Comparisons of Genetic Variability andGenome Structure Among MosquitoStrains Selected for Refractoriness to aMalaria ParasiteG. Yan, B. M. Christensen, and D. W. Severson

From the Department of Animal Health and BlomedlcalSdences, University of Wisconsin, 1655 Linden Drive,Madison, Wl 53706. We thank L Christensen, L Cul-berson-Smlth, V. Kassner, and Y. Zhang for technicalassistance. M. Ferdlg, C. Lowenberger, and D. Zaltllnprovided comments on the manuscript. This work wassupported by a National Research Service Award no.T32 (NIH grant AI07414) to G.Y., the Burroughs Well-come Fund to B.M.C., and NIH grant A133127 to D.WS.

Journal of Heredity 1997^8:187-194; 0022-1503/97/J5.00

Restriction fragment length polymorphism (RFLP) markers were used to evaluateAedes aegyptl genome structure and genetic variability within and between sub-strains selected for different levels of refractoriness to the malaria parasite, Plas-modlum galllnaceum. The MOYO-R substrain was previously selected for completerefractoriness and the MOYO-IS substrain for intermediate susceptibility from theMoyo-ln-Ory (MOYO) strain by selective Inbreeding (F = 0.5). Eighteen mappedRFLP markers were used to provide coverage of the mosquito genome. The twosubstralns showed reduced genetic diversity compared with the MOYO strain, in-cluding significant reductions In mean heterozygosity, number of alleles per locus,and proportion of polymorphic loci. Genetic differentiation between the two sub-stralns was statistically significant, as reflected by differences in allele frequencies.Significant palrwise linkage disequilibrium among the RFLP loci was detected inall three strains, most evidently in the MOYO strain. This is surprising because theRFLP loci examined are separated by large map distances, and therefore linkagedisequilibrium should decay to zero after many generations of laboratory culture.Our hypothesis to explain this phenomena is that lack of recombination, or lowrecombination rates in some regions of the A. aegyptl genome, is a result of chro-mosome Inversions. Finally, we used graphical genotyping, wherein whole genomegenotypic information for individual mosquitoes Is represented in a simple graphicformat, to illustrate genome structure and allellc variation within and among themosquito strains. Our analysis revealed an apparent chromosomal deletion onchromosome 3 for some individuals in the MOYO strain and MOYO-IS substrain.

Malaria, one of the most important para-sitic diseases today, infects about 400 mil-lion people worldwide and results inabout 2 million deaths per year, primarilyin children from tropical Africa (WorldHealth Organization 1995). The emergenceof pesticide resistance in mosquito vec-tors and antimalarial drug resistance inPlasmodium has significantly limited ma-laria control programs. Novel controlstrategies based on genetic disruption ofmosquito vector competence have beenproposed (Collins and Besansky 1994;James 1992). Knowledge of genome com-plexity and structure is essential for thesuccess of mosquito genetic manipulation.The development of DNA-based geneticmarkers provides the technology to elu-cidate the genetic mechanisms of vectorcompetence and to evaluate mosquito ge-nome structure at the molecular level(Severson 1994).

The Aedes aegypti-Plasmodium gallina-ceum system is an excellent model systemto investigate the genetic basis of mosqui-

to vector competence (Kllama and Craig1969; Severson et al. 1995b; Thathy et al.1994; Ward 1963). This host-parasite sys-tem is relatively easy to manipulate in thelaboratory, and a wealth of genetic infor-mation is available for the host species. P.gallinaceum also shares a close phyloge-netlc relationship with P. falciparum, themost serious malaria parasite infecting hu-mans (Waters et al. 1993). In addition,chromosomal conservation among mos-quito species both within and across sub-families has been suggested by severalstudies, using either isozyme markers(Matthews and Munstermann 1994) orDNA markers (Kumar and Rai 1993; Sev-erson et al. 1994b); therefore, rapid ad-vances in the molecular genetics of A. ae-gypti could facilitate our understanding ofthe genetic relationship between P. falcip-arum and its Anopheles vectors. A geneticlinkage map based on restriction fragmentlength polymorphism (RFLP) markers hasbeen constructed for A. aegypti (Seversonet al. 1993). Using these RFLP markers,

187

4

•p"IO

"3cro£oo

1 -

0.8-

0.6-

0.4-

0.2-

0-

n -\xr\\fr\ n

m MOYO-IS

• MOYO

w%,

11•p• ' / / / / ; ,

sr/f*/• 1 1iiHL_==Jlfi8B^_J_f=ML0 1-10 11-50

Number of oocysts per midgutFigure 1. Oocyst distribution among the three Aedes aegypti strains exposed to Ptasmodium gallinaceum. Samplesize Is 102 female Individuals for the MOYO strain, 177 for the substraln selected for refractoriness (MOYO-R),and 124 for the substraln selected for Intermediate susceptibility (M0YO-1S).

two putative quantitative trait loci (QTL)that significantly affect the susceptibilityof A. aegypti to P. gallinaceum subsequent-ly were identified (Severson et al. 1995b).Successful development of a physical mapfor A. aegypti (Brown et al. 1995) and con-struction of saturated linkage maps forQTL genome regions should facilitatemap-based cloning of the genes conferringrefractoriness and/or susceptibility to P.gallinaceum.

The mosquito genome consists of singleor low-copy DNA sequences and repetitiveDNA with short-period interspersion(Black and Ral 1988). Intraspecific and in-terspecific variation in abundance and dis-tribution of repetitive DNA has receivedconsiderable attention (Black and Rai1988; Kumar and Rai 1991; McLain et al.1986, 1987). In this study we focused onallelic variations of single- or low-copyDNA sequences (i.e., RFLP markers devel-oped from cDNAs) among A. aegyptistrains and then used these markers to ex-amine mosquito genome structure. UnlikeDrosophila and Anopheles species, thelack of well-developed polytene chromo-somes in A. aegypti obstructs detection ofchromosomal inversions, deletions, dupli-cations, and translocations (Sharma et al.1978). Information on genome organiza-tion and complexity is valuable for our un-derstanding of the population genetics ofgenes conferring refractoriness. For ex-ample, significant linkage disequilibrium(defined as nonrandom association be-

tween alleles of different genes) betweenrefractory genes and other genes couldwell affect the behavior of the refractorygenes in a population if other genes arealso subject to natural selection or artifi-cial selection (e.g., insecticides).

The use of RFLP markers to genotypeindividual mosquitoes within a populationat many loci distributed across the entiregenome may provide indirect evidencesupporting putative QTL locations and re-veal potential chromosomal abnormali-ties. These genotypic data can be repre-sented in a format termed graphical gen-otyping (Severson and Kassner 1995;Young and Tanksley 1989), and thereforewhole-genome genotypic information forindividual mosquitoes can be easily visu-alized. In this study we used graphicalgenotyping to illustrate allelic variationalong the whole genome in two A. aegyptisubstrains selected for different levels ofsusceptibility to P. gallinaceum, and to ex-amine mosquito genome structuralchanges associated with selective inbreed-ing. We also identified a chromosomal de-letion at one locus and found significantlinkage disequilibrium among several lociin these laboratory mosquito strains.

Materials and Methods

Mosquito StrainsTwo substrains (MOYO-R and M0Y0-1S)and a laboratory stock strain (Moyo-ln-Dry; MOYO) of A. aegypti were used in this

study. The MOYO strain was originally col-lected from Shauri Moyo Village, Momba-sa, Kenya, in 1974. We estimate that theMOYO strain has been maintained in thelaboratory for more than 150 generations.The MOYO-R and MOYO-IS were selectedfor complete refractoriness to P. gallina-ceum (i.e., number of oocysts per mosqui-to midgut = 0) and Intermediate suscep-tibility (number of oocysts per midgutranges from 0-10) from the MOYO strainby four generations of recurrent sib mat-ing (see Thathy et al. 1994). Therefore thetwo substrains have a theoretical inbreed-ing coefficient (F) of 0.5 (Crow and Kimura1970). Details for the selection of thesesubstrains are provided by Thathy et al.(1994). Mosquitoes were reared as previ-ously described (Christensen and Suther-land 1984). Following selection, each sub-strain has been maintained as a randommating colony with a minimum populationsize of 800. We examined generation 31 fol-lowing initial selection from the MOYOstrain. The relative susceptibilities of thethree strains to P. gallinaceum were peri-odically examined using the method de-scribed by Thathy et al. (1994), and allstrains received the same exposure to theparasite. Prevalence and mean infectionintensities were calculated based on six in-dependent samples.

RFLP and Probe SelectionDNA extraction from individual mosqui-toes, digestion with £coRI, Southern blot-ting, and hybridization were as previouslydescribed (Severson and Kassner 1995;Severson et al. 1993). Eighteen mappedRFLP markers (Mosquito Genomics 1996)were selected to provide coverage for theentire A. aegypti genome, with an averageresolution of 8.8 cM (Figure 1). All clonesused were random cDNA clones, exceptthat Mad is a known gene (James et al.1989). We examined a total of 50 femalesand 50 males each for MOYO and MOYO-IS, and 88 females and 74 males for MOYO-R.

Data AnalysisAmong-strain variation in infection inten-sity was analyzed using the Wilcoxonrank-sum test, and prevalence was ana-lyzed using the G test (Sokal and Rohlf1981). Infection intensity is defined as thenumber of oocysts per infected individual,and prevalence as the percentage of in-fected individuals. RFLP genotype datawere analyzed using the GENEPOP com-puter program (Raymond and Rousset1995). Analyses included computation ofallele frequencies, observed and expected

188 The Journal of Heredity 1997:88(3)

Table 1. Allele frequency, observed (H^J heterozygosity, and Fa estimates revealed by RFLP in Aedesaegypti strains susceptible (MOYO), refractory (MOYO-R), and Intermediate In susceptibility (MOYO-IS)to Piasmodiam galllnaceum

AllelesChromo-some Locus

1 LF9O

LF230

LF198

LF178

TY7

2 ARC1

LF138

LF124

LF282

LF98

LF250

LF115

3 LF352

LF261

LF168

Mal\

LF347

Strain*

MOYOMOYO-RMOYO-ISMOYOMOYO-R*MOYO-ISMOYOMOYO-RMOYCMSMOYOMOYO-RM0Y04S*MOYOMOYO-RMOYO-IS

MOYOMOYO-RMOYO-ISMOYOMOYO-RMOYCMSMOYO"MOYO-RMOYO-iSMOYOMOYO-RMOYO-ISMOYOMOYO-RMOYCMSMOYOMOYO-RMOYO-ISMOYO*MOYO-RMOYCMS

MOYOMOYO-RMOYO-ISMOYOMOYO-RMOYCMSMOYOMOYO-RMOYCMSMOYO***MOYO-RMOYCMSMOYOMOYO-RMOYCMS

•

0.3650.78003380.7780.3320.8000.3100.6370.6060.5050.3910.3740.275

—

0.106

0.4131.0000.7320.191

0.2780.6850.7600.3790.1100.3270.1050.2580.4390.0810.6000.6610.105

0.3650.6971.0000.192

0.0050.3920.4600.1051.0001.0000.2470.4130.505

2

0.6150.2200.6620.2220.6680.2000.4100.3630.3940.1450.6090 6260.7251.0001000

0.3941.00010000.582

0.2680.0571.0000.5760.0300.2400.4240.4100.6730.8950.6920.5610.9190.3750.3350.895

0.3600.303

0.50510001.0000 5250.6080.5400.395

0.4850.5870.495

3

0.020

0.280

0.205

0.343

0.005

0.340

0.230

0.1970.480

0.051

0.0250.003—

0.200

0.263

0460

0.450

0.134

—

4

0.115

0.157

0 412

0.1460.010

0.065

0.040

0.010

0.050

0.088

—

5

0.030——

0.005

—

0.010

0.046

—

6

0.040

w-0.5100.3790.4950.3230.3630.3000.6800.4770.5250.6500.4740.5860.3500.0000.000

0.7170.0000.0000.4690.0000.4140.5360.0000.5760.4500.4250 6770.5800 4320.2100 4950.4810.1620.4000.4370.170

0.6700.4780.0000.6260.0000.0000.6000.4260.5560.4400.0000.0000.5880.4270.424

V

-0.045-0.105-0.105

0.0640.182*0.062

-0.035-0.032-0.098

0.0270.004

-0.2520.123

—

-0.038—

0.047

-0.0560.206**

-0.0110.053

-0.164-0.061

0.0150.018

-0.117-0.095

0.024-0.087

0.1980.0310.096

0.033-0.133

0.017

—-0.170

0.107-0.119

0.299***

0.1300.1200.150

* P < .05; ** P < .01; *** P < .001.• Strain marked with asterisks Indicates that its allele frequencies at that locus do not conlorm to Hardy-Welnberg

expectations.*FB was computed as Weir and Cockerham (1984). A significant, positive Fa estimate indicates heterozygote defi-

ciency; negative Fa indicates excess of heterozygosity.

heterozygosity per locus, and a test forconformance with the Hardy-Welnbergequilibrium at each locus, using the exactHardy-Weinberg test (Guo and Thompson1992; Haldane 1954; Weir 1990). We furthertested whether distortion from Hardy-Weinberg equilibrium resulted from defi-cient or excessive heterozygosity (F^), us-ing the method described by Rousset andRaymond (1995). FG is denned as [1 - (ob-served heterozygosity/expected heterozy-gosity)] (Weir and Cockerham 1984). The

among-strain variation in heterozygositywas analyzed following the method of Weir(1990). Briefly, individuals were scored ateach locus as heterozygous or homozy-gous with a 1 or 0, respectively. Thesedata were analyzed with ANOVA usingstrains, individuals, loci, and interactionsof loci and individuals as factors. All fac-tors were treated as random effects exceptloci, because the same loci were repeat-edly scored (Weir 1990). The denomina-tors for each effect in the model were syn-

thesized by the SAS, JMP statistical pro-gram from linear combinations of the ap-propriate mean squares and have thesame expectation as the effect to be testedunder the null hypothesis. The degrees offreedom were adjusted for imbalance inthe experiment by Satterthwaite's method(see SAS 1994).

Genetic differentiation between MOYO-R and MOYCMS was estimated withWright's F statistics, based on the methodof Weir and Cockerham (1984), and usingthe FSTAT computer software (Goudet1995). The significance of the F statisticswas tested by jackknifing 1000 replicatesper locus (Goudet 1995; Weir 1990).

Because multiple alleles were found forsome loci, the normalized total linkage dis-equilibrium D' for each pair of loci wascalculated (Hedrick 1987; Lewinton andKojima 1960). D' for locus A with k allelesand locus B with / alleles is

k I

D' = £ 2P.<7/IA//A»J. (1)

where Dtj = Pu - p,<7,; £>„,„ = m i n O ^ , (1- Pt)(l ~ q}] when £„ < 0, or D^ . ^,(1— QJ)> (1 ~ PL)QJ] when £><, > 0; p, and qt arethe frequencies of alleles Ai and Bj at lociA and B. P^ is the frequency of a gametewith alleles Ai and Bj. D^ was estimatedfrom the genotypic data using the maxi-mum likelihood method (Weir 1990). Sig-nificance of the genotypic disequilibriumestimate was tested by chi-square statis-tics using the LJNKDIS computer programprovided by Garnler-Gere and Dillmann(1992). D' values range from 0 to 1 and areindependent of allele frequency (Hedrick1987). Therefore comparison of D' amongstrains or pairs of loci was possible. Weperformed linear regression between D'and map distance to test for the distanceeffect. RFLP data for the LF227 locus wereexcluded from the above analysis becausea chromosomal deletion at this locus wasdetected in MOYO and MOYO-IS (see be-low), and consequently the heterozygotegenotype (e.g., Ad) could not be differen-tiated from the homozygote genotype(e.g., AA).

Finally, graphical genotypes for all indi-viduals were prepared as described byYoung and Tanksley (1989), and graphs for15 individuals randomly selected fromeach of the three strains are presented.

ResultsVariation in Mosquito Susceptibility toParasitismThe relative susceptibility of the three Aaegypti strains to P. gallinaceum is shown

Yan et al • Genetic Variability of Mosquitoes 189

Table 2. Proportion of polymorphic l o d and population mean heterorygoslty (fl^J for the three Aedesaegypti (trains

Chromosome

1

2

3"

Overallloci

Numberof RFLPlociexam-ined

5

7

5

17

MOYO (n = 100)

Num-berofallelesperlocus

3.0

37

4.4

3.7

Propor-tion ofpoly-morph-Icloci

1.00

1.00

1.00

1.00

4*.(SD)

0503(0.165)0521(0.104)0.585(0.087)0.534(0.118)

MOYO-R (n = 1

Num-berofallelesperlocus

1.8

1.7

1.6

1.7

Propor-tion ofP°ly-morph-Iclocl

0.80

0.57

0.60

0.65

162)

/?«.(SD)

0339(0.196)0.254(0.238)0.266(0.244)0.282(0.217)

MOYO-iS (n =

Num-berofallelesperlocus

1.8

2.0

1.4

1.8

Propor-tion ofpoty-morph-Iclocl

0.80

0.86

0.40

0 71

100)

4*(SD)

0.381(0.238)0.316(0.246)0.196(0 272)0.299(0.247)

• RFLP marker LF227 was not included In the analysis. See text for details.

(Table 1), the proportion of polymorphicloci, and the observed average heterozy-goslty (Table 2). A total of 63 unique al-leles were identified for the 17 RFLP locianalyzed. Only 27 alleles (43%) were com-mon to the three strains, and 30 alleleswere unique to the MOYO strain (Table 1).An average of 3.7 alleles per locus was ob-served in the MOYO strain, but the MOYO-R and MOYO-IS substrains had only an av-erage of 1.7 and 1.8 alleles per locus, re-spectively (Table 2). The genotypes atseveral loci were in Hardy-Weinberg dis-equilibrium, apparently due to heterozy-gote deficiencies, because significant pos-itive Fa was found at these loci (Table 1).

in Figure 1. The MOYO strain showed aprevalence of 82.3%, significantly greaterthan either the MOYO-IS (45.9%; G = 33.1,df = 1, P < .001) or MOYO-R (11.3%; G =148.6, df = 1, P < .001) substrains. Themean intensity of the MOYO strain (16.1oocysts/midgut) was also significantlyhigher than MOYO-IS (8.6; x2 = 4.7, df =1, P < .05) or MOYO-R (3.8; X

2 = 13.2, df= 1, P < .001). The mean infection inten-sity and prevalence in MOYO-IS were sig-nificantly greater than the MOYO-R (x2 =5.0, df = 1, P < .05 for intensity; G = 33.1,df = 1, P < .001 for prevalence). The with-in-strain variation in infection intensitywas considerably higher in the MOYOstrain (SD = 20.6, n = 84) than for MOYO-R (SD = 3.7, n = 20) or MOYO-IS (SD =10.2, n = 57; Figure 1).

Genetic VariationWithin- and among-strain genetic variabil-ity can be quantified by allele frequencies

Table 3. S ta t i s t i c s analysis of polymorphic locifor the MOYO-R and MOYO-IS substrains

Chromo-some Locus F.' Fn F~

LF90LF230LF198LF178

LF282LF98LF250LF115

LF168LF347

- 0 1000.146*

-0.055-0.090

0.024-0.006

0.0100.047

0.0230.137*

0.268*0.449*

-0.057-0.094

0.203*0.117*0.268*0.495*

0.0280.147*

0.334**0.354**

-0.002-0.003

0.184"0.122**0.260**0.470**

0.0050.012*

" fB - fixation Indices of individuals relative to the total sub-populations; Fn = fixation indices of individuals relative tothe total populations. It takes into account both the effectsof nonrandom mating within subpopulations (F£ and theeffects of population subdivision (F^ Fs = coefficient ofpopulation differentiation, and It measures the between-pop-idabon component of standard goieOc variance.

*/»= 0.052.

* P < .05; ** P < .01; ' " P< .001.

Table 4. Palrwlse total linkage disequilibrium (DO among the three Aedes aegypti strains for the RFLP lod

LF90 LF230 LF198 LF178Chromo-some 1

MOYO

MOYO-R

MOYO-IS

Chromo-some 2

MOYO

MOYO-R

M0Y04S

Chromo-some 3

MOYO

MOYO-R

MOYO4S

LF230LF198LF178TY7LF230LF198LF178TY7LF230LF198LF178TY7

LF138LF124LF282LF98LF250LF115LF138LF124LF282LF98LF250LF115LF138LF124LF282LF98LF250LF115

LF261LF168ManLF347LF261LF168Mal\LF347LF261LF168ManLF347

0.2610.0490.4490.3890.479*0 0360.031

0.435*0.0540.245

ARC1

0.1960.2350.158

0.2750.095

0.644*0.641*

LF138

0.2730.121

0.055

0 3 0 8 "

LF124

0.223

LF282 LF98 LF25O

0.2040.1730.1810.1090.1720.143

0.236**0.087"0.1790.467"0.157

0.267*0.2490.2490.109

LF352

02290.12903160.2090.117

LF261

0.340"0.32303590.438

LF168

0.1790.194*0324

0.1470.0590.113

0.3970.629*0.634*

Man

0.1670.069*

0.252*0.012

0395*0.032

0.239*

0.240

0.113

0.152*0.0760.492*0.198

0.014

0.018

0.362"0549*0 2 4 1 "

0.119"0.182"

0.050

0.047

0.262*

' Indicates linkage disequilibrium estimate is significantly different from 0 at P < .05, ** P < .01, and * " P < .001.

• Linkage disequilibrium estimate is not available due to fixation of a locus In a population.

1 9 0 The Journal of Heredity 1997 88(3)

0 -p LF90

11.6 --LF230

19.5 --LF198

31.8 --LF178

44.0 -L TY7

7.7 -

20.6-

28 .2-

36.1 -39.4-

49 .9 -

59 .9 -

- ARC1

- LF138

- LF124

- L F j ^" LF98

- LF250

- LF115

8.6 - - LF352

16.4 - - LF261

25.2 - - LF168

39.0 - -35 3 _ _ Mai I

pgsP, MaimLF227

56.5 - L LF347

Figure 2. Relative map position of the 18 RFIP markers used In this study. Map distances are In KosambI cen-timorgans. The RFLP markers best defining the two quantitative trait loci for the susceptibility of Aedes aegypti toPtasmodium gallmaceum are Indicated (pgs[2, LF98] and pgs[3, Mall]; Severson et al. 1995b)

All loci were polymorphic in the MOYOstrain, but 35% and 29% of the loci werefixed in the MOYO-R and MOYO-IS sub-strains, respectively. The mean heterozy-gosity of the MOYO-R and MOYO-IS sub-strains was significantly reduced by 47%and 44%, respectively, compared to theMOYO strain (Table 2; ANOVA, F = 10.38,df = 2, 32, P < .001). This is apparentlydue to fixation of alleles at loci on chro-mosomes 2 and 3 and the reduced numberof alleles at other loci. The reduction inmean heterozygosity in the two substralnswas in accordance with the theoretical ex-pectation of 50% (Crow and Klmura 1970).

Although no significant difference wasdetected in the mean heterozygosity be-tween MOYO-R and MOYO-IS (Table 2;ANOVA, F = 0.149, df = 1, 16, P > .05), thetwo substralns exhibited significant differ-ences in allele frequency at most of thepolymorphic loci (Table 1). Significant F^estimates at those loci provide furthersupport for these results (Table 3). Genet-ic differentiation between the MOYO sub-strains may have resulted from geneticdrift, or selection, or both.

Genotyplc Linkage DisequilibriumGenotypic linkage disequilibrium amongpairs of polymorphic loci was evaluatedfor each of the three strains. The resultsare presented in Table 4, with marker lociarranged in chromosomal order. Signifi-cant pairwise linkage disequilibrium wasdetected in each strain, and it was re-markably evident for the loci on chromo-

somes 2 and 3 of the MOYO strain (Table4). Significant linkage disequilibrium (£>')was not expected in any of these strains.In theory the decay of linkage disequilib-rium in t generations relative to its initialvalue can be estimated as (1 - r)' for arandom mating population, where r is therecombination frequency between a pairof loci (Hedrick 1987). Thus D' should de-cay to zero for pairs of loci with fairlylarge recombination frequencies (e.g., thesmallest recombination frequency, r,among all pairs of loci studied is 0.0612 forLF282-LF98; see Figure 2) after long-termlaboratory culture (e.g., / > 150 genera-tions for MOYO). To test for the effect ofrecombination on the disruption of allelicassociations among those RFLP loci, linearcorrelations between D' estimates andtheir corresponding map distances weredetermined for each of the three strains.We found significant correlations with theMOYO-R substrain, but not with MOYOand MOYO-IS (Figure 3), suggesting thatrecombination did not effectively breakdown nonrandom allelic associations inMOYO and MOYO-IS. Time-dependent de-cay in linkage disequilibrium probably oc-curred in MOYO-R. This suggestion wassupported by the data that fewer pairs ofloci were in linkage disequilibrium forMOYO-R and by generally smaller D' esti-mates (Table 4).

Graphical GenotyplngGraphical genotypes for 15 randomly se-lected individuals from each of the three

0.8

0.6

0.2

•c

••= 0 6

c 02-

0 6 -

02-

A MOYO: r2 = 0.002, df = 40. P > 0.05

B. MOYO-R: r2 = 0.277, df = 14, P < 0.05

C. MOYO-IS: r2 = 0.076, df = 21, P > 0.05

0 10 20 30 40 50 60

Map distance (cM)

Figure 3. Correlation between pairwise linkage dis-equilibrium (C) and map distance for the three Aedesaegypti strains. The U estimates were obtained fromTable 3 and plotted against the map distance obtainedfrom Figure 1. Statistically significant disequlllbria areshown as solid triangles and nonsignificant dlsequill-bria as open circles.

strains are shown in Figure 4. We foundthat a chromosomal deletion event oc-curred around the LF227 locus in 10.9% ofthe individuals in the MOYO strain and16.2% in the MOYO-IS substrain (Figure4C). Southern blots of £coRI-digested DNAfrom those Individuals showed no hybrid-ization to the LF227 clone, while hybrid-ization was observed with all individualswith all other markers examined (Figure4C). Therefore, absence of hybridizationto LF227 with those individuals was notdue to incomplete DNA digestion or poorprobe conditions.

We have previously Identified two RFLPmarkers, LF98 on chromosome 2 and Mallon chromosome 3, which best define theQTL for A aegypti susceptibility to P. gal-linaceum (Severson et al. 1995b). TheMOYO strain exhibited high polymor-phism with both markers, and most indi-viduals were heterozygotes (Figure 4).The high level of genetic diversity of thisstrain may reflect large variation in sus-ceptibility to the malaria parasite (Figure1). Marker Mall was fixed in the MOYO-Rand MOYO-IS substrains (Figure 4). Mostindividuals of the two substrains were ho-mozygous for the same allele at the LF98locus (46% for MOYO-R and 79% for

Yan et al • Genetic Variability of Mosquitoes 191

A. Chromosome 1

MOYO MOYO-R MOYO-IS

TY7

Q a l a l I a2a2 |Heterozygote I 5cM

MOYO-IS

B. Chromosome 2MOYO MOYO-R

ARCl

LF138

LF124

LF282 V<

LF98

LF250

LF115

Figure 4. Graphical genotyping for 15 Individual mosquitoes representing the MOYO, MOYO-R, and MOYO4Sstrains. For each panel each column represents the chromosomal organization for a single mosquito. Individualmarkers reflect the half-recombinatlonal distance between adjacent markers or between an adjacent marker andthe end of the linkage group.

MOYO-4S), although a small proportionwas heterozygous (43% for MOYO-R and21% for MOYO-IS) or homozygous for asecond allele (11% for the MOYO-R and 0%for M0YO4S substrains).

Discussion

This study examined genetic variationwithin and among A aegypti strains se-

lected for refractoriness to P. gallinaceum,and evaluated mosquito genome struc-ture, particularly in response to selection,using RFLP markers developed from A ae-gypti (Severson et al. 1993). A. aegypti pop-ulation genetic variation has previouslybeen studied with isozymes (Tabachnick1982; Tabachnick et al. 1985) and random-amplified polymorphic DNA (RAPD) mark-ers (Ballinger-Crabtree et al. 1992). Iso-

zyme and RAPD techniques have the ad-vantage of relative simplicity; however,RFLP markers offer several advantages.First, by using RFLP markers we were ableto detect high levels of polymorphism inthe three A aegypti strains (Table 2). Be-cause RFLP markers segregate as codom-inant markers, heterozygotes can bescored directly and population allele fre-quencies can be accurately estimated (Ta-ble 1), and therefore the RFLP markers arevaluable for studying population geneticstructure (Table 3). The RFLP markersused here detected significantly highermean heterozygosity in our A aegypti pop-ulations (see Table 2) than did isozymemarkers in other A. aegypti, A albopictus(Tabachnick et al. 1985), and Anophelesspp. (Estrada-Franco et al. 1993; Fritz et al.1995; Manguin et al. 1995) populations orother Diptera groups (Graur 1985). Sec-ond, although chromosomal inversionsand translocations undoubtedly occurredduring mosquito evolution (Matthews andMunstermann 1994), these authors andother studies (e.g., Severson et al. 1994b)suggest that the basic genome structuremay be largely conserved for mosquitospecies within the subfamilies Culicinaeand Anophelinae. Support for this sugges-tion of genome structure conservationwas provided when Severson et al. (1995a)demonstrated complete conservation oflinkage group and linear order in A albo-pictus for 18 RFLP markers developed fromA aegypti. For these species, the sameRFLP loci may be examined for geneticvariation for populations within a speciesor between these species. In addition, theRFLP markers provide an opportunity forsampling the whole mosquito genome be-cause their relative map position is known(Figure 2; see also Mosquito Genomics1996). Therefore RFLP technology pro-vides a powerful tool to study mosquitopopulation genetic variation. Finally, be-cause our RFLP markers are single- or low-copy cDNA sequences, it is possible to usethem to reveal genomic anomalies such aschromosomal duplications and deletions.Our data suggest a chromosomal deletionaround locus LF227 for some individualsof the MOYO and MOYO-IS strains (Figure4).

We observed significant linkage disequi-librium between loci separated by largemap distances (i.e., 7.6-40.1 cM; see Fig-ure 3). Severson and Kassner (1995) alsoreported similar results in other popula-tions of A aegypti. In Drosophila melano-gaster, very little linkage disequilibrium isobserved between genes that are more

192 Ttie Journal of Heredity 1997.88(3)

C. Chromosome 3

MOYO

LF352

LF261

LF168

Mall

LF227

LF347

MOYO-R MOYO-IS

I a 2 a 2 | l a 3 a 3 fflDeletion HHeterozygote 15 cM

Figure 4. Continued.

than 10 cM apart (Langley et al. 1977,1978). Linkage disequilibrium is more like-ly to be found for closely linked genes,usually in genome regions less than 100 kb(e.g., Aquadro et al. 1992; Langley et al.1988; Macpherson et al. 1990; Schaefferand Miller 1993; Smit-McBride et al. 1988),or for functionally related genes (e.g., Ba-ker 1975; Bech-Hansen et al. 1983; Van derLoo et al. 1987). In a large, random-matingpopulation, linkage disequilibrium shouldbe nearly zero without epistatlc selectionor other genetic mechanisms. If a popula-tion is founded by a small number of in-dividuals in which significant linkage dis-equilibrium exists between a given pair ofloci, this linkage disequilibrium should de-crease with each successive generation ata rate inversely proportional to the recom-bination rate between them. Therefore ourmosquito populations were expected tobe at linkage equilibrium at all RFLP lociexamined in this study.

Linkage disequilibrium may result fromvarious genetic mechanisms other thanepistatic selection, such as low recombi-nation rate, genetic drift, population sub-division, migration, and hitchhikingamong the linked genes (Hedrick et al.1978; Hill and Robertson 1968; Lewinton1974; Nei 1987; Ohta and Kimura 1969).The rationale for epistatic selection is thatnatural selection may favor particularcombinations of alleles at different loci(coadapted gene complex), therefore thefrequencies of these allelic combinations

in the gametes are higher than expectedfrom the random combinations of theirfrequencies (Smit-McBride et al. 1988).Population subdivision and migration arenot applicable to our mosquito popula-tions. Genetic drift resulted in significantgenetic differentiation between the MOYO-R and MOYO-IS substrains (see Table 3);however, it is not likely to be an importantmechanism for the observed linkage dis-equilibrium in these populations becausegenetic drift alone is not likely to createsubstantial linkage disequilibrium in apopulation with rapid growth (Slatkin1994). We also argue that epistatic selec-tion and genetic hitchhiking are not majormechanisms because (1) the RFLP mark-ers are random cDNA sequences whoseputative functions are clearly not related(Severson and Zhang 1996), and (2) in theMOYO strain, linkage disequilibrium wasdetected in 48% of the pairs of loci onchromosomes 2 and 3, although they arenot tightly linked (Table 4). We hypothe-size that the lack of recombination or lowrecombination rates in some regions ofthe A aegypti genome Is probably a resultof chromosomal inversions. Nonsignifi-cant correlation between linkage disequi-librium and map distance in some mos-quito strains supports this hypothesis(e.g., MOYO and MOYO-IS; see Figure 3).Examples of linkage disequilibrium result-ing from paracentric inversions have beendocumented in Drosophila (Prakash 1977;Prakash and Lewinton 1968; Voelker et al.

1978). It is well known that chromosomalinversion polymorphisms occur in themosquito An. gambiae (Coluzzi et al.1985), and the presence of naturally oc-curring inversions in A aegypti have beensuggested (Macdonald and Sheppard1965; Severson and Kassner 1995). Directcytologlc evidence or genetic assay withsingle pair matings is required for verifi-cation of this hypothesis.

Finally, we used graphical genotyping toevaluate the A aegypti genome. The re-duction in allelic variation resulting fromselective inbreeding in the MOYO-R andMOYO-IS substrains, including reducedheterozygosity, the proportion of poly-morphic loci, and loss of alleles, can beeasily visualized from Figure 4. Using thismethod, Severson and Kassner (1995)found that different alleles were fixed inloci across a large distance in two A ae-gypti strains refractory and susceptible tothe filarial worm Brugia malayi, includingchromosomal regions influencing filarialworm susceptibility (Severson et al.1994a) and intensity (Beerntsen et al.1995). The large variation in the suscepti-bility to P. gallinaceum in the MOYO strainis reflected by elevated genetic heteroge-neity at the markers flanking the QTL onchromosomes 2 and 3 (Figures 1 and 4).One allele for the marker Mall most close-ly linked to the QTL on chromosome 3 wasfixed in the MOYO-R and MOYO-IS sub-strains. The chromosomal region contain-ing the QTL on chromosome 2 was poly-morphic in the two substrains (Figure 4).In addition, these two substrains exhibitedallelic variations in other genome regions.Whether such allelic variations affect mos-quito fitness and vector competence re-mains unknown. Graphical genotyping,therefore, provided us a method to easilyreveal genetic changes due to past evolu-tionary history, examine within- andamong-population genetic variation, andformulate hypotheses about the geneticbasis of vector competence.

References

Aquadro CF, Jennings RMJ, Bland MM, Laurie CC, andLangley CH, 1992. Patterns ol naturally occurring re-striction map variation, dopa decarboxylase activityvariation and linkage disequilibrium In the Ddc generegion of Drosophila melanogasler. Genetics 132:443-452.

Baker WK, 1975. Linkage disequilibrium over space andtime In natural populations of Drosophila monlana.Proc Natl Acad Scl USA 72:4095-4099.

Balllnger-Crabtree ME, Black WC, and Miller BR, 1992.Use of genetic polymorphisms detected by the random-amplified polymorphic DNA porymerase chain reaction(RAPD-PCR) for differentiation and Identification of Ae-

Yan et al • Genetic Variability ot Mosquitoes 193

des aegypli subspecies and populations. Am J TropMed Hyg 47:893-901.

Bech-Hansen NT, Linsley PS, and Cox DW, 1983. Restric-tion fragment length polymorphisms associated withImmunoglobulin C genes reveal linkage disequilibriumand genomic organization. Proc Natl Acad Sci USA 80:6952-6956.

Beerntsen BT, Severson DW, Kllnkhammer JA, KassnerVA, and Christensen BM, 1995. Aedes aegyplr. a quan-titative trait locus (QTL) Influencing filarlal worm In-tensity Is linked to QTL for susceptibility to other mos-quito-borne pathogens. Exp Parasitol 81:355-362.

Black WC and Ral KS, 1988. Genome evolution in mos-quitoes: intraspeclfic and Interspecific variation In re-petitive DNA amounts and organization. Genet Res 51:185-196.

Brown SE, Mennlnger J, Difilllpantonio M, Beaty BJ,Ward DC, and Knudson DL, 1995. Toward a physicalmap of Aedes aegypli. Insect Mol Blol 4:161-167.

Christensen BM and Sutherland DR, 1984. Bntgia pa-hangi: exsheathment and midgut penetration in Aedesaegypli. Trans Am Microsc Soc 103:423-433.

Collins FH and Besansky NJ, 1994. Vector biology andthe control of malaria in Africa. Science 264:1874-1875.

Coluzzi M, Petrarca V, and Dl Deco MA, 1985. Chro-mosomal Inversion Integradation and incipient specla-tion In Anopheles gambiae. Boll Zool 52:45-63.

Crow JF and KJmura M, 1970. An Introduction to pop-ulation genetics theory. New York: Harper & Row.

Estrada-Franco JG, Ma MC, Gwadz RW, Sakal R, LanzaroGC, Laughinghouse A, Galvan-Sanchez C, Cespedes JL,and Vargas-Sagarnaga R, 1993. Evidence through cross-mating experiments of a species complex in Anophelespseudopunclipennis sensu lato: a primary malaria vec-tor of the American continent. Am J Trop Med Hyg 49:735-745.

Fritz GN, Bermudez H, and Seawright JA, 1995. Geneticdifferentiation and diagnostic loci of Anopheles nunez-tooari. An. tnnhae, and An. rangeli (Diptera. Cullcldae).J Med Entomol 32:663-672.

Garnler-Gere P and Dillmann C, 1992. A computer pro-gram for testing pairwise linkage dlsequlllbria In sub-divided populations. J Hered 83:239.

Goudet J, 1995. FSTAT version 12: a computer programto calculate f-statistlcs. J Hered 86:485-486.

Graur D, 1985. Gene diversity In Hymenoptera. Evolu-tion 39190-199.

Guo SW and Thompson EA, 1992. Performing the exacttest of Hardy-Welnberg proportions for multiple alleles.Biometrics 48:361-372.

Haldane JBS, 1954. An exact test for randomness ofmating. J Genet 52:631-635.

Hedrlck PW, 1987. Gametic disequilibrium measures:proceed with caution. Genetics 117:331-341.

Hedrick PW, Jain S, and Holden L, 1978. Multllocus sys-tems In evolution. Evol Blol 11:101-182.

Hill WG and Robertson A, 1968. Linkage disequilibriumIn finite populations. Theor Appl Genet 38:226-231.

James AA, 1992. Mosquito molecular genetics: thehands that feed bite back. Science 25737-38.

James AA, Blacbner K, and Racioppi JV, 1989. A sali-vary gland-specific maltase-llke gene of the vector mos-quito, Aedes aegypli. Gene 75:73-83.

Kllama WL and Craig GB Jr, 1969. Monofactortal Inher-itance of susceptibility to Ptasmodium gallinaceum inAedes aegypti. Ann Trop Med Parasitol 63:419-432.

Kumar A and Ral KS, 1991. Organization of a clonedrepetitive DNA fragment In mosquito genomes (Dip-tera: Culiddae). Genome 34598-1006.

Kumar A and Rai KS, 1993. Molecular organization andevolution of mosquito genome. Comp Biochem Physlol106B:495-504.

Langley CH, Ito K, and Voelker RA, 1977. Linkage dis-

equilibrium In natural populations of Drosophila mela-nogasler. seasonal variation. Genetics 86:447^)54.

Langley CH, Shrimpton AE, Yamazaki T, Mlyashlta N,Matsuo Y, and Aquadro CF, 1988. Naturally-occurringvariation in the restriction map of the Amy region ofDmsophila melanogaster. Genetics 119:619-629.

Langley CH, Smith DB, and Johnson FM, 1978 Analysisof linkage disequilibria between allozyme loci In natu-ral populations of Drosophila melanogaster. Genet Res32:215-230.

Lewinton RC, 1974. The genetics basis of evolutionarychange. New York: Columbia University Press.

Lewinton RC and Kojlma K-t, 1960. The evolutionarydynamics of complex polymorphisms. Evolution 14:116-129.

Macdonald WW and Sheppard PM, 1965. Cross-overvalues In the sex chromosomes of the mosquito Aedesaegypli and evidence of the presence of Inversions. AnnTrop Med Parasitol 59:74-87.

Macpherson JN, Weir BS, and Brown AJL, 1990. Exten-sive linkage disequilibrium In the achaete-scute com-plex of Drosophila melanogaster. Genetics 126:121-129.

Manguln S, Roberts DR, Peyton EL, Fernandez-Salas I,Barreto M, Loayza RF, Splnola RE, Granaou RM, andRodriguez MH, 1995. Biochemical systematic* and pop-ulation genetic structure of Anopheles pseudopunclipen-nis, vector of malaria In Central and South America. AmJ Trop Med Hyg 53:362-377.

Matthews TC and Munstermann LE, 1994. Chromosom-al repattemlng and linkage group conservation in mos-quito karyotyplc evolution. Evolution 48:146-154.

McLaln DK, Ral KS, and Fraser MJ, 1986. Interspecificvariation In the abundance of highly repeated DNA se-quences in the Aedes scutellaris (Diptera: Cullcldae)subgroup. Ann Entomol Soc Am 79:787-791.

McLaln DK, Rai KS, and Fraser MJ, 1987. Intraspecificand Interspecific variation in the sequence and abun-dance of highly repeated DNA among mosquitoes ofthe Aedes albopictus subgroup. Heredity (Edinburgh)58:373-381

Mosquito Genomlcs, 1996. The Aedes aegypti databaseavailable from the World Wide Web at URL http://klab agscl.colostate.edu.

Nei M, 1987. Molecular evolutionary genetics. NewYork; Columbia University Press.

Ohta T and Klmura M, 1969. Linkage disequilibrium ata steady state determined by random genetic drift andrecurrent mutation. Genetics 63:229-238.

Prakash S, 1977. Further studies on gene polymor-phism in the malnbody and geographically Isolatedpopulations of Drosophila pseudoobscura. Genetics 85:713-719.

Prakash S and Lewinton RC, 1978. A molecular ap-proach to the study of genlc heterozygosity ID. Directevidence of coadaptation In gene arrangements of Dro-sophila. Proc Natl Acad Sci USA 59-^98-405

Raymond M and Rousset F, 1995. GENEPOP (version1.2): a population genetics software for exact tests andecumenldsm. J Hered 86:248-249.

Rousset F and Raymond M, 1995. Testing heterozygoteexcess and deficiency. Genetics 140:1413-1419.

SAS, 1994 JMP* user's guide. Cary, North Carolina: SASInstitute.

Schaeffer SW and Miller EL, 1993. Estimates of linkagedisequilibrium and the recombination parameter de-termined from segregating nucleotlde sites In the al-cohol dehydrogenase region of Drosophila pseudoob-scura. Genetics 135541-552.

Severson DW, 1994. Applications of molecular markeranalysis to mosquito vector competence. Parasitol To-day 10-336-340.

Severson DW and Kassner VA, 1995. Analysis of mos-quito genome structure using graphical genotyplng. In-sect Mol Blol 4:279-286.

Severson DW, Mori A, Zhang Y, and Christensen BM,1993. Linkage map for Aedes aegypti using restrictionfragment length polymorphisms. J Hered 84:241-247.

Severson DW, Mori A, Zhang Y, and Christensen BM,1994a. Chromosomal mapping of two loci affecting fi-larial worm susceptibility In Aedes aegypti Insect MolBlol 3:67-72.

Severson DW, Mori A, Zhang Y, and Christensen BM,1994b. The suitability of RFLP markers for evaluatinggenetic diversity among and synteny between mosqui-to species. Am J Trop Med Hyg 50:425^132.

Severson DW, Mori A, Kassner VA, and Christensen BM,1995a. Comparative linkage maps for the mosquitoes,Aedes albopictus and Ae. aegypti, based on commonRFLP loci. Insect Mol Blol 4:41-15.

Severson DW, Thathy V, Mori A, Zhang Y, and Christen-sen BM, 1995b. Restriction fragment length polymor-phism mapping of quantitative trait loci for malariaparasite susceptibility In the mosquito Aedes aegypti.Genetics 139:1711-1717.

Severson DW and Zhang Y, 1996. Generation of ex-pressed sequence tags and sequence-tagged sites asphysical landmarks In the mosquito, Aedes aegypti, ge-nome Genome 39:224-229.

Sharma GP, Mittal OP, Chaudhry S, and Pal V, 1978. Apreliminary map of the salivary gland chromosomes ofAedes (Stegomyla) aegypli (Culicadea, Diptera). Cyto-blos 22:169-178.

Slatldn M, 1994. Linkage disequilibrium In growing andstable populations. Genetics 137:331-336.

Smit-McBride Z, Moya A and Ayala FJ, 1988 Linkagedisequilibrium In natural and experimental populationsof Drosophila melanogaster. Genetics 120:1043-1051.

Sokal RR and Rohlf FJ, 1981. Biometry. San Francisco:W. H. Freeman.

Tabachnlck WJ, 1982. Geographic and temporal pat-terns of genetic variation of Aedes aegypti In New Or-leans. Am J Trop Med Hyg 31:849-853.

Tabachnlck WJ, Wallls GP, Aitken TH, Miller BR, AmatoGD, Lorenz L, Powell JR, and Beaty BJ, 1985. Oral in-fection of Aedes aegypti with yellow fever virus: geo-graphic variation and genetic considerations. Am JTrop Med Hyg 34.1219-1224.

Thathy V, Severson DW, and Christensen BM, 1994. Rein-terpretauon of the genetics of susceptibility of Aedes ae-gypti to Ptasmodium gallinaceum. J Parasitol 80:705-712.

Van der Loo W, Arthur CP, Richardson BJ, Wallace-Drees M, and Hamers R, 1987. Non-random allele as-sociations between unlinked protein loci: are the poly-morphisms of the Immunoglobulin constant regionsadaptive? Proc Natl Acad Sci USA 84:3075-3079.

Voelker RA, Cockerham CC, Johnson FM, Schalfer HE,Mukal T, and Mettler LE, 1978. Inversions fail to ac-count for allozyme cllnes. Genetics 88515-527.

Ward RA, 1963 Genetic aspects of the susceptibility ofmosquitoes to malaria Infection. Exp Parasitol 13:328-341.

Waters AP, Hlggins DC, and McCutchan TF, 1993. Evo-lutionary relatedness of some primate models of Ptas-modium. Mol Bio! Evol 10514-923.

Weir BS, 1990. Genetic data analysis. Sunderland, Mas-sachusetts: Slnauer.

Weir BS and Cockerham CC, 1984. Estimating F-statis-tlcs for the analysis of population structure. Evolution38:1358-1370.

World Health Organization, 1995. World Health Report1995—executive summary. Available on the WorldWide Web at URL http://www.who.ch.

Young ND and Tanksley SD, 1989. Restriction fragmentlength polymorphism maps and the concept of graph-ical genotypes. Theor Appl Genet 7755-101.

Received March 1, 1996Accepted September 19, 1996

Corresponding Editor Ross Maclntyre

194 The Journal of Heredity 1997 88(3)