Two Dobzhansky-Muller Genes Interact to Cause Hybrid Lethality in Drosophila

See allHide authors and affiliations

Science  24 Nov 2006:
Vol. 314, Issue 5803, pp. 1292-1295
DOI: 10.1126/science.1133953


The Dobzhansky-Muller model proposes that hybrid incompatibilities are caused by the interaction between genes that have functionally diverged in the respective hybridizing species. Here, we show that Lethal hybrid rescue (Lhr) has functionally diverged in Drosophila simulans and interacts with Hybrid male rescue (Hmr), which has functionally diverged in D. melanogaster, to cause lethality in F1 hybrid males. LHR localizes to heterochromatic regions of the genome and has diverged extensively in sequence between these species in a manner consistent with positive selection. Rapidly evolving heterochromatic DNA sequences may be driving the evolution of this incompatibility gene.

Plant and animal hybrids are often sterile or lethal as a result of interspecific genetic divergence. The Dobzhansky-Muller model proposes that hybrid incompatibilities (HIs), which contribute to speciation, evolve as a consequence of interactions between or among genes that have diverged in each of the hybridizing species (1). Dobzhansky-Muller incompatibility genes require three criteria: Each gene reduces hybrid fitness, has functionally diverged between the hybridizing species, and depends on the partner gene to cause HI (Fig. 1A). Major-effect HI genes have been discovered, and functional divergence of single genes has been demonstrated by genetic tests (2, 3) or suggested by patterns of molecular evolution (4). Although HI systems composed of complementary factors have been described (5, 6) and interacting genomic regions of hybridizing species identified (79), no pair of Dobzhansky-Muller genes has been reported. It remains unclear whether HI phenotypes can be explained even in part by two-locus interactions, or alternatively whether HIs require complex multilocus interactions (1012).

Fig. 1.

CG18468 encodes Lhr. (A) Model of Hmr and Lhr functional divergence and the interaction that causes hybrid lethality. Lethality results from the interaction between the D. melanogaster Hmr allele (Hmr-mel) and the D. simulans Lhr allele (Lhr-sim). (B) Map of Lhr region in D. simulans. Genetic markers and estimated cytological locations in D. melanogaster are shown below the line. Ten recombinants between jba and Lhr were selected and cross-overs mapped within a region of approximately 480 kb; the most proximal recombinant was between Sip1 and qkr54B (diagram not to scale). (C) CG18468 has a characteristic leucine zipper-like structure in all Drosophila species except D. simulans, D. mauritiana, and D. sechellia, which have a 16–amino acid insertion; this insertion is lacking in the D. simulans rescue strain Lhr2.(D) Map of CG18468 region at 54B7 and insertion in the Lhr1 strain. Black boxes, coding regions; unfilled boxes, UTRs; arrows, predicted translation start sites. The large triangle represents an insertion of ∼4 kb (triangle not to scale) located between nucleotides 11 and 13 of the predicted CG18468 mRNA in Lhr1. (E) The insertion in Lhr1 reduces the level of mRNA of CG18468 but not of CG6546. RT-PCR with larval RNA from a vermillion (v)–marked D. simulans strain and from the Lhr1 strain.

Interspecific crosses of D. melanogaster females to D. simulans males produce no sons. The D. simulans mutation Lhr1 suppresses the lethality of these F1 hybrid males (13). A mutant allele of the X-linked D. melanogaster gene Hmr similarly suppresses hybrid male lethality (14). Hmr encodes a rapidly evolving protein with sequence similarity to the myb/SANT-like domain in Adf-1 (MADF) class of DNA binding proteins (14). Genetic interaction studies (15, 16) suggested that Hmr and Lhr interact to cause lethality in a manner consistent with the Dobzhansky-Muller model (Fig. 1A).

Lhr1 maps 1.7 centimorgans from the visible marker jba on chromosome 2R (17). We identified recombinants between Lhr1 and jba and typed them using molecular markers that distinguish the Lhr1 and jba D. simulans strains (Fig. 1B). On the basis of the distribution of recombinants, Lhr1 is likely within several hundred kilobases centromere-proximal to qkr54B.

Because of the paucity of visible markers in D. simulans, we did not attempt to obtain a proximal limit for Lhr by mapping. Instead, we searched preliminary assemblies of the D. simulans genome for candidate genes on the basis of similarities to Hmr, namely higher-than-average divergence between D. melanogaster and D. simulans and a possible role in DNA or chromatin binding. Among ∼37 genes, we first examined CG18468 because it contains a boundary element–associated factor 32/Su(var)3-7/Stonewall (BESS) domain. The BESS domain is found in 21 Drosophila proteins, often associated with MADF domains, and mediates protein-protein interactions (18). Hmr is predicted to encode a protein with two MADF domains (14), and we detected a putative BESS domain (fig. S1), suggesting a possible functional relationship between CG18468 and Hmr.

Most predicted D. simulans proteins are >90% identical to their D. melanogaster orthologs. In contrast, CG18468 has only ∼80% identity, caused by amino acid divergence and by a16–amino acid insertion in D. simulans (Fig. 1C and fig. S2). The estimated average divergence of CG18468 is similar to Hmr (14), measuring 0.078 at nonsynonymous sites (KA) and 0.106 at synonymous sites (KS). This KA value, but not the KS value, is substantially higher than the average value between D. melanogaster and D. simulans (19). Again similar to Hmr, CG18468 is highly diverged outside the melanogaster subgroup (fig. S2), and both genes apparently lack orthologs outside of Drosophila.

We discovered that CG18468 is mutated in the Lhr1 rescue strain, which contains an insertion of ∼4 kb in the predicted 5′ untranslated region (UTR) (Fig. 1D) that appears to be a moderately repetitive retrotransposed sequence. This insertion is not found in any of the five strains from which genome sequence is available nor from 11 additional lines we sequenced (19). CG18468 is adjacent to and divergently transcribed from CG6546 (Fig. 1D), so the insertion in the Lhr1 rescue strain could potentially affect the transcription of either or both of these genes. Reverse transcription polymerase chain reaction (RT-PCR) products, derived from RNA from the critical early larval stage (15), showed that both genes are transcribed in a control strain but that CG18468 transcription is strongly reduced in the Lhr1 rescue strain (Fig. 1E). These data suggest that the Lhr1 phenotype is caused by reduced expression of CG18468.

We cloned the wild-type CG18468 gene from D. simulans and transformed D. melanogaster with P element vectors containing a D. simulans CG18468 cDNA under the control of Saccharomyces cerevisiae Upstream Activity Sequences (UAS), henceforth referred to as UAS-Dsim/Lhr. Expression was induced from a second, independently segregating transgene expressing the S. cerevisiae transcriptional activator GAL4 (Fig. 2). Control crosses using two different GAL4-expressing transgenes suggested that activation of UAS-Dsim/Lhr does not cause lethality in D. melanogaster, as evidenced by the similar numbers of progeny inheriting the GAL4 driver (“red eyed” in Table 1) compared to those that did not (“orange eyed” and “white eyed”) (Table 1 and table S1).

Fig. 2.

Complementation crosses to test for suppression of hybrid male rescue. Female parents are heterozygous for both transgenes, each marked with w+ producing intermediate levels of eye pigmentation. The GAL4-containing transformants have darker eye colors and are epistatic to the lighter-colored UAS-containing transformants. The red-eyed class is therefore potentially composed of two distinct genotypes.

Table 1.

Number of offspring recovered from complementation tests of hybrid rescue mutations by UAS-Dsim/Lhr expression. Full parental genotypes and female progeny are shown in table S1, crosses 9 to 12. UAS indicates D. simulans Lhr under yeast UAS transcriptional control; GAL4 indicates yeast GAL4 protein driven by the Actin5C promoter. In the absence of viability effects, the ratio of red-eyed:orange-eyed:white-eyed males will be 2:1:1. Deviations from this ratio were tested by χ2 tests. Results for the control cross and the cross with Df(1)Hmr- were not significantly different from this ratio (P > 0.05). Results for the crosses with Lhr1 and Hmr1 were significantly different from this ratio (P < 0.001).

ProgenyD. melanogaster control Lhr1 Df(1)Hmr- Hmr1
Red-eyed male UAS/+;GAL4/+ and +/+;GAL4/+ 485 89 169 22
Orange-eyed male UAS/+;+/+ 214 9 82 7
White-eyed male +/+;+/+ 262 94 95 33

This result demonstrated that we could introduce both UAS-Dsim/Lhr and the GAL4-expressing transgenes into hybrids from the D. melanogaster parent (Fig. 2). In contrast to the intraspecific control cross, when the same D. melanogaster females were crossed to D. simulans Lhr1 males, only half of the expected hybrid males carrying the GAL4-expressing transgene were obtained (Table 1). PCR-based genotyping confirmed our inference that Lhr1-rescued males carrying only the GAL4-expressing transgene are viable, whereas those carrying both transgenes and thus expressing UAS-Dsim/Lhr are lethal (table S1). The reduced viability in some crosses of Lhr1-rescued males containing only the UAS-Dsim/Lhr transgene is likely due to maternal inheritance of the GAL4 protein (table S1).

These results suggest that expression of UAS-Dsim/Lhr complements the Lhr1 hybrid rescue phenotype. We confirmed that UAS-Dsim/Lhr expression is not generally lethal to hybrids compared with D. melanogaster pure species by testing for effects in hybrid males rescued by a mutation in Hmr. We found that both D. melanogaster control males and male hybrids rescued by the null mutation Df(1)Hmr are fully viable when expressing UAS-Dsim/Lhr (Table 1). RT-PCR experiments demonstrated that UAS-Dsim/Lhr was expressed in these crosses (fig. S3). We concluded that UAS-Dsim/Lhr expression specifically complements the Lhr1 mutation in hybrids, that CG18468 is Lhr, and that Lhr is a major-effect hybrid lethality gene.

For Lhr to fit the Dobzhansky-Muller model of functional divergence, D. simulans Lhr, but not D. melanogaster Lhr, must cause hybrid lethality (Fig. 1A). Crosses with three different D. melanogaster Lhr deletions produced essentially only F1 hybrid females, demonstrating that removal of D. melanogaster Lhr does not suppress F1 male lethality (fig. S4 and table S2).

Genetic and molecular analyses have demonstrated that Hmr1 retains partial Hmr activity (14, 15). In contrast to the results shown in Table 1, which used the null allele Df(1)Hmr, we found that rescue of hybrids by the hypomorphic mutation Hmr1 is suppressed by D. simulans Lhr expression (Table 1 and table S1). These data suggest that the lethal effect of D. simulans Lhr requires the presence of D. melanogaster Hmr function. The deleterious effects of a D. melanogaster Hmr+ duplication are suppressed by Lhr1 (15, 16), results that suggest, based on our characterization of the Lhr1 mutation, that D. melanogaster Hmr requires a functional D. simulans Lhr to cause lethality. These reciprocal genetic interactions are consistent with the model of Hmr and Lhr forming a Dobzhansky-Muller pair of interacting genes (Fig. 1A).

Functional divergence between species led us to examine the evolutionary forces driving the sequence divergence of Lhr. The high KA/KS value of 0.731 between D. melanogaster and D. simulans is consistent with either positive selection or relaxed selective constraint. We sequenced multiple alleles of Lhr from D. melanogaster and D. simulans and performed a McDonald-Kreitman test (19). The results of this test rejected the null hypothesis that these genes are evolving neutrally (Fisher's Exact Test, P = 0.011) and suggested that there is an excess of nonsynonymous fixations between the species (table S3). Phylogenetic analyses further suggest that the KA/KS ratio has increased on branches leading to D. melanogaster and its sibling species (fig. S5).

The McDonald-Kreitman and KA/KS tests only consider alignable regions of the Lhr coding region. Lhr from D. simulans and its sister species D. mauritiana and D. sechellia each contain a 16–amino acid insertion, interrupting a potential leucine zipper domain, relative to D. melanogaster and outgroup species (Fig. 1C and fig. S2). Notably, we found that this insertion is precisely deleted in a second D. simulans stock named Lhr2, which also produces viable F1 hybrid males (20). Although Lhr2 contains additional amino acid substitutions relative to Lhr+ alleles, its hybrid rescue phenotype suggests that the functional divergence of D. simulans Lhr may be caused by the 16–amino acid insertion.

Heterochromatin Protein 1 (HP1) is a chromodomain-containing protein that localizes to heterochromatic regions of chromosomes and is required to maintain heterochromatic states (21). LHR was previously identified (as CG18468) as interacting with HP1 in a yeast two-hybrid assay (22). We confirmed this interaction (Fig. 3A), and discovered that D. simulans LHR also interacts with D. melanogaster HP1 (Fig. 3B). Because D. simulans HP1 is nearly identical to D. melanogaster HP1 (fig. S6) we hypothesize that D. simulans LHR also binds to D. simulans HP1. This apparent conservation of HP1-binding function of LHR suggests that the intraspecific function of Lhr is conserved between D. melanogaster and D. simulans, in contrast to the interspecific function for hybrid lethality, which we have shown is specific only to D. simulans Lhr.

Fig. 3.

LHR interacts and colocalizes with HP1. (A) D. melanogaster LHR interacts with D. melanogaster HP1. Yeast two-hybrid interactions were detected by activation of HIS3 and growth on media lacking histidine; loading controls [complete media (CM) -Leu -Trp] contain histidine. (B) D. simulans LHR interacts with D. melanogaster HP1. (C to E) Colocalization of D. melanogaster YFP::LHR and HP1 on salivary gland polytene chromosomes. Chromosomes from P{UAS-YFP::Lhr}168-3/+;P{GawB}C147/+ third-instar larvae were incubated with primary antibodies to GFP and HP1, which were detected using rhodamine red-X–conjugated (red) and cyanine-conjugated (green) secondary antibodies, respectively. (C) Antibody to GFP to detect YFP::LHR. (D) Antibody to HP1. (E) Merge with 4′,6′-diamidino-2-phenylindole signal (blue) to detect DNA. A predominant colocalization occurs at the chromocenter (long arrow), fourth chromosome (short arrow), and a telomere (arrowhead).

A D. melanogaster LHR–yellow fluorescent protein (YFP) fusion protein accumulated in a small number of foci (usually 1 to 2) in salivary gland nuclei (fig. S7), similar to HP1 (23). In polytene chromosomes, HP1 accumulates predominantly in the chromocenter and along the highly heterochromatic fourth chromosome as well as at telomeres and a number of bands along the euchromatic arms (24). LHR-YFP has a similar pattern and predominantly colocalizes with HP1 (Fig. 3, C to E). We suggest that Lhr may be coevolving with rapidly evolving heterochromatic repetitive DNAs, consistent with the hypothesis that the molecular drive inherent in repetitive DNAs contributes to hybrid incompatibilities and speciation (25, 26).

Hmr and Lhr cause F1 hybrid lethality because they are partially or fully dominant. The large number of HI genes estimated from other studies (27) may be mechanistically distinct because they are recessive and only cause HI when homozygous in F2 hybrids or in interspecific introgressions. However, our results also show that the interaction of Hmr and Lhr alone is insufficient to recapitulate hybrid lethality, because control crosses showed that expression of D. simulans Lhr does not cause lethality in a D. melanogaster pure-species background (Table 1 and table S1). Pontecorvo suggested that an interaction among the D. melanogaster X (which contains Hmr), D. simulans chromosome II (which contains Lhr), and D. simulans chromosome III causes hybrid lethality (28). Hybrid lethality may therefore be enhanced by a multilocus interaction involving additional genes. Alternatively, Hmr and Lhr may be the only major-effect genes, but their lethal interaction requires a hybrid genetic background. We suggest that altered chromosome morphology and chromatin structure in hybrids due to the cumulative effects of species-specific differences in satellites, transposable elements, and other repetitive DNAs cause this hybrid genetic background effect.

Supporting Online Material

Materials and Methods

Figs. S1 to S7

Tables S1 to S3


References and Notes

Stay Connected to Science

Navigate This Article