Genetic diversity and natural selection on the thrombospondin-related adhesive protein (TRAP) gene of Plasmodium falciparum on Bioko Island, Equatorial Guinea and global comparative analysis

This study was carried out in Malabo Regional Hospital, in private diagnostic clinics (PDCs) in different regions of Bioko Island, and in the clinic of the Chinese medical aid team to the Republic of Equatorial Guinea. Bioko is an island 32 km off the west coast of Africa, and the northernmost part of Equatorial Guinea. The island has a population of 334,463 inhabitants (2015 census, of which approximately 90% live in Malabo, the capital city of Equatorial Guinea) and a humid tropical environment. The launch of the BIMCP has had a marked impact on malaria transmission, with the decrease of parasite prevalence from over 45% in 2004 to 8.5% in 2016 and the reduction of entomological inoculation rate (EIR) from more than 1000 before 2004 to 14 in 2015 (www.​mcdinternational​.​org) in Bioko Island.

A total of 153 blood spot samples were collected from the patients with uncomplicated malaria during January 2016–December 2018. Consents were obtained from all participating subjects or their parents and the ethical approval was obtained from the Ethics Committee of Malabo Regional Hospital. Included patients were aged between 4 months and 80 years, were residents on Bioko Island. Malaria patients were classified into uncomplicated malaria states according to World Health Organization (WHO) criteria, which were defined as positive smear for P. falciparum and presence of fever (≥ 37.5 °C). Dried blood spots were collected on day zero of enrollment through finger prick bleeding spotted onto Whatman 903® filter paper (GE Healthcare, Pittsburgh, USA) for future use. Laboratory screening for malaria was done using rapid diagnostic test (RDT) and confirmed using microscopic examination of blood smears. For quality control, archived malaria-positive microslides were re-examined, and parasite density was recorded; The Plasmodium species were identified by a real-time PCR followed by high-resolution melting (HRM) [11]. The pGEM-T standard plasmids of the four human Plasmodium species, P. falciparum, Plasmodium ovale, Plasmodium malariae and Plasmodium vivax were used as control. In this study, although only clear non-superimposed signals of the sequences were included from electropherograms for further analysis, these sequences could derive from either monoclonal infections or the most predominant clones in these isolates.

Parasite genomic DNA was extracted from dried filter blood spots by Chelex-100 extraction method described in a previous article [12]. The DNA products were collected in sterile tubes and stored at − 80 °C in reserve.

The full-length TRAP gene (NCBI Gene ID: 814170) was divided into two segments and amplified by nested PCR. The primers designed for nested PCR are presented in Table 1. For the first round PCR, l μl of genomic DNA was amplified with 12.5 μl 2 × Master Mix (DNA Polymerase, dNTP Mixture, PCR buffer), 1 μl 10 nM outer forward primer, 1 μl 10 nM outer reverse primer, and sterile ultra-pure water to a final volume of 25 μl. Thermal cycling parameters for PCR were as follows: initial denaturation at 94 °C for 3 min; 30 cycles of 94 °C for 30 s, annealing at 55 °C for 30 s and extension at 72 °C for 2 min; followed by a final extension step at 72 °C for 5 min. For the second round PCR, 2 μl of the primary PCR product was amplified in a 50 μl reaction volume comprised of 25 μl 2 × Master Mix (DNA Polymerase, dNTP Mixture, PCR buffer), 2 μl 10 nM inner forward primer, 2 μl 10 nM inner reverse primer, and sterile ultrapure water to a final volume of 50 μl. All PCR products were analysed using 1.2% agarose gel electrophoresis, and then, they were purified and sequenced by using an ABI 3730 × L automated sequencer (Shanghai Yingjun Biotechnology Co., LTD, Guangzhou branch). To ensure the accuracy of the sequencing, each isolate was sequenced at least twice. Sequencing primers were the reverse primers of the second round PCR. All the sequences were analysed and integrated by MEGA6 software [13]. These nucleotide sequences have been deposited at NCBI under Accession Numbers (MK981410–MK981530).

The nucleotide and deduced amino acid sequences of PfTRAP were analysed using EditSeq and SeqMan in the DNASTAR package (DNASTAR, Madison, WI, USA). The PfTRAP sequence of the laboratory-adapted P. falciparum strain 3D7 (NC_004331.3) was included in the alignment for comparison as a reference sequence. The values of segregating sites (S), the number of haplotypes (H), haplotype diversity (Hd), and nucleotide diversity (π) were calculated using DnaSP version 5.10.00 [14]. In order to test the null hypothesis of neutrality of PfTRAP, the rates of synonymous (dS) and nonsynonymous (dN) substitutions were estimated and were compared using the Z-test (P < 0.05) in MEGA6 program [13] using Nei and Gojobori’s method [15] with the Jukes and Cantor (JC) correction of 1000 bootstrap replications. Tajima’s D test [16] were performed using DnaSP ver. 5.10.00 in order to evaluate the neutral theory of natural selection. The probability of recombination between adjacent sites per generation (Rb), and the minimum number of recombination events (Rm) were analysed using DnaSP ver. 5.10.00.

In order to analyse the genetic diversities and natural selection of PfTRAP among global P. falciparum isolates, a total of 1287 sequences from 13 malaria endemic regions (Bangladesh, Cambodia, Congo, Gambia, Ghana, Guinea, Laos, Malawi, Mali, Myanmar, Senegal, Thailand and Vietnam) were acquired by mining the MalariaGEN Pf3k Project database (release 5) [16, 17] using samtools [18] and vcftools [19]. Genetic polymorphism and tests of neutrality were calculated for each population using DnaSP ver. 5.10.00 and MEGA6 as described above. In order to investigate the genetic relationships among global PfTRAP haplotypes, the haplotype network for 1406 full-length sequences of PfTRAP from Bioko and other countries listed above was constructed by Network 10.1.0.0 program using Median-Joining method [20].

The 3-dimensional structure of PfTRAP (3D7 isolate) was predicted and modeled by I-TASSER online server (https://​zhanglab.​ccmb.​med.​umich.​edu/​I-TASSER/​) [21] and presented using YASARA software [22]. PolyPhen-2 [23] online serve was used to predict potential impact of amino acid substitutions on the structure or function. A prediction of probably damaging indicates that the query substitution is predicted to be damaging with high confidence, while a prediction of benign indicates that the query substitution is predicted to be benign with high confidence. A prediction of possibly damaging indicates that the query substitution is predicted to be damaging, but with low confidence. Using FOLDX plugin [24] in YASARA [22] to predict the changes in free energy before and after the mutations: ΔΔG(change) = ΔG(mutation)  − ΔG(wild-type). As a rule of thumb for use: ΔΔG (change) > 0: the mutation is destabilizing; ΔΔG (change) < 0: the mutation is stabilizing.

As the result of genetic polymorphism and natural selection analysis shown in Table 2, on Bioko Island, a total of 87 haplotypes were found among 119 samples, with the haplotype diversity (Hd) for 0.99. Furthermore, nucleotide diversity (π) of Bioko Island was detected as 0.01046, which was a relatively high value among the 14 countries in this analysis. As for the parameters related to natural selection, the value of dN–dS was 6.2231 (p < 0.05), which hinted at natural selection on Bioko Island. The value of Tajima’s D of Bioko Island was detected as − 0.41438, which was not significant based on Tajima’s D significance levels [16]. To analyse the recombination degree of PfTRAP on Bioko Island, the minimum number of recombination event (Rm) and the probability of recombination between adjacent sites (Rb) was detected as 22 and 119, respectively, which shows relatively high level among the 14 countries or areas included in this analysis.

Not only Bioko Island, but also 7 other African countries and 6 Asian countries were included in the genetic diversity and natural selection analysis of PfTRAP. 1287 global sequences mined from Pf3K database (23 for Bangladesh, 393 for Cambodia, 54 for Congo, 34 for Gambia, 228 for Ghana, 49 for Guinea, 42 for Laos, 106 for Malawi, 39 for Mali, 45 for Myanmar, 110 for Senegal, 104 for Thailand, 60 for Vietnam, shown in Additional file 4) and 119 Bioko sequences were applied in the global comparative analysis. According to the results, a total of 161 SNPs were detected from the 1406 PfTRAP sequences (7 SNPs for DomainI, 41 for DomainII, 23 for DomainIII, 83 for DomainIV, 3 for DomainV and 4 for DomainVI, respectively), among them there were 42 SNPs appeared only once. Due to the uncertainty, these single SNPs were not included in the following analysis (original SNP information was shown in Additional file 1). There were 7 SNPs (S91N, K130R, S179N, L287P, N317K, S419P and A489G) popular globally, with percentage over 80%, and notably, there are 60 SNPs unique to Africa, while there were 12 SNPs unique to Asia. Not surprisingly, no mutations were found in some important motifs including WSPCSVTCG in A-domain, metal ion-dependent adhesion site (MIDAS), IQQ motif, A glycoprotein with a cellular recognition function (RGD) and two important points (aa131 and aa162).

As shown in Table 2, both Hd and π show that the polymorphism in Africa is greater than in Asia (except Bangladesh). All the 14 countries and areas in this analysis shown evidences of natural selection with the positive value of dN–dS (p < 0.05). Except for Bioko, Congo, Guinea, Myanmar, Bangladesh, the Tajima’s D of other 9 countries shown positive value, which signified low levels of both low and high frequency polymorphisms, indicating a decrease in population size and/or balancing selection (Table 2). In the sight of the Tajima’s D of the full-length PfTRAP, on Bioko, the value of DomainII is positive and obviously deviated from 0, while most of the rest region got the negative value (Fig. 1). The overall trend is similar in Bioko and Africa, while in Asia, higher positive Tajima’s D was found in N-terminal of DomainII and C-terminal of DomainIV (Fig. 1). When turns to the recombination effect, the parameters (Rm and Rb) revealed that more frequent recombination events were taking place in Africa rather than Asia (Table 2).

For further exploration of the relationship of haplotypes, a haplotype network was constructed based on global PfTRAP sequences (excluded repeat region) using NETWORK software. A total of 661 haplotypes were detected among 1406 isolates (Fig. 2), among them, 667 Asian isolates clustered in 136 haplotypes and 739 African isolates clustered in 528 haplotypes, which indicated the higher heterogeneity found among African P. falciparum population. Generally, the vast majority of haplotypes were limited to one continent (Africa or Asia) and only several haplotypes (Hap_8, Hap_18, Hap_69, Hap_104 and Hap_370) were shared between two continents isolates (Detailed information about haplotypes was shown in Additional files 2, 3 and 5). In the meantime, it was found that the haplotypes of Bioko Island isolates were scattered. Obviously, haplotypes from Asian isolates were distributed relatively concentrated while the haplotypes from African isolates were in greater dispersion with long branch, which indicated that the recent mutations have been more active in Africa than Asia. In order to analyse the genetic differentiation among PfTRAPs from 14 malaria endemic regions, Fixation Index (FST) was calculated and shown in Table 3. According to Sewall Wright rules, FST range from 0 to 0.05 reflects little genetic differentiation; FST range from 0.05 to 0.15 means moderate genetic differentiation; FST range from 0.15 to 0.25 for great genetic differentiation, while FST over 0.25 means extremely high genetic differentiation. According to Table 3, it seems like the TRAP gene from Bioko Island have not so much genetic differentiation with other 7 African countries in this analysis according to the FST between them shows little or moderate level of genetic differentiation (p < 0.05). The FST between Bioko and Asian countries were over 0.15 (p < 0.05), which indicated a greater genetic differentiation. Generally, the differentiation of PfTRAP among African countries is less pronounced, and the same phenomenon was found among Asian countries, but the high degree of genetic differentiation between African and Asian countries cannot be ignored (p < 0.05).

In this study, all the proven T cell epitopes, B cell epitopes, as well as nonsynonymous substitutions were marked and presented in Fig. 3. As it shown, 98 amino acid substitutions were located at the B cell epitopes while 25 substitutions were at T cell epitopes. 24 substitutions located at the overlap region of T cell epitope and B cell epitope. Since previous reports shown that cellular immunity is more dominant than humoral immunity in the PfTRAP-based vaccine clinical trials, the following mutation effect prediction analysis was focus on the variations at T cell epitopes. In Table 4, among the 30 mutations in T cell epitopes, 13 of them were predicted as benign and 12 for possibly damaging. Not surprisingly, the high frequency (> 90%) mutations were all predicted as benign. It is worth noting that there were 5 mutations (I116T, L221I, Y128F, G228V and P299S) predicted as probably damaging and interestingly, all these 5 mutations were mostly or totally distributed in Africa region, with the occurrence frequency ranging from 0 to 28% worldwide.

Furthermore, the three dimensions (3D) structure of full-length 3D7 PfTRAP was predicted and modeled by using I-TASSER server and displayed in YASARA application (Fig. 4). Based on the predicted structure of TRAP, the changes in free energy difference before and after the mutations (ΔΔG) were calculated, and the result shows that these mutations L49V, R285G, R285S, P299S and K421N would lead to the destabilization because of the obviously increased free energy difference (ΔΔG > 1). Furthermore, the spatial conformational changes of the twelve relatively common (> 10% globally) mutations at T cell epitopes were predicted and presented in Fig. 5. According to it, 7 amino acid substitutions (E46Q, I116S, K119R, S123N, Y128F, L287P and R290W) were identified as would lead to the breakage of hydrogen bonding, which would further impair the stability of the protein structure after these mutations (Fig. 5).