Recovery of a Fully Viable Chimeric Human Parainfluenza Virus (PIV) Type 3 in Which the Hemagglutinin-Neuraminidase and Fusion Glycoproteins Have Been Replaced by Those of PIV Type 1

Abstract

The recent recovery of human parainfluenza virus type 3 (PIV3) from cDNA, together with the availability of a promising, highly characterized live attenuated PIV3 vaccine virus, suggested a novel strategy for the rapid development of comparable recombinant vaccine viruses for human PIV1 and PIV2. The strategy, illustrated here for PIV1, is to create chimeric viruses in which the two protective antigens, the hemagglutinin-neuraminidase (HN) and fusion (F) envelope glycoproteins, of an attenuated PIV3 variant are replaced by those of PIV1 or PIV2. As a first step, this has been achieved by using recombinant wild-type (wt) PIV3 as the recipient for PIV1 HN and F, engineered so that each PIV1 open reading frame is flanked by the existing PIV3 nontranslated regions and transcription signals. This yielded a viable chimeric recombinant virus, designated rPIV3-1, that encodes the PIV1 HN and F glycoproteins in the background of the wt PIV3 internal proteins. There were three noteworthy findings. First, in contrast to recently reported glycoprotein replacement chimeras of vesicular somatitis virus or measles virus, the PIV3-1 chimera replicates in LLC-MK2 cells and in the respiratory tract of hamsters as efficiently as its PIV1 and PIV3 parents. This is remarkable because the HN and F glycoproteins share only 43 and 47%, respectively, overall amino acid sequence identity between serotypes. In particular, the cytoplasmic tails share only 9 to 11% identity, suggesting that their presumed role in virion morphogenesis does not involve sequence-specific contacts. Second, rPIV3-1 was found to possess biological properties derived from each of its parent viruses. Specifically, it requires trypsin for efficient plaque formation in tissue culture, like its PIV1 parent but unlike PIV3. On the other hand, it causes an extensive cytopathic effect (CPE) in LLC-MK2 cultures which resembles that of its PIV3 parent but differs from that of its noncytopathic PIV1 parent. This latter finding indicates that the genetic basis for the CPE of PIV3 in tissue culture lies outside regions encoding the HN or F glycoprotein. Third, it should now be possible to rapidly develop a live attenuated PIV1 vaccine by the staged introduction of known, characterized attenuating mutations present in a live attenuated PIV3 vaccine candidate into the PIV3-1 cDNA followed by recovery of attenuated derivatives of rPIV3-1.

Human parainfluenza virus type 1 (PIV1), PIV2, and PIV3 are significant causes of serious lower respiratory tract disease in infants and children and account for approximately 18% of all hospitalizations of pediatric patients for respiratory tract infection (5, 22, 25). A vaccine has not been approved for the prevention of PIV disease, nor is there an effective antiviral therapy once disease occurs. Two promising live attenuated PIV3 vaccine candidates are undergoing clinical evaluation (18, 19). First, a bovine PIV3 (BPIV3) that is antigenically related to human PIV3 protects animals against PIV3 infection and has been found to be safe, genetically stable, and immunogenic in human infants and children (19). Second, a cold-adapted mutant has been generated by 45 serial passages of the JS wild-type (wt) strain of human PIV3 (PIV3/JS) in cell culture at low temperature (1). This attenuated mutant (cp45) is protective against PIV3 challenge in experimental animals and is satisfactorily attenuated, genetically stable, and immunogenic in seronegative human infants and children (15, 18).

PIV3 is a member of the Paramyxovirus genus of the Paramyxoviridae family in the order Mononegavirales (5). Its genome is a single-stranded, negative-sense RNA that is 15,462 nucleotides (nt) in length (12, 34). It encodes at least eight proteins: the nucleocapsid protein N, the phosphoprotein P, the C and D proteins of unknown functions, the matrix protein M, the fusion glycoprotein F, the hemagglutinin-neuraminidase glycoprotein HN, and the large polymerase protein L (5, 11, 33). The PIV3 P mRNA also contains an open reading frame (ORF) that might encode a V protein, although it is not known whether this is expressed (13). The M, HN, and F proteins are associated with the envelope, and the latter two are the neutralizing and protective antigens of PIVs (5). The significant sequence divergence in these two protective antigens among the PIVs is the basis for the type specificity of protective immunity (5).

Infectious recombinant PIV3 (rPIV3) has recently been recovered from cDNA (8, 16), and it is now possible to use this reverse genetics system to generate infectious virus bearing predetermined attenuating mutations. We demonstrated previously that the rPIV3/JS isolate that we recovered from cDNA manifests the wt phenotype for efficient replication in vitro and in vivo (8), and we are currently using this reverse genetics system to define the attenuating mutations present in the cp45 and BPIV3 candidate vaccine viruses as well as to construct a cDNA-derived live-attenuated PIV3 vaccine (32).

Comparable vaccine candidates or reverse genetics systems do not exist for human PIV1 and PIV2. Here we describe an alternative strategy for producing a live attenuated PIV1 vaccine that takes advantage of the available PIV3 system. Specifically, this involves construction of a chimeric antigenomic cDNA in which the F and HN ORFs in the existing PIV3 antigenomic cDNA are replaced with those of PIV1. A chimeric infectious recombinant virus, called rPIV3-1, which encodes the internal proteins of PIV3 and the HN and F glycoproteins of PIV1, was recovered by using the previously described rescue system (8). Remarkably, rPIV3-1 replicates as well as its parental viruses in vitro and in vivo. Thus, it will serve as a suitable substrate for the production of a live attenuated PIV1 vaccine by the introduction of those attenuating mutations found in the cp45 vaccine and BPIV3 candidate vaccine viruses that lie outside the HN and F coding regions.

MATERIALS AND METHODS

Viruses and cells.

The PIV1 strain used in this study, PIV1/Washington/20993/1964 (PIV1/Wash64), was confirmed previously to be a virulent wt virus in adult human volunteers (26). It was propagated in LLC-MK2 cells (ATCC [American Type Culture Collection] CCL 7.1) in Opti-MEM I (Life Technologies, Gaithersburg, Md.) with 50 μg of gentamicin sulfate per ml, 2 mM glutamine, and 0.75 μg of trypsin (catalog no. 3741; Worthington Biochemical Corp., Freehold, N.J.) per ml. The Greer strain of human PIV2 (catalog no. V-322-001-020; NIAID Repository, Rockville, Md.) used in the hemagglutination inhibition assay (HAI) was propagated in the same way. PIV3/JS and its recombinant derivative from cDNA (rPIV3/JS), both of which are wt, were propagated as previously described (8). The modified vaccinia virus Ankara (MVA) recombinant that expresses the bacteriophage T7 RNA polymerase was generously provided by L. Wyatt and B. Moss (38).

HEp-2 cells (ATCC CCL 23) were obtained from the ATCC and maintained in Opti-MEM I with 2% fetal bovine serum, 50 μg of gentamicin sulfate per ml, and 2 mM glutamine.

Construction of a cDNA encoding a complete chimeric PIV3-PIV1 antigenome.

cDNA clones of the F and HN genes of PIV1/Wash64 were generated from infected cell RNA by reverse transcription (RT) using random hexamers followed by PCR using synthetic primers that introduced NcoI-BamHI sites flanking the F cDNA and NcoI-HindIII sites flanking the HN cDNA (Fig. 1). The sequences of these primers were (with PIV-specific sequences in uppercase, restriction sites underlined, nucleotides altered from the wt sequence in lowercase, and start and stop codons in bold): upstream F, 5′-cgccATGgAAAAATCAGAGATCCTCTTCT-3′; downstream F, 5′-ctggatccTAATTGGAGTTGTTACCCATGTA-3′; upstream HN, 5′-aaccATGGCTGAAAAAGGGAAAA-3′; and downstream HN, 5′-ggtgaaGCTtAAGATGTGATTTTACATATTTTA-3′. It should be noted that the NcoI site in the glycoprotein F oligonucleotide changes the assignment of F amino acid 2 from Gln to Glu, which lies within the cleaved signal sequence.

FIG. 1.

Construction of cDNA encoding the chimeric PIV3-PIV1 antigenome in which the PIV3 HN and F ORFs were replaced by those of PIV1. First (starting from the bottom left), the PIV3 F and HN genes were subcloned from the full-length PIV3 cDNA clone p3/7-131.2G (A and B). PCR mutagenesis (C) was performed to delete the complete coding regions of the PIV3 F and HN genes and to introduce new restriction sites (boxed). Chimeric F and HN genes were constructed by importing the PIV1 F and HN ORFs into the PIV3 deletions (C to E). The chimeric F and HN were assembled together to generate pSE.PIV3-1.hc (F). The F and HN chimeric segment was introduced into full-length PIV3 clone p3/7-131.2G to generate pFLC.2G+.hc (G). The small boxes between genes represent the gene end, intergenic, and gene start regions, and the lines represent the noncoding regions. Shaded portions are from PIV1; open boxes are from PIV3. Large arrows depict the T7 promoter; black boxes depict the hepatitis delta virus (HDV) ribozyme.

The PIV3 F and HN genes were subcloned in several steps from the full-length clone of PIV3/JS (p3/7-131.2G) (8), and the PIV3 F and HN coding regions were deleted by PCR mutagenesis and replaced with NcoI-BamHI and NcoI-HindIII sites to accept the PIV1 F and HN cDNA described above (Fig. 1) (3). The two plasmids were amplified, cleaved by the restriction endonucleases, and used as recipients for the mutagenized PIV1 cDNA. The sequences of the positive-sense and negative-sense mutagenic primers were (i) for the PIV3 F cDNA, 5′-AAATAggatccCTACAGATCATTAGATATTAAAAT-3′ and 5′-cGcCATgGTGTTCAGTGCTTGTTG-3′, and (ii) for the PIV3 HN cDNA, 5′-ccacAAgCtTAATTAACCATAATATGCATCA-3′ and 5′-TTCCATggATTTGGATTTGTCTATTGGGT-3′. This mutagenesis deleted 3 nt immediately before the start codon of the chimeric HN gene so that the antigenomic cDNA will conform to the rule of six (4, 9). The PIV1 HN or F cDNAs described above were imported in as an NcoI-HindIII fragment for HN or as an NcoI-BamHI fragment for F, which generated pLIT.PIV3-1.HNhc and pLIT.PIV3-1.Fhc (Fig. 1). These two cDNAs were then joined into pSE.PIV3-1.hc, which was subsequently sequenced in its entirety. The BspEI-SphI fragment was then inserted into p3/7-131.2G to generate pFLC.2G+.hc (Fig. 1). The cDNA engineering was designed so that the final PIV3-1 antigenome conformed to the rule of six (4, 9). Its length value of 15,516 nt does not include two 5′-terminal G residues contributed by the T7 promoter, because it is assumed that they are removed during recovery.

Transfection.

HEp-2 cell monolayers were grown to confluence in six-well plates, and transfections were performed in HEp-2 cells essentially as described previously (8). Trypsin was added to a final concentration of 0.75 μg/ml on day 3 posttransfection, 1 day prior to harvesting. Cell culture supernatants were clarified and passaged (referred to as passage 1) onto fresh LLC-MK2 cell monolayers. After overnight absorption, the medium was replaced with fresh Opti-MEM I with 0.75 μg of trypsin per ml. Passage 1 cultures were incubated at 32°C for 4 days, and the amplified virus was harvested and passaged again under the same condition (referred to as passage 2).

Nucleotide sequence analysis.

DNA sequencing was done with a Circumvent sequencing kit (New England Biolabs, Beverly, Mass.). For isolation of virion RNA, virus was amplified in T75 flasks of LLC-MK2 cells and concentrated from the supernatant by polyethylene glycol precipitation (23). Viral RNA was purified and amplified by RT with random hexamer primers followed by PCR with PIV1- or PIV3-specific primer pairs. RT-PCR products were gel purified by electrophoresis onto, and elution from, strips of NA45 DEAE nitrocellulose membrane (Schleicher & Schnuell, Keene, N.H.) and were sequenced.

Replication of PIVs in LLC-MK2 cells.

Plaque enumeration on LLC-MK2 monolayers was performed as previously described except that 0.75 μg of trypsin per ml was added in the case of PIV1 and rPIV3-1 (15). After incubation at 32°C for 6 days, the agarose overlay was removed and plaques were identified by hemadsorption (HAD) with guinea pig erythrocytes (RBCs). The virus stocks were also characterized by 50% tissue culture infective dose (TCID₅₀) assay, in which cells were incubated in the presence of 0.75 μg of trypsin per ml at 32°C for 6 days. Virus titer was determined by direct observation of cytopathic effect (CPE) as judged by cell rounding and detachment and by subsequent HAD.

Growth of the PIV in tissue culture was evaluated by infecting confluent LLC-MK2 monolayers on 12-well plates with virus at a multiplicity of infection (MOI) of 0.01. Cells were incubated in the presence of 0.75 μg of trypsin per ml at 32°C for 6 days. At each 24-h interval, a 0.3-ml medium aliquot was removed from each well and was replaced with 0.3 ml of fresh medium with 0.75 μg of trypsin per ml. The titer of virus in the aliquots was determined in parallel at 32°C by HAD on LLC-MK2 cell monolayers, using fluid overlay as previously described (15), and the titers were expressed as log₁₀ TCID₅₀/milliliter.

Replication of PIVs in the respiratory tracts of hamsters.

Golden Syrian hamsters were inoculated intranasally with 0.1 ml of L-15 medium containing 10⁵ PFU of rPIV3/JS, rPIV3-1, or PIV1/Wash64. On days 4 and 5 postinoculation, six hamsters from each group were sacrificed, their lungs and nasal turbinates were harvested and homogenized, and virus titers in the samples were determined on LLC-MK2 cell monolayers at 32°C as described above. The titers were expressed as mean log₁₀ TCID₅₀/gram of tissue for each group of six hamsters.

Nucleotide sequence accession number.

The nucleotide sequence of the BspEI-SphI fragment containing the chimeric F and HN genes is in GenBank (accession no. AF016281).

RESULTS

Construction of a cDNA clone encoding a full-length, chimeric PIV3-1 antigenomic RNA.

The construction of the PIV3-PIV1 chimeric cDNA, in which the ORFs of the wt PIV3/JS HN and F glycoprotein genes were replaced by those of PIV1/Wash64, is described in the Materials and Methods and summarized in Fig. 1. The sequences of the junction points in the chimeric genes will be shown later. The final plasmid construct, pFLC.2G+.hc (Fig. 1), encodes a PIV3-1 chimeric antigenomic RNA of 15,516 nt, which conforms to the rule of six (4, 9).

Recovery and characterization of the recombinant chimeric virus rPIV3-1.

The pFLC.2G+.hc cDNA encoding the chimeric PIV3-1 antigenome was transfected onto HEp-2 cells together with the PIV3 N, P, and L support plasmids. The p3/7-131.2G cDNA encoding the wt PIV3/JS antigenome was transfected in parallel to generate a rPIV3/JS control parental virus. Virus was recovered from each transfection by two passages on LLC-MK2 cells, and studies were initiated to confirm that each recombinant virus was derived from cDNA.

Recombinant viruses rPIV3-1 and rPIV3/JS were first characterized for the presence of the PIV1 or PIV3 HN glycoprotein by HAI assay with serotype-specific anti-HN monoclonal or polyclonal antibodies. rPIV3/JS was shown to contain the introduced monoclonal antibody (MAb)-resistant mutation that marks this virus as being derived from cDNA (8), and the rPIV3-1 virus was shown to contain the PIV1 HN protein, as expected (Table 1).

TABLE 1.

rPIV3-1 possesses the HN glycoprotein gene of PIV1

Virus	HAI titera (reciprocal) of:
	PIV1b antiserum	PIV2c antiserum	PIV3 MAbd
			423/6	77/5
PIV1/Wash64	256	32e	≤50	≤50
rPIV3-1	64	≤2	≤50	≤50
rPIV3/JS	4	≤2	≤50	3,200
PIV3/JS	8	≤2	12,800	6,400
PIV2/Greer	8	512	≤50	≤50

^aChicken RBCs were used in HAI assay (36) for PIV1, PIV2, and rPIV3-1 because of their greater sensitivity, and guinea pig RBCs were used for PIV3/JS and rPIV3/JS because chicken RBCs do not react well with PIV3. 

^bPIV1 polyclonal rabbit antiserum purchased from Denka Seiken Co. Ltd., Tokyo, Japan (catalog no. 410-701). 

^cPIV2 polyclonal guinea pig antiserum obtained from the NIAID Repository (catalog no. V322-503-558). 

^dBiologically derived PIV3/JS contains epitopes recognized by both MAb 423/6 and MAb 77/5, whereas rPIV3/JS was engineered to lack reactivity with MAb 423/6 (8). 

^eThe PIV2 antiserum had some reactivity with PIV1 and therefore is not completely type specific. 

We next sought to confirm by nucleic acid analysis that the rPIV3-1 virus contained the engineered, chimeric PIV3-1 HN and F genes. As designed, the genetic structure of rPIV3-1 was unique in four junction regions (boxed in Fig. 2A) compared with either of its parents, PIV1/Wash64 or rPIV3/JS. These regions are the transition points at which the sequence switches from the PIV3 noncoding region to the PIV1 coding region and then from the PIV1 coding region back to the PIV3 noncoding region. Using primer pair A, specific to PIV3 M and L genes, or primer pair B, specific to the PIV1 M and HN genes (Fig. 2A), we generated RT-PCR products for virion-derived RNAs from rPIV3-1, rPIV3/JS, and PIV1/Wash64. Control reactions showed that the RT step was required for generation of RT-PCR products, confirming that the template was RNA rather than contaminating DNA (data not shown). As expected, the PIV3-specific primer pair A generated a 4.6-kb cDNA product from rPIV3-1 and rPIV3/JS that spans the F and HN genes, while a PIV1-specific primer pair amplified a similar-size product from the PIV1 control but not from rPIV3-1 (data not shown).

FIG. 2.

rPIV3-1 is a chimeric virus. (A) Diagram of the chimeric HN and F genes of rPIV3-1 (middle) in comparison with those of rPIV3/JS (top) and PIV1/Wash64 (bottom). The four junction regions containing the sequence transitions from PIV3 to PIV1 are boxed and numbered I to IV. Small boxes between genes represents the gene end, intergenic, and gene start regions, and the lines represent the noncoding regions. RT-PCR primers, specific to the PIV3 M and L genes (primer pair A; top) or to the PIV1 M and HN genes (primer pair B; bottom) used in RT-PCR, are depicted as arrows. (B) Sequences of PIV3-PIV1 junctions in the RT-PCR products of rPIV3-1. The sequence for each of the four junction regions (regions I to IV) is presented and aligned with the corresponding regions of rPVI3/JS (top line) and PIV1/Wash64 (bottom line), which were sequenced in parallel from RT-PCR products. Vertical bars indicate sequence identity, while the boxed regions indicate introduced mutations and restriction sites. The Gln-to-Glu codon change in the chimeric F gene is indicated by a shaded box. Start and stop codons are underlined.

The nucleotide sequences of the 4.6-kb RT-PCR product of rPIV3-1, rPIV3/JS, and PIV1/Wash64 were determined for the four junction regions (Fig. 2). Each sequence was in complete agreement with the cDNA from which it was derived (Fig. 2B). These data confirm that rPIV3-1 is a recombinant chimeric virus whose sequence structure is exactly as designed.

Trypsin dependence and cytopathicity of rPIV3-1 in vitro.

PIV1, like Sendai virus but unlike PIV3, requires trypsin for cleavage of its F glycoprotein in order to undergo multicycle replication on continuous lines of tissue culture cells (10). In addition, PIV1 is a noncytopathic virus whereas PIV3 readily produces extensive CPE (5). We compared rPIV3-1, rPIV3/JS, and PIV1/Wash64 on the basis of these properties. As shown in Table 2, rPIV3-1, like its PIV1/Wash64 parent, required trypsin for efficient replication in cultures with fluid overlay as well as for efficient plaque formation, consistent with the presence of the F glycoprotein of the PIV1 parent virus. On the other hand, rPIV3-1 produced CPE, as indicated by cell rounding and detachment in the virus-infected monolayers, almost to the same extent as its PIV3 parent. This finding suggests that this biological property is a function of PIV3 genetic information which lies outside the HN and F ORFs. Thus, rPIV3-1 possesses biological properties from both parents.

TABLE 2.

Comparison of HA titers, infectivities, and cytopathicities of parental and chimeric PIVs^a

Virus	HA titer		Infectious titerb (log10 TCID50/ml)				PFU/mlc (log10)
	Chicken RBCs	Guinea pig RBCs	CPE as endpoint		HAD as endpoint
			−	+	−d	+	−	+
PIV1/Wash64	16	8	≤2.5	≤2.5	4.8	6.3	<0.7e	5.8
rPIV3-1	64	16	≤2.5	5.8	5.5	7.8	<0.7e	7.1
rPIV3/JS	0	8	4.5	7.3	5.0	7.5	5.0	6.2

^aVirus stocks were grown in LLC-MK2 cells which were infected at an MOI of 0.01 and incubated for 6 days in the presence (PIV1/Wash64, rPIV3-1) or absence (rPIV3/JS) of trypsin (0.75 μg/ml). The resulting virus stocks were assayed in the presence (+) or absence (−) of trypsin as indicated. 

^bThe TCID₅₀ assay was read at 6 days by direct visualization of CPE or by HAD. 

^cPlaques were visualized by HAD after 6 days of incubation. 

^dThe HAD of PIV3-infected monolayers was grossly apparent, whereas that of PIV1 and rPIV3-1 was observable only under the microscope in which single cells with RBCs adsorbed were observed. 

^eThe lowest level of virus detectable was 10^0.7/ml. 

Comparison of the levels of replication of rPIV3-1 and its parental viruses in LLC-MK2 cells and hamsters.

The multicycle replication of rPIV3/JS, rPIV3-1, and PIV1/Wash64 viruses was evaluated following inoculation of LLC-MK2 tissue culture cells at an MOI of 0.01 (Fig. 3). It can be seen that the kinetics and magnitudes of replication of the three viruses are very similar, which indicates that the substitution of the HN and F genes of PIV1 for those of PIV3 did not attenuate the virus for replication in vitro. Also, rPIV3-1 is not temperature sensitive; i.e., it produced plaques at 32, 37, or 40°C with equal efficiency (data not shown). We next sought to determine if rPIV3-1 was attenuated in vivo, specifically for replication in the upper and lower respiratory tracts of hamsters (Table 3). It can be seen that the level of replication of rPIV3-1 was similar to or slightly higher than that of either parent in the upper and lower respiratory tracts of hamsters.

FIG. 3.

Multicycle growth of parental and chimeric PIVs in tissue culture. LLC-MK2 cell monolayers were inoculated with virus at an MOI of 0.01, and virus-infected cells were incubated at 32°C in the presence of trypsin. Tissue culture supernatants were harvested at 24-h intervals, frozen, and analyzed in the same TCID₅₀ assay, using hemadsorption to identify virus-infected cultures. Each point represents the mean titer of three separate cultures, with the standard error indicated. The dotted horizontal line indicates the lower limit of virus detection.

TABLE 3.

Levels of replication of parental and chimeric PIVs in the upper and lower respiratory tracts of hamsters^a

Virus	Virus titer (mean log10 TCID50/g ± SE) in:
	Nasal turbinates		Lungs
	Day 4	Day 5	Day 4	Day 5
PIV1/Wash64	5.2 ± 0.24	5.2 ± 0.12	5.0 ± 0.31	5.0 ± 0.38
rPIV3-1	4.9 ± 0.23	6.2 ± 0.17	5.8 ± 0.15	6.0 ± 0.09
rPIV3/JS	4.5 ± 0.09	5.0 ± 0.18	5.1 ± 0.26	5.0 ± 0.32

^aEach hamster was infected intranasally with 10^5.0 TCID₅₀ of the indicated virus, and lungs and nasal turbinates were removed on day 4 or 5 after infection. The titers are means for six animals per day. 

DISCUSSION

We have been developing live attenuated virus vaccines for respiratory syncytial virus and for PIV3 by using conventional techniques, and after more than two decades of work, several very promising candidate vaccines are under clinical evaluation (7). However, live attenuated virus vaccine candidates for PIV1 or PIV2 do not exist. Since development of a live attenuated respiratory virus vaccine by using conventional techniques such as passage in tissue culture or mutagenesis is a lengthy process with an uncertain outcome, the prospect of using such techniques to generate live attenuated virus vaccines for PIV1 or PIV2 is unlikely to be successful in a timely fashion. Rather, the technique of reverse genetics that has recently been developed for the single-stranded, negative-sense RNA viruses of the order Mononegavirales (6, 8, 14, 16, 20, 21, 28, 30, 37) offers a new approach for the rapid and rational development of well-defined paramyxovirus vaccines. However, reverse genetics systems do not exist for human PIV1 and -2, and even if these soon become available it is difficult to devise attenuating mutations a priori. The successful application of reverse genetics to developing PIV3 and respiratory syncytial virus vaccine viruses depends in part on the availability of mutations identified in biologically derived attenuated viruses (2, 8, 17). This advantage does not exist for PIV1 and -2. We therefore chose to explore the possibility that live attenuated vaccines for PIV1 could be generated in a timely fashion by replacing the protective HN and F antigens of PIV3 with those of PIV1 by using the established PIV3 rescue system. In this present study, we found that it indeed was possible to recover an rPIV3-1 chimeric virus in which the ORFs of the PIV1 HN and F glycoproteins were substituted for those of rPIV3. This chimeric virus replicated like its wt PIV1 and PIV3 parental viruses in vitro and in vivo, demonstrating that the substitution of the glycoprotein ORFs did not result in attenuation of rPIV3-1.

There are other recent reports of recombinant nonsegmented negative-strand viruses in which the homologous glycoprotein(s) was replaced with a heterologous one. In one case, the G glycoprotein of recombinant vesicular stomatitis virus (VSV) Indiana strain was replaced by that of the New Jersey strain (21). In addition, the HN and F proteins of recombinant measles virus were replaced by the VSV G protein (27, 29). However, these chimeras were restricted in cell culture at least 10- to 50-fold compared to their respective parents, indicative of a significant defect (27, 29). This level of reduction is comparable to that observed when the cytoplasmic domain of rabies virus G was deleted, when G was deleted altogether (24), or when the transport of the VSV G protein to the cell surface was blocked by a temperature-sensitive defect (31). Thus, while highly efficient virion formation commonly is thought to involve contacts between the M protein and the cytoplasmic tail(s) of the glycoprotein(s), or at least to require the presence of both homologous components, it has long been clear that a lower but still very substantial residual level of virion morphogenesis occurs independently of the homologous glycoprotein(s). It is tempting to conclude that the reduced level of virion morphogenesis by the above-mentioned previous chimeric viruses, each lacking the homologous glycoprotein(s), represents this residual, glycoprotein-independent level. In comparison, the rPIV3-1 chimera described here is noteworthy because replacement of both glycoproteins did not reduce the efficiency of production of infectious virus at all. Efficient replication in vitro is required for the eventual manufacturing of vaccines for human use. Importantly, this study is unique in that this was demonstrated not only in cell culture but also in vivo, during infection of the respiratory tracts of hamsters.

PIV1 and PIV3 represent two distinct serotypes, and their HN and F glycoproteins have 43 and 47% sequence identity, respectively. The transfer of the two glycoproteins together would, of course, obviate glycoprotein-to-glycoprotein incompatibility (35). Since it is generally thought that the glycoproteins interact with the M protein (which is 63% identical between PIV1 and PIV3) through their cytoplasmic (CT) or transmembrane (TM) domains, it is interesting that the degree of sequence identity between the HN and F proteins of the two serotypes in the TM and CT domains is low indeed: 30 and 22%, respectively, for the TM domain, and 9 and 11%, respectively for the CT domain. In light of this low level of sequence relatedness, we also had pursued a parallel strategy of constructing chimeric glycoproteins in which the PIV1 ectodomain of each glycoprotein was fused to the PIV3 TM and CT domains (results not shown). However, the successful recovery of the recombinant described here, and its unimpaired capacity for growth, rendered this alternative but more complicated strategy unnecessary. It might be that a conserved structure, such as a constellation of charged amino acids, is important for interaction with the M protein or other internal proteins rather than a conserved sequence. Alternatively, it might be that interaction of the TM and CT domains of the glycoproteins with internal proteins is not as critical as has been previously thought. It will be possible to examine this issue for rPIV3-1 by using the reverse genetics approach. Also, this issue will be revisited during work in progress to construct a PIV3 variant bearing HN and F of PIV2.

It was expected that rPIV3-1 would require trypsin for efficient replication in tissue culture since this is a property conferred by the PIV1 F glycoprotein, and this was found to be the case. However, it was interesting that rPIV3-1 caused CPE that more closely resembled that of the PIV3 parent, indicating that a PIV3 gene(s) other than the HN or F gene specifies this phenotype. It will be possible to explore which gene or genes of PIV3 specifies this phenotype by the exchange of additional gene(s) between the noncytopathic PIV1 and the cytopathic PIV3.

It is now possible to exploit the reagents and experience generated during two decades of PIV3 vaccine development for the rapid development of PIV1 and PIV2 vaccines. The first step, described here, was to successfully recover the rPIV3-1 recombinant. The second step, the construction and recovery of a comparable chimera between PIV3 and PIV2, rPIV3-2, is in progress. The third step will be to identify and characterize the basis for the attenuation of the cp45 and BPIV3 vaccine candidates. For example, the differences between the nucleotide sequences of cp45 and the wt JS virus have been determined, and these mutations are being examined by their introduction, singly and in combination, into the wt rPIV3/JS background (32). This will result in a menu of attenuating mutations which can then be introduced into rPIV3-1 and rPIV3-2 chimeras to achieve the desired level of attenuation and immunogenicity.