Patterns of Mitochondrial Variation Within and Between African Malaria Vectors, Anqpheles Gambiae and An

Coppight 8 1997 by the Genetics Society of America

Patterns of Mitochondrial Variation Within and Between African Malaria Vectors, Anqpheles gambiae and An. ambiensis, Suggest Extensive Gene Flow

Nora J. Besansky,*’tTovi Lehmann,*’+G. Thomas Fahey,* Didier Fontenille,* Lawrence E. 0. Braack,g Wfiam A. Hawlep* and Frank H. Collins*9t *Centen for Disease Control and Prevention, Division of Parasitic Diseases, Chamblee, Georgia 30341, +Departmentof Biology, Emmy University, Atlanta, Georgia 30322, ILaboratoire ORSTOM de Zoologie Medicale, Institut Pasteur, Dakar, Senegal, lKruge-r National Park, Skukuza, South Africa and **Kenya Medical Research Institute, Clinical Research Centre, Nairobi, Kenya Manuscript receivedJuly 10, 1997 Accepted for publication September 15, 1997

ABSTRACT Anopheles gambiae and An. arabiensis are mosquito species responsible for most malaria transmission in subSaharan Africa. They are also closelyrelated sibling speciesthat share chromosomal and molecular polymorphisms as a consequence of incomplete lineage sorting or introgressive hybridization.To help resolve these processes, this study examined the partitioning of mtDNA sequence variation within and between species across Africa, from both population genetic and phylogeographic perspectives. Based on partial gene sequences from the cytochrome b, ND1 and ND5 genes, haplotype diversity was high but sequences were very closely related. Within species, littleor no population subdivision was detected, and there was no evidence for isolation by distance. Between species, there were no fixed nucleotide differences, a high proportion of shared polymorphisms, and eight haplotypes in common over distances as great as 6000 km. Only one of 16 shared polymorphisms led to an amino acid difference, and there was no compelling evidence for nonneutral variation. Parsimony networks constructed of haplotypes from both species revealed no correspondence of haplotype with either geography or taxonomy. This trend of low intraspecific genetic divergence is consistent with evidence fromallozyme and microsatellite data and is interpreted in terms of both extensive gene flow and recent range expansion from relatively large, stable populations. We argue that retention of ancestral polymorphisms is a plausible but insuffi- cient explanation for low interspecific genetic divergence, and that extensive hybridizationis a contribut- ing factor.

nopheles gambiae and An. arabimis are important from one species have been introgressed into the other A components of a malaria-vectorial system encom- by backcrossing in the laboratory (DELLA TORREet al. passing most of subSaharan Africa. Of the six closely 1997). related Afrotropical sibling species collectively known Until quite recently, the only available phylogeny of as the An. gambiae complex, these two member species the An. gambiae complex was based on fixed paracentric arethe most widespread, occurring in sympatry inversion differences between species (COLUZZIet al. throughout most of subSaharan Africa and its offshore 1979; PAPE 1992). In spite of shared autosomal inver- islands. In spite of such extensive sympatry, the inci- sions and behavioral similarities, An. arabiensis and An. dence of naturally occurring interspecies hybrids (rec- gambiae were not predicted to be sister taxa. Their di- ognized chromosomally) is at orbelow 0.2% (reviewed vergent placement on the inversion tree was dictated in COLUZZIet al. 1979). The fate of hybrids in nature by species-specific inversions on the X chromosome. is unknown, but laboratory experiments have demon- Recent attempts at phylogenetic reconstruction using strated their potential as a bridge for gene exchange. DNA sequences have yielded conflicting results.Ge- Although male F1 hybrids are sterile, females are fertile netic distance measurements using DNA-DNA hybrid- and vigorous under laboratory conditions (COLUZZIet ization of total singlecopy DNA were unsuccessful be- al. 1979). Interestingly, An. gambiae and An. arabimis cause differentiation between species was below the lim- share cytologically identical paracentric inversions on its of resolution of this technique, underscoring their chromosome 2. If molecular analyses showthese inver- close relationship (N. BESANSKY,G. CACCONE and J. sions to bemonophyletic, their presence in bothspecies POWELL, unpublisheddata). However, phylogenetic can be explained either by retention of ancestral poly- analyses of sequences from mitochondrial, X-linked ri- morphism or by recent gene flow. The gene flow hy- bosomal, and two chromosome 2 genes (both outside pothesis is plausible in that chromosomal arrangements of the shared inversion) unanimously and strongly sup ported a sister taxa relationship between An. gambiae Cumspading author Nora J. Besansky, Department of Biological Sciences, 317 Galvin Life Sciences Bldg., University of Notre Dame, and An. arabiensis (BESANSKYet al. 1994; MATHIOPOU- Notre Dame, IN 46556. E-mail: [email protected] LOUS et al. 1995; CACCONE et al. 1996), in contradiction

Genetics 147: 1817-1828 (December, 1997) 1818 N. 1818 J. Besansky et al.

to the inversion phylogeny. Moreover,mtDNA phyloge- et al. 1993; GenBank accession # L20934) werePCR amplified niesbased on the ND4ND5 genes (BESANSKYet al. in 50-pl reactions containing 1 p1 of a 1:lO DNA dilution (or 1 pl undiluted if from legs alone), 50 pmol primers, 5 p1 and the control (CACCONE aZ. 1994) region et 1996) 1OX reaction buffer containing 15 mM MgC12, 200 p~ each from multiple laboratory strains revealeda paraphyletic dNTP (PerkinElmer), and 1.25 U Taq polymerase (Boehr- relationship between sequences from An. gambiae and inger Mannheim or GibcoBRL). Primers for the ND5 An. arabiensis. gene were 19CL (5'-CITCCACCAATTACTATAACAG3', This study attempts to distinguish the current and positions 6731-6755) and DMP3A(5"AGGATGAGATGG CTTAGGTT-3', positions 7680-7699). Primers for NDl-cytb historical relationships among thesetaxa, using a com- wereDM33C (5'-ACTCTAGCAAGTlTCGAGG, positions parative population geneticand phylogeographic analy- 11338-11356) and DMlOC (5'-GGTTTAGTCTGGCTAGCT, sis of mtDNA. It was premised on the expectation that positions 12000-12017). After 5 min denaturation at 94", 35 recent or ongoing hybridization leadingto interspecies cycles of 15 sec denaturation at 94", 15 sec annealing at 50", transfer of mtDNA mightbe distinguished from shared and 1 min extension at 72" were performed, followed by a 5 min final extension at 72". PCR products were purified using ancestral mtDNA polymorphism, given sufficient intra- the Wizard PCR Preps kit (Promega), and cycle-sequenced. specificdifferentiation over the entire geographic Manual sequencing was performed using the fmol DNA range of each species. The data presented hereprovide Cycle Sequencing System (Promega) and primers end-labeled the first detailed picture of the mitochondrial relation- withy-"P. For ND5, these primers were the same as those ships within and among these taxa acrossAfrica, a pic- used for PCR amplification. For NDl-cytb, the DM33C PCR primer did not perform well for sequencing and was replaced ture that implicatesongoing hybridization as a contrib with DMlC (5"GAGITCGAGGGACTTTA-3', positions uting factor in their evolutionary dynamics. 11751- 11767). Because both sequencing primers, DMlC and DMlOC, annealed to the same strand, NDl-cytb was only sequenced in one direction. MATERIALSAND METHODS Automated sequencing of ND5 was performed with the Sampling: Adult An. gambiae and An. arabiensis were col- PRISM Dye DeoxyTerminator Cycle Sequencing Kit (Applied lected from eight locations in Kenya and three locations in Biosystems), and Centriceppurified samples were run on the Senegal, and additional specimens of An. arabiensis were taken Applied Biosystems 377 DNA sequencer. Both strands were from two locations in South Africa that lie outside of the range sequenced, using 6848 (5'-ACTAACCGAAATGAATAACAT of An. gambiae (COETZEEet al. 1993; Figure 1).Morphological ACAG3', positions 6848-6872) and DMP3A. Sequences have identification of anophelines in the field followedGILLIES and been deposited in GenBank, accession numbers AF020965- DEMEILLON (1968); further identification of species in the AF021023. An. gambiae complex was achieved by a PCR assay (SCOIT et mtDNA analysis: The absence of insertions and deletions al. 1993). allowed for unambiguous sequence alignment. Basicse- In Kenya,gravid females were collected in Mayof 1987 quence statistics were computed with MEGA (KUMAR et al. by aspiration from the walls or bednets inside dwellings, as 1993). Haplotype diversity (heterozygosity) was calculated us- previously described (MCLAIN et al. 1989). The mosquitoes ing formula 8.5 of Nei(1987). Estimates of nucleotide variabil- analyzed werethe F1 progeny, which werereared to adulthood ity were not corrected for multiple substitutions because levels and stored in liquid nitrogen; only one progeny per field- of divergence were very low. A program written by T.L. in collected female was used. SAS language (SAS INSTITUTE1990) was initially used to esti- In Senegal, collections were made from Dielmo and Ndiop mate the average number of pairwise nucleotide differences between October and December of 1994 by indoor and out- within and between populations of each species (nucleotide door night captures on human volunteers as described pre- diversity T, equation 10.5 of NEI 1987) as well as the gross viously (FONTENILLEet al. 1997a,b). Collections at Barkedji and net divergence between them (d, and dA of NEI 1987, were made in September of 1995 by pyrethrum spraying of equations 10.20-21). These and other population genetics bedrooms and storerooms (LEMASSONet al. 1997). While still parameters were also computed by the program DnaSP 2.0 in the field, specimens were individually placed into 1.5 ml (ROZASand ROZAS,1997). The standard error (SE) of T was microcentrifuge tubes with desiccant; long-term laboratory the square root of the total (stochastic + sampling) variance storage was at -20". per nucleotide site (TAJIMA,1993). The parameter 0 (= 2Np) In a section of Kruger National Park, South Africa, at least was estimated from the number of segregating sites (com- 9 km from humanhabitation, a freshwater geothermal spring puted by DnaSP with equation 10.3 of NEI, 1987 and herein provides a permanent breedingsite for An. arabiensis (BRAACK called Os). Its SEwas the square root of the variance per et al. 1994). At this site (Malahlapanga; 31"03'E, 22'53's) in nucleotide site for no recombination (equation 4 of TAJIMA, December 1995, mosquitoes seeking a bloodmeal from hu- 1993). Alternatively, the phylogenetic estimate of O (UPBLUE man volunteers were captured by mouth aspiration. These of Fu, 1994 and herein called 0,) was computed at the Web were sealed in individual gelatin capsules and stored at room site of Y.-X. Fu, http://hgc.sph.uth.tc.edu/fu. temperature with desiccant. Additional An. arabiensis speci- To test for neutral mutation, the D statistics of TAJIMA(1989) mens from South Africawere obtained from a laboratory and Fv and h (1993) were computed using DnaSP. TAJIMA'SD strain established from material capturedin Mananga was calculated using the value of Os based on the number of (31°50'E, 25"56'S), 350 km south of Malahlapanga. segregating sites. The D statistic of Fu and LI required an out- DNA extraction, amplification,and sequencing: DNA from group. This consistedof ATD5 sequences fromtwo fieldcollected individual specimens, or legs from specimens, was extracted An. menu, another sibling in the species complex (GenBank (COLLINSet al. 1987) and resuspended in 50-100 pl water. A accession numbers AFO21024-AFO21025). 968-bp segment of the mitochondrial NADH dehydrogenase The extent of nucleotide differentiation between popula- subunit 5 (ND5) gene and a 679-bp segment spanning part tions was calculated by estimating FS~,where FST = 1 - [HJ of the NADH dehydrogenase subunit 1 (NDI),the complete Hb],and H," and Hbare theaverage number ofpainvise differ- tRNAsergene, and part of the cytochrome b (cytb) gene (BW ences within and between subpopulations, respectively (HUD- mtDNA Variation in An. gambiae and An. arnbiasis 1819

.escarpment

Kisian. *Wafhrego a"-L- d Ahero Muhroni W. KENYA

2 FIGUREI.-Map of col- Area of I Nyakoch. detail lection sites. Representa- tivesof both species were d collected at all sites except those in South Africa (An. arabiensis only) and two in Kenya (Ahero, An. am- =F biasis only; Nvakoch, An. .Barke gambiae only). 0

SON et al. 1992a, Equation 3). To account for differences in An. gambiae and 23 An. arabimsis sampled from Kenya. sample size, H,,, was calculated based on a weighted mean. These two gene regions were chosen because prior re- Significance of FS7.estimates was evaluated against the results from 500 random permutations (HUDSONet al. 1992b), exe- striction site surveys of Kenyan populations from both cuted by a program written in SAS by T. LEHMANN.Levels of species identified them as polymorphic (N. BESANSKY, gene flow (N,J were estimated by N,,, = HJ2( Ht>- H,,,) ( HUD- P. MEHAFFEYand F. COLLINS,unpublished data). All SON et al. 1992a, Equation 4). sites segregating within An. gambiae and An. arabiensis, Genealogical relationships among mtDNA haplotypes were andthe sequence variants (haplotypes)defined by represented in a hand-drdwn network that minimized the number of painvise mutational differences. The probability them, areshown in Figure 2. All polymorphic sites were of parsimonious (i.e., no unobselved) connections between either silent codon sites or were conservative replace-

haplotypes was calculated as in TEMPLETONet al. (1992). ment sites (position 7290, V + I; 7332, D + N; 7479, Allozyme analysis: The allozyme frequency data reported V + I; 7485, G + S; 7506, I + V). This, together with by MILES (19i8) for populations of both species across Africa were analyzed using BIOWS (SWOFFORDand SELAVDER,1989) the fact that direct sequencingrevealed no ambiguities as described previously bv LEHMANNet al. (1996) for An. gam- characteristic of heteroplasmy and that no insertion- bine. Populations of at least 14 mosquitoes from West Africa deletion differences were found within or between spe- (The Gambia: Keneba, Mandinari; Senegal: Hanene, Thies), cies, suggests that the sequenced segments were bona East Africa (Kenya:Kisumu, Chulaimbo), and SouthernAfrica fide mtDNA rather than nuclear-transposed copies. (South Africa: Pelindaba; Zimbabwe: Kanyemba) were examined using data from five to six loci. These were the a-naph- No sites were fixed at different nucleotides between thy1 acetate esterases (Est-], Est-3 for both species and Est- species. Of the 12 ND1-9th sites that were polymorphic 2 for An. gambine only), octanol dehydrogenase (Odh), and in either species, 11 and threewere polymorphic within phosphoglucomutase (Pgm-1, Pgm-2). An. gamhiae and An. arabimsis, respectively. More striking, of 39 ND5 nucleotide sites polymorphic in either RESULTS species, 28 and 25 werepolymorphic within An. gamhiae Inter- and intraspecific mitochondrialDNA polymor- and An. nrahiensis, respectively. Not only was there sub- phism: The nucleotide sequence of 665 bp of the ND5 stantial shared polymorphism in both species, but the gene (positions 6896-7560 in the An. gambiaereference distribution of ND5 haplotypes by species and geo- sequence of BEARDet al. 1993) was determined for a graphic origin indicated eight haplotypes common to total of 65 An. gamhiae and 56 An. arabimsis sampled both species over distances as great as 6000 km (Table from three countries: Kenya, Senegal, and South Africa 1). The average level of ND5 sequence divergence be- (Figure 1).In addition, the nucleotide sequenceof 531 tween species was only 0.46% per nucleotide site, with bp containing thetRNA5"' gene flanked by portions of a net difference of 0.04% after accounting for within the NDI and cytb genes (positions 11340-11870in species polymorphism. BEARDet al. 1993) were determined for the same 37 Within each species, no single sequence type pre- 1820 N. J. Besansky et al.

Nn5 m1 -cytb 666667777777777777777777777777777777777 111111111111 HF 999990000011122222223333333344444444445 111111111111 ar 112681247902303458890123568900234578880 455566777777 ~e 388725755959841052806872701628324093566 523957236689 #q RR RRR 057953993619 a1 (13) ACCAAACGTTTATTCGCTGAA&TAAGCCCCT TGGCAGCAATGT ...... A...... C...... G.A...... C ...... T...... A...... A.A ...... C ...... T. ..A...A...A ...... AG...... C.,...T...... C ...... A...... G...AG..... G..C...... C .....A. ..A...... C. ...A...... C ....C...... T...... CA. FIGURE 2.-Unique ...... A... ..A ...... Uk.*...... T. T...... A. ...A...... T ....C....A...... C...... T. ..AA.... G.A. mtDNA haplotypes found ...... A...... T.C...... T...... G... for the NL)5 and NL)l-cytb ...... A...... A...... C...... T. ..A...... A. gene fragments from both ....G...... A ...... T. CA.....T..A. species. Onlythe polymor- ...... A...... TC...... T. C.A...... A. phic positions are shown, ...... A...... C.G...... T. ...A ...... I4 and these arenumbered ...... A...T...... T...... A...A. with reference to the pub- ...... A..G...... C ...... G. C...G.....A. lished An. gambiae mtDNA ...... AG...... C ...... T.T. *gggaggg*ggg sequence (BEARDet al., ....G ...... A...... 1993; GenBank accession ...... A...... A...... C...... T.. no. L20934). Dots repre- ...... A...... C.G.....A. ..T. sent identity with respect ....G...... A...... C...... G. ...T. to the first sequence listed. ....G.T ...... A...... Each haplotype was as- ...... c...... signed a number that was TT ...... C. ...A...... C ...... T. preceded by the letter ora T ...... A...... C..G...... C g if it was unique to An. TT ...... A...... C...... A. ..T. arabiensis or An. gambiae, T ...... A...... C...... respectively;if shared by ...... A..G ...... C...... both species, it was pre- ...... A ...... TC...... ceded by an asterisk. The ...... A...... A. frequency, given in paren- ...... TA...... theses, indicates the num- ...... A...... ber of times the haplotype ...... A...... C...... T. was found in thetotal sam- ...... A...... C..T...... T. ple. The five positions in ...... A... ..G ...C...... T. ND5 in which bp substitu- ...... T ...... A...... C...... T. tions would result in an ...... A...... T. amino acid replacement ...... AG.....G .. C...... T.... are indicated by an R ...... C ....A...... T.C...... T. above the sequence. Be- ...... A .....G...C.....T...... low the list of sequences, ...... A ...... AG...... C...... T. an asterisk markspositions ...... A ...... A...... C...... thatare polymorphic in ...... C ..A...... T. both species; an a or g ...... AG ....G. ..C.....T...... marks positions that are ...... C ..A...... polymorphic only in An...... T. arabiensis or An. gambiae, ..T ...... A...... C...... T... respectively...... AG ...... C...... T...... A...... C...G...... T. ..T ...... A...... C...... A ....G...... C...... A...... C.C. ...A...... G ...... T...... TA ...... C ...... T...... A ...... C...... T. T...... A ....G. ...C..G ...... A ...... C...... A... T...... A ...... -...... T. **ggaga*gaggag*a*agag*g****a*Waa*gaa*g*a*g mtDNA Variation in An. gambiae and An. arabiensis 1821

TABLE 1 Geographic distribution of ND5 haplotypes from An. gumbiae and An. ambieds

An. gambiae An. arabiensis CollectionSamplesite mtDNAsize haplotypes"Sample mtDNAsize haplotypes" Kenya Asembo 8 2, 11, 32, 33,47, 48 (2), 49 1 4 (2), 2, 40 Escarpment 6 33, 42,50, 51, 52, 53 2 1, 41 Ahero - - 4 1, 11, 42, 43 Kisian 2, 4 11, 32, 54 2 44 1, Wathrego 6 2, 33 (3), 48, 53 5 1 (Z), 11, 43,45 Nyakoch 4 2, 11, 55,56 - - Jew 5 32 (2), 41 (2), 57 2 1 (2) Muhroni 4 4 3359 (Z), 58, 41,1, 2, 46 Senegal Dielmo 10 2 (4), 3 (2), 4,16, 35, 39 9 5, 17, 18, 19, 20, 21,22, 23, 26 Ndiop 9 2, 3 (Z), 6,7, 24, 25, 27, 34 9 1 (2), 10, 11, 12, 13, 14, 15, 37 Barkedji 9 2, 3 (4), 9, 28,29, 38 8 1, 2, 8, 33 (2), 34, 35,36 South Africa Malahlapanga - - 5 11, 30 (3), 32 Mananga - - 2 31 (2) Haplotypes shown in bold italic are present in both species. " Values in parentheses are frequency.

dominated. Estimated haplotype diversity was quite (Fu and LI, 1993) were consistent with neutral muta- high in all populations sampled, although certain hap tion of ND5 and NDl-cytb. Significant negative values lotypes (e.g., 2 and 11) were geographically widespread found in the total An. gambzm sample likely were due (Table 1). Among the 65 An. gambiae hD5 sequences, to inappropriate pooling of locations in Senegal and 33 haplotypes were found, 26 of which were singletons Kenya (see below), each with a high proportion of sin- (represented in a single individual). Similarly, 11 of 15 gletons. Interpretation of the significant negative value NDI-cytb haplotypes were singletons. In An. arabiensis, of Fu and LIS D from An. gambiae in Senegal will require the same pattern was detected. Among 54 ND5 se- sequence data from multiple unlinked loci (SIMONSEN quences (excluding thetwo derived from the Mananga et al. 1995). In the absence of compelling evidence to colony), 32 haplotypes were found, 25 of which were the contrary, we assume that variation detected in the singletons. The two mosquitoes from the Mananga col- mtDNA of both species is neutral and that selection is ony shared the same haplotype, which differed slightly not responsible for the maintenanceof shared polymor- from the 32 sampled from natural populations. For the phisms between these species. 23 An. arabiensis NDl-cytb sequences, there were no sin- Geographicstructure and gene flow within spe- gletons among four haplotypes, yet diversity was still cies: The locations sampled in Senegal and Kenya were relatively high. separated by -6000 km and were, respectively, -7000 This diversity of haplotypes was achieved with rela- and -2000 km removed from Kruger National Park, tively slight differentiation in pairwise comparisons (Ta- South Africa. Because theyrepresent opposite extremes ble 2). Thus, theaverage number of pairwisenucleotide of the species' distributions on the African continent, differences (T)for An. garnbim ND5 was 0.38% per site, it was expected that comparison of sequences sampled or 2.5 nucleotides (nt) per sequence. Similarly, T for from these locations would show the maximum levels An. arabiensis was 0.46% per site for ND5, or 3.1 nt per of genetic divergence. sequence. In neitherspecies did any pair of haplotypes The effect of distance on levels of gene flow was esti- differ by more than 7 nt. mated from pairwise FsT values. For both species and Tests of neutralmutation: Under the infinite-sites all comparisons among paired locations within 700 km model of neutral mutationsand a population at equilib- of each other, these values wereeither zero or insignifi- rium with respect to mutationand drift, 7r and 6 should cantly small, as judged by the results of 500 random both equal 2Np for mtDNA, where N is the effective permutations. For An. arabiensis, even the FsTvaluesfor population size and p is the rate of mutation per site South Africa vs. Kenya or Senegal were not significant. per generation (NEI, 1987). In An. gambiae and An. Only for Senegal vs. Kenya was significant genetic differ- arabiensis populations, values ofOs and 6"were generally entiation detected, with FST = 0.018 (P< 0.02) for An. higher than those for K (Table 2). Nevertheless, most arabiensis, and FsT = 0.085 (P< 0.002) for An. gambiae. values of TAJIMA'SD (TAJIMA,1989) and Fu and LIS D While significant, these values nevertheless corre- 1822 N. J. Besansky et al.

TABLE 2 Summary statistics for mtDNA polymorphism

nS h T os err Tajima’s D F and L’s D An. gambiae Total (AD3 65 28 0.948 0.0038 0.0089 0.0181 - 1.836* -2.46* (0.0021) (0.0028) (0.0039) Senegal (AD3 28 17 0.886 0.0038 0.0066 0.0107 - 1.467Ns -2.97** (0.0021) (0.0025) (0.0032) Kenya (ND5) 37 15 0.938 0.0035 0.0054 0.0093 -1.158NS -0.2SNS (0.0019) (0.0021) (0.0027) (ADl-Cytb) 37 11 0.856 0.0031 0.0050 ND -1.13gNS -0.7gNS (0.0019) (0.0020) An. arabiensis Total (AD3 25 55 0.937 0.0046 0.0082 0.0202 -.1413NS - 1 .4SNS (0.0024) (0.0027) (0.0044) Senegal (ND5) 23 26 0.988 0.0051 0.0091 0.0209 - 1.602NS - 1.84NS (0.0027) (0.0034) (0.0056) S. Africa (AD3 530.70 0.0023 0.0022 0.0026 0.262NS 0. 76Ns (0.0016) (0.0015) (0.0017) Kenya (AD3 23 8 0.806 0.0040 0.0033 0.0060 0.684Ns 1.42NS (0.0022) (0.0015) (0.0022) (ADl-qtb) 23 3 0.708 0.0026 0.0015 ND 1 .690NS 0.9gNS (0.0016) (0.0010)

~~ ~~ ~ ~~ ~~~ ~ n, the number of sequences; S, the number of polymorphic sites; h, haplotype diversity; T, Os, err,Tajima’s D, F and L’s D, are as defined in MATEUS AND METHODS; ND, not computed. The totalAn. arabiensis sample included a single specimen from the Mananga colonyas well as all field specimens from Malahlapanga, South Africa. Values in parentheses are SE. * P < 0.05; ** P < 0.02.

sponded to average levels of migration (N,) in excess BIOSYS. FSTanalyses were carried out on five enzyme of two per generation, above the threshold required loci from An. gambiae and four from An. arabiensis sam- for genetic differentiation by genetic drift (SLATKIN, pled from West Africa(Senegal and TheGambia), from 1987). East Africa (Kenya), and from an An. arabiensis popula- Evidence for isolation by distance was examined in tion from South Africa. These enzyme loci are autoso- both species for each pair of locations up to 100 km mal; the esterases and Pgm2 have been mapped tochro- apart, and for An. arabiensis over the entire continent, mosome 2 and Odhis linked to chromosome 3 (HUNT, after pooling the six samples from western Kenya and 1987; CREWS-OYEN et aZ. 1993). In addition, microsatel- the two from southwestern Senegal. Because N,is unde- lite data from five loci representing all three chromo- fined when FST = 0 (as it often was), pairwise FyT values somes are available from An. gambiae populations from rather thanN, values wereplotted against distance (Fig- Senegal and Kenya (LEHMANNet al. 1996, 1997)) and ure 3), and against log(distance) (not shown). Al- from 25 loci from both species in Kenya (muet al. though the highest FsTvalues wereassociated with com- 1997). Table 3 compares the FST estimates based on parisons across 2000 km, it was also true thatsome pairs mtDNA to those derived from allozyme and microsatel- of samples separated by 6000-7000 km showed FST = lite markers. Because sampling times and locations were 0. No simple nor convincing pattern of isolation by different for the different markers and the calculation distance was evident from these mtDNA sequences; vari- of FST or RsT differed by marker, exact numbers cannot ation in levels of differentiation was independent of be appropriately compared. However, the concordant distance. However, it should be noted that the coarse patterns of genetic differentiation indicated by the dif- geographical sampling scheme, and small sample sizes ferent markers strengthenthe view that large geo- at some locations (e.g., coastal Kenya), may have limited graphic distances have not imposed a barrier to gene the ability to detect isolation by distance. flow within species and emphasize the low level of diver- Analyses based on this single, maternally inherited gence between species. locus seemed to suggest that neither large geographic Pairwise difference distributions: If the frequency of distances nor species designations play an important mtDNA haplotype pairs that differ by i nucleotide sites, role in stratifymg variation in An. gambiae and An. ara- where i 3 0, is plotted, the shape of the distribution biensis. Evidence from allozyme and microsatellite mark- can provide information about the history of changes ers supports this view. Allozyme frequency data from in population size (SLATKINand HUDSON, 1991; ROG both species, collected from numerous locations ERS and HARPENDING, 1992). Simulation studies suggest throughout Africa (MILES,1978), were analyzed using that for a stable population, the shape is expected to be mtDNA Variation in An. gambiae and An. arabiensis 1823

A B 0.3 0.5 I3 0.25 + 0.4 + 0.2 0.3 u 0 2 0.15 + 0 0.2 0.1 0 0.1 0.05 1,U + AI-I m 0 I I 0 0 20 40 60 80 I00 2,000 4,000 6,000 8,000 distance distance 1 An. gambiae arabiensisAn. I FIGURE3.-Values of F,, plotted against distance (km) ona microgeographic scalefor both species (A), ora macrogeographic scale for An. arabiensis (B). In each case, the fit values were weighted by the harmonic mean of the sample sizes.

irregular because of stochastic lineage loss. By contrast, recent population expansion should produce a smooth TABLE 3 unimodal distribution that approximates Poisson. For both An. gambiae and An. arabiensis, the distribution of Geographic structure and geneflow indicated by mtDNA pairwise differences for ND5 sequences from Senegal (ND5), allozyme, and microsatellite markers and Kenya was plotted (Figure 4). Although all distribu- mtDNA"Allozymes Microsatellites tions except Kenyan An. arabiensis appeared unimodal, FST FST RST each strongly deviated from POISSON(P Q 0.01). This result is consistent with population stationarity but does An. gambiae not necessarily rule out morecomplex models of popu- 5-1000.017 (28.9) 0.006 (41.4) 0.013' (38.0) lation expansion (see ELLERand HARPENDING, 1996). 250-700 km 0.032 (15.1) 0.025 (9.8) 0.037d (9.8) 2000-7000 km 0.085 (5.4) 0.030 (8.1) 0.036' (3.4) mtDNA phylogeography: With only 25 polymorphic An. arabiensis sites that were informative for parsimony and more than 5-100 km 0 (Sl) -ND -ND twice that number of haplotypes, maximum parsimony 250-700 km 0 (Sl) 0.021(11.7) -ND analyses could not be usefully applied. Instead, a parsi- ND 2000-7000 km 0.044 (10.8) 0.038 (6.3) - mony network was constructed in which the quantity An. gambiae vs. minimized was not global tree length, but individual An. arabiensis 0.093 0.072' 0.153' mutational connectionsbetween haplotypes (Figure 5). Values in parentheses are Nm. ND, no data available. This approach better reflected the expectation that, for a Summary FyT values for each geographic range were an a collection of very closely related sequences, common average ofpairwise FSTvalues, with each pairwisevalue ancestral sequences will beextant in the collection weighted by the harmonic mean of sample sizes. In computing this average, negativepairwise Fsrvalues were not considered (CRANDALLet al. 1994). Theprobability of a parsimoni- as zero, to better account for random noise. However, any ous connection between haplotypes was supported negative summary values obtained after averaging were con- across the network at P 2 0.97 (equation 8 of TEM- sidered as zero. PLETON et al. 1992). However, the proportion of ambig- Average across four loci, excluding ODH. Including ODH, uous connections indicates that atleast some mutations FST = 0.191. are homoplasious. Indeed, although the network-based 'Average across five loci from LEHMANNet al. (1997). Average acrossnine loci (T. LEHMANN,unpublished data). mutational distances between pairs of haplotypes were 'Average across five loci (from LEHMANNet al. 1996). often in good agreement with observed distances, this 'Median R,~Tvalue for 25 loci (from KAMAU et al. 1997). was not always the case. For example, haplotypes 24 1824 N. J. Besansky et al. An. gambiae minal ones were not. However, there were three termi- A nal haplotypes from Senegal (8, 17, and 19) that were 150 connected to haplotypes found onlyin Kenya, even (D n e when all ambiguous connections were considered. It is 0 striking that haplotypes 8 and 17 were also from An. .-u) 100 arabiensis but connected to An. gambim haplotypes.

0E 0 DISCUSSION 6 50 z Intraspecific mtDNA variation: Previously published studies of mtDNA variation within anopheline species 0 01234567 have used restriction mapping of the entire molecule No. pairwise differences (eg., CONNet al. 1993, 1997; FREITAS-SIBAJEVet al. 1995; PERERAet al. 1995). With this technique, estimated lev- An. arabiensis els ofnucleotide diversity found for six species of nearc- 6 tic and neotropical anopheline mosquitoes ranged 100 from 0.0018 to 0.0085 (CONNet al. 1997). Nucleotide

(D diversity estimates obtained by sequencing the ND5 g 80 gene fragmentof An. gambim and An. arabiensis (0.0038 u) .-L and 0.0046, respectively) fall within this range. How- 2 60 ever, the pattern of mtDNA variation found in each 5 geographic location in the An. gambim complex is at 0 40 least superficially distinct from that described for the 0 z other anophelines, in that no haplotype predominates 20 and the majority of haplotypes are singletons. Further- more, cooccurring haplotypes commonly differ at four 0 01234567 to five nucleotide positions. This is a pattern consistent No. pairwise differences with relativelylarge continuous populations that do not experience severe dry season bottlenecks, a conclusion Kenya Senegal reached by TAYLORet al. (1993) based on frequencies of 0 chromosomal inversions in An. arabiensis. In the other anophelines examined to date, one or two haplotypes FIGURE4.-Pai&sedifference distributions from mtDNA predominate at each geographic location, with rare ND5 sequences of An. gumbiue (A) and An. arabiasis (B). Empirical distributions are shown as hatched bars (Senegal) ones differing by a single restriction site (CONNet al. or white bars (Kenya); their fit to a Poisson distribution is 1993, 1997; FREITAS-SIBAJEVet a[. 1995; PERERA et al. given by solid (Senegal) or dashed (Kenya) lines. 1995). A possible explanation for this distinction is the higher resolution of sequencing us. restriction map and 25 are seven steps apart in the network but only five ping. It remains to be seen whether underlying differ- steps apart when the sequences are compareddirectly. ences in population size and structure also contribute. In Figure 5, those haplotypes that were most frequent Despite high levels of allelic diversity, we detected in the collection were generally positioned at interior little to no geographic structure to themtDNAvariation nodes, from which many connections sprouted. In A, within An. gambim and An. arabiensis, even at distances where haplotypes are colorcodedaccording tospecies, of 6000 km. Although there are few obvious environ- it can be seen that most of the more frequent,ancestral mental barriers to geneflow, these results were unantic- haplotypes were shared between species, whereas termi- ipated. In part,this was because the population genetics nal haplotypes (those at the tips) were rarely shared. of these species has been studied mainly from the point So far, this pattern is most easily interpreted as sorting of view of chromosomal inversion polymorphisms (e.g., of shared ancestral polymorphism. However, assuming Co~uzzret al. 1985; TouRt et al. 1994), which are un- that terminal haplotypes were created by the most re- likely to beselectively neutral. For An. gambiue, in which cent mutations, it is worth noting that threesuch haplo- the frequencies of alternative chromosomal arrange- types from An. arabiensis (8,17, and26) were connected ments in West Africa showsignificant temporal and spa- most parsimoniously to haplotypes only found in An. tial heterogeneities strongly correlated with environ- gambim. In B, the same networkof haplotypes was mental factors (COLUZZIet al. 1985; TOU& et al. 1994), coded accordingto geographic origin (Senegal, Kenya, the impression was highly structured populations. How- or South Africa). This revealed a similar pattern, in ever, if natural selection were responsible for this pat- which the more frequentancestral haplotypes were gen- tern, it may “say more about theenvironmental condi- erally shared between geographic regions and the ter- tions than about the gene flow regime of the species” mtDNA Variation in An. gnmbinr and An. nrnbimsis 1825

FIGURE 5.--Parsimony network for 59 mtDNA iW5 haplotypes from An. gnmhiat and An. nrilipnsis. Num- bered haplotypes are en- closedincircles roughly proportional to theirfre- quency: “0” are missing in- termediates. Each connection represents one nucleo- tidedifference. Dashed linesindicate ambiguous connections.In A, haplotypes are colorcoded ac- cording to taxonomy, with white circles and black circles representinghaplo- types unique to An. gamhim or An. arainmis, respectively. Hatched circlesrepre B sent haplotypes common to C.._._. .-...... “...... “ bothspecies. In B, haplotypes are colorcoded ac- cording to geography, with whitecircles, black circles, and checked circles (haplotypes 30-31) representing haplotypes unique to Ke- nya, Senegal, and SouthAf- rica,respectively. Hatched circles (1 -2, 11, 33, 41-42) represent haplotypescom- mon to Senegal and Kenya; cros-hatched circle (32) represents a haplotype common to South Africa and Kenya.

...... “..___ ...... -... -.

(AWE, 1994). Indeed, notonly do patterns ofallozyme cently proposed for a similar pattern foundin the West- and microsatellite variation concur with mtDNA, but ern European house mouse(NACHMAN et al. 1994). The estimates of gene flow based on these different classes parsimony network (Figure 5) fulfills a prediction of of markers are remarkably similar in spite of mutation recent colonization in that the older, more frequent rates that may differ by several orders of magnitude interior haplotypes are the most widespread geographi- (LEHMANNet nl. 1996, 1997; Table 3). cally, whereas the newer terminal ones are geographi- One possible interpretation is that gene flow within cally limited (CASTELLOEand TEMPLETON,1994). both species is extensive enough to prevent differentia- Is the dispersal ability of these mosquitoes consistent tion across the African continent. An alternative expla- with that required to make geneflow a plausible explanation is that this pattern does not reflect current popu- nation for low differentiation across Africa? One mark- lation structure, but thehistory of a recent range expan- release-recapture study measured the meanflight range sion across Africa by both species. The shape of the at <1.6 km, with over 90% recaptured within 3 km mismatchdistributions (Figure 4) doesnot support of the release point (GILLIES, 1961). However, flight rapid population growth from a small founder popula- capacity of An. gambiae is as high as 7 km (HOLSTEIN, tion, but is still compatible with range expansion by 1954), and a recent mark-release study gave consider- large and relatively stable populations, a scenario re- ably higher estimates of 350-650 meters per day per 1826 N. J. Besansky et al. adult mosquito, whose daily probability of survival was plex (COLLINSet al. 1990).An. gambiae and An. arabiensis 0.8 (CONSTANTINIet al. 1996). NEICELand AVISE(1993) constitute another exception,since there were no fixed have modeledthe expected geographic ranges for nucleotide differences and multiple shared haplotypes mtDNA lineages dispersed by a multigeneration “ran- between them. dom walk.” Assuming a per generation dispersal dis How likely isit that these taxa share identical mtDNA tance of 2-10 km, 12 generations peryear, and mtDNA haplotypes solely on the basis ofretained ancestral poly- lineages 2-12 million generations old (using a molecu- morphism? The expectation that fixed mtDNA differ- lar clock of 1.1-1.2% per million years per lineage; ences will distinguish closely related species depends BROWER,1994), the expected geographic rangefor on the time since species splitting and female effective these lineages actually exceeds 6000 km (Figure 4 of population size. The shorter thetime and the larger the NEICELand AVISE,1993), so that it is at least theoreti- population size, the greater the likelihood of mtDNA cally possible that equilibrium exists between gene flow polyphyly or paraphyly between taxa (AVISE,1994). As- and genetic drift across Africa. Beyond the possibility suming that all gene flow between species permanently of active dispersion, it shouldalso be noted thatpassive ceased at the time of species splitting, the probability transport of An. gambiae over long distances by plane, that both taxa retain identical mtDNA haplotypes from boat, and truck has beenrecorded repeatedly both the ancestral population is veryslight beyond 4n genera- within and outside its normal distribution range (HOL tions, where n is carrying capacity or female population STEIN, 1954). When such transport leads to previously size (AVISEet al. 1984). The long-term effective popula- unoccupied territory with potential breeding sites, the tion size of An. gambiae has been estimated from micro- process of colonization can be remarkably rapid. An. satellite frequencies across nine loci to be -240,000, gambiae invaded Brazil in late 1929 or early 1930 and using a conservative average mutation rate of lop5 (T. within 10 years had spread into an area with a -230 LEHMANN,unpublished data). Using this estimate and km radius before it was eradicated ( SOPERand WILSON, assuming stable-sized populations and selective neutral- 1943). ity, it is very unlikely for ancestral mtDNA haplotypes Although active and passive dispersal may be suffi- to have persisted in either taxa beyond 480,000 genera- cient explanations for the distribution of mtDNA variations (-40,000 years). This is much less than the mean tion in both species, their behavior makes it difficult to time to coalescence for the most divergent haplotypes discount recent range expansion as an additional fac- in either species: -510,000 years witha standard devia- tor. An. gambiae and to a lesser degree An. arabiensis are tion of -180,000 years (calculated after TEMPLETON, “domestic” species, not only depending on humans for 1993, using a mutation rate of per year). Based on blood, but also for resting and breeding sites, since they a gamma distribution, the 95%confidence limits about rest in human habitations or storage sheds and breed these means, 219,000-914,000 years, do not include in bare sunlit pools created by human activity such as 40,000 years. However, there is no external calibration footprints, animal hoofprints, rice fields, and irrigation for the mtDNA mutation rate in An. gambiae, and if a ditches. The anthropophily of An. gambiae led COLUZZI 10-fold higher mutation rate is used, the mean time to et al. (1985, p 46) to observe that “it would be difficult coalescence is -51,000 years with a standard deviation to hypothesize its evolution and its wide diffusion in of18,000 years, and the 95% confidence limits of Tropical Africa in the absence of man.” On behavioral 21,900-91,400 years easily include 40,000 years. With grounds, it is tempting to speculate that An. gambiae conservative estimates of mutation rates, this coalescent is a relatively young species, whose origin may have approach does not rule out shared ancestral polymor- coincided with the thousand-fold increase of human phism as the explanation for shared haplotypes. How- populations in Africa followingthe arrival of agriculture ever, if retained ancestral polymorphism were accepted into West Africa within the last 9000 years (CAVALLI- as a sufficient explanation, nuclear gene polymor- SFORZAet al. 1993). phisms should be shared even more extensively than Interspecific mtDNA variation: Of the several dozen those of mtDNA(because of their larger effective popu- species of anophelines implicated in human malaria lation sizes), and this has not yet been found outside transmission, most are thought tobelong to sibling spe- of shared inversions (BESANSKYet al. 1994; MATHIO- cies complexes (COLLINSand PASKEWITZ,1995); thus POULOS et al. 1995; GARCIAet al. 1996). Therefore, we the ability to discriminate members of a complex has a believe that transfer of mtDNA across species bound- practical dimension that has motivated many studies. aries provides a plausible explanation for haplotypes Where restriction fragment length polymorphism sur- shared between these taxa. veys of mtDNA variation in other anopheline species Frequent examples of natural hybridization and in- complexes have been made, in An. dims (YASOTHORNS trogression exist in bothplants and animals (HARRISON, RIKUL et al. 1988), An. freebmi (COLLINSet al. 1990), An. 1993; AVISE, 1994;ARNOLD, 1997), particularly among quadrimaculatus (MITCHELLet al. 1992) and An. albitarsis closely related taxa. What is unusual in the case of An. (NARANGet al. 1993), fixed differences distinguished gambiae and An. arabiensis is the apparent absence of member species in all casesexcept the An. freebonzi com- any hybrid zones and the geographic scale of mtDNA mtDNA Variation in An. gambiae and An. arabiasis 1827 introgression (but see MASON et al. 1995), such that CAVALLI-SFORZA,L. L., P. MENOZZI and A. PIAZZA,1993 Demic expansions and human evolution. Science 259 639-646. without independent molecular or cytogenetic mark- COETZEE, M.,R. H. HUNT, L. E. 0. BUCK and G. DAVIDSON,1993 ers, these isomorphic taxa wouldnot be distinguishable. Distribution of mosquitoes belonging to the Anopheles gumbiue Although potentially fertile female hybrids have been complex, including malaria vectors, south of latitude 15"s. S. Afr. J. Sci. 89: 227-231. found at rates of one to two per thousand females exam- COLLINS,F. H., and S. M. PASKEWITZ,1995 Malaria: current and fu- ined (CoLuzz1 et al. 1979), ratestheoretically high ture prospects for control. Annu. Rev. Entomol. 40 195-219. enough to homogenize interspecific variation,their rel- COLLINS,F. H., A. MENDEZ,M. 0. "USSEN, P. C. MEHAFFEY,N. J. BESANSKYet ul., 1987 A ribosomal RNA probe differentiates ative fitness in natural populations has not been stud- member species of the Anopheles gambiue complex. Am. J. Trop. ied. Our data indicate that at least some proportion Med. Hyg. 37: 37-41. of hybrids must successfully backcross to either parent COLLINS,F. H., C. H. PORTERand S. E. COPE, 1990 Comparison of rDNA and mtDNA in the sibling species Anophelesfi-eebai and species, and depending upon the direction of the back- A. hermsi. Am. J. Trop. Med. Hyg. 42: 417-423. cross, transfer mtDNA across species boundaries. This COLUZZI,M., A. SABATINI,V. PETRARCAand M.A. DI DECO,1979 raises the possibility that portions of the nuclear ge- Chromosomal differentiation and adaptation to human environ- ments in the Anopheles gambiae complex. Trans. R. SOC.Trop. nome could betransferred as well. Thus, both ancestral Med. Hyg. 73: 483-487. polymorphism and more recent introgressive hybridiza- COLUZZI,M., V. PETRARCAand M. A. DI DECO,1985 Chromosomal tion challenge the interpretation of gene flow within inversion intergradation and incipient speciation in Anopheles gumbiue. Bull. Zool. 52: 45-63. and between An. gambiaeand An. arabiensis. The rewards Cow,J., A. F. COCKBURNand S. E. MITCHELL,1993 Population dif- of this challenge lie not only in informationabout the ferentiation of the malaria vector Anopheles aquasalis using mito- nature of divergence at the population-species inter- chondrial DNA. J. Hered. 84: 248-253. Corn, J., S. E. MITCHELLand A. F. COCKBURN,1997 Mitochondrial face, but in information critical tothe successful appli- DNA variation within and between two species of neotropical cation and evaluation of future genetic control strate- anopheline mosquiotes (Diptera: Culicidae). J. Hered. 88: 98- gies that target these dangerous vectors. 107. CONSTANTINI,C., S.-G. LI, A. DELLATORRE,N. SAGNON,M. COLUZZI et al., 1996 Density, survival and dispersal of Anopheles gambiue We are grateful to A. CORNELfor providing specimens of the An. complex mosquitoes in a West African Sudan savanna village. arabienris Mananga colony from the South African Institute for Medl- Med. Vet. Entomol. 10: 203-219. cal Researchand to A. HIGHTOWERfor SAS support. KEITH CRANDALL CRANDALL,K. A., A.R. TEMPLETONand C. F. SING,1994 Intraspe- kindly calculated the probability of parsimonious haplotype connec- cific phylogenetics:problems and solutions, pp. 273-297 in Mod- tions for the estimated cladograms. We thank BRIANHOLLOWAY and els in Phylogay Recmtruction, edited by R W. SCOTLAND,D. J. staff of the NCID Biotechnology Core Facility for synthesizing the SIEBERTand D. M. WILLIAMS.Clarendon Press, Oxford. oligonucleotide primers. T.L. was supported by an American Society CREWSOYEN,A. E., V. KUMARand F. H. COLLINS,1993 Association for Microbiology Fellowship.This work was supported by a grant from of two esterase genes, a chromosomal inversion, and susceptibil- the UNDP/World Bank/WHO Special Programme for Research and ity to Plmmodium cynomolgc in the African malaria vectorAnopheles gambiue. Am. J. Trop. Med. Hyg. 49 341-347. Training in Tropical Diseases to N.J.B. and a grant from the JOHN DELLATORRE, A,,L. MERZAGORA,J. R. POWELLand M. COLUZZI,1997 D. and CATHERINE T. MACARTHUR Foundation to F.H.C. Selective introgression of paracentric inversions betweentwo sibling species ofthe Anopheles gambiuecomplex. Genetics146: 239- 244. LITERATURE CITED ELLER,E., and H. C. HARPENDING,1996 Simulations show that neither population expansion nor population smtionarity in a West ARNOLD, M. L., 1997 NaturalHybridiultion and Evolution. Oxford African population can be rejected. Mol. Biol. Evol. 13 1155- University Press, New York. 1157. AWE, J. C., 1994 Molecular Mak, Natural Histq andEvolution. FONTENILLE, D.,L. LOCHOUARN, N. DIAGNE,C. SOKHNA,J. J. LE- Chapman and Hall, New York. SON et al., 1997a High annual and seasonal variationsin malaria AVISE,J. C., J. E. NEIGELand J. ARNOLD,1984 Demographic influ- transmission by anopheline and vector species composition in ences on mitochondrial DNA lineage survivorship in animal pop Dielmo, a holoendemic area in Senegal. Am. J. Trop. Med. Hyg. ulations. J. Mol. Evol. 20: 99-105. 56 247-253. BW,C. B., D. MILLS HA" and F. H. COLLINS,1993 The mito- FONTENILLE,D.,L. LOCHOUARN,M. DIATTA,C. SOKHNA,I. DIA et chondrial genome of the mosquito Anopheles gambiae: DNA se- al., 1997b A 4year entomological study of seasonal malaria in quence, genome organization, and comparisons with mitochon- Ndiop, a mesoendemic area in Senegal: large annual variations drial sequences of other insects. Insect Mol. Biol. 2 103-124. of transmission, close bionomics of Anopheles gambiue and An. BESANSKY,N. J., J. R. POWELL,A. CACCONE,D. M. HA", J. A. Scorn et arabiasis. Trans. Royal SOC.Trop. Med. Hyg. (in press). al., 1994 Molecular phylogeny ofthe Anopheles gambiue complex FREITAS-SIBAJEV,M. G. R., J. CONN,S. E. MITCHELL,A. F. COCKBURN, suggests genetic introgression between principal malaria vectors. J. A.SEAWRIGHT et ab, 1995 Mitochondrial DNA and morpho- Proc. Natl. Acad. Sci. USA 91: 6885-6888. logical analyses of Anopheles durlingi (Diptera: Culicidae) popula- BRAACK,L. E. O., M. COETZEE,R. H. HUNT,H. BIGGS,A. CORNELet tions from Brazil. Mosq. Syst. 27: 78-99. al., 1994 Biting pattern and host-seeking behavior of Anopheles Fu, Y.-X., 1994 A phylogenetic estimator of effective population size arabiensis (Diptera: Culicidae) in northeastern South Africa. J. or mutation rate. Genetics 136: 685-692. Med. Entomol. 31: 333-339. Fu, Y.-X., and W.-H. LI, 1994 Statistical tests of neutrality of muta- BROWER,A.V. Z., 1994 Rapid morphological radiation and conver- tions. Genetics 133: 693-709. gence among races of the butterfly Heliconius erato inferred from GARC~,B. A., A. CACCONE, K D. MATHIOPOULOS and J. R. POWELL, patterns of mitochondrial DNA evolution. Proc. Natl. Acad. Sci. 1996 Inversion monophyly in African anopheline malaria vec- USA 91: 6491-6495. tors. Genetics 143 1313-1320. CACCONE,A., B. A. GARCIAand J. R. POWELL,1996 Evolutionof GILLIES,M. T., 1961 Studies on the dispersion and survival of Anoph- the mitochondrial DNA control region in the Anopheles gambiae eles gambiae Giles in East Africa,by means of marking and release complex. Insect Mol. Biol. 5 51-59. experiments. Bull. Entomol. Res. 52 99-127. ~TELLOE,J., and A.R. TEMPLETON,1994 Root probabilities for GILLIES,M. T., and B.DE MEILLON,1968 The Anophelinue ofAfiZca intraspecific gene trees under neutral coalescent theory. Mol. south ofthe Sahara. Publications of the South African Institute for Phylogenet. Evol. 3 102-113. Medical Research No. 54, Johannesburg. 1828 N. J. Besansky et al.

HARRISON,R G., 1993 Hybrid Zones and the Evolutionaty Process. Ox- chondrial DNA restriction fragment length polymorphisms. Bie ford University Press, New York. chem. Genet. 31: 97-112. HOLSTEIN,M. H., 1954 Biology of Anopheles gadae. World Health NEI, M., 1987 Molecular Evolutionaty Genetia. Columbia University Organization Monograph Series No. 9, Geneva. Press, New York. HUDSON,R R, M. SLATKINand W. P. MADDISON,1992a Estimation NEIGEL,J. E., and J. C. AVISE, 1993 Application of a random walk of levels of gene flow from DNA sequence data. Genetics 132: model to geographic distributions of animal mitochondrial DNA 583-589. variation. Genetics 135 1209-1220. PAPE,T., 1992 Cladistic analyses of mosquito chromosome data in HUDSON,R. R., D. D. Boos and N.L. KAPLAN,1992b A statistical Anopheles subgenus Cellia (Diptera: Culicidae). Mosquito System- test for detecting geographic subdivision. Mol. Biol. Evol.9: 138- atics 24: 1-11. 151. PERERA,0. P., S. E. MITCHELL,A. F. COCKSURNand J. A. SEAWRIGHT, HUNT,R. H., 1987 Location of genes on chromosome arms in the 1995 Variation in mitochondrial and ribosomal DNA of An@& Anopheles gambiae group of species and their correlation to link- les quadrimaculatur species A (Diptera: Culicidae) across a wide age data for other anopheline mosquitoes. Med. Vet. Entomol. geographic range. Ann. Entomol. SOC.Am. 88: 836-845. 1: 81-88. ROGERS,A. R., and H. HARPENDING,1992 Population growth makes KAMAu, L., W. A. HAWLEY,T. LEHMANN,A. S. S. ORAGO,A. CORNEL waves in the distribution of paiMiise genetic differences. Mol. etaZ., 1997 Use of short tandem repeats for analysis of genetic Biol. Evol. 9: 552-569. variability in sympatric populations of Anopheles gambiae and ROZAS,J., and R. ROZAS,1997 DnaSP version 2.0: a novel software Anopheles arabiensis. Heredity (in press). package for extensive molecular population genetics analysis. KUMAR,S., IC TAMURAand M. NEI, 1993 MEGA: MolecularEvolution- Comput. Applic. Biosci. 13 307-311. a7y Genetics Analysis, version 1.01.The Pennsylvania State Univer- SAS INSTITUTEINC., 1990 SAS Language: Refmces. Version 6, 1st sity, University Park. edition. Cary, NC. LEHMANN,T., W. A. HAWLEY,L. KAMAu, D. FONTENILLE,F. SIMARD et SCOTT,J. A,, W. G. BROGDONand F. H. COLLINS,1993 Identification al., 1996 Genetic differentiation of Anopheles gambiae popula- of single specimens of the Anopheles gambiae complex by the tions from East and West Africa: comparison of microsatellite polymerase chain reaction. Am. J. Trop. Med. Hyg. 49 520-529. SIMONSEN,K L., G.A. CHURCHILLand C. F. AQUADRO,1995 Proper- and allozyme loci. Heredity 77: 192-208. ties of statistical testsof neutrality for DNA polymorphism data. LEHMANN,T., N. J. BESANSKY,W. A. HAWLEY,T. G. FAHEY, L.KAMAu Genetics 141: 413-429. etaZ., 1997 Microgeographic structure of Anopheles gambiae SLATKIN, M.,1987 Gene flow and the geographical structure of pop in western Kenya based on mtDNA and microsatellite loci. Mol. ulations. Science 236 787-792. EcoI. 6: 243-253. SMTKIN,M., and R. R HUDSON,1991 Pairwise comparisons of mito- LEMASSON,J. J., D. FONTENILLE,L. LOCHOUARN,I. DIA, F. SIMARD chondrial DNA sequences in stable and exponentially growing et al., 1997 Comparison of behaviour and vector efficiency of populations. Genetics 129 555-562. Anopheles gambiae and An. arabiensis (Diptera: Culicidae) in SOPER, F. L.,and D. B. WILSON,1943 Anopheles gambiae in Brazil 1930 Barkedji, a Sahelian area of Senegal.J. Med. Entomol. 34 396- to 1940. The Rockefeller Foundation, New York. 403. SWOFFORD,D. L., and R. B. SEIANDER,1989 BIOSYS-I. AComputer WON,D. J., R. IC BUTLINand P. GACESA,1995 An unusual mito- F’rogram for the Analysis ojAllelic Variation in Population Genetics and chondrial DNA polymorphism in the Chorthippw biguttulus spe- Biochemical Systaatics. Release 1.7. University of Illinois,Urbana. cies group (Orthoptera: Acrididae). Mol. Ecol. 4 121-126. TAJIMA, F.,1989 Statistical method for testing the neutral mutation MATHIOPOULOS,K. D., J. R. POWELLand T. F. MCCUTCHAN,1995 An hypothesis by DNA polymorphism. Genetics 123: 585-595. anchored restriction-mapping approach applied to the genetic TAJIMA,F., 1993 Measurement ofDNA polymorphism, pp. 37-59 in Mechanisms Molecular Evolution, edited by N. TAKAHATAand analysis of the Anopheles gambiae malaria vector complex. Mol. of A. G. CLARK.Sinauer Associates, Sunderland, MA. BIOI Ed. 12: 103-112. TAYLOR, C. E.,Y.T. TOURE,M. COLUZZIandV. PETRARCA,1993 Effec- MCLAIN,D. R, F. H. COLLINS,A. DAVID BRANDLINGBENNETTand tive population size and persistence of Anopheles arabiensis during J. B. 0. WERE,1989 Microgeographic variation in rDNA in- the dry season in West Africa. Med. Vet. Entomol. 7: 351-357. tergenic spacers of Anopheles gambiae in western Kenya. Heredity TEMPLETON,A. R., 1993 The “Eve” hypothesis: a genetic critique 62 257-264. and reanalysis. Am. Anthropol. 95: 51-72. MILES,S. J., 1978 Enzyme variation in the Anopheles gambiae Giles TEMPLETON,A. R., K. A. CRANDALLand C. F. SING,1992 A cladistic group of species (Diptera: Culicidae). Bull. Ent. Res. 68: 85-96. analysis of phenotypic associations with haplotypes inferred from MITCHELL,S. E., S. K NARANG,A. F. COCKBURN,J. A. SEAWRIGHTand restriction endonuclease mapping and DNA sequence data. 111. M. GOLDENTHAL,1992 Mitochondrial and ribosomal DNA vari- Cladogram estimation. Genetics 132: 619-633. ation among members of the Anopheles quadrimaculatus (Diptera: TOUR^, Y. T., V. PETRARCA,S. F. TRAORCA. COULIBALY,H. M. MAtu Culicidae) species complex. Genome 35: 939-950. et al., 1994 Ecological genetic studies in the chromosomal form NACHMAN,M. W., S. N. BOYER,J. B. SEARLEand C. F. AQUADRO,1994 Mopti of Anopheles gambiae s.str. in Mali, West Africa. Genetica Mitochondrial DNA variation and the evolution of Robertsonian 94: 213-223. chromosomal races of house mice, Mus domesticus. Genetics 136: YASOTHRONSRIKUL,S., S. PANYIMand R. ROSENBERG,1988 Diagnos- 1105-1120. tic restriction fragment patterns of DNA from the isomorphic species of Anopheles dims. Southeast Asian J. Trop. Med. Public NARANG,S. T. A. KLEIN,0. P. PERERA,J. B. LIMAand A. T. TANG, IC, Health 19: 703-708. 1993 Genetic evidence for the existence of cryptic species in the Anopheles albitarsus complex in Brazil: allozymes and mito- Communicating editor: J. HEY