Neighboring genes for DNA-binding proteins rescue male sterility in Drosophila hybrids

HMS1 refers to a region in the right arm of chromosome 3 of Drosophila mauritiana that is associated with hybrid male sterility between D. mauritiana and its sibling species D. simulans. When the HMS1 region from D. mauritiana (HMS1[mau]) is homozygous in an otherwise D. simulans genetic background, the males show a dramatic reduction in fertility. HMS1 is one of about 20 HMS factors in chromosome 3 originally mapped as quantitative-trait loci affecting male fertility (29). Most HMS factors have relatively small effects so that generally at least two or three different factors are required to produce complete sterility (29). In contrast, males homozygous for HMS1 alone are usually quasisterile (29), although the effect of HMS1 is strongly dependent on genetic background (30).

Originally localized to a region of 1.26 Mb between the genes Regena (Rga) and Antennapedia (Antp) in chromosome 3R within the overlap between P-element introgressions P32.110 and P45.6 (Fig. 1A) (29), the HMS1 interval was significantly narrowed by four successive rounds of recombination to a region of 9.2 kb (Fig. 1B) corresponding to band 84A1 in the polytene salivary-gland chromosomes (30). Within this region are four genes, two of which exhibit significant coding and noncoding differences between the species, making them candidate genes for hybrid male sterility (30). The candidate genes are Taf1 (CG17603) and agt (CG1303) (Fig. 1B), both of which encode DNA-binding proteins The genetic mapping and initial analysis left it unclear, however, which one of these genes, or possibly both, contributed to the male-sterility phenotype.

HMS1 male sterility is manifested in a D. simulans genetic background (strain SimB) into which the HMS1 region of D. mauritiana (strain Mau12) has been introgressed. For simplicity, we will refer to the genotype of the SimB introgression genotype as HMS1[mau]/HMS1[mau] (see refs. 29 and 30 for details of the SimB and Mau12 strains and the introgression methods). Homozygous HMS1[mau]/HMS1[mau] males are almost completely sterile: 97.9% of tested males are sterile (n = 109), and among those with progeny the average number is 1.5 ± 0.71 (Fig. 1C and Table S1). The sterility is restored in heterozygous HMS1[mau]/HMS1[sim] males, among which all are fertile (n = 183) with an average offspring number of 82.6 ± 52.1 (Fig. 1C and Table S1).

To ascertain the genetic basis of HMS1 sterility, we carried out germ-line transformation using various constructs inserted into the piggyBac vector (31). The strategy was to create HMS1[mau]/HMS1[mau] male-sterile genotypes carrying an extra, transgenic, piggyBac copy of all or part of the HMS1 region from SimB (HMS1[sim]) to look for evidence of complementation, which would be manifested as partial or complete recovery of male fertility. One problem with this strategy is that HMS1[mau]/HMS1[mau] is male sterile and so cannot be transformed directly. Instead, we transformed SimB with piggyBac constructs containing all or part of HMS1[sim] (symbolized pB::HMS1[sim]) (Fig. S1) to generate transgenic HMS1[sim]/HMS1[sim]; pB::HMS1[sim] animals. In each case, the location of the transgene in the genome was determined by the inverse PCR (32). The pB-HMS1[sim] transgene was then transmitted among genotypes by a stepwise crossing design to create the required HMS1[mau]/HMS1[mau]; pB::HMS1[sim] genotypes to assay for fertility (Fig. S2). The genetic markers for the HMS1[mau] introgression and also the pB::HMS1[sim] constructs all affect eye color; however, the copy number effects on phenotype usually allow the genotypes to be distinguished (Fig. S3). All assayed males, both fertile and sterile, were unambiguously genotyped postmating by means of PCR, and in the case of fertile males the reliability of the molecular genotyping was further verified by observing segregation of the eye color phenotypes among the progeny.

Among the piggyBac constructs created to dissect the genetic basis of HMS1 were agt[sim] and Taf1[sim] (Fig. S4). The former includes the entirety of the agt gene, and the latter, the entirety of Taf1, both derived from the SimB strain of D. simulans. We tested three independent insertions of pB::agt[sim] and two of pB::Taf1[sim] in a genetic background of HMS1[mau]/HMS1[mau] to detect possible position effects of the insertions. Although introgressed HMS1[mau]/HMS1[mau] males are virtually all sterile (Fig. 1C), all three pB::agt[sim] and both pB::Taf1[sim] transgenes restore fertility to some extent (Fig. 2). Among transgenic males that are fertile, the level of fertility is high and not significantly different overall among all pB::agt[sim] and pB::Taf1[sim] transformants (t test on progeny number: t = 0.369, df = 152, P = 0.713). The fertility rescue is only partial, however, as indicated by the relatively high proportions of sterile males among all transgenic lines (Fig. 2). The percentage of sterile males averages 42 ± 4% among pB::agt[sim] transformants and 54 ± 5% among pB::Taf1[sim] transformants (Fig. 2). For agt[sim], the number of progeny does not differ significantly for the three genomic transgene insertions (one-way ANOVA: F = 2.65, df = 2, P = 0.075) and the fractions fertile are also homogeneous [P = 0.172 (X² = 3.523, df = 2, n = 185)]. For Taf1[sim], the number of progeny differs somewhat from one insertion site to the next (t test on progeny number: t = –3.030, df = 43, P = 0.004) as indicated by the asterisk in Fig. 2, but importantly the fraction fertile remains homogenous [χ² test on number fertile: P = 0.063 (X² = 3.453, df = 1, n = 94)], which attests to the limited impact of positional effects at the sites of insertion studied. From the standpoint of rescuing complete sterility, agt[sim] is thus not significantly different from Taf1[sim] [P = 0.079 (X² = 3.076, df = 1, n = 279)]. In both cases, the partial rescue is specific to the agt[sim] or Taf1[sim] transgene, because control males carrying an empty piggyBac vector with no insert fail completely to rescue HMS1 sterility (Table S2).

To further investigate the mechanism by which fertility is restored in the transgenic lines, we assayed whether decreased expression levels would account for the sterility at either candidate locus. We first quantified mRNA levels in male reproductive tissue of both agt and Taf1 loci in fertile (HMS1[mau]/HMS1[sim]) and sterile (HMS1[mau]/HMS1[mau]) introgression males, which are isogenic except for a small region including HMS1 (SI Materials and Methods and Fig. S5A). These males yield baseline levels of agt and Taf1 expression for males bearing either one or two sim alleles of each gene. In each of two genetic backgrounds, SimB and w501, our quantitative PCR results indicate no significant difference in mRNA level of either Taf1 or agt between fertile (HMS1[mau]/HMS1[sim]) and sterile (HMS1[mau]/HMS1[mau]) males (Fig. S5B and Table S3). This finding supports the inference that the hybrid male sterility is not mediated by differences in gene expression of either agt or Taf1.

A previous analysis had suggested that an amino acid replacement in agt might contribute to hybrid male sterility (30). In D. simulans, amino acid position 121 is occupied by aspartic acid, whereas in D. mauritiana position 121 is an asparagine. This D121N difference is the only fixed amino acid difference then known between the species (30), hence making it a strong candidate for an incompatibility. To test this hypothesis, we created an agt[simD121N] sequence (Fig. S4) that is identical to agt[sim] except for a G-to-A nucleotide change in codon 121 resulting in D121N. A transgenic agt[simD121N] piggyBac transformant was obtained, crossed into an agt[mau]/agt[mau] genetic background (Fig. S2), and tested for male fertility. The results are shown in Fig. 3. If D121N were solely responsible for the agt[mau] incompatibility, then transgenic pB::agt[simD121N] males should be sterile. However, 27% of the males are fertile (Fig. 3), and their average number of offspring does not differ significantly from that observed for pB::agt[sim] (t = 0.184, df =121, P = 0.89) (Fig. 2). Nevertheless the fertility rescue of pB::agt[simD121N] is significantly less than that of pB::agt[sim], with 73 ± 7% sterile in the former and 42 ± 4% in the latter [P = 0.014 (X² = 6.09, df = 1, n = 225)]. These results apply to a single genomic insertion; however, the observation that position effects for degree of rescue are not observed for distinct genomic insertion sites for agt[sim] (Fig. 2), together with the absence of transcriptional difference in agt between HMS1[mau]/HMS1[sim] and HMS1[mau]/HMS1[mau] males, suggest that the reduced fraction of fertile males in pB::agt[simD121N] may be heavily attributable to the coding mutation.

The findings with agt[simD121N] suggest that the agt[mau] incompatibility is due only in part to D121N and in part to other nucleotide differences in agt between the SimB and Mau12 strains. A total of 16 nucleotide differences are positioned across the gene (Table 1), or 132 differences if those in the intergenic regions are also considered; hence there is considerable opportunity for interactions among sites. To test for such interactions, we created an agt[simP_mauCDS] construct that includes the promoter region from D. simulans and the ORF from D. mauritiana (agt contains no introns), and the reciprocal agt[mauP_simCDS] construct (Fig. S4). Two independent transgenic piggyBac lines were obtained and tested from each construct (Table S1), yielding the grouped results shown in Fig. 3. Both types of construct show levels of complete fertility rescue comparable to those of agt[sim] [P = 0.257 (X² = 5.309, df = 4, n = 335)] and, among the fertile males, only agt[simP_mauCDS] sire more progeny (one-way ANOVA: F = 2.998, df = 5, P = 0.012, Tukey post hoc test). These results support a model in which the incompatibility in agt[mau] results from epistatic interactions of nucleotide sites within or near the agt gene, interactions that the agt[simP_mauCDS] and agt[mauP_simCDS] constructs disrupt.

Differences between SimB and Mau12 in the Taf1 region give even more scope for complex interactions than in agt. Taf1 is a much larger gene than agt (9,412 vs. 758 nt). Across the Taf1 gene region, there are 127 nucleotide differences between the SimB and Mau12 strains (Table 1), and 164 differences if the 3′-untranslated and intergenic regions are also taken into account. To test for interactions among these sites, we created two chimeric constructs designated Taf1[simPEx10_mauEx10end] and Taf1[mauPEx10_simEx10end] (Fig. S4). The first construct joins a DNA fragment containing the SimB Taf1 promoter through an AvrII restriction site in exon 10 with another DNA fragment containing Mau12 Taf1 coding sequence from the AvrII restriction site through to the 3′ end of the gene. The AvrII site was chosen as the joining site because it is near where recombination analysis had suggested the approximate 5′ upstream boundary of HMS1 was located (30). The second construct is the reciprocal, joining the 5′ end of Mau12 Taf1 with the 3′ end of simB Taf1, again with the joining site at the AvrII site in exon 10.

Results with the reciprocal Taf1 transformants are shown in Fig. 3. The two constructs do not differ significantly in either the proportion of sterile males or the fecundity of fertile males, nor do they differ significantly from the Taf1[sim] insertions in Fig. 2 (χ² on number fertile: P = 0.913, X² = 0.182, df = 2, n = 189) or fecundity from Taf1[sim] insertion 5 (one-way ANOVA on fecundity: F = 0.612, df = 2, P = 0.545). Considering the limited contribution of position effects at the sites of insertion studied in terms of degree of rescue (although not globally for fecundity; i.e., lower progeny numbers in Taf1[sim] insertion 4), these results suggest that the Taf1[mau] incompatibility is due to interactions between sites in or near the gene. In the case of Taf1, between the strains SimB and Mau12, there are 5 nonsynonymous differences and 44 synonymous differences in the coding sequence preceding the AvrII site, and 9 nonsynonymous differences and 23 synonymous differences following the AvrII site; hence, opportunities for interactions among sites are plentiful.

To look for signals of positive selection, we sequenced the ORF of 32 alleles of agt and exons 10−14 of 15 alleles of Taf1 from genomic DNA extracted from individual flies from 17 diverse D. simulans populations from South, Central, and North Africa, Europe, Japan, Australia, and North America as well as 16 independent acquisitions of D. mauritiana (an island endemic), from Mauritius (Table S4). Statistical tests for selection included comparing the mean number of pairwise differences with the number of segregating sites [Tajima’s D (33)], as well as comparing old and recent mutations according to where they occur in the gene genealogies using mutation rates either estimated from nucleotide diversity [Fu and Li’s D (34)] or else estimated from nucleotide polymorphism [Fu and Li’s F (34)]. We also analyzed nonsynonymous versus synonymous polymorphism and divergence [McDonald–Kreitman test (35)], as well as site-specific models of selection using phylogenetic analysis by maximum likelihood (PAML) (36). None of these tests reached statistical significance, even when uncorrected for multiple testing (Table S4).

The fly food consists of agar (9.29 g; Affymetrix), torula yeast (32.35 g; Affymetrix), cornmeal (61.17 g; MP Biomedicals), dextrose (64.70 g; Affymetrix), molasses (47.05 mL; Genesee Scientific), 10% (vol/vol) tegosept solution in ethyl alcohol (9.41 mL; Genesee Scientific), propionic acid (5.88 mL; VWR), and distilled water (to 1 L).

For agt[sim], a 1.36-kb genomic region (Fig. S4A) was amplified using SimB genomic DNA as template, which comprises 681 bp upstream flanking DNA sequence including: (i) the predicted transcription start at position −270, (ii) the 579-bp intronless agt[sim] ORF, and (iii) 95 bp of the downstream intergenic region between agt and spase. This PCR fragment was inserted into a MW-FPNS piggyBac vector [pBac(3xP3-EGFPafm)::MCS::(pW8 miniwhite)] obtained from Dave Miller in the laboratory of Thomas Kauffman, Indiana University, Bloomington, IN. The intergenic flanking region between agt and spase totals 145 bp and is identical in sequence between Mau12 and SimB. A PCR QuikChange site-targeted mutagenesis (Stratagene) and modified primers were used to generate the agt[simD121N] point-mutation construct, following the manufacturer’s procedure. The chimeric constructs agt[simP_mauCDS] and agt[mauP_simCDS] were generated by blunt-end ligation between the agt[SimB] coding and UTR region and the agt[Mau12] promoter region, and conversely, by fusing the agt[Mau12] coding and UTR regions with the agt[SimB] promoter region (Fig. S4A). Specifically, oligonucleotide primers encompassing AvrII and StuI or StuI and FseI restriction sites were used to generate PCR fragments corresponding to the promoter or coding sequence (+95 bp downstream sequence) using SimB or Mau12 genomic DNA as template, respectively, followed by ligation in PCR TOPO 2.1 (Invitrogen) and transformation in Escherichia coli TOPO 10 competent cells. The internal StuI restriction site (AGGGCCT) was engineered in the oligonucleotide primer via a single-nucleotide change from the original SimB/Mau agt DNA sequence that is located 5 bp upstream from the agt 5′-UTR region. Individual plasmid DNA constructs were verified by double-restriction digestion (AvrII/StuI or StuI/FseI) and sequenced on both strands with ABI 3730xl automated capillary sequencing instruments and ABI PRISM BigDye chemistry to ensure that no mutations were introduced, before subcloning into the AvrII and FseI cloning sites of the linearized pB vector. Internal fusion between coding sequence and promoter regions was achieved in this ligation step by blunt-end ligation at the StuI restriction site. All final constructions encompassed the same gene region, only with variable portions of the SimB or Mau12 alleles of agt.

For Taf1[sim], oligonucleotide primers were designed encompassing the FseI and Sbf1 restriction sites (Fig. S4B) to amplify a 10.5-kb fragment of the SimB allele corresponding to: (i) 1,273 bp of upstream intergenic region comprising the 100-bp-long 5′-UTR and 1,173 bp of the promoter region including the predicted transcription start at position −1,098, (ii) the 8,883-bp region corresponding to the total Taf1[SimB] (16 exons and 15 introns), and (iii) 431 bp downstream sequence comprising the 3′-UTR and the entire intergenic region between Taf1 and gene CG1307. For the Taf1 chimeras, Taf1[simPEx10_mauEx10end] and Taf1[mauPEx10_simEx10end] (Fig. S4B), DNA fragments corresponding to each part were amplified using TaKaRa LA Taq (Clontech) from SimB or Mau12 genomic DNA templates, respectively, and religated at an internal AvrII site adjacent to a previously mapped molecular marker marking the HMS1 boundary (Fig. S4B) in a single subcloning reaction in the linearized pB vector. First, PCR fragments were ligated in TOPO XL PCR (Invitrogen) and transformed in E. coli TOPO 10 cells (Invitrogen), before plasmid purification and sequencing. Upon plasmid restriction digestion, the 5.9-kb Taf1[mau] or Taf1[sim] DNA fragment (including the promoter region and partial coding sequence) and delimited by FseI and AvrII was respectively fused at AvrII to the 4.6-kb Taf1[SimB] or Taf1[Mau12] fragment encompassing the coding sequence end and 3′-UTR. Each Taf1 chimera was religated in a single directional cloning step at the FseI and Sbf1 restriction sites of the linearized pB vector. The empty MW-FPNS piggyBac vector [pBac(3xP3-EGFPafm)::MCS::(pW8 miniwhite)] alone was used for germ-line transformation generating control lines. The MW-FPNS piggyBac vector, and final agt and Taf1 constructs were sequenced on both strands with ABI 3730xl automated capillary sequencing instruments and ABI PRISM BigDye chemistry to ensure that no mutations had been introduced before germ-line transformation.

An inverse PCR protocol adapted from the Berkeley Drosophila genome project (32) was used to determine the 5′- and 3′-end sequences flanking pB insertions. High-quality genomic DNA was extracted from five w⁺ males of each line and digested using Sau3A (5′ end) or HinP1 (3′ end), before overnight self-ligation using T4 DNA ligase (New England Biolabs). Ligation products were subjected to two successive rounds of PCR using the following cycling conditions: 94 °C for 5 min, 35 cycles of 94 °C for 30 s, 58 °C for 1 min, 72 °C for 2 min 30 s, and a final step at 72 °C for 10 min using primers designed based on the MW-FPNS plasmid on the boundaries of the left and right insertion sites (all primer sequences used in this study are available upon request). Strong unique bands were detected by agarose gel electrophoresis in all PCRs, and the PCR products were purified using Exo and SAP enzymes (Fermentas) before sequencing. DNA sequences were analyzed in Geneious, version 7.1.7, and searched against the D. simulans and D. melanogaster reference genomes using the flybase BLAST tool, altogether confirming that each transgenic strain has a single-copy gene inserted in one genomic location; locations are provided in Table S1.

Each P[w⁺]-element is a semidominant marker bearing a copy of the miniwhite w⁺ gene. Eye color is sensitive to the copy number and position of the P[w⁺]-insert, which allows to distinguish 1P or 2P heterozygotes from 2P and 3P homozygotes phenotypes. P45.6 males (HMS1[mau]/HMS1[sim]) typically have dark red eyes, whereas P32.110 males (HMS1[mau]/HMS1[sim]) have bright orange eyes. Corresponding virgin females have lighter eye coloration, orange and yellow, respectively; however, the eye colors darken with an individual’s age. Both P45.6 and P32.110 heterozygous lines carry a Drosophila mauritiana introgression that covers HMS1 and confers full sterility to males when homozygous. The homozygous 2P (P45.6/P32.110) flies, with two copies of P[w⁺]-inserts have a clearly different, bright red eye color. 3P rescue homozygotes are not distinguishable from homozygotes 2P, requiring progeny phenotyping and molecular genotyping of all P[w⁺]-elements.

Genomic DNA was extracted in 50 μL of fresh squishing buffer (10 mM Tris, pH 8.2, 1 mM EDTA, 25 mM NaCl, and 200 μg/mL Proteinase K), and the extract was incubated for 1 h at 37 °C and a 2-min inactivation at 95 °C. Single genotyping PCRs were run using 1 μL of male DNA as template and sets of specific primers targeting pB, P45, and P32 P-elements under the following cycling conditions: 94 °C for 4 min, 35 cycles of 94 °C for 30 s, 60 °C for 30 s, 72 °C for 1 min 30 s, and 72 °C for 10 min, before visualization of amplicon presence/absence on 1% agarose gels. HMS1[mau]/HMS1[mau]; pB::HMS1[sim] or 3P males were also genotyped to ensure no recombination at P45.6 or P32.110 had taken place. Whether pB::HMS1[sim] alleles were inserted on the second or third chromosome, males bearing the pB and P32.110 elements were crossed to P45 females. Accordingly, 3P fertile males were genotyped to ensure no recombination at P45. 3P fertile male progeny bearing pB on chromosome 3 were genotyped at P32.110 to ensure the absence of recombination in intermediate 2P females (P32.110 + pB or HMS1[mau]/HMS1[sim]; pB::HMS1[sim]).

For fertility assays in single-male mating tests, bright red-eye male progeny corresponding to HMS1[mau]/HMS1[mau]; pB::HMS1[sim] and also homozygous 2P genotypes obtained for crosses on chromosome 2 (HMS1[mau]/HMS1[mau]) were selected and crossed at 18 °C with three 2- to 3-d-old virgin females of the D. simulans tester w;e line, where the e (ebony) recessive marker is used to detect potential non virgins. Similarly, HMS1[mau]/HMS1[sim] (P45.6) dark red-eyed male progeny were crossed to record the progeny distribution of “fertile” control heterozygous. On the 10th day, females were cleared, and dental cotton was inserted in the medium to maximize pupation. Only sterile vials for which the male and at least one female were still alive at the time of collection were included in the analysis to prevent inclusion of males who might have died before mating. A test male was considered sterile if he produced zero progeny, and fertile if he produced one progeny or more. Progeny were counted on the 20th and 25th day. The male parent genotype was determined based both on the segregation of P-element eye coloration in the progeny, and based on PCR genotyping at all three P-elements. The latter is particularly critical to distinguish homozygous 2P from 3P parent males because both bear several genetic w⁺ elements in addition to the dominant red P45.6, and therefore display similar bright vermillion eye coloration. Results for the tested transgenic lines carrying pB::agt[simP_mauCDS] (insertions 7 and 8) and pB::agt[mauP_simCDS] (insertions 9 + 10) were grouped for clarity as presented in Fig. 3 after confirming that the degree and level of fertility rescue were similar between lines for each transgene, respectively [P = 0.236 (X² = 1.406, df = 1, n = 36); P = 0.889 (X² = 0.019, df = 1, n = 114)].

The reproductive organs of hybrid males that did not sire offspring on the 10th day were dissected in Ringer’s isotonic solution (111 mM NaCl, 5.6 mM KCl, 2.2 mM CaCl₂, and 2.4 mM NaHCO₃), and gently squashed preparations were visualized under phase contrast microscopy to assess the presence of elongated spermatids, individualized spermatocytes, and motile sperm during a 60-s observation period. Because sperm motility is very difficult to measure quantitatively, we did not define a sperm motility index based on the number of motile sperm, and characterized males as bearing motile sperm whenever one or more sperm were observed moving. Reproductive organs from an arbitrary subset of fertile males were observed and always showed evidence of abundant motile sperm. None of the examined sterile individuals showed evidence for smaller, absent, or abnormal testes (e.g., hybrid dysgenesis in the form of gonadal abnormalities), and elongated spermatocytes were observed in all cases. All pB lines considered, sperm motility was assessed in a subset of 35 2P sterile homozygous males and 256 sterile 3P males.

2P males were generated by crossing virgin 2-d-old P32.110 males (Sim B w;nt;III/HMS1[mau]) or P32.75 males (Sim B w;nt;III/HMS1[sim]) with 4-d-old virgin females from the P45.6 tester line (SimB w;nt;III/HMS1[mau]). In the w;nt;III/HMS1[sim] introgression line, the D. mauritiana portion of the chromosome does not span the HMS1 region (30). Two genetic backgrounds were used for the experiment, in which HMS1 was introgressed in the SimB or w501 strains.

Individual 2P male fertility phenotypes were assessed by crossing each male to three 4-d-old virgin w;e females following the procedure described under the fertility assay section. On the seventh day, 2P males were anesthetized and testes were dissected in sterile Ringer’s solution, immediately transferred in 20 μL of TRIzol and snap-frozen in TRIzol. Testes tissues in TRIzol were subsequently disrupted on dry ice using a sterile pestle rotor before storage at –80 °C. The corresponding individual male carcasses were used for DNA extraction and genotyping at P45.6 to ensure absence of recombination in females, before tissue pooling and RNA extraction.

For each 2P genotype category, RNA was extracted using the TRIzol procedure (Invitrogen) before overnight precipitation at –20 °C and with NaAc/ethanol purification and treatment with DNase I (New England Biolabs) from biological replicates representing a population of 10 testes samples for which males displayed fertile or sterile phenotypes. cDNAs were synthesized using the SuperScript II Reverse Transcriptase (Invitrogen), an Oligo d(T) 23 VN (New England Biolabs), and the RNase Out (Invitrogen) following procedures provided by the manufacturer.

The 25-μL quantitative PCRs (qPCRs) were run on an Applied Biosystems 7900HT using SYBR Green PC master mix (Applied Biosystems) and 20 ng of cDNA, 200 nM GSPs, and 50 nM Rox dye with the following cycling conditions: 50 °C for 2 min, 95 °C for 2 min, followed by 40 cycles of 95 °C for 30 s, 60 °C for 1 min, and 72 °C for 1 min. The dissociation curve analysis was as follows: 95 °C for 1 min, 60 °C for 30 s (and a gradual heating gradient to 95 °C at 0.01 °C/s. Target genes primer sets were designed in AlleleID software (PREMIER Biosoft International) as follows: Taf1: 5′-CAACGACGGCAAGGAATA-3′ and 5′-TGCGAACTGCTTGATGAA-3′ (amplicon size: 112 bp); and agt: 5′-CCAATTGCGACTTGGTCTTT-3′ and 3′-CCAGAACACCCAAGTTCGTT-3′ (amplicon size: 126 bp). Quantification reactions for each cDNA were run in duplicates in two separate qPCRs (n = 4 for genotypes with one biological replicate of 10 testes samples, or n = 8 for genotypes with two biological replicates of 10 testes samples). Relative expression was calculated using ∆Ct values after robust normalization against three stable endogenous control genes (RpL32, mRpL15, and Alpha-tubulin84D) targeting the following gene regions: RpL32 (Dsim/GD17388): 5′-ATCGGTTACGGATCGAACAA-3′ and 5′-GACGATCTCCTTGCGCTTCT-3′ (amplicon size: 165 bp); mRpL15 (Dsim/GD12186): 5′-GGTCATAGCGGCCATAGAGA-5′ and 5′-CAGCATGCGGCTAGGAATAG-3′ (amplicon size: 130 bp); and Alpha-tubulin84D (Dsim/GD19940): 5′-TGTCGCGTGTGAAACACTTC-3′ and 5′-AGCAGGCGTTTCCAATCTG-3′ (amplicon size: 96 bp). All primer pair reaction efficiencies were initially validated with standard curves calculated from running preexperimental qPCRs with serial dilutions of template cDNA. We initially tested several primer pairs and additional reference genes including that encoding actin, and the stability of the reference gene expression patterns chosen for our analysis was assessed postexperimentally by statistical analyses of variation using the Bestkeeper software program for reference gene selection (www.gene-quantification.com/bestkeeper.html).

Our findings demonstrate the genetic complexity of a small region of 9.2 kb designated HMS1 that is associated with hybrid male sterility when introgressed from D. mauritiana strain Mau12 into D. simulans strain SimB. The region contains two likely candidate genes for hybrid male sterility, agt and Taf1, and we find that both genes contribute to the incompatibility. In both cases, transgenes carrying a compatible allele rescue the fertility of otherwise sterile HMS1 males. The rescue phenotype has the unusual property that, although only about one-half of the transgenic males regain fertility, those that are fertile produce on average as many progeny as HMS1[mau]/HMS1[sim] heterozygous males (Figs. 1 and 2). We note that position effects cannot be fully excluded from having contributed to some of the results, even if the few replicates noted do not show statistically significant differences after multiple comparisons. In agt, a transgene carrying an amino acid replacement that had previously been identified as a candidate incompatibility factor (30) rescued HMS1 sterility less than controls; however, the full effect of agt appears to involve an interaction among two or more nucleotide sites along the gene (Fig. 3). Likewise, the Taf1 incompatibility seems to result from interactions among nucleotide sites, as chimeric constructs do ameliorate the HMS1 sterile phenotype (Fig. 3). Sequence analysis of allelic diversity among naturally occurring agt and Taf1 alleles gives no evidence of positive selection.

Our findings with HMS1 challenge a number of hypotheses regarding hybrid incompatibilities while they support others. The first is that hybrid incompatibilities are due to single genes of large effect. Although a number of large-effect genes are known that result in partial rescue of hybrid sterility or lethality (13, 16, 37, 38), HMS1 does not represent a single incompatibility gene but a nearby pair of incomplete incompatibilities that each afford only partial rescue. The partial incompatibility due to each gene is itself attributable to epistatic interactions among sites in or near the gene, as expected of ongoing evolution through slightly deleterious mutations compensated by other mutations elsewhere in the gene (39). The number of interacting sites could be as few as two, but experimental evidence implies that higher-order interactions are more likely (40, 41). Considering the within-gene interactions and the effects of environment and genetic background (30), hybrid male sterility in this case behaves like a classical quantitative trait affected by multiple genes with relatively small effects (42) as well as environmental factors. Although a number of genes associated with hybrid incompatibility show evidence of positive selection (3, 10, 43, 44), this is not the case for either agt or Taf1; however, our findings do not rule out strong selection in the ancient past or ongoing weak selection because these processes would not necessarily leave statistically detectable signals of selection on contemporary polymorphisms.

Fig. 4 illustrates a threshold model of HMS1 sterility treated as a quantitative trait subject to environmental effects. The horizontal axis represents a hypothetical fertility potential, and the vertical axis is proportional to the fraction of males in a genetically homogeneous population. Any male with a fertility potential less than some threshold (dashed vertical line) is completely sterile. The mau/mau curve on the Left corresponds to the recessive hybrid male-sterile HMS1[mau]/HMS1[mau] carrying two copies of the HMS1 introgression, whereas the sim/mau curve depicts the fertility potential of SimB male individuals carrying one copy of the HMS1 introgression (HMS1[sim]/HMS1[mau]). Because the incompatibility is recessive, sim represents the dominant allele conferring male fertility. Note that a small proportion of mau/mau males are fertile, and likewise a small proportion of sim/mau males are sterile. The sim + mau/mau curve in the Center represents the effects of the rescue transgenes pB::agt[sim] and pB::Taf1[sim] in Fig. 2 and the chimeric genes in Fig. 3. One copy of the dominant fertility allele is sufficient to restore fecundity in the normal range, on average, in 45% of the transgenic males. This kind of fecundity rescue would be expected if, in the transgenic males, any germ cells that survive the incompatibility proliferate more than they otherwise would, thereby compensating for germ cells that succumb to the incompatibility and restoring male fecundity to high levels. It is worth noting that a substantial proportion of males who fail to rescue fertility produce some motile but dysfunctional sperm (Table S1). The proportion of sterile individuals with motile sperm is relatively constant (50%) regardless of their transgenic genotype.

One hypothesis that the genetic analysis of HMS1 strongly supports is the prominence of internal genomic conflicts in hybrid incompatibilities, especially those conflicts involving genes for DNA- or chromatin-binding proteins (3, 10). In Drosophila, spermatogenesis proceeds through a mitotic phase of stem cell renewal and differentiation followed by meiosis and posttranscriptional spermatid maturation into fully motile, individualized sperm (45). Germ cell development is influenced by genotypic variability as well as microenvironmental perturbations that are ultimately reflected in quantitative differences in the function of mature spermatocytes and spermatozoa within and among individual males.

Several lines of evidence suggest that alternatively spliced isoforms of Taf1 transcripts regulate spermatogenesis in a tissue-specific manner. Taf1 encodes a TAF (TATA-box–binding associated factor) constituting one subunit of transcription factor TFIID broadly required for transcription by RNA polymerase II (46, 47). The product of Taf1 serves as a multifunctional transcriptional regulator operating in the normal cell cycle, the assembly of other TAFs and TBPs (TATA-binding proteins), promoter activities, histone acetylation, and other chromatin modifications that can modulate chromatin structure enabling it to access transcriptionally repressed chromatin (48). In mice, the product of the Taf1 homolog Bdrt associates with hyperacetylated histone H4 and functions in chromatin remodeling following histone hyperacetylation as postmeiotic germ cells mature into fully differentiated sperm (49). In Drosophila, the primary transcript of Taf1 is alternatively spliced into at least four isoforms whose protein products contain AT hooks that directly bind with DNA (50). One of the isoforms (Taf1-2) is enriched in Drosophila melanogaster testes where it accounts for 45% of Taf1 mRNA versus 10% of Taf1 mRNA in adult male whole flies (50). Among other possible functions, the Taf1-2 product can bind promoters of testes-specific genes including sperm-specific dynein intermediate chain (Sdic), heat shock protein 70bc (hsp70), β2-tubulin (β2t), and don juan (dj) (51), and the role of Taf1-2 as transcriptional activator in the testes is likely conserved across Drosophila. At 10 kb, Taf1 is a relatively large gene for Drosophila, and there are about 1.5% nucleotide sequence differences in Taf1 between SimB and Mau12 (Table 1). Several fixed nucleotide differences are located in the gene region around to exon 12a, which is a spliced exon required for isoform Taf1-2. The sequence divergence might therefore affect alternative splicing efficiency (52), promoter-binding affinity, chromatin modification, or other processes that could contribute to hybrid male sterility as a quantitative trait (53, 54). The role of Taf1 in hybrid male sterility is perhaps not surprising in light of the rapid evolution and functional diversification of testes-specific TAFs in Drosophila (55).

The mechanism by which agt contributes to HMS1 sterility is more obscure. The gene is induced by genotoxic stress, and the protein is an alkyl-cysteine S-alkyltransferase that removes alkyl groups from DNA, notably from O⁶-methylguanine (56), a repair process observed in eukaryotes, notably in postmeiotic male germ cells in Drosophila (57). A much smaller gene than Taf1, with an ORF of only 192 codons and no introns, agt is 2.4% different between SimB and Mau12 (Table 1). It is also widely divergent in the species subgroup, with nearly every species having a different amino acid at position 121 (30). The D121N difference between SimB and Mau12 clearly contributes to the hybrid male sterility (Figs. 2 and 3); however, interactions among sites in agt are also implicated (Fig. 3). Although the role of agt in hybrid male sterility in Drosophila is uncertain, in the mouse the DNA repair gene Ercc1 is essential for functional integrity of germ cell DNA and normal spermatogenesis (58), and a histone methyltransferase encoded by Prdm9 is necessary for meiosis (59). This emphasizes the intriguing relationships between gametogenesis and the variety of DNA repair enzymes functioning in parallel pathways in spermatogenesis. Also, it suggests that in eukaryotes dysfunction of genes involved in any DNA repair mechanism in male germ cell stages may influence the degree of fertility (60, 61).

It is unlikely that either Taf1 or agt was involved in the origin of reproductive isolation between D. simulans and D. mauritiana. Recent experimental evidence (62) and theory (63–67) imply that certain allele combinations causing partial reproductive incompatibility can be found segregating in natural populations, which suggests that, on an evolutionary timescale, reproductive isolation can evolve very rapidly from standing genetic variation for deleterious allelic mutations causing partial reproductive isolation, even in sympatric populations (68, 69).

A quick, heuristic calculation also suggests that the incompatibilities of Taf1 and agt are both likely to have evolved more recently than the species divergence time of 250,000 y. Between D. simulans and D. mauritiana, there are an estimated 40 autosomal genes that contribute to hybrid male sterility (29); however, the X chromosome, which evolves incompatibilities faster than the autosomes, has about 100 such factors (24). Each factor reduces male fertility by about 10%, and, taken together, they constitute 15 “HMS equivalents,” where each HMS equivalent includes a sufficient number of incompatibilities to result in hybrid male sterility (24). In effect, hybrid males of D. simulans and D. mauritiana are sterile 15 times over. The factors do not act independently, however: epistasis among incompatibility genes is pervasive (7, 29, 70).

If hybrid incompatibility factors evolve at a rate that is linear in time, and those in the X chromosome evolve at a rate 2.5 times that of those in the autosomes (24), then each HMS equivalent would entail, on average, 9.3 total incompatibility factors of which 2.7 would be autosomal. Because 40 autosomal factors have evolved in 250,000 y, then with a linear increase in number through time, 2.7 autosomal factors would be expected to have evolved in 17,000 y. This number is less than 7% of the total time since species divergence, so any incompatibility factor between D. simulans and D. mauritiana is very likely to have evolved since their divergence. On the other hand, population genetics theory suggests that the accumulation of incompatibility factors might be exponential rather than linear (9). In the exponential case, the situation for early evolution is a bit more optimistic, but not by much. With exponential increase from 0 to 40 autosomal factors in 250,000 y, 2.7 factors would be expected to accumulate in 27,000 y, or about 10% of the total time since divergence. Both estimates imply that most hybrid incompatibility factors identified in well-established, reproductively isolated species are likely to have arisen since speciation. Unveiling the genetic origins of species, “that mystery of mysteries” (71), requires an untangling of complex combinations of the evolutionary forces that create incompatibilities and the reproductive isolating mechanisms that result.

All flies were reared on cornmeal–molasses–agar medium sprinkled with yeast grains (SI Materials and Methods). The D. simulans stocks used in this study are w501 (University of California, San Diego, line 14021-0251.011), w;e (w;II:e), and simB (w;nt;III), where II and III represents isogenic second and third chromosomes, respectively, from D. simulans line (13w × JJ) (24). The D. mauritiana × D. simulans heterozygous introgression lines used in this study are designated P45.6 and P32.110 (72); the creation of these lines has been described in detail earlier (72, 73). P represents the immobile P-transposon element P[w⁺] marking the portion of D. mauritiana [Mau12, a white-eyed inbred laboratory stock (14021 0241.60)] on the right arm of the third chromosome. P[w⁺]-elements are semidominant markers sensitive to the location and copy number of the miniwhite w⁺ gene (Fig. S3 and SI Materials and Methods). All crosses were performed at room temperature (20–22 °C), except the cross that generated HMS1[mau]/HMS1[mau]; pB::HMS1[sim] (“3P”) progeny, which was performed at 18 °C. Absence of recombination in Drosophila males allows maintaining the original introgression lines by backcrossing each generation of P[w⁺]-males with virgin SimB females.

Germ-line injections were carried out by BestGene. D. simulans w⁻ embryos were injected with the purified MW-FPNS piggyBac (pB) plasmid carrying either agt[sim], Taf1[sim], agt[simD121N], agt[simP_mauCDS], agt[mauP_simCDS], Taf1[simPEx10_mauEx10end], Taf1[mauPEx10_ simEx10end], or a control MW-FPNS plasmid with no insert. Full details of cloning procedures are provided in SI Materials and Methods. Injected embryos were subsequently raised in the laboratory, and emerging individual virgin adult w⁻ flies were backcrossed to SimB. Progeny were screened via presence or absence of the w⁺ eye marker (74), and those containing the pB[w⁺] transposed DNA insert were selected to establish stable male transgenic lines. Altogether, the progeny of 429 (agt) and 160 (Taf1) injected adult flies was scored, all constructs considered, and yielded similar transformation efficiencies of 2.79% and 3.1%. Flanking genomic insertions were determined via inverse PCR (Table S1 and SI Materials and Methods). Stable male transgenic lines representing eight unique agt insertions, four unique Taf1 insertions, and two control lines were used to assess the role of each candidate gene in fertility rescue crosses (Fig. S1).

Rescue hybrid males bearing a copy of the SimB allele for either agt or Taf1 (HMS1[mau]/HMS1[mau]; pB::HMS1[sim]), or a chimeric agt or Taf1 construct, were assayed for their ability to sire progeny in single-mating fertility assays with three virgin w;e females before scoring of sperm motility and genotyping of all P[w⁺]-elements. Full details of these procedures are provided in SI Materials and Methods.

The genomic region comprising agt or Taf1 and their respective regulatory regions were amplified for D. mauritiana Mau12 and D. simulans SimB using overlapping oligonucleotide primers, sequenced, and analyzed in Geneious, version 9.0.5 (75). The sequences have been deposited in the GenBank database under accession nos. KX225407–KX225410.

Patterns of DNA divergence were calculated to detect departure from neutral models of molecular evolution among sequenced agt and Taf1 alleles from 17 D. simulans populations spanning all continents, as well as 16 independent isogenic lines of D. mauritiana collected on Mauritius island (see Results, Tests for Positive Selection). Various specific site models from the PAML, v4.8, package (36) were used to test for interspecific site-specific positive selection. Likelihood ratio tests were used to evaluate whether the model pairs allowing positive selection provided a significantly better fit to the data. Results from theses comparisons were consistent, all supporting neutral evolution in and around the HMS1 region.

The linear model used for the accumulation of 40 hybrid male-sterility factors in 250,000 y was f(t) = 40 × (t/250,000); the exponential model used was f(t) = 40 × [Exp(t/250,000) − 1]/[Exp(1) − 1].