Recruitment of both the ESCRT and autophagic machineries to ejecting Mycobacterium marinum

Cytosolic Mycobacterium marinum are ejected from host cells such as macrophages or the amoeba Dictyostelium discoideum in a non‐lytic fashion. As described previously, the autophagic machinery is recruited to ejecting bacteria and supports host cell integrity during egress. Here, we show that the ESCRT machinery is also recruited to ejecting bacteria, partially dependent on an intact autophagic pathway. As such, the AAA‐ATPase Vps4 shows a distinct localization at the ejectosome structure in comparison to fluorescently tagged Vps32, Tsg101 and Alix. Along the bacterium engaged in ejection, ESCRT and the autophagic component Atg8 show partial colocalization. We hypothesize that both, the ESCRT and autophagic machinery localize to the bacterium as part of a membrane damage response, as well as part of a “frustrated autophagosome" that is unable to engulf the ejecting bacterium.


| INTRODUC TI ON
Pathogen egress from a host cell needs to be an orchestrated and tightly controlled process to guarantee successful maintenance and spreading of an infection (Flieger et al., 2018;Friedrich et al., 2012).
As summarized in Flieger et al. (2018) exit strategies include both lytic and non-lytic processes.Often these are based on a blend of virulence strategies and host responses with unclear molecular mechanisms.Importantly, intracellular pathogens can invoke membranolytic stresses, which must be confronted with the hosts membrane repair machineries.
Mycobacterium marinum, like its close relative Mycobacterium tuberculosis, is taken up by professional phagocytes and then establishes a replication-permissive compartment.Both species are able to escape this compartment by rupturing its membrane and subsequently reside within the host cell cytosol (Hagedorn & Soldati, 2007;Houben et al., 2012;Simeone et al., 2012;van der Wel et al., 2007).The genetically tractable amoeba Dictyostelium discoideum, a professional phagocyte, is well established as a useful model system to dissect the basic interactions between host cells and their pathogens, such as M. marinum (Barisch & Soldati, 2017;Hagedorn & Soldati, 2007;Koliwer-Brandl et al., 2019).Using this system, we previously found that after translocation into the host cytosol, M. marinum can escape from the cell by locally rupturing the plasma membrane, without host cell lysis (Gerstenmaier et al., 2015;Hagedorn et al., 2009).This egress occurs through a barrel-shaped F-actin structure, called the ejectosome.In our previous study (Gerstenmaier et al., 2015) we demonstrated that the autophagic machinery is specifically recruited to ejecting bacteria and accompanies them on their way outwards.Closer analyses showed that an autophagic membrane accumulates at the distal pole of ejecting bacteria and protects the host cell from cytosolic leakage and lysis during the egress process.This requires the canonical autophagy machinery, including functional Atg1 to initiate the process, ubiquitination of the target site and recruitment of Atg8-positive membranes via adaptor proteins, such as p62.In D. discoideum, two paralogs of Atg8 can be found, Atg8a and Atg8b, both of which associate with autophagosomes but with distinct functions (Meßling et al., 2017).While Atg8b seems to colocalize with nascent autophagosomes, Atg8a, the isoform used in this study, appears later on autophagosomes and seems to resemble the mammalian GABARAP (GABA type A receptor-associated protein) family.
Ubiquitination is also found in the context of the invagination of intralumenal vesicles required to form multivesicular bodies, where it recruits the Endosomal Sorting Complexes Required for Transport (ESCRT) machinery (Wollert & Hurley, 2010).ESCRT is an evolutionary conserved multi-protein complex that is involved in membrane remodeling and also plays a role in membrane damage repair (Jimenez et al., 2014), as well as mycobacterial infection (Lopez-Jimenez et al., 2018;Mehra et al., 2013;Mittal et al., 2018;Philips et al., 2008).Four protein complexes (ESCRT-0, -I, -II, -III), as well as the AAA-ATPase Vps4 and multiple accessory proteins, such as Alix, make up the ESCRT machinery.Membrane deformation is driven by ESCRT-III components, including Vps32, and membrane scission, the final step is catalyzed by the AAA-ATPase Vps4 (Vietri et al., 2020).
ESCRT has been shown to repair small (less than 100 nm) wounds in the plasma membrane, possibly by membrane extrusion (Jimenez et al., 2014, Lopez-Jimenez et al. 2018).Interactions with ESCRT have been reported for many pathogens, for example, in Drosophila cells ESCRT has been shown to restrict M. smegmatis growth (Philips et al. 2008).The secreted virulence factor EsxH from M. tuberculosis directly binds and inhibits the ESCRT machinery in mammalian macrophages (Portal-Celhay et al., 2016).During infection of epithelial cells with Candida albicans membrane damage by candidalysin, a pore forming toxin, is repaired by ESCRTIII-dependent membrane extrusion (Westman et al., 2022).Finally, both the ESCRT and autophagic machinery have been shown to be recruited when M. marinum breaks its replication compartment, though they appear to localize in a spatially and temporally independent manner from each other (Lopez-Jimenez et al., 2018).Whilst they may serve distinct roles in bacterial compartment rupture, work in nematodes, flies and mammals indicates that ESCRT and autophagy may be functionally linked (summarized in Rusten & Stenmark, 2009) with recent work showing that ESCRT-III mediates phagophore closure (Takahashi et al., 2018;Zhen et al., 2020).Furthermore, Raykov et al. recently identified an E3-Ligase, TrafE, to function at the intersection of the ESCRT and autophagic pathway (Raykov et al., 2021).
In this study, we investigate the roles and interactions between ESCRT and the autophagic machinery during Mycobacterial ejection.
We find that in addition to the autophagic machinery (Gerstenmaier et al., 2015), ESCRT components localize to distinct regions of the ejecting bacteria partially dependent on an intact autophagic pathway.Therefore, multiple cellular pathways appear to respond in concert, as well as independently from each other to the bacterium breaking through the plasma membrane.

| ESCRT is recruited to bacteria engaged in ejection
To monitor whether the ESCRT machinery is recruited to ejecting bacteria, Dictyostelium cells expressing the ESCRT-I component GFP-Tsg101, the ESCRT-III effector GFP-Vps32, the AAA-ATPase Vps4-GFP or the ESCRT-associated protein Alix were infected with wildtype M. marinum.The F-actin cytoskeleton was visualized with fluorescently labeled phalloidin and bacteria engaged in ejection were imaged.Fluorescence microscopy analysis revealed that, similarly to the autophagic marker Atg8 (Figure 1a), each ESCRT component accumulated at ejecting bacteria (Figure 1b-e).Quantitative analysis (Figure 1f) showed, that accumulation of the visualized ESCRT proteins at ejecting bacteria was frequent (61%-100%) with the Vps4 protein being present at all analyzed ejectosomes.Alix, Tsg101, and Vps32 appeared to localize around the distal pole, reminiscent of Atg8 localization, or along the distal part of the bacterium.In contrast, Vps4 was most of the time concentrated closely at or even within the F-actin structure of the ejectosome (Figure 1d).By analyzing the fluorescence intensity along the bacterial axis (Figure 1e,g,h), the accumulation of signals along the bacterium relative to the F-actin signal of the ejectosome was mapped (Figure 1i).The comparison of the distance between ejectosome and marker signal confirmed that the localization of Vps4GFP close to the ejectosome structure was significantly different to the other investigated ESCRT components (Figure 1i).

| ESCRT and Atg8 recruitment to ejecting bacteria partially depend on each other
The autophagy and ESCRT machineries can interact on numerous levels (Raykov et al., 2021;Rusten & Stenmark, 2009) and both are known to play important roles in mycobacterial infection (Mehra et al., 2013;Mittal et al., 2018;Philips et al., 2008).We therefore decided to investigate the interplay between these dynamic complexes at the ejectosome.
The serine/threonine kinase Atg1 (human ULK1) is an essential regulator of autophagy induction.Therefore, to monitor if the localization of ESCRT components to ejecting bacteria depends on a functional autophagic process, we infected atg1 knock-out cells that express the fluorescently tagged ESCRT proteins described above with M. marinum.The accumulation of the ESCRT proteins at ejecting bacteria was quantified by fluorescence microscopy (Figure 2a) and compared to the accumulation in wild-type cells (Figure 1f).In general, the fusion proteins were present at lower frequency in the atg1-null strain, although this was only statistically significant for Alix and Vps4.This suggests that, in part, ESCRT recruitment appears to be dependent on an intact autophagy pathway.
Next, we examined ejectosome formation and GFP-Atg8 accumulation in cells that either have a disrupted ESCRT machinery or disruptions in both autophagy and ESCRT pathways.We found that in tsg101-null cells, the number of ejectosomes at 24 hpi is significantly reduced in comparison to wild-type cells (Figure S1).This phenotype was suppressed when atg1 was also disrupted (tsg101-null; atg1-null), with ejectosomes formed as efficiently as in wild-type cells (Figure S1) at 24 hpi, although lacking Atg8 recruitment.In addition, although half of ejecting bacteria were associated with Atg8 in wild-type cells, only 23% showed accumulation in cells lacking the ESCRT component Tsg101 (Figure 2b), indicating a reciprocal partial dependence between the autophagy and ESCRT machinery.

| ESCRT and Atg8 partially colocalize at ejecting bacteria
To further dissect the interplay between ESCRT and autophagy, we directly compared components of both pathways at the same time at ejecting bacteria by immunofluorescence staining for Atg8 combined with expression of fluorescently-tagged ESCRT proteins.In contrast to the other microscopy images, these analyses were performed using an Airyscan equipped LSM which allowed for higher resolution.As observed above, localization of both markers was restricted to the intracellular part of the bacteria while the bacteria were transiting through the ejectosome actin barrel structure Vps4 (as indicated by the small white arrow in Figure 2e″), which shows a gap that is filled by Atg8 signal.
Overall, we observe that ESCRT components have distinct localisations at the ejectosome and along the intracellular part of the bacterium, and only partially colocalize with autophagic markers.This is in accordance with our functional recruitment observations that show partial dependence on each other.

| Disruption of ESCRT severely reduces bacterial transmission
In our previous study, we showed that while the autophagic machinery is dispensable for efficient ejectosome formation, disruption of autophagy strongly impaired cell-to-cell transmission of bacteria (Gerstenmaier et al., 2015).We therefore determined how disruption of ESCRT impacted cell-to-cell transmission of bacteria over the first 32 h of infection.
Previously, the presence of bacteria in the acceptor strain was quantified by fluorescence microscopy.In order to increase the throughput, the assay was adapted to flow cytometry.The principle of this transmission assay is depicted in Figure 3a.In short, donor  The transmission rate was measured at 6 hpi as a starting point followed by 24, 28 and 32 hpi (Figure 3b).At 6 hpi, all strains show a transmission rate of "1" because at this time the acceptor strain is added and ratios of strains are adjusted to each other.Over time, transmission rates increased and were highest for the wild-type donor cells.As observed before, transmission rates from autophagy deficient (atg1-null) host cells was significantly reduced at all following timepoints.Strikingly, the tsg101-null mutant showed an almost complete lack of transmission at 24 hpi but recovered slightly at 32 hpi to about half the rate of the atg1-null mutant.A double mutant lacking Atg1 and Tsg101 exhibited lower transmission rates than wild-type donors, very similar to the atg1 mutant.No change over time was observed for the double mutant.This is consistent with our quantification of ejectosome numbers above, indicating a key role for ESCRT in bacterial transmission-but only in the presence of an intact autophagy machinery.
Interestingly, for all mutants we observe a slight drop in the transmission rate at 28hpi.It is not clear why this happens but there are several scenarios that can lead to fluctuations in the transmission ratio.For example, the division of infected donor-cells spreads the infection into daughter cells thereby increasing the number of infected donor cells and lowering slightly the infection rate.In addition, extracellular bacteria, which will be released by dying cells late in the infection cycle, will also be taken up by non-infected donor cells changing the ratio.

| DISCUSS ION
With this study, we expand our previous report in which we showed that the autophagic machinery is recruited to ejecting bacteria.
Here, we present evidence that components of the ESCRT-complex also accumulate along the intracellular part of bacteria engaged in ejection.During ejection, bacteria breach the integrity of the host's plasma membrane on their way outwards, thereby introducing large damage or wounds to the cell (Hagedorn et al., 2009).Topologically, the ejectosome wound represents a special situation with the damaging force pushing against the membrane from the intracellular side.Based on our observation in this and our previous studies, we hypothesize that multiple protective mechanisms respond in a concerted manner to this damage, thereby allowing the bacterium to pass through the plasma membrane and at the same time the host cell to survive bacterial egress (Gerstenmaier et al., 2015;Hagedorn et al., 2009).
As we have shown previously (Gerstenmaier et al., 2015), the autophagic machinery recruited to the distal pole of ejecting bacteria plays a role in keeping the ejectosome tight.Among many functions described for the ESCRT complex, one is to facilitate the final step in autophagosome genesis, the closure of the phagophore.Vps37a provides the link between the two processes and appears to be responsible for the recruitment of ESCRT to the closing phagophore (Takahashi et al., 2019).Here we observe that the ESCRT machinery also localizes to ejecting bacteria and that this is to a high degree dependent on a functional autophagic pathway.
Based on these observations we hypothesize that these two complexes represent a "frustrated autophagosome" with ESCRT being unable to perform the final sealing step, because the bacterium is "clamped" between intra-and extracellular space by the ejectosome structure (see model in Figure 4).Once the bacterium reaches the final steps of ejection, the stalled autophagosome is capable to close with ESCRT catalyzing the closure step.Vps4 is usually closely localized at the ejectosome structure (Figure 1i), possibly at the plasma membrane that often folds inwards around the ejecting bacterium (Hagedorn et al., 2009).It appears that Vps4 is ready to perform the final sealing step upon bacterial exit followed by the disassembly of the ESCRT components.It remains to be determined how Vps4 is localized and held at this position.However, it has been shown that the Vps4 recruitment to sites of membrane damage is dependent on the E3-ligase TrafE in D. discoideum (Raykov et al., 2021).Overall, the autophagosome structure accompanying the ejecting bacterium likely prevents leakage from the cell, in a so far unknown mechanism related to fusion.We observe that ESCRT components can also localize to ejecting bacteria independently of autophagy.We speculate that this is also due to membrane damage.Ejectosome formation depends on the secretion of the mycobacterial factor ESAT-6 (Hagedorn et al., 2009), a membranolytic effector common to pathogenic mycobacteria (Mittal et al., 2018).A mycobacterial mutant lacking ESAT-6 and its secretion system (delta RD1mutant), does not form ejectosomes but can be trans-complemented in Dictyostelium by cytosolic expression of the mycobacterial factor (Hagedorn et al., 2009).In addition, the continuous secretion of ESAT-6 by the immobilized, ejecting bacteria might damage not only the plasma membrane but also endomembranes in its vicinity, such as lysosomes.Subsequently, the ESCRT complex might respond in its role to seal small membrane damages.The mechanism, by which ESCRT components are recruited remains to be determined.However, ubiquitination, which is known to recruit Tsg101, has also been found on cytosolic M. marinum, as well as bacteria engaged in ejection (Gerstenmaier et al., 2015).Overall, the recruitment of the two major membrane remodeling and sealing machineries is in large parts reminiscent of the events that occur earlier in infection, when mycobacteria are attempting to exit their vacuole into the host cell cytosol.

F I G U R E
Based upon membranolytic pressure from the bacterium, autophagic and ESCRT components are recruited in a concerted but independent manner to the breaking vacuole (Lopez-Jimenez et al., 2018).Often, the vacuole breaks under pressure and bacteria are observed in the host cytosol (Hagedorn & Soldati, 2007;Lopez-Jimenez et al., 2018).However, there are also significant differences between the vacuolar escape and ejection scenarios.
In addition to the topological distinction, the bacterium is immobilized in the ejectosome which allows the cellular damage response machineries to accumulate on the bacterium.Furthermore, the reorganization of the actin cytoskeleton into the barrel-shaped ejectosome could provide an additional mechanism to restrict plasma membrane damage and facilitate its repair (by locally lowering the membrane tension).
Finally, in our system, functional ejectosomes are the major pathway of cell-to-cell transmission (Hagedorn et al., 2009).Flow cytometry-based measurements confirmed our previous observations that a non-functional autophagy pathway leads to reduced transmission (previously observed in Gerstenmaier et al., 2015), presumably due to a lack of proper sealing, but not to a reduction in ejectosome numbers.This is likely also true for the double mutant of atg1 and tsg101, which, as shown in this study, exhibits reduced transmission with no change in ejectosome numbers in comparison to wild-type cells.However, the similarity between the single atg1 and double atg1/tsg101 mutant also highlights that the role for autophagy in the sealing process outweighs the effect of a non-functional ESCRT machinery.Close to no transmission is observed from a mutant lacking Tsg101.This is probably due to a still functional autophagic pathway, which catches cytosolic bacteria and prevents them from transmitting.This is consistent with data from Lopez-Jimenez et al., who observed that in cells lacking functional Tsg101 M. marinum is caught by the autophagic system upon vacuole rupture, leading to strong restriction of cytosolic bacteria (Lopez-Jimenez et al., 2018).Accordingly, in the double mutant atg1/ tsg101 both the number of ejectosomes (Figure S1) and transmission (Figure 3c) are restored to levels of the atg1 mutant.
Overall, the ejectosome represents a topologically remarkable wounding system with the damaging agent breaking the plasma membrane from the intracellular side.We have shown here that several wound response mechanisms are active at the site of egressing bacteria.Critical questions remain to be addressed, e.g.what triggers the responses.

| Dictyostelium cell culture
Wild-type Dictyostelium cells (Ax2) were axenically cultivated in HL5c medium (Formedium) at 22°C.The Dictyostelium atg1-null cells were previously described (King et al., 2013), atg1/tsg101-null and tsg101-null mutants were generated in the T. Soldati group (Lopez-Jimenez et al., 2018).Expression constructs for the ESCRT components were kindly provided by T. Soldati (University of Geneva, Switzerland).The expression plasmid for Atg8a was a gift from J.

(
Figure 2c-e, ejectosomes indicated by arrowhead).Closer analysis revealed that signals of Atg8 and ESCRT partially overlap (as indicated by the white arrowhead in Figure 2c′-c‴, d′-d‴).However, at the same time there are areas in which the signals appear interspaced indicating a spatial separation (indicated by white arrow in the corresponding images).This was observed most clearly with

F
I G U R E 1 ESCRT components localize to ejecting bacteria.(a-e) Fluorescence micrographs of bacteria (blue) during ejection by the ejectosome indicated with a white arrowhead.GFP-tagged Proteins Atg8 (a), Alix (b), Tsg101 (c), Vps4 (d) and Vps32 (e) (green) accumulate at the bacterium (white arrow), often at the distal pole.Insets show the green channel alone.Scale bars indicate 2 μm.(f) shows quantification of ESCRT component accumulation at the ejectosome in the indicated strains.(g) and (h) show the exemplary quantification of markers in (d) and (e) with the quantified region indicated by rectangles and the direction shown by an orange arrows.The distance between the maxima for actin and marker signals is indicated with a black bar (in g and h).For (i), ejecting bacteria that are a minimum of 1 μm intracellularly were chosen and the distances (μm) between the ejectosome (actin) and the indicated marker proteins plotted.A scheme next to the graph indicates the orientation of the ejecting bacterium and direction of travel.The light blue color indicates the various lengths of the measured bacteria.One-way ANOVA analysis with Tukeys Post test was performed using GraphPad Prism (*p ≤ 0.05, **p ≤ 0.01).F I G U R E 2 Analysis of Atg8 and ESCRT recruitment to ejectosomes in the respective mutant strains.(a) The degree of ESCRT accumulation at ejectosomes in atg1 knock-out mutants.In (b) the recruitment of Atg8 is shown in wild-type cells, as well as a tsg101-null and a tsg101/atg1 double null mutant.(*p ≤ 0.05, **p ≤ 0.001, ANOVA, Bonferroni post test; all strains compared to the parental strain Ax2, see Figure 1f).(c-e) Fluorescence micrographs of ejectosomes with the colocalization of Atg8 and Alix, Vps32 and Vps4 respectively.(c-e) A quadruple staining with bacteria indicated in blue, actin in white, Atg8 in red and the respective ESCRT component in green.Arrowheads indicate ejectosomes, arrows the pole of the ejecting bacterium.(c′, d′, e′) The ESCRT-component (magenta, arrowhead) and the ejectosome (green).In (c″, d″, e″) Atg8 (cyan) and the ejectosome (green) are shown.(c‴, d‴, e‴) the merged images of only Atg8 (cyan) and ESCRT components (magenta).In (c′-c‴), a region of colocalization of markers is indicated by a white circle.In e′-e‴, white ovals encircle a region with distinct localization for Vps4 and Atg8.Scale bars represent 2 μm.

(
mutant) host cells were infected with green fluorescent bacteria and at 6 hpi mixed with a wild-type acceptor strain expressing cytosolic RFP.Flow cytometry dot-plots therefore allow four different cell populations to be distinguished from each other by their fluorescence and shape: infected donor, non-infected donor, non-infected acceptor, as well as acceptor cells carrying bacteria.Examples are shown in Figure 3b.Quantification of these populations allows the ratio of infected acceptors per infected donor cell to be calculated-a measure for cell-to-cell transmission efficiency.

F
Quantitative comparison of bacterial transmission from atg1, tsg101, and double knock-out mutants.(a) a schematic representation of the quantitative transmission assay (modified from Hagedorn et al., 2009).Donor cells are infected with green fluorescent bacteria and at 6 hpi mixed with wild-type acceptor cells.Over time, transmission can be viewed by fluorescence microscopy or quantified using flow cytometry.Schematic flow cytometry plots for no spreading and successful spreading are shown.(b) Examples of flow cytometry patterns for wild-type, atg1-null and tsg101-null cells are shown at 6 and 32 hpi.Respective populations are indicated in the graphs.(c) Transmission coefficients (normalized number of infected acceptor cells per infected donor cells (norm.# inf.acceptor/ # inf.donor cells)) at 6, 24, 28 and 32 hpi for wild-type, atg1-null, tsg101-null and atg1/tsg101 double null mutant cells are shown (triplicates, ±SEM).Two-way ANOVA was performed in Graph Pad prism with a Bonferroni post test (n.s.not significant, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, **** p ≤0.0001).
4 A model ejectosome.A schematic representation of a bacteria (gray) immobilized in an ejectosome (red) structure.While cytosolic Mycobacterium marinum are fully or partially engulfed in Atg8-positive double-layered membranes (green), ejecting bacteria show partial localization of Atg8-positive membrane.Full engulfment of ejecting bacteria is inhibited by the ejectosome structure (black arrows) which leads to accumulation of ESCRT components (blue dots) at the intracellular part forming a frustrated autophagosome.The orange arrow indicates the direction of egress.
Furthermore, cellular signals for membrane damage, such as flagging the inner leaflet of the damaged membrane with lectins can be envisaged.Possible candidates, members of the family of discoidins, are encoded the Dictyostelium genome and might trigger the assembly of ESCRT components.Finally, as mentioned above Raykov et al. (2021) TrafE, an E3 ubiquitin ligase, has been identified as a key player in the membrane damage response also in the context of M. marinum infection.TrafE appears to act upstream of both, the autophagic as well the ESCRT recruitment and thus might be the common denominator and coordinator of these processes.Similar to the intracellular damage that is induced during M. marinum infection, TrafE could well coordinate the response at the ejectosome site.