Role of the ESCRT‐III complex in controlling integrity of the Salmonella‐containing vacuole

Intracellular pathogens need to establish specialised niches for survival and proliferation in host cells. The enteropathogen Salmonella enterica accomplishes this by extensive reorganisation of the host endosomal system deploying the SPI2‐encoded type III secretion system (SPI2‐T3SS). Fusion events of endosomal compartments with the Salmonella‐containing vacuole (SCV) form elaborate membrane networks within host cells enabling intracellular nutrition. However, which host compartments exactly are involved in this process and how the integrity of Salmonella‐modified membranes is accomplished are not fully resolved. An RNA interference knockdown screen of host factors involved in cellular logistics identified the ESCRT (endosomal sorting complex required for transport) system as important for proper formation and integrity of the SCV in infected epithelial cells. We demonstrate that subunits of the ESCRT‐III complex are specifically recruited to the SCV and membrane network. To investigate the role of ESCRT‐III for the intracellular lifestyle of Salmonella, a CHMP3 knockout cell line was generated. Infected CHMP3 knockout cells formed amorphous, bulky SCV. Salmonella within these amorphous SCV were in contact with host cell cytosol, and the attenuation of an SPI2‐T3SS‐deficient mutant strain was partially abrogated. ESCRT‐dependent endolysosomal repair mechanisms have recently been described for other intracellular pathogens, and we hypothesise that minor damages of the SCV during bacterial proliferation are repaired by the action of ESCRT‐III recruitment in Salmonella‐infected host cells.

with the Salmonella-containing vacuole (SCV) form elaborate membrane networks within host cells enabling intracellular nutrition. However, which host compartments exactly are involved in this process and how the integrity of Salmonellamodified membranes is accomplished are not fully resolved. An RNA interference knockdown screen of host factors involved in cellular logistics identified the ESCRT (endosomal sorting complex required for transport) system as important for proper formation and integrity of the SCV in infected epithelial cells. We demonstrate that subunits of the ESCRT-III complex are specifically recruited to the SCV and membrane network. To investigate the role of ESCRT-III for the intracellular lifestyle of Salmonella, a CHMP3 knockout cell line was generated. Infected CHMP3 knockout cells formed amorphous, bulky SCV. Salmonella within these amorphous SCV were in contact with host cell cytosol, and the attenuation of an SPI2-T3SS-deficient mutant strain was partially abrogated. ESCRT-dependent endolysosomal repair mechanisms have recently been described for other intracellular pathogens, and we hypothesise that minor damages of the SCV during bacterial proliferation are repaired by the action of ESCRT-III recruitment in Salmonellainfected host cells.  (Brumell & Scidmore, 2007). The enteropathogen Salmonella enterica resides in a specialised membrane-bound compartment termed Salmonella-containing vacuole (SCV), which possesses late endosomal characteristics and allows bacterial survival and proliferation (Haraga, Ohlson, & Miller, 2008). What is crucial for intracellular survival of Salmonella is the function of the type III secretion system (T3SS) encoded on Salmonella pathogenicity island II (SPI2) (Hensel et al., 1998). The SPI2-T3SS translocates a set of effector proteins across the SCV membrane into the host cytosol in order to manipulate host cell functions (Figueira & Holden, 2012).
A specific characteristic of host manipulation by Salmonella is the extensive reorganisation of the endosomal system and the formation of endosome-derived tubular structures extending from the SCV (Liss & Hensel, 2015). These Salmonella-induced tubules comprise various structures whereas the best-studied tubules are termed Salmonella-induced filaments (SIF) which are characterised by the presence of lysosome-associated membrane glycoprotein 1 (LAMP1) (Garcia-del Portillo, Zwick, Leung, & Finlay, 1993;Schroeder, Mota, & Meresse, 2011). The induced formation of an extensive SIF network by Salmonella provides bacterial nutrition enabling proliferation in cell-based models (Liss et al., 2017). The process of SIF formation is dependent on specific SPI2-T3SS effector proteins and their fine-tuned interaction with host cell proteins. The precise involvement of Salmonella effector proteins and host interaction partners is still not fully elucidated, but undoubtedly the best understood T3SS effector is SifA (Stein, Leung, Zwick, Garcia-del Portillo, & Finlay, 1996). SifA interacts with various host proteins, among them SKIP (SifA and kinesin-interacting protein) (Boucrot, Henry, Borg, Gorvel, & Meresse, 2005;Dumont et al., 2010), also referred to as PLEKHM2 (pleckstrin homology domain-containing protein family member). SifA together with SKIP and kinesin-1 forms a complex that mediates elongation of SIF along microtubules. The lack of either SifA or SKIP leads to a deficiency in SIF formation, affecting SCV integrity and resulting in bacterial release into the host cell cytosol. This leads to cytosolic hyper-replication in epithelial cells and reduced intracellular proliferation in macrophages (Beuzon et al., 2000a;Boucrot et al., 2005;Zhao et al., 2015). The reorganisation of the host endocytic pathway leads to a recruitment of endosomal compartments to the SCV. Therefore, various host endosomal marker proteins, among these predominantly Rab GTPases, were found decorating the SCV and SIF network (Brumell & Scidmore, 2007).
To identify further host cell proteins that are required for SCV and SIF biogenesis, we previously performed an RNA interference (RNAi) knockdown (k/d) screen targeting ca. 500 host cell proteins involved in cellular logistics using small interfering RNAs (siRNA) (Kehl et al., 2019). The screen deployed live cell imaging (LCI) of Salmonellainfected host cells and evaluation of the siRNA-induced effects on SCV biogenesis and SIF formation. Among the hits of this screen were subunits of the ESCRT-III (endosomal sorting complexes required for transport) complex, which was investigated in detail in this study.
ESCRT comprise five distinct complexes (ESCRT-0, -I, -II, -III, and the Vps4 complex) which assemble into a multisubunit machinery that performs unique membrane bending and scission reactions away from the cytosol. The ESCRT complexes are known to act in concert to facilitate the multivesicular body (MVB) pathway, cytokinesis, and virus budding (McCullough, Colf, & Sundquist, 2013), but recently other functions were identified, such as nuclear envelope (NE) resealing (Hurley, 2015). Moreover, it was demonstrated that the ESCRT system is also involved in repair of endomembranes, in particular of endolysosomal damage, and therefore holds a protective function for the cell (Radulovic et al., 2018;Skowyra, Schlesinger, Naismith, & Hanson, 2018). The MVB pathway represents a process where proteins destined for degradation are incorporated into intraluminal vesicles that bud from the membrane of late endosomes creating a MVB. By fusion of MVB with lysosomes, the intraluminal vesicles are delivered to the degradative pathway (Piper & Katzmann, 2007).
Here, we show the impact of the ESCRT-III complex on Salmonella intracellular lifestyle by use of host cells devoid of functional ESCRT-III. For that purpose, we created, using the clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 system, a HeLa knockout (k/o) cell line lacking CHMP3, a core protein of the ESCRT-III complex. We demonstrate that a functional ESCRT system is necessary for SCV biogenesis, since mutant host cells showed a high prevalence of amorphous SCV with defects in separation of intracellular Salmonella from the host cell cytosol.

| Defects in ESCRT-III affect SCV formation
We previously performed an LCI-based RNAi k/d screen for host cell functions contributing to intracellular proliferation of Salmonella enterica serovar Typhimurium (STM), and formation of SCV and SIF in STM-infected HeLa cells (Kehl et al., 2019). Transfection with control siAllStars had no effect on SCV formation (see Figure 1(a) and Movie 1). The k/d of CHMP7 had moderate effects on SIF formation and intracellular proliferation of Salmonella in HeLa cells; however, an aberrant SCV morphology was observed. Whereas the SCV membrane is usually tightly enclosing Salmonella (Eswarappa, Negi, Chakraborty, Chandrasekhar Sagar, & Chakravortty, 2010), the CHMP7 k/d caused the SCV to expand during the course of infection, culminating in noticeably spacious SCV (see Figure 1(b) and Movie 2).
Interestingly, a similar phenotype was recently observed upon depletion of PLEKHM1, also resulting in reduced intracellular replication of STM (McEwan, Richter, et al., 2015).
Mammalian CHMP7 is a poorly characterised accessory protein of ESCRT-III (Horii et al., 2006). Though recently, a nonendosomal role specifically for CHMP7 was discussed (Bauer, Brune, Preiss, & Kölling, 2015), the numerous endosomal interactions of Salmonella (Knuff & Finlay, 2017;Liss & Hensel, 2015;Tuli & Sharma, 2019) hint at a possible role of the ESCRT pathway in SCV and/or SIF biogenesis in this context. In line with this, both isoforms of the AAA ATPase VPS4, responsible for disassembly of ESCRT-III, were high-ranking hits in the F I G U R E 1 Legend on next page.
RNAi screen, and ESCRT-0 component HGS/HRS was the highestranking hit (Kehl et al., 2019). However, so far, the only involvement of ESCRTs in Salmonella pathogenesis shown was the SopBdependent impairment of proper trafficking of the epidermal growth factor receptor (EGFR) to the lysosome, a process dependent on the ESCRT pathway (Dukes et al., 2006). Thus, we tested whether the k/d of another accessory ESCRT-III component or a core component would lead to similar phenotypes.
The efficient k/d of all analysed CHMP components was verified by RT-PCR (see Figure 1f). In fact, we observed that k/d of CHMP1B (see Figure 1(c) and Movie 3) or CHMP4B (Figure 1(d) and Movie 4) both resulted in SCV phenotypes comparable to k/d of CHMP7.
These observations strengthen the involvement of the complete ESCRT-III in SCV biogenesis.

| ESCRT-III subunits are present at membranes of SCV and SIF
To probe the potential interaction of ESCRT-III with Salmonellamodified membranes, we investigated the localisation of various components of ESCRT-III in Salmonella-infected host cells. HeLa cells expressing both fluorescent protein-tagged LAMP1 and subunits of the ESCRT-III complex were infected with Salmonella, and LCI was performed (Figure 2(a)). For CHMP3, we did not succeed to detect fusion proteins in transfected cells. We found that core subunits CHMP6 and CHMP4A both decorated SIF and an almost complete colocalization with LAMP1 was observed. Association with membranes of SIF and SCV was also observed for accessory component CHMP5 but not for CHMP1B and CHMP7.
In conclusion, we observed that core components of the ESCRT-III system responsible for membrane fission in the MVB pathway are recruited to SIF and SCV in Salmonella-infected cells.

| Generation and validation of monoclonal knockout HeLa cell lines
To further analyse the role of ESCRT-III in the intracellular lifestyle of Salmonella and Salmonella-induced endosomal remodelling, we generated a k/o HeLa cell line devoid of ESCRT-III. To achieve complete loss of ESCRT-III function, ablation of a core protein was required. CHMP3 represents a suitable target, as the protein had no isoforms in humans and is localised at the core of the filamentous protein chain representing the ESCRT-III complex (Figure 2(b)). We decided to also generate a k/o cell line of a known Salmonella effector interaction partner, the host protein SKIP. Monoclonal CHMP3 or SKIP k/o HeLa cell lines were created using the clustered regularly interspaced short palindromic repeats (CRISPR) system. We first created dedicated vectors expressing GFP, Cas9, and an single-stranded guide RNA (sgRNA) with specific target sequences against either the human CHMP3 or SKIP gene (see Figure S1). After transfection of HeLa cells, the cell population was sub- To validate that CRISPR/Cas9-treated cells show normal cellular functions, proliferation of monoclonal k/o cell lines was determined F I G U R E 1 Knockdown of CHMP subunits causes formation of amorphous SCV. LAMP1-GFP-expressing HeLa cells were reverse transfected with siRNA for negative control (AllStars, a), and for knockdown of CHMP7 (b), CHMP1B (c), or CHMP4B (d) and incubated for 72 hr. Subsequently, cells were infected with Salmonella WT constitutively expressing mCherry with an MOI of 15. Live cell imaging was performed from 1 to 7 hr p.i. on an SDCM. Grey arrowheads indicate normal SCV, and white and yellow arrowheads indicate development of representative spacious SCV. The SCV indicated by yellow arrowheads are shown in magnified details. Scale bar, 10 μm. Time lapse series are shown in Movie 1, Movie 2, Movie 3, and Movie 4. (e) Quantification of SCV morphologies in siRNA-transfected cells after infection with Salmonella WT. SCV containing single bacteria with LAMP1-positive membranes tightly enclosing single bacteria were scored as "normal SCV." SCV containing multiple bacteria, and/or loose association with bacteria, and/or increased luminal space were scored as "amorphous SCV." For each condition, at least 100 SCV were scored, combining images from two independent siRNA k/o screens. (f) Validation of CHMP knockdown. HeLa LAMP1-GFP cells were reverse transfected with AllStars or the indicated siRNA, and incubated for 72 hr. Total RNA was extracted and mRNA reverse transcribed. The generated cDNA was used in RT-PCR targeting the corresponding host factor with GAPDH serving as reference gene. Depicted is the mean with standard deviation of three biological replicates (n = 3) each performed in triplicates. Statistical analysis was performed with Student's t test and indicated as ***p < .001 using the carboxyfluorescein succinimidyl ester (CFSE) assay. CFSE (ex. 492/em. 517) stably labels nuclear DNA, and each cell division results in a sequential twofold dilution of fluorescence intensities (Lyons & Parish, 1994). Cells were treated with CFSE, and fluorescence was monitored by flow cytometry over the course of 3 days. As  To further validate the HeLa CHMP3 k/o line, an EGFR internalisation assay was performed. Upon binding its ligand EGF, EGFR undergoes internalisation and endocytic trafficking. Some receptors recycle back to the plasma membrane, whereas others are degraded via the endocytic pathway cycling from late endosomes to lysosomes (Sorkin & Goh, 2008). Depletion of CHMP3 leads to impaired ligandinduced degradation of EGFR, resulting in an arrested shuttling of EGFR from early to late endosomes/lysosomes (Bache et al., 2006). HeLa and HeLa CHMP3 k/o cells were stimulated with EGF, and after a 6-hr chase, cells were fixed and immunostained against EGFR, early F I G U R E 2 Colocalization of ESCRT-III subunits with SCV and SIF. (a) HeLa cells either stably transfected with LAMP1-GFP (green) or transiently transfected with LAMP1-mCherry (red) were cotransfected with various plasmids for the synthesis of fusion proteins of CHMP1B, CHMP4A, CHMP5, CHMP6, and CHMP7 with GFP (green) or mRuby (red). Cells were infected with STM WT expressing mCherry (red) and living cells were imaged by CLSM 6-9 hr p.i. Scale bars, 10 μm. (b) Schematic depiction of the multivesicular body pathway and the involvement of the ESCRT system. The subunits of ESCRT-III complex are shown in detail and levels of localisation with SIF are indicated by label colours: green, full colocalization; yellow, partial colocalization; red, no colocalization; black, not tested endosome antigen 1 (EEA1) and LAMP1 ( Figure S2). Confocal laser scanning microscopy (CLSM) revealed lack of EGFR-positive vesicles in HeLa cells. In contrast, colocalization of EGFR-positive vesicles with the early endosomal marker EEA1 but not with the late endosomal/lysosomal marker LAMP1 was found in HeLa CHMP3 k/o cells ( Figure S2).
These results indicate that the loss of subunit CHMP3 is sufficient to inhibit ESCRT-III-dependent endocytic protein degradation. We did not

| Attenuation of SPI2-T3SS mutant strains is reduced in CHMP3 k/o cells
We next investigated the role of CHMP3 for intracellular proliferation of Salmonella. HeLa or HeLa CHMP3 k/o cells were infected with F I G U R E 3 Validation of HeLa CHMP3 and SKIP k/o cell lines. (a) Western blot analysis of cell lysates were performed to determine protein levels of CHMP3 and SKIP in various CRISPR/Cas9-processed single cell-sorted clones. As loading control, blots were processed for detection of γ-Tubulin. Marked in red are clones chosen for further evaluation. (b) CFSE assays were performed to monitor the proliferation of HeLa and k/o cell lines. Overlays of percentages of the maximum fluorescence intensities measured by flow cytometry of CFSE-stained cells over the course of 3 days with or without nocodazole treatment are shown. (c) The factor of daily fluorescence decrease over 3 days was determined for cells either nontreated, or treated with nocodazole. Depicted are means and standard deviations of one biological replicate done in triplicates STM WT, ssaV, and sifA mutant strains. Intracellular CFU were determined at 1 and 16 hr p.i., and the x-fold intracellular replication was determined (Figure 4(a)). The sifA strain lacks an effector protein required for maintaining SCV integrity and remodelling of the host cell endosomal system (Beuzon et al., 2000a). SsaV is an integral subunit of the SPI2-T3SS and ssaV-deficient strains are impaired in T3SS assembly, cannot translocate effector proteins, and showed reduced intracellular replication in cell-based models (Hansen-Wester, Stecher, & Hensel, 2002;Hensel et al., 1998;Klein & Jones, 2001;Nikolaus et al., 2001). We observed a moderate reduction of replication of STM WT in HeLa CHMP3 k/o cells compared with HeLa cells.
In contrast, intracellular replication of the STM ΔssaV and ΔsifA strains was increased in HeLa CHMP3 k/o cells, and the proliferation of the ΔsifA strain exceeded that of STM WT in HeLa cells.
The competitive index (CI) assay is a method that allows to measure more subtle difference in virulence. Host cells were coinfected with STM WT and a mutant strain, and the intracellular replication after 16 hr p.i. was determined, with the ratio of WT to mutant strain representing the CI. As both strains are exposed to the same condi- In summary, the data suggest that in a mixed infection of HeLa CHMP3 k/o cells with STM WT and ΔssaV, the mutant strain gained advantage in intracellular replication, manifesting in a decreased CI index in comparison to parental HeLa cells.  for Salmonella proliferation. Further, the frequency of host cells with cytosol-exposed Salmonella was determined. As shown in Figure 8, the number of infected host cells with cytosol-exposed Salmonella was about twofold higher in HeLa CHMP k/o cells compared with HeLa cells. Ciprofloxacin inhibited intracellular proliferation (Figure 8 (b)) and led to decreased frequency of host cells harbouring cytosolexposed Salmonella in a time-dependent manner (Figure 8(c)). The increased frequency of cytosol-exposed Salmonella in HeLa CHMP k/o cells was increased at various time points with or without addition of ciprofloxacin.

| CHMP3 k/o HeLa cells form amorphous SCV after Salmonella infection
Taken together, these results show that in HeLa CHMP3 k/o cells, Salmonella is in contact with cytosolic G6P while still being confined by a SCV membrane. This indicates a membrane disintegration of the SCV without resulting in the outbreak and hyper-replication of Salmonella in the cytosol.

| DISCUSSION
Here, we describe the impact of the ESCRT-III complex in general, and of the core component CMHP3 in particular, on Salmonella intracellular lifestyle. An siRNA k/d screen revealed the involvement of ESCRT-III core proteins in proper SCV biogenesis. Such core proteins were recruited to SCV and SIF continuum during Salmonella infection. We demonstrated the role of ESCRT-III in SCV biogenesis using a CRISPR/Cas9-edited k/o cell line. HeLa CHMP3 k/o cells failed to induce ESCRT-III-dependent endocytic EGFR degradation, as reported for CHMP3 siRNA-mediated k/d (Bache et al., 2006), indicating lack of a functional ESCRT-III complex. We showed that ablation of CHMP3 and therefore disintegration of a functional ESCRT-III complex lead to aberrant SCV biogenesis. This manifested in a higher prevalence of malformed bulky SCV, and the increased contact of Salmonella to cytosolic components, while still being surrounded by SCV membranes. Quantitative flow cytometry measurements in comparison to HeLa WT cells revealed that a HeLa SKIP k/o line showed a high increase in cells harbouring cytosol-exposed Salmonella, and the cytosolic population increased over time. In contrast, a HeLa CHMP3 k/o showed a lower increase in cytosolic bacteria, and the cytosolexposed population did not increase over the course of the infection.
We concluded that CHMP3 k/o leads to impaired SCV membrane integrity, allowing cytosolic components to diffuse into the SCV. Nevertheless, the SCV did not become prone to rupture and bacteria were not able to escape the SCV to hyper-replicate in the cytosol, as observed for SKIP k/o host cells. The phenomena of "bag-like" F I G U R E 6 SCV phenotypes in HeLa CHMP3 k/o cells after infection by WT or ΔssaV strains. HeLa and HeLa CHMP3 k/o cells were infected with Salmonella WT (a) or ΔssaV strains (b). At 16 hr p.i., cells were then fixed with 3% PFA and immuno-stained against LPS (red) and LAMP1 (green) and imaged by CLSM. (c) Quantification of SCV phenotypes in HeLa and HeLa CHMP3 k/o cells. Per condition, at least 100 infected HeLa cells were identified and scored for normal or amorphous, bulky SCV phenotypes. Depicted are the means and standard deviation of three biological replicates done in triplicates. Statistical analyses were performed by Student's t test and significances are indicated as follows: **p < .01; ***p < .001. Scale bars, 1 and 5 μm in overview and details, respectively F I G U R E 7 Analysis of the presence of cytosol-exposed Salmonella in HeLa and HeLa k/o cells. (a) HeLa, CHMP3, and SKIP k/o cells were pulse-chased with dextran-Alexa647 (blue) and infected with Salmonella expressing a cytosolic reporter for 8 hr at an MOI of 50. Shown are representative SDCM images. Cytosol-exposed Salmonella (green) and total Salmonella (red) were quantified by flow cytometry. (b) Hela, CHMP3 k/o, and SKIP k/o cells were infected with Salmonella strains expressing a cytosolic reporter with MOI 5 and fixed 4, 8, and 16 hr p.i. with subsequent flow cytometry measurements of infected cells with sfGFP-positive (cytosol-exposed) STM. Shown are the means and standard deviation of three biological replicates done in triplicates. Statistical analyses were performed by Student's t test and significances are indicated as follows: *p < .05; **p < .01; ***p < .001. Scale bar, 10 μm enlarged SCV, very similar to the aberrant SCV described in this study, has been observed before. In host cells depleted of pleckstrin homol- Given the fact that Salmonella accumulates factors of the ESCRT system on SCV and SIF membranes, we surmised that these proteins can be specifically recruited to the SCV-SIF continuum. By disruption of the ESCRT-III complex, also a source for SCV extension by endomembranes, in particular MVBs, would be depleted. In this case, the amorphous SCV phenotype would represent a general consequence of shortage of endomembrane supply to SCV. Nevertheless, some recent studies suggest alternative explanations involving membrane repair mechanisms at various locations in the cell. On the one F I G U R E 8 Bacterial replication leads to increased SCV damage in CHMP3 k/o cells. HeLa cells and HeLa CHMP3 k/o cells were infected with STM harbouring cytosolic reporter p4889 at a multiplicity of infection (MOI) of 5. At 1, 2, 4, or 6 hr p.i., 100 ng x ml −1 ciprofloxacin (cipro.) were added to stop bacterial replication if indicated. Cells were fixed 8 hr p.i. and subjected to flow cytometry for quantification of infected HeLa cells with sfGFP-negative (SCV-bound bacteria) and infected cells with sfGFP-positive (cytosol-exposed) STM. Constitutively expressed DsRed serves as proxy for intracellular proliferation. (a) Shown are representative data for infections of HeLa and HeLa CHMP3 k/o cells. The gates indicate cells harbouring DsRed-and sfGFP-positive STM, thus been positive for cytosol-exposed STM. Means and standard deviations from three independent experiments are indicated. (b) Quantification of DsRed intensity (bacterial load) of infected cells. Means and standard deviations from triplicates are shown. (c) Quantification of at least 10,000 infected cells harbouring cytosol-exposed STM. Statistical analyses were performed by ANOVA comparing HeLa WT and CHMP k/o cells and significances are indicated as follows: n.s., not significant, *p < .05; **p < .01; ***p < .001 hand, the general repair of plasma membrane wounds (Jimenez et al., 2014), as well as specifically membrane pores caused by pyroptosis was accredited to the ESCRT-III complex (Ruhl et al., 2018). On the other hand, nuclear envelope (NE) resealing is dependent on ESCRT-III specifically recruited by CHMP7 (Vietri et al., 2015) via the inner nuclear membrane protein LEMD2/LEM2 (Gu et al., 2017).
Additionally, membrane instability due to ESCRT defects was attributed to endomembrane repair mechanisms by this system . In this recent study, repair of minor endolysosomal damage by the ESCRT system, in particular the ESCRT-III complex, was reported. This raises the question if pathogens also use the ESCRT system to repair damage on phagosomal membranes. In HeLa cells expressing CHMP4B-eGFP and infected with the obligate intracellular pathogen Coxiella burnetii, pathogen-containing vacuoles became positive for CHMP4B before later being negative for the marker (Radulovic et al., 2018). These phenomena repeated several times and vacuole expansion, a sign for an intact replicative niche, occurred after CHMP4B recruitment. Accordingly, the authors hypothesised that membranes of the Coxiella-containing vacuole are prone to sporadic ruptures, which are repeatedly repaired by the ESCRT machinery. Various ESCRT-III proteins are also recruited to phagosomes harbouring Mycobacterium tuberculosis (Mittal et al., 2018). Here, the authors speculate that the ESCRT-III system is responding to phagosomal membrane damage inflicted by the ESX-1 secretion system (T7SS) to maintain membrane integrity and allow

Moreover, RNAi screens and follow-up studies demonstrated that
ESCRT is important in mycobacterial growth restriction (Mehra et al., 2013;Philips, Porto, Wang, Rubin, & Perrimon, 2008;Philips, Rubin, & Perrimon, 2005). This could be shown for fast-growing In light of these findings, we propose that Salmonella exploits the ESCRT system in a similar way. Lesions of the SCV in infected cells could be mended by the specific recruitment and subsequent action of the ESCRT system (Figure 9(d)). This recruitment most likely does not only encompass the core ESCRT-III, as indicated by phenotypes of CHMP3 k/o cells, but also the accessory proteins CHMP7 and CHMP1B as indicated by the phenotypes caused by their k/d, even though both proteins did not show strong localisation with SCV or SIF. This discrepancy might be explained by only transient interactions of these proteins with Salmonella-modified membranes.
CHMP7 might thus play a similar role in recruitment of ESCRT-III as in NE resealing, in which possibly a Salmonella effector might substitute the host adapter LEMD2. Interestingly, whereas CHMP7 interacts with IST1 to recruit the microtubule (MT)-severing AAA ATPase spastin in NE sealing (Vietri et al., 2015), specifically CHMP1B recruits spastin in cytokinesis (Reid et al., 2005;Yang et al., 2008).
Hence, this additionally implies that MT participate in regulating SCV integrity as already demonstrated by several interactions of Salmonella effectors with the MT network (Leone & Meresse, 2011;Rajashekar & Hensel, 2011). Consequently, in HeLa CHMP k/d or k/o cells, membrane damages of the SCV cannot be repaired due to a dysfunctional ESCRT-III complex and SCV lesions remain, allowing the influx of cytosol (Figure 9(b)). We reason that STM inflicts minor SCV membrane damage, either by the T3SS or as a side effect of bacterial proliferation and resulting mechanical tension on the SCV.
However, proliferation of Salmonella WT in CHMP3 k/o cells was not impaired.
Using CI assays, we examined intracellular replication of mutant strains. Independent of the host cell genotype, the CI for mixed infections with WT and sifA mutant strain was app. one, showing a balanced replication. This phenotype is based on the capability of Salmonella sifA mutant strains to escape instable SCV and proliferate in the cytosol of epithelial cells (Beuzon, Salcedo, & Holden, 2002;Brumell, Tang, Zaharik, & Finlay, 2002). The knockout of either CHMP3 or SKIP in HeLa cells did not change CI indices, indicating Interestingly, we observed altered CI indices of HeLa k/o cells when coinfected with Salmonella WT and ΔssaV strains. Mutant strains lacking a functional SPI2-T3SS fail to induce SIF formation, resulting in nutrient starvation as the formation of SIF acts as means of nutrient supply for Salmonella (Liss et al., 2017). Because the ssaV mutant strain proliferates better in CHMP3-deficient than in CHMP3-proficient host cells, we speculate that a partially instable SCV membrane supports nutrient access to the ssaV mutant strain.
Additionally, proteomic analyses in macrophages revealed that ΔssaV mutant strain showed signatures of increased exposure to stress by host defence mechanisms, in particular reactive oxygen species (Noster et al., 2019). Therefore, the beneficial effect of a leaky SCV membrane might be a result of access to nutrients in the host cytosol, as well as reduced exposure to damaging factors in the SCV ( Figure (9). This hypothesis refutes that minor SCV lesions are caused by the SPI2-T3SS, as the secretion system is absent in the ΔssaV strain. Thus, we favour the model that minor damages of SCV membranes result from mechanical forces due to bacterial proliferation during later stages of infection. In contrast, SCV damage at early stages of Salmonella infection induced by SPI1-T3SS activity is not repaired by the ESCRT system but by autophagy-dependent membrane repair mechanisms (Kreibich et al., 2015;Owen & Casanova, 2015). In general, this indicates that SCV damage caused by the SPI1-T3SS is more severe than by bacterial proliferation. Endogenous endosomal membrane wounds recruit the autophagy or the ESCRT system, respectively, depending on whether the wound is large or small . Although defective endogenous membrane is degraded by autophagy, the SCV is repaired in the late stages of intracellular lifestyle.
Overall, we present evidence that the ESCRT system, in particu-

| Bacterial strains and growth conditions
Salmonella enterica serovar Typhimurium (STM) was used as wild-type (WT) strain and mutant strains were isogenic to WT (Table 1). STM strains were routinely grown on LB agar or in LB broth containing 50 μg × ml −1 carbenicillin for maintenance of plasmids at 37 C using a roller drum at 60 rpm for aeration.

| Cell lines and cultivation
Experiments were performed using a HeLa cell line (ATCC No. CCL-2), or a lentivirus-transfected HeLa line for stable expression of LAMP1-GFP (Krieger et al., 2014). Cells were routinely cultured in Dulbecco's modified Eagle's medium (DMEM) containing 4.5 g × l −1 glucose, 4 mM stable glutamine, and sodium pyruvate (Biochrom, Berlin, Germany) supplemented with 10% inactivated fetal calf serum (iFCS; Gibco, Darmstadt, Germany) in an atmosphere of 5% CO 2 and 90% humidity at 37 C.  Table S1. Next, a mixture of the transfection reagent HiPerFect (Qiagen, Hilden, Germany) and serum-free cell culture medium was applied, and this was incubated for 5-10 min at room temperature (RT). Then, 5,000 or 125,000 cells per well for 96-well plates and 6-well plates, respectively, in serum-containing medium were added and incubated for 72 hr at 37 C in a humidified atmosphere containing 5% CO 2 . For RT-PCR 1 μl of cDNA was used with the Thermo Scientific Maxima SYBR Green/Fluorescein qPCR Master Mix (2×). As reference gene, the housekeeping gene GAPDH was selected (Vreeburg, Bastiaan-Net, & Mes, 2011

MvP503
ΔsifA::FRT (Popp et al., 2015) for those as well as GAPDH are listed in Table S2. Primer concentration was 150 nM each and for each primer pair primer efficiency was determined. RT-PCR was performed in an iCycler (Bio-Rad, Munich, Germany) in triplicates in 96-well plates. Relative expression was determined using the 2 −ΔCt method (Kerr et al., 2010;Town et al., 2009) with GAPDH expression set as 100%. Results were graphed using SigmaPlot 13 (Systat Software, Erkrath, Germany).

| Construction of plasmids
Plasmids used in this study were either obtained from Addgene, kind gifts from various laboratories, or cloned by Gibson Assembly or restriction enzyme digests and are listed in Table 2. For the construction of plasmids encoding for host proteins fused to mRuby2 or eGFP, Gibson Assembly was performed. Oligonucleotides for cloning are listed in Table S2. First, N-or C-terminal mRuby2 vectors were cloned. The vectors pEGFP-C1 and pEGFP-N1 were amplified, and mRuby2 was exchanged by eGFP. Genes encoding for host proteins were amplified from vectors obtained from DNASU (

| Western blot analyses
Whole cell lysates were prepared using a lysis buffer (1% Triton X-100, 5% glycerol in phosphate-buffered saline (PBS)) from CRISPR/ Cas-treated monoclonal and nontreated HeLa control cell lines. After centrifugation, the protein concentration in the supernatant was determined by the BCA Protein Assay (ThermoFisher Scientific) and acetone precipitation for 1.5 hr at −20 C was performed. Protein pellets were dissolved in SDS-PAGE loading buffer containing 0.1% glycine/HCl, pH 2.2 and boiled for 5 min. Circa 15 μg of precipitated whole cell protein was loaded on a 15% SDS-PAGE and after electrophoresis, samples were blotted onto a 0.22 μm nitrocellulose membrane using a semi-dry electrophoretic transfer unit (BioRad). Blots were incubated with a primary antibody directed against CHMP3 (1:2,000) or SKIP (1:100), respectively, and γ-Tubulin (1:1,000; Table S3) as loading control. Secondary antibodies coupled to horseradish peroxidase were chosen according to donor species of primary antibodies and diluted 1:10,000. Detection was achieved by an ECL detection kit (ThermoFisher Scientific), and blots were visualised with a ChemiDoc imaging system (BioRad).

| CFSE proliferation assay
CFSE was used to monitor the proliferation of CRISPR/Cas-treated and nontreated HeLa cell lines. Cells were seeded in surface-treated 6-well plates (TPP) and stained with 5 μM CFSE in PBS for 20 min at 37 C and washed twice with PBS. As a control for blocked proliferation, cells were treated with 100 nM nocodazole for the ongoing

| Cytosolic reporter
HeLa cells either CRIPSR/Cas9-treated or nontreated were seeded in surface-treated 12-well plates (TPP). Cells were infected with Salmonella WT strain harbouring the cytosolic reporter plasmid p4889 (

| Epi-fluorescence microscopy
For confocal laser-scanning microscopy (CLSM) of fixed cells, untreated HeLa or HeLa k/o cells were seeded on coverslips (VWR) in surface-treated 24-well plates (TPP). Salmonella WT and ssaV mutant strains (Table 1) were grown overnight in LB broth at 37 C in a roller drum and subcultured 1:31 in fresh LB media for 3.5 hr. Cells were infected at MOI 50 for 25 min at 37 C. After washing thrice with PBS, cells were incubated with media containing 100 μg × ml −1 gentamicin for 1 hr to kill noninvading bacteria and then replaced with media containing 10 μg × ml −1 gentamicin for the rest of the experiment. Sixteen hours p.i. cells were fixed with 3% PFA and immunostained as described (Müller, Chikkaballi, & Hensel, 2012). In short, permeabilization was done using a blocking solution (2% goat serum, 2% BSA, 0.1% saponin) for 30 min at RT. Incubations with primary antibodies anti-Salmonella LPS Oantigen (1:500), anti-human LAMP1 (1:500), and secondary antibodies (-
After incubation for 10 min at RT, the transfection mix was added to the cells in DMEM with 10% iFCS for at least 18 hr. Before infection, the cells were provided with fresh medium without transfection mix.
Cells were either transfected with a plasmid encoding LAMP1-mCherry using the FuGENE ® HD transfection reagent (Table 2), or pulse-chased with dextran-Alexa647 (10,000 Da molecular weight, Invitrogen). For the latter, cells were pulsed with 200 μg × ml −1 dextran-Alexa647 1 hr p.i. for 4 hr and after washing twice with PBS, cells were chased for 1 hr directly before LCI. Infection was performed as described above with Salmonella WT either harbouring a plasmid encoding GFP or a cytosolic reporter ( Table 2). The medium contained 10 μg × ml −1 gentamicin to kill extracellular Salmonella.

| Competitive index assay
Competitive index (CI) assay was performed as previously described (Segura, Casadesus, & Ramos-Morales, 2004). Briefly, CRISPR/Cas-treated and nontreated HeLa cell lines were seeded in a surfacetreated 24-well plate (TPP) and grown to 80% confluency at the day of infection. WT and aph mutant strains were separately grown overnight in LB and LB containing 50 μg × ml −1 kanamycin and subcultured 1:31 in fresh LB media for 3.5 hr in a roller drum at 37 C.
Cells were infected with a 1:1 mixture of WT and mutant strains at an overall MOI of 1 for 25 min. Noninvading bacteria were killed using media containing 100 μg × ml −1 gentamicin for 1 hr and replaced by media containing 10 μg × ml −1 gentamicin for the ongoing experi-