Bacterial outer membrane vesicles provide an alternative pathway for trafficking of type III secreted effectors into epithelial cells

Outer membrane vesicles (OMVs) are proteo-liposomes universally shed by Gram-negative bacteria. Their secretion is significantly enhanced by the transition into the intra-host milieu and OMVs have been shown to play critical roles during pathogenesis. Enterohemorrhagic Escherichia coli O157 (EHEC), causes diarrheal disease in humans, and soluble toxins including Shiga-like toxins that contribute to disease severity and clinical complications including hemolytic uremic syndrome, have been shown to be OMV associated. In addition to Shiga-like toxins, EHEC produces a type III secretion system (T3SS), and T3SS effectors are associated with colonization and disease severity in vivo. Here, we show that type III secreted substrates including translocators and effectors are incorporated into OMVs independent of type III secretion activity. EHEC strains with non-functional type III secretion systems shed more OMVs and vesicles enter host cells with accelerated kinetics compared to vesicles shed from wild type EHEC. The T3SS effector translocated intimin receptor (Tir) is trafficked from OMVs into host cells and localizes to the membrane. However, its clustering on the host membrane and co-localization with bacterial pedestals is intimin-dependent. We further show that OMV-delivered Tir can cross-complement an effector-deficient EHEC strain, demonstrating that OMV-associated effectors reach the host cell in a biologically intact form. Finally, we observe that the non-LEE encoded E3 ubiquitin ligase effector NleL is also trafficked to host cells via OMVs, where it ubiquitinylates its target kinase JNK. Together, these data demonstrate that trafficking of OMV-associated effectors is a novel and T3SS-independent pathway for the delivery of active effectors to host cells.

Their secretion is significantly enhanced by the transition into the intra-host milieu and OMVs have been 23 shown to play critical roles during pathogenesis. Enterohemorrhagic Escherichia coli O157 (EHEC), 24 causes diarrheal disease in humans, and soluble toxins including Shiga-like toxins that contribute to 25 disease severity and clinical complications including hemolytic uremic syndrome, have been shown to be 26 OMV associated. In addition to Shiga-like toxins, EHEC produces a type III secretion system (T3SS), and 27 T3SS effectors are associated with colonization and disease severity in vivo. Here, we show that type III 28 secreted substrates including translocators and effectors are incorporated into OMVs independent of type 29 III secretion activity. EHEC strains with non-functional type III secretion systems shed more OMVs and 30 vesicles enter host cells with accelerated kinetics compared to vesicles shed from wild type EHEC. The 31 T3SS effector translocated intimin receptor (Tir) is trafficked from OMVs into host cells and localizes to 32 the membrane. However, its clustering on the host membrane and co-localization with bacterial pedestals 33 is intimin-dependent. We further show that OMV-delivered Tir can cross-complement an effector-34 deficient EHEC strain, demonstrating that OMV-associated effectors reach the host cell in a biologically 35 intact form. Finally, we observe that the non-LEE encoded E3 ubiquitin ligase effector NleL is also 36 trafficked to host cells via OMVs, where it ubiquitinylates its target kinase JNK. Together, these data 37 demonstrate that trafficking of OMV-associated effectors is a novel and T3SS-independent pathway for 38 the delivery of active effectors to host cells.

41
Enterohemorrhagic Escherichia coli (EHEC) are a leading cause of food-borne diarrheal disease world-42 wide. In some cases, gastrointestinal symptoms can be complicated by the development of hemolytic 43 uremic syndrome (HUS), which is linked with increased morbidity and mortality (1). Secreted toxins, 44 chiefly Shiga-like toxins, lead to the development of HUS, and the accessory toxins cytolethal distending 45 toxin V (CdtV) and hemolysin (Hly) are thought to contribute to HUS pathology (2, 3). However, many 46 more virulence factors are associated with primary colonization of the gastrointestinal tract, in particular 47 the locus of enterocyte effacement (LEE), which encodes for a type III secretion system (T3SS) and 48 associated effectors (4, 5). The T3SS is a needle-like conduit that translocates effector proteins into the 49 host cell where they manipulate host cellular signaling machinery, induce cytoskeletal rearrangements and 50 modulate immunity to facilitate infection. The T3SS needle is formed by the structural protein EspA (6), 51 and secretion activity is driven by the ATPase EscN (7). EscN directly interacts with T3SS chaperones as 52 well as secreted effectors (8). The T3SS initially translocates structural proteins, followed by the 53 translocon components EspD and EspB which perforate the host cell membrane (9, 10), and in response 54 to environmental signals, eventually switches to secrete effectors targeting host proteins (11,12).

55
The translocon components EspB and EspD form a complex (13) that embeds into and forms a pore in the 56 host membrane bilayer (14,15). EspB is both part of the translocon as well as an effector. It has been 57 shown to also be targeted to the host cell cytoplasm (10), where it interacts with catenin and myosin to 58 cause actin reorganization and contribute to microvillus effacement and inhibition of phagocytosis (16, 59 17) . Other translocated effectors include the translocated intimin receptor (Tir), which inserts into the 60 host membrane and contains both intracellular and extracellular domains. The protein is clustered in the 61 membrane via interactions with the adhesin intimin (18), and Tir clustering leads to downstream signaling 62 events that ultimately result in actin reorganization and pedestal formation (18,19). Tir is critical for 63 colonization and disease pathogenesis in the infant rabbit model (20).
In addition to type III secretion systems, which are pathogen-specific, Gram-negative bacteria also use a 65 range of generalized secretion systems to transport cargo. Outer membrane vesicles (OMVs) are thought 66 to be an alternative secretion system that can facilitate both inter-bacterial (21, 22) as well as bacteria-to-67 host cargo transfer (23). OMVs are proteo-liposomes of 10-200 nm diameter formed by budding of the 68 outer membrane, and can contain membrane-associated and soluble proteins, nucleic acids and small 69 molecules. OMV production is a constitutive process, but is also a protective response and increases in 70 the presence of environmental stresses and during infection (24)(25)(26)(27). Often, this increase in OMV 71 production is co-regulated with other virulence mechanisms (24, 28). Hly (29, 32), which have been shown to reach the target cells in a biologically active form and contribute 76 to pathogenesis (29, 31). However, localization and trafficking of T3SS components by OMVs has not 77 been studied. Here, we set out to investigate whether T3SS components could be trafficked by OMVs,78 and whether this process may act as an alternative pathway to facilitate effector delivery to host cells.

81
Type III secreted proteins localize to outer membrane vesicles in the presence and absence of 82 functional type III secretion. First, we set out to determine whether EHEC OMVs contain proteins that 83 are usually secreted through the type III secretion machinery. Type III secretion is a hierarchical process, 84 where needle components are secreted as early substrates, followed by secretion of the translocon protein 85 EspD and EspB, and finally the effectors. We measured the amount of the translocator EspB as well as 86 the T3SS effector translocated intimin receptor (Tir) in total cell lysates, culture supernatants, and purified OMVs from EHEC wild type strain NCTC 12900 and the isogenic secretion deficient mutant ∆escN, 88 which lacks the translocation ATPase. We also probed fractions of EHEC wild type strain TUV 93-0 and 89 the isogenic mutant ∆OI-148A. ∆OI-148A contains a deletion of the first half of the LEE pathogenicity 90 island, which removes the important regulator Ler required for expression of the T3SS (33). As previously 91 shown, all strains except the ∆OI-148A mutant produced EspB and Tir ( Figure 1). As expected, secretion 92 was only detected in both wild type strains, but not in the ∆escN and ∆OI-148A mutants. Purified OMVs

98
III secretion. Total cell lysates, supernatants, and OMVs were harvested from NCTC 12900 wt and an isogenic ∆escN deletion 99 strain, or a TUV 93-0 wt and an isogenic ∆OI-148A deletion strain. All fractions were normalized using OD600 readings at 100 harvest. For EspD blots, OMV concentrations were measured, and adjusted to concentrations of 10 11 (left lanes) and 2*10 11 101 OMVs/ml (right lanes) for both strains. All fractions were separated by SDS-PAGE and either probed by Western Blotting 102 using α-EspB, α-Tir, or α-EspD antibodies to visualize the translocon components EspB and EspD and the effector translocated 103 intimin receptor (Tir), or stained with Coomassie Blue to provide loading controls.
A type III secretion deficient EHEC strain produces excess outer membrane vesicles. To determine 105 the effect of T3SS activity on OMV production, EHEC wild type and a T3SS secretion deficient strain, 106 ∆escN were compared. The ∆escN strain produces T3SS structural components and effectors, but is unable 107 to secrete them since the associated T3SS ATPase is lacking (34). Both strains were grown in LB for 18 108 hours, culture densities were normalized, and OMVs were harvested and quantified using Nanosight 109 particle tracking analysis. The ∆escN mutant produced significantly more OMVs per cell than the NCTC 110 12900 wild type strain ( Figure 2A). The mean diameter of vesicles produced by the isogenic ∆escN strain 111 was slightly larger than of OMVs derived from the corresponding wild type strain ( Figure 2B). These data 112 show that in the absence of a functional type III secretion system, EHEC hypervesiculates and produces 113 OMVs of aberrant morphology.

121
Outer membrane vesicle uptake by host cells is accelerated in the absence of a functional type III 122 secretion system. To determine if loss of a functional type III secretion system would affect the kinetics 123 of OMV entry into host cells, we used a CCF2-AM based reporter assay as previously described (30).

124
This assay uses a vesicle-targeted beta-lactamase reporter (ClyA-Bla) to measure the fast entry kinetics of OMVs into the host cell cytoplasm in real time. Epithelial host cells were pre-loaded with the fluorescent 126 beta-lactamase substrate CCF2-AM, which upon enzymatic cleavage shifts from green to blue 127 fluorescence emission (30, 35). The ClyA-Bla reporter was expressed in either EHEC NCTC 12900 wild 128 type cells or the isogenic ∆escN mutant. OMVs were harvested and adjusted to a multiplicity of infection 129 (MOI) of 1000 OMVs per cell, using concentration data from particle tracking analysis. As previously 130 determined (30), this corresponds to an MOI of approx. 37 bacteria/cell, a physiologically relevant level.

131
OMVs were incubated with dye-loaded epithelial cells and uptake kinetics were followed for three and a 132 half hours ( Figure 3A, B). Both wild type and ∆escN derived OMVs were rapidly taken up by host cells, 133 but OMVs from ∆escN were taken up with higher efficiency ( Figure 3C) and significantly faster ( Figure   134 3D) than vesicles derived from the wild type strain. We conclude that in the absence of a functional T3SS,

135
OMV uptake is accelerated. Since we have previously established that the chemical structure of the 136 bacterial lipopolysaccharide plays a significant role in rate and efficiency of OMV uptake (30), we 137 analyzed the LPS structure of NCTC 12900 wild type and ∆escN strains. As a control, we also analyzed 138 the LPS of the EHEC strain TUV 93-0, and a corresponding isogenic mutant (∆OI-148A) that does not   OMVs derived from both wild type strains as well as the secretion deficient mutant ∆escN, Tir was present 192 in host cells. However, its localization was not focal but diffuse all over the plasma membrane (Fig. 5B,   193 D, F). Tir was absent following incubation with OMVs derived from the ∆OI-148A strain which does not 194 express the T3SS (Fig. 5H). We also tried to immunostain cells using an α-EspB antibody, but the signal 195 was too weak to draw conclusions, even after prolonged incubation. Together, our data suggests that 196 effectors usually secreted via the EHEC T3SS can also be delivered to intestinal cells via outer membrane 197 vesicles. OMV-delivered translocated intimin receptor reaches the host cell in a biologically active form. 208 Since we observed that the T3SS effector Tir could be trafficked into intestinal cells via OMVs, we next 209 asked whether Tir would reach the host cell in a biologically active form. To test the biological activity of 210 OMV-derived Tir, we infected intestinal epithelial cells with a Tir-deficient EHEC strain, which attaches 211 to host cells but fails to form pedestals (Fig. 6A). We then asked whether Tir-containing OMVs could 212 cross-complement and reconstitute pedestal formation for the Tir-deficient strain. Tir was absent from 213 cells infected with EHEC ∆tir (Fig. 6A), but present when infections were carried out in the presence of 214 OMVs derived from wild type or ∆escN strains (Fig. 6B, C). The presence of bacteria caused the 215 redistribution of OMV-derived Tir, which formed foci under the bacterial cells (Fig. 6B, C). No Tir or 216 Trafficking of T3SS effectors in OMVs | 12 pedestals were observed for uninfected cells (Fig. 6D). Overall, bacterial attachment was enhanced and 217 actin pedestal formation restored in the presence of OMV-derived Tir (Fig. 6E, F).   (Fig. 7). These data suggest that NleL, a non-LEE encoded 240 T3SS effector, is also trafficked to host cells in a biologically active state via outer membrane vesicles.

259
In addition to Shiga-toxins, type III secreted effector proteins are critical factors promoting EHEC 260 colonization and pathogenesis in vivo (4). T3SS effectors modulate key cellular signaling processes, 261 leading to cytoskeletal rearrangements and modulation of host immune responses. In this study, we set 262 out to investigate whether type III secreted proteins are also secreted via outer membrane vesicles, and 263 whether they would be biologically active within the host.

264
OMV biogenesis is thought to require local instabilities in the bacterial cell envelopeeither the 265 local loss of linkage between outer membrane and peptidoglycan layer, or local changes in the outer 266 membrane organization, such as changes in phospholipid content or curvature (41). Here, we found that 267 failure to assemble a functional type III secretion system also leads to increased vesiculation (Fig. 2). The

268
ATPase deficient ∆escN mutant assembles the type III secretion system basal body, but is incapable of producing a needle and secreting effectors. The T3SS structure traverses both membranes as well as the 270 peptidoglycan layer, and T3SS assembly is known to be accompanied by local remodeling of the 271 peptidoglycan layer (42,43). It is conceivable that the initiation of peptidoglycan remodeling by T3SS 272 associated lytic transglycosylases without concurrent full assembly of the T3SS leads to local changes in 273 envelope structure that result in increased vesiculation. The mechanistic details of this process have yet to 274 be elucidated. 275 We report here that EHEC OMVs contain T3SS associated factors (Fig. 1)  We further found here that the uptake of EHEC OMVs derived from a secretion deficient ∆escN 289 T3SS mutant was accelerated compared to OMVs from the wild type strain (Fig. 3). We have previously 290 described that the preferred uptake pathway, and thus entry kinetics of OMVs are shaped by the strain's 291 lipopolysaccharide composition (30). Although we identified minor changes in the long O-antigen of 292 T3SS deficient strains compared to wild type isolates (Fig. 3) which would be consistent with the accelerated uptake kinetics, it is unclear whether those changes are sufficient to explain the accelerated 294 entry kinetics we observed here. It has been described previously that LPS composition and T3SS activity 295 can impact each other (46,47), and that O-antigen length is inversely correlated with the amount of T3SS 296 expression (48). Additionally, it is possible that the T3SS needle structure facilitates targeting to the 297 clathrin-mediated endocytic route, which mediates slower trafficking to the cytoplasm, and that the 298 absence of the needle structure from both mutants drives preferential uptake via raft-mediated endocytosis, 299 which enables faster and more efficient trafficking of OMV contained substrates to the cytoplasm, as we 300 have described previously (30). 301 We observed that OMV-associated T3SS effectors reach the host cell cytoplasm, and retain their 302 endogenous biological activity in the context of an infection ( Fig. 5 and 6). The delivery of biologically 303 active virulence factors we observe here is in agreement with other studies that found OMV-delivered 304 Shiga toxin, hemolysin and cytolethal distending toxin V all retain biological activity within host cells, 305 despite being subject to endosomal trafficking (29). We found that OMV-delivered Tir preferentially 306 localizes to the host cell membrane, both when translocated through the T3SS and when delivered via 307 OMVs (Fig. 5).However, the characteristic focal pattern we observe upon whole cell infection (Fig. 5A 308 and E) is absent when the effector is delivered via OMVs, where we instead observe diffuse localization 309 on the host plasma membrane (Fig. 5B, D, F). However, co-localization with bacteria and with actin 310 pedestals is restored for vesicle-trafficked Tir, once bacteria attach to the host cell (Fig. 6). This indicates 311 that the effector remains biologically active, but its correct localization is dependent on the presence of     Rate estimation, efficiency of uptake, and statistical analysis. To estimate the gradients of the data (i.e. 377 rates of uptake), polynomials were fitted to each data set using the cubic spline function csaps in Matlab 378 R2017b. Numerical estimates of the gradients of the resulting polynomials were determined using the 379 gradient function. To ensure that the gradient estimates were as smooth as possible whilst also retaining 380 the overall shape and trend of the data, a small smoothing parameter was used. Analysis of variance 381 (ANOVA) was used to determine statistical significance, with a Brown Forsythe test to determine equal 382 variance (GraphPad Prism software). A p-value of <0.05 was considered statistically significant.

383
Efficiency of uptake was calculated as the absolute change in blue:green fluorescence intensity ratio 384 between 0 and 3 hours ([Em460/Em530]t=0hrs)/ [Em460/Em530]t=3hrs). Analysis of variance (ANOVA) was used to determine statistical significance, with a Brown Forsythe test to determine equal variance 386 (GraphPad Prism software). A p-value of <0.05 was considered statistically significant.

388
Immunofluorescence staining and microscopy. Infections were carried out as described above, with the 389 exception that RKO cells were used and seeded onto glass cover slips. Cells were infected with EHEC or 390 OMVs as described above for four hours, washed with PBS, and fixed with 3.2% paraformaldehyde in 391 PBS at 4 °C for 16 hours. Fixed cells were permeabilized with 1% Triton X-100 in PBS for 10 minutes, 392 and blocked using 1% BSA in TBST for one hour. Cells were then stained with α-Tir (LS-C500576 from 393 LSBio, 1:1000 in blocking buffer) for one hour at room temperature, washed three times with TBST, and 394 stained with α-rabbit secondary antibody conjugated to Alexa-488 (1:500 in blocking buffer) for one hour 395 at room temperature. Cells were washed three times with TBST, stained with Hoechst (1:1000 in PBS) 396 and phalloidin conjugated with Alexa-680 (1:100 in PBS) for 10 minutes and washed three times with 397 PBS and one with water prior to mounting in ProLong Antifade Gold. Slides were imaged after curing in 398 the dark for 16 hours, using an Olympus IX83 with a Fluoview FW3000 confocal system and a 100x oil 399 immersion objective.