], including membrane-damaging activity [9][10][11], thought to be implicated in phagosomal escape at later stages of infection [12][13][14][15][16]. Although this phenomenon is a matter of debate [2,[17][18][19][20], by use of a single-cell Fluorescence Resonance Energy Transfer (FRET)-based technology [21], we recently demonstrated that ESX-1-proficient Mtb and recombinant Mycobacterium bovis BCG::ESX-1 were able to induce phagosome rupture in human THP-1 macrophage (MF)-like cells [15]. This assay uses the ability of the surfaceexposed BlaC β-lactamase of Mtb [22,23] to cleave the FRET substrate CCF-4, which consists of a cephalosporin core linking 7-hydroxycoumarin to fluorescein that has also been used for exploring effector injection and intracellular localization of Gram-negative bacteria [21,24,25]. The ESX-1-induced rupture of the phagosomal membrane, which results in the exit of mycobacterial products from the endosomal pathway and in extra-phagosomal localization of bacilli [13][14][15][16] is of relevance for the outcome of the immune control and bacterial dissemination [26][27][28][29]. Phagosomes are reported to be specialized platforms for pathogen recognition [30] and there is also growing evidence of a link between the functionality of the ESX-1 secretion system and the presence of mycobacteria-associated molecular patterns in the host cytosol. Peptidoglycans [31,32] and extracellular mycobacterial DNA [33] were reported to be sensed by the cytosolic receptors of the innate system with multiple biological consequences. Indeed, the Mtb-mediated induction of Nucleotide binding Oligomerization Domain (NOD)-Like Receptor pathways, i.e., NOD2 / Receptor-interacting protein 2 kinase (Rip2) / TANK-Binding Kinase 1 (TBK1) / Interferon regulatory factor (Irf) 5, is responsible for a significant part of type I interferon (IFN) production [31,32]. On the other hand, the signaling through the Stimulator of IFN Genes (STING) / TBK1 / Irf3 pathway [33] leads to a type I IFN signature on which depends the expression of CCL5, CXCL10 and Nitric Oxide Synthase 2 [34,35]. The formation of Nucleotide-binding domain and Leucin-rich Repeat pyrin-containing Protein-3 (NLRP-3) / ASC (Apoptosis-associated Speck-like protein containing a carboxy-terminal CARD) / caspase-1 inflammasome complex, is required in humans for the processing of the pro-IL-1β into biologically active pleïotropic immune mediator IL-1β following Mtb infection [36,37]. Moreover, the ubiquitination of Mtb prior to its delivery to the autophagic machinery also necessitates the ESX-1-dependent translocation of extracellular Mtb DNA to the cytosol [16,33,38,39]. Thus, the events arising from mycobacterial cytoplasmic access may substantially influence both the immune control of Mtb and the inflammation-induced tissue damage.
The impact of selected components of the ESX-1 system on phagosomal rupture has recently been assessed [13,15,16], however, other potential intervening factors, including those from the host cell remain largely unexplored. Here, we have investigated the host parameters modulating the Mtb-mediated vacuolar breakage, by developing a CCF-4 FRET-based approach that can be used for the study of Mtb-infected cells by flow cytometry. This approach, which permits to combine the detection of phagosomal rupture with the analysis of numerous host cell phenotypic and functional parameters, allowed us to explore multiple phagocyte types, including those isolated from mouse airways. Our results provide first and unique evidence that Mtb-induced phagosomal rupture does occur in vivo inside the lungs and spleens of infected experimental animals and lasts over several days. Moreover, we here explore the impact of vacuolar acidification that constitutes a fundamental cellular defense mechanism [40] and demonstrate that the characteristic partial prevention of phagosomal acidification by Mtb is a prerequisite for phagosomal escape of the pathogen. Our study thus reveals novel details and presents a refined model of cellular events during infection with Mtb.

ESX-1-dependent Mtb-mediated phagosomal rupture detected by FRET-based flow cytometry
To evaluate mycobacteria-mediated phagosomal rupture in different phagocyte types and different physiological contexts, we adapted the previously used microscopy-based CCF-4 FRET technique [15] for flow cytometry. The latter approach not only allows monitoring of bacteriainduced phagosomal rupture or tracking of endosome-to-cytosol antigen translocation [25,41], but also permits the simultaneous inspection of surface markers and analysis of hundreds of thousands of host cells. At first, we infected differentiated THP-1 cells at a multiplicity of infection (MOI) of 1 either with Mtb H37Rv WT or the isogenic ΔESX-1 derivative, Mtb H37Rv-ΔRD1 [10], which both display similar β-lactamase activity [15]. These THP-1 cells were then incubated with CCF-4-AM, an esterified, lipophilic form of the CCF-4 substrate that can readily enter into cells, where it is converted by endogenous cytoplasmic esterases into negatively charged CCF-4, which is retained in the cytosol and emits green fluorescence (500-550 nm) upon stimulation at 320-380 nm, due to FRET from the coumarin moiety to the fluoroscein part (S1 Fig.). In the case of Mtb-induced phagosomal rupture, cleavage of CCF-4 by the intrinsic Mtb BlaC β-lactamase leads to loss of FRET and a change of the CCF-4 emission spectrum from green to blue coumarin fluorescence (410-470 nm). As depicted in Fig. 1A, the CCF-4 emission signals of CD11b + gated THP-1 cells, infected with wild-type (WT) Mtb H37Rv, showed a marked shift of the CCF-4 emission towards blue at 4 days post infection (dpi). In contrast, a much weaker shift of the CCF-4 spectrum was observed for Mtb ΔESX-1-infected cells, validating our experimental setup and confirming the fundamental virulence differences between the used ESX-1-proficient and ESX-1-deficient Mtb strains [10,15]. The residual blue shift in Mtb ΔESX-1-infected cells relative to non-infected cells is likely a consequence of paraformaldehyde (PFA) fixation prior to signal acquisition (S2A-B Fig.). These results were further corroborated by ratios of Mean Fluorescence Intensities (MFI) of blue vs. green signals (Fig. 1B), and blue MFI 447 nm (Fig. 1C). Moreover, we also used fluorescent Mtb (DsRed-Mtb H37Rv) to infect THP-1 cells, at a weaker initial dose (MOI = 0.3), and thereby observed that the CCF-4 blue emission shift selectively occurred in cells that had engulfed the bacteria (Fig. 1D). This approach thus allowed a quantitative study of phagosomal rupture in host cells that have engulfed Mtb, and whose subtype can be identified/determined by staining of the Critical Impact of Maturation-Control on Cytosolic Access of Mtb specific surface markers. Hence, our experimental setup was adapted to be used for various cell types and physiological situations, including the detection of vacuolar rupture in rare (infected) cells that were dispersed in a large and heterogeneous cell background population.

Mtb-induced phagosomal rupture in dendritic cells and macrophages, relationship with the infection rate and cell necrosis
Dendritic cells (DC) and MF do not play the same roles during the infection. DCs that have engulfed Mtb, are more prone to process and present pathogen-derived antigens and to prime T cells than Mtb-laden MF which are thought to initiate the inflammatory program and are considered as long-term Mtb reservoirs. We thus comparatively evaluated the potential of Mtb to induce phagosome rupture in bone-marrow-derived (BM)-DC and -MF At first, by using fluorescent DsRed WT and ΔESX-1 Mtb variants, we showed similar uptake and infectivity of both strains at the beginning of the infection ( Fig. 2A). Infection of BM-DC and BM-MF with WT Mtb then resulted in a strong blue shift at 3 dpi and thereafter, whereas for cells infected with the ΔESX-1 Mtb strain only a minor blue shift was detected (Fig. 2B). The relatively stable CCF-4 green signal and its progressively increasing blue shift for WT Mtb resulted in a blue/ green ratio of 15 in BM-DC and 10 in BM-MF respectively, at 6 dpi ( Fig. 2C). Similar as observed for THP-1 cells (Fig. 1D), infection of BM-DC with DsRed expressing Mtb showed that cells, which had engulfed DsRed Mtb, progressively increased their CCF4 blue shift over the observation period of 3 to 5 dpi (S3 Fig.). Together, these results suggest that ESX-1-dependent, Mtb-induced phagosomal rupture does occur in DC and MF.
To ascertain that the absence of FRET inhibition in cells infected with the ΔESX-1 Mtb mutant was not due to other molecular reasons than the absence of the ESX-1 secretion system, we complemented the Mtb ΔESX-1 strain with the integrative cosmid p2F9, containing 32 kb of the ESX-1 encoding genomic region from Mtb H37Rv [42]. This complementation reconstituted the ability of the resulting strain to induce phagosomal rupture, and thereby validated the ΔESX-1 mutants used throughout this study (S2C Fig.).
When uncontrolled inside the host cell, Mtb infection may lead to necrosis [27,43], which could theoretically allow exchanges between phagosome and cytosol and thereby establish a contact between mycobacterial β-lactamase located within the phagosome and CCF-4 located inside the cytosol. To investigate this key question, we determined whether the cytosolic access of Mtb was a consequence of host cell necrosis. In a dose-response experiment, changes in the FRET signal for the Mtb WT strain were seen as a function of the MOI (Fig. 2D). Except for an MOI below 1, the proportions of BM-DC displaying FRET inhibition were higher than the percentages of necrotic cells (Fig. 2E). In contrast, BM-DC infected with Mtb ΔESX-1 at the same MOIs displayed much weaker CCF-4 blue shifts. These data suggest that ESX-1-mediated phagosomal rupture progressively occurs in phagocytes in an MOI-dependent manner and that the resultant presence of mycobacterial β-lactamase activity in the host cell cytosol does not arise from host cell necrosis but rather precedes cell death.
Early minor levels of phagosome disruption and their full proportionality with type I IFN production So far, Mtb-induced phagosomal rupture has only been observed at later stages of infection, i.e, 3-5 dpi, a kinetic situation, which cannot explain the very early, ESX-1-dependent release of type I IFNs or IL-1β, that requires recognition of mycobacterial components by the host cytosolic sensors [44]. However, our highly sensitive approach allowed now detection of minor levels of FRET inhibition indicated by enhanced MFI 447 nm (blue), as early as 3 hours post infection (hpi) with WT Mtb (Fig. 3A-B). The blue shift then progressively increased at 24 and Comparison of these results with those from infection experiments using the Mtb ΔESX-1 deletion mutant, which overall showed much lower MFI 447nm (blue) values (Fig. 3B), suggests that Mtb-mediated phagosomal rupture begins already at such early time-points, likely caused by initial ESX-1-induced pore forming activity, and progresses into stronger phagosomal disassembly over time. These findings suggest that the time during which the Mtb-infected host cell displays phagosomal rupture and Mtb cytosolic access, prior to host cell death, is longer than previously estimated [15].
Considering the long Mtb replication time of 20h, such early initiation of Mtb-mediated phagosomal rupture suggests that this phenomenon does not depend on bacterial replication, but on the functions of the implicated bacterial virulence factors. The levels of phagosome disruption were entirely proportional to the amounts of secreted IFN-β (Fig. 3C). A partially ESX-1-dependent increase in the IFN-α secretion was also detected, which might be linked to the induction of Irf7 subsequent to IFN-β induction [45]. Therefore, minute levels of early phagosomal rupture are in direct correlation with the kinetics of the induction of type I IFN production. In contrast, no differences were found between ESX-1-proficient and ΔESX-1 Mtb strains when IL-1β secretion was studied (S4 Fig.), which is consistent with the inflammasome/ caspase-1-independent IL-1β secretion in mice during Mtb infection [46] and which is different to the situation in humans [36].
Link between the phagosomal environment and the ability to induce phagosomal rupture We next evaluated whether the characteristic Mtb-mediated partial inhibition of phagosome acidification was connected to the phenomenon of phagosome rupture. Given the previously established role of Natural resistance-associated macrophage protein (Nramp)-1, a phagosomal bivalent cation transporter, in phagosomal acidification and pH regulation [47][48][49], we evaluated its possible impact on mycobacteria-mediated phagosomal rupture. We thus used Mtb WT or ΔESX-1 strains to infect cells from the murine MF cell line Raw264.7, deficient in functional Nramp-1, which had been transfected with a non-functional nramp-1S (Sensitive) or a functional nramp-1R (Resistance) allele [50]. At 3 dpi, intense CCF-4 blue shifts were observed in WT Mtb-infected parental Raw264.7 cells and Raw264.7::Nramp-1S cells, whereas much less FRET inhibition was detected in Raw264.7::Nramp-1R cells ( Fig. 4A-C). As assessed for various MOI, the intracellular mycobacterial load inside parental, Nramp-1S-or Nramp-1Rtransfected Raw264.7 cells was comparable at 3 dpi, when the phagosomal rupture was monitored (Fig. 4D). Thus, the functional Nramp-1R seems to provide protection against Mtbinduced phagosomal rupture for the benefit of the host cell. The Nramp-1-mediated rescue of the host cells occurred at any MOI and independently of the host cell proliferation rate, which as we noticed, both influence the control of the infection (S5 Fig.). We obtained further confirmation of our results by using an nramp-1 gene silencing strategy in Raw264.7::Nramp-1R cells (Fig. 4E), which reversed the phenotype and promoted Mtb-mediated phagosomal rupture ( Fig. 4F-G).
We further treated Raw264.7::Nramp-1R cells or, as primary phagocytes, BM-DC from Sv129 (nramp-1R) mice with bafilomycin, a specific inhibitor of vacuolar proton ATPases, prior to infection with WT Mtb H37Rv. As shown in Fig. 5A-B, the bafilomycin-mediated in parallel, in cells from the same BM-DC cultures. The results are representative of 2 independent experiments. Of note, the decrease of the dead cell percentage for WT Mtb at very high MOI is likely due to generation of cell debris, not anymore measurable by cytometry.
doi:10.1371/journal.ppat.1004650.g002   reduction of phagosomal acidification resulted in enhanced phagosomal rupture in both cell types. This observation provides additional evidence for a link between restriction of phagosome acidification and the strength of observed phagosomal rupture. In this FRET-based method, the β-lactamase operates on CCF-4 located in the host cytosol, where the pH remains neutral [25,41]. However, to further ascertain that the micro-environmental acidity did not affect the functionality of mycobacterial BlaC, we tested the β-lactamase enzymatic activity of Mtb at different pH levels by the use of nitrocefin, a chromogenic β-lactamase substrate. These experiments confirmed that Mtb, grown at different pH, ranging from 5 to 7, preserves entirely its β-lactamase enzymatic activity (Fig. 5C).
Thus, acidification of the phagosomal lumen seems to be a critical host cell parameters, which exerts an antagonistic effect on Mtb-mediated phagosomal rupture in phagocytes. The finding that both phenomena are linked provides a new basis for elucidating the molecular key players that govern the host-pathogen interaction during Mtb infection.

ESX-1-dependent Mtb-mediated phagosome disruption in pulmonary phagocytes and in vivo in lungs and spleen of infected mice
Previous studies on vacuolar rupture and phagosomal escape of M. marinum [12,51] and Mtb [13,15,16] used infected MF or DC under in vitro conditions. To extend our investigations towards cells from the lung, we examined the Mtb-mediated phagosomal rupture in different phagocyte types of mouse airways. To this end, low-density cells isolated from mouse lung parenchyma were infected ex vivo at an MOI of 1 with ΔESX-1 or WT Mtb strains. CCF-4 signals obtained from monocytes/MF (CD11b hi CD11c -) and DC (CD11b int CD11c + ) were analyzed at 4 dpi, when changes in the FRET signal were detected in lung monocytes/MF and DC (Fig. 6A), showing the occurrence of Mtb-mediated phagosomal rupture in the primary lung phagocytes.
To assess the relevance of mycobacteria-mediated phagosomal rupture in phagocytes in vivo, in a first attempt we used T-/B-cell deficient recombination activation gene (rag) 2 knockout mice in which infection with Mtb is more persistent and the innate cell compartments more developed than in their immunocompetent counterparts. However, flow cytometric analysis of lung-or spleen-derived MF/monocytes, DC and neutrophils obtained from infected (1 x 10 6 CFU i.v. /mouse of WT or ΔESX-1 Mtb) or uninfected rag2°/°mice displayed indistinguishable CCF-4 blue profiles (S6 Fig.). The apparent failure in the detection of phagosomal rupture in this experimental setting seems to be related to the very low frequencies of mycobacteria-infected cells within each innate cell subset and/or a possible furtive feature of the phenomenon in vivo due to possible efferocytosis [52] of the primary phagocytes, in which phagosomal rupture and certain damage signals would have been initiated.
To distinguish infected and non-infected cells, we then used fluorescent DsRed-WT Mtb (1 x 10 6 CFU/mouse) for intravenous (i.v.) infection of C57BL/6 mice, which allowed us to focus on the relatively few Mtb-infected phagocytes present during the initial phase of chronic infection. At 3 weeks p.i. mice were sacrificed, the spleens homogenized and resulting cells enriched and subjected to flow cytometric analysis. We have focused on the phagocytes of the spleen because this organ is particularly targeted by the i.v. route of infection. When the CCF-4 blue signal of the innate immune cells that contained DsRed Mtb was compared to the other cells inside each cell subset in the spleen (Fig. 6B), a slight increase in CCF-4 blue signal was notably detected in Mtb-containing cells in the subsets of neutrophils (CD11b hi CD11c -Ly6G + ) and MF/monocytes (CD11b int CD11c -Ly6G -) (Fig. 6C), which suggests the occurrence of weak, albeit reproducible, levels of phagosomal rupture in these infected cells. Interestingly, no DsRed + cells were detected inside the CD11b lo CD11c + Ly6G -DC subset, which might be due to Critical Impact of Maturation-Control on Cytosolic Access of Mtb possible rapid turnover of infected DC or to their CD11b up-regulation. In this chronic infection model, it was however not possible to compare WT and ΔESX-1 Mtb strains, because of the non-persistence of the latter. To overcome this limitation we developed an alternative in vivo model whereby mice were instilled intra-nasally with cells that were infected with Mtb in vitro prior to transfer, and whose infection status in vivo could be specifically monitored. To this end, BM-DC from mice with CD45.1 hematopoietic allelic marker were infected in vitro with WT or ΔESX-1 Mtb, in conditions that allowed up to 70% of the cells to be infected ( Fig. 2A), whereas control cells were left uninfected. At 16 hpi, the cells were instilled into the airways of congenic CD45.2 recipients. At different time points post-transfer, the lung lowdensity cells were isolated and the CCF-4 blue shift in the CD11b + CD45.1 cell subset of the different experimental groups assessed (Figs. 7A and S7). Strikingly, at day 4 and day 6 post-transfer, in the CD11b + CD45.1 population infected with WT Mtb, a blue shift was detected in comparison to the non-infected or ΔESX-1-infected transferred cells (Fig. 7B-C).
Moreover, independent flow cytometric examination of cells extracted directly from surface lung granuloma tissue of Mtb-infected C57BL/6 mice revealed a small, distinct cell population that displayed a clear-cut blue signal and a CD11b + CD11c + phenotype (Fig. 7D), which points to the presence of innate cells in these lungs wherein Mtb-mediated phagosomal rupture had occurred.
Altogether, our data suggest that the Mtb-induced phagosomal rupture does indeed happen in vivo, in Mtb-infected cells in the organs of small laboratory animals. The detected phagocytes containing intracellular bacteria seem to have a life-time of several days, which however does not exclude the possibility that a portion of the total number of infected phagocytes might get eliminated by efferocytosis [52], as suggested by the relatively modest differences in blue shift observed in the in vivo settings.

Discussion
The pathogenic potential of Mtb is intimately linked to the interplay between the host defense and the persistence of the mycobacteria. The intracellular localization and cytosolic access of the bacterium has substantial consequences on the recognition of mycobacteria-associated patterns by the cytosolic receptors of the innate immunity that determine innate and adaptive immune responses and ultimately the fate of the host cell and the bacterium [27]. Subsequent to phagocytosis, in order to avoid the acidified environment generated by the phagosome-lysosome fusion, some specialized intracellular bacteria, such as S. flexneri, Listeria monocytogenes or Francisella tularensis, evolved to rapidly escape from phagosomes into the cytosol [21,53,54]. In contrast, Mtb has been described as a bacterium that resists degradation in the phagosome by inhibiting the fusion with lysosomes, a characteristic feature that seems to protect the bacilli from bactericidal mechanisms of the phagocytes and allows intracellular survival and multiplication [10,18,[55][56][57]. However, recent reports based on in vitro infection of phagocytes also suggest that at later stages of infection ESX-1-dependent vacuolar breakage might be an important requirement for the pathogenic potential of Mtb, given that ESX-1-deficient bacilli that are unable to perforate and lyse the phagosomal membrane are-in general-attenuated [13,15,16,18,[56][57][58][59].
In previous studies, Mtb-mediated phagosomal escape has only been reported at late time points like 2-5 dpi, a kinetic feature that was not reconcilable with the intracellular host immune events, like type I IFN induction, which require the early recognition of mycobacterial components by cytosolic sensors. Here, the use of highly sensitive FRET-based cytometry enabled us to highlight minor levels of cytosolic contact of Mtb and its products initiated as soon as 3 hpi, which is kinetically concordant and proportional with the amounts of IFN-β released In parallel to these cells, lung low-by DC. While we cannot exclude the possibility that some of this effect may have been caused by bacterial products translocating through permeable phagosomal membranes [30], the reproducible differences observed between the WT and the ΔESX-1 Mtb strains argue for a specific, ESX-1-mediated impact. We also noted that distinct cell types might display different susceptibility to phagosomal rupture, with THP-1 cells as the most susceptible ones, followed by BM-DC/BM-MF, and the Raw264.7 MF as the least affected cell types, tested.
Our results show that the phagosomal bivalent cation transporter Nramp-1 interferes with Mtb-induced phagosomal rupture as observed at 3 dpi, i.e., a time point at which mycobacterial loads were still comparable in Mtb-infected MF harboring Nramp-1S (non functional)-or Nramp-1R (functional) allelic forms. In line with that, the effect of bafilomycin, reported to inhibit phagosomal acidification [60], reconstituted in Nramp-1R-proficient phagocytes the capacity of Mtb to enhance phagosomal rupture to the level of Nramp-1S phagocytes. Thus, the partial inhibition of phagosome acidification emerges as a prerequisite to mycobacterial phagosomal rupture. Plausibly, only when phagosome acidification is partially inhibited, mycobacteria may survive, use their virulence factors and induce phagosomal membrane disruption.
Although cellular models may provide important new insights into cell biological mechanisms, evaluation of the accuracy of the findings in an in vivo model, i.e. in tissues or organs is of crucial importance to emphasize their relevance. Previous electron microscopy analyses of lung innate cells isolated from TB patients or mycobacteria-infected mice have led to discrepancies with regards to intracellular location [18]. In alveolar MF of TB patients and in granuloma or lung homogenates of infected mice, Mtb has been detected as single bacterium or pairs of bacilli inside phagosomes [61,62], whereas Mtb has also been observed in membrane-disrupted compartments or free in the cytosol in the mouse granulomas [63,64]. Moreover, heavily infected human alveolar MF [62] and damaged mouse MF of inflammatory sites [65] contain multiple mycobacteria per phagosome. In this context, our results from carefully designed in vivo infection experiments add new elements to the discussion. Although the strength of the FRET-inhibition was found weaker under in vivo conditions (Figs. 6 and 7) than observed for the cell culture-based infection assays ( Figs. 1 and 2), the reproducibility and complementarity of the results from the three distinct in vivo settings analyzed, point to biological relevance of mycobacteria-induced phagosomal rupture in the organs of Mtb-infected laboratory animals. It should be noted that in our experiment with BM-DC from mice with the CD45.1 hematopoietic allelic markers (Fig. 7), we cannot exclude that in the infected DC some minor cytosolic contact might develop already in vitro, prior to their instillation to the CD45.2 recipient mice. However, the finding that FRET inhibition remains detectable for several days after the transfer into the lungs of the CD45.2 recipients suggests that the phagocytes in which cytosolic access of Mtb progressively builds up, can survive in the host environment for some days. Together with ex vivo results from MF/monocytes and DC isolated from the lung parenchyma, the in vivo demonstration of cytosolic access of Mtb provides important new insights into the cellular events during infection inside the organs. Our data suggest that after infection, the concerned phagocytes may persist in the organs long enough to have a potential impact on host defense mechanisms that likely also include key cellular processes, such as autophagy, which requires Mtb ubiquitination in an ESX-1-dependent manner [16,33,38,39]. The intracellular localization of mycobacteria and mycobacteria-mediated phagosomal rupture have been subject of numerous controversies, which may be explained by the differences between the level of virulence of mycobacterial strains used, the MOI and the conditions of the mycobacterial cultures in vitro [18]. For the virulent strains, here we used WT and DsRed Mtb previously passaged in immunocompetent mice to maintain a normal degree of virulence and to remain as close to natural infection as possible. We only used mycobacterial cultures in midlog 10 growth phase to minimize bacterial mortality, and we cultured the bacteria in the presence of Tween 80 to avoid clumping, as phagocytosis of non-viable or clumped mycobacteria may lead to rapid phagosome-lysosome fusion and prevent visualization of phagosomal rupture [18]. In addition, we systematically compared the ESX-1-proficient and ESX-1-deficient mycobacterial strains and detected a relevant phagosomal rupture only with ESX-1-proficient strains.
Previous observations with numerous virulent and attenuated Mtb strains suggest that the capacity of a strain to induce phagosomal rupture in vitro is often correlated with its virulence [15,16]. Hence, the ESX-1-dependent, mycobacteria-induced phagosomal rupture emerges as a major characteristic feature of Mtb infection, which likely initiates the first damages caused by this intracellular pathogen to the host cell. Consequently, modulation of the parameters, which orchestrate this phenomenon, may constitute a promising base for vaccinal or therapeutic interventions against TB. For example, we have previously noticed that recombinant BCG and M. microti strains with a reconstituted ESX-1 secretion system showed enhanced protective efficacy [66,67]. More recently, a dedicated study identified small molecule inhibitors belonging to the benzyloxybenzylidene-hydrazine and the benzothiophene chemical classes, which interfered with ESAT-6 secretion and thereby protected host cells from Mtb-induced lysis [68]. Molecules belonging to closely related chemical scaffolds were also identified in a high content phenotypic screen as agents that interfered with the intracellular growth and the virulence of Mtb [69]. Hence, it is conceivable that future phenotypic library screening might identify novel pharmacological compounds that inhibit Mtb-mediated phagosomal rupture in the host cell. Such molecules would represent interesting anti-virulence compounds to be tested as addition to conventional treatment regimens against TB.
In conclusion, our study suggests that Mtb is not the passive pathogen that induces pathology only by the over-boarding reaction of the host immune system. We show that ESX-1-mediated phagosomal rupture contributes in a significant way to establish mycobacterial cytosolic contact, which is however only possible if the maturation / acidification of the phagosome is limited in a first process. In this direction, our study also opens new perspectives for future studies on the mycobacterial components involved in the modulation of phagosomal acidification such as the phthiocerol dimycocerosates and other mycobacterial factors, reported to intervene in this process [70,71].
The ESX-1 system might thus represent one of the final members in a chain of virulence factors that determine the pathogenicity of Mtb through the induction of phagosomal rupture, and its function might therefore have been evolutionary preserved [5,7]. As such, our work has the potential to reconcile the outcome of previous studies on mycobacterial virulence factors that interfere with vacuolar acidification [71][72][73][74] and studies on cellular localization of Mtb [13][14][15][16] and establishes Mtb-mediated phagosomal rupture as a basic biological mechanism involved in TB pathogenesis.

Materials and Methods
Animal infection model C57BL/6 mice, rag2°/°or CD45.1 were obtained from Animal Facilities of Institut Pasteur. C57BL/6 mice were purchased from Janvier Le Genest-Saint-Isle France). CD45.2 mice were anesthetized by i.p. injection of 100 mg/kg Ketamine (Lyon, France) and 10 mg/kg Xylazine (KCP Kiel, Germany) before cell transfer by i.n. route. Mouse infection with Mtb via aerosol route was performed as previously described [75]. Granuloma were recovered from the surface lung parenchyma of infected C57BL/6 mice at 6 weeks p.i.

Ethics statement
Mouse studies were approved by the Institut Pasteur Safety Committee, in accordance with French and European guidelines and regulations (Directive 86/609/CEE and Decree 87-848 of 19 October 1987) and the Animal Experimentation Ethics Committee Ile-de-France-1 (reference number 2012-0005).

Cell cultures
THP-1 cells (our laboratory stock collection, initially originating from ATCC provided cells) were maintained in RPMI, complemented with 10% heat-inactivated FBS and were treated with 20 ng/ml of Phorbol 12-Myristate 13-Acetate for 72h to induce their differentiation into MF. Raw264.7 cells transfected with nramp-1S or -1R allele (kind gift of Pr J. Blackwell) [50] were treated with 8 μg/ml of the selective antibiotic puromycin.

Mtb cultures and cell infection
Mtb H37Rv, WT, ΔESX-1 (kind gift of Pr. W. Jacobs) [10] or ΔESX::ESX-1 [42] were maintained in 7H9 medium supplemented with ADC (Difco). Seven-to-10 days before cell infection, bacteria were transferred into Dubos medium, which contains Tween 80, to avoid mycobacterial clumping. DsRed-WT or -ΔESX-1 strains were obtained by complementation with the pMRF plasmid containing a DsRed cassette, under the hsp60 promoter (kind gift of Dr. S. Cho) and were cultured in the continuous presence of 20 μg/ml of the selective antibiotic kanamycin. In in vivo experiments, we used an Mtb H37Rv strain with a plasmid containing the DsRed and hygromycin resistance genes (kind gift of Dr. O. Neyrolles). Only mycobacteria grown to mid-log 10 phase were used to minimize the frequency of death bacteria.
Raw264.7 cells were infected at various MOI with Mtb in complete antibiotic-free RPMI. At 3 dpi, equal numbers of cells were lysed by addition of 0.1% Triton X-100 in PBS and the intracellular CFU was determined by plating serial dilutions of cell lysates on 7H9 Agar medium and incubation at 37°C for 3 weeks.
Gene silencing siRNA transfection to cells was performed by using reverse transfection method. A pool of four Nramp-1-specific siRNAs, GGUCAAGUCUAGAGAAGUA, GAUCCUAGGCUGUCU CUUU, GGGCGACUGUGCUAGGUUU and GAAGUCAUCGGGACGGCUA, at final concentration of 50 nM, was mixed with 6 μl of lipofectamine (Invitrogen) in 500 μl of PBS in 6well plates. After 30 min incubation at RT, 3 x 10 5 cells contained in 2 ml of complete RPMI were added to the mixture and incubated for 3 days at 37°C. The efficiency of gene silencing was determined by qRT-PCR before the infection. One mg of total RNA was transcribed into cDNA. Then, 4 μL of cDNA was tested by qRT-PCR with LightCycler 480 SYBR Green using GCCACTGTGCTAGGTTTGCT and AATGGTGATCAGTACACCGC primers. All experiments were run in triplicate and the Livak method [76] was applied for relative quantification with β-actin.

Mycobacterial β-lactamase activity assay
The β-lactamase activity of Mtb, grown in Dubos broth with various pH, was measured by use of the chromogenic β-lactamase substrate, nitrocefin. Briefly 1 x 10 6 bacteria, re-suspended in 100 μl of Dubos broth at indicated pH, were incubated in 96-well plates with 50 μl of nitrocefin, reconstituted at 0.5 mg/ml in PBS which contained 5% DMSO. Absorbance by nitrocefin at 486 nm was measured after 3 hours of incubation at 37°C.

Enrichment of innate immune cells
Lungs or spleen were removed aseptically and were digested by treatment with 400 U/ml type IV collagenase and DNase I (Roche). Following a 45 min incubation at 37°C, single-cell suspensions were prepared by use of a Gentle Macs (Miltenyi) and by passage through 100-μm nylon filters (Cell Strainer, BD Falcon). When indicated, cell suspensions were enriched in low-density cells on iodixanol gradient medium (OptiPrep, Axis-Shield), according to the manufacturer's protocol. Notably this gradient only selects alive cells, as confirmed by blue Trypan exclusion assay. These cells were either used directly in flow cytometry analyses or were plated in 12 well culture plates in complete RPMI to be infected ex vivo with mycobacteria.
Supporting Information S1 Fig. Cartoon of the principle of the CCF-4 based FRET assay. In step 1, CCF4-AM (Life Technologies) represents a lipophilic, esterified form of the CCF4 substrate, which allows it to readily enter cells. As shown for step 2, upon entry, cleavage by endogenous cytoplasmic esterases rapidly converts CCF4-AM into its negatively charged form, CCF4, which is retained in the cytosol and thus cannot enter into the different cell organelles, including phagosomes containing bacteria. In case the bacteria remain engulfed in the intact phagosome, the endogenous bacterial beta-lactamase can not reach the CCF-4 substrate and upon stimulation at~409 nm, CCF-4 shows Fluorescence Resonance Energy Transfer (FRET) that leads to emmision of "green" light at~535 nm. As shown for step 3, in case of phagosomal rupture and cytosolic contact the mycobacterial β-lactamase (shown as red dots linked to the bacterium) gets in contact with the CCF-4 substrate trapped in the cytosol, and is inducing cleavage of the substrate and inhibiting FRET, thereby leading to emission of "blue" light at~450 nm. It is noteworthy that, compared to THP-1 cells, BM-DC or BM-MF, relatively low levels of phagosomal rupture were generally observable in Raw264.7 MF. Indeed, a CCF-4 blue shift is weakly detectable at 2 dpi, peaks at 3 dpi and then decreases as soon as 4 dpi. This feature seems to be linked to intense proliferative capacity of these cells despite the infection and also to their possible intense efferocytic capacity.