Assessment of the Epidemic Potential of a New Strain of Rotavirus Associated with the Novel G9 Serotype Which Caused an Outbreak in the United States for the First Time in the 1995-1996 Season

Rotavirus causes severe morbidity in developed countries andfrequent deaths ( $>=$ 500,000 per year) in less-developedcountries. Historically, four serotypes—G1, G2, G3, andG4—have predominated; they are distinguished by one oftwo surface neutralization antigens (VP7). However, in 1983and 1984 we described a new rotavirus serotype, designated G9,in five children hospitalized for diarrhea in Philadelphia,Pa. G9 rotavirus was not identified again in the Western Hemisphereuntil it caused ca. 50% of the rotavirus disease detected inPhiladelphia in the 1995-1996 season. This outbreak allowedus to question whether a rotavirus strain completely new toa well-studied community would target either very young infantsor older children, cause especially severe disease, or completelydisplace previously extant serotypes. We observed a significantexcess of G9 infections in younger infants (especially in those<6 months old) that might be attributed to the lack of G9-specificantibodies in mothers. Of further note, six of the seven oldestpatients with rotavirus diarrhea were infected with the G9 strains(not significant). However, the age distribution of childrenwith rotavirus did not differ over a 5-year study period regardlessof the infecting serotype. Patients with diarrhea associatedwith G9 strains did not have disease more severe than that causedby the G1, G2, or G3 serotype. G9 strains did not displace theother serotypes but were virtually completely replaced by G1or G2 serotypes in the three subsequent rotavirus seasons. Weconclude that the abrupt appearance of this novel rotavirusserotype did not present a special threat to public health inthe community.

INTRODUCTION

Top

Introduction

Rotavirus is the leading cause of severe diarrhea among childrenworldwide and a priority target for vaccines development (7).The scientific basis for these vaccines has been the observationthat immunity to rotavirus develops after natural infection(1, 22), and this immunity is at least partially dependent uponthe serotype of the infecting strain (21). Consequently, tobe effective, vaccines must protect against the main serotypesof rotavirus, and surveillance activities have been mountedby a number of groups to monitor strains and characterize theserotypes that are in circulation. In particular, vaccines preparedfrom animal strains of rotavirus of bovine or simian originhave been most effective when they contain reassorted gene productsthat encode for each individual human serotype (4, 19).

Rotaviruses are small (100-nm) icosahedral viruses with threeshells that encode 11 segments of double-stranded RNA. Serotypesare determined by two gene segments that encode two proteinsthat form the outer capsid of the virus: VP7, the glycoproteinor G protein, and VP4, the protease-cleaved or P protein. Historically,four common G serotypes designated G1, G2, G3, and G4 have beenidentified worldwide. Because of the segmented nature of thegenome, rotaviruses can also be characterized by the patternof migration of their segmented genome in polyacrylamide gelelectrophoresis (PAGE). For surveillance, the serotypes of rotavirusare most often determined by reverse transcription-PCR (RT-PCR)methods that provide a genetic profile that correlates wellwith serotype (9, 10).

As part of our effort to develop new rotavirus vaccines basedupon a bovine rotavirus strain, our group has been monitoringstrains of rotavirus circulating in Philadelphia from patientswith diarrhea who submitted fecal specimens to the virologylaboratory at the Children's Hospital of Philadelphia. Surveillancewas sporadic in the 1980s but became routine in 1992 and, eachseason since then, a sample of all rotavirus strains has beencharacterized for serotype and electropherotype. Each year,we have observed the circulation of the four historically commonstrains, most frequently G1 and G3. Then, in the 1995-1996 season,we noted the appearance of a new strain distinguished by itsuncommon short electropherotype that we designated P[6] G9.The G9 strain, W161 (P[8] G9), had been first observed and characterizedby our group in Philadelphia in 1983 but did not reappear after1984 (3). Since the mid-1980s, other G9 strains have been identifiedin Asia (strains AU32 [16] and F45 [11]), and in the 1990s theyhave been found with increased frequency worldwide (6, 14, 20).

The emergence of an entirely new serotype of rotavirus in Philadelphiain a population without previous exposure to this novel P[6]G9 strain provided a unique opportunity to question whetherthis major strain shift might, like influenza, cause an epidemicin a population without serotype-specific neutralizing antibodyor, alternatively, whether preexisting immunity might protectagainst diarrhea caused by this heterotypic strain. Specifically,we questioned whether this novel strain might cause more severedisease in infected children or affect older children withoutsolid immunity to this novel strain. We describe here the emergenceof this novel G9 strain in Philadelphia and our studies comparingthe age and clinical severity of disease in children infectedwith the G9 rotavirus to those of children infected with commonstrains.

MATERIALS AND METHODS

Top

Materials and Methods

Clinical specimens. From 1994 to 1999, fecal specimens submitted to the virologylaboratory at the Children's Hospital of Philadelphia in thewinter season (December to May) were screened for rotavirusby using a commercial enzyme-linked immunosorbent assay (Rotaclone;Meridian Diagnostics, Inc.). Most specimens (>80%) were obtainedfrom patients admitted to the hospital with diarrhea, althoughan average of 13% were from patients whose diarrhea was nosocomialand developed >72 h after admission.

Strain characterization. Specimens that tested positive for rotavirus were further characterizedby polyacrylamide gel electrophoresis (PAGE) and RT-PCR. Thestrains were tested by PAGE, followed by silver staining bymethods previously described (8), and strains whose 11 segmentshad the same migration profile were grouped for selective typing.A sample of strains from each PAGE group was then tested forG and P type by RT-PCR at the Centers for Disease Control andPrevention as reported previously (9, 12).

Clinical information. To assess whether patients with G9 versus patients with non-G9strains differed in disease severity or by age at presentation,we examined all available patient records from the 1995-1996season to collect standard information on the patient's durationand severity of diarrhea, vomiting, fever, and malaise, factorswhich were then scored on a scale of severity from 2 to 24 asdescribed previously (2). These observations were also comparedto those of the first 24 rotavirus disease cases identifiedin the previous and immediately subsequent rotavirus seasons.Scoring was performed twice, once by a physician and once bya microbiologist, and the values were then compared and averagedto provide a summary result for comparison.

RESULTS

Top

Results

Emergence of serotype G9. The distribution of rotavirus serotypes was plotted for a sampleof the specimens that were successfully electropherotyped forthe five winter seasons from 1994 to 1999 (Fig. 1A). Each year,the distribution of serotypes varied, with G1 strains beingthe most common overall (38%). In the 1995-1996 season, thenew serotype P[6] G9 strains with a short electropherotype appearedand made up a majority of the cases. The novel strain emergedwithout warning in January 1996, remained throughout the entire1996 winter season, and did not reappear at the beginning ofthe next winter season (Fig. 1B). The Philadelphia strain wasindistinguishable from the subsequently isolated strain US 1205by genogroup analysis as determined by Kirkwood et al. (13)and was extremely close to strain US 1205 by phylogenetic analysisof the VP7 gene as determined by Ramachandran et al. (18). Theseasonality of rotavirus disease in the 1995-1996 season didnot differ from the seasonality of rotavirus disease in thenon-G9 seasons. A single additional P[6] G9 isolate was observedin the 1997-1998 season (Fig. 1A). The P[6] genotype and electropherotypeof this isolate were identical to those of the G9 rotavirusdetected in the 1995-1996 season.

View larger version (39K):
[in this window]
[in a new window]
FIG. 1. (A) Distribution of rotavirus G types during five seasons (1994 to 1999). Values for mixtures and values not determined are not included in these pie graphs. In the 1995-1996 season only, all specimens (i.e., those with or without complete electropherotypes) were also serotyped by RT-PCR. (B) Monthly distribution of rotavirus strains by G type (for the year 1995-1996). Undetermined values were not included in the pie graphs.

Comparison of patients with G9 infections to patients with non-G9 infections. We hypothesized that since the community had no previous exposureor immunity to the G9 strains, children infected with the novelG9 strain might be younger or older or have more severe diseasethan children infected with the historically more common strains.Younger infants might be more susceptible because maternal immunitywould not include G9 specific neutralizing antibodies, and olderchildren might be overrepresented because previous infectionsmight not have induced active immunity to the novel strain.We therefore compared the cumulative frequency distributionsof ages (Fig. 2) and the severity scores of patients with G9infections versus patients with non-G9 infections (Fig. 3).We found an excess of G9 cases among children <12 monthsold that was concentrated in infants <6 months of age andwas statistically significant (P = < 0.01 [Kolmogorov-Smirnovgoodness-of-fit test]) (15). At the other extreme of age, sixof the seven oldest cases of rotavirus disease were in childreninfected with G9 strains, but this difference was not significant.Only 39% of all rotavirus cases fell within the age range from6 to 23 months, which is usually considered to be the age ofpeak susceptibility. The age distribution of cases was comparedfor all children during the five seasons (Fig. 4). Childrenin the 1995-1996 season, when G9 strains were predominant, tendedto be younger, with more children present who were <7 monthsof age, but when all infections (i.e., G9 and non-G9) were comparedto other seasons, the difference was not statistically significant.

View larger version (51K):
[in this window]
[in a new window]
FIG. 2. Age distribution of patients with rotavirus-induced diarrhea infected with G9 versus non-G9 serotypes in the 1995-1996 season.

View larger version (60K):
[in this window]
[in a new window]
FIG. 3. Severity score for four G4 types in the 1995-1996 season.

View larger version (56K):
[in this window]
[in a new window]
FIG. 4. Age distributions for infants from 0 to 24 months old for five seasons.

From our chart review (Fig. 3), we were able to develop severityscores from the records of 148 children hospitalized in the1995-1996 season representing four different serotypes. Theseverity scores of children infected with G1, G3, and G9 strainsdid not differ (mean score, 10.0 to 11.0), but children infectedwith G2 strains had disease that was slightly more severe (meanscore, 13.0) (15). The cumulative frequency curve of severityscores indicated that serotype G2 caused significantly moresevere disease than that caused by the serotype G9 virus. Theseverity of disease among 24 children infected at the beginningof the preceding season and at the beginning of the post-1995-1996season did not differ significantly from those infected in theentire 1995-1996 season.

DISCUSSION

Top

Discussion

This study documents the emergence of the novel G9 serotypeof rotavirus in the US in the 1995-1996 winter season. A newand distinct P[8] G9 strain appeared briefly in Philadelphiain 1983 (9), and different G9 strains were subsequently foundin India and throughout Asia in the late 1980s and early 1990s(12). These strains have now become endemic in the United Statesand have established themselves as the fifth most common serotypeworldwide (13, 17, 18). This emergence of a new serotype hasled to the need to include new diagnostic primers for G9 strainsin the RT-PCR assays to characterize rotaviruses worldwide.

The present study provides some comfort to our considering thepotential efficacy of vaccines against rotavirus currently inclinical trials. We hypothesized that, like influenza, the emergenceof a totally novel rotavirus strain, with both outer capsidproteins different from previously circulating strains, mightlead to disease of greater severity or to breakthrough infectionsamong older children already immune to the other common serotypes.Although we did identify an increased risk of infection in youngerinfants <6 months of age, perhaps due to an absence of maternalneutralizing antibody specific to this novel strain, childreninfected with G9 strains did not have disease of greater severityor duration than children infected with the non-G9 strains.Since initial rotavirus infections with every serotype apparentlyoccur only when maternal antibody has waned beyond protectivecapacity, this observation is not surprising but is reassuring.Nonetheless, six of the seven rotavirus infections among childrenmore than 5 years of age were caused by G9 infections. Our resultsare in contrast to those reported by Cubitt et al. in the UnitedKingdom, who found that children infected with G9 strains tendedto be older and have more severe dehydration than children withnon-G9 strains (5). We cannot explain this discrepancy, butwe did note that the distribution of ages and clinical severityscores of our patients did not differ significantly over fivewinter seasons of observation in which four different G serotypespredominated.

Our results should provide some reassurance about the heterotypicprotection conferred by one rotavirus infection against a secondinfection. Despite the complete antigenic difference in boththe G and P outer capsid proteins of this novel P[6] G9 strain,children infected with this strain, although predisposed toinfection at an earlier age, did not exhibit enhanced severityof their disease. Vaccines currently under development shouldexamine this cross-protection further.

ACKNOWLEDGMENTS

This study was supported by a grant from Merck & Co., Inc.,West Point, Pa.

We thank Philip Bennett of Merck Research Laboratories for assistancewith the statistical analysis.

FOOTNOTES

* Corresponding author. Mailing address: Children's Hospital of Philadelphia, Division of Infectious Diseases, 3516 Civic Center Blvd., Philadelphia, PA 19104. Phone: (215) 590-2044. Fax: (215) 590-2025. E-mail: clarkf{at}email.chop.edu.

REFERENCES

Top