1251
Evidence that Protection against Rotavirus Diarrhea after Natural Infection Is Not Dependent on Serotype-Specific Neutralizing Antibody Richard L. Ward, John D. Clemens, Douglas R. Knowlton, Malia R. Rao, Frederik P. L. van Loon, Nurul Huda, Faruque Ahmed, Gilbert M. Schiff, and David A. Sack
Division ofClinical Virology. J. N. Gamble Institute ofMedical Research. Cincinnati. Ohio; International Centre for Diarrhoeal Disease Research. Bangladesh. Dacca; Divisions of Epidemiology Statistics and Prevention Research. National Institute ofChild Health and Human Development. Bethesda. and Johns Hopkins School ofPublic Health. Baltimore. Maryland
Rotavirus infections are the most important cause ofdehydrating diarrhea in infancy and early childhood worldwide and are a major cause of mortality in developing countries. As a result, the development of vaccines against rotavirus has received a high priority [I]. Because natural rotavirus infection is known to confer protection against subsequent rotavirus disease [2-7] and because efforts to develop live, oral rotavirus vaccines with minimal side-effects have been notably successful [8-13], there has been optimism that a highly protective rotavirus vaccine would be imminently available as a public health tool. The last decade, which witnessed a succession of vaccines developed with the goal of stimulating serotype-specific immunity against rotavirus, has yielded conflicting results. In initial studies with bovine rotavirus vaccine strains, protection occurred in the absence of detectable neutralizing antibody to the circulating rotavirus strains [10, 14], but these vaccines were not consistently efficacious [7, 15-18]. Protection found during multiple trials with RRV, a rhesus rotavirus vaccine, has also been erratic [17, 19-22] and the failure of a nursery strain of human rotavirus to protect was
Received I April 1992; revised 22 July 1992. Informed consent was obtained from the parents or guardians of study subjects. Financial support: US Agency for International Development (contract 282-90-0019). Reprints or correspondence: Dr. Richard L. Ward, J. N. Gamble Institute of Medical Research, 2141 Auburn Ave., Cincinnati, OH 45219. The Journal ofInfectious Diseases 1992;166:1251-7 © 1992 by The University of Chicago. All rightsreserved. 0022-1899/92/6606-0008$01.00
particularly unsettling [23]. The relationship between serotype-specific neutralizing antibody responses to vaccines and protection in these trials has been inconsistent. Similar inconsistencies have been found in animal studies [24-33]. Thus, the basis of infection-induced immunity to rotavirus, particularly the role of neutralizing antibody, needs to be established. A diarrheal surveillance system in rural Bangladesh, in which all four major serotypes of rotavirus were found to cause illnesses in a single population [34], offered a unique opportunity to reexamine the importance of serotype-specificity of naturally induced serum neutralizing antibody for protection against clinically significant rotavirus diarrhea. In an earlier study, we reported that titers of serum rotavirus IgG were protectively correlated with the risk of developing rotavirus diarrhea [35]. In this study we evaluated whether protection was attributable to homologous versus heterologous serotype-specific serum neutralizing antibody titers.
Methods Overview. The relationship between serotype-specific serum immunity and the risk of clinically significant rotavirus diarrhea was evaluated within the framework of a case-control study. Cases were defined as children with rotavirus diarrhea of defined serotype that was detected during comprehensive surveillanceof persons presenting for care of diarrhea in a geographically defined population in rural Bangladesh during 1985-1986. Controls were children who were contemporaneously sampled from the same population in four serologic surveys. Levels of serum antibodies to the four major serotypes of human rotavirus were
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
This case-control study sought to determine whether protection against clinically significant rotavirus diarrhea in children aged 4-35 months correlated with titers of serum neutralizing antibody and, if so, whether this protection was serotype-specific. Titers of acute-phase sera from 156 cases of treated rotavirus diarrhea in rural Bangladesh were contrasted with titers from 312 contemporaneously selected, age-matched controls. Analyses of the culture-adapted rotaviruses from the cases revealed that 24%,15%,43%, and 17% belonged to serotypes 1-4, respectively. Titers of both homologous and heterologous neutralizing antibody in acute blood specimens of cases were significantly lower than those of matched controls. However, multivariate logistic regression models demonstrated that only antibody titers to heterotypic rotaviruseswereindependently associated with protection against rotavirus disease. These data, which indicate that the correlation of protection with neutralizing antibody titers is not serotype-specific, suggest that immunity to rotavirus disease may be mediated by other factors.
1252
Ward et al.
The reciprocal of the serum dilution that reduced the number of viral focus forming units (fib) by 60%was considered its titer. However, to assign titers to sera that had neutralizing antibody titers of < I00, sera that when diluted I: 100 reduced fib by 0-20% or 21%-50%were considered to have titers of25 and 50, respectively. Sera that when diluted I: 100 reduced ffu by > 50% were retitered using twofold serial dilutions starting at I: 100, and neutralizing antibody titers were calculated. Assignments of titers of 25 and 50 were based on the following observations. When neutralizing antibody titers of infant sera obtained in the United States were determined for the four prototype strains of human rota virus, dilutions of sera that reduced viral ffu by 5%20% caused ;;.60% reduction in viral ffu at the next lower dilution in none of 17 specimens tested. When the dilution of sera reduced viral ffu by 21%-30%, however, the next lower dilution reduced viral ffu by ;;.60% in 4 (31 %) of 13 specimens. Similar results were obtained with specimens obtained in Bangladesh. Titers less than the cutoff value are typically assigned titers of one-half of this value, which in this case would have been 50. Therefore, to be conservative in assigning sera titers of25, only sera that reduced viral ffu by ~20% at a dilution of I: 100 were assigned this titer. Laboratory personnel were kept unaware of the study hypotheses and of the case-control status of the tested sera to enable blinded evaluations. Definitions. Diarrhea was defined as 'an illness in which.at least three loose or liquid stools were passed in any 24-h period. Visits for the treatment of diarrhea were grouped into "episodes" if the date of the onset of symptoms before a visit was ~7 days after the date of discharge for the previous visit. An episode of rotavirus diarrhea was defined as a diarrheal episode in which rotavirus was detected in a stool specimen collected during any component visit of the episode, and the "onset" of such an episode referred to the onset of symptoms before the initial visit of the episode. Severe rotavirus diarrhea denoted an episode in which the child had an absent or feeble pulse, together with at least one additional objective sign of dehydration (depressed anterior fontanelle, tenting of skin, sunken eyes, dry mucous membranes). A child was classified as breast-fed if breast milk constituted any portion of the diet just before the onset of the episode for rotavirus cases and on the date of visit for children assembled in the field surveys. The date of selection referred to the date children with rotavirus diarrhea presented for care and to the date of home visits for subjects participating in the field surveys. Age was age on the date of selection. Assembly ofcases. An episode ofrota virus diarrhea was eligible to be a case if each of several criteria, formulated before the conduct of the study, was fulfilled: (I) The episode was associated with recovery ofrotavirus that could be serotyped; (2) the patient resided in one of the Matlab villages included in the cholera vaccine trial area and presented for care between I February 1985 and 30 June 1986; (3) a blood specimen was obtained on the date of presentation for care; (4) Vibrio cholerae 0: I, Shigella and Salmonella organisms, and enterotoxigenic Escherichia coli were not isolated; (5) diarrhea onset was ~3 days before presentation for care; and (6) the patient was 4-35 months old. The selection interval for assembling cases was chosen to coincide with four serologic surveys in Matlab done in conjunction
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
measured at the time that cases presented and on the date of a home visit for controls. With the assumption that the level of antibodies measured in serum collected at the time of presentation for care approximated the level that existed just before the rotavirus infection under study, the association between antibody titers and casecontrol status would refiect the association between levels of antibodies and the subsequent risk of rotavirus diarrhea. Accordingly, a higher titer of target antibodies in controls than in cases would correspond to an inverse (i.e., protective) relationship between the antibodies and the subsequent risk ofrota virus diarrhea. Surveillance for rotavirus diarrhea. The study was in the Matlab field studies area of the International Centre for Diarrhoeal Disease Research, Bangladesh. In conjunction with a field trial of oral cholera vaccines, clinical and microbiologic surveillance of Matlab residents treated for diarrhea was instituted in early 1985 at all three Matlab treatment centers that offer treatment for diarrhea [36-38]. To diagnose rotavirus infections, stools were initially screened with an ELISA, using reagents provided by the World Health Organization [39]. Positive specimens were confirmed by a more sensitive, specific ELISA [40], and cultivatable rotavirus isolates were serotyped, as described in detail elsewhere [34]. A fingerstick blood specimen (0.1 mL, diluted I: lOin saline) was obtained at the time of presentation for care. Serologic surveys. Four surveys were made in the field trial area during the vaccine field trial: April-May 1985 (survey I), August-September 1985 (survey 2), November-December 1985 (survey 3), and March-April 1986 (survey 4). In each survey, clusters (50 for the first and 70 for all others) ofgeographically contiguous families were randomly selected from census records ·of the Matlab Demographic Surveillance System [41]. Each cluster was unique and contained - 300 persons. Within each cluster, 37 persons were targeted for collection ofa O.I-mL fingerstick blood specimen. Blood specimens were diluted I: 10 in the field with sterile saline, placed on ice, and promptly transferred to the Matlab laboratory for separation of serum. Evaluation ofsera. Sera were frozen shortly after collection and stored for subsequent testing. In all laboratory evaluations, sera from different matching strata (see below) were tested in random order to prevent biases from secular trends in the results of assays, and sera from cases and controls within a matching stratum were tested concurrently. Titers of serum neutralizing antibody to prototype rotavirus strains representative of the four major human rotavirus serotypes were determined by a focus reduction neutralization assay as previously described [42]. The prototype strains included Wa, DS-I, P, and VA70, representative of serotypes 1-4, respectively. Titers of serum neutralizing antibody of cases were also determined against the cultureadapted rotavirus obtained from each matched subject. Because blood was diluted I: lOin the field and just over 50% of the volume of blood was recoverable as serum, a sixfold dilution of serum was considered to yield a final dilution of I: 100. Because of the small volumes available, the minimum initial dilution of stored sera used was threefold. This, coupled with a twofold dilution when mixing with virus, meant that the actual minimum dilution used in the study was I: 100.
lID 1992; 166 (December)
JID 1992; 166 (December)
Antibody Titers and Rotavirus Immunity
Table 1. Characteristics of 156rotavirus diarrheal episodesin the case-control study. Characteristic Severity of diarrheal symptoms Severe Nonsevere Diarrheal characteristics Watery Nonwatery Bloody Rotavirus serotype I
2 3 4
No. (%) 59 (38) 97 (62) 146 (94) 9 (6)
1(0.6) 38 (24) 24 (15) 67 (43) 27 (17)
Duration of symptoms (days)*
o I 2 3
33 (21) 73 (47) 33 (21) 17 (II)
*Time between onset ofsymptoms and when treatment was sought.
increasingly negative coefficients reflected progressively greater protective associations between the titers and the risk of rotavirus diarrhea. All statistical tests were interpreted in a twotailed fashion to estimate P values.
Results Descriptionofcases. Ofthe 156 cases that met the criteria for inclusion, 38% had severe diarrhea (table I). Almost all (94%) presented with watery or liquid diarrhea; the stool of I subject contained blood. All four serotypes were represented: The highest percentage. was serotype 3 (43%) and the lowest was serotype 2 (15%). Although subjects were included if the onset of symptoms had occurred up to 3 days before treatment was sought, serum antibody titers were unrelated to duration of symptoms within this 3-day period. Baselinefeatures ofcasesand controls. Cases and controls were similar in age distribution (table 2), although the median age of cases (II. 9 months) was slightly lower than controls ( 13.2 months). The cases also included a higher percentage of boys (P < .00 I), which could have been related to a genuine effect of gender on the risk of rotavirus diarrhea but more likely reflects the differential patterns of parental care of boys and girls previously described in Matlab [44]. In addition, a higher proportion of Muslims (P < .01) and, corresponding to their slightly younger ages, a greater percentage of breast-fed children (P < .05) were also found. Finally, the distribution of selection intervals differed in the two groups (P < .001), primarily because ofa greater proportion ofcases than controls in phase I. The imbalance of cases to controls with respect to age and interval resulted from a scarcity of controls available for selection in some of the strata. These
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
with the cholera vaccine trial. Only subjects with short symptomatic histories were studied to increase the likelihood that serum titers at presentation reflected preinfection titers. The age criteria limited the study to children who were old enough that maternally derived antibodies would have declined to negligible levels as we found in our earlier analyses of serum rotavirus IgG in this population [35], that is, all 6 cases in children aged 0-3 months had detectable serum rotavirus IgG (;;;.100 units/ml.), while only I (3%) of 33 cases in children aged 4-7 months had detectable rotavirus IgO (P < .000 I; Fisher's exact test). Subjects shedding rotavirus with short RNA electropherotypes were included in the group ofserotype 2 cases, since our earlier analyses had shown this electrophoretic pattern to be specific for serotype 2, and the cultivation of serotype 2 strains was somewhat less efficient than that for other serotypes [34]. A total of 341 episodes of rotavirus diarrhea, each in a different person, was found in subjects who presented for care at the treatment centers during the selection interval; 315 were in children 4-35 months old. Of these, 165 met all eligibility criteria for the study. However, an appropriate control could not be found for 9, leaving 156 cases for analysis. Assembly of controls. Controls were assembled from 7256 randomly selected participants who were bled in the four surveys. Controls were individually matched to cases according to age and date of selection. For this purpose, age was distributed within eight categories (4-7,8-11, 12-15, 16-19,20-23,2427, 28-31, and 32-35 months). Date of selection was classified into four phases (I February-3D June 1985; I July-l 5 October 1985; 16 October 1985-31 January 1986; I February-3D June 1986). The four date-of-selection phases corresponded to surveys 1-4, respectively. Up to three controls were randomly selected for each case according to these features. With this strategy, 312 controls were selected for the 156 cases. Analysis. In simple analyses, case-control contrasts of categorical variables were statistically appraised with the x 2 test or with Fisher's exact test when mandated by sparse data. For such odds ratios, 95%confidence intervals were estimated with testbased methods [43]. Contrasts of dimensional variables were evaluated statistically by Student's t test or the Mann-Whitney U test when parametric assumptions were not fulfilled. To fulfill parametric assumptions, serum antibody titers were log-transformed before statistical tests were done. To evaluate associations between serologic titers ofantibodies and the occurrence of rotavirus diarrhea after controlling for potentially confounding variables, we used logistic regression models [43]. These models employed a forward stepwise selection algorithm whereby case-control status was taken as the dependent variable, variables for potentially confounding factors were "forced" as independent variables, and variables for each serotype-specific antibody titer (log-transformed to improve fit) were stepped into the model as additional independent variables, if they were significantly (P < .05) associated with casecontrol status after controlling for other variables already in the model. The coefficient for an independent variable in these models reflected the association between the variable and the occurrence ofrotavirus diarrhea, adjusting for the confounding effects ofother independent variables in the model. For variables corresponding to serotype-specific antibody titers in these models,
1253
1254
Ward et al.
Table 2. Baseline features of cases and controls.
Feature'
Controls (n = 312)
29 (19) 50 (32) 28 (18) 32 (21) 9 (6) 8 (5) 105 (67)t 148 (95)~ 151 (97)1 80 (51) 46 (30)
58 (19) 80 (26) 60 (19) 68 (22) 22 (7) 24 (8) 155(50) 267 (86) 282 (90) 149(48) 101 (32)
52 (33)t 20 (13) 37 (24) 47 (30)
78 (25) 60 (19) 89 (29) 85 (27)
NOTE. Data are no. (%). , Breast-fed and phase of selection are defined in Methods. t p < .00 I, ~ < .0 I, 1 < .05 (two-tailed) for comparison of cases and controls. , Family compound had tube well or latrine.
differences were all accounted for during analyses of the data. Serotype-specific neutralizing antibody titers in cases versus controls. To determine the relationship between the titers of serotype-specific neutralizing antibody and risk of rotavirus diarrhea, comparisons of these titers were made between cases who experienced rotavirus diarrhea and matched controls who had not. Relative to matched controls, cases consistently had lower neutralizing antibody titers to all four serotypes of rota virus tested and, except for cases infected with serotype 2 strains, these differences were all statistically significant (table 3). To further evaluate whether the protective association be-
tween neutralizing antibodies and rota virus disease was primarily homotypic or heterotypic, we used the same data for several types ofmultivariate analyses, using forward stepwise selection algorithms for logistic regression models, as described previously [43]. This stepwise algorithm first selected the variable for the serotype antibody whose titer was most strongly associated with rota virus diarrhea of the cited serotype, and then sequentially selected additional variables for serotype-specific antibodies exhibiting significant associations after controlling for variables already included in the model by earlier selection or by intentional forcing. This statistical strategy attempted to sort out which antibody titers were independently associated with the risk of rotavirus diarrhea, as opposed to being associated simply by virtue ofintercorrelations of levels of different types of serotype-specific antibodies in the same individual. In the first set of models, we evaluated which serotype-specific antibodies were most significantly related to protection against rota virus diarrheal illnesses of each serotype. These models controlled for potentially confounding variables (age, sex, religion, breast-feeding status, and phase of selection), and they evaluated associations for variables corresponding to titers of antibodies directed against each serotype (table 4). Strikingly, only heterotypic antibodies were selected for inclusion in these models. Thus, only these antibodies were independently associated with the risk of rotavirus diarrhea. In a second series of models, we evaluated whether titers of heterotypic antibodies were associated with serotype-specific rotavirus infections, after controlling not only for the above-cited confounding variables but also for homotypic antibody titers. These models demonstrated that above and beyond any association between homotypic antibodies and their corresponding serotype-specific infections, heterotypic antibody titers were protectively related to serotype I (serotype 4 antibodies) and serotype 3 (serotype I antibodies) diarrheal illnesses. A possible explanation for these results is that the prototype strains of rotavirus used in the serologic analyses were
Table 3.
Geometric mean titers (GMTs) of neutralizing antibody to the four major serotypes of human rotavirus in cases versus controls.
NOTE.
Controls are classified by serotype of infecting strain of their matched cases.
, P < .00 I, t < .05, ~ < .01 (two-tailed) for comparison of cases and controls.
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
Age (months) 4-7 8-11 12-15 16-19 20-23 24-35 Male Muslim Breast-fed Tube well' Latrine' Phase of selection I 2 3 4
Cases (n=156)
lID 1992;166 (December)
lID 1992;166 (December)
Antibody Titers and Rotavirus Immunity
1255
Table 4.
Association between serotype-specific antibody and rotavirus diarrheal episodes of each serotype in stepwise logistic regression models.
I
2 3 4
(-) (-)
-4.18 t (-)
-1.26'
-2.79t
(-) (-) (-)
(-) (-) (-)
not representative of those causing infection in Bangladesh, and infected subjects may have had lower titers to the rotaviruses that caused their infections than to the prototype strains. When tested, however, titers of neutralizing antibody to the prototype strains were, in fact, consistently lower in the cases than to the viruses actually causing the illnesses (data not shown). Thus, we found no evidence to suggest that the use of prototype viruses in the serologic assays led to spurious associations.
Discussion Stimulation of protective serotype-specific neutralizing antibody against circulating rotavirus strains has been a major goal of the development of human rotavirus vaccines and has also been used as an index ofsuccessful vaccination. This practice is supported by animal and human studies in which both active and passive immunity to rotavirus appeared to be serotype-specific [5, 25-27, 45, 46]. Others have reported, however, that experimental infections with heterotypic rotaviruses also elicit protective responses [10, 14, 24, 28-32]. Thus, it is unclear whether serotype-specific neutralizing antibody is needed for protection and whether stimulation of these antibodies is a reliable indicator of successful vaccination. Natural rotavirus infection is known to protect against subsequent rotavirus illness [2-7], but the serotype-specificity of this response is unclear. To make this determination, subjects should be exposed to multiple rotavirus serotypes, and titers of neutralizing antibody to these serotypes should be known at the time of exposure. Both conditions were fulfilled in our surveillance study in rural Bangladesh [34], thus providing a unique opportunity to determine the importance of serotype-specific neutralizing antibody in protection against clinically significant rotavirus diarrhea. Subjects who experienced rotavirus diarrhea had significantly lower neutralizing antibody titers against all four ma-
jor serotypes of human rotavirus than matched controls. Thus, previous rotavirus infections that elicited these responses were associated with protection against subsequent rotavirus illness. Neutralizing antibody production is, however, only one component of the immune response to viral infections. Other immunologic components may be responsible for protection, and neutralizing antibody may only serve as an indicator of their presence. In support of this suggestion, it was found that protection associated with titers of neutralizing antibody was not serotype-specific. In fact, neutralizing antibody titers to a heterotypic rotavirus consistently provided the best correlate ofprotection to each of the four circulating serotypes of rotavirus. This does not rule out neutralizing antibody as an important component ofimmunity to rota virus in humans but indicates that other components are also involved. In a study of active immunity to rotavirus infection in a mouse model, we found that experimental neonatal infection with only certain strains of rotavirus resulted in protection against a subsequent murine rotavirus infection [47]. Although the cause of protection was not identified, there was strong evidence that it was unrelated to either serum or intestinal neutralizing antibody. Similarly, a recent study with respiratory syncytial virus, whose pathogenesis also involves a mucosal surface, revealed no evidence of association between protection and neutralizing antibody to the challenge virus [48]. These findings suggest that protection from viral disease at a mucosal surface involves mechanisms other than or in addition to viral neutralization by antibody. Other investigators have suggested that virus-specific cytotoxic T lymphocytes may have a role [49, 50]. Because the cytotoxic T lymphocyte response in mice was found to be heterotypic [51,52], protection by these cells is consistent with the lack of serotype-specific immunity observed in this study after natural infection of humans. Clearly, the mechanisms ofrotavirus immunity need to be better understood in order to develop reliable vaccines.
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
NOTE. In each model, variables for four serotype-specific neutralizing antibodies were evaluated with forward stepwise selection, after controlling for age, sex, religion, breast-feeding status, and phase of selction. Titers were transformed to 10glO base for analysis in models. Negative values, protective associations; (-), variable was not selected for inclusion in model at P < .05. , P < .05, t < .00 I, t < .0 I, (two-tailed) for association between antibody titer and cited group of cases.
1256
Ward et al.
References I. Institute of Medicine. Prospects for immunizing against rota virus. In: New vaccine development. Establishing priorities. Washington, DC: National Academy Press. 1986:D 13-1-12. 2. Ryder RW, Singh N. Reeves WC, Kapikian AZ, Greenberg HB. Sack RB. Evidence of immunity induced by naturally acquired rota virus and Norwalk virus infection on two remote Panamanian islands. J Infect Dis 1985;151:99-105. 3. Black RE. Greenberg HB. Kapikian AZ, Brown KH, Becker S. Acquisition of serum antibody to Norwalk virus and rotavirus in relation to diarrhea in a longitudinal study of young children in rural Bangladesh. J Infect Dis 1982; 145:483-9. 4. Bishop RF, Barnes GL. Cipriani E. Lund JS. Clinical immunity after neonatal rotavirus infection: a prospective longitudinal study in young children. N Engl J Med 1983;309:72-6. 5. Chiba S. Yokoyama T. Nakata S. et al. Protective effect of naturally acquired homotypic and heterotypic rotavirus antibodies. Lancet 1986;2:417-21. 6. Bernstein DI, Sander DS, Smith VE. SchiffGM, Ward RL. Protection from rotavirus reinfection: 2-year prospective study. J Infect Dis 1991; 164:277-83. 7. Bernstein DI, Smith VE. Sander DS. Pax KA, Schiff GM. Ward RL. Evaluation ofWC3 rotavirus vaccine and correlates of protection in healthy infants. J Infect Dis 1990;162: 1055-62. 8. Vesikari T. Isolauri E. Delem A, D'Hondt E, Andre FE, Zissis G. Immunogenicity and safety oflive oral attenuated bovine rota virus vaccine strain RIT 4237 in adults and young children. Lancet 1983;2:807II. 9. Losonsky GA. Rennels MB, Kapikian AZ. et al. Safety. infectivity. transmissibility and immunogenicity of rhesus rota virus vaccine (MMU 18006) in infants. Pediatr Infect Dis J 1986;5:25-9. 10. Clark HF, Borian FE. Bell LM, Modesto K, Gouvea V. Plotkin SA. Protective effect ofWC3 vaccine against rotavirus diarrhea in infants during a predominantly serotype I rota virus season. J Infect Dis 1988; 158:570-87. II. Clark HF. Borian FE. Plotkin SA, Immune protection of infants against
rotavirus gastroenteritis by a serotype I reassortant of bovine rotavirus WC3. J Infect Dis 1990; 161: 1099-104. 12. Perez-Schael I. Blanco M. Vilar M. et al. Clinical studies of a quadrivalent rotavirus vaccine in Venezuelan infants. J Clin Microbiol 1990;28:553-8. 13. Midthun K, Halsey NA. Jett-Goheen M, etal. Safety and immunogenicity of human rotavirus vaccine strain M37 in adults, children. and infants. J Infect Dis 1991;164:792-6. 14. Vesikari T. Isolauri E. D'Hondt E, Delem A, Andre FE, Zissis G. Protection of infants against rotavirus diarrhea by RIT 4237 attenuated bovine rotavirus strain vaccine. Lancet 1984; I:977-81. 15. Hanlon P, Hanlon L. Marsh V, et al. Trial of an attenuated bovine rotavirus vaccine (RIT 4237) in Gambian infants. Lancet 1987; I: 1342-5. 16. Lanata CF. Black RE, del Aguila R, et al. Protection of Peruvian children against rotavirus diarrhea of specific serotypes by one. two. or three doses of the RIT attenuated bovine rotavirus vaccine. J Infect Dis 1989;159:452-9. 17. Santosham M. Letson GW. Wolff' M, et al. A field study of the safety and efficacy of two candidate rotavirus vaccines in a native American population. J Infect Dis 1991;163:483-7. 18. Georges-Courbet MC, Monges J. Siopathis MR. et al. Evaluation of the efficacy of a low-passage bovine rotavirus (strain WC3) vaccine in children in Central Africa. Res ViroI1991;142:405-11. 19. Flores J, Perez-Schael I. Gonzalez M, et al. Protection against severe rotavirus diarrhoea by rhesus rotavirus vaccine in Venezuelan infants. Lancet 1987; I:882-4. 20. Christy C. Madore HP. Pichichero ME, et al. Field trial of rhesus. rotavirus vaccine in infants. Pediatr Infect Dis J 1988;7:645-50. 21. Vesikari T, Rautanen T. Varis T, Beards GM, Kapikian AZ. Rhesus rotavirus candidate vaccine. Am J Dis Child 1990;144:285-9. 22. Rennels MB. Losonsky GA. Young AE. Shindledecker CL. Kapikian AZ. Levine MM. An efficacy trial of the rhesus rotavirus vaccine in Maryland. Am J Dis Child 1990;144:601-4. 23. Vesikari T, Ruuska T. Hannu-Pekka K. Green KY. Flores J, Kapikian AZ. Evaluation of the M37 human rotavirus vaccine in 2- to 6month-old infants. Pediatr Infect Dis J 1991; 10:912-7. 24. Wyatt RG. Mebus CA, Yolken RH. et al. Rotaviral immunity in gnotobiotic calves: heterologous resistance to human virus induced by bovine virus. Science 1979;203:548-50. 25. Gaul SK. Simpson TF, Woode ON. Fulton RW. Antigenic relationships among some animal rotaviruses: virus neutralization in vitro and cross-protection in piglets. J Clin MicrobioI1982;16:495-503. 26. Woode GN. Kelso NE, Simpson TF. Gaul SK, Evans LE, Babiuk L. Antigenic relationships among some bovine rotaviruses: serum neutralization and cross-protection in gnotobiotic calves. J Clin MicrobioI 1983; 18:358-64. 27. Bohl EH, Theil KW. Saif LJ. Isolation and serotyping of porcine rotaviruses and antigenic comparison with other rotaviruses. J Clin Microbiol 1984; 19: 105-1 I. 28. Bishop RF, Tzipori SR. Coulson BS, Unicomb LE. Albert MJ. Barnes GL. Heterologous protection against rotavirus-induced disease in gnotobiotic piglets. J Clin Microbiol 1986;24: 1023-8. 29. Torres A, Lin JH. Diarrheal response of gnotobiotic pigs after fetal infection and neonatal challenge with homologous and heterologous human rota virus strains. J Virol 1986;60: 1107-12. 30. Hoshino Y, Torres A, Aves TM. et al. Neutralizing antibody response of gnotobiotic piglets after fetal infection and neonatal challenge with human rotavirus strains [letter). J Infect Dis 1987;156:103840. 31. Woode GN. Zheng S, Rosen BI. Knight N, Kelso-Gourley NE. Ramig RF. Protection between different serotypes of bovine rotavirus in gnotobiotic calves: specificity of serum antibody and coproantibody responses. J Clin Microbiol 1987;25: 1052-8.
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
The method used to measure neutralizing antibody titers to different serotypes of human rotavirus in this study was a focus reduction neutralization assay. Others have quantified antibody titers to specific epitopes on the VP7 and VP4 proteins by an epitope-blocking assay (EBA) using monoclonal antibodies to these proteins [53-57]. In some, but not all, subjects the ability of the EBA to detect antibody rises after vaccination or natural rota virus infection was comparable to that of the neutralization assay. Because the EBA is much less laborious, it could be the assay of choice for studies such as that reported here. The major limitation of this procedure, however, is that it detects antibody only to specific epitopes on the prototype strains used in the assay procedure. If the rotaviruses that elicit the antibody responses to be tested either lack or have little antibody response to these specific epitopes, the EBA will not provide a valid measure ofneutralizing antibody induction by these strains. Therefore, unless a series of monoclonal antibodies to each serotype is available and used in the EBA, the neutralization assay will remain the more reliable method to measure overall neutralizing antibody titers to specific rotavirus serotypes.
JID 1992; 166 (December)
JID 1992;166 (December)
Antibody Titers and Rotavirus Immunity
46. Green KY, Kapikian AZ. Identification ofVP7 epitopes associatedwith protection against human rotavirus illness or shedding in volunteers. J Virol 1992;66:548-53. 47. Ward RL, McNeal MM, Sheridan JF. Evidence that active protection following oral immunization of mice with live rotavirus is not dependent on neutralizing antibody. Virology 1992; 188:57-66. 48. Trudel M, Nadon F, Seguin C, Binz H. Protection ofBALB/c mice from respiratory syncytial virus infection by immunization with a synthetic peptide derived from the G glycoprotein. Virology 1991;185: 749-57. 49. Offit PA, Dudzik KI. Rotavirus-specific cytotoxic T lymphocytes passively protect against gastroenteritis in suckling mice. J Virol 1990;64:6325-8. 50. Dharakul T, Rott L, Greenberg HB. Recovery from chronic rotavirus infection in mice with severe combined immunodeficiency: virus clearance mediated by adoptive transfer of immune CD8+ T lymphocytes. J Virol 1990;64:4375-82. 51. Offit PA, Dudzik KI. Rotavirus-specific cytotoxic T lymphocytes crossreact with target cells infected with different rotavirus serotypes. J ViroI1988;63:3507-12. 52. Offit PA, Boyle DB, Both GW, et al. Outer capsid glycoprotein VP7 is recognized by cross-reactive, rotavirus-specific, cytotoxic T lymphocytes. Virology 1991;184:563-8. 53. Shaw RD, Fong KJ, Losonsky GA, et al. Epitope-specific immune responses to rotavirus vaccination. Gastroenterology 1987;93:941-50. 54'. Green KY, Taniguchi K, Mackow ER, Kapikian AZ. Homotypic and heterotypic epitope-specific antibody responses in adult and infant rota virus vaccines: implication for vaccine development. J Infect Dis 1990;161:667-79. 55. Beards GM, Desselberger U. Determination of rotavirus serotype-specific antibodies in sera by competitive enhanced enzyme immunoassay. J Virol Methods 1989;24: 103-10. 56. Taniguchi K, Urasawa T, Kobumichi N, et al. Antibody response to serotype-specific and cross-reactive neutralization epitopes on VP4 and VP7 after rotavirus infection or vaccination. J Clin Microbiol 1991;29:483-7. 57. Matson DO, O'Ryan ML, Pickering LK, et al. Characterization of serum antibody responses to natural rota virus infection in children by VP7-specific epitope-blocking assays. J Clin Microbiol 1992;30: 1056-61.
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
32. Bridger JC, Oldham G. Avirulent rotavirus infection protects calves from disease with and without inducing high levels of neutralizing antibody. J Gen ViroI1987;68:2311-7. 33. Hoshino Y, Saif Ll, Sereno MM, Chanock RM, Kapikian AZ. Infection immunity of piglets to either VP3 or VP7 outer capsid protein confers resistance to challenge with a virulent rotavirus bearing the corresponding antigen. J Virol 1988;62:744-8. 34. Ward RL, Clemens JD, Sack DA, et al. Culture adaptation and characterization of group A rotaviruses causing diarrheal illnesses in Bangladesh from 1985 to 1986. J Clin MicrobioI1991;29:19l5-23. 35. Clemens JD, Ward RL, Rao MR, et al. Serologic evaluation of anti bodies to rotavirus as correlates of the risk of clinically significant rotavirus diarrhea in rural Bangladesh. J Infect Dis 1992;165: 161-5. 36. Clemens J, Sack D, Harris J, et al. Field trial of oral cholera vaccines in Bangladesh. Lancet 1986;2:124-7. 37. Clemens J, Harris J, Sack D, et al. Field trial of oral cholera vaccines in Bangladesh: results of one year of follow-up. J Infect Dis 1988; 158: 60-9. 38. Clemens J, Sack D, Harris J, et al. Field trial of oral cholera vaccines in Bangladesh: results of three-year follow-up. Lancet 1990;335:270-3. 39. Beards GM, Campbell AD, Cottrell NR, et al. Enzyme-linked immunosorbent assays based on polyclonal and monoclonal antibodies for rotavirus detection. J Clin Microbiol 1984; 19:248-54. 40. Ward RL, Bernstein or, Young EC, Sherwood JR, Knowlton DR, SchiffGM. Human rotavirus studies in volunteers: determination of infectious dose and serological response to infection. J Infect Dis 1986; 154:871-80. 41. Ruzicka LT, Chowdhury AKMA. Demographic surveillance system. Matlab. Vol 2. Census 1974. Dacca: Cholera Research Laboratory, 1978 (Scientific report no. 10). 42. Bernstein or, McNeal MM, SchiffGM, Ward RL. Induction and persistence oflocal rotavirus antibodies in relation to serum antibodies. J Med Virol 1989;28:90-5. 43. Kleinbaum D, Kupper L, Morgenstern H. Epidemiologic research. Principles and quantitative methods. Belmont, MA: Lifetime Learning Publications, 1983. 44. Chen LC, Huq E, D'Souza S. Sex bias in the family allocation of food and health care in rural Bangladesh. Popul Devel Rev 1981;7:54-70. 45. Perez-Schael I, Garcia D, Gonzalez M, et al. Prospective study of diarrheal diseases is Venezuelan children to evaluate the efficacy of rhesus rotavirus vaccine. J Med ViroI1990;30:219-29.
1257
Evidence that Protection against Rotavirus Diarrhea after Natural Infection Is Not Dependent on Serotype-Specific Neutralizing Antibody Richard L. Ward, John D. Clemens, Douglas R. Knowlton, Malia R. Rao, Frederik P. L. van Loon, Nurul Huda, Faruque Ahmed, Gilbert M. Schiff, and David A. Sack
Division ofClinical Virology. J. N. Gamble Institute ofMedical Research. Cincinnati. Ohio; International Centre for Diarrhoeal Disease Research. Bangladesh. Dacca; Divisions of Epidemiology Statistics and Prevention Research. National Institute ofChild Health and Human Development. Bethesda. and Johns Hopkins School ofPublic Health. Baltimore. Maryland
Rotavirus infections are the most important cause ofdehydrating diarrhea in infancy and early childhood worldwide and are a major cause of mortality in developing countries. As a result, the development of vaccines against rotavirus has received a high priority [I]. Because natural rotavirus infection is known to confer protection against subsequent rotavirus disease [2-7] and because efforts to develop live, oral rotavirus vaccines with minimal side-effects have been notably successful [8-13], there has been optimism that a highly protective rotavirus vaccine would be imminently available as a public health tool. The last decade, which witnessed a succession of vaccines developed with the goal of stimulating serotype-specific immunity against rotavirus, has yielded conflicting results. In initial studies with bovine rotavirus vaccine strains, protection occurred in the absence of detectable neutralizing antibody to the circulating rotavirus strains [10, 14], but these vaccines were not consistently efficacious [7, 15-18]. Protection found during multiple trials with RRV, a rhesus rotavirus vaccine, has also been erratic [17, 19-22] and the failure of a nursery strain of human rotavirus to protect was
Received I April 1992; revised 22 July 1992. Informed consent was obtained from the parents or guardians of study subjects. Financial support: US Agency for International Development (contract 282-90-0019). Reprints or correspondence: Dr. Richard L. Ward, J. N. Gamble Institute of Medical Research, 2141 Auburn Ave., Cincinnati, OH 45219. The Journal ofInfectious Diseases 1992;166:1251-7 © 1992 by The University of Chicago. All rightsreserved. 0022-1899/92/6606-0008$01.00
particularly unsettling [23]. The relationship between serotype-specific neutralizing antibody responses to vaccines and protection in these trials has been inconsistent. Similar inconsistencies have been found in animal studies [24-33]. Thus, the basis of infection-induced immunity to rotavirus, particularly the role of neutralizing antibody, needs to be established. A diarrheal surveillance system in rural Bangladesh, in which all four major serotypes of rotavirus were found to cause illnesses in a single population [34], offered a unique opportunity to reexamine the importance of serotype-specificity of naturally induced serum neutralizing antibody for protection against clinically significant rotavirus diarrhea. In an earlier study, we reported that titers of serum rotavirus IgG were protectively correlated with the risk of developing rotavirus diarrhea [35]. In this study we evaluated whether protection was attributable to homologous versus heterologous serotype-specific serum neutralizing antibody titers.
Methods Overview. The relationship between serotype-specific serum immunity and the risk of clinically significant rotavirus diarrhea was evaluated within the framework of a case-control study. Cases were defined as children with rotavirus diarrhea of defined serotype that was detected during comprehensive surveillanceof persons presenting for care of diarrhea in a geographically defined population in rural Bangladesh during 1985-1986. Controls were children who were contemporaneously sampled from the same population in four serologic surveys. Levels of serum antibodies to the four major serotypes of human rotavirus were
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
This case-control study sought to determine whether protection against clinically significant rotavirus diarrhea in children aged 4-35 months correlated with titers of serum neutralizing antibody and, if so, whether this protection was serotype-specific. Titers of acute-phase sera from 156 cases of treated rotavirus diarrhea in rural Bangladesh were contrasted with titers from 312 contemporaneously selected, age-matched controls. Analyses of the culture-adapted rotaviruses from the cases revealed that 24%,15%,43%, and 17% belonged to serotypes 1-4, respectively. Titers of both homologous and heterologous neutralizing antibody in acute blood specimens of cases were significantly lower than those of matched controls. However, multivariate logistic regression models demonstrated that only antibody titers to heterotypic rotaviruseswereindependently associated with protection against rotavirus disease. These data, which indicate that the correlation of protection with neutralizing antibody titers is not serotype-specific, suggest that immunity to rotavirus disease may be mediated by other factors.
1252
Ward et al.
The reciprocal of the serum dilution that reduced the number of viral focus forming units (fib) by 60%was considered its titer. However, to assign titers to sera that had neutralizing antibody titers of < I00, sera that when diluted I: 100 reduced fib by 0-20% or 21%-50%were considered to have titers of25 and 50, respectively. Sera that when diluted I: 100 reduced ffu by > 50% were retitered using twofold serial dilutions starting at I: 100, and neutralizing antibody titers were calculated. Assignments of titers of 25 and 50 were based on the following observations. When neutralizing antibody titers of infant sera obtained in the United States were determined for the four prototype strains of human rota virus, dilutions of sera that reduced viral ffu by 5%20% caused ;;.60% reduction in viral ffu at the next lower dilution in none of 17 specimens tested. When the dilution of sera reduced viral ffu by 21%-30%, however, the next lower dilution reduced viral ffu by ;;.60% in 4 (31 %) of 13 specimens. Similar results were obtained with specimens obtained in Bangladesh. Titers less than the cutoff value are typically assigned titers of one-half of this value, which in this case would have been 50. Therefore, to be conservative in assigning sera titers of25, only sera that reduced viral ffu by ~20% at a dilution of I: 100 were assigned this titer. Laboratory personnel were kept unaware of the study hypotheses and of the case-control status of the tested sera to enable blinded evaluations. Definitions. Diarrhea was defined as 'an illness in which.at least three loose or liquid stools were passed in any 24-h period. Visits for the treatment of diarrhea were grouped into "episodes" if the date of the onset of symptoms before a visit was ~7 days after the date of discharge for the previous visit. An episode of rotavirus diarrhea was defined as a diarrheal episode in which rotavirus was detected in a stool specimen collected during any component visit of the episode, and the "onset" of such an episode referred to the onset of symptoms before the initial visit of the episode. Severe rotavirus diarrhea denoted an episode in which the child had an absent or feeble pulse, together with at least one additional objective sign of dehydration (depressed anterior fontanelle, tenting of skin, sunken eyes, dry mucous membranes). A child was classified as breast-fed if breast milk constituted any portion of the diet just before the onset of the episode for rotavirus cases and on the date of visit for children assembled in the field surveys. The date of selection referred to the date children with rotavirus diarrhea presented for care and to the date of home visits for subjects participating in the field surveys. Age was age on the date of selection. Assembly ofcases. An episode ofrota virus diarrhea was eligible to be a case if each of several criteria, formulated before the conduct of the study, was fulfilled: (I) The episode was associated with recovery ofrotavirus that could be serotyped; (2) the patient resided in one of the Matlab villages included in the cholera vaccine trial area and presented for care between I February 1985 and 30 June 1986; (3) a blood specimen was obtained on the date of presentation for care; (4) Vibrio cholerae 0: I, Shigella and Salmonella organisms, and enterotoxigenic Escherichia coli were not isolated; (5) diarrhea onset was ~3 days before presentation for care; and (6) the patient was 4-35 months old. The selection interval for assembling cases was chosen to coincide with four serologic surveys in Matlab done in conjunction
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
measured at the time that cases presented and on the date of a home visit for controls. With the assumption that the level of antibodies measured in serum collected at the time of presentation for care approximated the level that existed just before the rotavirus infection under study, the association between antibody titers and casecontrol status would refiect the association between levels of antibodies and the subsequent risk of rotavirus diarrhea. Accordingly, a higher titer of target antibodies in controls than in cases would correspond to an inverse (i.e., protective) relationship between the antibodies and the subsequent risk ofrota virus diarrhea. Surveillance for rotavirus diarrhea. The study was in the Matlab field studies area of the International Centre for Diarrhoeal Disease Research, Bangladesh. In conjunction with a field trial of oral cholera vaccines, clinical and microbiologic surveillance of Matlab residents treated for diarrhea was instituted in early 1985 at all three Matlab treatment centers that offer treatment for diarrhea [36-38]. To diagnose rotavirus infections, stools were initially screened with an ELISA, using reagents provided by the World Health Organization [39]. Positive specimens were confirmed by a more sensitive, specific ELISA [40], and cultivatable rotavirus isolates were serotyped, as described in detail elsewhere [34]. A fingerstick blood specimen (0.1 mL, diluted I: lOin saline) was obtained at the time of presentation for care. Serologic surveys. Four surveys were made in the field trial area during the vaccine field trial: April-May 1985 (survey I), August-September 1985 (survey 2), November-December 1985 (survey 3), and March-April 1986 (survey 4). In each survey, clusters (50 for the first and 70 for all others) ofgeographically contiguous families were randomly selected from census records ·of the Matlab Demographic Surveillance System [41]. Each cluster was unique and contained - 300 persons. Within each cluster, 37 persons were targeted for collection ofa O.I-mL fingerstick blood specimen. Blood specimens were diluted I: 10 in the field with sterile saline, placed on ice, and promptly transferred to the Matlab laboratory for separation of serum. Evaluation ofsera. Sera were frozen shortly after collection and stored for subsequent testing. In all laboratory evaluations, sera from different matching strata (see below) were tested in random order to prevent biases from secular trends in the results of assays, and sera from cases and controls within a matching stratum were tested concurrently. Titers of serum neutralizing antibody to prototype rotavirus strains representative of the four major human rotavirus serotypes were determined by a focus reduction neutralization assay as previously described [42]. The prototype strains included Wa, DS-I, P, and VA70, representative of serotypes 1-4, respectively. Titers of serum neutralizing antibody of cases were also determined against the cultureadapted rotavirus obtained from each matched subject. Because blood was diluted I: lOin the field and just over 50% of the volume of blood was recoverable as serum, a sixfold dilution of serum was considered to yield a final dilution of I: 100. Because of the small volumes available, the minimum initial dilution of stored sera used was threefold. This, coupled with a twofold dilution when mixing with virus, meant that the actual minimum dilution used in the study was I: 100.
lID 1992; 166 (December)
JID 1992; 166 (December)
Antibody Titers and Rotavirus Immunity
Table 1. Characteristics of 156rotavirus diarrheal episodesin the case-control study. Characteristic Severity of diarrheal symptoms Severe Nonsevere Diarrheal characteristics Watery Nonwatery Bloody Rotavirus serotype I
2 3 4
No. (%) 59 (38) 97 (62) 146 (94) 9 (6)
1(0.6) 38 (24) 24 (15) 67 (43) 27 (17)
Duration of symptoms (days)*
o I 2 3
33 (21) 73 (47) 33 (21) 17 (II)
*Time between onset ofsymptoms and when treatment was sought.
increasingly negative coefficients reflected progressively greater protective associations between the titers and the risk of rotavirus diarrhea. All statistical tests were interpreted in a twotailed fashion to estimate P values.
Results Descriptionofcases. Ofthe 156 cases that met the criteria for inclusion, 38% had severe diarrhea (table I). Almost all (94%) presented with watery or liquid diarrhea; the stool of I subject contained blood. All four serotypes were represented: The highest percentage. was serotype 3 (43%) and the lowest was serotype 2 (15%). Although subjects were included if the onset of symptoms had occurred up to 3 days before treatment was sought, serum antibody titers were unrelated to duration of symptoms within this 3-day period. Baselinefeatures ofcasesand controls. Cases and controls were similar in age distribution (table 2), although the median age of cases (II. 9 months) was slightly lower than controls ( 13.2 months). The cases also included a higher percentage of boys (P < .00 I), which could have been related to a genuine effect of gender on the risk of rotavirus diarrhea but more likely reflects the differential patterns of parental care of boys and girls previously described in Matlab [44]. In addition, a higher proportion of Muslims (P < .01) and, corresponding to their slightly younger ages, a greater percentage of breast-fed children (P < .05) were also found. Finally, the distribution of selection intervals differed in the two groups (P < .001), primarily because ofa greater proportion ofcases than controls in phase I. The imbalance of cases to controls with respect to age and interval resulted from a scarcity of controls available for selection in some of the strata. These
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
with the cholera vaccine trial. Only subjects with short symptomatic histories were studied to increase the likelihood that serum titers at presentation reflected preinfection titers. The age criteria limited the study to children who were old enough that maternally derived antibodies would have declined to negligible levels as we found in our earlier analyses of serum rotavirus IgG in this population [35], that is, all 6 cases in children aged 0-3 months had detectable serum rotavirus IgG (;;;.100 units/ml.), while only I (3%) of 33 cases in children aged 4-7 months had detectable rotavirus IgO (P < .000 I; Fisher's exact test). Subjects shedding rotavirus with short RNA electropherotypes were included in the group ofserotype 2 cases, since our earlier analyses had shown this electrophoretic pattern to be specific for serotype 2, and the cultivation of serotype 2 strains was somewhat less efficient than that for other serotypes [34]. A total of 341 episodes of rotavirus diarrhea, each in a different person, was found in subjects who presented for care at the treatment centers during the selection interval; 315 were in children 4-35 months old. Of these, 165 met all eligibility criteria for the study. However, an appropriate control could not be found for 9, leaving 156 cases for analysis. Assembly of controls. Controls were assembled from 7256 randomly selected participants who were bled in the four surveys. Controls were individually matched to cases according to age and date of selection. For this purpose, age was distributed within eight categories (4-7,8-11, 12-15, 16-19,20-23,2427, 28-31, and 32-35 months). Date of selection was classified into four phases (I February-3D June 1985; I July-l 5 October 1985; 16 October 1985-31 January 1986; I February-3D June 1986). The four date-of-selection phases corresponded to surveys 1-4, respectively. Up to three controls were randomly selected for each case according to these features. With this strategy, 312 controls were selected for the 156 cases. Analysis. In simple analyses, case-control contrasts of categorical variables were statistically appraised with the x 2 test or with Fisher's exact test when mandated by sparse data. For such odds ratios, 95%confidence intervals were estimated with testbased methods [43]. Contrasts of dimensional variables were evaluated statistically by Student's t test or the Mann-Whitney U test when parametric assumptions were not fulfilled. To fulfill parametric assumptions, serum antibody titers were log-transformed before statistical tests were done. To evaluate associations between serologic titers ofantibodies and the occurrence of rotavirus diarrhea after controlling for potentially confounding variables, we used logistic regression models [43]. These models employed a forward stepwise selection algorithm whereby case-control status was taken as the dependent variable, variables for potentially confounding factors were "forced" as independent variables, and variables for each serotype-specific antibody titer (log-transformed to improve fit) were stepped into the model as additional independent variables, if they were significantly (P < .05) associated with casecontrol status after controlling for other variables already in the model. The coefficient for an independent variable in these models reflected the association between the variable and the occurrence ofrotavirus diarrhea, adjusting for the confounding effects ofother independent variables in the model. For variables corresponding to serotype-specific antibody titers in these models,
1253
1254
Ward et al.
Table 2. Baseline features of cases and controls.
Feature'
Controls (n = 312)
29 (19) 50 (32) 28 (18) 32 (21) 9 (6) 8 (5) 105 (67)t 148 (95)~ 151 (97)1 80 (51) 46 (30)
58 (19) 80 (26) 60 (19) 68 (22) 22 (7) 24 (8) 155(50) 267 (86) 282 (90) 149(48) 101 (32)
52 (33)t 20 (13) 37 (24) 47 (30)
78 (25) 60 (19) 89 (29) 85 (27)
NOTE. Data are no. (%). , Breast-fed and phase of selection are defined in Methods. t p < .00 I, ~ < .0 I, 1 < .05 (two-tailed) for comparison of cases and controls. , Family compound had tube well or latrine.
differences were all accounted for during analyses of the data. Serotype-specific neutralizing antibody titers in cases versus controls. To determine the relationship between the titers of serotype-specific neutralizing antibody and risk of rotavirus diarrhea, comparisons of these titers were made between cases who experienced rotavirus diarrhea and matched controls who had not. Relative to matched controls, cases consistently had lower neutralizing antibody titers to all four serotypes of rota virus tested and, except for cases infected with serotype 2 strains, these differences were all statistically significant (table 3). To further evaluate whether the protective association be-
tween neutralizing antibodies and rota virus disease was primarily homotypic or heterotypic, we used the same data for several types ofmultivariate analyses, using forward stepwise selection algorithms for logistic regression models, as described previously [43]. This stepwise algorithm first selected the variable for the serotype antibody whose titer was most strongly associated with rota virus diarrhea of the cited serotype, and then sequentially selected additional variables for serotype-specific antibodies exhibiting significant associations after controlling for variables already included in the model by earlier selection or by intentional forcing. This statistical strategy attempted to sort out which antibody titers were independently associated with the risk of rotavirus diarrhea, as opposed to being associated simply by virtue ofintercorrelations of levels of different types of serotype-specific antibodies in the same individual. In the first set of models, we evaluated which serotype-specific antibodies were most significantly related to protection against rota virus diarrheal illnesses of each serotype. These models controlled for potentially confounding variables (age, sex, religion, breast-feeding status, and phase of selection), and they evaluated associations for variables corresponding to titers of antibodies directed against each serotype (table 4). Strikingly, only heterotypic antibodies were selected for inclusion in these models. Thus, only these antibodies were independently associated with the risk of rotavirus diarrhea. In a second series of models, we evaluated whether titers of heterotypic antibodies were associated with serotype-specific rotavirus infections, after controlling not only for the above-cited confounding variables but also for homotypic antibody titers. These models demonstrated that above and beyond any association between homotypic antibodies and their corresponding serotype-specific infections, heterotypic antibody titers were protectively related to serotype I (serotype 4 antibodies) and serotype 3 (serotype I antibodies) diarrheal illnesses. A possible explanation for these results is that the prototype strains of rotavirus used in the serologic analyses were
Table 3.
Geometric mean titers (GMTs) of neutralizing antibody to the four major serotypes of human rotavirus in cases versus controls.
NOTE.
Controls are classified by serotype of infecting strain of their matched cases.
, P < .00 I, t < .05, ~ < .01 (two-tailed) for comparison of cases and controls.
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
Age (months) 4-7 8-11 12-15 16-19 20-23 24-35 Male Muslim Breast-fed Tube well' Latrine' Phase of selection I 2 3 4
Cases (n=156)
lID 1992;166 (December)
lID 1992;166 (December)
Antibody Titers and Rotavirus Immunity
1255
Table 4.
Association between serotype-specific antibody and rotavirus diarrheal episodes of each serotype in stepwise logistic regression models.
I
2 3 4
(-) (-)
-4.18 t (-)
-1.26'
-2.79t
(-) (-) (-)
(-) (-) (-)
not representative of those causing infection in Bangladesh, and infected subjects may have had lower titers to the rotaviruses that caused their infections than to the prototype strains. When tested, however, titers of neutralizing antibody to the prototype strains were, in fact, consistently lower in the cases than to the viruses actually causing the illnesses (data not shown). Thus, we found no evidence to suggest that the use of prototype viruses in the serologic assays led to spurious associations.
Discussion Stimulation of protective serotype-specific neutralizing antibody against circulating rotavirus strains has been a major goal of the development of human rotavirus vaccines and has also been used as an index ofsuccessful vaccination. This practice is supported by animal and human studies in which both active and passive immunity to rotavirus appeared to be serotype-specific [5, 25-27, 45, 46]. Others have reported, however, that experimental infections with heterotypic rotaviruses also elicit protective responses [10, 14, 24, 28-32]. Thus, it is unclear whether serotype-specific neutralizing antibody is needed for protection and whether stimulation of these antibodies is a reliable indicator of successful vaccination. Natural rotavirus infection is known to protect against subsequent rotavirus illness [2-7], but the serotype-specificity of this response is unclear. To make this determination, subjects should be exposed to multiple rotavirus serotypes, and titers of neutralizing antibody to these serotypes should be known at the time of exposure. Both conditions were fulfilled in our surveillance study in rural Bangladesh [34], thus providing a unique opportunity to determine the importance of serotype-specific neutralizing antibody in protection against clinically significant rotavirus diarrhea. Subjects who experienced rotavirus diarrhea had significantly lower neutralizing antibody titers against all four ma-
jor serotypes of human rotavirus than matched controls. Thus, previous rotavirus infections that elicited these responses were associated with protection against subsequent rotavirus illness. Neutralizing antibody production is, however, only one component of the immune response to viral infections. Other immunologic components may be responsible for protection, and neutralizing antibody may only serve as an indicator of their presence. In support of this suggestion, it was found that protection associated with titers of neutralizing antibody was not serotype-specific. In fact, neutralizing antibody titers to a heterotypic rotavirus consistently provided the best correlate ofprotection to each of the four circulating serotypes of rotavirus. This does not rule out neutralizing antibody as an important component ofimmunity to rota virus in humans but indicates that other components are also involved. In a study of active immunity to rotavirus infection in a mouse model, we found that experimental neonatal infection with only certain strains of rotavirus resulted in protection against a subsequent murine rotavirus infection [47]. Although the cause of protection was not identified, there was strong evidence that it was unrelated to either serum or intestinal neutralizing antibody. Similarly, a recent study with respiratory syncytial virus, whose pathogenesis also involves a mucosal surface, revealed no evidence of association between protection and neutralizing antibody to the challenge virus [48]. These findings suggest that protection from viral disease at a mucosal surface involves mechanisms other than or in addition to viral neutralization by antibody. Other investigators have suggested that virus-specific cytotoxic T lymphocytes may have a role [49, 50]. Because the cytotoxic T lymphocyte response in mice was found to be heterotypic [51,52], protection by these cells is consistent with the lack of serotype-specific immunity observed in this study after natural infection of humans. Clearly, the mechanisms ofrotavirus immunity need to be better understood in order to develop reliable vaccines.
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
NOTE. In each model, variables for four serotype-specific neutralizing antibodies were evaluated with forward stepwise selection, after controlling for age, sex, religion, breast-feeding status, and phase of selction. Titers were transformed to 10glO base for analysis in models. Negative values, protective associations; (-), variable was not selected for inclusion in model at P < .05. , P < .05, t < .00 I, t < .0 I, (two-tailed) for association between antibody titer and cited group of cases.
1256
Ward et al.
References I. Institute of Medicine. Prospects for immunizing against rota virus. In: New vaccine development. Establishing priorities. Washington, DC: National Academy Press. 1986:D 13-1-12. 2. Ryder RW, Singh N. Reeves WC, Kapikian AZ, Greenberg HB. Sack RB. Evidence of immunity induced by naturally acquired rota virus and Norwalk virus infection on two remote Panamanian islands. J Infect Dis 1985;151:99-105. 3. Black RE. Greenberg HB. Kapikian AZ, Brown KH, Becker S. Acquisition of serum antibody to Norwalk virus and rotavirus in relation to diarrhea in a longitudinal study of young children in rural Bangladesh. J Infect Dis 1982; 145:483-9. 4. Bishop RF, Barnes GL. Cipriani E. Lund JS. Clinical immunity after neonatal rotavirus infection: a prospective longitudinal study in young children. N Engl J Med 1983;309:72-6. 5. Chiba S. Yokoyama T. Nakata S. et al. Protective effect of naturally acquired homotypic and heterotypic rotavirus antibodies. Lancet 1986;2:417-21. 6. Bernstein DI, Sander DS, Smith VE. SchiffGM, Ward RL. Protection from rotavirus reinfection: 2-year prospective study. J Infect Dis 1991; 164:277-83. 7. Bernstein DI, Smith VE. Sander DS. Pax KA, Schiff GM. Ward RL. Evaluation ofWC3 rotavirus vaccine and correlates of protection in healthy infants. J Infect Dis 1990;162: 1055-62. 8. Vesikari T. Isolauri E. Delem A, D'Hondt E, Andre FE, Zissis G. Immunogenicity and safety oflive oral attenuated bovine rota virus vaccine strain RIT 4237 in adults and young children. Lancet 1983;2:807II. 9. Losonsky GA. Rennels MB, Kapikian AZ. et al. Safety. infectivity. transmissibility and immunogenicity of rhesus rota virus vaccine (MMU 18006) in infants. Pediatr Infect Dis J 1986;5:25-9. 10. Clark HF, Borian FE. Bell LM, Modesto K, Gouvea V. Plotkin SA. Protective effect ofWC3 vaccine against rotavirus diarrhea in infants during a predominantly serotype I rota virus season. J Infect Dis 1988; 158:570-87. II. Clark HF. Borian FE. Plotkin SA, Immune protection of infants against
rotavirus gastroenteritis by a serotype I reassortant of bovine rotavirus WC3. J Infect Dis 1990; 161: 1099-104. 12. Perez-Schael I. Blanco M. Vilar M. et al. Clinical studies of a quadrivalent rotavirus vaccine in Venezuelan infants. J Clin Microbiol 1990;28:553-8. 13. Midthun K, Halsey NA. Jett-Goheen M, etal. Safety and immunogenicity of human rotavirus vaccine strain M37 in adults, children. and infants. J Infect Dis 1991;164:792-6. 14. Vesikari T. Isolauri E. D'Hondt E, Delem A, Andre FE, Zissis G. Protection of infants against rotavirus diarrhea by RIT 4237 attenuated bovine rotavirus strain vaccine. Lancet 1984; I:977-81. 15. Hanlon P, Hanlon L. Marsh V, et al. Trial of an attenuated bovine rotavirus vaccine (RIT 4237) in Gambian infants. Lancet 1987; I: 1342-5. 16. Lanata CF. Black RE, del Aguila R, et al. Protection of Peruvian children against rotavirus diarrhea of specific serotypes by one. two. or three doses of the RIT attenuated bovine rotavirus vaccine. J Infect Dis 1989;159:452-9. 17. Santosham M. Letson GW. Wolff' M, et al. A field study of the safety and efficacy of two candidate rotavirus vaccines in a native American population. J Infect Dis 1991;163:483-7. 18. Georges-Courbet MC, Monges J. Siopathis MR. et al. Evaluation of the efficacy of a low-passage bovine rotavirus (strain WC3) vaccine in children in Central Africa. Res ViroI1991;142:405-11. 19. Flores J, Perez-Schael I. Gonzalez M, et al. Protection against severe rotavirus diarrhoea by rhesus rotavirus vaccine in Venezuelan infants. Lancet 1987; I:882-4. 20. Christy C. Madore HP. Pichichero ME, et al. Field trial of rhesus. rotavirus vaccine in infants. Pediatr Infect Dis J 1988;7:645-50. 21. Vesikari T, Rautanen T. Varis T, Beards GM, Kapikian AZ. Rhesus rotavirus candidate vaccine. Am J Dis Child 1990;144:285-9. 22. Rennels MB. Losonsky GA. Young AE. Shindledecker CL. Kapikian AZ. Levine MM. An efficacy trial of the rhesus rotavirus vaccine in Maryland. Am J Dis Child 1990;144:601-4. 23. Vesikari T, Ruuska T. Hannu-Pekka K. Green KY. Flores J, Kapikian AZ. Evaluation of the M37 human rotavirus vaccine in 2- to 6month-old infants. Pediatr Infect Dis J 1991; 10:912-7. 24. Wyatt RG. Mebus CA, Yolken RH. et al. Rotaviral immunity in gnotobiotic calves: heterologous resistance to human virus induced by bovine virus. Science 1979;203:548-50. 25. Gaul SK. Simpson TF, Woode ON. Fulton RW. Antigenic relationships among some animal rotaviruses: virus neutralization in vitro and cross-protection in piglets. J Clin MicrobioI1982;16:495-503. 26. Woode GN. Kelso NE, Simpson TF. Gaul SK, Evans LE, Babiuk L. Antigenic relationships among some bovine rotaviruses: serum neutralization and cross-protection in gnotobiotic calves. J Clin MicrobioI 1983; 18:358-64. 27. Bohl EH, Theil KW. Saif LJ. Isolation and serotyping of porcine rotaviruses and antigenic comparison with other rotaviruses. J Clin Microbiol 1984; 19: 105-1 I. 28. Bishop RF, Tzipori SR. Coulson BS, Unicomb LE. Albert MJ. Barnes GL. Heterologous protection against rotavirus-induced disease in gnotobiotic piglets. J Clin Microbiol 1986;24: 1023-8. 29. Torres A, Lin JH. Diarrheal response of gnotobiotic pigs after fetal infection and neonatal challenge with homologous and heterologous human rota virus strains. J Virol 1986;60: 1107-12. 30. Hoshino Y, Torres A, Aves TM. et al. Neutralizing antibody response of gnotobiotic piglets after fetal infection and neonatal challenge with human rotavirus strains [letter). J Infect Dis 1987;156:103840. 31. Woode GN. Zheng S, Rosen BI. Knight N, Kelso-Gourley NE. Ramig RF. Protection between different serotypes of bovine rotavirus in gnotobiotic calves: specificity of serum antibody and coproantibody responses. J Clin Microbiol 1987;25: 1052-8.
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
The method used to measure neutralizing antibody titers to different serotypes of human rotavirus in this study was a focus reduction neutralization assay. Others have quantified antibody titers to specific epitopes on the VP7 and VP4 proteins by an epitope-blocking assay (EBA) using monoclonal antibodies to these proteins [53-57]. In some, but not all, subjects the ability of the EBA to detect antibody rises after vaccination or natural rota virus infection was comparable to that of the neutralization assay. Because the EBA is much less laborious, it could be the assay of choice for studies such as that reported here. The major limitation of this procedure, however, is that it detects antibody only to specific epitopes on the prototype strains used in the assay procedure. If the rotaviruses that elicit the antibody responses to be tested either lack or have little antibody response to these specific epitopes, the EBA will not provide a valid measure ofneutralizing antibody induction by these strains. Therefore, unless a series of monoclonal antibodies to each serotype is available and used in the EBA, the neutralization assay will remain the more reliable method to measure overall neutralizing antibody titers to specific rotavirus serotypes.
JID 1992; 166 (December)
JID 1992;166 (December)
Antibody Titers and Rotavirus Immunity
46. Green KY, Kapikian AZ. Identification ofVP7 epitopes associatedwith protection against human rotavirus illness or shedding in volunteers. J Virol 1992;66:548-53. 47. Ward RL, McNeal MM, Sheridan JF. Evidence that active protection following oral immunization of mice with live rotavirus is not dependent on neutralizing antibody. Virology 1992; 188:57-66. 48. Trudel M, Nadon F, Seguin C, Binz H. Protection ofBALB/c mice from respiratory syncytial virus infection by immunization with a synthetic peptide derived from the G glycoprotein. Virology 1991;185: 749-57. 49. Offit PA, Dudzik KI. Rotavirus-specific cytotoxic T lymphocytes passively protect against gastroenteritis in suckling mice. J Virol 1990;64:6325-8. 50. Dharakul T, Rott L, Greenberg HB. Recovery from chronic rotavirus infection in mice with severe combined immunodeficiency: virus clearance mediated by adoptive transfer of immune CD8+ T lymphocytes. J Virol 1990;64:4375-82. 51. Offit PA, Dudzik KI. Rotavirus-specific cytotoxic T lymphocytes crossreact with target cells infected with different rotavirus serotypes. J ViroI1988;63:3507-12. 52. Offit PA, Boyle DB, Both GW, et al. Outer capsid glycoprotein VP7 is recognized by cross-reactive, rotavirus-specific, cytotoxic T lymphocytes. Virology 1991;184:563-8. 53. Shaw RD, Fong KJ, Losonsky GA, et al. Epitope-specific immune responses to rotavirus vaccination. Gastroenterology 1987;93:941-50. 54'. Green KY, Taniguchi K, Mackow ER, Kapikian AZ. Homotypic and heterotypic epitope-specific antibody responses in adult and infant rota virus vaccines: implication for vaccine development. J Infect Dis 1990;161:667-79. 55. Beards GM, Desselberger U. Determination of rotavirus serotype-specific antibodies in sera by competitive enhanced enzyme immunoassay. J Virol Methods 1989;24: 103-10. 56. Taniguchi K, Urasawa T, Kobumichi N, et al. Antibody response to serotype-specific and cross-reactive neutralization epitopes on VP4 and VP7 after rotavirus infection or vaccination. J Clin Microbiol 1991;29:483-7. 57. Matson DO, O'Ryan ML, Pickering LK, et al. Characterization of serum antibody responses to natural rota virus infection in children by VP7-specific epitope-blocking assays. J Clin Microbiol 1992;30: 1056-61.
Downloaded from http://jid.oxfordjournals.org/ at University of Sussex on September 6, 2015
32. Bridger JC, Oldham G. Avirulent rotavirus infection protects calves from disease with and without inducing high levels of neutralizing antibody. J Gen ViroI1987;68:2311-7. 33. Hoshino Y, Saif Ll, Sereno MM, Chanock RM, Kapikian AZ. Infection immunity of piglets to either VP3 or VP7 outer capsid protein confers resistance to challenge with a virulent rotavirus bearing the corresponding antigen. J Virol 1988;62:744-8. 34. Ward RL, Clemens JD, Sack DA, et al. Culture adaptation and characterization of group A rotaviruses causing diarrheal illnesses in Bangladesh from 1985 to 1986. J Clin MicrobioI1991;29:19l5-23. 35. Clemens JD, Ward RL, Rao MR, et al. Serologic evaluation of anti bodies to rotavirus as correlates of the risk of clinically significant rotavirus diarrhea in rural Bangladesh. J Infect Dis 1992;165: 161-5. 36. Clemens J, Sack D, Harris J, et al. Field trial of oral cholera vaccines in Bangladesh. Lancet 1986;2:124-7. 37. Clemens J, Harris J, Sack D, et al. Field trial of oral cholera vaccines in Bangladesh: results of one year of follow-up. J Infect Dis 1988; 158: 60-9. 38. Clemens J, Sack D, Harris J, et al. Field trial of oral cholera vaccines in Bangladesh: results of three-year follow-up. Lancet 1990;335:270-3. 39. Beards GM, Campbell AD, Cottrell NR, et al. Enzyme-linked immunosorbent assays based on polyclonal and monoclonal antibodies for rotavirus detection. J Clin Microbiol 1984; 19:248-54. 40. Ward RL, Bernstein or, Young EC, Sherwood JR, Knowlton DR, SchiffGM. Human rotavirus studies in volunteers: determination of infectious dose and serological response to infection. J Infect Dis 1986; 154:871-80. 41. Ruzicka LT, Chowdhury AKMA. Demographic surveillance system. Matlab. Vol 2. Census 1974. Dacca: Cholera Research Laboratory, 1978 (Scientific report no. 10). 42. Bernstein or, McNeal MM, SchiffGM, Ward RL. Induction and persistence oflocal rotavirus antibodies in relation to serum antibodies. J Med Virol 1989;28:90-5. 43. Kleinbaum D, Kupper L, Morgenstern H. Epidemiologic research. Principles and quantitative methods. Belmont, MA: Lifetime Learning Publications, 1983. 44. Chen LC, Huq E, D'Souza S. Sex bias in the family allocation of food and health care in rural Bangladesh. Popul Devel Rev 1981;7:54-70. 45. Perez-Schael I, Garcia D, Gonzalez M, et al. Prospective study of diarrheal diseases is Venezuelan children to evaluate the efficacy of rhesus rotavirus vaccine. J Med ViroI1990;30:219-29.
1257
