MALARIA accounts for million clinical cases 1989; Zheng et al. 1997; Romans et al. 1999). The first

Transcription

1 Copyright 2005 by the Genetics Society of America DOI: /genetics Multilevel Analyses of Genetic Differentiation in Anopheles gambiae s.s. Reveal Patterns of Gene Flow Important for Malaria-Fighting Mosquito Projects Frédéric Tripet*,1 Guimogo Dolo and Gregory C. Lanzaro* *Vector Genetics Lab, Department of Entomology, University of California, Davis, California and Malaria Research Training Center, Département d Entomologie, Ecole Nationale de Médecine et de Pharmacie, Bamako, B.P. 1805, Mali Manuscript received January 15, 2004 Accepted for publication July 9, 2004 ABSTRACT Malaria control projects based on the introduction and spread of transgenes into mosquito populations depend on the extent of isolation between those populations. On the basis of the distribution of paracentric inversions, Anopheles gambiae has been subdivided into five subspecific chromosomal forms. Estimating gene flow between and within these forms of An. gambiae presents a number of challenges. We compared patterns of genetic divergence (F ST ) between sympatric populations of the Bamako and Mopti forms at five sites. We used microsatellite loci within the j inversion on chromosome 2, which is fixed in the Bamako form but absent in the Mopti form, and microsatellites on chromosome 3, a region void of inversions. Estimates of genetic diversity and F ST s suggest genetic exchanges between forms for the third chromosome but little for the j inversion. These results suggest a role for the inversion in speciation. Extensive gene flow within forms among sites resulted in populations clustering according to form despite substantial gene flow between forms. These patterns underscore the low levels of current gene flow between chromosomal forms in this area of sympatry. Introducing refractoriness genes in areas of the genome void of inversions may facilitate their spread within forms but their passage between forms may prove more difficult than previously thought. MALARIA accounts for million clinical cases 1989; Zheng et al. 1997; Romans et al. 1999). The first and million deaths/year, 90% of which occur strains carrying genes for refractoriness may soon be in Africa (World Health Organization 1993; Mar- produced in the laboratory. Parallel efforts are being shall 2001). To date, none of the classical approaches made to develop drive mechanisms that would ensure to malaria control proved effective in Africa and its in- that refractory genes were transmitted at higher rates cidence is still increasing (Marshall 2001). Among other than those expected simply through Mendelien genetfactors, this is due to the breakdown of mosquito control ics. Transposable elements, viruses, and the intracellular programs, which in turn is due to a lack of funds, a ban symbiont Wolbachia are receiving increasing attention on DDT (Taverne 1999), and the evolution of mosquito as potential genetic drive systems (Kidwell and Ribeiro resistance to insecticides (Chandre et al. 1999; Coetzee 1992; Ribeiro and Kidwell 1994; Turelli and Hoffet al. 1999). A number of research groups around the mann 1999; Sinkins and O Neill 2000). world are now working toward the development of mo- A major concern regarding the introduction and lecular-level techniques for the genetic control of ma- spread of refractoriness genes is the possibility that laria vectors (Rai 1996; Collins et al. 2000). The prime they cannot be integrated into natural malaria vector target species are Anopheles gambiae, the main vector of populations because of barriers to gene flow (Lanzaro malaria in tropical Africa, as well as An. arabiensis and and Tripet 2003). There is ample evidence that the An. funestus (Cohuet et al. 2004, accompanying article genetic structure of An. gambiae populations, especially in this issue), two other important vectors. One of the in West Africa, is unusually complex (Touré et al. most appealing approaches would be to decrease ma- 1998a,b; Powell et al. 1999). Five chromosomal forms laria transmission by spreading genes for refractoriness that may be partially or totally reproductively isolated to Plasmodium into wild vector populations via the re- have been identified on the basis of the distribution of lease of a genetically modified (GM) strain (Carlson six paracentric inversions, namely the j, b, c, u, and d 1996; James et al. 1999). Recent progress includes trans- inversions on the right arm of chromosome 2 and the forming anopheline germlines (Catteruccia et al. a inversion on the left arm (Coluzzi and Sabatini 2000) and mapping refractoriness genes (Vernick et al. 1967; Bryan et al. 1982; Touré et al. 1983; Coluzzi et al. 1985). Similarly, in An. funestus, two chromosomal forms 1 have been described that may be reproductively isolated Corresponding author: Department of Entomology, University of California, 1 Shields Ave., Davis, CA in some parts of Africa but not in others (Cohuet et al. ftripet@ucdavis.edu 2004). Paracentric inversions act as crossover suppres- Genetics 169: ( January 2005)

2 314 F. Tripet, G. Dolo and G. C. Lanzaro Figure 1. Potential genetic structures in complex populations of An. gambiae. In A, local gene exchanges between forms is so high that populations from two forms from site 1 are more similar to each other than are populations of either form from a second site. In B, extensive gene flow between sites within forms overrides local genetic exchanges between forms. The two structures depict two extreme situations, and intermediate patterns (i.e., no clear clustering by site of form) could be expected, depending on the amount of gene flow between forms and within forms and on the geographical distance between the sites under study. sion mechanisms and thereby may link and protect coadapted gene complexes (Mayr 1963; Dobzhansky 1970). In An. gambiae they are thought to confer selective advantages under different environmental conditions on their carriers (Coluzzi 1982; Touré et al. 1994). In agreement with that hypothesis, the spatial distribution of inversion karyotypes has been shown to strongly correlate with specific ecological zones (Coluzzi et al. 1985; Touré et al. 1994). In areas where chromosomal forms co-occur, the relative frequencies of inversions also change seasonally, providing further evidence for their adaptive nature (Touré et al. 1994). No signs of genetic differentiation between chromosomal forms have been found outside inversion areas except in the intergenic spacer (IGS) and transcribed spacer of the ribosomal DNA locus on the X chromosome (della Torre et al. 2001; Gentile et al. 2001; Mukabayire et al. 2001). Microsatellite loci near the ribosomal locus were also found genetically differenti- ated (Wang et al. 2001). In Mali, the so-called M IGS type was found in near-perfect linkage disequilibrium with the inversion karyotypes characteristic of the Mopti form while the S type is common to the Savanna and Bamako forms. This difference suggests that the Mopti form may be completely reproductively isolated from the other two forms (Favia et al. 1997, 2001). In other parts of West Africa the distribution of M and S molecular types is not linked to chromosomal forms (Gentile et al. 2002) but is still thought to characterize partially or completely isolated populations (della Torre et al. 2002; Wondji et al. 2002). Direct estimates of hybridization rates seem to argue in favor of semipermeable boundaries between the putative M and S taxa (reviewed in Lanzaro and Tripet 2003). In Mali, analyses of the distribution of the M and S types in adult females and the sperm extracted from their spermathecae showed that, despite assortative mating, rare cross-matings do occur between the M and S types (Tripet et al. 2001, 2003). A few larvae and adults exhibiting hybrid-like IGS patterns have also been observed (della Torre et al. 2001; Taylor et al. 2001; Edillo et al. 2002; Diabate et al. 2003). Although strong selection is expected against hybrids, some level of introgression may occur and this could explain the lack of genetic differentiation between forms found in most parts of the genome (Lanzaro et al. 1998; Lanzaro and Tripet 2003). Complete reproductive isolation between forms of An. gambiae would have important consequences for the GM mosquito project since it would multiply the number of transgenic strains required for transforming malaria vector populations in a given area. However, if residual gene flow between forms occurs, one could hope that introduced refractoriness genes could be spread between forms using areas of sympatry as corridors for gene flow (Lanzaro and Tripet 2003). Estimating gene flow between forms and distinguishing contemporary from historical gene flow or genetic similarity due to common decent is a complex endeavor (Chambers and MacAvoy 2000; Balloux and Lugon-Moulin 2002). A number of studies based on genetic markers have assessed the degree of genetic differentiation between forms (reviewed in Powell et al. 1999; Black and Lanzaro 2001; Lanzaro and Tripet 2003). They usually failed to find significant differentiation between forms outside inversion areas (Cianchi et al. 1983; Lanzaro et al. 1998; Wang et al. 2001) or found significant but low levels of differentiation (Wondji et al. 2002). The question remains, however: How much of that genetic similarity reflects historical vs. contemporary gene flow? In addition, co-ancestry and allele-size homoplasy i.e., allelic identity by state, but not by descent may bias gene flow estimates (Estoup and Cornuet 1999; Balloux and Lugon-Moulin 2002). To assess the relative level of gene flow, we investigated patterns of genetic differentiation among An. gambiae populations in an area of Mali, West Africa, where several forms co-occur. We focused on patterns of genetic divergence (F ST ) between 10 Bamako and Mopti populations of An. gambiae from five sites as well as one population of An. arabiensis, a sibling species of An. gambiae. Analyses were conducted using allelic frequencies at five microsatellite loci inside the j inversion, which is fixed in the Bamako form but absent from the Mopti form. We assumed no recombination involving Bamako/ Mopti genes at loci within the j inversion. This is because

3 Gene Flow in An. gambiae 315 Figure 2. Geographical position of Mali (inset) and of the villages of Banambani, Kalabankoro, Moribabougou, N Gabacoro Droit, and Selenkenyi, where mosquitoes were collected. The dashed lines delimit four zones characterized by the presence of different chromosomal forms of An. gambiae. In zone S and M, the Savanna and Mopti forms are, each time, the only forms to be found. In contrast, in the M and S and in the M, S, and B zones, the Mopti and Savanna or the Mopti, Savanna, and Bamako forms co-occur at various densities; hence these zones could potentially act as corridors for the passage of transgenes between forms. heterokaryotypes between the two forms are extremely the discontinuous habitat characteristic of the upper tributar- rare in nature (Touré et al. 1998b) and recombination ies of the Niger River, an area that features rocky outcrops and valleys, and enjoy higher rainfalls than the rest of the in heterokaryotypes occurs only through equally rare country. An. gambiae populations in that area are characterdouble crossing over (Krimbas and Powell 1992). A ized by high proportions of Bamako and Mopti forms while second set of analyses used five microsatellite loci on the Savanna form occurs at low frequency in Kalabankoro and the third chromosome, which is void of chromosomal Selenkenyi (0 5%) but is common in Banambani (10 40%). inversions and where recombination is expected to octhe Niger River at the start of a vast plateau that extends to Both Moribabougou and N Gabacoro Droit are located near cur freely. Rather than conducting isolated pairwise the northeast. They exhibit mosquito populations dominated comparisons between forms, we used a hierarchical ap- by the Mopti and Bamako forms and a few Savanna individuals proach to describe patterns of genetic differentiation (0 5%). An. arabiensis occurs at low frequencies throughout between forms as well as within forms and among sites. the country (Touré et al. 1998b). This approach allowed us to discriminate between (A) Identification of karyotypes and molecular forms: Polytene chromosome preparations were made from the ovaries of sema putative genetic structure in which extensive genetic igravid females and DNA was extracted from their head and exchange between forms at a single site results in sym- thorax using established protocols (Coluzzi 1968; Hunt 1973; patric populations of different forms being more similar Post et al. 1993). The chromosomal forms of An. gambiae and than populations of a single form from different localight An. arabiensis were distinguished by scoring chromosomes by microscopy. We also determined the IGS rdna type of tions and (B) a genetic structure in which local gene all samples using the PCR diagnostic developed by Favia et al. flow between forms is overridden by gene flow within (1997). One individual karyotyped as a Bamako form but exforms across locations (Figure 1). The potential effects hibiting a Mopti-like M IGS type was excluded from the analyses. of co-ancestry and homoplasy on estimates of gene flow The final numbers of karyotyped females used for this were evaluated by estimating gene flow in two situations study are shown in Table 1. where little is expected: first, between An. gambiae and Estimates of genetic diversity and differentiation: We esti- mated the amount of genetic differentiation between all popu- An. arabiensis, which exhibit premating and postmating lations using 10 microsatellite loci of GT repeats that were reproductive barriers and for which hybridization rates isolated and mapped by Zheng et al. (1996) and used in previous are better known; and second, between the Bamako studies (Lanzaro et al. 1995, 1998; Tripet et al. 2001). and Mopti forms of An. gambiae, using the set of loci Five loci were on the third chromosome and five were distrib- protected from recombination within the j inversion. uted within the j inversion of chromosome 2 (Figure 3). The exact location of the loci was checked by blast searching the This study has important implications for our under- An. gambiae genome with the microsatellite-dna sequence standing of patterns of gene flow and speciation in this of each locus using the National Center for Biotechnology species complex and, importantly, potential patterns of Information search engine located at transgene movement among related taxa. nih.gov/. The loci were PCR amplified using fluorescent prim- ers and a PTC-200 thermal cycler (MJ Research, Watertown, MA). PCR products were mixed with a Genescan (Perkin-Elmer, MATERIALS AND METHODS Norwalk, CT) size standard and run on an ABI 3100 capillary sequencer (Perkin-Elmer). The gels were analyzed using the Study sites: Adult females were collected from huts during ABI PRISM Genescan analysis software and Genotyper DNA the rainy season in July 1993 in Banambani, August 1993 in fragment analysis software (Perkin-Elmer). Estimates of allelic Selenkenyi, August 1996 in Moribabougou and Kalabankoro, richness (number of alleles), gene diversity H S, heterozygosity and August 1999 in N Gabacoro Droit (Figure 2). The villages H O, and inbreeding coefficient F IS were calculated using the of Banambani, Kalabankoro, and Selenkenyi are located in software package FSTAT (Goudet 2001) available at

4 316 F. Tripet, G. Dolo and G. C. Lanzaro TABLE 1 Geographical coordinates of the five villages from which sympatric populations of An. arabiensis and the Bamako and Mopti forms of An. gambiae were analyzed Location Coordinates Species or form No. of individuals Banambani N, 8 3 W An. arabiensis 40 Bamako form 23 Mopti form 40 Kalabankoro N, 8 0 W Bamako form 11 Mopti form 19 Moribabougou N, 7 57 W Bamako form 27 Mopti form 21 N Gabacoro Droit N, 7 50 W Bamako form 21 Mopti form 23 Selenkenyi N, 8 17 W Bamako form 39 Mopti form 40 Sample sizes are expressed as number of individuals. www2.unil.ch/izea/softwares/fstat.html. The same package tially similar to other approaches based on analyses of gene was used to conduct an exact test of genetic differentiation frequencies (Weir and Cockerham 1984; Long 1986). The among all populations (Raymond and Rousset 1995; Goudet model of genetic structure tested was used to partition the et al. 1996). Arlequin version (Schneider et al. 2001; total variance into covariance components that were then used available at was used to calcu- to compute fixation indices (Wright 1965). The significance late F ST values between all pairs of populations. F ST measurements were translated into estimates of the number of mi- permutation approach (see Excoffier et al for details). level of fixation indices was then tested using a nonparametric grants per generation (Nm) using the standard formula F ST The AMOVA analyses were conducted by locus, an option 1/(4Nm 1) (Wright 1931). that allows correcting the degrees of freedom for each locus Cluster analyses and analyses of molecular variance: We according to their level of missing data, thus maximizing the used SYSTAT 5.2 (Wilkinson et al. 1992) to perform neighbor- power of the analyses (Schneider et al. 2001). joining cluster analyses on the matrices of pairwise F ST s generated by Arlequin. After the tree was generated, we tested the statistical significance of major branching between two or RESULTS more members of adjoining taxonomic units by a hierarchical analysis of molecular variance (AMOVA). We used the AMOVA Indices of genetic diversity in relation to forms and developed in Arlequin (Excoffier et al. 1992) that is essen- locus: Comparisons of the mean number of alleles, gene diversity H S, and observed heterozygosity H O for loci within the j inversion yielded significant differences between the Bamako and Mopti forms of An. gambiae. In each case, Bamako populations exhibited lower genetic diversity than Mopti (permutation tests: P 0.05 in all cases). Lower allelic numbers and gene diversities were observed across all five loci. Heterozygosity at locus 197 was the only exception and was slightly higher in the Bamako form ( as opposed to ). There was no significant difference between forms in mean inbreeding coefficient F IS. When loci on the third chromosome were considered, no significant differences could be found between Bamako and Mopti forms in terms of mean allelic richness, H S, H O,orF IS (P 0.05 in all cases). Genetic differentiation, divergence, and gene flow Figure 3. Position of the 10 microsatellite loci used to between species and forms: The exact test of genetic estimate the amount of genetic differentiation between An. differentiation (Goudet et al. 1996) yielded significant arabiensis and the Bamako and Mopti chromosomal forms of values between An. arabiensis and An. gambiae popula- An. gambiae. The loci were located either outside inversion tions for loci both inside the j inversion and on the areas on the third chromosome or inside the j inversion that third chromosome (Table 2, underlined data). Genetic is fixed in the Bamako form and does not occur in the Mopti form. The b inversion is fixed in neither form while the c and differentiation between populations within forms of u inversions are fixed in the Bamako chromosomal form but An. gambiae was mostly nonsignificant for loci both in- are found as floating inversions in the Mopti form. side and outside the inversion (Table 2, data without

5 Gene Flow in An. gambiae 317 TABLE 2 Matrices of pairwise estimates of genetic divergence (F ST ) and significance levels of exact tests of genetic differentiation between 10 populations of the Bamako and Mopti forms of An. gambiae and a population of An. arabiensis from Banambani, Kalabankoro, Moribaboug, N Gabacoro Droit, and Selenkenyi Arab Ban Bam Ban Bam Kal Bam Mori Bam Ngaba Bam Sel Mop Ban Mop Kal Mop Mori Mop Ngaba Mop Sel j inversion Arab Ban Bam Ban 0.111*** Bam Kal 0.132*** 0.047, NS Bam Mori 0.089*** 0.013, NS 0.031, NS Bam Ngaba 0.093*** 0.016, NS 0.029, NS 0.014, NS Bam Sel 0.093*** 0.001, NS 0.053, NS 0.016, NS 0.029, NS Mop Ban 0.112*** 0.115*** 0.081*** 0.082*** 0.077*** 0.114*** Mop Kal 0.110*** 0.134*** 0.098*** 0.082*** 0.081*** 0.124*** 0.033*** Mop Mori 0.108*** 0.092*** 0.061*** 0.046*** 0.053*** 0.091*** 0.021** 0.014, NS Mop Ngaba 0.079*** 0.094*** 0.066*** 0.054*** 0.056*** 0.083*** 0.018* 0.000, NS 0.000, NS Mop Sel 0.094*** 0.089*** 0.058*** 0.050*** 0.059*** 0.079*** 0.020, NS 0.016, NS 0.004, NS 0.001, NS Third chromosome Arab Ban Bam Ban 0.128*** Bam Kal 0.133*** 0.011, NS Bam Mori 0.111*** 0.004, NS 0.012, NS Bam Ngaba 0.119*** 0.009, NS 0.015, NS 0.009, NS Bam Sel 0.133*** 0.005, NS 0.015, NS 0.008, NS 0.004, NS Mop Ban 0.135*** 0.016* 0.028, NS 0.017*** 0.022*** 0.011* Mop Kal 0.142*** 0.014, NS 0.013, NS 0.027, NS 0.026* 0.006, NS 0.001, NS Mop Mori 0.159*** 0.013, NS 0.035, NS 0.027, NS 0.028* 0.019, NS 0.010, NS 0.025, NS Mop Ngaba 0.133*** 0.027*** 0.020, NS 0.011** 0.014* 0.014** 0.011*** 0.012* 0.006, NS Mop Sel 0.13*** 0.010, NS 0.023, NS 0.009** 0.018, NS 0.006, NS 0.002, NS 0.004, NS 0.003, NS 0.006, NS The top matrix was based on loci inside the j inversion and the bottom one on loci on the third chromosome. Arab, Arabiensis; Bam, Bamako; Mop, Mopti; Ban, Banambani; Kal, Kalabankoro; Mori, Moribabougou; Ngaba, N Gabacoro Droit; Sel, Selenkenyi. Significant F ST s are in boldface type. Between-species comparisons are underlined and between-forms comparisons are in italic type. Significance levels are *P 0.05; **P 0.01; ***P 0.001; NS, nonsignificant.

6 318 F. Tripet, G. Dolo and G. C. Lanzaro TABLE 3 Hierarchical AMOVA based on microsatellite allele frequencies for loci inside the j inversion in An. gambiae s.s. Sum of Variance Percentage Source of variation d.f. squares components variation P-value Among forms Among sites within forms Within populations Total Analyses were performed using data from the five sites (Banambani, Kalabankoro, Moribabougou, N Gabacoro Droit, and Selenkenyi) where the Mopti and Bamako chromosomal forms occurred in sympatry. Fixation indices were equal to (among forms), (among sites within forms), and (within populations). An average degree of freedom is presented for simplicity but the AMOVA was conducted by locus with locusspecific degrees of freedom to better account for missing data (see materials and methods for details). underlining or italics). In contrast, genetic differentia- tions. For loci inside the j inversion the analysis yielded tion between forms was significant for loci inside the a tree in which forms clustered together, forming two inversion but, for the most part, not significant for loci deeply rooted groups (Figure 4A). For loci on chromosome on the third chromosome (Table 2, italicized data). 3, forms were grouped likewise but the branching F ST estimates between An. arabiensis and An. gambiae between forms was a lot shallower (Figure 4B). For both were high and significant for loci both inside the j inversion sets of loci, the clusters by form differed significantly as (mean equal to SD) and on the shown by an analysis of molecular variance (see below). third chromosome (mean equal to ; Table Allele frequencies from the 10 sympatric Mopti and 2, underlined data). F ST s within forms of An. gambiae Bamako populations from the five villages were fitted were generally low for loci both inside ( ) to a model of AMOVA partitioning the variance into and outside the inversion ( ) and varied components of within site, between sites within form, in their levels of significance (Table 2, data with no and between forms. For data based on loci inside the underlining or italics). In contrast, F ST s between forms j inversion, the between-form component explained were high and significant for loci inside the inversion nearly 7% of the variance (Table 3). For loci on the ( ; Table 2, italicized data). For loci on the third chromosome, this component explained only 1% third chromosome, F ST s were much lower ( ) of the variation (Table 4). The effect of form was highly and many comparisons were not significant (Table 2, significant in both models. The within-site and betweensite italicized data). within-form components of the models explained We also examined F ST estimates between species and a significant amount of variance for both areas of the forms and within forms, locus by locus, to assess variation genome. Removing locus 788 from the analyses did not among loci. Within-form comparisons were charac- change those patterns although it decreased the amount terized by low F ST for all loci (mean ). The of variance explained by the between-site within-form exception was locus 788 in j inversion, which yielded component of the model presented in Table 3 (data high F ST s ( ) between Bamako populations not shown). but not between Mopti populations ( ). Be- Estimates of gene flow between species and forms: tween-form comparisons yielded higher F ST values for We translated F ST s presented in Table 2 to obtain esti- all loci inside the j inversion ( ) compared mates of the number of migrants per generation Nm with those on chromosome 3 ( ). The high between pairs of populations. Concordant with our tests amounts of genetic divergence observed for locus 788 of genetic differentiation and F ST estimates, migration within the Bamako form generated elevated F ST comparisons between populations within forms was found to be very between forms for this locus ( ). high for both areas of the genome. Comparisons be- Between-species comparisons were characterized by a tween the An. arabiensis population and the An. gambiae high amount of variation in F ST values with some loci population yielded Nm s significantly higher than zero exhibiting very high amounts of divergence and others for loci inside the j inversion (mean: SD) and very little (range ; mean ). on the third chromosome (mean: ; Table 5). These patterns were observed for loci both inside the Estimates of migration between forms differed acinversion and on the third chromosome. cording to the area of the genome considered. For loci Cluster analysis and hierarchical analysis of molecular in the inversion area, Nm s were low and the variance variance: We used the matrices of F ST s based on the set was low (mean: ). For loci on chromosome 3, of loci inside the j inversion and on chromosome 3 Nm s were considerably higher and more variable (Table 2) to build neighbor-joining trees of all popula- (mean: ). Estimates of migration within forms

7 Gene Flow in An. gambiae 319 Figure 4. Dendrograms performed on the matrix of F ST s presented in Table 2, using the neighbor-joining method of cluster analysis. In A, the tree is based on the microsatellite loci inside the j inversion, and in B, the tree is based on microsatellite loci on chromosome 3. Asterisks (*) indicate a significant statistical difference (P 0.001) between the clusters of Bamako and Mopti populations by an analysis of AMOVA. See text for further details. based on the third chromosome ranged from 9.8 to infinity (Table 5). Assuming that Nm s obtained for loci inside the j inversion reflect biases from co-ancestry or allele-size homoplasy (see discussion), we subtracted these from chromosome 3 estimates to obtain conservative estimates of gene flow between forms. The resulting Nm s between populations of the different forms were still considerable (mean: ). The same analyses were performed without locus 788 with little effect on the outcome (mean: ). DISCUSSION Adaptive inversion and speciation: The high degree of genetic differentiation and the high F ST estimates between forms found for loci inside the j inversion are consistent with the view that this inversion links and TABLE 4 Hierarchical AMOVA based on microsatellite allele frequencies for loci on the third chromosome in the Mopti and Bamako chromosomal forms of An. gambiae s.s. Sum of Variance Percentage Source of variation d.f. squares components variation P-value Among forms Among sites within forms Within populations 498 a Total Fixation indices were equal to (among forms), (among sites within forms), and (within populations). a An average degree of freedom is presented for simplicity (see materials and methods for details).

8 320 F. Tripet, G. Dolo and G. C. Lanzaro TABLE 5 Matrices of gene flow estimates (Nm) between 10 populations of the Bamako and Mopti forms of An. gambiae and a population of An. arabiensis from the villages of Banambani, Kalabankoro, Moribabougou, N Gabacoro Droit, and Selenkenyi Arab Ban Bam Ban Bam Kal Bam Mori Bam Ngaba Bam Sel Mop Ban Mop Kal Mop Mori Mop Ngaba Mop Sel j inversion Arab Ban Bam Ban 2.0 Bam Kal Bam Mori Bam Ngaba Bam Sel Mop Ban Mop Kal Mop Mori Mop Ngaba Mop Sel Third chromosome Arab Ban Bam Ban 1.7 Bam Kal Bam Mori 2.0 Inf 21.0 Bam Ngaba Inf Bam Sel 1.6 Inf Mop Ban Mop Kal Mop Mori Mop Ngaba Mop Sel The top matrix was based on loci inside the j inversion and the bottom one on loci on the third chromosome. Arab, Arabiensis; Bam, Bamako; Mop, Mopti; Ban, Banambani; Kal, Kalabankoro; Mori, Moribabougou; Ngaba, N Gabacoro Droit; Sel, Selenkenyi; Inf, infinity. Between-species comparisons are underlined and between-form comparisons are in italic. protects co-adapted gene complexes from recombina- in diverging species. In An. gambiae, heterokaryotypes tion (Dobzhansky and Pavlovsky 1958; Dobzhansky between the Bamako and Mopti forms produce repro- 1970; Coluzzi 1982). The potential role of paracentric ductively normal progeny of both sexes (Persiani et al. inversions for the evolution of species and forms in the 1986). Without a direct involvement of the paracentric An. gambiae complex has been discussed by Coluzzi j inversion in the speciation process through a decrease (1982) and Coluzzi et al. (1985, 2002). Inversions are in fertility of inversion heterozygotes, it seems more considered chromosomal mechanisms that preserve plausible that the inversion facilitated speciation through gene associations arising in populations that are temporarily its role as recombination suppressor. If, as suggested isolated in geographically or ecologically marginal here, the Bamako form of An. gambiae is undergoing or habitats (Coluzzi 1982; Coluzzi et al. 1985, 2002). If has undergone speciation through such a process, it in such a marginal population an adaptive inversion probably did so from a Savanna-like ancestor with whom reaches fixation upon secondary contact with another it shares the S rdna type. The significantly lower genetic population, selection against hybrids between the two diversity found in the j inversion of the Bamako form populations may occur. The genetic content of the inversion as compared to the same area in the Mopti form suggests itself remains largely protected because of the that indeed the Bamako form may have evolved follow- rarity of crossing over in that area of the genome (Krimbas ing the fixation of the j inversion in a small marginal and Powell 1992). These two effects combined founding population. Whether the j inversion reached may promote the rapid evolution of premating barriers fixation under such a process of speciation or another, to hybridization and thus may have a fundamental role its content underwent a strong genetic sweep and this in speciation processes within the species complex. is likely to be the main cause of decreased genetic diver- Rieseberg (2001) and Hey (2003) recently discussed sity in that area of the genome. This initial loss of diversity the potential role of inversions in suppressing recombination has apparently not yet been regenerated by mutation and providing linkage between genes causing nor has it been recovered by genetic exchange between assortative mating and those under disruptive selection forms. In contrast, we observed comparable estimates

9 Gene Flow in An. gambiae 321 of genetic diversities between the two forms on the third may spread within and between populations at various chromosome. In this area of the genome, recombination spatial scales and to identify biological and physical fea- can occur freely and selection against recombinants tures of the environment that may interfere with their may be less intense. The lack of significant genetic differentiation movements (Lanzaro and Tripet 2003). F ST estimates and values of F ST s for that area of the genome can potentially be biased by allelic identity by descent strongly support the idea that introgression between (co-ancestry) and allelic identity by state (homoplasy; forms is responsible for those patterns. Recent studies Estoup and Cornuet 1999; Chambers and MacAvoy of D. pseudoobscura and D. persimilis have reported comparable 2000; Balloux and Lugon-Moulin 2002). Consequently, patterns of genetic differentiation within and assessing potential biases can facilitate our interpreta- outside fixed inversions (Wang et al. 1997; Rieseberg tions of F ST estimates and improve our understanding et al. 1999; Machado et al. 2002). Noor et al. (2001) of An. gambiae s population structure. Taylor et al. also found genes contributing to hybrid male sterility (2001), for example, compared estimates of gene flow and female mate preference in D. pseudoobscura and within the Mopti, Bamako, and Savanna forms derived D. persimilis that were associated with fixed inversions from F ST s with those measured by mark-release-recapdiffering between the two species. Taken together, these ture experiments (MRR) between two villages. F ST s results support the view that chromosomal inversions yielded an estimated Nm SD and were may promote speciation through their role in recom- markedly consistent with the Nm obtained bination and patterns of gene flow (Rieseberg 2001; by MRR (Taylor et al. 2001). In this study we conducted Hey 2003). two types of comparisons based on microsatellite loci Contrary to the Bamako form, the Mopti form could that could give us a better idea of the effects of conot have diverged from the Savanna-Bamako group un- ancestry and homoplasy on indirect estimates of gene der a simple scenario of speciation because it does not flow. F ST s between An. gambiae and its sister species, feature any characteristic fixed inversion (Touré et al. An. arabiensis, were used to estimate a migration rate, 1983; Coluzzi et al. 1985). As is the case among members Nm, of 2 migrants per generation. This estimate can of Drosophila species groups, models of speciation be compared to direct measures of hybridization rate involving inversion fixation can reasonably be assumed made by extensive cytogenetic studies, which have shown in certain cases e.g., D. pseudoobscura and D. persimilis the frequency of hybridization between the two sister (King 1993; Noor et al. 2001) but in pairs that do not species to be close to 0.05% (White 1970; Petrarca feature fixed characteristic inversions e.g., D. pseudo- et al. 1991; Touré et al. 1998b). Effective population obscura and D. pseudoobscura bogotana other processes sizes in a village such as Banambani ranged from 2000 of speciation must be involved (Powell 1997; Wang to 6000 (Taylor et al. 1993; Touré et al. 1998a). The et al. 1997). At present it is unknown how much gene product of the hybridization rate and effective popula- flow occurs through direct hybridization between the tion size gives an estimate of one to three hybrids, equivalent Bamako and Mopti forms or if introgression between to migrants produced per generation. the two occurs through the Savanna form. Cytogenetic These rough estimations suggest that neither co-ancestry studies seemed to support the latter by showing frequent nor homoplasy substantially biased Nm s between the hybrid-like heterokaryotypes between the Savanna and two species. It should be noted, however, that the high Bamako forms and the Savanna and Mopti forms but variance in F ST s observed between loci tends to indicate not between the Bamako and Mopti forms (Touré et al. that some loci may indeed be prone to homoplasy and 1994, 1998b). However, the rdna diagnostic applied that overall this may contribute to a slight underestimato these heterokaryotypes did not yield hybrid patterns, tion of F ST s and an overestimation of Nm s. The second thereby suggesting that these heterokaryotypes may be type of comparisons we used to assess potential biases the result of floating inversions atypical of each form were between-form F ST s based on loci inside the j inversion. occurring at low frequencies rather than the result of These yielded an average migration rate of three hybrids (Favia et al. 1997). Whether these atypical inver- individuals per generation. Assuming, conservatively, that sions are traces of rare introgression events or independent these three migrants solely reflect biases due to co-ances- events of chromosomal rearrangements remains try and/or homoplasy, we can conclude that these con- to be investigated (Lanzaro and Tripet 2003). founding factors do not seriously bias gene flow estimates Disentangling past and current gene flow: Estimating between the forms of An. gambiae based on microsatellites. current gene flow within and among forms of An. gambiae Differentiating ongoing gene flow from gene flow that is of critical importance in the context of the GM occurred in the recent past has been and remains the mosquito project. Population genetic studies are required main problem associated with the use of microsatellite for identifying discrete population groups across markers in the An. gambiae complex. Recent methods Africa, for determining their geographical distribution, based on coalescence theory and maximum likelihoods, and for evaluating the degree to which they are repro- such as those implemented in Migrate 1.5 (Beerli 2002) ductively isolated (Lanzaro and Tripet 2003). They and the combined haplotypes/short-tandem-repeat analy- are also necessary to estimate the rate at which genes ses recently proposed by Hey et al. (2004), offer the

10 322 F. Tripet, G. Dolo and G. C. Lanzaro possibility of estimating population sizes at divergence Consequences for malaria control programs: The patterns as well as gene flow since divergence. These methods of gene flow reported here have important implias can be useful for detecting dispersal between distant cations for the mosquito genetic control project. They populations but their assumptions of constant population stress the fact that gene inserts should be targeted to sizes and divergence followed by constant gene flow areas of the genome located outside inversions if they (Beerli and Felsenstein 1999; Hey et al. 2004) limit are to be spread efficiently. Although we focused on the their use for estimating and interpreting gene flow be- extreme case of one form possessing a fixed inversion tween forms of An. gambiae. Thus F-statistics currently and another not, inversions, fixed or floating, will inter- remain the best tool for studying population structure fere with the spread of transgenes between populations in An. gambiae despite the difficulties in interpreting both within and between forms. This is because recombi- pairwise F ST s between populations of different forms nation rates can be expected to be just as rare in withinforms (Lanzaro and Tripet 2003). Typically, the low levels as in between-form heterokaryotypes. Chromo- of genetic divergence between forms characterizing ar- some 3, which enjoys a higher recombination rate than eas of the genome away from inversions have been interpreted the X chromosome and the inversion-rich chromo- as evidence either for incomplete reproductive some 2, would be the logical location for gene inserts. isolation or for complete reproductive isolation, de- Our results also suggest that multiple transgenic strains pending on the authors (e.g., Lanzaro et al. 1998; Won- will be needed for effective manipulation of malaria dji et al. 2002; Lanzaro and Tripet 2003). Here, we vector populations. Our study clearly shows that local attempted to compensate for the low interpretative residual gene flow between forms is almost negligible power of isolated between-form comparisons by examin- compared to the extensive gene flow within forms. ing a large number of populations at a wider spatial Hence, if refractory mosquitoes of one form are released scale (125 km). If current gene flow is important, one in a given location, the genetic construct will spread over could expect that two sympatric populations of the Ba- a considerable geographical range before potentially mako and Mopti forms could locally be more similar passing to another form. As a result, malaria incidence than two distant populations of the same form. Thus is not likely to decrease significantly in the area of the one could expect a genetic structure intermediate be- release unless all species and forms of vectorial importween the structures A and B presented in Figure 1. In tance are targeted simultaneously. Furthermore, betheir early study based on 21 microsatellite loci spanning cause mosquitoes expressing refractory genes and carthe entire genome and Bamako and Mopti populations rying the genetic-drive system are likely to exhibit some from the villages of Banambani and Selenkenyi, Lan- fitness costs (Yan et al. 1997; Catteruccia et al. 2003), zaro et al. (1998) presented data for the third and the simply transforming the most abundant form would X chromosome suggesting that this might be the case. leave the opportunity for other mosquito vectors of ma- If current gene flow is low or absent, one can expect laria to expand their ecological niche and out-compete within-form gene flow to quickly override these patterns the refractory population (Curtis et al. 1999). and populations to cluster according to their forms We thank G. Caccone, J. Powell, C. Taylor, W. J. Black, D. El Naiem, (structure B in Figure 1). Our results show that current and M. Slotman for their comments on earlier versions of the manugene flow between forms is not high enough to counter- script. This work was supported by fellowship no. 823A from act the homogenizing effects of migration between popfrom the Swiss National Science Foundation to F.T. and grant no. AI40306 ulations within forms. This is true despite significant the National Institutes of Health to G.C.L. genetic differentiation between some populations within forms. Thus the question remains as to how much of the theoretical 13 migrants per generation estimated LITERATURE CITED here is due to the few hybrids reported between the M Balloux, F. and N. Lugon-Moulin, 2002 The estimation of populaand S forms of An. gambiae and how much is accounted tion differentiation with microsatellite markers. Mol. Ecol. 11: for by past gene flow. Lehmann et al. (2003) recently Beerli, P., 2002 MIGRATE: documentation and program, part of presented data showing that four distant M-form popu- LAMARC, Version 1.5. Distributed over the Internet (http: / lations from Senegal, Ghana, Cameroon, and the Demo- evolution.genetics.washington.edu/lamarc.html). cratic Republic of the Congo in West and Central Africa Beerli, P., and J. Felsenstein, 1999 Maximum-likelihood estima- tion of migration rates and effective populations using a coaleswere more similar to each other than to any S-form cent approach. Genetics 152: populations. Although the study included only one sym- Black, W. C., and G. C. Lanzaro, 2001 Distribution of genetic patric S-form population matching the M populations variation among chromosomal forms of Anophele gambiae s.s.: from the Democratic Republic of the Congo, these two Introgressive hybridization, adaptive inversions, or recent repro- ductive isolation? Insect Mol. Biol. 10: 3 7. populations did not cluster together. These results are Bryan, J. H., M. A. Di Deco, V. Petrarca and M. Coluzzi, 1982 consistent with our interpretation of low residual gene Inversion polymorphism and incipient speciation in Anopheles flow between forms and suggest that the patterns we gambiae s. str. in the Gambiae, West Africa. Genetica 59: Carlson, J. O., 1996 Genetic manipulation of mosquitoes: an apfound in the area of M/S sympatry in Mali could extend proach to controlling disease. Trends Biotechnol. 14: to larger spatial scales. Catteruccia, F., T. Nolan, T. G. Loukeris, C. Blass, C. Savakis

11 Gene Flow in An. gambiae 323 et al., 2000 Stable germline transformation of the malaria mos- applications to human mitochondrial DNA restriction data. Genetics quito Anopheles stephensi. Nature 405: : Catteruccia, F., H. C. Godfray and A. Crisanti, 2003 Impact Favia, G.,A.della Torre, M.Bagayoko, A.Lanfrancotti, N. of genetic manipulation on the fitness of Anopheles stephensi Sagnon et al., 1997 Molecular identifications of sympatric chro- mosquitoes. Science 299: mosomal forms of Anopheles gambiae and further evidence of their Chambers, G. K., and E. S. MacAvoy, 2000 Microsatellites: consen- reproductive isolation. Insect Mol. Biol. 6: sus and controversy. Comp. Biochem. Physiol. B Biochem. Mol. Favia, G., A. Lanfrancotti, L. Spanos, I. Sidén-Kiamos and C. Louis, Biol. 126: Molecular characterization of ribosomal DNA (rdna) poly- Chandre, F., S. Manguin, C. Brengues, J. D. Yovo, F. Darriet et al., morphisms discriminating among chromosomal forms of Ano Current distribution of a pyrethroid resistance gene (kdr) pheles gambiae s.s. Insect Mol. Biol. 10: in Anopheles gambiae complex from West Africa and further evi- Gentile, G., M. Slotman, V. Ketmaier, J. R. Powell and A. Caccone, dence for reproductive isolation of the Mopti form. Parassitologia 2001 Attempts to molecularly distinguish cryptic taxa in Ano- 41: pheles gambiae s.s. Insect Mol. Biol. 10: Cianchi, R., F. Villani, Y. T. Touré, V. Petrarca and L. Bullini, Gentile, G., A. della Torre, B. Maegga, J. R. Powell and A. Cac Electrophoretic study of different chromosomal forms of cone, 2002 Genetic differentiation in the African malaria vec- Anopheles gambiae s.s. Parassitologia 25: tor, Anopheles gambiae s.s., and the problem of taxonomic status. Coetzee, M., D. W. K. Horne, B. D. Brooke and R. H. Hunt, 1999 Genetics 161: DDT, dieldrin and pyrethroid insecticide resistance in African Goudet, J., 2001 FSTAT: a program to estimate and test gene diversimalaria vector mosquitoes: an historical review and implications ties and fixation indices, Version (http: /www2.unil.ch/ for future malaria control in southern Africa. S. Afr. J. Sci. 95: izea/softwares/fstat.html) Goudet, J., M. Raymond, T. de Meeüs and F. Rousset, 1996 Testing Cohuet, A., I. Dia, F. Simard, M. Raymond, F. Rousset et al., 2004 differentiation in diploid populations. Genetics 144: Gene flow between chromosomal forms of the malaria vector Hey, J., 2003 Speciation and inversions: chimps and humans. Bio- Anopheles funestus in Cameroon, Central Africa, and its relevance Essays 25: in malaria fighting. Genetics 168: Hey, J., Y. J. Won, A. Sivasundar, R. Nielsen and J. Markert, 2004 Collins, F. H., L. Kamau, H. A. Ranson and J. M. Vulule, 2000 Using nuclear haplotypes with microsatellites to study gene flow Molecular entomology and prospects for malaria control. Bull. between recently separated Cichlid species. Mol. Ecol. 13: World Health Organ. 78: Hunt, R. H., 1973 A cytological technique for the study of Anopheles Coluzzi, M., 1968 Cromosomi politenici delle cellule nutrici ovari- gambiae complex. Parassitologia 15: che nel complesso gambiae del genere Anopheles. Parassitologia James, A. A., B. T. Beerntsen, Mde. L. Capurro, C. J. Coates, N. J. 10: Coleman et al., 1999 Controlling malaria transmission with ge- Coluzzi, M., 1982 Spatial distribution of chromosomal inversions netically-engineered, Plasmodium-resistant mosquitoes: milestones and speciation in Anopheline mosquitoes, pp in Mechain a model system. Parassitologia 41: nisms of Speciation: Proceedings From the International Meeting on Mechabe used to drive disease refractoriness genes into vector popula- Kidwell, M. G., and J. M. Ribeiro, 1992 Can transposable elements nisms of Speciation, edited by C. Barigozzi. A. R. Liss, NewYork. Coluzzi, M., and A. Sabatini, 1967 Cytogenetic observations on tions? Parasitology Today 8: species A and B of the Anopheles gambiae complex. Parassitologia King, M., 1993 Species Evolution: The Role of Chromosomal Change. 9: Cambridge University Press, Cambridge, MA. Coluzzi, M., V. Petrarca and M. A. Di Deco, 1985 Chromosomal Krimbas, C. B., and J. R. Powell, 1992 Drosophila Inversion Polymorinversion intergradation and incipient speciation in Anopheles phism. CRC Press, Boca Raton, FL. gambiae. Boll. Zool. 52: Lanzaro, G. C., and F. Tripet, 2003 Gene flow among populations of Anopheles gambiae: a critical review, pp in Ecological Coluzzi, M., A. Sabatini, A. della Torre, M.A. Di Deco and V. Aspects for Application of Genetically Modified Mosquitoes (Wagen- Petrarca, 2002 A polytene chromosome analysis of the Anopheingen Frontis Series, Vol. 2), edited by W. Takken and T. W. les gambiae complex. Science 298: Scott. Kluwer Academic Press, Dordrecht, The Netherlands Curtis, C. F., H. V. Pates, W. Takken, C. A. Maxwell, J. Myamba (http: /library.wur.nl/frontis/malaria/toc.html). et al., 1999 Biological problems with the replacement of a vector Lanzaro, G. C., L. Zheng, Y. T. Touré, S. F. Traoré, F. C. Kafatos population by Plasmodium-refractory mosquitoes. Parassitologia et al., 1995 Microsatellite DNA and isozyme variability in a West 41: African population of Anopheles gambiae. Insect Mol. Biol. 4: della Torre, A., C. Fanello, M. Akogbeto, J. Dossou-yovo, G.Favo Lanzaro, G. C., Y. T. Touré, J. Carnahan, L. Zheng, G. T. Dolo et al., 2001 Molecular evidence of incipient speciation within et al., 1998 Complexities in the genetic structure of Anopheles Anopheles gambiae s.s. in West Africa. Insect Mol. Biol. 10: gambiae populations in west Africa as revealed by microsatellite della Torre, A., C. Costantini, N. J. Besansky, A. Caccone, V. DNA analysis. Proc. Natl. Acad. Sci. USA 95: Petrarca et al., 2002 Speciation within Anopheles gambiae the Lehmann, T., M. Licht, N. Elissa, B. T. A Maega, J. M. Chimumbwa glass is half full. Science 298: et al., 2003 Population structure of Anopheles gambiae in Africa. Diabate, A.,T.Baldet, C.Chandre, K.R.Dabire, P.Kengne et al., J. Hered. 94: Kdr mutation, a genetic marker to assess events of intro- Long, J. C., 1986 The allelic correlation structure of Gainj and gression between the molecular M and S forms of Anopheles gam- Kalam speaking people. I. The estimation of Wright s F-statistics. biae (Diptera: Culicidae) in the tropical savannah area of West Genetics 112: Africa. J. Med. Entomol. 40: Machado, C., R. M. Kliman, J. M. Markert and J. Hey, 2002 Infer- Dobzhansky, T., 1970 Genetics of the Evolutionary Process. Columbia ring the history of speciation from multilocus DNA sequence University Press, New York. data: the case of Drosophila pseudoobscura and its close relatives. Dobzhansky, T., and O. Pavlovsky, 1958 Interracial hybridization Mol. Biol. Evol. 19: and breakdown of co-adapted gene complexes in Drosophila Marshall, E., 2001 A renewed assault on an old and deadly foe. paulistorum and Drosophila willistoni. Proc. Natl. Acad. Sci. USA Science 290: : Mayr, E., 1963 Animal Species and Evolution. Harvard University Press, Edillo, F. E., Y. T. Touré, G. C. Lanzaro, G. Dolo and C. E. Taylor, Cambridge, MA Spatial and habitat distribution of Anopheles gambiae and Mukabayire, O., J. Caridi, X. Wang, Y. T. Touré, M. Coluzzi et al., Anopheles arabiensis (Diptera: Culicidae) in Banambani Village, 2001 Patterns of DNA sequence variation in chromosomally Mali. J. Med. Entomol. 39: recognized taxa of Anopheles gambiae: evidence from rdna and Estoup, A., and J.-M. Cornuet, 1999 Microsatellite evolution: infer- single copy loci. Insect Mol. Biol. 10: ences from population data, pp in Microsatellites: Evolution Noor, A. F., K. L. Grams, L. A. Bertucci and J. Reiland, 2001 Chroand Applications, edited by D. B. Goldstein and C. Schlötterer. mosomal inversions and the reproductive isolation of species. Oxford University Press, Oxford. Proc. Natl. Acad. Sci. USA 98: Excoffier, L., P. Smouse and J. Quatro, 1992 Analysis of molecular Persiani, A., M. A. Dideco and G. Petrangeli, 1986 Osservzioni variance inferred from metric distances among DNA haplotypes: di laboratorio su polimorfismi da inversione originati da incroci