Zika Virus Circulates at Low Levels in Western and Coastal Kenya

Abstract

Background

Zika virus (ZIKV) was discovered over 70 years ago in East Africa, but little is known about its circulation and pathogenesis there.

Methods

We screened 327 plasma samples collected 2–12 months after febrile illness in Western and coastal Kenya (1993–2016) for binding and neutralizing antibodies to distinguish ZIKV and dengue virus (DENV) responses, which we found were common in coastal Kenya.

Results

Two cases had durable ZIKV-specific antibodies and 2 cases had ZIKV antibodies at similar levels as DENV antibodies.

Conclusions

This suggests low-level ZIKV circulation in Kenya over 2 decades and sets a baseline for future surveillance efforts in East Africa.

Four cases of Zika virus infection were identified by screening 327 convalescent plasma samples collected from individuals who experienced Zika-like illness in 2 urban regions of Kenya, suggesting a low burden of Zika infections in Kenya over a 20-year period.

The recent unprecedented spread and severe pathogenic features of Zika virus (ZIKV) in the Americas have led to its recognition as a global threat and the need to understand the global prevalence of ZIKV, which was first discovered in Uganda in 1947. Seroepidemiological surveys conducted in the decades after its discovery suggested that ZIKV is endemic on the African continent [1]; however, there is little evidence for the sequalae associated with the American ZIKV outbreak. Prior reports of ZIKV seroprevalence in Africa are hampered by methodological challenges that do not account for extensive antibody cross-reactivity between ZIKV and other closely related flaviviruses, particularly dengue virus (DENV), which is common in Africa. As a result of these challenges and little concern about ZIKV before the American outbreak, its prevalence in Africa remains poorly defined, particularly in East Africa where ZIKV was first discovered. More recent studies of ZIKV seroprevalence in East Africa have collectively reported low ZIKV exposure (0.2%–0.8%) [2–5] and little evidence for ZIKV as a cause of febrile illness in East Africa [3]. However, a study in northwestern Kenya found that 7.7% of samples contained ZIKV-neutralizing antibodies (nAbs) [2], suggesting there may be regions of higher ZIKV prevalence in Kenya.

To better define the prevalence of ZIKV infections in East Africa, we focused on 2 of the most densely populated regions of Kenya where the mosquito vector for ZIKV circulates, including Mombasa, where there have been several arboviral outbreaks [6, 7]. We screened convalescent plasma from individuals who experienced ZIKV-like illness (fever and/or rash) in cohorts in Kisumu (2011–2013) and Mombasa (1993–2016), Kenya for evidence of ZIKV exposure and identified cases of prior ZIKV infection.

MATERIALS AND METHODS

Study Design and Plasma Samples

Individuals with ZIKV-like symptoms were identified in 2 cohorts of women in Kenya (Supplemental Methods). In a sex worker cohort from Mombasa (1993–2016), individuals with fever (temperature ≥37.5°C) on exam were identified, as well as individuals who had self-reported fever and rash either by self-report or on exam. In a pregnancy/postpartum cohort from Kisumu (2011–2013), individuals with fever and/or rash on exam were identified. The first available plasma sample collected 2–12 months after each study visit with ZIKV-like symptoms (“convalescent sample”) (Supplemental Table 1) was tested for antibodies, as were longitudinal samples in a subset of cases with evidence for ZIKV antibodies in the convalescent sample. Plasma from individuals with defined infection status were used as controls (Supplemental Table 2). This research was approved by the Kenyatta National Hospital Ethics and Research Committee and Institutional Review Boards of the University of Washington and the Fred Hutchinson Cancer Research Center.

Immune Assays

An anti-ZIKV nonstructural protein 1 (NS1) IgG ELISA (R&D Systems) was performed using the manufacturer’s protocol and specified cutoff for positive and equivocal samples (Supplemental Methods). Neutralization studies using ZIKV and 4 serotypes of DENV luciferase reporter virus particles (RVPs) were carried out in BHK-DCSIGN cells as described [8] (Supplemental Methods). The “% neutralization” was calculated for each sample by normalizing relative luciferase units to no-serum controls.

RESULTS

In total, 327 plasma samples (235 from Mombasa; 95 from Kisumu) from 245 individuals, including 71 individuals with >1 plasma sample, were examined for ZIKV-specific antibodies (Supplementary Table 1), starting with an enzyme-linked immunosorbent assay (ELISA) that detects antibodies targeting ZIKV NS1. The ELISA was first validated with plasma samples from ZIKV-infected, ZIKV-naive/DENV-infected, and flavivirus-naive individuals (Supplementary Table 2). All ZIKV-infected samples were positive by ELISA except 1 early-convalescent sample (50 835) that was negative (Supplementary Figure 1). Seventeen of 18 flavivirus-naive and ZIKV-naive/DENV-infected samples were negative by ELISA; 1 sample (negative 4) was equivocal. Seven (2.1%) Kenyan convalescent plasmas were positive by ELISA, 4 of which were from Mombasa and 3 were from Kisumu (Figure 1A). Forty-three (13.1%) samples were equivocal and 277 samples (84.8%) were negative.

Figure 1.

Zika virus (ZIKV) and dengue virus (DENV) antibody levels in convalescent plasma. (A) The net optical density (O.D.) value obtained in the anti-ZIKV NS1 IgG ELISA is shown for samples, grouped by whether they are from the Mombasa and Kisumu cohorts. The dotted lines correspond to the positive and negative thresholds of the assay, as recommended by the manufacturer. (B) The percentage of neutralization of ZIKV at 1:75 dilution of plasma is shown for all convalescent samples, grouped by the Mombasa and Kisumu cohorts. The dotted line corresponds to the highest level of ZIKV neutralization observed in the ZIKV-naive/DENV-infected control plasma (Supplementary Figure 2A). Samples that were positive in the anti-ZIKV NS1 ELISA are labeled with green dots. (C) The percentage of ZIKV neutralization in the Mombasa and Kisumu cohorts is shown compared with the highest level of percentage of DENV neutralization among the 4 serotypes. The dotted line corresponds to the highest level of ZIKV neutralization observed in the ZIKV-naive/DENV-infected plasma (Supplementary Figure 2A). Samples that were positive in the anti-ZIKV NS1 ELISA are labeled with green dots.

Neutralizing antibody activity was then defined using an RVP assay. Control samples with established ZIKV infection potently neutralized ZIKV, with 100% neutralization at a 1:75 dilution for all cases except 50 835, which showed lower neutralization that was dose-dependent (Supplementary Figure 2). The ZIKV-naive/DENV-infected control plasma had a significantly lower level of neutralization against ZIKV at 1:75 dilution (P < .0001; range, 3%–59%). Only 1 of the 8 flavivirus-naive samples showed detectable ZIKV neutralization (25%).

Twenty-one (6.4%) Kenyan convalescent samples neutralized ZIKV above the highest level detected in ZIKV-naive/DENV-infected plasma (59%), 4 of which were ELISA-positive and 2 were equivocal by ELISA (Figure 1B). To assess the specificity of the nAb responses in these 21 samples, ZIKV neutralization was compared with the highest level of DENV neutralization (Figure 1C). In each cohort, there was 1 sample (M-1, K-1) that had high ZIKV neutralization (>90%) and low DENV neutralization (<21%). The 19 remaining plasma samples, all of which were from Mombasa, had high DENV neutralization (>95%).

To attempt to distinguish whether the responses in these 21 samples were due to ZIKV and/or DENV exposure, the IC₅₀ of neutralization (NT₅₀) against ZIKV and each DENV serotype was determined for each sample (Table 1). Zika virus-only infection was assumed for 2 cases (M-1 and K-1) in which there was a ≥4-fold higher NT₅₀ against ZIKV than all DENV serotypes, a difference that was used to differentiate ZIKV and DENV infections in previous studies [9]. For 2 cases (M-2, M-4), we could not predict which infection led to the responses because the NT₅₀ for DENV and ZIKV were within 4-fold of each other, but 1 (M-4) also had evidence of ZIKV NS1 antibodies by ELISA. Seventeen samples had >4-fold activity against DENV versus ZIKV and thus were defined as only DENV-infected [9], although 3 (M-5, M-10, M-15) did have evidence of ZIKV NS1 antibodies.

Table 1.

NT₅₀ Results With the 21 Kenyan Plasma Samples That Neutralized ZIKV Above Levels Observed in ZIKV-Naive/DENV-Infected Plasma

				NT50
Cohort	Sample ID	Year	ZIKV NS1 ELISA Resulta	ZIKV	DENV-1	DENV-2	DENV-3	DENV-4	Interpreted Infection History
Mombasa	M-1	1996	+	823	<75	112	<75	<75	ZIKV
Kisumu	K-1	2013	+	1158	<75	<75	<75	<75
Mombasa	M-2	1997	-	562	1036	187	601	892	DENV and possible ZIKV
	M-4	1994	±	1538	2410	2160	1253	3819
	M-6	1994	-	<75	78	4392	216	592	DENV-2
	M-18	1997	-	<75	567	87	78	<75	DENV-1
	M-19	1995	-	<75	284	671	2707	524	DENV-3
	M-5	2006	+	570	6515	15 281	4502	6690	Secondary DENVb
	M-10	2013	+	380	3015	4875	3425	3488
	M-15	2016	±	129	1584	2659	4682	1172
	M-8c	1993	-	156	1050	417	3285	1817
	M-17	1996	-	145	804	141	590	245
	M-3	1994	-	3859	6332	14 827	56 482	78 027
	M-7	2006	-	326	2015	131	1232	236
	M-9	2006	-	336	12 774	659	8119	6178
	M-11	2012	-	112	2062	558	2172	1075
	M-12	2008	-	156	3499	2506	873	2845
	M-13	1994	-	<75	<75	149	672	875
	M-14	2008	-	82	449	651	408	530
	M-16c	1996	-	157	1043	202	1793	1233
	M-20	2006	-	<75	213	2018	4817	3856

Abbreviations: DENV, dengue virus; ELISA, enzyme-linked immunosorbent assay; ID, identification; NS1, nonstructural protein 1; NT₅₀, IC₅₀ of neutralization; ZIKV, Zika virus.

^aELISA results are shown as positive (+), equivocal (+/-), or negative (-).

^bSecondary DENV was defined as NT₅₀ values within 4-fold of each other for at least 2 DENV serotypes [9].

^cM-8 and M-16 samples are from the same subject who experienced 2 separate events of fever/rash ~2 years apart.

To examine the ZIKV nAb response over time in cases with evidence of either ZIKV binding or neutralizing responses, we tested available longitudinal plasma samples (Supplemental Figure 3). In the 2 cases defined as ZIKV-only infections, potent ZIKV-neutralizing activity was observed over time, including as far as 35.8 months before the reported time of a febrile illness (M-1) and at the reported time of febrile illness (K-1). The nAb titers against ZIKV in subject M-1 were durable over an ~10-year period, with a spike of nAb titers against ZIKV coincident with a lower spike of neutralizing activity against all serotypes of DENV at 67.8 months postfebrile illness.

For the 2 cases with similar levels of ZIKV and DENV nAbs (M-2, M-4), the levels of DENV nAbs were, in most cases, higher than ZIKV nAb levels in the subsequent longitudinal plasma. For M-2, there was an increase in DENV-1 and DENV-4 responses in plasma from 4 months after the convalescent sample. For M-4, there was a spike in neutralizing activity after febrile illness against 4 of 5 viruses tested that declined in subsequent samples for all viruses except DENV-1. For the 3 cases that were only positive or equivocal in the ELISA (M-5, M-10, M-15), there were potent responses to DENV and little evidence for ZIKV nAb activity at all time points tested.

Discussion

In this study, we screened convalescent plasma from subjects with reported febrile illness in Mombasa and Kisumu, Kenya for antibody responses to ZIKV and DENV to define ZIKV prevalence in these regions. We identified 2 subjects with ZIKV-only infections, and 2 cases with evidence for both ZIKV and DENV infections. The 4 cases span an ~20-year period (1994, 1996, 1997, 2013), suggesting a very low, persistent burden of ZIKV in 2 major urban areas in East Africa.

There was one ZIKV-only infection in each cohort (K-1, M-1). Both subjects had concordant binding and neutralizing responses against ZIKV without evidence for DENV antibodies. In Kisumu, there was little evidence for DENV infection overall, making it more straightforward to detect ZIKV infection. In the Mombasa cohort, where DENV antibodies were common, we detected 2 cases that suggested the individual had been infected with both ZIKV and DENV. In both cases (M-2, M-4), there were durable neutralizing responses to both ZIKV and DENV in longitudinal plasma samples. M-4 also had an equivocal ELISA result for ZIKV NS1 antibodies. This supports infection by both ZIKV and DENV in this case, although there was some evidence for a shift to a more dominant DENV-1 response at later time points as the other responses waned and thus we were not able to rule out a single DENV serotype infection with strong cross-reactive antibodies to ZIKV [10]. Case M-2 was ELISA-negative and did show some evidence for more potent DENV responses over time, making it possible but less clear if infection with both viruses occurred. Including these 2 possible cases of ZIKV infection with the 2 clear cases indicate a prevalence of 1.2% among samples tested.

We found a durable ZIKV-neutralizing response lasting over 10 years in ZIKV-positive subject M-1. To our knowledge, this is the first report of durable neutralizing responses against ZIKV over such a long time period. We were intriguingued to find that ~6 years after ZIKV-specific nAbs were detected, there was spike in ZIKV nAb titers along with a coincident low-level spike in those against DENV. This could represent an infection with a flavivirus other than ZIKV or DENV that triggered a cross-neutralizing response, as has been described previously [11].

The 3 subjects who had a positive or equivocal result for ZIKV NS1 binding antibodies with detectable but low-level ZIKV nAbs at multiple time points nonetheless had much higher levels of neutralizing responses against DENV. We defined these as cases of DENV infection with cross-reactive responses, although we cannot rule out that detection of an actual ZIKV nAb response is confounded by high DENV nAbs due to coinfection in these cases. We also cannot rule out this possibility for 3 other samples (M-3, M-7, M-9) that were classified as secondary DENV infections but also had high-titer anti-ZIKV nAb responses (NT₅₀ >300).

We identified 3 other samples that were ELISA-positive but did not have potent nAb titers against ZIKV or DENV (Figure 1B and 1C, green points). None of these cases present a strong case of ZIKV infection and suggest that the ZIKV ELISA has lower specificity than expected, at least for ZIKV strains in East Africa. This could reflect the development of this assay for the American ZIKV outbreak.

Our study has several limitations, including reliance on self-report of ZIKV-like symptoms in the Mombasa cohort, inclusion of only female individuals, and differential follow-up of study participants. As noted above, there is some imprecision in defining some cases, although given that the cases are few with imprecision in both directions, the prevalence estimates would have very little variation. We did not examine plasma activity against Spondweni virus (SPOV), which belongs to the same serogroup as ZIKV. Although we did not rule out SPOV infection, the activity of SPOV nAbs against ZIKV is generally much lower than against SPOV [12], and the titers observed here are high in the 2 cases that we defined as the clearest ZIKV infections (M-1, K-1); they also have concordant anti-NS1 binding activity.

Conclusions

Our findings concur with other recent studies in East and Central Africa and, together, demonstrate low population-level exposure to ZIKV in many regions on the African continent [2–5, 13]. They highlight that ZIKV does not have a large contribution to febrile illness in large urban areas of Kenya, which is in agreement with a recent prospective, hospital-based study in Uganda [3]. However, a higher prevalence of ZIKV (6.3%–7.7%) was reported in northwestern Kenya and West Africa, although it is unclear how cross-neutralizing responses were accounted for in northwestern Kenya and nAb responses were not examined in West Africa [2, 14]. In any case, the prevalence of ZIKV is not nearly as high in Kenya as suggested by early studies (as high as 52% reported in coastal Kenya) that used assays that were poorly discriminating against other flavivirus infections [1]. Given that ZIKV is likely to continue its geographic expansion, with reported outbreaks of Asian-lineage ZIKV and related pathogenesis in Cape Verde and Angola [15], our findings set an important baseline for future ZIKV surveillance in the major urban regions of Kenya where arbovirus infections are common.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. We thank William Messer for sharing flavivirus-naive sera, the National Institute of Allergy and Infectious Diseases’ Biodefense and Emerging Infectious Disease Resource Repository (BEI Resources) for providing well characterized Zika virus- and dengue virus-infected sera, and Leslie Goo and Daryl Humes for helpful discussions. Finally, we thank the participants and investigators of the Mombasa and Kisumu cohorts.

Financial support. This work was funded by the National Institutes of Health (R37AI038518, P01HD064915, T32AI083203, F30AI142870, K01AI116298) and funding from an Endowed Chair from the Fred Hutchinson Cancer Research Center.

Potential conflicts of interest. J. K. reports grants from National Institutes of Health, during the conduct of the study. G. J.-S. reports grants from National Institutes of Health, personal fees from University of Washington, during the conduct of the study; grants from National Institutes of Health, grants from Centers for Disease Control and Prevention, grants from Thrasher, personal fees from University of Washington, personal fees from UpToDate, grants from IMPAACT, outside the submitted work. R. S. M. reports grants from US National Institutes of Health, during the conduct of the study; grants from Hologic Corporation, personal fees from Lupin Pharmaceuticals, outside the submitted work. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Presented in part: Annual Meeting of the University of Nairobi Collaborative Centre for Research and Training in HIV/AIDS/STIs, January 2019, Nairobi, Kenya.