Copyright 0 1995 by the Genetics Society of Amrrica
Competition Between Mitochondrial Haplotypesin Distinct Nuclear Genetic Environments: Drosophilu pseudoobscura vs. D. persirnilis Carolyn M. Hutter and David M. Rand Department of Ecology and Evolutionary Biology, Brown University, Providence, Rhode Island 02912
Manuscript received September 12, 1994 Accepted for publication February 16, 1995 ABSTRACT A test for coadaptation of nuclear and mitochondrial genomes was performed using the sibling species, Drosophila pseudoobscura and D.persirnilis. Two lines of flies with “disrupted” cytonuclear genotypeswere constructed by repeated backcrossing of males from one species to females carryingmitochondrial DNA (mtDNA) from the other species. Each “disrupted” strain was competed in population cages with the original stock of each species from which the recurrent males were obtained during the backcrossing. As such, the two species’ mitochondrial types were competed reciprocally in the nuclear genetic environments of each species. The trajectories ofmtDNA haplotypes werefollowedin discrete-generation population cages using a PCR-four-cutter approach. A significant increase in the frequency of D.pseudoob scura mtDNA was observed in eachof four replicate cages with a D.pseudoobscura nuclear background. In the D.persirnilis nuclear background, one cage actually showed an increase in frequency of D.pseudoobscura mtDNA, although together the four replicate cages show little change in frequency. These resultswere repeated after frequency perturbations and reinitiation of each cage. An analysis of fitness components revealed that fertility selection greatly outweighed viability selection in these cytonuclear competition experiments. The asymmetry of the fitnesses of the mtDNA haplotypes on the two genetic backgrounds is consistent in direction with the previously reported asymmetry offemale fertility in backcrosses between these two species. While our experiments do not allow us to identify mtDNA asthe sole sourceof fitness variation, at a minimum the data indicate a fitness association between nuclear fertility factors and the D. pseudoobscura mtDNA on its own genetic background.
T
HE biosynthesis of a mitochondrial organelle re-
quires the coordinated expression of many different genes located in the nucleus and the mitochondrion (ATTARDIand SCHATZ1988). For example, the electron transport system of the inner mitochondrial membrane is assembled from >IO0 proteins encoded in the nucleus and 13 proteins encoded in the mitochondrial DNA (mtDNA) (WALLACEet al. 1988). Recognizingthatthisnuclear-mitochondrialinteraction has been evolving since the endosymbiotic origins of mitochondria much more than one billion years ago (GRAY1989), it seems likely that there has been strictsense reciprocal coevolution (e.g.,JANZEN 1980) of in(and for teracting nuclear and mitochondrial genes organelle genes in general). One might predict that the physiological state o r fitness ofa n organism carrying its own mtDNA would be superior to that of an organism carrying a foreign mitochondrial genome. Moreover, the degree to which the fitness of an organism might be depressed by carrying foreign mtDNA would presumably be a function of the amount of sequence divergencebetweenthe“native”andtheforeign mtDNAs. In light of the data presented here, and that of other published work, it is important to distinguish Corresponding author: David M. Rand, Dept. of Ecology and Evolutionary Biology, Box G W , Brown University, Providence, RI 02912. E-mail:
[email protected] Genetic.; 140: 537-548 (.lune, 1995)
between fitness studies of organisms carrying mtDNA of another genetically distinct speciesand fitness studies thatmixnuclearandmitochondrialgenomesfrom within a species. The former really address nuclear-mitochondrial coevolution, whereas the latter address the neutrality of mtDNA, or atmost, the incipientstages of nuclear-mitochondria1 coevolution. The issue of mtDNA neutrality is of considerable importance because most studies usingmtDNA as a marker in population biology and systematics assume the neutrality of mtDNA (WILSONet al. 1985; AVISEet al. 1987) . In recent years, several different studies have used functional approaches to address the neutrality of mtDNA (for evidence of nonneutral mtDNA evolution from statistical analyses of nucleotide polymorphisms see WHITTAM et al. 1986; EXCOFFIER 1990; BALLARD and KREITMAN 1994; NACHMAN et al. 1994; RAND et al. 1994). Thework (1988) most explicitly adof CLARKand LYCKEGAARD dresses the fitness effects of nuclearcytoplasmic interactions. CLARKand LYCKEGAARD (1988) studied segregation of second chromosomes in each of several different cytoplasmic backgrounds of Drosophila melanogaster. Significant nuclear-cytoplasmic interaction effects were detected only when the chromosomes and cytoplasms used were from diverse populations; n o effects were detected when chromosomes andcytoplasms were taken from the same population. This result fits nicely with theoretical
538
C. and M. Hutter
expectations of the fate of nonneutral, haploid variants within populations. Adaptive mtDNA variantsshould increase to fixation relatively rapidly and deleterious variants are not likely to persist long in populations. One consequence is that mtDNA variationamong populations is more likely to be of adaptive significance, as observed by CLARK and LYCKEGAARD ( 1988) . Although CLARK and LYCKEGAARD do caution against ascribing the interaction effects directly to mtDNA, their study serves as a clear illustration of the incipient stages of nuclear-cytoplasmic coadaptation. A number of additional studies have used population cage competition experiments to test the neutrality of mtDNA variants. MAcRAE and ANDERSON ( 1988) used subspecies of D. pseudoobscura (from Apple Hill, CA, and Bogota, Colombia) to test the neutrality of mtDNA. They observeda dramatic increase of the Bogota mtDNA over the Apple Hill haplotype in an initial cage on a mixed nuclear genetic background. This result was not repeated in subsequentcages and it has been suggested that unbalanced mating preferences, partial reproductive isolation and cytoplasmic incompatibility between the two strains may explain the apparent nonneutral trajectory in the initial cage ( SINGHand HALE 1990; but see MACand ANDERSON 1990; T. M. JENKINS, C . BABCOCK,D. M. GEISERand W. W. ANDERSON, unpublished data). Whatever the causes of the mtDNA frequency shifts observed by MAcRAE and ANDERSON (1988), it seems likely that the divergence between the subspecies of D.pseudoobscura is greater than that between the different populations of D.melanogaster used by CLARKand LYCKEGAARD ( 1988).As described above, the design used by MACRAE and ANDERSON ( 1988) may bear more on incipient coevolution of nuclear and mitochondrial genes than on theneutrality of intraspecific variants of mtDNA. Using intraspecific mtDNAvariants of D. simulans (restriction variants within the si11 mtDNAtypeofthis species, as described by SOLIGNAC et al. 1986), NIGRO and PROUT( 1990) observed frequency shifts of mtDNA haplotypes in replicate population cages. During their experiment, it was discovered that one of the strains harbored aWolbachia endosymbiont that was responsible for cytoplasmic incompatibility between the two strains. NIGRO and PROUT(1990) determined that the fitness effects of this incompatibility could have generated the same trajectories of mtDNA frequencies and that directselection on mtDNA haplotypes did not need to be invoked. Further evidence that Wolbachia incompatibility agents can affect the fate of mtDNA haplotypes in population experimentswas provided by KAME HAMPATI et al. ( 1992). In population cages involving compatible strains of Aedes albopictus with different mtDNA variants, changes in mtDNA haplotye frequencies could be accounted forby random genetic drift. In similar experiments using strains with a unidirectional cytoplasmic incompatibility caused by Wolbachia sp.,
D. M. Rand
one mtDNAvariant went to fixation in populationcages by the F2 generation. Using D.subobscuru, FOSet al. ( 1990) reported significant frequency shifts of mtDNA haplotypes in experimental populations. While no effort was made to eradicate Wolbachia-likeincompatibility agents, Fos et al. ( 1990) took advantage of naturally occurring inversions to manipulate the nuclear genetic backgrounds of two distinct mtDNA strains of flies. Their data did show significantly different trajectories of mtDNA haplotypes on different nuclear genetic backgrounds suggesting that nuclear-cytoplasmicinteractions may underlie theapparent nonneutral behavior ofmtDNA haplotypes under experimental conditions. Despite the problems with experimental controls reviewed above, it is perhaps inap propriate to consider population cages as definitive experiments addressing the question of mitochondrial neutrality. An answer to the question “Is mtDNA variation neutral?” may be best provided by statistical analyses of static samplesof mtDNA sequence variants (see RAND et al. 1994) . The relevant question for functional studies is “How do specific mtDNA variants behave in a given context of nuclear genetic variation?” Here we report the results of an experiment where mtDNA variants derived from different species (D. pseudoobscura and D.persirnilis) were competed against one another in the nucleargenetic background of each species. Whilewe recognize that species are notuniversally defined biological units, the motivation for the experiment is that the genetic differences between species should provide material for a greater perturbation of nuclear-mitochondrial fitness interactions than comparable perturbations done between subspecies or within species. As a logical extension of the work of MAcRAE and ANDERSON ( 1988),we have chosen D.pseudoobscura and D. persirnilis and employed controls not used in earlier population cage experiments. This species pair produces fertile female hybrids in both directions of the cross ( DOBZHANSKY and EPLING1944) providing a means to manipulatethe cytonuclear genotype by backcrossing. We detected a strongfitness advantage of the D.pseudoobscura mtDNAtype on itsown nuclear background, and counterintuitively, a slight but nonsignificant fitness advantage of the D. pseudoobscura mtDNA on the D. persirnilis background. In both cases, fertility selection is the primary source of the fitness difference. We discuss the results in light of nuclearmitochondrial coadaptationand argue thatinterspecies cytonuclear competition experiments using additional geographic strains of these (and other)species offer a fruitful means of dissecting the geneticbasis of nuclearmitochondrial fitness interactions. MATERIALS AND METHODS
Fly strains: Fly stocks were obtained from the National Drosophila Species Resource Center, Bowling Green, OH. The D. persirnilis stock is numbered 14011-01 11 .O (from Cold Creek,
Fitness Drosophila of mtDNAs in
C A ) , the D. pseudoobscura stock is numbered 14011-0121.0 (from Tuscon,A Z ) . Although the southernlimit of D.persirnilis is not known with certainty, the D.pseudoobscura line used most likely lies outside the range of D.persirnilis reducing the likelihood that mtDNA introgression hasoccured between the two lines (POWELL1983). Moreover, H A L E and BECKENBACH ( 1985) found noevidence for mtDNA introgression in areas where these species are sympatric. We recognize that there may be genetic variation among lines of D. pseudoobscura and D.persirnilis for cytonuclear fitness interactions. The lines used were selected as an initial attempt to determine if any cytonuclear fitness effects are evident. The lines were grown for one generation on 0.3% tetracycline to remove tetracycline-sensitive cytoplasmically transmittedmicroorganismssuch as Wolbachia ( HOFFMANN et al. 1986).Taking advantage of the maternal inheritance of mtDNA and the fact that D. pseudoobscura and D. persirnilis produce fertile F, hybrid females from both reciprocal hybrid crosses of male and female parents, two additional strains were constructed that carried the mtDNA of one species on the nuclear genetic background of the other (strains with “disrupted” cytonuclear genotypes). Hereafter Sp will refer to thestrain of flies with the D.pseudoobscura nuclear chromosomes and the D. persirnilis mtDNA, while Ps will refer to the strain with D. persirnilis nuclear chromosomes and D. pseudoobscura mtDNA, i.e., the uppercase letter refers to the nuclear background and the lower case letter to the mtDNAtype. These two strains were constructed by nine generations of backcrossing. The Sp strain was constructed using D. pseudoobscura as the recurrent male (e.g., female D. persirnilis X male D. pseudoobscura = F1; FI female X male D.pseudoobscura = B, , etc. for nine generations). The Ps strain was constructed using D.persirnilisas the recurrentmale (e.g., female D.pseudp obscura X male D . persirnilis = F, ; F, female X male D. persirnilis = B,, etc. for nine generations). Thebackcrosses were done using 10 virgin females and 10 males each generation, with one exception. At generation B, in the construction of the Ps strain, only a single fertile female survived for mating to D. persirnilis males. In subsequent generations, 10 virgin females and males were used as described above. Using this backcross scheme, the nuclear genome of the strain becomesincreasingly homozygous for thealleles of one species. After nine generations,assuming independent inheritance of alleles, the nuclear genome of such a backcrossed strain will carry 99.8% of the alleles of the recurrent male (i.e., of thegenome will derive fromthe original female parent). Because of linkage it is likely that thelevels ofhomozygosity will be higher than thatpredicted from the expected halving of existing genetic variation with each backcross generation. While recombination can occur during thebackcross process, the X , second and thirdchromosomes differbetween D. pseudoobscura and D.persirnilis by fixed inversions (TAN 1935 ) . These inversions will serve to suppress recombination resulting in large effective linkage groups that will accelerate the homogenization process (see DISCUSSION) . We recognize that residual genetic variation for fitness traits may be liberated by recombination and exist within the backcross lines ( S p and Ps) or between these lines and the original species stocks ( S s and P p ) . However, the cages are initiated and perturbed so as to randomize this variation with respect to cytoplasms (see below). Themain goal of this study is to test the hypothesis that there are repeatable fitness effects between strains with disrupted cytonuclear genotypes. The backcrossing scheme described above will accomplish a sufficient genetic perturbation for a robust test of this hypothesis. Initiation of cages: After completion of the backcrosses, the Sp and Ps lines and the original stocks of D. pseudoobscura ( Ss) and D. persirnilis ( P p ) were grown to large numbers in
-
539
bottles. The mtDNA haplotypes of allfour lines were checked before startingthe cages using aPCR assay ( see below) Competition between mtDNAtypes in each of the two nuclear genetic backgrounds was initiated by placing equal number of eggs carrying the two mtDNAs into 200ml culture bottles with 60 ml of standard cornmeal-yeast extract-sucrose D r e sophila medium, each of which were placed in replicate population cages. The two types of eggs in the D. pseudoobscura nuclear background were obtained by crossing 80 virgin Sp females with 80 Ss males in one bottle, and in a different bottle crossing 80 virgin Ss females with 80 Sp males. The flies were allowed to mate and lay eggs for 48 hr and then were transferred to fresh bottles for additional egg laying. The eggs in the two initial mating bottles were collected and divided into four equalsamples from each of the two mating bottles. Equal numbers of eggs of the two mtDNA types were placed into each of four fresh replicate culture bottles and these bottles were placed inseparate Plexiglas population cages (20 X 18 X 30 cm) . This procedure was repeated for three additional 24hr egg laying periods, resulting infour replicate population cages each holding four culturebottles with equal frequencies of the two mtDNA types. The population cages in the D. persirnilis background were initiated in asimilar manner, starting with reciprocal crosses between Ps and Pp strains. The initiation of the cages from eggs derived from reciprocal crosses between the disrupted and undisrupted strains is intended to randomize mtDNAs with respect to any residual nuclear variation between the backcrossed strain and the undisrupted strain ( S s or Pp) . Maintenance and sampling of cages: The cages were maintained at 23” in a discrete generation, 25day cycle. On the 17th day after the eggs were placed in bottles, the caps were removed and adult flies were allowed to emerge into the population cages. Adult emergence lasted for fourdays (until day 21 ) at which point old bottles were discarded and four fresh 200-ml bottles with 60 ml of standard food were placed ineachreplicate cage. The flies were allowed to lay eggs in the new bottles for four days. After four days the bottles containing the next generation of eggs were cleared of any adults, capped and moved to a clean cage. Population sizes were -1000 adults. The adults were frozen at -80” for later analysis. Adult flies were sampledeach generationand mtDNA haplotype frequencies were estimated every other generation. Estimation of mtDNA haplotype frequencies: mtDNA h a p lotype frequencies were estimated using a squish-amplify-digest protocol. Frozen adult flies were placed individually in wellsof a 96-well flat-bottom microtiter tray and 50 pl of “squishing buffer” [IO mM Tris ( p H 8 . 2 ) ,1 mM EDTA and 25 mM NaCl and proteinase K added to a concentration of 200 pg/ml ( GLOORand ENGE1.S 1991) ] was added to each well. Flies were squished with a 96-pronged pestle designed to fit precisely intothe flat-bottom microtiter tray ( R A N D 1992). The tray of squished flies was incubated at 37” for 20 min and then at 95” for 2 min to denature the proteinase K and any other proteins. This homogenate was used directly as template in PCR amplification reactions. A section of the NADH dehydrogenase (ND5 ) gene was amplified with the primer pair 880R 5 ’ C W G A G G C A TATCACTJ ’ and 2230L: 5 ’AGCTATAGCTGCTCCTACAC3’ as described in RAND et al. 1994 [number = 3’ nucleotide of primer based on the published sequence of GARESSE( 1988) and the letter indicates the direction of elongation with respect to GARESSE(1988), Figure 2 3 . For each sample a 10p1 aliquot of the amplification reaction was digested directly with NlaIII in 20 PI reactions following the manufacturer’s specifications (New England Biolabs,Beverly, M A ) . The 1390-bp amplification product of the ND5 gene of D.pseudoob-
.
540
C . M. Hutter and D. M. Rand
FIGURE I.-Photograph of a rcprcsentativc gel indicating the restriction pattern differences between D. pspurloohsrum (oncecut) and 1). pmsimilis (uncut). The BRL 1-kb ladder is in the left-most lane. A 1390-bp fragment was amplified, digested with NlaIII and run out on 1.5%agarosegels. Frequency estimatesfor each data point were obtained by counting the numbers of the two haplotypes on each gel. scura has an NlnIII site that is lacking in that of D. persimilis (see Figure 1 ) . The NlnIII restriction digest generates two fragment5 -1100 and 290 bp in length that canbeeasily distinguished from the undigested D. persimilis fragments by electrophoresis through 1.5% agarose gelsand ethidium bromide staining ( SAMRROOK d nl. 1990). Thefrequency of the two mtDNA haplotypesfor a particular generation is estimatedfrom the number of digested and undigested fragments in a sample of 48 individuals (see Figure 1 ) Experimental perturbations: After four generations, each cage from both nuclear genetic backgrounds was perturbed back to the starting frequencies of 50% of the two mtDNA types. This was accomplished by collecting 96 mated females from each cage, allowing them to lay eggs individuallyin vials for four days and then determining the mtDNA haplotype of each founding female. This procedure also provided the frequency estimate for the fourth generation (sample size = 96 per cage). With knowledge of the mtDNA haplotype of each female line, the perturbation cages could be initiated with a mixture ofvials carrying offspring of either mtDNA type. Shortly before hatching, the number of pupal cases in each vial was counted. The target frequency of 50% of each mtDNA typewas approximated by placing the appropriate mixture of vials carrying each of the two mtDNA types into each perturbation cage so that an equal number of pupal casesof the two mtDNA typeswas present. The vialswere uncorked beforehatching so that sibmatingwas not forced on the offspring within each vialand random mating of offspring among vialswas allowed. This procedure should randomize mtDNk5 with respect to remaining nuclear variation. Both the original cagesand the new perturbation cages were maintained and sampled as described above. Tests of viabilityandfertility selection: Using isofemale lines ofknown mitochondrial type obtained as described above, a series of specific crosses were made to examine the fertility and viability of flies carrying each of the two mtDNA types within each of the nuclear genetic backgrounds. The experimental design was a three-way analysis of variance using the mtDNA of the female parent, the mtDNA of the male parent and replicate cage as main efrects. Identical designs and analyses were done independently within each nuclear background; for simplicity we will describe only the design
.
for the D. ps~udoobscwanuclear background. Two isofemale lines of each mtDNA type were sampled from each of cages 1 and 2 (replicates 1 and 2 of the D. p.sa~d:ool,sczrranuclear background cages). Eight malesand eight virgin females were isolated from each isofemale line. Four of the virgin females were sihmated individuallytoindividualmales,while the other four virgin females were mated individually to males of the other mitochondrial type from within that cage (hence, four of the eight males from each vial were used in the sib matings and four were mated to females of the other mitochondrial type). Thus, considering one isofemale lineof each mitochondrial type, four replicates of each of the following crosses represent one block of the mating design: Ss female X SS sibling male, Ss female X Sp male, Sj, female X X s male and Sp female X Sp sibling male. These crosses were then repeated in a second block using eight virgin females and eight males from the other isofemale lines of each mtDNA type. Hence, twoblocks containing four replicates of four two typesofcrosseswere done using isofemale lines from replicatecages (cages 1 and 2 ) . Thissamedesign was repeated in the D.persirnilis nuclear background ( i x . , using two isofemale linesof each mtDNA type from each of two replicate cages, Le., cages 5 and 6 ) . The matings were done in vials and the flies were allowed to lay eggs in two successive P-day samples. At day 2 and day 4,the flies were transferred to new vials and the number of eggs laid was counted as an estimateof fertility. The vials were kept at 23" and the date and number of flies hatching out of each vial was recorded. Viability was scored as the percent of eggs hatching from the initial number of eggs laid in each vial. Differences in fertility and viability wereexamined through an analysis of variance with one degree of freedom for each of the main effects (mtDNA type of the mother, mtDNA type of the father, replicate cage) and for each of the interaction effects (mother X father, mother X cage, father X cage). Fertility selection, sf,was estimated from these data using the equation s, = 1 - (average number of eggs produced by females with D. pmirni1i.s mtDNA) / (average number of eggs produced by females with D. pseudoohcum mtDNA) . Viability the equation selection, s,.,was estimated from these data using s, = 1 - (percent ofeggs hatched carrying D. persirnilis mtDNA) / (percent of eggs hatched carryingD. pspucioobsczcm mtDNA) . Hence both viability and fertility selection are exflies carrying pressed in terms of a selection coefficient against the D. pmsimi1i.s mtDNA. Statistical analysesof population cagedata: Statistical analysisof selection on mitochondrial haplotypes i n each cage et was done using the two methods developed by SCHAFFER al. ( 1977). Thefirst method generates a chi-square test statistic, with the degrees of freedom equal to one less than the number of generations observed, againstthe null hypothesis of random genetic drift alone being responsible for the o h servedfrequencychanges. The second method can detect smaller selection coefficients and generates a chi-square test statistic, with one degree of freedom, for the presence of a linear trend of frequency changes in the data. Both methods require an estimateof the effective population size of mtDNA for each generation of data. Thesewere estimated by weighing the sample of flies obtained from each cage,determining the number of femalesin the samplebased on the different weight of males and females and then reducing this female census size by 75%. CROWand MORTON ( 1955) estimate that the effective population size in Drosophila population cages is at last 75% of the census size. From these calculations we applied an estimateof Ne = 200, whichwill make the statistical tests conservative given that the actual N, is probably a bit larger than 200. Two approaches were used for estimating selection coeffi-
Drosophila
in
mtDNAs
of
Fitness
541
D. pseudoobscum background
D. persirnilis background
0.8
0.6 0.4
I
Perturbation
Generation FIGURE 2.-Trajectories of mtDNA haplotype frequencies in replicate populationcages. The ordinate shows the frequency of D. pseudoobscura mtDNA and the abscissa shows the number of generations. One plot(left) shows the results of the cages in the D. pseuahobscura nuclear background (cages 1-4) and the other is for the cages in the D. persirnilis nuclear background (cages 5-8). The original cages (thin lines) and perturbation cages (thick lines starting at generation4) are plotted on thesame set of axes. The observed frequencies in the sampleof adults hatched from the initial F1 cross was taken as generation 0 for each original cage ( i.e., the target frequencies of p = q = 0.5 at the egg stage were not used in any analyses).
cients from changes in haplotype frequencies. One method uses a model of haploid selection where the change in frequency is a function of the selection coefficient and the product of the frequenciesof the alternative haplotypesat generation t + l and t :
a p = sp*+1qr.
(1)
This equation is rearranged to solve for the selection coefficient, s: s=
Ap/pt+l q t .
(2)
Alternatively, sis estimated from the slope of the linear regression of ln[p( t ) / q( t ) ] against time, t (see DYKHUIZEN and HARTL 1980). The latter methodis fine fora range of intermediate frequencies, butcan give biased estimates of s when or loss due to the inflafrequency estimates approach fixation tion or deflation of the logarithmof a ratio containing a number close to0. Neither expression can be evaluated when a frequency estimateis at fixation or loss. RESULTS
D. pseudoobscum nuclear background The patterns of mtDNA haplotype changes in the D.pseudoobscura nuclear background (cages1-4) andin the D.persirnilis nuclear background (cages5-8), are presented in Figure 2. In the D.pseudoobscura nuclear background, the D.pseudoobscura mtDNA haplotype increased from -50 to 80% in the first three generations in the four replicate cages (cages 1-4) . A test for a linear trend in the data ( SCHAFFER et al. 1977) between generations 0 and three is significant for each individual cage, and highly significant when the test statistics from the four repli-
cates cages are pooled (see Table 1) . A test of the null hypothesis that random genetic drift can account for the frequency changes is rejected in three of the four cages individually (cages 2-4), and strongly rejected when the teststatistics from thefour replicates are pooled (see Table 1) . This general pattern of frequency changes was repeated in four new population cages after mtDNA frequencies were perturbed back down to the starting frequencies of 50% in generation four (see Figure 2 ) . In two generations, the D.pseudoobscura mtDNA increased from -50 to 80% (averaged across the four replicate perturbation cages, 1P-4P).All cages showed significant increases in the D.pseudoobscura mtDNA using both the test for genetic drift and that for a linear trend. The pooled statistics for all fourperturbation cages are highly significant for both tests (see Table 1) . During this postperturbation interval the original cages ( 1-4) continued to show an increase in the frequency of the D.pseudoobscura mtDNA, which appeared to have gone to fixation in three of the cages by generation seven (Figure 2 ) . D. persirnilis nuclear background: In the D.persirnilis nuclear background a slight increase in the frequency of the D. pseudoobscura mtDNA was observed in three of the four replicate cages ( cages 5-8; see Figure 2 ) . Averaged across cages, the D.pseudoobscura mtDNA increased from -50 to 65%. Only one of the replicate cages (cage 6 ) showed a significant increase using the test for a linear trend.The pooled statistics for all four cages showa significant linear trend for generations03 and 0-7. However, this result is due almost entirely
C. and M. Hutter
542
D. M. Rand
TABLE 1 Significance tests of mitochondrial haplotype frequency changes model Cage
Linear
Generations
drift d.f.
Random Chi square
P
d.f. square
P
Chi -
D. pseudoobscura background 1 2 3 4 1-4 1P 2P 3P 4P 1P-4P 1 2 3 4 1-4 D. persirnilis background 5 6 7 8 5-8 5P 6P 7P 8P 5P-8P 5 6
7 8 5-8
0-3 0-3 0-3 0-3 0-3 4-7 4-7 4-7 4-7 4-7 0-7
0-7
3 3 3 3 12 3 3 3 3 12
7 7 7 7
0-7 0-7 0-7
28
0-3 0-3 0-3 0-3 0-3 4-7 4-7 4-7 4-7 4-7
3 3 3 3 12 3 3 3 3 12
0-7
7
0-7 0-7 0-7
7
0-7
7 7 28
5.35 8.09 34.08 9.44 56.96 14.99 26.67 22.76 20.93 85.35 25.03 19.79 39.67 34.61 119.10
