The biogenesis of extracellular vesicles from Staphylococcus aureus and their application as a novel vaccine platform

Gram-positive bacteria secrete extracellular vesicles (EVs) that package diverse bacterial antigens and play key roles in bacterial pathogenesis. However, the mechanisms underlying EV production in Gram-positive bacteria are poorly understood. We purified and characterized EVs from a community-associated methicillin-resistant Staphylococcus aureus isolate (USA300) and investigated mechanisms underlying EV production. Native EVs contained 165 proteins, including cytosolic, surface, and secreted proteins, autolysins, and numerous cytolysins. Staphylococcal alpha-type phenol-soluble modulins (surfactant-like peptides) promoted EV biogenesis, presumably by acting at the cytoplasmic membrane, whereas peptidoglycan crosslinking and autolysin activity were found to increase EV production by altering the permeability of the staphylococcal cell wall. To address the immunogenicity of EVs, we created engineered EVs (eng-EVs) by expressing detoxified proteins HlaH35L and LukE in EVs generated from a nontoxic S. aureus ΔagrΔspa mutant. Eng-EVs exhibited no cytotoxicity in vitro, and mice immunized with the eng-EVs produced toxin-neutralizing antibodies and showed reduced lethality in a mouse sepsis model. Our study reveals novel mechanisms underlying S. aureus EV production and highlights the usefulness of EVs as a novel S. aureus vaccine platform.


11
S. aureus global regulator. We subsequently deleted spa (the gene encoding protein A) in the EV 245 host strain since an agr mutant overexpresses Spa, which binds to the Fcγ domain of 246 immunoglobulin and dampens antibody development by cross-linking the Fab domain of VH3-type 247 B cell receptors, resulting in apoptotic collapse of these cells 48 . The JE2∆agr∆spa double mutant 248 served as our S. aureus EV vaccine producing host strain. Consistent with previous reports, we 249 demonstrated that the JE2 agr mutation significantly inhibited mRNA expression of hla and the 250 genes encoding all nine leukocidin subunits (Fig. S1a). EVs from WT JE2, but not the 251 JE2∆agr∆spa mutant, contained native Hla as assessed by western blotting (Fig. S1b). Similarly, 252 by using an antibody reactive with both LukS-PV and LukE subunits, we showed that only EVs 253 from WT JE2 had detectable leukocidin reactivity (Fig. S1b). To further validate our results, we 254 analyzed the protein content of EVs purified from JE2∆agr∆spa by MS. Notably, many of the 255 extracellular proteins that were present in JE2 WT EVs were not detectable in EVs from 256 JE2∆agr∆spa. However, some antigens such as MntC and FhuD2 that have been shown to 257 protect mice against experimental S. aureus infections 49-52 were still present in EVs from the 258 mutant strain. Neither protein A nor the cytotoxins Hla, Luk-PVL, LukED, HlgCB, SelX, and PSMs 259 were detectable by LC-MS/MS in EVs purified from the JE2∆agr∆spa mutant (Table S2). Although 260 LukAB was still present in EVs from JE2∆agr∆spa, there was ≥86% reduction in the number of 261 peptides detected in the mutant strain (Table S1 and Table S2). Moreover, as indicated below, 262 EVs recovered from the mutant strain showed no residual toxicity toward human leukocytes. 263 To investigate whether the detoxified JE2 EVs were immunogenic and protective against 264 infection, we immunized mice with 5 µg EVs from JE2∆agr or JE2∆agr∆spa mutants; control mice 265 were given PBS. EVs from either mutant elicited a serum antibody response against sonicated 266 WT EVs, although the antibody level elicited by ∆agr EVs was higher than that elicited by 267 ∆agr∆spa EVs (Fig. S2a). To examine the antigen profiles from EVs that elicited antibody 268 responses after immunization, a bacterial lysate from the USA300 FPR3757 strain was subjected 269 to SDS-PAGE and immunoblotted with pooled sera from mice immunized with either ∆agr EVs or 270 12 ∆agr∆spa EVs. Notably, sera from ∆agr∆spa EVs-immunized mice reacted with more bacterial 271 antigens than sera from ∆agr EVs-immunized mice (Fig. S2b), suggesting that ∆agr∆spa EVs 272 elicited a greater diversity of antibodies than ∆agr EVs. To further evaluate the protective efficacy 273 of EVs, the immunized mice were challenged with WT USA300 strain FPR3757. Immunization of 274 mice with EVs from JE2∆agr∆spa, but not EVs from JE2∆agr, provided significant protection 275 against lethal sepsis (Fig. S2c). Preliminary studies indicated that immunization with higher doses 276 of EVs mixed with alum did not enhance immunogenicity (Fig. S2d). antibodies. Hla is a major secreted staphylococcal cytotoxin, and its production has been 280 associated with severe infections caused by community-acquired MRSA 53 . Immunization 281 against a nonpore-forming Hla variant (HlaH35L) prevents experimental S. aureus pneumonia, 282 skin abscesses, and lethal peritonitis 54-56 . To enhance the protective efficacy of detoxified EVs 283 from JE2∆agr∆spa, we engineered JE2 to package nontoxic HlaH35L 57 and the LukE monomer 284 within eng-EVs. LukED is a member of the S. aureus family of bicomponent leukotoxins and is 285 detected in 82% of blood isolates and 61% of nasal isolates 58 . LukED targets both human and 286 murine neutrophils, macrophages, T cells, dendritic cells, NK cells, and erythrocytes 59,60 . 287 We expressed nontoxic HlaH35L and LukE in strain JE2∆agr∆spa under control of the spa 288 promoter. Because the activity of the spa promoter is enhanced in an ∆agr genetic background, 289 the mRNA levels of Hla H35L and LukE expressed in JE2∆agr∆spa were dramatically increased 290 compared to expression in JE2∆agr∆spa or JE2∆agr∆spa with the empty vector (Fig. S1c). As 291 predicted, both HlaH35L and LukE were detected by Western blot in engineered EVs (eng-EVs) 292 isolated from recombinant strain JE2∆agr∆spa (pHlaH35L-LukE) (Fig. S1b). 293 To evaluate the relative toxicity of EVs prepared from WT strain JE2 and JE2∆agr∆spa vs. 294 eng-EVs from JE2∆agr∆spa (pHlaH35L-LukE), we incubated the EVs in vitro with three different 295 13 cell types. A549 cells are susceptible to Hla-mediated cytolysis, and WT strain JE2 EVs were 296 toxic at concentrations as low as 1 µg/ml. In contrast, JE2∆agr∆spa mutant EVs and the eng-EVs 297 from JE2∆agr∆spa (pHlaH35L-LukE) exhibited negligible toxicity (Fig. S3a). HL60 cells are resistant 298 to Hla-mediated lysis, but they are susceptible to the cytolytic activity of the S. aureus leukocidins 299 (including HlgAB, HlgCB, PVL-SF, LukED, LukAB, and phenol soluble modulins [PSMs]). EVs 300 isolated from strain JE2, but not the ∆agr∆spa mutant or eng-EVs, were cytolytic for HL60 cells 301 at concentrations as low as 1 µg/ml (Fig. S3b). Rabbit erythrocytes are susceptible to Hla, PSMs, 302 and the leukocidins HlgAB and LukED 61,62 . EVs isolated from WT strain JE2 exhibited significant 303 hemolytic activity at concentrations as low as 1 µg/ml, but no hemolytic activity resulted from EVs 304 prepared from the ∆agr∆spa mutant or eng-EVs, even at 20 µg/ml (Fig. S3c). These data 305 demonstrate that the eng-EVs were nontoxic in vitro for mammalian cells. 306 We immunized mice on days 0. 14, and 28 with 5 µg EVs from JE2∆agr or JE2∆agr∆spa 307 mutants; control mice were given 5 ug bovine serum albumin (BSA). Whereas sera from mice 308 immunized with both eng-EVs and ∆agr∆spa EVs, but not BSA, reacted by ELISA with sonicated 309 WT JE2 EVs (Fig. 6a), only mice given the eng-EVs responded with antibodies to purified Hla 310 ( Fig. 6b) or LukE (Fig. 6c). These data indicate that recombinant proteins packaged within 311 S. aureus EV are immunogenic. 312 To examine whether the antibodies elicited in mice by the eng-EV vaccine were functional, 313 toxin neutralizing assays were performed. Sera from mice immunized with eng-EVs effectively 314 neutralized Hla at dilutions ranging from 1:20 to 1:80 (Fig. 6d). In contrast, neutralizing antibodies 315 were low or undetectable in serum from mice given BSA or ∆agr∆spa EVs. Similarly, sera from 316 mice immunized with eng-EVs, but not BSA or ∆agr∆spa EVs, were able to effectively neutralize 317 LukED at dilutions ranging from 1:10 to 1:20 (Fig. 6e). Sera from mice immunized with eng-EVs 318 also neutralized leukocidin HlgAB (Fig. 6f), but not PVL-SF or HlgCB leukotoxins. 319 The immunized mice were challenged with USA500 strain NRS685, a PVL-negative MRSA 320 14 bacteremia isolate. We chose this strain because the PVL-S and PVL-F subunits can interact with 321 LukE and LukD to form inactive hybrid complexes, and this influences LukED-mediated S. aureus 322 virulence in mice 63 . As shown in Fig. 6g, immunization with eng-EVs, but not ∆agr∆spa EVs, 323 protected 50% of the mice in the lethal sepsis model. 324

Discussion 325
Membrane vesicles, produced by mammalian cells, fungi, and bacteria, is an evolutionarily 326 conserved secretory pathway that allows cell-free intercellular communication 64-66 . Microbial EVs 327 encapsulate cargo that include lipids, proteins, glycans, and nucleic acids, which have been 328 shown to play roles in microbial physiology, pathogenesis, and the transmission of biological 329 signals into host cells to modulate biological processes and host innate immune 330 responses 64,65,67,68 . In Gram-negative bacteria, EVs are generated by pinching off the outer 331 membrane, but the mechanism(s) by which EVs escape the thick cell walls of Gram-positive 332 bacteria, mycobacteria, and fungi is unknow. Once shed, S. aureus EVs can undergo cholesterol-333 dependent fusion with host cell membranes to deliver their toxic cargo 28 . S. aureus EVs have 334 been shown to be produced in vivo during experimental pneumonia in mice 28 . In this report, we 335 demonstrate unique properties associated with EV production by JE2, a S. aureus USA300 strain 336 that is representative of the CA-MRSA clone that has rapidly disseminated in the United States. 337 Similar to EVs characterized from other S. aureus isolates 16,17,28 , JE2 EVs encapsulate an array 338 of bacterial antigens, including lipoproteins, exotoxins, and cytoplasmic proteins. 339

340
In an effort to better understand the multiple stages of EV biogenesis in S. aureus, we 341 evaluated putative factors that modulate the membrane and PGN related steps of EV release. The S. aureus cell envelope is comprised of a thick, highly cross-linked PGN layer, 359 proteins, and glycopolymers like WTA and CP. When we assessed the role of the PGN layer on 360 EV release, we found that highly crosslinked PGN serves as a barrier for EV biogenesis. 361 Treatment of S. aureus with sublethal concentration of penicillin G or genetic inactivation of pbp4 362 or tagO, which result in reductions in PGN cross-linking, resulted in a significant increase in EV 363 production, as well as the average size of released EVs. This inverse correlation between PGN 364 cross-linking and EV yield was not unique to USA300 strain JE2 but was also observed with 365 S. aureus strains MW2, COL, and Newman. WTA has been shown to be critical for PGN-366 crosslinking by regulating PBP4 localization to the septation site 42 . A secondary mechanism by 367 which WTA regulates EV production is via its ability to control the activity of Atl and Sle1 -not only 368 by preventing their binding to S. aureus cell wall PGN 72,73 , but also by creating an acidic milieu 369 that limits Atl PGN hydrolase activity 74 . Consequently, autolytic activity is not localized to the the entire septal surface, Atl localized only at the external (surface-exposed) edge of the 395 septum 77 . How autolysins modulate EV release from the cell wall or whether this process is 396 spatially or temporally regulated remains to be determined. 397 Newman and MN8 in our study. We reported that S. aureus CP was shed from broth-grown 399 S. aureus cells 78 , and it is feasible that EVs could serve as a vehicle to liberate CP from the cell 400 envelope. The S. pneumonia capsule was reported to hinder EV release in this pathogen 7 , 401 whereas no effect was observed on EV yield in strains with or without the hyaluronic capsule of 402 S. pyogenes 8 . Whether these streptococcal CPs are present as EV cargo in these pathogens was consequently diminished EV production. Our study elucidates certain mechanisms whereby 458 S. aureus produces and sheds EVs (Fig. 7) and will ultimately further our understanding of 459 bacterial physiology and pathogenesis. We designed and created eng-EVs as a novel vaccine 460 platform against S. aureus infection. Detoxified EVs that over-produced HlaH35L and LukE were 461 immunogenic, elicited toxin neutralizing antibodies, and protected mice in a S. aureus lethal 462 sepsis model, indicating that these naturally produced vesicles have potential as a noval vaccine 463 platform. 464

Materials and Methods 465
Bacterial strains and plasmids. S. aureus isolates (listed in Table S3)  Antibiotics were added in the following concentrations: penicillin G (penG; 0.2 μg ml -1 ), ampicillin 470 20 (Amp; 100 μg ml -1 ), erythromycin (Em; 5 μg ml -1 ), chloramphenicol (Cm; 10 μg ml -1 ), kanamycin 471 (Kan; 50 μg ml -1 ), or tetracycline (Tet; 5 μg ml -1 ). 472 DNA manipulation. Fey  To construct a shuttle vector for expression of HlaH35L and LukE, the spa promoter, hlaH35L, 484 and lukE genes were amplified from S. aureus strains JE2, DU1090 (pHlaH35L), and FRP3757, 485 respectively. To drive the expression of hlaH35L, its sequence was fused to the 3' terminus of the 486 spa promoter containing the ribosome binding site by overlapping PCR. The Pspa-hlaH35L fusion 487 sequence was cloned into the shuttle plasmid pCU1 with restriction enzymes HindIII and SalI. 488 The amplified lukE sequence containing a ribosome binding site was inserted into pCU1 with 489 restriction enzymes SalI and EcoRI. The resulting plasmid pCU1-Pspa-hlaH35L-lukE was verified by 490 enzyme digestion and DNA sequencing. To construct JE2∆spa∆agr expressing nontoxic HlaH35L 491 and LukE, pCU1-Pspa-hlaH35L-lukE was transformed into RN4220 by electroporation and then 492 transduced with ϕ80α to JE2∆spa∆agr, selecting for Cm resistance. 493 21 Isolation and purification of EVs. Isolation of EVs from S. aureus was performed as previously 494 described 7,16 with minor modifications. S. aureus was cultivated in TSB with shaking to an OD650 nm 495 of 1.2. The culture supernatant was filtered and concentrated 25-fold with a 100-kDa tangential 496 flow filtration system (Pall Corp.). The retentate was filtered again before centrifugation at 497 150,000 g for 3 h at 4°C to pellet the vesicles and leave soluble proteins in the supernatant. The 498 EV pellet was suspended in 40% Optiprep density gradient medium (Sigma) and overlaid with 499 gradient layers of Optiprep ranging from 35% to 10%. After centrifugation at 139,000 g for 16 h at 500 4°C, 1 ml fractions were removed sequentially from the top of the gradient. Each fraction was 501 subjected to SDS-PAGE and stained with a Thermo Fisher silver staining kit. Fractions with a 502 similar protein profile on SDS-PAGE were pooled, and the Optiprep medium was removed by 503 diafiltration with phosphate-buffered saline (PBS; 10 mM Na2HPO4, 2 mM KH2PO4, 2.7 mM KCl, 504 and 137 mM NaCl, pH 7.4) using an Amicon Ultra-50 Centrifugal Filter Unit. The diafiltered 505 retentate was filtered (0.45 µm) and stored at 4°C. EV protein concentrations were determined by 506 using a Protein Assay Dye Reagent (Bio-Rad). The size distribution and diameter of vesicles was 507 measured using a ZetaPALS dynamic light scattering detector (Brookhaven Instruments Corp.). 508 Nanoparticle tracking analysis (NTA) was performed by purifying EVs from 100 ml bacterial 509 cultures, as described above. The number of EV particles recovered from individual cultures (and 510 suspended in 1 ml PBS) was determined using a Nanosight NS300 Sub Micron Particle Imaging 511 System (Malvern), as previously described 87 . 512 Electron microscopy of S. aureus EVs. Five microliters of S. aureus EVs were adsorbed for 513 1 min to a carbon coated grid that was made hydrophilic by a 30-sec exposure to a glow discharge. 514 The samples were stained with 0.75% uranyl formate for 30 sec and examined in a JEOL 1200EX 515 or a TecnaiG² Spirit BioTWIN transmission electron microscope. Images were recorded with an 516 AMT 2k CCD camera. For EV dot blotting assays, intact or sonicated EVs were applied to nitrocellulose 546 membranes using a 96 well dot blotter system (Bio-Rad). To block the staphylococcal IgG binding 547 proteins Spa and Sbi, the membranes were blocked with PBST + 5% skim milk and incubated 548 overnight at 4°C with an irrelevant human IgG1 monoclonal antibody (10 µg/ml) in PBST + 1% 549 skim milk. The membrane was washed with PBST and incubated overnight at 4°C with sera 550 (diluted 1:1000 in PBST + 1% skim milk) pooled from mice immunized with EVs (see below) or 551 murine mAbs 78 to CP5 (4C2; 1.2 µg/ml) or CP8 (5A6; 1.2 µg/ml). After washes with PBST, the 552 membrane was incubated with alkaline phosphatase (AP)-conjugated goat anti-mouse antibody 553 (1:15000 dilution in PBST + 1% skim milk) at RT for 2 h. The membrane was washed with PBST 554 and developed with AP membrane substrate (KPL). 555 EV cytotoxicity. The relative toxicity of S. aureus EVs (1 to 20 µg/ml) toward human A549 lung 556 epithelial cells, neutrophil-like HL60 cells, and rabbit erythrocytes was assessed. A549 lung 557 epithelial cells grown in a 96-well plate were incubated overnight at 37°C with EVs or 1 µg/ml of 558 purified Hla. Toxicity was assessed using an LDH cytotoxicity assay kit (ThermoFisher Scientific). Swiss Webster mice (4 weeks old; Charles River) were immunized by the subcutaneous (s.c.) 569 route on days 0, 14, and 28 with 5 µg/ml of ∆agr EVs, ∆agr∆spa EVs, or eng-EVs. Control animals 570 were immunized similarly with bovine serum albumin (BSA; Sigma). Blood was collected from the 571 mice by tail vein puncture before each vaccination and again before challenge. Sera were diluted 572 1:100 and tested by ELISA on 96-well plates coated with 5 µg/ml sonicated WT EVs, 5 µg/ml 573 LukE, or 1 µg/ml Hla. Immunized mice were inoculated with ~2 x 10 8 CFU S. aureus by 574 intravenous (IV) tail vein injection two weeks after the third vaccination. Survival was monitored 575 up to 14 days post-challenge, and the data were analyzed using the log-rank test. 576 Toxin neutralization assays (TNAs). For Hla TNAs, the assay was performed as we previously 577 described 88 . For leukocidin TNAs, human blood was collected from healthy volunteers giving 578 written informed consent, as approved by the Institutional Review Board of The Brigham and 579 Women's Hospital (Human Subject Assurance Number 00000484). Neutrophils were isolated 580 from 10 ml human blood using Polymorphprep (Accurate Chemical), washed, and suspended in 581 RPMI (Invitrogen) containing 5% fetal bovine serum (Invitrogen). Sera from immunized mice were 582 serially diluted and mixed with toxin concentrations yielding ~75% cell lysis (12.5 µg/ml LukED, 583 2.5 µg/ml PVL, 1 µg/ml HlgAB, or 2 µg/ml HlgCB (1∶1 S and F subunits). Samples were pre-584 incubated with leukocidins for 30 min at RT before the addition of neutrophils (1.2×10 5 cells). After 585 2 h at 37°C in 5% CO2, the cells were harvested by centrifugation and suspended in fresh medium. 586 Cell viability was evaluated using CellTiter kit (Promega) according to the manufacturer's 587 recommendations. Percent neutralization was calculated using the formula: [% Viability of (serum 588 + leukocidin + neutrophils) -% Viability of (leukocidin + neutrophils)]. 589 25 Author contributions 590 X.W. initated the project, and X.W., C.W. and J.C.L designed experiments. X.W. performed 591 experiments. X.W. and J.C.L analyzed data, and X.W., C.W., and J.C.L wrote the manuscript. 592

Acknowledgements 593
We are grateful to Drs. Michael Otto for providing the S. aureus psm mutants, Jianxun Ding for 594 providing assistance with DLS and NTA experiments, and Matthew Waldor for use of the 595 StepOnePlus Real-Time PCR System. Christopher Thompson and Anthony Yeh provided expert 596 technical assistance. 597

Data availability 598
Mass spectrometry proteomics data were deposited in the ProteomeXchange Consortium 599 (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository 89 with the data 600 set identifier PXD007953. Additional data that support the findings of this study are available 601 from the corresponding author upon request. 602 Competing interests 603