Potential impact of host immunity on malaria treatment outcome in Tanzanian children infected with Plasmodium falciparum

Background In malaria endemic areas children may recover from malaria after chemotherapy in spite of harbouring genotypically drug-resistant Plasmodium falciparum. This phenomenon suggests that there is a synergy between drug treatment and acquired immunity. This hypothesis was examined in an area of moderately intense transmission of P. falciparum in Tanzania during a drug trail with sulphadoxine-pyrimethamine (SP) or amodiaquine (AQ). Methods One hundred children with uncomplicated malaria were treated with either SP or AQ and followed for 28 days. Mutations in parasite genes related to SP and AQ-resistance as well as human sickle cell trait and alpha-thalassaemia were determined using PCR and sequence-specific oligonucleotide probes and enzyme-linked immunosorbent assay (SSOP-ELISA), and IgG antibody responses to a panel of P. falciparum antigens were assessed and related to treatment outcome. Results Parasitological or clinical treatment failure (TF) was observed in 68% and 38% of children receiving SP or AQ, respectively. In those with adequate clinical and parasitological response (ACPR) compared to children with TF, and for both treatment regimens, prevalence and levels of anti-Glutamate-rich Protein (GLURP)-specific IgG antibodies were significantly higher (P < 0.001), while prevalence of parasite haplotypes associated with SP and AQ resistance was lower (P = 0.02 and P = 0.07, respectively). Interestingly, anti-GLURP-IgG antibodies were more strongly associated with treatment outcome than parasite resistant haplotypes, while the IgG responses to none of the other 11 malaria antigens were not significantly associated with ACPR. Conclusion These findings suggest that GLURP-specific IgG antibodies in this setting contribute to clearance of drug-resistant infections and support the hypothesis that acquired immunity enhances the clinical efficacy of drug therapy. The results should be confirmed in larger scale with greater sample size and with variation in transmission intensity.


Background
Plasmodium falciparum resistance to commonly available antimalarial drugs such as chloroquine (CQ), amodiaquine (AQ) sulphadoxine-pyrimethamine (SP) is now widespread in most malaria-endemic areas, including Tanzania [1,2]. It has been established that polymorphisms in the parasite dihydrofolate reductase (dhfr), dihydropteroate synthetase (dhps) and chloroquine resistance transporter (Pfcrt) genes are associated with SP and CQ resistance, respectively in vitro [3,4]. Point mutations at positions N51I, C59R and S108N in the dhfr gene [5,6] and at positions A437G and K540E in the dhps gene [7,8] have shown to predict a reduced efficacy to SP in vivo. Likewise, the K76T mutation in the Pfcrt gene is a well described predictor of reduced parasite susceptibility to CQ [9], and to a lesser extent AQ [10]. The prevalence of these mutations has increased as a result of high drug pressure in most sub-Saharan countries in recent years (reviewed in [2,11]).
Patients infected with P. falciparum parasites carrying such drug-resistant mutations sometimes overcome infection after treatment [12]. The ability to recover has been associated with host age [13,14] and transmission intensity [15,16], reflecting an effect of acquired host immunity. From animal models it has also been established that immunity enhances the efficacy of malaria drug treatment [17]. Moreover, haemoglobinopathies, such as sickle cell trait, has been related to increased efficacy of SP treatment of uncomplicated falciparum malaria in Kenya [18]. Therefore, recovery from malaria may depend on the drug efficacy and parasite drug-resistance, as well as a complex interaction with host factors like acquired immunity and innate resistance e.g. haemoglobinopathies.
Various studies have investigated the relationship between potential immune mechanisms, such as antibody responses, and therapeutic efficacy. It has been demonstrated that increased amounts of anti-RESA and anti-NANP antibodies in patients treated with CQ were associated with better clearance of resistant parasites [19,20], whereas other studies could not establish evidence for elevated anti-MSP1 and anti-AMA1 antibody levels in patients recovering after treatment with CQ, SP or AQ [21][22][23]. These observations are however difficult to compare, when factors like patient age, innate resistance, intensity of transmission and level of drug resistance vary substantially between these studies and may influence treatment outcome.
The objective of this study was to evaluate factors influencing outcome of antimalarial treatment to uncomplicated P. falciparum malaria, such as acquired immunity, haemoglobinopathies and genotypic markers of drug resistance. The study likewise wanted to investigate the applicability of drug efficacy trails in testing the importance of antibodies to different vaccine-candidates in patients receiving drugs with reduced efficacy, as suggested previously [14,22]. Patients were children below five years of age exposed to low-to-moderate levels of malaria transmission in Tanzania, treated with either SP or AQ for episodes of uncomplicated febrile malaria.

Study population and samples
The study was done as part of an annual clinical drug-efficacy trial under the East Africa Network for Monitoring Antimalarial Treatment (EANMAT) in collaboration with the National Malaria Control Programme in Tanzania. The trial was conducted during the rainy season between February and July 2005 in Chamwino village, Dodoma region, which is an area characterized by low-to-moderate malaria transmission of mainly P. falciparum. The study protocol was approved by the Ethical Committee of the National Institute for Medical Research and Ministry of Health, Tanzania.
The efficacy study were designed to enrol 100 patients aged 6-59 months presenting with uncomplicated malaria if they met the criteria as defined in the standard efficacy testing protocol by WHO [24]: i) monoinfection with P. falciparum at parasite densities above 2,000/μl and below 200,000/μl, ii) axillary temperature ≥ 37.5°C, iii) haemoglobin > 5 g/dl, iv) absence of severe malnutrition, v) absence of general danger signs, severe and complicated malaria, and vi) informed written consent from parents/guardians. Patients were randomly allocated to receive either sulphadoxine-pyrimethamine (SP) (25 mg/kg sulphadoxine and 1.25 mg/kg pyrimethamine) as a single dose (Fansidar ® , Roche, Switzerland) or amodiaquine (AQ) (10 mg/ kg), once daily for three days (Pfizer, Senegal). Patients were followed up clinically and parasitologically on days 0, 1, 2, 3, 7, 14, and 28 and on any other day if the patient felt unwell.
Treatment outcome during 28 days of follow-up were classified as early treatment failure (ETF), late clinical failure (LCF), late parasitological failure (LPF) or adequate clinical and parasitological response (ACPR) according to the standard protocol of WHO [24]. For the purpose of this study, patients with ETF, LCF and LPF were grouped as treatment failures (TF). LPF were defined as patients with presence of parasitaemia on any day from day 7 to day 28 and axillary temperature < 37.5°C without previously having ETF or LCF. Patients failing treatment or with signs of severe malaria were given quinine and referred to the nearby district hospital.
Parasitaemia was determined microscopically using thin smear of fingerprick blood and stained with 3% Giemsa. Packed cell volume (PCV) was measured in haematocrit capillary tubes. Plasma samples were collected upon centrifugation, and stored at minus 20°C until further analysis.
DNA and genetic analysis DNA was extracted from fingerprick blood collected on filter paper, as previously described [25]. Nested polymerase chain reaction (PCR) assays were applied to differentiate recrudescence from new P. falciparum infections by comparing PCR-generated msp2 genotype patterns in regions FC27 and IC-1 from samples collected on day 0, 7, 14 and 28 [26]. Analysis of the msp2 genotypes was done according to Cattamanchi et al [27]: An outcome was defined as recrudescence if a subsequent sample on any day from day 7 to day 28 post-treatment contained identical alleles or a subset of the alleles present in the day 0 sample and the patient were classified as having late parasitological treatment failure (LPF). An outcome was defined as re-infection if a subsequent sample on any day from day 7 to day 28 post-treatment contained only new alleles compared to the day 0 sample and the patient were classified as having an adequate clinical and parasitological response (ACPR) [24].
Plasmodium falciparum single nucleotide polymorphisms (SNPs) at dhfr (position 50/51, 59 and 108), dhps (position 436/437, 540, 581 and 613) and Pfcrt (position 72-76) were determined on day 0 by sequence-specific oligonucleotide probes (SSOP) and Enzyme-Linked ImmunoSorbent Assay (ELISA)-based technique of PCR amplified fragments as described in [27] Briefly, biotinconjugated nested PCR amplified DNA were fixed on streptavidin-coated ELISA plates and mixed with digoxigenin-labelled oligonucleotide probes with specificity for the SNP's of interest. The mixtures were washed with high stringency at set temperatures before incubated with peroxidase-conjugated anti-digoxigenin antibodies and visualized by o-phenylene-diamine (OPD). The SSOP's enables detection of mixed haplotypes with high specificity. The SNP's were constructed into haplotypes. Infections with mixed parasite haplotypes were only classified as 'mixed' when none of the polymorphisms in the same codon were dominant as determined by OD values of differences above 50%. Sickle cell trait were detected by screening the human β-haemoglobin gene for the A18T mutation by SSOP-ELISA [28] and the African alpha 3.7 deletion variant of alpha + -thalassaemia was determined as previously described [29].

Antibody levels estimated by ELISA
The plasma samples collected at day 0 were tested for the presence of IgG to the eleven antigens mentioned above by ELISA as previously described [35]. Wells of Maxisorp microtiter plates (Nunc, Roskilde, Denmark) were coated with 100 μl of the recombinant protein diluted in 0.1 M glycine-HCl (pH 2.75) (apart from CSP that was diluted in phosphate-buffered saline [PBS], pH 7.2) by overnight incubation at 4°C. Plasma collected from 30 residents of a Tanzanian village where malaria is holoendemic were used as a reference positive plasma pool and plasma from 19 healthy Danish donors who have never been exposed to malaria were used as negative controls. To account for day-to-day variation between assays, antibody responses were calculated as arbitrary units (AU) with the following formula: [(OD sample plasma -OD background )/(OD positive plasma -OD background )] × 100. The cut-of for a positive antibody response was defined as the mean level plus two standard deviations of the antibody reactivity among the negative controls.

Measurement of VSA-specific IgG by flow cytometry
Plasma levels of IgG with specificity for variant surface antigens (VSA) on two parasite isolate lines with different VSA-expression profiles, 3D7 unselected (VSA1) and 3D7 selected on transformed human bone marrow endothelial cells (THrMEB) (VSA2) [35] were tested by a flow cytometric assay previously described [38]. Plasma from six residents of a holoendemic Ghanaian village and six Danish donors served as positive and negative controls, respectively. For each plasma sample, the mean fluorescence index (MFI) was recorded and used as a measure of the VSA-specific antibody level. Antibody responses were calculated as arbitrary units (AU) with the following for-mula: [(MFI sample plasma -MFI background )/(MFI positive plasma -MFI background )] × 100. The antibody positivity cut-of was determined as the mean antibody levels of the plasma from the mean of the six Danish donors plus two standard deviations.

Statistics
Statistical analyses of data were performed with Stata/SE version 8.2 (Stata Corp., Texas, US). χ 2 test and Fisher's exact test were used to compare differences in proportions of antibody responders, presence of parasite resistant genotypes and genes assessing haemoglobinopathies in patients with adequate clinical and parasitological response (ACPR) and treatment failure (TF). Non-parametric Mann Whitney rank sum test was used to analyse differences in median antibody levels between treatment groups. Associations between quantitative variables as age, parasite densities and haemoglobin levels were assessed by linear regression. For multivariable analysis, logistic regression models were used to assess predictors of treatment failure. In these models, parasite density (log parasite density pre-treatment), age and age-squared were used as continuous exposure variables, while presence or absence of parasite resistant genotypes, human antibodies and genes causing haemoglobinopathies were included as categorical exposure variables. Possible effect modification between the exposure variables was considered by testing for interaction between presence of parasite mutant haplotypes and IgG antibodies. P < 0.05 was defined as significant.

Characteristics of the study groups and outcome of treatment
The baseline characteristics of the 100 patients enrolled in the study are shown in Table 1. Age, sex, parasitaemia, temperature and haemoglobin levels on the day of enrolment were similar in the SP and AQ treatment groups.
Of the 50 patients receiving SP, none presented with early clinical failure (ECF), three patients presented with late clinical failures (LCF), 31 patients presented with late parasitological failures (LPF) and the remaining 16 patients had an adequate clinical and parasitological response (ACPR). Of the 50 patients receiving AQ, one presented with ECF, two patients presented with LCF, 16 patients presented with LPF and the remaining 31 patients had an ACPR. All samples were adjusted for new infections by PCR genotyping. Before PCR corrections, 32 patients receiving SP and 22 patients receiving AQ presented with LPF, respectively. All patients with either ECR, LCF or LPF were grouped together as treatment failures (TF). Thus, the treatment failure rates were 68% in the SP group and 38% in the AQ group. There were no significant differences with respect to age, sex, temperature, and haemoglobin level between patients with ACPR or TF in any of the treatment groups. However, in patients receiving SP, the parasite density on day 0 was higher in the TF than ACPR group (P = 0.03). Levels of haemoglobin, parasite density and failure rates showed no relation to age (data not shown).

Associations between parasite mutant haplotypes and treatment outcome
The haplotypes based on the single nucleotide polymorphisms at codon 50/51, 59 and 108 of the P. falciparum dhfr gene, and codon 436/437, 540, 581 and 613 of the P. falciparum dhps gene, were determined on day 0 for parasite infections in patients receiving SP (Table 2).
The majority of parasites in patients receiving AQ expressed the Pfcrt mutant haplotype CVIET (68%) and the remaining samples were the CVMNK wildtype haplotype. There was a trend towards higher prevalence of the CVIET haplotype in patients failing treatment (84%) compared to patients with ACPR (58%), although not statistically significant (χ 2 = 3.2, P = 0.08).
Thus, a higher prevalence of parasites expressing triple mutant dhfr haplotype in patients failing SP treatment and a tendency for a higher prevalence of the mutant CVIET haplotype in patients failing AQ treatment were observed, indicating some association between presence of parasite resistant genotypes and treatment outcome.

Associations between anti-malaria antibodies and treatment outcome
Prevalence and levels of IgG antibodies with specificity for the eleven malaria antigens and the surface of the two parasite isolates were measured on day 0 in patients treated with SP or AQ ( Figure 1 and Table 3). In univariate analysis, patients with ACPR had a higher prevalence of anti-GLURP-R0 IgG (χ 2 = 20.9, P < 0.001) and anti-GLURP-R2 IgG (χ 2 = 12.7, P < 0.001) as well as higher levels of anti-GLURP-R0 and R2 IgG (P < 0.001) compared to patients with TF. Such statistically significant associations were not observed for any of the other antigen-specific IgG responses. Furthermore, the associations between ACPR and prevalence and levels of GLURP-R0 and R2 antibodies were significant when the SP and AQ groups were analysed separately ([SP]: R0: χ 2 = 11.8, P = 0.001; R2: χ 2 = 9.4, P = 0.002. [AQ]: R0: χ 2 = 11.3, P = 0.001; R2: χ 2 = 3.3, P = 0.07). The prevalence of the combined IgG responses to GLURP-R0 and R2 was also correlated to ACPR in both the SP (χ 2 = 10.1, P = 0.002) and the AQ (χ 2 = 11.3, P = 0.001) group. By contrast, no association between antibody responses to any other antigen, analysed individually or in combination, and treatment outcome in neither the SP nor the AQ group was demonstrated. Approximately 25% of patients with IgG against GLURP R0 did not have IgG against R2 and vice versa, indicating that responding to R0 not necessarily means responding to R2. Thus, among the thirteen IgG specificities, only anti-GLURP-R0 and R2 IgG proved to be associated with ACPR.

Impact of haemoglobinopathies on anti-malaria antibodies, parasitaemia and treatment outcome
In the study population, the overall frequency of homozygote and heterozygote alpha + -thalassaemia was 38% and the frequency of sickle cell trait (HbAS) 12%. No sickle cell anaemic patients (HbSS) were found. The prevalence of these traits was similar in the two treatment outcome groups for both SP and AQ (see Table 1). Prevalence and levels of antibodies did not differ between patients with and without sickle cell trait or with and without alpha +thalassaemia for any of the malaria antigens. A lower parasite density was observed at enrolment in patients with alpha + -thalassaemia receiving AQ (P = 0.007), but not for SP (P = 0.41). Haplotypes defined as amino acid sequences based on the following substitutions in dhfr: S108N (single), C59R + S108N or N51I + C59R (double) and N51I + C59R + S108N (triple), and in dhps: A437G (single) and A437G + K540E (double) were assessed. For the crt gene, the K76T amino acid substitution was assessed. *Classified as mixed when none of the haplotypes were dominant.

Potential predictors of treatment outcome
Patients were divided into four groups based on infections with or without mutant haplotypes (the dhfr triple mutant haplotype in the SP group, and the CVIET haplotype in the AQ group), and presence or absence of IgG antibodies to each antigen. In Table 4, only data for GLURP-R0 and R2 are shown.
As expected, patients with anti-GLURP IgG infected with a sensitive haplotype (group 1) were more likely to have ACPR compared to patients with no anti-GLURP IgG infected with a resistant parasite haplotype (group 4). In multivariate regression models, the effect of IgG antibodies, parasite haplotype and parasite density on treat-ment outcome was analysed ( Table 5). Regardless of parasite density and age, anti-GLURP IgG were the strongest predictor of treatment outcome in the SP and AQ groups separately and combined. Presence of resistant parasite haplotypes and high parasitaemia were also associated with treatment failure in patients receiving SP, but not as strongly as the presence of GLURP-R0 and R2 IgG and not in the AQ group.

Discussion
This study evaluated the contribution of some antimalaria antibodies and parasite resistant haplotypes on the outcome of SP and AQ treatment of patients under five years of age with uncomplicated malaria. The rate of SP treatment failure observed (68%) is unexpectedly high for this area of endemicity, but similar SP failure rates (43%-74%) have been shown in other parts of Tanzania the last 10 years [39][40][41]. Not surprisingly patients failing SP and AQ treatment were generally infected with parasites expressing drug resistance-related mutations. However, increased risk of SP treatment failure was associated with the triple dhfr (CIRN) irrespective of the dhps haplotype status as demonstrated in other studies [5,42]. In the Prevalence of IgG antibody in patients at day 0 to different malaria antigens in relation to treatment outcome AQ treatment group, only a trend towards an increased risk of treatment failure was found in patients infected with the Pfcrt-CVIET mutant haplotype. Although the latter could be explained by involvement of other putative genes related to AQ resistance, such as Pfmdr [43], and that the assumptions are based on a relatively small sample size, these findings indicate that the association between parasite resistant genotypes and SP or AQ treatment outcome is not absolute. Interestingly, the prevalence and level of anti-GLURP antibodies was associated with a substantial decreased risk of treatment failure irrespective of drug regimen. This association could not be demonstrated for any of the other antigens studied, neither individually nor combined. Similarly, although there was an association between parasite density and patients failing SP treatment, the treatment outcome was independent of age, alpha +thalassaemia (homozygote or heterozygote) and sickle cell trait, and for AQ treatment outcome, also parasite density. Furthermore, in regression analysis, the presence of GLURP R0 or R2 specific IgG antibodies influenced the risk for failing SP and AQ treatment more strongly than presence of parasite resistant haplotypes. This is in line with a recent study from Uganda, where transmission intensity, rather than the parasite genotype, influenced treatment outcome [16]. This indicates that the degree of immunity in the population in combination with the number of resistant-related parasite mutations predicts ACPR, suggesting a synergistic effect between the antimalarial and host immunity, as also demonstrated in mice [17].
It has recently been proposed that therapeutic responses to resistant malaria infections can provide a mechanism for measuring effective clinical immunity [14], i.e. drug trials can be useful in identifying antigens and vaccine candidates affording protection [22]. Investigations of the impact of immunity on treatment outcome have however resulted in conflicting findings [19][20][21][22][23]. In line with the observations from the present study, level of anti-IgG to MSP1 19 was not associated with CQ [22] SP [23] or AQ [21,23] treatment outcome, and prevalence of anti-AMA1 [22] and anti-MSP3 [19] IgG was not associated with reduced risk for CQ treatment failure. However, in contrast to the findings from the present study, the prevalence of IgG to MSP1 19 was higher in patients clinically recovering after CQ treatment [22], and IgG reactivity to multiple, but not single, K1 and MAD20 alleles of MSP1 were related to efficacy of AQ treatment [21]. In addition, prevalence of anti-NANP and anti-RESA IgG has been demonstrated to be higher in patients recovering successfully from CQ [19] and artesunate [20] treatment. The inability to demonstrate an association between prevalence and levels of antibodies against antigens other than GLURP in  Similar tendencies were observed when restricting the analysis to those with a measurable antibody level only, and when analysing each drug arm separately. For definitions of antigens, see Figure 1 and in the method section.
the present study may be explained by the relatively small sample size (50 patients in each group); alternatively by allelic diversity as shown in the study from Gabon (20), or by variable IgG subclass composition of the patient sera. For instance, it has previously been described that only IgG3 against MSP3 is associated with protection against clinical malaria [44], although this was not observed in another study where total IgG was a strong predictor of protection [45]. Finally, one could speculate if prevalence or level of IgG antibodies to these other specific antigens are not primary responsible for, or associated with, host immune protection against uncomplicated malaria but rather severe malaria -which none of the patients had. However, in patients with established clinical malaria, sporozoite-blocking anti-CSP antibodies were, however, not expected to influence treatment outcome.
Interestingly, at day 0, all patients had IgG antibodies to at least one of the antigens tested, but no patient had IgG antibodies to all the antigens. This may reflect substantial variation in response to infections. Previous immuno-epidemiological studies have demonstrated that GLURP-specific IgG antibodies are associated with protection against high parasitaemia [46], clinical disease [47] and inhibition of parasite growth in vitro [48], but this is not direct evidence of IgG antibodies against GLURP increase the ability to clear uncomplicated malaria infections, nor of a causal relationship between treatment success and IgG antibodies against GLURP. However, it was shown that patients clinically recovering after treatment are more likely to carry anti-GLURP IgG antibodies and the presence of such antibodies were associated with reduced risk of failing drug-resistant parasite infections.

Conclusion
This study suggests that presence or level of GLURP-specific IgG antibodies in some settings are better predictors of SP and AQ treatment outcome than the parasite density at enrolment and drug-resistant related parasite mutations. This confirms the hypothesis that acquired immunity enhances the efficacy of antimalarial treatment, and supports the use of drug trials to identify markers of immunological importance. In addition, the findings suggest that drugs with reduced efficacy, such as SP, may still be effective for treatment of uncomplicated malaria in individuals with some acquired immunity, such as older children and adults living in areas of intense malaria transmission.