Segregation Distorter (SD) is an autosomal meiotic drive gene complex found worldwide in natural populations of Drosophila melanogaster. During spermatogenesis, SD induces dysfunction of SD+ spermatids so that SD/SD+ males sire almost exclusively SD-bearing progeny rather than the expected 1:1 Mendelian ratio. SD is thus evolutionarily “selfish,” enhancing its own transmission at the expense of its bearers. Here we review the molecular and evolutionary genetics of SD. Genetic analyses show that the SD is a multilocus gene complex involving two key loci—the driver, Segregation distorter (Sd), and the target of drive, Responder (Rsp)—and at least three upward modifiers of distortion. Molecular analyses show that Sd encodes a truncated duplication of the gene RanGAP, whereas Rsp is a large pericentromeric block of satellite DNA. The Sd–RanGAP protein is enzymatically wild type but mislocalized within cells and, for reasons that remain unclear, appears to disrupt the histone-to-protamine transition in drive-sensitive spermatids bearing many Rsp satellite repeats but not drive-insensitive spermatids bearing few or no Rsp satellite repeats. Evolutionary analyses show that the Sd–RanGAP duplication arose recently within the D. melanogaster lineage, exploiting the preexisting and considerably older Rsp satellite locus. Once established, the SD haplotype collected enhancers of distortion and suppressors of recombination. Further dissection of the molecular genetic and cellular basis of SD-mediated distortion seems likely to provide insights into several important areas currently understudied, including the genetic control of spermatogenesis, the maintenance and evolution of satellite DNAs, the possible roles of small interfering RNAs in the germline, and the molecular population genetics of the interaction of genetic linkage and natural selection.
Mendelian inheritance is a marvelous device for making evolution by natural selection an efficient process.... The Mendelian system works with maximum efficiency only if it is scrupulously fair to all genes. It is in constant danger, however, of being upset by genes that subvert the meiotic process to their own advantage. James F. Crow (1979)
SEGREGATION Distorter (SD) is a selfish, coadapted gene complex on chromosome 2 (an autosome) found at low frequency in nearly all natural populations of the fruit fly, Drosophila melanogaster. In heterozygous males carrying SD and a typical wild-type second chromosome (SD/SD+), most SD+-bearing spermatid nuclei fail to complete the histone-to-protamine transition during spermiogenesis, so that primarily SD-bearing spermatids develop properly and go on to fertilize eggs. SD/SD+ males thus sire almost exclusively SD-inheriting progeny. This distortion of classic Mendelian ratios has intrigued geneticists and evolutionary biologists for more than 50 years—and for good reason. As we describe below, SD is a newly evolved system that subverts one of the fundamental laws of inheritance by exploiting an ancient molecular pathway. In this review, we describe how SD was discovered and first characterized; the genetic components and features of the SD system; the molecular basis by which SD is thought to achieve its transmission advantage; theoretical and molecular population genetics analyses of the dynamics and evolutionary history of the SD system; and, finally, some of the still-unsolved questions concerning SD. The answers, we believe, have important implications for our understanding of spermiogenesis and the regulation of a ubiquitous, but still poorly understood, class of selfish sequences—genomic satellite DNAs.
Six Strange Chromosomes
In 1956, a first-year graduate student in James F. Crow’s laboratory, Yuichiro Hiraizumi, tackled a straightforward, if laborious, population genetics problem of the day: in natural populations of D. melanogaster, “recessive” lethal mutations were found to segregate at lower frequencies than expected, raising the possibility of nonnegligible heterozygous fitness effects (Hiraizumi and Crow 1957, 1960; Hiraizumi et al. 1960). Hiraizumi set out to estimate those heterozygous effects for second chromosomes extracted from a natural population in Madison, Wisconsin. For one collection, this involved testcrossing 183 +/cn bw males to cn bw females (where + is a wild-derived chromosome, cn bw is a standard tester chromosome, and cn and bw are recessive eye color markers that together produce white eyes when homozygous). For these crosses, healthy wild-derived second chromosomes will produce half wild-type progeny (+/cn bw) and half white-eyed progeny (cn bw/cn bw). Sickly wild-derived second chromosomes might, however, produce a few less wild-type progeny than otherwise expected. In a footnote to the article, Hiraizumi and Crow (1960) made reference to an altogether unexpected class of second chromosomes:
Among the chromosomes examined, six (five lethal and one control) behaved in a strange way, giving 0.80–0.95 wildtype flies. This turned out to depend on distorted segregation ratios in heterozygous males… .
These six strange chromosomes were the first SD chromosomes, the starting materials for analyses to be undertaken together by Hiraizumi and a new postdoc, Larry Sandler. Just before arriving in Crow’s lab, Sandler had published a seminal paper with his Ph.D. advisor, Ed Novitski, “Meiotic drive as an evolutionary force” (Sandler and Novitski 1957). Biased transmission, they argued, may be a pervasive, underappreciated cause of allele frequency change in natural populations over evolutionary time scales. (Although Sandler originally defined meiotic drive in the strict sense to refer to biased transmission resulting from disturbance of meiosis per se (Sandler and Novitski 1957), he would later use meiotic drive in the broad sense to refer to biased transmission resulting from disturbances at any stage of gametogenesis (Zimmering et al. 1970).) With Hiraizumi, Sandler had a chance to tackle the basis of a naturally occurring meiotic drive system in a well-studied genetic model species. In the few years that followed, the two would publish eight articles on SD together (reviewed in Crow 1991; Ganetzky 1999).
Initial Characterization of SD
In the first article on SD, Sandler et al. (1959) got most essential facts about the system right. They showed that the original six SD chromosomes fell into two classes. The five “lethal” SD-5-type chromosomes were allelic for the same recessive lethal mutation and carried two nonoverlapping paracentric inversions on 2R, one proximal and one distal; the one “control” SD-72 chromosome was lethal free and had only the distal inversion in common with SD-5 (Figure 1). All six SD chromosomes caused very strong distortion, with k ≥ 0.95 (where k is the proportion of SD-bearing progeny sired by heterozygous SD/SD+ males), when SD-bearing males were heterozygous against cn bw, or heterozygous against other marked or wild-derived second chromosomes. Distortion was male specific, as transmission through heterozygous females was normal. As recombinant SD chromosomes lacking the inversions still caused distortion, SD, they concluded, must be a form of genic meiotic drive rather than chromosomal drive. Finally, preliminary mapping placed the SD locus in or near the centromeric heterochromatin of chromosome 2.
Two findings from this early work led to reasonable but ultimately incorrect conclusions. First, Sandler et al. (1959) found that SD chromosomes failed to distort when heterozygous against two different pericentric inversion-bearing chromosomes, In(2LR)Cy and In(2LR)Pm2. From this, they inferred that chromosomal synapsis—at the centromeric SD region in particular—was required for distortion. [It would later be shown, however, that the In(2LR)Cy and In(2LR)Pm2 chromosomes are simply insensitive to distortion (Hartl 1975a).] Second, SD is not a single locus. While Sandler and Hiraizumi (1960) and Hiraizumi and Nakazima (1967) would later infer that the SD locus involved two separable loci, their results were complex. Part of the problem was that Sandler and Hiraizumi (1960) had been unlucky in choosing SD-72 (rather than SD-5) for genetic mapping, unaware that it carried a pericentric inversion that prevented recombination in the critical region (Lewis 1962).
Together, three articles ushered in the modern understanding of the genetics of the SD complex. First, Sandler and Carpenter (1972) settled a key question concerning the action of SD: Was the presence of SD essential for spermatid development in SD males? Or did SD somehow alter SD+ homologs in a way that caused SD+ spermatid lethality? Using clever genetic trickery, they showed that spermatids bearing SD+ chromosomes are gamete lethal. This finding led them to infer that SD is a trans-acting factor that acts at a cis-acting “sensitive-receptor” site on the SD+ chromosomes. SD chromosomes must therefore possess an insensitive-receptor. Taking the pericentric inversion of SD-72 into account, the old mapping data of Sandler and Hiraizumi (1960) now made sense. The SD system can indeed be decomposed into two loci: SD, the distorter, resides on chromosome arm 2L between 37B2 and 39–40 and the receptor, the target of distortion, resides between 39–40 and 42A but is not the centromere itself.
Second, Dan Hartl (1974), who began work on SD as a student in Crow’s lab, used recombination to cleanly partition the SD region of a SD-5-type chromosome, which lacks the pericentric inversion, into two major loci—a leftmost distorter locus, which he termed Segregation distorter (Sd), and a rightmost sensitivity locus, which he named Responder (Rsp). Thus SD refers to chromosomes that have the genotype, Sd Rspi, whereas SD+ chromosomes have the genotype Sd+ Rsps, where Rspi and Rsps are insensitive and sensitive to segregation distortion, respectively. Heterozygous SD/SD+ males thus transmit Sd Rspi-bearing sperm to >95% of progeny as the action of Sd disrupts development of Sd+ Rsps-bearing sperm. As a final proof of principle, Hartl (1974) constructed suicidal Sd Rsps recombinant chromosomes: in heterozygous Sd Rsps/Sd+ Rspi males, Sd Rsps are unable to distort against their insensitive homologs and instead distort against themselves.
Third, Barry Ganetzky (1977), who began work on SD as a student in Sandler’s lab, used X rays to mutagenize SD-5 chromosomes. For the first time, the chromosomal locations of the SD loci would begin to be known with precision. The X-ray-induced lesions would also reveal important properties of the SD loci (discussed below). Ganetzky’s experiments showed that Sd falls between cytological regions 37D2–7 and 38A6–B2 on chromosome arm 2L, and Rsp resides in the centric heterochromatin of chromosome arm 2R. These findings are consistent with those of Sandler and Carpenter (1972) but much refined. But then came a surprise. Ganetzky (1977) discovered a third major locus on SD chromosomes, one only hinted at in previous analyses (Hartl 1975a): a strong upward modifier of distortion, which Ganetzky named Enhancer of Segregation Distorter, E(SD), exists in the centric heterochromatin of chromosome arm 2L of SD chromosomes. After more than 15 years, the genetic structure of the SD complex finally emerged (Figure 1).
Ganetzky’s (1977) experiments not only mapped Sd but provided information on its properties. In particular, he distinguished two possibilities: if Sd represents a loss of wild-type Sd+ function, then mutagenesis ought to be able to induce wild-type chromosomes to become distorters; conversely, if Sd is a gain-of-function mutation, then mutagenesis ought to be able to revert SD chromosomes to nondistorters. While Ganetzky was unable to induce distorter mutations on wild-type chromosomes, he recovered four SD-5 chromosome revertants—SD chromosomes that lost the ability to distort due to deletion or inactivaton of the Sd locus. Sd behaves as a neomorph.
The discovery that Sd is neomorphic, and hence dispensable for wild-type function, likely helped perpetuate the sense that there was something foreign about it. Historical context was also a factor. SD was discovered at a time when transposable elements were just coming into focus, in particular with McClintock’s maize work, which had a profound and lasting effect on Sandler’s (and others’) thinking about SD:
The phenomenon of segregation–distortion is, directly, a case in which a genetic entity located on some chromosome influences the behavior of an element or elements located on a different chromosome. The system is thus analogous to “Ac-Ds-like” systems in maize (McClintock 1951, 1956)… . The obvious analogy is that SD is an Ac-like element and the point or points on the SD+ chromosome which are subject to (the action of SD) are DS-like elements (Sandler et al. 1959).
While this complexity may simply mean that segregation–distortion is polygenic, it may also very well be that the elements involved are not conventional, and therefore, that standard genetic interpretations, such as those we have employed here may be misleading (Sandler and Carpenter 1972).
To call these “mutations” is the conservative assumption: there is no specific evidence against the possibility that they are minute chromosomal rearrangements or deletions or insertions of a virus-like or foreign element…. (Hartl 1974).
The answer may lie in the hints that the elements of distortion are extrinsic and perhaps transposable (Sandler and Golic 1985).
In 1983, Brittnacher and Ganetzky would do another X-ray mutagenesis screen on SD chromosomes other than SD-5, including an SD chromosome collected in Italy (Brittnacher and Ganetzky 1983). Two advances would come from this work. First, the experiments showed that other SD chromosomes, even ones from geographically disparate populations, had the same Sd locus with the same properties in the same chromosomal position. Second, the experiments precisely narrowed the location of the Sd locus to a small interval 37D2–6 (Figure 2A; Brittnacher and Ganetzky 1983), which made feasible the molecular analyses that would finally dispel any notion that Sd is a foreign element.
The precise localization of Sd to such a small chromosomal interval allowed for the molecular dissection of the region and the discovery that SD flies from around the world all share a ∼5-kb tandem duplication in the Sd region (Powers and Ganetzky 1991). In transformation experiments, a ∼12-kb genomic DNA fragment that includes the duplication can cause full-strength segregation distortion in the appropriate genetic background (McLean et al. 1994). Sd thus involves extra endogenous DNA.
Two nested genes are duplicated within the 12-kb fragment: one encodes heparan sulfate 2-O-sulfotransferase (Hs2st) and the other Ran GTPase activating protein (RanGAP; Figure 2B). Sd and Sd+ are structurally equivalent in the proximal half of the duplication, but the duplication junction cleaves the 3′-end of the distal copy of RanGAP, introducing a nonsense mutation that results in a truncated RanGAP protein that lacks at most 234 amino acids from its COOH-terminus (Figure 2C; Merrill et al. 1999). Expression of this truncated RanGAP alone is sufficient to reproduce segregation distortion in the appropriate genetic backgrounds (Merrill et al. 1999). The duplicate gene encoding the truncated distorting protein is hereafter called Sd–RanGAP.
The hope of course is that the discovery of Sd–RanGAP might, after 40 years, finally reveal the cellular basis of segregation distortion. Wild-type RanGAP functions in many cellular venues but is known best for its role in nuclear transport. Below we consider how production of a truncated Sd-RanGAP protein might affect nuclear transport and other Ran-mediated systems, specifically during spermatogenesis, to induce dysfunction of sensitive Rsp-bearing spermatids. But first, we turn to the question of the genetic and molecular nature of the Rsp locus itself.
Being buried in the heterochromatin of 2R, Rsp has proved to be a challenge to study. Rsp changed names several times during its first 14 years. Sandler and Hiraizumi (1960) first designated Rsp as Activator of SD [Ac(SD)] and hypothesized that Ac(SD) was required in cis to “activate” SD function to cause distortion (here Ac(SD) is the genotypic equivalent of Rspi). They proposed that SD induced breakage on its homolog during synapsis and speculated that Ac(SD) and SD resided within a chromosome aberration, an insertion or duplication, that was later shown to be a pericentric inversion (Lewis 1962). Hiraizumi and Nakazima (1967) later posited that Ac(SD) was the locus conferring insensitivity to the SD chromosome. The experiments of Sandler and Carpenter (1972) revealed Ac(SD) as the direct target of distortion, which they renamed receptor: SD chromosomes have an insensitive-receptor allele and SD+ have a sensitive-receptor allele. Finally, Hartl (1973) renamed the locus to Responder to fit his hypothesis that Sd encodes a regulator protein that binds Rsp.
Rsp is located in the proximal 2R heterochromatin (Brittnacher and Ganetzky 1989) but is not the centromere (Lyttle 1989). Sensitive Rsp alleles, like Sd, behave as a neomorphs: deletions of 2R heterochromatin containing Rsp convert sensitive chromosomes into insensitive chromosomes (Ganetzky 1977). Sensitive Rsp alleles act in cis to cause spermatid dysfunction: moving Rsp to a different chromosome makes that chromosome sensitive to SD (Brittnacher and Ganetzky 1989; Lyttle 1989). To explain its frequency in populations, it is usually assumed that sensitive Rsp alleles have some function in spermatogenesis, although that function must be nonessential as homozygous deletions of Rsp are viable and fertile (Ganetzky 1977). As we discuss below, the Rsp locus does not correspond to a conventional protein-coding gene but instead corresponds to a class of DNA sequences whose function we know little about: satellite DNA.
Responder corresponds to a block of satellite DNA repeats
The first clue that Rsp might have atypical genetic properties came from the discovery that most natural populations harbor a continuous range of sensitivities to segregation distortion (Sandler and Carpenter 1972; Martin and Hiraizumi 1979; Hiraizumi et al. 1980; Hiraizumi and Thomas 1984; Temin and Marthas 1984; Lyttle 1986). Three other kinds of data, appearing nearly simultaneously, then converged on the finding that Rsp comprises a repetitive satellite DNA locus in the proximal heterochromatin of 2R.
In cytological preparations of mitotic chromosomes from Rsps and Rspi strains, Pimpinelli and Dimitri (1989) discovered a large size difference in the heterochromatic band h39 on chromosome 2 (Figure 3A). The size of the h39 band correlated with sensitivity to SD: the h39 band is largest on super-sensitive Rspss chromosomes, intermediate sized on sensitive Rsps chromosomes, and small or absent on Rspi chromosomes (Pimpinelli and Dimitri 1989; Figure 3B).
Guided by the assumption that Rsp was repetitive (and by Pimpinelli and Dimitri’s cytological evidence; see above), Wu and colleagues screened a λ-phage library for DNA fragments that hybridized to Rsps flies better than Rspi flies (Wu et al. 1988). One of the clones, called H0, contained a characteristic XbaI restriction site. Digestion with XbaI revealed a 240-bp repeat. Sequencing this clone indicated that the XbaI repeat was a dimer of two related 120-bp A+T-rich repeats, corresponding to a satellite repeat (Figure 3C). The sensitivity to segregation distortion correlates with the number of Rsp repeats: Rspss alleles have ∼2500 copies, Rsps alleles have ∼700 copies, Rspi alleles associated with SD+ chromosomes have 100–200 copies, and Rspi alleles associated with SD chromosomes have <20 copies of the dimeric repeat (Wu et al. 1988; Lyttle 1991).
By moving blocks of 2R heterochromatin to the Y chromosome, Lyttle was able to shift the sensitivity to SD to the Y chromosome (Lyttle 1989). Importantly, however, in one translocation, both chromosome 2 and the Y chromosome had some degree of sensitivity indicating that the breakpoint occurred within the Rsp locus and, therefore, that the sensitivity of Rsp was itself physically divisible (Lyttle 1989), consistent with Rsp being identified as a large block of satellite DNA at heterochromatin band h39 (Wu et al. 1988; Pimpinelli and Dimitri 1989).
Lyttle’s translocations were particularly informative as he was able to construct males that produced sperm with zero, one, or two copies of a Rsps allele by using males carrying two different Rsps alleles: one on the Y chromosome (the Dp(2;Y)Rsps allele) and one on chromosome 2 (cn bw), in an SD/SD+ background (Lyttle 1989). Rather than measuring the frequency of SD sperm as k, Lyttle measured the proportion of Rsps-bearing spermatids escaping the action of SD as a measure of Rsps survival probability. While the survival probability correlated with the number of Rsps alleles carried by a spermatid, it appeared that the survival of a sperm with a single Rsps allele is increased if another Rsps allele is present in the genome. These findings imply that Sd corresponds to some limiting gene product that gets diluted in the presence of multiple sensitive targets (but see Hiraizumi 1990). However, sperm bearing two copies of Rsps made up a smaller fraction of sperm than expected if each Rsps affects spermatid dysfunction independently, suggesting an epistatic effect of multiple Rsps copies in the same spermatid (Lyttle 1989).
The molecular structure of the Rsp locus
Rsp repeats are organized as dimers with “left” and “right” 120-bp A+T-rich repeats that have a XbaI restriction site (TCTAGA) between each dimer (Figure 3C; Wu et al. 1988). Canonical Rsp repeats (i.e., those identified by the H0 clone of Wu et al. 1988) occur as head-to-tail tandem repeats dispersed throughout region h39; but these comprise only ∼15% of the total size of the Rsps locus (Cabot et al. 1993; Houtchens and Lyttle 2003). Rather than being interspersed with transposable elements and other kinds of satellite sequences, the canonical Rsp repeats are interspersed with other, more divergent variants of Rsp repeats (Houtchens and Lyttle 2003; Figure 3C). The left and right Rsp repeats each form monophyletic clades with ∼4% divergence within each repeat type and ∼16% between the two repeat types (Cabot et al. 1993; Houtchens and Lyttle 2003). The noncanonical repeat sequences are considerably more variable than the canonical ones, with sequence divergence among noncanonical variants ranging from 8 to 32% (Houtchens and Lyttle 2003).
Recombination at Rsp and variability in sensitivity to distortion
The large variability in Rsp repeat number, and hence sensitivity, between individuals could be generated by unequal exchange events (Wu and Hammer 1991). Cabot et al. (1993) found evidence for recombination events between left and right copies of Rsp and all of the exchanges were restricted to a 29-bp region, where left and right copies are the most similar. Cabot et al. (1993) compared Rsp repeats cloned and sequenced from three chromosomes: lt pk cn bw (Rspss, ∼2500 copies), cn bw (Rsps, ∼700 copies), and Canton-S (Rspsemi-sensitive, ∼300 copies). To their surprise, Rsp repeats within a chromosome differed more than Rsp repeats between chromosomes (Cabot et al. 1993). There are two possible explanations for this result: recombination between homologs at the Rsp locus may have homogenized Rsp repeats between chromosomes; alternatively, recent selection near the centromere may have dragged a Rsp allele to high frequency. The latter explanation would lead to the homogenization of Rsp repeats between chromosomes because different Rsp alleles would share a recent common ancestor.
Enhancer of Segregation Distorter
Genetic deletion of Sd typically eliminates distortion altogether, whereas deletion of the enhancer, E(SD), reduces the strength of distortion [from k = 0.99 to ∼0.70; (Ganetzky 1977; Brittnacher and Ganetzky 1984; Sharp et al. 1985)]. These observations imply that Sd is the major distorting locus, while E(SD) functions as a secondary modifier. But as it turns out, E(SD) has its own capacity for distortion, even in the absence of Sd. Sharp et al. (1985) discovered that recombinant SD-5 chromosomes lacking Sd—i.e., Sd+ E(SD) Rspi —cause modest distortion against sensitive Rsps alleles (k = 0.6) and stronger distortion against supersensitive Rspss alleles (k = 0.85). Temin (1991) further showed that E(SD) distorts against sensitive Rsp chromosomes in a dose-dependent manner. Furthermore, Sd and E(SD) are both suppressible by the same genetic modifiers (see below), suggesting that they distort via the same mechanism (Temin 1991).
The independent distorting ability of E(SD) raises an intriguing possibility: Was E(SD) the first distorter in the system to evolve on a Rspi chromosome, followed by the later acquisition of Sd (Lyttle 1993)? To address this question directly, the molecular identity of E(SD) must be determined. We can, however, at least exclude the possibility that E(SD) and Sd are both recent duplicates of the RanGAP gene. For one, their genetic properties differ: E(SD) has a dose-dependent effect on the strength of distortion; Sd does not. More directly, Southern blot analysis reveals no evidence for sequence homology between Sd–RanGAP and E(SD) (Powers and Ganetzky 1991).
Other Modifiers of the SD System
For a segregation distorter to invade a population it must be genetically linked to its target, with insensitive alleles occurring in cis and sensitive target alleles in trans (Prout et al. 1973; Charlesworth and Hartl 1978). Free recombination between distorter and target would regularly generate suicide combinations. Nonrecombining sex chromosomes and the low-recombination centromeric regions are thus particularly susceptible to the invasion of multilocus drive systems. Once a drive system is established, however, it may recruit further modifiers of recombination (see below).
Genetic enhancers and suppressors of segregation distortion have different linkage requirements: enhancers will evolve in cis with the distorter, whereas suppressors can evolve anywhere in the genome in trans with the distorter (Thomson and Feldman 1974; Hartl 1975b; Crow 1991). The regions of the genome under strongest pressure to evolve suppressors are those directly targeted by the distorter (e.g., the target locus should evolve from a sensitive to an insensitive state). But when a distorter like SD is at equilibrium frequency, even unlinked suppressors are expected to increase in frequency (Hartl 1975b; Crow 1991). To explain why, Crow (1991) offered the “grandchild” argument. Imagine two scenarios, one in which SD is challenged by a suppressor, reducing k, and one in which SD is not challenged by a suppressor, leaving k close to 1. The number of progeny sired by SD/SD+ males is the same with or without the suppressor—only the ratio of SD to SD+ alleles changes. As SD chromosomes (being usually homozygous sterile or lethal) come with a fitness cost, any modifier that reduces k will increase the fraction of fitter SD+ alleles in the population and thus leave more grandchildren (Crow 1991). Unlinked modifiers that suppress distortion will therefore increase in frequency in the population because they contribute more progeny to future generations.
As we discuss next, the SD complex and its system of recombination modifiers, enhancers, and suppressors provide evidence consistent with the rich population genetic theory of drive systems. We argue later that the molecular identities and functions of these enhancers and suppressors may prove essential to solving the problem of the mechanistic and molecular basis of SD.
Chromosomal inversions on SD
The three major SD loci are clustered about the centromeric region of chromosome 2, where the rate of crossing over per physical unit is strongly reduced (Figure 1). In addition to exploiting this centromeric effect on crossing over, most SD chromosomes have also recruited genetic modifiers of recombination—chromosomal inversions that further tighten the linkage among its interactors (Sandler and Hiraizumi 1960; Hartl and Hiraizumi 1976; Crow 1988; Lyttle 1991; Lyttle 1993). SD-72 chromosomes (and its derivatives), for example, carry a pericentric inversion that captures E(SD) and Rspi and facilitates their linkage disequilibrium with Sd, thereby preventing wasteful production of nondriving or worse, suicidal, recombinants.
Most SD chromosomes also carry one or more inversions on chromosome arm 2R (Figure 1 and Table 1). SD-72 and SD-5 carry In(2R)NS (52A;56F)—a common cosmopolitan inversion—and SD-5 carries an additional inversion, In(2R)45C-F;49A (Lewis 1962). In Africa, the locally common SD chromosome, SD-Mal, carries the two overlapping inversions, In(2R)44F3-12;54E3-10 and In(2R)51B6-11;55E3-12 (Presgraves et al. 2009). On SD-5 and SD-Mal chromosomes, paracentric inversions prevent crossing over along ∼40% and ∼50% of the length of arm 2R, respectively. None of these paracentric inversions overlap the three major SD loci, and none are required for distortion. Indeed, recombinant SD-5 like chromosomes fail to provide any evidence that its two paracentric inversions (or the alleles therein) contribute to distortion intensity (Sandler et al. 1959; Hiraizumi and Nakazima 1967).
What, then, is their function? In addition to the three major loci, SD chromosomes have collected numerous, mostly uncharacterized enhancers along 2R (discussed below). The commonly accepted hypothesis is that the 2R inversions help to hold SD and its multiple enhancers together as a single distorting haplotype. There is, however, considerable physical distance between the major SD loci clustered about the centromere and the paracentric inversions on 2R (Figure 1): the inversions can be readily separated from the three major SD loci. Their strong association with SD therefore suggests that strong epistatic selection—i.e., selection favoring a particular multilocus genotype and disfavoring recombinants—also reinforces the long-distance linkage disequilibrium that spans from Sd on 2L, across the centromeric heterochromatin, and (in many cases) across most of 2R.
Stabilizer of Segregation Distorter
Ordinarily, there is little variation in segregation distortion strength from male to male among SD/SD+ heterozygotes. In analyses of SD-5 and SD-72 recombinants, Sandler and Hiraizumi (1960) discovered a major modifier of SD, the loss of which reduces the mean strength of distortion, from k = 1.0 to 0.71 and, interestingly, increases the variance in k among males of the same genotype. As the loss of this modifier makes segregation distortion ratios “unstable,” it was dubbed, Stabilizer of Segregation Distorter [St(SD)]. Sandler and Hiraizumi (1960) showed that St(SD) must reside beyond the distal-most breakpoint of the inversion (56F), at the tip of 2R. Without the inversions on 2R, St(SD) would be effectively unlinked from the major SD loci near the centromere.
Modifier of Segregation Distorter
Hiraizumi et al. (1980) revealed evidence of an additional modifier of segregation distortion in a genetic mapping analysis of a recombinant SD-36 chromosome (an SD-5-like chromosome; Figure 1) for which the two paracentric inversions on 2R and St(SD) were removed (Hiraizumi and Nakazima 1967). A locus outside of the SD complex, somewhere in the euchromatic region proximal to 43E on arm 2R, has a major effect on the strength of segregation distortion, boosting it from k = 0.66 to 0.92 (Hiraizumi et al. 1980). They named the locus Modifier of Segregation Distortion [M(SD)].
Unlinked modifiers: Suppressors of Segregation Distorter
Unlinked suppressors of SD [Su(SD)] can segregate at high frequencies in natural populations on both the X (Kataoka 1967; Hiraizumi and Thomas 1984) and third chromosomes (Trippa and Loverre 1975). Su(SD) loci also occur in laboratory strains such as the X chromosome balancers, Muller-5 and FM6 (Sandler and Rosenfeld 1962; Nicoletti and Trippa 1967) and the third chromosome balancer, TM6 (Lyttle 1986). Curiously, the action of the Su(SD)3 (on chromosome 3) appears specific to particular SD chromosomes: it enhances the distortion of some SD chromosomes but suppresses distortion of others (Trippa and Loverre 1975). It is unclear, if these suppressors represent adaptations to the presence of SD in natural populations or incidental genetic variation that happens to affect the distortion phenotype. The frequency of Su(SD)X, for instance, does not correlate with SD frequency (e.g., no SD chromosomes were identified in a Texas population, but Su(SD)X was found at a frequency of ∼75%; Hiraizumi and Thomas 1984). Once the Su(SD) loci are identified, molecular screens will allow more comprehensive (and less labor intensive) surveys of suppressor allele frequencies.
Mechanism of Segregation Distortion
While the identities of the distorter, Sd, and the target, Rsp, have been established for the SD system, the mechanism of distortion has remained elusive. In the following sections, we first review the pertinent aspects of normal spermatogenesis, discuss the segregation distortion phenotype, and then present current models of SD function.
Normally during Drosophila spermatogenesis, each diploid spermatogonial cell undergoes four mitotic divisions and then grows substantially before entering meiosis as a primary spermatocyte (Figure 4A; reviewed in Fuller 1993). Each of the 16 primary spermatocytes undergoes two meiotic divisions to produce 64 round spermatids that then differentiate into mature sperm. Because each cell division has incomplete cytokinesis, cytoplasmic bridges connect the developing germ cells. During spermiogenesis, the nucleus must undergo a radical remodeling as it shrinks ∼200-fold from round spermatid to the needle-shape of mature sperm (reviewed in Fuller 1993). To accomplish this hypercompaction of chromatin, preelongation spermatids replace canonical histones with sperm-specific protamines, a process known as the histone-to-protamine transition (Rathke et al. 2007; Miller et al. 2010). The spermatids then elongate and individualize and move to the seminal vesicle as mature sperm (see Fuller 1993).
Sandler and colleagues originally proposed that SD causes distortion by rendering SD+-bearing spermatids dysfunctional or, in their specific hypothesis, dead (Sandler et al. 1959). While effectively correct, the particulars of the model were later refuted, as early cytological studies would reveal that meiosis in SD/SD+ testes was normal, at least at the resolution of light microscopy (Peacock and Erickson 1965; Hartl et al. 1967; Nicoletti et al. 1967), ruling out hypotheses of SD-induced chromosome breakage during synapsis (Sandler et al. 1959; Crow et al. 1962). Peacock and Erickson (1965) next proposed the “functional pole” hypothesis, building on Novitski and Sandler’s suggestion that only half of male meiotic products become functional (analogous to the polar bodies of female meiosis; Novitski and Sandler 1957). Finding no gross abnormalities in SD/SD+ testes, Peacock and Erickson reasoned that SD must affect the orientation of chromosomes at metaphase I so that SD chromosomes orient to the functional pole to the exclusion of SD+ chromosomes (Peacock and Erickson 1965). This model predicts that SD males suffer no sperm dysfunction and hence no reduced fertility. But two subsequent studies would show unequivocally that SD males sire fewer progeny than wild-type males, refuting the functional pole hypothesis and suggesting instead some form of sperm dysfunction (Hartl et al. 1967; Nicoletti et al. 1967). Later, Sandler and Carpenter (1972) were able to generate males that produced individual spermatids bearing both SD and SD+ chromosomes; their experiments showed that SD will destroy itself when in the same spermatid as SD+, confirming the sperm dysfunction hypothesis.
The nature of the sperm dysfunction was finally revealed by ultrastructure analyses of SD/SD and SD/SD+ testes. The first ultrastructural studies showed that roughly half of the developing spermatids in the cysts of SD/SD+ heterozygotes had abnormal sperm tail formation (Nicoletti 1968). Later studies showed that the first visible difference between SD and SD+ spermatids occurs in early postelongation cysts as a difference in chromatin condensation (Tokuyasu et al. 1977; Figure 5, A–C) and that, usually, these spermatids fail to individualize or coil (Peacock et al. 1972; Tokuyasu et al. 1977). The large, uncondensed SD+ spermatid nuclei that fail to individualize end up in the waste bag (Peacock et al. 1972; Tokuyasu et al. 1977), and those few SD+ spermatids that do individualize break down later in the testicular duct, the seminal vesicle (Peacock et al. 1972), or the female reproductive tract (Hartl 1969; Childress and Hartl 1972). SD is a sperm killer (and thus causes meiotic drive in the broad sense; Zimmering et al. 1970).
Motivated by the observation that Sd genetically mapped near the histone locus and might therefore correspond to the histones themselves, Kettaneh and Hartl (1976) examined the histone-to-protamine transition in SD homozygotes and inferred that the transition does not take place in these testes. SD is temperature sensitive—both heat and cold treatment reduce the intensity of distortion in most cases (Mange 1968; Denell et al. 1969; Hihara 1971; Matthews and Mortin 1983)—allowing temperature-shift experiments to pinpoint the timing of SD action during spermatogenesis. As the developmental time course of spermatogenesis in D. melanogaster is known (Khishin 1955), the timing of SD action can be inferred by shifting SD/SD+ males to different temperatures at various development stages and measuring its effect on the strength of distortion. The first temperature-shift assays suggested that the critical action of SD occurs during early meiosis I (Mange 1968), but the response to temperature varies among different SD chromosomes (Mange 1968; Hihara 1971, 1974) and inconsistencies exist in the inferred timing of SD action, with some inferring a later, postmeiotic critical stage (e.g., SD-72 acts postmeiotically; Matthews and Mortin 1983). Temperature shift experiments on X-linked suppressors of SD found the critical stage to be postmeiotic, in early spermatids (Hihara 1971; Hiraizumi 1993). A postmeiotic critical stage, if correct, would correspond neatly with the histone-to-protamine transition.
Models of Segregation Distortion
The SD system has been a merciless destroyer of otherwise elegant hypotheses, as the pages of this review recount, and it would be too much to expect that the present hypothesis will fare much better. (Hartl and Hiraizumi 1976)
With the recent identification of the Sd gene product as a mutated version of one of the key players in nuclear transport, the end game may at last be in sight… . But it would be a mistake to underestimate this wily prey. (Ganetzky 1999).
We have already seen the chromosome breakage and functional pole hypotheses for SD falsified. In the following sections, we present a brief summary of other hypotheses on the mechanism of segregation distortion, including our own speculations.
Direct interaction between Sd and Rsp
On the basis of a complementation analysis of 11 SD chromosomes, Hartl (1973) proposed that Sd encodes a multimeric protein (a “regulator”) that, during normal spermatogenesis, complexes at the Rsp locus. In SD/SD+ males sperm dysfunction results from the inability of Sd hetero-multimers (Sd and Sd+ products) to complex at the Rsp locus on the SD+ chromosome (Hartl 1973). As spermatid dysfunction in SD/SD+ testes happens after meiosis, when only limited transcription occurs, Hartl (1973) proposed that the function of Rsp might be to silence transcription of a gene or genes after meiosis.
Ganetzky’s 1977 discovery that Sd and Rsp were neomorphic—deletions of Sd fail to distort and deletions of Rsp create insensitive Rsp alleles—refuted the idea that different Sd alleles correspond to deletions (Peacock and Miklos 1973). His discovery also required a modification of Hartl’s model. Ganetzky (1977) proposed that the Sd product and the Rsp locus do not interact during normal spermatogenesis (as both loci are dispensable) but rather sperm dysfunction is a specific consequence of their interaction. Both Ganetzky’s and Hartl’s models imply that sensitive and insensitive Rsp alleles compete for binding with a limited amount of Sd product. But Hartl’s model predicts that insensitive Rspi alleles bind with higher affinity to Sd, whereas Ganetzky’s model predicts that sensitive Rsps alleles bind with higher affinity to Sd. Ganetzky’s model was later modified by Hiraizumi et al. (1980), who suggest instead that the product of Sd prevents a direct, essential interaction between the product of M(SD) and Rsp.
Disrupted nuclear transport in SD males
The discovery that Sd is in fact Sd-RanGAP, which encodes a truncated RanGAP protein, naturally led to the proposal that segregation distortion disrupts nuclear transport, one of the well-known functions of wild-type RanGAP. Normally, the RanGAP protein is tethered to the cytoplasmic side of the nuclear pore complex (Figure 6, A and B), where it stimulates hydrolysis of Ran–GTP to Ran–GDP; its cofactor, Ran guanine exchange factor (RanGEF) is chromatin bound in the nuclear compartment where it converts Ran–GDP back to Ran–GTP by nucleotide exchange. The RanGAP–RanGEF system thus maintains a steep Ran–GTP concentration gradient across the nuclear envelope (high in nucleus, low in cytoplasm) that stimulates and provides directionality to nuclear transport (Gorlich and Kutay 1999).
Sd–RanGAP is enzymatically wild type but has an abnormal subcellular distribution (Figure 6, C and D). Unlike wild-type RanGAP, Sd–RanGAP is found inside the nucleus, at least for a subset of primary spermatocytes, presumably owing to the loss of one of its nuclear export sequences (NES; Figure 2C; Kusano et al. 2001). Sd–RanGAP is also diffusely distributed in the cytoplasm of some primary spermatocytes (Figure 6, C and D), perhaps because the sites responsible for tethering wild-type RanGAP to the cytoplasmic face of the nuclear pore by SUMOylation are missing in the truncated protein (Figure 2C; Kusano et al. 2001).
Kusano et al. (2001) hypothesize that the mislocalization of enzymatically active Sd–RanGAP to the nucleus affects nuclear transport, thereby preventing some factor required for chromatin condensation from entering the nucleus and ultimately causing the dysfunction of SD+ spermatids. Sd–RanGAP requires enzymatic activity and nuclear localization to cause segregation distortion: mutating sites required for enzymatic activity or appending a NES to Sd–RanGAP both disrupt segregation distortion (Kusano et al. 2001). Two lines of evidence suggest that Sd–RanGAP causes segregation distortion by perturbing Ran-mediated nuclear transport. First, if Sd–RanGAP localization in the nucleus causes excess Ran–GDP to accumulate in the nucleus, then overexpressing Ran and/or RanGEF should increase the concentration of Ran–GTP in the nucleus, thus restoring the gradient of Ran–GTP/Ran–GDP across the nuclear membrane. Second, Kusano et al. (2001) showed that, indeed, overexpression of either Ran or RanGEF in the germline suppresses segregation distortion in Rsps and Rspss backgrounds. Second, Kusano et al. (2001) used GFP–NLS–NES reporters of nuclear transport activity to show that, at least in salivary glands, nuclear export is disrupted in SD flies.
This functional work on Sd–RanGAP and its role in nuclear transport is groundbreaking. There are, nevertheless, several caveats concerning the disrupted nuclear transport model of distortion. The first concerns the protein in question: RanGAP features in a wide range of functions in the cell (see below), and it is difficult to know whether nuclear transport is the critical phenotype or correlated with the critical phenotype. Second, the results of the nuclear transport assay are not entirely consistent with segregation distortion phenotypes (e.g., the NES–GFP was still “trapped” in the nucleus in some SD revertants but not others; Kusano et al. 2001). The fact that Kusano et al. (2001) performed the nuclear transport assay in salivary glands presents another difficulty, as SD affects postmeiotic spermatid development. The apparent inconsistencies between segregation distorter phenotypes and the results of the transport assay might be attributable to differences in nuclear transport between salivary glands and testes.
Some experiments suggest that segregation distortion is due to the mislocalization of Sd–RanGAP rather than some functional difference between wild-type RanGAP and Sd–RanGAP (Kusano et al. 2002). Kusano et al. (2002) show that even wild-type RanGAP can cause segregation distortion: a 10- to 20-fold overexpression of wild-type RanGAP in the germline causes segregation distortion and shows aberrant localization similar to Sd–RanGAP. Interestingly, extra doses of E(SD), known to cause segregation distortion (Sharp et al. 1985; Temin 1991), do not affect Sd–RanGAP or RanGAP expression but instead lead to an accumulation of RanGAP in the nucleus of primary spermatocytes. Its cytoplasmic signal around the nuclear membrane, however, looks normal (Kusano et al. 2002) unlike Sd–RanGAP (Kusano et al. 2001). The mislocalized RanGAP appears to affect the Ran–GTP/Ran–GDP gradient: overexpression of Ran or RanGEF suppresses segregation distortion caused by both wild-type RanGAP and extra doses of E(SD) (Kusano et al. 2002).
Unanswered questions under the nuclear transport model
There are two unanswered questions under the nuclear transport model. First, why do only Rsps spermatids die? It is possible that nuclear Sd–RanGAP preferentially accumulates in Rsps nuclei, perhaps because Sd–RanGAP binds Rsp and so the nuclear transport defect occurs only in these spermatids (Pimpinelli and Dimitri 1989; Kusano et al. 2001; Kusano et al. 2002). Kusano et al. (2001, 2002) report on the distribution of Sd–RanGAP in premeiotic primary spermatocytes, but the subcellular localization of Sd–RanGAP in postmeiotic spermatids is unknown, leaving an important question unanswered: Is Sd–RanGAP mislocalized in only half of the developing spermatids—those with Rsps (Figure 7A)? Alternatively, Sd–RanGAP could be mislocalized in all spermatids, disrupting the Ran–GTP/Ran–GDP gradient, but Rsps-bearing spermatids are especially sensitive to this perturbation (Kusano et al. 2001, 2002). For example, Rsp may bind some limited chromatin condensation factor. Therefore, in spermatids with many Rsp repeats (Rsps and Rspss), the Rsp locus may act as a sink for the chromatin factor preventing the proper condensation of other chromatin (Figure 7B).
The second difficulty, and one of the most critical questions of all, for the nuclear transport model is this: Why is the only observable phenotype in SD/SD+ heterozygotes specific to the postmeiotic stages of spermatogenesis? The NLS–NES–GFP assays show disrupted nuclear transport in salivary gland cells, but how could all of the cells of the fly function normally with such a defect in an essential nuclear transport pathway? Spermiogenesis is unlike any other developmental process. Given the radical nature and rapid pace of spermatid nuclear remodeling, one can easily imagine that developing spermatids are especially sensitive to perturbations in the Ran cycle (Kusano et al. 2001, 2002, 2003).
What else does RanGAP do?
The cycling of Ran–GTP and Ran–GDP does more than facilitate nuclear transport. Ran is involved in mitotic spindle assembly, nuclear envelope assembly, chromosome segregation, and processes occurring during anaphase and cytokinesis (Joseph et al. 2002; reviewed in Quimby and Dasso 2003; Renshaw and Wilde 2011; Figure 8). RanGAP and RanGEF are critical in establishing the gradient of Ran–GDP and Ran–GTP for these other cellular functions and not only across the nuclear envelope (Figure 8A). During interphase, the Ran gradient functions after the breakdown of the nuclear envelope to spatially organize spindle assembly, directing microtubule growth toward chromatin (Figure 8B). A high level of Ran–GTP at chromatin also directs nuclear envelope reassembly after mitosis (reviewed in Renshaw and Wilde 2011; Figure 8C). If RanGAP is mislocalized and causes a reduction in nuclear Ran–GTP, it could plausibly affect any one of these other functions as well. In Schizosaccharomyces pombe, RanGAP has a role in the nucleus in addition to the cytoplasm: nuclear RanGAP interacts with histones (H3) and histone methyltransferases (e.g., Clr4, the yeast ortholog of D. melanogaster Su(var)3–9) to mediate heterochromatin assembly (Nishijima et al. 2006). Given that the establishment of a Ran–GTP/Ran–GDP gradient has many critical functions throughout the cell cycle, segregation distortion may be caused by something other than disruption of nuclear transport.
It is also possible that Sd–RanGAP interferes with some other cellular process specific to male germ cells. The gene ran-like is a testis-specific retroduplicate of Ran, encoding a testis-specific GTPase whose functions are currently unknown. ran-like originated at least 5 million years ago and has a history of rapid evolution in the melanogaster group, but appears to have functionally degenerated in some species—observations that led Tracy et al. (2010) to speculate that ran-like is a gene that spread to fixation by distortion. We suggest the possibility that, at least in D. melanogaster, ran-like may have acquired an important male germline-specific function—one that SD may disrupt to cause spermatid dysfunction. In particular, if the mislocalization of Sd–RanGAP inappropriately hydrolyzes ran-like-GTP in the testis to cause segregation distortion, it could explain why the effects of SD are specific to spermatogenesis.
Segregation distortion and small RNAs
A truncated Sd–RanGAP makes it easy to imagine how a disrupted Ran (or ran-like) gradient might affect spermatid development, but the connection to the repetitive satellite DNA locus, Rsp, remains unclear. Considering the still newly discovered roles of small RNAs and RNA interference (RNAi) in the germline may provide important insights. The well-studied Stellate (Ste)–Suppressor of Stellate [Su(Ste)] system of D. melanogaster, which some have argued is an ancient cryptic X–Y meiotic drive system (Hurst 1996), involves RNAi (Aravin et al. 2004). In this system, the X-linked products of the Ste locus are silenced by small antisense RNAs corresponding to Y-linked repeats of Su(Ste). Another study has implicated RNAi in the Winters sex-ratio distortion system of Drosophila simulans, for which most Y-bearing spermatids are dysfunctional owing to a postmeiotic chromatin condensation problem reminiscent of SD (Tao et al. 2007). Ordinarily the X-linked driver, called Distorter on the X (Dox), is suppressed by an autosomal gene, called Not much yang (Nmy). Nmy is a retrogene that originated from Dox and that now gives rise to a dsRNA transcript that Tao et al. (2007) hypothesize silences Dox via RNAi. Tao et al. (2007) speculate that the SD system might also involve RNAi.
Some repetitive sequences like satellite DNAs and transposable elements produce 23- to 30-bp repeat-associated small interfering RNAs or rasiRNAs (Aravin et al. 2003). Some rasiRNAs are implicated in silencing heterochromatin in plants, fungi, and Drosophila, where ribonucleoprotein (RNP) complexes associated with rasiRNAs recruit silencing factors to the chromatin (Pal-Bhadra et al. 2004; Verdel et al. 2004; Buhler and Moazed 2007; Kavi and Birchler 2009; Wang and Elgin 2011). The most famous rasiRNA pathway in the germline is the Piwi pathway including P-element-induced wimpy testis (Piwi), Aubergine (Aub), and Argonaute 3 (Ago3). While the role of rasiRNAs associated with the Piwi pathway (called piRNAs) in controlling transposable elements is well studied (Brennecke et al. 2007), the role of rasiRNAs in controlling satellite DNAs is less clear. Rsp generates rasiRNAs in Drosophila ovaries (Saito et al. 2006), and consistent with the involvement of RNAi in SD, we have recently determined that among rasiRNAs bound by Ago3 and Aub in the testis (Nishida et al. 2007; Nagao et al. 2010), Rsp rasiRNAs are enriched (A. Larracuente and D. Presgraves, unpublished results).
If SD/SD+ heterozygotes suffer disrupted nuclear transport during spermatogenesis, then Rsp rasiRNAs, after associating with RNP complexes in the cytoplasm, may not be efficiently imported back into the nucleus for targeting to the Rsp satellite (Figure 7C). Presumably the rasiRNAs direct chromatin remodeling complexes after meiosis in the testis, similar to RNAi-dependent heterochromatin formation in somatic tissues (Pal-Bhadra et al. 2004; Kavi and Birchler 2009). Because Rsps alleles have many (and Rspi alleles have few) repeats, Rsps-bearing spermatids might be disproportionately affected by the dearth of Rsp rasiRNAs in the nucleus and as a result fail to condense their pericentric heterochromatin (Ferree and Barbash 2007; Tao et al. 2007). While the Ran cycle is involved in nuclear export of miRNAs (Ohrt et al. 2006; Mtango et al. 2009), its role in rasiRNA transport is unknown. It is possible then that neither Ran nor ran-like are responsible for the postmeiotic import of rasiRNAs to their target satellites. One alternative that relies on the Ran (or perhaps ran-like) cycle, but not nuclear transport per se, involves the Ran-GTP/Ran-GDP gradient present during cell division. During mitosis, chromosomes “signal” their location in the cell with high concentrations of Ran-GTP (Renshaw and Wilde 2011; Kiyomitsu and Cheeseman 2012; Figure 8C). The initiation of postmeiotic chromatin condensation might similarly require high Ran-GTP (or ran-like-GTP) concentrations at chromosomes.
The simple fact is that we still lack a cogent model for how Sd and Rsp interact to cause distortion. The key problems are that: (1) RanGAP is likely pleiotropic, raising the possibility that nuclear transport is not the critical phenotype but rather correlated with the critical phenotype; (2) we are surprisingly ignorant about the genetic control of spermiogenesis in Drosophila; and (3) we are ignorant about the molecular basis of the maintenance of satellite DNAs in the Drosophila male germline.
Population Dynamics and Evolution of SD
The frequency of SD components in natural populations
SD is found at low frequencies (1–5%) in most natural populations of D. melanogaster surveyed throughout the world, including North America, South America, Europe, the Pacific islands, Australia, Asia, and Africa (Hiraizumi and Nakazima 1967; Hartl 1975a; Hiraizumi and Thomas 1984; Temin and Marthas 1984; Hao et al. 2000; Presgraves et al. 2009). The remarkably consistent frequency of SD among different geographic populations strongly suggests a stable deterministic equilibrium between drive and negative selection, one that is robust to variation in local climate and population demography. However, given the strength of SD’s distortion (≥95%), its low population frequency seems puzzling.
Several factors may hold SD back. First, SD chromosomes tend to be homozygous sterile and, often, lethal. But when SD is at low frequency these effects are easily swamped out by the strength of meiotic drive as homozygotes are so rarely formed (Temin and Marthas 1984). Indeed, the predicted frequency of a strong distorter that is homozygous sterile is ∼50%— much higher than that observed (Crow 1979). Second, as SD/SD+ males have reduced fertility (Hartl et al. 1967), it seems likely that they suffer reduced sperm competitive ability. Third, the appreciable frequency of Rspi alleles segregating in some natural populations may also keep SD in check (Kataoka 1967; Hartl 1970, 1977; Hihara 1974; Hartl and Hartung 1975; Trippa and Loverre 1975; Charlesworth and Hartl 1978; Hiraizumi and Thomas 1984; Temin and Marthas 1984). The frequency of Rspi alleles ranges between 3% (Kataoka 1967), ∼12–30% (Temin and Marthas 1984), 45% (Hartl and Hartung 1975), and even up to 86% (Trippa et al. 1980), depending on both the population and the screen used to identify these alleles. Simple population models predict that SD chromosomes can be stably maintained at low frequency (Charlesworth and Hartl 1978; Crow 1979). The models make plausible assumptions about the strength of distortion and the fitness of the Sd Rspi chromosome; they further assume that Sd+Rspi chromosomes suffer a small reduction in fertility compared to Sd+Rsps chromosomes to facilitate maintenance of sensitive alleles at Rsp. As these fitness measures are not known with any precision for this potentially broad range of genotypes (see Hauschteck-Jungen and Hartl 1978), the extent to which these assumptions hold in natural populations remains to be determined (Temin and Marthas 1984).
Early inferences about the history of SD
Studies on the evolutionary history of SD have led to conflicting notions about its age. Two of the earliest observations seemed to support the idea that SD must be old (e.g., Sandler and Hiraizumi 1960; Lyttle 1991; Lyttle 1993). First, the intrinsic complexity of the SD system—with its multiple enhancers, modifiers, stabilizers, and (usually) inversions—seemed to imply a history long enough for its piecemeal evolution. Indeed, wherever SD is found, it seems to have acquired local chromosome 2 inversions (Hartl and Hiraizumi 1976; Crow 1979; Presgraves et al. 2009). Second, the worldwide distribution of SD implies that the system has been with D. melanogaster throughout much of its dispersal history.
Other kinds of observations, however, suggested the possibility of a more recent origin. While most SD chromosomes are associated with inversions, both inversion-bearing and inversion-free—and thus the presumed ancestral type—SD chromosomes occur in Spain and Italy, suggesting that SD may have originated in the Mediterranean basin and subsequently spread throughout the world (Temin et al. 1990; Lyttle 1991; Wu and Hammer 1991). As D. melanogaster is a sub-Saharan species thought to have a single out-of-Africa expansion into Europe ∼15,000 years ago (Baudry et al. 2004; Li and Stephan 2006; Thornton and Andolfatto 2006; Hutter et al. 2007), the Mediterranean origins hypothesis would seem to place a ∼15,000-year upper limit on the age of SD.
Molecular evolution and population genetics of Sd–RanGAP
Sd–RanGAP appears to be specific to D. melanogaster (Powers and Ganetzky 1991), raising two possibilities. Did Sd–RanGAP originate after the split between D. melanogaster and its sister species of the D. simulans species complex, ∼3 MYA? Or did Sd–RanGAP originate before the species split only to be lost from other (non-melanogaster) lineages? With the molecular identification of Sd–RanGAP, the molecular evolutionary history of the gene responsible for distortion can be studied directly. The amount of nucleotide divergence between Sd–RanGAP and its parent gene, RanGAP, provides information on the timing of the origin of the duplicate and, hence, the age of the distorter. Within D. melanogaster, Sd–RanGAP and RanGAP divergence is more than an order of magnitude less than the divergence observed between the RanGAP orthologs in D. melanogaster and D. simulans. Sd–RanGAP thus arose within the D. melanogaster lineage well after its split from D. simulans.
The first molecular population genetic analyses of SD chromosomes revealed two haplotypes, one represented by the inversion-free SD chromosomes and the other by all other SD chromosomes (Wu and Hammer 1991). Population genetic analyses at four loci on chromosome 2, however, found little or no nucleotide differentiation between SD and SD+ chromosomes, consistent with a recent Mediterranean origin and rapid geographic expansion of SD (Palopoli and Wu 1996). Either the SD complex evolved so quickly that nucleotide differentiation has not had time to accumulate (Palopoli and Wu 1996) or recombination (gene conversion) between SD and SD+ chromosomes has prevented their differentiation. Recently, SD chromosomes were found in D. melanogaster populations distributed throughout Africa at frequencies typical of other worldwide populations, 1–5%, raising doubts about Mediterranean origins model and suggesting that SD may have originated earlier than ∼15,000 years ago (Presgraves et al. 2009).
Rapid turnover among SD chromosomes
The Charlesworth and Hartl (1978) theory shows that one kind of SD chromosome can displace another, while the overall population frequency of SD remains unchanged. Two different studies provide evidence for precisely this kind of rapid replacement. The first was observed directly in a longitudinal study in Madison, Wisconsin. In 1956, among the original six SD chromosomes sampled by Hiraizumi in Madison, five were lethal-bearing SD-5 chromosomes and one was a lethal-free SD-72 chromosome. In 1979, among 44 SD chromosomes sampled in the same Madison populations by Temin and Kreber (1981), one was a SD-5 chromosome and 19 were SD-72 chromosomes. Thus, in ∼25 years, SD-72, with its pericentric inversion acting to tighten the genetic linkage among Sd, E(SD), and Rspi, replaced SD-5 as the predominant SD chromosome.
The second replacement event was inferred from molecular population genetic data. In Africa, a new class of SD chromosomes, SD-Mal, causing perfect distortion (k = 1.0, for over 10,000 progeny scored) and bearing two overlapping paracentric African-endemic inversions on 2R, swept across the continent very recently (Presgraves et al. 2009). The expansion of SD-Mal chromosomes across east and west Africa was so recent and rapid that a region spanning Sd–RanGAP at 37E on 2L to region 55B on 2R comprises a single haplotype lacking any nucleotide variability (95% upper confidence limit on the age of the sweep ∼3400 years). An alternative class of SD chromosomes still segregrates in Africa, one lacking inversions and, surprisingly, lacking the ability to cause detectable distortion (Presgraves et al. 2009). Why these latter SD chromosomes fail to distort remains unclear, but they will, presumably, soon be lost from the population.
Molecular evolution of Responder
The DNA sequence divergence between the left and right halves of the Rsp repeats is greater than the average divergence between D. melanogaster and its sibling species D. simulans, strongly suggesting the dimeric structure of the Rsp repeats formed before the speciation event (Wu et al. 1988; Cabot et al. 1993). The Rsp repeats thus appear to have been in place when the Sd duplication arose. The dearth of Rsp repeats from the sibling species of D. melanogaster (at least using hybridization-based approaches; Moschetti et al. 1996; reference to C. I. Wu, unpublished results, in Lyttle 1991; and Cabot et al. 1993) is most likely due to either their rapid divergence, or loss, rather than Rsp being a novelty in the D. melanogaster genome. Understanding the evolutionary history of the Rsp repeats is, however, somewhat complicated by the discovery that these repeats also occur outside of the h39 locus, near the pericentromeric region of the third chromosome at cytological band 80C (Moschetti et al. 1996; Houtchens and Lyttle 2003) and elsewhere (A. Larracuente, unpublished results).
Is Responder a functional element maintained by natural selection?
The only function that can be attributed to Rspsens is that it causes a chromosome carrying it to be distorted by SD, a peculiar reason to exist. (Ganetzky 1977).
Why do Rsps alleles persist in natural populations? SD frequencies of 3–4% are sufficient to cause the fixation of Rspi alleles because of their insensitivity to segregation distortion (Charlesworth and Hartl 1978; Crow 1979). Large blocks of Rsp repeats might provide some benefit to flies that outweighs the negative effects of being distorted by SD. To test for such a fitness benefit, Wu et al. (1989) used large cage experiments to compete Sd+Rspi flies bearing a deletion for the Rsp locus (R16; Ganetzky 1977) against the standard Rsps strain (cn bw) in the presence of SD and, separately, in the absence of SD. Not surprisingly, in the presence of SD, the R16 chromosome outcompeted cn bw, due to its insensitivity to SD. In the absence of SD, however, cn bw outcompeted R16— the frequency of R16 decreased each generation. The fitness effect of the R16 deletion was on viability, not male fertility (Wu et al. 1989). If correct, these experiments would provide some of the first evidence for the functional significance of a satellite DNA.
However, while the fitness detriment of the R16 deletion could reflect the deletion of the Rsp locus, the details of the construction of the R16 chromosome raise doubts. R16 was constructed by irradiating cn bw chromosomes (Ganetzky 1977), which could have induced other, deleterious, mutations. Cytological inspection confirms that the deleted region of 2R heterochromatin is confined to band h39—the site of the Rsp locus (Pimpinelli and Dimitri 1989). While the R16 deletion is unlikely to remove unique DNA sequences corresponding to conventional genes (Wu et al. 1989), it does remove a cluster of Bari-I repeats. The Bari-I repeats encode transposon-like elements found in euchromatin and a single unique heterochromatic location at h39, adjacent to the Rsp repeats (Caizzi et al. 1993). The Bari-I repeats may evolve under evolutionary constraints, as the repeats at h39 have intact open reading frames and invariant array length (Caizzi et al. 1993). Therefore, this region of 2R heterochromatin is likely to have several functional elements that may account for the results of the cage experiments.
While the cage experiments may not satisfactorily demonstrate functionality of the Rsp repeats (Wu et al. 1989), it is still likely that Rsp has a function outside of its role in segregation distortion. Rsp is a type of satellite DNA— tandemly repeated sequences originally recognized as forming a “satellite” band on a CsCl gradient owing to their A+T- or G+C-rich nucleotide composition. Although once thought to be junk DNA, it seems that satellite DNA may have functional roles that are still not well understood (e.g., Karpen et al. 1996; Aravin et al. 2003; Pal-Bhadra et al. 2004; Brennecke et al. 2007). Similar to other satellite repeats, Rsp repeat DNA is bent, as revealed by its slowed mobility on polyacrylamide gels (Doshi et al. 1991). Additionally, nucleosome spacing at the Rsp locus is ∼240 bp compared to ∼190 bp of bulk heterochromatin (Doshi et al. 1991). The bending of satellite DNA (Radic et al. 1987) and nucleosome spacing differences may have a role in facilitating the compaction of heterochromatin (Doshi et al. 1991). Satellite sequences also generate rasiRNAs, whose role is not fully understood (Aravin et al. 2003; Brennecke et al. 2007); however, one likely role is in directing heterochromatin formation (Pal-Bhadra et al. 2004; Kavi and Birchler 2009).
Several early articles report that SD induces a heritable X-linked suppressor effect (“conditional distortion”; Sandler and Hiraizumi 1959; Sandler and Hiraizumi 1961; Sandler and Rosenfeld 1962); that certain insensitive second chromosomes can reduce the sensitivity of previously sensitive chromosomes (“translocal modification”; Sandler et al. 1959); and that there is a high “mutability ” revealed by St(SD) (Sandler and Hiraizumi 1959; Sandler and Hiraizumi 1960). These phenomena may be explained by the failure to control for the many then-unappreciated modifiers of SD, of both small and large effect, segregating in the genetic background (Miklos and Smith-White 1971; Hartl 1975a).
A few observations, however, lack satisfactory explanation under any of the proposed models. First, negative distortion has been reported, in which the SD chromosome is recovered at a frequency significantly less than 50% (Hiraizumi 1989, 1990, 1993; Hiraizumi et al. 1994). This negative distortion appears to be restricted to two genotypes: a weak recombinant SD chromosome (Sd E(SD) Rspi M(SD)+ St(SD)+) in the presence of an X-linked suppressor of segregation distortion (Su(SD)X; Hiraizumi 1989, 1990) or a strong SD chromosome (SD-72) in the presence of the Su(SD)X chromosome but at high temperature. Second, SD has an age effect (older males exhibit reduced segregation distortion), an observation that, like any other, requires explanation. What is especially intriguing, however, is that the age effect appears to be heritable: the sons of old SD/SD+ males also exhibit reduced distortion, an effect that persists for multiple generations (Sandler and Hiraizumi 1961; Hiraizumi and Watanabe 1969). Some RNAi-mediated phenomena show heritable age effects, including P-M hybrid dysgenesis. Specifically, the age-related recovery of fertility in P-M dysgenic hybrid females is mediated by novel P-element insertions into piRNA clusters. These novel, heritable piRNA cluster insertions give rise to piRNAs that silence P-elements, protecting against further dysgenesis (Khurana et al. 2011). It is unclear what parallels, if any, exist between piRNA-mediated silencing of transposons and the heritable age effect of SD. Third, among SD+ progeny escaping the action of SD, there is a female-biased sex ratio (Hiraizumi and Nakazima 1967; Denell et al. 1969). The success of a particular SD+ spermatid appears to depend on whether or not it cosegregates with an X or Y chromosome. This sex-ratio effect suggests either that sensitive Rsp-like sequences occur on some Y chromosomes or that SD has off-target effects on non-Rsp sequences on the Y (consistent with observations of Enns 1970). Any complete model of the molecular basis of SD will have to account for these phenomena.
The >50 years of continuous, intensive study of the genetics, molecular biology, and evolution of SD has made it one of our best known meiotic drive systems. The genetic properties of Sd, E(SD), Rsp, M(SD), and St(SD) have been established and the molecular identities of both the major driver, Sd–RanGAP, and the target, Rsp, have been determined. The early work on SD provided an important example at a time when the idea of selfish genetic systems was still unfamiliar. Its discovery and characterization helped establish the real and persistent susceptibility of Mendelian segregation to selfish genetic elements (Sandler and Novitski 1957) and, at the same time, the impetus for the evolution of genetic modifiers that restore fair segregation (Crow 1991). The work on SD further shows that, far from requiring some foreign genetic invader of the genome, even an ancient endogenous molecular pathway (i.e., Ran cycling) with multiple essential functions in the cell can be coopted by selfish gene systems. SD has also provided some of the most striking examples of the interaction between selection and recombination: the core distorting Sd E(SD) Rspi genotype was able to only invade natural populations of D. melanogaster because of its fortuitous clustering around the low-recombination centromeric region of chromosome 2, and its subsequent evolution involved the rapid recruitment of secondary modifiers of recombination (chromosomal inversions) that further tighten the genetic linkage among its interactors. The short-term evolutionary benefit of reduced crossing over on SD chromosomes, however, comes with a long-term evolutionary cost: most inversion-bearing SD chromosomes have accumulated recessive lethal mutations that they cannot now shed.
Despite these insights into the evolution of fair segregation, selfish genes and coadapted gene complexes, many questions about SD remain. First, despite having the molecular identities of both the driver and the target, a basic understanding of how Sd and Rsp interact to cause segregation distortion continues to elude. Second, we know little about Rsp and its function in spermatogenesis. Why is the Rsp satellite in particular, and not others, sensitive to Sd-mediated distortion? And why do sensitive alleles of Rsp persist in populations? Third, how is it that the presence of mislocalized, enzymatically active Sd–RanGAP seems to affect only spermiogenesis without inducing other major phenotypic consequences? Fourth, the molecular identities of the many modifiers of SD in the genome are still unknown. E(SD) is a strong enhancer, has the capacity to drive on its own and, in extra doses causes wild-type RanGAP to be mislocalized to the nucleus. The two other major modifiers, M(SD) and St(SD), are less well characterized, having been largely neglected despite their large effects on the expression of segregation distortion. Identifying and characterizing these modifiers and the segregating suppressors of SD may well provide the clues necessary to pin down the molecular basis of segregation distortion and some of the unexplained phenomena, like negative distortion (see above Hiraizumi 1989, 1990, 1993; Hiraizumi et al. 1994).
Going forward, work on SD seems certain to provide important insights into the genetics, genomics, evolution, and basic biology of spermatogenesis in Drosophila. So far SD’s mode of action has remained obscure, in part, because we know surprisingly little about the genetic control of spermiogenesis, nuclear remodeling, the role of small RNAs during spermatogenesis, and the maintenance, function, and evolution of genomic satellite DNAs. The SD system thus presents an entrée into these notoriously difficult problems. Determining, for instance, the connection between Sd and Rsp, as well as the molecular identities and functions of the various modifiers of SD, will provide information on the genetic control of spermiogenesis, including the role of RanGAP. Determining how large blocks of Rsp satellite (Rsps alleles) lead to spermatid dysfunction in the presence of SD will undoubtedly have implications for how pericentromeric heterochromatin is regulated during spermatogenesis and, possibly, for the role of small RNAs in the male germline.
We thank Colin Meiklejohn for helpful discussions. We also thank Michael Turelli and two anonymous reviewers who provided comments on the manuscript and Michael Turelli for inviting us to write it. Our work on Segregation Distorter is supported by funds from the Alfred P. Sloan Foundation, the David and Lucille Packard Foundation, and the University of Rochester.
Communicating editor: M. Turelli
- Received April 21, 2012.
- Accepted June 8, 2012.
- Copyright © 2012 by the Genetics Society of America