Characterization of a novel inhibitory human monoclonal antibody directed against Plasmodium falciparum Apical Membrane Antigen 1

Malaria remains a major challenge to global health causing extensive morbidity and mortality. Yet, there is no efficient vaccine and the immune response remains incompletely understood. Apical Membrane Antigen 1 (AMA1), a leading vaccine candidate, plays a key role during merozoite invasion into erythrocytes by interacting with Rhoptry Neck Protein 2 (RON2). We generated a human anti-AMA1-antibody (humAbAMA1) by EBV-transformation of sorted B-lymphocytes from a Ghanaian donor and subsequent rescue of antibody variable regions. The antibody was expressed in Nicotiana benthamiana and in HEK239-6E, characterized for binding specificity and epitope, and analyzed for its inhibitory effect on Plasmodium falciparum. The generated humAbAMA1 shows an affinity of 106–135 pM. It inhibits the parasite strain 3D7A growth in vitro with an expression system-independent IC50-value of 35 μg/ml (95% confidence interval: 33 μg/ml–37 μg/ml), which is three to eight times lower than the IC50-values of inhibitory antibodies 4G2 and 1F9. The epitope was mapped to the close proximity of the RON2-peptide binding groove. Competition for binding between the RON2-peptide and humAbAMA1 was confirmed by surface plasmon resonance spectroscopy measurements. The particularly advantageous inhibitory activity of this fully human antibody might provide a basis for future therapeutic applications.


Selection of PBMC donor and EBV-transformation and screening.
In order to choose a promising candidate for the generation of an AMA1-specific human monoclonal antibody, plasma of 31 adult Ghanaian blood donors were screened by indirect ELISA for IgG reactivity against different recombinant variants of AMA1, including the allelic variant of Plasmodium falciparum strain 3D7, as well as a mixture of three artificial diversity covering variants of AMA1 (described by Remarque et al. 53 ). Against the single allele recombinant antigen AMA1 (3D7), 23 out of 31 (74%) of the Ghanaian plasma samples showed a positive immune response (Fig. 1A). The reactivity of the plasma against AMA1 (DiCo1-3) has been shown before 54 . Here, 30 out of 31 samples showed a positive reaction (reactivity superior of the negative control plus 2 standard deviations), corresponding to 97%.
The reactivities of the Ghanaian plasma samples against the variants AMA1 (3D7) and AMA1 (DiCo1-3) showed a strong positive correlation (p < 0.0001, Fig. 1B). Next, these plasma IgG were tested for recognition of conformational epitopes of AMA1. Therefore, ELISA using the same plasma samples were carried out similarly, but additionally using reduced/alkylated version of AMA1 (3D7). The results indicate that the majority of the antibodies are directed against conformational epitopes. The median reactivity of the plasma samples to reduced vs. the reactivity to the native antigen decreased by a factor of 17 (p < 0.0001, Wilcoxon matched pairs test, Fig. 1C). as well as a European control pool of two donors which had never been exposed to Malaria (NIP), ELISA was performed to estimate the reactivity against 3D7 AMA1. As positive control, a pool of four highly reactive Ghanaian donors was used. Values reflect relative reactivity in comparison to this positive control's reactivity, which was set to 1 (A). The reactivity of the samples to AMA1 (3D7) and AMA1 (DiCo 1-3) (previously published in Feller et al.) 54 were compared and the degree of correlation was estimated using the Spearman rank sum test (B). The reactivity of the plasma samples against both the native antigen and the reduced/alkylated antigen was estimated by ELISA, and shown as a box plot (C). The p-values were estimated using the Wilcoxon matched pairs test.
Scientific RepoRts | 6:39462 | DOI: 10.1038/srep39462 Only 2 of the 31 samples showed a reactivity against the reduced antigen AMA1 (3D7) and reactivities of these two were only slightly above the threshold of two standard variations above the negative control reactivity.
The sample of donor G10 had the highest reactivity against AMA1 (DiCo1-3) and is also among the samples with highest reactivity against AMA1 (3D7). This is why PBMCs of donor G10 were selected for the following steps of antibody generation. The PBMCs were thawed, antigen-specific B-cells were stained (IgG + /CD22 + / anti-AMA1 (DiCo1-3) + ) and sorted by flow cytometry. The supernatants of the cultured lymphoblastoids were tested for specific antibody production after two weeks of culture, the supernatant of well 1E4 showed high reactivity against AMA1 (DiCo1-3). Subsequently, the sequences coding for variable domains of the antibody light and heavy chain were amplified from cDNA from these cells, sequenced and cloned into expression vectors for recombinant production.
Recombinant expression of humAbAMA1. The expression cassettes for the antibody production in plants and mammalian cell culture are shown in Fig. 2A. HumAbAMA1, the recombinant antibody of EBV-clone 1E4, was produced as a full-size human IgG1:κ by Agrobacterium-mediated transient protein expression in N. benthamiana plants. After harvest and homogenization of the plant material, the antibodies were purified by Protein A chromatography. The final yield of the pure and intact plant produced antibody was 70 mg/kg fresh leaf material. Similarly, secretory production of the antibody was performed by transient transfection of HEK293-6E cells. The cell culture supernatants were purified by Protein A chromatography. Here, the productivity of the antibody after 6 days of antibody production was approx. 10 mg/l cell culture supernatant. Integrity and purity of both antibodies was assessed by SDS-PAGE (Fig. 2B).
HumAbAMA1 recognizes a conformational epitope and binds to it with high affinity. The binding specificity of the antibody humAbAMA1 was first tested in indirect ELISA. The reactivity of the antibody was concentration dependent, with a detection limit of approx. 6 ng/well (Fig. 2C). AMA1 possesses eight disulfide bridges, which strongly stabilize its tertiary structure 55 . Rodent-derived anti-AMA1 mAbs, such as the murine mAb 1F9 and the rat mAb 4G2 recognize conformational, reduction-sensitive epitopes 43,56 . This is also the case for humAbAMA1, which does not react with reduced/alkylated AMA1 in indirect ELISA (data not shown). Surface plasmon resonance (SPR) spectroscopy measurements were used to compare the affinity of hum-AbAMA1 produced in plants and in HEK293-6E cells towards the recombinant antigen AMA1 (3D7), revealing very high affinities (K D ) of 135 ± 8 pM and 106 ± 5 pM for the plant and HEK-derived antibodies, respectively . The binding of humAbAMA1to AMA1 (3D7) was tested by indirect ELISA. Dilutions of the antigen (AMA1 (3D7)) were coated onto ELISA plates, then human antibodies were applied followed by goat anti-human antibody coupled to alkaline phosphatase (Dianova, Hamburg, Germany). Black diamonds: Semi-immune (premune) plasma pool; green circles: humAbAMA1 (N. benthamiana); Red open squares: humAbAMA1 (HEK293); orange inverted triangles: mock control antibody (2G12).
Scientific RepoRts | 6:39462 | DOI: 10.1038/srep39462 (Fig. 3). Additionally, the affinity of the antibody humAbAMA1 produced in N. benthamiana against two further AMA1 antigen variants representing the strains HB3 and FCR3 have been measured. The binding towards AMA1 (HB3) is in the similar range (667 ± 7 pM), whereas the binding to AMA1 (FCR3) is two orders of magnitude lower (73 ± 0.2 μ M). The kinetics of the antibodies are more precisely summarized in Table 1.

HumAbAMA1 recognizes AMA1 in its native context on the surface of merozoites and schizonts.
Next, humAbAMA1 was tested for binding native AMA1 located on merozoites and schizonts. To this end, immunofluorescence assays (IFAs) were carried out using double staining with the humAbAMA1, as well as the MSP4-specific antibody 2.44 (Fig. 4, upper panel). We observed a characteristic punctiform staining at one end of ripening and free merozoites, which is in line with the apical localization of AMA1. In some cases we also saw a staining of the outer membrane of free merozoites, indicating colocalization with antibody 2.44 (Fig. 4, lower panel). This is consistent with the partial redistribution of AMA1 once the merozoites have been released upon schizont rupture 57 . The data obtained strongly resemble immunofluorescence images using AMA1-specific antibodies published earlier 38,57 . Plasmodium falciparum strains 3D7A, HB3 and FCR3 are strongly inhibited by humAbAMA1.
The inhibitory potential of humAbAMA1 was tested in in vitro growth inhibition assay (GIA) with P. falciparum 3D7A, HB3 and FCR3. We found that P. falciparum strain 3D7A was reproducibly inhibited by plant-produced humAbAMA1 with an IC 50 value of 35 μ g/ml (95% confidence interval (CI): 33-37 μ g/ml). Inhibition assays using the antibodies produced in mammalian cell culture showed the same IC 50 value of 35 μ g/ml (95% CI: 34-37 μ g/ ml) (Fig. 5A). As controls, we expressed the rodent-derived antibodies described in the literature (4G2 and 1F9) as chimeric antibodies. The IC 50 values of these antibodies were 105 μ g/ml (95% CI: 99-111 μ g/ml) and 292 μ g/ml (95% CI: 250-342 μ g/ml) for the chimeric antibodies 4G2 and 1F9, respectively, thus being three to eight times higher than the IC 50 value of humAbAMA1. The IC 50 -value of the humAbAMA1 against the parasite strains HB3  Table 1. Surface plasmon resonance measurement results for antibody affinity against AMA1 variants of P. falciparum strains 3D7, HB3 and FCR3.
and FCR3 was also 3-fold and 4-fold lower than the chimeric antibody 4G2, respectively. The second chimeric control antibody inhibited the parasite strains HB3 and FCR3 at values above 1000 μ g/ml, thus the estimated values were 7 to > 13-fold higher than the antibody humAbAMA1. The assays are represented in Fig. 5 and the estimated IC 50 -values summarized in Table 2.
Next, combinations of the human monoclonal antibodies humAbAMA1 and humAb10.1 (anti-MSP10, previously isolated in our laboratory) 52 were tested for synergistic effects. The antibodies were mixed in five different ratios, added in serial dilutions to the parasite culture (P. falciparum strain 3D7A) and the IC 50 of each combination was determined. Generally, two calculations are performed to evaluate the degree of synergy, the sum of FIC and HEWLETT's S. Values corresponding to sum of FIC below 1 and HEWLETT's S above 1 are regarded as synergistic, FIC below 0.5 and HEWLETT's S above 2 reflect strongly synergistic effects. As shown in the isobologram, the combination of the AMA1-specific antibody and the MSP10 antibody humAb10.1 shows synergistic inhibition of the parasite strain 3D7A (Fig. 6). The calculated FIC and HEWLETT's S are 0.68 and 1.64, respectively. In comparison, two AMA1-specific antibodies (humAbAMA1 and 4G2) rather act in an additive way, as shown in Fig. 6 (sum of FIC = 0.92 and HEWLETT's S = 1.11).
HumAbAMA1 partially shares an epitope with mAb 1F9. After the confirmation of the strong inhibitory response in GIAs, humAbAMA1 was tested for potential epitope overlapping with the well-characterized inhibitory antibodies 1F9 or 4G2. To this end, a "pairwise epitope mapping" was carried out. In a multiple capture setup AMA1 (3D7) was first caught on a mAb1D7 (AMA1 domain III-specific antibody) surface. Then, the antibodies to be tested were injected singly or consecutively. It was found that humAbAMA1 and 1F9 partially share an epitope, i.e. 1F9 could not bind efficiently anymore if humAbAMA1 had been injected first. However, their epitopes appeared not to be completely identical since a small proportion of 1F9 could still recognize AMA1 (3D7) (Fig. 7, cycle 6). In cycle 3 (Fig. 7), it became clear that humAbAMA1 and 4G2 could bind simultaneously. Thus, their corresponding epitopes neither overlap nor show signs of allosteric competition. Subsequently, the epitope was more specifically determined using the CLIPS (Chemical Linkage of Peptides onto Scaffolds) technology (Pepscan Presto BV, Lelystad, The Netherlands), in which peptides covering the complete sequence of AMA1 (3D7) were screened. The advantage of using this method in comparison to standard peptide-screening methodology is that the three-dimensional structures of the original protein are simulated by introducing various peptide loop patterns 58 . The results suggest that the antibody specifically binds to the loop of the peptide 194-TLDEMRHFYKDNK-206 within the domain I of AMA1, which is in close proximity to the hydrophobic RON2 binding groove (Fig. 8). The binding site of humAbAMA1 overlaps largely with the epitope of 1F9 56 . RON2sp1 and humAbAMA1 compete for binding to AMA1 (3D7). HumAbAMA1 was capable of inhibiting P. falciparum 3D7A with an IC 50 value of 35 μ g/mL (Fig. 5) and competed with 1F9 for binding to AMA1 on the same epitope. We were thus interested if humAbAMA1 might as well interfere with the binding of RON2, the natural ligand of AMA1, which had been previously described for both 1F9 and 4G2 42,43,56,59 . AMA1 (3D7) and RON2sp1 peptide, the portion of RON2, which binds to the hydrophobic trough of AMA1 60 , were pre-incubated in different molar ratios. In SPR measurements it was then analyzed to which extent humAbAMA1 could still bind AMA1. As can be seen in Fig. 9 the more RON2sp1 had been binding to AMA1 (3D7), the lower was the proportion of humAbAMA1 recognizing AMA1 (3D7). This indicates that humAbAMA1 binds to an epitope on AMA1 which is at least partially obscured once RON2sp1 has bound to AMA1.

Discussion
The presented work describes the first human antibody recognizing the malaria vaccine candidate AMA1. HumAbAMA1 reflects a significant degree of affinity maturation in vivo (88-96% germ line homology), recognizes a conformational variable epitope lying in very close proximity to the RON2 binding site, and inhibits   various P. falciparum strains with IC 50 values ranging from of 35 μ g/ml for the strain 3D7A to 195 μ g/ml for the strain FCR3. Most individuals who are premune to malaria have developed antibodies directed at AMA1 10,54,61,62 and the recognition of conformational determinants of AMA1 is believed to be crucial for a protective AMA1-specific response [63][64][65] . Rodent antibodies such as 1F9 and 4G2 with inhibitory potential had been reported about fifteen to twenty years ago 66,67 . Here, for the first time, a similar human monoclonal antibody is described. The natural affinity maturation of humAbAMA1 resulted in a binding strength of < 150 pM, which is close to the range of 10 −11 M that is considered the lower limit to be achievable in vivo [68][69][70] . Only few human antibodies have been disclosed which possess K D values of as low as ≤ 10 −10 M 71,72 . Moreover, as compared to 1F9 and 4G2 exhibiting K D values of 1.3 × 10 −9 M and 3.3 × 10 −9 M, respectively 73 , the affinity of humAbAMA1 to AMA1 (3D7) is 7-to 18-times stronger.
In analogy to the higher affinity, the inhibitory potential of humAbAMA1 described by its IC 50 value of 35 μ g/ml is well below the IC 50 values reported for 1F9 (ca. 850 μ g/ml) 42 and AMA1-specific polyclonal human IgG (70-240 μ g/ml) 38,74 and to our knowledge, it is the lowest IC 50 -value for a blood-stage antibody described so far. Differences of the values we obtained here and the inhibitory concentrations described in the literature might be  Using SPR spectroscopy, pairwise epitope mapping was performed. The chip was functionalized with antibody mAb1D7, which binds to domain III of AMA1. This chip was regenerable. Subsequently, in each cycle AMA1 (3D7) (1750 RU) was caught. Then, humAbAMA1, 1F9, or 4G2 were injected singly (cycles 1, 2, 4 and 5) or consecutively (cycles 3 and 6). In cycles 1-3, 1 μ g/ml of each antibody was applied, whereas in cycles 4-6, the concentration of each antibody was 10 μ g/ml. In case spectral resonance is increasing with the binding of the secondary antibody, no competition is detectable, in case the signal is not increased after binding of the secondary antibody, the epitopes either overlap or binding is compromised by allosteric competition. based on differences in the assays (parasitemia, hematocrit, readout) as well as the fact that we used recombinant chimeric antibodies, whereas in the publications stated above, the original rodent antibodies have been used.
Very efficient murine mAbs directed at other plasmodial targets such as reticulocyte-binding protein homologue 5 (Rh5) show IC 50 values of 50-100 μ g/ml 44 , and AMA1-specific polyclonal rabbit IgG generated by hyperimmunization protocols could achieve IC 50 values in a similar range of 34-180 μ g/ml 73,75 . A direct comparison between polyclonal and monoclonal antibodies is difficult as in vivo, an entire spectrum of antibodies contributes to the defense against plasmodial infections and a successful control is likely due to synergistic effects of these antibodies. Therefore, we used humAbAMA1 in combination with humAb10.1, which showed the best IC 50 value of all MSP10-specific antibodies we described 52 . We found a weak synergistic effect, characterized by a HEWLETT's S of 1.6. Nonetheless, the effect is inferior to the synergistic effect that has been described for other polyclonal antibody combinations, e.g. murine Rh4-and Rh5-specific mAbs 76 , or combinations of anti-plasmodial chemotherapeutics 77,78 .
Although humAbAMA1 inhibits the parasite strains HB3 and FCR3 with IC 50 -values of 126 and 195 μ g/ml, which is 3.6 to 5.6-fold higher than the respective IC 50 -value measured in the strain 3D7A, it is still superior to Its localization pertains to domain I of AMA1 (yellow) and is highlighted in red. The published epitope of 1F9 within domain I is highlighted in cyan, the region in which the epitopes of humAbAMA1 and 1F9 overlap is highlighted in magenta 56 . The domain II of AMAI is shown in blue. The RON2 peptide (green) is bound to its specific hydrophobic binding groove in domain I. The structural data was obtained from the protein data base PDB, accession number 3ZWZ 102 . The structure in the second image is a 90° rotation of the first image. Figure 9. Competition of humAbAMA1 and RON2sp1 peptide. The determination of the epitope region recognized by humAbAMA1 using the peptide array, as well as the binning experiments suggests a possible steric interference between humAbAMA1 and RON2sp1. To investigate the potential competition between RON2sp1 and humAbAMA1 for AMA1-binding AMA1 (3D7) was preincubated with various concentrations of the peptide RON2sp1 for 1 hour. Subsequently, the AMA1 (3D7):RON2sp1 solutions were injected over a surface featuring a defined amount (400 RU) of humAbAMA1 caught on a Protein A-functionalized chip using a BIAcore T200 SPR instrument. For visualization of the results the measured response units were plotted against the concentration of RON2sp1 (μ M) used as competitor.
Scientific RepoRts | 6:39462 | DOI: 10.1038/srep39462 the inhibitory capacity of the well-described antibodies 4G2 and 1F9. It is a very interesting observation that the antibody showed a direct relation between the affinity to the antigen variant and the inhibitory capacity to the respective parasite strain. This needs to be reconfirmed with an increased number of parasite strains and antigen variants.
AMA1-specific antibodies may inhibit plasmodia by different means, i.e. (1) the blocking of epitopes that are important for the interaction with other proteins such as RON2, (2) the prevention of structural rearrangements within the AMA1 molecule required for efficient RON2-binding, (3) the impairment of the proteolytic processing of AMA1, and (4) the interference of the redistribution of the cleavage products 79 . Moreover, a fourth mode of action could be the premature induction of the phosphorylation of serine 610 in the cytoplasmic tail of AMA1 and the untimely initiation of downstream signalling events, e.g. by imitating the binding of its natural ligand. The phosphorylation of S610 was shown to be critical for invasion 37,80,81 . Based on the current data at hand we cannot conclusively say how humAbAMA1 exerts its inhibition of P. falciparum 3D7A. However, there are several clues. The first evidence is that 1F9 and humAbAMA1 competed for binding to AMA1 (3D7), whereas 4G2 and humAbAMA1 does not (Fig. 7). 4G2 binds to a region of domain II which is distant from the hydrophobic trough of AMA1 34,43 , but until now, the crystal structure of 4G2 binding to AMA1 could not be resolved. 1F9 on the other hand binds close to the hydrophobic cleft of AMA1 and to adjacent regions in domain I, including residues P188, M190, P192, E197, H200, F201, Y202, D204, M224, I225, N228 in AMA1 (3D7) 56 . This region shares a significant overlap with the epitope region of humAbAMA1 identified by peptide array analysis . Additionally, we showed either a clear intereference with the binding site of RON2, since humAbAMA1 could not bind efficiently to AMA1:RON2sp1 complexes in an SPR setup (Fig. 9). This implies that by the binding of RON2sp1 the epitope of humAbAMA1 was either masked or structurally rearranged by RON2sp1. The evidence given strongly argues for an interference of humAbAMA1 with the binding of RON2 to AMA1 on the parasite, thus preventing the formation of the AMA1:RON2 complex, leading to the inhibition we observed in GIAs. However, this mode of inhibition remains to be proven.
In this study we took a closer look at P. falciparum strains 3D7A, HB3 and FCR3 as well as their respective AMA1 alleles. It is well known that AMA1 is subject to diversifying and balancing selection in order to evade the host's immune system as well as to maintain its structure and function [82][83][84] . Therefore, future experiments will have to address if humAbAMA1 is capable of inhibiting even more P. falciparum strains than the presented ones. Nevertheless, the data presented here already shows that the antibody has some cross-strain inhibitory activity, even though it binds to a variable region of the parasite. The determination of the structures of a Fab fragment of humAbAMA1 and of the Fab:AMA1 complex will likely help to shed more light on the nature of the epitope and the mode of action of the new antibody we presented.
Since transient expression in N. benthamiana is a well-established antibody 52 and vaccine candidate 85 expression platform in our lab, we used it for the production of humAbAMA1 and compared expression levels and antigen-binding properties with humAbAMA1 produced by transient transfection of HEK293-6E cells. As shown by SPR-analysis, the plant and HEK produced molecules show no significant differences in AMA1 binding.
Besides elucidating the epitope targets and parasite growth inhibitory mechanisms of naturally occurring human Pf-specific mAbs, such molecules are also of interest for therapeutic use. Of crucial importance for a therapeutic antibody is its ability to inhibit a brought set of parasites occurring naturally. The set of parasites chosen here represent a number of variants of the estimated epitope. All strains could be inhibited by humAbAMA1, albeit at varying concentrations. Nevertheless, further investigations are necessary in order to estimate the full therapeutic potential of humAbAMA1. Even though the majority of monoclonal antibodies used in therapy is targeting tumors or chronic inflammatory diseases, there are certain applications for antibodies and antibody cocktails also in other fields like the treatment of toxin mediated 86 and infectious diseases. Post-exposure treatment of rabies infections includes the injection of rabies virus specific antibodies isolated from horse serum 87 , but various groups are working on the development of recombinantly produced humanized or human monoclonal antibodies to replace such sera. Humanized monoclonal antibodies have been proposed as therapeutics for West Nile Virus infections 88 and a cocktail of three humanized antibodies 89 targeting the Ebola virus EBOV glycoprotein has received some attention in the context of the latest outbreak of EBOV in several West African countries. The antibody-based Ebola-emergency treatment ZMapp used in a small number of human patients has been produced in N. benthamiana plants by transient expression 90 . In the case of malaria antibodies could be used for short-term prevention and/or therapy in clinical cases with drug resistant strains 91 . Based on antibodies such as the one presented here in combination with affordable expression systems, the cure of malaria infections or the generation of passive vaccination strategies becomes feasible.

Materials and Methods
Ethics statement and blood donors. This study was conducted according to the principles expressed in the Declaration of Helsinki. Ethical approval was given by the Committee on Human Research Publication and Ethics (CHRPE) of the Kwame Nkrumah University of Science and Technology in Ghana. After the goals of the study had been carefully explained, written informed consent was obtained from all participants. By the time of blood donation, the 31 volunteers were between 20 and 45 years old and had not undergone any clinical phase of malaria for at least two years. The Ghanaian study, as well as the inclusion and exclusion criteria had been described previously 92 . Each volunteer donated 50 ml of heparinized blood which was then used for the preparation of plasma and PBMCs as described below.

Fluorescent labeling of recombinant AMA1 (DiCo 1-3). AMA1 (DiCo 1-3) was kindly provided
by Jürgen Drossard (Fraunhofer IME, Aachen, Germany) with the kind permission of Ed Remarque (BPRC Rijswijk, The Netherlands). The proteins were chemically labeled with "AlexaFluor488-TFP" (Life Technologies, Darmstadt, Germany) as recommended by the manual's instructions. On average, the proteins were found to be labeled with 5.9 AlexaFluor 488 moieties (mol/mol) as determined by a spectrophotometric measurement.
Reduction and alkylation of AMA1 (3D7). For testing of conformational dependency of epitope recognition by polyclonal AMA1-specific antibodies and by humAbAMA1, AMA1 (3D7) was irreversibly reduced. To this end, it was treated with 5 mM DTT at 56 °C for 45 min while shaking. The reaction was stopped by adding 15 mM iodoacetamide. After incubation for 30 min in the dark 5 mM DTT were added to quench excessive iodoacetamide.

Epstein-Barr virus transformation.
After a viability staining with Trypan blue specific IgG + /CD22 + B lymphocytes were infected with Epstein-Barr virus (B95-8) and subjected to virocrine transformation as described 52,93 . Enzyme-linked immunosorbent assays. ELISA procedures for the various parts of the study were carried out as described before to: (A) identify promising B cell cultures 52 , (B) assess the specificity of recombinant human antibodies 52 , and (C) screen the Ghanaian plasma samples for reactivity against AMA1 (3D7) 54 .
Rescue and cloning of immunoglobulin variable regions. cDNA of promising cultures was generated with the "SuperScript III CellsDirect" kit (Life Technologies, Darmstadt, Germany) according to the manufacturer's instructions using oligonucleotides for the rescue and cloning of human immunoglobulin variable regions (Vh and Vl) as previously described 94 . Primers for the cloning into pTRAkt vectors were designed to correct any mutations introduced by the degenerate primer set used for nested PCRs -thereby reconstituting the authentic sequences of 1E4 Vh and 1E4 Vl. Vh and Vl were cloned individually into the respective pTRAkt vectors for apoplastic expression of human full-size IgG1:κ antibodies in plants 95 .

Expression of full-size humAbAMA1 IgG in Nicotiana benthamiana. Expression and purification
of full-size IgG by Agrobacterium-mediated transient transfection of N. benthamiana plants was performed using agrobacteria transformed with the vectors pTRAkt_IgG1_H_1E4 (humAbAMA1 IgG heavy chain) and pTRAkt_ Igκ _1E4 (humAbAMA1 Igκ light chain), respectively, as described elsewhere 95 . Expression of full-size humAbAMA1 IgG in HEK293-6E. Expression of humAbAMA1 in HEK293-6E cells was performed with the help of vector pTT5, into which the expression cassette used in pTRAkt vectors described above was inserted. Transfection, cultivation of the cells as well as ProteinA-based purification was performed as described before 96 . Plasmodium falciparum culture and growth inhibition assays. P. falciparum strains 3D7A, HB3 and FCR3 were cultured as described 97 . Parasites were maintained at a hematocrit of 5% in RPMI 1640, 25 mM HEPES, 2 mM L-glutamine, 50 μ g/ml gentamycin, 10% Albumax II.
In vitro growth inhibition of the various antibodies was assessed with P. falciparum 3D7A, HB3 and FCR3 as described before 53 . HumAbAMA1 was applied in concentrations ranging from 4.35 mg/ml to 4.25 μ g/ml. Polyclonal rabbit-anti-AMA1 IgG (BG98) in a concentration of 6 mg/ml was taken as a positive control 98 . Serum IgG of a naïve rabbit was taken as a corresponding negative control (6 mg/ml). HumAb 2G12, an HIV-1-specific IgG, was taken as a plant-expressed non-specific control. 2G12 was kindly provided by Jürgen Drossard. Prior to the setup of the assay P. falciparum strains had been synchronized three times by sorbitol treatments 99 . Parasites were seeded at schizont stage at a parasitemia of 0.4% in 2% hematocrit. After an incubation period of 40-46 h, the GIA was harvested and frozen at − 80 °C for at least 16 h. Parasitaemia was read out by assessing the pLDH activity 100 . Relative growth inhibition was calculated as follows: % inhibition = 100 − 100 × [A 655nm (sample) − A 655nm (erythrocyte control)]/[A 655nm (schizont control) − A 655nm (erythrocyte control)]. In order to test for a synergy of invasion inhibition both humAbAMA1 and a MSP10-specific antibody, humAb10.1 (described in ref. 52), were employed. To this end, the antibodies were used singly and in ratios proportional to their respective IC 50 values. The ratios were 1:1, 1.5:1, 4:1, 1:2, and 1:4. Data were analyzed by calculating HEWLETT's S and the sum of FIC, a quantitative index of the extent of synergy 76 Immunofluorescence assay. For immunofluorescence assays P. falciparum 3D7A was synchronized by sorbitol treatments. Parasitized erythrocytes were washed once with RPMI 1640 before being taken up in FCS (low bovine IgG) and smeared onto object slides. Parasites were air-dried and then fixed in methanol at − 20 °C for 10 min. Slides were incubated with 5 μ g/ml humAbAMA1 and 20 μ g/ml mouse-anti-MSP4(EGF) (mAb 2.44, kindly provided by Alexander Boes, Fraunhofer IME, Germany) 95 in 1% FCS (low bovine IgG) in PBS in a moist box at 25 °C for 1 h. After five washing steps in PBS, the slides were treated with 1:100 goat-anti-mouse IgG(H + L)-AlexaFluor488 (Life Technologies, Darmstadt, Germany), 1:1000 goat-anti-human IgG(H + L)-Cy3 (Jackson ImmunoResearch/Dianova, Hamburg, Germany), and 10 μ g/ml Hoechst 33342 in a moist box at 25 °C for 1 h. Five washing steps were carried out. Mounting medium ("ProLong Gold Antifade Reagent", Life Technologies, Darmstadt, Germany) was used to conserve the staining. Images were taken with a Leica TCS SP8 microscope at a 630-fold magnification.
Surface plasmon resonance spectroscopy. In our study SPR measurements fulfilled three purposes. (A) Determination of the affinity of humAbAMA1 to AMA1 (3D7), (B) pairwise epitope mappings with monoclonal rodent antibodies 1F9 and 4G2, and (C) RON2-competition measurements. All measurements were performed by using a Biacore T200 instrument (Biacore, GE Healthcare) and CM5-S-Series sensor chips at 25 °C with HBS-EP (10 mM 4-(2-hydroxyethyl)-1piperazineethanesulfonic acid, 150 mM NaCl, 3 mM EDTA, 0.005% (w/v) polysorbat-20) as running buffer. (A) The kinetic analysis of the humAbAMA1 interaction was performed as previously described 73 . (B) To roughly map the epitope of humAbAMA1 in relation to the binding sites of previously described AMA1 specific monoclonal antibodies 1D7, 1F9 and 4G2, AMA1 (3D7) was captured (1600 RU) using a CM5 surface functionalized by EDC-NHS chemistry with 7000 RU of the rat monoclonal 1D7. Mab1D7 is specific for the ectodomain III of AMA1, recognizes an epitope, which is distant from the hydrophobic cleft, and thus does not interfere with 1F9 or 4G2 binding. The antibodies 1F9, 4G2 and humAbAMA1 (at concentrations of 1 μ g/ml and/or 10 μ g/ml) were separately and consecutively injected for 180 s and binding levels were compared. (C) In order to rule out if humAbAMA1 and RON2 compete for binding to AMA1, AMA1 (3D7) and RON2sp1 were mixed in different molar ratios and incubated for 1 h at 25 °C. (RON2sp1 corresponds to the portion of RON2 which binds to the hydrophobic through of AMA1 102 ). The extent to which humAbAMA1 could bind to AMA1:RON2sp1 complexes was then determined. To this end, first, humAbAMA1 was caught on a Protein A-functionalized chip and then the different pre-incubated AMA1:RON2sp1 solutions were injected.
Epitope mapping using Pepscan screening. The epitope mapping of humAbAMA1 was carried out by Pepscan Presto BV (Lelystad, The Netherlands) as described before 58 . For the screening, four sets of peptides were used, all based on the AMA1 sequence of the parasite strain 3D7. Set 1 consisted of linear peptides of the length of fifteen amino acids. Set 2 consisted of constrained peptides of the length of 17 amino acids, with cysteine residues inserted at positions 1 and 17, and amino acids 2-16 representing AMA1 sequences. Cysteine residues on positions 1 and 17 were linked using the CLIPS technology as described 58 . Set 3 was designed to mimic α -helices of AMA1. These peptides consisted of 16 amino acids with cysteine residues inserted at positions 1 and 5. Set 4 constituted peptides with lengths of 22 amino acids and with cysteine placed on position 1 and 22. In this peptide set 4, amino acids on positions 11 and 12 had been replaced by proline and glycine, respectively. In all non-linear peptide sets described, cysteine residues occurring naturally the AMA1 sequence were chemically protected by acetamidomethyl in order to prevent unintentional reaction.