Key words: Drosophila, population genetics, transposable elements ... polymorphism of transposable elements from various kinds of populations (natural ...
Genetica 86: 67-84, 1992. © 1992 Kluwer Academic Publishers. Printed in the Netherlands.
Population genetics of transposable DNA elements A D r o s o p h i l a point o f view
C. Birmont
Laboratoire de Biomdtrie, G~n~tique et Biologie des Populations, URA 243, Universit6Lyon 1, 69622 Villeurbanne, France Received 2 March 1992
Accepted 26 March 1992
Key words: Drosophila, population genetics, transposable elements Abstract
This paper is an attempt to bring together the various, dispersed data published in the literature on insertion polymorphism of transposable elements from various kinds of populations (natural populations, laboratory strains, isofemale and inbred lines). Although the results deal mainly with Drosophila, data on other organisms have been incorporated when necessary to illustrate the discussion. The data pertinent to the regions of insertion, the rates of transposition and excision, the copy number regulation, and the degree of heterozygosity were analysed in order to be confronted with the speculations made with various theoretical models of population biology of transposable elements. The parameters of these models are very sensitive to the values of the transposable element characteristics estimated on populations, and according to the difficulties of these estimations (population not at equilibrium, particular mutations used to estimate the transposition and excision rates, trouble with the in situ technique used to localize the insertions, undesired mobilization of TEs in crosses, spontaneous genome resetting, environmental effects, etc.) it cannot be decided accurately which model better accounts for the population dynamics of these TEs. Tendencies, however, emerge in Drosophila: the copia element shows evidence for deficiency of insertions on the X chromosomes, a result consistent with selection against mutational effects of copia insertions; the P element repartition does not significantly deviate from the neutral assumption, in spite of a systematic copy number of insertions higher on the X than on the autosomes. Data on other elements support either the neutral model of TE containment, neither of the two models, or both. Prudence in conclusion should then be de rigueur when dealing with such kind of data. Finally the potential roles of TEs in population adaptation and evolution are discussed.
Introduction
The discovery that transposable elements (TE) are common components of genomes has stimulated passionate debate about the nature of the forces that affect their distributions in genomes from populations. As a result of their giant salivary gland chromosomes, most of the data currently available are from Drosophila in which the indentification of the insertion sites of transposable elements is permitted by the technique of in situ hybridization. Although this technique is somewhat coarse, in that the positions of the sequences can only be established to the
level of the salivary chromosome bands, it provides a useful picture of the general properties of the elements in the population. Moreover, the entire genome (though only the euchromatin) can be studied in this way, as opposed to the restriction mapping of small genomic portions. Although there is now an increasing accumulation of data on insertion site polymorphism, most of the Drosophila data in which sites have been precisely localized are from the X chromosomes with only four studies of natural populations (Montgomery & Langley, 1983; Ronsseray & Anxolabrhrre, 1986; Leigh Brown & Moss, 1987; Charlesworth &
68 Lapid, 1989). There have been a few studies in which the sites of insertion were localized on all the major chromosomes, for a number of elements. These studies have made it possible to compare the properties of chromosomes and of various elements. They were done, however, with inbred lines or laboratory strains (Belyaeva et aL, 1984; Pasyukova et aL, 1988; Biemont, 1986; Birmont & Gautier, 1988; Birmont et al., 1990b). All other studies with wild populations assessed copy numbers only and did not establish the site positions (Montgomery et al., 1987; Ronsseray et aL, 1989). Theoretical and empirical studies yield to the conclusion that the transposable elements are maintained in populations as a result of a balance between the transpositional increase in copy number and some opposing forces. These forces include regulation of the rate of transposition with increasing copy number, selection against insertional mutations, and induction of deleterious chromosome rearrangements by unequal recombination. Thus far, in view of the natural population data alone, it seems impossible to decide which of these forces play a major role in maintaining the element copy number in populations. Various syntheses have been published on different aspects of the transposable element characteristics (Brookfield, 1986; Charlesworth, 1985, 1988; Ajioka & Hartl, 1989; Echalier, 1989; Finnegan, 1989; Mackay, 1989) and the readers should refer to these works. Here I only analyse the whole set of data available in the literature on transposable element distribution from either natural or artificial populations, and I discuss the way these data are obtained, in order to search for any tendency that could emerge and then help us better understand the way these elements behave in populations. Although this literature compilation mainly deals with the restricted knowledge acquired on Drosophila, I also use data from other organisms when relevant to the discussion either to corroborate the conclusion or to refute the arguments, and in all cases to enlarge the discussion.
The observed data
The regions of insertion The transposable element DNA sequences represent about ten percent of the Drosophila genome.
With an average TE size of about 5 kb, the Drosophila genome contains about 3000 copies of TEs of 30-50 different families. A view of the giant chromosomes from the salivary glands could thus show that these TEs are integrated into almost every band (Ananiev et al., 1984). It is thus of importance to understand how such a high number of insertions are maintained in the genome without producing too high a pressure of mutations or chromosomal rearrangements. There is a large number of potential sites into which members of each family of TE can integrate. These insertion sites are apparently dispersed at random along the chromosomes; even so some TEs may insert in a site-specific fashion. Indeed, the 297 element shows site specificity for the sequence TATA (Ikenaga & Saigo, 1982) while Tom prefers TATAT (Tanda et al., 1988), gypsy inserts into TACATA and generates a duplication of the sequence TACA (Freund & Meselson, 1984), the P element inserts with a high degree of location site specificity in rich-GC consensus sequence GCCCAG (O'Hare & Rubin, 1983) or in genes expressed in the germ line (Bowes, 1990), but mdg1 and copia, that belong to copia-like elements bounded by 5' TG .... CA 3', and resemble vertebrate retrovirus provirus, insert into the chromosome without obvious site specificity (Inouye et aL, 1984). At first it was a widely accepted view that viral sequences integrated at random into the genome of infected cells. Integrated forms of retroviruses have been shown, however, to be located in a limited number of sites used as targets at very high frequency (Shih et al., 1988). Similarly, Salinas et al. (1987) observe that some viruses integrate preferentially in isochore compartments having a matching base composition, and retroviral integration does not occur at random but is influenced by the chromatin conformation of the host genome (Taruscio & Manuelidis, 1991). Retroviruses seem then to preferentially integrate into open, hypersensitive chromatin regions, for example, near the DNAse I-hypersensitive site (Gridley et al., 1987). It is striking, as shown by Voelker et al. (1990), that the regions into which the Drosophila P insertiotls occurred contain a number of DNAse I hypersensitive subregions. These authors thus suggest that the selection of insertion sites by P elements is based on some criterion other than simple nucleotide sequences.
69 The visual inspection of the distribution of various elements on the chromosomes suggests that each element has its preferential regions of insertion (Belyaeva et aL, 1984; Biemont & Gautier, 1988; Ronsseray & Anxolabrhrre, 1986), and that some particular sites may be nests of transposons (Gvozdev, 198 I). The division 42B for example is very impressive. All four nomadic segments of Young & Schwartz (1981), the four elements mdg1, I, copia and P in the study of Birmont and Gautier (1988) and many other sequences (297, 412, copia, mdg-I, mdg-3, roo; see Potter et al., 1979; Strobel et al., 1979; Pierce & Lucchesi, 1981; Ananiev et al., 1984; Belyaeva et aL, 1984; Birmont, 1986; Bi~mont & Aouar, 1987; Leigh Brown & Moss, 1987; Montgomery et al., 1987; Birmont & Gautier, 1989) hybridize to this chromosomal region. Moreover, Ananiev et al. (1978), Belyaeva et al. (1984), Gvozdev ( 1981) postulate that a considerable proportion of the mdg-1 copies are found in regions of intercalary heterochromatin (regions showing ectopic pairing, late replication, and presence of weak points). Such results have not been reported for other elements, however (Ronsseray & Anxolabrhrre, 1986), although insertions in ectopic fibers can often be observed. Hence the copy number distribution of most elements is now considered to be usually Poisson (Charlesworth & Lapid, 1989) with low frequencies at individual chromosomal sites, although the variance of the P and 297 element copy numbers in populations analysed by Yamaguchi et al. (1987) were highly significantly greater than those expected under the Poisson distribution. The distribution of Ac-like sequences in inbred maize also does not fit the expected Poisson distribution. There is a 'non-random' chromosomal distribution of these sequences over the chromosomes with a large number of elements found on chromosome 4 (Jones et al., 1990). Although it is evident that the behaviour of an element in a genome may interact with the behaviour of others, the data on site co-occurrences are very scarce. In a sample of X chromosomes from North Carolina, Montgomery and Langley (1983) found that the elements 297 and 412 tend to cooccur at chromosomal sites more frequently than on the basis of random expectation, and an analysis of insertions of four mobile elements (mdg-1, copia, I and P) in genomes of highly-inbred lines of D.
melanogaster reveals that the P and mdg-1 elements tend to avoid each other, whereas mdg-1 and copia, two copia-like elements, tend to insert preferentially in sites where one of them is already present; I and mdg-1 share the same bulk of potential insertion sites without any kind of interaction (Biemont & Gautier, 1989). It is thus proposed that some regions of the genomes may have particular chromosomal environments (DNA sequence, local chromatin structure) suitable or unsuitable for insertion of different elements. Charlesworth and Lapid (1989) found a correlation between different elements families with respect to the identities of the sites occupied at least once in the chromosome sample; the absence in their study, however, of a within-chromosome effect suggests that the correlation in occupied sites reflects some degree of shared site-specificity for insertions, or protection in certain regions against excision. The rates of transposition and excision The experiments aimed at estimating the rate of transposition (the probability of insertion in a new site) in laboratory lines suggest that this rate is fairly low and ranges approximately from 10 -3 to 10-4 per element per generation (Young & Schwartz, 1981; Pierce & Lucchesi, 1981) or even to 10 -6 (Aquadro et al., 1990; Harada et al., 1990). Transposition appears, however, to be discontinuous in time, to vary in rate among different transposon families. It is not equally active in different Drosophila laboratory populations (Junakovic et al., 1987; Birmont et aL, 1987). Transpositions could follow the mutation rate which has been found to differ among regions of the mammalian genome (measured by variation in the silent substitution rate and variation in base composition, Wolf et al., 1989). Junakovic et al. (1984) think that some kind of induction (variation in food, fly density, temperature) could even periodically produce a reshuffling of the TEs or increase previously low rates of ~ransposition (See article by C. Di Franco, D. Galuppi and N. Junakovic in this volume), similar to what is thought to be the case in hybrid dysgenesis (Bregliano & Kidwell, 1983). It is commonly admitted that rates of excision (the probability of loss of an element in a site, i.e., the element is no longer detectable by in situ hybridization, or a reversed mutation is observed) are
70 ten times smaller than rates of transposition. The idea that the loss of elements from their chromosomal locations is very rare comes, however, from the observation of the great stability of the majority of mutations caused by TE insertions. We must consider the fact that such observations concern only mutations that were detected visually and that result from a TE insertion. These mutations do not concern the whole set of insertions on the chromosomes contrary to what is seen in studies dealing with location of all chromosomal sites by in situ hybridization. In studies on highly inbred lines of Drosophila melanogaster, the transposition and excision rates were found similar and equal to 2 x 10.3 and 1.6 X 10-3 per P element, per generation, respectively (Birmont et al., 1990b) and to 1.4 X 10 -3 and 4.6 x 10 -3 per element, per generation, respectively, for the mdg-1 element (Birmont & Aouar, 1987). These rates were equal to 2.5 x 10 -4 and 3 x 10 -4 for the FB element (Ising & Block, 1981), 7.7 X 10 -6 and 3.4 x 10 -6 for the BN element (Harada et al., 1988), and 5 X 10-5 and 7.5 x 10-5 for a set of 12 families of elements (Eggleston et al., 1988). It is true, however, that in situ hybridization can not discriminate between a true deletion (a precise removal of elements) and an imprecise one in which as much as an LTR (Long Terminal Repeat) can remain as a landmark. But, that the excision rate should concern only precise deletions or could include imprecise ones is not at all evident because we do not know much about the real mechanism of excision of most elements. The homologous recombination between the LTRs of the retroviruslike elements mimics excision and indeed an LTR is left behind; the choice between precise or imprecise removals will of course influence the estimation of this deletion parameter. Other considerations may influence these estimations: for example, in Antirrhinum majus there are many stable mutations due to the insertions of the Tam2 element; their stability may be due, according to Hudson et al. (1987), to the absence in the genome of activating factors and to the presence of repressors; and in Drosophila a high-frequency P element loss is homolog dependent (Engels et aL, 1990). The analysis of populations not at equilibrium for their genomic copy number could also explain the differences sometimes reported between the transposition and excision rates (Preston & Engels, 1984).
The copy number regulation One important observation for mobile element biology is that despite great variability in location of elements among strains, the upper-limit copy number of each family remains fairly constant, and tentative experiments to change copy number were generally not successful. In an attempt to increase I copy number during a chromosomal contamination process which follows a dysgenic I-R cross, Prlisson and Brrgliano (1987) were unable to increase copy number beyond 10-15, suggesting thus that I element copy number is very precisely regulated. One way to increase our knowledge of copy number regulation in genomes and populations is to gather information on the distribution over the chromosome arms of various transposable elements. A simple analysis of such data can consist of calculation of correlation coefficients over chromosome arms. By doing this in Drosophila, Montgomery etal. (1987) found a significant association in one of the eighteen cases tested between the element roo/2L vs. roo/2R, but they rightly concluded that this may not be meaningful in view of the large number of tests done. No significant correlation between the copy number of insertions in the different chromosome arms was observed for the I element in another Drosophila population analysed by Birmont (1986), whereas the copy number for mdg-1 elements on the 3R chromosome arm was negatively correlated with the copy number on the X chromosome. Note, however, that such a correlation was not found in a previous experiment with mdg-1 suggesting strongly that such a relationship was only fortuitous. Moreover, Ronsseray and Anxolab~h+re (1986) noticed a correlation between the X and the 3R for the P elements, and between the 3R and the 2L for the I elements. All these various observations can only lead to the conclusion that the correlation is not meaningful in view of the large number of tests done, which makes the probability of obtaining a statistically significant test quite high. To overcome this difficulty due to multiple individual tests, Birmont and Gautier (1988) and Biemont et al. (1990b) proposed a nonparametric statistic to analyse their data on inbred lines. They worked with the variances of the number of copies per inbred lines and the variances of the number of copies per chromosome arms; they assumed that in
71 the absence of a genomic control on copy number, the variables 'number of element copies in each chromosome arm' must be statistically independent, implying that the observed variance of the total copy number equals the sum of the variances of copy numbers in each arm. It thus appeared that the retrovirus-like transposable elements copia and mdg-1, and the I element, all of which have a DNA sequence for a reverse transcriptase, differ in their copy number regulation from P and hobo, which code for a putative transposase (Birmont et aL, 1988). Only the copy number of the former class of elements is submitted to genomic control (i.e., a compensatory effect among chromosome arms for the number of insertions; if one or more chromosome arms have many copies of an element, other chromosomes have few copies), and a correlation between the numbers of copies of elements from these families was observed. With the P and hobo elements, however, lines with high or low numbers of insertions of all the chromosome arms have been obtained, and contrary to the retrovirus-like elements, these two elements appeared to be independently regulated. When dealing with copy number regulation, however, it is generally assumed that all the inserted sequences are submitted to the regulation, whatever their region of localization on the chromosomes, and whatever their structure or their active or inactive state. We have no reason, however, to consider all the insertions sites as equivalent, and the P elements are good illustrators of this problem with their nonautonomous and autonomous elements that can in addition be inserted in some regions with specific characteristics (Biemont et aL, 1990b; Ronsseray et al., 1991). The total number of copies may thus not reflect the true number of copies that are submitted to the regulation process. The knowledge of the proportion of active and inactive copies may not, however, solve the problem; defective nonautonomous copies can indeed be mobilized by even one active copy, as it is the case with the P elements (Ronsseray et al., 1991). Data on chromosome arm regulation of copy number are unfortunately not yet available for natural populations. We thus need more data concerning the precise locations of transposable elements on the entire genome (all the chromosome arms in Drosophila) from various natural populations.
The degrees of heterozygosity in insertion sites So far we do not have any direct estimation of the degree of heterozygosity (the proportion of sites with one insertion in a chromosome and none on the homologous one) of genomes from natural populations, and little idea whether this heterozygosity differs depending on the transposition and excision rates. A current point of view comes from localizations of insertions in the X chromosome in Drosophila males, in which the very low values of frequencies at individual chromosomal sites suggest a high degree of insertion site heterozygosity (Montgomery & Langley, 1983; Leigh Brown & Moss, 1987; Charlesworth & Lapid, 1989). Two studies, however, either from inbred lines or from the population from which the inbred lines originated, gave indirect estimations of the degree of heterozygosity of the entire genome for the mdg-1, I, copia, and P transposable elements (Birmont & Gautier, 1987; Biemont & Gautier, 1988). The estimated average proportions of heterozygous sites per individual were equal to 0.86, 0.92, 0.90 and 0.94 for the mdg-1, L copia, and P elements, respectively. There was no difference in the estimates obtained either from the theoretical hybrids resulting from crossing the inbred lines or from the original population, suggesting that although inbreeding during many generations purges the population of mutant alleles (Barrett & Charlesworth, 1991) or can change the insertion patterns because of recombinations, the information obtained from the inbred lines is not such a bad reflection of what exists in the original population. In the study of Biemont and Gautier (1987, 1988) no significant correlation was found between the four mobile element families for the heterozygosity levels of the individuals. One mobile element thus does not give a precise idea of the overall genome heterozygosity. It may be concluded that the genome is not homogeneous and that its degree of heterozygosity depends on the mobile element family concerned (and thus on the transposition and excision rates). In such cases, it is quite confusing to calculate the degree of heterozygosity by lumping together various elements in a single ensemble, as incorrectly done many times with enzymatic data (see Lewontin, 1974 for a discussion). Genetic diversity measured in terms of mobile elements is thus not a good estimate of the actual
72 heterozygosity of the entire individual genome (in the hypothesis that this total heterozygosity has a real meaning), and the divergence among lines, measured by this approach, may not reflect the potential divergence determined from enzymatic polymorphism. Because of transposition, recombination, copy number regulation, and genome reshuffling (Bi6mont et aL, 1987), the genome may indeed encounter some changes independent of the enzymatic locus background. This could shed new light on the divergent relationships reported between enzymatic heterozygosity and morphological or physiological characteristics (Frei et aL, 1986; Strauss, 1986). These relationships have been widely debated, but there is still no consensus as to what kind of genes or how many loci, can actually give a good indication of the heterozygosity of the genome. We must, however, raise the point that any level of partial inbreeding in a population will lead to a significant correlation between the degrees of heterozygosity calculated on different TEs. This is because the increased homozygosity of some genomes will affect the insertion polymorphism of the various elements in the same way (here a decrease in heterozygosity); a correlation between the estimated values of TE insertion heterozygosity will thus result when all the individuals (inbreds and non inbreds) of the population submitted to partial inbreeding are considered.
The models for copy number maintenance The models and their tests
To account for the copy number regulation of most TEs in populations, many attempts that describe the frequency distributions of these transposable elements have been proposed, and parameters of the models estimated (Charlesworth & Charlesworth, 1983; Kaplan & Brookfield, 1983; Langley et al., 1983; Charlesworth, 1985). Statistical analyses of population data suggests that elements are maintained by a balance between transpositions and forces removing them from the populations. A first model (generally called the 'neutral model') proposes that element copy number is regulated by copy number dependent transpositions (the transposition rate per element declines with the number of copies) or excision independent of chro-
mosomal location. With this model an equiproportionality between the X and the autosomes is expected in the Drosophila genome. A second model (called the 'transposition/selection model') proposes that elements are eliminated by natural selection against mutational effects of their insertion into the chromosomes. This model assumes equal negative selective effects of all insertions (this could be approximately true when the frequency of each site is low: Charlesworth, 1985); then insertion will be mainly heterozygous, and a stable copy number is attained at (~-v)/s, where is the transposition rate, v is the probability of element excision, and s is the selection coefficient associated with the TE insertions. Since p~and v are low, a small selection coefficient is supposed to be sufficient to maintain equilibrium copy number of 20-50 elements per genome. This model predicts that in Drosophila the proportion of the elements on the X chromosomes will be smaller than the proportion on the autosomes, since recessive or partially recessive deleterious elements will be selected against in the hemizygous male carriers. An alternative to this model is that selection could act against dominant lethal chromosome mutations produced by unequal exchange between insertions of TEs over non homologous regions of the chromosomes (Langley et aL, 1988). Favorable mutations and occasional beneficial effects on fitness, as demonstrated for TEs in bacteria (see Ajioka & Hartl, 1989), have not been taken into account in theoretical models. Indeed, although such effects may in the long run be of great importance for population evolution or adaptation, their selective advantages (resistance to antibiotics, increased growth rate in bacterial chemostats, for example) seem too low in intensity and duration to greatly affect the short-term population dynamics of TEs in populations not submitted to artificial selection. Moreover, that such effects involve mechanisms directly related to transposition or mutation, or are simply the result of a generalized increase in the mutation rate or of any kind or physiological phenomenon, is not at all clear. The same cautious considerations hold for the observations of relationships between viability values and mobile element copy numbers in some experiments with Drosophila populations (Pasyukova et al. 1986; Biemont & Terzian, 1986; Bi6mont et al., 1989).
73 Montgomery et al. (1987) have modelized these neutral and negative selection hypotheses and predicted a frequency of elements on the X of 0.11 with selection against deleterious insertions, and of 0.17 if insertions are neutral and the copy number is stabilized by a balance between regulated transpositions and excisions. These values became 0.133 and 0.199 when corrected by taking into account the presence of two X chromosomes in a female and only one in a male (Langley et al., 1988). Note that with a non-negligible rate of excision the frequency expectation in the transposition/ selection model is closer to that of 0.17 (or 0.199) given by the neutral hypothesis. To test these two models we have gathered all the data available from the literature, from natural populations, massmated populations, isofemale lines, and inbred lines, on which statistical analysis could be done. We thus got information of elements as diverse as
mdg-1, mdg-3, copia, 412, 297, roo, L FB-NOF, and P. As seen in Table 1, although the P element follows neither of the two models (neutral or selected) under the first estimation of the model parameters (XZa values), it fits the neutral hypothesis (the equiproportional model) when the corrected coefficients are used (XZbvalues). The proportion of P insertions on the X chromosomes does not then show any significant excess as compared with the autosomes, at least under these present models. The FB-NOF element fits neither of the two models, but the limited number of populations analysed is not sufficient to conclude with great confidence. For the elements mdg-l, mdg-3, 412, 297, roo, and L the tendency of the ×2a values is in favor of the neutral model, even though a few cases sustain the selected model. The X2bvalues, however, show disagreement with both models. This shows that no firm conclusion can be drawn from a study on only one population, and this should weaken the idea that 412 obeys the selected model as postulated by Montgomery et al. (1987). The value of the chi2 and hence its statistical significance depends, moreover, on the sample size. So, although we have recorded in the Table only data of reasonable and close sample size, smaller sizes may account for the non-significance of some of the chi2 values. Copia should, however, attract more of our attention. Indeed, in the four studies none favors the neutral model, and in all experiments the chi2 values are
higher for the neutral model than for the selected model. In all, copia shows the tendency to follow the selected model. Before trying to imagine biological explanations for this particular behavior, we propose that more experiments be done to verify whether insertions of copia lead to detrimental mutations eliminated by natural selection more than other elements do.
The validity of the models The above transposition/selection model assumed that insertions into all sites have similar selective effects, if any, and that extremely small selection coefficients (less than the transposition rate values o f 10 -3 to 10-6) are sufficient to contain the TE copy number in a genome (Charlesworth & Langley, 1989). This is hardly compatible with the calculated mean selection coefficient of 0.02 against lethal and semilethal heterozygous mutations for either newly arisen mutants or population at equilibrium (see Simmons & Crow, 1977; Crow & Simmons, 1983). We must keep in mind, however, the fact that in some cases the viability of newly arisen heterozygous mutants can be reduced by 2.5% at 25°C and 17°C but increased by 1.5% at 29°C (Tobari & Murata, 1970). In addition, a lethal second chromosome of D. pseudoobscura may confer a viability advantage on their heterozygous carriers if the lethals were tested in a native genetic background, while in an alien background the viability of heterozygous lethals is lowered by 5 % (Anderson, 1969). It is quite amazing that recently Suh and Mukai (1990) observed a homozygous detrimental load of chromosome 2 higher in a foreign background than in a native background; they concluded that it was possible that new mutations were induced by dysgenesis in the process of extracting the second chromosome (this latter point is emphasized in the discussion section below). First of all we will accept that mutations affecting viability of populations reduce fitness by about 2% in the heterozygous condition, a value much higher than the estimated rates of transposition per element per generation. We can, however, question the validity of the selection coefficient for viability since this estimate is an average value including all mutations affecting viability and not only those induced by transposable elements. Although P elements seem to induce mainly lethal and severely
383 285 493 220 303 105 511 124 115.3 411 92.4 454 1225 427 298 487 326 60 316 721 690 659 533 522 523 144
mdg-1 mdg- 1 mdg-1 mdg-3 copia copia copia copia 412 412 297 297 roo I I I I FB-NOF P P P P& P(P) P(Q) P(M') P
0.16 0.17 0.09 0.23 0.09 0.12 0.12 0.10 0.17 0.13 0.24 0.20 0.19 0.16 0.19 0.11 0.17 0.30 0.22 0.24 0.22 0.19 0.23 0.24 0.22 0.23
Proportion on X
0.1 0.0 21.7"** 5.1" 14.0"** 3.1 10.7"** 4.1" 0.0 4.9* 3.0 2.2 2.5 0.4 0.7 23.6*** 0.0 7.19"* 6.7** 54.2*** 13.1"** 1.5 14.4"** 17.4*** 7.71"* 3.5
t 1.6"** 9.9** 1.8 30.9*** 1.4 0.4 0.5 0.1 3.9* 1.0 15.5"** 31.2"** 66.8*** 10.6"* 18.5"** 0.0 13.7"** 22.1"** 42.5*** 255.5*** 88.0*** 40.7*** 80.6*** 88.1"** 59.6*** 20.6***
neutral neutral selected neither selected both selected selected neutral selected neutral neutral neutral neutral neutral selected neutral neither neither neither neither neutral neither neither neither neutral
2.9 1.7 35.9*** 1.1 23.0*** 7.3** 23.3*** 7.3** 0.7 12.6"** 0.9 0.0 1.1 4.2* 0.2 49.2*** 1.3 3.8* 1.3 16.0"** 2.2 0.5 3.6 5.1 * 0.9 0.8
Neutral (0.199)
Inference
Neutral (0.17)
Selected (0.11)
X 2b
X ~-a
3.3 3.1 7.4** 16.9"** 5.1" 0.1 1.0 1.0 1.2 0.06 8.9** 15.6"** 30.9*** 2.6 7.8** 4.3* 4.8* 14.5"** 23.0*** 145.8"** 47.1"** 17.1"** 45.0*** 50.4*** 31.0"** 11.4"**
Selected (0.133) both both neither neutral neither selected selected selected both selected neutral neutral neutral selected neutral neither neutral neither neutral neither neutral neutral neutral neither neutral neutral
Inference
masse-mated population inbred lines laboratory strains laboratory strains inbred lines natural population natural population laboratory strains laboratory strains isofemale lines laboratory strains isofemale lines isofemale lines masse-mated population inbred lines isofemale lines natural population laboratory strains inbredlines isofemalelines isofemale lines isofemale lines natural population natural population naturalpopulation selected lines
Population
Bi6mont 1986 Bi6mont & Gautier 1988 Belyaeva et al., 1984 Belyaeva et al., 1984 Bidmont & Gautier 1988 Bi6mont et aL (unpublished) Yamaguchi et al., 1987 Strobel et al., 1979 Strobel et aL, 1979 Montgomery et aL, 1987 Strobel et al., 1979 Montgomery et aL, 1987 Montgomery et al., 1987 Bidmont 1986 Biemont & Gautier 1988 Ronsseray & Anxolab6h6re 1986 Ronsseray et al., 1989 Harden & Ashburner 1990 Bidmont & Gautier 1988 Ronsseray & Anxolabdh~re 1986 Eanes et al. 1988 Eanes et al. 1988 Ronsseray et al., 1989 Ronsseray et al., 1989 Ronsseray et al., 1989 Shrimpton et al., 1990
References
×2a: chi square values calculated according to Montgomery et aL (1987); ×2b: chi square values calculated according to the models in Langley et al. (1988) which take into account the presence of two X chromosomes in a female and only one in a male; P~': the same isofemales lines as above but after having excluded the high frequency tip site 1A known to bear P elements capable of eliciting the P cytotype (Bi6mont et al., 1990b; Ronsseray et al., 1991); *P
