Changes in Mumps Virus Gene Sequence Associated with Variability in Neurovirulent Phenotype

Abstract

Mumps virus is highly neurotropic and, prior to widespread vaccination programs, was the major cause of viral meningitis in the United States. Nonetheless, the genetic basis of mumps virus neurotropism and neurovirulence was until recently not understood, largely due to the lack of an animal model. Here, nonneurovirulent (Jeryl Lynn vaccine) and highly neurovirulent (88-1961 wild type) mumps virus strains were passaged in human neural cells or in chicken fibroblast cells with the goal of neuroadapting or neuroattenuating the viruses, respectively. When tested in our rat neurovirulence assay against the respective parental strains, a Jeryl Lynn virus variant with an enhanced propensity for replication (neurotropism) and damage (neurovirulence) in the brain and an 88-1961 wild-type virus variant with decreased neurotropic and neurovirulent properties were recovered. To determine the molecular basis for the observed differences in neurovirulence and neuroattenuation, the complete genomes of the parental strains and their variants were fully sequenced. A comparison at the nucleotide level associated three amino acid changes with enhanced neurovirulence of the neuroadapted vaccine strain: one each in the nucleoprotein, matrix protein, and polymerase and three amino acid changes with reduced neurovirulence of the neuroattenuated wild-type strain: one each in the fusion protein, hemagglutinin-neuraminidase protein, and polymerase. The potential role of these amino acid changes in neurotropism, neurovirulence, and neuroattenuation is discussed.

Mumps virus is a 15.3-kb enveloped, nonsegmented, negative-strand RNA virus in the Paramyxoviridae family. The mumps virus genome organization is 3′ NP-P-M-F-SH-HN-L 5′, in which the abbreviations represent the genes encoding the following proteins: NP, nucleoprotein; P, phosphoprotein; M, matrix protein; F, fusion protein; SH, small hydrophobic protein; HN, hemagglutinin-neuraminidase protein; and L, polymerase (16). The virus causes mumps, an acute communicable viral disease typical of childhood. Mumps virus is transmitted through oropharyngeal secretions, with primary virus replication in the respiratory mucosa. Following the development of viremia, mumps virus infects a number of targets, including the central nervous system (CNS), and was a common cause of viral meningitis in the prevaccine era (6). The most common neurological manifestation of CNS infection is aseptic meningitis (23, 37). More rarely, clinical manifestations involve encephalitis (31, 38), cerebellar ataxia (12), and transverse myelitis and poliomyelitis-like disease (29, 33). While symptoms are typically transient, CNS infection can produce significant morbidity (in such manifestations as, e.g., deafness [20, 36, 57]) and mortality, particularly in the case of encephalitis (38). Despite the clinical significance of mumps virus and the widespread international use of a number of mumps virus vaccines, little is known about the genetic basis for mumps virus attenuation or virulence.

In this study, we sought to identify genetic changes associated with attenuation-neurovirulence of two mumps virus strains, Jeryl Lynn (JL) and 88-1961. These two virus strains represent the extremes of the mumps virus neurovirulence spectrum. JL, a highly attenuated chicken embryo-adapted vaccine strain, is not causally associated with CNS complications in vaccinees (5) and induces no significant neuropathology in rat neurovirulence assays (44). In contrast, strain 88-1961, a wild-type strain isolated from a patient with CNS symptoms, causes severe, devastating neurological damage (e.g., hydrocephalus) in the rat neurovirulence assay (44). Strains JL and 88-1961 (accession numbers AF345290 and AF467767, respectively) are members of different mumps virus genotypes (3, 54) and differ at 989 bases. Thus, the specific genetic changes responsible for attenuation-neurovirulence phenotype cannot be distinguished by a simple comparison of genomes. Therefore, to better examine the molecular basis for neurovirulence-attenuation, these parental strains were passaged in conditions designed to select for variants with increased or reduced neurovirulence (as confirmed in the rat neurovirulence assay) and a genetic comparison was then made between the parental and the neuroadapted or neuroattenuated variant virus strains.

MATERIALS AND METHODS

Virus stocks.

The following mumps virus strains were used: (i) JL parental and JL vaccine (Merck and Co., Inc., West Point, Pa.) (22); (ii) JL-SY5Y, a variant of the JL parental strain generated by 10 serial passages on human SH-SY5Y neuroblastoma cells; (iii) JL-Vero, a variant of the JL parental strain generated by 10 serial passages on Vero cells; (iv) 88-1961 parental, a wild-type strain isolated from a throat swab of a patient with symptoms of CNS infection which was subjected to one passage in MK2 cells and two passages on Vero cells (kindly provided by Nando K. Chatterjee, New York State Department of Health); and (v) 88-1961-CEF, a laboratory derivative of the 88-1961 parental strain generated by four serial passages on chicken embryo fibroblast cells. Serial passages of viruses was carried out at a 1:20 dilution for each passage. Infectivity titers were measured by plaque assay as described previously (43). All viral stocks were prepared in minimal essential medium (MEM; Gibco BRL, Gaithersburg, Md.) supplemented with 5% fetal bovine serum (FBS; Gibco BRL), with the exception of JL parental, which was taken from the final marketed vaccine product.

In vitro virus characterization.

Monolayers of Vero cells in 12-well plates were inoculated in triplicate with the JL parental, JL-Vero, JL-SY5Y, 88-1961 parental, or 88-1961-CEF strain at a multiplicity of infection of 0.1 PFU/cell in MEM with 7% FBS. After 1 h of incubation at 37°C, inocula were removed and the monolayers were washed twice in MEM and incubated with 1 ml of MEM with 7% FBS at 37°C. Supernatants were removed from each well every 24 h over a 6- to 10-day period. Monolayers were washed twice with MEM between supernatant harvests, thereby allowing the measurement of 24-h de novo virus production at each time point. The titers of each virus harvest were determined by plaque assay (43). At the time of virus harvest, monolayers were examined by light microscopy and qualitatively assessed for the extent of cytopathic effects (cell-to-cell fusion, syncytium formation, and lysis).

Rats and virus inoculation.

Lewis rats (Harlan, Indianapolis, Ind.) (1 day old) were inoculated intracerebrally with 20 μl of MEM containing 100 PFU of the JL parental (n = 43), JL-SY5Y (n = 40), JL-Vero (n = 40), 88-1961 parental (n = 39), or 88-1961-CEF (n = 48) strain. As a control, two litters of rats were inoculated with an equal volume of uninfected Vero cell supernatant (n = 20). Animals were examined every other day, and those exhibiting clinical signs of disease (including lethargy, ataxia, and paresis) were euthanized immediately by CO₂ inhalation (all brains were saved). All other rats were kept until day 25 postinoculation (p.i.), when the experiment was terminated.

Virus titration and antigen distribution in the brain.

On days 2, 4, 6, and 8 p.i., the brains from two uninoculated control rats and six virus-inoculated rats from each group were removed aseptically. One control brain and three virus-inoculated brains from each group of rats were individually examined by plaque assay for virus content per gram of brain (43). The remaining brains were flash frozen in −80°C methyl butane and serially sectioned in the sagittal plane at 16-um intervals on a cryostat for immunohistochemical analysis. Brain sections were fixed in acetone, blocked with 3% normal goat serum in phosphate-buffered saline, and incubated with mouse anti-mumps virus monoclonal antibody (Chemicon International Inc., Temecula, Calif.). Some brain sections were dual stained with rabbit anti-glial fibrillary acidic protein (anti-GFAP) polyclonal antibody (Dako Corp., Carpenteria, Calif.) for identification of astrocytes or with rabbit anti-microtubule-associated protein-2 (MAP-2) (Chemicon International Inc.) for identification of neurons. Following the addition of Cy3-conjugated anti-mouse immunoglobulin G and/or fluorescein isothiocyanate-conjugated anti-rabbit immunoglobulin G secondary antibodies (Chemicon International Inc.), tissues were visualized under a fluorescent confocal microscope.

Histology.

Duplicate brain sections obtained as described above (on days 2, 4, 6, and 8 p.i.) were stained with hematoxylin and eosin and observed under light microscopy for evidence of neuropathology.

For evaluation of hydrocephalus severity, 15 to 24 brains from each group of virus-inoculated rats were removed on day 25 p.i. and immersion fixed in 10% neutral buffered formalin. Fixed brains were processed, embedded in paraffin, and sectioned sagittally at 8-um intervals. All sections were stained with hematoxylin and eosin and observed under light microscopy. The severity of hydrocephalus was assessed as described previously (44). Briefly, using Image Pro Plus image analysis software (Media Cybernetics, Silver Spring, Md.), the percentage of the total brain cross-sectional area (excluding the cerebellum) occupied by the lateral ventricle was measured in sagittal sections taken at a standard distance from either side of anatomical midline. The severity of hydrocephalus was determined as the mean percentage of hydrocephalus in each experimental group of rats.

Mumps virus genome sequencing.

Mumps virus RNA was extracted with a QIAamp Viral RNA Mini kit (Qiagen, Valencia, Calif.) and reverse transcribed using random hexanucleotide primers (Rd6) (New England Biolabs, Boston, Mass.) and a Superscript II kit (Invitrogen, Carlsbad, Calif.) according to the instructions of the manufacturers. Mumps virus-specific primers were deduced on the basis of the published sequences for the JL-1 (accession number AF338106) and 88-1961 (accession number AF467767) strains and were used to generate overlapping 1-kb PCR fragments spanning the entire length of the JL parental, JL-SY5Y, 88-1961 parental, and 88-1961-CEF strains. The first-round PCR products were sequenced, and new gene-specific primers were deduced from the generated sequence data to amplify the rest of the genome with a proofreading ThermalAce polymerase (Invitrogen). To determine the exact 3′ and 5′ termini of the viruses with a three- to fivefold redundancy, 3′ and 5′ RACE commercial kits (Stratagene, La Jolla, Calif.) were used. The cDNA at the 3′ end was prepared by adding a poly-A tail to the viral RNA and reverse transcribing with a poly-T adapter primer. In a second step, a gene-specific primer (in combination with a truncated adapter primer) was used to generate the PCR product. The cDNA at the 5′ end was prepared by adding a poly-C tail to the cDNA followed by PCR amplification with a combination of the truncated adapter primer and a gene-specific primer. Both gene-specific primers for the 3′ and 5′ ends were deduced from the Urabe AM9 mumps virus strain (accession number AB 000388). DNA sequencing was performed using an Applied Biosystems Prism 310 genetic analyzer with the AB1 prism Big Dye terminator V3.1 cycle sequencing kit for fluorescence detection. A total of 80 to 150 ng of each PCR fragment and 3 pmol of gene-specific primers was used in 10 μl of reaction mix to generate 500 to 600 bases of overlapping sequence information from both strands. Alignments of viral sequences were performed using Auto Assembler software (Applied Biosystems).

Comparisons were made between the complete nucleotide and amino acid sequences of the JL parental and JL-SY5Y strains and between the 88-1961 parental and 88-1961-CEF strains. GCG-lite DNA and protein analysis tools (National Institutes of Health), SRS-EMBL software, and the ExPASy proteomics server (Swiss Institute of Bioinformatics) were used for analyses of genetic changes.

Mutant analysis by PCR and restriction enzyme cleavage (MAPREC).

The JL parental strain is composed of two substrains, JL1 and JL2, which differ at 414 nucleotides, leading to 87 amino acid substitutions (1, 4). Here, using quantitative mutant analysis as described previously (4), nucleotide 12997 in the polymerase gene was used to differentiate JL1 (C) from JL2 (T). Briefly, viral genomic RNA was isolated from the JL parental, JL-Vero, and JL-SY5Y strains by phenol-sodium dodecyl sulfate extraction and reverse transcribed into cDNA with random dN6 primers by using SuperScript II RT (Life Technologies). A region spanning nucleotide 12997 of the polymerase gene was then radiolabeled by second-strand DNA synthesis with a 5′-[³²P]-labeled antisense partially mismatched primer and TaqDNA polymerase. Use of this primer created a Hinf1 restriction site for JL1 and an EcoR1 restriction site for JL2. The double-strand DNA was digested separately with the respective restriction endonucleases and separated by polyacrylamide gel electrophoresis. Areas of radioactive DNA bands containing both the full-length DNA (upper bands) and the restriction fragments (lower bands) were quantitated using a Storm 860 PhosphorImager (Molecular Dynamics). The percentage of JL1 or JL2 in the virus stocks was calculated on the basis of the relative radioactivity of the restriction fragment compared to the total measured radioactivity.

Statistical analysis.

The in vivo and in vitro virus titration data were analyzed using two-way analysis of variance (ANOVA) followed by a post hoc all-pairwise multiple comparison (Tukey test). Hydrocephalus severity scores were analyzed using Student's t test. All values are expressed as means ± standard errors of the means. Statistical significance was assigned for P values < 0.05.

RESULTS

Virus replication and cytopathic effect in vitro.

To assess growth of these five viruses in vitro, virus production and cytopathology were assessed in Vero cells. The kinetics of virus production results for all three JL-based viruses were similar, with peak production occurring on day 3 p.i. (Fig. 1). A two-way ANOVA comparison between the JL parental and JL-SY5Y virus strain titers and between the JL parental and JL-Vero virus strain titers revealed that the differences between the effects of various times p.i. [F(5, 35) = 46 and F(5, 35) = 35; all P < 0.001] and of various virus strains [F(1, 35) = 19 and F(1, 35) = 220; all P < 0.001] were significant. A post hoc all-pairwise comparison showed significant differences (all P < 0.01) between JL parental and JL-SY5Y virus strain titers and between JL parental and JL-Vero virus strain titers at all time points until day 6 p.i., when the titers of the JL parental and JL-SY5Y virus strains dropped to zero due to complete cell lysis. The course of cytopathology for these viruses paralleled that of virus production, with early stages of cell-to-cell fusion and syncytium formation occurring by day 2 p.i. for all three JL-based viruses. All monolayers were fused by day 4 p.i. and monolayers infected with the JL parental and JL-SY5Y virus strains were completely lysed by day 5 p.i., whereas the monolayers infected by JL-Vero remained intact through day 10 p.i., presumably reflecting a persistent infection by this virus strain, although this remains to be confirmed.

FIG. 1.

Results of 24 h of de novo virus production by Vero cells inoculated with the JL parental, JL-SY5Y, or JL-Vero strain.

The kinetics of virus production for the 88-1961 parental strain was similar to that of the 88-1961-CEF strain, with peak virus production for both virus strains occurring on day 3 p.i. (Fig. 2). A two-way ANOVA comparison of the 88-1961 parental and 88-1961-CEF virus strain titers revealed that the differences between the effects of various times p.i. [F(9, 59) = 865, P < 0.001] and of various virus strains [F(1, 59) = 235, P < 0.001] were significant. A post hoc all-pairwise comparison showed significant differences between 88-1961 parental and 88-1961-CEF virus strain titers at all time points (all P < 0.01) except for days 9 (P = 0.50) and 10 (P = 0.60) p.i. No cytopathic effects were noted in cultures infected by either of the strain 88-1961-based viruses, possibly indicating persistent infections.

FIG. 2.

Results of 24 h of de novo virus production by Vero cells inoculated with the 88-1961 parental or 88-1961-CEF strain.

JL-based virus infection of rats.

None of the rats inoculated with the JL parental (0/19), JL-SY5Y (0/16), or JL-Vero (0/16) strain developed signs of disease. Nonetheless, hydrocephalus of the lateral and third ventricles, the hallmark of mumps virus infection in rats (43, 44), was evident in some rats. The incidence and severity of hydrocephalus in each group of virus-infected rats is shown in Fig. 3. The severity of hydrocephalus, calculated as the ratio of the cross-sectional area of the brain to that of the lateral ventricle, was significantly greater (P = 0.03) in rats inoculated with strain JL-SY5Y (1.74% ± 0.8%) than that seen with rats inoculated with the JL parental strain (0.07% ± 0.07%). The differences in the severity of hydrocephalus calculated between strain JL-Vero and JL parental strain infections were not significant (P = 0.82).

FIG. 3.

Incidence (A) and severity (B) of hydrocephalus in rats inoculated with the JL parental, JL-SY5Y, or JL-Vero strain.

The concentrations of JL parental, JL-SY5Y, and JL-Vero virus in rat brain at various times p.i. are shown in Fig. 4A. No differences were observed at any time point in the amount of mumps virus recovered from rats inoculated with strain JL-Vero versus the JL parental strain (day 2, P = 0.84; day 4, P = 0.78; day 6, P = 1.0; day 8, P = 1.0). In contrast, peak production of the JL-SY5Y virus continued in the brain between days 2 and 4 p.i., while JL parental strain peak titers were recovered at day 2 p.i. and gradually declined thereafter. Although the increase in JL-SY5Y virus strain load in the brain on day 4 p.i. was small, it was statistically significant [F(2, 35) = 0.2, P = 0.04].

FIG. 4.

(A) Virus growth in rat brains inoculated with the JL parental (solid circle), JL-Vero (triangle), or JL-SY5Y (open circle) strain. (B) Virus antigen expression in a rat brain stained indirectly with fluorescently labeled anti-mumps virus monoclonal antibody. Note the restriction to the ventricular ependymal cell lining (arrow).

Immunohistochemical staining revealed that expression of all three JL-based viruses was limited to the ependymal cell lining of the lateral and third ventricles (Fig. 4B). There was no indication of differences in the extent of infection among the different virus substrains.

Strain 88-1961-based virus infection of rats.

In contrast to results of inoculation with the JL-based viruses, inoculation with the strain 88-1961-based viruses resulted in significant clinical disease, including lethargy, ataxia, and paresis. The incidence of clinical disease was much greater in rats inoculated with the 88-1961 parental strain (9/16 [56%]) than in rats inoculated with strain 88-1961-CEF (3/24 [12%]). The incidence and severity of hydrocephalus in each group of virus-infected rats are shown in Fig. 5. The severity of hydrocephalus was significantly greater in rats inoculated with the 88-1961 parental strain (12.2% ± 2.0%) than in rats inoculated with 88-1961-CEF (4.38% ± 1.3%) (P = 0.002). In some rats with severe hydrocephalus, interstitial edema of the white matter of the brain, distortion of the hippocampal formation, Chiari type 1 malformation of the cerebellum, and cerebellar neuronal migration abnormalities were also observed as reported previously (43, 52). As was the case for rats inoculated with the JL-based viruses, there was no evidence of ventriculitis, meningitis, or encephalitis at any of the time points evaluated.

FIG. 5.

Incidence (A) and severity (B) of hydrocephalus in rats inoculated with the 88-1961 parental or 88-1961-CEF strain.

The amount of infectious virus recovered at all time points was lower in rats inoculated with strain 88-1961-CEF than in rats inoculated with the 88-1961 parental strain (Fig. 6A). A two-way ANOVA comparison of the 88-1961 parental and 88-1961-CEF strain virus titers revealed that the differences in the effects of the virus strains were significant [F(1, 29) = 0.2, P < 0.001].

FIG. 6.

(A) Virus growth in rat brains inoculated with the 88-1961 parental (solid circle) or 88-1961-CEF (open circle) strain. (B) Virus antigen expression in a rat brain stained indirectly with fluorescently labeled anti-mumps virus monoclonal antibody. Note the spread of virus from the ependymal cell lining to areas deeper in the surrounding brain parenchyma (arrow).

Immunohistochemical staining revealed that unlike the JL-based viruses, the 88-1961 parental and 88-1961-CEF strains were able to spread from the ependymal cell lining (the apparent initial site of infection) to areas deeper in the surrounding brain parenchyma (Fig. 6B). There was no indication of differences in the extent of infection between the 88-1961 parental and 88-1961-CEF strains. Double labeling with antibodies to GFAP or MAP-2 demonstrated that the virus-infected cells in the periventricular area were predominantly neurons and not astrocytes (Fig. 7).

FIG. 7.

Confocal laser scanning microscopy of the subventricular region of a rat infected with strain 88-1961, showing colocalization of virus antigen with neurons and not glial cells. (A and D) 88-1961-positive cells (red). (B) Neuron-specific MAP-2-positive cells (green). (C) Coincident confocal image of panels A and B, demonstrating virus antigen expression in neuronal cells (yellow). (E) GFAP-positive cells (green). (F) Coincident confocal image of panels D and E, demonstrating the absence of virus antigen expression in glial cells.

Genome comparison of JL variants.

The nucleotide substitutions and predicted amino acid changes between the JL parental and JL-SY5Y strains are shown in Table 1. There were a total of seven nucleotide substitutions, resulting in three predicted amino acid changes: Phe→Pro⁴⁶⁷ in the NP, Val→Ala⁸⁵ in the matrix protein, and Glu→Asp¹¹⁶⁵ in the polymerase. Amino acid changes in the NP and matrix protein were not predicted to affect charge or hydrophobicity and were not located in any of the known structural motifs or functional domains. In contrast, the Glu→Asp¹¹⁶⁵ substitution in the polymerase was located in a highly conserved domain believed to be important in RNA synthesis (14). All other nucleotide substitutions were silent, and none were in the 3′ noncoding region (NCR).

TABLE 1.

Genome comparison of JL variants

Viral genea	Nucleotide change (JL parental→JL-SY5Y)	Amino acid change (JL parental→JL-SY5Y)
NP	1402/03 (TT→CC)	467 (Phe→Pro)
M	253 (T→C)	85 (Val→Ala)
HN	1380 (T→C/T)	None
L	1197 (A→G)	None
	2689 (C→T)	None
	3495 (G→T)	1165 (Glu→Asp)

^aM, matrix protein; L, polymerase.

Genome comparison of strain 88-1961 variants.

The nucleotide substitutions and amino acid changes between the 88-1961 parental and 88-1961-CEF strains are shown in Table 2. There were a total of four nucleotide changes, of which one was silent and the remaining three were predicted to result in the following amino acid changes: Pro/Thr→Thr⁹¹ in the fusion glycoprotein, Ser→Asn⁴⁶⁶ in the HN glycoprotein, and Ile→Val⁷³⁶ in the polymerase protein. Amino acid position 91 in the fusion glycoprotein, which in the parental strain is a heterogeneous mixture of proline (nonpolar hydrophobic) and threonine (polar uncharged), becomes homogeneous for threonine upon passage in CEF cells. Not only does this change alter polarity at this position, but it is also predicted to lead to a homogeneous population of viruses with an N-glycosylation site at amino acids 89 to 92. In the HN glycoprotein, the Ser→Asn⁴⁶⁶ substitution was predicted to result in loss of an N-glycosylation site (amino acids 464 to 467). The Ile→Val⁷³⁶ substitution, while not predicted to result in changes in charge or hydrophobicity, does lie within a possible functional motif within a highly conserved domain of the polymerase protein (42).

TABLE 2.

Genome comparison of 88-1961 variants

Viral genea	Nucleotide change (88-1961 parental→88-1961-CEF)	Amino acid change (88-1961 parental→88-1961-CEF)
M	1068 (C→A)	None
F	271 (A/C→A)	91 (Pro/Thr→Thr)
HN	1398 (G→A)	466 (Ser→Asn)
L	2206 (A→G)	736 (Ile→Val)

^aM, matrix protein; F, fusion protein; L, polymerase; HN, hemagglutinin-neuraminidase protein.

MAPREC quantification of JL1 and JL2 substrains in JL-based viral stocks.

Results of MAPREC analysis are shown in Fig. 8. Phosphorimager analysis of the gel demonstrated that the JL1:JL2 ratios in the JL parental, JL-SY5Y, and JL-Vero strains were 5:1, 1:0, and 9:1, respectively.

FIG. 8.

MAPREC analysis at nucleotide position 12997, demonstrating variations of the JL1 and JL2 substrains in JL parental, JL-Vero and JL-SY5Y strains. Each MAPREC quantitation included a negative control (Uncut) and a restriction enzyme digest that discriminated between JL1- and JL2-specific nucleotide sequences.

DISCUSSION

Serial passage of neurotropic-neurovirulent viruses in nonneuronal substrates to achieve neuroattenuation for the preparation of vaccines has been successfully performed for a number of viruses, including polio virus and mumps virus (9, 17, 18). This process apparently selects for mutant viruses with an increased ability to replicate under specific conditions, with a concomitant reduction in the ability to infect the natural target tissue. Neuroattenuation is evaluated on an empirical basis, with passage continued until sufficient attenuation is achieved. The reverse process, that of neuroadaptation, is conceptually similar. Prolonged successive passages of virus in neuronal tissues or cells are used to select for mutant viruses with an increased tropism for those substrates. This technique has been employed for research purposes to generate neuroadapted variants of several viruses, including Sindbis virus, Newcastle disease virus, respiratory syncytial virus, yellow fever virus, influenza virus, and measles virus (7, 11, 24, 30, 45, 48). In this study, we exploited the neuroadaptation and neuroattenuation processes in combination with virus genome sequencing to identify specific genomic changes that might influence mumps virus neurotropism (the ability of the virus to infect and replicate in the brain) and neurovirulence (the ability of the presence of the virus to lead to damage in the brain).

Neuroadaptation of the JL parental strain, a nonneurotropic, nonneurovirulent mumps vaccine strain (5, 43, 44), was achieved by serial passage on human neuronal cells (SY5Y). The resulting variant virus strain, JL-SY5Y, appeared to have a slight replication advantage in rat brain (neurotropism) and, more significantly, was more neurovirulent (judged on the basis of an increased incidence and severity of hydrocephalus) than the JL parental strain. Despite an increased ability of JL-SY5Y to replicate and cause damage in brain, the gross tissue distributions of the JL parental and JL-SY5Y strains appeared to be similar (both being limited to the ventricular ependymal cells). Of note, the increased ability of strain JL-SY5Y to grow in brain is not likely a simple reflection of the fact that JL-SY5Y is mammalian cell adapted whereas the JL parental strain is chick cell adapted, since no differences in virus growth were observed between rats inoculated with the JL parental strain versus those inoculated with strain JL-Vero, another mammalian-cell-adapted variant of the JL parental strain.

In comparing the genomic sequence between the JL parental and JL-SY5Y strains, seven nucleotide changes were found, all in coding regions. Three of the changes were predicted to result in amino acid substitutions: one each in the NP (Phe→ Pro⁴⁶⁸), matrix protein (Val→Ala⁸⁵), and polymerase (Glu→Asp¹¹⁶⁵). As the NP and the polymerase are involved in replication and transcription and the matrix protein is believed to be involved in the virion assembly process (10), it is quite plausible that changes in these proteins are responsible for the observed in vitro and in vivo findings of enhanced growth and virulence of strain JL-SY5Y. Notably, amino acid changes in the NP, matrix protein, and polymerase in other viruses have been associated with changes in neurovirulence phenotype (40, 47, 59, 60). While in some cases, changes in protein sequence have been predicted to result in structural-functional changes, e.g., elimination of a phosphorylation site in the rabies virus NP (59) or disruption of the pH-dependent association-disassociation between the matrix protein and the ribonucleoprotein complex in influenza virus (47, 60), the precise mechanism by which these changes affect neurovirulence of these and other viruses is not clear. In this study, neither of the amino acid changes in the NP and matrix protein of JL-SY5Y was predicted to result in structural-functional changes. However, not all structural-functional motifs have been identified for these proteins; therefore the possibility of the NP and matrix protein amino acid changes affecting virus phenotype cannot be ruled out. Unlike the NP and matrix protein amino acid changes, evidence does exist for a functional role of the Glu→Asp¹¹⁶⁵ substitution in the polymerase. This particular amino acid change is located in the highly conserved V domain of the protein (14, 42, 46). Nearby mutations in this domain in Sendai virus, a close relative of mumps virus, have been shown to significantly affect RNA synthesis in vitro and in vivo (14). Thus, it is possible that the Glu→Asp¹¹⁶⁵ substitution in the JL polymerase can affect virus growth. Confirmation of an effect on replication, transcription, or virulence caused by the NP, matrix protein, and polymerase amino acid changes identified here will ultimately require application of a reverse genetics system.

An alternate possible explanation for the observed differences in JL parental and JL-SY5Y strain replication in vitro and replication and virulence in vivo might lie in the relative proportions of JL1 and JL2, the two substrains known to comprise the JL strain (1). Since it has been demonstrated that the JL1:JL2 ratio is host substrate dependent (4), we measured the JL1:JL2 ratios in the three JL-based viruses used here. Our results showed that JL2 was comprised of 15% JL parental strain and 10% strain JL-Vero, whereas JL-SY5Y was entirely comprised of JL1. Thus, it is tempting to speculate that relative to JL2, JL1 is of increased replication competence, which may manifest as heightened neurotropism and neurovirulence in the rat assay. At this point, however, potential effects of JL1:JL2 ratio changes cannot be separated from potential effects of the NP, matrix protein, and polymerase amino acid changes.

To examine genomic differences following neuroattenuation, the 88-1961 parental strain, a highly neurotropic and neurovirulent strain (44), was passaged on an avian fibroblast cell line (CEF), mimicking classical techniques used to produce live, attenuated vaccine from wild-type mumps virus (9). As predicted, relative to the 88-1961 parental virus, the 88-1961-CEF variant showed a reduced ability to replicate in rat brain, a reduced incidence and severity of hydrocephalus, and a lower incidence of clinical disease. Consistent with these in vivo findings, strain 88-1961-CEF also showed a reduced ability to replicate on Vero cells, suggesting that reduced tropism for mammalian cells contributed to the observed decrease in infectivity in rats. Again, however, simple differences in replication competence in mammalian cells cannot fully explain the neurovirulence outcomes, since strains 88-1961-CEF and JL-SY5Y replicated to similar titers in mammalian cells in vitro yet had vastly different neurovirulence outcomes in vivo.

Unlike the JL-based virus stains, replication of the strain 88-1961-based viruses was not restricted to the ependymal cells. Numerous virus-infected neurons were also detected in the surrounding brain parenchyma. These data suggest that infection of cell types other than the ependymal cells can contribute to the heightened incidence and severity of hydrocephalus and clinical disease. However, while differences in neuroanatomic or cellular sites of replication may account for some of the variations in hydrocephalus severity between JL and 88-1961 strains, those differences may not explain the observed variations in neurovirulence between 88-1961 parental and 88-1961-CEF, since no obvious variations were observed in the neuroinvasiveness or distribution of virus among rats infected with these viruses. Certainly, a better understanding of the precise mechanism(s) by which mumps virus leads to hydrocephalus will ultimately be required for a better appreciation of the role of virus-infected subventricular neurons in hydrocephalus development. To date, the preponderance of data suggests that hydrocephalus is the result of accumulation of virus-infected cell debris in the aqueduct of Sylvius, blocking cerebrospinal fluid egress (27, 28, 49). A similar mechanism is believed to be responsible for mumps virus hydrocephalus in humans (21, 41, 55). However, hydrocephalus has also been observed prior to, or in the total absence of, aqueductal stenosis (26, 50, 51, 58), suggesting that stenosis of the aqueduct can be a secondary consequence of external compression by surrounding edematous tissue in already hydrocephalic animals and not causally related to the pathogenesis of hydrocephalus.

In comparing the genomic sequences of 88-1961 parental and 88-1961-CEF strains, four nucleotide substitutions were found, three of which were predicted to result in amino acid substitutions: one each in the polymerase (Ile→Val⁷³⁶), fusion (Pro/Thr→Thr⁹¹), and HN (Ser→Asn⁴⁶⁶) proteins.

All three of the predicted amino acid substitutions, individually or in combination, are good candidates for affecting virus replication and virulence. The polymerase Ile→Val⁷³⁶ substitution is located within a putative functional motif in the highly conserved domain III region of the polymerase protein (42, 46). This motif, identified by Poch et al. as motif B, is believed to represent an important element of the active site for template recognition and/or phosphodiester bond formation (49). If such a function exists for this region, changes in its amino acid sequence could lead to changes in virus growth and, by extension, virulence.

The predicted amino acid substitutions in the fusion and HN glycoproteins are also located in known functional regions. The Pro/Thr→Thr⁹¹ change in the fusion protein, a protein responsible for fusion of viral and cellular membranes, is located within a region involved in the disulfide bond linkage between the F₁ and F₂ active subunits of the F₀ precursor protein (61). In addition to possibly affecting the stability of the fusion protein, this change also impacts the polarity-hydrophobicity at this position and gives rise to a N-glycosylation site at amino acids 89 to 92. While this amino acid population change therefore has the potential to affect protein function, no differences were observed between levels of fusogenicity of the 88-1961 parental strain and 88-1961-CEF strain in vitro. Other more sensitive measurements of fusion protein function are planned.

Of the strain 88-1961 amino acid changes, the Ser→Asn⁴⁶⁶ substitution in the HN glycoprotein, the protein that promotes viral attachment to cellular receptors, is most promising. The HN protein has been shown to play a role in mumps virus neurovirulence. For example, a change at position 360 in the highly neurovirulent hamster brain-adapted Kilham mumps virus strain was associated with a decreased ability to infect neurons in hamsters and attenuated disease (32, 35), and amino acid changes at positions 335, 464, and 498 of the Urabe-AM9 mumps virus vaccine strain have been associated with meningitis in vaccinees (2, 8). The HN protein has two apparently opposing activities, adsorption to sialic acid-containing cell surface molecules (hemagglutination) and enzymatic cleavage of sialic acid (neuraminidase activity). Recent studies have indicated that both functions are located in a single sialic acid recognition site (13, 15). A recent crystal structure study of the HN glycoprotein of a closely related paramyxovirus, Newcastle disease virus, indicates that amino acid position 466 may be at or near this active site (15). Thus, the Ser→Asn⁴⁶⁶ substitution in the HN glycoprotein of strain 88-1961-CEF, which is predicted to result in a loss of an N-glycosylation site (amino acids 464 to 467), has the potential to affect virus tropism and virulence. Whether or not the loss the N-glycosylation site translates to a change in secondary structure is not clear. While the Garnier, Osguthorpe, and Robson algorithm predicted a significant change in the secondary structure (turns and β-sheets) of this protein, the Chou-Fasman algorithm predicts no structural changes. Nonetheless, loss of N-glycosylation sites, regardless of effects on secondary structure, have been linked to changes in the neurovirulence in other viruses, including influenza virus (34, 56), lactate dehydrogenase-elevating virus (19) and yellow fever virus (39).

Considering that JL and 88-1961 strains represent different genotypes (25, 54), it is not surprising that none of the amino acids associated with JL neuroattenuation were identified in the attenuated 88-1961-CEF virus strain and none of the amino acids associated with the enhanced neurovirulence of the JL-SY5Y virus were identified in the neurovirulent 88-1961 parental virus strain. In a more expansive intergenotypic comparison of several mumps virus strains representing a wide range of neuropathogenicity for humans, no attenuation or neurovirulence-specific clusters were identified (unpublished data).

The results presented here demonstrate that the affinity of mumps virus for the brain can be increased or decreased by passage on neuronal or nonneuronal-nonmammalian cells, respectively. The relative abilities of these viruses to replicate in rat brains could not completely account for the relative levels of neurovirulence of the viruses, since both strain JL-SY5Y and strain 88-1961-CEF replicated to similar titers and yet strain 88-1961-CEF was significantly more neurovirulent than strain JL-SY5Y. Remarkably, despite multiple passages in vitro on cells of differing classes (avian versus mammalian), a small number of genetic changes, individually or in combination, appeared to account for these observed differences. Further studies, including reverse genetics, are required for confirmation of the specific role each of these protein changes in neurovirulence.