Background
In Thailand, control efforts have been highly effective in curbing malaria nationwide [
1], but malaria remains a significant public health problem along the hilly and forested areas of the country’s borders with Myanmar and Cambodia [
2‐
4]. High geographical heterogeneity in malaria endemicity and the presence of multiple
Plasmodium species that cause human malaria (
Plasmodium falciparum,
Plasmodium vivax,
Plasmodium malariae,
Plasmodium ovale and
Plasmodium knowlesi) are characteristics of malaria epidemiology in the region.
The Government of Thailand aims to achieve malaria elimination by 2030 [
5], thus identification (including parasite speciation) and accurate treatment of asymptomatic and symptomatic infections are critical to accomplish this goal. However, this remains a major challenge because light microscopic analysis of blood smears, the gold standard in malaria diagnosis in Thailand, is insensitive at detecting low-level parasitaemia [
6]. It is known that as malaria transmission declines, an increasing proportion of individuals are found to have asymptomatic and submicroscopic malaria infections [
7], such as in many countries in Southeast Asia and the Amazon [
8‐
12]. This is important because asymptomatic and submicroscopic malaria infections are known to contribute to transmission [
7,
13,
14].
Our study site in the Myanmar-border province of Tak, once Thailand’s highest malaria burden region, experienced a drastic reduction in transmission recently [
15,
16], and based on microscopy-estimated parasite prevalence (<1 %), it is now considered to be a low-transmission area. However, in our previous study in the Thai–Myanmar border area, a significant number of asymptomatic and submicroscopic malaria infections amongst the adult population were detected by quantitative polymerase chain reaction (qPCR) [
17]. Moreover, amongst adult malaria patients attending a malaria clinic, a large number of cryptic mixed-species infections went undetected and untreated [
17].
In the present study, the molecular screening and antibody profiling by microarray were expanded to include samples from all ages and three seasonal time points to obtain a more accurate assessment of the current epidemiology of falciparum and vivax malaria in Tak. For blood samples collected during community mass blood surveys and passive case detection activities at the hospital and clinic, the presence of all five Plasmodium species found in Thailand was investigated by qPCR. The population’s antibody responses against P. falciparum and P. vivax proteins were profiled to examine the development of antibody acquisition relative to age and season, and the targets of antibody responses in persons with asymptomatic and febrile malaria in an age-related manner.
Methods
Study sites and samples
The study was conducted in the village Mae Salid Noi (17° 28′ 4.7202″, 98° 1′ 48.5106″) and the town of Mae Tan (17° 13′ 49.0146″, 98° 13′ 55.6212″) in Tak Province, northwestern Thailand along the Thai–Myanmar border [
17]. The sites are in a low and unstable transmission area, with higher transmission in the rainy season from May to October [
3]. The four human malaria parasites, as well as the simian malaria species
P. knowlesi, are known to infect humans in the region [
18,
19], but
P.
vivax and
P. falciparum are vastly predominant [
16,
20,
21].
Whole blood samples were collected during three cross-sectional mass blood surveys (MBS) in the study village Mae Salid Noi, in March (n = 485) and August (n = 398) of 2013, and January 2014 (n = 464) from individuals ranging from 0.7 to 92 years of age (median, 14; mean 22, 95 % CI 21–23). The health history, including absence or presence of malaria symptoms as described below, was recorded weekly for each study participant from October 2012 to June 2014. The sample size (n = 1347) enabled us to determine parasite prevalence by qPCR with 2.6 % margin of error, using alpha 0.05.
Additionally, 297 whole blood samples were collected from individuals with suspected malaria during passive case detection (PCD) at the Mae Tan malaria clinic and hospital in March (n = 67) and August (n = 53) 2013, and January 2014 (n = 177). Age of PCD participants ranged from 0.8 to 84 years old (median, 23; mean 25, 23–27).
Blood sample preparation for DNA analysis
From each study participant, 300 μL of whole blood was collected from finger prick into a Microvette CB300 capillary blood collector with lithium–heparin (Sarstedt, Newton, NC). Samples were centrifuged to separate cellular and plasma fractions, then immediately frozen at −80 °C for shipment to University of California Irvine for analysis. Upon thawing, plasma was removed and stored at −80 °C until use. Total genomic DNA was isolated from 100 μL of pelleted cellular fraction using DNeasy Blood and Tissue kit in the QIAcube automated system (Qiagen, Valencia, CA), using the Blood and Body Fluid Spin Protocol with Manual Lysis V1. Purified genomic DNA samples were eluted with 200 μL of Buffer AE and kept at −20 °C until use.
Sample analysis by microscopy and quantitative PCR
Field microscopy was performed by local trained staff who provided the first result, and positive cases were treated per national malaria treatment guidelines. Individuals found positive in subsequent expert microscopy and molecular screenings were not treated. Expert microscopic examination was conducted at Mahidol University by an expert microscopist with over three decades of experience. Molecular detection of Plasmodium parasites was performed at University of California Irvine using a two-tier strategy for qPCR consisting of an initial screening of all 1644 samples for the presence of Plasmodium genus-specific products using SYBR Green, followed by TaqMan assays to determine the Plasmodium species in samples positive in the genus-specific assay.
For the Plasmodium-genus screening, primers were designed to hybridize with a region of 18S rRNA gene conserved amongst P. falciparum (Pf), P. vivax (Pv), P. malariae (Pm), P. ovale (Po) and P. knowlesi (Pk). The forward primer sequence was 5′-GTATTCAGATGTCAGAGGTG-3′, and the reverse primer was 5′-CCTACTCTTGTCTTAAACTAGT-3′. Amplification was performed in 20 μL reactions containing 2 μL of genomic DNA, 10 μL FastStart SYBR Green qPCR Master Mix (Roche, Indianapolis, IN), 0.2 μM of each primer and 3 mM MgCl2, in a CFX96 Touch Real-Time PCR Detection System (BIORAD, Hercules, CA). After initial denaturation at 95 °C for 10 min, 45 cycles of 94 °C for 30 s, 60 °C for 30 s, 68 °C for 1 min were followed by a final step of 95 °C for 10 s and a melting curve from 65 to 95 °C with 0.5 °C increments for 5 s. Samples were considered positive if Cq was smaller than 41 and there was a product with melt peak in the temperature ranging between 74 and 75.5 °C. All assays included positive and negative controls. The detection limit of this method was determined to be 0.05 parasites/μL using serially diluted P. falciparum cultures.
Samples positive in the genus screening were examined in uniplex TaqMan assays using Plasmo 1 and Plasmo 2 primers and species-specific probes for
P. falciparum,
P. vivax, P. malariae and
P. ovale as described in Rougemont et al. [
22] and for
P. knowlesi as described in Divis et al. [
23], with the modification that probes were either FAM/ZEN/Iowa Black FQ (for Pf, Pv and Pk reactions) (Integrated DNA Technologies, San Diego, CA) or FAM/MGBNFQ (for Pm and Po reactions) (Applied Biosystems, Foster City, CA). Amplification was performed in 20 μL reactions containing 2 μL of genomic DNA, 10 μL TaqMan Universal Master Mix II, no UNG (ThermoFisher, Grand Island, NY), 0.2 μM of each primer and 80 ηM of probe, in a CFX96 Touch Real-Time PCR Detection System (BIORAD, Hercules, CA). Samples were tested in duplicate and considered positive if it generated a Cq value smaller than 40.
Sample classification
Samples were classified into four major categories, according to collection site and presence or absence of Plasmodium DNA by qPCR. Samples collected at Mae Salid Noi village during community MBSs were classified as (1) community malaria (if sample was qPCR-positive for Plasmodium), or (2) community healthy (if sample was qPCR-negative). Samples collected at the hospital/malaria clinic from suspected malaria patients were classified as (3) febrile malaria (if qPCR-positive), or (4) non-malaria illness (if qPCR-negative). Malaria symptomatology is defined as fever (>37.5 °C), fatigue, myalgia, headache and nausea, occurring alone or in combination.
Plasmodium falciparum and P. vivax protein microarray
The protein microarray used in this study, named Pf/Pv500 (Antigen Discovery Inc., Irvine CA), was described in detail in Baum et al. [
17], and on NCBI’s Gene Expression Omnibus under platform accession number GPL21194. Briefly, the array includes 500
P. falciparum and 515
P. vivax polypeptides printed as in vitro transcription translation (IVTT) reactions, that were down-selected from larger microarray studies based on seroreactivity and antigenicity to humans. Gene accession numbers follow annotation published on PlasmoDB [
24]. For large proteins printed on the microarray as overlapping polypeptides or individual exons, the exon position relative to the full molecule and the segment of the ORF are indicated where applicable.
Probing of plasma samples on the Pf/Pv500 microarray
From the community MBS, 298 plasma samples were probed: 41 community malaria, and 257 community healthy samples, including the corresponding longitudinal samples of community malaria samples, and samples from individuals age-, gender-, and season-matched to the community malaria samples. Of these, 150 samples collected in March 2013 were randomly selected for the seroconversion study, 15 samples for each of 10 age groups. Additionally, 83 samples collected during PCD at the malaria clinic and hospital were probed, including 32 malaria fever and 51 age- and gender-matched non-malaria illness patients. As unexposed controls, 16 samples from healthy blood donors from the United States, with no travel history to malaria endemic regions, were used for serology comparisons. All microarray probing was performed on the same day, and as previously described in Baum et al. [
17].
Data analysis
For analysis of antibody binding to
P. falciparum and
P. vivax polypeptides on the microarray the following steps were taken: (1) the median background signal of antibody binding to 24 spots of IVTT reaction without DNA template (no template control, NTC) was calculated for each individual sample; (2) the raw values of antibody binding to
P. falciparum and
P. vivax polypeptides were divided by their corresponding median NTC value, generating fold-over-control (FOC) values; (3) FOC values were log
2-transformed for data normalization. Normalized data was used for statistical analyses and for figure representations of the data. (4) To determine which polypeptides were seroreactive to plasma from the Thai sample cohort, Significance Analysis for Microarrays (SAM) [
25] was performed comparing the intensity of antibody binding to the proteins on the array between the samples collected in Thailand from individuals over 15 years-old (
n = 211) and 16 USA unexposed controls. The test was performed using MeV 4.8.1, with the following parameters: median and 90th percentile of false discovery rate (FDR), 1.2 and 9.3 %, respectively; median and 90th percentile of number of false significant genes, 3.9 and 30.5, respectively. This resulted in 326 polypeptides being considered significantly seroreactive in exposed Thai samples, and all further analyses considered only this set. Individual plasma samples were considered seropositive for a polypeptide if the sample’s signal intensity value was above the upper 99 % confidence interval value of the unexposed control group. Breadth of response was determined by the number of antigens an individual or group of samples were seropositive to, based on the above criteria for seropositivity. For analysis of antibody responses, the non-parametric multiple comparison Steel–Dwass and the Wilcoxon test were used for pairwise comparisons of means, using JMP11.2. Significance tests were 2-sided and set at 0.05 level for type I error. Z-scores were calculated as the number of standard deviations above or below the mean of the unexposed group. Annual rates of seroconversion and seroreversion for the immunogenic polypeptides were calculated using Systat 11 (Systat Software, Chicago, IL) by fitting age-specific seroprevalence data to a reversible catalytic model using the maximum-likelihood method that assumes binomial error distribution:
\(P_{t} = \lambda /\lambda + \rho (1 - e( - (\lambda + \rho )*t))\) [
26]; where
P
t
is the proportion of seropositive individuals in each age group
t, λ is the annual rate of seroconversion and ρ is the annual rate of reversion to seronegativity. The model was fitted to seroprevalence curves to each seroreactive antigens using ten age groups, shown in Additional file
1. Individuals below 2 years of age were excluded to eliminate pre-existing maternal antibodies from analyses.
T test for comparison of slopes of linear regression lines was performed as per Zar [
27]. A two-proportion Z-test was used to compare the proportion of samples in different age categories between two populations. The geometric mean value and the 95 % confidence interval of data are reported in parenthesis, i.e. (0.73, 0.72–0.74).
Discussion
In a pilot study in the village of Mae Salid Noi and town of Mae Tan, a high level of submicroscopic infections was found amongst asymptomatic and symptomatic adults [
17]. Now, from an age-stratified cohort six-times larger, new observations from this low transmission area are added regarding the frequency of asymptomatic and submicroscopic infections, differences in rates of antibody acquisition to
P. falciparum and
P. vivax antigens, and antibodies profiles associated with seasonal changes and with clinical and subclinical malaria.
While the detection limit of expert microscopy is approximately 100 parasites per μL [
6], the qPCR technique used here detected as few as 0.05 parasites per μL of whole blood. In the community mass blood surveys,
Plasmodium infection rates estimated by qPCR were nearly 14-times higher than those estimated by expert microscopy, and despite the overall low parasite prevalence in the population (3.0 %), over 90 % of infections were both asymptomatic and submicroscopic at the time of the survey. This is consistent with the mounting evidence from other low transmission areas in the Amazon, Africa and Southeast Asia, where a previously unrecognized reservoir of
Plasmodium infection is rising in the form of asymptomatic submicroscopic parasitaemia [
8‐
12,
28‐
31]. In areas of very low transmission submicroscopic carriers are estimated to be the source of 20–50 % of all human-to-mosquito transmissions [
14]. Although the importance of submicroscopic infections for malaria transmission is still unclear [
32], the Malaria Eradication Research Agenda suggests that any parasitaemia, no matter how small, may be potentially a source of transmission and thus a threat to malaria elimination efforts [
7].
Mixed-species infections were common in both the community survey and passive case surveillance in the hospital and clinic, where 25 % of asymptomatic infections and 68 % of febrile malaria cases were caused by co-infections of primarily
P. falciparum and
P. vivax. All mixed-species infections detected by qPCR were misdiagnosed as single-species infections by light microscopy, primarily by not detecting cryptic
P. falciparum in the blood smears. The inability to detect and treat malaria infections is a serious threat to the health care of individual patients and to malaria control and elimination. Molecular diagnosis is not practical in the field and hospital or clinic settings, therefore training of hospital technicians and improvement of malaria rapid diagnostic tests are urgently needed, particularly in light of the presence of multidrug-resistant parasites in Thailand [
33,
34].
Notwithstanding the small number of parasitaemic samples found in the three MBSs of the community by qPCR, serological measures indicated exposure to malaria parasites in all individuals tested. However, the youngest children showed limited responses, reacting to a small number of antigens, a serological indication of the success of malaria control efforts in the area because very low antibody titers are expected to correlate with reduced parasite exposure events. These children are an attractive target group for serosurveillance because detection of antibodies in this group, after drastic reduction or elimination of local transmission, indicates resurgence of malaria in a population [
35,
36]. Serology provides a view of present and past parasite exposures, and seroprevalence rates can be used to define malaria endemicity in an area [
37,
38].
It could be expected that the low levels of antibodies against
Plasmodium in the youngest group would render it especially susceptible to malaria infection and disease, and indeed, in this study half of community malaria cases were of persons below 15 years old. However, despite the lack of high titers or broad antibody repertoires against the parasite, the vast majority of the infections were submicroscopic and asymptomatic, and remained so for at least 5 months. This challenges the notion that individuals with little exposure to malaria have limited ability to suppress parasitaemia and would become symptomatic, but is consistent with epidemiological findings from other low transmission areas [
8,
14,
39,
40]. In fact, high levels of antibodies against
P. falciparum in children from Papua New Guinea were not associated with protection from malaria, rather with higher exposure to parasites [
41]. The high prevalence of submicroscopic and asymptomatic infections in the 3–15 years old cohort demonstrates that despite the low antibody titers, efficient parasite-controlling immunity developed after few exposures.
Experimental and natural infection studies have shown that complete strain-specific immunity to vivax malaria may be acquired after two or three exposures, but the rate of acquisition of disease protective immunity differs between
P. falciparum and
P. vivax (reviewed in [
42]). Abundant evidence showed that the number of clinical episodes of vivax malaria decreased earlier in life than that of falciparum malaria, suggesting a faster acquisition of protective immunity against
P. vivax [
42]. In this study, the intensity and breadth of antibody response in the youngest age groups was significantly higher against
P. vivax, but a reversal in species-specific immunodominance occurred in adults. Two possible reasons can be suggested to explain this age-related difference in immunodominance, which are not necessarily mutually exclusive. First, the shift in species-specific response with increasing age may result from differences in the biology of the falciparum and vivax infections, such as relapses in
P. vivax and higher parasitaemia in
P. falciparum. The estimated SCR and SRR showed that antibodies against
P. falciparum accumulated continuously and decayed at a slow rate throughout life, whereas antibodies against
P. vivax were quickly acquired but were less stable. Second, the historical relative prevalence of the two species also could play a role in this difference. From the 1970s to the mid-1990s, the prevalence of
P. falciparum exceeded
P. vivax in the region [
15], a trend that was reverted in the recent past years, resulting in
P. vivax currently exceeding
P. falciparum by 2:1. Hence, adults in the cohort would have historically experienced more falciparum infections than the children, who now are comparatively more frequently exposed to
P. vivax. Additionally, antibody cross-reactivity between
P.
falciparum and
P. vivax could add confounding factors to this analysis, but this could not be addressed in this study due to the co-endemicity of these two species in the region.
As transmission levels drop further, serology is being recognized as a valuable tool to monitor malaria epidemiology [
7]. Examination of age-specific seroprevalence profiles can be used to detect transmission hot-spots [
43] and detect changes in local transmission [
35,
38]. Furthermore, absence of antibodies against
Plasmodium has been used to show the success of elimination programmes in several countries [
44‐
46]. As the longevity of antibodies against
Plasmodium varies between antigens [
47,
48] and has both short- [
49‐
51] and long-lived [
52‐
54] components, it is imperative to choose antigenic markers with the appropriate kinetics to detect recent parasite exposure, rather than memory responses from past infections.
The kinetics of antibody responses to hundreds of proteins were explored in the longitudinal cohort, but there was little change at the level of individual markers in the 10 months studied. Children showed the most changes across time, indicative of reinforcement of memory responses or acquisition of new antibodies with time, while there were minimal changes in the adults’ responses. This is similar to what was observed in Kenya, where significant increase in breadth and intensity of antibody response after the malaria season was observed in children below 8 years old, but not in adults [
55]. Nonetheless, there was significant overall increase in antibody levels from before to after the rainy season in both adults and children, suggesting that some may have had asymptomatic infections in the intervening months between sampling times. Thus, in order to accurately determine antibody kinetics for use in serological surveillance, more frequent sampling accompanied by molecular screening is required, such as in the study recently published by Helb et al. [
56].
Another aim of this study was to identify antigenic targets that elicited higher antibody levels in asymptomatic carriers when compared to febrile patients, and that could be suggested as to play a role in protective responses. However, antibody titers were significantly higher in those who experienced clinical symptoms, especially in the younger age groups. Here, it may be once again the case that high levels of antibodies to
Plasmodium are indicative of higher exposure rather than of protection to malaria [
41]. Not only did symptomatic patients have, in the majority of cases, parasite levels that were detectable by microscopy while asymptomatic carriers had subpatent parasitaemia, the majority of symptomatic cases also carried multi-species infections. These factors likely would augment the antibody response in febrile patients, while asymptomatic carriers are capable of controlling parasite levels effectively by some other mechanism. Interestingly, recent studies showed evidence of upregulation of inhibitory receptors eliciting exhaustion-related phenotypes on T cells associated with defective effector function during chronic parasitaemia or regular reinfections with
P. falciparum [
57] and
P. vivax [
58]. The exhaustion of CD4 T cells may be associated with failure to develop a fully differentiated B cell response in these cases, such as in the subpatent asymptomatic infections common in our cohort, and result in lower levels of antibodies [
59]. Nonetheless, other studies designed to discover markers of disease protection in malaria in high and low transmission settings have identified several serological correlates of protection [
55,
60,
61].
In summary, the findings of the present study have implications for surveillance, control and elimination of malaria in Thailand. In particular, the data suggests children aged 3–15 years old are a sensitive group for serosurveillance monitoring changes in transmission due to malaria interventions. Additionally, in the community 100 % of
P. falciparum and mixed species infections, and 90 % of
P. vivax infections went unrecognized and untreated. Amongst the febrile cases, microscopy failed to detect 75 % of
P. vivax infections; and of the mixed-species infections, all went unrecognized, being misdiagnosed as a single-species infection or completely undetected. Because current surveillance methods focus on case management, malaria transmission cannot be interrupted if asymptomatic infections are infectious to mosquitoes and contribute to transmission [
7,
11]. Currently, implementation of molecular screening methods for population-wide surveillance is not feasible in developing countries, and the field applicable serological surveillance tools are deficient because of lack of optimal serological markers with the appropriate kinetics [
36]. The mounting evidence of the shortcomings of the current malaria surveillance, detection and treatment methods argues for a major change in malaria control approaches. For example, mass drug administration to largely healthy populations has been controversial and resistance to this approach is considerable [
28,
62,
63]. However, there is an urgent need for discussion and exploration of such alternative approaches by the scientific and public health communities.
Authors’ contributions
EB performed all experiments (except microarray probing), statistical analysis, and wrote the manuscript. JS, JS, KK collected blood samples and donor symptom history. AJ and OT performed microarray probing. ML database management. HD contributed to some data analysis. LC, PF, GY helped conceive the study and reviewed drafts of the manuscript. All authors read and approved the final manuscript.