Non-falciparum malaria infection and IgG seroprevalence among children under 15 years in Nigeria, 2018

Ethics

All participants provided informed consent/assent before enrollment into the NAIIS survey. For children age <10 years, consent for biomarker testing and consent to future testing was granted by parent or guardian; assent was received from children age 10–14 years. Secondary laboratory testing for malaria biomarkers as part of the Nigeria Multi-disease Serologic Surveillance using Stored Specimens (NMS4) project was approved by the National Health Research Ethics Committee of Nigeria (NHREC/01/01/2007) and determined to not involve human research by the Centers for Disease Control and Prevention Human Subjects office (project 0900f3eb819d4c63).

Survey design

NAIIS 2018 was a national HIV population-based household survey led by the Government of Nigeria under the Federal Ministry of Health, and survey methods have been previously reported⁴⁰. The sampling design utilized a stratified two-stage cluster sample with 37 strata (the 36 states and Federal Capital Territory, FCT). Enumeration areas (EAs) were selected within each state in the first stage with probabilities proportionate to estimated population size according to the projected 2018 number of households (based on the 2006 census data). When calculating the number of EAs to be selected per state, state variability of household size was considered. This sampling frame produced 662,855 EAs, of which 4035 were selected for NAIIS. For the second stage, a household listing exercise was carried out in all selected EAs, and a random sample of 28 households per EA were chosen for inclusion. The resulting sample included 101,580 households to be approached for enrollment of household participants into NAIIS.

Laboratory data collection for malaria biomarkers

This report from the NMS4 project is inclusive of data from all children (persons <15 years of age) who provided a DBS for the NAIIS survey. Priority was given this this subset of samples as children represent the segment of the population most susceptible to adverse outcomes from malaria infection. Collection of data for persons >15 years of age is still ongoing, and an upcoming report will outline those findings. All multiplex antigen and IgG detection assays were performed at the National Reference Laboratory (NRL) of the Nigeria Centre for Disease Control (NCDC) in Gaduwa, Nigeria; DNA assays were performed at at the US Centers for Disease Control and Prevention in Atlanta, GA. Multiplex antigen and IgG detection were conducted for all available DBS specimens for children <15 years of age, and DNA assays performed for only samples selected as described below.

Multiplex bead-based Plasmodium antigen detection assay

To rehydrate blood samples from DBS, a 6 mm punch (~10 µL whole blood) was taken from each DBS sample and blood eluted to a 1:40 concentration in blocking buffer (Buffer B: Phosphate Buffered Saline (PBS) pH 7.2, 0.5% Bovine Serum Albumin (BSA), 0.05% Tween 20, 0.02% sodium azide, 0.5% polyvinyl alcohol, 0.8% polyvinylpyrrolidone and 3 µg/mL Escherichia coli extract)⁴¹. Elutions were stored at 4 °C until testing. As described previously²³, samples were assayed in singlet at the 1:40 whole blood dilution for histidine-rich protein 2 (HRP2), pan-Plasmodium aldolase (pAldolase), pan-Plasmodium lactate dehydrogenase (pLDH), and P. vivax LDH (PvLDH). Assay plates were read on MAGPIX™ instruments (Luminex Corp, Austin, TX), and with a target of at least 50 beads/region, the median fluorescence intensity (MFI) was generated for each analyte by Luminex xPonent® software version 4.2 (Luminex Corp). For each assay plate, the MFI signal for sample dilution buffer (buffer background) was subtracted for each target’s MFI signal to provide a MFI-bg assay signal for analysis. Plate-to-plate variation was accounted for by assessment of controls on each plate, inclusive of antigen negative blood (at 1:40 dilution) and a 4-point titration curve of a recombinant antigen positive pool. As described previously²³, assay plates ‘failed’ if the negative controls showed a positive signal, or if the positive controls for both pLDH and HRP2 were both 2 standard deviations outside of the cumulative moving average of the MFI-bg signal for assay plates. Previous analysis found 1.2% of initial antigen detection plates failed²³, and failed plates were re-run to obtain valid data. To determine the assay signal threshold for antigen positivity, a finite mixture model approach was employed to estimate the ‘antigen negative’ population within these data and then establishing a cutoff using the mean + 2 standard deviations from this distribution⁴².

Multiplex bead-based IgG detection assay

The IgG-detecting multiplex bead assay (MBA) was performed as described previously⁴¹, and included targets for a variety of infectious diseases, as well as vaccine-preventable disease targets. Assay plates were run on same machines utilizing the same software as the antigen detection assay. From the DBS elution described above, an approximate final serum dilution of 1:400 was used for the assay with samples run in singlet. Analysis in this current study was restricted to the Schistosoma japonicum glutathione-S-transferase (GST) internal control antigen and the four Plasmodium merozoite surface protein 1, 19kD (MSP1) antigens from the human malarias: P. falciparum, P. malariae, P. ovale, and P. vivax³⁷. Plate-to-plate variation was accounted for by assessment of controls included on each assay plate, and a similar ‘pass/fail’ structure as listed above for the multiplex antigen assay. Assay controls were inclusive of a 1:400 dilution of a sera pool from U.S. residents (no malaria exposure), and a positive control pool of persons residing in various P. falciparum transmission settings. In a similar manner as described previously⁴³, Levey-Jennings charts were generated for the positive control to track deviations in assay signals over time, and an assay plate ‘failed’ if negative controls showed a positive signal, or positive controls were 2 standard deviations outside of the cumulative moving average. Of 523 IgG assay plates, 31 (0.6%) failed and needed to be repeated to obtain valid data.

Sample selection for DNA assays

A two-part selection strategy was utilized based on the Plasmodium antigen data to select samples for further DNA assays, and to provide a representative estimate of the burden of non-falciparum malaria in Nigeria. Results from both sampling designs for DNA testing were considered when presenting national and state-level active infection estimates. Estimates are presented either as non-falciparum single-species infection, or a P. falciparum infection mixed with another species, or as a combination of these two (i.e., ‘any P. ovale present’, etc.).

As described in previous studies^44,45,46, and illustrated in Supplementary Fig. 2, the first set of samples was selected based on the ratio of HRP2 antigen to the other three Plasmodium targets. If any specimen was positive to a non-HRP2 target, but found to have a low or negative HRP2 signal relative to other Plasmodium antigens by visual identification, this sample was selected for suspicion of non-P. falciparum malaria infection (or potentially P. falciparum deleting either pfhrp2 and/or pfhrp3 genes)^44,46. This selection strategy would favor the identification of single species non-P. falciparum infections, or infections where P. falciparum was not the dominant Plasmodium species present.

The second selection criteria involved specimens positive to HRP2, which would indicate active P. falciparum infection. For each of the six zones in Nigeria (North West, North East, North Central, South West, South East, South South) a range of HRP2 levels (100–1000 ng/mL) was chosen from DBS from children in each zone, with 100 DBS randomly selected to geographically represent P. falciparum infections for each zone (target sample size of 600), and ultimately the country as a whole. This selection strategy allowed assessment by DNA assays of P. falciparum infections that may also harbor another Plasmodium species (mixed infections). As only P. falciparum infections were selected from the second selection strategy, the percentages of mixed infections were calculated for each Nigerian state and extrapolated to the total number of samples with antigen data from that Nigerian state.

Photo-induced electron transfer (PET) PCR assays

One 6 mm DBS punch was placed into a 1.5 mL tube for processing according to manufacturer’s instructions. The QIAamp DNA blood mini kit (Qiagen, Valencia, CA, USA) was used to elute whole DNA in 150 µL of elution buffer, aliquoted, and stored at −20 °C until use. The PET-PCR reaction was performed with primer targets for the 18 S ribosomal RNA gene for Plasmodium genus and for P. falciparum as described previously⁴⁷. PET-PCR primer targets for the P. ovale reticulocyte binding protein 2 (rbp2) gene have been published previously⁴⁸, as well as the PET-PCR dihydrofolate reductase-thymidylate synthase (dhfr-ts) gene for P. malariae and P. vivax⁴⁹. Primer sequences for all targets listed above are shown in Supplementary Data 1. Reactions were performed in a 20 μL reaction containing 2X TaqMan Environmental buffer 2.0 (Applied BioSystems, Grand Island, NY, USA), 125 nM each of forward and reverse primers except for the P. falciparum HEX-labelled primer which was used at a 62.5 nM. Each sample was run with 2 μL of DNA in the PCR reaction with the following cycling parameters: initial hot-start at 95 °C for 10 min, followed by 45 cycles of denaturation at 95 °C for 10 s, annealing at 60 °C for 40 s. Fluorescence channels were selected to detect each fluorescently labelled primer-set and cycle threshold (CT) values recorded at the end of each annealing step. All assays were performed using Agilent Mx3005pro thermocyclers (Agilent technologies, Santa Clara, CA, USA) with Agilent Aria 2.0 analytical software. The PET-PCR criteria for DNA positivity for any primer set was a CT value < 40.0.

Statistical analysis

Survey questionnaire data cleaning was conducted using Census and Survey Processing System (CSPro) version 7.7.2 (U.S. Census Bureau; census.gov/data/software/cspro) and SAS version 9.4 (SAS Institute, Cary, NC). For the IgG analysis, DBS with glutathione-S-transferase (GST) MFI-bg reads above 500 (indicating non-specific binding) were excluded from the analysis. To dichotomize seropositivity to each species MSP1 antigen, a 2-component finite mixture model (FMM) of the log₁₀-transformed MFI-bg signal was fit for each IgG response³⁷, and the positivity cutoff was determined by the lognormal mean of the first component (presumed seronegative) plus 2 standard deviations⁴². A 6-component model was created for PfMSP1 to more accurately define the seronegative population. FMM plots were generated in SAS version 9.4.

Remaining analyses were conducted in R version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria) or Microsoft® Excel® version 2208. Estimation of malaria exposure was adjusted while taking into account the complex sample design of the NAIIS using the R survey package⁵⁰, and data were weighted using normalized survey weights. Demographic, socioeconomic, and behavioral risk factors were considered in relation to malaria infection and exposure, including sex, age group, wealth quintile, and living in a household with at least 1 mosquito net per 1.8 household members. Since there was no data collected on mosquito net usage per child, a proxy net coverage variable was created using the ratio of nets to household members, based on the WHO criteria recommended for procurement purposes to ensure universal net coverage in households⁵¹. Cluster level covariates included: place of residence (urban/rural) and community net coverage. Community net coverage was defined as the proportion of individuals in the cluster with adequate household net coverage (1 net per 1.8 household members). Individual net coverage was normalized by subtracting the average community net coverage from the individual-level variable⁵². Three multivariate, mixed-effects logistic regression models were fit to determine factors associated with malaria exposure (measured as IgG positivity to PmMSP1, PoMSP1, or PvMSP1) and included a random intercept for the cluster variable. For the malaria infection risk factor analysis, multivariate logistic regression models were fit without random effects since there were only 1204 observations with over 800 clusters represented (average of 1.4 persons per cluster). Firth’s penalized-likelihood logistic regression was performed for the P. malariae and P. ovale infection models to reduce the bias due to small sample size and quasi-complete and complete separation of the wealth covariate for the P. malariae and P. ovale models, respectively⁵³.

To model the rate of change from seronegative to seropositive for each of the three antigens (PmMSP1, PoMSP1, and PvMSP1), serocatalytic conversion models were fit to the seropositivity by age data for each antigen^54,55. The model estimated the seroconversion rate (SCR or λ) which is the mean annual rate of conversion from any individual in the population from seronegative to seropositive, and the seroreversion rate (SRR or ρ) which is the mean annual rate of reversion from seropositive to seronegative. Children under age 1 year were excluded to remove the potential effect of maternally-derived anti-malarial antibodies. The parameters were estimated using Markov Chain Monte Carlo (MCMC) with 10,000 MCMC iterations. R code to fit these models was utilized from the code provided by Dr. Michael White, Serology Github Repository⁵⁶. The geospatial analysis was conducted in R version 4.1.1 and QGIS version 3.16.3, and spatially smoothed sero-prevalence maps were created using local Empirical Bayes spatial smoothing. All code is available at: https://github.com/cleonard297/nonPf_seroPCR_code.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.