This study analyzed genetic diversity and population structure of Plasmodium falciparum in 5 populations in the Brazilian Amazon region. Microsatellite markers were analyzed in 196 parasite isolates. There was significant multilocus linkage disequificance within populations, particularly those with fewer mixed infections. However, most isolates had unique multilocus genotypes, indicating genetic diversity. Genetic divergence between populations was substantial but did not correlate simply with geographical distance. The findings suggest distinct population structures and minimal gene flow between foci in the region.

1. MAJOR ARTICLE
Genetic Structure of Plasmodium falciparum
Populations in the Brazilian Amazon Region
Ricardo L. D. Machado,1,4 Marinete M. Povoa,1 Vanja S. P. Calvosa,1 Marcelo U. Ferreira,3 Andrea R. B. Rossit,4
´
Eduardo J. M. dos Santos,2 and David J. Conway5
1
Malaria Laboratory, Evandro Chagas Institute and National Health Foundation, and 2Para University, Brazil, Belem/Para, and 3Department
´ ´ ´
of Parasitology, Institute of Biomedical Sciences, University of Sao Paulo, Cidade Universitaria, and 4Faculty of Medicine of Sao Jose
˜ ´ ˜ ´
do Rio Preto, Center for Investigation of Microrganisms, Department of Dermatological, Infectious, and Parasitic Diseases, Sao Paulo, Brazil;
˜
5
Department of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, London, United Kingdom
After a major increase in incidence between the 1970s and the 1990s, the Brazilian Amazon region now
accounts for the most cases of Plasmodium falciparum malaria in the Americas. Polymorphism of 10 micro-
satellite loci in the P. falciparum genome was studied in 196 isolates obtained from 5 populations in the region.
There was significant multilocus linkage disequilibrium, particularly within populations with lower proportions
of mixed-genotype infections. However, most multilocus genotypes in different isolates were distinct, and
there was no evidence of any recent epidemic expansion of particular clones. Genetic divergence between
populations was very substantial but did not fit a simple model of isolation by distance. Thus, different foci
of P. falciparum in Brazil are quite independent, with distinct population structures and minimal gene flow,
a finding that has implications for strategies to control infection and to contain the spread of drug resistance
at a regional level.
A large proportion of all malaria cases in South America or whether they mainly emerge from local endemic pop-
occur in the Brazilian Amazon region, where the in- ulations [3]. Insecticide spraying has been useful in re-
cidence increased greatly between the 1970s and the ducing transmission in some areas, but the main strategy
early 1990s, with ∼500,000 cases now reported annually of control is the prompt diagnosis and treatment of cases
[1, 2]. Although Plasmodium vivax is the most common [1]. However, because the epidemiological factors are
malaria parasite in this region, P. falciparum accounts variable and complex and because resistance to different
for ∼30% of cases overall and is a greater cause of antimalarials has emerged [4], an overall understanding
morbidity and mortality. The distribution of P. falci- of the population genetic structure of the parasite is im-
parum infection is focal, being more common than P. portant as a biological basis for considering control strat-
vivax in some areas but very rare or absent in others. egies in the future. In Asia and Africa, particular drug-
The emergence of epidemic and endemic foci has been
resistance alleles have spread on a continental scale [5–
influenced by colonization of different areas by diverse
7], but these same alleles have not been detected in the
groups of people, but it is unclear whether parasites are
Americas, where the geographical spread of resistance
commonly spread from one area to another by migrants
appears to be more complex [4, 5].
Microsatellite loci that contain polymorphic, simple
sequence repeats are highly informative genetic mark-
Received 22 March 2004; accepted 13 May 2004; electronically published 21
September 2004. ers. The P. falciparum genome contains an abundance
Financial support: British Council (BRAS/881/161); Conselho Nacional de of such loci, predominantly having dinucleotide (TA)n
Desenvolvimento Cientifico e Tecnologico (grants 910119/95.3 and 300543/98-0);
´
UK Medical Research Council (grant G9803180); Evandro Chagas Institute/Fundacao ¸˜ or trinucleotide (TAA)n repeats, with an average of 1
de Saude.
´ microsatellite locus every 2–3 kb [8, 9]. These markers
Reprints or correspondence: Dr. David J. Conway, Dept. of Infectious and Tropical
Diseases, London School of Hygiene and Tropical Medicine, Keppel St., London have been used in studies to map traits that are inher-
WC1E 7HT, UK (david.conway@lshtm.ac.uk). ited in the progeny of an experimental cross [10] and
The Journal of Infectious Diseases 2004; 190:1547–55
ᮊ 2004 by the Infectious Diseases Society of America. All rights reserved.
to analyze and compare the genetic structures of pop-
0022-1899/2004/19009-0004$15.00 ulations in different parts of the world [11, 12]. Three
P. falciparum Genetics in Brazilian Amazon • JID 2004:190 (1 November) • 1547

2. populations in South America (from Colombia, Bolivia, and Bra-
zil) from which samples were obtained each showed statistically
significant, nonrandom associations among alleles of 12 different
microsatellite loci that are widely scattered in the genome, a
pattern of multilocus linkage disequilibrium that indicates less-
frequent recombination in these populations than in populations
in other parts of the world [11]. There was also a very high
degree of divergence in microsatellite allele frequencies between
these South American populations, indicating that they were
genetically very isolated from each other. Some other studies,
which used a small number of polymorphic antigen gene loci,
have described extreme patterns of low genetic diversity and
linkage disequilibrium in P. falciparum populations elsewhere in
South and Central America, including French Guiana [13], Hon-
duras [14], and the Venezuelan Amazon region [15, 16]. How-
ever, Brazilian populations have not shown such extremely low Figure 1. Map of the Brazilian Amazon region, showing the locations
levels of allelic diversity in antigen genes [17–19]. of the 5 populations from which isolates of Plasmodium falciparum studied
Here, 10 microsatellite loci are studied in 196 P. falciparum here were obtained: Rio Branco (Acre state [AC]), Porto Velho (Rondonia
ˆ
state [RO]), Tailandia and Maraba (Para state [PA]), and Serra do Navio
ˆ ´ ´
isolates obtained from 5 endemic populations in the Brazilian
(Amapa state [AM]).
´
Amazon region. Levels of allelic diversity differ only slightly
between the populations, but levels of multilocus linkage dis-
equilibrium are markedly different. As expected, the multilocus scopic examination of thick blood films or by the Parasight F
linkage disequilibrium is highest in populations with the lowest diagnostic test for circulating antigen (P. falciparum histidine-
proportions of mixed-clone infections, so it appears to be a rich protein 2). From each subject, a 5-mL blood sample for
sensitive reflection of the degree of inbreeding in the different parasite DNA analysis was collected, after informed consent
populations. This is not accounted for by recent “epidemic” was obtained and with approval from the ethical committees
expansion of identifiable clones, since most isolates have unique of the Evandro Chagas Institute and the National Health Foun-
multilocus genotypes, but it is indicative of ongoing reduced dation in Brazil. Samples were treated with proteinase K, and
levels of recombination in these populations. Allele frequency nucleic acids were extracted by use of 2 rounds of phenol:
patterns do not indicate recent severe genetic bottlenecks in chloroform:isoamyl alcohol (25:24:1), 1 round of chloroform,
any of the populations. The degree of genetic divergence be- and 1 round of ether, followed by ethanol precipitation. The
tween pairs of the populations does not correlate with the extracted nucleic acid samples were dissolved in sterile, pure
geographical distance between them. deionized water and stored at Ϫ20ЊC before use.
Genotyping of P. falciparum microsatellites. Ten poly-
POPULATIONS, MATERIALS, AND METHODS morphic P. falciparum microsatellite loci were studied (chro-
mosomal locations in parentheses) [9]: Polya (Chr4), TA42
Study populations and parasite isolates. The 196 P. falci- (Chr5), TA81 (Chr5), TA1 (Chr6), TA87 (Chr6), TA109 (Chr6),
parum isolates studied here were collected from 5 geographical ARA2 (Chr11), PfPK2 (Chr12), Pfg377 (Chr12), and TA60
populations in the Brazilian Amazon region: 39 isolates were (Chr13). These 10 loci are among those used in 2 large analyses
from Maraba (Para State) in 1998 and 1999, 36 were from
´ ´ of global populations [11, 12]. Seminested polymerase chain
Tailandia (Para State) in 1998, 55 were from Porto Velho (Ron-
ˆ ´ reaction amplification was performed with a dye-labelled inner
donia State) between 1996 and 1998, 34 were from Rio Branco
ˆ primer for each locus, and the sizes of allelic fragments were
(Acre State) in 1996 and 1997, and 32 were from Serra do analyzed, by electrophoresis, on an automated sequencer (ABI
Navio (Amapa State) between 1989 and 1992. The isolates from
´ 377), with scoring and checking by use of GENESCAN and
Porto Velho comprise the majority of a sample set that has GENOTYPER programs (Applied Biosystems). We used the
been previously genotyped and reported in an analysis of pop- method described by Anderson et al. [20], with modifications
ulations from different countries [12]. Figure 1 shows the geo- in the combination of florescent dyes used [12].
graphic location of each of the populations from which samples Statistical analyses. In each isolate, the single or predom-
were obtained. The 2 most closely situated locations were Tai- inant allele at each locus was scored and counted toward the
landia and Maraba (330 km apart), and the 2 most distant were
ˆ ´ population sample for that location. If an additional allele was
Tailandia and Rio Branco (2250 km apart). Individual subjects
ˆ present (due to the presence of a second parasite clone in the
with P. falciparum infections were initially identified by micro- infection), this was recorded, but the allele did not count toward
1548 • JID 2004:190 (1 November) • Machado et al.

3. Table 1. Proportion of mixed-clone infections, mean nos. of alleles, and allelic diversity in 5 Plasmodium falciparum
populations in which isolates were typed at 10 microsatellite loci.
Proportion of No. of alleles
Proportion of isolates all individual in the population Diversity
No. of isolates with 11 allele at locus scores per locus, (virtual heterozygosity),
Population studied any of the 10 loci with 11 allele mean ‫ ע‬SE mean ‫ ע‬SE
Porto Velhoa 55 0.31 0.058 4.5 ‫ע‬ 0.5 0.52 ‫ע‬ 0.04
Serra do Navio 32 0.19 0.028 4.4 ‫ע‬ 0.4 0.46 ‫ע‬ 0.07
Maraba ´ 39 0.38 0.059 4.1 ‫ע‬ 0.5 0.44 ‫ע‬ 0.07
Tailandia
ˆ 36 0.14 0.019 3.2 ‫ע‬ 0.3 0.39 ‫ע‬ 0.07
Rio Branco 34 0.17 0.015 4.4 ‫ע‬ 0.7 0.51 ‫ע‬ 0.07
a
The isolates from Porto Velho comprise the majority of a sample set that has been described elsewhere [12].
the population sample. The within-population diversity at each to identify more loosely related haplotypes that may not be
locus was summarized by use of the virtual heterozygosity in- assumed to belong to a clonal complex (identical to others at
dex, H, equal to 1 minus the sum of the squared allele fre- 7 or 8 of the 10 loci studied).
S
quencies [21]. Between-population measurements of differ- The standardized index of association (IA ) among alleles from
entiation for each locus (and values averaged over all loci) were all 10 loci was calculated to test for evidence of multilocus
performed by use of the fixation indices FST and RST. The FST linkage disequilibrium in each population. This index is a mod-
index is based on the differences in allele frequencies in each ification of the index of association originally described else-
population, giving a potential range from 0 (no frequency dif- where [27], which has been standardized for the number of
ferences) to 1.0 (completely fixed differences with no shared loci used so that comparisons are possible among studies using
S
alleles). Values of the v version of FST [22] were calculated by different numbers of loci. The IA values were calculated by use
use of the DOS-based FSTAT program (version 1.2) [23], which of the LIAN 3.0 program [28], with permutation testing of the
S
performs calculations in the same way as a more accessible null hypothesis of complete linkage equilibrium (IA p 0 ). For
Windows-based version (2.9.3.2) (http://www2.unil.ch/popgen each population, the analysis was performed in 3 different ways,
/softwares/fstat.htm). This FST index was calculated for each of following an example from a previous study [11], as explained
the 10 microsatellite loci studied here. The standardized RST in Results.
index is based on the variance in numbers of copies of a simple A test for evidence of a bottleneck in each population was
sequence tandem repeat (microsatellite length) among alleles applied, by comparing the virtual heterozygosity, H, at each
within a population and between populations, as performed locus with that expected under mutation-drift equilibrium,
with the RSTCALC program [24]. The RST index was calculated given the number of different alleles in each population sample.
for 6 of the microsatellite loci studied here (TA81, TA87, ARA2, It is considered that a bottleneck will result in observed values
PfPK2, Pfg377, and TA60), in which the polymorphism has been of virtual heterozygosity being greater than expected from the
considered to conform to a stepwise mutation model (SMM) numbers of alleles (since the numbers of alleles will have been
[25]. Correlations between genetic distance (FST or RST) and disproportionately reduced). The expectations were generated
geographical distance between pairs of populations were ex- by use of 2 different models for the relationships among alleles:
plored by use of nonparametric correlation coefficients, and (1) the infinite alleles model (IAM) applied to all 10 loci and
the significance was tested with a partial Mantel test of matrix (2) the SMM applied to the 6 loci noted above, by use of the
correlation by use of GENEPOP with a Web interface (http:// BOTTLENECK program (version 1.2.02) [29]. The proportions
wbiomed.curtin.edu.au/genepop/). of loci with observed values higher or lower than expected were
Genetic relationships among the 10-locus microsatellite hap- compared with random expectations, by use of an exact test.
lotypes were analyzed by use of the eBURST algorithm (http:
//eburst.mlst.net). eBURST identifies groups of related haplo- RESULTS
types on the basis of a simple evolution model in which an
ancestral or “founding” haplotype increases in frequency in the Allele frequency distributions and within-population diver-
population and, while doing so, begins to diversify to produce sity. Each of the 196 P. falciparum isolates from 5 different
a cluster of closely related haplotypes [26]. A stringent defi- populations in the Brazilian Amazon region was typed for all
nition was used to identify groups of haplotypes that are be- 10 microsatellite loci. A minority of isolates apparently con-
lieved to descend from a single founding haplotype (“clonal tained 11 P. falciparum genotype (with 11 allele at any of the
complexes” of haplotypes that are identical to others at 9 of 10 loci). The proportion of such isolates differed significantly
the 10 loci studied), and a less-stringent definition was used among populations (P ! .05 ), ranging from 14% in Tailandia
ˆ
P. falciparum Genetics in Brazilian Amazon • JID 2004:190 (1 November) • 1549

4. Table 2. Allele frequencies of 10 Plasmodium falciparum microsatellite loci in
5 Brazilian populations.
Porto Velho Serra do Navio Maraba Tailandia Rio Branco
Locus, allelea (n p 55)b (n p 32) (n p 39) (n p 36) (n p 34)
G377
95 0.473 0.156 0.103 0.194 …
98 0.527 0.844 0.897 0.806 0.735
104 … … … … 0.265
TA109
160 … 0.094 0.051 … 0.029
163 0.127 0.063 0.026 0.028 0.647
166 0.764 0.781 0.923 0.972 0.324
169 … 0.031 … … …
178 0.036 0.031 … … …
197 0.055 … … … …
202 0.018 … … … …
TA60
68 … … … … 0.029
71 … 0.063 0.026 … …
74 0.545 0.281 0.231 0.528 0.676
77 … 0.031 0.026 … 0.059
80 0.073 0.063 … … …
83 0.218 0.469 0.462 0.083 0.206
86 0.164 0.094 0.256 0.389 …
89 … … … … 0.029
TA81
110 … … … … 0.029
113 … 0.031 … … …
116 … … … … 0.029
119 0.018 … … … …
122 0.673 0.313 0.077 0.806 0.706
125 0.309 0.625 0.923 0.139 0.206
128 … 0.031 … 0.056 …
134 … … … … 0.029
ARA2
63 … … 0.077 … …
66 0.709 0.281 0.615 0.222 0.147
69 0.236 0.438 0.256 0.500 0.206
72 0.055 0.156 0.051 0.028 0.088
75 … 0.063 … … 0.382
81 … 0.063 … … 0.059
84 … … … 0.250 0.088
90 … … … … 0.029
Polya
138 … … … … 0.029
141 … … … … 0.029
144 … … … … 0.088
150 … … … … 0.059
153 0.164 0.063 0.128 0.028 0.588
156 … … … … 0.029
165 0.055 0.188 0.128 0.750 …
172 … … … … 0.029
177 0.073 … … … …
180 … 0.031 0.103 … …
183 0.509 0.719 0.615 0.222 0.118
189 0.200 … 0.026 … 0.029
(continued)

5. Table 2. (Continued.)
Porto Velho Serra do Navio Maraba Tailandia Rio Branco
Locus, allelea (n p 55)b (n p 32) (n p 39) (n p 36) (n p 34)
TA42
188 0.836 0.781 0.769 0.972 0.265
191 0.073 0.031 0.103 … …
203 0.073 0.188 0.128 0.028 0.676
206 0.018 … … … 0.059
TA1
163 0.018 … … … …
169 0.018 … … 0.056 …
172 0.036 0.156 0.231 0.389 0.059
175 0.673 0.656 0.667 0.528 0.206
178 0.127 0.094 … … 0.647
181 0.055 … … … …
184 0.073 0.063 0.103 0.028 0.088
187 … 0.031 … … …
PfPK2
163 0.036 0.063 … … …
166 0.018 … 0.052 0.028 …
169 0.273 0.094 0.128 0.389 0.029
172 0.145 0.281 0.077 0.083 …
175 0.509 0.563 0.718 0.500 0.971
178 0.018 … … … …
184 … … 0.026 … …
TA87
97 0.018 0.031 0.026 … 0.088
100 0.382 0.469 0.077 … 0.176
103 … 0.094 … … …
106 0.018 0.125 0.205 0.028 …
112 0.091 0.156 0.205 0.139 0.324
115 0.473 0.125 0.436 0.833 0.412
118 0.018 … 0.051 … …
a
Allele names refer to the base-pair sizes of the allelic products from the polymerase chain
reaction.
b
The isolates from Porto Velho comprise the majority of a sample set that has been described
elsewhere [12].
to 38% in Maraba (table 1). Most of these mixed-genotype
´ First, all majority allele scores for an isolate, for the 10 loci,
infections had mixed alleles at only 1 or 2 of the 10 loci typed, were considered to indicate an infection haplotype (admitting
so that the overall proportion of all individual locus scores that that this could technically result in some “false” haplotype pro-
had 11 allele was very low, ranging from 1.5% in Rio Branco files in isolates that have mixed alleles at у2 loci). Second, the
to 5.9% in Maraba (table 1). In all cases, there was a clear
´ mixed-clone isolates were excluded so that only the single-clone
majority allele on the electropherogram, and this was counted infections were analyzed (giving slightly smaller sample sizes
toward the sample of allele frequencies. Among the 5 popu- but absolute confidence in the haplotype profiles). Third, any
lations, there were only very slight differences in the mean multilocus genotypes that occurred in 11 isolate were only
number of alleles per locus or the gene diversity (virtual het- counted once in the analysis (thus collapsing the data slightly
erozygosity) index (table 1). The allele frequencies for each of and thereby removing the possible effect of recent epidemic
the populations are given in table 2. expansion of particular clones).
Tests for multilocus linkage disequilibrium within popu- The analysis of all infection haplotypes indicates significant
S S
lations. The IA was used to test for evidence of multilocus multilocus linkage disequilibrium (IA 1 0 ) in all 5 of the pop-
linkage disequilibrium in each of the populations, with per- ulations (table 3). Similar results were obtained when only sin-
S
mutation testing of the null hypothesis of IA p 0 (equilibrium gle-clone infections were analyzed, except that the value for 1
of multilocus frequencies). This test was applied to the data population was no longer significant (table 3). The results were
from each population in 3 ways, as in a previous study [11]. also very similar when each haplotype was counted only once,
P. falciparum Genetics in Brazilian Amazon • JID 2004:190 (1 November) • 1551

6. Table 3. Multilocus linkage disequilibrium (index of associ- and little or no linkage disequilibrium in 2 populations (Porto
ation IS]) among 10 Plasmodium falciparum microsatellite loci in
A Velho and Maraba). The significant multilocus linkage dis-
´
5 Brazilian populations. equilibrium in the first 3 populations cannot be attributed to
recent epidemic expansion of particular clones, since signifi-
Unique
All infections Single clones haplotypes cance remains in the analysis that counts each distinct hap-
lotype only once. As expected, a comparison of results in tables
Population No. IS
A No. IS
A No. IS
A
a a
1 and 3 shows that the populations with the fewest mixed-
Porto Velho 55 0.016 38 0.021 34 0.014
genotype infections (Tailandia and Rio Branco) had the strong-
ˆ
Serra do Navio 32 0.042b 26 0.043
b
25 0.034
a
Maraba ´ 39 0.032
a
24 0.023 21 0.005
est linkage disequilibrium, and those with the most mixed-
Tailandia
ˆ 36 0.094
c
31 0.107
c
23 0.073
c genotype infections (Porto Velho and Maraba) had the weakest
´
Rio Branco 34 0.113c 29 0.124c 29 0.124c linkage disequilibrium. Overall, the proportion of all locus
a
P ! .05, by test of randomness (IS p 0).
scores with 11 allele (table 1) and the multilocus linkage dis-
A
b
P ! .01, by test of randomness (IS p 0).
A equilibrium of unique haplotypes (table 3) are the population
c
P ! .001, by test of randomness (IS p 0).
A
variables that show the strongest and most-significant negative
correlation (Spearman’s R p Ϫ1.0; P ! .01).
except that the values for 2 populations were then no longer Genetic relationships among isolates. Most isolates were
significant (table 3). Overall, this indicates strong linkage dis- genetically distinct from all others. Among the 196 isolates
equilibrium in 2 populations (Tailandia and Rio Branco), mod-
ˆ studied, there were 172 different combinations of predominant
erate linkage disequilibrium in 1 population (Serra do Navio), alleles at each of the 10 microsatellite loci within an isolate
Figure 2. Relationships among 165 different 10-locus microsatellite infection haplotypes in 189 isolates collected in the Amazon Basin of Brazil,
as defined by an eBURST analysis. Each line connects haplotypes that are identical at 9 of the 10 loci, and all connected haplotypes are considered
to be possible descendants from a single founding haplotype, forming 11 different “clonal complexes” (with 2–32 haplotypes each). The majority of
haplotypes are shown as isolated points, since they differ by у2 microsatellite loci from all others (though they are shown in the figure because
they share at least 7 alleles with у1 others). Seven other haplotypes, each recovered from a single isolate, are not shown, since they differ by 13
alleles from all others. The size of the circles representing haplotypes is proportional to the number of isolates in which the haplotype was seen,
ranging from 1 (in most cases) to 6 (for the most-common haplotype, 89). The large haplotype cluster in the center of the figure contains isolates
only from Rio Branco, whereas the 2 large clusters at the left of the figure each contain isolates from different populations.
1552 • JID 2004:190 (1 November) • Machado et al.

7. Table 4. Genetic differentiation (FST and RST) and geographical tween populations are not significantly correlated with geo-
distance between populations. graphical distances between them (table 4). Rio Branco was the
most differentiated of the populations (compared with other
Geographical FST RST
populations: mean FST p 0.27; mean R ST p 0.23).
Population pair distance, km (mean, 10 loci) (mean, 6 loci)
Test for evidence of genetic bottlenecks. A parasite popu-
MRB-TAI 330 0.23 0.22
lation with very unstable endemicity might be prone to a genetic
SNV-TAI 390 0.20 0.19
PVL-RBR 450 0.23 0.28 bottleneck due to fluctuations of effective population size down
SNV-MRB 600 0.05 0.15 to occasional very low levels. Evidence of a recent bottleneck
PVL-MRB 1575 0.12 0.17 can be sought by comparing the gene diversity (virtual het-
PVL-SNV 1725 0.08 0.08 erozygosity) derived from the observed allele frequency distri-
PVL-TAI 1830 0.16 0.16
bution at each locus, compared with the theoretically expected
MRB-RBR 2025 0.30 0.30
SNV-RBR 2175 0.24 0.20
diversity under mutation-drift equilibrium, given the observed
TAI-RBR 2250 0.32 0.17 number of alleles at each locus. Rare alleles are disproportion-
ately likely to be lost during a bottleneck, whereas the virtual
NOTE. MRB, Maraba; PVL, Porto Velho; RBR, Rio Branco; SNV, Serra
´
do Navio; TAI, Tailandia. Spearman’s test (FST vs. geographical distance: Spear-
ˆ heterozygosity index, which relies mainly on common alleles,
man’s r p 0.430, P 1 .05, by Mantel test of matrix correlation; RST vs. geo- is less severely affected [29]. Thus, a postbottleneck population
graphical distance: Spearman’s r p Ϫ0.091, P 1 .05, by Mantel test).
that has been purged of rare alleles will tend to have a higher
virtual heterozygosity than the equilibrium expectation, given
(termed infection haplotypes), and, among the 148 isolates rep- the (reduced) number of alleles. The observed numbers of al-
resenting single-clone infections, there were 131 different hap- leles and virtual heterozygosity, for each locus, were determined
lotypes (strictly definable as being “true” haplotypes). Each of for each of the 5 populations here (from the full data set in
the haplotypes present in 11 isolate was seen in only 1 of the table 2). The theoretical expectations were derived from an IAM
populations, except for 3 haplotypes (39, seen in 2 isolates from applied to each of the 10 loci or an SMM applied to the 6 loci
Maraba and 1 isolate from Porto Velho; 67, seen in 2 isolates
´ that were previously considered to potentially conform to this
from Serra do Navio and 1 isolate from Porto Velho; and 89, mode of evolution [25].
seen in 5 isolates from Maraba and 1 isolate from Tailandia).
´ ˆ Table 5 shows the proportions of the loci for which the ob-
Five haplotypes were seen in isolates from different years. served virtual heterozygosity was greater than or less than the
Relationships among closely related haplotypes were analyzed theoretically expected value, for each of the 5 populations. By
by use of the eBurst algorithm, which was applied to infection use of the IAM, in all populations, the numbers of loci with
haplotypes here, to cover as much of the diversity as possible values that are higher or lower than expected are evenly split,
(figure 2). Haplotypes connected by single lines are identical and, thus, there is no significant departure from the expected
at 9 of the 10 loci studied, and these connected haplotypes values. By use of the SMM, 2 of the 5 populations (Maraba and ´
form groups that may be termed clonal complexes (incorpo- Serra do Navio) have a significant excess of loci, with values of
rating between 2 and 32 haplotypes that are putatively related virtual heterozygosity lower than expected from the numbers of
to 1 founding haplotype). The clonal complex of 17 haplotypes different alleles (this trend is opposite that expected from a bot-
located at the center of the figure includes isolates from Rio tleneck effect). Thus, on the basis of this test, there is no evidence
Branco only (none of these haplotypes are seen elsewhere), of recent severe genetic bottlenecks in any of the populations.
whereas the 2 large clonal complexes to the bottom left of the
figure each include isolates from 11 population. This suggests
that there is some connection between the different populations Table 5. Proportions of loci with excess or deficiency in virtual
heterozygosity, compared with expected values, at mutation-drift
but that the Rio Branco population might be the most distinct.
equilibrium (by use of the infinite alleles model [IAM] and step-
Genetic divergence among the different populations. Al- wise mutation model [SMM]), in 5 Brazilian populations.
lele frequencies show considerable variance among populations,
with 19% of the overall allelic variation being attributable to IAM (10 loci) SMM (6 loci)
differences among the 5 populations (mean FST p 0.19, for all Population Hexcess Hdeficiency P Hexcess Hdeficiency P
10 loci; range for individual loci, 0.10–0.33; all FST values were Porto Velho 7 3 .279 2 4 .213
significantly greater than zero). The interpopulation differen- Serra do Navio 6 4 .544 1 5 .051
tiation was very similar to the RST index for the 6 loci that were Maraba ´ 6 4 .473 0 6 .007
analyzed by use of the SMM (mean R ST p 0.19 , for the 6 loci; Tailandia
ˆ 6 4 .445 3 3 .502
range for individual loci, 0.09–0.31; all RST values were signif- Rio Branco 6 4 .475 2 4 .221
icantly greater than zero). Pairwise differentiation indices be- NOTE. Tests performed by use of the BOTTLENECK program [29].
P. falciparum Genetics in Brazilian Amazon • JID 2004:190 (1 November) • 1553

8. DISCUSSION is an ongoing reduced recombination rate in these populations
due to inbreeding, rather than recent epidemic expansion of
This is the first reported study of the genetic structure of P. particular clones. However, the existence of networks of related
falciparum involving analysis of microsatellite loci in isolates multilocus haplotypes indicated that some parasite isolates were
from multiple populations in a single country. Since the Bra- closely related, and these could be considered as belonging to
zilian Amazon is a huge and diverse region in which the epi- clonal complexes that have each emerged from a single ancestral
demiology of malaria is complex and geographically variable, haplotype. The strength of linkage disequilibrium within pop-
it was particularly relevant to examine the population genetic ulations was inversely related to the proportion of mixed-ge-
structure there. notype infections within the populations, as was expected.
There was a high degree of genetic divergence among pop- Geographical variation in the extent of parasite inbreeding
ulations, indicating as much divergence in allele frequencies may have consequences for the success of potential malaria-
among some of the Brazilian parasite populations as has been control strategies. The degree of inbreeding modifies the ef-
seen among populations in different countries in South Amer- fective recombination rate [31, 32] and may affect the rate of
ica and in different continents [11]. The pairwise FST genetic increase of drug resistance when 11 gene locus is involved [31,
distances between populations differed widely, but a matrix 33, 34]. There is widespread resistance to chloroquine and qui-
correlation analysis showed that these were not significantly nine [35, 36], Fansidar [4], and mefloquine [37] in different
associated with the geographical distances between the popu- parts of Brazil. Under some conditions [31], it is possible that
lations. Rio Branco was the most divergent from the other the higher degree of inbreeding in some populations (such as
populations, as indicated by higher FST values and by the ex- Rio Branco, Serra do Navio, and Tailandia) could allow the
ˆ
istence of a complex of related haplotypes that were restricted more-rapid increase of multilocus drug-resistant genotypes,
to this population. This highlights the fact that there are very compared with other populations (such as Porto Velho and
distinct geographical foci of P. falciparum endemicity in the Maraba). The heterogeneity in population structure, effective
´
Brazilian Amazon region and that the genetic relationship recombination rates within populations, and geographical pop-
among these is not determined by a simple model of isolation ulation subdivision are likely to affect the evolution of P. falci-
by distance. Frequencies within a population are quite stable, parum in different parts of the Brazilian Amazon region. This
as assessed by analysis of microsatellite genotypes obtained over is consistent with the view that P. falciparum malaria in the
a longer period of time (110 years) in Porto Velho [30]. In the largest endemic region of the Americas should not be seen as
present study, the population in Serra do Navio, from which a single entity, and different strategies for prevention and con-
samples were obtained from 1989 to 1992, was not an outlier, trol may be optimal for its diverse endemic locations.
compared with the other populations, from which samples were
obtained during the late 1990s.
Globally, linkage disequilibrium has been inversely associated Acknowledgments
with transmission intensity in P. falciparum, since the parasite We thank Jose Maria de Souza Nascimento, Zsuszanna Mikes, and Jennie
´
has a mixed mating system (in which fertilization may occur Lloyd, for laboratory assistance, and the staff of the Brazilian National
between parasite gametes of either same or different genotypes Health Foundation, for help in field work.
in mosquitoes that feed on blood containing multiple geno-
types). Higher levels of transmission increase the proportion
References
of infections that have mixed genotypes, so inbreeding (and
consequently linkage disequilibrium) predominates in low- 1. Akhavan D, Musgrove P, Abrantes A, d’A Gusmao R. Cost-effective
transmission areas, whereas higher levels of outbreeding occur malaria control in Brazil. Cost effectiveness of a malaria control pro-
gram in the Amazon Basin of Brazil, 1988–1996. Soc Sci Med 1999;
in high-transmission areas [11]. The present study has shown 49:1385–99.
that a spectrum of population structures exists within the Bra- 2. Roberts DR, Laughlin LL, Hsheih P, Legters LJ. DDT, global strategies,
zilian Amazon region. and a malaria control crisis in South America. Emerg Infect Dis 1997;3:
295–302.
Within each of the local parasite populations, significant de- 3. Camargo LMA, Noronha E, Villalobos Salcedo JM, et al. The epidemi-
viations from random associations among loci were observed ology of malaria in Rondonia (Western Amazon region, Brazil): study
when the data on all of the isolates were analyzed. To investigate of a riverine population. Acta Trop 1999; 72:1–11.
4. Cortese JF, Caraballo A, Contreras CE, Plowe CV. Origin and dissem-
whether populations had an epidemic structure, in the sense
ination of Plasmodium falciparum drug-resistance mutations in South
proposed by Maynard Smith et al. [27] and previously applied America. J Infect Dis 2002; 186:999–1006.
to P. falciparum population data by Anderson et al. [11], we 5. Wootton JC, Feng X, Ferdig MT, et al. Genetic diversity and chloroquine
then treated multiply represented genotypes as individuals, after selective sweeps in Plasmodium falciparum. Nature 2002; 418:320–3.
6. Roper C, Pearce R, Bredenkamp B, et al. Antifolate antimalarial re-
which significant linkage disequilibrium remained only in Rio sistance in southeast Africa: a population-based analysis. Lancet 2003;
Branco, Tailandia, and Serra do Navio. This indicated that there
ˆ 361:1174–81.
1554 • JID 2004:190 (1 November) • Machado et al.

9. 7. Nair S, Williams JT, Brockman A, et al. A selective sweep driven by 22. Weir BS, Cockerham CC. Estimating F statistics for the analysis of
pyrimethamine treatment in Southeast Asian malaria parasites. Mol population structure. Evolution 1984; 38:1358–70.
Biol Evol 2003; 20:1526–36. 23. Goudet J. FSTAT (version 1.2): a computer program to calculate F
8. Su X-Z, Wellems TE. Towards a high resolution Plasmodium falciparum statistics. J Hered 1995; 86:485–6.
linkage map: polymorphic markers from hundreds of simple sequence 24. Goodman SJ. RST Calc: a collection of computer programs for calculating
repeats. Genomics 1996; 33:430–44. estimates of genetic differentiation from microsatellite data and deter-
9. Su X-Z, Ferdig MT, Huang Y, et al. A genetic map and recombination mining their significance. Mol Ecol 1997; 6:881–5.
parameters of the human malaria parasite P. falciparum. Science 1999; 25. Anderson TJC, Su X-Z, Roddam A, Day KP. Complex mutations in a
286:1351–3. high proportion of microsatellite loci from the protozoan parasite Plas-
10. Su X-Z, Kirkman LA, Fujioka H, Wellems TE. Complex polymorphisms modium falciparum. Mol Ecol 2000; 9:1599–608.
in an 330 kDa protein are linked to chloroquine-resistant P. falciparum
26. Feil EJ, Li BC, Aanensen DM, Hanage WP, Spratt BG. eBurst: inferring
in Southeast Asia and Africa. Cell 1997; 91:593–603.
patterns of evolutionary descent among clusters of related bacterial
11. Anderson TJC, Haubold B, Williams JT, et al. Microsatellites reveal a
genotypes from multilocus sequence typing data. J Bacteriol 2004; 186:
spectrum of population structures in the malaria parasite Plasmodium
1518–30.
falciparum. Mol Biol Evol 2000; 17:1467–82.
27. Maynard Smith J, Smith NH, O’Rourke M, Spratt BG. How clonal are
12. Conway DJ, Machado RLD, Singh B, et al. Extreme geographical fix-
bacteria? Proc Natl Acad Sci USA 1993; 90:4384–8.
ation of variation in the Plasmodium falciparum gamete surface protein
gene Pfs48/45 compared with microsatellite loci. Mol Biochem Parasitol 28. Haubold B, Hudson RR. Lian 3.0: detecting linkage disequilibrium in
2001; 115:145–56. multilocus data. Bioinformatics 2000; 16:847–8.
13. Ariey F, Chalvet W, Hommel D, et al. Plasmodium falciparum parasites 29. Piry S, Luikart G, Cornuet JM. BOTTLENECK: a computer program
in French Guiana: limited genetic diversity and high selfing rate. Am for detecting recent reductions in the effective population size using
J Trop Med Hyg 1999; 61:978–85. allele frequency data. J Hered 1999; 90:502–3.
14. Haddad D, Snounou G, Mattei D, et al. Limited genetic diversity of 30. Tonon AP, Hoffmann EHE, da Silveira LA, et al. Plasmodium falciparum:
Plasmodium falciparum in field isolates from Honduras. Am J Trop sequence diversity and antibody recognition of the merozoite surface
Med Hyg 1999; 60:30–4. protein–2 (MSP-2) in Brazilian Amazonia. Exp Parasitol (in press).
15. Laserson KF, Petralanda I, Almera R, et al. Generic characterization of 31. Dye C, Williams BG. Multigenic drug resistance among inbred malaria
an epidemic of Plasmodium falciparum malaria among Yanomami Am- parasites. Proc R Soc Lond B Biol Sci 1997; 264:61–7.
erindians. J Infect Dis 1999; 180:2081–5. 32. Conway DJ, Roper C, Oduola AMJ, et al. High recombination rate in
16. Tami A, Grundmann H, Sutherland C, et al. Restricted genetic and an- natural populations of Plasmodium falciparum. Proc Natl Acad Sci USA
tigenic diversity of Plasmodium falciparum under mesoendemic trans- 1999; 96:4506–11.
mission in the Venezuelan Amazon. Parasitology 2002; 124:569–81. 33. Paul REL, Packer MJ, Walmsley M, et al. Mating patterns in malaria
17. Ferreira MU, Liu Q, Kaneko O, et al. Allelic diversity at the merozoite parasite populations of Papua New Guinea. Science 1995; 269:1709–11.
surface protein–1 locus of Plasmodium falciparum in clinical isolates 34. Curtis CF, Otoo LN. A simple model of the build up of resistance to
from the southwestern Brazilian Amazon. Am J Trop Med Hyg 1998;59: mixtures of anti-malarial drugs. Trans R Soc Trop Med Hyg 1986; 80:
474–80. 889–92.
18. Hoffman EHE, da Silveira LA, Tonhosolo R, et al. Geographical patterns
35. Couto AA, Calvosa VS, Santos MA, Souza JM. The evolution over
of allelic diversity in the Plasmodium falciparum malaria vaccine can-
time of the resistance of Plasmodium falciparum to antimalarial drugs
didate, merozoite surface protein–2. Ann Trop Med Parasitol 2001; 95:
in 2 areas of the Brazilian Amazonia with distinct socioeconomic and
117–32.
geographic characteristics. Rev Soc Bras Med Trop 1995; 28:357–65.
19. Creasey A, Fenton B, Walker A, et al. Genetic diversity of Plasmodium
falciparum shows geographical variation. Am J Trop Med Hyg 1990; 36. Zalis MG, Pang L, Silveira MS, Milhous WK, Wirth DF. Characteri-
42:403–13. zation of Plasmodium falciparum isolated from the Amazon region of
20. Anderson TJC, Su X-Z, Bockaire M, Lagog M, Day KP. Twelve mi- Brazil: evidence for quinine resistance. Am J Trop Med Hyg 1998; 58:
crosatellite markers for characterisation of Plasmodium falciparum from 630–6.
finger prick blood samples. Parasitology 1999; 119:113–25. 37. Calvosa VS, Adagu IS, Povoa MM. Plasmodium falciparum: emerging
21. Nei M. Molecular evolutionary genetics. New York: Columbia Uni- mefloquine resistance in vitro in Para State, north Brazil. Trans R Soc
versity Press, 1987. Trop Med Hyg 2001; 95:330–1.
P. falciparum Genetics in Brazilian Amazon • JID 2004:190 (1 November) • 1555