Mammalian orthoreovirus can exit cells in extracellular vesicles

Several egress pathways have been defined for many viruses. Among these pathways, extracellular vesicles (EVs) have been shown to function as vehicles of non-lytic viral egress. EVs are heterogenous populations of membrane-bound structures released from cells as a form of intercellular communication. EV-mediated viral egress may enable immune evasion and collective viral transport. Strains of nonenveloped mammalian orthoreovirus (reovirus) differ in cell lysis phenotypes, with T3D disrupting cell membranes more efficiently than T1L. However, mechanisms of reovirus egress and the influence of transport strategy on infection are only partially understood. To elucidate reovirus egress mechanisms, we infected murine fibroblasts (L cells) and non-polarized human colon epithelial (Caco-2) cells with T1L or T3D reovirus and enriched cell culture supernatants for large EVs, medium EVs, small EVs, and free reovirus. We found that both reovirus strains exit cells in association with large and medium EVs and as free virus particles, and that EV-enriched fractions are infectious. While reovirus visually associates with large and medium EVs, only medium EVs offer protection from antibody-mediated neutralization. EV-mediated protection from neutralization is virus strain- and cell type-specific, as medium EVs enriched from L cell supernatants protect T1L and T3D, while medium EVs enriched from Caco-2 cell supernatants largely fail to protect T3D and only protect T1L efficiently. Using genetically barcoded reovirus, we provide evidence that large and medium EVs can convey multiple particles to recipient cells. Finally, T1L or T3D infection increases the release of all EV sizes from L cells. Together, these findings suggest that in addition to exiting cells as free particles, reovirus promotes egress from distinct cell types in association with large and medium EVs during lytic or non-lytic infection, a mode of exit that can mediate multiparticle infection and, in some cases, protection from antibody neutralization.

. After adsorbing L cells with T1L or T3D reovirus or 137 medium alone (mock), we evaluated plasma membrane disruption using trypan blue staining 138 every 24 h for 96 h. Compared to T1L-infected and mock-infected cells, significantly more T3D-139 infected cells were trypan blue positive, with plasma membranes of nearly all cells disrupted by 140 96 h p.i. (Fig. 1B). In contrast to T3D infection, T1L infection yielded low levels of trypan blue-141 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint positive cells comparable to mock infection at most time points, indicating minimal plasma 142 membrane disruption. Thus, although T1L and T3D both replicate efficiently in L cells, these 143 strains exhibit significant differences in their capacity to disrupt cell membranes. Therefore, we 144 suspected these viruses may employ different egress strategies 145 146 To evaluate reovirus association with fractions enriched for EVs, we adsorbed L cells with T1L 147 or T3D and used sequential differential centrifugation to fractionate EV populations. The 148 centrifugation conditions chosen enrich for certain sizes of EVs; 2,000 × g enriches for large 149 EVs, and 10,000 × g enriches for medium EVs (Fig. 1C) (6). Centrifugation at 100,000 × g is 150 anticipated to pellet a mixed population of small EVs and free reovirus particles (6, 50). These 151 fractions are not "pure" populations of any one type of EV; rather, they represent an enrichment 152 based on size. However, we anticipate that apoptotic blebs would primarily be enriched in the 153 large-EV fraction, microvesicles in the medium-EV fraction, and exosomes in the small-EV 154 fraction (6). Due to their size and density, free reovirus particles are not anticipated to pellet at 155 2,000 × g or 10,000 × g unless they are directly associated with larger structures (50). To 156 determine whether reovirus protein associates with each EV-enriched fraction, we harvested 157 supernatant from infected L cells every 24 h for 96 h and enriched for large EV, medium EV, 158 and small EV/free virus fractions. We resolved equal volumes of each sample by SDS-PAGE 159 and immunoblotting and quantified the reovirus λ3 protein signal associated with each fraction.

160
We found that reovirus protein associated with fractions enriched for each EV size, and 161 association increased with infection time (Fig. 1D-G). By 96 h p.i., we detected T1L protein in 162 approximately equivalent proportion in association with large EV, medium EV, and small EV/free 163 virus fractions (Fig. 1E). Likewise, at 96 h p.i., we detected T3D protein associating 164 approximately equivalently with medium EVs and with the small EV/free virus fraction, though 165 T3D protein association with the large EV fraction was comparatively lower (Fig. 1G). Thus, 166 although some strain-specific protein association differences exist between the reovirus strains,

167
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

170
To determine whether the reovirus protein associated with EV fractions represented infectious 171 reovirus, we used a plaque assay to determine the titers of T1L and T3D associated with large 172 EV, medium EV, and small EV/free virus fractions. We chose the 72 h p.i timepoint for this 173 analysis because there are high amounts of detectable reovirus-EV association (Fig. 1E, 1G) 174 without the near complete plasma membrane disruption induced by T3D at 96 h p.i. (Fig. 1B).

175
Infectious reovirus was detected in all EV fractions (Fig. 1H)  variable and sometimes lower than those present in the small EV/free-virus fraction, though 180 titers were generally high, and differences were not statistically significant (Fig. 1H)  size of interest but were not homogenous (Fig. 2). Large EVs purified from reovirus-infected cell 191 supernatants contained enveloped structures hundreds to more than a thousand nanometers in 192 diameter with membranes that often appeared thin and non-uniform, potentially due to a loss of 193 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint the contents within the large EVs ( Fig. 2A-B). In some cases, the EV structures in this fraction 194 were smaller, appeared to have thicker membranes, and formed aggregates. We observed 195 reovirus particles measuring about 80 nm in diameter adhered to, or in some cases possibly 196 enclosed within, these structures. When we visualized medium EVs purified from supernatants 197 of reovirus-infected cells (Fig. 2C-D), we observed vesicles measuring ~ 600 nm in diameter.

198
These EVs and tended to have rounder, more uniform shapes with well-defined membranes.

199
We observed reovirus particles associating with medium EVs, though it was often unclear 200 whether particles were on the interior or exterior of the EVs. Occasionally we observed single 201 viral particles, pairs of particles, or multiparticle clusters ( Fig. 2A-D). Overall, these findings 202 suggest that centrifugation enriches for large and medium EVs, although these fractions do 203 appear to contain at least partially heterogenous EV populations. Furthermore, T1L and T3D 204 reovirus both associate with large and medium EVs; however, it is unclear whether the reovirus 205 particles are bound on the exterior of the EVs or whether they are packaged internally.

207
Extracellular reovirus fails to associate with small EVs. We anticipated that the final step in 208 the sequential centrifugation protocol enriched for a mixed population of small EVs and free 209 reovirus particles. To determine whether small EVs could be separated from free reovirus 210 particles, we infected L cells with T1L or T3D for 72 h, harvested the supernatant, and 211 concentrated large EV-depleted and medium EV-depleted supernatant on an iodixanol cushion 212 (Fig. 3A). We applied the resulting small EV/free virus pellet to an iodixanol gradient and 213 centrifuged overnight. We collected 12 × 1ml fractions, with fraction 1 representing the top of the 214 gradient and fraction 12 representing the bottom of the gradient. We resolved collected fractions 215 and immunoblotted for reovirus proteins and a protein marker of small EVs, CD81 (Fig. 3B-E)

216
(6, 51). Gradient-separated T1L-infected cell supernatants yielded a strong CD81-positive small 217 EV signal in fractions 7-9, which was distinct from the reovirus protein signal detected in 218 fractions 10-12 ( Fig. 3B-C). Gradient-separated T3D-infected cell supernatants exhibited a 219 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint similar phenotype; we detected CD81 in fractions 7-10 and reovirus proteins in fractions 9-12, 220 with peak signals of each in distinct fractions ( Fig. 3D-E). For T1L and T3D, fraction 7 contained 221 small EVs, which resembled exosomes based on their small size and cup-shaped morphology 222 (52, 53); we did not detect reovirus particles in this fraction (Fig. 3F-G). Fraction 10 contained 223 mainly protein aggregates, with some small EVs scattered sparsely throughout (Fig. 3H-I). We 224 did not detect any T1L particles in this fraction, and although we did detect T3D particles, we did 225 not observe physical association of the reovirus particles with the small EVs. Fraction 11 226 contained free T1L and T3D virus particles, with no small EVs (Fig. 3J-K). These findings    with EVs can shield reovirus from antibody-mediated neutralization, we employed a plaque 244 reduction neutralization assay. We enriched large EV, medium EV, and small EV/free virus 245 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint fractions, as well as iodixanol gradient-separated free virus, from the supernatants of L cells 246 infected with T1L or T3D for 72 h (Fig. 4A). We treated each fraction with reovirus strain-247 specific neutralizing antiserum or with medium alone and determined titers by plaque assay. We 248 hypothesized that if reovirus is present as free particles or as particles adhered to the EV 249 exterior, then the virus would be sensitive to neutralization, and the treated sample titer would 250 be reduced relative to the untreated sample titer. However, if reovirus particles are enclosed 251 within EVs, then the virus would be protected from neutralization, and the treated sample titer 252 would be comparable to the untreated sample titer. We found that when reovirus associated 253 with large EV or small EV/free virus fractions, both T1L and T3D were neutralized to similar 254 levels as free reovirus, with titers reduced on average by 100-fold ( Fig. 4B-E). In contrast, when 255 associated with the medium EV fraction, T1L and T3D titers were unaffected, demonstrating 256 robust protection from neutralization. These findings suggest that T1L and T3D particles 257 released from L cells are specifically packaged inside medium EVs, but not inside large EVs.   (Fig. 4F). We adsorbed fresh L cell monolayers with serially diluted intact 265 EV fractions or free reovirus particles and isolated individual plaques, which we define here as 266 single infectious units. In the case of a free virus particle, a single infectious unit is likely to be 267 an independent WT particle or an independent BC particle. If an EV bundles multiple particles 268 together, then a single infectious unit could contain multiple WT, multiple BC, or multiple WT 269 and BC particles. We genotyped individual viral plaque infectious units using HRM analysis, 270 which distinguishes WT and BC RNA based on differences in melt temperature conferred by 271 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint genetic polymorphisms in the barcode (Fig. 4G)

309
To determine whether reovirus particles are packaged inside Caco-2-derived EVs and protected 310 from antibody neutralization, we used our plaque reduction neutralization assay (Fig. 4A). We 311 found that T1L was neutralized to similar levels as free reovirus when it was associated with 312 Caco-2-derived large EVs and small EVs/free virus but was protected from neutralization when 313 associated with medium EVs (Fig. 5G-H). Interestingly, T3D was efficiently neutralized to levels 314 similar to free T3D reovirus when associated with any Caco-2-derived EV fraction ( Fig. 5I-J).

315
These findings suggest that T1L particles are packaged inside of medium EVs released from 316 two cell types. However, EV-mediated protection appears to be virus-strain-and cell-type-317 dependent, as T3D is efficiently protected when associated with L cell-derived medium EVs, but 318 T3D is inefficiently protected when it associates with Caco-2-derived medium EVs.

321
EV release on a whole-cell level, we adsorbed L cells with medium (mock) or T1L or T3D 322 reovirus, incubated cells for 72 h, and used sequential differential centrifugation to enrich cell 323 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint culture supernatants for large EV, medium EV, and small EV/free virus fractions, resuspending 324 each in an equal volume. To compare relative amounts of protein present in each EV fraction, 325 we used SDS-PAGE and resolved equal sample volumes (Fig. 6A). Compared with uninfected 326 cells, T1L-or T3D-infected cells released material containing significantly increased total protein 327 signal in most EV fractions (Fig. 6B). On average, we detected an approximately two-fold 328 increase in released protein in each fraction for infected cells compared to mock-infected cells 329 (Fig. 6C). To select for EVs, we subjected large EV, medium EV, and small EV/free virus 330 fractions to further separation using annexin V nanobead immunoprecipitation. Annexin V binds 331 phosphatidylserine, which is present on the exterior of most EVs but is not displayed on the 332 surface of healthy cells. After subjecting equal volumes of immunoprecipitated samples to SDS-333 PAGE, we found that reovirus infection increased the protein amount associated with released 334 large EV, medium EV, and small EV/free virus fractions relative to mock (Fig. 6D). Although 335 only the difference for the large EV fraction was statistically significant, on average we detected 336 a three-fold to more than four-fold protein signal increase in EV fractions released from infected 337 cells over the corresponding fractions from mock-infected cells ( Fig. 6E-F

345
In a parallel approach to determine the effects of reovirus infection on EV release, we used a 346 lipophilic dye, DiI, to quantify the EVs released from reovirus-infected cells compared to 347 uninfected cells. Following adsorption with T1L, T3D, or medium (mock), we enriched large EV, 348 medium EV, and small EV/free virus fractions. We mixed fractions with DiI to stain membranes,

349
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint then imaged ten randomly selected fields of view by confocal microscopy and used the 350 EVAnalyzer FIJI plugin to count DiI-positive puncta, which likely represent EVs ( Fig. 6G 371   7). Presently, we are unable to delineate the specific EV subpopulations with which reovirus 372 associates, though T1L and T3D have been visualized inside EVs that resemble microvesicles 373 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint ( Fig. 2 and S1 Fig). Molecular and imaging approaches will reveal additional insights into the 374 detailed mechanisms of EV-associated reovirus egress in the future.

375
We envision a model in which reovirus can use at least three distinct pathways to exit 376 infected cells: (i) by membrane lysis, (ii) enclosed within medium EVs, or (iii) using a mechanism 377 involving "sorting organelles" and "membranous carriers" (38, 48). More than one reovirus 378 egress pathway may function in each cell, and the pathways used may vary by cell type.

379
Bluetongue virus exits cells using both lytic and non-lytic strategies (15). In addition to inducing 380 cell lysis, bluetongue virus also buds non-lytically from the plasma membrane in multiparticle 381 EVs that carry markers of lysosomes and exosomes (15, 16). In L cells and Caco-2 cells, T3D 382 reovirus may employ a similar strategy in which free virus particles are released through lysis, 383 while additional virus exits cells enclosed within medium EVs. In contrast to T3D, T1L reovirus 384 egress occurs in the near-complete absence of membrane disruption (Fig. 1B, 5C), which is 385 consistent with enclosure in medium EVs but fails to explain free virus release. We propose that 386 an additional host cell-assisted mechanism, perhaps akin to the non-lytic reovirus egress

393
Virus release in association with EVs may protect reovirus from host immune defenses.

394
T1L and T3D reovirus released from L cells are shielded from antibody-mediated neutralization 395 and, thus, are likely enclosed within medium EVs (Fig. 4B-4E). Our findings echo those

410
In this scenario, T3D could be packaged internally in the L cell-derived medium EV subtype and 411 bound externally on the Caco-2-derived medium EV subtype. Precedent for a "dual EV egress" 412 strategy exists for encephalomyocarditis virus, which exits cells in two distinct EV subtypes, one 413 of which carries markers derived from the plasma membrane, and one of which carries markers 414 associated with secretory autophagosomes (61). However, given that we are currently unable to 415 enrich solely for reovirus-containing EVs, further study is needed to discern differences between 416 T1L-associated and T3D-associated medium EVs derived from L cells or Caco-2 cells. Another 417 possibility, though somewhat refuted by the observation that infectious T3D reovirus associated 418 efficiently with all EV fractions released from Caco-2 cells (Fig. 5F), is that the efficiency with 419 which T3D interacts with EV biogenesis pathways differs between cell types and may result in a 420 difference in T3D's capacity to orchestrate internal or external medium EV packaging. The  (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

477
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint Altogether, our work suggests that in addition to exiting as free independent particles, in 478 a virus strain-and cell type-dependent manner, reovirus egresses from two distinct cell types 479 enclosed in medium-sized, immune-protective EVs that can promote multiparticle infection.

546
debris was pelleted and discarded following centrifugation at 300 × g for 10 min. The resulting 547 supernatant was centrifuged at 2,000 × g for 25 min to pellet large EVs, followed by 548 centrifugation at 10,000 × g for 30 min to pellet medium EVs, and then at 100,000 × g for 2 h to 549 pellet a mixed population of small EVs and free virus particles. Pelleted EV fractions were re-550 suspended in EV storage buffer (5M NaCl, 1M MgCl2, 1M Tris pH 7.4) and stored briefly at 4°C 551 or used immediately for assays.

552
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint Iodixanol Gradient Separation of Small EVs and Free Virus. Cell debris, large EVs, and 554 medium EVs were cleared from supernatants by sequential differential centrifugation, as 555 described above. The resulting supernatant was concentrated on a 2 ml 60% iodixanol cushion 556 in 0.25 M sucrose and 10 mM Tris, pH 7.5 at 100,000 × g for 4 h (94). Following     (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint   (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint

760
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint

787
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint small EVs were harvested from supernatants using sequential centrifugation, as previously 798 described. Then, EVs were immunoprecipitated using annexin V nanobeads, which bind to 799 phosphatidylserine. Equal volumes of immunoprecipitated material were resolved by SDS-PAGE    . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.29.555250 doi: bioRxiv preprint