Background
Malaria remains a global health emergency, with an estimated 228 million cases and over 400,000 deaths occurring worldwide in 2018 [
1]. While control measures, such as insecticide-treated nets, sensitive diagnostics tools and effective case management, have helped reduce the number of cases and deaths, progress has stalled in recent years [
1]. New tools, including novel drugs and vaccines, are urgently needed, particularly in the areas of highest endemicity.
Recently, pilot implementation programmes to deliver the RTS,S vaccine in select areas of three African countries began [
2]. RTS,S is effective in young children, but efficacy wanes over time [
3]. RTS,S elicits responses to the circumsporozoite protein (CSP), the major surface protein on the infecting sporozoite [
4]. Antibodies against the NANP repeat regions of CSP induced by RTS,S have been linked to its protective efficacy and these wane after vaccination with kinetics that mimic the decline in efficacy [
5]. A second vaccine that targets CSP, R21, has also shown efficacy in nonclinical studies [
6] and recently in clinical testing [
7]. Radiation attenuated sporozoites, PfSPZ, administered by intravenous (i.v.) injection have shown efficacy in some populations [
8] and induce both antibody responses to CSP and cellular responses to multiple antigens expressed during the liver stage [
8].
Monoclonal antibodies (mAbs) to CSP have been proposed as new infection-blocking interventions [
9‐
12], and protection by mAbs can be expected to wane with the half-life of the antibody in serum. While altering the sequence of mAb Fc regions has been used to extend serum half-life of mAbs [
13], an important additional approach for improving the durability of protection is to identify mAbs effective at lower concentrations. Identification and development of such potent antibodies will require assays that reliably measure their functional activity.
New molecular tools have aided identification and development of human mAbs that bind to CSP. [
11,
14,
15] from humans naturally exposed to
Plasmodium falciparum [
10], volunteers vaccinated with RTS,S [
16], or from whole sporozoites [
17]. The functional activity of these antibodies varies, with potent antibodies binding to the NANP repeat region [
10,
11,
16] and to a region just upstream of the repeat domain termed the bridging peptide or junctional region [
17,
18].
The current study focuses on the use of two in vivo models of
Plasmodium sporozoite infection to measure the functional activity of anti-PfCSP mAbs.
Plasmodium parasites are highly host species specific, with
Plasmodium berghei efficiently infecting mice while
P. falciparum fails to do so.
Plasmodium berghei parasites can be engineered to replace the native CSP with
P. falciparum parasites [
19]. These parasites can be further engineered to express luciferase upon liver infection, allowing simple measurement of infection by measuring luciferase-induced luminescence from infected liver cells. Parasites can be administered by i.v. injection and mosquito bite challenge. Functional antibody activity can be measured either as the ability to inhibit liver infection measuring a luminescence endpoint or to block infection entirely as monitored by the appearance of parasitaemia in the blood. Methods used to conduct these assays were recently described [
20]. An alternative model has also been reported in which mice have been engineered to contain human hepatocytes which allows infection by
P. falciparum sporozoites [
21,
22].
The experimental conduct of in vivo protection assays is inherently complex, requiring the consistent execution of multiple steps including preparation of the infectious challenge dose as either isolated sporozoites or infected mosquitoes, consistent delivery of the protective mAb in animal models, as well as monitoring of the infection endpoint. As the overall goal for such assays is to reliably identify antibodies with high potency, results were compared for multiple experiments conducted with the same methods and reagents. To assess each assay’s utility in selecting the most potent mAbs results were collected on intra- and inter- experiment consistency using two mAbs with high protective potency [
16]. The results demonstrate that consistent information on mAb potency can be obtained using the procedures applied and provide guidance for the selection of superior mAbs as clinical candidates. The results may also be extendable to the testing of improved CSP targeted vaccines.
Methods
Materials
Animals: Female mice C57Bl/6 6–7 weeks of age were purchased from Charles River Laboratories, Frederick MD USA. All studies were performed under the protocol MOI7H369, approved by the ACUC at JHU.
Antibodies: AB311 and AB317 are human immunoglobulin G 1 (IgG1) mAbs isolated from an experimental clinical trial of RTS,S, MAL071 [
23] clinical trial and both mAbs bind to NANP repeats [
16]; mAb1245 is also a human IgG1 mAb, isolated from a Kymab mouse, and binds to a
P. falciparum ookinete protein Pfs25, and thus is used as a negative isotype control [
24]. All mAbs were expressed by transient transduction 0.5 L TunaCHO cultures followed by protein A purification at Lake Pharma Inc. Belmont, CA.
Parasites: Transgenic sporozoites in
P. berghei expressing
P. falciparum CSP, green fluorescent protein and luciferase reporter gene, used in all studies, has been previously described [
20] and methods for parasite preparation have been described in detail [
20]. Briefly, 5-day old adult
Anopheles stephensi mosquitoes were allowed to feed on mice carrying 1 to 2% transgenic parasites. Twenty to 22 days post murine blood meal, transgenic sporozoites were collected from salivary glands and used within 60 min. for i.v. infection for liver burden studies. For parasitaemia studies, infectious mosquitoes were used directly.
Reduction in parasite liver burden assay
As previously described [
20], mice were injected intravenously in the tail-vein with test and control antibodies at the indicated concentration in PBS (200 µL). Challenge with transgenic sporozoites occured 16 h following mAb administration. Forty-two hours after parasite challenge, parasite load in the liver was measured by bioluminescence in an in vivo imaging system (IVIS Spectrum, Perkin Elmer). Mice were injected intraperitoneally with 100 µL of
d-luciferin (30 mg/mL) and immediately anesthetized with isoflurane for 5 min prior to IVIS. Groups of five anesthetized mice were placed in the imager and the radiance measurements recorded by the live imager software, version 4.5.1. The total flux reading for each mouse was recorded individually. Background reading was verified for each study with two naïve mice that received only the
d-luciferin substrate; these are reported in Table
1 as naïve controls.
Table 1Assay consistency across studies
600 | Not applicable (N/A) | 5.38 | 0.05 | 7.40 | 0.11 | 5.31 | 0.26 | 5.40 | 0.13 | 5.31 | 0.05 |
300 | 5.83 | 0.24 | 5.84 | 0.22 | 5.84 | 0.25 | 5.56 | 0.17 | 7.43 | 0.16 | 5.82 | 0.18 | 5.77 | 0.10 | 5.58 | 0.17 |
100 | 6.41 | 0.20 | 6.57 | 0.17 | 6.53 | 0.35 | 6.46 | 0.21 | 7.32 | 0.08 | 6.01 | 0.21 | 6.25 | 0.23 | 6.16 | 0.15 |
30 | 7.06 | 0.11 | 7.05 | 0.12 | N/A | 7.03 | 0.11 | 7.47 | 0.06 | 6.84 | 0.12 | 6.52 | 0.15 | 6.64 | 0.16 |
10 | | | | | | | | | | | 6.95 | 0.12 | | | | |
Untreated | 7.62 | 0.06 | 7.48 | 0.05 | 7.46 | 0.08 | 7.47 | 0.15 | 7.47 | 0.15 | 7.22 | 0.10 | 7.25 | 0.06 | 7.20 | 0.15 |
Naïve | 5.09 | 0.09 | 5.11 | 0.03 | 5.03 | 0.18 | 5.08 | 0.13 | 5.07 | 0.12 | 5.24 | 0.03 | 5.25 | 0.02 | 5.25 | 0.03 |
Protection from mosquito bite challenge assay
As previously described [
20],
An. stephensi mosquitoes were fed on mice infected with chimeric
P. berghei parasites encoding full-length
P. falciparum CSP. In order to determine the proportion infected, 19–20 days after blood feeding on mice, a few mosquitoes were dissected to determine whether sporozoites were present in salivary glands.
Sixteen hours after administration of mAbs, C57Bl/6 mice were anesthetized with 2% Avertin and exposed to five infected mosquitoes for ~ 10 min. The number of mosquitoes that fed on blood was determined by observation of a red abdomen. Days 4–12 post challenge, blood smears from the tip of the mouse’s tail were collected, stained with 10% Giemsa, and examined by light microscopy to determine the presence of blood stage parasites.
Monoclonal antibody serum pharmacokinetics
Blood was collected from mice via retro-orbital sinus one hour prior to challenge (fifteen hours after prophylaxis). Blood was processed to serum using standard methods and stored at 4 °C until assayed. Three-fold dilution series against Nunc MaxiSorp™ plates (ThermoFisher Scientific) coated with 5 ng/mL recombinant CSP were assayed in standard ELISA assays using HRP-conjugated Goat anti-Human IgG (H + L) secondary antibody (Jackson ImmunoResearch) and ABTS peroxidase substrate (KPL). Human IgG concentration in serum was determined by comparison to mAb standard curves.
Analysing parasite liver burden assay repeatability
Overall inter- and intra- study liver burden assay variability was assessed by analysing the standard deviations (SD) of log10 total flux estimated using random effects models with log10 flux as the outcome and a random intercept specified for each study. This model was fit to the naïve, infected control data to assess both assay and infectivity variability. For treated animals (mice receiving potent mAbs), a fixed effect for dose was also included to control for treatment effect (i.e., a linear mixed effects model was fit). From these models, inter-experiment variance was estimated as the variance of the random effect and intra-experiment variance was estimated as the residual variance. Estimated standard deviations of log10 flux were transformed back (unlogged) to be interpreted as fold-change deviations.
Dose–response modelling and estimating ID50 and IC50
The relationships between dose or circulating mAb and assay outcomes were modelled using four-parameter logistic (4PL) models with the following functional form: y = L+(U − L)/(1 + (x/ID50)h).
Where:
L is the minimum value of the assay reached as the protective dose or circulating mAb levels increase toward infinity, or measured in negative control animals (i.e., lower limit or lower saturation);
U is the maximum value of the assay for dose or circulating mAb of 0, or measured in naïve, infected animals (i.e., upper limit or upper saturation);
ID50 is the dose (IC50 for circulating antibody concentration) where the outcome is 50% reduced relative to U and L (the point of inflection); and
h is the Hill slope determining steepness in the linear section of the curve.
For the liver burden assay, the outcome modelled was log10 flux. The upper limit was estimated by including the naïve, infected controls as equivalent to a dose or circulating mAb concentration of 0 in the given experiment. Thus, each dose–response curve within an experiment was adjusted for the infectivity of the challenge strain for that study. The lower limit was fixed to 5 log10 flux based on all of the negative control data. The ID50 (or IC50) then represents the dose (or circulating mAb concentration) at which flux is 50% reduced relative to the upper and lower log10 flux for that study.
For the protection from parasitaemia assay, the lower and upper limits were fixed ranging from 0 to 100% infected. As only the ID50 or IC50 and the slope were fitted, this is a 2PL model. Here, the ID50 (or IC50) represents the dose (or circulating mAb concentration) where animals have a 50% probability of infection (or are 50% protected).
Statistical methods for comparing liver burden reduction
To compare mAb potencies using the liver burden assays, 4PL models were fit to both mAbs within a study and the estimated ID50s and IC50s (on the log-scale) were compared using a t test. Comparisons were performed by pooling the data across the studies where both mAbs were tested.
Statistical methods for protection from parasitaemia
Protection from parasitaemia was determined by either 1) the proportion of uninfected animals remaining after the observation period post-challenge, or 2) the observed infection day following challenge. As the liver burden assay established higher potency of AB317 compared to AB311, all statistical comparisons were one-sided testing for superiority of AB317 over AB311. Within a study, at a single dose, protection was compared using Barnard’s exact test (Z-pooled method, [
25]) for each dose with at least partial protection observed in any study (100 μg and 300 μg). Survival curves were also fitted for each experiment, grouping by antibody and dose. To test the null hypothesis that there were no differences in survival time between mAbs, a log-rank test using the exact conditional distribution with the Hothorn-Lausen tie-method [
26] was conducted by dose group.
Two approaches were used to compare protection between the antibodies across all of the doses: (1) comparison of ID50s or IC50s (on the log-scale, one-sided) from fitted 2PL models with a common Hill slope for both mAbs using a z-test; and (2) testing for a significant odds ratio between mAbs using a logistic regression with log dose and mAb as predictor.
The 2PL model and logistic regression including log dose are equivalent models if a common Hill slope for both mAbs is fitted in the 2PL model or an interaction term between mAb and log dose is specified for the logistic regression. For the logistic regression, a small sample size correction (Firth-correction) was used to robustly estimate the odds ratio and one-sided 95% confidence intervals through profiling [
27]. A one-sided test was then performed for AB317 superior potency (odd ratio < 1) by comparing the upper bound of the CI to 1. Effectively, the logistic regression approach represents a strategy to accommodate for the small sample size constraints of these data by assuming differences in potency are represented strictly by shifts in the ID50 or IC50 and not by steeper or shallower curves.
Power analysis and sample size determination for study design
Power calculations were performed under the framework that candidate mAbs would be compared to AB311 as a reference in a single dose study for both the liver burden and parasitaemia assays. For the liver burden assay, it was assumed that the chosen dose would represent the ID50 of the reference mAb (AB311) and the candidate mAb would be tested against the reference using a t-test comparing log10 flux. The power for the t-test comparison was then calculated by varying effect size and a sample size. Effect size was determined based on a range of theoretical ratios of flux between mAbs (differences in log10 flux) and variation measured within the study. The range of flux ratios represents potential increasing potencies for candidate mAbs. Three levels of variation were chosen using the range of standard deviations from the studies to represent low, average, and high variability scenarios. Sample size varied from 5 to 12 per group to represent an experimentally practical range for total mice within a single study. Tests were two-sided with an alpha = 0.05 (5% false positive rate).
To calculate power for the protection from parasitaemia assay, a single dose of 300 µg was chosen to optimally incorporate the empirical data and the Barnard’s exact test was then used to compare protection, matching the data analysis. Potential candidate mAbs were considered with potencies ranging from full protection (100% protection) to equivalent reference AB311 protection at 300 µg (averaged protection across the studies). Power was then calculated across a sample size of 6–12 animals. Tests were one-sided for superiority of the candidate mAb with an alpha = 0.05.
For additional interpretation in the power analysis, ID50s were also estimated from the assay endpoints for the candidate mAbs across the range of tested potencies. For the liver burden assay, this was done by estimating a log
10 flux based on the fold reduction at the given dose and then deriving the ID50 from the 4PL model using the upper bound (7.22 log
10 flux) and Hill slope [
1] parameters estimated for AB317, a proxy for potential candidate mAbs. Similar for the protection from parasitaemia analysis, the ID50 was derived from the 2PL model using the protection estimate at 300 µg and a fixed Hill slope (common estimate for both AB311 and AB317). Once the candidate mAb ID50 was estimated, the ratio was taken relative to the mean AB311 ID50 by assay.
Statistical software
Statistical analysis was performed using the R programming language [
28] Linear mixed effects models for repeatability analysis were fit using the lme4 package [
29]. The model fitting for 4PL and 2PL models and the ID50 and IC50 comparisons were performed using the drc package [
30]. The log-rank test was implemented with the coin package [
31]. Comparisons and power calculations using Barnard’s exact test (Z-pooled method) were performed using the Exact package [
32]. Visualizations were generated using the ggplot2 package [
33].
Discussion
In order to identify and select novel vaccine and mAb interventions that target the CSP protein of P. falciparum, it is important to understand how preclinical assays can be used to measure and compare the functional potency of antibody preparations. This study analyses the consistency of results obtained in two in vivo protection assays performed in one laboratory with standardized procedures. Well-defined mAbs were used in order to standardize the input test article and it should be recognized that results described here will need to be confirmed using polyclonal sera to assess vaccine potency.
The reduction in parasite liver burden model has the advantage of providing quantitative information on the amount of liver infection while the protection from parasitaemia assay measures a more clinically relevant outcome. In both assays, challenge of mice used an engineered
P. berghei parasite in which the natural CSP gene has been replaced by that of
P. falciparum [
34]. This challenge system may not fully replicate the desired biology of
P. falciparum infection of the liver [
35‐
37]; however, it can test
P. falciparum interventions efficiently at relatively low cost in immunocompetent mice [
34]. Assays using a similar approach with transgenic sporozoite expressing
Plasmodium yoelii strain specific CSP have been published by other groups [
38]. A system in which mice are rendered susceptible to
P. falciparum infection, via engrafted human liver cells [
22] may also have application to the testing of mAbs and vaccines. Direct comparisons of results for the determination of functional potency of mAbs among these model systems are warranted.
The current study strengthens comparative testing of CSP-specific mAbs by evaluating intra- and inter-assay variability. The results show these in vivo model systems can be highly reproducible. The ability to consistently prepare and deliver sporozoites for i.v. challenge is indicated by the low intra- and inter-experiment variation in log10 total flux for untreated control and mAb prophylaxis groups. The mosquito bite protection model also showed good consistency between three experiments with five infected mosquitoes causing infection in all mice in controls as well as allowing complete blockade by high doses of functional mAbs. These results indicate that infectious mosquitoes can be prepared to produce a consistent parasite challenge. Inter-assay consistency was also evident in the calculations of potency of the tested mAbs. Using the liver burden model, seven independent determinations of the ID50 for AB311 (Fig.
1) had inter-experimental overlapping confidence intervals. For AB317, consistent ID50 measurements were also determined. Similar conclusions can be drawn from the analysis of the protection from parasitaemia model with consistent ID50s calculated across three experiments for each antibody.
The results provide guidance on the ability of the assays to discriminate the relative functional activity of test articles. Multiple comparisons of AB311 and AB317 in both assays consistently showed more potency for AB317. In the liver burden assay, approximately a 1.5-fold difference in activity may be close to the limit of the assay to discriminate differences in functional activity using experimental designs reported here. This is inferred from the fact that a difference in ID50s between AB311 and AB317 could be detected (with P < 0.05) in one of two experiments and in the pooled data (Table
2). In the protection from parasitaemia assay, it was difficult to discriminate ID50 between the mAbs in all experiments and in the pooled data using only the sterile protection endpoint. Log-rank comparisons of time to patency between the two mAbs could discriminate between mAbs in most but not all experiments. Again, this suggests it may be challenging to reliably discriminate differences of the level exemplified by AB311 and AB317 in experiments with these group sizes. The results suggest that considering time to infection in the protection from parasitaemia assay as a measure of protective activity may be helpful in discriminating differences in potency; however, larger group sizes are needed given the highly significant discrimination of functional activity seen in the pooled result (Table
7).
Calculation of IC50 using circulating concentration of mAb at the time of challenge did not aid in the detection of differences in functional activity of the mAbs. Both AB311 and AB317 were produced as human IgG1, and the serum concentration was similar between the mAbs at a given dose and the 15-h timepoint monitored. It is important to recognize that similar pharmacokinetics may not occur for all mAbs tested, and consequently, measurement of serum concentration does add to the information on relative potency of mAbs compared.
The analysis of results reported here also provide guidance for how these assays can be used in the future to screen for mAbs with higher functional potency. For the parasite liver burden model, experiments can be designed to detect differences in activity exemplified by AB317 and AB311 using a single administered dose and modest group sizes of six with the important caveat that a low level of intra-assay variability is minimized. The power analysis shows that larger group sizes are required to detect differences in antibody potency when there is an increase in the observed standard deviation of measurements of infection within groups. This has important implications for establishing this assay at a new testing laboratory site.
The power analysis based on the protection from parasitaemia assay has two important implications for the design of experiments intending to identify more potent CSP antibodies. First, it will be important to identify a dose under which the control mAb (as exemplified by AB311) is consistently partially protective. Second, substantial group sizes (N > 10) will be needed to identify a more potent mAb that confers a protection increase from 40% in the reference mAb to 84% for the novel mAb (equivalent to approximately a 1.5 difference in ID50).
It is encouraging that the liver burden model and parasite protection model gave similar results when comparing the two antibodies tested. ID50 ratios between the mAbs were consistent across assays: 1.5-fold for liver burden and 1.4-fold in the protection from parasitaemia assay (comparing the pooled results (Tables
2 and
4). ID50 estimates were about threefold higher in the protection from parasitaemia compared to reduction in parasite liver burden for both mAbs. Based on these findings, it seems reasonable to expect the antibodies that perform well in reducing parasite liver burden will also perform well in a mosquito bite challenge model. Given the ease of experimentation and lower group sizes, it seems practical to screen for mAbs using the liver burden assay and confirm using the protection from parasitaemia assay. However, this relationship may not extend to antibodies that have different mechanisms of action, such as those that bind outside the NANP repeat region of CSP. A mechanistic understanding of why AB311 and AB317 display different functional activity is outside the scope of this study. It is clear that these antibodies bind differently as determined by structural studies using X-ray crystallography and by electron microscopy [
16,
39]. However, the association of binding properties with functional activity remains a very active area of research and such investigations should be aided by the information in the current report.
Conclusion
The results reported here indicate that in vivo assays measuring reduction in liver burden and protection from parasitaemia following mosquito bite challenge can provide relevant information for understanding comparative functional activity of mAbs when performed using standardized reagents and protocols. For full assay qualification and validation, which will be useful for the broader malaria research field, additional data will need to be generated and analysed. For example, in this study variation introduced by inter-laboratory or inter-personnel differences or the robustness of the assays to variation in protocol and reagents were not explored. Until such data is generated and analysed, it is recommended that standardized comparator reagents be used across laboratories; the antibodies described in this manuscript could serve this purpose. In conclusion, the results reported here indicate that these two assays can be a valuable tool to assess any newly isolated mAbs as well as to probe specific structural elements of mAbs for their impact on function. It is also anticipated that these assays will be useful as mAbs are designed and optimized for desired product characteristics for use in malaria prevention.
Acknowledgements
The authors acknowledge many contributions on experimental planning, design and interpretation by Holger Kanzler and Jacqueline Kirchner (Gates Foundation) and for many helpful suggestions on the manuscript contributed by Randall MacGill and Ashley Birkett (PATH). The authors also thank the Insectary and Parasitology Core Facilities of Johns Hopkins Malaria Research institute. FZ, YFG, HJ, SH and DPR thank the Bloomberg Philanthropies for continued support”.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.