Zika virus seroprevalence declines and neutralization antibodies wane in adults following outbreaks in French Polynesia and Fiji

Background Serosurveys published following major outbreaks of Zika virus (ZIKV) have so far shown a high level of seroprevalence from samples collected within 12 months of the first confirmed case. A common assumption is that ZIKV infection confers long-term protection against reinfection, preventing ZIKV from re-emerging in previously affected areas for many years. However, the long-term immune response to ZIKV following an outbreak remains poorly documented. Methods We compared results from eight serological surveys, with sample sizes ranging from 49 to 700, before and after known ZIKV outbreaks in the Pacific region: five from cross-sectional studies of schoolchildren and the general population in French Polynesia over a seven-year period; and three from a longitudinal cohort in Fiji over a four-year period. Findings We found strong evidence of a decline in seroprevalence in both countries over a two-year period following first reported ZIKV transmission. In the cohort in Fiji, there was also a significant decline in antibody titres against ZIKV. However, the decline in seroprevalence was concentrated in adults, while high seroprevalence persisted in children. Interpretation The observed patterns of long-term anti-ZIKV antibody levels following outbreaks in the Pacific could be an early indication of the dynamics of population immunity in Latin America. Given that ZIKV antibody levels can wane substantially over time, follow-up seroprevalence studies and prospective clinical trial designs in Latin America may need to be revised, and assumptions about the potential for ZIKV to re-emerge may need to be revisited. Funding Pacific Funds, ANR, MRC, Wellcome, Royal Society.


SUMMARY Background
Serosurveys published following major outbreaks of Zika virus (ZIKV) have so far shown a high level of seroprevalence from samples collected within 12 months of the first confirmed case. A common assumption is that ZIKV infection confers long-term protection against reinfection, preventing ZIKV from re-emerging in previously affected areas for many years. However, the long-term immune response to ZIKV following an outbreak remains poorly documented.

Methods
We compared results from eight serological surveys, with sample sizes ranging from 49 to 700, before and after known ZIKV outbreaks in the Pacific region: five from cross-sectional studies of schoolchildren and the general population in French Polynesia over a seven-year period; and three from a longitudinal cohort in Fiji over a four-year period.

Findings
We found strong evidence of a decline in seroprevalence in both countries over a twoyear period following first reported ZIKV transmission. In the cohort in Fiji, there was also a significant decline in antibody titres against ZIKV. However, the decline in seroprevalence was concentrated in adults, while high seroprevalence persisted in children.

Interpretation
The observed patterns of long-term anti-ZIKV antibody levels following outbreaks in In countries with known ZIKV outbreaks, the few serological surveys that have been published have demonstrated a high level of ZIKV seroprevalence following the outbreak. In French Polynesia, a population-representative cross-sectional serological survey at the end of the outbreak in 2014 suggested a seroprevalence of 49% 11 . In Martinique, a study of blood donors showed a post-outbreak seroprevalence of 42·2% in 2015 12 . In Salvador, Northeastern Brazil, a serosurvey in 2016 of prospectively sampled individuals including microcephaly and non-microcephaly pregnancies, HIV-infected patients, tuberculosis patients, and university staff, found a postoutbreak seroprevalence of 63·3% 13 . Finally, in paediatric and household cohort studies in Managua, Nicaragua, ZIKV seroprevalence was estimated to be 46% in households following the outbreak in 2016 14 .
It has been suggested that infection with ZIKV confers long-term immunity, which lasts several years. If so, the high level of seroprevalence in affected countries may provide herd immunity such that the current ZIKV epidemic is over in many locations and the virus will not be able to re-emerge for decades to come 2,9,13,15 . Recent evidence suggests that neutralizing antibodies can distinguish between ZIKV and dengue virus (DENV) -a closely related Flavivirus -and that the immune response following ZIKV infection can persist over a year 16,17 . Early data also suggests that primary ZIKV infection confers protective immunity 18 . However, ZIKV serosurveys conducted at the end of the outbreak and 18 months later in French Polynesia found a drop in seroprevalence in the Society Islands, the archipelago where over 85% of the inhabitants of French Polynesia reside 11 . Therefore, the long-term immune response following a ZIKV outbreak remains unclear. Consequently, so too does the potential for new outbreaks.
Here, we explore short-and long-term antibody responses against ZIKV as well as neutralizing response against ZIKV following two ZIKV outbreaks in the Pacific region. We compared results from five serological surveys in the Society Islands, French Polynesia, over a seven-year period, and three serial serological surveys in the same cohort of individuals in Central Division, Fiji, over a four-year period. These surveys span the pre-and post-outbreak period in each country, allowing us to examine temporal changes in antibody responses and hence, herd immunity, following a ZIKV outbreak.

French Polynesia
Four separate ZIKV serosurveys were previously conducted in the Society Islands (Table 1). As reported previously 11,19 , a first serosurvey (!=593) was conducted in

Fiji
Two serial serosurveys were previously conducted in Fiji (Table 1)

French Polynesia
The five serosurveys were approved by the Ethics Committee of French Polynesia

French Polynesia
Serum samples collected from blood donors between July 2011 and October 2013 and samples collected from the general population and schoolchildren in 2014 were all tested for presence of IgG antibodies against ZIKV and each of the four DENV serotypes using a recombinant antigen-based indirect ELISA as reported previously 11,19 . Samples collected from the general population in 2015 and from schoolchildren in 2018 were tested by microsphere immunoassay (MIA) using the same recombinant antigens as for the ELISA 7,11 .

Fiji
All serum samples collected in Fiji were tested using MIA to detect IgG antibodies against ZIKV and all four DENV serotypes as previously reported 7,11,21 .
As an additional validation, a subset of samples collected in 2013 and 2015, and all samples collected in 2017 were tested for the presence of neutralizing antibodies against ZIKV and each of the four DENV serotypes using a neutralization assay as previously described 7 . Among the 14 individuals that tested seronegative against all viruses in 2013 and were re-tested in 2015, there was no evidence of an association between changes in ZIKV titre and changes in any of the DENV titre, suggesting that the changes in ZIKV titre were unlikely to be influenced by DENV cross-reaction (Supp. Fig 1).

Statistical analysis
For data from Fiji, where serial samples were collected from the same individual, changes in seroprevalence between studies were tested using McNemar's test. In French Polynesia, chi-squared tests were performed to test for evidence of a change in seroprevalence between two cross-sectional surveys. Changes in mean titre level between groups were analyzed using a t-test.
All data and code used in the analysis are available at: https://github.com/a-henderson91/zika-sero-pacific.git.

Role of the funding source
The authors declare that the funders had no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication.

RESULTS
ZIKV seroprevalence estimates from blood samples collected before and after the first  Table 1). In contrast seroprevalence in children remained stable over the period at 19% (Figure 1), with five seroconversions and four Table 1).
In order to assess whether the decline in ZIKV seroprevalence was also observed for other circulating Flaviviruses, the seroprevalence pattern against each of the four DENV serotypes was analyzed in both countries, by age group (Supp. Figure 2). In   . Table 3). After adjusting for this sampling bias, there was no evidence of a change in seroprevalence for any of the four DENV whereas the decline in ZIKV was still present (Supp. Table 3).
To explore dynamics of antibody waning at the individual level, we performed neutralization assays on a subset of 46 participants from Fiji for whom sufficient sera were available from all three collection periods. We found that in the 32 individuals who were ZIKV seropositive in 2015, anti-ZIKV antibody levels waned significantly in 2017, with an average decline in log titre of -1·94 (t-test, p<0·001) (Figure 2A and Supp. Table 4).
To test whether the dynamics of anti-ZIKV antibody waning were different from the response to DENV infection, we compared results for ZIKV to the neutralization  Table 4).
However, while the mean antibody log titre for ZIKV declined between 2015 and 2017 (-1·94) (t-test, "<0·001), it increased for DENV-3 (0·89) over the same period (t-test, p=0·04) (Figure 2A and Supp. Table 4).  This cross-sectional study found that all participants had neutralizing antibodies against ZIKV 12 to 19 months after infection. This study also found that at least 37% of the participants had no evidence of past DENV infection, which supports the hypothesis that anti-ZIKV immune responses may persist longer in populations that have had less exposure to DENV. More data are therefore needed to test the hypothesis about the potential impact of pre-existing anti-DENV immune response on anti-ZIKV antibody waning.

DISCUSSION
Although we found evidence of waning antibodies following two ZIKV outbreaks, the implications for susceptibility to further ZIKV infection remain unclear. Given the antigenic similarity of DENV and ZIKV 25 , it is commonly assumed that the immune response to ZIKV infection will be similar to that following DENV infection. High levels of neutralizing antibodies to DENV have been shown to correlate with protection from symptomatic infection 26 . Moreover, infection with a single DENV serotype can confer lifelong immunity to the infecting serotype as well as a transient period of cross-neutralization against heterologous serotypes 27 . However, it is unclear in the context of DENV, as well as ZIKV, what the relationship is between a specific titre value and susceptibility to further infection. Despite these caveats, our results show that anti-ZIKV antibody levels can wane substantially over time and long-term antibody dynamics following a ZIKV outbreak are distinctive to those following a DENV outbreak in the same location during the same period.
There are some additional limitations to our analysis. First, we did not have polymerase chain reaction (PCR) confirmation of ZIKV infection in individuals sampled in this study. We have presented analysis of representative serological surveys in two locations with known, PCR-confirmed ZIKV outbreaks 5,28 . However, PCR confirmation for ZIKV at the individual level remains difficult to obtain, in particular from blood samples, and there have been relatively few confirmations globally 28 , let alone analysis of long-term antibody dynamics in PCR confirmed patients.
Our analysis was also limited by study design. In French Polynesia, surveys were cross-sectional, so we were unable to examine temporal antibody dynamics at the individual level. However, both cross-sectional studies of the general population were conducted using population representative cluster sampling 11 in the same remote island locations with stable population, which enabled robust comparisons of overall seroprevalence. We did identify one potential source of sampling bias with different DENV exposure profiles in the two surveys, but our conclusions of declining seroprevalence for ZIKV persisted once we adjusted for this bias (Supp. Fig 3,   Tables 1-2). We also used different serological analyses between the studies in French Polynesia in 2014 and 2015, however both used the same recombinant antigens and it has been shown that there was good agreement between ELISA and MIA in the 2014 samples 11 . In Fiji, a strength of our study was the collection of longitudinal samples from the same individuals at three time points. However, our sample size was limited given the logistical challenge of recontacting participants twice over a four-year period. These data provided very strong evidence that ZIKV seroprevalence declined over the two-year period following first reports of circulation, but our sample size was insufficient to fully explore the potential effect of anti-DENV pre-existing immunity on anti-ZIKV antibody waning once we stratified individuals by previous DENV exposure.
The global ZIKV epidemic began in the Pacific islands in 2013 before spreading in Central and South America from 2015. Seroprevalence studies following ZIKV epidemics in Latin America have been reported but data have either been nonrepresentative 13 or not enough time had elapsed since the outbreak to observe longterm dynamics 14 . To our knowledge, these are the first studies of seroprevalence over a long-term period following a ZIKV outbreak. Therefore, patterns observed in

Pacific islands may be an early indication of what might happen in Latin America
where ZIKV outbreaks began two to three years after the French Polynesia epidemic 3,29 .
In the short-term, our findings have implications for the design of follow up studies of ZIKV in Latin America. Specifically, our results provide evidence that levels of seroprevalence one to two years following ZIKV circulation may be lower than previously expected and study designs may need to be adapted to reflect this. For example, estimates of microcephaly risk may be inflated if derived from long-term seroprevalence data that underestimate the true extent of infection within the population. There may also be a stronger case for pursuit of vaccine development if low seroprevalence and waning antibody levels against ZIKV are associated with susceptibility to reinfection, and hence future outbreaks. However, designs of vaccine trials 30 could be affected if post-outbreak seroprevalence levels are lower than originally expected. Potential for waning of long-term responses would also need to be considered during vaccine development, as would the DENV exposure history of vaccine recipients. Finally, our results could have implications for public health planning in countries that have experienced ZIKV epidemics. Predictive modelling has typically assumed that infection confers lifelong immunity against the virus 2,9,15 , but the resulting conclusions about potential outbreak duration may need to be revised if waning of long-term ZIKV antibodies leads to a decline in herd immunity over time. In particular, more work is needed to establish the relationship between waning antibody levels and susceptibility to reinfection, and to be vigilant for potential reemergence of ZIKV in areas that previously had high post-outbreak seroprevalence.

DECLARATION OF INTERESTS
We declare no competing interests.

ACKNOWLEDGMENTS
We are grateful to the minister for education of French Polynesia and to the directors, teachers, nurses and schoolchildren from the elementary and junior high schools selected for the serosurvey in 2018.
We greatly thank all the participants and community leaders in Fiji who generously contributed to the study over the three visits. We would like to acknowledge the work   Year Seroprevalence

Supp. Figure 1: Correlation between rise in DENV and ZIKV log titres between 2013-2015 for individuals who were initially seronegative to all five viruses in 2013.
There is significant correlation between DENV-1 and DENV-3 viruses (top row, p=0.0012), which are known to be antigenically similar 1 , suggesting likely crossreactive responses. However, changes in ZIKV titres were not associated with responses to any of DENV viruses (far right column), which strongly indicates that the ZIKV results were genuine infections.