Background

Malaria, the cause of 627,000 of deaths in 2020, is concentrated in Africa, with the majority of cases being caused by Plasmodium falciparum [1]. According to World Health Organization (WHO), malaria-control programmes are expected to rely on test-and-treat strategies, where the effectiveness of diagnosis is essential to ensure a prompt treatment [2]. In this regard, rapid diagnostic tests (RDTs) are becoming increasingly popular as they provide a simple and cheap diagnosis [3].

RDTs are based on detection of specific parasitic proteins [4], and the majority of RDTs for P. falciparum rely on the detection of P. falciparum histidine rich protein-2 (PfHRP2), which reports cross-reactions with P. falciparum histidine rich protein-3 (PfHRP3) [5, 6]. PfHRP2 is the preferred protein for RDTs due to its stability, specificity and abundant expression during the blood stage [7].

The pfhrp2 and pfhrp3 genes comprise two exons, namely exon 1–2, which contains signal and cleavage sequences, and exon 2, which is the encoding sequence [8]. Deletions in pfhrp2 and pfhrp3 could be the cause of misdiagnosis and of false-negative P. falciparum results in HRP2-based RDTs [9, 10]. This could have an impact on delay or lack of treatment, with consequences for both the patient and for the maintenance of malaria transmission [11, 12]. In addition, pfhrp2 and pfhrp3 are highly variable genes with tandem repetitions [13, 14].

Variations in amino acid repeat motifs have also been reported to be a factor that may modify the effectiveness of HRP2-based RDTs [15, 16]. In this regard, although the main epitopes of PfHRP2 have been identified previously [17], the protein structures of PfHRP2 and PfHRP3 are still unknown, which complicates prediction of the epitope availability or the effect of polymorphisms [18, 19].

The dynamics of the evolution of these genes is not well understood. Thus, although there is evidence of selective pressure on pfhrp2 and pfhrp3 [20, 21], including in Ethiopia [22], multiple origins of parasites with deletions have also been reported suggesting spontaneous and multiple deletion events [23, 24].

Ethiopia is the most populous country in eastern Africa and reported approximately 260,000 malaria cases in 2019, with 5626 deaths [25]. However, it is considered to be a country in the process of consolidating malaria control, with this disease being heterogeneously distributed [26]. Amhara has a low-moderate malaria transmission, with a Plasmodium sp. prevalence of 0.8% in 2015 [27].

HRP2-based RDTs were implemented in Ethiopia in 2004 and have been used since 2010 as evidence for treatment as part of a test-and-treat policy [28]. Since then, however, deletions in pfhrp2 and pfhrp3 genes have been reported in some regions of Ethiopia [22, 29, 30], and neighbouring countries, such as Kenya [31] and Eritrea [10], thereby jeopardizing the efficacy of test-and-treat policies and efforts made by National Malaria Programmes in their implementation. As such, an increased understanding of the status of these genes and their genomic dynamics could help us to better understand their dynamics and design future guidelines. Moreover, there are no studies in Ethiopia focussing on the structural diversity of HRP2 and HRP3 proteins.

In light of the above, the aims of this study were to characterize the deletion prevalence and patterns of exons 1–2 and 2 of the pfhrp2 and pfhrp3 genes in the Amhara region and to study their genetic diversity, using the parameters diversity and haplotype networks, along with the amino-acid sequence variations in exon 2. The predicted sensitivity of HRP2-based RDT was also assessed using a theoretical model and the frequency and localization of epitopes on predicted protein structure models. Also assessed was the possible force driving selection and the evolution of these genes by performing a phylogenetic and genetic evolution analysis.

Methods

Sample collection and study area

The study was conducted in the Amhara region, Bahir Dar Department, Ethiopia (East Africa). During the malaria transmission season in 2014 (September to December), samples from symptomatic malaria infections from outpatients at health facilities were collected. Febrile or otherwise malaria-symptomatic patients aged ≥ 16 years with positive P. falciparum diagnosis by microscopy were eligible to participate. Samples were collected at four health facilities: Bahir Dar city, the principal city of the region near to Tana Lake (population: 170,000); Tis Abay, a town surrounded by farming areas close to the Blue Nile Falls waterfall; Meshenti, a medium-sized city to the southwest of Bahir Dar city; and, Zenzelima, a rural and surrounding area to the northeast of Bahir Dar city (Fig. 1). This region has a sub-tropical highland oceanic climate, with temperatures ranging between 18 and 24 °C (an average temperature of 20 °C), and abundant precipitation concentrated in the rainy season (rainfall around 1850 mm per year).

Fig. 1
figure 1

Map of collection sites in Ethiopia. The red dots correspond to the location of the four cities’ health centres where samples were collected

Full size image

Blood samples were taken on Whatman 903TM paper (GE Healthcare Bio-Sciences Corp.) by finger prick and stored in double zip-lock plastic bags with silica gel at − 20 °C, and while they were being analysed they were stored at 4 °C. They were then transported to the National Centre for Tropical Medicine, Institute of Health Carlos III (Madrid, Spain) for molecular and genetic analysis.

DNA extraction and molecular diagnosis of Plasmodium sp.

DNA was extracted from dot blood spots on filter paper using the chelex–saponin method [32]. A 5-mm diameter punch containing 10 μL of blood was used [33]. Malaria molecular diagnosis was performed on all samples using nested multiplex PCR targeting the 18S small sub-unit ribosomal RNA (ssRNA) to confirm P. falciparum infections [34, 35]. Amplification of other P. falciparum genes (pfdhfr, pdhps, pfmdr1, pfcrt) was performed to ensure that the samples were free of PCR-inhibiting elements and thus confirm that when the pfhrp2 and pfhrp3 genes did not amplify it was due to a true deletion and not inhibition.

pfhrp2 and pfhrp3 gene deletions

Amplification of exon 2 of pfhrp2 and pfhrp3 was performed by semi-nested PCR following a previously described procedure [10] and amplification of exon 1–2 of pfhrp2 and pfhrp3 was performed following a protocol with slight modifications [36], such as a lower annealing temperature (55 °C) and more cycles for both PCRs (35 cycles). A summary of the workflow is described in Fig. 2. PCR amplicons were separated and visualized on a 2% agarose gel stained with Pronasafe (Pronadisa SA, Spain). If no amplification was observed, the PCR experiment was repeated, and, only for exon 2 of both genes, if the result was still negative, it was repeated a third time with a lower elongation temperature (60 °C) [37]. If there was no amplification in any of the experiments, the maintenance of DNA quality was confirmed by nested multiplex PCR, and if amplification was observed, the result was considered to be a deletion event.

Fig. 2
figure 2

Methodological flow scheme. Microscopy diagnosis was confirmed by nested multiplex PCR, distinguishing P. falciparum, Plasmodium vivax, Plasmodium Ovale, and Plasmodium malaria. Four independent PCRs were run for the P. falciparum samples to detect deletions in exon 1–2 and exon 2 of pfhrp2 and pfhrp3. The deletion of any exon was confirmed by an absence of amplification after three PCR repetitions and the confirmation of maintained DNA quality. Samples that could not be confirmed for DNA quality were excluded from the analysis for that gene. Subsequent data analysis for the combined results was performed only with those samples included in the independent analysis for both genes. Additionally, a sub-sample lacking the deletion for exon 2 of both genes was sequenced and the genetic diversity and variation in amino-acid sequences analysed. Finally, a sub-sample of these sequences was used to predict the protein structure model and to locate the epitopes in the structure

Full size image

Samples with deletion in exon 1–2 but not in exon 2, or vice versa, were also amplified using the forward primer for exon 1–2 and the reverse primer for exon 2, and the previously reported PCR protocols for exon 2 of both genes. Samples that did not amplify with that PCR were assumed to have had some part of the exon that did not amplify also not previously deleted. Plasmodium falciparum clone 3D7 was used as a positive control in all PCRs, with clone Dd2 as negative control for the pfhrp2 gene amplification. This clone contains a pfhrp2 gene deletion. Similarly, clone HB3, with pfhrp3 deleted, was used as negative control for the pfhrp3 gene amplification.

Genetic diversity of exon 2 from pfhrp2 and pfhrp3 genes

Sequencing and genetic analysis of exon 2 from pfhrp2 and pfhrp3

A random sub-sample of PCR-positive samples for both genes (pfhrp2 and pfhrp3) was sequenced. The PCR products were purified using Ilustra Exoprostar 1-step (GE Healthcare Life Sciences) following the manufacturer’s instructions. Samples were sequenced in both directions using forward and reverse primers for the second PCR of exon 2 at a concentration of 6 pmol/μL, sequencing using a standard dye terminator in an ABI PRISM 3730 XL Analyser (Big Dye Terminator v3.1 Cycle Sequencing kit).

Bio Edit Sequence Alignment Editor Software v7.1.3.0 was used for sequence editing and solving undetermined sites. Mega7 v7.0.26 was used for multi-sequence alignment with ClustalW tool, a progressive alignment method recommended for data sets with varying degrees of divergence [38]. Polymorphism was then analysed by determining the diversity of haplotypes, nucleotide diversity (π) and genetic diversity (Θ) using DNAsp v6.12.03. Diversity was analysed through pairwise comparison, where gaps/missing presented in each comparison were excluded but only for that comparison. Haplotype networks were constructed using the Median Joining Network method, which is phylogeographical analysis with lack of rooting, in PopART v.1.7. The randomness of the distribution of haplotypes between populations was assessed using the Exact test for population differentiation based on Haplotype Frequencies calculated in Arlequin v3.5.2.2. Neutrality tests, namely Tajima’s D, which is based on allele frequencies, and Fu and Li’s D*, which analyses synonymous and non-synonymous sites, were calculated using DNAsp v6.12.03 as a means of detecting selection.

Phylogenetic analysis

The evolutionary relationships of exon 2 pfhrp2 and pfhrp3 were assessed for the same sub-sample of nucleotide sequences sequenced for genetic analysis using the Datamonkey server [39]. First, genetic recombination was evaluated using the Genetic Algorithm for Recombination Detection (GARD) tool [40]. GARD is a pre-processing step for selection inference because it groups unique phylogenetic histories for each recombination block. The Branch-site Unrestricted Statistical Test for Episodic Diversification (BUSTED) tool was then applied to detect gene-wide positive selection at at least one site on at least one branch [41].

To assess phylogenetic relationships combined with clustering by geographical location. Phylogenetic trees were performed using Bayesian Inference by MrBayes v3.2.7a and Maximum Likelihood by Mega7 v7.0.26 as inference methods to estimate evolutionary distances between all sequences. Both methods used substitution models GTR + T + I, where missing data or alignments gaps were not considered, so loci with gaps or missing data were eliminated of the analysis. In this analysis, 33 homologous sequences of each gene from countries bordering Ethiopia and other regions of Sub-Saharan Africa were included, published in GenBank (Additional file 1).

Study of amino acid sequences and protein structure

Variation in amino acid repeats patterns

All the sequenced samples were used to amino acid repeats characterization with a number according to Baker et al. [15] and recent repeats characterized in Kenya [42]. When two or more amino acid repeats overlapped, only the largest repeat was counted. Analysis of the frequency and organization of amino acid repeats was carried out with a custom code using Python 3.9 Software (Additional file 2).

In silico translation of exon 2 of pfhrp2 and pfhrp3

Prediction of the secondary and tertiary structure of the protein was performed using two different platforms to obtain more consistency in the results: (1) RaptorX (http://raptorx.uchicago.edu); and (2) Protein Homology/analogy Recognition Engine v2.0 (PHYRE2) (http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi?id=index). Both analyses applied a hierarchical approach with a combination of derived-template and de novo protein prediction to generate full-chain models. Firstly, amino acidic sequences were submitted in PHYRE2 for a batch-processing, then sequences with higher quality and confidence were selected for following analysis, structure prediction by RaptorX. RaptorX showed the best results for secondary and tertiary structures. Model quality was estimated using the RMSD (root mean square deviation) value and the model with lowest RMSD was selected as this indicates the deviation observed between the native and predicted positions [43]. Additionally, PHYRE2 was used for structure-based function prediction, according to the alignments.

Description and localization of epitopes in the protein

The presence of 11 linear epitopes, previously described as the epitopes for optimal monoclonal antibodies potentially used in HRP2-based RDTs, as information about mAbs used in commercial RDT kits is unavailable, was confirmed in protein sequences and the frequency of each epitope per sequence was analysed. The median repetition of each epitope, and its frequency among the sequences, was calculated. The PyMOL2 molecular graphic system v.2.5 (https://pymol.org/2/) was used to locate the epitopes in the predicted tertiary structure and APBS [44] and PDB2PQR [45] were used to calculate the Solvent Accessible Surface area.

Results

Deletions in pfhrp2 and pfhrp3 genes

A total of 354 filter paper samples were collected from symptomatic patients with malaria (diagnosed by microscopy) from Tis Abay (N = 200), Bahir Dar (N = 81), Zenzelima (N = 35), and Meshenti (N = 38). Of these, 52 samples were excluded due to low DNA quality. Plasmodium falciparum was confirmed in most of the samples (294/302; 97.35%), P. vivax was identified in 2.65% (8/302) and no mixed infections were detected.

Plasmodium falciparum samples were used for amplification of exon 1–2 and exon 2 of pfhrp2 and pfhrp3 genes. After the confirmation of deleted samples (i.e., deletion in pfhrp2 and/or pfhrp3 genes) 280 samples were included for the pfhrp2 analyses and 238 samples for the pfhrp3 analyses (Fig. 3). Negative samples that could not be confirmed to maintain sufficient DNA quality were excluded from the analysis (n = 5 for pfhrp2 analysis only, n = 47 for pfhrp3 analysis only, and n = 9 for both pfhrp2 and pfhrp3 analysis). For joint outcomes of both genes, 219 samples were included as they were only samples that were not excluded from any of the analyses.

Fig. 3
figure 3

Summary of results of molecular analysis of pfhrp2 and pfhrp3. Figure represents the different results in the independent experiments for exon 1–2 and 2 of both genes. Deletions samples: samples that did not amplify in the first PCR for the exon. Confirmed deletions: samples that did not amplify in three independent PCR experiments for the same specific gene, process called ‘repetition pfhrp2/3 PCR’ in the figure. Presence: samples that amplified at least one for a specific exon

Full size image

Frequency of deletion in exons 1–2 and 2 of pfhrp2 and pfhrp3 genes

Deletions were detected in exons 1–2 and 2 for both genes, with a higher frequency of deletions in exon 1–2 (19.29% (54/280) for pfhrp2 and 68.49% (163/238) for pfhrp3) than in exon 2 (4.64% (13/280) for pfhrp2 and 18.49% (44/238) for pfhrp3). Candidate partial deletion was considered when only exon 1–2 or 2 have amplified but ‘partial deletions’ could not be confirmed. Full-length PCR was performed for partially deleted candidate samples, but only 5 of 51 of pfhrp2 and 14 of 123 pfhrp3 amplified with that PCR showing different fragment sizes than expected (Fig. 4).

Fig. 4
figure 4

Gel images of the full-length PCR of pfhrp2 and pfhrp3. Electrophoresis gel showed the result of the amplification for PCR combining exon 1–2 and exon 2, the expected fragment was around 900 base pairs as it could be observed in the positives controls placed at the end of the gel: 3D7, Dd2 and HB3. Negative controls for the first and second PCR were placed in the first two wells of the gel

Full size image

Pfhrp2 presented at least one deleted exon in 21.08% of isolates and 2.86% of them had the pfhrp2 gene completely deleted (any exon amplified) (Table 1). A high proportion of isolates presented some deletion in pfhrp3 (73.95%), with complete deletion of this gene being detected in 18.49% of isolates. There were significant differences in the frequency of exon 2 of pfhrp3 deletion by location (p-value < 0.05), but not for exon 2 of pfhrp2. Frequencies of exon 2 of pfhrp2 deletion are similar between locations but pfhrp3 deletion showed significant differences by location (Additional file 3).

Table 1 Deletion patterns for exon 1–2 and exon 2 of pfhrp2 and pfhrp3 genes
Full size table

Deletion patterns in pfhrp2 and pfhrp3

Overall, double-negative samples (pfhrp2 and pfhrp3 completely deleted) were found in 1.37% (3/219) of cases and both complete genes were identified in 30.59% (67/219) (Table 2).

Table 2 Frequencies of deletion patterns based on deletion/presence of exon 1–2 and exon 2 of pfhrp2 and pfhrp3 genes
Full size table

Genetic diversity of exon 2 of pfhrp2 and pfhrp3 genes

A total of 95 pfhrp2 and 79 pfhrp3 exon 2 P. falciparum sequences were analysed and translated successfully into amino acids (aa). These sequences were submitted to the GenBank database and the homology with the pfhrp2 and pfhrp3 genes of P. falciparum checked for further analysis. The size of the exon 2 sequence ranged from 400 to 750 bp in pfhrp2 and 350 to 700 bp in pfhrp3, with an average of 210 and 188 amino acids, respectively.

A total of 92 haplotypes were identified in the pfhrp2 sequences, with 93.68% of sequences being unique (89/95). All the haplotypes shared by more than one sample include only two sequences (Additional file 4). A total of 77 haplotypes were detected for the pfhrp3 sequences, with two haplotypes shared by two samples from different health centre. All haplotype sequences were submitted to the GenBank database (accession numbers: OL961136–OL961227 and OL961228–OL961304).

The haplotype analysis of pfhrp2 showed a higher diversity, in terms of haplotype distribution and frequency, than the haplotype network for pfhrp3, for which lower distribution for the different haplotypes were observed (Additional file 4). These results are consistent with the genetic diversity analysis, which revealed a higher nucleotide diversity for pfhrp2 (π = 0.158) than for pfhrp3 (π = 0.027) (Table 3).

Table 3 Genetic diversity of pfhrp2 and pfhrp3 by geographical location
Full size table

Test statistics for neutrality were applied to assess the hypothetical evolution of the genes. The neutrality tests for pfhrp2 were both negative (Fu and Li’s D* = –2.43 and Tajima’s Neutrality Test D = − 0.96), but only Fu and Li’s D* was statistically significant (p-value < 0.05). In contrast, the neutrality tests for pfhrp3 were both negative (Tajima’s D = − 2.23 and Fu and Li’s D* test statistic = − 6.05) and statistically significant (p-value < 0.05).

Phylogenetic and evolution analysis

The ClustalW alignment resulted in an alignment of 916 base pairs for pfhrp2 and 618 sites for pfhrp3.

Potential recombination breakpoints between the pfhrp2 and pfhrp3 sequences were detected by GARD analysis and were considered for a subsequent evolution analysis. BUSTED analysis found evidence for gene-wide episodic diversifying selection at at least one site on at least one branch for both pfhrp2 and pfhrp3, with a synonymous rate variation (ω) > 1. These results were confirmed by repeating the same analysis including homologous sequences from other regions of sub-Saharan Africa from GenBank.

The Bayesian inference phylogenetic tree showed an unclear clustering by health centre, country or region for both genes (Fig. 5). The Maximum Likelihood Inference tree showed similar results. Sequences from the same health centre tend to form small clusters but without being truly isolated. Similarly, sequences from countries bordering Ethiopia (Kenya and Sudan) were more closely related to sequences from the present study. However, sequences from other African regions are also relatively closely related to some sequences, thus indicating an unclear geo-clustering.

Fig. 5
figure 5

Bayesian inference phylogenetic tree for pfhrp2 (a) and pfhrp3 (b) sequences. The nucleotide sequences of exon 2 for both genes, with fragments ranging from 300 to 700 base pairs, were analysed. This analysis combined nucleotide sequences from this article, represented with a dot (n = 95 for pfhrp2 and n = 79 for pfhrp3) and sequences from bordering countries published in previous studies, represented with dots with different shapes (n = 33 for pfhrp2 and n = 33 for pfhrp3). Coloured dots indicate the geographical origin of the samples. Phylogenetic trees were performed using Bayesian inference

Full size image

Variation in amino-acid repeat patterns

Differences in the frequency and organization of amino-acid repeats were detected, with the most frequent amino acids in pfhrp2 and pfhrp3 being alanine (37.32%), histidine (36.16%) and aspartate (9.59%), and histidine (31.18%), alanine (30.19%), asparagine (9.55%), and aspartate (9.11%), respectively.

Variation in amino-acid repeats in pfhrp2

A total of 20 previously reported amino-acid repeat types were found, with repeat types 2 and 7 being observed in all isolates and with the highest average repetition frequency (type 2: 8.8-times and type 7: 4.3-times) (Table 4). The next most frequent repeats were type 3 (83%) and type 8 (90%), which appeared in the range 0–2-times per sequence, and type 4 (84%), for which the repetition times per sequence ranged from 0- to 18-times. New or modified repeat amino-acid motifs were detected and classified (Additional file 5). Organization of the amino-acid repeats in the sequences followed similar patterns, although with some variations (Additional file 6).

Table 4 Occurrence and number of amino-acid repeat types in pfhrp2 and pfhrp3 sequences
Full size table

Variation in amino-acid repeats in pfhrp3

A total of 15 previously reported amino-acid repeat types were found in the sequences (Table 4). Repeat type 16 was included in the majority of isolates (99%) and was the most repeated type in the sequences (average time: 11.15). The following more frequent types had markedly fewer repeats per sequence. For example, types 7 and 20 were found in 95% and 88% of samples, respectively, but only one time per sequence. The organization of amino-acid repeats was similar for all samples and in agreement with previously reported results (Additional file 7).

Structure prediction and epitope localization

In silico structure prediction

The secondary and tertiary protein structure was predicted for a sub-set of samples (n = 22), selected based on sequence quality, using RaptorX. For most samples, the majority of the secondary structure was a helix, with some coils placed principally in intermediate positions. Batch Phyre2 structure-based function prediction analyses were carried out for 96 samples, with a significant result being found for only 10 samples: significant results are needed for structure prediction and homology. Among these, the main structure matches were with proteins with a hydrolase function (Additional file 8).

Epitopes of HRP2 and HRP3 targeted by monoclonal antibodies

Seven of the 11 major antibody epitopes were identified in more than 84% of pfhrp2 sequences (n = 81) (Table 5). Two epitopes, namely 3A4 and C1-13, were present in all sequences with the highest median frequencies: 14- and 13-times per sequence, respectively. Epitopes 2G12-1C12, N7 and PTL-3 were present in 99, 99 and 95% of sequences, respectively, but at lower frequencies (2–4-times per sequence).

Table 5 Frequency of major epitopes in pfhrp2 and pfhrp3 exon-2 targeted by MAbs in HRP2-based RDT
Full size table

Among pfhrp3 sequences, epitope 3A4 was the most common (70%), but with only a low frequency (median = 1), followed by 1E1-A9, which was found in 65% of sequences (median = 1).

The eight tertiary structure models were used to locate the main epitopes in the structure. These were then compared with the solvent-accessible surface area, which has a mainly negative electrostatic potential. Epitopes 3A4 and C1-13 were found to be the most accessible as they were widely distributed over the protein surface, whereas other epitopes were located in only a specific region of the surface (Fig. 6). Additionally, epitopes recognized by 3A4 were also found in the PfHRP3 models studied, but with less accessibility (Fig. 7).

Fig. 6
figure 6

Secondary structure, solvent surface area and map** of epitopes in the PfHRP2 protein structure models. Model of structure was assessed by RSMD measured with Å. Solvent surface area is coloured gradually according to electrostatic potential from − 5 kT/e (red) to + 5 kT/e (blue). NA: Not applicable due to absence of epitope in the sequence

Full size image
Fig. 7
figure 7

Secondary structure, solvent surface area and map** of epitopes in the PfHRP3 protein structure models. Model of structure was assessed RSMD measured with Å. Solvent surface area is coloured gradually according to electrostatic potential from − 5 kT/e (red) to + 5 kT/e (blue). NA: Not applicable due to absence of epitope in the sequence

Full size image

Discussion

An understanding of pfhrp2 and pfhrp3 gene deletions and genetic diversity is key to the design of future interventions in RDT-based malaria test-and-treat strategies. This study describes the frequency of deletions and unique data regarding the sequence variation of pfhrp2 and pfhrp3 from the Amhara region in Ethiopia.

pfhrp2 and pfhrp3 deletions

This study showed that the detected frequency of the exon-2 deletion for pfhrp2 in the Amhara region of Ethiopia (4.64%) is worryingly close to the threshold (5% at the lower end of the 90% CI for false negatives by HRP2-based RDT) established by the WHO to change the use of HRP2-based RDTs [46]. Although this study did not use results from RDT, a higher prevalence of deletion among HRP2-based RDT false-negatives than for samples diagnosed by microscopy, where false negatives and true positives by HRP2-based RDT could be mixed, is to be expected. Taking this into account, the higher frequency of deletion (11.5%) reported for RDT false-negatives in 2018 in Amhara suggests a gradual increase [22]. In contrast, a higher frequency of deletion has been reported in other regions of Ethiopia, for example 18% among samples including true positives and false negatives and 42% among HRP2-based RDT false-negatives in the Assosa region, with high/moderate transmission [29], and 100% among general samples in Oromia, with low transmission [30]. The differences between reported prevalence could be related to the association between deletion frequency and low malaria transmission, where it is less probable to find a mixed infection of parasites with and without pfhrp2 that masks the deletion [20]. Another explanation for the wide range of deletion prevalence reported could be the differences in genoty** approach, for example, a recent study in Ethiopia only analysed samples with low or without HRP2 signal by multiplex bead assay [47]. That makes it essential to generalized technical strategies across studies [48]. Potential false negatives caused by pfhrp2 deletions could be avoided by cross-reaction with pfHRP3 as this can be detected by HRP2-based RDT in the absence of pfHRP2 [4]. However, in the light of this study’s results, in Ethiopia this is unlikely to be possible due to the high frequency of pfhrp3 deletions (18.5%) and isolates with deletions in both genes.

With regard to patterns of deletions, the findings are in agreement with the literature, which shows that deletions in exon 1–2 and in pfhrp3 are more common than in exon 2 and in pfhrp2, respectively [49, 50]. This may suggest the possibility of sequential deletion, starting with deletion in exon 1–2 but ending with complete deletion [51]. But, deletions in only one exon, called in this manuscript ‘candidate partial deletion’, could not be confirmed and could be simply due to worse amplification in exon 1–2. However, deletions in different fragments of these genes have also been reported, suggesting that the full gene is not always deleted [22]. Evolution and selection of pfhrp gene deletions will be closely associated with their fitness cost [52]. Although this remains to be clarified, previous studies in vitro have suggested the selection of strains without deletions and slower growth in strains with deletions [21, 53]. Moreover, in Ethiopia the extended used of combination HRP2/Pv or Pan—LDH-based RDT [54], due to high prevalence of P. vivax, could weaken the selection pressure of pfhrp-deleted parasites, unlike in other sub-Saharan countries.

Genetic diversity and evolution analysis of exon 2 from pfhrp2 and pfhrp3 genes

The dynamics of the spread or selection of pfhpr2-deleted populations remains unclear. However, as HRP2-based RDT use has increased exponentially in recent years, a selective pressure for parasites with pfhrp2 deletions could be expected [31]. Indeed, linkage-specific clonal expansion of populations with pfhrp2 deletions has been reported [55]. In contrast, the results of the BUSTED test showed diversifying expansion in the present study. This diversity, combined with a lack of clear clustering, could be related to an increased diversity step, when gene deletions occur, followed by clonal expansion promoted by selective pressure. In fact, recent strong selection of pfhrp2-deleted parasites have been reported from Ethiopia in a study carried out some years later, when this clonal expansion, boosted by test-track-treat policies, could be expected to happen [22]. Multiple spontaneous origins of deletions in these genes, which could be part of the diversifying expansion, have been reported previously [22, 56].

Variation in amino-acid repeat patterns and its implication for the effectiveness of HRP2-based RDT

The role of a variation in amino-acid repeats in RDT is unclear. Previous studies have reported both an influence on RDT results, increasing its sensitivity or causing misdiagnosis [57, 58], or a lack of association [59]. Parasites could maintain the gene but undergo changes in some specific amino-acid repeats, thereby becoming undetectable [60, 61]. Some repeat types could have more influence in HRP2-based RDT results, but there is not found a clear association [13]. This suggests the need to monitor the effectiveness of RDT.

The pfhrp2 sequences showed 16 previously described repeat types, including some variations. The recent increase in reported variations and new domains is consistent with a diversifying expansion [40, 62]. In line with other studies, the most common repeats were types 2 and 7, the frequency of which markedly influences RDT detection [15, 63].

Pfhrp3 sequences showed less diversity in terms of amino-acid sequences than pfhrp2, as reported previously [60]. Similar to previous studies, the most frequent and repeated type was type 16 [64, 65], and type 17 showed a marked variation, as in Kenya [61], thus suggesting a high antigenicity and that it may be useful for RDT detection.

Protein structure prediction and epitope localization

The relative similarity between the occurrence and organization of amino-acid repeats, either with geographically close parasites or those from other continents, contrasts with the high diversity detected at a nucleotide level. However, this could imply the temporal persistence of PfHRP2 epitopes despite the genetic diversity [61]. For instance, the results from the Amhara region are similar to those reported previously in other countries for the 12 available PfHRP2-specific MAbs. Although information about the antibodies used in commercial RDTs is unavailable, it is reasonable to assume that many of these epitopes are targeted [17]. The epitopes recognized by MAbs 3A4 and C1-13 were the most frequent [62, 66], and the epitope for 3A4 was also the most common in pfhpr3 sequences [24]. Additionally, the epitope recognized by C1-13 was identified for the first time in the pfhrp3 sequence in this paper. This detection of the usual epitopes increases the probability of parasite detection, even though they were present at only low frequencies.

Protein-structure models showed broad surfaces with solvent accessible surface areas on which the majority of linear epitopes were found and shown to be topographically available. In contrast, few of the linear epitopes were detected among HRP3 protein sequences, thus suggesting that, in this protein, the epitopes recognized by HRP2-RDT could be conformational rather than linear.

Pfhrp2 and pfhrp3 deletions might have a different impact depending on malaria endemicity. Low malaria diversity settings, which are related to a low prevalence of malaria, could lead to an easier selection of deleted parasites [67]. As both the competition and recombination between different strains of parasites are lower, deletion will be easier to maintain in following generations. The exception to this is parasites with both anti-malarial resistance and deletions, as they could be more easily selected irrespective of the conditions [20]. As such, surveillance of pfhrp2 and pfhrp3 is particularly important in settings in which malaria is decreasing as a result of prevention and control efforts, such as Ethiopia [68, 69]. Although HRP2-based RDTs are still considered useful for malaria diagnosis in endemic regions [70], WHO guidance recommends countrywide change when the threshold was exceeded in any region. Meanwhile, combination of pLDH-RDTs with targets other than HRP2 are preferable [9, 71].

There are some limitations to this study. For example, samples were collected based on microscopy smears positive for P. falciparum but RDT results or parasite densities were not available, thus limiting the power to detect the effect of deletions in malaria diagnosis. Furthermore, representativeness and comparisons between the different health centres was not possible epidemiologically due to the low number of samples. Also, candidate partial deletions could not be confirmed by genetic analysis. Moreover, genetic analysis of repetitive genes required the exclusion of large number of sites affecting genetic diversity results. However, the study still provides a more in-depth genetic characterization of pfhrp2 and pfhrp3 genes and their amino-acid composition. Subsequent studies should assess the implications of variations in sequences or epitopes and their effect on RDT results. Also, full-length of ‘partially deleted’ samples will provide insightful information about this partial deletions.

Conclusion

Pfhrp2 and pfhrp3 deletions have been identified in P. falciparum isolates from Amhara region in Ethiopia. The high frequency of exon 2 pfhrp2 deletion and the detection of isolates with double deletions in both exon 2 of pfhrp2 and pfhrp3 could jeopardize malaria-control strategies based on diagnosis with HRP2-based RDTs.

The high genetic diversity and high polymorphism of the amino-acid sequences found in pfhrp2 and pfhrp3 may have an impact on RDT results and malaria diagnosis, especially given the relatively low frequency of repetition types 2 and 7 detected.

This study also provides insights into the localization of epitopes in the protein structure of pfhrp2 and pfhrp3, with the use of combinations of MAbs 3A4 and C1-13 being recommended to increase the reactivity of HRP2-based RDTs. However, a better understanding of protein structure could be useful for the design and evaluation of monoclonal antibodies for RDT.

In light of all these results and the malaria situation in Ethiopia, this study highlights the importance of more surveillance studies to increase our understanding of the dynamics of these genes.