Study population and samples analyzed
Pf infection was confirmed by qPCR in 408 (6.7%) of the 6109 residents in Magude District from whom DBS were collected (101 out of 1035 [9.8%] in May 2015; 168 out of 1322 [12.7%] in November 2015, and 139 out of 3752 [3.7%] in May 2017). Ninety-nine (58.9%) and 107 (76.9%) of the Pf-positive samples from November 2015 and May 2017, respectively, were available for pfama1 and pfcsp sequencing after confirming enough parasite DNA material through a nested PCR targeting pfcrt. Among these samples, 85 (85.9%) and 93 (86.9%) amplified both genes and were included in the library for deep sequencing. Seventeen samples were excluded due to having less than 25 reads either in pfama or pfcsp, and 2 were further discarded due to incomplete demographic data (Fig. 1). Ninety (89.1%) and 80 (57.6%) of the Pf-positive samples from May 2015 and May 2017, respectively, were available in RNA protect and therefore used for the analysis of gametocyte-specific markers. Among these, 51 (56.7%) and 57 (71.2%) amplified the HK gene (Fig. 1), and one pre-intervention sample was excluded due to incomplete demographic data. Study participants before and after the intervention from whom genetic and transcriptional data was available were similar in area of residence, age, sex and qPCR-quantified Pf densities (Table 1). Also, the selected participants were representative of the overall population of Pf-infected individuals detected by qPCR, except for parasite densities which were higher among those with successfully analyzed Pf infections (Appendix table s4).
Table 1
Main characteristics of study participants.
| | Genetic structure | | Gametocyte markers |
| | Nov 2015 | May 2017 | | | May 2015 | May 2017 | |
| | (n = 78) | (n = 81) | p | | (n = 50) | (n = 57) | p |
Period | Nov 2015 | May 2017 | | | May 2015 | May 2017 | |
Study Area, n (%) | | | | | | | |
| Magude Sede | 47 (60) | 45 (55) | 0.654 | | 34 (68) | 31 (54) | 0.099 |
| Motaze | 14 (18) | 15 (19) | | | 11 (22) | 12 (21) | |
| Panjane | 10 (13) | 8 (10) | | | 5 (10) | 6 (11) | |
| Mahele | 2 (3) | 6 (7) | | | 0 (0) | 3 (5) | |
| Mapulanguene | 5 (6) | 7 (9) | | | 0 (0) | 5 (9) | |
Age, Mean (SD) | 14.2 (1,2) | 11.5 (1.7) | 0.198 | | 8.5 (10.9) | 10.5 (12.8) | 0.400 |
Gender, n (%) | | | | | | | |
| Female | 39 (50) | 40 (49) | 1.000 | | 22 (44) | 29 (51) | 0.562 |
| Male | 39 (50) | 41 (51) | | | 28 (56) | 28 (49) | |
Pf density, GM (SD) | 106.3 (254.6) | 165.4 (535.1) | 0.330 | | 169.2 (466.3) | 508.2 (1537.8) | 0.053 |
Pf density as assessed by qPCR; GM, Geometric mean; SD, Standard deviation |
Pf haplotype distribution
Across all successfully genotyped infections with complete demographic data (159 out of 178 [89%]), read coverage was similar in samples collected in 2015 and 2017 (Appendix table s5), with a mean of 6221 (SD 5416) and 7965 (SD 5908) reads for pfama1 and pfcsp, respectively. Parasite densities of samples that were successfully genotyped tended to have higher parasite densities (110 parasites/µl, IQR: 17–825), compared to those that failed the genotyping (41 parasites/µl, IQR: 5-798, p = 0.089; Appendix figure s1). After haplotype assignment and quality filtering of reads, we identified 30 SNPs for pfama1 and 24 for pfcsp. In total, 45 distinct haplotypes were detected at the pfama1 locus and 57 at the pfcsp locus (Fig. 2A and B), with 9 pfama1 haplotypes and 14 pfcsp haplotypes being detected only once (rare haplotypes). The number of haplotypes detected per locus was higher in samples collected from 2015 (43 for pfama1 and 49 for pfcsp) than in samples from 2017 (23 for pfama1 and 40 for pfcsp). Out of the 45 pfama1 haplotypes, 22 were shared in both periods and 23 were private of study period (21 in 2015 and 2 in 2017). Out of the 57 pfcsp haplotypes, 32 were shared and 25 were private (17 in 2015 and 8 in 2017). The proportion of Pf isolates with rare haplotypes was 10% (8 out of the 78) in 2015 and 0.1% (1 out of 81) in 2017 (p = 0.039) for pfama1 and 18% (13 out of the 78) in 2015 and 5% (4 out of 81) in 2017 (p = 0.024; Fig. 2C). Pre-intervention private haplotypes were detected in less samples than shared haplotypes (Fig. 2D).
Within-host Pf diversity
Most study participants had polygenomic infections, with single haplotypes detected in 48% (76/159; pfama1) and 33% (52/159; pfcsp) of genotyped infections. Of 83 participants with polygenomic infection at the pfama1 locus, 67 (81%) also had more than one pfama1 haplotype (and vice versa—63%). Additionally, within individual participants, the number of pfama1 and pfcsp haplotypes was highly correlated (Appendix figure s2; ρ = 0.83, p < 0.001, Spearman Rank test). The mean number of pfama1 haplotypes detected per sample was lower in 2017 (1.9, SD: 1.4) than in 2015 (2.6, SD: 2.2; p = 0.012; Fig. 3A). Such a difference was not observed for pfcsp haplotypes (3.0, SD: 0.23 in 2015 and 3.1, SD: 0.32 in 2017; p = 0.734). Mean Shannon index was lower in 2017 than in 2015 for both loci (pfama1: 0.36, SD: 0.05 in 2015 and 0.14, SD: 0.04 in 2017, p = 0.002; pfcsp: 0.43, SD: 0.06 in 2015 and 0.26, SD: 0.05 in 2017, p = 0.018; Fig. 3A and B). The proportion of pfama1 monogenomic infections was higher in 2017 (56%; 45/81) than in 2015 (40%, 31/78; p = 0.047). A similar trend, although not statistically significant, was observed for pfcsp (29% [23/78] in 2015 and 36% [29/81] in 2017, p = 0.397; Fig. 3B). No statistically significant association was observed with area of sample collection and age of study participants (Appendix table s6), except for the proportion of pfcsp monogenomic infections which were higher among the older individuals (18/37; 49%) than among younger ones 34/122, 28%, p = 0.027; Appendix table s7-8). The differences in within-host diversity metrics were maintained after adjusting the regression analysis by parasite density and age of study participants (Appendix table s8).
Between-infection Pf similarity indices
We employed 3 metrics that describe genetic similarity between pairs of Pf infections: binary sharing, Jaccard distance and Pearson CC. There was no evidence of statistically significant differences in the proportion of sample pairs that shared at least one haplotype (binary sharing; Appendix figure s3 and Appendix table s9). Pearson CC and Jaccard distance (Fig. 4A, B, C and D) of pfama1 increased from 2015 to 2017. Similarly, the fraction of sample pairs with high Pearson CC (> 0.9999) increased after the intervention (Fig. 4E) for both pfama1 (from 1.03–4.97%, p < 0.001) and pfcsp (from 1.03–3.4%, p = 0.019). Similar trends but not statistically significant were observed for highly similar parasites defined on the basis of Jaccard distance (Fig. 4F). No correlation was observed between genetic relatedness and gender, area, use of antimalarial, age and parasite density (Appendix figure s4).
Pf gametocyte-specific transcripts
Pf isolates with RT-qPCR and complete demographic data (n = 107) had higher parasite densities (546 parasites/µL, IQR: 23-2243) than those that failed (5, IQR: 3-1180, p < 0.001; Appendix figure s1). The proportion of Pf isolates with measurable gametocyte-specific transcripts ranged between 94% (101/107) for pfs25 and 61% (46/107) for pfs230, with intermediate values for pfap2g (75%; 80/107) and pfgexp02 (69%; 74/107) (Fig. 5A and Appendix table s10). Relative levels of female-specific gametocyte transcripts (pfs25) increased from 2015 to 2017, as did the proportion of parasites showing detectable pfap2g and pfgexp02 transcripts and their levels (Fig. 5A and B and Appendix table s10). The statistical significance was maintained in multivariate regression models (Appendix table s11). Both the univariate and the multivariate regression models (Appendix table s11) showed that pfap2g and pfgexp02 transcript levels increased with increasing parasite densities, while the opposite trend was observed for pfs25. No difference was observed by the area where the Pf isolates were collected (Appendix table s6).