Adult An. albimanus specimens were collected from March 2005 to November 2007 in seven localities on the Caribbean and the Pacific coasts of Colombia (Figure 1, Table 1): Los Achiotes (ACH), Moñitos (MON), Santa Rosa de Lima (SRL), Nuquí (NUQ), Pizarro (PIZ), Buenaventura (BUE) and Tumaco (TUM). For details on sampling strategy and species identification see Gutiérrez et al 4. Additional specimens collected in 2007 (Gutierrez et al, unpublished data) from Turbo (TUR), Antioquia, were also included (Figure 1, Table 1). Human landing catches of adults were conducted under an informed consent agreement using a protocol and collection procedures that were reviewed and approved by a University of Antioquia (SIU-UdeA) Institutional Review Board. ACH, SRL and MON are in the Maracaibo biogeographic province, NUQ, PIZ, BUE, and TUM are in the Chocó biogeographic province and TUR is in the Magdalena biogeographic province 6. For COI sequencing samples from all sites were used, with individuals from Alto Guandipa in Mosquera (MTU) and La Ensenada in Santa Bárbara (STU), both sites in Tumaco, analysed as a single population (indicated as TUM; Figure 1A). For MS analysis, PIZ was excluded because there were not enough specimens with successful PCR amplification; STU and MTU from Tumaco were analysed separately because they are ~50 km apart (Figure 1B). Approximately 50% of the mosquitoes (randomly selected) included in this study were confirmed as An. albimanus using an ITS2-PCR-RFLP based assay 4041.

DNA was extracted from individual mosquito abdomens following a salt precipitation protocol 42. The DNA was then resuspended in 25 μL of TE buffer (10 mM Tris, 1.0 mM EDTA) and stored at -20°C. A 1,300-bp segment of mtDNA COI gene was amplified by PCR using primers UEA3 5'-TAT AGC ATT CCC ACG AAT AAA TAA-3' and UEA10 5'-TCC AAT GCA CTA ATC TGC CAT ATT A-3' 43. The PCR was performed in 25 μL containing 2 μL of DNA, 0.2 μM each primers, 1× reaction buffer, 1.5 mM MgCl₂, 0.1 mM each dNTP, and 1 U Taq polymerase (Bioline, London, UK). The cycling conditions were: 5 min denaturation at 94°C, followed by 36 cycles of 1 min denaturation at 94°C, 1 min annealing at 50.2°C, and 1 min 15 seconds extension at 72°C, ending with a final extension at 72°C for 7 min. The PCR products were purified using the Wizard SV Gel and PCR Clean-Up System (Promega, Madison, WI), following the protocol recommended by the manufacturer. Both strands of the purified PCR products were sequenced, with a region of ~600-bp overlap, for 220 individuals. Only DNA sequence segments with higher than 30 Phred values 4445 were used in the analyses. The forward and reverse chromatograms were manually corrected using the electropherogram viewer Chromas Lite^© 46. Sequences from each mosquito were assembled by pairwise alignment using BioEdit Sequence Alignment Editor 47, then multiple alignment was performed using ClustalX 48. Unique haplotypes were determined using DAMBE, version 4.5.68 49; identical sequences were considered to be a single haplotype.

The An. albimanus mosquitoes genotyped from each locality were selected randomly. Four di-nucleotide microsatellite (MS) loci described for An. albimanus by Molina-Cruz et al 29, were genotyped: 6-41, 1-90, 2-14 and 2-25. Each locus was amplified by PCR performed in a 25 μL volume containing 1 μL of DNA, 0.25 μM of each primer, 1× reaction buffer, 2.5 mM MgCl₂ (6-41, 1-90, 2-14) or 2.0 mM MgCl₂(2-25), 0.2 mM for each dNTP, and 1 U Taq polymerase (Bioline, London, UK). PCR additives were used as follows: 5% DMSO for reactions 1-90 and 2-25 and 0.5 μg/ml BSA for reaction 2-14. For all reactions, the cycling conditions were: 5 min denaturation at 95°C, followed by 35 cycles of 30 sec denaturation at 94°C, 20 sec annealing at 62°C (6-41), 50°C (1-90), 55°C (2-14), 58°C (2-25), and 30 sec extension at 72°C, ending with a final extension at 72°C for 10 min. Amplified fragments were separated by electrophoresis on DNA denaturing 6% polyacrylamide sequencing gels, and MS alleles were visualized by silver staining. To estimate allele sizes the length of bands was compared to a 10 bp DNA ladder (Invitrogen Inc., Carlsbad, CA, USA) over the migration/size of each MS allele using Quantity One^® software (Biorad Laboratories, Hercules, CA, USA). The most frequent homozygote MS alleles per locus were sequenced to provide reference sizes to estimate the number of repeats of other alleles. Allele sequences are available in GenBank [GenBank: FJ785408-FJ785419].

A 1,058-bp sequence of the An. albimanus COI gene corresponding to positions 1,802-2,859 of Anopheles gambiae s.s. mitochondrion complete genome and aligned with RefSeq NC 002084 84 was analysed for 220 An. albimanus mosquitoes from eight localities (Figure 1A, Table 2). All 112 haplotype sequences are available in GenBank under the accession numbers: FJ015158-FJ015269. Sequences contained no missing data such as ambiguous base pairs and alignment revealed 107 variable sites, 67 of which were parsimony informative. A total of 115 nucleotide substitutions (with a majority predicted to introduce synonymous amino acid changes), 101 transitions and 14 transversions were observed. The best-fit DNA substitution model selected by the hierarchical likelihood ratio test (hLRT) was TrN+I+G 85, with invariable sites (I = 0.7675) and Gamma distribution shape parameter (G: 0.9555). Using Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), the best-fit model was the Transitional model: TIM+I+G 86, with invariable sites (I = 0.7664) and Gamma distribution shape parameter (G: 0.8906), A-T composition of the COI sequences was 39.7% and 29.5%, respectively, similar to that found in other Anopheles species 23, and the above parameters were adjusted to construct the split networks from inferred distance matrices and the neighbour-joining analyses.

Although diversity values were low overall, Pacific region populations had higher nucleotide diversities (0.0098, range 0.0078-0.0152) than Caribbean populations (0.0035, range 0.0028-0.0038), and haplotype diversity values were similar. There were 112 unique haplotypes (Table 2), 16 shared between different localities (named in order of frequency using capital letters), with the remainder exclusive to a particular geographic site (Figure 2, Table 2). In concordance with the parsimony-based network, the distance-based haplotypes networks (data not shown) and neighbour-joining tree (Figure 3) illustrated mainly two distinctive groups, corresponding to samples from the Caribbean or the Pacific regions.

For MS analyses, 280 An. albimanus were genotyped and the four microsatellite loci were polymorphic in all collection sites studied (Figure 1B). The number of alleles per locus ranged from three to twelve: locus 6-41 showed the lowest value and 1-90 the highest (Table 3). Allelic richness (Rs) per locus ranged between three in MON and MTU (6-41) to eleven, also, in MTU (2-25). STU showed the lowest average allelic richness (6.125), while BUE showed the highest (8.144) compared to all other localities. The expected heterozygosity (He) across all samples ranged from 0.503 (ACH) to 0.87 (MTU), while the average expected heterozygosity ranged from 0.681 (STU) to 0.784 (TUR). Of 32 tests performed for HWE, 12 (37.5%) remained significant after the sequential Bonferroni correction (p < 0.01); all of them were associated with heterozygote deficits, in each locus as follows: 6-41 (STU), 1-90 (TUR and BUE), 2-14 (ACH, MON, NUQ and BUE) and 2-25 (ACH, SRL, TUR, NUQ and BUE). Of 48 tests conducted, no locus was at linkage disequilibrium after the sequential Bonferroni correction (p < 0.05) (Table 3). The frequency of null alleles at each locus was compared; however, there were no significant changes in comparison with the initial results. Under LD the average Ne calculated for all populations was 308 individuals (CI 215-493) and the average Ne for all Caribbean populations (including TUR) was an infinite number of individuals (CI 1,613- ∞), while for all Pacific populations it was 256 individuals (CI 136- ∞). SRL showed the lowest value (231 individuals), and TUR, NUQ, STU and MTU showed an infinite result (∞). Under the HE model all of populations showed infinite values (See Additional file 1: Estimates of effective population size (Ne) and heterozygosity tests based on MS data for An. albimanus).

In the AMOVA analyses using haplotype frequencies with the eight populations separated into two groups, including TUR in the Caribbean region, 5.85% (p < 0.05) of the total variance was explained at the among regions level. When populations were separated into two groups, including TUR in the Pacific region, only 3.86% (p < 0.05) of the total variance was due to the variation among regions (Table 4).

The F_ST value estimated from COI sequences for the comparison between the Caribbean and Pacific regions was 0.07 (p < 0.05), and most of the pairwise comparisons of F_ST between localities were significant; however, no significant F_ST values at p < 0.01 were observed in comparisons between SRL, MON and TUR from the Caribbean coast, or between NUQ, PIZ and BUE from the Pacific coast (Table 5). Estimates of pairwise genetic differentiation (F_ST) and gene flow (Nm) among populations using MS data are shown in Table 6. Overall significant F_ST values were observed between population belonging to the same region, Caribbean or Pacific. Of 28 differentiation tests, 18 were significant (p < 0.01). Comparisons between ACH and STU showed the highest degree of differentiation. An. albimanus samples collected from TUR, located in the middle of the collection range, showed lower differentiation with its nearest localities: NUQ (Pacific) and MON and SRL (Caribbean). Nm estimates among An. albimanus populations ranged from 4 to 59 individuals, presenting an average value of 21 individuals.

The SAMOVA was used to estimate heterogeneity among populations from the Caribbean and Pacific regions. This test maximized the proportion of total genetic variance between populations at six groups: 1) ACH, SRL and MON, 2) TUR, 3) NUQ, 4) BUE, 5) STU, 6) MTU (Figure 4A). In agreement with SAMOVA data, the UPGMA dendogram based on genetic distances showed two distinctive groups in which ACH, SRL and MON constituted a cluster (82% support) and TUR, NUQ, BUE, STU and MTU were included in a second one at 61% support (Figure 4B). Given the number of MS loci tested in this study and in the low support values for the UPGMA dendogram, the possibility that data on the apparent relationship between populations from the Caribbean and Pacific regions may be biased cannot be excluded. However, BAPS clustering was also congruent with the results obtained by SAMOVA and UPGMA analyses, which proposed two groups without admixture among them, except for TUR which showed signs of admixture with both. AMOVA was conducted with both the COI and MS data to test the Isthmus of Panama as a putative barrier between the Caribbean and Pacific regions, including and excluding TUR. The AMOVA corresponding to two groups: the Caribbean and the Pacific samples (excluding TUR), showed the highest percentage of the total variance and F_ST value, 4.3% (p < 0.05) and 0.06049 (p < 0.05), respectively (Table 4). Results of the assignment statistics showed that on average 28.6% (80 of 280) of the individuals were correctly assigned to their original reference site. Samples from STU and BUE presented the highest proportion of correctly assigned individuals (38%). In general, mis-assignments occurred between samples of the same region, either the Caribbean or the Pacific (Table 7).

A correlation test between genetic distance based on MS loci, measured by F_ST/(1- F_ST), and the geographic distance (Ln) was statistically significant, although with low values, for the whole dataset (Mantel tests: R² = 0.36, p = 0.003). The IBD model best explains genetic differentiation among populations from the Caribbean region: ACH, SRL and MON (excluding TUR), but it could not explain the low differentiation observed within the Pacific region populations (NUQ, BUE, STU and MTU). No statistically significant correlation was detected when tests of IBD were carried out separately for the Caribbean (including TUR) or the Pacific region (with or without TUR).

The most common haplotypes were A (n = 27), exclusively Caribbean, and B (n = 24), exclusively Pacific. Three localities (TUR, NUQ and BUE) share haplotypes H and M from both regions (Figure 2, Table 2). Haplotypes from the two regions differed by more than 13 mutational steps. Interestingly, the shared haplotypes (H, M) and the Pacific haplotypes J, P38, B357, B857 and N2865 clustered with the Caribbean region (Figure 2). Haplotypes A and B were the most common interior haplotypes and so are most likely to be ancestral 87. The majority of haplotypes in both regions were tip alleles, considered to be more recently derived and geographically restricted 7787. In addition, private haplotypes located peripherally are suggestive of a demographic expansion 7275 within the Caribbean region with subsequent limited gene flow between the two regions. In general, the parsimony-based network (Figure 2), distance-based haplotype networks and neighbour-joining tree (Figure 3) consistently show two distinctive groups, with a large star-shaped Caribbean network surrounding haplotype A, a smaller network around haplotype C, and a small Pacific network surrounding haplotype B. In the NJ tree, the Caribbean haplotypes were supported at 86% and the Pacific haplotypes at 98% (Figure 3), and the bootstrap support level was 99.9% for the branch that connects Caribbean and Pacific haplotypes in the An. albimanus neighbour-net network.

Both neutrality tests found the Caribbean populations have significant negative values, and one test was significant for the Pacific populations (see additional file 2: Results of neutrality tests based on COI sequences of An. albimanus from Colombia), which suggested a possible past demographic expansion event. The distribution of the number of differences (mismatches) between pairs of DNA sequences from Caribbean populations demonstrated the expected unimodal distribution for Caribbean all populations together and also for each population. The expected distribution did not differ significantly from the sudden-expansion model. The distribution for each locality from the Pacific region and for the grouped Pacific region showed a multimodal distribution typical of populations at equilibrium. Although the r (0.009; p = 0.91) and SDD (0.002; p = 0.75) values were not significant for Caribbean populations, the estimated values were small, also supporting a population expansion. As in previous Anopheles studies 2331, the Drosophila (D. melanogaster and D. yakuba) mutation rate values of 10^-8/site/year 88 and 10 generations/year 89 were assumed to estimate the time for the Caribbean population expansion. For An. albimanus, using τ = 4.654, this is approximately 21,994 years ago (95% CI, 8,969-34,347) during the late Pleistocene.

The heterozygote test performed under SMM, IAM and TPM using MS data showed different results. Using IAM, heterozygote excess was detected for all of the populations tested; however, none were statistically significant. Under SMM four populations showed heterozygote deficits: STU, BUE, NUQ and ACH; only STU was statistically significant. For TPM three populations presented heterozygote excess: MON, SRL and NUQ; only NUQ was statistically significant. STU showed heterozygote deficits that were not statistically significant.

Nucleotide diversity values for An. albimanus COI sequences, including synonymous and non-synonymous changes, were similar to those found for An. darlingi COI sequences from Central and South America 23, An. gambiae 84 and other species within the Insecta class. In particular, all are characterized by having an adenine-thymine rich mitochondrial genome. Anopheles albimanus COI sequences presented a combined frequency of ~70% AT. The results show that An. albimanus populations from the Caribbean and Pacific regions in Colombia have moderately high genetic diversity, in contrast to the lower diversity in Pacific localities detected using isozymes 30. This difference is likely a result of the greater sensitivity of DNA markers to detect genetic variability, since DNA polymorphic sites are not necessarily seen at the protein level 90. Nucleotide diversity was higher in most of the Pacific populations; this may indicate that different An. albimanus maternal lineages invaded these two Colombian regions at different times and (or) that the Pacific populations are older relative to the Caribbean ones, perhaps most closely related to populations from Cuba and Central America, which had similar values using ND5 sequence 29.

Panmictic populations of An. albimanus (within ~665 km) had been observed in Central America, Costa Rica and Panama, Colombia and Venezuela. It appears that An. albimanus populations have had different periods of isolation followed by secondary contact throughout the species range in America 272829. The data showed some haplotypes from Pacific populations clustering with the Caribbean group (Figure 2), possibly a genetic signature of a panmictic gene pool that existed before the late Pleistocene. Multiple factors may be responsible for the observed distribution and frequency of haplotypes from the Caribbean and Pacific regions in Colombia. Most haplotypes are not shared between regions, perhaps because of the distance between populations (>200 km), distinctive demographic history, human impact (insecticide use could have led to local haplotype extinctions) and/or distinctive ecological conditions in each region. Previous mtDNA analysis of An. albimanus did not detect a significant correlation between genetic and geographic distance 29. Although, genetic differences and geographical distances were not directly compared because the populations were not in MDE, distance did not appear to be a significant factor affecting the data.

The neutral model is rejected for the Caribbean populations, indicating possible previous population expansion and/or natural selection. In general, all tests illustrate two distinctive groups, corresponding to haplotypes from the Caribbean and from the Pacific regions. The data did not show significant differences between An. albimanus populations from TUR, located in the Magdalena province as described by Morrone 6, and the other populations from the Caribbean coast. The relatively strong mtDNA support for expansion in the Caribbean An. albimanus populations may indicate different demographic histories for this species in these two regions of Colombia, and the AMOVA support for variance resulting from these two groups is significant, albeit low. Population expansion for An. albimanus from the Caribbean coast in Colombia is estimated to the late Pleistocene, similar to An. darlingi 23.

Microsatellite loci used in this study have not been physically mapped in An. albimanus polytene chromosomes, consequently, their location on chromosome inversions is unknown and neutrality cannot be assumed 91. Chromosomal inversions have not been detected in An. albimanus and it is conspecific along its distribution 3092; however, the microsatellite loci analysed in this study were highly polymorphic (Table 3), showing their usefulness for evaluating the population structure of An. albimanus in Colombia.

Significant departures from HW equilibrium were detected associated with heterozygote deficits (Table 3), similar to that reported in other anopheline microsatellite studies 2529349394. Heterozygote deficits are attributed to either significant inbreeding, Wahlund effect, natural selection or the presence of null alleles 249596. Inbreeding as the possible cause of heterozygote deficits is not considered due to the fact that it affects all loci equally, which is not compatible with the heterogeneity detected in this study (Table 3). The Wahlund effect refers to reduction of heterozygosity in a population caused by subpopulation structure 97. In the data, SAMOVA, UPGMA and BAPS based on MS loci identified some degree of population subdivision, mainly among the Caribbean and the Pacific regions, and within the Pacific region, thus a part of the heterozygote deficits detected could be the result of Wahlund effect. High levels of heterozygote deficits could also be the result of null alleles, as a result of accumulation of mutations in the primer binding sites 34. In this study, there was failure of amplification of different loci for some specimens; however poor DNA quality was discarded as a cause, similarly, Molina-Cruz et al 29, observed similar cases of amplification failure. In summary the most likely causes of heterozygote deficits in the present study were null alleles and Wahlund effect.

The moderate level of population differentiation detected (F_ST) was not observed with isozyme markers 30. In theory, the analysis of a low number of loci could potentially increase population differentiation values (F_ST), but at the same time, the F_ST is the estimate presenting the lowest differentiation bias 98. The SAMOVA, UPGMA and BAPS detected two distinctive clusters: ACH, SRL, MON and TUR, NUQ, BUE, STU and MTU. Unlike the COI data, the TUR population clustered with the Pacific region populations. Low admixture was detected between the cluster represented by ACH, SRL and MON from the Caribbean region and the other five populations. Thus, the largest genetic differentiation was observed in comparisons among the Caribbean and the Pacific region, and SAMOVA analyses placed TUR, NUQ, BUE, STU and MTU in five different groups perhaps suggesting a specific barrier that reduces gene flow between these populations.

Data could suggest that features related to the three different biogeographic provinces, Maracaibo, Magdalena and Chocó 6, such as vegetation, weather, moisture, precipitation and temperature (Figure 1B-E) 5, constitute a semi-permeable barrier that reduces gene flow of An. albimanus populations at the inter-regional or the inter-biogeographic provinces level. Also the differences in N_e detected among the populations may have contributed to the genetic differentiation of these regions 99. Because some differentiation was observed between populations in the Chocó province (Pacific region) it is possible that other processes besides ecological conditions may affect free gene flow between An. albimanus populations.

The level of differentiation (F_ST), among the Caribbean and the Pacific region of Colombia is comparable to that reported among countries in South America and between Cuba and continental An. albimanus populations (F_ST = 0.057 and 0.059, respectively) 29. Similar levels of differentiation were also found among Anopheles nuneztovari s.l from Cordoba, Norte de Santander and Valle in the east, west and southeast of Colombia (F_ST = 0.024-0.06), using RAPDs (Randomly Amplified Polymorphic DNA) 100. In this study, the authors suggested IBD could explain their results. In a study of An. darlingi populations from Cordoba, Meta and Chocó in the northwest and west of Colombia, using AFLPs (Amplified Fragment Length Polymorphism) and RAPDs (RAPD F_ST = 0.084, AFLP F_ST = 0.229), results suggested that the observed genetic differences could be the result of the biogeographic characteristics of each particular region 101. The data showed that in addition to the differences in the demographic history of each region, the presence of semi-permeable biogeographic barriers, could contribute to the differentiation observed using MS.

In general, in the Caribbean region, there was low genetic differentiation, partly explained by the high Ne, that could have increased gene flow and decreased population genetic structure 24. IBD was the model that best explained differentiation among ACH, SRL and MON; however, if TUR was included in the analysis, the resulting data did not fit this model. TUR, the most genetically distant Caribbean region population, (Figure 4B), could be influenced by its location in Magdalena biogeographic province 6, and by differences in the effective population sizes. Similarly, low non-significant differentiation was observed among the Pacific region populations between NUQ-STU (362 Km), NUQ-MTU (371 Km) and BUE-MTU (217 Km) (Table 5) and the level of differentiation observed was not congruent with IBD.

A significant level of differentiation among the populations closer to BUE was detected, probably influenced by ecological characteristics 5 that could reduce gene flow with NUQ and STU (Figure 1B-E), in addition to their differences in effective population sizes. Ecologic and climatic variation with appropriate conditions for mosquito development have been recognized among the main causes for peaks of mosquito abundance and subsequent peaks of malaria cases 102. Therefore, further studies would be essential in BUE in comparison with others localities from the Pacific region. Even though there was restricted gene flow with respect to adjacent populations, BUE presented the highest genetic diversity, characterized by a high proportion of private alleles in low frequency. This could be explained by the characteristics of BUE, the main Colombian port city, where human migration and activities may promote gene flow among mosquito populations with different genetic pools, thereby increasing variability, as has been reported for An. gambiae 35103104. The levels of differentiation observed between MTU and STU could also be due to ecological differences 5; however, an overestimation of differentiation (F_ST), is not discarded given the low sample number 97105.

Most population structure analyses assume equilibrium, however, in several studies on anopheline vectors this has been violated, for example for An. gambiae and Anopheles arabiensis 106, Anopheles dirus 94, Anopheles moucheti 36 and An. darlingi 34. In this study, the heterozygosity test did not detect a significant departure from equilibrium for most of the populations. Nevertheless, under the TPM model, there was evidence of a bottleneck for NUQ which was not sustained by the SMM and IAM models and in disagreement with the effective population size observed, perhaps because of the mutational model and low number of analysed loci 83. Under the SMM model, the STU population appears to be expanding, consistent with the effective population size.

The Ne calculated under the HE model assumes equilibrium 107. Some of the Caribbean and Pacific populations were at H-W disequilibrium for some loci; it is possible that the departure from equilibrium influenced the values detected with HE, nevertheless, the confidence intervals obtained with LD agreed with those of HE. In general, high Ne values were obtained for all localities and on average, the effective population size (Ne = 308 individuals) was three fold higher than in the study with An. albimanus from Central and South America (Ne = 96 individuals) 29. For TUR and the Pacific region populations, except BUE, an infinite Ne was obtained under both models. According to Donnelly et al. 99 heterogeneity in the Ne could contribute to the levels of population genetic differentiation seen in this study. Similarly, Ne heterogeneity has been reported for An. darlingi 2434 and within the An. gambiae complex 108109110111. In An. gambiae, the Ne heterogeneity observed in the savana cytotype was demonstrated to be region specific 111. In the present study, high Ne might be associated with a wide availability of breeding places that are suitable for An. albimanus, such as lakes, ponds, mine excavations and fish ponds 3112. The high precipitation levels characteristic of the Pacific region 5, could contribute significantly to the availability of a variety of breeding sites for An. albimanus.