Structural basis of antigenic escape of a malaria
Sheetij Dutta*†, Seung Yeon Lee*, Adrian H. Batchelor‡, and David E. Lanar§
*Department of Epitope Mapping,§Division of Malaria Vaccine Development, Walter Reed Army Institute of Research, Silver Spring, MD 20910; and
‡University of Maryland School of Pharmacy, Baltimore, MD 21201
Edited by Louis H. Miller, National Institutes of Health, Rockville, MD, and approved June 8, 2007 (received for review February 16, 2007)
Antibodies against the malaria vaccine candidate apical membrane
antigen-1 (AMA-1) can inhibit invasion of merozoites into RBC, but
antigenic diversity can compromise vaccine efficacy. We hypoth-
esize that polymorphic sites located within inhibitory epitopes
function as antigenic escape residues (AER). By using an in vitro
model of antigenic escape, the inhibitory contribution of 24 poly-
morphic sites of the 3D7 AMA-1 vaccine was determined. An AER
cluster of 13 polymorphisms, located within domain 1, had the
highest inhibitory contribution. Within this AER cluster, antibodies
primarily targeted five polymorphic residues situated on an ?-
helical loop. A second important AER cluster was localized to
domain 2. Domain 3 polymorphisms enhanced the inhibitory con-
could be split, such that chimeras containing domain 1 of FVO and
domain 2 ? 3 of 3D7 generated antisera that showed similarly high
level inhibition of the two vaccine strains. Antibodies to this
chimeric protein also inhibited unrelated strains of the parasite.
Interstrain AER chimeras can be a way to incorporate inhibitory
epitopes of two AMA-1 strains into a single protein. The AER
clusters map in close proximity to conserved structural elements:
the hydrophobic trough and the C-terminal proteolytic processing
site. This finding led us to hypothesize that a conserved structural
basis of antigenic escape from anti-AMA-1 exists. Genotyping
high-impact AER may be useful for classifying AMA-1 strains into
inhibition groups and to detect allelic effects of an AMA-1 vaccine
in the field.
Plasmodium ? apical membrane antigen-1 ? invasion
malaria vaccine, containing merozoite surface protein-2 (msp-2)
of Plasmodium falciparum 3D7 strain as one of its components,
successfully reduced the prevalence of the 3D7 msp-2 genotype
but had no impact on the prevalence of parasites with the FC27
msp-2 genotype (2). Understanding the molecular basis of strain
specificity and the resulting antigenic escape is therefore impor-
tant for vaccine development.
Apical membrane antigen-1 (AMA-1) is one of the leading
malaria vaccine candidates. Immunization with AMA-1 induces
antibodies that inhibit invasion, conferring protection in animals
are currently in efficacy human trials (4, 5). Despite the strong
preclinical evidence favoring its vaccine candidacy, there are
?60 polymorphic sites on AMA-1 protein. Among the 50 Thai
isolates sequenced, there were 27 haplotypes. Similarly, of the 50
Nigerian sequences there were 45 haplotypes, and of the 68
The strain variability of AMA-1 is a cause of concern to
Strain-specific differences are reported among field antisera
and invasion inhibition (GIA) (11). Allelic replacement exper-
iments have directly implicated sequence polymorphism in an-
tigenic escape (12), and cross-strain GIAs suggest that the extent
of escape correlates sequence distance between the vaccine and
ntigenic diversity has been implicated in the failure of
several licensed and test vaccines (1). The ‘‘Combination B’’
the target strain (13). In the rodent malaria challenge model,
polymorphism of AMA-1 has been unequivocally linked to
vaccine failure (14). Human sera against the WRAIR 3D7
AMA-1 vaccine, which inhibits invasion of the homologous 3D7
strain, showed little or no inhibition of the heterologous FVO
strain (5). In an attempt to overcome the polymorphism prob-
lem, one group is following a coimmunization strategy, and
antibodies to a bi-allelic 3D7?FVO vaccine show high-level
inhibition of both the vaccine alleles (4, 13). However, the extent
of global haplotype diversity within AMA-1 has hindered the
rational selection of haplotypes for the multiallelic mixture
approach and is likely to complicate allelic shift analyses in the
upcoming efficacy trials unless the most important escape res-
idues are identified.
The nature and distribution of AMA-1 polymorphisms seems
to have strong structural basis. Only ?10% of AMA-1 residues
are polymorphic, and these polymorphisms are concentrated in
a relatively small hypervariable region on domain 1 (6, 7).
Distant polymorphisms cluster in three dimensional space and
are located on one side of the AMA-1 crystal structure: ‘‘the
polymorphic face’’ (15–17). Additionally, all of the polymorphic
sites do not have an equal contribution toward antigenic escape.
to inhibition by anti-3D7 AMA-1 antisera, despite the 9-aa
differences between 3D7 and D10 AMA-1 (11, 13). We hypoth-
esize that polymorphisms located within important inhibitory
epitopes confer most of the escape advantage to the parasite. We
term these critical polymorphic sites as ‘‘antigenic escape resi-
dues’’ (AER). The objective of this study is to determine the
relative inhibitory contribution of various polymorphic clusters
to map the structural location of AER of the 3D7 AMA-1
There are 24-aa differences between 3D7 and FVO strain
3. To determine the relative inhibitory contribution of these 24
polymorphic sites, we produced chimeric FVO AMA-1 proteins
displaying 3D7 specific polymorphic clusters. The chimeras were
used to selectively deplete 3D7-recognizing antibodies, in GIA
reversal experiments, and the resulting reversal of inhibition was
used as a readout to map the AER.
Chimeric AMA-1 Proteins Had Comparable Purity and Contained
Elements of Correct Structure. Domains 1, 2, 3, and 2 ? 3 were
selectively switched from FVO to 3D7 type in chimeric proteins
Author contributions: S.D. and A.H.B. designed research; S.D. and S.Y.L. performed re-
search; S.D. and A.H.B. analyzed data; and S.D., A.H.B., and D.E.L. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Abbreviations: AMA-1, apical membrane antigen-1; GIA, growth and invasion inhibition
assay; AER, antigenic escape residues; RE, relative escape.
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2007 by The National Academy of Sciences of the USA
July 24, 2007 ?
vol. 104 ?
D1, D2, D3, and D2?3, respectively. Proteins C1, C2, C3,
C1?C2, and C1-L were chimeras in which clusters of 3D7-
specific polymorphisms within domain 1 were substituted for the
FVO type. Polymorphisms included within a cluster were based
on the spatial proximity in three-dimensional space [Fig. 1 and
supporting information (SI) Table 1]. The recombinant chimeric
proteins were expressed in Escherichia coli, refolded, and puri-
fied to homogeneity. All nine chimeras, the 3D7 and the FVO
AMA-1 proteins, migrated as monomeric bands at ?54 kDa
(Fig. 2A). Proteins were ?95% pure by densitometry of Coo-
massie blue-stained bands. Positive reactivity with a P. falcipa-
rum AMA-1, conformation dependent, monoclonal antibody
4G2dc1, suggested correct folding of the chimeras (28) (Fig. 2B).
Furthermore, the presence of disulphide bonded tertiary struc-
tures were confirmed by mobility difference of the proteins
observed by SDS/PAGE, under reducing and nonreducing con-
ditions (SI Fig. 8).
Domain 1 and Domain 2 Are the Prime Targets of Strain-Specific
Antibodies. The chimeras were screened for gain in reactivity to
a 3D7 specific polyclonal antibody reagent by using Western blot
(Fig. 2C). A densitometric scan of the blot was also done (SI Fig.
9). The 3D7 specific antibodies showed minimal cross-reactivity
with the FVO AMA-1 scaffold. Among the domain chimeras D1
and D2 showed high reactivity as compared with D3. The D2?3
chimera showed higher reactivity than D2 or D3 (SI Fig. 9).
Among the cluster chimeras, C1?C2 and C1 showed high
reactivity as compared with C1-L, C2, and C3. The FVO
AMA-1-based chimeras were also tested for loss of reactivity to
a strain-specific anti-FVO reagent (Fig. 2D). Although cross-
reactivity of anti-FVO with the FVO AMA-1 scaffold (common
in all chimeras) could be a complicating factor, chimeric proteins
reacted similarly with anti-FVO, suggesting that strain-specific
anti-FVO are more cross-reactive than strain-specific anti-3D7.
Domain 1, Domain 2, and Domain 2 ? 3 Polymorphic Sites Are Located
Within Inhibitory Epitopes. By displaying 3D7 polymorphisms onto
the FVO AMA-1 scaffold, we determined the inhibitory con-
tribution of various polymorphic clusters, as indicated by their
ability to deplete strain-specific antibodies and reverse parasite
growth inhibition. Fig. 3 shows results from a typical reversal
assay, showing domain-wise inhibitory contribution of 3D7
mediated by 7% anti-3D7 AMA-1 serum pool against the 3D7 parasite target.
P. berghei AMA-1 (PbAMA) was used as the negative control antigen.
Five serial dilutions (x axis) of the 3D7, the FVO, or the domain
polymorphic differences between 3D7 and FVO AMA-1. Dark lines separate
domain boundaries. (Right) Polymorphic face of P. falciparum AMA-1 (15, 16,
28). Residues selectively switched from FVO to 3D7 type are shown as solid
amino acids. Polymorphisms were switched domain-wise: D1 (red plus blue
clusters within domain 1 were also switched: C1 (red plus blue), C2 (orange),
C1-L (red), C3 (purple), and C1?C2 (red plus blue plus orange). The C-terminal
hydrophobic trough residues (green) (15) are also shown. Domain 1 residues
not included in the cluster chimeras are colored sky blue. Figures were made
by using PyMOL software (www.pymol.org).
Clusters of AMA-1 polymorphic sites. (Left) The table shows the 24
Western blot with mAb 4G2dc1. Anti-3D7 and anti-FVO AMA-1 rabbit sera
pools (1:10,000 dilution) were mixed with an equal volume of 0.5 mg/ml of
FVO or 3D7 AMA-1 proteins, respectively. The cross-reactive antibodies were
blocked, and the free strain-specific anti-3D7 (C) and anti-FVO (D) antibodies
were used in a Western blot. Densitometric analysis of Fig. 2C blot is shown in
SI Fig. 9. Two batches of C1?C2 were loaded on these gels.
Dutta et al.PNAS ?
July 24, 2007 ?
vol. 104 ?
no. 30 ?
specific polymorphisms. At 7% concentration, an anti-3D7
asites in a GIA (Fig. 3). Addition of the rodent Plasmodium
berghei AMA-1 had no effect on inhibition. Homologous 3D7
AMA-1 protein depleted both cross-reactive and strain-specific
antibodies against AMA-1 and resulted in a dose dependent and
complete reversal of inhibition. In comparison, the FVO
depleted only the cross-reactive antibodies, causing significantly
inhibition and antibody concentration is nonlinear (SI Fig. 10).
The 80–60% reduction in inhibition results in ?60% drop in
antibody concentration. Thus, ?40% of antibodies in the poly-
clonal serum pool were 3D7 strain-specific.
The D3 chimera caused baseline reversal that was comparable
with FVO protein (Fig. 3; D3 compared with FVO) therefore,
the 60% strain-specific inhibition was due to antibodies against
domain 1 and domain 2 or domain 2 ? 3 polymorphisms. D1
chimera caused complete reversal of inhibition, similar to 3D7
AMA-1 (Fig. 3; D1, 3D7). This observation suggested that all of
the polymorphic inhibitory epitopes were located on domain 1.
However, this was not the case as the depletion of domain 2 and
domain 2 ? 3 antibodies by D2 and D2?3 also showed reversal
observation suggested that anti-AMA-1 antibody-mediated par-
asite inhibition was complex and resulted from different anti-
body components: 25–30% inhibition was mediated by domain
1 specific antibodies, whereas 30–35% inhibition resulted from
a combination of domain 1 and domain 2-specific antibodies.
Also, the slightly higher reversal by D2?3 as compared with D2,
and the higher reactivity of D2?3 chimera as compared with the
D2 and D3 (Fig. 2C and SI Fig. 9), suggested a role of domain
3 polymorphisms when displayed along with domain 2 polymor-
phisms. Overall the domain 1 polymorphisms played a dominant
role in binding to inhibitory antibodies.
Strain-Specific Inhibitory Antibodies to Domain 1 Target a Cluster of
13 Polymorphic Residues. The 18 domain 1 polymorphisms were
further divided into 3 clusters C1, C2, and C3 (Fig. 1) and tested
in a GIA reversal assay. Fig. 4 shows results from one such assay.
Chimera C1?C2 mediated reversal that was similar to D1 and
3D7 AMA-1 protein, indicating that the majority of the domain
1 inhibitory epitopes were in the C1 and C2 clusters (Fig. 4,
compare C1?C2, D1, and 3D7). C3, because of its low strain-
specific antigenicity (Fig. 2C), showed reversal that was slightly
greater than baseline reversal by FVO protein (Fig. 4, compare
FVO and C3). Further fragmentation of the C1?C2 cluster into
C1 (eight polymorphisms) and C2 (five polymorphisms) showed
significantly less reversal as compared with C1?C2. Mixing
experiments showed that equimolar C1 and C2 proteins caused
significantly less reversal than C1?C2 (data not shown). There-
fore, inhibitory strain-specific antibodies to domain 1 recognize
a surface footprint that incorporates both C1 and C2. This
observation may also explain why a previous study (12) con-
cluded that polymorphisms on the C1 cluster alone (residues
178–260) did not significantly contribute to the absence of
cross-strain inhibition in the 3D7-W2mef model.
An additional chimera, C1-L, displaying a short 3D7 loop
containing five highly polymorphic sites, was also tested. Sur-
prisingly, C1-L (Fig. 4; C1-L) showed reversal that was at least
as high or in some experiments higher than C1. This observation
indicated that polymorphisms on the C1-L loop had the highest
inhibitory contribution within C1?C2. Residues 167 and 300 (in
sky blue in Fig. 1) that were a part of D1 chimera were not
included in domain 1 cluster chimeras because of their relative
isolation on primary and tertiary structure. Because C1?C2
polymorphisms accounted for the majority of the strain-specific
inhibitory activity of D1, the inhibitory role of residues 167 and
300 was presumed to be minor.
AER Map of 3D7 AMA-1. Depletion of strain-specific antibodies by
recombinant proteins prevent their binding to the parasite
resulting in reversal of inhibition. In the field, polymorphic
differences between vaccine and challenge strain are thought to
reduce binding of strain-specific antibodies, resulting in anti-
genic escape. Therefore the in vitro reversal of inhibition by
chimeric proteins is an indirect strategy to map the AER.
Because 3D7 AMA-1 protein depletes antibodies to all of the 24
3D7 polymorphic sites, its corresponding reversal represents
100% ‘‘relative escape’’ (RE). Likewise, FVO AMA-1 does not
deplete any 3D7-specific antibodies (it depletes only cross-
reactive antibodies); its reversal represents 0% RE. Fig. 5 shows
reversal experiments, plotted on the 100–0% RE scale. RE per
polymorphic residue was also calculated by dividing the mean
target. P. berghei AMA-1 (PbAMA) was used as the negative control antigen.
Five serial dilutions (x axis) of the 3D7, FVO AMA-1, or the domain 1
was reversed by a 1.75 ?M concentration of 3D7, FVO, or the chimeric AMA-1
parasitemia in adjuvant control well). RE (y axis) was determined by subtract-
ing % invasion in the presence of FVO AMA-1 antigen from % invasion in the
presence of the test antigen and then adjusting RE for 3D7 AMA-1 to 100%
and FVO AMA-1 to 0% (not plotted). This calculation allowed the reversal of
inhibition due to cross-reactive antibodies to be canceled out. Mean RE ? SD
of three independent experiments is shown. RE/residue (dots) ? mean RE/
number of residues displayed by a particular chimera.
www.pnas.org?cgi?doi?10.1073?pnas.0701464104Dutta et al.
RE by the number of 3D7-specific residues being displayed by
There was no significant difference between the RE of 3D7,
D1 and C1?C2 (3D7 vs. D1, P ? 0.99; 3D7 vs. C1?C2, P ? 0.5).
This observation suggested that C1?C2 contained the most
important AER of the 3D7 AMA-1 vaccine, because depletion
of antibodies to its 13 polymorphic sites accounted for ?100%
escape in the 3D7-FVO model. Within C1?C2, the loop C1-L
had the highest impact, 51% RE, which was higher than 36% RE
for C1; however, this difference failed to reach statistical sig-
nificance (P ? 0.2). The five polymorphic sites within C1-L were,
therefore, most likely the most important component of the
C1?C2 AER cluster. Outside of domain 1, the 65% RE for
D2?3 was slightly higher than the 52% RE for D2, but not
significantly different (P ? 0.8). This observation confirmed that
the four polymorphic sites within D2 form the second most
important AER cluster, and that domain 3 polymorphic sites
enhanced the overall inhibitory contribution of domain 2 poly-
morphisms. RE for C2, C3, and D3 were not significantly
different from heterologous FVO AMA-1 (FVO vs. D3, C2, C3;
P ? 0.6, 0.3, and 0.9, respectively).
When RE per residue was considered (Fig. 5; dots), D2,
(?10% RE per residue). D1, C1?C2, C1, and D3 AER had
mid-level impact (5–7%), and C2 and C3 clusters contained
polymorphisms with the lowest impact (?3% RE per residue).
Because the D3 polymorphisms by themselves are not contrib-
uting to inhibition, the most significant AER can be assigned to
D2 and C1-L.
Map. Groups of three rabbits were vaccinated with 100 ?g each
3D7 AMA-1, FVO AMA-1, chimeras D1, D2, D2?3, or the
Sera from three rabbits within the group were pooled, and
AMA-1-specific IgG were affinity purified and concentrated to
an equivalent 4 mg/ml. A SDS/PAGE confirmed similar purity
of the IgG fractions (SI Fig. 11). Three independent GIA
experiments were conducted against both the 3D7 and FVO
target (SI Fig. 12), and the average ED50(effective dose for 50%
invasion inhibition) was plotted (Fig. 6, left y axis). ELISA was
also performed, by using 3D7 or FVO AMA-1-coated plates.
ELISA unit was defined as ng/ml IgG needed for OD415? 1.0
(Fig. 6, right y axis).
Inhibition of 3D7 target. The anti-3D7 AMA-1 showed the highest
reactivity against the homologous 3D7 antigen (lowest ELISA
units). Comparatively, antibodies against the domain chimeras
showed lower ELISA reactivity. Anti-3D7 also had the lowest
ED50against the 3D7 strain (84 ?g/ml) by GIA. Anti-D1 (170
?g/ml), anti-D2 (223 ?g/ml), and anti-D2?3 (211 ?g/ml) showed
significantly higher ED50 (P value compared with anti-3D7,
?0.001). There was no significant difference in the ED50 for
anti-D2 and anti-D2?3. As predicted by the AER map, the
high-level strain-specific inhibition of 3D7 parasite is primarily
due to inhibitory targets on domain 1 and domain 2. The biallelic
vaccine showed the best inhibitory activity, despite the lower
Inhibition of FVO target. Anti-FVO showed the highest reactivity
against the homologous FVO antigen by ELISA. Interestingly,
IgG against domain chimeras containing FVO-specific domain
1 (i.e., anti-D2 and anti-D2?3) showed ELISA reactivity com-
parable with anti-FVO. Not surprisingly therefore, the ED50for
anti-D2 (214 ?g/ml) and anti-D2?3 (215 ?g/ml) were statisti-
cally not different from the 127 ?g/ml ED50for anti-FVO (P ?
0.5). In comparison, antisera against vaccines that did not
contain FVO domain 1, i.e., anti-3D7 (824 ?g/ml) and anti-D1
(649 ?g/ml), showed significantly higher ED50 (P ? 0.0001).
Therefore, unlike the 3D7 target, the inhibition of FVO parasite
depends only on the domain 1 genotype of the vaccine. A GIA
reversal experiment further corroborated this finding, where
only the D1 chimera (but not D2, D2?3, or D3) showed reduced
ability to reverse anti-FVO AMA-1-mediated inhibition of the
FVO parasite (SI Fig. 13). As with the 3D7 target, the biallelic
vaccine showed superior inhibitory activity; despite the lower
ELISA reactivity, its ED50 was comparable with the FVO
Chimeric Proteins as Vaccine Candidates. Efficacy of an AMA-1
vaccine primarily depends on the quantity and quality of the
induced antibodies. Whole sera from the rabbits were tested in
a GIA against unrelated P. falciparum strains CAMP, 7G8, HB3,
DD2, and M24 (Fig. 7 and SI Fig. 14). Observations in Aotus
trials suggest that ?60% inhibition in a GIA correlates protec-
tion (18). We have made similar observations in a 3D7 AMA-1
vaccine trial, in Aotus monkeys, against a closely homologous
FCH4 strain challenge (our unpublished data). By using ?60%
inhibition at 20% serum concentration, as a cutoff inhibition
which correlates protection (dotted line, Fig. 7), the GIA data
inhibition (ED50) was determined against 3D7 and FVO target parasites. The
Right y axis (circles), ELISA units (ng/ml IgG that resulted in OD415? 1.0)
determined by using the homologous 3D7 or FVO AMA-1 coat antigens.
Left y axis (bars), concentration of IgG required for 50% invasion
Dutta et al.PNAS ?
July 24, 2007 ?
vol. 104 ?
no. 30 ?
were analyzed. Anti-3D7 inhibited only the 3D7 strain. Anti-
FVO inhibited the FVO strain and also CAMP and 7G8 strains.
Anti-3D7?FVO (biallelic vaccine), inhibited 3D7, FVO,
CAMP, and 7G8 strains. Among the chimeric vaccines, anti-
D2?3 induced antibodies that almost matched the inhibitory
spectrum of the biallelic vaccine, inhibiting the 3D7, FVO,
CAMP, and 7G8 strains. A confirmatory GIA with affinity
purified anti-AMA-1 confirmed the high-level cross-strain in-
hibitory activity of anti-D2?3 (SI Fig. 15). These results dem-
onstrate that it is possible to generate cross-reacting antiserum
by mixed vaccination, or by using AMA-1 chimeras.
Within the sequence limits of the AMA-1 constructs, the
CAMP, HB3, 7G8, M24, and DD2 strains differ from 3D7
AMA-1 at 19, 24, 26, 24, and 22 residues, respectively (SI Table
2). These strains are closer to FVO AMA-1, differing at 16, 21,
16, 22, and 19 residues, respectively. This is the likely reason for
higher cross-strain inhibitory activity of anti-D2 and D2?3 (and
not the anti-D1 chimera).
Determining the inhibitory contribution of 24 polymorphic
residues of the 3D7 AMA-1 vaccine has direct relevance to
vaccine development efforts. The highest impact AER on 3D7
domain 1 mapped to an ?-helix-containing loop, C1-L. This loop
includes residue 197 that is critical for the binding of an
inhibitory mAb 1F9 (19). Parasite strains that differ from a
vaccine within the C1-L polymorphisms are likely to show
maximal escape. Mapping of a second important AER cluster on
domain 2 was surprising, because population analyses suggest
that only domain 1 and domain 3 are under strong diversifying
selection (6–8). Domain 2 and C1-L genotyping may provide an
important readout of allelic effects in upcoming vaccine trials.
Domain 3 polymorphisms, by themselves, do not seem to be a
major target of inhibitory antibodies, particularly when the
whole ectodomain is used as an immunogen, and similar findings
observed that domain 3 polymorphisms enhanced the overall
inhibitory activity of domain 2 polymorphisms. Critically, do-
strains to generate chimeric proteins that induced inhibitory
response against both the vaccine strains.
To determine whether the high-impact AER genotype can
predict antigenic escape, cross-strain GIA data in Fig. 7 was
correlated with AMA-1 sequences (SI Table 2). Anti-FVO
showed biologically significant inhibition against FVO, 7G8 and
CAMP, these strains belong to the ‘‘FVO inhibition group’’ (Fig.
7). Within the ‘‘high-impact’’ AER of C1-L and domain 2
clusters, the 7G8 and CAMP share a high level of similarity
compared with the escape strains. Although no other members
of the ‘‘3D7 inhibition group’’ were identified, it is reported that
results, none of the 9-aa differences between 3D7-D10 strains
map to C1-L, and only one difference is within the domain 2
AER cluster (SI Table 2). These data suggest that examination
of AER genotypes will be useful for strain selection of multial-
lelic vaccines, or to predict escape frequency during efficacy
C1?C2 AER are present on loops with their hydrophilic
side-chains extending away from the conserved PAN domain
core (Fig. 1). Domain 2 AER also localized to loops, although
much shorter loops (Fig. 1). Structural plasticity allows loops to
folding. Presumably, radical changes can most effectively dimin-
ish antibody binding, and this requirement coupled with the AT
codon bias of P. falciparum leaves only a few highly restricted
choices of mutations that are positively selected for by the
immune system. This reasoning could explain the limited rep-
ertoire of amino acid substitutions (most sites are dimorphic)
and the resulting balancing selection reported among field
isolates (6, 22).
AER within the C1?C2 cluster map adjacent to the hydro-
phobic trough of AMA-1 (Fig. 1). The domain 2 AER map in
close spatial proximity to a proteolytic processing site at residue
517 (Fig. 1) (23). We have previously shown that bivalent
polyclonal antibodies to AMA-1 inhibit invasion by steric block-
of AMA-1 due to their cross-linking effect (24, 25). It is possible
that strain-specific antibodies to domain 1 block access to the
hydrophobic trough and the domain 2 ? 3 antibodies cross-link
AMA-1 and inhibit the C-terminal cleavage. If steric blocking of
with cross-linking), the dominant role of domain 1 AER can be
Although AMA-1 has ?60 polymorphic sites, mapping the
AER in the 3D7-FVO model provides a model to rationalize the
basis of antigenic escape from antibodies against a vaccine strain
of AMA-1. In future efficacy trials, genotyping the escape
variants would determine whether the high-impact AER clusters
identified here universally mediate antigenic escape. We also
demonstrate the feasibility of combining epitope mapping with
protein engineering approaches to improve the cross-strain
inhibitory activity of AMA-1 vaccines.
Immunization and Affinity Purification of IgG. Groups of three
rabbits were administered three doses of 100 ?g each of antigen
along with the adjuvant Montanide ISA720 adjuvant (Seppic,
Inc., Paris, France) (26). Sera from all three rabbits in a group
were pooled (0.7 ml each). The 2.1-ml serum pool was passed
over a protein G column (Amersham, Piscataway, NJ). IgG
binding and elution buffers (Pierce, Rockford, IL) were used
according to the manufacturer’s instructions. After neutraliza-
tion with 1 M Tris (pH 8.0), the IgG were dialyzed against PBS.
An AMA-1 affinity column was prepared by binding 5 mg/ml
(Amersham). IgG were passed over this column, and the flow-
strains (SI Fig. 14). The mean invasion (?SD) of three rabbits at 20% serum dilution was plotted.
www.pnas.org?cgi?doi?10.1073?pnas.0701464104Dutta et al.