Genetic Analysis of the Hybrid male rescue Locus of Drosophila

Corresponding author: H. Allen Orr, Department of Biology, University of Rochester, Rochester, NY 14627., aorr{at}mail.rochester.edu (E-mail)

Communicating editor: R. S. HAWLEY

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Several hybrid rescue mutations—alleles that restore the viability of normally lethal hybrids—have been discovered in Drosophila melanogaster and its relatives. Here we analyze one of these genes, Hybrid male rescue (Hmr), asking two questions about its role in hybrid inviability. (1) Does the wild-type allele from D. melanogaster (Hmr^mel) cause hybrid embryonic inviability? (2) Does Hmr^mel cause hybrid larval inviability? Our results show that the wild-type product of Hmr is neither necessary nor sufficient for hybrid embryonic inviability. Hmr^mel does, however, appear to lower the viability of hybrid larvae. The data further suggest (though do not prove) that Hmr^mel acts as a gain-of-function poison in hybrids. These findings support previous claims that hybrid embryonic and larval lethalities are genetically distinct and suggest that Hmr^mel is at least one of the proximate causes of hybrid larval inviability.

ONE of the most important and surprising findings in the genetics of speciation has been the discovery of "hybrid rescue" mutations. At least since DOBZHANSKY 1937 , evolutionary biologists have imagined that speciation involves the gradual accumulation of many slight incompatibilities between populations. Reproductive isolation should therefore be essentially irreversible: because taxa will come to differ at many sets of genes, hybrids will suffer many developmental problems and it is implausible to think that evolution could ever retrace its steps, restoring the fitness of hybrids. In MULLER's (1939) words, genetic divergence between populations leads to an "ever more pronounced immiscibility" that cannot be undone.

The discovery by Watanabe and by Ashburner and their colleagues of at least five different hybrid rescue mutations comes as a surprise, then (reviewed in HUTTER et al. 1990 and SAWAMURA et al. 1993). These mutations, when introduced singly into species hybrids, restore the viability of normally lethal hybrids. Some mutations appear to have a nearly complete effect on hybrid fitness, i.e., a lethal sex is restored to a 50:50 sex ratio. Most of the known rescue mutations occur in the Drosophila melanogaster group (HUTTER 1997 ), rescuing hybrids formed when D. melanogaster ("mel") is crossed to its relatives D. simulans, D. mauritiana, or D. sechellia (these three sibling species are hereafter referred to as "sib"). Rescue mutations will presumably turn out to be widespread. The fact that they are best known in Drosophila surely reflects the intense scrutiny to which these hybrids have been subjected.

The discovery of rescue mutations raises several questions. First, how many different genes can rescue hybrids? Although the screens performed so far have not been exhaustive, four (and perhaps five) rescue loci have been found and each new screen seems to uncover new ones, e.g., SAWAMURA et al. 1993A , SAWAMURA et al. 1993B discovery of maternal hybrid rescue (mhr) and Zygotic hybrid rescue (Zhr). Second, are rescue genes alleles of "speciation genes"? In other words, are rescue genes mutant alleles of the genes that normally kill hybrids? This question is important because, if the answer is yes, the study of rescue genes may provide a shortcut to the molecular characterization of the genes causing speciation. Rescue mutations need not, however, be alleles of speciation genes. They might instead be second-site suppressors—loci that, while not causing hybrid inviability, can, when mutated, override the effects of the genes actually killing hybrids. (Rescue genes might, for instance, allow use of an alternative metabolic pathway that sidesteps the primary problematic pathway.) Under the second-site hypothesis, rescue mutations may even occur at loci that have not diverged molecularly between the relevant species.

Third and most important, does the recovery of rescue genes imply that reproductive isolation has a simple developmental basis? After all, if inviability reflects many independent developmental problems, the chance of recovering single mutations that simultaneously correct all of these problems would seem vanishingly small. A simple developmental basis, in turn, suggests a simple genetic basis. If many genes cause hybrid inviability, why do they all affect the same developmental pathway?

HUTTER et al. 1990 argued that hybrid inviability in the mel-sib species cross is, in fact, simple and proposed an elegant genetic model to explain it. This model had two important features. First, it posited that hybrid inviability is caused by wild-type alleles at known rescue genes (i.e., rescue mutations are alleles of speciation genes). Second, it posited that hybrid inviability reflects a single incompatibility between the wild-type alleles of two hybrid rescue genes, Hybrid male rescue (Hmr) and Lethal hybrid rescue (Lhr).

The Hutter et al. model elegantly explains the complex results obtained when mel and sib are crossed. Postzygotic isolation between these taxa involves both larval and embryonic lethality. When mel females are crossed to sib males, only hybrid females appear; males die as third-instar larvae. The reciprocal cross produces only hybrid males, with females dying as embryos (HADORN 1961 ). Interestingly, the cross of mel attached-X females to sib males produces only males, with females dying as late larvae. [See HUTTER et al. 1990 for a summary of these and other crossing results.] Because all hybrids, viable and inviable, from these crosses carry a haploid set of autosomes from each species, differences in survival must involve the sex chromosomes and/or cytoplasm. It is now clear that hybrid larval and embryonic lethality involve genes on the mel X chromosome (ORR 1991 ; YAMAMOTO 1992 ). Hutter et al. suggested that both forms of lethality involve the same X-linked locus, the rescue gene Hybrid male rescue (Hmr). According to their model, "the product of the l⁺ [i.e., Hmr^mel] allele would lead to death" (HUTTER et al. 1990 , p. 918) when brought together in hybrids with the wild-type allele of another rescue gene, Lethal hybrid rescue (Lhr), from sib. They further suggested that the rescue allele, Hmr, is a loss-of-function mutation: removal of the lethal Hmr^mel product restores hybrid viability. (Their data suggest that Hmr is a hypomorph, not an amorph.)

Later work, however, called this simple model into question. SAWAMURA et al. 1993C extensive analysis of several rescue mutations suggested that embryonic and larval lethalities have different genetic bases, with Hmr rescuing larval fitness only. Sawamura and colleagues thus proposed an alternative model of mel-sib hybrid inviability. This model also had two salient features. First, like Hutter et al.'s model, it posited that postzygotic isolation is caused by wild-type alleles at hybrid rescue genes. But second, it posited that one set of loci (mhr^sib and Zhr^mel) causes hybrid embryonic lethality, while another set (Hmr^mel and Lhr^sib) causes larval lethality.

The Sawamura model explains a remarkable range of experimental results and is now widely accepted. Several findings, however, suggest that this "separate compartment" view of hybrid embryonic vs. larval lethality might be too simple. First, HUTTER et al. 1990 presented strong evidence that addition of Hmr^mel as a duplication to hybrid males causes embryonic, not larval, lethality. Moreover, several groups have repeatedly found that the mutation Hmr appears to rescue hybrid male embryonic lethality, at least at low temperatures. In the cross of D. simulans attached-X females to mel males, X^mel-bearing hybrid males die as embryos (SAWAMURA et al. 1993A ). But ORR 1991 , HUTTER et al. 1990 , and SAWAMURA et al. 1993A all showed that such males are weakly rescued in crosses to Hmr. The latter authors thus concluded that "[a] possibility remains that Hmr can sometimes rescue ... embryonic lethality" (SAWAMURA et al. 1993A , p. 305).

These findings might be legitimately explained away under the Sawamura model. [For example, the Hmr stock might by chance carry a Zhr mutation. This does not, however, seem likely as several stocks were used; see also SAWAMURA et al. 1993B , whose results suggest that their Hmr stock was Zhr⁺.] In any case, such special pleading is unnecessary. As we will see, definitive tests of Hmr's role in hybrid embryonic vs. larval lethality are possible.

One of the sources of uncertainty about the precise roles of rescue genes in hybrid inviability is clear. Much of the literature involves attempts to infer the action of wild-type alleles from the behavior of a small number (usually one per locus) of poorly characterized mutant alleles. (We sometimes do not know if a mutation is an amorph, hypomorph, hypermorph, or neomorph.) Although this strategy may often succeed, it is of course entirely possible that the wild-type allele at, e.g., Hmr may affect both embryos and larvae, while the particular lesion producing a Hmr mutation only restores larval viability. In such a case, inferences from the behavior of the mutant allele would mislead.

Here we attempt to obtain clearer evidence about the role of rescue genes in postzygotic isolation by manipulating wild-type alleles at a rescue gene. In particular, we investigate the role of Hmr^mel in hybrid inviability by testing whether its removal (by deficiency) from species hybrids rescues embryonic viability and whether its addition (by duplication) to species hybrids causes embryonic or larval lethality. Our results provide strong additional support for the Sawamura model. In particular, they reveal that wild-type product of Hmr is neither necessary nor sufficient for hybrid embryonic inviability. Hmr^mel does, however, appear to affect hybrid larval viability. The wild-type allele at Hmr may therefore be a "speciation gene."

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Mutations and stocks:
Hmr, which was described by HUTTER and ASHBURNER 1987 ; HUTTER et al. 1990 and ZHANG et al. 1999 , resides at 32.0 on the X chromosome within polytene bands 9D1-9E4.

The experiments below use three X chromosome deletions. The first is Df(1)HC133 (9B9; 9F2-5), which Hutter et al. argue includes Hmr. The distal breakpoint of this deficiency extends far to the left of the likely position of Hmr. The proximal breakpoint is to the right of fliK (9F3-5) which, according to the mapping of HUTTER et al. 1990 , is to the right of Hmr (see also ZHIMULEV et al. 1987 and LINDSLEY and ZIMM 1992 ). Df(1)HC133 includes raspberry (ras; 1-32.35), which is very tightly linked to Hmr (ZHANG et al. 1999 ). Two other deficiencies, Df(1)ras-v-17 (9D1-2; 10A2-3) and Df(1)v^-L15 (9B1-2; 10A1-2), were also used. The breakpoints of these large deficiencies suggest they include Hmr (also D. BARBASH, personal communication).

Hutter et al. also argue that the duplication Dp(1;2)v^+75d (9A2; 10C2) includes Hmr. This duplication extends far past Hmr's likely position on both the left and right. More important, Hutter et al. show that males bearing the rescue allele Hmr are not rescued if they also carry Dp(1;2)v^+75d, a result that we confirm below.

D. mauritiana Synthetic is a mixture of six isofemale lines collected by O. Kitagawa on Mauritius in 1981; the lines were pooled in 1983. This stock, as well as D. mauritiana v^C, D. simulans v m, and D. simulans Ro/+ were kindly provided by J. A. Coyne. All other mutations are described by LINDSLEY and ZIMM 1992 .

Crosses:
Although Hmr rescues the viability of hybrids from all of the mel-sib species crosses, the present experiments focus on the D. melanogaster-D. mauritiana ("mel-maur") and D. melanogaster-D. simulans ("mel-sim") hybridizations. These species crosses differ in one important respect. While the maur female x mel male cross gives unambiguous results (females almost never appear), the sim female x mel male cross is leaky, i.e., hybrid females sometimes appear, depending on temperature and the particular sim stock used (ORR 1993A ; SAWAMURA et al. 1993).

Some of the species crosses below involved the CyO balancer. Because Cy sometimes overlaps wild type, we scored the phenotype of 217 Df(1)HC133; Dp(1;2)v^+75d/CyO males at 18° and 22°; all but one were phenotypically Cy. Cy's reliability was further confirmed in several preliminary crosses, e.g., mel v females x Df(1)HC133; Dp(1;2)v^+75d/CyO males. All phenotypically Cy males (69) were vermillion-eyed while all phenotypically Cy⁺ males (91) were wild-eyed, as expected. This result of course also confirms the presence of the v⁺-bearing duplication in the Df(1)HC133; Dp(1;2)v^+75d/CyO stock.

Flies were typically aged for several days before species crosses were set up. All crosses were performed on standard cornmeal medium at 22° or 18°, except as indicated.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Deletion/duplication tests:
A direct test of Hmr^mel's role in embryonic lethality requires its deletion from species hybrids. If, as Hutter et al. suggested, Hmr^mel product causes hybrid inviability—and the rescue allele Hmr is a loss-of-function mutation—deletion of Hmr^mel must rescue embryonic lethality. Conversely, addition of the putatively poisonous Hmr^mel product to hybrids that do not normally carry it must cause lethality.

Fig 1 shows how Hmr^mel can be deleted from normally lethal hybrids and added to other, normally viable hybrids. In particular, one can delete Hmr^mel by deficiency from hybrid females (which normally die as embryos), giving them an Hmr genotype that is identical to that of viable hybrid males. Simultaneously, one can add Hmr^mel by duplication to hybrid males (which normally survive), giving them an Hmr genotype that is identical to that of lethal hybrid females.

View larger version (14K): In this window In a new window Download PPT slide   Figure 1. Deletion/duplication test of Hmr in sib-mel hybrids. HUTTER et al. 1990 model predicts that hybrid females who have had Hmrmel deleted should be viable, while Dp-bearing hybrid males should be embryonic lethal. In most cases, "Df(1)Hmr-" = Df(1)HC133, while "Dp(1;2)Hmr+" = Dp(1;2)v+75d (see Table 2).

We first confirm that the normal species crosses behave as expected. Consider first the results of crosses between mel and maur. Table 1 shows that all control crosses behave as expected at both 18° and 22°. The cross of mel female x maur male produces only hybrid daughters, with males dying as late larvae. The reciprocal cross produces only hybrid males, with hybrid females dying as embryos, although a few individuals hatch and die as first-instar larvae (at 18°, one escaper survived to adulthood). Table 1 also shows that the Hmr mutation rescues normally inviable hybrid males. Preliminary crosses also confirmed that the Df(1)HC133 stock carries a deficiency in the expected region (not shown).

We now consider the deficiency/duplication tests. The within-species control cross of mel females x mel Df(1)HC133; Dp(1;2)v^+75d/CyO males behaves as expected: all genotypes of males and females are recovered in reasonable numbers. The critical cross of maur females x mel Df(1)HC133; Dp(1;2)v^+75d/CyO males, however, behaves quite differently. While hybrid females who carry the duplication are lethal as expected, females who carry the balancer chromosome remain completely inviable despite the fact that Hmr^mel has been deleted (Table 2). Moreover, these females die as embryos: a large number of brown eggs (and a few dead first-instar larvae) were observed. This result shows that Hmr^mel is not necessary for hybrid embryonic lethality. Table 2 further shows that hybrid males who have had Hmr^mel added by duplication remain viable. Thus Hmr^mel is also not sufficient for hybrid embryonic lethality. Although hybrid females carrying one maur X and one mel X (in a maur cytoplasm) are invariably lethal, hybrid males carrying one maur X and one piece of the mel X including Hmr^mel (in a maur cytoplasm) are not.

It is, however, worth noting a point that will become increasingly important. Although males carrying a duplication of Hmr^mel are not invariably lethal, they do appear to suffer some inviability: Dp-bearing males are typically rarer than non-Dp-bearing males. Although the effect is frustratingly variable, recovery ratios as extreme as 210:34 are not uncommon. This recovery bias does not reflect a trivial property of the duplication because no such shortage of Dp-bearing flies occurs in within-species controls. Indeed Dp vs. balancer males appear in almost perfect Mendelian ratios within mel (Table 2, lines 1 and 2). Duplications of Hmr^mel may therefore cause some hybrid larval lethality. We will return to this point below.

The deficiency/duplication test was extended in one trivial way: it was repeated using another deficiency, Df(1)ras-v-17 (9D1-2; 10A2-3), that extends farther proximally. Although the required cross (maur Synthetic females x mel Df(1)ras-v-17 ; Dp(1;2)v^+75d/+ males) proved extremely difficult and only 10 hybrids were recovered, all were, once again, males.

Table 2 also reveals that all of the above results (both intra- and interspecific) remain essentially unchanged at the two temperatures studied (18° and 22°).

The mel-sim hybridization:
Because species crosses between mel and sim proved far more difficult (especially in the critical sim female x mel male direction), some hybridizations were performed at only one temperature, 22°. Table 1 again shows that control crosses behave as expected. Note that the sim female x mel male cross is leaky. Although female hybrids are rarer than males, they occur at appreciable frequencies.

Does deletion of Hmr^mel boost the frequency of these weakly viable females to much higher "male-like" values? The answer is no. Table 2 shows that females who carry no Hmr^mel appear (at most) only one-third as often as males. Once again, hybrid males who have had Hmr^mel added to their genome are recovered fairly readily. Also once again, however, these males appear to suffer some decreased fitness as Dp-bearing males are consistently rarer than their balancer brothers, although the effect is small and variable.

Table 2 also reports the results of a deletion/duplication test in which a much larger X chromosome deletion, Df(1)v^-L15 (9B1-2; 10A1-2), was employed. The results are nearly identical to those obtained with Df(1)HC133.

Effect of Hmr^mel on hybrid larval viability:
The possibility that Hmr^mel affects larval viability was tested further by crossing mel attached-X; Dp females to sib males and scoring the recovery of Dp- and non-Dp-bearing hybrid sons. These crosses also allow us to test whether Hmr^mel's apparent fitness effect depends on the species origin of the cytoplasm. This is important because, although all of the above crosses produce hybrids carrying sib cytoplasm, normally lethal F₁ males carry mel cytoplasm.

Table 3 shows that X^sib hybrid males who carry a duplication including Hmr^mel are again rarer than their balancer brothers, with dead late larvae and pupae common. Once again, though, the results are variable. In crosses to maur, for instance, Dp-bearing adults make up ~35–40% of males. In crosses to sim, they make up ~10–20% of males.

We can draw two conclusions. First, Hmr^mel appears to cause some hybrid larval inviability. Indeed, summing across the hybridizations shown in Table 2 and Table 3, Dp-bearing males are rarer than their balancer brothers in 11 of 12 species crosses, a pattern that is highly significant (Wilcoxon's signed rank test: z = 2.845, P < 0.005). Second, this fitness effect does not depend on the species identity of the cytoplasm.

These results are consistent with HUTTER et al. 1990 finding that addition of Hmr^mel as a duplication to otherwise rescued hybrid males carrying the Hmr mutation reverses rescue, a result that suggests (but does not prove) that Hmr^mel produces a product that kills hybrid males. We verified this finding by crossing mel y Hmr v/Df(1)HC133; Dp(1;2)v^+75d/+ females to maur Synthetic males. We recovered 901 hybrid females and 78 non-Dp-bearing Hmr males (phenotypically v). No Dp-bearing males (phenotypically v⁺) were recovered. As expected, many dead third-instar larvae/pseudopupae were seen. This inviability is not an artifact of an unusual interaction between the Hmr mutation and Dp(1;2)v^+75d because no such lethality occurs within species: the cross of mel y Hmr v females to mel Df(1)HC133; Dp(1;2)v^+75d/CyO males produced 73 non-Dp-bearing Hmr males (phenotypically Cy and v) and 101 Dp-bearing males (phenotypically Cy⁺ and v⁺), as well as many females. Together with the above, these results suggest that Hmr^mel causes (or at least contributes) to hybrid larval lethality between D. melanogaster and species belonging to the D. simulans clade.

These data are also consistent with Hutter et al.'s suggestion that Hmr is a loss-of-function mutation. Put conversely, Hmr^mel appears to act in a gain-of-function manner.

Interaction with Lhr:
The hypothesis that Dp(1;2)-v^+75d's deleterious effect reflects the action of Hmr^mel leads to a simple prediction. This effect should be reversible by the hybrid larval rescue mutation, Lhr. If, on the other hand, the observed deleterious effect is due to some other factor(s) carried on the duplication chromosome (which acts only in hybrids), the presence vs. absence of Lhr should be irrelevant. We tested this prediction by crossing mel C(1)DX, y f; Dp(1;2)v^+75d/CyO females to control sim v m and to experimental Lhr males. The crosses proved extremely difficult, but 155 hybrids were obtained in the former cross and 209 in the latter (after months of crossing). As Table 4 reveals, the results were consistent with our prediction: Lhr allows recovery of significantly more Dp-bearing males than in controls (² = 4.24, P < 0.05), although the effect is small.

The effect of extra doses of Hmr^mel:
Although the above results suggest that Hmr^mel plays a role in larval but not embryonic hybrid inviability, at least one loose end remains. Hutter et al. presented suggestive evidence that Hmr^mel affects embryonic viability. They argued that hybrid males that are normally destined to die as larvae will instead die as embryos if forced to carry an extra copy of Hmr^mel. In particular, they showed that the cross of mel Dp(1;2)v^+75d/CyO females x maur males—which is expected to produce males that die as late larvae—instead produces many lethal embryos (HUTTER et al. 1990 , their Table 7). They guessed that these dead embryos were Dp-bearing hybrid males; lethal embryos were not seen when using other duplications that do not include Hmr.

We performed a similar experiment. Through a series of crosses, we produced mel females of the genotype y² cv v f/y² cv v f; Dp(1;2)v^+75d/+. In the experimental cross, these Dp-bearing mel females were crossed to maur males (Table 5, line 1). This cross, like Hutter et al.'s, should produce dead embryos. In an intraspecific control, Dp-bearing mel females were crossed to mel males (Table 5, line 2). This cross controls for any deleterious effect the duplication (or any other factor in the duplication stock) may have within species. In a second control, non-Dp-bearing mel females were crossed to maur males (Table 5, line 3). This cross controls for the effect of the species cross per se on embryonic viability (i.e., the traditional description of this hybridization could be wrong, with lethal embryos routinely appearing).

Our results confirm that lethal embryos appear when Dp-bearing females are crossed to maur males (Table 5). However, dead embryos also appear when Dp-bearing females are crossed within species (Mann-Whitney U-test against the intraspecific control: z = -0.76, P > 0.40). As expected, lethal embryos are very rare in the "normal" species cross in which no Dp is used (Mann-Whitney U-test comparing lines 1 and 3 in Table 5: z = -4.27, P < 0.0001). Although these results must be viewed as rough (as we have no guarantee of equal fertilization rates), the embryonic lethality seen in Hutter et al. and in the present Table 5 (line 1) may have little to do with the genetics of hybrid inviability. Instead, some factor in the Dp(1;2)v^+75d stock clearly causes embryonic lethality, whether in hybrids or within species (although we cannot rule out the possibility that different levels of lethality occur within vs. between species). This embryonic lethality, however, does not appear to involve the duplication per se as intraspecific controls show that duplication and balancer flies appear in Mendelian ratios, at least among males (Table 2, lines 1 and 2).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Two conclusions follow from these experiments. First, the wild-type allele Hmr^mel does not cause hybrid embryonic lethality. This result is essentially proved by the finding that hybrid females who have had the putatively poisonous Hmr^mel allele deleted remain embryonic lethal. [It is worth noting that Hutter et al. also introduced deletions of Hmr^mel into hybrids. Unfortunately, they only considered the mel female x sib male direction of the cross (their Table 8). Because hybrid females are already viable in this direction of the cross, deletion of Hmr^mel, for our purposes, is uninformative.] The present finding thus militates against the original Hutter et al. model of hybrid inviability, providing strong support for the later SAWAMURA et al. 1993C one. Although a good body of data already suggested that Hmr (as well as Lhr) play no role in embryonic lethality, this inference was based on the behavior of a small set of mutant alleles. The present results, on the other hand, derive solely from manipulations of the wild-type allele at Hmr. Fortunately, the mutant and deficiency/duplication approaches reach identical conclusions: hybrid embryonic lethality has some cause (or causes) other than Hmr^mel.

Second, Hmr^mel does appear to cause hybrid larval lethality. This conclusion, unlike the first, must remain tentative. Our results show that addition of Hmr^mel to hybrid males who do not ordinarily carry it significantly lowers their viability. Although the magnitude of this effect is frustratingly variable, its direction is consistent. Moreover, the effect is independent of the species origin of the cytoplasm (mel vs. sib). D. BARBASH, J. ROOTE and M. ASHBURNER (personal communication) have recently documented a similar effect: addition of an extra dose of Hmr^mel to hybrid females lowers their viability relative to balancer controls. Although it is difficult to prove that these duplication viability effects are due to the Hmr locus per se, this seems likely for two reasons. First, duplications harboring Hmr^mel have no such viability effect within species (Table 2), suggesting that the effect is tied to the genetics of hybrid inviability, not to some trivial effect of the duplication itself. Second, the deleterious effects of Hmr^mel-bearing duplications can be reversed by the rescue mutation Lhr (Table 4), although the effect is small.

This work supports two key features of the Sawamura model. First, the two forms of hybrid inviability characterizing the mel-sib species crosses appear genetically distinct. Hmr (and Lhr) apparently affect only hybrid larval viability, while mhr and Zhr apparently affect only hybrid embryonic viability. Second, hybrid lethality appears to result from the action of the wild-type allele at the Hmr locus. Our results thus support the "speciation gene" hypothesis (rescue mutations are alleles at the genes whose wild-type product causes hybrid inviability) and militate against the second-site suppressor hypothesis (rescue mutations are second-site suppressors at genes whose wild-type product does not cause hybrid inviability).

These results also touch on several other properties of Hmr that merit discussion. First, the fact that duplications carrying Hmr^mel have variable (but deleterious) effects on hybrid viability suggests that hybrid lethality may be a threshold character that is very sensitive to subtle environmental change (see also ORR 1996 ). This interpretation is strengthened by the fact that duplication viability effects vary considerably even when performing the same cross with the same stocks at different times (see the repeated crosses in Table 2).

Second, Hutter et al. argued that the rescue allele Hmr is a partial loss-of-function mutation. Conversely, the wild-type allele Hmr^mel acts in a gain-of-function manner. The present findings are consistent with this suggestion. Similar results appear to hold at the Zhr gene (SAWAMURA et al. 1993B ) as well as at the Tumor (Tu) locus, which causes inviability among certain swordtail-platyfish backcross hybrids (WITTBRODT et al. 1989 ).

It is important to understand what such gain-of-function behavior does and does not imply. MULLER 1942 suggested that the genes causing postzygotic isolation often behave as recessives in hybrids. He further claimed that this behavior may explain Haldane's rule [the preferential sterility/inviability of the hemizygous (XY) sex hybrids]. This idea eventually led to the modern dominance theory of Haldane's rule (TURELLI and ORR 1995 ). But as ORR and TURELLI 1996 emphasized, "[t]he dominance theory says nothing about why the alleles causing hybrid problems act as partial recessives. It merely says that if they are recessive, Haldane's rule will follow" (p. 613). Early on, however, ORR 1993B speculated that the recessivity of speciation genes might imply that they act in a loss-of-function manner when placed on a hybrid background. The above findings suggest that this speculation is wrong. But this conclusion does not affect the dominance theory per se. The dominance theory holds only that the fitness of l/+ hybrids (where l represents a hybrid lethal/sterile) is closer to that of +/+ than l/l hybrids, regardless of the biochemical basis of this relationship. In short, "speculation about mechanism can be distinguished from the dominance theory per se" (ORR and TURELLI 1996 ).

In summary, the deletion/duplication tests reported here allow us to rule out the simplest models of how Hmr^mel may cause postzygotic isolation. They instead provide strong support for the alternative, and reigning, SAWAMURA et al. 1993C model. Given that Hmr^mel may be a direct cause of hybrid larval inviability, its genetic and molecular characterization may well provide a valuable window on the genetic causes of speciation.

	ACKNOWLEDGMENTS

We thank M. Ashburner, D. A. Barbash, A. Davis, M. M. Green, R. S. Hawley, C. D. Jones, C. H. Langley, D. C. Presgraves, J. Roote, and two anonymous reviewers for helpful discussions and comments. We especially thank M. Ashburner, J. Roote, and D. A. Barbash for sharing many of their unpublished results with us. This work was supported by National Institutes of Health grant GM-51932.

Manuscript received November 1, 1999; Accepted for publication January 27, 2000.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

DOBZHANSKY, T., 1937 Genetics and the Origin of Species. Columbia University Press, New York.

HADORN, E., 1961  Zur Autonomie und Phasenspezifitat der Latalitat von Bastarden zwischen Drosophila melanogaster und Drosophila simulans.. Rev. Suisse Zool. 68:197-207.

HUTTER, P., 1997  Genetics of hybrid inviability in Drosophila.. Adv. Genet. 36:157-185[Medline].

HUTTER, P. and M. ASHBURNER, 1987  Genetic rescue of inviable hybrids between Drosophila melanogaster and its sibling species. Nature 327:331-333[Medline].

HUTTER, P., J. ROOTE, and M. ASHBURNER, 1990  A genetic basis for the inviability of hybrids between sibling species of Drosophila. Genetics 124:909-920[Abstract].

LINDSLEY, D. L., and G. G. ZIMM, 1992 The Genome of Drosophila melanogaster. Academic Press, San Diego.

MULLER, H. J., 1939  Reversibility in evolution considered from the standpoint of genetics. Biol. Rev. Camb. Phil. Soc. 14:261-280.

MULLER, H. J., 1942  Isolating mechanisms, evolution and temperature. Biol. Symposia 6:71-125.

ORR, H. A., 1991  Genetic basis of postzygotic isolation between D. melanogaster and D. simulans.. Dros. Info. Serv. 70:161-162.

ORR, H. A., 1993a  Haldane's rule has multiple genetic causes. Nature 361:532-533[Medline].

ORR, H. A., 1993b  A mathematical model of Haldane's rule. Evolution 47:1606-1611.

ORR, H. A., 1996  The unexpected recovery of hybrids from a Drosophila species cross. Genet. Res. 67:11-18[Medline].

ORR, H. A. and M. TURELLI, 1996  Dominance and Haldane's rule. Genetics 143:613-616[Medline].

SAWAMURA, K., T. TAIRA, and T. K. WATANABE, 1993a  Hybrid lethal systems in the Drosophila melanogaster species complex. I. The maternal hybrid rescue (mhr) gene of Drosophila simulans. Genetics 133:299-305[Abstract].

SAWAMURA, K., M.-T. YAMAMOTO, and T. K. WATANABE, 1993b  Hybrid lethal systems in the Drosophila melanogaster species complex. II. The Zygotic hybrid rescue (Zhr) gene of D. melanogaster. Genetics 133:307-313[Abstract].

SAWAMURA, K., T. K. WATANABE, and M.-T. YAMAMOTO, 1993c  Hybrid lethal systems in the Drosophila melanogaster species complex. Genetica 88:175-185[Medline].

TAKAMURA, T. and T. K. WATANABE, 1980  Further studies on the Lethal Hybrid Rescue (Lhr) gene of Drosophila simulans.. Jpn. J. Genet. 55:405-408.

TURELLI, M. and H. A. ORR, 1995  The dominance theory of Haldane's rule. Genetics 140:389-402[Abstract].

WITTBRODT, J., D. ADAM, B. MALITSCHEK, W. MAUELER, and F. RAULF et al., 1989  Novel putative receptor tyrosine kinase encoded by the melanoma-inducing Tu locus in Xiphophorus.. Nature 341:415-421[Medline].

YAMAMOTO, M.-T., 1992  Inviability of hybrids between D. melanogaster and D. simulans results from the absence of simulans X not the presence of simulans Y chromosome. Genetica 87:151-158[Medline].

ZHANG, Y. Q., J. ROOTE, S. BROGNA, A. W. DAVIS, and D. A. BARBASH et al., 1999  stress sensitive B encodes an adenine nucleotide translocase in Drosophila melanogaster.. Genetics 153:891-903[Abstract/Free Full Text].

ZHIMULEV, I. F., G. V. POKHOLKOVA, A. B. BGATOV, G. H. UMBETOVA, and I. V. SOLOVJEVA et al., 1987  Fine cytogenetical analysis of the band 10A1-2 and the adjoining regions in the Drosophila melanogaster X-chromosome. V. Genetic characteristics of the loci in the 9E-10B region. Biol. Zent. Bl. 106:699-720.

This article has been cited by other articles:

				D. A. Barbash Nup96-Dependent Hybrid Lethality Occurs in a Subset of Species From the simulans Clade of Drosophila Genetics, May 1, 2007; 176(1): 543 - 552. [Abstract] [Full Text] [PDF]

				N. J. Brideau, H. A. Flores, J. Wang, S. Maheshwari, X. Wang, and D. A. Barbash Two Dobzhansky-Muller genes interact to cause hybrid lethality in Drosophila. Science, November 24, 2006; 314(5803): 1292 - 1295. [Abstract] [Full Text] [PDF]

				H. A. Orr The genetic basis of reproductive isolation: Insights from Drosophila PNAS, May 3, 2005; 102(suppl_1): 6522 - 6526. [Abstract] [Full Text] [PDF]

				H. A. Orr and S. Irving Segregation Distortion in Hybrids Between the Bogota and USA Subspecies of Drosophila pseudoobscura Genetics, February 1, 2005; 169(2): 671 - 682. [Abstract] [Full Text] [PDF]

				D. A. Barbash, D. F. Siino, A. M. Tarone, and J. Roote A rapidly evolving MYB-related protein causes species isolation in Drosophila PNAS, April 29, 2003; 100(9): 5302 - 5307. [Abstract] [Full Text] [PDF]

				D. C. Presgraves A Fine-Scale Genetic Analysis of Hybrid Incompatibilities in Drosophila Genetics, March 1, 2003; 163(3): 955 - 972. [Abstract] [Full Text] [PDF]

				H. A. Orr and S. Irving Complex Epistasis and the Genetic Basis of Hybrid Sterility in the Drosophila pseudoobscura Bogota-USA Hybridization Genetics, July 1, 2001; 158(3): 1089 - 1100. [Abstract] [Full Text] [PDF]