ABSTRACT
BACKGROUND Plasmodium falciparum resistance to chloroquine (CQ), the most widely used antimalarial drug, has historically posed a major threat to malaria control in Angola and throughout the world. Although Angola replaced CQ with artemisinin combination therapy (ACT) as a frontline treatment in 2006, malaria cases and deaths have recently been rising. CQ-resistance mutations may still be a contributing factor, given that (1) some also modulate resistance to ACT partner drugs and (2) ACT is not yet consistently implemented across Angola. It is important to continue monitoring all known resistance alleles in P. falciparum, but no studies have done so in Angola since 2012.
METHODS We sampled P. falciparum DNA from the blood of 50 hospital patients in Cabinda, Angola in 2018. Each infection was genotyped for 13 alleles in the genes crt, mdr1, dhps, dhfr, and kelch13, which collectively confer resistance to six common drugs. To analyze frequency trajectories over time, we also collated P. falciparum genotype data published from across Angola in the last two decades.
RESULTS The two most important alleles for CQ resistance, crt 72-76CVIET and mdr1 86Y, have both declined in frequency from respective highs of 98% in 1999 and 73% in 2003. However, the former remains at 71% frequency in this sample while the latter has dropped to just 7%. Of seven possible alleles for sulfadoxine-pyrimethamine (SP) resistance in dhps and dhfr, the average total number per isolate increased from 2.9 in 2004 to 4.4 in 2018. Finally, we detected no non-synonymous polymorphisms in kelch13, which is involved in artemisinin resistance in Southeast Asia.
CONCLUSIONS Changes in drug policy in Angola since 2006 appear to have exerted strong selection on P. falciparum drug resistance alleles. Resistance to CQ is declining, but due to functional tradeoffs and novel selection at mdr1 loci, resistance to ACT partner drugs appears to be rising. More haplotype-based studies at mdr1 will be needed to understand the changing efficacy of multiple drugs. Finally, SP resistance has jumped rapidly since 2014, consistent with widespread use of intermittent SP treatment during pregnancy. These data can be used to support effective drug policy decisions in Angola.
BACKGROUND
Antimalarial drugs have long been important tools for malaria control1. However, their efficacy is constantly threatened by the evolution of drug resistance in Plasmodium falciparum2. Multiple P. falciparum genes are involved in drug resistance, and selection on them varies by allele, genetic background, and drug environment3–5. Therefore, frequent monitoring of resistance alleles is crucial to predicting the spread of drug resistance. This is especially true in the West African country of Angola, where malaria cases and deaths are on the rise6.
The first anti-malarial drug to enjoy widespread use in Angola was chloroquine (CQ) in the 1950s7. CQ resistance was first confirmed in Angola in the 1980s, and by the early 2000s, CQ failure rates exceeded 80%8,9. As a result, CQ was discontinued in Angola in favor of artemisinin-based combination therapy (ACT) starting in 200610. To discourage the evolution of artemisinin resistance, artemisinin is used in combination with the longer-acting partner drugs lumefantrine (LMF) or amodiaquine (AQ), which is chemically related to CQ11. Artemisinin resistance has not yet appeared in Angola, although many resistant kelch13 mutations have emerged in Southeast Asia5,12. Nonetheless, occasional ACT treatment failures have been reported in Angola due to partner drug resistance10.
Strong P. falciparum resistance to CQ and AQ is caused by crt K76T, a lysine to threonine substitution at codon 76 of the chloroquine resistance transporter (Table 1). A meta-analysis found this allele to be 7.2-fold overrepresented in CQ treatment failures13, reflecting its selection by CQ and AQ in many clinical studies (Table 1). In Angola, K76T is found on the haplotype crt 72-76 CVIET, which is of Asian origin14. CQ resistance has also evolved independently through the haplotype crt 72-76 SVMNT in South America and Papua New Guinea15.
Although CQ has been discontinued in Angola, CQ-resistance loci are also involved in resistance to the ACT partner drugs AQ and LMF. Numbers in the header indicate amino acid position. For mdr1, incomplete haplotypes are shown as reported in the literature. *These alleles are unlikely to confer resistance directly, but they are less deleterious than the alternate allele in the presence of drug. *This allele is unlikely to confer resistance directly, but it is linked to other functional alleles. Additional details and references are available in Table S1.
The N86Y allele of mdr1, or multidrug resistance gene 1, also confers resistance to CQ and AQ13. Although this specific polymorphism dominated early studies of mdr1 and CQ resistance, the evolution of mdr1 is complicated by linkage between position 86 and other functional polymorphisms16. Precise mdr1 haplotypes vary among P. falciparum populations and drug settings, but in Angola alone, at least six alleles at three mdr1 positions have been proposed to modulate resistance to CQ, AQ, and the ACT partner drug lumefantrine (LMF) (Table 1; Table S1).
The drug sulfadoxine-pyrimethamine (SP) has also been in widespread use in many African countries since the 1960s17. P. falciparum quickly began evolving partial resistance to SP, mediated by numerous substitutions in dhps and dhfr18. The risk of SP treatment failure increases with the number of mutant alleles present, with “quintuple mutants” at codons 437/540 of dhps and codons 51/59/108 of dhfr of particular concern19–21. By the early 2000s, these alleles were common in Angola and 25-39% of P. falciparum infections failed to respond to SP treatment9. SP has since been discontinued as a frontline therapy, but it is still administered to pregnant women to reduce common complications from malaria22. Although this approach is generally still useful in Africa23,24, its efficacy is waning as additional dhps mutations continue to emerge18,25–27. In one recent example from Tanzania, a novel mutation at dhps 581 was both selected by SP treatment and associated with worse pregnancy outcomes28. Because SP is still in widespread use, it is critical to continue monitoring its effectiveness along with variation in its target genes.
In this work, 50 P. falciparum infections from Cabinda, Angola were genotyped for 13 markers of drug resistance in the genes crt, mdr1, dhps, dhfr, and kelch13. Similar allele frequency data were also gathered from studies published on Angolan P. falciparum in the last two decades. For every locus but kelch13, we found temporal patterns of allele frequency change that are consistent with changes in drug policy. This work can inform future decisions on drug administration in Angola, particularly given rapid increases in SP resistance.
RESULTS
Genotyping success and MOI
Each sample was successfully genotyped at an average of 12 out of 13 loci (Table S3). The kelch13 locus had the highest success rate (100%), while crt had the lowest success rate (78%). Although the crt primers have performed well on other Angolan samples29, in this cohort, even the nested protocol amplified products of multiple sizes (Fig S1).
Fifteen of 50 samples had sequence diversity (i.e., peaks of two bases) in at least one resistance marker site. Assuming that double peaks indicated the presence of two strains, the overall multiplicity of infection (MOI) was 1.3.
Very little polymorphism in kelch13
No kelch13 polymorphisms were observed at codons 578-580, which have been associated with ACT resistance in Southeast Asia and Uganda12. Moreover, with the exception of one synonymous variant in one sample, no polymorphism was observed across all 261 kelch13 codons sequenced in this study.
Markers of CQ resistance and LMF susceptibility are declining
The CVIET haplotype at crt codons 72-76, which confers strong resistance to CQ, was detected at 71% frequency in this study (Fig 1). This represents a significant decline from a peak of 98% in 1999 (p = 0.03), although individual estimates have been noisy over time.
The solid circle indicates new data from this study. Historical data were obtained from29,31–35. The diamond indicates the combined presence of two resistant haplotypes (CVIET and SVMNK) at crt 72-76 in29; all other points represent CVIET only. τ is the Kendall rank correlation between time of sampling and frequency of resistance. The dashed line shows the year that CQ was officially discontinued in Angola.
The mdr1 allele 86Y, which also confers resistance to CQ, was detected at just 6.5% frequency in this study (Fig 2A). This marker has declined rapidly and steadily from ~80% frequency in 2003 (p = 0.007). Accordingly, the alternate allele 86N—which is both CQ-sensitive and LMF-resistant (Table 1) — has increased in frequency to 93.5% (Fig 2B, p = 0.006). The linked polymorphism mdr1 184F, which is also preferred in the presence of LMF (Table 1), has been rising in frequency at a similar rate (Fig 2B, p = 0.087), although it remains less common than 86N. A single sample contained the additional CQ-resistance allele mdr1 1246Y (Table 1), which occurred on an 86Y/184Y background (Table S3).
Markers of SP resistance markers have become more common
One third (33%) of P. falciparum isolates sampled here were “quintuple mutants” for five dhfr and dhps alleles that confer strong SP resistance (Fig 3). Compared to samples from migrant workers collected around 201438, this represents a 2.8X increase of quintuple mutants in Angola in less than five years. Three “sextuple mutants” were also observed for the first time in Angola, including resistance alleles at dhps codons 436 (22% frequency) and 581 (8.2% frequency). The average number of combined dhfr/dhps resistance alleles per isolate has increased sharply over time, from 2.9/7 in 2004 to 4.4/7 in this study (t = −9.71, p < 2.2 x10-16).
Each point represents the total number of resistance allels in both genes from a single isolate. Solid points indicate new data from this study; historical data are from32,38,41. Points are jittered horizontally and vertically for clarity. Resistance alleles were counted at dhfr codons 51, 59, and 108 and dhps codons 436, 437, 540, and 581. The dashed box surrounds a subset of 2014 isolates that carried 0-3 mutant alleles, but for which more precise data were not available. Variance explained (R2) and p-value are shown are from a linear regression excluding all 2014 data.
DISCUSSION
The official withdrawal of CQ in Angola since 2006 has likely contributed to the decline of CQ-resistance alleles in crt (Fig 1) and mdr1 (Fig 2A). This result is similar to other African countries that have discontinued CQ, including Malawi, the Gambia, Kenya, Ethiopia, Tanzania, and Grand Comore39. In Malawi, clinical CQ sensitivity largely returned after the prevalence of crt K76T declined from 85% in 1992 to 13% in 200040. In Angola, however, the prevalence of the CVIET haplotype in Angola remains high at 71% (Fig 1). Although the exact fraction of resistant parasites may vary by locality (Fig 1), these results imply that CQ resistance via crt is still standard in Angola. In contrast, the rate of decline of mdr1 86Y—the second-most important CQ-resistance allele—is sharp enough to suggest its disappearance from Angola within a few years (Fig 2A). This stark difference between the evolution of crt and mdr1 may be best explained by the differential effects of LMF, an ACT partner drug, on their CQ-sensitive alleles (Table 1). Specifically, wild-type crt K76 is only passively selected in the absence of CQ, while wild type mdr1 N86 directly confers LMF resistance (Table 1). LMF resistance is highest when mdr1 N86 co-occurs with mdr1 184F, which is rapidly spreading in Angola (Fig 2B), and mdr1 1246D, which is nearly fixed (Table S3). Consequently, the rise of LMF resistance could soon challenge the success of ACT as currently implemented in Angola.
The most rapid change observed in this study was the increase in total SP-resistance alleles per isolate (Fig 3). Similar increases over time have been reported in a number of other African countries42–44, likely in response to the implementation of WHO recommendations for SP during pregnancy. It is now clear that more than five mutations in dhfr/dhps contribute to SP resistance: in our sample, seven such mutations were present, and three infections (7.0%) carried six of them. Because intermittent SP treatment currently recommended by the WHO does not eliminate parasitemia45, it is strongly expected to select for additional SP resistance. The benefits of SP in pregnancy have outweighed these costs in the past, but the present rate of resistance evolution implies that these benefits may be eroding27,28. Further research will be required to weigh the impact of novel resistance haplotypes against other factors impacting SP treatment efficacy. To help accomplish this goal, we emphasize the importance of reporting complete haplotype information for all combined dhfr/dhps alleles in each sampled infection.
Finally, we detected no signs of artemisinin-resistant alleles in kelch13. This result is consistent with the high efficacy of ACT in Angola10, and overall, there is little evidence that artemisinin resistance alleles are spreading in Africa46. Monitoring of kelch13 in Africa is nonetheless important, as artemisinin is the only drug for which resistance alleles are not already widespread.
CONCLUSIONS
Changes in drug policy since 2006 have had clear impacts on the frequencies of several drug resistance alleles in Angola. Markers of SP resistance are rapidly becoming more common, which endangers the efficacy of intermittent treatment during pregnancy. Resistance to CQ is declining, but resistance to LMF appears to be rising. More frequent monitoring and drug policy adjustments will likely be necessary to regain control of P. falciparum malaria in Angola.
METHODS
Sample collection and ethics statement
Patients reporting to the Hospital Regional de Cabinda in 2018/2019 with fever, chills, or other malaria symptoms were offered the option to be consented to this study. Sample collection followed protocols approved by Stanford University (IRB #39149) and the Medical Ethics Committee of the University 11th of November in Cabinda. Consented participants’ blood was drawn from a vein and screened under a microscope for P. falciparum parasites. If positive, whole blood was filtered through cellulose columns to remove leukocytes47. The filtered red blood cells were spotted on Whatman FTA cards (Sigma Aldrich), dried, and stored for at least 6 months.
DNA extraction and genotyping
To elute DNA, saturated circles were cut out of the Whatman FTA cards and incubated in 800 uL TE buffer (10 mM Tris-Cl, 1 mM EDTA, pH 8.0) with 20 uL Proteinase K (Invitrogen) for 2 hours at 65°C. DNA was extracted from the liquid supernatant using a phenol-chloroform protocol with phase-lock gel tubes48.
PCR amplification of the P. falciparum genes crt, mdr1, dhfr, dphs, and kelch13 was performed with previously published primers29,49,50. Cycling protocols were based on manufacturer recommendations for OneTaq Hot Start 2X Master Mix (NEB) and/or Phusion High-Fidelity PCR Master Mix with HF Buffer (NEB) (Table S2). Reactions were visualized in 1% agarose gels, and if successful, cleaned with ExoSAP-IT (ThermoFisher) and Sanger sequenced (Elim Bio). Sanger chromatogram data were compared to PlasmoDB reference P. falciparum sequences using Benchling. Amino acid substitutions were identified in the following positions: mdr1 86, 184, and 1246; crt 72-76; dhfr 50, 51, 59, and 108; dhps 436, 437, 540, and 681; and kelch13 578-580.
MOI and allele frequency calculations
For each sample, a double infection was inferred if the sequencing chromatogram showed equally sized, double peaks for any of the 13 analyzed loci. Multiplicity of infection (MOI) was calculated as the total number of infections divided by the total number of samples, as previously described51. Similarly, the frequency of each allele was determined based on the total number of infections, with double infections at any locus contributing two genotypes at every locus. Samples without missing data at dhfr or dhps were also assessed for the presence of up to seven SP-resistance alleles (dhfr-51I, dhfr-59R, dhfr-108G, dhps-436, dhps-437G, dhps-540E, dhps-581)52,53.
Collection of historical data
Publications reporting allele frequencies for drug-resistance loci anywhere in Angola since 1995 were gathered from the Worldwide Antimalarial Resistance Network (WWARN) Molecular Surveyor tool (http://www.wwarn.org/molecularsurveyor/), facilitated by a recent review7. The original data published in these studies were used to calculated alleles frequencies as described above. For studies that spanned multiple years, the average year was used for time-course analysis (below). Studies that did not provide linked data for dhfr and dhps (e.g., reported the two genes separately) could not be included.
Statistical analysis
To evaluate changes in mdr1 and dhfr/dhps alleles over time, linear models were fit to the frequency or count data using the lm function in R. To avoid a bias from incomplete data, all 2014 samples were excluded from the dhfr/dhps timecourse analysis. For crt, the relationship between CVIET frequency and time was not linear; therefore, Kendall’s rank correlation was applied using the cor.test function in R.
DECLARATIONS
Ethics approval and consent to participate
Ethics approval for this study was obtained from Stanford University IRB (#39149) and the Medical Ethics Committee of the University 11th of November in Cabinda.
Consent for publication
Prior to participation, all study subjects and/or their parents consented in writing to the publication of study results in the scientific literature.
Competing interests
The authors declare that they have no competing interests.
Funding
This study was supported with grants from the Stanford Center for Computational, Evolutionary, and Human Genomics to S.B. and E.R.E.; an MRC award (MR/M01987X/1) to S.B.; and an NIH award (5R35GM118165-05) to D.A.P.
Author’s Contributions
E.R.E., D.A.P., and S.B. designed the study. F.R. and S.B. supervised the study. E.R.E. collected data, analyzed data, and wrote the manuscript. All authors have approved the final manuscript.
Acknowledgements
We are immensely grateful to the study participants and staff of the Hospital Regional de Cabinda. Logistic support in sample collection was provided by Dr. Maria das Dores Sungo and Dr. Francisco Casimiro Lubalo, Rector and Vice-Rector of the Faculty of Medicine, University 11th of November, Cabinda. We also thank Rachael Madison, Barbara Baro Sastre, and Elizabeth Egan for their assistance in preparing and testing sampling materials.