Cytonuclear Genic Incompatibilities Cause Increased Mortality in Male F₂ Hybrids of Nasonia giraulti and N. vitripennis

Abstract

The haplodiploid wasp genus Nasonia is a promising model for studying the evolution of genic incompatibilities due to the existence of interfertile species and haploid males. The latter allows for significantly reducing the sample size required to detect and map recessive dysfunctional genic interactions. We exploited these features to study the genetics of intrinsic hybrid inviability in male F₂ hybrids of Nasonia giraulti and N. vitripennis. Analyzing marker segregation in 225 hybrid embryos, we inferred a linkage map with 38 framework markers. The markers were tested for marker transmission ratio distortion (MTRD) and interchromosomal linkage disequilibrium in populations of embryonic and adult hybrids. We found evidence for four transmission ratio distorting loci (TRDL). Three TRDL showed a deficit of the N. giraulti allele in hybrids with N. vitripennis cytoplasm. A separate TRDL exhibited a deficiency of the N. vitripennis allele in hybrids with N. giraulti cytoplasm. We ascribe the observed MTRD in adult hybrids to cytonuclear genic incompatibilities causing differential mortality during development since hybrid embryos did not show MTRD. The identified cytonuclear genic incompatibilities in F₂ hybrids with N. vitripennis cytoplasm account for most of the intrinsic hybrid inviability in this cross. The high mortality rate in F₂ hybrids with N. giraulti cytoplasm cannot be explained by the single identified TRDL alone, however.

UNDERSTANDING the genetic basis of reproductive barriers is one of the major challenges in evolutionary genetics and in speciation genetics in particular (Coyne and Orr 2004). Two frequently observed reproductive barriers that isolate species are intrinsic hybrid inviability and sterility. They can be caused by differences in ploidy levels, chromosomal organization, infection by endosymbionts, and genic incompatibilities (e.g., Stebbins 1950; Werren 1998; Johnson 2000; Delneri et al. 2003). So far, only a few studies have identified genes that are incompatible in hybrids and provided insights to how negative genic interactions account for hybrid inviability and sterility (e.g., Wittbrodt et al. 1989; Schartl et al. 1994; Malitschek et al. 1995; Ting et al. 1998; Barbash et al. 2003; Presgraves et al. 2003; Brideau et al. 2006). In general, however, little is known about the nature of genic incompatibilities that cause intrinsic postzygotic reproductive isolation (Hutter 1997; Coyne and Orr 2004; Orr et al. 2004; Orr 2005). Three problems may account for this lack of knowledge: the difficulty of carrying out genetics between populations that are reproductively isolated (Lewontin 1974; Orr et al. 2004; Orr 2005), the complexity of genic incompatibilities that frequently involve more than just two genes (e.g., Muller 1942; Dobzhansky 1975; Cabot et al. 1994; Orr and Irving 2001), and the fact that genic incompatibilities tend to be recessive (Muller 1942; Tao and Hartl 2003; see also Coyne and Orr 2004).

The parasitic wasp genus Nasonia has two particular features that allow us to overcome these obstacles and facilitate speciation genetic studies: a haplodiploid sex determination and the ease with which its species can be crossed in the laboratory (Whiting 1967; Skinner and Werren 1980; Breeuwer and Werren 1990; see also Beukeboom and Desplan 2003). In arrhenotokous species, females develop from fertilized eggs and are diploid whereas males develop parthenogenetically from unfertilized eggs and are haploid. Since there are no intralocus interactions in haploid males, alleles can easily be tested individually for their phenotypic effect in this sex. This significantly reduces the population sizes required to detect and map recessive alleles and epistatic loci in haploid F₂ hybrid males (Gadau et al. 1999, 2002; Weston et al. 1999; Wilfert et al. 2007). The genetics of genic incompatibilities can easily be explored in the genus Nasonia, since interspecific F₁ hybrid females are viable and fertile while a certain percentage of their F₂ hybrid male offspring are not (Breeuwer and Werren 1995).

Reproductive isolation between species of Nasonia (Darling and Werren 1990; Campbell et al. 1993) has been studied extensively. Under natural conditions, gene flow is impeded due to the infection of each Nasonia species with different strains of endosymbiotic Wolbachia bacteria (Breeuwer and Werren 1990; Bordenstein et al. 2001). These Rickettsia-like bacteria are maternally transmitted in the cytoplasm and are known to alter host reproduction (Werren 1997). In Nasonia, Wolbachia causes cytoplasmic incompatibility (Breeuwer and Werren 1990). Since nearly all individuals of Nasonia are doubly infected with two species-specific Wolbachia strains, the cytoplasmic incompatibility is bidirectional (Breeuwer and Werren 1990, 1993; Bordenstein and Werren 1999). All Nasonia species hybridize, however, once they are cured of their Wolbachia endosymbionts. Interspecific sexual isolation occurs, but it is asymmetrical and incomplete (Drapeau and Werren 1999; Bordenstein et al. 2000; Velthuis et al. 2005). While F₁ hybrid females are viable and fecund, haploid F₂ hybrids from crosses between Nasonia vitripennis and N. giraulti or N. longicornis suffer from increased mortality (Breeuwer and Werren 1995; C. Pietsch, L. Beukeboom and J. Gadau, unpublished results).

Hybrid breakdown, the increased sterility and mortality in F₂ and later hybrid generations, is thought to be caused by the recombination of two differentially coadapted parental genomes (Stebbins 1950; Coyne and Orr 2004). In Nasonia, hybrid breakdown was first investigated by Breeuwer and Werren (1995) in F₂ hybrids of N. giraulti and N. vitripennis using the inbred strains RV2T and AsymC. F₂ hybrid males in this cross suffer from increased mortality during larval development. F₂ hybrid males with N. giraulti cytoplasm experience additional mortality during the pupal stage. Diploid F₂ hybrid females obtained by backcrossing the F₁ hybrid females with males of either species suffer only marginally from hybrid breakdown, suggesting that the underlying genic incompatibilities are recessive and can be amended by adding a complete genome complement. Breeuwer and Werren (1995) suggested that both nuclear–nuclear and nuclear–cytoplasmic (=cytonuclear) genic incompatibilities are involved in N. giraulti × N. vitripennis F₂ hybrid breakdown.

To map the nuclear–nuclear genic incompatibilities that had been hypothesized by Breeuwer and Werren (1995), Gadau et al. (1999) searched for significant deviations from expected recombinant and nonrecombinant genotypes between markers from different linkage groups (LGs) in haploid N. giraulti × N. vitripennis F₂ hybrids with N. vitripennis cytoplasm. However, the authors chose to use a N. giraulti strain with N. vitripennis cytoplasm (i.e., R16A) for their cross experiment instead of the N. giraulti inbred strain RV2T with the idea of eliminating the impact of cytonuclear genic incompatibilities on their analysis. Interestingly, many markers in their study showed a significant bias toward N. vitripennis alleles (i.e., marker transmission ratio distortion, or MTRD), despite the fact that the N. giraulti nuclear genome was already introgressed in N. vitripennis cytoplasm and showed no incompatibilities within the R16A strain. The cause of the observed bias—potentially cytonuclear genic incompatibility due to a deficient introgression of the N. giraulti nuclear genome in the N. vitripennis cytoplasm—remained to be investigated.

MTRD, the preferential inheritance of one parental allele, is frequently observed in interspecific crosses (e.g., Hall and Willis 2005; Moyle and Graham 2006 and references therein), and several mechanisms that can cause a bias in the recovery rate of parental alleles are known. Selfish genetic elements, for example, are known to transmit themselves at the expense of their homologs (Lyttle 1991; Hurst 1993; Hurst and Werren 2001). Well-studied examples are the Segregation distorter in Drosophila (Ganetsky 1977; Wu et al. 1988; Merrill et al. 1999) and the t haplotypes in mice (Silver 1993). The underlying process that causes the corresponding alleles to be overrepresented in the gametes is referred to as meiotic drive, although the mechanisms of this process can differ significantly. Meiotic drive has been reported to drive the evolution of reproductive isolation (e.g., Merçot et al. 1995; Cazemajor et al. 1997; Montchamp-Moreau and Joly 1997; Tao et al. 2001; Fishmann and Willis 2005; Orr and Irving 2005). MTRD can also occur because of postzygotic viability differences due to an incompatibility of nuclear alleles with cytoplasmic factors (e.g., mitochondrial-encoded genes; Fishmann et al. 2001; Fishmann and Willis 2006). In most cases, however, the mechanism causing MTRD is unknown (Pardo-Manuel De Villena et al. 2000).

In this study, we investigate the occurrence and cause of MTRD in F₂ hybrid males of N. giraulti and N. vitripennis, considering both directions of the cross. We address the following questions:

1. Is MTRD also observed in crosses between the inbred strains RV2X(U) and AsymCX of N. giraulti and N. vitripennis, respectively? These strains are descended from AsymC and RV2T, which had been studied by Breeuwer and Werren (1995), and are assumed to be genetically identical with them.

2. Does MTRD differ between the reciprocal crosses?

3. How many transmission ratio distorting loci (TRDL) have to be assumed to explain the extent of MTRD in F₂ hybrid males that differ in their cytoplasm, and where in the genome are they located?

4. How much of the observed mortality can the TRDL explain?

5. Which TRDL allele is more frequently transmitted to adult F₂ hybrid males with a given cytoplasm? For example, is the N. vitripennis allele of a TRDL more frequently found in adult F₂ hybrids males with N. vitripennis cytoplasm than the N. giraulti allele?

6. Do male F₂ hybrid embryos already exhibit MTRD?

By answering these questions, we have taken the first step in identifying the genes that cause MTRD in Nasonia. The recently sequenced Nasonia genome (http://www.hgsc.bcm.tmc.edu/projects/nasonia/) provides the means for continuing this line of research and makes this a goal within reach.

MATERIALS AND METHODS

Stocks:

We used the Nasonia strains AsymCX and RV2X(U) for the cross experiments. AsymCX is derived from the Wolbachia-infected N. vitripennis wild-type strain LBii (LabII); its geographical origin is Leiden, The Netherlands (Breeuwer and Werren 1995). RV2X(U) is descended from the Wolbachia-infected N. giraulti wild-type strain RV2, which had been collected in Rochester, New York (Breeuwer and Werren 1995). We chose these two strains because they are highly inbred (i.e., homozygous for every marker tested so far), antibiotically cured of their endosymbiotic Wolbachia bacteria (allowing the formation of interspecific F₁ hybrids), and were used for the genome sequencing (Breeuwer and Werren 1990, 1993; Breeuwer et al. 1992; http://www.hgsc.bcm.tmc.edu/projects/nasonia/). The two Nasonia laboratory strains and their hybrids were cultured in an incubator on pupae of the flesh fly (Sarcophaga bullata) at 25° and with permanent light.

Cross experiments:

We began the cross experiments by collecting males and females of the two strains AsymCX and RV2X(U) as virgin pupae and raised each individual in a separate vial. Once the wasps had eclosed, we provided them with a 3:1 solution of honey to water and kept them in the incubator for another 2 days. We then placed each female together with a single male of the other strain in a small glass tube (12 × 75 mm; Fischer Scientific, Fair Lawn, NJ). After 24 hr, each pair was provided with two host puparia. We collected the F₁ hybrid females as virgin pupae 10 days later and placed them in separate vials. Their male siblings were not hybrids, since they developed from unfertilized eggs. Eclosed F₁ hybrid females were fed with a 3:1 solution of honey to water and provided with a host puparium 2 days later. Since the F₁ hybrid females were virgins, they laid unfertilized eggs, which developed into haploid (hemizygote) F₂ hybrid males. We opened a fraction of parasitized host puparia after 12–16 hr and, using a sterile pin, removed 120 F₂ hybrid embryos each from the cross AsymCX (♀) × RV2X(U) (♂) and the reciprocal cross AsymCX (♂) × RV2X(U) (♀). Each embryo was transferred into a separate 0.2-μl PCR tube and stored at −70° for subsequent procedures. The remaining host puparia were left intact for 10 days. We then carefully opened them with a pair of forceps to facilitate the emergence of the adult wasps, because F₂ hybrid males are not capable of opening their host puparium themselves (our personal observation). Within 36 hr after their eclosion, we collected 120 adult F₂ hybrid males each from the cross AsymCX (♀) × RV2X(U) (♂) and the reciprocal cross AsymCX (♂) × RV2X(U) (♀). The adult wasps were transferred into 95% ethanol and stored at −70° for subsequent procedures. To verify TRDL (see below) predicted from analyzing the two populations of adult F₂ hybrid males, we repeated the cross experiments and collected an additional 120 adult F₂ hybrid males from each cross within 36 hr after they hatched. These hybrids were also stored in 95% ethanol at −70°.

Molecular procedures:

DNA from adult wasps was extracted using a Chelex extraction protocol described by Niehuis et al. (2007). Due to the small amount of genomic DNA in early F₂ hybrid embryos, DNA was preamplified for use in subsequent procedures using the GenomiPhi DNA amplification kit (Amersham Biosciences, Piscataway, NJ). Each embryo was first homogenized with the aid of a sealed sterile 10-μl pipette tip in 1 μl alkaline lysis solution (400 mm KOH, 100 mm DTT, 10 mm EDTA). After 10 min incubation, we added 1 μl neutralization buffer (400 mm HCl, 600 mm Tris–HCl, pH 6.0) to the lysate and the lysate was briefly mixed well. Each sample tube was then filled up to 10 μl with sample buffer (included in the kit) and heated for 3 min at 95°. After the samples were cooled to 4° on ice, we amplified the template DNA with the GenomiPhi kit reaction premix following the manufacturer's recommendations. The amplification products were subsequently purified using a standard ethanol precipitation protocol. Quantity and quality of Chelex DNA extracts and of the GenomiPhi amplified DNA were assessed with a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE).

Marker segregation data were collected using polymerase chain reaction (PCR)-based methods and consisted of three types of molecular genetic markers for sampling the nuclear genome: length polymorphic (LP) markers, present/absent (P/A) markers, and single nucleotide polymorphism (SNP) markers. Most of the LP markers and all P/A markers were adopted from Pietsch et al. (2004), Rütten et al. (2004), Pietsch (2005), and Niehuis et al. (2007); markers published by Rütten et al. (2004) were chromosomally anchored. Additional LP markers had been reported by Gadau et al. (2008), but lacked a formal characterization so far (see appendix). SNP markers were adopted from Niehuis et al. (2007) and Gadau et al. (2008) (see appendix).

APPENDIX.

Characterization of nuclear and mitochondrial molecular markers in this investigation

Locus	Type	Primer sequences (5′ → 3′)	Amplicon lengtha	SNP positiona	GenBank accession no.a	PCR temperature profilec	Reference
Chromosome 1
NvC1-20	LP	F: TGTAAAAGTAGTCCGCTTCG	∼117/∼125		–	I	Rütten et al. (2004)
		R: TATTTATATATGGAAAAAGAGG			–
Cu/Zn SOD	SNP	F: AGTTCATGCCTTTGTGATGC	253	152 (A/G)	EF190557	III	Niehuis et al. (2007)
		R: CACCGACAAACAGATTCAGC			EF190556
NvC1-21	P/A	F: GTAACAGTGAGATAAATGTG	−/∼148		–	I	Rütten et al. (2004)
		R: TAGCAACGATAGTCCACG			–
NvC1-6	LP	F: GGTTGCTTTTAAGTCTTTGC	∼143/∼141		–	I	Rütten et al. (2004)
		R: CTGGTCTTCTGCATAATGG			–
NvC1-22	LP	F: GCAGAGTCGAGGCAAG	∼214/∼206		–	I	Rütten et al. (2004)
		R: TTACCGGAGTTCGTTAAC			–
EF-1α F1	LP	F: GGTCATCGTCCTCAATCATC	219/214		EF190561	III	Niehuis et al. (2007)
		R: TTGTTTTACCTTTTCCCAACC			EF190560
NvC1-12	LP	F: TCGCATTTTACATCTCTTTC	∼91/∼146		–	I	Rütten et al. (2004)
		R: GAGATAAACGAATCAAAAGAC			–
NvC1-13	P/A	F: TAAAAGTATTAGACCTTTGGG	−/∼115		–	iv	Rütten et al. (2004)
		R: AGTGGCTGAGCTTGGC			–
Chromosome 2
Hex70b	SNP	F: GCAGCTTCTCAGGAGTCAGG	185	88 (G/T)	EF638397b	III	Gadau et al. (2008)
		R: AGGCATCATCTTCAGCAAGG		163 (C/T)	EF638396b		and this article
UQCRFS1	SNP	F: GCAACCTGAATGGTTGATTG	142	82 (A/T)	EF190571	III	Niehuis et al. (2007)
		R: TGGACCTTTCCTAATTCTTCC			EF190570
Hsc70	SNP	F: AGGCCAGCATTGAAATCG	208	137 (T/C)	EF638394b	III	Gadau et al. (2008)
		R: TTCTGGATCTTGGGGATACG			EF638395b		and this article
Nv-20	LP	F: TGACGAAGTATCCGAGAAG	∼87/∼103		–	II	Rütten et al. (2004)
		R: TCGAAAAACGATATTGCTCG			–
PTEN	LP	F: ACGGGTGTGATGGTATGTTG	232/235		EF638400b	III	Gadau et al. (2008)
		R: ACATACCTCCTTTGGGATGG			EF638401b		and this article
Nv-23	P/A	F: CAGCATACTCAAGCAAGC	−/∼217		–	II	Rütten et al. (2004)
		R: GATACCTGAAGTTTGATGC			–
Nv-36	LP	F: TCGATCCAGATGAAGAGG	∼140/∼168		–	I	Pietsch (2005)
		R: AGAGAATTAAGAGAAAGTCCG			–
EF-1α F2	SNP	F: CCAGGACGTCTACAAGATTGG	249	106 (T/C)	EF638393b	III	Gadau et al. (2008)
		R: GGGTTGTTCTTGGAGTCACC		151 (A/T)	EF638392b		and this article
Vg	SNP	F: GGTCCTGACCAGAGAGAACG	191	25 (C/T)	EF638398b	III	Gadau et al. (2008)
		R: GCTCGAAGATTTCGAAGACG		75 (C/G)	EF638399b		and this article
TpnC	SNP	F: CTGAGGCTCTGGAGAAGGAG	257	121 (A/G)	EF190573	III	Niehuis et al. (2007)
		R: AAGTCGACGGTGCCAGAG			EF190572
Nv-17	LP	F: AAGAATGTATCAAGTATGAGCC	∼234/∼215		–	I	Pietsch (2005)
		R: TCAGTTCTTGAAACGTTGC			–
Chromosome 3
INSR	SNP	F: AAGATGTGGTCAAAGGCAAG	203	29 (T/G)	EF638386b	III	Gadau et al. (2008)
		R: ATTAAAGCCCCTGCAAAAAG		53 (T/C)	EF638387b		and this article
				124 (A/G)
				163 (C/A)
				178 (C/T)
Hsp83	SNP	F: GCTCTTCCTCGTCGATTCC	215	178 (T/C)	EF190563	III	Niehuis et al. (2007)
		R: GAAGCACCTCGAGATCAACC			EF190562
apoLp-III	LP	F: CCTTCCTGGAACTGCTTCTG	271/270		EF190555	III	Niehuis et al. (2007)
		R: TTCGTCAACAACGTCCAGAG			EF190554
NvC3-18	P/A	F: GCCCAAATCATGCTTTCG	−/∼105		–	II	Rütten et al. (2004)
		R: GTTGTTCTTAAATGTGTATTCC			–
RPS2	SNP	F: TCAGCGTACTCGTTTCAAGG	256	141 (T/C)	EF190553	III	Niehuis et al. (2007)
		R: GCTTACCGATCTTGTTACCC			EF190552
SDR	LP	F: ACTGCTCTGCGCAATATGAC	321/307		EF638383b	III	Gadau et al. (2008)
		R: ATCAGAACAGCATCCGCTAC			EF638382b		and this article
TOR	LP	F: TCTTCCCATTCAACATCCAC	174/175		EF638390	III	Gadau et al. (2008)
		R: TGACCATAAGCTGCGTCAC			EF638391		and this article
Chromosome 4
Nv-26	LP	F: TTCGCAGCTTTCCTTTGC	∼117/∼141		–	II	Rütten et al. (2004)
		R: AGCAGCTAGTATGAACCG			–
Bb-77L07e	LP	F: GCCCGAAGATCTACATACGC	227/259		EF638385b	I	Gadau et al. (2008)
		R: CGATCCCAAATTACCTGCAC			EF638384b		and this article
Nv-24	LP	F: CCGAAATCCACATAGACC	∼98/∼118		–	II	Rütten et al. (2004)
		R: AGGAACTCATCAAGACGG			–
PPAF-III	SNP	F: TGTCAAGCTCAAGGTCAACG	237	49 (C/T)	EF190559	III	Niehuis et al. (2007)
		R: TCCATACCACAGGGTGACG			EF190558
ACT	LP	F: TGCGAAACCAATTTCTTCTG	86/90		EF190569	III	Niehuis et al. (2007)
		R: AATGTGTATTAAAAGCACTTTCG			EF190568
NvC4-15	P/A	F: GCAGGGCTTTGTTATAGC	−/∼111		–	I	Rütten et al. (2004)
		R: CGACGAAACCGAAGTGG			–
COXVa	SNP	F: AATTCAATGTAAGGCTCGTTTC	144	77 (C/G)	EF190567	III	Niehuis et al. (2007)
		R: TGGATGAACTGGGTATCAGC			EF190566
Chromosome 5
26-kDa lectin	SNP	F: CGTATTTTTCAAGCCTCTCG	153	35 (C/T)	EF190565	III	Niehuis et al. (2007)
		R: TCGCAGATAAAAGCTCGTTC			EF190564
Nv-46	LP	F: TTACGTCAAGGTATAGCTGC	∼303/∼332		–	I	Pietsch (2005)
		R: AATAAGTGGCTGAAAGTTCC			–
Nv-25	P/A	F: GTAAGTCTGCGGTAGCTG	−/∼238		–	II	Rütten et al. (2004)
		R: TTGACGGAGTAGTTCCAG			–
Ft	SNP	F: CAGCCTTCTATGCCAGTGTG	156	64 (A/G)	EF638389b	III	Gadau et al. (2008)
		R: CTGGTCGAAATGTAGAAAAAGC		84 (G/C)	EF638388b		and this article
				87 (T/C)
Nv-27	LP	F: AATACTCGCTGTTCAATCG	∼168/∼192		–	II	Rütten et al. (2004)
		R: CGCTAGATCGGATTTCCG			–
Mitochondrion
COI	SNP	F: CAACATTTATTTTGATTTTTTGG	385	70 (G/A)	EF638403b	V	Modification from
		R: GCWACWACRTAATAKGTATCATG		190 (A/T)	EF638402b		Gadau et al. (1999)

LP, length polymorphic marker; SNP, single nucleotide polymorphism marker; P/A, present/absent marker.

^aThe two sizes/alleles/accession numbers are for N. giraulti and N. vitripennis, respectively.

^bNewly submitted sequences.

^cI: 5 min at 95°, 10 cycles of 1 min at 94°, 1 min at 55° minus 1° each cycle, 1 min at 72°, followed by 25 cycles of 1 min at 94°, 1 min at 49°, 1 min at 72°; final 10 min at 72°. II: 5 min at 94°, 15 cycles of 1 min at 94°, 1 min at 60° minus 1° each cycle, 1 min at 72°, followed by 20 cycles of 1 min at 94°, 1 min at 50°, 1 min at 72°; final 10 min at 72°. III: 5 min at 95°, 30 cycles of 1 min at 95°, 1 min at 55°, 1 min at 72°; final 10 min at 72°. IV: 5 min at 95°, 30 cycles of 1 min at 95°, 1 min at 50°, 1 min at 72°; final 10 min at 72°. V: 4 min at 95°, 35 cycles of 1 min at 95°, 1 min at 45°, 1 min at 68°; final 4 min at 68°.

The genotype of LP and P/A markers was inferred by separating their PCR products on a denaturing polyacrylamide gel. The PCRs were performed in 12.5-μl volumes [1× colorless GoTaq reaction buffer, 0.625 units GoTaq polymerase (Promega, Madison, WI), 1.6 mm dNTP mix, 0.4 μm of each primer, 10 ng DNA] using an Eppendorf epGradient Mastercycler (Eppendorf, Hamburg, Germany). To facilitate the detection of the PCR products, we used fluorescently (IR700/800) labeled oligonucleotide primers. The applied PCR temperature profiles are given in the appendix. All fluorescently labeled PCR products were separated on a LI-COR 4300 DNA analysis system (LI-COR, Lincoln, NE) and analyzed using the SAGA Generation 2 software (LI-COR). SNP markers were genotyped with the recently described Ecotilling technique (Comai et al. 2004), applying the protocol given by Niehuis et al. (2007).

The genotype of the mitochondria in hybrids was inferred by sequencing a 385-bp-long fragment of the mitochondrial-encoded gene Cytochome oxidase subunit I (COI). Since all F₁ hybrid females in our cross experiments had laid unfertilized eggs, we genotyped only one representative F₂ hybrid male from each F₁ hybrid female's offspring. The COI fragment was amplified via PCR [25 μl volume (1× colorless GoTaq reaction buffer, 0.625 units GoTaq polymerase (Promega), 1.6 mm dNTP mix, 0.4 μm of each primer (see appendix), 20 ng DNA] using a PTC-100 Peltier Thermal Cycler (MJ Research, Watertown, MA). The applied PCR temperature profile is given in the appendix. After checking the PCR products on 1.5% agarose mini-gels, we enzymatically cleaned them up using the ExoSAP-IT kit (USB, Cleveland). Both strands of the DNA fragments were then sequenced in the Arizona State University sequencing facility on a 3730 DNA Analyzer (Applied Biosystems, Foster City, CA) using BigDye Terminator v3.1 cycle sequencing kit (Applied Biosystems). We assembled the complementary strands with the Sequencher 4.6 software (Gene Codes, Ann Arbor, MI) and aligned the contigs in BioEdit 7.0.5.3 (Hall 1999). The sequences are deposited in GenBank under the accession nos. EF638404–EF638434.

Linkage map reconstruction:

Marker segregation data were studied with the program MultiPoint (http://www.multiqtl.com), which infers marker order and map distances using a multipoint likelihood approach (Mester et al. 2003a,b, 2004). To minimize the effect of negatively interacting loci causing mortality on the linkage map reconstruction due to a phenomenon commonly referred to as quasi- or pseudolinkage (Korol et al. 1989; Lorieux et al. 1995a,b; Liu 1998; Weston et al. 1999; Peng et al. 2000), we considered only F₂ hybrid embryos for linkage analysis. We first evaluated the mapping population (N = 225) by analyzing markers for missing information and MTRD. We subsequently searched for putative sets of linkage groups by calculating pairwise recombination fractions (rf_s) for all pairs of markers using a maximum-likelihood estimation procedure as implemented in the software. We chose rf_s = 0.35 as a reasonable threshold level as it prevented fusion of anchored markers from different chromosomes or disruption of anchored markers on single chromosomes. The optimal marker order in individual linkage groups was found by minimizing the total length of the multi-locus map while iteratively testing for marker-order stability using a Jackknife approach (sampling 90% of the total population, 1000 iterations) and successively removing markers with uncertain position. Markers violating a monotonic increase of rf_s along linkage groups were finally deleted by applying the control for monotony function. After a consistent linear order of markers was reached, we calculated map distances in centimorgans from the recombination fractions using Haldane's mapping function (Haldane 1919). We refrained from using Kosambi's mapping function (Kosambi 1944), since the Bayesian approach that we used to map TRDL (see below) assumes a simple noninterference crossover model (i.e., Haldane's mapping function; Vogl and Xu 2000).

Genome length and coverage estimation:

The length of individual linkage groups was estimated employing “method 4” described by Chakravarti et al. (1991). Thus, assuming a uniform distribution of markers on the linkage groups, the map distance between the most distal markers on each linkage group was inflated by multiplying the factor (m_i + 1)/(m_i − 1), where m_i is the number of markers on the linkage group (Chakravarti et al. 1991). We tested for a uniform marker distribution with the Kolmogorov–Smirnov (K–S) goodness-of-fit test for continuous data (Zar 1999) and employed the correction for small sample sizes expounded by Harter et al. (1984) and Khamis (1990, 1993). The total genome length, L, was calculated as sum of the estimated lengths of all linkage groups. To estimate the genome coverage, we considered two approaches: (i) applying the formula c = 1 − e^−2nd/L, provided by Lange and Boehnke (1982), with c being the fraction of the genome within d cM of a marker on the linkage map and n being the total number of markers on the linkage map; it assumes that markers are randomly distributed on the linkage map; (ii) dividing the estimated total genome length through the summed lengths of all linkage groups. To test for a random distribution of markers on the linkage map, we compared the distribution of marker intervals of consecutive markers along linkage groups with the null hypothesis that they follow a normal distribution applying the Kolmogorov–Smirnov one-sample test and Lilliefors' significance correction (Lilliefors 1967). We calculated the average marker spacing by dividing the summed lengths of all linkage groups by the number of marker intervals.

Analysis of marker segregation ratios:

Markers in the data set were first explored in Microsoft Excel 2003 (Microsoft, Redmond, WA) with the χ² test for goodness of fit (Zar 1999) to the expected 1:1 segregation ratio of parental alleles (d.f. = 1) and applying Yates' correction for continuity (Yates 1934). We applied a Bonferroni correction to lower the number of false positives by dividing α = 0.01 by the number of linkage groups (i.e., five). Our criterion for rejecting the null hypothesis (i.e., no MTRD) was thus P < 0.002. We further applied a Bayesian multipoint mapping approach developed by Vogl and Xu (2000) and implemented in the software ANITA (provided by Claus Vogl) to estimate the number, location, and effect of TRDL on each linkage group. The maximum number of TRDL per linkage group was set to four and the position of analyzed markers predefined by adopting the inferred linkage map after removing redundant markers (i.e., NvC1-22 and Nv-23; these markers map at the same location as EF-1α F1 and Nv-36, respectively). For each linkage group, we ran one Markov chain with random start parameters and 21,000 iterations. The first 1000 iterations were conservatively discarded to account for a possible burn-in period of the Markov chain. The remaining samples of the Markov chain were subsequently analyzed in Microsoft Excel 2003 and SPSS 14.0 (SPSS, Chicago). To measure the change of the odds between the prior and posterior probability for the number of TRDL on a given chromosome, we calculated Bayes factors (Lavine and Schervish 1999). The density distributions of the posterior probability for position and effect of TRDL were approximated by grouping sample values in successive bins of 1 cM and of 1% N. vitripennis alleles, respectively. For specifying intervals, we used a square bracket notation. Specifically, [x,y[ indicates an interval ranging from x to y, including the value x and excluding the value y. As an estimate of the most likely position and effect of TRDL, we took the mode of the estimated posterior probability distributions. Selected markers close to hypothesized TRDL were finally tested for distorted segregation ratios in a second population with the same cytoplasm as the one in which the effect was initially observed. Segregation data were further explored for negative nuclear–nuclear and nuclear–nuclear–cytoplasmic (i.e., between a cytoplasmic and two nuclear loci) genic interactions by testing for deviations between the expected and the observed frequency of recombinant and nonrecombinant genotypes between markers from different linkage groups (see also Gadau et al. 1999; Sawamura et al. 2006), assuming an independent assortment of markers on different chromosomes. The tests were performed with the help of a Perl script using the χ² test for goodness of fit (d.f. = 1). To lower the number of false positives, we applied a Bonferroni correction by dividing α = 0.01 by the number of possible chromosome pairs (i.e., 10). Our criterion for rejecting the null hypothesis was thus P < 0.001. As for the hypothesized TRDL, selected markers close to possible negatively interacting nuclear loci were subsequently tested for linkage disequilibrium in a second population with the same cytoplasm as the one in which the effect was initially observed.

RESULTS

Mapping populations:

We obtained six mapping populations from our cross experiments. Three of them had N. giraulti cytoplasm (hereafter indicated by [g]) and three had N. vitripennis cytoplasm (hereafter indicated by [v]). The genotype of the cytoplasm was confirmed by sequencing a fragment of the mitochondrial-encoded gene COI in representative F₂ hybrids (N = 31). For each direction of the N. giraulti × N. vitripennis cross, there were thus two populations of adult F₂ hybrid males, one for identifying TRDL (N_[g] = 120, N_[v] = 116) and one for verifying them (N_[g] =120, N_[v] = 120), as well as an additional population with <25-hr-old F₂ hybrid male embryos (N_[g] = 112, N_[v] = 113).

Linkage map:

To exclude pseudolinkage problems, we considered only segregation data from hybrid embryos for the generation of a linkage map. The linkage analysis is based on segregation data of 47 markers in 225 haploid F₂ hybrids (i.e., 112 F₂ hybrid embryos with N. giraulti cytoplasm plus 113 F₂ hybrid embryos with N. vitripennis cytoplasm). The phase of the markers was known from prior screening representatives of each of the two parental highly inbred strains [i.e., RV2X(U), AsymCX]. Nine of the 47 markers were removed from the data set during the mapping process because they significantly reduced marker-order stability and/or violated the monotony criterion.

The inferred N. giraulti × N. vitripennis linkage map consists of 38 molecular markers (characterized in the appendix), which mapped into five linkage groups. By considering chromosome anchored microsatellite markers (Rütten et al. 2004), it was possible to homologize individual LGs with chromosomes (chr) (Figure 1).

Figure 1.—

Linkage map inferred from marker segregation data in a N. giraulti × N. vitripennis F₂ hybrid population of male embryos (N = 225). Recombination distances are shown in Haldane centimorgans. Bars near chromosomes specify the 95% confidence limits for the position of predicted TRDL; arrowheads point to the region with the highest posterior probability for the position of TRDL. Percentages indicate the estimated mortality caused by each TRDL in a population of F₂ hybrid males with [g] = N. giraulti and [v] = N. vitripennis cytoplasm, respectively.

Estimated chromosome lengths and genome coverage:

The N. giraulti × N. vitripennis chromosome map spans 381.5 cM (Haldane mapping function). The lengths of the chromosomes are 113.6 cM (chr 1), 105.4 cM (chr 2), 54.3 cM (chr 3), 87.0 cM (chr 4), and 21.2 cM (chr 5), respectively. Since the distribution of the markers on chromosomes 1, 2, 4, and 5 does not significantly differ from a uniform distribution (K–S test, P > 0.5), we estimated the chromosome lengths by inflating the observed chromosome lengths with the formula described by Chakravarti et al. (1991) under “method 4.” The estimated lengths are 146.1 cM (chr 1), 126.5 cM (chr 2), 116.0 cM (chr 4), and 31.8 cM (chr 5). The most distant markers on each chromosome would accordingly span 78, 83, 75, and 67%, respectively, of its actual length. Applying the same method for chromosome 3 provided an estimated length of 72.4 cM, and the most distant markers on this chromosome would span 75% of its total length. However, the distribution of the markers on chromosome 3 significantly differed from a uniform distribution (K–S test, P < 0.01).

The estimated total genome length, L, as sum of the estimated lengths of all linkage groups is 492.8 cM. Our N. giraulti × N. vitripennis chromosome map would consequently span 77% of the total genome, but this estimate includes the questionable value for chromosome 3. We also considered the method proposed by Lange and Boehnke (1982) to estimate the genome that is covered within 10 and 20 cM, respectively, of a marker on the map. The corresponding values are c_10cM = 79% and c_20cM = 95%. This calculation assumes a random distribution of markers on the chromosome map and an accurate estimate of the total genome length. The distribution of marker intervals of consecutive markers along the linkage groups, however, shows that markers tend to cluster (Figure 2). Compared to an expected average marker spacing of 11.6 cM, there is a disproportionally high frequency of intervals with a length of 0–5 cM (Figure 2). Testing the observed distribution of marker intervals against the null hypothesis that they follow a normal distribution showed this difference as significant (K–S test, P < 0.01).

Figure 2.—

Distribution of the interval size between consecutive markers in the N. vitripennis × N. giraulti chromosome map. The superimposed graph shows a normal distribution with a mean of 11.7 cM against which the observed interval size distribution was tested (K–S test, P < 0.01).

Marker transmission ratio distortion:

None of the markers in the two F₂ hybrid populations of male embryos showed a significant bias from a 1:1 Mendelian segregation ratio (χ² test; P > 0.05). The same was true for adult F₂ hybrid males with N. giraulti cytoplasm (χ² test; P > 0.05). By contrast, 11 of the 38 markers in adult male hybrids with N. vitripennis cytoplasm exhibited MTRD (χ² test at P < 0.002): NvC1-12 (chr 1), Nv-20, Nv-23, Nv-36, EF-1α F2, PTEN, Vg (chr 2), Nv-27, Nv-46, Ft, 26-kDa lectin (chr 5).

The Bayesian analyses of segregation ratios in adult wasps corroborated the existence of TRDL in hybrids with N. vitripennis cytoplasm on chromosomes 1, 2, and 5 (Table 1); the posterior probabilities (hereafter abbreviated pP) for these assumptions were 99.5, 97.89, and 97.58%. The pP for no TRDL was >50% for the remaining chromosomes in hybrids with N. vitripennis cytoplasm and for all chromosomes in hybrids with N. giraulti cytoplasm, except for chromosome 4, in which the pP for the presence of TRDL was 55.09%.

TABLE 1.

Posterior probabilities for the number of TRDL on individual chromosomes in four mapping populations of F₂ hybrids of N. giraulti and N. vitripennis

		Posterior probabilityb for no. of TRDL and corresponding Bayes factorsc
Populationa	Chromosome	0 (%)	B12 (0)	1 (%)	B12 (1)	2 (%)	B12 (2)	3 (%)	B12 (3)	4 (%)	B12 (4)
Embryonic F2 hybrids [g] (N = 112)	1	66.85	1.3	28.82	0.9	3.88	0.5	0.45	0.4	0.01	0.1
	2	87.70	4.6	10.96	0.3	1.21	0.1	0.13	0.1	0.00	0.0
	3	88.87	5.2	9.62	0.2	1.38	0.2	0.13	0.1	0.01	0.1
	4	83.05	3.2	14.64	0.4	2.09	0.3	0.21	0.2	0.02	0.1
	5	86.15	4.0	11.67	0.3	1.97	0.2	0.22	0.2	0.01	0.1
Adult F2 hybrids [g] (N = 120)	1	73.24	1.8	18.38	0.5	8.75	1.2	1.53	1.2	0.12	0.8
	2	85.20	3.7	12.79	0.3	1.71	0.2	0.29	0.2	0.02	0.1
	3	89.08	5.3	9.43	0.2	1.28	0.2	0.21	0.2	0.02	0.1
	4	44.93	0.5	48.15	2.1	6.24	0.8	0.66	0.5	0.04	0.3
	5	90.30	6.0	8.43	0.2	1.09	0.1	0.18	0.1	0.01	0.1
Embryonic F2 hybrids [v] (N = 113)	1	87.31	4.5	11.23	0.3	1.32	0.2	0.15	0.1	0.01	0.1
	2	88.28	4.9	10.66	0.3	0.97	0.1	0.10	0.1	0.00	0.0
	3	88.09	4.8	10.24	0.3	1.49	0.2	0.17	0.1	0.03	0.2
	4	84.34	3.5	13.96	0.4	1.49	0.2	0.21	0.2	0.01	0.1
	5	54.35	0.8	38.62	1.4	6.20	0.8	0.81	0.6	0.03	0.2
Adult F2 hybrids [v] (N = 116)	1	0.50	0.0	80.86	9.7	16.48	2.4	2.03	1.6	0.13	0.8
	2	2.11	0.0	72.79	6.1	21.75	3.4	3.21	2.6	0.15	0.9
	3	78.13	2.3	15.09	0.4	5.65	0.7	1.01	0.8	0.13	0.8
	4	80.71	2.7	17.20	0.5	1.88	0.2	0.20	0.2	0.03	0.2
		5	2.43	0.0	77.70	8.0	16.45	2.4	3.19	2.6	0.24	1.5

^aGenotype of cytoplasm is indicated by [g] for N. giraulti and [v] for N. vitripennis, respectively.

^bPosterior probability values calculated from 20,000 Markov chain Monte Carlo samples taken from the stationary phase and assuming a Poisson prior (λ = 0.5) for the no. of TRDL. Highest posterior probabilities are in italics.

^cBayes factors (B₁₂) indicate the change of the odds between the prior and posterior probability for the tested hypotheses.

The calculated Bayes factors (Table 1), which are a measure of the change of the odds between the prior and posterior probability for hypotheses, suggested that only one TRDL on each of the four chromosomes mentioned above (i.e., chr 1 [v], chr 2 [v], chr 4 [g], and chr 5 [v]) would likely be responsible for the bias of markers. For the three chromosomes in F₂ hybrid males with N. vitripennis cytoplasm, this assumption was reasonably well supported as indicated by Bayes factors of 9.7 (chr 1), 6.1 (chr 2), and 8.0 (chr 5). For chromosome 4 in F₂ hybrids with N. giraulti cytoplasm, however, the data set changed the odds in favor of one TRDL by only 2.1.

The Bayesian multipoint mapping suggested that the putative TRDL on chromosome 1 is with 95% confidence in the interval [66,99[ cM, with the highest pP between [81,83[ cM. The obtained density function for the position of the putative TRDL on chromosome 2 was more complex, however, and the 95% confidence interval is discontinuous: [10,14[ cM + [15,21[ cM + [22,23[ cM + [24,25[ cM + [29,92[ cM. The highest pP was found for the interval [47,48[ cM. The putative TRDL on chromosome 4 is with 95% confidence within the two intervals [0,61[ cM and [62,64[ cM, with the highest pP between [8,9[ cM. The putative TRDL on chromosome 5, finally, is with 95% confidence in the interval [0,8[ cM + [11,21.2[ cM; the highest pP was found for the interval [0,1[ cM. The results are summarized in Figure 3A.

Figure 3.—

Density distributions of the posterior probability for (A) position and (B) effect of TRDL on chromosomes in adult male F₂ hybrids of N. giraulti and N. vitripennis. Each distribution was calculated from 20,000 Markov chain Monte Carlo samples taken from the stationary phase and assuming a Poisson prior (λ = 0.5) for the number of TRDL. Shown are density distributions for one TRDL on each of the depicted chromosomes; the posterior probabilities for these assumptions were 80.86, 72.79, 48.15, and 77.70% (see Table 1). The genotype of the cytoplasm of the population, in which marker transmission ratio distortion was seen, is indicated by [g] = N. giraulti and [v] = N. vitripennis, respectively. Underlined markers were tested for segregation ratio distortion in a second population of adult F₂ hybrid males to confirm the TRDL (see Figure 4).

We further used the Bayesian mapping approach to estimate the effect of putative TRDL in terms of the N. vitripennis allele recovery rate at a TRDL. For the putative TRDL on chromosome 1, the analysis suggested with 95% confidence a recovery rate of N. vitripennis between 59 and 82%; the highest pP was between 67 and 68%. For chromosome 2, the recovery rate with 95% confidence at the TRDL is between 55 and 78%; the highest pP is between 63 and 64%. For the TRDL on chromosome 4, the 95% confidence limits for the recovery rate of N. vitripennis are 27–51%, with the highest pP between 41 and 42%. The confidence limits for the effect on chromosome 5 are 55–77%, with the highest pP obtained for a recovery rate between 63 and 64%. The results are summarized in Figure 3B.

On the basis of the preceding analysis of MTRD in the first two populations of adult F₂ hybrid males, we selected a total of six markers close to the most likely positions of the four hypothesized TRDL to test their segregation ratios in two independent populations of adult F₂ hybrid males (Figure 3A). Specifically, we chose NvC1-12 (chr 1), Nv-20, Nv-36, Vg (chr 2), and 26-kDa lectin (chr 5) for testing in the second set of F₂ hybrid males with N. vitripennis cytoplasm and Nv-26 (chr 4) for testing in the second set of F₂ hybrid males with N. giraulti cytoplasm. Since only a limited number of tests had to be performed and because these tests were meant to confirm the existence and direction of the hypothesized TRDL, we applied a conservative Bonferroni correction by dividing α = 0.01 by 5 while testing for a segregation ratio distortion of the five markers in F₂ hybrid males with N. vitripennis cytoplasm. The criterion for rejecting the null hypothesis accordingly was P < 0.002. The criterion for rejecting the null hypothesis in F₂ hybrid males with N. giraulti cytoplasm was P < 0.01. The results of the tests for MTRD are illustrated in Figure 4. As predicted, we found a significant distortion in at least one of the selected markers on each of the four chromosomes. Furthermore, the specific direction of the distortions corresponded exactly with those predicted from the first two mapping populations. By contrast, none of the deviations from the expected 1:1 segregation ratios in these markers was significant in embryos, and in two instances (NvC1-12 and 26-kDa lectin) the specific direction of the observed bias in embryos is opposed to that found in adult wasps.

Figure 4.—

Deviation of selected markers linked to predicted transmission ratio distorting loci in six populations of male F₂ hybrids of N. giraulti and N. vitripennis from the expected 50% N. vitripennis alleles. The genotype of the cytoplasm of a population is indicated by [g] = N. giraulti and [v] = N. vitripennis. Solid bars: deviation of the markers in the two populations ([g] and [v] cytoplasm, respectively) of adult wasps in the original experiment, which aimed to identify possible TRDL (asterisks indicate significance). Shaded bars: deviation of the same markers in two independent populations ([g] and [v] cytoplasm, respectively) of adult wasps studied in a second (confirmation) experiment (asterisks indicate significance). Open bars: deviation of markers in early (<16 hr old) hybrid embryos.

Linkage disequilibrium:

A comparison between the expected and observed frequency of recombinant and nonrecombinant genotypes between markers from different chromosomes in the first set of adult F₂ hybrids indicated significant deviations in five pairs of markers in F₂ males with N. vitripennis cytoplasm (χ² test at P < 0.001): RPS2/Nv-24, RPS2/ACT, SDR/Nv-24, SDR/Bb-77L07e, and SDR/ACT. However, none of these pairs was found significantly biased in the second population of adult F₂ hybrid males with N. vitripennis cytoplasm (χ² test, P > 0.3). We further compared the expected and the observed frequency of recombinant and nonrecombinant genotypes between markers from different linkage groups in the F₂ hybrid populations of adult males combined, but none of the tests were significant (χ² test at P < 0.001).

DISCUSSION

This study investigated the genetic basis of F₂ hybrid breakdown in the parasitic wasp genus Nasonia. Hybrid breakdown contributes to the integrity of diverged genomes from closely related and occasionally hybridizing species and thus may play an important role in speciation. The results in this investigation revealed that cytonuclear genic incompatibilities contribute significantly to hybrid breakdown in male F₂ hybrids of N. giraulti and N. vitripennis. MTRD, the preferential inheritance of one parental allele, was found to be a direct consequence of cytonuclear genic incompatibility. Not all cytonuclear genic incompatibilities in F₂ hybrids of N. giraulti and N. vitripennis appear to manifest in MTRD, however.

MTRD in male F₂ hybrids of N. giraulti and N. vitripennis was first reported by Gadau et al. (1999). Since the authors considered only one direction of the cross (i.e., N. giraulti [v] ♀ × N. vitripennis ♂), it remained unclear whether or not the phenomenon is restricted to F₂ hybrids with N. vitripennis cytoplasm. Gadau et al. (1999) furthermore used a N. giraulti introgression strain with N. vitripennis cytoplasm (R16A) to conduct their cross experiment. It was therefore possible that MTRD occurs exclusively in combination with the R16A strain. Our data revealed that this is not the case: MTRD is also observed in adult male F₂ hybrids when the N. vitripennis and N. giraulti inbred strains AsymCX and RV2X(U) are crossed. Also, MTRD is not restricted to hybrids with N. vitripennis cytoplasm but it is found in adult F₂ hybrid males with N. giraulti cytoplasm as well (Figures 3 and 4).

The estimated number and position of TRDL required to explain the biased recovery rate of parental alleles of markers in the investigated adult F₂ hybrid males differed significantly between the reciprocal crosses. While we found evidence for only one TRDL on chromosome 4 in hybrids with N. giraulti cytoplasm, at least three TRDL, one each on chromosomes 1, 2, and 5, had to be assumed in hybrids with N. vitripennis cytoplasm. Since we used highly inbred strains to conduct the cross experiments, the F₁ hybrid females from reciprocal crosses had identical nuclear genomes. There are no sex chromosomes in haplodiploid wasps. The reciprocal crosses thus differed only in their cytoplasm. The significant differences in the recovery rate of parental alleles in the reciprocal crosses therefore indicate that factors in the cytoplasm are pivotal for the MTRD in the adult F₂ hybrid males.

In contrast to the adult F₂ hybrid males, their embryos did not exhibit significant MTRD. This excludes meiotic drive as an explanation for the observed MTRD in the adult hybrid wasps since this process leads to an overrepresentation of certain alleles in the gametes and therefore should already be detectable in the embryos. The fact that we found MTRD in adult F₂ hybrid males but not in their embryos suggests that the MTRD is the result of postzygotic viability differences of the F₂ hybrid males contingent upon their cytoplasm and nuclear TRDL genotypes. All four identified TRDL exhibited a deficit of the paternal allele. This result is consistent with an incompatibility between paternal TRDL alleles and maternal factors in the cytoplasm. A maternal effect seems unlikely since the F₁ hybrid females from the reciprocal crosses had identical nuclear genomes, and maternal-effect genes generally affect early developmental processes in the offspring. The experiments by Breeuwer and Werren (1995) showed that hatching rates of male F₂ hybrids of N. giraulti and N. vitripennis do not significantly differ from those of the parental lineages. Rather, the increased mortality of F₂ hybrid males occurs during their larval development and during metamorphosis in the pupal stage. The asymmetry and direction of MTRD in the reciprocal crosses suggests genic incompatibilities between the nuclear genome and a maternally inherited genetic factor in the cytoplasm, most likely the mitochondrial genome. This interpretation is consistent with the Dobzhansky–Muller model for the evolution of genic incompatibilities (Dobzhansky 1934; Muller 1939, 1940, 1942).

The Dobzhansky–Muller model for the evolution of genic incompatibilities predicts that hybrid incompatibilities are asymmetrical (Muller 1942; Orr 1995; Turelli and Moyle 2007). While, for example, a N. vitripennis allele at locus A might be incompatible with a N. giraulti allele at locus B, the N. vitripennis allele of locus B cannot be incompatible with a N. giraulti allele of locus A (Coyne and Orr 2004). This is because the latter represents an intermediate step in the divergence of the two species. Assuming that the N. giraulti alleles at loci A and B represent the ancestral state, the derived state, i.e., N. vitripennis alleles at loci A and B, would probably not have evolved had the intermediate state not been compatible. In the aforementioned example, one can easily replace “locus A” with one of the identified TRDL in N. giraulti × N. vitripennis F₂ hybrid males with N. vitripennis cytoplasm and “locus B” with a maternally inherited locus of the mitochondrial genome. The resulting pattern predicted by the Dobzhansky–Muller model for genic incompatibilities matches with our results.

If we assume that the observed MTRD is solely due to an increased mortality of individuals with the genotype in deficit, we can infer the mortality rate caused by a TRDL by calculating , where x is the estimated frequency of the parental allele occurring in excess. The survival rate s is then 1 − m. Considering the Bayesian estimates for the MTRD of TRDL (Figure 3B), the mortality rate in F₂ hybrid males with N. vitripennis cytoplasm caused by the TRDL on chromosome 1 is with 95% pP of 15–39%, with the highest pP of 25–26%. The corresponding values for the TRDL on chromosomes 2, 4, and 5 are 9–36% (21–22%), 0–32% (14–15%), and 9–35% (21–22%), respectively (Figure 1). Since tests for significant deviation from the expected distribution of recombinant and nonrecombinant genotypes between markers linked to TRDL on chromosomes 1, 2, and 5 (TRDL1, TRDL2, TRDL5, respectively) in hybrids with N. vitripennis cytoplasm did not indicate TRDL interdependency, the overall mortality rate, m₀, of F₂ hybrid males with N. vitripennis cytoplasm caused by the three TRDL is 1 − [(1 − m_TRDL1) × (1 − m_TRDL2) × (1 − m_TRDL5)]. Using the above-mentioned values, the three TRDL account with 95% pP for an overall mortality of 30–86%, with highest pP of 53–55%. In contrast, the TRDL on chromosome 4 in F₂ hybrid males with a N. giraulti cytoplasm explains 0–32% (95% pP; highest pP of 14–15%) mortality.

Breeuwer and Werren (1995) inferred estimates of the survival rate of male F₂ hybrids of N. giraulti and N. vitripennis. According to their study, F₂ hybrid males with N. vitripennis cytoplasm have a survival rate of ∼47% (±21 SD), while F₂ hybrid males with N. giraulti cytoplasm exhibit a survival rate of ∼18% (±6 SD). A smaller study in our own laboratory revealed almost exactly the same values (results not shown). The mortality rate of F₂ hybrid males with N. vitripennis cytoplasm caused by the three identified TRDL was estimated to be between 30 and 86% (highest pP of 53–55%). Thus, the mortality caused by the three TRDL in F₂ hybrid males with N. vitripennis cytoplasm is sufficient to explain the observed survival rate of 47%. The mortality rate of the single identified TRDL in F₂ hybrid males with N. giraulti cytoplasm was estimated to be between 0 and 32%. This low rate stands in stark contrast to the low survival rate of the F₂ hybrid males with N. giraulti cytoplasm found by Breeuwer and Werren (1995). The lower survival rate of male F₂ hybrids with N. giraulti cytoplasm in comparison to male F₂ hybrids from the reciprocal cross suggests that additional cytonuclear genic incompatibilities account for the increased mortality rate rather than pure nuclear–nuclear genic incompatibilities. There are several possible explanations of why we missed identifying these cytonuclear genic incompatibilities. First, additional TRDL may exist but are in regions of the genome that are not covered by our genome map. The methods applied to estimate the genome coverage of the N. giraulti × N. vitripennis linkage map in this study suggested that the linkage map spans 77% of the total genome and that 79 and 95% of the nuclear genome lay within 10 and 20 cM, respectively, of a mapped marker. However, these estimates had to be inferred under assumptions that the experimental data did not strictly fulfill (e.g., uniform distribution of markers). The values therefore should be interpreted with caution. Second, the increased hybrid mortality may well be caused by TRDL that are in regions covered by our genome map, but the MTRD that they cause might be too small to be detected with the studied sample sizes. And third, the incompatibilities might be more complex and involve additional nuclear loci. Tests for a deviation from the expected distribution of recombinant and nonrecombinant genotypes between markers from different linkage groups in F₂ hybrids with a given cytoplasm did not reveal evidence for three-way interactions. However, three-way interactions, in which the two nuclear loci are on the same chromosome, were not tested. More complex interactions, involving more than two nuclear loci, are imaginable as well.

Gadau et al. (1999) reported that pure nuclear–nuclear genic incompatibilities are involved in N. giraulti [v] × N. vitripennis F₂ hybrid breakdown as well. The authors compared the expected and the observed frequency of recombinant and nonrecombinant genotypes between markers from different chromosomes and found evidence for four possible nuclear–nuclear genic incompatibilities. Using the exact same approach, we failed to find comparable results in our study. Although an initial test of 120 F₂ hybrid males with N. vitripennis cytoplasm revealed notable deviations in five pairs of markers, we were not able to confirm them in a second hybrid population of the same size. The combined analysis of F₂ hybrid males from reciprocal crosses did not provide evidence for possible nuclear–nuclear genic incompatibilities either. In the latter, the sample size was equivalent to the one in the study by Gadau et al. (1999). One important difference between the two studies concerns the used strains. Gadau et al. (1999) used the N. giraulti introgression strain R16A with N. vitripennis cytoplasm for their experiment while we used the cured and highly inbred N. giraulti strain RV2X(U) with N. giraulti cytoplasm. R16A was created by backcrossing males of the N. giraulti strain RV2T for 16 generations to N. vitripennis females of the strain AsymC (Breeuwer and Werren 1995). Thus, although the nuclear genomes of R16A and RV2X(U) go back to RV2T and should theoretically have an identical nuclear genome, we assume that the R16A genome genetically changed due to intergenomic co-adaptation as a response to a heterospecific mitochondrial genome. A direct comparison of the results from these two studies is therefore difficult.

This study revealed that cytonuclear genic incompatibilities contribute significantly to hybrid breakdown in male F₂ hybrids of N. giraulti and N. vitripennis. Which specific genes are incompatible remains to be investigated. Mishmar and Gershoni (2007) recently pointed out that pathways that are vital and undergo tight-and-fast co-evolutionary processes might be of particular significance for speciation. The mitochondrial oxidative phosphorylation (OXPHOS) pathway, the primary energy-generation system in aerobic metazoans, has characteristics that suggest its possible major role in speciation (Mishmar and Gershoni 2007). Four of the five OXPHOS enzyme complexes are composed of nuclear- and mitochondrial-encoded subunits. Disruption of the delicate interactions between the nuclear- and mitochondrial-encoded subunits introduced by hybridization is likely to affect mitochondrial ATP production capacity with far-reaching effects on the viability of the hybrids (Blier et al. 2001; Rand et al. 2004). Since the mitochondrial genome evolves much faster than the nuclear genome (Boore 1999; Lynch et al. 2006), incompatibilities are also expected to evolve rapidly between them (Mishmar and Gershoni 2007). Nuclear genes of the OXPHOS system are therefore promising candidates for being incompatible with a heterospecific mitochondrion. However, so far only a few studies have addressed the possible role of cytonuclear genic incompatibilities in hybrid breakdown (e.g., Hutter 2002, 2007; Rawson and Burton 2002; Sackton et al. 2003; Burton et al. 2006; Ellison and Burton 2006; Harrison and Burton 2006; Willett 2006). The short generation time of Nasonia species, its haplodipolid sex determination, and availability of the genome sequence makes Nasonia a suitable model system for studying these interactions in further detail.