Copyright Ó 2009 by the Genetics Society of America DOI: 10.1534/genetics.108.097857
Note Recombination Load in a Chromosomal Inversion Polymorphism of Drosophila subobscura Mauro Santos1 Departament de Gene`tica i de Microbiologia, Grup de Biologia Evolutiva (GBE), Universitat Auto`noma de Barcelona, 08193 Bellaterra (Barcelona), Spain Manuscript received October 21, 2008 Accepted for publication November 18, 2008 ABSTRACT Chromosomal inversions suppress recombination in heterokaryotypes and may help to maintain positive epistatic interactions among groups of alleles at loci contained in the inversion. Here I evaluate the protective effect of inversions on recombination when different chromosomal segments, or even the whole chromosome O of Drosophila subobscura, can be effectively prevented from undergoing recombination in several naturally occurring heterokaryotypes. The fitness of flies made homozygous for recombinant chromosomes was generally lower when compared to their nonrecombinant counterparts, thus suggesting that segregating gene arrangements in this species hold together favorable combinations of alleles that interact epistatically.
C
HROMOSOMAL inversion polymorphisms occur in many plant and animal species, including humans. Three not mutually exclusive balancing selection mechanisms have been suggested for the maintenance of these polymorphisms: coadaptation, supergene selection, and local adaptation. The idea of coadaptation was proposed by Dobzhansky (1950) and entails two components. First, gene arrangements host a set of allelic combinations that interact harmonically within a local population and promote the superiority of heterokaryotypes. Haldane (1957) made clear that coadaptation here means cumulative heterosis and requires at least two loci linked to polymorphic inversions that carry alternative alleles with positive epistasis such that the fitness of heterokaryotypes is higher than expected from the contributions of the individual loci. Second, genetic exchange among chromosomes from different populations would disrupt locally adapted allele complexes and produce unfit offspring. Dobzhansky therefore viewed inversions as suppressors of recombination that help to maintain positive epistatic interactions within local populations. Wasserman (1968) elaborated on the idea of coadapted gene complexes (‘‘supergenes’’) and suggested that if different coadapted combinations of alleles coexist within one gene arrangement in a population, then
1 Address for correspondence: Departament de Gene`tica i de Microbiologia, Grup de Biologia Evolutiva (GBE), Facultat de Biocie`ncies, Edifici Cn, Universitat Auto`noma de Barcelona, 08193 Bellaterra (Barcelona), Spain. E-mail:
[email protected]
Genetics 181: 803–809 (February 2009)
homokaryotypes will suffer an extra disadvantage due to disruption of these favorable epistatic combinations since their chromosomes can freely recombine. Under this (supergene selection) mechanism the proportional reduction of offspring fitness obviously increases with the frequency of the corresponding chromosomal arrangement class. The local adaptation scenario has been recently proposed by Kirkpatrick and Barton (2006). Under this hypothesis inversions are favored even without epistasis since reduced recombination in inversion heterokaryotypes binds together locally adapted alleles and stabilizes them against gene exchange with migrants from other populations or genetic backgrounds. Inversions thus establish themselves in the populations simply as a byproduct of the genetic load introduced by gene flow (migration load). Although epistasis might be unnecessary for the evolution of chromosome inversions, Kirkpatrick and Barton’s (2006) model does not negate it. Therefore, unfit offspring resulting from recombinant chromosomes is not proof against their hypothesis. Recombination load is the loss of fitness because recombination breaks up associations between beneficial combinations of interacting alleles (Charlesworth and Charlesworth 1975). Most of the experiments carried out in Drosophila have compared the fitness of chromosomes derived from males (which do not have recombination) with that of chromosomes derived from females (e.g., Spassky et al. 1958; Dobzhansky et al. 1959; Charlesworth and Charlesworth 1975). The abso-
804
M. Santos Figure 1.—Schematic of chromosome O in Drosophila subobscura. The centromere is placed on the left (solid circle) and the telomere is on the right. The three gene arrangements used in the experiment are labeled on the right side (O314 consists of two overlapping inversions), and the vertical lines conventionally divide the chromosome into two segments: the distal segment (segment I) and the proximal segment (segment II). The fraction of chromosome O that is covered by inversion(s) O314 is 0.268, and by inversion O7 is 0.361. rI (rII ) indicates that recombination can occur in O314 =O31417 ðOst =O314 Þ heterozygous mothers in segment I (segment II). The whole chromosome is protected from recombination in Ost =O31417 females.
lute magnitude of recombination load, as measured by these experiments, is not great but strongly suggests a role for epistasis. However, none of them was specifically designed to test for the protective effect of an inversion on recombination. As far as I am aware only two such experiments were carried out by Wasserman, the first in Drosophila subobscura (Wasserman 1972) and the second one in D. pseudoobscura (Wasserman and Koepfer 1975), with mixed outcomes. A shortcoming of the experiments was a lack of control of the genetic background, which might have seriously hampered the interpretation of results for two reasons. First, interchromosomal assortment can produce variation in fitness (Spiess and Allen 1961) and downwardly bias the estimation of intrachromosomal recombination load (Allen 1966). Second, although recombination is reduced in the region covered by the inversion in heterokaryotypic females, it can increase in other parts of the genome through an unknown mechanism (Lucchesi and Suzuki 1968; Portin and Rantanen 2000). It is my contention here that recombination load in the context of supergene selection as envisaged by Wasserman (1968) has never been properly estimated. The results of an experiment designed to evaluate the extent of recombination load when different chromosomal segments, or even the whole chromosome O of D. subobscura, can be effectively prevented from undergoing recombination in several naturally occurring heterokaryotypes are presented. This species harbors one of the richest inversion polymorphisms in the genus and its O chromosome is by far the most polymorphic, with 24 described inversions in the Palearctic region that form complex gene arrangements with overlapping and nonoverlapping inversions (Krimbas and Loukas 1980). Latitudinal clines in the frequency of many gene arrangements were well documented in native populations (Krimbas and Loukas 1980), and the discovery of parallel clinal patterns a few years after the American
invasion (Balanya` et al. 2003) provided compelling evidence that the clines evolved by natural selection. D. subobscura therefore offers a wonderful scenario to study the evolutionary significance of inversion polymorphisms. The components of fitness examined here were egg-to-adult viability and virility, although the experiment also yielded some information on female fecundity. The results indicate that (within-population) recombination load does seem to be important, at least when flies are made homozygous for recombinant chromosomes, thus suggesting epistatic interactions for fitness within karyotypes. METHODS AND RESULTS
The 18 isochromosomal lines for the O chromosome in an otherwise highly homogeneous genetic background used in this study have been described elsewhere (Santos et al. 2005b). They included 6 independent lines for each of the three gene arrangements Ost , O314 , and O31417 (Figure 1). All lines had a quasi-normal viability according to the recorded proportions of wildtype flies raised in the final crosses to obtain the isochromosomal lines (Ferna´ndez Iriarte et al. 2003). Noninbred virgin F1 females (Table 1) were crossed in population cages to males coming from the Va/Ba (Varicose/Bare) balancer stock (Sperlich et al. 1977), which has the same genetic background as the isochromosomal lines. From the selected offspring a set of 12 crosses was designed with two purposes in mind (Table 2): (i) to estimate egg-to-adult viability for the six different karyotypes (crosses A, B, C, J, K, and L) and (ii) to estimate recombination load by comparing the fitness of flies that inherited chromosomes that had recombined at a particular chromosomal segment, or chromosomes that had freely recombined at any region, against their matching nonrecombinant counterparts
Note
805
TABLE 1 Crosses of F1 females to obtain recombinant offspring for the different segments in the O chromosome of Drosophila subobscura Population cage [cross type] 1 2 3 4 5 6
Offspring selected ($$ 1 1##1 )Va=OrstI 1rII I 1rII ($$ 2 1##2 )Va=Or314 rI 1rII ($$ 3 1##3 )Va=O31417 II ($$ 4 1##4 )Va=OrstII 1Va=Or314
½$$Oist =Ojst 3 ##Va=Ba j ½$$Oi314 =O314 3 ##Va=Ba j ½$$Oi31417 =O31417 3 ##Va=Ba i i ½$$Ost =O314 3 ##Va=Ba ½$$Oist =Oi31417 3 ##Va=Ba ½$$Oi314 =Oi31417 3 ##Va=Ba
($$ 5 1##5 ) Va=Ost 1Va=O31417 I I ($$ 6 1##6 )Va=Or314 1Va=Or31417
Noninbred homokaryotypic F1 females (crosses 1–3) were derived from the isochromosomal lines by a set of cyclically permuted reciprocal crosses O1Di 3 O2Di ; O2Di 3 O3Di ; . . . ; O6Di 3 O1Di ðDi ¼ st; 314; 31417Þ. The three different heterokaryotypes (crosses 4 6) were similarly constructed: O1Di 3 O1Dj ; O2Di 3 O2Dj ; . . . ; O6Di 3 O6Dj ; Di 6¼ Dj. Superscript rI (rII ) means that offspring can inherit a recombinant wild-type O chromosome in segment I (segment II) (see Figure 1).
(crosses D, E, F, G, H, and I). I define rI as the probability of recombination in segment I of chromosome O (Figure 1) and ‘I ¼ 1 (wI0 =wI1 ) as the recombinant load (Charlesworth and Charlesworth 1975) in that segment, where wI1 is the mean fitness of an offspring individual when crossing over was suppressed in inversion heterozygotes for segment I and wI0 is the corresponding mean fitness when derived from a female who was homozygous for that segment. rII and ‘II ¼ 1 (wII0 =wII1 ) are similarly characterized. In all crosses the egg-to-adult viability (Table 3) significantly deviated from the expected 2=3 Va=1: 1=3
1=1 ratio ðGP valuesÞ, with wild-type flies having a higher viability (all values remain statistically significant after a sequential Bonferroni test; Rice 1989). The replicated vials were generally homogeneous for the proportions of both phenotypes, but in several crosses some heterogeneity was detected (GH values). Viabilities were clearly different among the progenies from the various crosses (Table 4). Recombination load (‘‘viability load’’) was estimated from the average viabilities of the appropriate crosses and statistically tested as a linear contrast involving those crosses (Table 4). Note from Table 2 that the expected proportions of the various offspring
TABLE 2 Set of crosses performed to estimate karyotype selection and recombination load in the O chromosome
Cross (flies from offspring in Table 1) A: $$ 1 Va=OrstI1rII 3 ##1 Va=OrstI1rII B: C: D:
I1rII I1rII $$ 2 Va=Or314 3 ##2 Va=Or314 I1rII I1rII $$ 3 Va=Or31417 3 ##3 Va=Or31417 r II II $$ 4 (Va=OrstII 1Va=O314 ) 3 ##4 (Va=OrstII 1Va=Or314 )
E: $$ 5 (Va=Ost 1Va=O31417 ) 3 ##5 (Va=Ost 1Va=O31417 ) I I I I F: $$ 6 (Va=Or314 1Va=Or31417 ) 3 ##6 (Va=Or314 1Va=Or31417 ) I1rII G: ($$ 1 1$$ 2 )(Va=OrstI1rII 1Va=Or314 ) 3 (##1 1##2 ) I1rII (Va=OrstI1rII 1Va=Or314 )
Expected offspring (2=3 Va=1; 1=31=1) with the wild-type phenotype OrstI1rII =OrstI1rII I1rII I1rII Or314 =Or314 I1rII I1rII Or31417 =Or31417
rII rII rII 1 rII 1 rII 1 rII 4Ost =Ost ; 2Ost =O314 ; 4O314 =O314 1 1 1 4Ost =Ost ; 2Ost =O31417 ; 4O31417 =O31417 rI rI rI 1 rI 1 rI 1 rI 4O314 =O314 ; 2O314 =O31417 ; 4O31417 =O31417 1 rI1rII I1rII I1rII I1rII =OrstI1rII ; 12OrstI1rII =Or314 ; 14Or314 =Or314 4Ost
I1rII H: ($$ 1 1$$ 3 )(Va=OrstI1rII 1Va=Or31417 ) 3 (##1 1##3 ) rI1rII rI1rII (Va=Ost 1Va=O31417 )
1 rI1rII I1rII I1rII I1rII =OrstI1rII ; 12OrstI1rII =Or31417 ; 14Or31417 =Or31417 4Ost
I1rII I1rII I: ($$ 2 1$$ 3 )(Va=Or314 1Va=Or31417 ) 3 (##2 1##3 ) rI1rII rI1rII (Va=O314 1Va=O31417 )
rI1rII 1 rI1rII rI1rII rI1rII 1 rI1rII 1 rI1rII 4O314 =O314 ; 2O314 =O31417 ; 4O31417 =O31417
I1rII J: $$ 1 Va=OrstI1rII 5##2 Va=Or314
I1rII OrstI1rII =Or314
I1rII K: $$ 1 Va=OrstI1rII 5##3 Va=Or31417
I1rII OrstI1rII =Or31417
L:
I1rII $$ 2 Va=Or314 5##3
I1rII Va=Or31417
I1rII I1rII Or314 =Or31417
Virgin females and males were let to mate for 3 days. Up to 55 sets with five randomly chosen females each were then used to set up replicated vials for each cross to estimate egg-to-adult viability. 5 means that reciprocal crosses were performed.
806
M. Santos TABLE 3 Results of the egg-to-adult viability from the different crosses in Table 2 Average no. of flies per vial (6 SE)
Cross A B C D E F G H I J K L
Vials 49 48 45 55 50 55 54 51 55 47 47 44
Va=1 88.39 75.94 81.33 82.73 92.52 82.60 76.78 83.37 87.69 102.57 84.00 78.93
6 6 6 6 6 6 6 6 6 6 6 6
4.53 3.49 4.62 4.42 3.55 4.39 4.11 3.54 3.74 4.81 3.72 4.13
1=1
G-tests
6 6 6 6 6 6 6 6 6 6 6 6
GP ¼ 9:99**; GH ¼ 45:89; NS GP ¼ 25:86***; GH ¼ 45:11; NS GP ¼ 52:71***; GH ¼ 25:83; NS GP ¼ 65:84***; GH ¼ 75:31* GP ¼ 79:28***; GH ¼ 51:39; NS GP ¼ 46:91***; GH ¼ 37:23; NS GP ¼ 40:20***; GH ¼ 72:92* GP ¼ 29:76***; GH ¼ 64:29; NS GP ¼ 27:32***; GH ¼ 72:71* GP ¼ 27:83***; GH ¼ 39:40; NS GP ¼ 24:07***; GH ¼ 54:64; NS GP ¼ 10:09**; GH ¼ 38:81; NS
47.94 43.69 49.51 50.38 57.28 48.85 45.19 47.92 49.73 58.23 47.85 43.23
2.38 2.06 3.09 2.57 2.23 2.46 2.29 2.32 2.33 2.54 2.45 2.57
Viability (6 SE) 1.0873 1.1534 1.1950 1.2469 1.2407 1.1943 1.1959 1.1400 1.1334 1.1427 1.1314 1.0804
6 6 6 6 6 6 6 6 6 6 6 6
0.0306 0.0324 0.0245 0.0366 0.0317 0.0266 0.0396 0.0362 0.0352 0.0251 0.0330 0.0324
Heterogeneity G-tests (GP ; G pooled; GH ; G heterogeneity) were carried out as described in Sokal and Rohlf (1995, pp. 715– 724). In each vial viability was estimated as y ¼ 2 3 (no: wild-type flies)=(no: Va flies 1 1), the 1 being to correct a bias from averaging ratios (Haldane 1956). NS; P . 0:05; *P , 0:05; **P , 0:01; ***P , 0:001.
genotypes (karyotypes) in the crosses used to test for recombination load were the same, the only difference being the extent of recombination that had occurred in the F1 females. The (orthogonal) linear contrasts are therefore not biased by the competitive success of a genotype being dependent on the genotypes it competes against. It is clear that average viabilities differ in the expected direction: D . G, F . I, and E . H. The G D load estimates are ‘I ¼ 1 (yCross =yCross ) ¼ 0:0409 I I (loss of mean viability 0.0510; 95% confidence interval 0:0361; 0:1381) for segment I, ‘II ¼ 1 I F (yCross =yCross ) ¼ 0:0510 (loss 0.0609; 95% confidence II II interval 0:0258; 0:1476) for segment II, and ‘I1II ¼ H E 1 (yCross =yCross I1II I1II ) ¼ 0:0812 (loss 0.1007; 95% confidence interval 0:0102; 0:1911) for the whole chromosome. When both (cross E) or none (cross H) of the chromosomal segments were protected from recombination in F1 females, the differences were statistically significant; however, we could even conclude that the
second contrast (F vs. I) was also significant since the lower one-tailed 95% limit was 0:0004. There were slight differences among the average viabilities of the different karyotypes. Using the ‘‘moiety’’ technique in Li (1976), the variation in viability for this multiple-allele (gene arrangement) system was decomposed into a linear and a deviation component (Table 4). Deviation from a linear model was detected, but it was not due to heterosis (actually, the viability of the O314 =O31417 heterokaryotype was lower than those of the corresponding homokaryotypes). The important point here, however, is to note that the difference in mean viability between the fittest (O31417 =O31417 ) and the less fit (O314 =O31417 ) karyotype was 0.1146 (95% confidence interval 0:0182; 0:2110). Therefore, the experiment also provided enough evidence to conclude that the magnitude of the recombination load all along chromosome O of D. subobscura seems to be comparable to the fitness (viability) differences among karyotypes.
TABLE 4 Analysis of variance for egg-to-adult viability Source
d.f.
MS
F
P (parametric)
P (permutation)
Crosses ‘I: D vs: G ‘II: F vs: I ‘I1II: E vs: H Karyotypes Linear Deviations Error
11 1 1 1 5 2 3 588
0.006360 0.003048 0.006616 0.012544 0.004421 0.000005 0.007366 0.002319
2.74 1.31 2.85 5.41 1.91 0.42 3.18
0.0018 0.2521 0.0917 0.0204 0.0914 0.5191 0.0238
0.0018 0.3137 0.0870 0.0318 0.0510
pffiffiffi The response variable was expressed as arcsin p , with arcsin in radians and p ¼ (no: wild-type flies)=(no: total flies) in each vial. Permutation methods were carried out following Manly (1997, pp. 117–125). Each permutation test used 10,000 random permutations.
Note
807
TABLE 5 Results of the ‘‘virility load’’ tests with wild-type offspring males derived from the appropriate crosses in Table 2 No. of flies per bottle (6 SE) Cross
Non-VA flies
D E F G H I
67.70 73.10 72.20 69.20 69.30 68.07
6 2.22 6 2.53 6 2.37 6 1.83 6 2.30 6 2.21
VA flies 146.70 149.53 154.47 157.00 155.20 148.93
6 6 6 6 6 6
4.14 4.42 4.30 3.80 3.89 4.50
Virility (6 SE)
G-tests GP ¼ 9:02**; GH ¼ 29:67; NS GP ¼ 0:75; NS; GH ¼ 44:65* GP ¼ 6:76**; GH ¼ 43:06* GP ¼ 23:28***; GH ¼ 31:64; NS GP ¼ 18:65***; GH ¼ 32:67; NS GP ¼ 11:44***; GH ¼ 22:65; NS
0.9231 0.9817 0.9379 0.8842 0.8913 0.9126
6 6 6 6 6 6
0.0256 0.0332 0.0301 0.0259 0.0250 0.0218
In each case 500 virgin wild-type males were competed in population cages against 500 virgin males from the Va/Ba balancer stock for access to 500 1-week-old virgin Va/Ba females. After 3 days eggs were collected and placed in bottles at low density (30 bottles per cage) to score the progeny as to the type of male the females had mated. Since the expression of the Ba gene is highly variable, the virility of the experimental males relative to the control (Va/Ba) males was estimated as a function of the frequency of non-Va offspring to Va offspring (relative expectations 1=3 : 2=3, respectively): c ¼ 2 3 (no: non-Va flies)=(no: Va flies 1 1). NS; P . 0:05; *P , 0:05; **P , 0:01; ***P , 0:001.
Wasserman’s (1968) model of supergene selection, when specifically applied to Drosophila, requires that recombinant chromosomes produced by homokaryotype females have a dominant deleterious effect on the nonrecombinant chromosomes coming from males. This assumption was tested by crossing offspring progeny in I 1rII Table 1 as follows: ($$ 1 1$$ 3 )(Va=OrstI 1rII 1Va=Or31417 )3 ##5 (Va=Ost 1Va=O31417 ), which somewhat mimics the real situation in Drosophila. From 55 vials set up as described in the footnote of Table 2, the mean viability (6 SE) of wild-type flies relative to that of lethal-carrying (Va=1) heterozygotes was 1:2196 6 0:0367. The ‘‘dominance’’ effect of recombination load can be tested by comparing this number with the viability in cross E (where offspring inherited nonrecombinant O chromosomes; Table 3), which was 2% higher. The difference is, however, not statistically significant (t ¼ 0:578; P ¼ 0:564). Therefore, no conclusive evidence for a dominant deleterious effect of recombination was obtained. Although the egg-to-adult viability experiment was designed mainly to introduce strong competition among developing larvae (there was probably substantial density-dependent mortality in the crowded cultures), some information can be obtained on female fecundity. The average number of offspring per vial in cross E was 149:80 6 5:52 and in cross H was 131:29 6 5:63 (F1;588 ¼ 4:29; P ¼ 0:039). We could tentatively conclude that the egg hatchability was higher in those vials where Va=1 females carried nonrecombinant chromosomes since a similar recombination effect had been previously detected by Wasserman (1972) in chromosome U of D. subobscura. Parenthetically, the average number of offspring per vial in the additional cross performed to test for a dominant deleterious effect of recombination (see above) was 151:58 6 7:73. Again, there is no indication of dominance. The recombination load for virility (virility load) was also estimated following a similar rationale to that used
to estimate viability load. Thus, the actual performance of males was assessed under competitive conditions (Table 5), and virility loads were statistically tested as linear contrasts involving the appropriate crosses. The Spearman rank correlation between virility and viability (Table 3) values is 0.37 (P . 0.05). In most cases virility deviated from the relative expectations (GP values) and its average over all crosses was 0:9218 6 0:0112, with a standard deviation of 0.1503. One-way ANOVA (repffiffiffi sponse variable expressed as arcsin p ) showed that the heterogeneity among crosses was not statistically significant (F5;174 ¼ 1:58; P ¼ 0:168). However, the linear contrast E vs. H (F1;174 ¼ 5:12; P ¼ 0:025) clearly suggested that the average virility of males who inherited recombinant O chromosomes from F1 females was lower than that of males who inherited nonrecombinant ones: H E ‘I1II ¼ 1 (cCross =cCross I1II I1II ) ¼ 0:0921; loss of mean virility ¼ 0.0904 (95% confidence interval 0:0146; 0:1662). DISCUSSION
The method of measuring viability can be criticized on the grounds that it compares 1=1 with Va=1 flies, and the wild-type chromosome in the lethal-carrying flies varies among crosses (Table 2). The heterozygous Va/Ba could have been used as a yardstick against which the fitness of the wild-type flies could be assessed, i.e., by mating $$ Va=1 3 ## Ba=1 derived from the offspring of F1 females (Table 1). However, the expression of the Ba gene (a mutant that reduces the number of macrobristles on scutum, scutellum, and head in heterozygous flies) is highly variable, and the amount of modifier variability is dependent on the O gene arrangement (Alvarez et al. 1981, 1990). Since the Va/Ba stock used here displays a high number of bristles, sorting out the offspring from the previous cross into the right genotypes would introduce serious errors because 1=1 and Ba=1 flies can show the same
808
M. Santos
phenotype. It is difficult to evaluate the source of bias when Va=1 flies are used as a standard, but it seems reasonable to assume that the effect of a recombinant wild chromosome in combination with a Va one is lower than that in combination with another recombinant wild chromosome (actually, there was no clear indication that recombination has a deleterious dominant effect). Some insight can be gained from the pooled data published by Sperlich et al. (1977) on the relative viabilities of wild-type O chromosomes derived from several European localities. In their Table 1 they give the total number of offspring from Va=1 3 Ba=1 crosses, and there is a 92% correlation between 1=1 viabilities as measured using Va/Ba or Va=1 as controls. This piece of evidence suggests that the bias introduced in the present experiment may not be serious. The results from crosses D–I (Table 2) were clear cut: flies carrying recombinant O chromosomes in those regions unprotected from recombination by inversions generally had a lower average fitness than their nonrecombinant counterparts. Statistical significance was detected only when contrasting crosses E vs. H, thus suggesting that recombination load somewhat scaled with the extent of recombination along the chromosome. It is unlikely to have obtained these results if there were not a real and substantial effect of recombination in the base population from which the isochromosomal lines were derived. The careful control of their genetic background probably helped in the detection of a relatively large recombination load effect when compared with the amount of karyotype selection. It therefore seems clear that within-karyotype epistasis is an important ingredient for the evolution of chromosomal inversion polymorphisms in D. subobscura. Deterministic theories assuming epistasis with respect to fitness between the genes located in a chromosomal segment covered by an inversion reveal that there should be linkage disequilibrium in the initial population for selection on the newly arisen inversion to be effective (Charlesworth and Charlesworth 1973). Nonrandom associations between alleles at genetic markers and polymorphic inversions are frequently observed (Krimbas and Powell 1992) and often taken as good evidence of epistatic interactions for fitness. The snag is, however, that they may represent remnants of associations accidentally established at the origin of inversions since initial linkage disequilibrium can decay at a low rate when recombination between two different gene arrangements is small (Ishii and Charlesworth 1977; Nei and Li 1980). If the genes tied up within inversions do not interact epistatically, then differentiation among gene arrangement types is expected to persist only near the breakpoints because double crossovers and gene conversion events in heterozygotes will break up the different haplotypes (Navarro et al. 2000). In D. subobscura, extensive genetic differentiation between O chromosome gene arrangements, extending
over distances as great as 4 Mb, has been detected (Munte´ et al. 2005). This suggests that selection can maintain coadapted gene complexes or, alternatively, that double crossovers are not produced between those gene arrangements in the chromosomal region analyzed. Evidence of within-arrangement epistasis comes from the seasonal variation of linkage disequilibrium between allozyme loci, occurring exclusively in Ost (Fontdevila et al. 1983; see also Rodrı´guez-Trelles 2003). Again, this interpretation is not without caveats because latitudinal clines for Ost occur (Krimbas and Loukas 1980), and this species is known to have a high dispersal rate (Ayala et al. 1989). It could thus be plausible to imagine that (some) gene arrangements are differentiated among geographic populations and that periodical fluctuations in population mixing induce cyclical changes that might result in a greater contribution of migration to the linkage disequilibrium observed than epistasis. In natural populations of D. subobscura it is common to observe a large number of segregating gene arrangements in a given chromosome. Under the local adaptation scenario (Kirkpatrick and Barton 2006) it is easy to envisage that chromosomal arrangements are adapted to diverse environments that can keep them from becoming fixed: the clinal patterns found worldwide (Krimbas and Loukas 1980; Balanya` et al. 2003) provide strong evidence that this may be indeed the case. The idea that variable selection in space is maintaining genetic diversity can be tested by placing flies in a constant laboratory environment. Loss of inversion polymorphism has certainly been observed in cage populations of D. melanogaster (e.g., Inoue 1979). In contrast, a recent laboratory natural selection experiment in D. subobscura showed that although chromosome inversion frequencies consistently shifted according to experimental temperature, every one of the replicated population cages maintained all the initial chromosomal diversity, namely, a total of 16 polymorphic gene arrangements (5 in chromosome O) segregating in the five major chromosomes (Santos et al. 2005a). How this can happen is unclear, but it is intuitively tempting to presume that selection in heterogeneous environments is just one piece in the puzzle of the maintenance of inversion polymorphisms in this species, albeit not necessarily the most important one. Frequency-dependent male mating success also seems to be operating (Santos et al. 1986) and, as has been shown here, supergene selection can play its role too. I thank Walkiria Ce´spedes and Pedro Ferna´ndez Iriarte for their help in obtaining the isochromosomal lines; Hafid Laayouni and Montse Peiro´ for assistance throughout the experiments; Sergey Gavrilets, Carla Rego, and Luis Serra for helpful comments and discussions; and two anonymous reviewers for their constructive criticisms on earlier drafts that helped to significantly improve the manuscript. This research was supported by grants CGL2006-1342301/BOS from the Ministerio de Ciencia y Tecnologı´a (Spain) and 2005SGR 00995 from Generalitat de Catalunya to the Grup de Biologı´a Evolutiva.
Note LITERATURE CITED Allen, A. C., 1966 The effects of recombination on quasinormal second and third chromosomes of Drosophila melanogaster from a natural population. Genetics 54: 1409–1422. Alvarez, G., P. Martı´nez, C. Zapata, M. Santos and A. Fontdevila, 1981 Genetic analysis of modifier variability in Drosophila subobscura. Experientia 37: 1150–1151. Alvarez, G., P. Martı´nez and C. Zapata, 1990 Genetic variation in a modifier system affecting the expression of the Bare mutant of Drosophila subobscura. Heredity 64: 55–66. Ayala, F. J., L. Serra and A. Prevosti, 1989 A grand experiment in evolution: the Drosophila subobscura colonization of the Americas. Genome 31: 246–255. Balanya`, J., L. Serra, G. W. Gilchrist, R. B. Huey, M. Pascual et al., 2003 Evolutionary pace of chromosomal polymorphism in colonizing populations of Drosophila subobscura: an evolutionary time series. Evolution 57: 1837–1845. Charlesworth, B., and D. Charlesworth, 1973 Selection of new inversions in multi-locus genetic systems. Genet. Res. 21: 167– 183. Charlesworth, B., and D. Charlesworth, 1975 An experiment on recombination load in Drosophila melanogaster. Genet. Res. 25: 267–274. Dobzhansky, Th., 1950 Genetics of natural populations. XIX. Origin of heterosis through natural selection in populations of Drosophila pseudoobscura. Genetics 35: 288–302. Dobzhansky, Th., H. Levene, B. Spassky and N. Spassky, 1959 Release of genetic variability through recombination. III. Drosophila prosaltans. Genetics 44: 75–92. Ferna´ndez Iriarte, P., W. Ce´spedes and M. Santos, 2003 Quantitative genetic analysis of wing form and bilateral asymmetry in isochromosomal lines of Drosophila subobscura using Procrustes methods. J. Genet. 82: 95–113. Fontdevila, A., C. Zapata, G. Alvarez, L. Sanchez, J. Me´ndez et al., 1983 Genetic coadaptation in the chromosomal polymorphism of Drosophila subobscura. I. Seasonal changes of gametic disequilibrium in a natural population. Genetics 105: 935–955. Haldane, J. B. S., 1956 Estimation of viabilities. J. Genet. 54: 294–296. Haldane, J. B. S., 1957 The conditions for coadaptation in polymorphism for inversions. J. Genet. 55: 218–225. Inoue, Y., 1979 The fate of polymorphic inversions of D. melanogaster transferred to laboratory conditions. Jpn. J. Genet. 54: 83–96. Ishii, K., and B. Charlesworth, 1977 Associations between allozyme loci and gene arrangements due to hitchhiking effects of new inversions. Genet. Res. 30: 93–106. Kirkpatrick, M., and N. Barton, 2006 Chromosome inversions, local adaptation and speciation. Genetics 173: 419–434. Krimbas, C. B., and M. Loukas, 1980 The inversion polymorphism of Drosophila subobscura. Evol. Biol. 12: 163–234. Krimbas, C. B., and J. Powell (Editors), 1992 Drosophila Inversion Polymorphism. CRC Press, Boca Raton, FL. Li, C. C., 1976 First Course in Population Genetics. The Boxwood Press, Pacific Grove, CA.
809 Lucchesi, J. C., and D. T. Suzuki, 1968 The interchromosomal control of recombination. Annu. Rev. Genet. 2: 53–86. Manly, B. F. J., 1997 Randomization, Bootstrap and Monte Carlo Methods in Biology, Ed. 2. Chapman & Hall, London. Munte´, A., J. Rozas, M. Aguade´ and C. Segarra, 2005 Chromosomal inversion polymorphism leads to extensive genetic structure: a multilocus survey in Drosophila subobscura. Genetics 169: 1573–1581. Navarro, A., A. Barbadilla and A. Ruiz, 2000 Effect of inversion polymorphism on the neutral nucleotide variability of linked chromosomal regions in Drosophila. Genetics 155: 685–698. Nei, M., and W.-H. Li, 1980 Non-random associations between electromorphs and inversion chromosomes in finite populations. Genet. Res. 35: 65–83. Portin, P., and M. Rantanen, 2000 The occurrence of double strand DNA breaks is not the sole condition for meiotic crossing over in Drosophila melanogaster. Genetica 108: 87–90. Rice, W. R., 1989 Analyzing tables of statistical tests. Evolution 43: 223–225. Rodrı´guez-Trelles, F., 2003 Seasonal cycles of allozyme-bychromosomal-inversion gametic disequilibrium in Drosophila subobscura. Evolution 57: 839–848. Santos, M., R. Tarrı´o, C. Zapata and G. Alvarez, 1986 Sexual selection on chromosomal polymorphism in Drosophila subobscura. Heredity 57: 161–169. Santos, M., W. Ce´spedes, J. Balanya`, V. Trotta, F. C. F. Calboli et al., 2005a Temperature-related genetic changes in laboratory populations of Drosophila subobscura: evidence against simple climatic-based explanations for latitudinal clines. Am. Nat. 165: 258–273. Santos, M., P. Ferna´ndez Iriarte and W. Ce´spedes, 2005b Genetics and geometry of canalization and developmental stability in Drosophila subobscura. BMC Evol. Biol. 5: 7. Sokal, R. R., and F. J. Rohlf, 1995 Biometry, Ed. 3. W. H. Freeman, New York. Spassky, B., N. Spassky, H. Levene and Th. Dobzhansky, 1958 Release of genetic variability through recombination. I. Drosophila pseudoobscura. Genetics 43: 844–867. Sperlich, D., H. Feuerbach-Mravlag, P. Lange, A. Michaelidis and A. Pentzos-Daponte, 1977 Genetic load and viability distribution in central and marginal populations of Drosophila subobscura. Genetics 86: 835–848. Spiess, E. B., and A. C. Allen, 1961 Release of genetic variability through recombination. VII. Second and third chromosomes of Drosophila melanogaster. Genetics 46: 1531–1553. Wasserman, M., 1968 Recombination-induced chromosomal heterosis. Genetics 58: 125–139. Wasserman, M., 1972 Factors influencing fitness in chromosomal strains in Drosophila subobscura. Genetics 72: 691–708. Wasserman, M., and H. R. Koepfer, 1975 Fitness of karyotypes in Drosophila pseudoobscura. Genetics 79: 113–126.
Communicating editor: M. Aguade´
