In vitro and in vivo inhibition of malaria parasite infection by monoclonal antibodies against Plasmodium falciparum circumsporozoite protein (CSP)

Plasmodium falciparum malaria contributes to a significant global disease burden. Circumsporozoite protein (CSP), the most abundant sporozoite stage antigen, is a prime vaccine candidate. Inhibitory monoclonal antibodies (mAbs) against CSP map to either a short junctional sequence or the central (NPNA)n repeat region. We compared in vitro and in vivo activities of six CSP-specific mAbs derived from human recipients of a recombinant CSP vaccine RTS,S/AS01 (mAbs 317 and 311); an irradiated whole sporozoite vaccine PfSPZ (mAbs CIS43 and MGG4); or individuals exposed to malaria (mAbs 580 and 663). RTS,S mAb 317 that specifically binds the (NPNA)n epitope, had the highest affinity and it elicited the best sterile protection in mice. The most potent inhibitor of sporozoite invasion in vitro was mAb CIS43 which shows dual-specific binding to the junctional sequence and (NPNA)n. In vivo mouse protection was associated with the mAb reactivity to the NANPx6 peptide, the in vitro inhibition of sporozoite invasion activity, and kinetic parameters measured using intact mAbs or their Fab fragments. Buried surface area between mAb and its target epitope was also associated with in vivo protection. Association and disconnects between in vitro and in vivo readouts has important implications for the design and down-selection of the next generation of CSP based interventions.

. Structural analysis of mAb binding to protective P. falciparum Circumsporozoite protein (CSP) epitopes. (a) Schematic representation of full-length CSP (not to scale). (b) Antibody paratope structures depicting monoclonal antibody binding to representative repeat or junctional peptides: 311 (PDB code 6AXK), 317 (6AXL) 18 , CIS43 bound to peptide 21 (6B5M) and peptide 29 (6B5O) 12 , germline 580 (6AZM) and 663 (5BK0) 19 , MGG4 (6BQB) 21 . The contact area between the peptides and the complementarity determining regions, or the buried surface area (BSA), shown under each structure were calculated using the AREAIMOL program in the CCP4 software suite 43 . Heavy and light chains are colored blue and light grey, respectively. Peptides are depicted in stick conformations (green carbons). Hydrogen bonds are shown as black dashes. Type I β and 3 10 turns are indicated by black and red circles, respectively. www.nature.com/scientificreports/ such that mAb 580 has a lower buried surface area (BSA) and forms a lower affinity complex. For less common high affinity mAbs with a larger peptide recognition surface, rather than there being a core NPNA motif, two or three motifs are recognized. For example, mAb 317 recognizes three successive NPNA type I turns in an extensive interaction that involves many heavy as well as light chain contacts 18 . Mab CIS43 preferentially recognizes the junctional sequence and has DPNA NPNV containing two type I β turns at the center of the peptide-mAb complex 12 . Although, dual-specific mAb CIS43 preferentially recognizes the junctional sequence DPNA NPNV, it also binds to NPNA NPNA albeit with ~ tenfold lower affinity 12 .
Sporozoites delivered under the skin through the bite of an infected mosquito reach the liver where they traverse through hepatic tissue before invading a hepatocyte. Traversal inhibition assays [26][27][28] , exoerythrocytic stage development inhibition assays 19,29,30 and humanized mouse liver parasite burden inhibition assays 31 have been used to dissect the mechanism of anti-CSP activity. Tan et al. showed that mAbs like MGG4 that recognized the junctional sequence were potent neutralizers 21 and Kisalu et al. showed high level inhibition of hepatocyte invasion and protection by the dual-specific mAb CIS43 12 . Murugan et al. screened a panel of such cross-binders (including mAbs 4493 and CIS43) along with epitope-specific mAbs (including mAb 317) to demonstrate that high affinity to (NANP) n determined the traversal inhibition activity and there was no significant difference in the in vivo protective response of high affinity cross-binding, dual-specific and epitope-specific mAbs against transgenic parasite challenge 22 .
We functionally characterized a panel of repeat region mAbs to enable rational vaccine and immune-prophylactic development. MAbs 317, 311, CIS43, MGG4, 580 and 663 were compared head-to-head by direct ELISA (quantity, binding strength and specificity), avidity ELISA (binding strength), biolayer interferometry assay (kinetic parameters and affinity), in vitro inhibition of liver stage development assay (ILSDA), in vivo protection (transgenic parasite challenge) and the sporozoite reaction assay (precipitation of CSP on live sporozoites). MAb 317 emerged as most protective in vivo but mAb CIS43 was a better inhibitor of sporozoite invasion into human hepatocytes in vitro. Overall, our data strongly supports the continued evaluation of CSP mAbs as immuno-prophylactics.
To classify binding specificities, a peptide ELISA was conducted comparing binding to the repeat peptide NANPx6, peptide-21 (NP-DPNA-NPNV-DPNA-N) 12 , and peptide NPDP-19 (KQPADGNP-DPNA-NPNV-DPN) 21 (Fig. 3). MAbs 317 and 580 were confirmed as NPNA-specific (binding NANPx6 but not peptide 21 or NPDP 19 ) 18,19,32 . MAbs MGG4, 663, and 311 bound NANPx6 and NPDP-19 but not peptide 21, and were categorized as "cross-binders" according to Murugan et al. 22 . MAbs CIS43 bound all three peptides confirming that it was truly "dual-specific" for the junctional and (NPNA) n sequence as described by Kisalu et al. 12 . Affinity by biolayer interferometry (BLI). Affinities of mAbs within this panel were reported previously against a variety of peptide and protein targets (Table S1). To directly compare their affinities head-to-head, the monovalent Fabs and bivalent mAbs were tested by a BLI-based assay to determine the association rate constant (k a ), dissociation rate constant (k d ) and equilibrium dissociation constant (K D ) against the NANPx6 peptide and CSP antigen (raw data: Figs. S2, S3, Table S2).
Rate constants k a (10 4 to 10 5 M −1 s −1 ) and k d values (10 -1 to 10 -7 s −1 ) yielded wide ranging K D values for Fab fragments binding to the NANPx6 antigen ( Table 1). The highest affinity to NANPx6 peptide was obtained for Fab 317 with the slowest dissociation rate (k d = 1.7 × 10 -7 s −1 ) and lowest dissociation constant (K D = 3.0 pM). The lowest affinity to NANPx6 was obtained for Fab 580, with the fastest dissociation rate and the highest dissociation constant (K D = 2.6 μM) ( Fig. 4A; y-axis). The affinity of the dual-specific Fab CIS43 for NANPx6 (K D = 41 nM) was four orders of magnitude lower than the NPNA-specific Fab 317 (Table 1). A parallel BLI analysis against the nearly full-length FL-CSP antigen also confirmed the highest affinity for Fab 317 (K D = 2 pM); 800-fold higher than Fab CIS43 and 2 × 10 5 -fold higher than Fab 580 ( Fig. 4A; x-axis). For Fab fragments, the ratios of K D values against NANP/CSP were ≥ 1. In particular, for the cross-reactive Fabs CIS43 and 311, the binding affinities to CSP were 24 and 36-fold higher than to NANPx6, respectively (Table 1). Pearson correlation analysis (Table S3) showed a positive association between Fab affinity to NANPx6 and to the FL-CSP antigen (r = 0.98; p = 0.001), suggesting (NPNA) n was likely a preferred binding epitope on FL-CSP. Bivalent (intact) mAb affinities were also determined against the FL-CSP antigen 33 (Table 1). Association rate constant k a for the bivalent mAbs were similar (10 5 -10 6 M −1 s −1 ) and approached the diffusion limit for biological molecules 34 . In-line with the Fab data, intact mAb 317 affinity (K D = 30 pM) was the highest and the difference between mAb 317 affinity and that of mAbs CIS43, 663 and 580 were statistically significant (all p values < 0.05) (Fig. 4B). Pearson correlation analysis for bivalent versus Fab affinities to CSP showed a positive association (r = 0.99; p < 0.001) (Table S3). Remarkably, bivalent forms of three low affinity mAbs 580, 663 and MGG4 bound to CSP with 83 to 275-fold higher affinity than their respective Fab fragments, whereas conversely, the highest affinity Fab 317 bound to CSP with 12-fold higher affinity than its bivalent form ( Avidity by ELISA. The term avidity refers to the average strength of vaccine-induced polyclonal antibody binding to a target antigen 8,[35][36][37][38] . In an ELISA, the proportion of mAb that remained bound to CSP or NANPx6 (avidity index; AI), after a 2 M sodium thiocyanate (NaSCN) wash, was estimated on a 0-100% scale 39 (Fig. 5A, B). In an alternative ELISA methodology, 0-6 M NaSCN wash solutions were used, and the molarity of NaSCN that reduced optical density by half-maximal provided an estimate of avidity (Fig. 5C, D). Average avidity deter-   Table S3). The two avidity assay methodologies confirmed BLI data against both FL-CSP and NANPx6 target antigens that mAb 317 bound its target epitope with the greatest strength. Avidity differences between mAb 317 versus mAbs 580, CIS43, MGG4 and 311 were all statistically significant ( Fig. 5; p values < 0.05). MAb CIS43 showed a disproportionately low avidity by ELISA against the NANPx6 peptide (Fig. 5B, D) which was consistent with its preference for the junctional region of CSP. Overall, a slower dissociation rate and a lower dissociation constant corresponded to higher avidity as was alluded to by Dennison et al. 33 (r = − 0.71, p < 0.0001; Fig. S4C, D; Table S3).
In vivo protection. Intravenous transfer of mAbs in C57BL/6 mice was followed by challenge with 1000 transgenic P. berghei sporozoites, expressing P. falciparum CSP, via the intravenous route 40 . The absence of blood stage parasites over the next 14 days indicated sterile protection. To discern differences between mAbs, contingency table analyses were performed by Fisher's exact test and Kaplan-Meir survival curves were compared by a Log-rank test.
Experiment 1: 100 µg of each mAb was transferred and the mice were challenged at 24 h (n = 5) or 48 h (n = 5) post transfer (Fig. 7A, B). At 24 h, mAb 317 protected 100% mice compared to 0% of controls. MAbs 311 and 663 protected 60% mice and mAb CIS43 protected 40%. Fisher's test showed higher frequency of protected mice for mAb 317 compared to the controls, mAb 580 and mAb MGG4 group (p values = 0.004) ( Fig. 7A; Table S8). The five remaining mice were challenged at 48 h and mAb 317 showed 80% protection. MAb CIS43 protected 40% mice which was also consistent with the 24 h challenge. Unlike 0% and 30% protection for mAbs MGG4 and 311 in the 24 h challenge respectively, 60% and 80% protection were observed at 48 h challenge. The higher protection was likely due to the low mouse numbers (n = 5) and reduced sporozoite infectivity at 48 h challenge time-point. Fisher's test on protection frequency for mAbs 311 and 317 groups showed significantly higher protection than the controls and mAb 580 (p values = 0.024). Log-rank test confirmed these inferences ( Fig. 7B; Table S8).  Table S8). Compared to no protection in the controls, mAb 317 showed the best protection (90%) and CIS43 protected half of the challenged mice (50%). MAb MGG4-mediated protection (50%) was consistent with the 48 h challenge result from Experiment 1. MAb 311 protection (30%) was lower than mAb 317 similar to that observed in the 24 h challenge of Experiment 1. Protection frequency compared by Fisher's test showed mAb 317, CIS43 and MGG4 protection were higher than the controls (all p < 0.02), and mAb 317 protected significantly better than mAb 311 (p = 0.01). Log-rank test had consistent inferences (Table S8). Experiment 3: MAbs 317 and CIS43 were tested head to head at 50 and 15 µg (n = 10 per group) and mice were challenged at 24 h post transfer (Fig. 7D). Compared to no protection in the controls, mAb 317 showed 80% and 50% protection while mAb CIS43 protected 50% and 60% mice in the high and low dose groups respectively. Fisher's test showed both mAbs displayed significantly higher protection than the controls. At this lower dose, no difference in protection outcome between mAbs or dose levels was observed (Table S8). Log-rank test had consistent inferences.
Overall, the control mice were always infected, mAb 317 showed the best protection and mAb CIS43 protected about half the mice even at the lower doses.

Sporozoite binding assay.
To study the interaction of the mAbs with the transgenic sporozoites used during the challenge studies, a direct ELISA was performed against transgenic sporozoites. In two independent experiments, mAbs 317 and 311 bound to the sporozoites more efficiently than mAbs MGG4, 663 and CIS43, while mAb 580 bound least efficiently (Fig. 8A, B). CSP mAb binding to live sporozoites can cause a CSP reaction observed microscopically as a tail-like precipitate 41,42 . MAbs 311 and 317 (at 100 ng/mL) showed a positive CSP reaction, whereas no clear reaction was observed at this concentration with the other CSP mAbs (Fig. 8C).
Associations with in vivo protection. An exploratory Pearson's correlation analysis was performed on raw in vitro data and the corresponding in vivo protection frequency from independent challenges performed as part of Experiment-1 and 2 (Table S3). Several strong (correlation coefficients |r|> 0.7) to moderate linear associations (|r|= 0.4-0.7) are listed in Table 2. Direct ELISA reactivity (ng/ml mAb needed for OD = 1) against a short repeat peptide (NANPx6), showed a consistent negative association. Inhibition of sporozoite invasion, measured by ILSDA at 0.5 µg/ml, showed a consistent positive association. Strong and consistent associations were found with FL-CSP and NANPx6 avidity assays; likewise, strong negative associations were observed for K D values, measured by BLI using bivalent and Fab fragments. Finally, an estimate of the size of the interface between  www.nature.com/scientificreports/ mAbs and their epitopes called the Buried Surface Area (BSA) has been shown to be associated with CSP mAb affinity 19 . We calculated BSA from mAb crystal structures (Fig. 1) using AREAIMOL program 43 and found a strong positive association with protection. The most protective mAb 317 showed the highest BSA (> 500 Å 2 ).

Discussion
Short unstructured peptides are not particularly immunogenic due to a conformational entropy cost of stabilizing flexible epitopes 44,45 , but the repetitive and flexible repeat region of CSP is quite immunogenic 16 . In addition to cross-linking the B-cell receptors 15 , NPNA, DPNA and NPNV repeats are rich in asparagine residues whose sidechain has lower flexibility and a propensity to hydrogen bond back to the main chain 46 . This narrows the main chain flexibility of CSP repeats to the terminal alanine and valine residues between the short structured motifs: like the type I and '3 10 ′ turns 32 . As a result, the partly constrained (NPNA) n and the junctional sequence can bind an assortment of antibodies by an induced fit process [47][48][49] . Six such repeat mAbs were tested head-to-head, in immuno-chemical, biological and challenge experiments with transgenic P. berghei expressing P. falciparum CSP. Since this study was initiated, other potent P. falciparum CSP mAbs have been reported from human recipients of PfCVac, PfSPZ vaccine and recombinant CSP vaccinated transgenic mice (Kymab) 32,50,51 . While some recently reported mAbs were not included, the comparative biophysical and immunological assessment of 6 mAbs with a wide range of affinities provided key insights into CSP antibody effectiveness and function. First, we found that steady-state assays with long NPNA repeats (like CSP ELISA) were not predictors of protective efficacy likely due to the flexibility of the target repeat epitope. Second, mAb 317 whose Fabs and bivalent forms showed the highest affinity were most protective in vivo; while mAb 580 with the lowest affinity, originally derived from an IgM 19 , was not protective. The high affinity of mAb 317 was driven primarily by its low dissociation rate. Third, antibodies like CIS43 bound to the full-length CSP with higher affinity than the NANPx6 peptide, suggesting targets outside the major (NPNA) n repeats were also important. Forth, bivalent forms of some lower-affinity mAbs (such as 663, MGG4, 580) bound more strongly than their respective Fab fragments. While Fab affinity plays an important role in repeat antibody mediated protection, cooperative interactions between Fabs do not always augment immune-protection 20,24 .
The complex interplay between Fabs and intact IgG binding to CSP reported here has important implications for vaccine design. For example, a blood stage merozoite enters a RBC within a minute 52 , but a sporozoite transit from skin to the liver can take up to 30 minutes 53 . We hypothesize that CSP retains the highly repetitive and flexible central region during evolutionary selection to allow certain low affinity antibodies to bind and compete with the truly high affinity antibodies, diluting the overall effectiveness of host immunity. Mechanistically a bivalent antibody, but not Fab fragments, precipitate CSP on sporozoites 54 . Lower affinity bivalent mAbs have a lower tendency to precipitate CSP as they may be coalescing along a single CSP molecule through homotypic interactions. In contrast, higher affinity mAbs may be cross-linking adjacent CSP molecules disrupting motility and impairing sporozoite cell traversal. In future experiments it may be interesting to engineer vaccines that elicit antibodies (Fabs) that use multiple weak interactions to bind to CSP and compare them to vaccines that promote 317-like antibody properties. Indeed, restricting epitope flexibility using short loops displayed on the Tobacco Mosaic Virus-like particle has been shown to improve immunogenicity, avidity and protective efficacy of a CSP repeat-based vaccine 16 .
Compared to the standard dose regimen, RTS,S delayed and fractional third dose (DFD) resulted in higher somatic hypermutation, avidity, and protection in humans 8 . Not surprisingly, the RTS,S-induced mAb 317  www.nature.com/scientificreports/ displayed the highest affinity within the panel, it showed efficient sporozoite binding by ELISA, a positive CSP reaction and low ID 80 < 0.5 µg/mL. Most importantly, mAb 317 protected 100% of mice at ~ 5 mg/kg (100 µg in a 20 g mouse) which is within the therapeutic dose range recommended for prophylactic mAb products against Ebola (Zmapp; ~ 50 mg/kg 55 ) and Respiratory Syncytial Virus (Palivizumab; ~ 15 mg/kg 56 ). In our model, mAb 317 protection dropped to 80% and 50% at 50 µg and 15 µg doses. Others have also reported mAb 317 protection to wane from 100% (at 300 μg) down to 80% and 0% (at 100 or 30 μg respectively) 32,57 . Like RTS,S, the efficacy of a NANP-specific prophylactic mAb may wane rapidly as the circulating mAb concentration drops. Overall, the RTS,S mAb 317 emerged as prototype of a highly protective mAb in our mouse protection model; mutations to improve its half-life and serum retention need to be evaluated. Previously mAb CIS43 was shown to protect 100% of mice at 300 µg dose against transgenic parasite challenge along with potent inhibition of liver stage development in vitro 12 . In our studies, mAb CIS43 consistently protected about half of the challenged mice at 100 µg dose, despite being the most potent mAb in the ILSDA assay. The lower in vivo protection of mAb CIS43 (compared to mAb 317) was reflected in its 20-fold reduced Fab affinity and low avidity for the NANPx6 peptide. A biological explanation for the disconnect between the in vitro and in vivo readouts for mAb CIS43 could be that the transgenic parasites do not fully recapitulate the molecular events that occur between P. falciparum sporozoites and human hepatocyte 6 . Alternatively, a fraction of transgenic sporozoites may be resistant to CIS43 mediated inhibition in vivo. Tan et al. and Kisalu et al. have hypothesized that junctional sequence mAbs have biological activities in addition to inhibition of sporozoite motility and traversal 12,21 . We also observed that a junction epitope cross-binder mAb MGG4 had the secondhighest ILSDA activity (after mAb CIS43) at the lowest mAb concentration tested (0.05 μg/mL). A clinical trial for mAb CIS43 containing a half-life extension mutation 58 is reported to be underway using 5, 20, and 40 mg/ kg doses and our ILSDA and efficacy data supports determining the protective role of junctional vaccines and mAbs in humans 51 .
Despite the impressive biological activity reported for some cross-reactive mAbs 12 , Oyen et al. showed that a cross-reactive mAb 668 was less protective than the (NPNA) n -specific mAb 317 32 . A definitive study using 200 mAbs (that included mAb CIS43 and 317), Murugan et al. concluded that high affinity, and not epitope specificity was the primary driver of repeat mAb protection and in vitro function 22 . In agreement with Murugan et al., we found that protection outcome was highly dependent on mAb affinity and avidity readouts. While we cannot assign a cut-off affinity for a highly protective antibody, mAb 317 emerged as an excellent benchmark for future attempts to improve anti-CSP affinity. A positive correlation between BSA and protection observed here suggested that in silico directed modifications may be useful in improving CSP mAbs. Extremely high affinity can however present other liabilities. For example, the RSV prophylactic Motavizumab, that had higher affinity and neutralization profile than Palivizumab (Synagis; Medimmune), also showed an unforeseeable non-specific binding to human tissue 59 . Indeed, a report on differences in circulating serum concentrations of some CSP mAbs caution against off-target binding of CSP mAbs (e.g. the cross-binder mAb 4493) 22 .
The WRAIR intravenous challenge model bypasses skin immunity 60 and protection may rely on a multitude of anti-parasitic effects of CSP antibodies including inhibition of motility and traversal 61 , Fc mediated opsonization 62 , complement mediated killing 63 , blocking of proteolytic processing of CSP 64 , or blocking of a putative receptor-ligand interaction 65 . This model measures protection in the blood which requires inactivation of 100% of the parasites in the inoculum and is sensitive to the biological and technical variability between challenge experiments for example the variation in protection outcomes between 24 and 48 h challenge with mAbs MGG4 and 311. While we used only one of several transgenic parasite lines and in vivo models available 66,67 , our data showed mAb 317 consistently elicited the highest protection at 100 µg dose across 3 independent challenges (100%, 80%, 90%) and mAb CIS43 consistently protected about half the mice (40%, 40%, 50%). Our data provide support for a CHMI trial comparing efficacy of a high affinity (NPNA) n -specific mAb (like 317) and a potent dual-specific mAb (such as CIS43). Furthermore, the associations between in vivo protection and immune-chemical assays reported here call for prioritizing affinity and avidity assays on Fab and bivalent mAbs in addition to ILSDA and in vivo challenge for the down-selection of future CSP-based interventions.

Methods
Production of monoclonal antibodies. Antibody fragment (Fab) sequences for monoclonal antibodies (mAbs) 311, 317, CIS43, MGG4, 580, and 663 were obtained from the literature and synthesized with a human IgG Fc sequence 12,18,19,21 . Plasmids containing codon optimized full-length IgG genes (ATUM, Newark, CA) were transformed into Top10 E. coli cells (Thermo Fisher, Waltham, MA). Plasmid DNA was harvested and purified using the EndoFree Plasmid Purification Kit (Qiagen, Germany) following the manufacturer's instructions. Equal amounts of heavy chain and light chain DNA were transfected into HEK293 cells, followed by antibody purification using Protein G (GE healthcare). Protein purity was confirmed by SDS-PAGE under reduced and non-reduced conditions (Fig. S1).
Western blot. Plasmodium falciparum sporozoites (NF54 strain ~ 20,000 per well) were separated using SDS-PAGE gel and transferred onto nitrocellulose membranes using the IBlot system (Thermo Fisher, Waltham, MA). Membranes were blocked with casein overnight at 4 °C and incubated with diluted mAb (1 µg/mL) for 1 h. Blots were washed three times with PBS-Tween solution (1X PBS + 0.05% Tween) and Goat anti-human antibody conjugated to Alkaline phosphatase (1:5000) was added and incubated for 1 h. Following the final wash NBT/BCIP substrate was added to visualize the bands.
Immunofluorescence assay. Slides were prepared by adding 10,000 P. falciparum sporozoites diluted in 1% BSA/RPMI per well and fixed with methanol. After blocking with 1% goat serum for 30 min, mAbs were Whole sporozoite ELISA. The assay was essentially performed as described previously 68 . Transgenic P.
berghei sporozoites expressing P. falciparum CSP were added to Immulon 2HB 96-well plates (Thermo Fisher, Waltham, MA) at a concentration of 25,000/well and incubated at room temperature for 2 h. Plates were fixed with methanol and blocked with 1% BSA/PBS for 1 h. MAbs were serially diluted fourfold down plates starting at 2 µg/mL and incubated at room temperature for 2 h. After washing thrice with PBS, the assay proceeded similarly as described above. After washing with PBS-Tween, the assay proceeded similarly as described above. Linear regression was used to determine the concentration of NaSCN that reduced OD 415nm to half maximal of the OD in the 0 M NaSCN wells.

Avidity index ELISA.
Fab binding kinetics. Fab fragments were produced by incubating human mAbs (1 mg/mL) with Endoproteinase Lys-C at a ratio of 1:2500 antibody:enzyme at 37 °C for 2-3 h. Enzyme digestion was assessed by SDS-PAGE and upon completion, the reaction mixture was passed through Protein-A beads (0.5-1 mL beads) three times. Inhibition of liver stage development assay (ILSDA). The ILSDA was performed as previously described 30 . Briefly, P. falciparum (NF54 strain) sporozoites were incubated with NFS1 (positive control mAb) or human mAb at room temperature for 20 min. A primary human hepatocyte culture (BioReclamation IVT, Baltimore, MD) was inoculated with the sporozoite-antibody mixture and incubated at 37 °C for 3 h to allow sporozoites to infect hepatocytes. The culture was washed following the first incubation period to remove noninvaded sporozoites, and washed again at 24 h post-inoculation. Total RNA was isolated from hepatocytes 72 h after the second wash and purified for downstream quantitative real-time PCR (qRT-PCR) of Pf 18S rRNA. RNA levels were quantified in quadruplicate to determine the percent inhibition of liver stage development by each antibody compared to negative control (no antibody). To generate the standard curve, fixed number of Pf sporozoites (20. 60, 180, 540, 1620, 4860 were added to the wells seeded with cryopreserved primary human hepatocytes, the mixture was harvested and RNA are extracted for downstream RT-PCR analysis. Unknown sample parasite burden was determined using a standard curve between CT values and sporozoite. A representative standard curve used for a mAb ILSDA experiment is shown in Fig. S6. ILSDA experiments used to compare mAbs (0.5, 0.25 and 0.05 µg/ml) were repeated at least 3 times.
In vivo protection. All procedures were reviewed and approved by the Walter Reed Army Institute of Research's Animal Care and Use Committee and performed in accordance with the Animal Welfare Act and other federal statutes relating to animals and experiments. The study was carried out in compliance with the ARRIVE guidelines (http:// www. nc3rs. org. uk/ page. asp? id= 1357). Female C57BL/6 J mice (The Jackson Laboratory, Bar Harbor, ME) were intravenously injected with 100 µL of each mAb in sterile PBS. Sterile protection of animals by mAb was assessed using transgenic P. berghei sporozoites expressing functional, full-length P. falciparum CSP 40,66 . Animals were challenged 24 or 48 h post mAb transfer with 1000 sporozoites intravenously. Blood-stage parasitaemia was detected by microscopy of giemsa-stained thin blood smears. Percent survival was determined using the absence parasitaemia during the two-week observation period immediately following challenge. Fisher's exact and Log-rank tests were performed to compare survival outcomes and curves for each experiment, respectively.
CSP reaction. Biotinylated mAbs at 1000, 100 and 10 ng/mL were mixed with freshly dissected transgenic P. berghei sporozoites for 10 min at room temperature. The antibody-sporozoite mixtures were placed drop-wise onto microscope slides, fixed with methanol, and allowed to air dry. Slides were blocked with casein and incubated with an Avidin-FITC secondary antibody (Thermo Fisher, Waltham, MA) before being visualized at 40X under a Fluorescence microscope (Olympus). A positive sporozoite reaction was recorded if tail-like projections were seen on mAb incubated sporozoites. A negative control mAb 3G1 was used to observe for background reaction.
Statistical methods. All the raw data used in statistical analyses is shown in Table S9. Repeat measurements on sporozoite invasion inhibition by ILSDA were modeled using a Generalized Linear Model (GLM). Generalized Estimating Equation (GEE) was used to determine the effect of dose and mAb type on inhibition of sporozoite invasion. Least Square (LS) means (Table S4); pairwise comparison of LS means against negative control mAb 3G1 (Table S5), and paired comparison between the CSP mAbs (Table S6) were used to discern statistically significant effects. The ILSDA inhibition data were fitted to a three-parameter Michaelis-Menten model f(x) = c + d−c 1+ e x + ε 69 and F test was used to determine if regression parameters were equal and whether fitted curves coincide. Fitted curves from the equation were shown in Fig. 6E. The estimated inhibitory mAb dose needed for 80% inhibition from the fitted model were listed in Table S7. In vivo protection data on day 14 were analyzed by contingency table analyses. Fisher's Exact test (Table S8) was used to examine the difference in infection rates. Kaplan Meir survival curves were analyzed using both the log-rank and Wilcoxon tests and showed similar p values. Paired association analyses between in vitro assays and in vivo protection was performed using raw data on 18 variables. Linear regression or Generalized Linear model, was applied with and without repeated observations. The standardized regression coefficient was equivalent to the Pearson's Correlation coefficient for each pair-wise comparison and p values (Table S3). For all other comparisons, one-way ANOVA was used and p values corrected for multiple comparisons using the Tukey's method. P-values less than 0.05 were considered to be significant. We analyzed the data using SAS/STAT Software, Version 9.4 of the SAS System TS1M5 (Cary, NC). Graph Pad Prism 8 (GraphPad Prism software, La Jolla, CA) and Microsoft Excel were used for plotting.
Ethics approval. Animal procedures were conducted in compliance with the Animal Welfare Act and other federal statutes and regulations relating to animals and experiments involving animals and adhere to principles stated in the Guide for the Care and Use of Laboratory Animals, NRC Publication, 2011 edition.

Data availability
The datasets and reagents used during the current study are available from the corresponding author on reasonable request.