Listeria monocytogenes switches from dissemination to persistence by adopting a vacuolar lifestyle in epithelial cells

Listeria monocytogenes causes listeriosis, a foodborne disease that poses serious risks to fetuses, newborns and immunocompromised adults. This intracellular bacterial pathogen proliferates in the host cytosol and exploits the host actin polymerization machinery to spread from cell-to-cell and disseminate in the host. Here, we report that during several days of infection in human hepatocytes or trophoblast cells, L. monocytogenes switches from this active motile lifestyle to a stage of persistence in vacuoles. Upon intercellular spread, bacteria gradually stopped producing the actin-nucleating protein ActA and became trapped in lysosome-like vacuoles termed Listeria-Containing Vacuoles (LisCVs). Subpopulations of bacteria resisted degradation in LisCVs and entered a slow/non-replicative state. During the subculture of host cells harboring LisCVs, bacteria showed a capacity to cycle between the vacuolar and the actin-based motility stages. When ActA was absent, such as in ΔactA mutants, vacuolar bacteria parasitized host cells in the so-called “viable but non-culturable” state (VBNC), preventing their detection by conventional colony counting methods. The exposure of infected cells to high doses of gentamicin did not trigger the formation of LisCVs, but selected for vacuolar and VBNC bacteria. Together, these results reveal the ability of L. monocytogenes to enter a persistent state in a subset of epithelial cells, which may favor the asymptomatic carriage of this pathogen, lengthen the incubation period of listeriosis, and promote bacterial survival during antibiotic therapy.


Introduction
Over the past three decades, the study of the facultative intracellular pathogen L. monocytogenes has provided important insights into infection biology, immunity and host cell biology [1][2][3][4][5]. L. monocytogenes is a Gram-positive bacterium that belongs to the category of pathogens that escape from an internalization vacuole and proliferate in the cytosol of mammalian cells [6][7][8]. Following phagocytosis or receptor-mediated endocytosis, bacteria disrupt the invasion vacuole and enter the cytoplasm, where they replicate and use the actin cytoskeleton to propel themselves from cell-to-cell, thereby escaping intracellular autophagy and extracellular humoral immune responses [4,5,9]. To achieve this cytosolic lifestyle, Listeria deploys virulence effectors that specifically control a repertoire of host molecules and hijack cellular processes [4,5]. For instance, the invasion proteins InlA and InlB promote bacterial entry into non-phagocytic cells by interacting with the surface receptors E-cadherin and c-Met, respectively [10]; the cytolysin listeriolysin O (LLO) disrupts the internalization vacuole and triggers various host cell responses [11,12]; the listerial locomotion factor ActA mimics the host's actin nucleation machinery to stimulate actin polymerization, thereby promoting bacterial motility and intercellular spreading [13]. ActA also protects cytosolic bacteria from autophagic recognition by masking the bacterial surface [14,15].
Despite potent virulence factors and the capacity to invade a variety of cell types, L. monocytogenes causes a relatively rare disease. Invasive listeriosis is characterized by severe clinical manifestations (septicemia, meningitis and miscarriages) and a high fatality rate (20-30% of cases), but its incidence remains low (1 to 10 cases per one million individuals [16]). The proposed explanation for this low incidence is that cell-mediated immunity is highly protective in healthy individuals. Accordingly, conditions that weaken the immune system, such as pregnancy, aging or immunosuppressive treatments, predispose individuals to listerial infection. However, even in susceptible individuals, the incubation period of listeriosis may be long, particularly in pregnancy-associated cases for which it can last up to three months [17]. In addition, L. monocytogenes is ubiquitously distributed in the environment and exposure to this pathogen appears to be relatively common, with 1-5% of people shedding these bacteria in their feces without developing symptoms [18]. Together, these data suggest that colonization of humans by L. monocytogenes involves an asymptomatic silent phase.
Until now, only a few studies support the existence of asymptomatic L. monocytogenes reservoirs. Immunodeficient SCID mice develop a chronic Listeria infection, particularly in hepatic macrophages, in which bacteria reside within large vacuoles [19]. Spacious Listeriacontaining vacuoles have also been described in murine bone-marrow derived and RAW 264.7 macrophages cultured in vitro [20,21]. These vacuoles, termed Spacious Listeria-containing phagosomes (SLAPs), are phagosomes that do not mature into phagolysosomes and thus derive from a subpopulation of bacteria that never enter the cytosol. SLAPs are proposed to constitute a niche for L. monocytogenes survival in the host. The bone marrow and gall bladder are also potential asymptomatic reservoirs of L. monocytogenes, as revealed by bioluminescence imaging studies in mice [22,23].
However, all these observations were made in murine models. Although studying Listeria infection in the mouse has led to the discovery of most L. monocytogenes virulence factors [4], this model has some limitations since murine listeriosis does not recapitulate all the characteristics of human listeriosis. In the mouse, oral infection with L. monocytogenes is not very efficient and bacteria do not appear to have an elective tropism for the central nervous system and the fetoplacental unit. The low efficiency of L. monocytogenes in crossing murine epithelial barriers is in part due to the lack of interaction between the invasion protein InlA and mouse Ecadherin, impairing bacterial entry in murine epithelial cells [24]. In addition, differences in mucosal immunity between mice and humans [25] may change the way these species respond to L. monocytogenes infection. For instance, L. monocytogenes activates expression of IFN-λ in human hepatocytes [26] and modulates IFN-λ responses [27], but hepatocytes do not respond to IFN-λ in the mouse [28]. Such species-specific differences may prevent the characterization of some of the mechanisms involved in prolonged L. monocytogenes infection in humans, particularly in epithelial tissues.
Human cells grown in vitro can be used to model the behavior of this pathogen during a long-term infection. However, rapid multiplication and dissemination of Listeria induces cell death and detachment, thereby forming lytic plaques. Thus, in vitro infections with L. monocytogenes have been restricted to short time courses (usually from a few minutes to one day). Here, we have developed a new experimental protocol for studying L. monocytogenes during prolonged infections of non-phagocytic human cells and found that, in hepatocytes and trophoblast cells, bacteria cease to polymerize actin and are enclosed in vacuoles after the actindependent motility phase. This phenotypic switch is correlated with a decreased expression of ActA at the bacterial surface. In addition, ΔactA mutants enter a viable but non-culturable (VBNC) state during long-term infection. The formation of LisCVs could potentially enable the persistence of this pathogen in epithelial tissues.

L. monocytogenes localizes to LAMP1-associated compartments during a prolonged infection of human hepatocytes and trophoblast cells
In order to study long-term infection of epithelial cells by L. monocytogenes, the standard invasion protocol [29] was modified to reduce cytotoxicity. Bacterial inocula were prepared at stationary phase instead of exponential phase, which prevented rapid replication of bacteria at the onset of infection and increased the expression of invasion proteins InlA and InlB [30]. This enabled the use of a lower multiplicity of infection (MOI) and reduced cell damages caused by the bacterial toxin LLO. In addition, a high concentration of the membrane-impermeant antibiotic gentamicin was used for a short time (100 μg/ml for 10 minutes) to rapidly eliminate extracellular bacteria after the initial bacterial uptake. At this concentration, gentamicin does not enter the cells for at least 60 minutes [31]. Cells were then incubated in complete medium containing 25 μg/ml gentamicin, which eliminates extracellular bacteria and prevents cell reinfection by bacteria released from dead cells. Intracellular bacteria were enumerated by colony forming unit (CFU) at different time points for up to three days. Using this protocol with L. monocytogenes strain EGDe [32], marked differences were observed between the infection of HeLa cells and HepG2 hepatocytes (Fig 1A). Bacterial internalization was 100-fold higher in HepG2 cells than in HeLa cells, in agreement with the fact that both InlA and InlB mediate L. monocytogenes entry in HepG2 hepatocytes, whereas only the InlB pathway is functional in HeLa cells [33]. Following entry, bacterial loads sharply increased for 48h and then dramatically dropped in HeLa cells. In contrast, bacterial counts increased for 24h Micrograph of primary human hepatocytes infected with L. monocytogenes 10403S for 72h (MOI~5) and stained with LAMP1 (red in the overlay) and Listeria (green in the overlay) antibodies. Bar: 1 μm. E. Quantification of 10403S bacteria in different phenotypes at 72h p.i. in primary hepatocytes from three human donors. "n" indicates the number of scored bacteria. F. Ultrastructure of representative LisCVs at 72 h p.i. observed by TEM in HepG2 (images 1-3) or JEG3 cells (images 4-9). The nucleus (Nuc.), the nuclear envelope (Nuc. Env.), the membrane of the vacuole (Mb. LisCV), mitochondria (Mito.) and membranous structures (Mbs) are indicated. Images 1-3: a cluster of three Listeria (Lm.) sectioned along their short axis is enclosed within a single-membrane vacuole (LisCV). Three magnifications are shown: scale bars: 1 μm (1), 500nm (2) and 100nm (3). Image 4: three rod-shaped bacteria sectioned along their long axis within a LisCV. Bar: 500nm. Images 5-6: two LisCVs near the nucleus. The septum of a dividing bacterium is pointed with a white arrow and only slightly decreased thereafter in HepG2 cells. No major differences in the number of infected cells per well were observed between these two cell lines (Fig 1A).
Microscopic examination of infected HepG2 cells revealed a striking decline in the number of bacteria associated with filamentous actin (F-actin) from 6h to 72h post-infection (p.i.) ( Fig  1B). At the same time, there was an increase in the number of bacteria that colocalize with the late endosomal/lysosomal marker LAMP1 (Fig 1B). The percentage of bacteria associated with actin clouds or tails (Actin + ) dropped from 70% to 5% between 6h and 72h p.i., while the percentage of LAMP1 + bacteria increased from 3% to 80% (Fig 1C). In contrast, the percentage of Actin + bacteria did not change over time (30-50% of Actin + ) and the percentage of LAMP1 + bacteria remained consistently low (5-10%) in HeLa cells ( Fig 1C and S1A Fig). These results suggested that during long-term infection of HepG2 cells, bacteria stopped polymerizing actin and were incorporated into LAMP1 + vacuolar compartments.
Similar results were obtained in Huh7 hepatocytes (S1B Fig) as well as in JEG3 and BeWo trophoblast cells (S1C and S1D Fig), but not in HEK293 human embryonic kidney cells. This phenomenon was not strain-specific and was also observed with 10403S, the other widely used L. monocytogenes model strain [32], as well as with a listeriosis outbreak strain (CLIP63713, responsible for materno-fetal infections [34]) (S1C Fig): at 72h p.i in JEG3 cells, we quantified 80±5% of 10403S and 92±6% of CLIP63713 bacteria in LAMP1 + compartments. Thereafter, strains EGDe and 10403S were used for the study of this intracellular phenotype of L. monocytogenes.
L. monocytogenes also localizes to LAMP1 + compartments after longterm infection of primary human hepatocytes In order to rule out the possibility that this phenotypic switch was specific to transformed cell lines, experiments were performed in primary cells. Hepatocytes from the livers of three human donors were infected with strain 10403S. The efficiency of L. monocytogenes entry in primary hepatocytes was lower than in immortalized hepatocytes or HeLa cells (S1E Fig), as noticed for other primary cells [35], but bacterial loads were increased 200-fold after three days of infection (S1F Fig). Microscopic examination of hepatocytes infected for 72h with 10403S revealed that the majority of bacteria were enclosed in LAMP1 + compartments ( Fig 1D  and S1G Fig). More specifically, 63±3% of intracellular bacteria were positive for LAMP1, while most of the other bacteria were LAMP1-negative, but actin-free (36±2% of Actin -LAMP1 -). Only 1±0.5% of bacteria were associated with actin filaments (Actin + LAMP1 -) ( Fig 1E). These results indicate that L. monocytogenes stops polymerizing actin and entered LAMP1 + compartment during long-term infection of primary human hepatocytes.

Ultrastructure of the Listeria-containing vacuole
Transmission electron microscopy (TEM) analysis of cells infected for 72h revealed that LAMP1 + compartments were single-membrane bound vacuoles found in close proximity to the nucleus (Fig 1F, images 1-3). These Listeria-Containing Vacuoles (LisCVs) enclosed various amounts of bacteria (varying from one to twelve), which were rod-shaped in their longitudinal axis (Fig 1F, image 4). Only rare bacteria showed a division septum (Fig 1F, images 5-6). These vacuoles also contained damaged bacteria, electron-dense heterogeneous materials and membranous structures (Fig 1F, images 7-9). Of all intracellular bacteria observed with this technique (n = 591), 80% were in single-membrane vacuoles (mean ± SD of 3 ± 1 bacteria per vacuole). The other bacteria were cytosolic, either actin-free or associated with actin filaments, in membrane protrusions or within secondary vacuoles (Fig 1G). Examples of bacteria of each category are shown in S2 Fig. Importantly, 73% of vacuolar bacteria were morphologically intact, 1% exhibited a division septum, and 26% appeared degraded (Fig 1G). Of note, LisCVs were often close to compartments resembling electron-dense granular secondary lysosomes and to mitochondria (Fig 1F and S2 Fig).
Observation by confocal microscopy of GFP-expressing bacteria revealed staining of GFP in the bacterial cytosol, which indicates that the integrity of the envelope was preserved in most vacuolar bacteria ( Fig 1H). As with TEM, rare bacteria had a division septum. Hence, vacuolar Listeria have the ability to divide, albeit slowly. This is consistent with the observation that LisCVs emerged after the phase of active cytosolic proliferation, which occurs on the first day of infection (Fig 1A).
LisCVs are formed after the actin-based motility stage of L. monocytogenes Experiments were then performed to investigate whether LisCVs are formed in the first infected cells, such as SLAPs in macrophages, or in cells that are secondarily infected following cell-to-cell spread. JEG3 cells were preferentially used to address this question as these cells grow in a monolayer, which is suitable to observe cell-to-cell spread with microscopy. The effects of different strains and MOIs on the viability of JEG3 cells was first examined. As shown in S3A Fig, the best condition to limit cytotoxicity over time was to infect monolayers of JEG3 cells with strain 10403S using a very low MOI (0.1). At the onset of infection (2h p. i.), the mean number of intracellular bacteria per cell was 0.05±0.01. This low internalization rate allowed monitoring bacterial dissemination from isolated infected cells to neighboring uninfected cells. After 2h of infection, only a few cells were infected by at most one intracellular bacterium, as shown with a double-immunofluorescence staining that distinguishes between extracellular and intracellular bacteria ( In order to get a dynamic view of the process, infection of JEG3 cells by fluorescent mCherry-expressing 10403S bacteria was monitored using live cell imaging. The results confirmed that bacteria disseminated from the initially infected cells to the entire cell monolayer during the first day of infection. Thereafter, bacteria became progressively confined to perinuclear regions (S1 Movie). Highly infected cells rounded up and detached from the coverslip.
In contrast, secondarily infected cells spread on the coverslip and harbored perinuclear bacteria, as observed in fixed cells (three representative cells are indicated with arrows in S1 Movie, from 28h to 73h p.i.). These results suggested that LisCVs are not formed in the first infected cells and, consequently, do not originate from internalization vacuoles.
To confirm that LisCVs are not internalization vacuoles, we also studied the fate of a hly mutant strain, which does not produce the cytolysin LLO and fails to escape the entry vacuole in most cell types [36]. The strain 10403S-Δhly entered JEG3 cells as efficiently as wild type (WT) bacteria (Fig 2A), but the intracellular load of 10403S-Δhly bacteria declined after 6h and was about 100-fold less than that of wild-type bacteria at 72h p.i. Low-magnification microscopy confirmed that the 10403S-Δhly mutant did not disseminate in the cell monolayer ( Fig 2B). The few Δhly bacteria remaining after 3 days of infection were isolated bacteria enclosed in LAMP1 + compartments (Fig 2C and 2D). In contrast, WT bacteria that had spread in the cell monolayer ( Fig 2B) were enclosed in several LisCVs (6 ± 2 LisCVs per cell), each LisCV containing several bacteria (mean of 3 bacteria per LisCV, Fig 2C and 2D). These results indicate that LLO-deficient bacteria remain confined in entry vacuoles, which fuse with LAMP1 + compartments as in macrophages [37], but do not form LisCVs.
To demonstrate that bacteria present in LisCVs have entered the cytosol at some point during the infection process, we employed the CBD-YFP probe that specifically detects cytosolic L. monocytogenes, as described previously [37][38][39]. At 6h pi, CBD-YFP-labeled bacteria were efficiently detected in transfected cells (S4A Fig), as well as in surrounding non-transfected cells, indicating that the probe remained at the surface of motile bacteria during cell-to-cell spread. Accordingly, infection foci also contained CBD-YFP-labeled bacteria after 24h of infection. At 72h p.i., the CBD-YFP signal associated with intracellular bacteria decreased, but CBD-YFP dots remained visible on the surface of bacteria found within LAMP1 + compartments, particularly at bacterial poles (S4A and S4B Fig). Taken together, these results show that L. monocytogenes escapes into the host cytosol and spreads from cell to cell before being engulfed in vacuoles.
The formation of LisCVs coincides with the disappearance of ActA from the bacterial surface The expression of the gene encoding the actin-nucleating factor ActA is significantly induced when bacteria escape from the primary vacuole and enter into the host cytosol [40], but the fate of ActA during a prolonged infection is unknown. Since bacteria trapped in LisCVs were not associated with F-actin, we hypothesized that these bacteria no longer express ActA. Immunostaining for ActA revealed that the percentage of ActA-positive (ActA + ) bacteria decreased from about 70% to 5% in both HepG2 and JEG3 cells between 6h and 72h p.i. (Fig 3A and 3B). At 72h p.i., no LAMP1 + bacteria were labeled with ActA antibodies (Fig 3C). The decrease in the number of ActA + bacteria over time matched the decline in Actin + bacteria and began at 24h p.i., before the drastic increase in LAMP1 + bacteria observed between 48h and 72h p.i. (Fig 1C). These results suggest that bacteria enter the cytosol and non-synchronously cease producing ActA, before being incorporated in LAMP1 + compartments.

ΔactA mutant bacteria also occupy vacuoles after long-term infection
The transition of L. monocytogenes from the host cytosol to LAMP1 + compartments correlated with a decrease in ActA expression during long-term infection. However, it was possible that LisCVs derived from secondary vacuoles following cell-to-cell spread. To test this hypothesis, long-term infections were performed with 10403S-ΔactA, a strain defective in actin-based motility and cell-to-cell spread. CFU counts indicated that actA-deficient bacteria replicated for the first 6 hours of infection. Then the number of intracellular bacteria decreased in the same proportion (40-fold) as the WT bacteria at 72h p.i. (Fig 3D). Surprisingly, observation of bacteria by microscopy did not show any decrease in the number of ΔactA bacteria labeled with Listeria antibodies. Moreover, though non-motile, ΔactA bacteria were not constrained to the initially infected cells and spread to a few neighboring cells (S5 Fig), forming small infectious foci at 72h p.i. These foci showed bacteria trapped in LAMP1 + compartments, which were often larger and contained more bacteria than those generated during infection by the WT strain ( Fig 3E). Electron microscopy confirmed that ΔactA bacteria were present within single-membrane vacuoles, enclosing mixed populations of intact and damaged bacteria ( Fig  3F). A possible cause of the slight intercellular dissemination of ΔactA bacteria could be a transport of vacuoles during the division of host cells, as suggested by the detection of LAMP1 + WT and ΔactA bacteria in some mitotic cells ( Fig 3G). Together, these results indicate that, during a prolonged infection, actA-deficient bacteria become trapped in LisCVs, which may be transferred from cell-to-cell during mitosis.

LisCVs have characteristics of lysosome-like organelles
The sequestration of cytosolic Listeria by host endomembranes resembled an autophagic process. This hypothesis was supported by the fact that ΔactA mutants are targeted by autophagyrelated processes in MDCK cells [14] and macrophages [15]. In addition, ΔactA mutant have been shown to replicate as efficiently as their parent strains within the first day of infection in macrophages, suggesting that, if taken by the autophagy pathway, they avoid destruction in autolysosomes [41]. Since the link between Listeria and the autophagy machinery has not yet been studied in human epithelial cells, we examined the colocalisation of the autophagy marker LC3 with 10403S or EGDe. Strain 10403S displayed negligible recruitment of LC3 throughout the three-day infection in JEG3 cells (S6A and S6B Fig). Similarly, at 72h p.i., EGDe bacteria did not recruit LC3 in HepG2 cells (99±0.5% of bacteria were LC3-negative; n = 3). In JEG3 cells expressing GFP-LC3, a fusion protein that marks autophagic membranes [14,15,38], there was no noticeable overlap between GFP-LC3 and LAMP1 + EGDe bacteria (S6C Fig). The role of autophagy in the formation of LisCVs was also tested by knocking down BECLIN1 or ATG7, which are involved in the nucleation and elongation steps of autophagosome formation, respectively [42]. Transfection with BECLIN1 or ATG7 siRNA was efficient in reducing the expression levels of Overall, these results suggest that canonical autophagy is not involved in the formation of LisCVs, but do not exclude a role for a non-canonical pathway.
LisCVs are bound by a single membrane and contain electron-dense heterogeneous materials and membranes structures ( Fig 1F and S2 Fig), suggesting that these compartments have fused with lysosomes. In addition, TEM studies highlighted some bacteria that may be in the process of being captured by electron-dense organelles resembling secondary lysosomes (S2 Fig, lane 3). To further characterize these vacuoles, we used LysoTracker, a dye that marks acidic organelles including lysosomes [44], and an antibody that detects the lysosomal protease cathepsin D. In fixed JEG3 cells infected for 72h with 10403S, 85% and 89% of LisCVs were stained with LysoTracker and the cathepsin D antibody, respectively ( Fig 4A). Since lysosomes are degradative organelles, we also asked whether L. monocytogenes could survive in LisCVs. As previously noticed, the bacterial GFP fluorescence was preserved in many vacuolar bacteria, indicating that these bacteria were not were not lysed ( Fig 1H; S7A Fig). At the same time, quantitative TEM indicated that 26% of vacuolar bacteria were degraded ( Fig  1G). The viability of intravacuolar bacteria was assessed with the LIVE/DEAD BacLight assay. This assay is based on the discriminative labeling of two fluorescent nucleic acid-stains: the green membrane-permeant dye SYTO9 and the red membrane-impermeant dye propidium iodide (PI), which only crosses damaged cell membranes. Both dyes penetrate dead cells, but SYTO9 fluorescence is reduced in presence of PI. As such, live bacteria appear green, while those with damaged membranes appear red [45,46]. To allow the dyes to reach vacuolar Listeria, this staining protocol was performed in the presence of 0.1% Triton X100, which permeabilizes host cell membranes, including vacuoles containing bacteria [47]. The efficiency of permeabilization was demonstrated by the bright staining of host cell nuclei with PI, as the dye crossed the double-membrane of the nuclear envelope. Both wild type and ΔactA strains formed clusters of green and red bacteria in perinuclear regions, indicating that LisCVs contain mixed populations of bacteria with intact and damaged membranes (Fig 4B), in agreement with quantitative TEM data ( Fig 1G). These results reveal that a significant portion of bacteria that reside in LisCVs resist the degradative function of these lysosome-like organelles.
L. monocytogenes can return to an active state of replication and motility following subculture of host cells We then sought to determine whether L. monocytogenes has the ability to survive in LisCVs for a longer period of time. Considering that human cells need to be subcultured regularly in order to remain alive for a long time, a fluorescence-activated cell sorting (FACS)-based protocol was developed to enrich and subculture cells hosting EGDe-GFP in LisCVs (Fig 5A and  5B). Sorted HepG2 infected cells were plated in complete medium containing 25 μg/mL gentamicin, which prevents reinfection by extracellular bacteria. One hour after passaging, all sorted cells mainly contained LAMP1 + bacteria (85±5%; n = 3) (Fig 5C). After 8h of growth, cells were surprisingly divided into two distinct populations: those that still contained a majority of LAMP1 + bacteria, and those that contained a majority (over 50%) of actin-polymerizing bacteria ( Fig 5D). Since bacterial degradation was not observed in the cytosol of these cells, it was concluded that bacteria had escaped LisCVs and re-entered the cytosolic stage of their intracellular lifecycle. These cells were subsequently monitored for three days (d6), during which only remained a population of cells containing a majority of LAMP1 + bacteria (80±3%, n = 3) ( Fig  5E). After another cell passage and 24h of growth (d7), there were again two cell populations,  5F). By reproducing this experiment five times, we found that, intriguingly, the proportion of each cell population varied widely from one experiment to another (from 10% to 50%). These results showed that vacuolar bacteria could survive after more than three days of infection, but also had the ability to exit vacuoles and return to the active motility phase. Similar results were obtained in JEG3 cells infected with EGDe and subcultured following detachment with trypsin (without the FACS procedure). At day 7, from 20% to 80% of cells contained large amounts of actin-associated bacteria. Confocal microscopy highlighted bacteria with small actin tails mostly concentrated at the leading edge of the cell (Fig 5G). These results reveal that subculture of infected cells promotes a heterogeneous cycling of L. monocytogenes between vacuolar and cytosolic phases.

ActA-deficient bacteria persist in host cells as VBNC forms
In contrast to wild type bacteria, ΔactA bacteria cannot switch back to actin-based motility during cell subculturing. By studying the behavior of ΔactA mutants within cells propagated via several passages, one may study the long-term fate of vacuolar L. monocytogenes in the long term. JEG3 cells infected with EGDe-ΔactA or 10403S-ΔactA bacteria were grown for three cycles of three to four days (Fig 6A). On day ten, many cells remained infected by ΔactA bacilli associated with LAMP1 ( Fig 6B). Some of these vacuolar bacteria were observed in dividing cells at different stages of mitosis (S9 Fig), as previously observed after three days of infection ( Fig 3G). These results suggested that bacteria in vacuoles might be transmitted upon host cell division.
We quantified the number of vacuolar ΔactA bacteria at day 10 by plating cell lysates on BHI agar plates. This generated very few (EGDe-ΔactA) or no (10403S-ΔactA) bacterial colonies, even after one week of incubation at 37˚C. However, at the same time intact intracellular bacilli were observed by microscopy. In comparison to the quantification of bacteria using microscopy, CFU counts severely underestimated the number of intracellular bacteria (Fig 6C). These results suggested that actA-deficient bacteria were still viable but unable to grow on BHI plates. The inability of bacteria to grow on standard culture media, although they are still viable and maintain a low metabolic activity, defines the VBNC state [48,49]. The viability of intracellular ΔactA bacteria was assessed with the use of the LIVE/DEAD BacLight assay, which detect VBNC bacteria [50]. At day 10, most intracellular ΔactA bacteria stained green, indicated that these bacteria were not damaged ( Fig 6D). Yet, bacteria present in cell lysates from parallel experiments did not grow on BHI agar plates or in BHI liquid medium.
Assuming that if these bacteria were alive, they should not be eliminated after another cell passage, infected cells were further propagated for a total of 13 days of infection (4 cycles of subculturing, Fig 6A). LAMP1 + bacteria were still observed at this time-point, but no colonies were detected on BHI agar plates. Quantification of intracellular bacteria by immunofluorescence staining indicated a mean of 0.051 ± 0.007 bacteria per cell (~2000 cells were analyzed in each independent experiment; n = 3) for a total of about 50 000 bacteria per well. Of these bacteria, 53% were found in bacterial clumps, 46% were isolated and 0.5% showed a division septum (Fig 6E). This suggested that ΔactA bacteria found in cells after 13 days of infection had a low metabolic activity and divided slowly. Infected cells were subsequently propagated for many generations in order to determine whether VBNC bacteria are ultimately eliminated along cell passages. Rod-shaped L. monocytogenes bacteria were still detected in host cells after 25 days of subculture (seven passages), but no colonies were detected on BHI agar plates. These results indicate that ΔactA L. monocytogenes persisted in host cells in a VBNC state. Quantification of wild type intracellular L. monocytogenes by CFU or immunofluorescence-labeling of bacteria. HepG2 cells were infected with strain EGDe (same experiment as in Fig 1A) and JEG3 cells were infected with strain 10403S (same experiment as in S3C Fig). We assessed whether this phenotype was reversible upon re-expression of the actA gene in the EGDe-ΔactA mutant, by site-specific chromosomal integration of actA in the tRNA Arg locus [51,52]. We noticed that in this strain (EGDe-ΔactA+actA), actA was not regulated as in the wild type EGDe strain, since after 3 days of infection in JEG3 cells, most EGDe-ΔactA +actA bacteria produced ActA at the bacterial surface and polymerized actin, in contrast to the wild type strain (S10 Fig). This discrepancy in ActA expression between EGDe and the complemented strain might be explained by the complex regulation of actA expression [51]. This strain that produced sustained levels of ActA was used to assess the importance of ActA repression for the acquisition of an intracellular VBNC state by L. monocytogenes. In contrast to EGDe-ΔactA bacteria, EGDe-ΔactA+actA bacteria were still culturable on BHI plates after 13 days of cell propagation (Fig 6F) and continued replicating, polymerizing actin and spreading from cell-to-cell leading to many infected cell foci and lytic plaques (Fig 6G). Together, these results demonstrate that the repression of ActA is required for the acquisition and maintenance of the VBNC state during long-term infection of host cells by L. monocytogenes.

Evidence for the emergence of VBNC bacteria during prolonged infection of epithelial cells with wild-type L. monocytogenes
Similarly to ΔactA mutants, we speculated that wild-type strains of L. monocytogenes also enter a VBNC state during long-term intracellular infection. Since cellular subculturing induced the reactivation of wild-type L. monocytogenes, propagation of cells was avoided and the number of intracellular bacteria was estimated after a three-day infection, both by CFU counts and by microscopy. The number of EGDe bacteria found in HepG2 cells as evaluated by CFUs was 7-fold lower than that determined by microscopy. The discrepancy between these two quantification methods was even larger during the infection of JEG3 cells by 10403S (i.e. on the order of 2 logs) ( Fig 6H). These data support the hypothesis that LisCVs contain bacteria in different physiological states, including bacteria that grew on plates, bacteria that have reached a VBNC state and dead bacteria.

Gentamicin selects persistent forms of L. monocytogenes
The long-term infection protocol described above made use of a relatively high concentration of gentamicin (i.e. 25 μg/mL, "Genta-25") in the culture medium, in order to avoid the extracellular survival of bacteria and re-internalization events, as well as to limit the cytoxicity of the infection. However, it was possible that at this concentration, gentamicin accumulated in epithelial cells and reduced the survival of intracellular bacteria, as previously reported in macrophages [53]. This hypothesis was examined in JEG3 cells infected with L. monocytogenes 10403S without ("Genta-0") or with lower concentrations of extracellular gentamicin (1 μg/ mL, "Genta-1" or 5 μg/mL, "Genta-5"). In addition, we used a low MOI (0.1), which enables to skip the 10-minute pretreatment of cells with gentamicin 100 μg/mL at the onset of infection. As shown in Fig 7A, the lower the antibiotic dose, the more bacteria were present in the extracellular medium and the more the viability of host cells was altered. Infection reduced cell viability by 100% with Genta-0, 95% with Genta-1, 84% with Genta-5, but only 18% with Genta-25 (Fig 7A). At Genta-1 or Genta-5, cells that survived infection contained more bacteria than those grown with Genta-25, but these bacteria were either actin-free in the cytosol or in LisCVs, indicating that the phenotypic switch occurred independently of the concentration of gentamicin (Fig 7B). Similar results were obtained with EGDe (S11A Fig) and the switch was associated with the loss of ActA (S11B Fig). We performed a quantitative assessment of the effect of decreasing gentamicin concentration from 25 μg/mL to 5 μg/mL on the number and phenotypes of bacteria at 72h pi. There were about 5-fold more bacteria in cells cultured with Genta-5 than with Genta-25 (Fig 7C), indicating that at a high dose, gentamicin penetrated within cells and affected the growth of intracellular bacteria. However, the LisCV phenomenon was not due to the presence of a high concentration of gentamicin since with the low dose (Genta-5), there were 4-fold more LisCVs  Student t-test). G. Emergence of VBNC bacteria during the subculture of 10403S-ΔactA-infected cells grown in 5 or 25 μg/mL gentamicin (as in Fig 6). Infected cells were propagated for 13 days with passages at d3, d6 and d10. Cell lysates were plated before each passage and at d13. Each dot represents the number of bacteria forming colonies (CFU) in the lysate of a well. nd: not detectable. The results are from triplicate experiments (two wells per experiment).
https://doi.org/10.1371/journal.ppat.1006734.g007 than with the high dose (Genta-25) (Fig 7D). Interestingly, the number of bacteria per LisCV was not significantly affected by a high concentration of gentamicin (Fig 7E), which suggested that bacteria that resided in these vacuoles were protected from the action of gentamicin. Moreover, the proportion of vacuolar bacteria was higher with Genta-25 (79%) than with Genta-5 (56%), while the proportion of actin-free cytosolic Listeria was conversely lower (Fig 7F).
Altogether, these results show that during prolonged infections with L. monocytogenes, gentamicin controls the growth of cytosolic bacteria but is not responsible for the formation of LisCVs. These results also suggest that gentamicin selects for vacuolar bacteria by inhibiting the growth of bacteria that replicate in the cytosol of host cells.
We hypothesized that a high dose of gentamicin also promotes the selection of VBNC bacteria during the subculture of cells infected with actA-deficient bacteria. The subculture experiment described in Fig 6 was therefore reproduced with strain 10403S-ΔactA, this time propagating the infected cells in the presence of Genta-5 or Genta-25. As shown in Fig 7G, the VBNC phenotype still occurred despite the decrease in the concentration of gentamicin at 5 μg/mL, but in a less reproducible manner than with 25 μg/mL gentamicin. There was an inter-experimental and inter-well variation in the ability of actA-deficient L. monocytogenes to enter the VBNC state, when host cells were subcultured in presence of Genta-5. At day 13, the VBNC phenotype appeared in 4 out of 6 experiments, whereas in 2 of these experiments, the bacteria did not enter the VBNC state ( Fig 7G). Accordingly, JEG3 cells propagated with Genta-5 contained either LAMP1 + bacteria that did not form colonies on BHI plates, as in Genta-25 (S12A Fig), or LAMP1-negative bacteria that massively replicated in the cytosol and were culturable when plated on BHI agar plates (S12B Fig). Taken together, these results indicate that a high concentration of gentamicin does not cause, but selects, the vacuolar and VBNC phenotypes, by specifically inhibiting the growth of cytosolic bacteria.

Discussion
This study presents the first in vitro model of intracellular persistence of L. monocytogenes in human cells, and particularly describes a vacuolar phase in the infectious process of L. monocytogenes in hepatocytes and placental cells (Fig 8). During intercellular dissemination, bacteria non-synchronously stop producing ActA and polymerizing actin. Actin-free cytosolic bacteria are engulfed in lysosome-like vacuoles, where they enter a slow/non-replicative state. Bacteria have the capacity to exit this resting phase when the host cells are subcultured. However, if the ActA deficiency is maintained, a bacterial population survives in a quiescent state known as VBNC. The antibiotic gentamicin selects vacuolar persistent forms of L. monocytogenes by eliminating cytosolic bacteria. This in vitro model should be valuable to study the molecular determinants of the asymptomatic carriage L. monocytogenes in epithelial tissues and to establish tests for the detection of dormant Listeria in clinical samples.

Listeria life in vacuoles
Several species of bacterial pathogens proliferate in the cytosol of mammalian cells and use actin-based motility to propagate from cell to cell [8]. L. monocytogenes is one of the most studied species of cytosolic bacterial pathogens. However, the intracellular lifecycle of this bacterium needs to be revisited in light of studies showing that L. monocytogenes evolved means of living in vacuoles. The initially described mechanism takes place in murine macrophages, where a subset of bacteria remains confined in non-degradative phagosomes, known as SLAPs [20,21,39]. Here, we report another mechanism leading to the formation of Listeria-containing vacuoles (LisCVs), which could represent an adaptive strategy for this pathogen to establish long-term quiescent infections in epithelial tissues. In contrast to SLAPs, which are coupled to phagocytosis at the onset of infection, LisCVs are formed at a later stage (48h-72h) following the engulfment of actin-free bacteria by endomembranes. In addition, SLAPs are not acidic compartments, in contrast to LisCVs that display lysosomal features.
The segregation of cytosolic bacteria into LisCVs may occur by a mechanism related to xenophagy, a selective autophagy process that restricts the growth of intracellular microbes [44,54]. Intracellular pathogens have evolved several strategies to bypass or subvert xenophagy and proliferate in host cells [54][55][56][57][58][59]. In particular, L. monocytogenes uses various mechanisms to evade killing by xenophagy at the early onset of an intracellular infection, such as manipulation of the mTOR pathway [54], stalling of the phagophores by bacterial phospholipases [15,60] and camouflage of the bacteria by surface proteins, particularly by ActA [14,15,38]. While xenophagy avoidance is an important means for L. monocytogenes to promote its intracellular growth, hijacking the host autophagy machinery might also promote intracellular persistence,  (1), bacteria escape into the cytoplasm, multiply and induce expression of the actin-polymerizing factor ActA (2). Actin recruitment and polymerization promotes bacterial motility (3) and cellto-cell spread via the generation of membrane protrusions from the primary infected cell to neighbor cells (4). After resolution of the protrusions (5), bacteria are in double-membrane secondary vacuoles, from which they escape and start a new cycle of infection (6). 7-8. The phenotypic switch. During the dissemination stage, bacteria stop producing ActA by an unknown asynchronous mechanism. ActAfree cytosolic bacteria (beige bacteria) multiply in the cytosol (7). A xenophagy-like process captures actin-free cytosolic bacteria into tertiary Listeria-containing vacuoles (LisCV) (8). 9. The LisCV stage. LisCVs are lysosome-like compartments, in which subpopulations of bacteria resist stress and degradation and enter a slow/non-replicative state (dark orange bacteria), others are sensitive to stress and die (white bacteria with a "*"), and others enter a VBNC state (yellow bacteria). 10-11. The reactivation stage. Unidentified stimuli induce reactivation of bacteria, which exit from the vacuoles (10). Production of ActA at the bacterial surface re-initiates a novel cycle of actin-polymerization and spreading (11). 12-14. Behavior of ActA-deficient bacteria. actA mutants escape from the primary vacuole, replicate in the cytosol (12) and are captured by a xenophagy-like process (13). In LisCVs, bacteria enter a slow/non-replicative state (dark orange), or die (*) or enter a VBNC state (yellow) (14). 15. The dormant stage. ΔactA bacteria are propagated during host cell divisions as VBNC contaminants. In absence of a reactivation signal, wild type bacteria may behave similarly to ΔactA bacteria and acquire the VBNC state. 16. Gentamicin selects vacuolar and VBNC L. monocytogenes by inhibiting the growth of cytosolic bacteria.
https://doi.org/10.1371/journal.ppat.1006734.g008 as suggested for SLAPs [39]. However, although LisCVs are formed by the entrapment of cytosolic bacteria in membrane-bound compartments, we did not find any evidence that supports a role of the canonical autophagy machinery in the formation of these vacuoles. It is interesting to note that the cytosolic pathogen Mycobacterium marinum can also be captured in vacuoles with lysosomal characteristics by a mechanism independent of canonical autophagy [61]. LisCVs could therefore be formed through an uncharacterized immune process that allows host cells to sequester microbial invaders from the cytosol. Future work will aim at characterizing the mechanism of LisCVs formation as well as its relevance to other infectious diseases.

Emergence of VBNC L. monocytogenes during intracellular infection
LisCVs contain mixed populations of stress-sensitive and stress-resistant bacteria, as evidenced by the coexistence of damaged and intact bacteria. The latter survive the harsh conditions of lysosome-like organelles, like the intravacuolar pathogen Coxiella burnetii [62]. This resistance probably results from the ability of L. monocytogenes to withstand various stresses, including low pH conditions [63]. Many stresses trigger the VBNC state of bacteria [50]. Here, we provide the first evidence that L. monocytogenes can enter a VBNC state during infection, a property that had only been observed for Listeria in environmental conditions [64][65][66][67]. This conclusion is based on the phenotype of actA mutants, which model the behavior of Listeria unable to switch back to a motile state. L. monocytogenes VBNC bacteria may represent a dormant form of intracellular parasites that could be passively propagated during mitosis or by other mechanisms, such as direct cytosolic transfer between two adjacent cells [68]. Importantly, wild type L. monocytogenes retain the ability to exit from LisCVs and return to the active proliferation and dissemination phase upon subculture of host cells. We suggest that this phenotypic switching allows L. monocytogenes to disseminate or hide in tissues during the acute or asymptomatic phases of the infection, respectively. The identification of the molecular determinants required for the transition to the VBNC state during infection could lead to the development of therapeutics that target asymptomatic infections.
Antibiotic tolerance and pathogenic potential of persistent forms of L. monocytogenes Vacuolar pathogens may persist in the host for long periods, sometimes for the entire life of the host. For instance, the persistence of Mycobacterium tuberculosis in lungs, designated as "latency" or "dormancy", is responsible for the silent carriage of this pathogen by nearly two billion people worldwide [69,70]. L. monocytogenes is not described as a pathogen leading to subclinical infections, but it cannot be excluded that this bacterium might hide and persist in tissues. It has been shown that high-risk factors for sporadic non-perinatal listeriosis are mainly immunosuppressive diseases rather than the ingestion of particular foods [71]. In light of our results, the reactivation of dormant intracellular L. monocytogenes, long after the ingestion of contaminated foods, should be examined as a potential cause of sporadic listeriosis. The existence of a dormant subpopulation of L. monocytogenes might also be involved in recurrent listeriosis [72][73][74] and recurrent spontaneous miscarriages during the first weeks of pregnancy [75]. Moreover, not only silent infection may affect humans, but also animals. The phenotype described here may have relevance for animal reservoirs that could be ignored because of an asymptomatic intracellular carriage of L. monocytogenes.
Entry into an intracellular persistent state may promote tolerance of L. monocytogenes to antibiotic treatment during infection, as shown for several bacterial pathogens [76,77]. The occurrence of non-replicating bacteria that survive exposure to antibiotics is proposed to account for the ability of pathogens to cause difficult-to-treat infections. Here, we found that gentamicin alters the intracellular growth of cytosolic L. monocytogenes but does not seem to act on vacuolar Listeria. The acidic LisCV could be a niche where Listeria is protected against gentamicin and other antibiotics. It is of note that aminoglycosides, such as gentamicin, are almost devoid of activity at the acidic pH of lysosomes [78]. Bacteria found within LisCVs might also share characteristics with small-colony variants, which are tolerant to aminoglycosides [79], and/or with Listeria persisters that emerge during bacterial growth in cell-free broth supplemented with bactericidal antibiotics [80]. The results presented here should have a significant clinical impact since the primary treatment for listeriosis consists of a combination regimen of gentamicin and ampicillin [81]. Beside tolerance to antibiotics, it is possible that cells carrying dormant L. monocytogenes are not effectively presenting Listeria antigens and are not efficiently recognized and destroyed by immune cells.
The physiological relevance of vacuolar forms of L. monocytogenes in epithelial tissues now needs to be demonstrated by in vivo studies. It is worth mentioning that hepatocytes of mice infected for 24h with L. monocytogenes mostly contain actin-free cytosolic bacteria [82], which could be in the early process of being sequestered in LisCVs. However, due to the low internalization rate of Listeria in murine epithelial cells [24] and to differences in mucosal immunity between mice and humans [25], the murine model of listeriosis may not be optimal for investigating the persistence of L. monocytogenes in epithelial tissues. Moreover, asymptomatic persistent forms of Listeria are likely to be in a very low number in tissues and difficult to observe. In conclusion, there is a need to develop animal models and easy-to-use diagnostic methods for the detection and study of VBNC Listeria in tissue samples. The possibility that L. monocytogenes persistent forms could be tolerant to antibiotics and undetectable by routine culture methods would clearly be a concern for public health safety assessment.

Ethics statement
Primary human hepatocytes were purchased from Life Technologies. These liver cells were derived from tissue obtained from accredited institutions. These institutions obtained consent from the donors for use of the tissue and its derivative for research purpose.

Bacterial infections, bacterial viability and LysoTracker assays
Human cells were seeded into 6-or 24-well plates (with or without coverslips) 2-4 days prior infection in order to reach 50-80% confluency (for all cell lines except JEG3, and for transfection assays) or 90-100% confluency (JEG3 monolayer assays). A detail protocol for studying LisCVs in JEG3 cell monolayers is available in protocol.io (dx.doi.org/10.17504/protocols.io. kwmcxc6). Bacterial colonies from freshly streaked agar plates were cultured in BHI broth and incubated overnight at 37˚C in a shaking incubator. Stationary-phase bacteria were washed once in PBS and diluted in culture medium without serum. After a washing step, cells were infected with bacteria at a multiplicity of infection (MOI) of 0.1 to 5 and centrifuged 2 min at 300g to synchronize the bacterial uptake. Cells were infected for 1h. When using MOI of 1-5, cells were exposed to a 10-min treatment with a medium containing 100 μg/ml gentamicin, which ensures that extracellular bacteria are rapidly killed. For the lower MOI of 0.1, this step can be skipped and replaced by two washing steps. Cells were washed once and subsequently incubated in complete medium with 25 μg/ml gentamicin. When indicated, lower concentrations of gentamicin were applied (0, 1 or 5 μg/ml). Cells were processed for immunofluorescence (see below) or lysed at various time points using cold distilled water. Viable bacterial counts of intracellular bacteria were determined by plating serial dilutions onto BHI agar and numbering colony-forming units (CFU). Cells were enumerated in parallel wells following trypsin detachment and trypan-blue staining. To distinguish live bacteria with intact membranes from dead bacteria with compromised membranes, the Live/Dead BacLight Bacterial Viability kit (Life Technologies) was used with a protocol adapted from [47] and [46] with some modifications. JEG3 cells infected with bacteria in 6-well plates containing coverslips were gently washed twice in MOPS/MgCl 2 (0.

Epifluorescence microscopy and quantifications
Cells containing coverslips were fixed in 4% paraformaldehyde (PFA) in 1X PBS for 30 min at room temperature, gently washed in 1X PBS, incubated in blocking solution (2% BSA in PBS, pH 7.4), or in methanol for 10 min on ice when using the cathepsin D antibody, permeabilized using 0.4% Triton X-100 in PBS for 4 min, washed three time in PBS and processed for immunofluorescence with the indicated antibody diluted in 2% BSA. Fluorescent phalloidin and DAPI (or Hoechst) were added with the secondary antibodies to label F-actin and nuclei, respectively. To discriminate intracellular from extracellular bacteria, a dual staining was performed, whereby extracellular bacteria were stained with anti-Listeria antibodies prior to cell permeabilization. Samples were mounted on glass coverslips and analyzed with fluorescent microscopes (Carl Zeiss Axiovert 135, AxioObserver.Z1, KEYENCE BZ-X710 or Yokogawa CSU-X1 spinning disk confocal system). Images were acquired with a 10x non-immersion objective, or 40x, 63x or 100x oil immersion objectives, and images were processed with Zen (Carl Zeiss), MetaMorph (Universal Imaging) or Image J softwares. For time-lapse microscopy, JEG3 cells were seeded in a 35mm dish (MatTek) and infected with 10403S-mCherry bacteria (MOI 0.1). Imaging was performed in media without phenol red containing ProLong Antifade Reagents for Live Cells (Life Technologies) and using the KEYENCE fluorescent microscope BZ-X710 equipped with an incubation chamber (37˚C and 5% CO 2 ). Images were acquired with a non-immersion 40x objective. To quantify the number of bacteria that associated with the indicated markers, as well as the number of nuclei, 10 to 50 microscopic fields were examined from coverslips of at least three independent experiments. Due to the frequent overlap of bacterial immunofluorescence signals, the number of bacteria per infected cell was estimated by the ratio (T/I)/C, where "T" is the total immunofluorescence signal of the area occupied by bacteria, "I" is mean intensity values of representative single bacteria and "C" is number of cells/nuclei is the area of interest.

Cell sorting and subculturing of infected cells
For the FACS assays, HepG2 cells infected 3 days with GFP-expressing EGDe bacteria were trypsinized, washed once in PBS, resuspended in DMEM and subjected to cell sorting using a FACSAria II Cell Sorter (BD Bioscience). For each of three independent experiments, about 5x10 5 GFP-positive cells were collected and seeded onto polylysine-coated 6-well plates (1x10 5 cells per well) in complete medium containing 25 μg/ml gentamicin and were examined 1h, 8h or 3 days after FACS. For the propagation assays using trypsin detachment, cells were dissociated with trypsin (200 μl of Trypsin-EDTA solution, Life Technologies), counted, diluted 5-or 10-fold in complete medium supplemented with gentamicin 25 μg/ml and seeded in 6-well plates (containing or not coverslips). Cells were then incubated for the indicated time and either fixed to proceed to immunofluorescence assays, lysed with distilled water for CFU counting, or detached and subcultured again. It is of note that trypsinization and flow sorting may select for cell subpopulations. This is a limit of this assay.

Transmission electron microscopy
First analytic experiments were performed at the Ultrapole platform of Institut Pasteur, as follows. Cells were fixed overnight with 2.5% Glutaraldehyde in 0.1M sodium cacodylate buffer at pH 7.2. Then samples were rinsed in 0.1M sodium cacodylate (pH 7.2) and postfixed in a 1% osmium tetroxyde at room temperature with the same buffer for 1h. Samples were rinsed with water and dehydrated with a graded ethanol concentrations (10-100%), followed by a mixture of ethanol and embedded in an epoxy resin. Ultrathin sections (50-60 nm) were performed with an ultramicrotome « Ultracut UC7 » (Leica Microsystems, Vienna, Austria), stained with uranyl acetate and Reynold's lead citrate, and then observed with a tecnai T12 (FEI company) at 100-kV accelerating voltage. Images were recorded using us 4000 gatan camera and digital micrograph software or Eagle camera and tia software. Quantitative experiments were performed at the MIMA2 platform (Jouy-en-Josas) or ImagGif platform (Gif-sur-Yvette), with similar procedures and the following modifications: cells were fixed overnight with 2% Glutaraldehyde, counter stained on block with 0.5% OTE (Oolong Tea Extract) in 0.1M sodium cacodylate buffer and post-fixed in a solution of containing 1% osmium tetroxyde and 1.5% potassium ferrocyanate; samples were rinsed with water and dehydrated with a graded 30-100% ethanol concentrations. Ultrathin sections were observed with Hitachi HT7700 or Jeol 1400 (Jeol-Allemagne) at 80-kV accelerating voltage.  experiments). B. A representative image of the immunolabeling of LC3 and Listeria at 72h p.i. C. JEG3 cells expressing GFP-LC3 were infected with Listeria EGDe and processed for immunofluorescence assays with Listeria and LAMP1 antibodies. The arrows point regions, which are shown at a higher magnification below. Bars: 10 μm and 1 μm. D-F JEG3 cells were infected with 10403S bacteria, exposed to two successive siRNA treatments with ATG7, BECN1 or control siRNAs, one at 24h p.i., and one at 48h p.i. At 72h p.i., samples were processed for transcript quantification (D), or immunofluorescence assays with LAMP1 and Listeria antibodies and DAPI (E), or CFU and JEG3 cell counting (F). In D, the fold change of transcript level is relative to values normalized to GAPDH reference gene. YWHAZ was used as a control gene.  Fig  5) were detached with trypsin, diluted to 2 x 10 5 cell/mL and grown for 24h in complete medium with gentamicin 25μg/mL. The micrograph, which is representative of 5 experiments, is an overlay image of cells stained with Listeria and LAMP1 antibodies, fluorescent phalloidin to label F-actin and Hoechst to mark nuclei. Bar: 50 μm. Examples of cells in each of the two populations, either carrying persistent vacuolar Listeria ("P") or carrying active Actin-associated Listeria ("A"), are framed and represented at a higher magnification below. Bar: 5 μm.