Recoding of the Vesicular Stomatitis Virus L Gene by Computer-Aided Design Provides a Live, Attenuated Vaccine Candidate

ABSTRACT

Codon pair bias (CPB), which has been observed in all organisms, is a neglected genomic phenomenon that affects gene expression. CPB results from synonymous codons that are paired more or less frequently in ORFeomes regardless of codon bias. The effect of an individual codon pair change is usually small, but when it is amplified by large-scale genome recoding, strikingly altered biological phenotypes are observed. The utility of codon pair bias in the development of live attenuated vaccines was recently demonstrated by recodings of poliovirus (a positive-strand RNA virus) and influenza virus (a negative-strand segmented RNA virus). Here, the L gene of vesicular stomatitis virus (VSV), a nonsegmented negative-sense RNA virus, was partially recoded based on codon pair bias. Totals of 858 and 623 silent mutations were introduced into a 5′-terminal segment of the viral L gene (designated L1) to create sequences containing either overrepresented or underrepresented codon pairs, designated L1^sdmax and L1^min, respectively. Analysis revealed that recombinant VSV containing the L1^min sequence could not be recovered, whereas the virus with the sdmax sequence showed a modest level of attenuation in cell culture. More strikingly, in mice the L1^sdmax virus was almost as immunogenic as the parental strain but highly attenuated. Taken together, these results open a new road to attain a balance between VSV virulence and immunogenicity, which could serve as an example for the attenuation of other negative-strand, nonsegmented RNA viruses.

IMPORTANCE

Vesicular stomatitis virus (VSV) is the prototypic rhabdovirus in the order Mononegavirales. A wide range of human pathogens belong to this family. Using a unique computer algorithm and large-scale genome synthesis, we attempted to develop a live attenuated vaccine strain for VSV, which could be used as an antigen delivery platform for humans. Recombinant VSVs with distinct codon pair biases were rationally designed, constructed, and analyzed in both cell culture and an animal model. One such recombinant virus, L1^sdmax, contained extra overrepresented codon pairs in its L gene open reading frame (ORF) and showed promise as an effective vaccine candidate because of a favorable balance between virulence and immunogenicity. Our study not only contributes to the understanding of the underlying mechanism of codon pair bias but also may facilitate the development of live attenuated vaccines for other viruses in the order Mononegavirales.

INTRODUCTION

Nonsegmented, negative-sense (NNS) RNA viruses include a wide range of established and emerging mammalian pathogens. For most of them, neither vaccines nor antiviral therapeutics with proven efficacy are available (1–3). Vesicular stomatitis virus, an NNS virus of the Rhabdoviridae, is an arthropod-borne pathogen that primarily causes acute vesicular disease in livestock (4). However, rare cases of human infection have been reported. Individuals infected with VSV usually develop mild flu-like symptoms, but a single case of encephalitis in a 3-year-old child has been related to VSV infection (5). Nevertheless, VSV serves as an attractive model to study a large number of mammalian pathogens of the Rhabdoviridae, including rabies virus (6).

Except for the members of the Orthomyxoviridae (e.g., influenza viruses), whose RNA polymerase is a tripartite complex (7), all the negative-strand RNA viruses encode a single large polymerase protein, L, which contains all the enzymatic activities involved in nucleotide polymerization (8) and mRNA modification (9–11). Amino acid sequence comparison of L proteins from multiple NNS viruses revealed six conserved regions (CR I to VI), which led to the notion that the large single L protein might be organized into a series of domains that retain independent catalytic activities (12).

VSV has been used extensively during the last four decades as a laboratory tool to study molecular virology and cell biology. Its relatively innocuous phenotype in human infection makes this virus attractive for clinical application, a potential realized only recently. However, VSV is an infectious agent with neurotoxicity in experimental animals, which poses serious safety concerns for direct human use (13, 14). Here, we report attempts to develop a live attenuated VSV strain, making use of a genetic phenomenon known as codon pair bias (CPB) (15, 16). CPB, which has been observed in ORFeomes of all organisms studied so far, describes an unexpected irregularity of synonymous codon pairing in genomes of a specific phylum, independent of its genome-wide codon usage. We previously reported that codon pairs are either overrepresented or underrepresented in the genome of a particular organism (17). The nature of a codon pair can be defined by calculating its codon pair bias score (CPS), which is the natural logarithm of the ratio of the observed frequency of the codon pair to the expected frequency. The average of the sum of all CPSs in an ORF indicates the codon pair bias of the entire gene. In our previous experiments involving recoding of viral genomes, we have shown that the viral gene expression can be intentionally altered by a rational design of a coding sequence, based on codon pair bias (17–20). More importantly, the alteration of gene expression is likely a result of the numerous synonymous substitutions, which makes phenotypic reversion nearly impossible (21). This specific property of virus attenuation caused by changing codon pair bias is critical for the development of vaccine strains against RNA viruses, like VSV, described here, whose spontaneous mutation rates are much higher than those of DNA viruses and of other microorganisms (22).

In this study, we examined the impact of recoding the VSV L gene on virus replication in cell culture and on pathogenicity in mice. By using computer-aided design and chemical cDNA synthesis as described before (17, 19, 20), we constructed two recombinant versions of VSV, containing an excess of either underrepresented (“min”) or overrepresented (“sdmax”) codon pairs. Because of the large size of the L ORF, we focused on the N-terminal segment of the L gene, designating the recoded segments L1^min and L1^sdmax. The L1^min design led to a nonviable virus, probably because of poor expression of the L gene. Surprisingly, the L1^sdmax design caused modest attenuation in cell culture, whereas this variant was highly attenuated in the mouse model. Moreover, the L1^sdmax virus exhibited potent immunogenicity that fully protected vaccinated mice from lethal challenge with wild-type (WT) VSV. These observations suggest that the L1^sdmax virus could be developed into a promising vaccine candidate or a vector for the delivery of immunologically important antigens.

RESULTS

Construction and characterization of L recoded VSV variants in tissue culture.

We recoded a large N-terminal segment of the L gene ORF, L1, by computer algorithm-aided design (Fig. 1A). Without alteration of codon usage or amino acid sequence, we rearranged the positions of existing codons to generate either underrepresented or overrepresented codon pairs, resulting in hundreds of synonymous nucleotide substitutions (see Table S1 in the supplemental material). Compared to its WT counterpart, L1^min has a much lower codon pair score, whereas L1^sdmax has a significantly higher one (Fig. 1B). We took into account a possible increase in the frequency of XXCpGXX, a dinucleotide formed between two codons that may play a role in inducing an innate immune response (23), and of XXUpAXX, a dinucleotide that at high frequency has been proposed to reduce mRNA stability (24). We particularly avoided generating extra AAAUU (AU-rich) sequences, which are signals for mRNA degradation (25). The distribution of overrepresented and underrepresented codon pairs in these designs is graphed in Fig. S1 in the supplemental material.

FIG 1 .

Genome organization of VSV and construction of L variants. (A) The VSV genome consists of 2 untranslated regions (3′ leader and 5′ trailer; red boxes) and 5 viral genes (N, P, M, G, and L). Like the other four viral genes, the VSV L ORF is flanked by the untranslated regulatory sequences (white boxes) that are essential for transcription initiation and termination. The L1 region (blue box) is between nt 4961 and nt 7534, corresponding to amino acids 76 to 934 of the L protein. L1^sdmax and L1^min sequences were synthesized de novo and incorporated into the parental plasmid pVSV1(+) to replace the WT sequence, using HpaI and AvaI sites. (B) Codon pair bias of human genes and the L1 derivatives plotted against their amino acid lengths.

A recombinant VSV bearing a “deoptimized” L gene (L1^min) is expected to be highly attenuated and possibly nonviable. On the other hand, based on our previous experience in “maximizing” the poliovirus P1 region, we anticipated that sdmax might not affect L gene expression negatively (17). We initially included the sdmax design in our study in order to rule out the possibility that certain unidentified RNA sequences critical for virus replication were accidentally modified during the large-scale recoding (26). As shown in Table S1, about 33% of the nucleotides of the L1 region have been mutated in the L1^sdmax design; therefore, theoretically, any important RNA signal or highly structured element longer than 3 or 4 nucleotides (nt) would have been inactivated.

cDNAs corresponding to these recoded regions were chemically synthesized de novo and cloned into the parental antigenomic plasmid pVSV1(+) to replace the WT sequence. The reverse genetics of these recombinant viruses was carried out using the vaccinia virus-mediated system (27). The L1^sdmax design yielded infectious virus after one blind passage, whereas L1^min failed to generate a viable virus.

The growth phenotypes of the recombinant L1^sdmax virus was then analyzed in tissue culture. Compared to the WT virus, the plaque phenotypes revealed slightly smaller sizes in BHK-21 cells and A549 cells (Fig. 2A and B). Multistep growth kinetics, analyzed in BHK-21 or A549 cells by inoculation at a low multiplicity of infection (MOI) of 0.1 or 0.001, showed maximal titers of L1^sdmax between 24 and 48 hpi that were comparable to that of the WT virus. Viral replication of L1^sdmax, however, was slightly delayed, independent of the cell type it infected. A quantitative reverse transcription-PCR (RT-PCR) assay that measured the level of viral full-length RNA in infected BHK-21 cells supported the delay of genome synthesis (Fig. 2C). To measure the yield of viral proteins and expand the range of cell types, we then determined the synthesis of N protein by immunocytochemistry in HeLa cells. In L1^sdmax-infected cells, at 4 hpi, much less viral N protein was detected than in WT virus-infected cells (see Fig. S2 in the supplemental material).

FIG 2 .

Multistep growth kinetics and plaque phenotypes of WT and L1^sdmax virus in BHK-21 cells (A) and A549 cells (B) at low MOIs. (C) Quantitative RT-PCR was conducted to quantitate the full-length viral RNA in WT- and L1^sdmax-infected BHK-21 cells (MOI = 3).

Reduced expression of the L protein encoded by the L1^sdmax sequence.

During the early stages of virus infection, the slower growth of the L1^sdmax virus led to a reduced yield of all the viral products (see Fig. S3 in the supplemental material). We were unable to explain the decreased level of L protein synthesis by simply comparing the yields of viral protein to those of mRNA. To circumvent this problem, we constructed two L expression plasmids (designated pL^L1sdmax and pL^L1min) (see Fig. S4 in the supplemental material) by replacing the cognate WT L open reading frames of pL with the synthetic L derivatives (L1^sdmax and L1^min). We then examined the production of full-length L protein after transiently transfecting BHK-21 cells that were previously infected with vTF7-3. In this experimental setting, the entire L gene coding sequences are transcribed in the cytoplasm by T7 polymerase.

To increase the stability of the large polymerase polypeptide, we cotransfected into BHK-21 cells a plasmid expressing the VSV P protein together with the plasmid harboring the modified L sequence (28). In comparison to translation of pL^WT, translation of the L1^sdmax gene produced only a weak signal (Fig. 3A), even though equal amounts of L transcripts were detected in pL^WT- and pL^L1sdmax-transfected cells (Fig. 3B). This result was unexpected given that codon pairing in L1^sdmax has been designed to yield mRNA favorable for translation (17). The synthesis of L1^min was even more inefficient than that of L1^sdmax (see Fig. S4B in the supplemental material). This was not totally unexpected, although we do not know as yet whether the poor L1^min synthesis is due to an inhibition of translation per se or a rapid degradation of mRNA prior to or during translation.

FIG 3 .

The expression of L1^sdmax sequence was examined by transfecting pL^L1sdmax plasmid into BHK-21 cells that were previously infected with T7 polymerase (Pol)-expressing vaccinia virus vTF7-3. (A) The L protein expression in transfected cells was examined by Western blotting. (B) The synthesis of L mRNA was examined by quantitative RT-PCR (qRT-PCR). pt, posttransfection.

Another interesting observation is that the L1^sdmax virus showed an exquisite sensitivity to the pharmaceutical deprivation of Hsp90 (see Fig. S5 in the supplemental material), a host chaperone molecule that has been identified to be essential for the proper folding of the newly synthesized VSV L protein (29). It raises the possibility that an enhanced chaperone engagement may be required for the L1^sdmax virus.

Diminished virulence of the L1^sdmax virus after intranasal inoculation of mice.

As mentioned above, the poliovirus variant P1^max in which the capsid coding region was codon pair optimized, was as lethal in CD155 transgenic mice as WT poliovirus (17). We therefore tested the pathogenesis of L1^sdmax virus in a mouse model. Five- to six-week-old male BALB/c mice were infected with different doses of WT or L1^sdmax viruses intranasally. WT-infected mice lost weight rapidly and exhibited typical symptoms of infection starting around 3 dpi. Under our experimental conditions, the 50% lethal dose (LD₅₀) for WT VSV was 10⁵ PFU for intranasal infection. In contrast, neither mortality nor severe morbidity was observed in any mice infected with the L1^sdmax virus, even at a very high infection titer (2 × 10⁸ PFU) (see Fig. S6B in the supplemental material). Taking these data together, we conclude that L1^sdmax virus is highly attenuated in BALB/c mice. This was a surprise, but it matched our unpublished results obtained with recoded dengue virus variants that phenotypes observed in tissue culture with a codon pair-recoded virus do not necessarily predict phenotypes of the same recoded virus when it is tested in experimental animals.

Virus burdens in the infected organs.

The efficacy of live attenuated vaccines depends on reduced viral replication in vivo but a successful stimulation of immunological memory (18). Moreover, virulence and virus load in infected organs are usually positively correlated (19). Therefore, we examined virus burden in the infected animals by both plaque-forming assays and quantitative RT-PCR. At 3 dpi, the virus load in the L1^sdmax-infected organs was at least two orders of magnitude less than that in their WT-infected counterparts. On day 6, the infectious L1^sdmax virus could not be recovered in any of the mice lungs that we examined and could be detected in only one brain sample (n = 3) (Fig. 4A). In contrast, in WT-virus-infected animals, a virus burden of approximate 10⁴ PFU/g was observed in brain tissues up to 6 dpi, which ultimately resulted in the death of all the remaining animals by 9 dpi (data not shown). We also examined the viral RNA load in mouse brains using quantitative RT-PCR (Fig. 4B). Total RNA was extracted from infected brains, and a 150-bp fragment corresponding to the viral N gene was amplified. Viral RNA was present in every brain of the infected animals, but the viral RNA load of L1^sdmax-infected animals was lower than that of WT virus-infected ones.

FIG 4 .

Attenuation of L1^sdmax VSV in mice. (A) BALB/c mice were infected with WT or L1^sdmax, and viral loads in organs were determined by plaque assay. At 6 dpi, the infectious viruses were undetectable in all but one WT-infected lung, and the presence of infectious virus could be detected in only one L1^sdmax-infected brain (n = 3). The dashed line indicates the limit of detection (30 PFU/g) for plaque assay. (B) The viral RNA loads in mice brains were measured using quantitative RT-PCR (**, P < 0.01 [two-tailed Mann-Whitney U test]).

Disparate histopathology of WT- and L1^sdmax-infected brains.

Even the limited replication of L1^sdmax in the brains of infected animals raises the question of whether such viral variants could serve as vaccine candidates. We therefore carried out histological studies. BALB/c mice were infected intranasally with 10⁷ PFU of viruses, and brains were harvested for hematoxylin-and-eosin (H&E) staining after 5 days. Brain tissues infected with WT VSV exhibited severe immunopathological characteristics, evidenced by multifocal perivascular cuffing, mononuclear cell infiltration in the parenchyma, meningeal mononuclear cell infiltration, and hemorrhage (Fig. 5), consistent with previous studies (14, 30). Tissues collected from the L1^sdmax group exhibit only mild, if any, inflammatory changes upon virus infection.

FIG 5 .

Histopathology after VSV infection. Mice were inoculated with PBS (A and B), 10⁷ PFU WT (C and D), or L1^sdmax (E and F) virus intranasally, and the horizontal brain sections containing cerebral cortex were stained with H&E at 5 dpi. Representative brain histology is shown for each group (magnification, ×40). In panel C, solid yellow arrows indicate the perivascular cuffing, and dashed yellow arrows indicate the mononuclear cell infiltration in the brain parenchyma. In panel D, solid black arrows indicate meningeal hemorrhage, and dashed black arrows indicate the meningeal mononuclear cell infiltration.

VSV infection compromises the integrity of the blood-brain barrier (BBB) (31). The destruction of BBB accompanied by the intensive immunological infiltration causes severe immunopathogenesis that leads directly to the death of experimental animals. Therefore, we used the well-established Evans blue assay to assess the integrity of the BBB after virus infection, a method providing an additional indicator of virus pathogenesis (32). As shown in Fig. S6C in the supplemental material, at 6 dpi the brain of the animal infected with L1^sdmax showed no sign of BBB damage, whereas a widespread, intensive parenchyma staining was apparent in the WT-infected brain, which is a clear sign of diminished BBB integrity.

The immune response in L1^sdmax-infected mice.

We next determined whether the restricted replication of this viral variant provides effective immunological protection against challenge with WT virus. We vaccinated 5- to 6-week-old BALB/c mice with different doses of L1^sdmax virus intranasally. After 3 weeks, all mice received 10⁷ PFU of WT virus intranasally, representing an approximately 100-fold LD₅₀ of WT VSV. None of the L1^sdmax-immunized mice exhibited signs of disease, and their body weights increased steadily, whereas age-matched naive mice succumbed to challenge after 6 to 8 days (Fig. 6A).

FIG 6 .

Immunogenicity of the L1^sdmax virus. (A) Dynamics of L1^sdmax-immunized mice body weights after a lethal WT challenge at day 21. (B) Titers of virus-specific neutralizing antibody (NAb) in WT- and L1^sdmax-immunized mice 21 days after the primary exposure. For this experiment, n was 5 per group. A larger number of animals (n = 10) for the groups infected with 10⁵ PFU of the WT were initially immunized, because the infection titer is close to the LD₅₀ of WT virus. The sera from the survivors (4/10) were subsequently collected at 21 dpi, and the titers of virus-specific neutralizing antibody were determined as described in Materials and Methods.

We then asked whether the L1^sdmax immunization elicited a B cell-mediated humoral immunity. Sera were collected 21 days postvaccination from tail vein blood, and titers of virus-specific neutralizing antibodies (NAb) were measured. The L1^sdmax virus was capable of inducing a robust protective antibody response (Fig. 6B). A similar level of NAb was detected in mice that were immunized with the same amount of WT or L1^sdmax virus, whereas the control group did not develop a detectable humoral response against VSV infection. Taken together, these data indicate that a single-dose vaccination of the L1^sdmax virus can elicit a potent immune response that fully protects mice from a subsequent virulent challenge.

Prolonged induction of proinflammatory cytokines and chemokines in WT- but not in L1^sdmax-infected brains.

During VSV infection, mouse susceptibility correlated positively with extensive cytokine/chemokine expression in the central nervous system (CNS) (14). To determine the nature of the inflammatory response during L1^sdmax infection, we profiled the transcription levels of several proinflammatory cytokines and chemokines in virus-infected brains. Compared to their WT-infected counterparts, MCP-1 and RANTES were induced to a lesser extent in L1^sdmax-infected mice (Fig. 7A). Additionally, at 3 dpi, L1^sdmax-infected mice showed levels of tumor necrosis factor alpha (TNF-α) and interleukin 6 (IL-6) similar to those of WT-infected animals (Fig. 7B). WT-infected mice retained a high level of cytokine production until 6 dpi, while L1^sdmax-infected mice showed a significant decline of cytokine synthesis by then. This suggests that in contrast to the highly attenuated L1^sdmax virus, WT virus infection in the CNS cannot be controlled without a sustained inflammation, which eventually leads to immunopathogenesis and irreversible neuron damage.

FIG 7 .

Cytokine/chemokine profile after virus infection. (A) Quantitative RT-PCR to measure chemokine (MCP-1 and RANTES) expression relative to the mouse β-actin after virus infection. (B) Proinflammatory cytokine induction in PBS-, WT-, and L1^sdmax-inoculated mice. Data were collected from 4 to 6 mice per group. Statistical significance is indicated by asterisks (*, P < 0.05; **, P < 0.01 [two-tailed Mann-Whitney U test]).

DISCUSSION

Impact of codon pair deoptimization on gene expression.

In this study, we evaluated the consequences of large-scale L gene recoding upon VSV replication and pathogenesis. The L1 region comprises the coding sequences of domains I to III and part of domain IV, which primarily forms the core ring structure of the VSV RNA-dependent RNA polymerase (RdRp) and catalyzes RNA polymerization (33). The L1^min sequence yielded only trace amounts of full-length L protein after transient expression in tissue culture cells, which is reminiscent of data obtained with variants of codon pair-deoptimized poliovirus and influenza virus (17–19).

Codon pair deoptimization appears to affect several mechanisms that interfere with gene expression in a context-dependent manner. For example, it might be difficult for the ribosome to read through a string of “rare” codon pairs, a predicament leading to less protein production per mRNA. In addition, it is known that the dinucleotide frequency of CpG and UpA in eukaryotic RNA viruses, as well as in mammalian mRNA in general, is lower than expected (34–36), a phenomenon that must be advantageous for viruses. Generating rare codon pairs by rearranging existing synonymous codons inadvertently leads to an enrichment of UpA and CpG dinucleotides in the mRNA sequences. For instance, the L1^min but not the L1^sdmax design contains significantly more XXUpAXX and XXCpGXX dinucleotides than the WT sequence (see Table S1 in the supplemental material), with the increase mapping to between codons, as shown, since codon use remained unchanged. Recent studies suggest that the dinucleotide composition in RNA virus genomes influences RNA virus replication, possibly because of an unclear host-pathogen interaction (37) and/or increased instability of the mRNA (23). Although we cannot exclude the possibility that strengthened host recognition of UpA/CpG dinucleotides contributed to the attenuation phenotype of our codon pair-deoptimized viruses, the reverse genetics of L1^min was conducted in BHK-21 cells, a cell line well known for its deficiency in innate immune response (38). Hence, the nonviable phenotype of the L1^min design cannot be simply attributed to an enhanced host response.

Impact of codon pair “scramble-max” on protein synthesis.

The most striking phenotype that we observed with L1^sdmax is its attenuation that results from an enrichment of overrepresented codon pairs. Our previous experiments in which we “maximized” the encoding of the poliovirus P1 region suggested that such genetic alteration enhanced the expression of the corresponding viral ORF without affecting virulence (17). This appears to be contrary to the observation in this study. The attenuation of L1^sdmax is not host range restricted, because the virus forms smaller plaques in all cell lines tested. Considering that the transcriptional regulatory sequences in the untranslated region of the L gene were never altered, it becomes less likely that the L1^sdmax sequence is inadequately transcribed by the viral polymerase. Indeed, the poor expression of the L1^sdmax sequence does not seem to occur at the transcription level, because (i) a high level of L mRNA was detected when the L1^sdmax sequence was placed under the control of the T7 promoter and transfected into cells expressing T7 polymerase (Fig. 3B) and (ii) during L1^sdmax virus infection, the same kinetics of mRNA accumulation were found for the L and N genes (see Fig. S3 in the supplemental material) (ranging from 1 hpi to 6 hpi).

Considering the possibility that an enhanced chaperone engagement might be required for the L1^sdmax virus (see Fig. S5 in the supplemental material), we hypothesized that the alteration of the translation elongation rate, presumably caused by the accumulation of overrepresented codon pairs, might interfere with the coordination between translation and other events closely related to translation rate, such as protein folding. Recent reports have shown that clustered rare codons are frequently distributed between defined protein domains, especially for large multidomain proteins. This distribution pattern of “slow” codons might collectively reduce the read-through activity of the ribosome, which turns out to be beneficial for proper protein folding (39, 40). An early study conducted in a bacterial system also suggests that codon pair utilization influences protein folding by altering translational efficiency, although only a small number of codon pairs was included and analyzed at that time (41). Herein we speculate that the folding efficiency of a multidomain protein, like the VSV L protein described in this paper, might be perturbed by the large-scale genome recoding. The relatively low codon pair bias in the WT L1 region (Fig. 1B) might be a compromise between translation accuracy and folding efficiency.

L1^sdmax serves as a candidate for VSV-based vaccines and vectors.

VSV has been studied as an agent with unique oncolytic properties aided by its broad tissue tropism and exquisite sensitivity to innate immunity, which result in its preferential replication in tumors. In addition, VSV might serve as an excellent vaccine delivery platform, because of its easily manipulated genome and the capacity to stimulate a robust cytokine response via mucosal delivery. In this study, we generated and identified a recombinant virus, L1^sdmax, which is highly attenuated (its LD₅₀ is at least 2,000-fold more than that of the WT). Our results showed that L1^sdmax underwent a limited replication and dissemination in vivo that barely caused any morbidity, but this self-restricted infection was highly immunogenic. The L1^sdmax virus appears to be an attractive live vaccine candidate applicable to other viruses in the order of Mononegavirales that encode very large polymerase polypeptides. In the development of vaccines, conventional strategies usually lead to a limited number of mutations in viral genomes, whereas L1^sdmax carries in its L-encoding segment 858 point mutations. We have passaged the L1^sdmax virus 5 times in BHK-21 cells at low MOIs, and we did not observe any phenotypic reversion (data not shown), but long-term passaging is required to confirm genetic stability. Of course, the current construct may have to be further attenuated to avoid all proliferation in the CNS.

The recent Ebola outbreak in West Africa and the sporadic cases in the Western Hemisphere have elicited international concern. One of the most promising vaccine candidates against Ebola infection, VSV-ZEBOV, is a VSV chimera that expresses Ebola genetic information (42). It will be of interest to further evaluate the safety and efficacy of L1^sdmax as a platform to deliver foreign antigens.

In summary, the present study describes the first large-scale recoding of VSV, a rhabdovirus that belongs to a large family of human pathogens. The outcome of the experiments is surprising but nevertheless may herald a unique approach to develop live attenuated strains for VSV and possibly other NNS viruses.

MATERIALS AND METHODS

Cells and viruses.

BHK-21, Vero, A549, and HeLa cells were maintained in Dulbecco’s modified Eagle medium (DMEM) plus 10% fetal bovine serum at 37°C. WT VSV is recovered from the cDNA clone pVSV1(+), using the well-established vaccinia virus-mediated system (27). Briefly, to construct the cDNA clone of L1^sdmax or L1^min virus, the HpaI-to-AvaI fragment of pVSV1(+) was replaced with the corresponding synthetic fragment L1^sdmax or L1^min (GenScript, NJ, USA) in which overrepresented or underrepresented codon pairs were substantially enriched. The recovered viruses were plaque purified, and their full-length sequences were confirmed by RT-PCR.

Virus load and protection assay.

Animal studies were carried out in accordance with a protocol that adhered to the Guide for the Care and Use of Laboratory Animals of the NIH (43) and was reviewed and approved (approval number 354543) by the Stony Brook University Institutional Animal Care and Use Committee (IACUC). Five- to six-week-old male BALB/c mice (Taconic Farms, Inc.) were intranasally infected with 5 × 10⁶ PFU viruses, and at 3 and 6 dpi, their lungs and brains were harvested for determinations of virus titers by plaque-forming assay. For viral RNA load analysis, total RNA was extracted from brains of mice (TRIzol; Life Technologies) that were infected with 5 × 10⁶ PFU viruses, and the abundance of viral RNA was subsequently evaluated by real-time RT-PCR using a primer pair that recognizes the viral N gene (VSV N primer) (see Table S2 in the supplemental material). In parallel, qPCR directed against mouse β-actin mRNA was performed to control for any differences in the amount of total RNA inputs (mouse β-actin primer) (Table S2 in the supplemental material). For protection assay, mice were vaccinated intranasally with various doses of L1^sdmax virus and challenged with 10⁷ PFU of WT virus at 21 days after vaccination. A group of age-matched virus-naive mice (n = 3) were challenged as well, as a negative control. After the virulent challenge, the mice body weights and neuronal disorder symptoms were closely monitored for another 14 days.

Neutralizing antibody assay.

The mice were immunized with various doses of either WT or L1^sdmax virus intranasally. The serum was isolated from tail vein blood 21 days after vaccination, followed by a 2-fold serial dilution in DMEM starting at a dilution of 1:20. Fifty microliters of the diluted serum was incubated with 100 PFU WT VSV in 50 µl DMEM at 37°C for 1 h. One hundred microliters of Vero cell suspension was then combined with the mixture and seeded on 96-well plates in triplicate. Cell morphology was checked daily, and after a 48-h incubation at 37°C, the remaining cells were stained with 1% crystal violet. The reciprocal of the dilution that gives 100% protection from VSV CPE was recorded as the virus-specific neutralizing-antibody titer.

Histopathology.

Mice were infected with 10⁷ PFU virus by intranasal inoculation and euthanized at 5 dpi. Mice were transcardially perfused with 10% paraformaldehyde (PFA)-phosphate-buffered saline (PBS) and their brains were harvested and fixed in 10% PFA-PBS, followed by 50% ethanol processing and paraffin embedding. The brain tissues were subsequently sectioned at 5 µm and stained with H&E. The histopathological results were examined independently by two neuropathologists at the School of Medicine, Stony Brook University.

Quantitative RT-PCR to analyze cytokine/chemokine inductions.

Total RNA was extracted from mice brains inoculated with PBS, WT, or L1^sdmax at both 3 and 6 dpi and then treated with RNase-free DNase I (Life Technologies) for 30 min at 37°C before a reverse transcription was conducted with oligo(dT)₂₀ (SuperScript III; Life Technologies). Quantitative PCR was subsequently performed to evaluate the induction of cytokines and chemokines using gene-specific primers (see Table S2 in the supplemental material). All reactions were subsequently normalized to the RNA levels of mouse β-actin and reported as increases in induction over that in PBS-inoculated animals.

Quantitative RT-PCR to measure viral genome replication.

BHK-21 cells were infected with viruses at a MOI of 3. At various time points, total RNA was extracted and reverse transcribed by random hexamer priming. Quantitative PCR was carried out with a primer pair spanning the region between the VSV G and L genes (see Table S2 in the supplemental material) to measure the viral genome replication. The reaction results were normalized to the mRNA levels of the BHK-21 GAPDH gene and reported as the increase (fold) over the value at 0 hpi.

Plasmid construction and transient protein expression.

The corresponding region of the WT L gene in the expression plasmid pL (27) was replaced by two recoded derivatives using the XhoI and XbaI digestion sites. For transient expression, 10⁶ BHK-21 cells were infected with vTF7-3 for 1 h and then cotransfected with 1.5 µg pP and 1.25 µg pL derivatives. Whole-cell lysates and total RNA were harvested at various time points, and the yield of viral products was subsequently determined by Western blotting and quantitative RT-PCR.

Statistical analysis.

Statistical analysis was performed using GraphPad Prism software version 5.0c (San Diego, CA). Data are presented as means with standard deviations (SDs). Two-tailed Mann-Whitney U tests were used for Fig. 4B and 7.

SUPPLEMENTAL MATERIAL

Figure S1 

The distribution of codon pair bias in WT, L1^sdmax, and L1^min coding sequences. The red dashed line highlights the location of the L1 region in the full-length L gene ORF. The position of each codon pair is marked on the x axis, and the codon pair score (the bar graph) represents the frequency of individual codon pairs in the human genome. The nonbias codon pair score is set to 0.0. More and less frequently used codon pairs are scaled up and down accordingly. Download

Figure S2 

Immunocytochemistry of L1^sdmax-infected cells. HeLa cells were infected with WT or L1^sdmax virus at an MOI of 10, and the synthesis of VSV N proteins was examined by fluorescence microscopy at both 0.5 and 4 hpi. Images taken at 0.5 hpi were exposed for 1,500 ms, while images taken at 4 hpi were exposed for 200 ms. Download

Figure S3 

BHK-21 cells were infected with viruses at an MOI of 10, and total cytoplasmic RNAs were extracted at the indicated time points. The abundance of viral RNAs was further determined by qRT-PCR and normalized to the mRNA level of the GAPDH gene. Download

Figure S4 

(A) The full-length L gene containing the synthetic L1^sdmax or L1^min sequence was placed under the control of the T7 promoter, and the newly constructed expression plasmids were designated pL^L1sdmax and pL^L1min, respectively. (B) The expression of L1^min sequence was examined by Western blotting in BHK-21 cells that were previously infected with the T7 Pol-expressing vaccinia virus vTF7-3. Download

Figure S5 

The inhibitory effect of Hsp90 inactivation upon VSV replication. HeLa cells were pretreated with increasing concentrations of the Hsp90 inhibitor 17AAG for 15 min and then infected with WT or L1^sdmax at an MOI of 0.01. At 12 hpi, the supernatants were harvested, and virus titers were determined in HeLa cells by plaque-forming assay. The asterisks indicate that the titers of L1^sdmax virus were undetectable (<30 PFU/ml) in cultures when the drug concentration was 0.5 µM or higher. Download

Figure S6 

(A and B) Survival rate of 5- to 6-week-old male BALB/c mice intranasally infected with various doses of WT (A) and L1^sdmax (B) virus. (C) Evans blue assay to evaluate the integrity of the BBB at day 6 after virus infection (5 × 10⁶ PFU per mouse). One representative brain from each group is shown. Download

Table S1 

Characteristics of L1^sdmax and L1^min sequences.

Table S2 

Summary of primer sequences.