INTRODUCTION
Human respiratory syncytial virus (RSV), a member of the
Paramyxoviridae family, infects essentially everyone worldwide early in life and causes at least 33.8 million pediatric lower respiratory tract infections and 199,000 pediatric deaths worldwide each year (
1,
26). Despite a well-recognized public health need for RSV vaccines, there is no licensed vaccine or effective antiviral therapy for RSV (
9), although infants and young children at high risk for severe RSV disease can be substantially protected by passive immunoprophylaxis with RSV-neutralizing antibody (
24,
38). A long-standing goal has been the development of a pediatric live-attenuated intranasal vaccine that is safe and well tolerated yet satisfactorily immunogenic in the target population, young infants under 6 months of age. Several biologically derived vaccine candidates have been tested in clinical trials, beginning in the 1960s, but were found to be unsatisfactorily attenuated and in some cases exhibited genetic instability (
2,
19,
20,
35,
37). More recently, reverse genetics has been used both to identify existing and create new attenuating mutations and to make new cDNA-derived vaccine candidates containing desired combinations of mutations (
7,
18,
36).
The most promising vaccine candidate to date is a cDNA-derived virus called rA2
cp248/404/1030ΔSH, a name that summarizes its attenuating mutations (
18). This virus contains five independent attenuating elements that were identified in biologically derived viruses or created by reverse genetics and subsequently combined: (i) a set of five amino acid substitutions in the N, F, and L proteins that were identified in a cold-passaged (
cp) RSV (
16,
17,
32), (ii) the substitution of leucine for glutamine at amino acid 831 in the L protein (a mutation originally called
248) (
14), (iii) the nucleotide substitution of C for T (positive sense) at position 9 of the gene start (GS) transcription signal of the M2 gene start (a mutation originally called
404) (
10), (iv) the substitution of asparagine for tyrosine at amino acid 1321 of the L protein (a mutation originally called
1030) (
10,
12), and (v) deletion of the SH gene (
5,
29). The original
248,
1030, and
404 designations were based on plaque number during the original isolation of the mutants rather than on sequence position. These mutations have been evaluated in some detail in preclinical studies (
10,
11,
29,
31,
32). The
248,
404, and
1030 mutations each render RSV temperature sensitive (
ts): individually, they shift the shutoff temperature from >40°C for wild-type (wt) RSV to 38°C [
248 and
1030 (
11,
22,
30,
31)] or 37°C [
404 (
30)]. The
cp and ΔSH mutations are also attenuating (
29,
32) but do not confer a
ts phenotype to RSV. The combination of these five independently attenuating elements by reverse genetics resulted in the highly attenuated and highly temperature-sensitive vaccine candidate rA2
cp248/404/1030ΔSH, with a shutoff temperature of 35°C. Clinical studies showed that this virus was well-tolerated and immunogenic in 1- to 2-month old infants and was protective against a second vaccine dose (
18).
When rA2
cp248/404/1030ΔSH was evaluated in children and infants in the initial clinical study, analysis of specimens recovered from nasal washes in the days following vaccination showed that more than one-third of the isolates exhibited a partial loss of the
ts phenotype (
18,
21). Sequence analysis of a limited number of the recovered isolates identified two types of genetic changes, namely, loss of either the
248 or the
1030 mutation, with 80% of the observed changes involving
1030 (L amino acid 1321) (
18,
21). The wt assignment at amino acid 1321, tyrosine (
TAT), and the
ts/attenuating
1030 mutation, asparagine (
AAT), differ by a single nucleotide (underlined). In clinical trial specimens, reversion at this position was due to asparagine (
AAT) being replaced with the wt assignment of tyrosine (
TAT) or, in one case, with histidine (
CAT) (
18). Reversion to the wt or histidine assignment could account for the partial loss of the
ts phenotype. Since reversion at the
1030 mutation was the most frequent change observed in the clinical trial samples, it was of interest to investigate whether this mutation could be stabilized.
The rA2
cp248/404/1030ΔSH virus remains the most promising RSV vaccine candidate. A second, biologically equivalent version of this virus is presently being further evaluated in a phase 1/2 clinical study (ClinicalTrials.gov identifier NCT00767416) under the name Medi-559. The first and second versions will hereinafter be called rA2
cp248/404/1030ΔSH and Medi-559, respectively. They differ by 37 nucleotide substitution point mutations scattered throughout the genome, none of which affect amino acid coding. The nucleotide differences are due to naturally occurring variability in the wt RSV used as the template for reverse transcription (RT)-PCR and to differences in restriction site markers. All of the attenuating mutations of rA2
cp248/404/1030ΔSH are present in Medi-559, except that the codon for the 248 mutation (Q831L) is TTA in rA2
cp248/404/1030ΔSH and CTG in Medi-559. (Note that the nomenclature lists wt amino acid assignments to the left of the sequence position [e.g., Q831] and mutant assignments to the right [e.g., 831L]). Both of these codons readily reverted during a temperature stress test (
22) in which virus bearing the mutation of interest is passaged multiple times at progressively increasing temperature to select for revertants. Thus, it is unlikely that there is a significant difference in the genetic stability of rA2
cp248/404/1030ΔSH and Medi-559. The available data also indicate that these two viruses are indistinguishable with regard to replication in cell culture, temperature sensitivity, and attenuation
in vivo.
We previously described a strategy to increase the phenotypic and genetic stability of attenuating amino acid substitutions that are based on a single-nucleotide substitution, such as the
248 and
1030 mutations (
22,
23). This strategy is based on increasing the number of nucleotides that must be changed in a given mutant codon in order to encode an amino acid assignment conferring deattenuation, which can involve direct reversion to the wt assignment or change to another assignment that confers a wt-like phenotype. Substitution at any single-nucleotide position in RNA viruses can occur at a relatively high rate, ∼10
−4, thus providing for relatively frequent deattenuation if only a single nucleotide is involved. However, if deattenuation at the amino acid level requires changes at two or, preferably, three positions within the codon, the frequency will be much less, ∼10
−8 and ∼10
−12, respectively. As the first step in this analysis, the position of interest is changed systematically to encode each of the 20 possible amino acids on the wt background, resulting in a panel of viruses that are then evaluated to identify the
ts and attenuation phenotypes associated with each amino acid assignment. Then, based on the genetic code, the sequences of all the possible codons encoding attenuating amino acids are compared on paper to the wt-like codons with the goal of identifying a codon that would require changes at two or three positions in order to revert to a wt-like codon. This strategy can reduce deattenuation at a given codon to below the level of detection, although it is not always possible to identify a reversion-resistant codon (
22,
23). In the present study, we used this strategy to increase the genetic and phenotypic stability of the
1030 mutation involving position 1321 in the L protein. This mutation was used to construct an improved version of the Medi-559 vaccine candidate, called cps2, that has increased genetic stability and remains satisfactorily attenuated in seronegative chimpanzees.
DISCUSSION
RSV has long been recognized as a ubiquitous, important cause of respiratory tract disease worldwide, particularly in the very young and in the elderly, and remains a vaccine priority. Subunit vaccines are contraindicated in RSV-naïve recipients, based on the experience in the 1960s with a formalin-inactivated disease that primed for enhanced RSV disease and on apparently similar effects observed in experimental animals in response to subunit RSV vaccines (
8,
25). Live-attenuated vaccines did not prime for disease enhancement in experimental animals or in clinical studies (
34) and, thus, are suitable for use in RSV-naïve recipients. However, development of a live-attenuated RSV vaccine has been difficult due to the relatively poor growth of RSV
in vitro, the filamentous and physically unstable nature of the particle, the relative lack of predictive and permissive animal models, the immunologic immaturity of the young infant, the heightened concerns for vaccine safety in young infants, and the immunosuppressive effects of maternally derived RSV-specific serum antibodies present in this population (
35). The first live-attenuated RSV vaccine to be evaluated in infants and young children (cpRSV) was reported more than 40 years ago (
20). Initial efforts to produce biologically derived vaccine candidates (
10,
11,
14,
19,
35) were superseded by candidates designed by reverse genetics (
28–32), and the first vaccine candidate to be well tolerated and immunogenic in young infants, rA2
cp248/404/1030ΔSH, was described in 2005 (
18). The present study describes work to improve the genetic and phenotypic stability of this promising RSV vaccine candidate, specifically, the version called Medi-559 (ClinicalTrials.gov identifier NCT00767416).
When the original rA2
cp248/404/1030ΔSH virus was evaluated in RSV-naive infants and young children, a substantial frequency of reversion was observed at either of two of the five attenuating features of the virus, namely, the
248 (831L) and
1030 (1321N) mutations (
18), which are single-nucleotide amino acid substitutions that each contributes to the
ts and attenuation phenotypes of the virus (
31). Reversion at either or both of these sites was also observed previously during passage
in vitro (
21). Both
in vitro and
in vivo, reversion was more frequent for the
1030 mutation. While these revertants sometimes became the predominant species in vaccine recipients, they were not associated with enhanced replication or disease, reflecting the fact that by retaining four of the five original attenuating features, the revertants remained highly attenuated. However, since phenotypic reversions were observed in one-third to one-half of the isolates obtained from vaccines, and since the loss of more than a single attenuating mutation was observed during passage
in vitro (
21), it would be desirable to increase the genetic stability of this vaccine candidate.
In previous work, we attempted to stabilize the
248 Q831L mutation present in rA2
cp248/404/1030ΔSH (
18), using the same strategy as in the present study. It was possible to achieve a modest increase in stability for the
248 Q831L mutation by introducing the leucine codon TTG, which was the most stable codon based on temperature stress tests (
18).
In the present work, we investigated stabilization of the
1030 mutation at codon Y1321, which was the most frequent site of reversion in the previous clinical study (
18). In this case, we were able to identify several alternative amino acids that were associated with levels of attenuation similar to that of the original
1030 1321N(AAT) mutant. Specifically, analysis of the possible amino acid assignments for position 1321 identified four (glycine, lysine, glutamic acid, and proline) that conferred a
ts phenotype (
TSH = 38°C) and an attenuation phenotype similar to those of 1321N. Of note, there was a positive correlation between
TSH and virus replication attenuation in mice (
Fig. 1), and thus, both parameters could be used to assess assignments at position 1321. Incidentally, given the knowledge of the
ts and attenuation phenotypes associated with each amino acid assignment at position 1321, it also may be possible to incrementally increase or decrease the
ts/attenuation phenotype of the rA2
cp248/404/1030ΔSH or Medi-559 vaccine viruses by choosing appropriate amino acid assignments. However, this was outside the scope of the present study, whose goal was increased stability without affecting the
ts/attenuation phenotypes.
The codon options for the chosen four alternative amino acid assignments (glycine, lysine, glutamic acid, and proline) at position 1321 were examined to predict the outcomes of all possible single-nucleotide substitutions, using the knowledge of the
ts/attenuation phenotype associated with each amino acid assignment at that position. This theoretical analysis identified four codons that each would require at least two nucleotide changes to yield any possible amino acid assignment specifying a wt-like phenotype. There were no possibilities that would require three nucleotide changes to yield a wt-like assignment and would be the most refractory to reversion, although a previous study with human parainfluenza virus type 1 showed that sometimes this is possible (
23). This analysis suggested 1321G(GGA) as the most phenotypically stable
ts mutation, followed by mutants 1321K(AAA) and 1321E(GAA)/(GAG). According to this theoretical analysis, any of these four codons would be significantly more stable than 1321N or 1321P or the other possible codons, 1321G and 1321K.
In vitro stress tests showed that 1321G (GGA or GGT) and 1321K (AAA) mutants were stable under conditions in which virus with the original 1321N(AAT) mutant assignment exhibited reversion in every culture. The 1321E(GAA) mutant was somewhat less stable, with deattenuation in 20% of the populations, and virus with the 1321P(CCT) mutation, which was included as coding for an alternative amino acid that was not predicted to be more stable, exhibited deattenuation in 90% of the populations, similar to the 1321N mutant. However, following passage in the stress tests, we frequently detected a serine-to-cysteine missense mutation, S1313C, in the G, K, E, and P viruses. Analysis of recombinant virus in which this 1313C assignment was combined with attenuating 1321 assignments confirmed that this mutation was a compensatory second-site mutation that alleviated the ts/attenuation phenotype conferred by the 1321 assignment.
The ability of the S1313C mutation to dramatically compensate for attenuating mutations at position 1321 was unexpected. When placed in the wt RSV background, this S1313C mutation had no detectable phenotypic effect on either temperature sensitivity
in vitro or attenuation in mice. Thus, serine and cysteine appear to be interchangeable at this position in the wt background. The idea that a serine-to-cysteine substitution would be conservative also is suggested by the structural similarity of these residues: specifically, serine and cysteine are similarly small and nucleophilic, differing in a hydroxyl group (serine) versus a sulfhydryl group (cysteine). However, recent work with beta-lactamase showed that this seemingly conservative substitution can be substantially destabilizing due to alterations in hydrogen bonding and can have a substantial impact on protein folding (
27). It also is interesting that, in the present study, the ability of 1313C to cause deattenuation associated with position 1321 was largely independent of the structure of the 1321 assignment: 1313C appeared to be compensatory for glycine, lysine, glutamic acid, and proline (but apparently not asparagine). In general, the effects of substitutions at positions 1313 and 1321 in the RSV L protein were not readily predictable or interpretable by consideration of amino acid structure.
Given the frequent occurrence of this S1313C compensatory mutation, it was necessary to stabilize this position in order to maintain the attenuating effects of the assignment at codon 1321. While the S(AGC) codon present in wt RSV could mutate to encode cysteine by a single-nucleotide substitution, two other serine codons, S(TCA) and S(TCG), would require two substitutions. Therefore, we combined the 1321K(AAA) and S1313(TCA) mutations to create a stabilized version of the 1030 mutation. In the wt RSV background, the 1321K(AAA)/S1313(TCA) combination was indistinguishable from the 1321K(AAA) mutation with regard to ts phenotype, and the combination was completely stable in temperature stress tests.
Based on these results, we modified Medi-559 to yield cps2 by the following modifications: (i) codon 831 in the L protein (the 248 mutation) was changed from L(TTA) to L(TTG); codon 1313 (the second-site mutation) was changed from S(AGC) to S(TCA); and (iii) codon 1321 (the 1030 mutation) was changed from N(AAT) to K(AAA). This involved a total of 5 nucleotide substitutions (underlined) and one amino acid substitution. In vitro temperature stress tests combined with sequence analysis provided evidence of increased stability: reversion at position 831 was reduced by one-third, and reversion at positions 1313 and 1321 did not occur. One flask had evidence of mutation of position 1321 to arginine. However, this residue appeared to be lethal when we attempted to recover it in the wt background. Its presence in the stress test probably reflects genomes that encoded defective L protein but could be replicated by complementation by nondefective virus present due to the high multiplicity of infection used during the sequential passages, similar to the well-known situation with defective interfering genomes. Thus, this population of virus is probably nonviable and so does not represent deattenuation.
Analysis of virus from the published clinical study of rA2
cp248/404/1030ΔSH (
18,
21) indicated that the 248 mutation (position 831) reverted in 1 of 7 subjects examined who shed partially revertant virus, while the
1030 mutation (position 1321) reverted in 4 of 7 subjects with revertant virus. This suggests that 4 of 5 reversions involve a change in the
1030 mutation. Thus, stabilizing this position would be predicted to reduce the reversion rate by 80%. As noted, the modification at position 831 would be predicted to reduce the reversion rate by a further third. This estimates that the reversion rate will be reduced by ∼87%. However, it will be necessary to evaluate stability in a clinical study in seronegative individuals. The cps2 virus was indistinguishable from Medi-559 with regard to both
ts phenotype and replication in seronegative chimpanzees: this comparability indicates that cps2 can replace Medi-559 as the lead RSV vaccine candidate.