Mapping of Genetic Determinants of Rubella Virus Associated with Growth in Joint Tissue

Abstract

Rubella virus (RV) strains vary in their abilities to replicate and persist in cell cultures derived from human joint tissue (synovial cells [SC]), and this arthrotropism appears to be linked to their association with joint symptoms in vivo. In order to map the genetic determinants of arthrotropism, an infectious clone of the Cendehill vaccine strain of RV was constructed, as well as two chimeric clones containing cDNAs from both Cendehill and Therien (wild-type) strains. Replacement of the entire structural gene region of Therien in the infectious clone pROBO302 with the corresponding region of Cendehill did not affect growth in SC. A further observation that Cendehill bound equally well to SC and the permissive Vero cell line indicated that restriction was not at the level of receptor binding, a function of the envelope proteins. Mutations that affected growth in joint cells were mapped to two locations in the nonstructural gene region. The first of these (nucleotides 2803 and 6416) resulted in a 10-fold decrease in yield of progeny virus from SC. This region contained five mutations, at nucleotides 2829, 3060, 3164, and 3528 (near the carboxy terminus of P150 where the protease domain is located) and at nucleotide 4350 in p90. Further substitution of the sequence representing nucleotides 1 to 2803 to give a complete Cendehill infectious clone restricted growth in SC by a further 100-fold to less than 10 PFU/ml. This region contains three mutations, at nucleotides 34, 37, and 55, within the 5′ stem-loop structure. In conclusion, the Cendehill-specific mutations believed to be determinants of joint cell growth are located in two regions, the 5′ nontranslated region and in a sequence that encodes the carboxy-terminal region of p150 extending into the helicase domain of p90.

Rubella virus (RV), the etiologic agent of German measles, belongs to the family Togaviridae and is the only member of the genus Rubivirus. Natural infection in childhood causes a systemic illness characterized by a short-lived maculopapular rash and mild fever (50). The disease is generally benign, and infection is often asymptomatic. It is the teratogenic potential of rubella that brought the virus to the forefront of public health interests and provided the impetus for isolation of the virus and subsequent vaccine development (35, 49). The current vaccine strain RA27/3 has been very effective in reducing the incidence of congenital rubella syndrome in North America, where it is given to all children between 12 and 18 months of age. However like the wild-type strains and the earlier vaccine strain, HPV77/DE5, it is reported to be associated with acute and late-onset joint and neurological symptoms (20, 46, 47, 50).

The association of RV with acute, transient joint manifestations, after both natural infection and vaccination, has been recognized for many years (14, 23, 34, 46, 47). Rubella-associated arthritis (RAA) is usually short-lived, although a number of patients go on to develop chronic or recurrent pauci- or polyarticular symptoms which can persist for some time (7, 8, 22, 43, 45, 47). Studies to define the mechanism of pathogenesis of RAA have been limited by the fact that humans are the only natural host for RV and there is presently no animal model of infection. However, the frequency and intensity of clinical symptoms reported for wild-type (wt) and vaccine strains correlate directly with the ability of the infecting strain to propagate in organ cultures of human synovial tissue, suggesting that tropism for joint tissue is a measure of viral arthritogenicity (31). Although RV strains are genetically around 98% homologous, they display striking phenotypic variation in growth characteristics and plaque morphology as well as tropism for joint tissue (9, 31). The wt strains, such as Therien, which have the highest association with persistent joint symptoms (30%) (47), commonly replicate to titers of 10⁶ to 10⁷ PFU/ml in organ cultures of human joint tissue, comparable to the yields from the most permissive cell lines for RV. The vaccine strain RA27/3, which is associated with much lower levels of recurrent arthritis (4%) (47), is severely restricted in these cultures and does not attain titers greater than 10³ PFU/ml. However RA27/3, like the wt Therien strain, was found to persist in joint culture for over 3 months (31). In contrast, no replication of the European vaccine strain, Cendehill, was detected in human joint tissue in this study. Cendehill strain is reported to have a very low association with acute arthritis and none with chronic joint manifestations (4). These results indicate a correlation between the arthrotropism of a specific RV strain and its ability to induce joint symptoms and lends support to the hypothesis that recurrent RAA is triggered by reactivation of virus which has established a persistent infection in the joint.

A similar growth trend was observed in primary cultures of dissociated synovial cells (SC), although these cells are slightly more susceptible to all RV strains than joint tissue (31). For example, in these cells, growth of Cendehill virus was highly restricted, but even after repeated changes of the growth medium over 10 to 14 days, low titers of virus were detected (<10² PFU/ml), indicating that the virus was replicating at a very low level. In comparison, the RA27/3 strain grew to intermediate levels (10⁴ PFU/ml) and the wt strains (Therien and M33) produced high titers (10⁶ to 10⁷ PFU/ml) of progeny virus. In Vero cells, on the other hand, infection with either Cendehill or RA27/3 results in approximately 10-fold-lower yields than the wt strains, indicating that they possess attenuating mutations which limit their growth. However both vaccine strains must also possess additional mutations to account for their severely restricted growth in SC.

These findings indicate that both human joint fragments in organ culture and dissociated synovial cells can serve as model systems to investigate the arthrotropism of different RV strains. The dissociated SC cultures have the advantage of being easier to manipulate and providing greater consistency for comparative experimentation than primary organ cultures. In the absence of an animal model system, analysis of the properties which restrict virus growth in human SC provides a system to define what makes certain strains of RV nonarthritogenic.

The present study involved the mapping of genetic differences between the Cendehill (vaccine) and Therien (wt) strains associated with this variation in joint cell growth. To accomplish this, an infectious clone of the Cendehill strain has been constructed by insertion of Cendehill-specific cDNAs comprising the entire genome into a Therien cDNA clone pROBO302, provided by T. K. Frey, Atlanta, Ga. (37, 38, 48). Two chimeric cDNA clones consisting of part Therien and part Cendehill have also been produced, and the virus derived from these was compared with the parental strains for phenotypic properties of interest.

In addition the ability of each of the parental strains (Therien and Cendehill) to attach to human joint cells and their ability to replicate following transfection of viral RNA have been examined in order to address which stage(s) of the infectious cycle is involved in growth restriction of the vaccine strain. Finally, the sequence of the full-length infectious clone of the Cendehill strain was derived and compared with the published wt sequences of Therien and M33 strains in order to identify nucleotide changes in Cendehill strain in regions of the genome found to be associated with growth restriction in human joint cells.

MATERIALS AND METHODS

Cells, viruses, and bacteria.

The Therien strain of RV was originally derived from a wt strain isolated in the United States by A. Schluederberg and was passaged in Finland to obtain a derivative which gave more cytopathology and grew to higher titer than the original strain (33). The Cendehill vaccine strain was produced by Rohm Pharma, Weiterstadt, Germany. Vero (African green monkey kidney) and BHK (baby hamster kidney) cells were obtained from the American Type Culture Collection. Primary cultures of fetal human SC were prepared from autopsy material. WM1100 electrocompetent Escherichia coli cells were purchased from Bio-Rad Inc.

Plasmids and cloning vectors.

pCL1921 is a plasmid vector encoding spectinomycin resistance, constructed by Lerner and Inouye (27), which contains the polycloning site of pUC19 (51) in place of the polylinker region of pGB2 (12). pSUPC is a plasmid vector constructed by inserting a polycloning site containing restriction sites known to be single-cut sites in RV cDNA into the pSU8 vector (1) to facilitate the subcloning of RV cDNAs (30). pROBO302 is an infectious clone of the Therien strain in the low-copy-number vector pCL1921 and was generously supplied by T. K. Frey. Transcripts from this clone give yields of 10⁴ PFU/μg of RNA on transfection into BHK/21 cells.

Cell and virus culture.

Vero cells were cultured in medium 199 (M199; Gibco BRL) supplemented with 10% fetal bovine serum (FBS). BHK cells were grown in Dulbecco modified Eagle medium-F12 supplemented with 10% FBS. Following infection, all cell lines were incubated with medium (as described above) containing 5% heat-inactivated FBS (HIFBS) and 1% gentamicin in humidified incubators at 35°C in an atmosphere of 5% CO₂.

Primary culture of SC.

Human fetal knee joints were dissected a maximum of 6 h after autopsy as previously described (31). Sections extending approximately 2 mm to either side of the joint were excised, minced into 1-mm pieces, and cultured overnight. SC were dissociated from the matrix by incubation with 1% collagenase (Boehringer Mannheim) for 4 h at 37°C and 5% CO₂, followed by digestion with 0.25% trypsin–0.2% EDTA. SC were collected by centrifugation and were cultured in RPMI-1640 supplemented with 10% FBS–1% gentamicin. SC were passaged by trypsinization once a week and were used up to the fourth passage.

Plaque titration of virus.

Virus titers were determined by plaque assay as described by Fogel and Plotkin (19). Six days following inoculation the monolayers were examined for visible microfoci or plaques (Therien) and were then overlaid with 2 ml of M199–0.5% agarose containing neutral red dye (150 μg/ml) and observed for clearings after 24 h (Cendehill).

Virus concentration.

Culture fluid from virus-infected cells was collected and treated in one of two ways: the medium was either centrifuged for 4 h at 80,000 × g in a Beckman L870 M, SW27 rotor, at 4°C or collected and layered onto 25% sucrose in phosphate-buffered saline (PBS). Virions were pelleted through the sucrose cushion by ultracentrifugation at 150,000 × g for 2 h at 4°C.

Virus growth assay.

Subconfluent cells in 60-mm-diameter dishes were inoculated with virus at 0.1 PFU/cell. After a 1-h adsorption at 35°C, the cells were washed twice with 5 ml of PBS and supplemented with 5 ml of M199. At various times 500-μl aliquots were removed (and replaced with fresh medium) and stored at −70°C for plaque titration. Zero-time samples were taken immediately following the postadsorption washes for determination of the amounts of virus remaining from the inoculum. To determine the amount of intracellular virus the cells were washed twice with 5 ml of PBS and then freeze-fractured in 1 ml of M199 at −70°C. Prior to plaque titration the suspension was centrifuged for 2 min at 13,000 rpm in an Eppendorf Microfuge to remove cell debris.

Virus neutralization.

Virus neutralization was carried out with polyclonal rabbit anti-Therien antiserum at a 1:100 dilution for 1 h at 4°C. Following this pretreatment the virus stocks were serially diluted and plated as described above for plaque titration, except that the overlay contained a 1:500 dilution of anti-Therien antibody.

Western blotting and SDS-polyacrylamide gel electrophoresis.

Virus samples were prepared by sedimentation of the supernatant medium from infected cells (20 ml for Therien strain and 150 ml for RA27/3, Cendehill, and JCND strains) for 4 h at 80,000 × g. The pellets were resuspended in solubilization buffer (0.0625 M Tris [pH 6.8], 2% sodium dodecyl sulfate [SDS], 10% glycerol, 0.001% bromophenol blue) and heated at 100°C for 3 min. The proteins were separated by electrophoresis through discontinuous 10% polyacrylamide gels containing 0.1% SDS at 32 mA for 2.5 h (26).

Proteins were transferred to an Immobilon polyvinylidene difluoride (PVDF) membrane (Millipore) with a sodium carbonate buffer system containing 20% methanol (18). Transfer was carried out at 50 mA for 30 min, followed by 300 mA for 3 h in a water-cooled apparatus. PVDF membranes were blocked with 3% gelatin (40°C) and incubated with a 1:100 dilution of polyclonal anti-RV antiserum for 1 h at room temperature with gentle agitation. The bound antibodies were detected with the Vectastain ABC kit (Vector Laboratories), with 4-chloro-1-naphthol as the substrate.

Comparative binding assay.

Virus-infected Vero cells were labelled from 48 to 64 h postinfection (Therien) or from 72 to 88 h postinfection (Cendehill) with 30 μCi of trans-[³⁵S]cysteine-methionine (ICN) in RPMI 1640 deficient in cysteine and methionine (Sigma)–5% HIFBS. Labelled virus particles were purified by centrifugation through 25% sucrose and were resuspended in M199–5% HIFBS. Controls were prepared with labelled supernatant from uninfected Vero cells treated in an identical manner. The stocks were adjusted to pH 7.2 by the addition of 0.1 volume of 500 mM HEPES (Sigma) buffer. Lightly confluent monolayers of either Vero or SC cells in 1.9-cm² wells (24-well plates) were inoculated in triplicate with 80 μl of each stock and allowed to equilibrate in an atmosphere of 5% CO₂ for 5 min. The plates were then sealed with Parafilm and incubated a further hour at 25°C with gentle agitation. The cells were washed twice with ice-cold PBS and then solubilized in 0.5 ml of 0.5% SDS in H₂O. Each sample was suspended in 9 volumes of Cytoscint scintillation cocktail (ICN) and counted with a Beckman LS6000IC liquid scintillation counter. An average value for each set of triplicates was used to determine the amount of virus binding to SC or Vero cells. This was normalized to compensate for nonspecific binding of similarly labelled material from uninfected cells. Replicate variability within an experiment was less than 15%.

Transfection.

Cells were electroporated following the general procedure outlined previously (28). A Bio-Rad Gene Pulser with pulse controller was used to shock the cells in Bio-Rad electroporation cuvettes with a 0.2 mm electrode gap. The cells were suspended in ice-cold PBS (10⁷ cells/ml) and were mixed with 0.5 μg of RNA, estimated by visualization of an aliquot on ethidium bromide-stained agarose gels. They were subjected to two pulses, at 75 kV for SC or at 1.5 kV for BHK cells, with the pulse controller set at infinity. Twenty-four hours after electroporation the cells were washed with PBS, the medium was replaced, and the cells were incubated at 35°C in a humidified CO₂ incubator.

Electrocompetent WM1100 cells were electroporated by a modification of the method of Dower et al. (16) according to the procedure outlined in the Bio-Rad product description. The mixture was subjected to one pulse of 1.8 kV and 25 μF, with the pulse controller set at 200, in cuvettes with a 0.1-cm electrode gap. The cells were incubated for 1 h at 37°C prior to being spread on Luria-Bertani plates.

Plasmid isolation.

Plasmids were isolated by the alkaline lysis method (5) and were purified by polyethylene glycol precipitation as described by Sambrook et al. (41).

Ligation.

Ligations were carried out at 15°C for 20 h with T4 DNA ligase at 1 U/μg (BRL). The 5× ligase buffer supplied by the manufacturer contained 250 mM Tris-HCl (pH 7.6), 50 mM MgCl₂, 5 mM ATP, 5 mM dithiothreitol, and 25% (wt/vol) polyethylene glycol-8000.

Infectious clone construction. (i) Isolation of viral RNA.

Supernatant virus was concentrated by centrifuging the medium from virus-infected cells (harvested 72 to 98 h postinfection) for 4 h at 80,000 × g in a Beckman SW27 rotor. Virions were solubilized in 20 to 100 μl of 1% SDS, and viral RNA was isolated by extraction with guanidinium hydrochloride (11) or with Trizol (Gibco BRL/Life Technologies Inc.) according to the manufacturer's instructions.

(ii) Reverse transcription.

Specific primers complementary to the published sequence of the Therien strain (15) were used to initiate the first strand of DNA synthesis. The primers used were no. 16, 38, and 125 (Table 1). A mixture of primers and viral RNA in diethyl pyrocarbonate-water was heated for 3 min at 90°C. cDNA was synthesized with 200 U of Superscript II (Gibco BRL) in a reaction mixture for 1 h at 42°C. The standard reaction mixture contained 1 mM deoxynucleoside triphosphates (dNTPs), 10 mM dithiothreitol, and 10% dimethyl sulfoxide. The volume was brought to 100 μl by the addition of Tris-EDTA buffer and heated to 90°C to inactivate the reverse transcriptase. Enzyme, primers, and excess nucleotides were removed by extraction of the mixture with 1 volume of phenol-chloroform-isoamyl alcohol (25:24:1 [vol/vol/vol]), followed by 1 volume of chloroform.

TABLE 1.

Oligonucleotides used to prime first-strand synthesis and thermal cycling

Primer	Sequence	Positionc
F1	5′-CGCGAATTCTTTTTTTTTTTTTTTTTTTTCTATACAGCAACAGGTGC-3′a
F2	5′-TCGAAGCTTATTTAGGTGACACTATAG-3′b
9	5′-TGCAGCGTTCGACGCAAACG-3′	2133–2153
10	5′-TCCGAGTGCCGTTGCGATC-3′	2243–2262
16	5′-GCGTTCTTGATGTCGATATCGCG-3′	4410–4431
18	5′-CTCACTGATGTCTACACGCAGATG-3′	5281–5763
38	5′-GGAATTCCACTAGTTTTTTTTTTTTCTATACAGCAAC-3′	RV 3′ end complement
46	5′-CAACCACCTCGGGAATGC-3′	3241–3260
125	5′-TAGTCTTCGGCGCTTGG-3′	5747–5763
251	5′-TTTGCCAACGCCACGGC-3′	2603–2618

^aThe EcoRI site is in boldface; the RV 3′ end complement is underlined. 

^bThe HindIII site is in boldface; the SP6 sequence is underlined; the RV 5′ end sequence is overlined. 

^cLocation in the RV genome. 

(iii) Thermal cycling amplification.

Products of the first-strand reactions were amplified with specific primers (Table 1) and repeated cycles of incubation with Deep Vent (NEB) thermostable polymerase with 3′–5′ proofreading exonuclease activity. The standard reaction mixture contained 400 μM dNTP, 2 mM MgSO₄, 0.5 μM primer, and 1 U of polymerase. After the cycling reaction, unincorporated dNTP, proteins, and oil were removed with a QIAquick spin column (Qiagen).

pROC3 (see Fig. 2).

FIG. 2.

Construction of JCND. Cendehill double-stranded (ds) cDNA was produced by reverse transcription of Cendehill RNA followed by thermal cycling amplification. The ds cDNAs were cut with appropriate restriction enzymes and inserted sequentially into the analogous region of the Therien infectious clone (pROBO302), which had been similarly digested.

The first strand was synthesized with oligonucleotide 38 and was then amplified by thermal cycling with oligonucleotide 18 as the forward primer and F1 as the reverse primer. F1 contained a T₂₀ tract and additional nucleotides for an EcoRI restriction site downstream of the viral sequence. The cycling conditions were 98°C for 30 s, 60°C for 30 s, 75°C for 30 s, and 80°C for 4 min (34 cycles). The final product was a 4,478-bp cDNA homologous to the Cendehill genomic RNA from nucleotide (nt) 5281 to the 3′ end of the Cendehill genomic RNA. This fragment also contained an EcoRI site downstream of the terminal poly(A) sequence to facilitate cloning. The fragment was digested with EcoRI and BglII (nt 5355) and ligated into pROBO302, which had been similarly digested. The products were electroporated into WM1100 cells to yield an infectious clone in which the Therien cDNA had been replaced by the analogous Cendehill sequence from nt 5355 to the 3′ end.

pROC3M (see Fig. 2).

Oligonucleotide 125 was used to prime reverse transcription. The first strand was then amplified by thermal cycling with oligonucleotide 251 as the forward primer and oligonucleotide 125 as the reverse primer. Cycling conditions were 98°C for 30 s, 52°C for 30 s, and 72°C for 2.5 min (34 cycles). The product was a 3,160-bp cDNA homologous to nt 2603 to 5765 of the Cendehill genomic RNA. This was digested with NheI (nt 2803) and BglII (nt 5355) and subcloned into pSUPC. A mixture of fragments from four subclones was prepared for ligation into pROC3. Since pROC3 contains two sites for NheI (nt 2803 and nt 8690), an appropriate deletion was created by digesting the plasmid completely with BglII and then partially with NheI to obtain the desired 11,448-bp vector. The ligation product was electroporated into WM1100 cells to yield an infectious clone in which the Therien cDNA had been replaced by the analogous Cendehill sequence from nt 2805 to the 3′ end.

pJCND (see Fig. 2).

Oligonucleotide 16 was used to prime reverse transcription. Two fragments were amplified separately from the product of the first-strand reaction. Oligonucleotides F2 (forward) and 10 (reverse) were used to amplify a fragment from the 5′ end to nt 2262. A HindIII restriction site and an SP6 RNA polymerase promoter sequence were included in primer F2 upstream of the viral sequences. Cycling conditions were 98°C for 30 s, 50°C for 10 s, 75°C for 30 s, and 80°C for 2 min (34 cycles). Oligonucleotides 9 (forward) and 46 (reverse) were used to amplify a second fragment from nt 2133 to 3260. The cycling conditions were 98°C for 30 s, 54°C for 30 s, and 78°C for 1.5 min (34 cycles). These two cDNAs were digested with PvuI, which cuts at nt 2246 (and nt 9475). The cDNA fragment produced by using primers 9 and 46 was dephosphorylated to inhibit self-annealing, and the two fragments (1.5 μg) were ligated overnight at 15°C. Following ligation the mixture was digested with HindIII and NheI and the fragment was purified by electrophoresis on an agarose gel.

Since pROC3M contains two NheI sites (nt 2803 and 8690), to prepare for insertion of the 5′-terminal Cendehill cDNA, it was first digested with HindIII and BglIII to create a deletion from nt 1 to 5357 of the pROC3M sequence. This fragment was ligated to the fragment from nt 2803 to 5357 isolated from a second pROC3M digest, and the ∼11-kb fragment was gel purified. The product was pROC3M with a deletion from nt 1 to 2803, which was ligated overnight at 15°C with the HindIII-NheI Cendehill fragment and transformed into WM1100 cells by electroporation to yield the Cendehill infectious clone, pJCND, in which all of the Therien cDNA had been replaced by corresponding Cendehill sequences.

Screening of constructs.

The sequences of the selected clones were compared with the sequences of Cendehill cDNAs taken directly from thermal cycling reactions to confirm that the correct fragments had been inserted. The Cendehill infectious clone and the chimeric clones were all screened for infectivity. To accomplish this, the plasmids were linearized by digestion with EcoRI at the 3′ terminus of the viral sequence. Positive-polarity viral RNA was generated by transcription from the SP6 promoter, and the products were transfected into BHK21 cells by electroporation. After 2 days the supernatants were transferred to Vero cells, which were not amenable to electroporation but which gave higher and more-consistent yields of virus. Samples were removed for plaque titration 5 days later.

Sequencing.

Automated sequencing was carried out by the University of British Columbia, Nucleic Acid and Protein Sequencing Unit (Biotechnology Laboratory). Amplitaq dye terminator cycle sequencing (ABI) reagents were used, and fluorescent products were analyzed spectrophotometrically.

Nucleotide sequence accession number.

The sequence obtained in this study has been assigned GenBank accession no. AF188704.

RESULTS

Growth of Therien and Cendehill strains in SC.

To confirm previous results on the relative permissiveness of SC to the Therien and Cendehill strains (31), SC cultures were infected with each strain and washed rigorously following the adsorption period. After 4 days the levels of intracellular and supernatant virus were assayed (Fig. 1). A baseline of 5 × 10² PFU of virus/ml remaining after removal of the inoculum was determined by titration of medium sampled immediately after the adsorption period. As found previously, Therien replicated well and produced high titers of both intracellular and extracellular virus while Cendehill strain was severely restricted in its replication; titers above the basal level remaining from the inoculum were not detected in either the cellular or supernatant fraction, and no cytopathology was observed.

FIG. 1.

Viral titers detected in the supernatant medium or intracellularly in SC, 4 days postinfection with Cendehill (Cend) and Therien strains of RV. The level of residual virus remaining following removal of the inoculum and three washes in PBS is shown as the baseline of residual virus.

Comparative binding.

In order to determine whether the block to Cendehill strain replication in SC is due to impaired receptor interaction, we examined the binding of ³⁵S-labelled Therien and Cendehill virus to SC. Since virus preparations vary in such factors as the particle/PFU ratio and intensity of ³⁵S labelling, the ability of each strain to bind to SC relative to its ability to bind to Vero cells, which are highly permissive to both strains, was quantitated. Equal amounts of infectious virus (PFU) were added in each case. The specific activities of the Cendehill preparations were consistently lower than those for Therien, reflecting the former strain's markedly slower growth rate. For this reason, the absolute numbers of counts, for virus bound to either cell type, were lower for Cendehill. The results of a typical experiment are shown in Table 2.

TABLE 2.

Binding of sucrose density-purified ³⁵S-labelled supernatant virus to SC and Vero cells^a

Cells	Binding (cpm) by strain:
	Cendehill	Therien
SC	6,695	9,123
	6,831	9,490
		11,221
Avg	6,763	9,945
Vero cells	6,990	20,615
	7,553	20,662
	8,850	21,367
Avg	7,798	20,881

^aResults for a typical experiment (triplicate determinations for each cell type) are shown. The ratios of SC binding to Vero cell binding were 0.86 and 0.47 for Cendehill and Therien strains, respectively. 

The Cendehill strain was found to bind equally well to both cell types (average ratio of binding [SC/Vero cells] = 0.9; range, 0.80 to 1.1). We had also expected comparable binding to Vero cells and SC for the Therien strain, since it replicates to high titers in both cell types. Surprisingly, we found that the level of Therien strain binding to SC was approximately half that of its binding to Vero cells (average ratio of binding [SC/Vero cells] = 0.48; range, 0.43 to 0.52).

Since the levels of binding of the Cendehill strain to Vero cells and SC were nearly equivalent it seems unlikely that its growth restriction in SC occurs at the level of receptor recognition and binding. Moreover, because Therien grows well in both cell types, the lower level of binding to SC does not appear to interfere with its ability to establish a productive infection in these cells.

Electroporation of viral RNA into SC.

The comparative-binding studies described above suggested that the restriction of Cendehill in SC was not a result of impaired attachment. However these experiments do not rule out the possibility that Cendehill binds to an alternative receptor on SC that does not facilitate uptake. In order to circumvent any restrictions on entry and uncoating and to examine the intrinsic ability of the genomic RNA to initiate replication in SC, RNA isolated from Cendehill and Therien virus was electroporated into SC (∼10 ng per 5 × 10⁶ cells) and viral yields were determined. The cultures were harvested 5 days postelectroporation and were assayed for the presence of intracellular and extracellular virus. The Therien RNA was found to initiate a highly productive infection, with titers of 4.0 × 10⁴ PFU/ml detected in the medium and 4.0 × 10⁵ PFU/ml found intracellularly. This was in marked contrast with Cendehill RNA, which gave no detectable virus in the intracellular sample and only 10 to 20 PFU/ml in the medium, indicating that replication was severely restricted. Increasing the amount of RNA used for electroporation raised the yield such that low levels of Cendehill virus were detected in the cells (10 to 20 PFU/ml) as well as in the medium (10 to 100 PFU/ml). However, the yield of Cendehill was consistently 10³-fold lower than the corresponding yield for the Therien strain.

Construction of the Therien/Cendehill chimeras and Cendehill infectious clone.

In order to localize the regions of the genome involved in the restriction of Cendehill virus growth in joint tissue, two chimeric viruses and a full-length infectious clone of Cendehill were constructed by replacing portions of the Therien cDNA of pROBO302 with corresponding regions of Cendehill cDNA. Details of the cloning strategy are given in Materials and Methods and are depicted in Fig. 2. In pROC3 the sequence from nt 5355 to the 3′ end (nt 9762) of pROBO302 was replaced by the analogous Cendehill cDNA; this included the entire structural gene region. In pROC3M the Therien sequence of pROBO302 comprising nt 2805 to the 3′ end was replaced by the analogous Cendehill cDNA. Finally, the pROBO302 cDNA was completely replaced by the analogous Cendehill sequence to produce the full-length infectious clone of the Cendehill strain, pJCND. These clones were used as templates for transcription of infectious, positive-polarity viral RNA. After electroporation into BHK cells the progeny viruses were subsequently passaged three times in Vero cells to obtain high-titer virus, and the resultant stocks were designated the ROC3, ROC3M, and JCND strains. All of the clones replicated in Vero cells at levels comparable to that of pROBO302, giving titers of 10⁵ to 10⁶ PFU/ml.

Screening of constructs.

In order to confirm the identity of the progeny of the Cendehill infectious clone (JCND), several properties were compared. (i) Comparison of the sequence of pJCND cDNA and those of cDNAs derived directly from the Cendehill reverse transcription-PCR reaction mixture showed no differences. (ii) The ability of polyclonal RV antiserum to neutralize each strain was examined. Neutralization of the parental Cendehill strain was 92.5 to 99.95%, while that of JCND was 89.5 to 99.5% in three different experiments. (iii) The plaque morphology of Cendehill virus obtained from infected Vero cell supernatants was compared to that for JCND (Fig. 3). Plaques from both the parent virus and the construct showed the characteristic “bull's-eye” appearance previously described for the Cendehill strain and readily differentiated from the microfoci produced by the Therien strain (9, 42). (iv) The structural polypeptides of Cendehill and JCND were compared on Western blots with those of the RA27/3 vaccine strain (Fig. 4). The protein patterns for all three strains were similar, showing strong representation of the E2 protein and less of the E1 and C proteins, which were detected only as dimers in the Therien strain. The E2 band appears to migrate slightly faster for the Cendehill and JCND strains. Sequence analysis has shown that Cendehill lacks one of four potential glycosylation sites found in the E2 coding region, and therefore the fully processed molecule would be expected to have a slightly lower molecular weight. These results indicate that the infectious clone (pJCND) encodes infectious virus with phenotypic properties similar to those of the Cendehill strain.

FIG. 3.

RV plaques were allowed to develop for 7 days postinoculation and were visualized by adding an overlay containing neutral red dye. (A) Mock-infected; (B) Cendehill strain; (C) JCND strain. The bulls-eye plaques, characteristic of Cendehill strain, can be seen in panels B and C.

FIG. 4.

The structural proteins of Cendehill and JCND were visualized following nonreducing SDS-polyacrylamide gel electrophoresis and transfer to a PVDF membrane by using polyclonal antiserum to RV. Therien and RA27/3 strains were included for comparison. The locations of E2 and E1 monomers are marked (E2 is typically present as a series of glycosylation variants). The unglycosylated C protein appears exclusively as a dimer in nonreducing gels.

Electroporation of infectious transcripts into SC.

Full-length infectious RNA transcripts from the Therien and Cendehill infectious clones, as well as from the chimeric pROC3 and pROC3M clones, were tested for their abilities to replicate in SC. RNA transcripts (0.5 μg/5 × 10⁶ cells) from each of the infectious clones were electroporated into SC and assayed on day 5 as described previously. The results are shown in Table 3.

TABLE 3.

Ability of viral transcripts to initiate productive infection in SC^a

RV strain	Virus titer (PFU/ml) in:
	Cells	Medium
pROBO302	1.6 × 102	1.9 × 103
pROC3	1.6 × 102	2.5 × 103
pROC3M	2.1 × 101	2.4 × 102
pJCND	2	0

^aRNA transcripts (0.5 μg) from pROBO302, the Therien infectious clone, from the Therien/Cendehill chimeric infectious clones (pROC3 and pROC3M), and from the Cendehill infectious clone (pJCND) were transfected into SC by electroporation. SC were examined 5 days postelectroporation for intracellular and extracellular virus titers. 

Intracellular and extracellular titers of ROC3 were found to be equivalent to those of ROBO302. This demonstrated that substitution of the Cendehill structural genes into ROBO302 did not produce a restrictive effect on replication. In contrast, the titers of ROC3M showed a 10-fold reduction in extracellular virus and an 8-fold reduction intracellularly. Thus it appears that mutations in the region of nt 2803 to 5355 play a role in growth restriction of Cendehill in SC. Finally, pJCND transcripts showed essentially no production of progeny virus in SC (a further 100-fold reduction in virus yield), indicating that mutations in the 5′ terminus, from nt 1 to 2803, play a major role in the growth restriction of Cendehill virus in SC.

Comparative analysis of the nonstructural gene regions of Cendehill and Therien.

The sequence of the entire genome of the Cendehill strain has been determined (GenBank accession no. AF188704). In order to characterize nucleotide differences which might account for the phenotypic differences between the strains, the sequence of the Cendehill nonstructural gene region was compared with the published sequences of the Therien and M33 strains. The Therien sequence used was that reported by Dominguez et al. (15), with corrections as noted by Pugachev et al. (38). These changes include insertion of an arginine codon at nt 1268 (CGC) and GC at nt 6262. A consensus M33 sequence was used, as reported by Pugachev et al. (38).

Cendehill contains 81 substituted bases in the 6,426 nt of the nonstructural region relative to Therien strain, a frequency of 1.3%. Mutations found in the Cendehill sequence which were shared with the M33 (wt) strain were not considered relevant to growth restriction in joint tissue since M33 grows well in organ cultures of joint tissue and in dissociated SC (31). Mutations at the third position of a codon (thus not resulting in an amino acid substitution) were also not considered likely to be involved in growth restriction and, with the exception of mutations within putative control regions or known protein binding domains, are not included in this analysis. The results identify five amino acid substitutions and three mutations in control regions in the Cendehill nonstructural gene region which were not present in either Therien or M33 strains. Figure 5 and Table 4 show the relative locations of these mutations.

FIG. 5.

Mutations in Cendehill relative to Therien strain are noted on a schematic of the RV nonstructural gene region. With the exception of the mutations nearest the 5′ terminus (nt 34, 37, 55, and 358), only those mutations within the nonstructural protein ORF which result in amino acid changes are noted. (Note that mutations which were found in common between Cendehill and M33 (wt) are not shown.) Regions of homology with known functional domains are noted: MTR, methyltransferase; PRO, protease; HEL, helicase; POL, polymerase.

TABLE 4.

Strain Cendehill mutations^a

nt position	nt or aa in strain:
	Therien		Cendehill
	nt	aa	nt	aa
34	C		U
37	U		C
55	A		G
358	U		C
2829	G	Cys	A	Tyr
3060	A	Asp	G	Gly
3164	U	Thr	C	His
3528	C	Ala	U	Val
4530	C	Thr	U	Ile

^aThe nucleotide and amino acid changes found in the 5′ NTR and nonstructural gene region of the Cendehill strain in comparison with the equivalent nucleotides and amino acids in the Therien strain are shown. Only changes which occurred within putative control regions or which resulted in amino acid changes have been included. 

5′ SL.

Cendehill contains three mutations within the predicted 5′ stem-loop (SL): a C-to-U transition at nucleotide 34, a U-to-C transition at nucleotide 37, and an A-to-G substitution at nucleotide 55 (Table 4). These result in an alteration in the predicted structure involving an enlargement of the medial loop from 6 to 8 nt, an increase in the bulge at nt 46 to 51, and an increase in the size of the terminal loop from 6 to 11 nt (Fig. 6). These mutations could result in altered binding of host factors associated with translation (5′ positive-strand SL) or positive-strand replication (3′ negative-strand SL). Interestingly, the mutation at nt 55 results in the deletion of a stop codon for the first open reading frame (ORF) of the virus (UAG to UGG). This ORF is defined by the first AUG, which begins at nt 3 and has the potential to encode a 16-amino-acid peptide in the Therien strain.

FIG. 6.

SL structures in the 5′ nontranslated region of Cendehill and Therien strains predicted by the RNA folding program mfold (52).

P150.

The first 1,300 codons from the 5′ terminus of the genome, starting at AUG41, comprise P150. One change was found in Cendehill at nt 358, within the region encompassing nt 347 to 375 reported to be essential for the binding of genomic RNA to the capsid protein (29). No changes in this binding domain were found in M33, RA27/3, or any of the other six strains examined by Johnstone et al. (24).

The first amino acid change in the nonstructural gene region unique to Cendehill was found at residue 929, where a G-to-A substitution results in the replacement of cysteine with tyrosine. This occurs within the region of homology with the alphavirus NSP3 (15). Replacement of the SH-containing cysteine might alter intra- or intermolecular bonding and consequently the structure of the mature protein. A second change, from asparagine to glycine, is found at residue 1006. This mutation may be included in the same domain or may be in part of the beginning of the protease domain, the 5′ limit of which has not yet been determined.

Another substitution at residue 1041 substitutes tyrosine for histidine. Although both have bulky aromatic side groups, histidine is partly positively charged at neutral pH (pK = 6.5) while tyrosine possesses an uncharged polar side chain. The substitution may therefore affect intra- or interchain bonding. The fourth alteration results in replacement of alanine with valine at position 1162. This is a conservative substitution but is located 11 amino acids (aa) from the catalytic cysteine (aa 1151) of the protease (10).

P90.

This protein contains two putative functional domains. Homology to a global helicase motif has been identified between amino acids 1300 and 1600, near the amino terminus of the protein, and homology to a replicase domain between residues 1871 and 1971 has also been identified (15). Cendehill has only one unique change within P90, located at aa 1496: replacement of C with U at nt 4527 results in a change from a polar threonine to a nonpolar isoleucine. The amino acid sequence of P90 is highly conserved, and Cendehill, RA27/3, and M33 collectively (13, 38) contain only 6 substituted amino acids of the 905 amino acids of the protein, a putative replicase (frequency of mutation, 0.7%). In comparison, P150 was found to contain a total of 43 substituted amino acids collectively in M33, RA27/3, and Cendehill, relative to Therien. This is an overall frequency of 3.3%, indicating that a higher degree of variability in P150 is tolerated.

DISCUSSION

Although there is only one serotype of RV, there are many different isolates that have been laboratory adapted and also a number of vaccine strains that have been attenuated in various ways. While these cross-neutralize (21), variation in phenotypic characteristics such as plaque morphology (25), temperature sensitivity, and cell tropism (9) has been noted. In particular, significant variation in the abilities of different strains to replicate and persist in joint tissue was found, and this arthrotropism was found to correlate closely with the association of different strains with the induction of joint symptoms in vivo. More specifically, the Cendehill vaccine strain was found to be completely inhibited in its growth in organ cultures of human joint tissue and highly restricted in SC cultures (31). In contrast, the wt strains, Therien and M33, replicated to high titers in both these systems, and the RA27/3 vaccine strain showed intermediate growth characteristics.

Mapping the genetic differences associated with the variation in joint cell tropism has involved the production of an infectious clone of the Cendehill vaccine strain (pJCND) and also two infectious Therien/Cendehill chimeric cDNA clones, pROC3 and pROC3M. A comparative analysis of the abilities of RNA transcripts derived from these infectious clones, relative to that of the Therien clone, pROBO302, to initiate a productive infection in SC has indicated that the nonstructural gene region is primarily involved in growth restriction of the Cendehill strain in joint tissue.

Attenuating mutations in Cendehill virus have been mapped to two of the nonstructural gene regions, the 5′ SL and the sequence encoding the carboxy-terminal region of P150 extending into the amino terminus of p90. The concept that more than one locus can contribute to a phenotypic property is well documented in virology (6, 36, 40). For example, two mutations in E1, and one in E2, of Sindbis virus have been found to participate in the attenuation of neurovirulence. Alteration of E1 or E2 alone produced a reduction in neurovirulence, but alteration of both yielded a greater degree of attenuation (36). As another example, the polymerase protein of the P1/Sabin strain of poliovirus contains three amino acid changes that are required to produce temperature sensitivity. None of these mutations alone is sufficient to achieve a temperature-sensitive phenotype (6). Conversely, the same phenotypic property can be generated by completely different mutations. Two separate determinants of neurotropism were found to reside in either the 5′ nontranslated region or in the E2 glycoprotein of Sindbis virus (17). These studies indicate that one phenotypic property might be generated through a variety of mutations. That this occurs in RV is suggested by our observations that Cendehill and RA27/3 vaccine strains did not contain any mutations in common that were not also found in wt strain M33, although both share many phenotypic properties (9). This means that phenotypic traits shared between the two vaccine strains have been generated by different mutational events.

Mutations at nt 34, 37, and 55, within the 5′ SL structure, have been identified as important components of the restriction of Cendehill virus growth in joint tissue. This control region is believed to play a role in both replication and translation of the RV genome; therefore, these mutations could affect the rate or specificity of either process. The mutation at nt 55 might also disrupt the production of a small peptide from the first short ORF. Elimination of the stop at nt 54 to 56 would produce a protein extended to 28 amino acids in length with a molecular mass of approximately 3 kDa. Although it appears that this first ORF is not essential for viral replication (39), read-through in this area might interfere with normal protein synthesis or with synthesis from the third AUG. Alternatively, it is possible that a small protein is produced which, although not essential, might affect virus replication or translation in a cell-specific manner.

Another possible effect of mutations in the 5′ SL is interference with initiation of positive-strand RNA synthesis from the negative-strand replicative intermediate. Alterations to the analogous structure in Sindbis virus had deleterious effects on viral replication (32). A similar phenomenon may occur with RV, although one report has suggested that viral protein synthesis is affected by alterations in the 5′ SL to a much greater degree than RNA synthesis (39). Using the pROBO302 infectious clone, Pugachev and Frey (39) have shown that a variety of single point mutations within the SL loop are tolerated by the virus. Many of these changes had an effect on plaque morphology, and some caused a small reduction in virus titer. However none of these effects could be correlated with specific perturbations in the SL structure. Tolerance for substitutions within the SL was supported by the report of five separate alterations found at various sites in the 5′-terminal sequences of six other strains of RV (24). Mutations in this region are therefore generally not lethal, although they likely account for differences in replicative efficiency between the strains and also, from the present study, tissue tropism.

Four additional nonconservative amino acid substitutions due to mutations between nt 2828 and 4531 in the nonstructural genes appear to contribute further to the nonarthrotropic phenotype. One of these occurs in a region with homology to the essential NSP3 protein of Sindbis virus, two occur near the protease domain, and one occurs in the viral helicase domain. However, the functional consequences of such substitutions may reach outside of the proteins in which they are found. In the alphaviruses the RNA polymerase activity resides in a separate protein, NSP4, but appears to require the helicase and all of the other nonstructural proteins, as well as host factors, to form an active replication complex (2, 3). In addition, it is postulated that a complex of all of the alphavirus nonstructural proteins may be required to regulate the synthesis of positive- and negative-strand genomic RNAs (44). In place of the four nonstructural proteins found in the alphaviruses, the helicase and polymerase are combined in P90 of RV, while the methyltransferase, protease, and NSP3-homologous domains are combined in P150 (44). By analogy with the alphaviruses, these domains likely interact both within and between the two RV nonstructural proteins, and a change in the helicase domain, for example, might affect the function of a putative replicase complex.

Binding of Cendehill strain to SC did not appear to be impaired. Moreover, substitution of the structural gene region of Cendehill virus into the Therien infectious clone, pROBO302, did not have any effect on the replication of progeny in SC, suggesting that these genes and control regions do not play a role in determining tropism for joint tissue. In addition, although we have not directly examined the ability of the virus to successfully enter and uncoat following binding, the observation that Cendehill replication is highly restricted following electroporation of genomic RNA into SC (while Therien is not) gives weight to the idea that Cendehill restriction occurs subsequent to attachment and entry.

In conclusion, the restriction of growth of the Cendehill strain of RV in SC has been mapped to several mutations unique to this strain in the nonstructural gene region. The precise role of each of these individual mutations in determining the resulting phenotype is currently under investigation.