Ensemble modeling highlights importance of understanding parasite-host behavior in preclinical antimalarial drug development

Abstract

Emerging drug resistance and high-attrition rates in early and late stage drug development necessitate accelerated development of antimalarial compounds. However, systematic and meaningful translation of drug efficacy and host-parasite dynamics between preclinical testing stages is missing. We developed an ensemble of mathematical within-host parasite growth and antimalarial action models, fitted to extensive data from four antimalarials with different modes of action, to assess host-parasite interactions in two preclinical drug testing systems of murine parasite P. berghei in mice, and human parasite P. falciparum in immune-deficient mice. We find properties of the host-parasite system, namely resource availability, parasite maturation and virulence, drive P. berghei dynamics and drug efficacy, whereas experimental constraints primarily influence P. falciparum infection and drug efficacy. Furthermore, uninvestigated parasite behavior such as dormancy influences parasite recrudescence following non-curative treatment and requires further investigation. Taken together, host-parasite interactions should be considered for meaningful translation of pharmacodynamic properties between murine systems and for predicting human efficacious treatment.

Figure 1

Schematic representations of the mechanistic within-host parasite growth models for P. berghei (a) and P. falciparum (b), with summary model details (c). The base model by is represented in black with model modifications added in color, for all models erythrocytic parasite stage was split into n age compartments (n = 12). Model a to e for P. berghei mainly capture processes dictated by the host-parasite system such as reactions of the host to increasing infection in model b (bystander)) and model c (comp. erythr.), changes in parasite dynamics over the course of infection model d (impaired maturation), and host cell preferences of the parasite model e (reticulocyte). In turn, model f to h for P. falciparum dynamics are primarily influenced by the experimental set-up of continued human RBC injections. Whereas model f (const. RBC decay) and g (dd. RBC decay) additionally explicitly model mouse RBCs, model h (human RBC) only captures human RBC populations. RBC or parasite transitions are represented with solid lines and influencing processes with colored dashed lines. (c) Selected index numbers characterizing the growth of parasites in the respective mechanistic mouse models for the experiment shown in Fig. 3. Anemia is defined as the percentage of RBCs compared to values prior to infection.

Figure 2

Estimated values of the infectivity parameter β by model for both murine experimental systems. Each symbol represents the value estimated for one experiment. (a) Values estimated for murine P. berghei infection. Model a (base), c (comp. erythr.) and e (reticulocyte) show similar results whereas higher values were estimated for model b (bystander) and d (impaired maturation) (b) Estimated β –values for infection of humanized mice with P. falciparum using the mechanistic models f (const. RBC decay), g(dd. RBC decay), and h (human RBC). (c) Parasite growth rate p_gr and parasitemia at start of the exponential growth phase P₀ (72 h post-infection) estimated for model i (exponential). The laboratories are denoted by different symbols (not identified here).

Figure 3

Representative fit of the within-host models to data. (a) Data (•) and model predictions (- -) of infection with P. berghei with an inoculum of 2 × 10⁷ infected RBCs (i.v.) show a steep increase in parasitemia three days after inoculation. (b) Model output for unobserved total numbers of RBCs show an increase in infected RBCs (- -) with a simultaneous decrease in uninfected RBCs (▬) resulting in anemia. However, the total number of human and murine RBC populations differs between model predictions (compare model b (bystander)), given that the estimated percentage of infected RBCs is compared to observed. Further differences in models become apparent comparing predicted time of, and total parasite numbers at, peak parasitemia P_max (see Fig. 1c). (c) Infection of SCID mice with P. falciparum through an inoculum of 3.5 × 10⁷ infected RBCs (i.v.). Human RBCs (∆) are injected daily until day seven post-infection increasing total human RBC counts (▬). (d) As uninfected RBCs (▬) increase the predicted number of mouse RBCs (• •) decrease due to random clearance of excess RBCs. After RBC injections are ceased, the model predicts a steep decline in human RBCs. Data (•) and models f to h (- -) show lower values of predicted peak parasitemia compared to model i.

Figure 4

Comparison of drug efficacy estimates found for P. berghei in normal mice (a–c) and P. falciparum in SCID mice (d–f). EC₅₀ [ng/mL], E_max [1/h] and the clearance half-life [h] are illustrated for each drug and parasite growth model. The drug action model showing the best fit to data was chosen based on ΔOFV (AIC), visual assessment of model fit and biological plausibility for each parasite growth model (with Turnover-model (Turn), drug action through an effect compartment (Eff) and delayed clearance of dead parasites (Cl)). See Supplementary Table S9 for parameter values.

Figure 5

Representative fits of drug action models in SCID mice infected with P. falciparum at day 0 with an inoculum of 2 × 10⁷–3.5 × 10⁷ infected RBCs. The models were fitted to data of all administered doses with model predictions for the respective doses portrayed here. Treatment commenced three days after inoculation in dosing intervals of 24 hours. Mice were treated with 4 × 30 mg/kg ACT-451840, 4 × 50 mg/kg CQ, 2 × 10 mg/kg MMV390048 or 2 × 10 mg/kg OZ439. The cessation of human RBC injections in ACT-451840 and CQ experiments seven days after treatment leads to a decay of human RBCs and therefore also parasitemia 10–15 days after treatment (a,b). The horizontal dashed line represents the lower limit of quantification with 0.01% parasitemia. n = 2 mice for all doses shown.

Figure 6

Schematics of P. falciparum parasite dynamics in SCID mice after treatment and potential factors explaining variance and uncertainty. (a) The mechanistic models (orange) presented in this paper assume parasite growth characteristics remain constant throughout treatment and are therefore not capturing late recrudescence. This is in contrast to the exponential model (gray) that compensates for late recrudescence by shifting the curve to low parasite and drug concentrations. Alternative to our mechanistic models, we propose some hypothetical parasite recrudescence curves (blue and green), that include additional phenomena such as altered parasite maturation and parasite dormancy offering possible explanations for late recrudescence. We cannot capture these mechanisms with models without additional data. MIC estimates important for experimental interpretation and translation to humans are shown by yellow square points and are likely to be very different given assumptions about parasite growth behavior after treatment. (b) We hypothesize and extended the sources of variance and uncertainty of the parasite treatment curve described in to schematically illustrate parasite phenomena during growth, treatment and recrudescence for antimalarial experiments (murine and possibly human). These extended phenomena include altered parasite maturation, dormancy, and stochastic extinction occurring below the lower limit of quantification hindering estimation.