High-throughput analysis of antimalarial susceptibility data by the WorldWide Antimalarial Resistance Network (WWARN) in vitro analysis and reporting tool

Abstract

Assessment of in vitro susceptibility is a fundamental component of antimalarial surveillance studies, but wide variations in the measurement of parasite growth and the calculation of inhibitory constants make comparisons of data from different laboratories difficult. Here we describe a Web-based, high-throughput in vitro analysis and reporting tool (IVART) generating inhibitory constants for large data sets. Fourteen primary data sets examining laboratory-determined susceptibility to artemisinin derivatives and artemisinin combination therapy partner drugs were collated from 11 laboratories. Drug concentrations associated with half-maximal inhibition of growth (IC50s) were determined by a modified sigmoid Emax model-fitting algorithm, allowing standardized analysis of 7,350 concentration-inhibition assays involving 1,592 isolates. Examination of concentration-inhibition data revealed evidence of apparent paradoxical growth at high concentrations of nonartemisinin drugs, supporting amendment of the method for calculating the maximal drug effect in each assay. Criteria for defining more-reliable IC50s based on estimated confidence intervals and growth ratios improved correlation coefficients for the drug pairs mefloquine-quinine and chloroquine-desethylamodiaquine in 9 of 11 and 8 of 8 data sets, respectively. Further analysis showed that maximal drug inhibition was higher for artemisinins than for other drugs, particularly in ELISA (enzyme-linked immunosorbent assay)-based assays, a finding consistent with the earlier onset of action of these drugs in the parasite life cycle. This is the first high-throughput analytical approach to apply consistent constraints and reliability criteria to large, diverse antimalarial susceptibility data sets. The data also illustrate the distinct biological properties of artemisinins and underline the need to apply more sensitive approaches to assessing in vitro susceptibility to these drugs.

Fig 1

Example of growth inhibition data for dihydroartemisinin (DHA) and mefloquine (MQ), obtained by a tritiated-hypoxanthine incorporation assay of a sample from a traveler studied at the Centre National de Référence du Paludisme, Paris, France. In this case, a paradoxical increase in apparent growth is observed at higher concentrations of mefloquine (MQ) but not DHA. This phenomenon results in two distinct parameters of drug efficacy for MQ (see Table S1 in the supplemental material).

Fig 2

Effects of differing reliability criteria on interdrug correlation. Each series of interconnected symbols represents the Pearson correlation coefficients for chloroquine-desethylamodiaquine (CQ-DQ) (8 data sets) and mefloquine-quinine (MQ-QN) (11 data sets). Three levels of exclusion criteria were studied: ^a, default IVART criteria; ^b, IVART criteria with additional exclusion of fixed-gamma assays with a growth ratio of 2 to 3; ^c, IVART criteria with additional exclusion of fixed-gamma assays with a growth ratio of 3 to 5. Filled triangles, significant (P < 0.05) correlation coefficients; open triangles, nonsignificant (P > 0.05) correlation coefficients. Individual correlation coefficients, P values, and numbers of samples are shown in Table S3 in the supplemental material.

Fig 3

Growth ratios in nonmicroscopic forms of readout according to drug. Boxes show medians, while quartiles and whiskers indicate ranges. Dark shaded boxes, artemisinin derivatives; light shaded boxes, ACT partner drugs; open boxes, chloroquine (CQ) and quinine (QN). Asterisks indicate significant reductions in the growth ratio from that with DHA (P, 0.05 by the Mann-Whitney test with Bonferroni's correction). Data sets are those listed in Table 1 and were obtained by measurement of hypoxanthine incorporation (A to D), HRP2 (E to H), LDH (I to K), or SYBR green (L). Only assays passing core criteria were included.

Fig 4

Difference in proportional inhibition of growth {[G(C₀) − G_min]/G(C₀)} between dihydroartemisinin (DHA) and mefloquine (MQ) (as illustrated in Fig. 1). Medians and interquartile ranges are shown. Positive values indicate greater efficacy of DHA than of MQ. Data sets (indicated along the y axis) are those listed in Table 1 and were obtained by measurement of hypoxanthine incorporation (A to D), HRP2 (E to H), LDH (I to K), or SYBR green (L). Only assays passing core criteria were included.

Fig 5

Excerpt from an output PDF file showing results for a Ugandan sample tested with six drugs. IC₅₀s are given in nanomolar concentrations. Growth values for the zero-drug wells are plotted at the left-hand ends of the graphs (at 2 log units below the lowest drug concentration used) for display purposes.