Copyright 0 1991 by the Society of America

Lack of Underdominance in a Naturally Occurring Pericentric Inversionin and Its Implications for Chromosome Evolution

Jerry A. Coyne,* Sylvie Aulardt and Andrew Berr? *Dekartment of Ecology and Evolution, The University of Chicago, Chicago, Illinois 60637, TLaboratoire de Biologie et Ginitique Evolutives, Centre Nationale de la Recherche Scient$que 91 198Gq-sur-Yvette Cedex, France, and Service de Ginitique, Universiti de P. et M. Curie, Paris, France, and ?Department of Ecology and Evolutionary Biology, Princeton University, Princeton, NewJersey 08544 Manuscript received April 22, 1991 Accepted for publication August 2, 1991

ABSTRACT In(2LR)PL is a large pericentric inversion polymorphicin populations of Drosophila melanogaster on two Indian Ocean islands.This polymorphism is puzzling: because crossing over in female heterokar- yotypes produces inviable zygotes, such inversions are thought to be underdominant and should be quickly eliminated from populations.The observed fixation forsuch inversions among related species has led to the idea that geneticdrift can cause chromosome evolutionin opposition to . We found, however, that In(2LR)PL is not underdominant for fertility, as heterokaryotypic females produce perfectly viable eggs. Genetic analysis shows that the lack of underdominance results from the nearly complete absence of crossing over in the inverted region. This phenomenon is probably caused by mechanical and not genetic factors, because crossing overis not suppressed in In(2LR)PL homokaryotypes. Our observations do not support the idea that the fixationof pericentric inversions among closely related species implies the action of genetic drift overcoming strong natural selection in very small populations. If chromosome arrangements vary in their underdominance, it is those with the least disadvantage as heterozygotes, like In(2LR)PL , that will be polymorphic or fixed in natural populations.

HEstrongest evidence thatgenetic drift can sufficiently small, drift can increase the frequency of T overcome natural selection is the existence of a new underdominant arrangement above the unsta- fixed differences among related species for chromo- ble equilibrium frequency, after which it is fixed by some rearrangements that are underdominant (ie., selection. [This corresponds to a shift between adap- deleterious as heterozygotes).Heterokaryotypes for tive peaks, animportant part of WRIGHT’S(1970) many such rearrangements, including pericentric in- “shiftingbalance” theory of evolution.]When the versions, centric fissions, translocations and centric or underdominance is severe, this process requires ex- tandem fusions, may be semisterile because aberrant tremely small effective population sizes (on the order meiotic segregation or recombination within the ar- of 10; LANDE1979, 1984; HEDRICK1981; WALSH rangementsproduce aneuploid gametes (WHITE 1982). The fixationof chromosome arrangements 1973) . Because such arrangements always originate thus implies small effective species sizes: “It is not as heterokaryotypes, they should be eliminated im- unreasonable to consider much of the corpus of cy- mediately by naturalselection in randomlymating togeneticdata primafacie evidence that speciation populations. Indeed, these arrangements are rarely occurs by the geographic isolation of small popula- found as polymorphisms in nature. In Drosophila, for tions”(FUTUYMA and MAYER1980, pp. 262-263). example, there are thousands of polymorphisms for WHITE’S(1 968, 1978) theoryof stasipatric speciation, paracentric inversions, which are not underdominant, which proposes that chromosome arrangements are but almost no polymorphisms for pericentric inver- animportant cause ofreproductive isolation, also sions (PATTERSONand STONE1952). requires genetic drift in very small populations. Although polymorphisms for underdominant chro- There are two other explanations for the fixation mosomes are rare, closely related species may some- of chromosomes that are underdominant forfertility: times be fixed for suchdifferences (STONE 1955; natural selection and meiotic drive. If a new chromo- WHITE 1973). This fixation implies that populations somerearrangement containsfavorable , the or species have undergone an evolutionary transition net fitness of the arrangementmay be positive despite from one arrangement to another, a transition that underdominance for fertility. This couldcause either would be opposed by natural selection. polymorphism or fixation (DOBZHANSKY195 1 ; CHAR- The most widely acceptedexplanation for such LESWORTH 1985; BICKHAMand BAKER 1979). transitions is genetic drift.In populations thatare Alternatively, if a new arrangement was linked to

Genetics 129 791-802 (November, 1991) 792 J. A. Coyne, S.and Aulard A. Berry an that segregated preferentially in meiosis, it could be fixed even if opposed by natural selection (WHITE1978; HEDRICK198 1). While possible, meiotic ,. drive seems less plausible than drift. Genetic analysis has shown no evidence for fixed meiotic-drive alleles among closely related species, and theparapatric geo- graphic distribution of some chromosome arrange- ments could not occur if they were meiotically driven (COYNE1986, 1989). Here we describe the genetic properties of a chro- mosome arrangement whose polymorphismin nature was surprising because of its supposed underdomi- nance. In 1985, a pericentric inversion was found in 1 populations of Drosophila melanogaster on two Indian '/ ! Ocean islands: Mauritius, about 1000 km east of Mad- agascar, and Rodriguez, about 600 km east of Maur- itius (AULARD1990). This inversion is very large, so recombination should severely reduce the fertility of heterokaryotypic females. Yet the inversion's occur- rence on two islands in moderate frequencies (3 and 7%, respectively) suggested that it must have been present for at leastseveral generations. With the exception of the very small inversion associated with the Segregation-distorter allele (HARTL andHIRAIZUMI 1976), this is the only pericentric inversion reported from more than one population of D. melanogaster. Because D.melanogaster is rare on these islands (DAVIDet al. 1989), we originally suspected that this inversion might be strongly underdominant yet main- FIGURE1 .-Polytene second chromosomes inlarval salivary tained by genetic drift. As weshow below, genetic glands. Asterisk marks the chromocenter, and arrows show the analysis disproved these expectations, for the inver- position of the breakpoints. A, 2LR/ST heterokaryotype. B, 2LRM0"/ sion was not underdominant for fertility. This lack of 2LR'" homokaryotype, showing cytological identity of the inver- underdominance leads us to question the idea that sions from the two islands. chromosome evolution provides strong evidence for (LINDSLEYand ZIMM1985; ASHBURNER199 1). the ability of genetic drift to overcome natural selec- The isofemale lines from Mauritius were maintained in tion in small populations. the laboratory until 1989, when we began this analysis. All but one of the isofemale lines from Rodriguez were com- MATERIALSAND METHODS bined in 1986 and maintained asa single stock. Drosophila stocks used in the analysis: The inversion: In August, 1985, MICHEL SOLICNACcol- CyO: Second-chromosome balancer In(ZLR)O, dp'"' Cy pr lected D. melanogaster on Mauritius and Rodriguez, sam- cn2/In(2LR)bwV'. pling both wild and domestic habitats (DAVIDet al. 1989). FM7a: X chromosome balancer In(I)sc8 + ln(1) dl-49,$Id D.melanogaster was relatively rare on bothislands, occurring sc8 wo v'v B. only in warehouses in theharbors. Aulard (1990) kary- Compound-2 entire: C(2)EN bw sf, a stock having both otyped single chromosomes from 14 isofemale lines from second chromosomes fused into a single entity. Produced Rodriguez and30 from Mauritius. One line fromeach by NOVITSKI(1 976), this chromosome was used to measure populationcontained a largepericentric inversion with nondisjunction. breakpoints at cytological positions 3 1 F and 51C on the vg(ST): A stock homozygous for thesecond chromosome- second chromosome (Figure 1A). We name this inversion mutation vestigial (2-67.0) and made homokaryotypic for In(2LR) Port Louis, abbreviated In(ZLR)PL, after the town the Standard arrangementof the second chromosome. in Mauritius from which it came. In the rest of this paper, b cn(ST): A stock homozygous for the second chromo- however, we refer to it simply as 2LR. some-mutations black (2-48.5) and cinnabar (2-57.5) and 2LR is therefore large, covering nearlyhalf the cytological made homokaryotypic for the Standard arrangementof the length of the chromosome, and roughlysymmetrical around second chromosome. the centromere. It also includes the well known loci black, bcn vg(ST): This stock,also homokaryotypic forthe purple, cinnabar, alcohol dehydrogenase and vestigial, which Standard arrangement,was used to measure doublerecom- have been mapped to small regions of the polytene chro- bination in the heterokaryotype. mosomes (LINDSLEYand ZIMM 1985, 1990). We estimate w m: This stock, containing the X-linked mutations white the recombinational length of the inversion to be about 30 (1-1.5) and miniature (1-36. l), aswell as the ST sequence of centimorgans, because daughterless (map position 41.3) has the second chromosome, was used to measure recombina- been cytologically located between3 1C and 32A, andcurved tion on theX chromosome. (map position 75.5) is just outside the inversion at 52D all(ST): This stock, containing the mutationsa1 (2-0.01). Evolutionary Genetics of an Inversion 793 dp (2-13.0), b (2-48.5), pr (2-54.5), c (2-75.5) px (2-100.5) period were replaced, and if the female died the data were and sp (2-107.0), as well as the ST arrangement of the discarded. second chromosome, was used to measure recombination Adh RFLP study of invertedand standard chromo- on the second chromosome. somes: Using the Cy0 balancer, we extracted asingle second ru h th: This stock, containing the mutations roughoid (3- chromosome from each of ten isofemale linesfrom the island O.O), hairy (3-26.5), and thread (3-43.2), as well as the ST populations: three lines from Rodriguez (one 2LR and two sequence of the second chromosome, was used to measure ST ) and eight from Mauritius (one 2LR and seven ST). recombination on the third chromosome. Those lines carrying 2LR could not be made isochromoso- nodDm: A dominant allele of the X-linked nod locus (1- malbecause they carried recessive lethals, so these were 36), nodDmcauses nondisjunction of nonexchange chro- balanced over a chromosome, A178, which carries a dele- mosomes, with a similar but much weaker effect on chias- tion that includes the Adh structural locus. DNA was ex- matic bivalents (WRIGHT 1974; ZHANG and HAWLEY 1990; tracted according to theprocedure of KREITMAN and ZHANGet al. 1990; RAWLYet al. 199 1). Drosophila females AGUADE(1 986). Approximately 3 pg of DNA was digested have two meiotic pairing systems: exchange pairing, which overnight in sealed microtiter plates with 0.5 pg of RNAase ensures normal disjunction of recombinant chromosomes, I and 5 units of the following enzymes: AluI, DdeI/BamHI, and distributive pairing, which ensures normal disjunction HaeIII, HhaI, HinfI, MspI, Sau3AI and TagI. DNA was of nonrecombinant chromosomes such as the fourth (see precipitated with NaOAc and isopropanol, washed in 70% GRELL1976). Like other alleles at this locus, nodDmdisturbs ethanol, dried andresuspended in 3 pl of formamide loading distributive pairing, causing nondisjunction of nonexchange buffer (KREITMANand AGUADE1986). Heat-denatured sam- chromosomes. The amount of nodDm-inducednondisjunc- ples were electrophoresed on a 30 cm X 40 cm X 0.4 mm tion is hence a good measure of the frequency of nonex- 5% polyacrylamide/7 M urea sequencing gel with a NaOAc change chromosomes. electrolyte gradient (SHEENand SEED1988) and electropho- Originally called TW-6" by WRIGHT (1974), nodD* was retically transferred to nylon membranes, UV cross-linked, hybridized toa 2.7-kb fragment incorporating the Adh kept balanced in males in the stock C(I)Dx yf/@Fy X nodDm 1(1)P/BSY,and was used obtain indirect measurements of structural locus, washed and autogradiographed according to KREITMANand AGUAD~(1986). recombination in genotypes lacking morphological markers. From the published Adh sequence (KREITMAN1983), we cn bw RspS: A stock homozygous for the second chromo- know the expected fragment lengths and can therefore some mutations Responder-sensitive (2-56.6), cinnabar (2- identify exactly the positions of restriction site gain/losses. 57.5) and brown (2-104.5) (BRITTNACHERand GANETZKY The positions and sizes of insertions and deletions, however, 1983). When SD alleles are present in Rsp"/Rspi heterozy- must beestimated. Overall, these nine enzymes surveyabout gotes, meiotic drive largely eliminates the chromosome car- 19% of all possible substitutions within the probed region rying Rsp'. [i.e., 19 X 2700 = 5 13 base pairs of sequence (KREITMAN Iu: A stock homokaryotypic for the Standard arrange- and AGUADE,1986)l and all insertions and deletions. ments on all major chromosomes. It was founded in 1975 by combining the progeny of 200 isofemale lines collected by P. IVESin Amherst, Massachusetts. In 1977, B. CHAR- RESULTS LESWORTH extracted isofemalelines from thisstock and Extraction of theinversion: Karyotypicanalysis founded a new stock by combining 21 of these lines that were that were homokaryotypic for the Standard arrange- showed thatthe isofemale line from Mauritius and the ment on all chromosomes. mixture of isofemale lines from Rodriguez still con- Rearing of flies: All crosses were made on standard agar- tained a high frequency of 2LR chromosomes after yeast-cornmeal medium, and reared in 8-dram vials at 24" five yearsof laboratory culture. Second chromosomes on a 12-hr light/dark cycle. from each strain were extractedusing the Cy0 balan- Chromosome analysis: Salivary glands from third-instar cer. Fourteenof the chromosomes from the Mauritius larvae were dissected in Ringer's solution, squashed, and orcein-stained according to standard protocol (ASHBURNER isofemale line produced homozygous-viable offspring, 1989). and six lines produced no homozygous offspring.The Egg hatchability: Virgin females used to lay the eggs for lethal chromosomes were kept balanced against CyO. scoring were placed invials for 2days withan equal number Karyotypic analysis showed that all of the homozy- of males. Flies were then transferred toegg-laying chambers gous-viable second chromosomes carried theStandard containing a small portion of colored medium. After 17-20 hr, unhatched eggs were picked from the medium, washed (ST)arrangement, while the homozygous-lethal chro- in 70% ethanol, and placed in groups of 40 on small squares mosomes were all 2LR. Crossing tests showed that all of black paper. Each square was placed in an 8-dram vial of these lethals from Mauritius wereallelic. containing food and incubated at 24 O . Egg hatch was scored Eight chromosomes were extracted from the Rod- after 24 hr. Preliminary tests showed that all viable eggs riguez line; six of these were homozygous viable and hatched during this period. two were homozygous lethal. All of the viable chro- Measuringsegregation, recombination, and nondis- junction: Two or three females (depending on the stock) mosomes carried the ST arrangement, while the two were crossed to an equal number of males in 8-dram food- lethal chromosomes again carriedZLR, and thelethals containing vials. Adults were removed after five days, and were allelic. offspring scored periodically until 8 days afterthe first The pericentricinversions from the different is- eclosion. lands did not, however, carry allelic lethals, because Measuring fecundity: One male and one newly hatched numerous wild-typeoffspring were produced upon virgin female were placedin a food vial to mate, and transferred 2days later to an8-dram vial containing colored intercrossing the CyOI2LR balanced strains from the food. The pairs were transferred daily andtheir eggs two islands. The viability of homozygous 2LRMaUrifiU/ counted over the next 7 days. Males who died during this 2~~RodrIguezoffspring allowedus to determine that the 794 J.Berry A. Coyne, A. S.and Aulard

TABLE 1 Egg hatchability of heterokaryotypic females in two experiments, one using a 2LR from Maritius,and the other a 2LR from Rodriguez

Experimental STfZLR X Iu Control Iu X STfPLR

Inversion Eggs Fraction hatched Eggs Fraction hatched XP 95% C.I. Expt.-Control

A. 2LR'"""' 2360 0.961 2560 0.956 0.840.005 rt 0.01 1 B. 2LRRod 3180 0.9770.974 2960 0.0030.74 rt 0.008 Females are given first in the cross description. pericentric inversions in both populations were cyto- erokaryotypic females have at most only 2% as much logically identical, showing perfect pairing in salivary crossing over in the region as ST/ST females. We gland preparations (Figure 1B). confirmed the suggestion of reduced recombination To determine if the nonallelic lethals could be re- in the inverted region with the following experiments. moved by recombination in homokaryotypes, we be- Is there crossing over in heterokaryotypes? Al- gan a stock with the wild-type 2LRMaurit'us/2LRRod'~ezthough single crossovers in theinverted region of progeny and reextracted second chromosomes after heterokaryotypes lead to aneuploid gametes, two- fifteen generations.Eight of the eighteenhomozygous strand double-recombinants yield normal gametes. If chromosomes were viable and fertile, indicating that the high hatchability of eggs produced by 2LRIST the lethals had been removed.Five of these lines were females results from restricted crossing over in the tested cytologically and confirmed to be 2LR/2LR. inverted region, this region should also show a sub- These results imply that 2LR originally arose on a stantial reduction of double crossing over. To test this lethal-free chromosome The subsequent accumula- possibility, we crossed female Cy0/2LRMa""Li"sto b cn tion of two lethals also implies that 2LR was probably vg(ST) males. The non-Curly F1 females were hetero- segregating in nature for tens to hundreds of gener- karyotypic and heterozygous for the three markers. ationsbefore it was discovered,a suggestion sup- These were backcrossed to homozygous b cn vg fe- ported by its occurrence on two distant islands. Fi- males and double-recombination scored in the prog- nally, the removal of the lethals through recombina- eny. As a control, we used ST/ST instead of 2LR/ST tion shows that they are not associated with the females in the original cross. inversion breakpoints. Only two double-recombinants were recovered Is 2LR underdominant for fertilityin heterokar- among 16,919 offspring of the 2LR/ST, a frequency ytoypes? To determine the effect of the inversion on of 1.18 X (As expected,there were no single the production of aneuploid eggs, we produced het- recombinants). The control cross produced 70 recom- erokaryotypic females by crossing males from Cy01 binants among 15,171 offspring, a frequency of 4.6 2LR balancer stocks to ST/ST females from the hom- X 1O-3. (The expected frequency of double-recombi- okaryotypic Iv strain. The non-Curly F1 females (ST/ nation in ST/ST homokarytypes, based on map dis- 2LR ), were backcrossed to Zv males, and scored for tances and assuming no interference, is 8.6 X lo-'). egg hatch. The control experiment consisted of ST/ The difference betweenthe estimates from ST/ST and 2LR males (from the same F1 cross) backcrossed to ST/2LR is highly significant (P < using Fisher's STIST Zv females. The offspring of these two crosses exact test). There is a 40-fold difference in the amount are genotypically nearly identical,differing only in of double-recombination between the two classes of the source of their X chromosome. Only the hetero- females, substantiating the predictions fromegg-hatch karyotypic females, however, are expected to produce experiments. Further measurements of recombina- aneuploid eggs, as males have no recombination. This tion are described below. experiment was performed twice, once using the per- This largereduction in recombination was con- icentric inversion from Mauritius and once the inver- firmed in aseparate experiment using the nodDrW sion from Rodriguez. allele, which causes nondisjunction of nonrecombi- The hatchability of eggs laid by heterokaryotypic nantchromosomes and a small percentageof ex- females in both experiments was over 95% (Table l), change chromosomes. C(1)Dx y f/BsY X nodDrW 1(1) and did not differ significantly from that of control y/BsYmales, which have the Standard arrangement eggs. The95% confidence intervals show thatthe of the second chromosome, were crossed to Cy01 maximum possible reduction of egg hatchability of 2~~MauriItusfemales. The wild-type female offspring heterokaryotypic females compared to controlsis only are ST/2LR and heterozygousfor thedominant 0.6%. This figure corresponds to the maximum fre- nodDrW. These virgin females were crossed to C(2)EN quency of single crossing over in the inverted region- bw sp males at a density of three males and three that could have gone undetected in our study. As the females per vial. Offspring will appear only when the inverted region is about 30 centimorgans long, het- heterokaryotypic mother has no crossing over on the E volutionary GeneticsEvolutionary Inversion of an 795

TABLE 2 TABLE 3 Measurement of nondisjunction when homokaryotypicor Measurement of nondisjunction when homokaryotypicor heterkaryotypic females (both heterozygous for the dominant heterokaryotypicfemales and males are crossed to C(2)EN bw allele nodDnv)are crossedto C(2)EN bw sp males sp individuals of the opposite sex

Totaloffspring Mean female fecundity Karyotype and sex Total offspring Female fecundity Female (wild-type) (SE) tested (wild type) (SE)

STIPLR 802 (750) 106.1 (2.45) STISLR male 23 (15) STIST 100 (94) 105.2 (4.95) STIZLR female 21 (19)90.27 f 2.32 Total ST/2LR 44 (41) Twenty-five vials, each containing three malesand three females, STIST male were scored for each karyotype. Mean female fecundity is given for 39 (25) the total number of eggs laid over the 7-day test period (1 1 females STIST female 20 (16) 64.73 f 3.84 tested per genotype). Total ST/ST 59 (4 1) Seventy-fivevials, each containing three males and threefemales, second chromosome, leading to nondisjunction and were made for each determination. Average female fecundity is complementation of the nullo-2 or diplo-2 sperm pro- estimated as the total number of eggs laid over a 7-day period (1 1 females were measured for each genotype). Number of wild-type duced by the compound-2 fathers. offspring (as opposed to bw SF)are given for each total. To determine the baseline level of nondisjunction in STIST homokaryotyes lacking nodDrW,we made the offspring. Seventy-five replicate vials, each containing same cross using mothers from an STIST isochromo- three males and three females, set up for each of the soma1 line from Mauritius. To determine whether any two reciprocal crosses. The number of offspring pro- difference in offspring number between control and duced by homo- and heterokaryotypic females were experimental females was due only to a difference in nearly equal (Table 3), butoffspring from heterokar- female fecundity and not the rate of nondisjunction, yotypic males had significantly fewer progenythan we measured egg production of both type of females homokaryotypic males = 4.12, P < 0.05under when mated to males from the Zv strain. expectation of equal progeny number). Summing both Table 2 shows that the heterokaryotypic mothers sexes to yield a total frequency of nondisjunction, we produce eighttimes more offspring thanST/ST moth- find no significant difference between karyotypes un- ers, a highly significant difference when compared to derthe assumption of equal fertility = 2.18). the null expectation of equal nondisjunction ($11 = However, homokaryotypic females produced only 546.3, P < [For bothgenotypes, many more two-thirds as many offspring as heterokaryotypes, a wild-type than C(2) bw sp progeny were found, which significant difference (comparison of the two groups is expected because compound-2 males produce more of 11 females for weeklong fecundity gives t[2o1 = nullo-2 than compound-2 sperm(S. HAWLEY,personal 5.67, P < This reduction in fecundity, also communication).] The difference in progeny number described below, could be due to theST/ST females' is not due to inherent differences in fecundity, as the isogenicity for the entire second chromosome. two genotypes of females do not differ (Table 2; an The higher fecundity of the heterokaryotype im- unpaired comparison of the fecundity of females from plies that this genotype has slightly less nondisjunction the two groups gives t[221= 0.151, P = 0.88). than homokaryotypes. The lack of increased nondis- The eightfold difference in nondisjunction implies junction of 2LRIST individuals alsogives evidence an eightfold reduction in recombination in STI2LR that the distributive pairingsystem operates in nature. females compared to ST/ST females. In ST/2LR fe- Does 2LR carry agene that reduces crossing over? males, the reduction in recombination is measurable The lack of recombination in heterokaryotypic fe- only outside the inverted region, because single-re- males could be due to eithermechanical factors, such combinants within the region are not recovered. This as a lack of proper meiotic synapsis, or to genetic reduction of recombination in heterokaryotypes is, factors, such as a linked allele that reduces crossing however, far more than one expects simply from the over. Because recombination eliminated the lethals in reduced length of the chromosome in which recom- 2LRMa"'/2LRRod homokaryotypes, inversion homokar- bination can be observed. Additional tests described yotypes must undergo atleast some crossing over. We below demonstratethat heterokaryotypic females examined the effect of 2LR on recombination outside show reduced recombination over the entire second the inversion as well as on other chromosomes. chromosome. In the former experiment, we made 2LRIST het- Do heterokaryotypes have increased nondisjunc- erokaryotypic females in which the ST chromosome tion? It is possible thatalthough heterokaryotypic was the multiply marked all genotype. Black and pur- females do not produce a high frequencyof aneuploid ple lie within the invertedregion (LINDSLEYand ZIMM gametes by recombination, they maystill do so by 1985,1990), while aristaless,dumpy, plexus, curved having increased nondisjunction. We tested this pos- and speck are outside the inversion (Ashburner, 1991). sibility by crossing the STI2LR heterokaryotypes of A control cross estimated recombination rates in all both sexes to theC(2)EN c bw tester stock and counting heterozygotesthat were homokaryotypic for the 796 J. A. Coyne, S. Aulard and A. Berry

TABLE 4 Recombination in females heterozygous for theall chromosome when homokaryotypicor heterokaryotypic

Map distance Map in reduction Proportional ST/2LR fe- Map distance in in heterokaryotypes Region males Region ST/ST females (column P/column 3) 1 (al-dp) 0.020 0.108 0.19 2 (dP-4 0.031 0.288 0.1 1 3 (b-Pr) 0.0003" 0.057 0.005 4 (Pr-4 0.0003" 0.178 0.002 5 (C-PX) 0.040 0.217 0.18 6 (PX-SP) 0.030 0.055 0.55 Sample size 3384 1675 Frequency of doubles in inverted re- 0.0003 0.012 0.025 gion (3 and 4)

a A single double-recombinant (al dp bc px sp).

TABLE 5 Effect of the second-chromosome inversionon recombination on other chromosomes

Recombination in Recombination in Cross andregion ST/2LR females ST/ST females Grly 0.4 12 0.375w-m 12 0.4 7.14* ru h th + ST nod,t cy0 I nod,f cy0 + 2LR 0.273 0.217 5.47** " I" " Region 1 (ru-h) " X" + ST + . 2LR Region 2 (h-th) 0.254 0.206 5.08** * The sample size in the w m cross (X chromosome) was 1806 for It ST/ZLR females and 1661 for ST/ST females. The sample size in the ru h th cross (third chromosome) was 1045 for STIZLR females and 719 for STIST females. I a Significance levels: *P = 0.02; ** P = 0.008. I I 4 Count progeny I Count progeny Standard sequence. Table shows the results of both I crosses. FIGURE2.-Crossing scheme to construct all three chromosomal Heterokaryotypes again produced no single recom- genotypes carring the nodDTWallele. Only the X, Y and second binants in the inverted region, and double-recombi- chromosome are shown. The two genotypes marked with an asterisk nants were found only 2% as frequently as in homo- were intercrossed to produce nodDTW/+;STIZLR females. All three genotypes were then crossed to C(2)EN bw sp males, as shown, to karyotypes. As is commonly observed (ALEXANDER provide a measure of nondisjunction and hence crossing over. See 1952; ROBERTS 1967), the reduction of recombina- text for further details. tion in heterokaryotypes is more severe inside than outside the inversion. From the egg hatchability tests, portsa mechanical us. a genic explanationfor the we know that STI2LR females have at least a 50-fold reduced crossing over. reduction of recombination in the invertedregion Finally, a more indirect test of recombination was compared to STlST females. Outside the region,how- made using the nodDTWallele in heterokaryotypes and ever, recombination is reduced only 2-lO-fold, with the two homokaryotypesaccording to the crossing the highest levels between markers farthest from the scheme shown in Figure 2. Recombination in the breakpoints. nodDTW-containingfemale genotypes was estimated by The greatreduction of recombination inside as crossing them to C(2)EN bw sp males and counting the compared to outside the inversion supports a mechan- offspring, almost allof which areproduced when ical and not a genetic cause for the reduced crossing there is no crossing over on the second chromosome. over: if a gene were involved, its effects would have The females were also crossed to Zv males and scored to diminish sharply outside the inversion breakpoints. for fecundity during 1 week. Wealso compared STIST us. STIBLR karyotypes In this analysis, the heterokaryotype produces the for their effects on recombination on theX and third most offspring, again indicating reducedcrossing over chromosome (Table 5).In both cases, the presence of (Table 6). The 2LR/2LR homokaryotype, however, the heterokaryotype slightly but significantly increases produces only one-third as many offspring as STIST recombination on the otherchromosomes. This is the (an unpaired t test on the number of offspring pro- well known "interchromosomal effect" observed for duced in individual vials gives t[ssl = 8.57, P < 0.001). many arrangements (LUCCHESI1976), andagain sup- This implies that crossing over is even higher in 2LR/ Evolutionary GeneticsEvolutionary of an Inversion 797

TABLE 6 TABLE 7 Number of offspring produced by three karyotypescontaining Recombination across the invertedregion in the two nodDnVwhen each is crossed to C(2)EN bw sp males homokaryotypes and the heterokaryotype(see text for details)

Total offspring Fecundity, Region 1 Region 2 Region 3 Genotype (wild-type)eggs/week (SE) Genotype al-dp dp-px px-sp N

nodnw; STIST 372 (342) 78.8 (7.4) 2LR a2 dp px sp/2LR 0.1 10 0.486 0.048 1013 nodDw; STI2LR 1660 (1515) 101.3 (9.2) ST a1 dp px sp/ST 0.1 13 0.443 0.045 1276 nodDm; 2LR/2LR 116 (105) 94.2 (8.8) 2LR a1 dppxsp/ST 0.019 0.030 0.016 880

~~ ~~ Forty-five vials (each containing three mated pairs) were scored Recombination fractions for each region are calculated from the for each genotype. Egg production is given as average fecundity of sixteen genotypes segregating in each cross (data notshown). Actual each female over a week of measurement (six females were scored map lengths of each region from (LINDSLEYand ZIMM 1985, 1990) per genotype). The proportion of bw sp offspring produced by the are: region 1, 12.6 cM; region 2, 86.5 cM; region 3, 6.5 cM. three genotypes (data not shown) is homogeneous (Gp] = 0.28, P = 0.87). proportion of individuals in the 16 genotypic classes 2LR than in ST/ST females. The difference in progeny (data not shown; heterogeneity, xf151= 23.23, P = 0.08). Again, there is no evidence that 2LR contains production is not due to a lower fecundity of 2LR/ a gene that reduces crossing over. In this cross, as 2LR females, as they are significantly more fecund opposed to the cross using the nodDTWallele, there is than ST/ST females over1 week (tlo = = 3.42, P no sign of higher recombination in 2LR than in ST 0.007). We therefore find no evidence for reduced homokaryotypes. This difference may be due to either recombination in 2LR/2LR ascompared to ST/ST different genetic backgrounds in the two tests or the homokaryotypes. mosaic nature of the a1 dp 2LR px sp chromosome. The above experimentgives an indirect estimate of Because heterokaryotypes do not decrease recom- recombination in homozygotes for 2LR chromosomes bination onother chromosomes, because crossing extracted from nature. A more direct test of recom- over in heterokaryotypes is reduced far more inside bination can be made by placing mutant markers on than outside the inverted region, and because 2LR/ a 2LR-containing second chromosome,although in 2LR females do not show less recombination than ST/ this case the 2LR chromosome will be a mosaic of wild ST females, we conclude that crossing over in heter- and laboratory genomes. okaryotypes is reduced by mechanical and not genetic We constructed females heterozygous for2LR and factors. the multiply marked all chromosome containing the Does 2LR show segregation distortion?Although ST sequence. These females were backcrossed to all 2LR is not obviously underdominant, it does carry a males. From approximately 20,000 backcross off- lethal and hence shouldeventually be eliminated from spring, we collected one al dp 2LR @x sp/al dp b pr c laboratory cultures (this should also be true in nature px sp ST male, having two markers on either side of if all 2LRs carry lethals). Three obvious explanations the inversion. The Cy0 marker chromosome, which for its persistence are genetic drift, overdominance of contains pr, was used to extract 2LR and produce a the inversion (i.e., 2LR/ST genotypes are more fit stock homozygous for a1 dp 2LR px sp. A similar cross than the ST/ST and lethal 2LR/2LR karyotypes), and was used to manufacture a stock homozygous for the segregationdistortion, such thatheterokaryotypes a1 dp ST px sp chromosome. preferentiallyproduce gametes carrying the inver- We used these two stocks to estimate recombination sion. Several tests were done tocheck this last hypoth- among the four markers in all three possible kary- esis. otypes. To measure recombination in 2LR homozy- Because the Segregation-distorter locus is in the in- verted region (LINDSLEYand ZIMM 1990), oneobvious gotes, we crossed a1 dp 2LR px sp males to the stock homozygous for the unmarked, extracted 2LR chro- possibility is that 2LR contains the SD allele, which segregates at very high frequencies when heterozy- mosome. The a1 dp 2LR px sp/al+ dp+ 2LR px+ sp+ gous with a second chromosome carrying thesensitive females were backcrossed to all males andthe 16 allele at the linked Responder locus (Rsp'; HARTLand possible genotypes scored in the offspring. A similar HIRAIZUMI1976). To test this possibility, we produced estimate was made for ST/ST homozygotes using a1 a genotype heterozygous for the 2LR chromosome dp ST PX sp/aL+ dp+ ST ex'sp' females. Finally, we and a Standard-sequence chromosome carrying Re- measured recombination in heterokaryotypes using a1 sponder-sensitive and two recessive markers (2LRIST dp 2LR px sp/al+ dp+ST px+ sp' females. Rsp' cn bw). If 2LR males carry the SD allele, nearly As expected, the amount of recombination in ST/ all of their sperm should contain2LR and they should 2LR heterokaryotypes was much less than in homo- produce nearly all wild-type offspring when crossed karyotypes (Table 7). The 2LR/2LR and ST/ST horn- to cn bw females. okaryotypes, however, showed nearly equal frequen- These males, however, produced progeny in Men- cies of recombination and showed no difference in the delian proportions: the proportion of cn bw offspring 798 J. A. Coyne,A. S.and Aulard Berry

TABLE 8 Segregation ratiosof 2LR versus ST chromosome from heterokaryotypes

Proportion Female parent Male parent Total mutant G, b cnlbA. b cn b cn (ST)/b+ cn+ (STMau) 0.4616316 1.15 (n.s.) b cn/b b cn b cn cn+(ST)/b+ (2LRM"") 2802 0.472 B. vg(ST)/vg+(ST""") vglvg 2332 0.4844.32 (p < 0.05) vg(ST)/vg+(2LRMaU) %I% 2784 0.513 C. vg(ST)/vg+ (STRnd) uglvg 1649 17.980.445 (p < vg(ST)/vg+ (2LRRod) vglvg 1292 0.523 was 0.470 (N= 967), not significantly different from combinants thus gives the ratio of ST to 2LR sperm the frequency of 0.447 (N = 664) obtained in the produced by heterokaryotypic males. control reciprocal cross (GLll= 0.86, P = 0.36). We Every female hatching fromthree vials ofeach cross also determined in a separate cross that 2LR from was collected as a virgin and backcrossed to two b cn both islands carry the RspS allele (data not shown). uglb cn ug males. With the exception of the few females Three other tests of segregation ratios from heter- who escaped or died, all of the females produced okaryotypic males or females were conducted using more than the 55 offspring neededto determine with heterozygotesfor 2LR chromosomes and Standard 99% confidence whether the chromosome was ST or chromosomes containing recessive mutations. These 2LR. (Thisfrequency was determined in separate heterozygotes were backcrossed to stocks homozygous estimates of recombination in ST b cn ug/ST b+ cn+ug+ for the mutations;the ratios of mutant us. nonmutant heterozygotes.) progeny give the segregation ratios of ST us. 2LR Heterokaryotypesfrom both populations showed chromosomes. Control tests for viability examined the Mendelian segregation. In the cross using Mauritius segregation ratios in STIST individuals. females, the proportion of 2LR gametes was 0.495 (N One of the three tests showed no deviation from = 275), not significantly different from the expected Mendelian ratios (Table 8), while theother two 0.5 = 0.03, P > 0.75). In Rodriguez females, the showed slight but significant underrepresentation of proportion of 2LR gametes was 0.514 (N = 253), 2LR in the progeny. These deviations may be due to again not differing from Mendelian expectation (x711 = 0.19, P 0.5). differences in fitnesses of the uestigaal homozygote on > We therefore find no evidence for segregation dis- different genetic backgrounds. At any rate, these dif- tortion of 2LR over ST chromosomes from the same ferences do not indicate a meiotically driven 2LR population. Of course, our power todetect small chromosome because the inversion is underrepre- amounts of distortion is limited by the restricted sam- sented in the offspring of STI2LR females. ple of offspring. The upper 95% confidence limit for It is possible that although the inversion does not the proportion of 2LR sperm from heterokaryotypes show meiotic drive against the marked tester chro- is 0.575 for the Rodriguez cross and 0.553 for the mosomes, it might do so against ST chromosomes Mauritius cross. taken from the same natural population. It is difficult Do the inversions from the two islands have a to test this possibility, as it requires measuring segre- single origin?A limited RFLP study of the Adh region gation ratios of unmarked chromosomes. To do so, examined the likelihood of a single origin of the we constructed 2LRIST genotypes by crossing Cy01 inversions from the two islands. We analyzed 11 lines: 2LR females to males isochromosomal for ST chro- seven ST second chromosomes from Mauritius, one mosomes extracted from the same population as the 2LR chromosomefrom Maritius, two ST chromo- 2LR. (Inversions from the two islands were studied somes fromRodriguez andone 2LR chromosome separately.) The 2LRIST males were crossed to homo- from Rodriguez. The ST chromosomes came from zygous b cn ug(ST) females. We tested only heterokar- different isofemale lines. yotypic males because meiotic drive is nearly always There was striking uniformity among these chro- limited to that sex (ZIMMERING, SANDLERand NICO- mosomes, all having identical RFLP patterns except LETTI 1970). These males produce two types of off- for two STs from Mauritius [one with a loss of Ah1 spring: b cn ug(ST)/b+ cn+ ug+(ST), and b cn ug(ST)/b+ 423 and the other a loss of AluI 423 and a loss of cn+ ug+(2LR). Females of the former genotype will MspI 503 (numbering after KREITMAN1983)] and the produce recombinants when backcrossed crossed to two ST lines from Rodriguez (bothhaving loss of AluI homozygous b cn ug males; females of the latter gen- 423, gain of HaeIII 817, and a 35-bp insertion in otype produce no recombinants.The ratio of females fragment 498-573). Homozygosity is higher than has producing no recombinants to those producing re- been seen in other tested populations of D. melano- Evolutionary GeneticsEvolutionary of an Inversion 799 gaster: using the same techniques in a survey of pop- inance (with egg hatch reduced by 25 and 4396, re- ulations on the east coast of the United States, A. spectively), as well as frequent double-recombination BERRYand M. KREITMAN(personal communication) within theinverted region. Heterokaryotypes for found a homozygosity of Adh 4-cutter haplotypes of Zn(3LRj190, however, showed no reduction in fertility only 0.08 (1 13 haplotypes were seen in 1533 lines and virtually no double recombination. Roberts spec- from 25 populations). This contrasts with the homo- ulated that this inversion was notunderdominant of 0.383 for thecombined lines from the two because it spanned a regionof the chromosome having islands. The limited RFLP variation suggests that the low recombination. flies have restricted population sizes on the islands, Other anecdotal evidence that such inversions may which is further supported by their rarity. not be underdominant comes from a failure to find While the RFLP results do not rule out the possi- semisterility in mice heterokaryotypicfor both X- bility that 2LR has originated more than once, the linked and autosomalpericentrics (NACHMAN and similar RFLP patterns for the two inversions despite MYERS 1989), although sample sizes were very small. the differences among ST chromosomes supports the In addition, polymorphism for pericentric inversions idea of a unique origin. The cytological identity of has been reported in Drosophila robusta by CARSON the breakpoints of the 2LRs fromthe two islands and STALKER(1947), who speculated that the poly- (Figure 1) would also be a remarkable coincidence if morphism may have been permittedby reduced cross- the inversion arose more than once. ing over in heterokaryotypic females. Populations of other animal species are sometimes polymorphic for DISCUSSION pericentric inversions, and in many of these cases chiasmata are notobserved in the inverted region(see Despite the widespread idea that heterozygotes for below). pericentric inversions are semisterile, we find no such The most important implication of our results re- underdominance for fertility in 2LR. The lack of such lates to the widespread idea that fixed differences underdominance almost certainly comes from a dras- among taxa for “underdominant”inversions necessar- tic reduction of crossing over in the inverted region, ily imply extreme population bottlenecks and shifts so that heterokaryotypic females produce almost no between adaptive peaks. We emphasize first that most aneuploid eggs. Heterokaryotypes also show no in- newly arisingpericentric inversions in Drosophila crease in nondisjunctionover homokaryotypes, an must certainly be underdominant. In contrast to par- observationmade onother inversions (STONEand acentric inversions, they are almost never polymor- THOMAS1935; STURTEVANTand BEADLE 1936; MORAN198 1). phic in natural populations, and are also fixed much The reduced crossing over is not caused by a gene less frequently. STONE(1955) estimated that in the linked to the inversion, because 2LR homokaryotypes 650 Drosophila species known at that time, only 32 show at least as much recombination as ST homokar- pericentric inversions were either polymorphicor had yotypes, and the presence of 2LR does not reduce been fixed, compared to between 6,100 and 36,500 recombination on other chromosomes. Instead cross- paracentric inversions. Unless the two types of inver- ing over is apparently reduced by mechanical factors; sions occur at drastically different rates, which seems we suspect heterosynaptic meiotic pairing or a disrup- unlikely, this difference must reflect an innate differ- tion of pairing sites (see below). ence in fitness. The inversion’s lack of underdominance for fertility If, however, pericentric inversions vary in their rate undoubtedly contributes to its persistence in nature of recombination (and hence their fertility) as heter- and in laboratorycultures. With our present data, okaryotypes, it is those with the least underdominance however, we cannot discriminate between the possi- that will become fixed or polymorphic. Because this bilities that 2LR is neutral with respect to ST, that it study and others have shown such variation, the ob- is deleterious as a homozygote but maintained in the servation of species fixed for differentinversions need population by drift, or that it is overdominant because not imply a peak shift or strong episode of genetic it contains advantageous alleles but cannot be fixed drift. Such a conclusion requires that the supposed because of the included lethal. We can conclude, underdominance be testedby direct studiesof meiosis however, that stong drift in very small populations is and fertility in hybrids. Moreover, these studies are not required to explain thepolymorphism. complicated by the fact that genic as well as chromo- There is one previous report of a pericentric inver- somal differencescan cause meiotic problems and sion in which reduced crossing over in heterokary- hybrid sterility (DOBZHANSKY195 1). otypic females leads to nearly normal fertility. ROB- If such inversions are not underdominant,we need ERTS (1967) measured fertility and crossing over in not invoke special mechanisms such as meiotic drive heterokaryotypes for three X-ray-induced pericentric to explain their fixation. Indeed, thesegregation stud- inversions on the third chromosome of D. melano- ies show not the slightest evidence for meiotic drive gaster. Two inversions showed substantial underdom- of 2LR. Normal Mendelian segregation was also found 800 J. A. Coyne, S.and Aulard A. Berry in species hybrids heterokaryotypic for a pericentric unexplained phenomenon of reduced crossing over inversion fixed between two phylads of the Drosophila in females heterokaryotypic for paracentric inversions virilis group (COYNE1989). (NOVITSKIand BRAVER1954; ISHII and CHARLES- Our failure to find genes linked to 2LR that reduce WORTH 1977). In such genotypes, double exchange is crossing over in heterokaryotypes implies the involve- often reduced dramatically inside the inversion and ment of mechanical factors. It also militates against single-exchange reducedoutside the inversion; the the idea thatpericentric inversions persist because latter effect often reaches farbeyond the breakpoints. they are linked to genesreducing recombination, We suggest that such chromosomes do not synapse either fortuitously or as a result of selection (CARSON properly. and STALKER 1947). There may, of course, be other explanations for The mechanical causes of reduced crossing over are reduced recombination in heterokaryotypes for peri- unclear. One possibility is heterosynapsis: during centric inversions. The inverted sections, for example, meiosis in female heterokaryotypes, the homologues maysimply fail to pair because loopformation is may pair without forming inversion loops, leading to difficult in germ cells. Such explanations will be test- synapsis of nonhomologous DNA. Chiasmata are ap- able when it becomes possible to observe meiotic parently not formedin heterosynaptic regionsbecause pairing in female Drosophila. molecular recombination requires homology between The lack of underdominance of polymorphic peri- paired segments of DNA (WATSONet al. 1976). centric inversions may also apply to other arrange- Heterosynaptic pairing of paracentric and pericen- ments such as Robertsonian fusions, tandem fusions, tric inversions is often seen in other species, including and translocations. While it is clear that heterokary- corn (MCCLINTOCK1933), fungi (BOJKO1990), mice otypes for Robertsonian fusions or reciprocal trans- (ASHLEY,MOSES and SOLARI1981; MOSES et al. 1982; locations can be semisterile because of aberrant seg- GREENBAUMand REED1984; GREENBAUM, HALE andregation, the degree of sterility differsamong ar- FUXA 1986; HALE 1986), chironomids (MARTIN rangements, with some showing normal segregation 1967), orthopterans (COLEMAN1948; NUR 1968; andno underdominance (STONE 1949; LEWISand MORAN1981), and humans (GABRIEL-ROBEZet al. JOHN 1957; BRUEREand ELLIS1979; ELDERand PA- 1988). Some inversions initially pair as a loop and THAK 1980;PORTER and SITES 1985; MORITZ 1986). then “adjust” to straight heterosynaptic pairingduring Summarizing the data onsegregation of heterozygous pachytene (MOSES et al. 1982; BOJKO 1990), others arrangements in hybrid zones, SHAW(1981) found pair heterosynaptically throughout meiosis (HALE little evidence that naturally occurring arrangements 1986), while still others vary in their pairing behavior disrupted meiosis. among cells (MCCLINTOCK1933; NUR 1968; GREEN- As with pericentric inversions, each putative case of BAUM and REED1984). Cell-to-cell variation in pairing underdominance must be demonstrated by direct in- could account for the very low frequency of double- vestigation of meiosisand hybrid fertility. Such studies recombination seen in our studies. Unfortunately, it must also ensure that themeiotic problems result from is difficult to determine whether heterosynaptic pair- the chromosomal differences and not from genic di- ing occurs in Drosophila because it is impossible to vergence, and also that heterokaryotypes have low- get good chromosome preparationsof female meiosis. ered fertility (in some cases, abnormal gametes may Another explanation for reduced crossing over is be eliminated before fertilization)(BRU~RE andELLIS that the breakpoints of 2LR may lie in regions impor- 1979). Other requirements for demonstrating under- tant for synapsis and exchange. Using X-rays to pro- dominancefor fertility are reviewed bySITES and duce a number of 2-4 and 3-4 reciprocal transloca- MORITZ (1 987)and BAKERand BICKHAM(1 986). tions, ROBERTS(1970) found that the position of the If such arrangements are not underdominant, then breakpoints had a large effect on recombination in of course they cannotcause reproductive isolation and heterozygotes, with some translocations lowering re- speciation as proposed by WHITE(1968, 1978).The combination throughout the entirechromosome arm. most vigorous advocate of chromosomal speciation, Breakpoints with the largest effect occurred roughly WHITEbased much of his theory on fixed and poly- in the middle of each autosomal arm (ROBERTS1976, morphic pericentric inversions in orthopterans. But Figure 7). It is notable thatboth breakpoints of as JOHN (198 1, p. 43) noted, “In all these cases there Zn(2LR)PL fall within these regions, as do those of is straight, nonhomologous pairing of the relatively Zn(3LR)190, ROBERTS’(1 967) inversion discussed invertedsegments at male meiosis andno reverse above. HAWLEY(1980) found similar area effects on looping. Because such straight pairing precludes the recombination in heterozygotes forX-4 translocations, production of unbalanced gametes in the heterozy- although in this case the reduced recombination oc- gotes, it also precludes them from generating hybrid curred oversmaller regions of the chromosome. sterility. Indeed, there is no case that I am aware of Either heterosynapsis or chromosome breakage at where fixed differencesinvolving genuine pericentric pairing sites could account for the well known but inversions do lead to reproductive isolation.” Evolutionary Genetics of an Inversion 80 1 If chromosome evolution does not constitute strong BOYCOT",A. E.,C. DIVER,S. GARSTANGand F. M. TURNER, 1930 The inheritance of sinistrality in Limnaea peregra (Mol- support for peak shifts innature, are we left with any lusca, Pulmonata). Phil. Trans. R. SOC.Lond. B 219 51-131. evidence that genetic drift can overcome natural se- BRITTNACHER,J. G., and B. GANETZKY,1983 On the components lection? The one convincing case is that of coiling in of segregation distortion in Drosophila melanogaster. 11. Dele- snails. In some species,right- and left-hand coiling are tion mapping and dosage analysis of the SD locus. Genetics 103: 659-673. controlled by single Mendelian alleles showingmater- BRU~RE,A.N., and P. M. ELLIS,1979 Cytogenetics and repro- nal inheritance (BOYCOTTet ai. 1930; MURRAYand duction of sheep with multiple centric fusions (Robertsonian CLARKE1966). In a population fixed for one coiling translocations). J. Reprod. Fertil. 57: 363-375. allele, an individual of opposite coiling is at a disad- CARSON,H. L., and H. D. STALKER,1947 Gene arrangements in natural populations of Drosophila robusta Sturtevant. Evolution vantage because it has great difficulties achieving cop- 1: 113-133. ulation (LIPTONand MURRAY1979; GITTENBERGER CHARLESWORTH,B., 1985 Recombination, genome size, andchro- 1988). Indeed, polymorphisms for coiling are quite mosome number, pp. 489-513 in Evolution of Genome Sire, rare in nature. Nevertheless, populations of some edited by T. CAVALIER-SMITH.JOhn Wiley & Sons, Chichester. COLEMAN,L. C., 1948 The cytology of some western species of species are fixed for different coiling morphs, imply- Trimerotropis (Acrididae). Genetics 33: 5 19-528. ing that evolution has produced transitions between COYNE,J. A., 1986 Meiotic segregation and male recombination adaptive peaks. Such switches in coilingmay be initi- in interspecific hybrids of Drosophila. Genetics 114: 485-494. ated by genetic drift, which occasionally increasesthe COYNE,J. A., 1989 A test of the role of meiotic drive in chromo- some evolution. Genetics 123: 241-243. frequency of a rare allele above the unstable equilib- DAVID,J. R., S. F. MCEVEY,M. SOLICNACand L. TSACAS, rium frequency, after which it becomes fixed by selec- 1989 Drosophila communities on Mauritius and theecological tion (ORR 1991). This scenario is facilitated by the niche of D. mauritiana(Diptera, Drosophilidae). Rev. Zool. Afr. limited mobility of snails,their small populations, and 103: 107-1 16. DOBZHANSKY,T. 1951 Geneticsand the Origin of Species, Ed. 3. the maternal inheritance of the alleles, which allows Columbia University Press, New York. them to increase in frequency without producing ELDER,F. F. B., and S. PATHAK,1980 Light microscopic obser- many individuals having unfit phenotypes (JOHNSON vations on the behavior of silver-stained trivalents in pachytene 1982; GITTENBURGER1988). cells of Sipodon&lviventer (Rodentia, Muridae) heterozygous A lone example, however, cannot buttress strong for centric fusion. Cytogenet. Cell Genet. 27: 31-38. FUTUYMA,D. J., and G. C. MAYER,1980 Non-allopatric speciation claims about evolutionary change. Much more evi- in animals. Syst. 2001. 29: 254-271. dence is required if modelsof evolution involving GABRIEL-ROBEZ,O., C. RATOMPONIRINA,M. CROQUETTE,J. Cou- genetic drift, such as the shifting balance theory of TURIER and Y. RUMPLER,1988 Synaptonemal complexes in a WRIGHT(1970), are tobe taken seriously. subfertile man with a pericentric inversion in chromosome 21. Cytogenet. Cell Genet. 48: 84-87. GITTENBERGER,E., 1988 Sympatric speciation in snails; a largely This work was supported by a fellowship from the John Simon neglected model. Evolution 42: 826-828. Guggenheim Foundation and National Institutes of Health grant GREENBAUM,I. F., D. W. HALEand K. P. FUXA,1986 Synaptic GM38462 toJ.A.C. Wethank L. ANDREWS,J. BULL, A.CARPENTER, adaptation in deer mice: a cellular mechanism for karyotypic B. CHARLESWORTH,J. DAVID, M. KREITMAN, F. LEMEUNIER,S. orthoselection. Evolution 40: 208-21 3. HAWLEY,U. NUR,H. A. ORR, T. PROUT,J. W. SITES,JR., P. GREENBAUM,I. F., and M. J. REED,1984 Evidence for heterosy- SNIEGOWSKI,M. SOLIGNACand J. SFWFFORDfor advice and com- naptic pairing of the inverted segment in pericentric inversion ments, L. BROOKS,s. HAWLEY,F. LEMEUNIERand C.-I. Wu for fly heterozygotes of the deer mouse (Peromyscus maniculatus).Cy- stocks, J, RUX for technical assistance, and JEAN DAVIDand F. togenet. Cell Genet. 38: 106-1 11. LEMEUNIERfor their hospitality in France. GRELL,R. F., 1976 Distributive pairing, pp. 435-486 in The Genetics and Biology of Drosophila, Vol. la, edited by M. ASH- BURNER and E. NOVITSKI. Academic Press, New York. LITERATURECITED HALE,D. W., 1986 Heterosynapsis and suppression of chiasmata ALEXANDER,M. L., 1952 The effect of two pericentric inversions within heterozygous pericentric inversions of the Sitka deer upon crossingover in Drosophila melanogaster.Univ. Texas Pub. mouse. Chromosoma 94: 425-432. 5204: 219-226. HARTL,D. L., and Y. HIRAIZUMI,1976 Segregation distortion, ASHBURNER,M., 1989 Drosophila, a LaboratoyManual. Cold pp. 616-666 in The Genetics and Bzologv of Drosophila, Vol. lb, Spring Harbor Laboratory, Cold Spring Harbor, N.Y. edited by M. ASHBURNERand E. NOVITSKI. Academic Press, ASHBURNER,M., 1991 Drosophila Genetic Maps. Drosophila In- New York. form. Serv. 69. HAWLEY,R. S., 1980 Chromosomal sites necessary for normal ASHLEY,T., M. J. MOSES and A. J. SOLARI,1981 Fine structure levels of meiotic recombination in Drosophila melanogaster. I. and behaviour of a pericentric inversion in the sand rat, Psam- Evidence for andmapping of the sites. Genetics 94: 625-646. momys obesus. J. Cell Sci. 50 105-1 19. HEDRICK,P. W., 1981 The establishment of chromosome var- AULARD,S., 1990 Polymorphisme chromosomique de Drosophila iants. Evolution 35: 322-332. melanogaster, en Afrique et dans les iles de I'Ochn Indien. ISHII, K., and B.CHARLESWORTH, 1977 Associations between Thesis, UniversitC. Pierre et Marie Curie, Paris. allozyme loci and gene arrangementsdue tohitch-hiking effects BAKER,R. J., and J. W. BICKHAM,1986 Speciation by monobra- of new inversions. Genet. Res. 30 93-106. chial centric fusions. Proc. Natl. Acad. Sci. USA 83: 8245- JOHN, B., 198 1 Chromosome change and evolutionary change: a 8248. critique, pp. 23-51 in Evolution and Speciation. Essays in Honor BICKHAM,J. W.. and R. J. BAKER,1979 Canalization models of of M. J. D. White, edited by W. R. ATCHLEYand D. S. WOOD- chromosome evolution. Bull. Carnegie Mus. Nat. Hist. 13: 70- RUFF. Cambridge University Press, Cambridge. 84 JOHNSON,M. S., 1982 Polymorphism for direction of coilin BOJKO, M., 1990 Synaptic adjustment of inversion loops in Neu- Partula suturalis: behavioural isolation and positive frequency rospora crassa. Genetics 124: 593-598. dependent selection. 49: 145- 15 1. 802 J. A. Coyne, S. Aulard and A. Berry

KREITMAN,M., 1983 Nucleotide polymorphism atthe Alcohol Robertsonian heterozygotes from a highly polymorphic lizard Dehydrogenase locus of Drosophila melanogaster. Nature 304 population. Cytogenet. Cell Genet. 39 250-257. 412-417. RASOOLY,R. S., C. M. NEW, P. ZHANG,R. S. HAWLEY andB. S. KREITMAN,M., and M. AGUAD~,1986 Genetic uniformity in two BAKER,1991 The genetic analysisof microtubule mtors in populations of Drosophilamelanogaster as revealed by filter Drosophila melanogaster. 11. A dominant mutation at the nod hybridization of four-nucleotide-recognizing restriction en- locus is associated with a single base change in the putative zyme digests. Proc. Natl. Acad. Sci USA 83: 3562-3566. ATP-binding domain. Genetics (in press). LANDE,R., 1979 Effective deme sizes during long-term evolution ROBERTS, P. A,, 1967 A positive correlation between crossing estimated from rates of chromosomal rearrangement. Evolu- over within heterozygous pericentric inversions and reduced tion 33: 234-25 1. egg hatch of Drosophila females. Genetics 56: 179-187. LANDE,R., 1984 The expected fixation rate of chromosomal ROBERTS,P. A., 1970 Screening for X-ray-induced crossover sup- inversions. Evolution 38 743-752. pressors in Drosophila melanogaster: prevalence and effective- LEWIS,K. R., and B. JOHN, 1957 Studies on Periplaneta americana. ness of translocations. Genetics 65: 429-448. 11. Interchange heterozygosity in isolated populations. Hered- ROBERTS,P. A., 1976 The genetics of chromosome aberrations, ity 11: 11-22. pp. 67-184 in The Genetics and Biology of Drosophila, Vol. la, LINDSLEY,D., andG. ZIMM, 1985The genome of Drosophila edited byM. ASHBURNERand E. NOVITSKI. Academic Press, melanogaster. Part 1: Genes A-K. Drosophila Inform. Serv. 62. New York. LINDSLEY,D., and G.ZIMM, 1990The genome of Drosophila SHAW,D. D., 1981 Chromosomal hybrid zonesin orthopteroid melanogaster. Part 2: Genes L-Z, balancers, and transposable insects, pp. 146-1 70 in Evolution and Speciation.Essays in Honor elements. Drosophila Inform Serv. 68. of M. J. D. White, edited by W. R. ATCHLEYand D. S. WOOD- LIPTON,C. S., and J. MURRAY, 1979 Courtship of land snails of RUFF. Cambridge University Press, Cambridge. the genus Partula. Malacologia 19: 129-146. SHEEN,J., and B. SEED,1988 Electrolyte gradient gels for DNA LUCCHFSI,J. C., 1976 Interchromosomal effects, pp. 315-329 in sequencing. Biofeedback 6: 942-944. The Genetics and Biology of Drosophila, Vol. Ib, edited by M. SITES,J. W. JR., and C. MORITZ, 1987 Chromosomal evolution ASHBURNERand E. NOVITSKI.Academic Press, New York. and speciation revisited. Syst. Zoo]. 36: 153-174. MARTIN, J.,1967 Meiosisin inversion heterozygotes in Chiron- STONE,W. S., 1949 The survival of chromosomal variation in omidae. Can. J. Genet. Cytol. 9 255-268. evolution. Univ. Texas Pub. 4920: 18-2 1 MCCLINTOCK,B., 1933 The association of non-homologous parts STONE,W. S., 1955 Genetic and chromosomal variability in Dro- of chromosomes in the mid-prophase of meiosis in Zea mays. Z. sophila. Cold Spring Harbor Symp. Quant. Biol. 20 256-270. Zellforsch. Mikrosk. Anat. 19 191-237. STONE,W., and I. THOMAS,1935 Crossover and disjunctional MORAN,C., 1981 Spermatogenesis in natural and experimental properties of X chromosome inversions in Drosophila melano- hybrids between two chromosomally differentiated taxa of gaster. Genetics 17: 170-184. Caldeiu captiva. Chromosoma 81: 579-591. STURTEVANT,A. H., and G. W. BEADLE,1936 The relations of MORITZ,C., 1986 The population biology of Gehyra (Gekkoni- inversion in the X chromosome of Drosophila melanogaster to dae): chromosomal change and speciation. Syst. 2001. 35: 46- crossing over and disjunction. Genetics 21: 554-604. 67. WALSH,J. B., 1982 Rate of accumulation of reproductive isolation MOSES,M. J., P. A. POORMAN,T. H. RODERICKand M. T. DAVISSON, by chromosome rearrangements. Am. Nat. 120: 510-532. 1982 Synaptonemal complex analysis of mouse chromosomal WATSON,J. D., N. H. HOPKINS,J. W. ROBERTS,J. A. STEITZand rearrangements. IV. Synapsis and synaptic adjustment in two A. M. WEINER,1976 Molecular Biology of the Gene, Ed. 4. paracentric inversions. Chromosoma 84 457-474. Benjamin Cummings Publishing Co., Menlo Park, Calif. MURRAY,J., and B. CLARKE,1966 The inheritance of polymor- WHITE,M. J. D., 1968 Models of speciation. Science 159: 1065- phic shell characters in Partula (Gastropoda). Genetics 54 1070. 1261-1277. WHITE,M. J. D., 1973 Animal Cytology and Evolution. Cambridge NACHMAN,M. W., and P. MYERS,1989 Exceptional chromosomal University Press, Cambridge. mutations in a rodent population are not strongly underdomi- WHITE, M. J. D., 1978 Modes of Speciation. Freeman Company, nant. Proc. Natl. Acad. Sci. USA 86 6666-6670. San Francisco. NOVITSKI,E., 1976 The construction of an entire compound two WRIGHT,S., 1970 Random drift and the shifting balance theory chromosome, pp. 562-568 in The Genetics and Biology of Dro- of evolution. pp. 1-31 in MathematicalTopics in Population sophila, Vol. lb, edited by M. ASHBURNERand E. NOVITSKI. Genetics, edited by K. KOJIMA. Springer-Verlag. Berlin. Academic Press, New York. WRIGHT,T. R. F., 1974 A cold-sensitive zygotic lethal causing NOVITSKI,E., and G. BRAVER,1954 An analysis of crossing over high frequencies of nondisjunction during meiosis I in Drosoph- within a heterozygous inversion in Drosophilamelanogaster. ila melanogaster females. Genetics 76: 5 1 1-536. Genetics 39 197-209. ZHANG,P., and R. S. HAWLEY1990 The genetic analysis of distributive segregation in Drosophila melanogaster. 11. Further NUR,U., 1968 Synapsis and crossing over within aparacentric genetic analysis of the nod locus. Genetics 125: 115-127. inversion in the grasshopper, Camnula pellucida. Chromosoma ZHANG,P., B. A. KNOWLES,L. S. B. GOLDSTEINand R. S. HAWLEY, 25: 198-214. 1990 A kinesin-like proteinrequired for distributive chro- ORR,H. A,, 1991 Is single-gene speciation possible? Evolution 45: mosome segregation in Drosophila. Cell 62: 1053-1062. 764-769.. ZIMMERING,S., L. SANDLERand B. NICOLETTI, 1970 Mechanisms J. W. S. Evoluiion in the Genus PATTERSON,T., and STONE,1952 of meiotic drive. Annu. Rev. Genetics 4: 409-436. Drosophila. Macmillan, New York. PORTER,C. A., and J. W. SITES, JR., 1985 Normal disjunction in Communicating editor: A. G. CLARK