Gene Flow Among Anopheles Albimanus Populations in Central America, South America, and the Caribbean Assessed by Microsatellites and Mitochondrial Dna

GENE FLOW AMONG ANOPHELES ALBIMANUS POPULATIONS IN CENTRAL AMERICA, SOUTH AMERICA, AND THE CARIBBEAN ASSESSED BY MICROSATELLITES AND MITOCHONDRIAL DNA

ALVARO MOLINA-CRUZ, ANA MARI´AP.DEME´ RIDA, KATHERINE MILLS, FERNANDO RODRI´GUEZ, CAROLINA SCHOUA, MARI´A MARTA YURRITA, EDUVIGES MOLINA, MARGARITA PALMIERI, AND WILLIAM C. BLACK IV Universidad del Valle de Guatemala, Guatemala City, Guatemala; Department of Microbiology, Immunology and Pathology, Colorado State University, Fort Collins, Colorado

Abstract. Gene flow was examined among Anopheles albimanus populations from Cuba, Mexico, Guatemala, El Salvador, Nicaragua, Costa Rica, Panama, Colombia, and Venezuela by examining variation at four microsatellite (MS) loci and a mitochondrial DNA (mtDNA) marker. There was little variation among Central American populations and weak isolation by distance was only observed with the MS loci. There was moderate to large variation between Central and South American populations, suggesting a barrier to gene flow between Central and South America. However, Panamanian and Pacific Costa Rican populations differed with respect to western Central America, suggesting that there may be another barrier within Central America. There was small to moderate variation among Caribbean and continental populations. Phylogenetic and diversity analyses of mtDNA indicate that more ancestral and diverse haplotypes were present in the Caribbean population, suggesting that current continental An. albimanus populations may have originated from the Caribbean.

INTRODUCTION larger An. albimanus collections from throughout Central and South America and the Caribbean. Anopheles albimanus Wied. is the primary coastal vector of malaria from southern Mexico to northern Peru, and in the MATERIALS AND METHODS Greater Antilles.1 The species is found mainly at an altitude < 100 meters in a wide range of larval habitats that vary from Mosquito collections and extraction of DNA. The loca- hoof prints to lakes and brackish water.2 Anopheles albi- tions, collectors, and sample sizes of An. albimanus collec- manus populations vary considerably in their vector compe- tions in Central and South America and the Caribbean are tence for human malarias, their biting behavior, and host listed in Table 1, and the geographic locations of all sampling preference,2 even though the species is cytologically3 and sites are shown in Figure 1. Some of the collections from morphologically4 constant throughout its range. Knowledge Guatemala and one from Costa Rica were used in a previous of population structure can help make predictions of migra- study.8 Mosquitoes were collected in cattle corrals and kept tion among vector populations, give insights into the epide- alive for 24 hours in cardboard cartons with sugared water. miology and transmission of malaria, and help in the design of Afterwards, they were frozen and placed in 70% ethanol more effective vector control,5,6 including the possible release awaiting extraction of DNA.8 The DNA samples of An. bel- of genetically modified vectors.7 lator and An. cruzi individuals collected on the island of Trin- In a previous study,8 we examined the distribution of mi- idad and in Sao Paulo State, Brazil, respectively, were kindly tochondrial DNA haplotypes among An. albimanus collec- provided by Dr. Richard Wilkerson (Museum Support Cen- tions in Guatemala, to test for gene flow barriers using a ter, Smithsonian Institution, Suitland, MD). 390-basepair region of the mitochondrial NADH dehydroge- Microsatellite loci isolation, amplification, and identifica- nase subunit 5 (ND5) gene. Phylogenetic analysis among the tion. Anopheles albimanus genomic DNA was digested with 15 most common haplotypes did not detect clades associated Mbo I and subjected to electrophoresis on a 1.5% agarose gel. with geographic regions. Collections from different regions of DNA between 300 and 1,000 basepairs was separated and Guatemala were genetically similar, as were collections from purified from the gel. Digested DNA was ligated to Bam the same locations across three seasons. These results sug- HI-digested pBluescript plasmid (Stratagene, La Jolla, CA). gested that an earlier study of the An. albimanus ribosomal Recombinant colonies containing MS loci were identified by 9 32 DNA intergenic spacer (IGS) had overestimated genetic dif- hybridization with P-labeled (AG)20 and (AC)20 probes. ferences between Atlantic and Pacific populations, possibly Selected MS clones were sequenced and polymerase chain due to concerted evolution.8 Evidence from independent reaction (PCR) primers were designed and tested with An. nuclear markers was therefore required to support the results albimanus DNA from different geographic regions. The MS obtained with the ND5 marker. The earlier ND5 study8 also loci selected contained mostly AG repeats except for MS 1-90 suggested barriers to gene flow in Costa Rica and Panama and 6-41, which were composite microsatellites containing with respect to western Central America, and that ND5 hap- both AC and AG repeats. lotype frequencies in South America differed significantly Microsatellite locus 1-90 was amplified in single mosquitoes with respect to Central America. using primers 1-90+ (5Ј-GCA TAA ATA ATA GCC AA The present study expands on our previous findings on the CA-3Ј) and 1-90- (5Ј-GTC ACA CTT CCG ACT ACA AA- population structure and phylogenetic relationships among 3Ј). Microsatellite locus 2-14 was amplified with primers 2-14+ An. albimanus populations of Central and South America. In (5Ј-GCC CTT GCC AAG ATA AAA TGG AAA-3Ј) and this study, four microsatellite (MS) markers have been used, 2-14- (5Ј-TCA AAT AAT CCT AAA ACA CCG TCC-3Ј). in addition to the ND5 mitochondrial marker, to characterize Microsatellite locus 2-25 was amplified using primers 2-25+ 350 GENE FLOW AMONG AN. ALBIMANUS POPULATIONS 351

TABLE 1 Locations, collectors, regions, dates, and samples sizes of Anopheles albimanus collections*

Region Country (collector) Sub-region ND5 MS Collection site date (n for ND5, n for MS) n n Caribbean 146 170 Cuba 146 170 1) La Habana (I. Garcia) 5/19/99 146 170 Central America 1,575 1,110 Mexico, Chiapas 143 150 2) Zapata 12/1/98 47 50 3) Cossalapa 12/1/98 48 50 4) N. Independencia 12/19/98 48 50 Guatemala (N. Padilla, C. C. de Roslaes, P. Peralta, J. Garcia) 923 422 Northern Guatemala 298 115 5) Champona 3/6/95 (12, 0), 7/6/95 (32, 0), 9/5/95 (43, 0), 3/11/96 (0, 37) 8737 6) Nahua 3/9/95 (9, 0), 7/3/95 (43, 0), 9/6/95 (43, 0) 81 24 7) S. Luis Peten PE 7/10/95 (43, 0), 3/12/01 (0, 31) 43 31 8) S. Luis Peten BV 7/10/95 (43, 0), 9/4/95 (44, 0), 3/12/01 (0, 23) 87 23 Southern Guatemala 443 211 9) Cuto 2/6/95 (42, 0), 6/5/95 (26, 0), 4/5/95 (89, 0), 3/27/96 (0, 48) 147 48 10) Lauro 6/7/95 (4, 0), 4/4/95 (19, 0), 2/9/95 (31, 0), 3/27/96 (0, 14), 6/7/96 (0, 13) 54 27 11) Mango 3/15/95 (22, 0), 2/28/96 (33, 32), 3/7/96 (37, 0) 92 32 12) Ruperto 2/8/95 (16, 0), 6/7/95 (16, 0), 4/3/95 (9, 0), 10/4/95 (0, 48), 3/27/96 (0, 16) 41 64 13) Tallado 6/5/95 (12, 0), 2/7/95 (40, 40), 8/2/95 (38, 0), 6/6/95 (9, 0) 99 40 Eastern Guatemala 182 96 14) El Motor 3/20/96 45 48 15) Puente Blanco 7/11/95 (46, 0), 2/21/96 (46, 48), 9/11/95 (45, 0) 137 48 El Salvador (H. Francia) 125 149 16) San Alfredo 10/21/98 42 50 17) San Diego 10/21/98 39 50 18) Sta Lucia 10/21/98 44 39 Nicaragua (E. Lugo) 136 150 19) Corral 1 9/24/98 46 50 20) Corral 2 9/24/98 47 50 21) Corral 3 9/30/98 43 50 Costa Rica (T. Solano) 159 120 Atlantic Ocean Coast 95 120 22) Bananito 4/13/99 46 60 23) Batan 4/13/99 49 60 Pacific Ocean Coast 64 ND 24) Puntarenas 95 64 ND Panama (A. Ying) 89 119 25) Corral 1a 2/23/99 49 59 26) Corral 1b 2/23/99 40 60 South America 144 131 Colombia 47 49 27) El Carmen (V P. Howley) 91 47 49 Venezuela (Y. Rangel) 97 82 28) Corral Magdalena 8/99 30 50 29) Corral Puertas Negra 8/99 67 32 Total 1,865 1,411 * Samples sizes are indicated in parentheses when multiple collections were taken at a site. .not determined ס microsatellite; ND ס NADH dehydrogenase subunit 5, MS ס ND5

(5Ј-GGT TTC CAG CCT CCA TTC TC-3Ј) and 2-25- (5Ј- Mitochondrial gene amplification and haplotype identifica- CCT TAC TGT GCT GGA ACA CG-3Ј). Microsatellite lo- tion. The ND5 gene was amplified in individual mosquitoes cus 6-41 was amplified using primers 6-41+ (5Ј-CGG CAT using primers ND5P1 (5Ј-TWG CSC CTA ATC CKG CTA CCA TCC TTT CTC TG-3Ј) and 6-41- (5Ј-GAC CTC GCG TA-3Ј) and ND5M2 (5Ј-YTW GGA TGA GAT GGS TTA ,CorG ס purine, S ס pyrimidine, R ס CCT TGT CAT AA-3Ј). GG-3Ј), where Y -A or T. The amplified regions corre סGorT,andW ס Amplified MS alleles were size fractionated by electropho- K resis on denaturing DNA sequencing gels and visualized by spond with nucleotides 7,282-7,671 in An. quadrimaculatus silver staining.10 On each gel, the reciprocal of the length of (GenBank #L04272) and nucleotides 7,169-7,558 in An. gam- markers in a DNA ladder were regressed on the reciprocal of biae (GenBank # L20934). Amplification was done in an MJ their mobility.11 The mobility of An. albimanus MS alleles Research (Watertown, MA) thermocycler with the following was then entered into the regression equation to estimate size. conditions: 95°C for five minutes, min, 80°C on hold while In addition, the identity of the MS alleles was confirmed by one unit of Taq polymerase was added to each tube, then 10 sequencing, in triplicate, two of the most frequent alleles for cycles of 92°C for one minute, 48°C for one minute, and 72° each MS locus. Sequenced alleles served as references to es- for 1.5 minutes; this was followed by 32 cycles of 92°C for one timate the number of repeats in other alleles. minute, 54°C for 35 seconds, and 72°C for 1.5 minutes; a final 352 MOLINA-CRUZ AND OTHERS

Statistical analysis of haplotype and allele frequencies. Variation in mtDNA haplotype and MS allele frequencies was examined using the analysis of molecular variance (AMOVA) procedure on Arlequin version 1.1.12 The AMOVA was initially performed on collections from Central American countries and then among South American countries. The AMOVA next partitioned variation between Cen- tral and South America, and lastly between Cuba and continental populations. The significance of the variance compo- nents associated with each level of partitioning was tested using non-parametric permutation tests.12 Arlequin version

1.1 was also used to compute FST and RST, standardized mea- sures of variation in haplotype frequencies,13 among all collections and pairwise between all possible pairs of collection.

Effective migration rates (Nm) were estimated from FST or 14 FIGURE 1. Map of the Caribbean region showing the approximate RST. locations of Anopheles albimanus collections (see Table 1 for site and Pairwise FST values were transformed to FST/(1−FST) and country names); mountains higher than 300 meters are indicated in regressed on pairwise geographic distances among collections gray. to determine if geographic distance serves as a barrier to gene flow.14 Geographic distances were obtained by the GIS system using program ATLAS-GIS 3.0 (Environmental System extension was done at 72°C for seven minutes. The PCR Research Institute, Inc., Redlands, CA). This regression was products were analyzed by single-strand conformational poly- repeated using a natural logarithm of geographic distance.15 morphism (SSCP) analysis.10 The ND5 PCR products of the Transformations, regression analyses, and Mantel’s test16 26 most common ND5 haplotypes were sequenced along both were performed with Arlequin version 1.1.12 The reciprocal strands and 390 basepairs, primers excluded, were used in the of the estimated slope provides an estimate of the average 15 analysis. effective population size (Ne). Pairwise transformed FST

TABLE 2 Estimates of variability in the mitochondrial (ND5) haplotype sequences and the microsatellite (MS) markers*

ND5 MS 1-90 MS 2-14 MS 2-25 MS 6-41

No. of Nucleotide diversity Theta No. of No. of No. of No. of ␲ ␪ Sample Sample size† haplotypes† ( ) ( )Hobs Hexp alleles Hobs Hexp alleles Hobs Hexp alleles Hobs Hexp alleles La Habana 146 (145) 8 (7) 0.0159 ± 0.0084 6.19 0.78 0.82 12 0.81 0.84 15 0.69* 0.83 19 0.56 0.53 6 Zapata 47 (46) 10 (9) 0.0051 ± 0.0033 1.99 0.86 0.89 17 0.76 0.86 13 0.86 0.83 16 0.62‡ 0.74 7 Cosalapa 48 (47) 9 (8) 0.0034 ± 0.0024 1.31 0.80 0.86 13 0.76 0.86 15 0.90 0.85 16 0.78 0.74 9 Nva Ind 48 (47) 10 (9) 0.0025 ± 0.0019 0.99 0.78 0.87 15 0.78‡ 0.89 14 0.78 0.86 18 0.74 0.72 6 Champona 87 (80) 19 (14) 0.0058 ± 0.0036 2.27 0.84 0.89 11 0.81 0.85 11 0.81 0.83 10 0.59 0.62 4 Nahua´ 81 (77) 16 (12) 0.0036 ± 0.0025 1.39 0.67 0.81 8 0.63‡ 0.88 10 0.83 0.86 12 0.38‡ 0.75 6 SL Pete´n PE 43 (38) 13 (10) 0.0034 ± 0.0024 1.33 0.65‡ 0.85 9 0.73 0.76 11 0.74 0.81 12 0.71 0.62 5 SL Pete´n BV 87 (74) 18 (11) 0.0020 ± 0.0017 0.80 0.87 0.90 11 0.83 0.83 11 0.70 0.85 10 0.78‡ 0.63 4 Cuto 147 (152) 17 (13) 0.0027 ± 0.0020 1.07 0.85 0.85 11 0.85 0.86 12 0.75 0.83 14 0.67 0.66 7 Lauro 54 (51) 14 (12) 0.0041 ± 0.0027 1.59 0.78 0.87 10 1.00 0.90 14 0.78 0.84 12 0.70 0.67 7 El Mango 92 (89) 16 (13) 0.0033 ± 0.0023 1.29 0.72 0.81 9 0.81 0.85 10 0.88 0.87 15 0.69 0.76 6 Ruperto 41 (38) 12 (9) 0.0032 ± 0.0023 1.26 0.84 0.83 11 0.81 0.82 16 0.75 0.83 16 0.70 0.69 6 Tallado 99 (98) 13 (12) 0.0036 ± 0.0025 1.41 0.78 0.83 13 0.38‡ 0.86 12 0.82 0.83 12 0.58 0.63 6 El Motor 45 (43) 9 (7) 0.0020 ± 0.0016 0.78 0.79 0.86 10 0.77 0.89 13 0.77 0.80 13 0.69 0.69 8 Pte Blco 137 (134) 12 (9) 0.0031 ± 0.0022 1.22 0.75 0.84 12 0.79 0.87 14 0.79 0.84 15 0.60 0.60 5 Sn Alfredo 42 (40) 7 (6) 0.0013 ± 0.0012 0.49 0.92 0.90 16 0.84 0.86 16 0.84 0.84 16 0.72 0.70 6 Sn Diego 39 (39) 9 (9) 0.0028 ± 0.0021 1.09 0.71‡ 0.85 14 0.84 0.89 14 0.81‡ 0.85 16 0.65 0.77 7 Sta Lucia 44 (41) 11 (9) 0.0048 ± 0.0031 1.87 0.84 0.86 13 0.82 0.88 17 0.76 0.84 16 0.72 0.68 6 Corral 1 46 (43) 11 (9) 0.0038 ± 0.0026 1.47 0.84 0.89 15 0.80‡ 0.88 16 0.72 0.85 14 0.82 0.69 6 Corral 2 47 (47) 10 (10) 0.0035 ± 0.0024 1.35 0.72* 0.87 15 0.67‡ 0.87 18 0.80 0.85 15 0.56 0.66 5 Corral 3 43 (42) 8 (7) 0.0028 ± 0.0021 1.11 0.80 0.87 13 0.84 0.88 16 0.84 0.85 14 0.70 0.69 8 Bananito 46 (46) 9 (9) 0.0037 ± 0.0025 1.44 0.83 0.89 13 0.60 0.73 16 0.75 0.78 13 0.57 0.66 4 Batan 49 (48) 8 (7) 0.0085 ± 0.0049 3.33 0.62‡ 0.84 11 0.55‡ 0.61 10 0.57‡ 0.70 10 0.58 0.64 4 Puntarenas 64 (45) 12 (9) 0.0104 ± 0.0059 4.05 ND ND ND ND ND ND ND ND ND ND ND ND Corral 1A 49 (45) 8 (7) 0.0062 ± 0.0038 2.43 0.78 0.83 12 0.83 0.77 9 0.66 0.73 12 0.53 0.60 4 Corral 1B 40 (38) 8 (6) 0.0077 ± 0.0046 3.00 0.85 0.82 12 0.68 0.74 9 0.70 0.77 12 0.68‡ 0.64 6 El Carmen 47 (45) 9 (7) 0.0080 ± 0.0047 3.10 0.81 0.84 13 0.65 0.74 7 0.76 0.84 12 0.71 0.60 4 Magdalena 30 (14) 8 (4) 0.0047 ± 0.0032 1.84 0.70 0.67 6 0.61 0.61 4 0.66 0.63 7 0.09‡ 0.33 5 Puerta Negra 67 (33) 9 (5) 0.0051 ± 0.0033 1.99 0.70 0.69 8 0.73* 0.65 5 0.72 0.72 9 0.16‡ 0.24 6 Mean 64 (59) 11 (9) 0.0047 ± 0.0030 1.76 0.78 0.84 11.9 0.75 0.82 12.4 0.77 0.81 13.4 0.62 0.64 5.8 ס ס ס *Hobs observed heterozygosity; Hexp expected heterozygosity; ND not determined. † Total sample size (sample size sequenced for ND5) included in the analysis. ‡ Significantly different from Hexp (P < 0.05). GENE FLOW AMONG AN. ALBIMANUS POPULATIONS 353

FIGURE 2. Frequencies of microsatellite (MS) alleles by size and geographic region for loci 1-90 (A), 2-14 (B), 2-25 (C), and 6-41 (D). Alleles .basepairs ס with frequencies <0.01 are not shown. The repeat number (R) for sequenced alleles is indicated. bp values were entered into a distance matrix and used to con- RESULTS struct a dendrogram among all collections by cluster analysis using unweighted pair group method using averages Microsatellite allele frequencies among collections. The (UPGMA) analysis17 in the NEIGHBOR procedure in MS loci had an average of 10.9 alleles with a range of 4−19 PHYLIP3.5C.18 alleles per locus per collection. The average unbiased het- Phylogenetic and nucleotide diversity analysis of ND5 hap- erozygosity24 was 0.78 with a range of 0.33−0.90 (Table 2). lotype sequences. Haplotypes of the ND5 gene were manu- The number of individuals in which amplification failed (pre- ally aligned without gaps according to codon. Phylogenetic sumably due to mutations in the primer annealing sites) was relationships among haplotypes were estimated with small (17, 20, 12, and 2 of a total of 1,411 mosquitoes for MS PAUP4.0b10 using maximum parsimony, maximum likeli- 1-90, 2-14, 2-25, and 6-41, respectively). Of a total of 112 tests hood,19 and distance/neighbor joining.20,21 of goodness-of-fit to Hardy-Weinberg proportions (4 geno- For each collection, the nucleotide sequences and the fre- types × 28 collections), only 20 were not in equilibrium, with quency of each haplotype were entered into Arlequin version most of them due to heterozygote deficiency (Table 2). 1.1. This analysis could not be completed for all individuals The MS loci 1-90, MS 2-14, MS 2-25, and MS 6-41 loci, because we sequenced only the 26 most common haplotypes. respectively, contained 23, 25, 39, and 13 alleles. Most alleles For each collection, we estimated nucleotide diversity (␲), the obtained for each of the four MS loci differed from each other average number of nucleotide differences per site between by a multiple of two basepairs, but at least two low-frequency two sequences (equation 10.5),22 and the number of differ- alleles differed by one basepair. Sequencing the most fre- ences between two randomly chosen alleles, theta (␪) (equa- quent alleles at each MS locus confirmed the identity of the tion 1.4a).23 MS and the differences in repeat number (Figure 2). Allele 354 MOLINA-CRUZ AND OTHERS

frequency profiles at the four MS loci are shown in Figure 2. The FST/(1 − FST) was computed between all pairs of col- The countries in Central America west of Panama had simi- lections and the matrix of pairwise differences between col- lar MS allele frequencies, but differed from Panama (Figure lections was subject to cluster analysis (Figure 3A). All the 2B and C). Larger differences in allele frequencies were ob- Central American collections west of Costa Rica clustered served between Central and South America (Figure 2B, C, together but were separate from the cluster containing Costa and D). The Cuban population differed to the greatest extent Rican and Panamanian collections. The Venezuelan collec- from the other populations (Figure 2A, B, and C). tions clustered together and were the most distant relative to The MS allele frequencies were compared using AMOVA the Central American collections, followed by Cuba and Co- (Table 3). Similar results were obtained for the four MS loci lombia. with the infinite allele mutation model and the stepwise mu- Regression analysis of pairwise FST/(1 − FST) against geo- tation model. In all analyses, variation among individuals in graphic distance for Central American collections yielded a collections accounted for most (84−97%) of the total vari- small but significant correlation (Figure 4A), as did regression ance. Among Central American countries, AMOVA esti- analysis against the natural logarithm of the geographic dis- ∼ mated only 3% of the variance among collections from Chia- tance (Figure 4B). The FST did not increase until geographic pas (Mexico) to Panama (Table 3). More variation was de- distances approached ∼665 km (∼e6.5). The average effective ∼ tected among South American countries ( 6−8%) and among population size (Ne) was 96 individuals among collections collections within countries (∼1−4%) (Table 3). Between (Figure 4B). Central and South America, AMOVA estimated large vari- Mitochondrial DNA haplotype frequencies among collec- ance (∼8−11%) and ∼3% among collections within regions tions. Fifty different ND5 haplotypes were detected by SSCP (Table 3). These results suggest large genetic differences be- among 1,865 An. albimanus. There was an average of 11.1 tween Central and South American collections. The MS allele haplotypes detected per population with a range of 7−19 frequencies were lastly compared among Cuba and continen- (Table 2). Sequencing was done on 26 of the most common tal populations (Table 3). Significant variation occurred haplotypes. among regions (∼6−11%) and similar variation was detected The level of nucleotide diversity in the ND5 sequence, in- among collections (∼5%). This suggests significant genetic dif- cluding silent and non-silent sites, was moderate in most Cen- ferentiation between Cuban and continental collections. tral American populations (0.0013−0.0058), increased in

TABLE 3 Analysis of molecular variance of mitochondrial (ND5) and microsatellite markers (MS) among Anopheles albimanus collections*

ND5 MS (IAM)† MS (SMM)‡ Source of variation % variation % variation % variation Variation among countries in Central America Among countries 5.44 2.00 1.76 Among collections within countries 2.08 1.36 2.11 Within collections 92.47 96.64 96.14 Fixation indices FST FST RST (13.6 ס Nm) §0.018 (12.3 ס Nm) §0.020 (8.8 ס F (Countries) 0.054§ (Nm F (Collections in countries) 0.022§ 0.014§ 0.021§ F (All collections) 0.075§ 0.034§ 0.039§ Variation among countries in South America Among countries 24.88 5.74 8.22 Among collections within countries −1.83 3.71 1.40 Within collections 76.95 90.55 90.38 Fixation indices FST FST RST (3.0 ס Nm) 0.082 (4.1 ס Nm) 0.057 (1.5 ס F (Countries) 0.249 (Nm F (Collections in countries) −0.024§ 0.039§ 0.015 F (All collections) 0.231§ 0.094§ 0.096 Variation between Central and South America Between Central and South America 16.27 11.44 7.78 Among collections within regions 5.38 2.81 3.44 Within collections 78.35 85.75 88.78 Fixation indices FST FST RST (3.0 ס Nm) §0.078 (1.93 ס Nm) §0.114 (2.6 ס F (Regions) 0.163§ (Nm F (Collections in regions) 0.064§ 0.032§ 0.037§ F (All collections) 0.216§ 0.143§ 0.112§ Variation between Cuba and continental populations Between Cuba and continental populations 4.03 5.92 11.17 Among collections within regions 9.02 5.26 5.09 Within collections 86.96 88.82 83.74 Fixation indices FST FST RST (2.0 ס Nm) 0.112 (4.0 ס Nm) 0.059 (11.9 ס F (Regions) 0.040 (Nm F (Collections in regions) 0.094§ 0.056§ 0.057§ F (All collections) 0.130§ 0.112§ 0.163§ .effective migration rate סNm* † Infinite allele model. ‡ Stepwise mutation model. § Statistically significant by permutation test (1,023 permutations). GENE FLOW AMONG AN. ALBIMANUS POPULATIONS 355

FIGURE 3. Unweighted pair group method using averages cluster analysis of pairwise FST/(1-FST) relationships between collections for (A) microsatellite loci and (B) NADH dehydrogenase subunit 5 (ND5) mitochondrial DNA (mtDNA).

Panama and South America (0.0047−0.0080), and was great- (Table 3). The AMOVA indicated that ∼4% of the variation est in Cuba (0.0159) (Table 2). Of a total of 31 segregating arose between Cuba and continental populations (Table 3). sites, six resulted in amino acid replacements. Pairwise mtDNA FST/(1 − FST) were subject to cluster Haplotype frequency profiles of all 50 ND5 haplotypes are analysis (Figure 3B). As with the MS loci, all of the Central shown by country in Figure 5. Haplotype frequencies were American collections clustered together, except for the Costa similar for most countries in Central America (including Rican collection of Puntarenas and the Panamanian collec- Chiapas, Mexico) but differed substantially with those from tions. These collections separated from the Central American Cuba, the Pacific region of Costa Rica, Panama, Colombia, clusters as did the Cuban and South American collections, and Venezuela (Figure 5). Haplotype 1 was the most frequent suggesting major differences between western Central in both Cuba and Central America (Figure 5). America and Panama and South America.

There were eight haplotypes in Cuba and three were The FST/(1 − FST) estimates for Central American collec- unique to Cuba at a frequency Ն0.1; the other five were tions were regressed against geographic distances to test for shared with Central America and one was shared with Cen- isolation by distance. No correlation was found among col- ס Mantel probability ,0.01 ס tral and South America. There were 40 haplotypes present in lections west of Panama (R2 Central America, of which 15 were unique but in low fre- 0.207), but the inclusion of Panamanian collections in the quencies in Guatemala. Two unique haplotypes were present analysis resulted in increase in slope and an apparent corre- .(Figure 4C) (0.001 ס Mantel probability ,0.52 ס at low frequency in Panama, one in Mexico, and one in El lation (R2 Salvador. Central America shared seven haplotypes with South The mtDNA variation therefore indicates that the Central America. There were 14 haplotypes in South America, three American collections west of Panama seem to be panmictic. were unique to Colombia and two were unique to Venezuela. Pairwise MS FST/(1 − FST) and ND5 FST/(1 − FST) for all Haplotype frequencies were compared among Central collections were regressed and a significant correlation was -suggesting that MS and mtDNA mark ,(0.36 ס American countries using AMOVA (Table 3). As with the detected (R2 MS alleles, variation among individuals in collections ac- ers are giving similar patterns of variation. counted for most (77−92%) of the total variance. Approxi- Phylogenetic relationships among ND5 haplotypes. The mately 5% occurred among countries and ∼2% among col- ND5 sequences of An. albimanus were manually aligned with lections within countries. The variation detected among coun- the homologous regions of An. gambiae, An. quadrimacula- tries decreased to 0.9% when only those collections west of tus, An. bellator, and An. cruzi as outgroups. Phylogenetic Panama were considered. This pattern suggests a panmictic analysis identified a well-supported clade containing the three population from Chiapas (Mexico) to Costa Rica. Among unique Cuban haplotypes that was basal to a clade with countries in South America, 25% of the variation arose mainly South American haplotypes, and this in turn separated among countries (Table 3). A large amount (16%) of variance with slight bootstrap support from a clade formed by the rest occurred between Central and South American collections of haplotypes (Figure 6). 356 MOLINA-CRUZ AND OTHERS

America, both markers detected minor genetic differences between populations from Chiapas to Atlantic Costa Rica. There was no evidence of isolation by distance with the mtDNA markers, and MS markers indicated weak isolation by distance. The MS markers suggested that An. albimanus populations in Central America that are within ∼665 km of one another are panmictic. A significant isolation by distance was only detected with the mtDNA when including Panama in the regression analysis. However, the AMOVA suggests that it is not distance but rather some discrete barrier that causes the significant correlation. This also suggests that our previously inferred8 isolation by distance using the ND5 marker was in error, probably confounded by including genetic differences between Central and South America. That analysis was completed on all populations because of small sample sizes outside Central America. Within Guatemala, An. albimanus populations were genetically homogeneous between Atlantic and Pacific regions. This confirms our prior conclusion8 that the differences observed between these re- gions9 were probably due to concerted evolution of the IGS. Pacific and Atlantic populations exchange genes at a rate sufficient to homogenize the frequencies of microsatellite alleles and mitochondrial haplotypes, but not so rapidly as to overcome the homogenizing forces of molecular drive in local An. albimanus populations. Barriers to gene flow were evident between Central and South American An. albimanus populations. However, in the present study the mtDNA marker suggests that one barrier probably occurs within Central America, west into Panama and Costa Rica. This was not detected in our prior study because of sparse geographic sampling in Panama and South America. The variation between Atlantic and Pacific Costa Rican populations increases toward Panama. This genetic difference may have been caused in part by a population con- traction in Panama (e.g., due to more intense insecticide control). There were a smaller number of ND5 haplotypes and MS alleles in Panama compared with other Central American populations (Table 2). This barrier might be the mountain range that crosses Costa Rica and Western Panama, separat- ing the Atlantic from the Pacific regions (Figure 1). The mountains reach close to both coasts in some areas. Further sampling of An. albimanus in that area would be required to confirm this as a barrier. As in our previous study, An. albimanus populations in

FIGURE 4. Regression analysis of pairwise FST/(1 − FST) estimates South America were genetically heterogeneous. This genetic against pairwise geographic distance or against natural logarithms of differentiation is higher than the one previously detected us- geographic distances among collections in Central America using (A ing 25 allozymes in 11 populations of An. albimanus in Co- and B) microsatelllite (MS) loci or (C) NADH dehydrogenase sub- 25 unit 5 (ND5) mitochondrial DNA (mtDNA). Pairwise estimates be- lombia. In that case, northern populations of An. albimanus tween Panamanian collections and collections from other countries clustered separately from the southern populations but with (᭺) and between collections from Central American countries other Nei distances24 less than 0.05. The high level of population ᭹ from Panama ( ) are indicated. Regression analysis of ND5 pairwise structure found in South American An. albimanus also re- estimates excluding Panamanian collections (C-1) and including probability. quires further examination. Our study documents for the first ס .Panamanian collections (C-2) are indicated. Prob time small to moderate genetic differences between Caribbe- an and continental An. albimanus populations, suggesting DISCUSSION that the Caribbean ocean represents a partial barrier to gene flow. It is interesting to note that between Cuban and the Patterns of variation in both mtDNA and MS markers were continental collections, MS markers exhibited larger variance largely consistent with one another and with patterns ob- than the mtDNA marker, but MS markers detected less vari- tained with the mtDNA marker in our previous study.8 Al- ance than the mtDNA marker within Central America and though few microsatellite loci were analyzed, the consistency between Central and South America collections. The MS among themselves and with the mitochondrial marker sug- markers may exhibit less variance at larger geographic dis- gests that they represent genome wide effects. Within Central tances due to size homoplasy and size constraints.26−29 Nucle- GENE FLOW AMONG AN. ALBIMANUS POPULATIONS 357

FIGURE 5. Frequencies of the 50 NADH dehydrogenase subunit 5 (ND5) haplotypes in Anopheles albimanus populations grouped by country. otide diversity was greatest in the Cuban collections. Further- compared with An. gambiae (∼103).33 The small effective more, phylogenetic analysis of ND5 sequences indicated that population size of An. albimanus in Central America suggests three of the unique Cuban haplotypes were basal in An. al- a bottleneck in its natural history. This does not seem to be bimanus. Both observations are consistent with a hypothesis caused by seasonal fluctuations since no significant difference that continental An. albimanus populations originated in the in genetic composition was found between the rainy and the Caribbean islands. In addition, the basal Cuban clade was dry season.8 Although insecticide control in the region could more similar to the South American clade than to the larger have contributed to a bottleneck in the population, a founder clade containing predominantly haplotypes from Central effect of mosquitoes migrating from the Caribbean to Central America. This pattern is difficult to understand given the America seems more consistent with the data. present distribution of An. albimanus. The species is not Our results and the present distribution of An. albimanus present in the Lesser Antilles or in eastern Venezuela (Figure are consistent with the gene flow pattern shown in Figure 7. 7), regions that would be the closest link between the Greater The current An. albimanus populations originated in the Antilles and South America. Greater Antilles and moved across the Caribbean ocean to The level of genetic differentiation detected with MS loci Central and South America by different routes. The Carib- for An. albimanus among Central American and South bean ocean represents a partial barrier to gene flow; An. al- American populations is comparable with the one detected bimanus populations in the continent may have lower nucle- with MS in Africa among An. gambiae populations separated otide diversity due to genetic drift. Putative barriers to gene by the Rift Valley complex in Kenya30,31 and by 400−500 km flow located in Costa Rica and Panama decrease gene flow in western Africa.32 However, An. albimanus differs consid- among Central and South American populations. erably from An. gambiae in having a smaller effective popu- The gene flow pattern deduced for An. albimanus has sev- lation size (96 individuals) in Central America (Figure 4B) eral implications for vector control. The population west of 358 MOLINA-CRUZ AND OTHERS

Financial support: This project was supported by the UNDP/World Bank/World Health Organization Special Program for Research and Training in Tropical Diseases (TDR), grant no. 971171 to Ana María P. de Mérida, and training award no. M8/181/4/M.422 to Alvaro Mo- lina-Cruz. Authors’ addresses: Alvaro Molina-Cruz, Laboratory of Malaria and Vector Research, National Institute of Allergy and Infectious Dis- eases, National Institutes of Health, Twinbrook III, 12735 Twinbrook Parkway, Rockville, MD 20852, E-mails: [email protected] and [email protected]. Ana MaríaP.deMérida, Katherine Mills, Fernando Rodríguez, Carolina Schoua, María Marta Yurrita, Edu- viges Molina, and Margarita Palmieri, Medical Entomology Research and Training Unit, Universidad del Valle de Guatemala, 15 Avenida 11-95, Zona 15, VH III, Apartado Postal No. 82, 01901, Guatemala City, Guatemala, Telephone: 502-364-0336, Fax: 502-364-0052. Wil- liam C. Black IV, Department of Microbiology, Immunology and Pathology, Colorado State University, Fort Collins, CO 80523, Tele- phone: 970-4916136, Fax: 970-4911815, E-mail: wcb4@cvmbs. colostate.edu.

FIGURE 6. Maximum likelihood tree showing phylogenetic relationships among individual NADH dehydrogenase subunit 5 haplo- REFERENCES types. Bootstrap support using maximum parsimony analysis appears above each branch, while bootstrap support using Tamura-Nei ge- Rubio-Palis Y, Zimmerman RH, 1997. Ecoregional classification .1 ס netic distance/neighbor joining appears below each branch. An. of malaria vectors in the neotropics. J Med Entomol 34: 499– ס Guatemala; SA ס .Central America; Guat סAnopheles;CA .510 ס ס ס South America; Mex. Mexico; Pan. Panama; Colomb. Co- 2. Frederickson EC, 1993. Bionomics and Control of Anopheles al- lombia. bimanus. Washington, DC: Pan American Health Organiza- tion, Pan American Sanitary Bureau Regional Office of the World Health Organization. Costa Rica in Central America appears to be panmictic. Ge- 3. Keppler WJJ, Kitzmiller JB, Rabbani MG, 1973. The salivary gland chromosomes of Anopheles albimanus. Mosq News 33: netically modified mosquitoes released in these areas would 42–49. not experience barriers to gene flow. The inferred barriers to 4. Faran ME, 1980. Mosquito studies (Diptera, Culicidae) XXXIV. gene flow in Costa Rica and Panama might predict larger A revision of the Albimanus section of the subgenus Nyso- differences in vector capacity between An. albimanus popu- rhynchus of Anopheles. Contrib Am Entomol Inst Ann Arbor lations from Central, South America, and the Greater Anti- 15: 1–215. 5. Tabachnick WJ, Black WC IV, 1995. Making a case for molecular lles than among populations within each region. population genetic studies of arthropod vectors. Parasitol To- day 11: 27–30. Received September 5, 2003. Accepted for publication January 7, 6. Donnelly MJ, Simard F, Lehmann T, 2002. Evolutionary studies 2004. of malaria vectors. Trends Parasitol 18: 75–80. 7. Collins FH, Besansky NJ, 1994. Vector biology and the control of Acknowledgments: We thank all of our collaborators in Latin Ameri- malaria in Africa. Science 264: 1874−1875. can who provided mosquito collections. Dr. Mark Benedict (Centers 8. de Merida AM, Palmieri M, Yurrita M, Molina A, Molina E, for Disease Control and Prevention, Atlanta, GA) kindly provided us Black WC IV, 1999. Mitochondrial DNA variation among with primer sequences for MS 2-25. Dr. Richard Wilkerson kindly Anopheles albimanus populations. AmJTropMedHyg61: provided us with DNA samples of An. bellator and An. cruzi. 230–239. 9. de Merida AM, De Mata MP, Molina E, Porter CH, Black WC IV, 1995. Variation in ribosomal DNA intergenic spacers among populations of Anopheles albimanus in South and Cen- tral America. Am J Trop Med Hyg 53: 469–477. 10. Black WCI, DuTeau NM, 1997. RAPD-PCR and SSCP analysis for insect population genetic studies. Crampton JM, Beard CB, Louis C, eds. The Molecular Biology of Insect Disease Vectors: A Methods Manual. New York: Chapman and Hall, 361−373. 11. Schaffer HE, Sederoff RR, 1981. Improved estimation of DNA fragment lengths from agarose gels. Anal Biochem 115: 113– 122. 12. Excoffier L, Smouse PE, Quattro JM, 1992. Analysis of molecular variance inferred from metric distances among DNA haplotypes: application to human mitochondrial DNA restriction data. Genetics 131: 479–491. 13. Weir BS, 1996. Genetic Data Analysis II: Methods for Discrete Population Genetic Data. Sunderland, MA: Sinauer Associ- ates. 14. Slatkin M, 1993. Isolation by distance in equilibrium and non equilibrium populations. Evolution 47: 264–279. 15. Rousset F, 1997. Genetic differentiation and estimation of gene flow from F-statistics under isolation by distance. Genetics 145: 1219–1228. 16. Mantel N, 1967. The detection of disease clustering and a gener- FIGURE 7. Inferred gene flow (arrows) and its partial natural bar- alized regression approach. Cancer Res 27: 209–220. riers (blocks)forAnopheles albimanus populations in the Americas. 17. Sohkal RR, Sneath PHA, 1963. Principles of Numerical Tax- The current distribution of An. albimanus is shown in gray. onomy. San Francisco: W. H. Freeman and Co. GENE FLOW AMONG AN. ALBIMANUS POPULATIONS 359

18. Felsenstein J, 1993. PHYLIP: Phylogeny Inference Package. Ver- 27. Estoup A, Jarne P, Cornuet JM, 2002. Homoplasy and mutation sion 3.5C. Seattle, WA: University of Washington. model at microsatellite loci and their consequences for popu- 19. Hasegawa M, Kishino H, Yano T, 1985. Dating of the human-ape lation genetics analysis. Mol Ecol 11: 1591–1604. splitting by a molecular clock of mitochondrial DNA. J Mol 28. Donnelly MJ, Townson H, 2000. Evidence for extensive genetic Evol 22: 160–174. differentiation among populations of the malaria vector 20. Tamura K, Nei M, 1993. Estimation of the number of nucleotide Anopheles arabiensis in Eastern Africa. Insect Mol Biol 9: 357– substitutions in the control region of mitochondrial DNA in 367. humans and chimpanzees. Mol Biol Evol 10: 512–526. 29. Walton C, Thelwell NJ, Priestman A, Butlin RK, 1998. The use of 21. Saitou N, Nei M, 1987. The neighbor-joining method: a new microsatellites to study gene flow in natural populations of method for reconstructing phylogenetic trees. Mol Biol Evol 4: Anopheles malaria vectors in Africa: potential and pitfalls. 406–425. J Am Mosq Control Assoc 14: 266–272. 22. Nei M, 1987. Molecular Evolutionary Genetics. New York: Co- 30. Lehmann T, Licht M, Elissa N, Maega BT, Chimumbwa JM, lumbia University Press. Watsenga FT, Wondji CS, Simard F, Hawley WA, 2003. Popu- 23. Watterson GA, 1975. On the number of segregating sites in ge- lation Structure of Anopheles gambiae in Africa. J Hered 94: netical models without recombination. Theor Popul Biol 7: 133–147. 256–276. 31. Lehmann T, Hawley WA, Grebert H, Danga M, Atieli F, Collins 24. Nei M, 1978. Estimation of average heterozygosity and genetic FH, 1999. The Rift Valley complex as a barrier to gene flow for distance from a small number of individuals. Genetics 89: 583– Anopheles gambiae in Kenya. J Hered 90: 613–621. 590. 32. Carnahan J, Zheng L, Taylor CE, Toure YT, Norris DE, Dolo G, 25. Narang SK, Seawright JA, Suarez MF, 1991. Genetic structure Diuk-Wasser M, Lanzaro GC, 2002. Genetic differentiation of of natural populations of Anopheles albimanus in Colombia. Anopheles gambiae s.s. populations in Mali, west Africa, using JAmMosqControlAssoc7:437–445. microsatellite loci. J Hered 93: 249–253. 26. Nauta MJ, Weissing FJ, 1996. Constraints on allele size at mic- 33. Krzywinski J, Besansky NJ, 2003. Molecular systematics of rosatellite loci: implications for genetic differentiation. Genet- Anopheles: from subgenera to subpopulations. Annu Rev En- ics 143: 1021–1032. tomol 48: 111–139.