Ethical approval for the Ugandan trial and follow-up study were obtained from the ethical institutional review committees of Med Biotech Laboratories (Ref# IRB-00003990-MBL-BIOMEDICAL, IRB-00003995-MBL-BIOMEDICAL), Uganda National Council for Science and Technology (Ref# HS635, HS866), Osaka University (Ref# 20-3, 287), and Research Foundation for Microbial Diseases of Osaka University.

Approval for the Burkina Faso trial was obtained from Comité d’Éthique pour la Recherche en Santé du Burkina Faso (Ref# 2014-12-144), Comité Institutionnel de Bioéthique du INSP/CNRFP (previous name: CNRFP) (Ref# N°2014/071/MS/SG/CNRFP/CIB, N°2016/000008/MS/SG/CNRFP/CIB), Agence Nationale de Régulation Pharmaceutique (ANRP, previous name: Direction Générale de la Pharmacie, du Médicament et des Laboratoires [DGPML], Ref# N°2015_658/MS/CAB), Scientific Committee/Institutional Review Committee of the Research Institute for Microbial Diseases (Ref# 26), Osaka University (Ref# 574); and London School of Hygiene and Tropical Medicine Research Ethics Committee (Ref# 9175).

Consent, either by signature or thumbprint, was obtained from the children and/or children’s parents or guardians prior to sampling and other trial-related procedures. In Uganda, for subjects 8–17 years old, assent by the child took precedence over consent from the parents or guardians.

All studies were conducted in compliance with the protocol, International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH), Good Clinical Practices, the Declaration of Helsinki 2013, and country-specific laws and regulations.

The details of the randomized, single-blind, safety and immunogenicity phase 1b trial (Current Controlled trials ISRCTN71619711 https://doi.org/10.1186/ISRCTN71619711) and follow-up study in Lira, northern Uganda have been previously described (Palacpac et al., 2013; Yagi et al., 2016). The clinical trial site, Lira Medical Centre, is in a region with intense transmission and bimodal rainfall pattern from April to May and September to October (Proietti et al., 2011). Blood samples and data on malaria incidence were obtained from healthy children and young adults aged 6–20 years old (n = 84). Participants in 3 age cohorts (6–10, 11–15, and 16–20 years old) were randomized to be administered twice with either half- (0.5 mL, n=11) or full- (1.0 mL, n=6) dose BK-SE36 or saline (n=6). Subcutaneous vaccination was performed 21-days apart. During the follow-up study (130–440 days post-second vaccination), 50 additional children (as unvaccinated controls) were recruited. Recruitment of the additional control group used the same inclusion and exclusion criteria as in the clinical trial and was age-, gender-, and locality-matched as much as possible. The paired control participants visited the trial center on the same scheduled day as trial participants to undergo assessments for vital signs, physical examination, monthly questionnaire, and blood smear. Blood samples for filter blots, thick and thin blood smears, were obtained at 28-day intervals (active surveillance) or whenever the child was sick (passive surveillance). In this study, any blood smear-positive samples (any parasitemia) were used for analyses.

The Burkina Faso trial was a double-blind, randomized, controlled, age de-escalation safety and immunogenicity trial and follow-up study conducted in Banfora, south western Burkina Faso (Pan African Clinical Trials Registry: PACTR201411000934120; https://pactr.samrc.ac.za/TrialDisplay.aspx?TrialID=934 ) (Bougouma et al., 2022). The trial site, Unité de Recherche Clinique de Banfora, is in a region where transmission occurs throughout the year with a peak for clinical malaria occurring within the 4 months (June–September) of the rainy season of May–November (Tiono et al., 2014). Samples and data of clinical malaria episodes were obtained from healthy children aged 12–60 months (n = 108). Participants in two age cohorts (12–24 and 25–60-months-old) were randomized for administration of Synflorix^® (0.5 mL, control arm, n = 18); or BK-SE36 (1.0 mL) via either the subcutaneous route (n = 18), or the intramuscular route (n = 18). Two administrations were performed 28 days apart, and a third dose was administered at week 26 (Day 182). The control group was vaccinated with Synflorix^® at Days 0 and 182, and with saline at Day 28 to comply with the manufacturer’s recommendation of at least a 2-month interval between the 2 primary Synflorix^® doses while ensuring that the trial was performed in a double-blinded manner. Clinical malaria episodes, defined as ≥5000 parasites/µL + tympanic temperature ≥38°C, from Day 56 (4 weeks after dose 2) until the final visit (Day 477, 42 weeks after dose 3) were assessed. Two thick and thin blood smears were prepared for malaria diagnosis and blood samples spotted onto Whatman™ 903 Protein Saver Card (GE Healthcare Life Sciences, MA, USA) were dried and stored until use.

The BioRobot EZ1 DNA investigator kit (QIAGEN, Hilden, Germany) was used to extract parasite DNA from the filter paper blood spots. Four pieces of a 3-mm-diameter circle, corresponding to a 20-μL volume of blood, were punched out from 1-2 blood spots for downstream reactions. The extracted DNA was resuspended in 50 μL TE solution, and stored at -20°C until use.

Exon regions II-IV of the SERA5 gene, covering approximately 3.3 kb, were amplified using specific primers sera5-F1 and sera5-R1 ( Supplementary Table S1 ). Exon I encoding the signal peptide was not analyzed because of sequencing difficulties. Amplification was carried out in a 25-μL reaction mixture containing 0.4 μM each of forward and reverse primers, 0.4 mM each of dNTP, 0.5 units of KOD FX Neo polymerase (TOYOBO, Osaka, Japan), 12.5 μL of 2x PCR buffer, and 1 μL of genomic DNA solution. PCR conditions were as follows: 95°C for 2 min, 35 cycles of 95°C for 15 sec, 59°C for 30 sec, and 68°C for 2 min. A 2-μL aliquot of the PCR product was used as template for a second PCR amplification in a 25-μL reaction mixture using the primers sera5-F2 and sera5-R2 ( Supplementary Table S1 ) under the same thermocycler conditions. In case of failure with the first set of primers, other primer pairs were also tested: sera5-F1 and sera5-R1 for the 1st PCR and sera5-F3 and sera5-R2 for the 2nd PCR or sera5-F2 and sera5-R1 for the 1st PCR and sera5-F3 and sera5-R2 for the 2nd PCR. PCR products were purified using the QIAquick PCR Purification kit (QIAGEN) according to the manufacturer’s instructions. Purified DNA fragments were eluted in 30 μL TE. Optical density was measured with a NanoDrop (Thermo Fisher Scientific, Waltham, MA, USA) and the DNA concentration was adjusted to 0.026 μg/μL using TE. At this concentration, 1 μL was suitable for performing one sequencing reaction.

DNA sequencing was performed directly using the BigDye^® Terminator v3.1 Cycle Sequencing Kit and 3130xI Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). Sequencing primers were designed to cover target regions in both directions ( Supplementary Table S1 ). If inconsistencies were obtained after two independent amplifications, a third round of PCR/sequencing reaction was performed. Only isolates showing a single genotype infection, without overlapping peaks on the electropherograms, were used for further analysis. To compare the nucleotide diversity of SERA5 with other antigens and housekeeping genes, the sequences of apical membrane antigen 1 (ama1), circumsporozoite protein (csp), Ca2+-transporting ATPase (serca), and adenylosuccinate lyase (adsl) genes were determined using the Ugandan isolates. The PCR and sequencing strategies were the same as those used for sera5 except that the PCR conditions were adjusted to be suitable for each gene. The PCR sequencing primers are listed in ( Supplementary Table S1 ).

The newly determined nucleotide sequences in this study have been deposited in the DNA Data Bank of Japan (accession nos. LC580441– LC581218). Samples from a previous study were used to compare polymorphisms found in Africa. The Tanzania and Ghana isolates were described previously (Tanabe et al., 2010). Briefly, Tanzania blood samples (n = 55) were collected from residents of Nyamisati village in the Rufiji River Delta, eastern coastal Tanzania in February and March 1993, 1998, and January 2003. Asymptomatic donors had a mean age of 14.2 years (range, 1–78), 16.8 years (range, 1–63), and 13.8 years (range, 10–19) in 1993, 1998, and 2003, respectively. Ghana isolates (n = 33) were collected during malaria surveys in 0–15-year-old children from three villages near Winneba (Okyereko, Mpota, and Apam), a western coastal region, in November 2004. The accession numbers of sera5, ama1, csp, serca, and adsl are summarized in Supplementary Table S2 .

The nucleotide sequences of sera5 were aligned using CLUSTALW implemented in GENETYX^® ver. 15 (GENETYX Corporation, Tokyo, Japan) with manual corrections. According to our previous analyses (Tanabe et al., 2012), the SERA5 sequence was categorically divided into an octamer repeat (OR) region, serine repeat (SR) region, and non-repeat regions (2,562 bp) ( Figure 1 ). The number of haplotypes, haplotype diversity (Hd), and nucleotide diversity (θ _π) were calculated using DnaSP v5.10.01 (Librado and Rozas, 2009). The difference between the numbers of synonymous substitutions per synonymous site (dS) and nonsynonymous substitutions per nonsynonymous site (dN) was calculated by the Nei and Gojobori method (Nei and Gojobori, 1986) with Jukes and Cantor correction as implemented in MEGA X (Kumar et al., 2018). The statistical significance of the difference between dN and dS was estimated with MEGA Z-test. If dN was greater than dS, positive selection was predicted. Genetic differentiation of SERA5 among the parasite population was examined using Fst, the Wright’s fixation index (Wright, 1965) of inter-population variance in allele frequencies. Pairwise Fst between parasite populations was calculated using Arlequin v3.5 (Excoffier and Lischer, 2010). These programs had built-in statistical analyses which were performed automatically, however, when statistical analyses were not supported, Mann-Whitney test for differences between two independent groups, the Kruskal-Wallis test for differences between multiple groups, and the chi-square (χ2) test for proportions between group were used.

In the trial conducted in Uganda, of the 247 infection events recorded, sequence information from 172 infections was obtained (75 were mixed infections and excluded from further analyses). Parasitemia ranged from 16 to 151,680 parasites/µL and 4 samples had P. falciparum gametocyte. Of the 172 available sequences, 33 were from subjects who had only one infection and 66 were from participants who had at least two infections during the follow-up period. In summary, 77 isolates were from the BK-SE36 vaccine arm and 95 from the control group. The samples were collected from March 2011 to February 2012, covering 2 malaria transmission seasons.

For the Burkina Faso trial, of 78 clinical malaria events, sequence information was obtained from 54 cases (24 were mixed infections and excluded from further analyses). Parasitemia ranged from 6,527 to 253,900 parasites/µL. Of the 54 available sequences, 29 sequences were from subjects who had only one infection and 10 were from subjects who had at least two infections during the follow-up period. Twenty-nine sequences were from the BK-SE36 vaccine arm and 25 from the control group. Samples were obtained from July 2015 to January 2017, covering the rainy season of May–November. Each subject had a total follow-up period of 16 months.

P. falciparum isolates from the four African countries were used to compare polymorphisms in 3 antigen genes (sera5, ama1 and csp) and 2 housekeeping genes (serca and adsl) ( Figure 2 , Supplementary Table S3 ). As expected, the degree of polymorphism varied between antigen-coding genes and housekeeping genes. Notably, the level of polymorphism in all African isolates within each gene was similar. The Hd of the three antigen genes ama1, csp, and sera5 ( Supplementary Table S4A ) were extremely high (close to 1.0), demonstrating that most sequences were distinct from each other. The Hd of the concatenated housekeeping genes of adsl and serca (adsl+serca) was slightly lower than that of the antigen genes. In the antigen genes, the number of sequence variations was almost the same in both the nucleotide and amino acid sequences; however, in the housekeeping genes, the amino acid sequence showed far fewer variations compared to the nucleotide sequences. Nucleotide substitutions in the antigen genes mainly occurred at non-synonymous sites which resulted in amino acid changes, whereas in housekeeping genes, nucleotide substitutions tend to occur at synonymous sites without amino acid changes.

Nucleotide diversity was estimated as θ_π , the average pairwise nucleotide difference ( Figure 2A ). The θ_π of the antigen genes ama1 and csp were very high, whereas that of the antigen gene sera5 was much lower and comparable to the concatenated gene sequences of adsl+serca. Among the three antigen genes, ama1 and csp showed a signature of diversifying selection with a significant excess of dN over dS ( Figure 2B ). The high values of θ_π and dN in ama1 and csp indicate large differences between the sequences, and dN>dS is likely due to host immune evasion. In contrast, the values of θ_π , dN, and dS of sera5 were much lower than those of ama1 and csp. Although dN was significantly larger than dS in sera5 from the Ghana isolates, the differences between the amino acid sequences were small compared to either ama1 or csp ( Figure 2B ). For each of these genes, analyses for θ_π , dS, and dN were performed to compare Hd with ( Supplementary Table S4A ) or without regions containing insertions and deletions ( Supplementary Table S4B ). The comparison provides insight into the various processes involved in generating alleles in antigen genes. The excluded regions would refer to the ‘NANP’ repeat region in ama1; the eight-mer amino acid repeat units in csp; the serine repeat region, the 13-mer insertion/deletion region, and the 17-mer dimorphic region in sera5; and the asparagine repeat region in adsl+serca. When Hd of sera5 was analyzed using the same dataset which excluded insertions and deletions, the value of Hd was much smaller than that obtained using full-length sera5 ( Supplementary Table S4 ). The Hd of ama1 was similar, as there were no insertions or deletions in the gene. The Hd of csp was also similar despite the exclusion of NANP repeat regions. The lower Hd of sera when regions of insertions and deletions are removed suggests that sequence variations in sera5 are mainly introduced by recombination rather than by point mutation. In ama1, all polymorphisms are produced by point mutation, whereas in csp, both point mutation and recombination resulting in different numbers of NANP units act to generate sequence variations.

The consensus sequence of SE36 in SERA5 was inferred from 314 sequences obtained from four African countries: Uganda (n = 172), Burkina Faso (n = 54), Tanzania (n = 55), and Ghana (n = 33). The consensus sequence in African isolates (Af-cons) was aligned with the SE36 sequence based on Honduras-1 and two representative laboratory strains, 3D7 and FCR3. The BK-SE36 candidate vaccine was designed with reference to the SERA5 sequence of the Honduras-1 strain (Sugiyama et al., 1996). As shown in Figure 3 , there were several sequence mismatch regions, particularly between the vaccine-type SE36 variant and Af-cons.

At the N-terminal, of the 314 SERA5 sequences, there were 76 variations in the OR region ( Supplementary Figure S1 ). The number of octamer units varied widely: seven octamer units (23 OR variations) were detected in 196 SERA5 sequences across Africa (62.4%), eight octamer units (16 OR variations) were found in 47 sequences (15.0%), and six octamer units (8 OR variations) were found in 14 sequences (4.5%). In all four African countries, seven octamer units were the most frequently detected. The number of OR units detected from SERA5 sequences was not significantly different among the four countries (p > 0.05, Kruskal-Wallis test). Vaccine-type SE36 has six octamer units, made up of one Ib subunit; and an identical OR sequence was found only in 5 P. falciparum isolates (1.6%): 4 from Uganda and 1 isolate from Ghana ( Table 1 ; Supplementary Figure S1 ). The seven octamer units found predominantly in Af-cons contain two Ib subunits ( Figure 3 ).

The SR region has three distinct sequence components: a 13-mer insertion/deletion, stretch of serine repeats, and a 17-mer dimorphic region (Morimatsu et al., 1997; Tanabe et al., 2012). A total of 69 amino acid sequence variations was identified in the SR region ( Supplementary Figure S2 ). The 13-mer deletion in the SR region of the FCR3 strain ( Figure 3 ) was observed in Oceania and South America, but not in Africa (Tanabe et al., 2012). In vaccine-type SE36 and Af-cons, the 13-mer sequence contained a serine/proline variation at position 203: a proline residue was found in the vaccine-type and 3D7 strain but a serine residue was mostly found in all isolates from the four African countries. SERA sequences did not differ significantly among the four African countries in terms of serine/proline ratios at position 203 (p > 0.05, χ²-test) ( Table 2 ).

For the stretch of serine tandem repeats from which SERA5 was named, vaccine-type SE36 contains only two serine residues ( Figure 3 ). This stretch of serine tandem repeats was mostly deleted from the vaccine construct to improve the hydrophilicity of the protein and make it amenable for large-scale Good Manufacturing Practices (GMP) production (Palacpac et al., 2011). Among the 314 African isolates, the number of poly-serine residues varied from 5 to 43, with 21 repeats as the most common in all 4 African countries ( Table 3 ). The difference in the number of serine residues among the four countries was statistically significant (p < 0.05, Kruskal-Wallis test). Notably, the amino acids isoleucine, asparagine, glycine, and arginine were also found in the serine stretch in eight of the 314 African isolates ( Supplementary Figure S2 ).

Just after the poly-serine repeat is a 17-mer dimorphic region ( Figure 3 ) containing four major sequence variations (Tanabe et al., 2012): ‘VNPPANGAGSTPDAKKK’ (Type I), ‘ESLPANGPDSPTVKPPR’ (Type IV), and recombination forms ‘VNPPANGAGSTPDAKKR` (Type II) and ‘VNPPANGPDSPTVKPPR’ (Type III). Vaccine-type SE36 has the Type IV sequence, whereas all African isolates had Type I as dominant sequence ( Table 4 ). For the four African countries, the ratio of Type I to Type IV sequence was significantly different (p < 0.01, χ2-test); and notably, no Type II and Type III sequences were found in Burkina Faso isolates. There were also sequence variations in Type IV. Substitutions at position 241 of proline to leucine and position 244 of proline to leucine were found in one Ugandan and Tanzanian isolate, respectively ( Supplementary Table S2 , Supplementary Figure S2 ).

In non-repeat regions, amino acid mismatches were observed in Af-cons at five positions ( Figure 3 ). Those were at positions 159, 275, 308, 330, and 383 of the SERA5 amino acid sequence of P. falciparum strain 3D7. Among them, glutamic acid at position 159 and leucine at position 330 were also found in Uganda (n=1), Tanzania (n=1), and Ghana (n=1) isolates, although their frequencies were very low ( Figure 4 ). In the vaccine-type SE36, at positions 275, 308, and 383, isoleucine, serine, and lysine were found, whereas African isolates exclusively showed leucine, asparagine, and lysine, respectively.

According to the above results, there are multiple differences between the vaccine-type SE36 and the corresponding SERA5 sequence from African isolates. The occurrence of identical sequences to the vaccine-type variant was rare.

If BK-SE36 confers allele-specific immune response or shows efficacy targeting only vaccine-type parasites or closely related variants, then vaccinated individuals may be selectively infected with P. falciparum strains whose SERA5 sequences are less identical/heterologous to vaccine-type variant. Additionally, if sequence-dependent selection occurs, the haplotype diversity of SERA5 in the vaccinated group may be lower than that of SERA5 in the control group. To explore these possibilities, the haplotype diversity of SE36; OR and SR regions of SERA5 were compared between isolates from the BK-SE36 arm and isolates from the control arm in Uganda and Burkina Faso trials ( Supplementary Table S5 ). The haplotype diversity of SERA5 was similar between the vaccinated and control groups at the two trial sites. In the OR and SR regions, haplotype diversity was slightly higher in the vaccinated group than in the control group.

Four Ugandan isolates showed the same amino acid sequence as the OR region of vaccine-type SE36; three were from the BK-SE36 vaccine arm and one from the control arm. No isolate from Burkina Faso showed an identical OR sequence ( Supplementary Figure S1 ). In terms of the number of OR repeat units, seven Ugandan isolates and five Burkina Faso isolates had six OR units, similar to the vaccine-type variant ( Table 5 ). There was no significant difference in the frequency of the six OR units between the vaccinated and control groups, both in Uganda and Burkina Faso isolates (p > 0.05, χ2-test). As inferred above, from the African samples, seven OR units were the most common in both Uganda and Burkina Faso isolates, although overall no marked differences in OR frequencies can be seen between the vaccinated and control groups.

For the serine/proline substitution in the 13-mer insertion/deletion part of the SR region, the frequency of proline substitution was not biased between the vaccinated and control groups in both the Uganda and Burkina Faso isolates (p > 0.05, χ2-test) ( Table 6 ). For the 17-mer dimorphic region where the vaccine type is Type IV, the frequency of Type IV isolates did not greatly differ from those obtained from the BK-SE36 vaccine group and control subjects (p > 0.05, χ2-test) ( Table 6 ). Considering the number of serine residues between the vaccinated and control groups ( Figure 5 , Supplementary Table S6 ), the variation in the length of the serine stretch also did not differ between the two treatment arms (BK-SE36 or control arms) (p > 0.05, Mann-Whitney test). However, it is noted that the number of serine residues was relatively greater (i.e. longer stretch of serine residues) in the Ugandan isolates than those isolated from Burkina Faso (p < 0.05, Mann-Whitney test).

Genetic differentiation in the SE36 region of SERA5 was examined using Fst, the Wright’s fixation index of inter-population variance in allele frequencies (Wright, 1965) ( Table 7 ). The Fst value suggested a significant difference in the SE36 region of SERA5 and SR region between the Ugandan and Burkina Faso isolates (p < 0.05). However, no significant differentiation was detected between isolates from two vaccination arms (BK-SE36 or control). The OR region did not significantly differ between the Uganda and Burkina Faso isolates.