Mycobacterial escape from macrophage phagosomes to the cytoplasm represents an alternate adaptation mechanism

Survival of Mycobacterium tuberculosis (Mtb) within the host macrophage is mediated through pathogen-dependent inhibition of phagosome-lysosome fusion, which enables bacteria to persist within the immature phagosomal compartment. By employing ultrastructural examination of different field isolates supported by biochemical analysis, we found that some of the Mtb strains were in fact poorly adapted for subsistence within endocytic vesicles of infected macrophages. Instead, through a mechanism involving activation of host cytosolic phospholipase A2, these bacteria rapidly escaped from phagosomes, and established residence in the cytoplasm of the host cell. Interestingly, by facilitating an enhanced suppression of host cellular autophagy, this translocation served as an alternate virulence acquisition mechanism. Thus, our studies reveal plasticity in the adaptation strategies employed by Mtb, for survival in the host macrophage.

This marked distinction that segregated the strains into two groups was somewhat surprising. To verify this further we prepared lysates from H37Rv-and JAL2287-infected cells by using a protocol that leaves the membrane components intact 8 . TEM examination of these lysates revealed that, while the H37Rv bacteria were contained within vesicles, those of JAL2287 indeed remained free of any such bounded compartments (Fig. 1B). Unambiguous confirmation was then sought through electron tomographic 3D imaging of representative fields in sections from H37Rv-and JAL-2287-infected, cells. Electron tomography provides a powerful volume rendering of ultrastructural components and has been extensively used for 3D visualization of fine structure, as well as host-pathogen interaction events 14 . Analysis of the tomogram data through isovolume rendering in the two instances clearly confirmed the cytoplasmic existence of JAL2287 without any artifacts (Fig. 1C, lower panel). Moreover, the reconstruction revealed no connectivity of the JAL2287 bacterium through any specialized structure of the host cytoplasm but, rather, highlighted its presence in the free form (Fig. 1C, lower right panel). In contrast, analysis of the tomogram data for H37Rv only served to further reinforce its vesicle-enclosed nature (Fig. 1C, top panel). Figure 1D shows the distribution between membrane-free versus vacuolar bacteria that was obtained for the different strains in THP-1 cells, whereas Fig. 1E compares the bacillary load per cell in each case. No correlation was detected between these two features, which is consistent with earlier findings that phagosomal escape of bacteria does not at least entirely depend on the infection load 11,15 .
To confirm that the observed differences in intracellular niche preferences truly represented intrinsic properties of the individual strains, rather than a derivative of the host cell line used, we next infected mice with each of the Mtb strains. At 15 days later these mice were sacrificed and the lungs removed for subsequent processing and examination by TEM. We specifically monitored the infected alveolar macrophages, based on morphological criteria, and the results thus obtained reaffirmed our findings in THP-1 cells. That is while the H37Rv, JAL2261, BND433, BND320 and 1934 bacilli remained enclosed within vacuoles ( Fig. 2A, Fig. S1C), between 55-67% of the JAL2287, MYC431 and 2549 bacteria were found to be extra-vacuolar (Fig. 2B, Fig. S1D). This is highlighted by the representative high-magnification image for JAL2287 where the continuity of bacterial cell wall with the cytoplasm of lung macrophage is clearly evident, as compared to the situation for the vesicle-enclosed H37Rv ( Fig. 2C-(i), C-(ii)). Importantly, we could again confirm that this distinction in sub-cellular localization was not due to differences in bacterial loads. A parallel enumeration of bacterial counts in lungs of the corresponding mice revealed an overlap in the range of CFU values obtained for the two groups (Fig. 2D).
Thus our above results highlight the marked differences in apparent proficiencies exhibited by the individual Mtb isolates, in terms of translocating from phagosomes to the cytosol. Since only single time point observations were made however, it was unclear whether these differences reflected quantitative differences in the capacity to translocate, or, variations in the timing at which the individual strains initiated this process. In this context our finding that JAL2287, 2549, and MYC431 bacteria were already localized primarily in the cytoplasm by 24 hr after THP-1 cell infection ( Fig. 1B is especially significant. It contrasts with earlier suggestions that mycobacteria are generally contained within endosomes in the initial stages of infection 10,11,15 .

Rapid initiation of cytoplasmic translocation in JAL2287-infected cells.
To further probe the possible occurrence of phagosome escape in the early stage of infection, we next studied cells infected with JAL2287 for periods ranging from 2 to 48 hours. For the purposes of comparison, a parallel set of H37Rv-infected cells was also included in these experiments. Surprisingly, extra-vacuolar JAL2287 bacteria were evident even as early as 2 hrs after infection (Fig. 3A). Indeed, inspection of > 200 bacteria at each of the time points revealed that near peak levels of cytosolic bacteria was already achieved by this time point, with only a minor change thereafter (Fig. 3B). This was in spite of the fact that the cellular bacterial load increased significantly over this period (Fig. 3B). In contrast to this, H37Rv-infected cells predominantly displayed vacuolar bacteria across all the time points studied (Fig. 3A).
To resolve the above observation further we next generated sections from cells exposed to JAL2287 for only 1 hr, and analyzed them by TEM. Representative images are shown in Fig. 3C,D. It is obvious from Fig. 3C that, although vesicle-localized bacteria were present, several cytosolic bacteria also already appeared by this time. Figure 3D further elaborates this observation by depicting a phagosome where one of the bacteria is partly extruded out of the vesicle, with its membrane now fused to the phagosome membrane. Thus the results in Fig. 3C,D together suggest that, at least for JAL2287, phagosomal escape likely occurs in tandem with the infection process.
Phagosomal escape is mediated through cytosolic phospholipase A2 activation. Previous studies have established that cytosolic translocation of mycobacteria is mediated by ESAT-6 [9][10][11]15 . The membrane lysing activity of the protein has been implicated in this function, and more recent studies have delineated the functional domain to reside at its C-terminal end 9 . Although ESAT-6 is undoubtedly essential it alone, however, was not sufficient to explain the observed strain-specific differences in vacuolar versus cytosolic distribution of the bacteria. This was because, as previously shown 13 , all of these strains secrete ESAT-6 in the host cell milieu. Therefore, an additional pathway likely complemented with ESAT-6, to regulate strain-specific variability in the extent and kinetics of phagosomal escape. Such an inference would also be consistent with earlier observations that even BCG, which does not express ESAT-6, shows a limited capacity for permeabilizing phagosomes 16 .
The cytoplasmic phospholipase A 2 (cPLA2) enzymes play a critical role in both phagosomal trafficking, and export of cargo from the various endocytic compartments [17][18][19] . These enzymes cleave the sn2 acyl bond of phospholipids, releasing arachidonic acid and lysophospholipids. Previous studies have shown that cPLA 2 expression was enhanced in Mtb-infected macrophages 20 , and that recruitment of the subsequently derived arachidonic acid into the 5-lipoxygenase-dependent pathway inhibited macrophage apoptosis and cross-presentation of Mtb antigens by dendritic cells 21 . In the context of the present study though, cPLA 2 has also been shown to permeabilize the endosomal membrane in Mtb-infected macrophages 22 . Therefore, to test whether such activities may additionally contribute to Mtb cytosolic translocation we employed PED-6, a dye-labeled substrate for PLA 2 . PED-6 is a glycerophosphoethanolamine probe with a BIODIPY dye-labeled acyl chain at the sn2 position, and with the head group modified with a dinitrophenyl quencher. Cells infected with each of the eight strains were loaded with a liposomal preparation of the PED-6 probe as described earlier 22 and cPLA 2 activation was determined by intensity of the resulting fluorescence signal. The latter was a result of the unquenching of BIODIPY fluorescence, due to phospholipase (cPLA2)-dependent cleavage of the sn2 acyl bond.
Although some degree of cPLA 2 activation was evident in all cases, significantly higher levels were obtained in cells infected with the three strains that also displayed a bias towards cytosolic residence (Fig. 3E). Notably, a regression analysis between PED-6 fluorescence intensity and the proportion of cytoplasmic bacteria obtained in the differently infected cells yielded an excellent positive correlation (Fig. 3F), implying a link between phospholipase (cPLA2) activation and phagosomal escape of Mtb. This interpretation was further supported by the results of our subsequent experiments employing arachidonyl trifluoromethyl ketone (AACOCF3), an inhibitor of cPLA 2 23,24 . Mtb infection of cells in the presence of AACOCF3 led to the expected suppression of PED-6 fluorescence by 24 hr (Fig. S2A). Significantly, this treatment also resulted in a marked reduction in cytosolic population of the JAL2287, 2549, and MYC431 bacteria (Fig. 4A). This reduction was not due either to a loss in host cell viability, or any significant reduction in the cellular bacterial load (see Fig. 4B). Instead, it appeared to derive from an enhanced restriction of bacilli within the phagosomal compartment (Fig. 4A). Thus, at one level, these findings support that cPLA 2 activity also contributes towards regulating phagosomal escape of Mtb. In addition,   3F also supports that the observed strain-specific differences in extent of translocation likely derive from corresponding differences in ability of the Mtb strain to activate host cPLA2 in the intracellular milieu.
As opposed to the effects of short-term treatment, extended exposure to AACOCF3 of cells infected with either of the cytosol-preferring strains (JAL2287, 2549, MYC431) led to a progressive reduction in bacterial counts in all cases (Fig. 4B). This effect, however, was selective since the bacillary load in cells harboring any of the other, vesicle-localized, isolates remained relatively insensitive to the treatment (Fig. 4B). We subsequently also confirmed that AACOCF3 did not influence viability of either bacteria when grown in extracellular cultures, or of the infected host cells (Fig. S2B,C). Rather, a TEM analysis revealed that AACOCF3-treatment of infected cells caused an increased vesicular degradation of the cytosol-preferring, but not the endosome-localized, Mtb strains  Fig. 1D. The positive correlation implies a link between host cytosolic phospholipase A2 activation and phagosomal escape of Mtb. (Fig. S2D,E). Thus, taken together with the findings in Fig. 4A, these results indicate that a sustained prevention of phagosomal escape renders the cytosol-preferring strains more vulnerable to the anti-bacterial mechanisms of the host macrophage.
Cytosolic residence confers resistance to autophagy. Intriguingly, treatment of infected cells with rapamycin produced distinct effects on the vesicular-versus the cytosol-resident groups of Mtb strains. Whereas intracellular viability of the former was markedly compromised, the latter group proved to be significantly more resistant (Fig. 4C). Rapamycin is a bacterial macrolide that inhibits the mammalian target of rapamycin (mTOR) and, thereby, stimulates autophagy 25 . Studies from several laboratories have established that autophagy represents a key anti-mycobacterial mechanism of the macrophage, and the ability to suppress this process has been intricately linked to mycobacterial virulence 12,26,27 . Thus our finding in Fig. 4C suggests that the cytosol-preferring Mtb strains gain in virulence function by acquiring resistance to autophagy. The interpretation that cytosolic residence confers resistance to autophagy is additionally supported by the strong negative correlation seen between the proportion of cytosolic bacteria obtained for the different strains, and their relative insensitivity to rapamycin treatment (Fig. 4D). In subsequent experiments we found that rapamycin-dependent activation of autophagy was attenuated to a greater extent in cells infected with either of the cytosol residing Mtb strains, as opposed to those displaying a bias towards endosomal localization (Fig. 4E, Fig. S3). To further verify that intracellular localization of Mtb differentially influences autophagic responses of the host cell, we also tested under the physiologically more relevant condition of IFN-γ stimulation. Several studies have documented a central role for IFN-γ in the host immune response to Mtb infection [28][29][30] . Among other effects, the reported mode of action of this cytokine includes activation of autophagy and phagosome maturation in the infected macrophage 31,32 . Thus, infected cells were stimulated with IFN-γ and then monitored for autophagy induction. The results obtained are shown in Fig. 4F. It is evident here that, while cells infected with either of the vesicle-localized Mtb strains displayed a strong autophagy response the effect of IFN-γ was-however-significantly attenuated in cells harboring the cytosol-preferring bacteria. Importantly though in the latter instance, sensitivity to IFN-γ was markedly improved when cytosolic translocation of the bacteria was prevented by treatment of cells with AACOCF3 (Fig. S4). Thus, while the mechanism remains unknown, our cumulative findings confirm that cytosolic localization of Mtb is associated with an enhanced capacity for suppressing activation of host cellular autophagy.

Discussion
Our study provides an added perspective to the accumulating support for Mtb translocation from phagosomes to the cytosol of infected macrophages. Contrary to the prevailing notion that translocation always occurred several days post-infection, we found that the time course of this process was in fact strain dependent. Thus, in some strains, cytosolic translocation was evident within hours after macrophage entry. Indeed in the case of JAL2287, phagosomal escape appeared to be co-initiated with the infection process and a dominance of the cytoplasmic over the endosomal pool of bacteria was evident by as early as two hours of infection. Also notable is our finding that ESAT-6 alone could not account for the observed differences in translocation competency of the various Mtb strains. Instead, this latter feature showed a strong correlation with the degree of cPLA 2 activity that was induced within infected cells. Thus, host cPLA 2 activity likely complements ESAT-6 function to regulate phagosomal escape of Mtb. The basis for differential cPLA 2 activation by the individual Mtb strains, however, awaits future clarification.
Intriguingly, for the Mtb strains that preferentially localized to the cytoplasm, prevention of phagosomal escape through lipase inhibition rendered them vulnerable to killing by the host macrophage. In contrast, cPLA 2 inhibition had only a marginal effect on intracellular viability of those Mtb strains that were primarily restricted to the endosomal compartment. At least based on our preliminary findings, this distinction in sensitivity to lipase inhibition appeared to reflect a differential capacity between these two groups of strains to withstand phagosomal stresses. Consequently, it is tempting to speculate that early translocation into the cytoplasm presents an alternate survival mechanism for such strains that are less adapted to the endosomal environment. While this proposition is yet only a hypothesis and requires verification, the possibility that cytosolic translocation provides relief from microbicidal mechanisms of the macrophage has been previously suggested 8,15 .
Earlier reports have indicated that, relative to the phagosome, the cytoplasm provides a more conducive environment for mycobacterial replication 11,33 . Further, by 8 subsequently inducing necrotic death of the host cell, cytosolic translocation has also been suggested to facilitate escape of the bacteria from the intracellular environment altogether 10,34 . In addition to these properties though, our study identified yet another pro-survival trait that was acquired by the bacilli as a result of cytosolic translocation. In comparison with Mtb strains that predominated in the endosomes, those biased towards cytosolic residence were found to be significantly more adept at inhibiting the activation of autophagy in the host cell. Given that autophagy serves as a key pathway for clearance of intracellular bacteria by macrophages 12,26,27 , cytosolic translocation could well then constitute a 'restoration-of-virulence' strategy for those Mtb strains that are less adapted to survive within phagosomes. That is, the susceptibility resulting from an inherent deficit in the latter property was recompensed by a concomitant gain in resistance to autophagy.
The broad range of bactericidal responses unleashed by the macrophage requires the pathogen to acquire a portfolio of countermeasures, in order to survive and persist in the intracellular milieu. Such countermeasures include prevention of phagosomal maturation, suppression of oxidative responses, prevention of apoptosis, accession of nutrients, and capacity to enter into a non-replicative state among others 35 . In such a scenario then, 'pathogen fitness' could likely represent a combinatorial outcome of the relative contribution from these individual defense pathways. Our present study is especially relevant in this context as it uncovers an additional level of plasticity to the adaptation mechanisms resourced by Mtb, for surviving within macrophages.

Methods
Infection, Treatment, and CFU Determination. Infection of THP-1 and HuMφ cells with at a multiplicity of infection of 10:1, and determination of CFUs were done as described earlier 12,13 . Where required, AACOCF3 was added at a final concentration of 25 μM and Rapamycin at 5 μM. THP-1 cells were originally obtained from "ATCC" and maintained in our laboratory. For mice experiments, animal were obtained from the small animal facility of ICGEB and were infected with Mtb strains through aerosol exposure by delivering between 100-150 bacteria per lung. We confirmed the comparable bacterial uptake in all the groups by measuring the cfu in lungs of a parallel set of animals (3/group) at 24 hrs after infection. Two weeks from infection the mice were sacrificed and lungs were removed. Lungs from two mice in each group were taken for TEM imaging and remaining were plating for cfu determination as previously reported 12,13 . Transmission electron microscopy. Infected cells, or pieces of infected lungs were washed with sodium cacodylate buffer (pH, 7.4) and fixed in 3% glutaraldehyde. Cells were then osmicated and dehydrated prior to embedding in EPON 812 resin 36 . Polymerized blocks were cut into 80-100 mm ultrathin sections and collected on 200 or 400 mesh size copper grids. Sections were stained with Uranyl acetate (10 min-1 hr), followed by Reynolds lead citrate (2 min) and then imaged under 80 and 100 KeV operating voltages of a Tecnai 12 Biotwin transmission electron microscope (FEI Co, The Netherlands). Images were captured using a side mounted 2 k × 2k CCD camera (Megaview III, SIS Germany). Size calibration and morphometry was done using the TIA image analysis software (FEI Co, The Netherlands).
Extensive control imaging was done to ensure that the images collected were free of artifacts. This included monitoring for the natural preservation of nucleus, nuclear envelope, mitochondria and other cyto-organelles. Further, an adequate number of cell profiles and imaging fields were seen before any interpretation was made. Finally, we also ensured that our interpretations were based on published literature and standard reference Scientific RepoRts | 6:23089 | DOI: 10.1038/srep23089 material on ultrastructural pathology. For bacterial detection and enumeration, only those displaying morphologically relevant features as described in the literature were considered. Electron tomography. The stained grids imaged under 200 KeV operating voltage in a Tecnai-120 transmission electron microscope. Single axis serial tilt tomograms were recorded from + 50° to − 55° using a 2° tilt interval and 0.5° interval in high range were collected. The acquisition was done using a automated tomography acquisition software (XPLOR 3D TM suite, FEI, Co, The Netherlands). The goniometer stage centering, eucentric height adjustment, holder calibrations, autofocus function adjustments, cross-correlation feature tracking, were all automated and controlled by the automated-tomography software. Tomogram series were stored as MRC files in a networked Terastation (Buffalo computers USA). Tomogram sets were recorded at magnification of 200 nm.
For image processing and volume reconstruction, the acquired tomogram sets were subsequently processed using the Inspect 3D software (FEI Co., The Netherlands), stacks aligned and de-noised. The volume reconstructions of areas of interest were carried out using both Weighted Back Projection (WBP) and SIRT method with 20 iterations. Visualization was done using the AMIRA 5.0 software (AMIRA, France).