Hemolysin liberates bacterial outer membrane vesicles for cytosolic lipopolysaccharide sensing

Inflammatory caspase-11/4/5 recognize cytosolic LPS from invading Gram-negative bacteria and induce pyroptosis and cytokine release, forming rapid innate antibacterial defenses. Since extracellular or vacuole-constrained bacteria are thought to rarely access the cytoplasm, how their LPS are exposed to the cytosolic sensors is a critical event for pathogen recognition. Hemolysin is a pore-forming bacterial toxin, which was generally accepted to rupture cell membrane, leading to cell lysis. Whether and how hemolysin participates in non-canonical inflammasome signaling remains uncovered. Here, we show that hemolysin-overexpressed enterobacteria triggered significantly increased caspase-4 activation in human intestinal epithelial cells (IECs). Hemolysin promoted LPS cytosolic delivery from extracellular bacteria through dynamin-dependent endocytosis. Further, we revealed that hemolysin was largely associated with bacterial outer membrane vesicles (OMVs) and induced rupture of OMV-containing vacuoles, subsequently increasing LPS exposure to the cytosolic sensor. Accordingly, overexpression of hemolysin promoted caspase-11 dependent IL-18 secretion, gut inflammation, and enterocyte pyroptosis in orally-infected mice, which was associated with restricting bacterial colonization in vivo. Together, our work reveals a concept that hemolysin promotes noncanonical inflammasome activation via liberating OMVs for cytosolic LPS sensing, which offers insights into innate immune surveillance of dysregulated hemolysin via caspase-11/4 in intestinal antibacterial defenses. Significance Sensing of lipopolysaccharide (LPS) in the cytosol triggers non-canonical inflammasome-mediated innate responses. Recent work revealed that bacterial outer membrane vesicles (OMVs) enables LPS to access the cytosol for extracellular bacteria. However, since intracellular OMVs are generally constrained in endosomes, how OMV-derived LPS gain access to the cytosol remains unknown. Here, we reported that hemolysin largely bound with OMVs and entered cells through dynamin-dependent endocytosis. Intracellular hemolysin significantly impaired OMVs-constrained vacuole integrity and increased OMV-derived LPS exposure to the cytosolic sensor, which promoted non-canonical inflammasome activation and restricted bacterial gut infections. This work reveals the role of hemolysin in promoting non-canonical inflammasome activation and alerting host immune recognition, which provides insights into the more sophisticated biological functions of hemolysin upon infection.


Introduction 52
The host innate immune system can sense invading bacteria by detecting pathogen-associated 53 molecular patterns (PAMPs) (Vance et al, 2009). Lipopolysaccharide (LPS), a component of the 54 outer cell membrane of Gram-negative bacteria, is one of the strongest immune activators 55 (Rosadini & Kagan, 2017). Extracellular and endocytosed LPS is recognized by the 56 transmembrane protein Toll-like receptor 4 (TLR4), leading to gene transcriptional regulation in 57 response to infection (Kagan et al, 2008). Recent studies showed that host can detect LPS in the 58 cytosol via a second LPS receptor, caspase-11 in mice and caspase-4/5 in humans (Hagar et al, 59 2013;Kayagaki et al, 2013;Shi et al, 2014). Caspase-11/4/5 directly binds cytosolic LPS (Shi et 60 al, 2014), leading to its own activation, which thus cleaves gasdermin D to induce pyroptotic cell 61 death and activate non-canonical activation of NLRP3 to release interleukin-1β (IL-1β) or IL-18 62 (Shi et al, 2015;Yang et al, 2015). Therefore, compartmentalization of LPS receptors within cells 63 allows host to respond differentially and sequentially to LPS at distinct subcellular locales, which 64 function in concert to constitute host noncanonical inflammasome defenses. 65 Caspase-11/4/5, as cytosolic sensors, only recognize LPS that has entered the host cell 66 cytoplasm; however, the mechanism by which LPS from invading bacteria gains access to the 67 cytosolic sensors remains unclear. For intracellular bacteria, although some bacteria such as 68 Burkholderia (Aachoui et al, 2013) are cytoplasm-residing and easily expose LPS to the cytosolic 69 sensors, many other bacteria such as Salmonella typhimurium (Broz et al, 2012) or Legionella 70 pneumophila (Case et al, 2013) are predominantly constrained in pathogen-containing vacuoles 71 (PCVs), probably masking LPS from cytosolic innate sensing. In addition to vacuolar bacteria, 72 many extracellular bacteria, including Escherichia coli (Kayagaki et al, 2011), Vibrio cholerae 73 (Kailasan et al, 2014), Citrobacter rodentium (Meunier et al, 2014), and Haemophilus influenzae 74 (Rathinam & Fitzgerald, 2016) are thought to rarely access the cytoplasm, but induce caspase-11 75 dependent pyroptosis and cytokine release in cells. Thus, LPS entry into cell cytoplasm is a 76 critical event for recognition of vacuolar or extracellular bacteria by non-canonical inflammasome. 77 First, vacuolar bacteria may shed their LPS from endosome into the cytosol (Garcia-del et al, 78 1997). Lipopolysaccharide-binding protein (LBP) is also implicated in facilitating intracellular 79 LPS delivery (Kopp et al, 2016). Alternatively, guanylate-binding protein 2 (GBP2) induces lysis 80 of PCVs and promotes LPS leakage into the cytoplasm (Finethy et al, 2015;Meunier et al, 2014;81 Pilla et al, 2014). Recently, Vijay A.K. Rathinam and his colleagues improved the understanding 82 of how non-canonical inflammasomes detect extracellular Gram-negative bacteria. Briefly, the 83 outer membrane vesicles (OMVs) of extracellular bacteria enter cells by dynamin-dependent 84 endocytosis, enabling LPS to access the cytosol by escaping from early endosomes (Vanaja et al, 85 2016). However, how OMVs gain access from early endosomes to the cytosol remains unknown. 86 Hemolysin belongs to the pore-forming protein family, rupturing the cell membrane and leading 87 to cell lysis at high doses (Wiles & Mulvey, 2013). Recently, hemolysin was found to participate in 88 modulating cell death pathways at sublytic concentrations (Bielaszewska et al, 2013;Ristow & 89 Welch, 2016;Wiles & Mulvey, 2013), representing more sophisticated toxin activity in contrast to 90 outright pore-forming function. Uropathogenic Escherichia coli (UPEC) isolate CP9 activates 91 caspase-3/7 and stimulates rapid cell apoptotic death in vitro; this phenotype was lost in a ΔhlyA 92 mutant (Russo et al, 2005), indicating the involvement of hemolysin in the cell apoptosis pathway. 93 Recently, increasing evidence suggests that hemolysin promotes activation of inflammasome 94 signals during infection. For example, entero-hemolysin of enterohemorrhagic E. coli (EHEC) 95 O157:H7 triggered mature IL-1β secretion in human macrophages (Zhang et al, 2012), and 96 α-hemolysin of UPEC CFT073 mediated NLRP3-dependent IL-1β secretion in mouse 97 macrophages (Schaale et al, 2016). Strikingly, overexpression of hemolysin in UPEC UT189 98 activated significantly increased caspase-4 dependent cell death and IL-1α release than the 99 controls (Nagamatsu et al, 2015). This is the first evidence demonstrating the relevance of 100 hemolysin in caspase-4 activation, indicating that hemolysin might contribute to non-canonical 101 inflammasome activation. 102 In this study, we first demonstrated that hemolysin in various enterobacteria significantly 103 promoted caspase-4 dependent pyroptosis and IL-18 secretion in IECs. Further, we provided 104 insights into the mechanism of hemolysin-mediated increase in the sensitivity of non-canonical 105 inflammasome to invading bacteria. We showed that hemolysin internalizes into cells via binding 106 to OMVs and promotes rupture of OMV-containing vesicles, thereby releasing OMV-derived LPS 107 into the cytoplasm and eventually triggering significant activation of non-canonical 108 inflammasomes in IECs. Oral infection of mice showed that abnormal expression of hemolysin in 109 vivo alerts the immune system and induces caspase-11-dependent enterocyte pyroptosis and IL-18 110 secretion, which significantly constrains bacterial infection in the gut. Collectively, our results 111 reveal that hemolysin enables OMV-mediated LPS cytosolic delivery for caspase-11/4 sensing, 112 which alarms intestinal innate immune surveillance in vivo, providing insights into the 113 manipulation of non-canonical inflammasome signals by invading bacteria. 114 115

Hemolysin promotes caspase-4 dependent cell death and IL-18 secretion during infection. 117
To screen for bacterial factors involved in regulating non-canonical inflammasome activation, a 118 gene-defined mutant library of Edwardsallar tarda (E. tarda), an enteric pathogen infecting hosts 119 from fish to human (Chen et al, 2017;Wang et al, 2009), was used to identify mutants that 120 induced significantly increased pyroprosis in HeLa cells. Compared to the wild-type strain 121 (EIB202), one of the mutants (0909I) greatly increased LDH release ( Fig. S1A and S1B) and detected in wild-type cells infected with 0909I compared to those infected with EIB202, which 128 were counteracted in Caspase4 -/cells. These data indicate that E. tarda mutant 0909I promotes 129 caspase-4 dependent inflammasome activation in non-phagocyte cells. 130 Next, bioinformatics analysis revealed that the transposon insert site within 0909I is located 131 upstream of a non-RTX hemolysin-encoding gene, ethA (Wang et al, 2010). In agreement with the 132 upregulated transcription level of ethA (Fig. S2A), 0909I showed higher EthA expression (Fig. 133 S2B) and bacterial hemolytic activity (Fig. S2C) than EIB202, which were abolished in the strain 134 of 0909IΔethA. Further, deletion of ethA significantly impaired the ability of 0909I to increase 135 caspase-4 activation in Caco-2 ( Fig. 1A-1C) and HT-29 cells (Fig. S1E-S1G). These data suggest 136 that 0909I promotes non-canonical inflammasome activation by increasing hemolysin expression. 137 To explore whether hemolysin also upregulates non-canonical inflammasome activation in 138 other enteric bacteria, we expanded the investigation to the best-known RTX hemolysin, HlyA in 139 UPEC, enterohemorrhagic E. coli (EHEC), enteropathogenic E. coli (EPEC) and E. coli K12. 140 Similarly, the E. coli strains, containing HlyA expression plasmid, showed significantly higher 141 hemolytic activity (Fig. S2D), which elicited a higher level of caspase-4 dependent LDH release 142 Together, these results suggest that hemolysin plays critical roles in promoting caspase-4 144 activation during Gram-negative bacterial infection. 145 Hemolysin promotes LPS cytoplasmic release through dynamin-dependent endocytosis. 146 Host immunity senses bacterial LPS in the cytosol via caspase-11/4/5 (Hagar et al, 2013;147 Kayagaki et al, 2013;Shi et al, 2014). Because extracellular LPS cannot simply diffuse across the 148 membrane, and most Gram-negative bacteria are not cytosolic, delivery of LPS from bacteria to 149 the cytoplasm is thus very critical for non-canonical inflammasome activation. Because our data 150 demonstrated that hemolysin increased caspase-4 activation in IECs, it is reasonable to speculate 151 that hemolysin may promote cytosolic release of bacterial LPS during infection. We extracted the 152 cytosol of uninfected or infected Caco2 cells using digitonin and assessed LPS levels in a limulus 153 amebocyte lysate (LAL) assay. Digitonin is commonly used to isolate cytosol (Ramsby & 154 Makowski, 2011), and the use of an extremely low concentration of digitonin (0.005%) for a very 155 short duration allows extraction of cytosol devoid of plasma membrane, early and late endosomes, 156 and lysosomes (Vanaja et al, 2016). The LAL assay showed that LPS was present in the cytosol of 157 infected cells, but not in uninfected cells ( Fig. 2A). Significantly higher levels of cytosolic LPS 158 was detected in 0909I-infected cells than in EIB202-infected cells, and it was greatly reduced by 159 deletion of ethA in 0909I ( Fig. 2A). These data demonstrate that hemolysin promotes LPS delivery 160 into the cytoplasm upon E. tarda infection. 161 Edwardsallar tarda is an intracellular pathogen, that replicates within a E. tarda-containing 162 vacuole (ECV) (Cao et al, 2018;Hou et al, 2017;Vanaja et al, 2016), and has no direct access to 163 the cytoplasm. We explored how hemolysin promotes LPS delivery from this bacterium to the 164 cytosol. First, E. tarda strains showed similar internalization capacities between 0909I, EIB202, 165 and 0909IΔethA (Fig. S4A), indicating that hemolysin did not affect cellular uptake of E. tarda. 166 GBP2-induced vacuole destabilization was reported to release vacuolar bacteria for cytosolic LPS 167 recognition (Meunier et al, 2014) . However, no bacteria were recovered by agar plating in the 168 cytosol extracted from E. tarda-infected cells, while legionella pneumophila ΔsdhA was used as a 169 control of vacuole-escaping bacteria (Fig. S4B), suggesting that hemolysin did not promote 170 vacuole-constrained E. tarda to enter the cytosol. Further, to discriminate that hemolysin promotes 171 LPS release to the cytoplasm from vacuolar or extracellular bacteria, three endocytosis inhibitors, 172 cytochalasin D (CD), 5-(N-ethhyl-n-isopropil)-amiloride (EIPA), and dynasore (Dyn) were used to 173 inhibit the internalization of E. tarda in Caco-2 cells. Clearly, CD and EIPA inhibited the 174 internalization of 0909I (Fig. 2B), but did not reduce either cytoplasmic LPS release (Fig. 2C) or 175 caspase-4 activation ( Fig. 2D and Fig. S5A), suggesting that hemolysin-mediated LPS delivery 176 predominantly depends on extracellular bacteria. Unexpectedly, dynasore similarly inhibited 177 cellular uptake of E. tarda (Fig. 2B), but significantly suppressed LPS delivery ( influence on them. These results suggest that hemolysin promotes LPS cytosolic release from 183 extracellular bacteria through a dynamin-dependent endocytosis process. 184

Hemolysin-mediated caspase-4 activation is dependent on association with OMVs. 185
Outer membrane vesicles (OMVs) are spherical, bilayered nanostructures constitutively released 186 by growing bacteria (Schwechheimer & Kuehn, 2015). The association of bacterial toxins with 187 OMVs protects them from inactivation or degradation during infection, representing a highly 188 efficient mechanism of bacteria modulating host defenses (Kaparakis-Liaskos & Ferrero, 2015). culture into pellets, OMV-free supernatants and OMVs. Notably, over 60% of the total EthA was 195 detected in the fraction of OMVs (Fig. 3A), indicating that OMV-associated EthA is the major 196 form of this toxin in E. tarda. In accordance with the increased hemolytic activity in 0909I (Fig. 197 S2C), significantly higher levels of EthA and hemolytic activity were detected in 0909I OMVs 198 than in EIB202 or 0909IΔethA OMVs (Fig. 3B). In contrast, when subject to proteinase K (PK) 199 digestion, in which proteins inside OMVs are protected from degradation, OMV-associated EthA 200 was completely degraded, which agrees with the decreased hemolytic activity (Fig. 3B), 201 suggesting that EthA is exposed on the exterior of E. tarda OMVs. 202 A recent study suggested that OMVs were responsible for delivering LPS into the cytosol and 203 of LPS. Galectin-3 is a β-galactoside binding protein, that is specifically recruited to disrupted 226 pathogen-containing vacuoles (Paz et al, 2010). We assessed the recruitment of GFP-tagged 227 galectin-3 in cells incubated with OMVs. Indeed, 0909I OMVs induced more galectin-3 specks 228 than EIB202 OMVs ( Fig. S7A and Fig. 4A), and galectin-3 showed clearly association with OMV 229 specks within wild-type cells (Fig. 4B). In contrast, removal of hemolysin from 0909I OMVs by 230 proteinase K degradation or deleting ethA significantly decreased the intracellular galectin specks 231 in wild-type cells (Fig. 4A). These data suggest that hemolysin promotes rupture of 232 OMV-containing vacuoles. 233 Next, we explored the contribution of hemolysin to damaging the membrane of OCVs during E. 234 tarda infection. As expected, 0909I triggered significantly increased signal of galectin aggregation 235 in DMSO-treated cells, compared to EIB202 or 0909IΔethA ( Fig. S7B and S7C). In the presence 236 of EIPA, which inhibited the internalization of bacteria, but not OMVs into cells, 0909I induced 237 obvious cytoplasmic galectin specks ( Fig. S7B and S7C), indicating that 0909I-induced galectin 238 aggregation is mainly associated with internalized OMVs rather than bacteria. Further, 0909I 239 trigged a significant increase in cytoplasmic galectin signal, compared to EIB202 or 0909I ΔethA, 240 which was greatly suppressed by pretreating the cells with Dyn ( Fig. S7B and S7C). These results 241 indicate that hemolysin contributes to the destruction of OMV-residing vesicles during bacterial 242

infection. 243
Because hemolysin triggered significant membrane rupture of OCVs within cells, we evaluated 244 whether vesicle lysis facilitates LPS exposure to cytosolic sensors. We extracted the cytosol and 245 quantified the cytosolic LPS in Caspase-4 -/cells upon incubation with purified OMVs. Although 246 EIB202, 0909I, and 0909IΔethA OMVs showed comparable LPS contents (Fig. 4C) and uptake 247 efficiencies ( Fig. 3C and 3D), significantly increased cytosolic LPS was detected in the cytosol of 248 0909I OMV-incubated cells than in the controls. This difference was remarkably reduced by either 249 eliminating OMV-bound hemolysin or damping OMV internalization (Fig. 4D). Furthermore, 250 purified OMVs from E. coli strains were incubated with Caspase-4 -/cells for the cytosolic LPS 251 assay. Accordingly, HlyA + OMVs induced more LPS release into the cytoplasm than their 252 correspondent wild-type OMVs (Fig. 4E), indicating that hemolysin promotes cytosolic release of 253 OMV-LPS. Collectively, these data suggest that hemolysin-mediated membrane lysis of OCVs 254 represents an important means to liberate OMV-LPS for cytosolic sensing in Gram-negative 255 bacteria. 256

Hemolysin increases recognition of invading bacteria by caspase-11 inflammasome in vivo. 257
Intestinal epithelium cells are at the forefront of the host gut defense system. Accumulating data 258 indicate the importance of IEC inflammasomes in shaping intestinal immune defense against 259 bacterial infection via inducing IL-1α/β or IL-18 secretion or enterocyte pyroptosis (Sellin et al, 260 2015). As hemolysin was verified to promote non-canonical caspase-4 inflammasome activation in 261 IECs, we further assessed the involvement of hemolysin in noncanonical inflammasome activation 262 in vivo. C57BL/6 wild-type mice were orally infected with E. tarda strains. Compared to EIB202, 263 0909I showed significantly reduced bacterial burdens in the colon, cecum and lumen (Fig. 5A), 264 but not at the systemic sites (Fig. S8A), indicating that that over-expressed hemolysin restricts E. 265 tarda colonization in the mouse gut. Subsequently, it is interesting to explore the in vivo relevance 266 of hemolysin-mediated gut infection restriction with non-canonical inflammasome activation. Further, we demonstrated that 0909I induced remarkably higher mucosal and serum IL-18 level 273 than EIB202 in wild-type mice ( Fig. 5B and 5C). Tissue pathology analysis revealed that 0909I 274 evoked prominent intestinal inflammation in wild-type mice, typically featured by focal filtration 275 of inflammatory cells and epithelia cell shedding ( Fig. S8B and S8C). In contrast, these 276 and cecum of 0909I-infected wild-type mice, but absent in Caspase-11 -/mice gut (Fig. 5D), 282 indicating that 0909I induced membrane-rupturing pyroptosis in IECs and thus significantly 283 increased enterocyte extrusion. Together, these results indicate that dysregulation of hemolysin in 284 vivo alerts caspase-11 dependent intestinal defenses and restricts bacterial gut infection. promote the entry of OMVs into the cytosol, but directly target cytosolic OMVs and facilitate the 296 interaction of LPS with caspase-11. In addition to host factors, whether bacterial factors, such as 297 OMV-associated bacterial components, are involved in promoting non-canonical inflammasome 298 activation remains unknown. Here, we demonstrate that hemolysin binds OMVs and promotes the 299 lysis of OMV-residing vesicles, which facilitates cytosolic release of OMV-LPS and eventually 300 triggers significant non-canonical inflammasome signals (Fig. 6). Our results suggest that 301 hemolysin represents a biologically important mechanism for releasing endosome-constrained 302 OMV-LPS to cytosolic sensors. In addition to hemolysin, whether other pore-forming proteins

Galectin assay 361
The plasmid expressing GFP tagged galectin-3 was constructed and introduced into HeLa cells by 362 lentivirus transfection. Transfected cells were challenged with different E. tarda strains at a MOI 363 of 50 or incubated with purified OMVs at 100 μg/10 5 cells. Galectin speck formation within cells 364 was observed under a confocal microscope. 365

Cytosolic LPS quantification 366
Caspase-4 -/-HeLa or Caco-2 cells post-incubation with the indicated strains or purified OMVs 367 were treated with 0.005% digitonin. The extracted cytosol and residual fractions were subjected to 368 the Limulus Amebocyte Lysate (LAL) assay to quantify LPS. 369

Figure 5. Hemolysin increases recognition of invading bacteria by caspase-11 in vivo. (A) 591
Bacterial counting by agar plating in the colon, caecum, and lumen of wild-type or 592 Caspase-11-/-mice orally-infected by EIB202 or 0909I (5 × 10 7 cfu/g) at 24 hpi. (B and C) 593 Quantification of IL-18 in the colon, cecum (B) and serum (C) of the mice described in a. (D) 594 Confocal laser scanning of pyroptotic cells in the gut sections stained by anti-GSDMD 595 antibody (green) without permeabilizing cells. The actin was stained in red and nuclei in blue; 596 magnification = 20 ×; the boxes indicate the details of GSDMD signals in the guts. Graphs 597 depict 6-10 mice per genotype and are representative of two to three independent experiments. 598 *p < 0.05, **p < 0.01, ***p < 0.001; NS, not significant (one-way ANOVA).