Complexity and ultrastructure of infectious extracellular vesicles from cells infected by non-enveloped virus

Enteroviruses support cell-to-cell viral transmission prior to their canonical lytic spread of virus. Poliovirus (PV), a prototype for human pathogenic positive-sense RNA enteroviruses, and picornaviruses in general, transport multiple virions en bloc via infectious extracellular vesicles, 100~1000 nm in diameter, secreted from host cells. Using biochemical and biophysical methods we identify multiple components in secreted microvesicles, including mature PV virions; positive-sense genomic and negative-sense replicative, template viral RNA; essential viral replication proteins; and cellular proteins. Using cryo-electron tomography, we visualize the near-native three-dimensional architecture of secreted infectious microvesicles containing both virions and a unique morphological component that we describe as a mat-like structure. While the composition of these mat-like structures is not yet known, based on our biochemical data they are expected to be comprised of unencapsidated RNA and proteins. In addition to infectious microvesicles, CD9-positive exosomes released from PV-infected cells are also infectious and transport virions. Thus, our data show that, prior to cell lysis, non-enveloped viruses are secreted within infectious vesicles that also transport viral unencapsidated RNAs, viral and host proteins. Understanding the structure and function of these infectious particles helps elucidate the mechanism by which extracellular vesicles contribute to the spread of non-enveloped virus infection.


Results
Extracellular vesicles serve multiple roles for cells, including cell signaling and transport of functional proteins, coding RNAs, and/or non-coding RNAs [14][15][16] . Vesicles with a diameter of 100-1000 nm secreted from PV-infected cells were shown to carry PV virions 10 . Therefore, we analyzed extracellular vesicles isolated by size using the well-established fractionation method of differential centrifugation for microvesicles (100-1000 nm) 17 . Phosphatidylserine (PS) -containing microvesicles were purified for additional analysis, as diagrammed in Fig. 1a (see also Materials & Methods).
To follow PV-infection of HeLa cells and determine efficiency of viral infection, control (mock-infected) and PV-infected cells were harvested at 3 and 7 hours post infection (hpi) for control cells, and 3, 4, 5, 6, and 7 hpi for PV-infected cells. Cytoplasmic lysates were then probed for viral proteins. The non-structural viral protein 3CD was produced by PV-infected HeLa cells at 3 hpi, and its levels stayed constant from 3 to 7 hpi; production of the  sense viral RNA (vRNA) from infectious microvesicles (ImV) and mock microvesicles (MmV) that were collected at 8 hpi (see Methods for details). vRNA was measured after extracellular vesicles underwent: 1) no treatment (labeled ImVs or MmVs), or 2) freeze-thaw & detergent (1% sodium deoxycholate) & RNase treatment to break open vesicles and degrade unencapsidated RNAs (labeled DRImV for infectious or DRMmV for mock-infected samples). RT-qPCR relative quantification was calculated as ΔC t where ΔC t = (C t of endogenous control gene (GAPDH)) -(C t of gene of interest (vRNA)), using GAPDH of whole cells for normalization. (b) Schematic of the experimental design for determining the infectivity of untreated infectious microvesicles (ImVs) as compared to infection by ImVs treated by either freeze/thaw alone (FT) or FT, detergent, and RNase (DR). Serial dilutions of sample were used in plaque assays to quantify the number of infectious sites by visual inspection for cell death. intracellular contents in this study, cell viability greater than 90% was confirmed for both mock-and virus-infected cells (Fig. 1b). Samples included all vesicles secreted after the cells were washed and non-FBS-containing media was added at 4 hpi, the time of peak replication. Liquid chromatography-mass spectrometry (LC-MS) of 100-1000 nm diameter vesicles (called here 'microvesicles') secreted from infected cells resulted in the detection of the non-capsid, essential viral replication 2A-, 2C-, 3C-and 3D-containing proteins, in addition to viral capsid proteins VP1-VP3 (Fig. 1c). Consistent with the LC-MS data from size-fractionated microvesicles, western blot analyses ( Fig. 1d; full western blots for all samples are shown in Supplementary Fig. S2) further confirmed the presence of non-capsid viral proteins in microvesicles that contained the membrane phospholipid PS, a sub-population of microvesicles that had been shown to be infectious 10 . Specifically, the PV proteins 2BC, 2 C, 3 A, 3 AB, 3CD, and 3D were identified in PS-containing microvesicles secreted from PV-infected cells but not in those secreted from mock-infected cells (Fig. 1c,d). We did not see 2BC, 3 AB, and 3A in the LC-MS data, likely due to the loss of membrane-associated proteins 2B and 3A in the LC-MS sample preparation.
Replicative and genomic RNA were present in secreted infectious microvesicles. The initiation of PV replication for packaging (+) vRNA within virions requires the presence of template, replicative (−) vRNA. Therefore, we examined the content of PS-positive infectious microvesicles with regard to the presence of vRNA by RT-PCR (Materials & Methods). No amplicons were detected above the threshold set at a fixed signal intensity (0.475) for all experiments. To confirm the absence of DNA contamination from the total RNA extraction/purification process, SuperScript III Reverse Transcriptase was omitted from the "master mix" for reverse transcription, and no amplification occurred in total RNA from PV-infected cells, when either positive-sense (genomic) or negative-sense (anti-genomic, replicative) viral RNA was used as the probe ( Supplementary Fig. S3b,c). To test for nonspecific binding introduced by primer dimers and secondary structures of the primers, RT-qPCR was performed on RNase/DNase-free water including primers either against positive-or negative-sense RNA ( Supplementary Fig. S3d,e). In contrast, significant amplifications corresponding to either (+) or (-) vRNA occurred in PV-infected cells when the proper primers, enzymes, and reaction agents were present ( Supplementary Fig. S3f,g). In addition to these system controls, we analyzed melt curves of PV-infected samples to test whether the dye qPCR assay (SYBER) produced single, specific products. The appearance of single peaks indicated one melting event, corresponding to the positive target amplicon ( Supplementary  Fig. S3h) or the negative target ( Supplementary Fig. S3i). No amplification was observed in mock-infected cells ( Supplementary Fig. S3j). Consistently, both (+) vRNA and (-) vRNA were detected in PV-infected cells, and were absent in mock-infected cells ( Supplementary Fig. S3k). We found significant decreases (p ≤ 0.001) in the raw/unnormalized cycle threshold, C t (number of PCR cycles for the signal to exceed background) for PV ( + ) vRNA (genomic vRNA that can be used for both translation and as a template for (-) RNA synthesis) and for (-) vRNA (anti-genomic vRNA) within infectious microvesicles (ImVs) when compared to non-specific amplification from mock-infected microvesicles (MmVs) (Fig. 2a, ImVs vs. MmVs). These results established the presence of both (+) and (−) vRNA in ImVs.
The vRNA within secreted vesicles had been expected to be entirely virion-encapsidated vRNA because virion assembly is tightly regulated to encapsidate only (+) vRNA 19 . We therefore tested whether the newly identified intravesicular (−) vRNA was "free", or was packaged within virion capsids that were inside infectious microvesicles. We exposed microvesicles prepared as diagrammed in Fig. 1a to a series of further treatments (Fig. 2b) that included membrane disruption to release the vesicle content by freeze-thaw and detergent (sodium deoxycholate), followed by RNase treatment to remove any unprotected (intravesicular yet unpackaged in capsids) RNA. Because membranes are disrupted, and capsids are resistant to and unaffected by freeze-thaw, detergent, or RNases ( Supplementary Fig. S4), consistent with the literature [20][21][22] , the majority of (+) vRNA was still present in post-treated infectious microvesicles, as expected, whereas (−) vRNA was almost undetectable (C t ≥ 37, a value determined based on the negative reverse transcription and non-template controls) (Fig. 2a). Interestingly, however, there was a small but significant decrease (Fig. 2a, ImVs vs. FTDR ImVs, p ≤ 0.05) in the abundance of (+) vRNA in post-treated infectious microvesicles. These data indicate that not all (+) vRNA was protected inside assembled virions, as capsids are well known to protect their internal RNA from RNase-mediated degradation 19,22 . The status of this non-encapsidated RNA, whether single-or double-stranded, is not known, as under the conditions of the experiments both forms would be digested by RNaseA.
Host cell proteins identified by mass spectrometry analysis. Our data showed that less than 10% of the LC-MS identified protein peptides in microvesicles from PV-infected cells were viral proteins. Studying the components in more detail, we observed that the host cell protein components in microvesicles from mock-infected cells exhibited a much smaller diversity of proteins than infectious microvesicles from PV-infected cells (Supplementary Table S1). We identified a total of 65 host cell protein matches that were present in both independent experiments on infectious microvesicles, with three technical replicates combined per experiment, and a threshold cutoff of 0.999 probability. All five proteins identified in mock-infected sample (MmVs) matched proteins identified in ImVs (8% of the ImV proteins).
Although exosomes are typically smaller sized vesicles, some overlap between microvesicles and exosomes can be expected from a differential centrifugation separation 28,29 . Therefore, it is not surprising that components of the exosomal pathway were also identified in our size-fractionated infectious microvesicles. These components include PDCD6IP (programmed cell death 6 interacting protein, also known as ALIX), which participates in ESCRT-III recruitment 30 , syndecan binding protein (SDCBP), which is involved in the biogenesis and cargo loading of exosomes 31 , and several classical exosomal markers such as the ESCRT-III associated factor, increased sodium tolerance 1 (IST1) 32 and CD9 33 (Table S1).
The detection of these exosomal components led us to question whether PV infection might also exploit smaller exosome-like vesicles to transport virions and viral proteins from cell to cell, as has been shown for exosome-like virion-containing vesicles from HAV-infected cells 7 . To investigate their role in PV spread, exosomes shed from mock-and PV-infected cells were collected, using established size-based fractionation of 40-100 nm diameter vesicles and further purified using antibodies against the exosomal marker CD9 (see Materials and Methods), to avoid contamination of the exosome fraction with similarly sized free virions (28 nm diameter) and microvesicles, and used for functional infectivity characterization (plaque assays) and cryo-ET.

Intact extracellular vesicles were infectious and the carried contents altered cellular conditions.
Previous studies demonstrated that microvesicles transport virions. Classic infectivity plaque assays showed that intact infectious microvesicles produced fewer infectious centers than intravesicular virions released from these vesicles by freeze/thaw 10 . Confirming these previous results, our quantification of plaque assays revealed a 10-fold increase in plaque-forming units (pfu) after disruption of the infectious microvesicle membrane and release of intravesicular virions by freeze/thaw (FT) prior to infection (Figs. 2b, 3a, ImV vs. FT ImVs). We also saw a greater than 3-fold increase in pfu for infection with freeze/thawed-detergent/RNase-treated infectious microvesicles as compared to intact infectious microvesicles. We note that the number of infectious centers (pfu) was not significantly different when cells were infected with intravesicular PV virions released by either method: freeze/ thaw or freeze/thaw/detergent/RNase (FTDR) (Fig. 3a, FT vs. FTDR). Analogous experiments were completed on CD9-positive exosomes purified from the media of PV-infected cells. Exosomes from PV-infected cells were

Ingenuity Canonical Pathways -log(p-value) Ratio
Secreted infectious microvesicle (ImV) sample from PV-infected cells www.nature.com/scientificreports www.nature.com/scientificreports/ shown to be infectious, exhibiting a three-fold increase in plaque formation after disruption of the vesicles' membranous structures by freeze/thaw (Fig. 3b).
A linear increase in the MOI by PV is known to result in a proportional increase in (+) vRNA production in infected cells 34 . This suggests that if the additional vesicle contents here identified by LC-MS have no effect on viral replication, the (+) vRNA produced by intact vesicles or by the virions released from broken vesicles would be expected to be comparable. For example, suppose one infectious vesicle that contains 10 virions is added onto 10 cells, either as an intact vesicle or first freeze/thawed to release the intravesicular virions. Whether 1 cell is infected with 10 virions (local MOI = 10 in one cell, and 9 cells are uninfected) by this intact infectious vesicle, or whether 10 cells are each infected with 1 virion released from the broken infectious vesicle (local MOI = 1 in one cell, 10 infected cells in total), the measured vRNA production would be the same. A larger increase in (+) vRNA production by intact infectious vesicles would suggest the presence of additional virulence factors provided within the vesicle.
Thus, to evaluate possible function(s) of the complex content of infectious vesicles during early-stage infection, we performed experiments to test whether intact vesicles resulted in an increase in (+) vRNA production in newly infected cells beyond equal viral replication for infection with treated (broken) or untreated (intact) vesicles. The total collected infectious microvesicles were divided into three equal aliquots and used to infect the cells with 1) untreated, intact infectious microvesicles, 2) freeze/thawed, broken microvesicles to release free virions, or 3) freeze/thawed, broken microvesicles that were subsequently subjected to detergent and RNase treatment. A time course of (+) vRNA production by host cells was measured using RT-qPCR. At an early infection stage (3 hpi), intact infectious microvesicles initiated a more rapid onset of viral replication, evidenced by an over 2-log increase in (+) vRNA production in host cells compared to both freeze/thawed or freeze/thawed-detergent/ RNase treated groups (Fig. 3c, p ≤ 0.001). This result shows that intact infectious microvesicles accelerate RNA production in host cells at 3 hpi.

Ultrastructural analysis by cryo-electron tomography revealed multiple classes of infectious microvesicles. A remaining unanswered question was the 3-D structures of virus-induced extracellular ves-
icles that, as shown above, accommodate not only viral capsids, but also unencapsidated viral RNA, cellular proteins, and viral replication proteins. To visualize vesicle ultrastructure, PS-positive infectious microvesicles and purified CD9-positive exosomes secreted by PV-infected cells from 4 to 8 hpi were preserved by plunge-freezing, and then imaged by cryo-EM and cryo-ET. To enrich the infectious extracellular vesicle population for low throughput cryo-EM, we imaged PS-positive microvesicles and CD9-positive exosomes. PS-positive infectious microvesicles showed a wide size distribution, with diameters ranging from 70 nm to 820 nm. Approximately 90% of these microvesicles were 100 to 300 nm in diameter, with a median of 170 nm (n = 210 vesicles, Fig. 4a). The microvesicles displayed neither a strong uniformity in size, nor a strong correlation between size and number of included virions. 3-D reconstructions computed using cryo-ET tilt series revealed that 70% of the vesicles (n = 180 vesicles) contained virions. Quantification of virions per infectious microvesicle shows that 83% of infectious microvesicles carry 1 to 20 virions and, on average, each infectious microvesicle transports 10 virions (n = 150 vesicles from cryo-ET data, Fig. 4b).
Based on their internal morphology, we categorized infectious microvesicles into three classes (Fig. 4c). As shown in cryo-EM images in Fig. 5, Class I contained densely packed virions (Fig. 5a-d), Class II contained clustered virions and low density regions (Fig. 5e-f), and Class III included internal vesicular structure(s) (Fig. 5g). In addition to virions, all classes contained high density material that resembles a mat of threads or noodles (modeled in purple in Figs. 5 and 6). As with size versus number of virions per infectious microvesicle discussed above, there was no correlation between infectious microvesicle size and the packing arrangement of virions (i.e. infectious microvesicle class). For example, a relatively large infectious microvesicle with a diameter of 200 nm could have Class I morphology (Fig. 5c,d) or Class II features (upper right vesicle in Fig. 5e,f).
As expected, no virions were observed within any vesicles isolated from mock-infected (control) cells (Fig. 5h-k). In contrast to infectious microvesicles, the major recognizable structural components observed in control microvesicles were disordered or bundled actin filaments (Fig. 5h-j, green arrows), which is consistent with the mass spectrometry data showing an abundance of actin and actin-binding proteins in microvesicles secreted from mock-infected cells (Table S1). In contrast, actin filaments were rarely observed within infectious microvesicles; in the few examples where infectious microvesicles contained actin filaments, multiple virions were dispersed within the bundled actin ( Supplementary Fig. S5).
Three-dimensional reconstructions were computed using cryo-ET data from PS-positive infectious microvesicles and purified CD9-positive exosomes secreted by PV-infected cells from 4 to 8 hpi. Virions and density with a "mat-like" morphology were observed in the lumen of the microvesicles and exosomes (Figs. 6, 7, and Supplemental Movie). To confirm the presence of virions within the vesicles, we computed a subtomogram average of 118 structures from 1,022 particles we identified as virions, resulting in a reconstruction of the PV capsid at 6.9 nm resolution (Fig. 6e). Consistent with data from our 2-D cryo-EM images (Fig. 5), over 90% of the 3-D reconstructed infectious microvesicles showed either the Class I (Fig. 6a, deposited in the wwPDB under accession code EMD-7873) or Class II (Fig. 6b, wwPDB accession code EMD-7872) morphology, whereas Class III vesicles comprise only 10% of the population (Fig. 6c, wwPDB accession code EMD-7871, and Supplemental Movie). Infectious microvesicles were rarely spherical, and showed a single membrane enclosing an irregularly shaped structure, often with a scalloped outer contour, as seen in Fig. 5e,e' , (between the brown arrows). Class II infectious microvesicles (Fig. 6b), showed "empty" (low density) regions encompassing up to 90% of the infectious microvesicle volume, in addition to virions and mat-like structures. Class III infectious microvesicles contained inner vesicular structures as shown in Figs. 6c and 7, each representing a single-membrane vesicle entrapped in the lumen of an infectious microvesicle. The average ratio of diameter of the inner vesicle to that of www.nature.com/scientificreports www.nature.com/scientificreports/ its infectious microvesicle carrier was 0.5 ± 0.06 (n = 17). Less dense features were observed within inner vesicles (e.g. yellow region in Fig. 6c), as compared to the above-described mat-like structures of infectious microvesicles (Fig. 7, purple arrows). This suggests that structures inside the larger compartment of infectious microvesicles and those within inner vesicles may be distinct from each other. We also observed protein structures with globular 'heads' on a stalk in the membranes of both ImVs and their inner vesicles (Fig. 7, cyan arrows).
3-D structural analysis of CD9-positive exosomes secreted by PV-infected cells displayed a relatively uniform size, with an average diameter of 80 nm ± 27 nm and an average of 7 ± 3 virions per exosome ( Fig. 4d; n = 69; data from nine reconstructed tomograms). The predominant morphology of exosomes from infected cells corresponded to features seen in Class I infectious microvesicles, with a densely packed interior volume (Fig. 6d, and deposited in the wwPDB, accession code EMD-7879). www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
It has long been observed that non-enveloped viruses, such as those in the enterovirus genus, can propagate infection prior to lysis of host cells 35,36 . However, the mechanism for exit (and entry) of these viruses without disruption of the cell's plasma membrane is not well understood. Only recently has unconventional secretion of www.nature.com/scientificreports www.nature.com/scientificreports/ virion-containing extracellular vesicles been described [7][8][9][10]37 . Delivery of such a 'payload' containing multiple virions has the potential to propagate infection that might otherwise be hindered by frequent detrimental mutations in the viral genome that are present due to the inherently error-prone replication of RNA 34,38 . Such a diversity of virions can be achieved by either increasing the number of infecting free virions per cell (MOI) or by entry of a vesicle containing multiple virions into one cell.
Here, we have presented data showing that in addition to infectious virions, extracellular microvesicles secreted from PV-infected cells contain a complex mixture of unencapsidated (+) vRNA as well as (−) vRNA, ready-made viral replication proteins and host proteins. Additionally, CD9-positive exosomes are involved in the non-lytic cell-to-cell transmission. In our new model, we speculate that vesicles comprising virions and www.nature.com/scientificreports www.nature.com/scientificreports/ macromolecules contribute to enhanced viral transmission and establishment of new infection, because in addition to virions, infected cells receive all the components necessary to begin infection prior to translation of the vRNA that is packaged into virions. A schematic of conventional cellular secretion (Fig. 8a) is compared to unconventional secretion, diagrammed by the interplay between viral replication and packaging of contents for non-lytic exit in Fig. 8b. In this figure we designate structures in infected cells as "autophagosome-like" and "microvesicle-like", because while markers for these vesicles have been identified in PV-infected cells, it has not been unambiguously determined that these are either precisely autophagosomes or microvesicles. www.nature.com/scientificreports www.nature.com/scientificreports/ Vesicle components 1: viral RNAs and proteins. As has been shown previously 8,10,36 , we provide direct evidence that secreted infectious microvesicles transport virions from cell to cell (Figs. 3a, 5-6). We further show that exosomes are also infectious and involved in PV nonlytic cell-to-cell transmission (Fig. 3b). It has been proposed that multiple capsid-enclosed viral genomes that are transferred by vesicles en bloc into a single cell are the sole contributor to replication kinetics in the new host 10 . Indeed, this is the case for exosomes secreted from HAV-infected cells 37 . However, the delivery of (+) and (−) vRNA and viral replication proteins by extracellular vesicles into cells suggests that viral replication of PV may be facilitated by the delivery of (+) RNA, (−) RNA, and the viral proteins necessary for replication (e.g. the virus polymerase 3Dpol) together into new host cells. Support for these delivered components facilitating initial viral replication is our RT-qPCR data (Fig. 3c) showing an increase in (+) vRNA production in host cells at 3 hpi that is faster than infection with the same number of free naked virions 34 . However, because there are data showing a dependence of viral replication on cis-translation of a central region of the PV genome 39 , the question arises of roles for vesicle-transported non-structural proteins in trans. Perhaps the proximity of the proteins and RNA upon delivery overcomes the requirement of new translation, by eliminating the problem of diffusion-limited localization, such that the proteins are positioned as they would be, were they newly synthesized. This could result in production of a specific quaternary structure of vRNA necessary for replication 39 , providing a possible mechanism for subverting the cis-translation requirement for replication. Additionally, it remains to be explored whether this intact-ImV-induced jump-start to replication results in an overall increased vRNA production throughout the whole infection.
Vesicle components, 2: host proteins. For infection propagation, viruses alter cell processes, facilitating the viral life cycle. This includes utilization of host cell proteins as well as ribosomes for translatation of viral proteins. Examples include 1) virus-induced reorganization of the actin cytoskeleton for viral entry 40,41 and for transport of viral components within the cell 42 , and 2) utilization of host proteins during PV infection to shut down cap-dependent protein translation, so that protein production continues almost exclusively for (cap-independent) viral proteins 43 . With less than 10% of the proteins identified by mass spectrometry in secreted infectious vesicles being virus proteins, transported host cell proteins may have roles in advancing PV infection after cell-to-cell spread by vesicles. For example, our mass spectrometry data identified a significant enrichment of proteins from the glycolytic pathway in infectious microvesicles (p < 10 −8 , Table 1) that were not present in control microvesicles, including a strong and unique presence of pyruvate kinase PKM, the critical glycolysis rate-limiting enzyme. It is known that the presence of glucose during PV infection causes an average 170-fold increase in viral output 44 . Thus, a vesicle-infected host cell that has received PKM may be better prepared for viral replication than cells infected by free virions or infectious exosomes; PKM may assist in producing the large increase in glucose (or fructose) and glutamine known to be required for maximal PV replication 44,45 . Structures of infectious vesicles. Structure and function are often linked. Therefore, it is not surprising that after we identified the large number of components in infectious microvesicles and exosomes using biochemical methods, we were also able to visualize significant structural complexity in infectious vesicles that clearly contained internal components in addition to virions. Extracellular vesicles from PV-infected cells included dense structures with a mat-like morphology (Figs. 5-7), often seen adjacent to membranes, and/or in close proximity to virions. This juxtaposition suggests a possible involvement of protein-lipid interactions that results in the observed spatial arrangement of the cargo. While identification of the components that comprise the mat-like structures is not yet known, in our model these mats are composed of the non-capsid proteins identified by mass spectrometry (Table S1) and western blots (Fig. 1d); and vRNA identified by RT-qPCR (Fig. 2a). We note that unencapsidated (+) and (-) vRNA could be present as dsRNA, which would provide additional protection from the host immune system. Class III infectious vesicles, which are defined by the presence of internal membrane structures within the infectious microvesicle lumen, comprised only a minor population. It is currently unclear if Class III infectious vesicles are produced accidentally when a smaller vesicle happens to become entrapped, or if they originate from a cellular mechanism similar to the formation of multivesicular bodies, where the inclusion of additional components advances virus replication. Fig. 8a, is altered in response to intracellular stress, such as an invading pathogen 46 . Autophagosomes can fuse with early or late endosomes/MVBs to form amphisomes 47 , and unconventional secretion can occur via secretory autophagy 48 or lysosomal exocytosis 49 . Therefore, the endosomal, exosomal, lysosomal, and autophagic pathways are overlapping under some cellular conditions. Thus, it is not surprising that we identified both endocytic and exosomal marker proteins in infectious vesicles. We did not, however, identify autophagosome-associated lipidated LC3 (LC3-II). LC3-II was previously shown within cells, in autophagosome-like double membrane vesicles that contain virions; these autophagosomes are thought to be the precursor of secreted infectious vesicles [7][8][9][10]36 . This omission in our results, and previously in exosomes from HAV-infected cells 37 , is likely due to technical limitations in detecting lipidated proteins by LC-MS 50 .

cellular origins of infectious vesicles. Cellular transport in uninfected cells, diagrammed in
In our model (Fig. 8), the endosomal and autophagosomal pathways both feed into and provide exits from viral replication factories that then utilize unconventional secretion for non-lytic spread of infectious vesicles, thereby accelerating viral replication in new cells. Components within the MVBs are then either exported as exosomes using (endosomal) ESCRT machinery, or fuse with lysosomes, where contents are degraded. Distinct from the endosomal pathway, autophagy is a mechanism to entrap and degrade specific cellular compartments and invading pathogens within double-membrane vesicles. Termed autophagosomes, these vesicles then fuse with lysosomes and the contents are degraded. In uninfected cells stimulated by, e.g. stress or starvation, lysosomes are the convergence point for the autophagic and endosomal pathways. (b) As PV infection progresses, membrane remodeling and lipid synthesis produce replication factories (purple patches) for vRNA synthesis. Prior to cell lysis, infectious exosomes and infectious microvesicles are formed within PV-infected cells. Infectious exosomes are present within late endosomes (also called multivesicular bodies, or MVBs; green). Exosomes are secreted either directly from MVBs, or after their fusion with autophagosome-like double membrane vesicles (pink). Infectious microvesicle-like structures can either bud directly from the cytoplasm, as in uninfected cells (see a), or they can be secreted after fusion of autophagosome-like vesicles with MVBs. Packaged inside exported vesicles are virions (brown/yellow hexagons), viral proteins, host proteins, host RNA, and both template and genomic viral RNA (together depicted as purple patches). These virion-containing extracellular vesicles get internalized into a neighboring cell. A rapid initiation of viral replication is achieved by this transport and internalization of all components needed for replication: virions, viral proteins, cellular proteins, ribosomes, viral RNA.

Materials & Methods
cell culture and pV infection. HeLa cells (ATCC, Manassas, VA) were infected with Mahoney PV (gift from Dr. Karla Kirkegaard, Stanford University) using protocols from Burrill, Strings, and Andino 51 . Briefly, HeLa cells were cultured in low-glucose DMEM medium supplemented with 5% fetal bovine serum (FBS) and 1% penicillin-streptomycin-glutamine (supplDMEM; Atlanta Biologicals, Cat. S11150; ThermoFisher, Cat. 10378016, respectively) and grown at 37 °C (5% CO 2 ) to 60 to 80% confluence. After three washes with PBS + (PBS supplemented with 0.01 mg/ml MgCl 2 and 0.01 mg/ml CaCl 2 ), cells were infected either with PBS + or PBS + and PV stock at a multiplicity of infection (MOI) of 30 virions per cell, titrated by classic plaque assay (see below). After 30 min at 37 °C, cells were washed in PBS + and grown in supplDMEM. For the infectious microvesicle-induced infection assay shown in Fig. 3, infection was first synchronized by addition of infectious microvesicles on ice for 30 min to promote adherence, then washed with PBS+ , as in 52,53 prior to growth at 37 °C in supplDMEM. Because PV replication and release of infectious extracellular vesicles decreased when cells were cultured in non-bovine serum media 51 , we first infected HeLa cells in FBS-containing media for 4 h, during which the RNA replication rate reaches its maximum. We then minimized contamination of the extracellular vesicles by components from bovine serum (FBS) by washing (3×) and then growing the cells in non-FBS media before collecting vesicles at 8 hpi, when cells and/or the supernatant were collected for subsequent analyses.
To test for cell viability at the time of vesicle collection (Fig. 1b), HeLa cells were infected with PV at an MOI of 30 for 8 h, then stained in situ with a 0.2% Trypan Blue solution in PBS. Persistently PV-infected K562 cells were collected and stained with a 1:1 dilution of the cell suspension in 0.4% Trypan Blue-PBS. In each case, samples were double-blinded and two randomly selected fields of approximately 200 cells from each of three replicate wells were counted. Percentage viability was calculated as number of blue-staining cells divided by the total cells counted x 100.
RT-qPCR. RNA from whole cells or extracellular vesicles was extracted using the RNeasy Plus Mini Kit (Qiagen, # 74134), as in 61 . The full-length PV genome and its negative-sense template were amplified, in duplicate, through two-step RT-qPCR, as previously described 51 . Briefly, cDNA was synthesized using the SuperScript III RT (Life Technology, # 18080-093) system; RT 34 , with Tag primer to increase binding specificity, full-length production, and efficiency. The qPCR was performed using a master mix (Fast SYBR Green master mix system; Life Technology, # 4385610). Detection of RNA was performed in a two-step RT-qPCR using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, # 4374966) and the SYBR green master mix. In order to demonstrate the specificity of amplification, we conducted a series of controls including negative reverse transcription control and non-template controls, and melt curve analyses. As described in 62 , results were normalized to an endogenous control GAPDH of whole cells presented as ΔC t (where ΔC t = (C t of endogenous control gene (GAPDH)) -(Ct of gene of interest)) 62,63 . The mean of two technical replicates per cDNA sample was used to obtain raw C t and ΔC t for quantitative gene expression. The statistical analysis was obtained from six