In vitro evolution of high-titer, virus-like vesicles containing a single structural protein

Nina F. Rosea, Linda Buonocorea, John B. Schella, Anasuya Chattopadhyaya, Kapil Bahla, Xinran Liub, and John K. Rosea,1

aDepartment of Pathology, Yale University School of Medicine, New Haven, CT 06510; and bCenter for Cellular and Molecular Imaging, Yale University School of Medicine, New Haven, CT 06510

Edited* by Robert A. Lamb, Northwestern University, Evanston, IL, and approved October 22, 2014 (received for review August 6, 2014)

Self-propagating, infectious, virus-like vesicles (VLVs) are gener- RNA. These proteins form a complex that directs replication of ated when an alphavirus RNA replicon expresses the vesicular the genomic RNA to form antigenomic RNA, which is then stomatitis virus glycoprotein (VSV G) as the only structural protein. copied to form full-length positive strand RNA and a subgenomic The mechanism that generates these VLVs lacking a protein mRNA that encodes the structural proteins. The capsid protein has remained a mystery for over 20 years. We present evidence encases the genomic RNA in the cytoplasm and then buds from that VLVs arise from membrane-enveloped RNA replication facto- the cell surface in a membrane containing the SFV glycoproteins. ries (spherules) containing VSV G protein that are largely trapped Alphavirus RNA replication occurs inside light-bulb shaped, on the cell surface. After extensive passaging, VLVs evolve to membrane-bound compartments called spherules that initially grow to high titers through acquisition of multiple point muta- form on the cell surface and are then endocytosed to form cyto- tions in their nonstructural replicase proteins. We reconstituted pathic vacuoles containing multiple spherules (9). The replicase these mutations into a plasmid-based system from which high- proteins appear localized near the cytoplasmic side of the spher- ules (10). Positive-strand genomic RNA produced in the spherules titer VLVs can be recovered. One of these mutations generates is packaged into nucleocapsids before SFV budding. Alphavirus a late domain motif (PTAP) that is critical for high-titer VLV RNA replicons lacking structural protein genes can replicate ef- production. We propose a model in which the VLVs have evolved ficiently inside a cell, but they are incapable of propagating beyond in vitro to exploit a cellular budding pathway that is hijacked by the cell. many enveloped viruses, allowing them to bud efficiently from the We have been interested in developing VLVs as a vaccine cell surface. Our results suggest a basic mechanism of propagation platform (5, 6). However, the relatively low titers generated were MICROBIOLOGY that may have been used by primitive RNA viruses lacking capsid a major limitation of the system. We undertook the extensive proteins. may have evolved later to allow more efficient serial passaging studies described here to determine whether packaging of RNA, greater virus stability, and evasion of innate VLVs could evolve in culture to grow to high titers. We succeeded immunity. in generating VLVs that grow to at least 1,000-fold higher titers. In the process of studying these VLVs, we generated data sug- VSV glycoprotein | evolution | SFV replicon | late domain gesting the mechanism of VLV formation. In addition, our data suggest that the high-titer VLVs have evolved through passaging nveloped RNA viruses have highly organized structures. One to use a cellular budding machinery that is exploited by many Eor more capsid proteins encase their RNA, matrix proteins enveloped viruses to drive efficient budding. often lie between the capsid and the membrane, and one or more Many enveloped RNA viruses use components of a cellular vesicular budding machinery to drive efficient budding from the transmembrane glycoproteins can interact with the matrix or cell surface (11–13). Short sequence motifs called late domains capsid proteins to direct efficient particle assembly (1). Once the in their structural proteins recruit cellular protein complexes particles are released from cells, one or more glycoproteins in called ESCRT (endosomal sorting complex required for trans- the bind cellular receptors and catalyze membrane port). The ESCRT complexes are normally involved in budding fusion to allow the viruses to enter new cells (2). Vesicular stomatitis virus (VSV) is a negative-strand RNA virus Significance that encodes a single membrane glycoprotein (G), a matrix protein, and a nucleocapsid protein as well as two proteins that form the viral polymerase (3). Remarkably, when cells are transfected with All known membrane-enveloped RNA viruses have capsid an alphavirus RNA replicon encoding only the alphavirus non- proteins that encase their RNA genomes. This paper shows that structural replicase proteins and the VSV G protein, infectious, infectious, membrane-enveloped, virus-like vesicles with RNA self-propagating membrane-enveloped vesicles containing the VSV genomes can evolve in vitro to grow to high titers without G protein are generated (4). These infectious, virus-like vesicles a capsid protein. The infectious vesicles are apparently gener- (VLVs) grow to only low titers of 104 to 105 infectious units (i.u.) ated from RNA replication factories called spherules that bud per mL, but propagate like a virus in tissue culture cells. The from the cell surface. They evolve in vitro to bud with high vesicles contain the genomic RNA and VSV G protein, but unlike efficiency through the acquisition of multiple mutations in the known enveloped RNA viruses, they lack a capsid protein encasing non-structural replicase proteins. One mutation generates their RNA resulting in a low buoyant density (4). The mechanism a critical motif found in many viral structural proteins. This by which these VLVs are generated has never been determined. A motif is involved in recruiting cellular machinery to drive effi- nonspecific packaging was postulated because the VLVs contained cient budding. Prior to the evolution of capsid proteins, prim- both genomic RNA and subgenomic mRNA (4) and the titers itive RNA viruses may have used this budding mechanism. could be increased by sonication of the cells. Despite the low titers, VLVs expressing other proteins have proven useful as experi- Author contributions: N.F.R. and J.K.R. designed research; N.F.R., L.B., J.B.S., A.C., K.B., X.L., mental vaccines (5, 6). and J.K.R. performed research; N.F.R., L.B., J.B.S., and J.K.R. analyzed data; and N.F.R., L.B., The alphavirus replicon used in the studies described above and J.K.R. wrote the paper. was derived from Semliki Forest Virus (SFV), a positive-strand, The authors declare no conflict of interest. membrane-enveloped RNA virus that encodes four nonstructural *This Direct Submission article had a prearranged editor. proteins called nsP 1–4 and three structural proteins: capsid, and 1To whom correspondence should be addressed. Email: [email protected]. – the E1 and E2 transmembrane glycoproteins (7, 8). The nsP 1 4 This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. proteins are translated from the first two-thirds of the genomic 1073/pnas.1414991111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1414991111 PNAS Early Edition | 1of6 Downloaded by guest on September 24, 2021 of vesicles into cellular multivesicular bodies (MVB) as well as We compared the growth rate of the VLVs to VSV using tra- other cellular processes (12). , paramyxoviruses, filo- ditional one-step growth curves and found that the maximal VLV viruses, and rhabdoviruses have been shown to use components of titers (∼108 pfu/mL) were typically reached by between 12 and 24 h the ESCRT pathway to drive budding (11–13). However, other after high multiplicity infection. In contrast, maximal VSV titers 9 enveloped RNA viruses including SFV and virus do not (∼10 pfu/mL) were reached between 10 and 12 h after infection. use the ESCRT pathway in budding (14–16). Purified p50 VLVs Contain VSV G as Their Structural Protein. To de- Results termine what structural proteins were in the p50 VLVs, we × 6 Extensive Passaging Generates High-Titer VLVs. A plasmid with an infected 2 10 BHK cells with p50 VLVs or with VSV at a efficient promoter driving synthesis of SFV replicon RNA inside multiplicity of infection (MOI) of 10. After 24 h when the cells showed extensive cytopathic effect (c.p.e.), the VSV particles or cells was described previously (17). After transfection of this DNA the VLVs in the medium were purified through two rounds of onto cells, the replicon RNA is synthesized in the nucleus, and ultracentrifugation. The proteins in the VLVs and VSV particles then moves to the cytoplasm where it is translated and begins the were then analyzed by SDS/PAGE followed by silver staining. As replication cycle. We previously constructed a derivative of this shown in Fig. 2A, VSV virions contained the five proteins N, P, M, vector with two subgenomic SFV mRNA promoters (6). The first G, and L, whereas the purified p50 VLVs showed VSV G protein promoter drives expression of VSV G mRNA and the second as the only apparent structural protein. There was no trace of the promoter can drive expression of mRNAs encoding other antigens. SFV capsid protein (29.8 kDa) consistent with the loss of its ex- The infectious VLVs derived from these constructs are potent pression during early passages. The total yield of VLV protein experimental vaccines, but they released relatively low titers of only from 5 × 106 BHK cells infected with the p50 VLVs was ∼20 μg, about 104 to 105 infectious units (i.u.)/mL These titers could be whereas the same number of cells infected with VSV yielded increased about 100-fold by sonicating the cells (6). about 60 μg of total . We did not recover measurable To determine whether VLVs might be able to evolve to grow to protein yields from medium derived from uninfected BHK cells. high titers through passaging in tissue culture, we began with the DNA construct diagramed in Fig. 1. The SFV replicon RNA Characterization of the p50 VLVs by Electron Microscopy. The puri- generated by this construct expresses VSV G from the first sub- fied VLVs used for the protein characterization were first imaged genomic promoter and the SFV capsid protein from the second by transmission electron microscopy (TEM) following glutaral- promoter. We derived infectious VLVs from this construct after dehyde fixation and negative staining with uranyl acetate. Examples are shown in Fig. 2 B–D. Many of the vesicles appeared transfection of BHK cells and found low titers typical of VLVs B C (∼105 i.u./mL). We then continued passaging of the VLVs on to have protein spikes on their surfaces (Fig. 2 and , arrows). Some of the vesicles prepared for TEM by this method also BHK cells over a period of one year. In the initial passages we D transferred 5% of the medium from cells showing cytopathic ef- appeared to be broken and releasing nucleic acid (Fig. 2 ). We next performed TEM on VLVs that were unfixed, and fect (CPE) onto fresh BHK cells on 6 cm dishes, and strong CPE – labeled with antibody to VSV G followed by a secondary anti- typically developed in 3 5 d. Expression of the SFV capsid pro- body conjugated to 12-nm gold particles. The vesicles were then tein was completely lost by passage eight suggesting that its ex- negative-stained with phosphotungstic acid (Fig. 2E). The TEM pression was inhibiting particle production. With continued image shows clear surface labeling of numerous vesicles with passaging we noted increasing VLV titers and a much more rapid a halo of gold particles indicating the presence of VSV G on CPE developing in less than 24 h. By passage fifty we found that their surfaces. Vesicle diameters were in the 50–200 nm range. the VLVs had evolved to the point where they formed plaques on The size distribution was similar to that observed for the low titer 7 BHK cells within two days and attained titers of >5 × 10 plaque VLVs described previously (4). forming units (pfu)/mL At this point we picked a large VLV plaque and used this to generate a cloned VLV stock that we call Sequence of the p50 VLV Genome and Reconstruction into a Plasmid p50 VLV. This p50 VLV grows to titers of ∼108 pfu/mL, an in- DNA. To determine a consensus sequence of the p50 VLV ge- crease of ∼1000-fold over the starting vector. nome we performed reverse transcription and PCR to generate

Fig. 1. Large deletion and eleven amino acid changes occurred in high-titer evolved VLVs. A diagram of the plasmid pCMV-SFVG-CAP) used to derive VLVs for passaging is shown. The indicated promoter (CMV) drives expression of the positive strand replicon RNA encoding the nonstructural proteins. The subgenomic SFV mRNA promoters (small green arrows) drive expression of two mRNAs encoding VSV G or the SFV capsid protein from the full-length antigenomic RNA. After 50 passages of VLVs the consensus sequence of p50 VLV RNA derived from a single plaque showed the large deletion illustrated removing the second SFV promoter and the capsid gene. The x’s indicate the approximate positions of the 11 nucleotide changes that change amino acids in nsP proteins and in VSV G.

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1414991111 Rose et al. Downloaded by guest on September 24, 2021 TEM of Infected Cells Indicates a Mechanism for Generation of the High-Titer VLVs. In our original studies of the VLVs generated by expression of VSV G protein from the SFV replicon, the mechanism that generated the low-titer infectious vesicles was not determined. A possible explanation was that they were de- rived from SFV replication complexes that had been seen largely in membrane invaginations called spherules within cytopathic vacuoles (CPVs), but also occasionally on the cell surface (10, 18). More recent studies have shown that SFV spherules initially form at the cell surface, and are then endocytosed, eventually accumulating in the cytoplasmic CPVs (9). To examine the budding of the low vs. high titer VLVs, we prepared VLVs from the construct containing the 3′ deletion (pCMV-SFVG-Δ1672) and from the R,1,2,3 construct (pCMV- SFVG-p50R) containing the deletion and the 16 point mutations (Fig. 3). We used these VLVs to infect cells and performed TEM Fig. 2. Characterization of p50 VLVs. (A) Photograph showing SDS/PAGE of fixed, thin sections of cells at seven hours after infection. The analysis of VSV virions and p50 VLVs. Purified VSV virions or p50 VLVs (1 μg plasma membranes of cells infected with the low titer (Δ1672) of protein each) were fractionated by SDS-10% PAGE (Invitrogen) and the VLVs were studded with numerous ∼60- to 70-nm spherical proteins bands were stained and developed using a Pierce Silver Stain kit. structures that often appeared to be attached to the cell surface The positions and molecular weights of the VSV structural proteins N, P, M, by a thin “neck” (Fig. 4 A and B, arrows). We also found occa- G, and L are indicated. (B and C) TEM images of fixed and stained VLVs sional intracellular vacuoles that are the typical CPVs containing showing spikes protruding from the vesicle membrane. (D) TEM image of SFV spherules attached by a neck to the membrane (Fig. 4C). fixed and stained VLV apparently releasing nucleic acid. (E) TEM image of The structures on the cell surface appeared virtually identical to large field of unfixed and purified VLVs labeled with an anti-VSV G mouse the spherules seen in the CPVs, often including a dense central monoclonal antibody (I1) and a secondary anti-mouse antibody conjugated spot that is thought to be RNA (10). with 12nm gold particles. Scale bars for all TEM images are as indicated. In contrast to the results with the low titer VLVs, cells infected with the high titer p50 VLVs showed only occasional vesicles that appeared to be budding from the cell surface (Fig. 4 D and E). In overlapping DNA fragments covering the entire genome, and MICROBIOLOGY sequenced these fragments. The assembled sequence revealed many cases these were connected by a neck-like structure (arrows). a large deletion of 1,672 nucleotides (Fig. 1), which removed the These results suggested that the p50 VLVs might be budding ef- ficiently, whereas the low-titer VLVs remained mostly attached at second SFV promoter, the entire capsid gene, and all except 187 the cell surface. nucleotides of SFV sequences preceding the poly(A). In addition to this deletion, there were 16 single-base changes (Table 1). Ten A Late Domain Motif Evolved in the p50 VLV nsP1 Protein. Many of these changed amino acids in all four of the SFV non- enveloped viruses that bud from the cell surface recruit a cellular structural proteins, one changed an amino acid in VSV G, and budding machinery (ESCRT complex) that cells use to sort cargo four mutations were silent (Fig. 1 and Table 1). into vesicles that bud into multivesicular bodies (12, 13). Viruses To reconstitute high-titer p50 VLV sequence into a plasmid that use this pathway have sequence motifs called late domains DNA and begin to analyze the critical sequence changes, we began in their internal structural proteins that recruit components of a stepwise reconstruction (Fig. 3). First we introduced the 1,672- the ESCRT complex to the budding site. Mutations in these nucleotide deletion into the plasmid using a synthetic DNA motifs result in incomplete budding of virus particles from the fragment containing the deletion. We inserted this fragment be- cell and retention by a membranous neck (13). Because the low tween the unique Bpu10I and Spe I sites of pCMV-SFV-GCAP to titer VLVs appeared to retain large numbers of spherule-like generate pCMV-SFVG-Δ1672.TheVLVtitersobtainedafter structures on the cell surface, we examined the sequence changes 5 transfection of this DNA onto cells were only 2 × 10 per mL (Fig. in the p50 VLV genome to determine whether the passaging 3) indicating that the deletion had only a small effect on increasing VLV titers. We then used reverse transcription and PCR of p50 VLV RNA to generate DNA fragments spanning the indicated Table 1. Nucleotide and amino acid changes in p50 VLVs restriction sites, EcoRV-BstEII, BstEII-BglII, and BglII-Bpu10I. Nucleotide Protein These fragments containing the p50 VLV mutations were then change Amino acid change and context* affected used to replace the corresponding fragments in the pCMV-SFVG- Δ1672 plasmid DNA to generate the recombinants labeled R1, G-4700-A G-106-E (AASEKVL) nsP1 R2, and R2,3. All constructs were sequenced, and we chose only A-5424-G None those that matched the p50 VLV consensus. Interestingly, after G-5434A V-351-I (ATDITPE) nsP1 transfection of these initial recombinants onto cells, they yielded T-5825-C L-481-S (KRESIPV) nsP1 low and variable titers of VLVs (Fig. 3). However, when we put T-5930-C I-516-T (LVPTAPA) nsP1 all of the mutations together in recombinant R1,2,3 (plasmid A-6047-G D-555-G (QPNGVLL) nsP2 designated pCMV-SFVG-p50R), the high-titer phenotype was G-6783-A None recovered. These results indicated that some combination of G-6963-A None evolved mutations was required to generate the high-titer phe- G-7834-A A-1151-T (ALVTEYK) nsP2 notype. Also, one construct with a subset of the enhancing T-8859-A None mutations (R1 recombinant) grew poorly. We also determined in T-8864-C M-1494-T (AIDTRTA) nsP3 later constructs that the mutation N34D in the VSV G ectodo- G-9211-A A-1610-T (ERITRLR) nsP3 main was not required for generation of the high titer VLVs. A-10427-G N-2015-S (TLQSVLA) nsP4 To determine whether the enhancing mutations were affecting G-11560-A E-2393-K (SRYKVEG) nsP4 replicon RNA synthesis we performed Northern blots to detect A-11871-G N-34-D (NWHDDLI) VSV G replicon RNA or subgenomic G mRNA in cells infected with T-11978-C None VLVs derived from pCMV-SFVG-Δ1672 or pCMV-SFVG-p50R. We saw no significant differences in the amounts of RNA or in *Amino acids are numbered in the SFV nsP1-4 polyprotein, or in the VSV G the ratio of replicon to subgenomic RNA in the cells (Fig. S1). protein. Bold underlined text indicates the amino acid that was changed.

Roseetal. PNAS Early Edition | 3of6 Downloaded by guest on September 24, 2021 Fig. 3. Reconstruction of the p50 VLV genome shows that multiple mutations are required to generate the high-titer phenotype. The diagram illustrates the pathway of reconstruction of the p50 VLV genome sequence into the pCMV vector used to derive VLVs. The VLV titers obtained at 40 h after transfection of each construct into BHK cells are given on the corresponding line. Range of titers in two experiments is given.

might have selected for a late domain sequence in any of the and the replicon. Through extensive passaging we have been able SFV nonstructural proteins. We found that one of the four to derive VLVs that grow to titers of over 108 i.u./mL Multiple mutations in nsP1 changed the amino acid sequence PIAP to mutations in the SFV nsPs of the evolved VLVs are required for PTAP near the C terminus of nsP1 (Table 1). P(T/S)AP is generation of the high titers including a late domain motif that a common motif in internal viral structural proteins that binds to evolved near the C terminus of the SFV nsP1 protein. This motif the protein TSG101 to initiate recruitment of an ESCRT protein was necessary, but not sufficient for production of the high titers. complex and facilitate virus budding (19). SFV nsP1 protein contains an amphipathic membrane anchor The PTAP Motif Is Necessary but Not Sufficient for High-titer Particle Release. The low VLV titers obtained with the R1 recombinant (Fig. 3) suggested that the PTAP motif was not sufficient to generate the high-titer VLVs. However, this interpretation is confounded by the presence of three other mutations in nsP1 and one in nsP2 that might be deleterious in the absence of the other mutations that evolved in the nsP protein complex. To determine whether the PTAP mutation was necessary for high-titer particle production, we mutated just the PTAP motif in the p50 DNA vector (pCMV-SFVG-p50R) back to the original PIAP sequence. We then performed a one-step growth curve using the otherwise isogenic VLVs containing either the PIAP or PTAP sequences to analyze the kinetics of infectious particle release. The results of this experiment are shown in Fig. 5. Both constructs showed similar kinetics of growth reaching plateaus by 24 h. However, p50-PIAP mutant titer was reduced 50- to 100-fold compared with p50-PTAP at the later time points. These results indicate that the presence of the PTAP motif is important for generating the high titer. To determine whether the PTAP mutation was sufficient for generating high titer VLVs, we introduced the mutation gener- ating the PTAP motif alone into the pCMV-SFVG-Δ1672 plas- mid (Fig. 3) and derived VLVs. The VLV titers obtained after transfection of this DNA onto cells were only 1.1 × 105 per mL, indicating that the PTAP motif alone was not sufficient to drive efficient VLV production. Discussion We have described evolution of simple, infectious, VLVs that are Fig. 4. Thin-section TEM of cells infected with VLVs. (A and B)Examplesof generated in cells expressing a single structural protein, VSV G, large numbers of spherule-like structures on the surface of cells infected with from an RNA replicon (4). The RNA replicon was derived from low-titer VLVs derived from cells transfected with the pCMV-SFVG-Δ1672 an SFV expression vector (8) and encodes the nonstructural plasmid. (C) A typical CPV seen in a cell infected with VLVs. (D and E)Rare proteins that form the RNA-dependent RNA replicase, but none examples of what appear to be particles budding from cells infected with high- of the SFV structural proteins. Cells containing this replicon titer VLVs derived from cells transfected with the pCMV-SFVG-p50R plasmid. express VSV G protein on the cell surface and release low titers Arrows indicate examples of necks apparently attaching spherule-like struc- 5 (∼10 i.u./mL) of infectious vesicles containing VSV G protein tures to the membrane.

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1414991111 Rose et al. Downloaded by guest on September 24, 2021 For some viruses there is evidence that these compartments are lined with a shell of replicase proteins (25). In the case of SFV, there is no evidence that the spherules contain a replicase pro- tein shell. The nsPs appear to be located on the cytoplasmic side of the spherule neck (10). The membrane invaginations formed in the generation of multivesicular bodies by the ESCRT path- way lack an internal protein shell (26) and resemble the SFV spherules (27). Thus, it is possible that some components of the ESCRT pathway are normally involved with SFV spherule for- mation, but the components required for membrane scission are lacking or inhibited. The mutations selected in the p50 VLVs could result in recruitment of additional components required for scission or might release the inhibition of scission. By silver stain (Fig. 2) we did not detect structural proteins other than VSV G in the p50 VLVs suggesting that there is no major component forming a protein shell within them. It is interesting to speculate that at some early stage in the evolution of enveloped RNA viruses, spherules may have been Fig. 5. PTAP motif is required for high-titer VLV production. Cells were the direct precursors of primitive virus particles that budded infected at an MOI of 10 with high-titer p50VLVs derived from cells transfected with the pCMV-SFVGp50 plasmid or from a mutated form of the plasmid with the PTAP motif in nsP-1 reverted to PIAP. After a 30-min adsorption of VLVs at 37° the cells were washed twice in PBS and the medium was replaced with fresh DMEM containing 5% FBS. Small samples of the medium were then collected immediately (0 time) and at the indicated times thereafter, placed on ice and titered BHK cells using the infectious center assay.

sequence for the replicase complex of nsP1-4 and presumably tethers the nsP1-4 complex near or in the base of the spherules in MICROBIOLOGY which RNA replication is occurring (20, 21).

Model for Generation of the Low- and High-Titer VLVs. Fig. 6 shows a model for generation of the high-titer VLVs. SFV replicons not encoding VSV G are initially formed at the plasma membrane in the light-bulb shaped spherules with the replicase proteins near the base of the spherule neck (Fig. 6A). These spherules are then rapidly endocytosed and ultimately accumulate in the classic cy- topathic vacuoles (CPVs) seen in cells containing SFV replicons (9). In the case of SFV replicons that encode VSV G, endocytosis of the spherules may be inhibited by VSV G (22, 23), resulting in trapping of spherules on the cell surface (Fig. 6B). These spher- ules cannot bud efficiently, but are occasionally released from the cell surface to infect neighboring cells. They can also be released by sonication of the cells (4). After extensive passaging of VLVs, multiple mutations were selected including one generating a late domain PTAP motif near the C terminus of the SFV nsP1 protein. The mutations in addition to PTAP are perhaps required to create an appropriate structural context allowing access to binding of the ESCRT protein complex and rapid scission of the spherules from the cell surface (Fig. 6C). The requirement for more than one mutation to generate the high-titer VLV phenotype is consistent with the extensive passaging required to evolve it. Our model is based in part on our observation that large numbers of ∼60- to 70-nm spherule-like structures are present on the surfaces of cells infected with low titer VLVs, whereas cells infected with VLVs containing the PTAP and other mutations showed only occasional vesicles apparently budding from the cell surface. A similar accumulation of HIV-1 and other viruses is seen when viruses have mutations in their late domain motifs (13, 24). If the VLVs were derived exclusively from the budding of spher- ules, one might expect the purifed p50 VLVs to have a more uniform diameter. However, the purified VLVs that we observed Fig. 6. Model explaining origin of low-titer and evolved, high-titer VLVs. here and previously by TEM (4) ranged in diameter from about (A) SFV replicons not expressing VSV G first form in the light-bulb shaped 50–200 nm. One possible explanation for the range of sizes is spherules at the cell surface and are then rapidly endocytosed, forming classic CPVs. (B) Expression of VSV G protein from the SFV replicon inhibits fusion of two or more VLVs, potentially catalyzed by VSV G, to endocytosis of the spherules resulting in spherule accumulation at the cell generate larger vesicles. It is also possible that only the smaller surface. Occasional release of spherules containing VSV G generates low- VLVs are derived from spherules, and that the larger vesicles are titer, infectious VLVs that can infect new cells. (C) Extensive passaging of the generated through some other mechanism. VLVs selects for mutations that cause rapid VLV release from the cell surface. All positive-strand RNA viruses use some type of membrane- These high-titer VLVs evolved multiple mutations including a late domain bound compartments to sequester their RNA during replication. (PTAP) motif in nsP1 that promotes efficient VLV budding and high titers.

Roseetal. PNAS Early Edition | 5of6 Downloaded by guest on September 24, 2021 from cells and lacked capsids. Capsids may have evolved later to 24 h at 37°. The medium was then centrifuged at 800 × g for 5 min to allow more efficient packaging of RNA, greater virus stability, remove cells and cell debris. Ten ml of medium was then layered onto 28 mL and shielding from recognition by innate immune mechanisms. of 10% (wt/vol) sucrose in PBS and centrifuged at 25,000 rpm for 1.25 h in Our results led us to search the literature to determine whether a Beckman SW28 rotor. The pellet of VLVs was resuspended in 200 μL of PBS there are any known membrane-enveloped RNA viruses that lack and then layered onto 4.5 mL of 10% sucrose in PBS and centrifuged at a nucleocapsid. Interestingly several positive-strand RNA viruses 40,000 rpm for 1 h in a Beckman SW41 rotor. The pellet was resuspended in in the Pegivirus genus of the Flaviviridae family do not appear to 100 μL of PBS. VSV virions were grown and purified in parallel to provide encode a capsid protein (28, 29). However, biophysical evidence markers for SDS/PAGE gel electrophoresis. indicates that they acquire a capsid, perhaps a cellular protein, through an unknown mechanism (30). Electron Microscopy of VLVs and VLV-Infected Cells. VLVs prepared as above In parallel studies we have been examining the immunoge- (∼5 μL) were applied to carbon-coated EM grids made hydrophobic by nicity and potential pathogenicity of the p50 VLVs. We find that plasma ionization. Grids were fixed in 1% glutaraldehyde for 2 min, fol- these VLVs retain immunogenicity and lack pathogenicity. Thus, lowed by washing in PBS (pH 7.4) and water, then stained with 2% (wt/vol) these evolved VLVs have significant potential as vaccine vectors. uranyl acetate and air-dried. For immunogold labeling, the samples were applied to grids that had been preincubated with PBS containing 1% BSA. Materials and Methods Unfixed samples on girds were incubated directly with a mouse monoclonal Reconstruction of the p50 VLV Genome into a Plasmid DNA. The SuperScript III antibody to VSV G, washed with PBS containing 1% BSA and incubated RT-PCR kit from Life Technologies and 15 DNA primer pairs were used with RNA with donkey anti-mouse IgG conjugated to 12nm gold particles (Abcam). from p50VLVs to generate overlapping dsDNA fragments covering the p50 VLV Grids were then washed and stained with 2% (wt/vol) phosphotungstic acid genome. Sequences of the fragments were determined (Yale Keck Facility) and and then dried. Grids were viewed in a FEI 120kV transmission electron assembled using DNAstar software. The plasmid pCMV-SFVG-CAP (Fig. 1) was microscope. The digital images were acquired with a Gatan 4k × 4k generated by replacing the SIV gag gene from plasmid pBK-SFVG-E660Gag (6) CCD camera. with a synthetic gene encoding the SFV capsid protein. The 1,672-nucleotide BHK cells infected with VLVs were fixed at room temperature for 1 h in deletion found in the p50 VLV genome sequence was introduced in pCMV-SFVG- 0.1 M sodium cacodylate buffer (pH 7.4) containing 2% (wt/vol) glutaral-

CAP (Fig. 3) using a synthetic DNA fragment spanning the Bpu10I-SpeI sites. The dehyde. After rinsing with the same buffer, cells were postfixed in 0.5% OsO4 remaining mutations were introduced step-wise using the indicated restriction at room temperature for 30 min. Specimens were then stained en bloc with fragments prepared by RT-PCR from p50 VLV RNA. Specific point mutations were 2% aqueous uranyl acetate for 30 min, dehydrated in a graded series of generated using the Quik-Change II mutagenesis kit from Agilient Technologies. ethanol to 100%, and embedded in Poly/bed 812 for 24 h. Thin sections (60 nm) were cut with a Leica ultramicrotome and poststained with uranyl ace- Titering of VLVs. The p50 VLVs were routinely titered using serial dilutions and tate and lead citrate. Sample grids were examined in a FEI Tencai Biotwin a standard 2-d plaque assay on BHK-21 (ATCC CCL 10) cell monolayers. Low- transmission electron microscope at 80 kV. Images were taken using a Morada titer VLVs plaques were counted using a dissecting microscope or were titered CCD camera fitted with iTEM (Olympus) software. using indirect immunofluorescence microscopy to detect VSV G protein ex- pression in infectious centers. ACKNOWLEDGMENTS. We thank Robert Means and Michael Robek for helpful discussions and suggestions. We thank Gunilla Karlsson, Peter Liljestrom and Purification of p50 VLVs. Approximately 107 BHK-21 cells in DMEM with 5% Margaret Kielian for plasmids and reagents. This work was supported by NIH (wt/vol) FBS were infected with p50 VLVs at an MOI of 10 and incubated for Grants R37AI-040357 and R01AI-045510 (to J.K.R.)

1. Rossmann MG (2013) Structure of viruses: A short history. Q Rev Biophys 46(2): 16. Watanabe R, Lamb RA (2010) Influenza virus budding does not require a functional 133–180. AAA+ ATPase, VPS4. Virus Res 153(1):58–63. 2. Marsh M, Helenius A (2006) Virus entry: Open sesame. Cell 124(4):729–740. 17. Karlsson GB, Liljeström P (2004) Delivery and expression of heterologous genes in 3. Rose J, Whitt M (2001) Rhabdoviridae: The Viruses and Their Replication. Fields’ Vi- mammalian cells using self-replicating alphavirus vectors. Methods Mol Biol 246:543–557. rology, eds Knipe D, Howley P (Lippencott-Raven, Philadelphia), pp 1221–1240. 18. Grimley PM, Berezesky IK, Friedman RM (1968) Cytoplasmic structures associated with 4. Rolls MM, Webster P, Balba NH, Rose JK (1994) Novel infectious particles generated by an arbovirus infection: Loci of viral ribonucleic acid synthesis. J Virol 2(11):1326–1338. expression of the vesicular stomatitis virus glycoprotein from a self-replicating RNA. 19. Garrus JE, et al. (2001) Tsg101 and the vacuolar protein sorting pathway are essential Cell 79(3):497–506. for HIV-1 budding. Cell 107(1):55–65. 5. Rose NF, Publicover J, Chattopadhyay A, Rose JK (2008) Hybrid alphavirus-rhabdovirus 20. Spuul P, et al. (2007) Role of the amphipathic peptide of Semliki forest virus replicase – propagating replicon particles are versatile and potent vaccine vectors. Proc Natl Acad protein nsP1 in membrane association and virus replication. J Virol 81(2):872 883. Sci USA 105(15):5839–5843. 21. Varjak M, Zusinaite E, Merits A (2010) Novel functions of the alphavirus nonstructural – 6. Schell JB, et al. (2011) Significant protection against high-dose simian immunodefi- protein nsP3 C-terminal region. J Virol 84(5):2352 2364. ciency virus challenge conferred by a new prime-boost vaccine regimen. J Virol 85(12): 22. Whitaker-Dowling P, Youngner JS, Widnell CC, Wilcox DK (1983) Superinfection ex- clusion by vesicular stomatitis virus. Virology 131(1):137–143. 5764–5772. 23. Wilcox DK, Whitaker-Dowling PA, Youngner JS, Widnell CC (1983) Rapid inhibition of 7. Barth BU, Wahlberg JM, Garoff H (1995) The oligomerization reaction of the Semliki pinocytosis in baby hamster kidney (BHK-21) cells following infection with vesicular Forest virus membrane protein subunits. J Cell Biol 128(3):283–291. stomatitis virus. J Cell Biol 97(5 Pt 1):1444–1451. 8. Liljeström P, Garoff H (1991) A new generation of animal cell expression vectors based 24. Göttlinger HG, Dorfman T, Sodroski JG, Haseltine WA (1991) Effect of mutations af- on the Semliki Forest virus replicon. Biotechnology (N Y) 9(12):1356–1361. fecting the p6 gag protein on human immunodeficiency virus particle release. Proc 9. Spuul P, Balistreri G, Kääriäinen L, Ahola T (2010) Phosphatidylinositol 3-kinase-, actin-, Natl Acad Sci USA 88(8):3195–3199. and microtubule-dependent transport of Semliki Forest Virus replication complexes from 25. den Boon JA, Ahlquist P (2010) Organelle-like membrane compartmentalization of – theplasmamembranetomodifiedlysosomes.JVirol84(15):7543 7557. positive-strand RNA virus replication factories. Annu Rev Microbiol 64:241–256. 10. Froshauer S, Kartenbeck J, Helenius A (1988) Alphavirus RNA replicase is located on the 26. Wollert T, Hurley JH (2010) Molecular mechanism of multivesicular body biogenesis – cytoplasmic surface of endosomes and lysosomes. JCellBiol107(6 Pt 1):2075 2086. by ESCRT complexes. Nature 464(7290):864–869. 11. Rossman JS, Lamb RA (2013) Viral membrane scission. Annu Rev Cell Dev Biol 29: 27. Kallio K, et al. (2013) Template RNA length determines the size of replication complex – 551 569. spherules for Semliki Forest virus. J Virol 87(16):9125–9134. 12. Votteler J, Sundquist WI (2013) Virus budding and the ESCRT pathway. Cell Host 28. Simons JN, Desai SM, Schultz DE, Lemon SM, Mushahwar IK (1996) Translation initi- – Microbe 14(3):232 241. ation in GB viruses A and C: Evidence for internal ribosome entry and implications for 13. Weiss ER, Göttlinger H (2011) The role of cellular factors in promoting HIV budding. genome organization. J Virol 70(9):6126–6135. J Mol Biol 410(4):525–533. 29. Stapleton JT, Foung S, Muerhoff AS, Bukh J, Simmonds P (2011) The GB viruses: A 14. Bruce EA, et al. (2009) Budding of filamentous and non-filamentous influenza A virus review and proposed classification of GBV-A, GBV-C (HGV), and GBV-D in genus occurs via a VPS4 and VPS28-independent pathway. Virology 390(2):268–278. Pegivirus within the family Flaviviridae. JGenVirol92(Pt 2):233–246. 15. Taylor GM, Hanson PI, Kielian M (2007) Ubiquitin depletion and dominant-negative 30. Xiang J, et al. (1998) Characterization of hepatitis G virus (GB-C virus) particles: Evi- VPS4 inhibit rhabdovirus budding without affecting alphavirus budding. J Virol dence for a nucleocapsid and expression of sequences upstream of the E1 protein. 81(24):13631–13639. J Virol 72(4):2738–2744.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1414991111 Rose et al. Downloaded by guest on September 24, 2021