Background
Malaria still thrives in the African highlands, in spite of low vector density exposure [
1]. The western Kenya highlands are an area of particular interest based on the fact that on a relatively small spatial scale, there is considerable variation in altitude, water accumulation, and land-use patterns. As a consequence, the epidemiology of malaria varies markedly. For example, small differences in altitude have been noted to lead to large differences in suitability and availability of vector breeding habitats, and consequently, differing risks of malaria transmission and prevalence [
2,
3]. These patterns of malaria reflect heterogeneities in vector distribution, human vector-contact, and human host factors [
4]. Identified risk factors for malaria transmission include distance to known mosquito breeding sites [
5,
6], household construction methods [
7], and personal protection measures against mosquito bites [
8]. Moreover, altitude and environmental landscape, i.e., topography have also been correlated with risk of malaria infection [
2,
4,
9‐
11].
Assessing variation in malaria endemicty at different altitudes across regions with differing malaria transmission intensities can be achieved directly by determining exposure to malaria-infected mosquitoes, the entomological inoculation rate (EIR) [
12], or indirectly by evaluating serological evidence of malaria exposure in the human population [
13,
14]. Direct measure of the EIR becomes difficult when absolute numbers of mosquitoes and sporozoite rates are low, particularly when EIR is below the detection limits of commonly used trapping methods [
15,
16]. The situation is further complicated when the mosquito densities show marked heterogeneity, because spatial and temporal variations in mosquito densities necessitates long-term intensive and extensive sampling to be accurate [
15‐
17]. Direct determination of malaria parasite prevalence in the human population as an indicator of malaria transmission intensity has limited sensitivity when transmission is low [
18‐
20], furthermore, the sensitivity of the tools used in routine detection of parasitemia; microscopy and
PfHRP2 based rapid diagnostic test (RDTs) presents additional challenges at low parasite densities.
Prevalence of antibodies to
Plasmodium falciparum has been explored as a marker of human exposure to malaria [
13,
14,
21‐
24]. Measurement of serum antibodies is a useful index of malaria transmission intensity when the focus is on evaluation of malaria exposure over time, since anti-malarial antibodies develop after repeated exposures and can persist for months to years after infection [
14]. Seroprevalence reflects cumulative exposure and thus it is less affected by seasonality or unstable transmission due to the longer duration of the specific antibody response. Additionally the longevity of antibody response generates a seroprevalence that is higher than equivalent parasite rates, making it a more sensitive measure. Therefore, immunological markers may be useful to detect malaria exposure in areas of low endemicity [
21,
24]. Seroconversion rates are related to the force of infection of malaria as refracted through the immune responses of exposed individuals [
24‐
26]. Thus the seroconversion rates provide measures of malaria exposure that compares with the malaria transmission intensity [
13,
14,
27]. Additionally, antibody responses have been shown to have a tight correlation with EIR and offer the potential to detect recent changes in malaria transmission intensity [
13,
14,
27].
However, the use of inadequate serological markers may underestimate exposure by virtue of their lack of sensitivity. For instance, in the case of circumsporozoite protein (CSP), sporozoites injected by infected mosquitoes have a relatively short life-span in the blood. Some rapidly develop into liver stages and others are taken up by macrophages, processed, and presented to the immune system [
23]. The amount of antigenic material and the time of contact with immuno-competent cells are relatively shorter than blood stage antigens and thus may underestimate malaria exposure in low transmission settings. Druilhe and others could not detect CSP antigens in children in low transmission area even though 78% of them had detectable blood stage antigens, and from these data, they concluded that CSP is not a reliable marker of malaria endemicity when the total EIR in the area is less than 10 infectious bites per person per year [
28]. In hyperendemic areas, however, CSP has been reported to give reliable estimates of malaria endemicity and reflects the seasonal dynamics of transmission [
23,
29] and may be sufficiently sensitive to evaluate the protective efficacy of anti-vector devices in transiently exposed travelers to endemic countries [
30]. On the other hand, surface proteins of merozoites like apical merozoite antigen-1 (AMA1), are highly immunogenic and tend to saturate detectable antibody responses in the population in low to moderate transmission settings. Alternatively, Drakeley and others [
14] described MSP -1
19 as the most suitable immunological marker for assessing malaria endemicity at varying altitudes along transmission intensity cline.
MSP-1 is secreted as a 195-kDa precursor anchored via glycosylphosphatidylinositol [
31]. This is cleaved by proteases into fragments of 83, 28-30, 38-45 and 42 kDa [
32]. During merozoite invasion, the 42 kDa fragment (MSP-1
42) is further cleaved to produce a 33-kDa fragment (MSP-1
33) and a 19 kDa C-terminal fragment (MSP-1
19); with the later remaining attached to the merozoite surface and present on ring forms in newly invaded erythrocytes [
33]. MSP-1
19 is thus a recognized target of protective immunity [
34].
This study was therefore conducted to investigate serologic evidence of malaria exposure at a highland site along a malaria transmission intensity cline to characterize differences in malaria endemicity. The current study reports an improved understanding of the micro-geographic variation on malaria endemicity in the complex highland eco-system; and potentially identifies vulnerable groups in the event of hyper-transmission. This study confirms that sero-epidemiology provides valuable information for planning effective malaria control strategies and surveillance systems.
Discussion
Using age-specific MSP-1
19 seroprevalence, seroconversion rates (SCR), and total IgG titers, together with malaria infection prevalence, we have observed considerable variation in human exposure to malaria in western Kenya. Seroprevalence of residents at the valley bottom was almost two fold higher than that of the uphill residents. Comparison of the median total IgG titers in the study population revealed a13-fold difference between the uphill and valley bottom residents. Similar trends have been observed in entomological and parasitological studies, which have consistently found higher transmission intensity, vector abundance, and parasite prevalence at the valley bottom, in comparison to the top of the hill [
4,
17,
38]. This may partly be explained by the differences in water accumulation between the two sites, as the valley bottom residents live along River Yala which provides permanent breeding habitats for malaria vectors all year round, resulting in a higher risk of exposure to infected mosquitoes compared to residents uphill [
4,
10,
17,
38].
Immunity to malaria is a function of exposure challenge; it develops gradually as a consequence of experiencing multiple parasite exposures or persistent infection for years. Consequently, seroprevalence reflects cumulative exposure and, thus, is less affected by seasonality or unstable transmission due to the longer duration of the specific antibody response [
14,
21,
41]. It may be limited in detecting discrete seasonal variation in transmission but is a good indicator of long term transmission potential. And antibody persistence generates seroprevalence that are higher than equivalent parasite rates making it more sensitive [
14,
27]. Thus seroprevalence at this site particularly in the valley residents indicate possible frequent or persistent exposure to parasite infection. An earlier study at this same site reported that, 38.2% of asymptomatic individuals harbor infections that persist for 2-5 months and a further 14.2% of them harbored asymptomatic infections from 6-12 months [
17]. A follow up study revealed high infection turn-over rate i.e. frequent clearance and acquisition of infection, with the average infection duration of single parasite genotypes being 1.1 months, and the longest genotype persistence being 3 months [
10]. Whether these infections are recrudescence or re-infections they have the ability to maintain seropositivity and thus lead to acquisition and maintenance of significant high levels of antimalarial antibody responses. Moreover, it has been observed that persistent or even sub-patent infections are sufficient to maintain seropositivity and partial immunity consistent with the concept of premunition [
42].
Age-specific seroprevalence has been used to estimate seroconversion rates (SCR) as a measure of malaria transmission intensity. Earlier studies in neighboring Tanzania have shown that these estimates are tightly correlated with EIR measurement [
13,
14,
27]. Age sero -prevalence curves reflect different levels of transmission intensity. In low transmission settings development of antibodies is slow and is mainly exhibited by the adult population, whereas in a high transmission area, much of the population will be seropositive even at a younger age [
43]. This phenomenon is clearly demonstrated in our age seroprevalence curves (Figure
3). In the uphill population the seroprevalence reaches a peak of 73.3% only at age 20; this is in sharp contrast to that seen in the valley, where seroprevalence reaches a peak of 91.2% by the age of 10. These observations reveal the difference in the intensity of malaria transmission between the two localities, suggesting a higher intensity of malaria transmission in the valley area than the uphill area. These findings are further corroborated by other evidence such as the vector density variation between valley bottom and uphill in our study site [
4,
17] and by the repeated infections observed in the valley area [
10,
17].
In very low transmission settings, where parasite prevalence and EIR are insensitive, serological measures offer a way of accurately assessing endemicity and identifying focal areas of transmission supporting the potential for elimination [
43]. Historically, several studies have used serology in this context. In determining endemicity and evaluating eradication campaigns in Tanzania, repeated cross-sectional serological surveys were used to assess approximately 1,500 individuals, and the serological data accurately distinguished between areas of different transmission intensity [
44]. Similarly in Surinam, supporting an elimination program, approximately 2,000 individuals of all ages were serologically assessed; and the antibody responses in these individuals reflected the epidemiological situation at the time of sample collection and accurately defined areas which had eliminated malaria [
45]. Furthermore, in Mauritius and Tunisia, immuno-florescent antibody assays (IFAT) were used to follow reducing seroprevalence over several years in eradication campaigns until no seropositives were detected in children under 5 and 15 years, respectively, thus confirming successful eradication [
46,
47].
When we compared age-specific IgG titers from the two sites, there was a highly significant difference in the IgG titers between the valley bottom and the uphill residents, indicating a considerable variation in malaria endemicity within the highland area. The observed seasonal variation in the IgG titres in the 5-14 year group suggests that the level of exposure may not result in the development of stable humoral responses by the age 14. If this is solely due to seasonal exposure to malaria parasites, then it may represent a vulnerable group in the valley bottom area that may pose a public health problem in an event of hyper malaria transmission. As individuals remain seropositive for several years, the level of antibody response (IgG titers) can reflect fluctuations in recent exposure. It is known [
43] that antibody levels tend to be higher in actively infected individuals with a concomitant decline as the parasites are cleared. This seems to be the case in the less than 15 years age group in the valley who also had highest parasite prevalence. The valley bottom is characterized by persistent infections, a single parasite genotype has been observed to persist for 3 months [
10], detailed examination of IgG titer with age revealed that on the average, titer level at age 5 in valley is similar to that at age 20 in uphill (data not shown). This should not be surprising as at the age of 10 seroprevalence in the valley exceeds 90%.
The observed parasite prevalence in the current study is lower compared to earlier studies conducted in the same area [
17,
40]. Munyekenye and others [
43], reported a mean annual parasite prevalence of 47.0% in children of 1-9 years and 9.5% in > 19 year olds. Baliraine and others [
17] using the combination of microscopy and PCR technique observed parasite prevalence of 34.4%, 34.1% and 9.1% in 5-9, 10-14 and > 15 years, respectively. The current conducted two cross-sectional surveys across all age groups of participants, and observed a mean infection prevalence of 14.0% in < 5 and 5-14 year olds and 6.8% in > 15 years. Consistent with all these studies is the finding that parasite prevalence generally decreases with age and distance from the valley bottom. Parasite prevalence in adults has not exceeded 10% for several years, probably due to their ability to clear and suppress parasites through their acquired immunity or from clearance using antimalarials. On the whole the mean parasite prevalence in the valley population and that of uphill population were 16.3% and 6.3% respectively. However the equivalent seroprevalence were 79.56% and 46.34% for valley and uphill respectively. This implies that looking at parasite prevalence alone about 84% and 94% of valley and uphill residents would be considered unexposed. Seroepidemiology thus presents a more sensitive tool in describing the malaria endemicity of a population under low to moderate transmission.
We observed a spectrum of MSP-1
19 responses that are highly varied within the same age group and altitude transects. Other studies have observed similar variation in exposure, susceptibility and even disease patterns at the individual level [
11,
48,
49]. This finding may partly be explained by factors such as host genetic polymorphism, [
50], MSP-1 polymorphism [
51] or antigenic sin. Nevertheless, in a population that is predominantly of one sub-tribe, variation in exposure to mosquito bites is likely to play an important role [
7]. Intrinsically individual factors such as household structure, use of ITN [
8], and proximity to breeding sites [
15,
22] may be important determinants of this variation. A follow-up study is underway to test this hypothesis by testing individual responses to
An. gambiae salivary gland protein (gSG6-P1) previously validated as a marker of mosquito bite exposure [
52].
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
KB carried out the field survey, the serological experiments, and the serological analysis and wrote the first draft of the manuscript. GY, YA and AG developed the study protocol, YA led the field operations, data and laboratory specimen collection, and GZ supported field team operations and data analysis. JW, JL, EA and AS coordinated laboratory procedures and serological testing, EA provided the plate antigen and participated in the study design and implementation, GY, AG, DJP and JMO conceived the study, participated in its design and implementation and supervised fieldwork, laboratory investigations, data analysis and manuscript development. All authors read and approved the final manuscript.