This article is corrected by:

  1. Errata: ERRATUM Volume 69, Issue 5, 1360, Article first published online: 30 March 2015


Speciation is responsible for the vast diversity of life, and hybrid inviability, by reducing gene flow between populations, is a major contributor to this process. In the parasitoid wasp genus Nasonia, F2 hybrid males of Nasonia vitripennis and Nasonia giraulti experience an increased larval mortality rate relative to the parental species. Previous studies indicated that this increase of mortality is a consequence of incompatibilities between multiple nuclear loci and cytoplasmic factors of the parental species, but could only explain ∼40% of the mortality rate in hybrids with N. giraulti cytoplasm. Here we report a locus on chromosome 5 that can explain the remaining mortality in this cross. We show that hybrid larvae that carry the incompatible allele on chromosome 5 halt growth early in their development and that ∼98% die before they reach adulthood. On the basis of these new findings, we identified a nuclear-encoded OXPHOS gene as a strong candidate for being causally involved in the observed hybrid breakdown, suggesting that the incompatible mitochondrial locus is one of the six mitochondrial-encoded NADH genes. By identifying both genetic and physiological mechanisms that reduce gene flow between species, our results provide valuable and novel insights into the evolutionary dynamics of speciation.

The vast diversity of life on earth is the result of countless episodes of speciation, in which populations diverge and become reproductively isolated from one another. Although successful mating may never take place between many species due to various prezygotic isolating mechanisms, interspecific hybridization has been estimated to occur in up to 10% of known animal species (Mallet 2005; Schwenk et al. 2008). When this happens, reproductive isolation may still occur if the hybrid offspring are sterile or inviable due to intrinsic factors, such as deleterious interactions between genes derived from the two parental species, that is, due to Bateson–Dobzhansky–Muller incompatibilities (Bateson 1909; Dobzhansky 1937; Muller 1942). The case in which the hybrid F1 generation is largely unaffected by these incompatibilities whereas subsequent generations suffer severely from them is referred to as F2 hybrid breakdown (=hybrid breakdown). Intrinsic postzygotic isolation, such as hybrid breakdown, has historically been difficult to study in many traditional model organisms because crosses between species often produce no viable or fertile offspring, hindering genetic analysis of the isolating mechanisms (Coyne and Orr 2004). Despite this difficulty, some progress has been made, and a number of genes have been implicated in such cases of isolation, referred to as speciation genes (see Maheshwari and Barbash 2011, for examples). However, the proximate causes for hybrid breakdown are still poorly understood and more examples in experimentally tractable organisms are needed to determine what general principles might be involved in this process (Maheshwari and Barbash 2011).

Interspecific hybrid breakdown has been well documented within the pteromalid wasp genus Nasonia (Darling and Werren 1990; Breeuwer and Werren 1995; Gadau et al. 1999; Ellison et al. 2008; Niehuis et al. 2008; Koevoets et al. 2011). Nasonia is well suited for genetic studies in the laboratory due to its haplo-diploid genetics, the ease with which virgins of both sexes can be collected before they eclose, and the availability of genomic resources (Werren et al. 2010; Muñoz-Torres et al. 2011). Nasonia vitripennis is the only species of its genus with a cosmopolitan distribution and is widely sympatric with N. giraulti in eastern North America (Darling and Werren 1990). Species-specific strains of Wolbachia endosymbionts prevent gene flow between most known Nasonia species through bidirectional cytoplasmic incompatibility (Breewer and Werren 1990; Bordenstein and Werren 1998; Bordenstein et al. 2001). However, Wolbachia-free strains can be produced from each species through antibiotic treatment and they can then be crossed to produce F1 hybrid females (Breeuwer and Werren 1990). It is important to note that Nasonia males develop parthenogenetically from unfertilized eggs, and thus the first hybrid males are produced in the F2 generation from unfertilized eggs produced by the F1 hybrid females.

Breeuwer and Werren (1995) found that in crosses between Wolbachia-free strains of N. vitripennis and N. giraulti, F1 hybrid females do not suffer any major genic incompatibilities that manifest as sterility or inviability. They discovered, however, that F2 hybrid males suffer from an increased mortality rate during larval development. Mortality ranges between ∼53% and ∼82%, depending on whether the hybrids possess a N. vitripennis or N. giraulti cytoplasm (the two types of hybrids are subsequently denoted GV[V] and VG[G], respectively, with the letter in brackets specifying the origin of the cytoplasm: [G] = N. giraulti and [V] = N. vitripennis). Although it has been known that incompatibilities in GV[V] and VG[G] F2 hybrid males lead to larval mortality, the specific growth rates and developmental trajectories of these larvae have so far not been studied. To gain a better understanding of the developmental and physiological deficiencies that F2 hybrid males suffer from, we examined the number of larvae and their sizes at two stages of development and genotyped them for markers in the region of a newly identified incompatibility locus on the left arm of chromosome 5.

Niehuis et al. (2008) mapped regions of the genome involved in larval mortality in hybrids of N. vitripennis and N. giraulti by searching for loci that exhibit a conspicuous bias in the frequency of the parental alleles in adult wasps. The authors showed that there is significant, cytoplasm-specific marker transmission ratio distortion (MTRD) of nuclear markers in the F2 hybrid adults, but not in hybrid embryos, suggesting that the additional mortality of hybrids occurs during their larval development. They also mapped loci that could explain all of the additional mortality that GV[V] hybrids suffer from, but discovered only one locus that explained mortality in VG[G] hybrids, leaving approximately 60% of the observed mortality in VG[G] hybrids unexplained (Niehuis et al. 2008). One hypothesis put forth to explain this discrepancy was that regions accounting for the remaining observed mortality of VG[G] hybrid larvae lie in parts of the genome that were unmapped at that time. Indeed, the map only covered approximately 70% of the genome (Niehuis et al. 2008). To address this, Niehuis et al. (2010) remapped the Nasonia genome using the newly available genomic resources and found a region of ∼10 Mbp on the left arm of chromosome 5 that was missing from previous genetic maps. Werren et al. (2010) assessed whether this new map revealed any additional regions exhibiting MTRD, by pooling 100 adult F2 hybrid males for each reciprocal cross and hybridizing their DNA onto a competitive genotyping microarray (Desjardin et al. 2013). They found a locus within the new region on the left arm of chromosome 5 that had nearly 100% N. giraulti alleles in VG[G] hybrids, but with a sample size of one (due to the pooling of the samples’ DNA) it was not possible to statistically evaluate the result. In addition to this statistical problem, another shortcoming of this approach was that it did not allow for assessment of the genotype of each marker within a single individual. As a result, it was not possible to use individual recombination events in hybrid males to narrow down the genomic region that harbors the genetic factor causally involved in the mortality. In this investigation, we use a microarray capable of genotyping individual samples across 1536 single nucleotide polymorphism (SNP) markers to statistically assess the strength of MTRD in VG[G] F2 hybrid males and to analyze recombination events in individual hybrids to better map and characterize the region and genes on chromosome 5 that exhibit and cause MTRD (Goldengate Genotyping Assay; Illumina, Inc., San Diego, CA; Niehuis et al. 2010).

A genotyping approach is one method of narrowing down the causal genetic elements underlying mortality. A complementary approach is to determine, a priori, pathways that may be involved in mortality and to identify and investigate candidate genes within this pathway, both in terms of their role in the pathway and their position within the genome. Ellison et al. (2008) showed that adult VG[G] F2 hybrid males exhibit reduced activity of their oxidative phosphorylation (OXPHOS) enzymes relative to that of parental males, making this pathway a strong candidate for the deleterious interactions in these Nasonia hybrids due to genic incompatibility. In addition, incompatibilities of individual components of the OXPHOS pathway could produce a pattern of cytoplasm-specific MTRD, because it is the only pathway to incorporate both nuclear- and mitochondrial-encoded proteins. Using this candidate pathway approach, we hypothesize that the OXPHOS system is involved in hybrid breakdown in Nasonia and we therefore predict that either a single gene or multiple genes belonging to the OXPHOS pathway lie within the above-mentioned region of chromosome 5.

Here we present results demonstrating that VG[G] F2 hybrid male larvae show a genotype-dependent size distribution. We found that larval size is strongly related to the genotype of a small genomic region on the left arm of chromosome 5 and we show that this region contributes to a large proportion of the observed mortality during larval development, resulting in an extreme paucity of N. vitripennis alleles within this region. Finally, we evaluated candidate gene(s) that could be involved in the observed mortality of VG[G] F2 hybrid larvae by studying individual recombination events at the distal-most end of the left arm of chromosome 5 in adult hybrid males, and we propose a nuclear-encoded OXPHOS gene in this region as a likely candidate for the incompatibility.

Materials and Methods


We used two Wolbachia-free Nasonia strains, AsymCX and RV2X(U), in our cross experiments. AsymCX is derived from a wild-type strain of N. vitripennis (LBii or LabII) collected in Leiden, The Netherlands (Breeuwer and Werren 1995). RV2X(U) is derived from a wild-type strain of N. giraulti (RV2) collected in Rochester, New York (Breeuwer and Werren 1995). Both strains were chosen because they are highly inbred, their genomes are sequenced (Werren et al. 2010), and they have been previously shown to suffer from hybrid inviability (Niehuis et al. 2008). These strains, as well as their hybrids, were reared under constant light in an incubator (25°C) on pupae of the flesh fly, Sarcophaga bullata.


We produced F2 hybrid males following the procedures described by Niehuis et al. (2008). Briefly, virgin wasps were collected and allowed to mate with one heterospecific partner, producing two types of F1 hybrid females (i.e., GV[V] and VG[G]). These F1 females were set as virgins on hosts to produce F2 males, which were collected at either their embryo stage (12–16 hours old) or their adult stage (2 days after eclosion). For larval measurements, females were kept individually and were initially given a single host for feeding purposes. After this initial host was discarded, the females were left without a host for 12 hours and were then given hosts in 12-hour intervals interspersed with 12 hours without a host for a maximum of two hosts per female. All samples were stored at −70°C until DNA extraction was performed. Note that we used DNA of embryos from a previous study by Niehuis et al. (2008). See supplementary information for more specific information on sample sizes (File S1).


DNA from larvae and adult wasps was extracted using a Chelex extraction protocol described by Niehuis et al. (2007). Chelex extracted DNA quantity and quality was assessed with a NanoDrop ND-1000 Spectrophotometer (NanoDrop Products, Wilmington, DE). We studied two types of molecular markers: microsatellite and SNP markers. Microsatellite markers were amplified using polymerase chain reaction (PCR) and were designed specifically for this study (File S2). The PCR products were separated and visualized on a Li-Cor 4300 DNA Analyzer (Li-Cor, Lincoln, NE) and analyzed with the SAGA Generation 2 software (Li-Cor). We additionally analyzed 1134 SNP markers spanning the entire genome and described in detail by Niehuis et al. (2010).


Infected hosts were opened either 3 or 6 days post-oviposition (i.e., halfway through or at the end of normal larval development time of Nasonia; Whiting 1967), and all larvae were placed over a size standard and photographed using an 11.2 Color Mosaic digital microscope camera (Diagnostic Instruments Inc., Sterling Heights, MI) connected to a Leica MZ125 stereomicroscope (Leica Microsystems, Heerbrugg, Switzerland). We measured the body width of each larva (Fig. 1) and subsequently pooled larvae based on their size (within each female / host cohort). Larval size differences between hybrids and their parental species were tested using a Wilcoxon signed-rank test with a Bonferroni-corrected critical P-value to reject the null hypothesis of < 0.0042.

Figure 1.

Body sizes of male larvae of Nasonia vitripennis, N. giraulti, and F2 hybrids of these two species with either N. vitripennis or N. giraulti cytoplasm (GV[V] and VG[G], respectively) at day 3 (top) and day 6 (bottom) of their development.


We performed a X2 goodness-of-fit test (Zar 1999) to test whether the segregation ratios of the markers in the dataset were consistent with an expected 1:1 Mendelian segregation ratio of parental alleles. Yates’ correction for continuity was applied (Yates 1934) as well as Bonferroni correction for multiple testing (241 markers on chromosome 5 resulted in a critical value to reject the null hypothesis of P < 0.00021).


To fine-map the genomic region causally involved in hybrid mortality on the left arm of chromosome 5, we analyzed pairs of markers of chromosome 5 and searched for a significant difference between the expected and actually observed number of recombinant adult VG[G] hybrid males. We determined the expected number of recombinant VG[G] hybrid males by accounting for the mortality incurred during development. Thus, our population of 333 adult hybrids are the result of ∼18% survival (Breeuwer and Werren 1995) of an original population of ∼1850 individuals. Based on the genetic distances of markers given by Niehuis et al. 2008, we then calculated the number of recombinant individuals for individual pairs of markers that we would have expected to observe in the population if there was no hybrid-specific mortality (File S3). We focused on two markers, M1 and M3 (File S3), that are separated by ∼3 Mbp to ensure a sufficient number of recombinant individuals to test, although we used an additional 15 SNP markers that span the entire distance (9.49 cM) between M1 and M3 (see File S3). There are two classes of individuals that result from a recombination event between markers M1 and M3. Each class was, a priori, expected to occur at equal frequency. We designate these two classes as M1v:M3g and M1g:M3v, differing in whether the N. vitripennis allele is distal (M1v:M3g) or proximal (M1g:M3v) to the centromere of chromosome 5. We expected that individuals with a recombination between these markers that still possess the N. vitripennis allele of a gene that is causally involved in hybrid mortality will be more likely to die during development, but that individuals with a recombination that brings in the compatible N. giraulti allele will be rescued and will be observed at the expected number of instances. By comparing the observed and expected number of individuals of each recombinant class, we were able to better assess the exact location of the incompatibility locus on the left arm of chromosome 5. We performed a X2 goodness-of-fit test (Zar 1999) to test whether the observed number of recombinant individuals significantly deviated from the expected number. Bonferroni correction for 16 different tests in this interval resulted in a critical value to reject the null hypothesis of P < 0.003. We used this method to reduce the size of region that we subsequently searched for candidate genes using the official gene set for Nasonia (OGS 1.2; Muñoz-Torres et al. 2011).


Candidate genes were analyzed for evidence of selection with the program MEGA5 (Tamura et al. 2011). We aligned the complete coding sequence of candidates (from translation start to stop) from N. vitripennis and N. giraulti and calculated the proportions of both synonymous (dS) and nonsynonymous (dN) substitutions using the Nei–Gojobori model (Nei and Gojobori 1986).



N. vitripennis females produced more larvae per host than N. giraulti or either type of hybrid (VG[G] and GV[V]) when compared at both 3 and 6 days post-oviposition (Table 1). F2 hybrid males with either cytoplasm were significantly smaller than either parental species at both 3 and 6 days post-oviposition (Fig. 1; Wilcoxon signed-rank test, W > 532, P < 0.0042).

Table 1. Number of male offspring per host produced by females of Nasonia vitripennis, N. giraulti, and hybrids of these two species
Species/ hybridAge (days)Average number of offspring (±SD)Sample size (N)
  1. Abbreviations: N. = Nasonia; GV[V] = N. giraulti × N. vitripennis F2 hybrid males with N. vitripennis cytoplasm; VG[G] = N. vitripennis × N. giraulti F2 hybrid males with N. giraulti cytoplasm.

N. vitripennis320.2±3.012
N. giraulti313.9±3.311

At 3 days post-oviposition, the N. giraulti larvae have a unimodal size distribution, whereas the N. vitripennis larvae as well as the F2 hybrid larvae may have bimodal distributions (Fig. 1, top). As these larvae develop from 3 to 6 days, the GV[V] hybrid larvae and the larvae of both parental species continue to grow in size and their distributions appear to shift toward unimodality. Despite this growth, the GV[V] hybrid larvae remain smaller than age-matched larvae of either parental species. In VG[G] F2 hybrid males, however, the two modes of the distribution become more distinct from one another, with the initially larger mode continuing to get larger over time (similar to what is seen in GV[V] F2 hybrids) and the initially smaller mode remaining similar in size (Fig. 1, bottom).


Analysis of SNP markers in 333 adult VG[G] F2 hybrid males confirmed a region of very strong distortion on the left arm of chromosome 5, with the distal-most marker on this chromosome exhibiting less than 5% N. vitripennis alleles in adult hybrids (observed = 14, expected = 165.5; X = 277.4, P < 0.00021; Fig. 2A and File S1). The bias was not detected in embryos of the same cross using five evenly spaced markers spanning the whole chromosome (File S2). Neither embryos nor adults from the reciprocal cross show significant MTRD in this region (Niehuis et al. 2010). However, Niehuis et al. (2008) reported MTRD in GV[V] on the opposite arm of this chromosome that we are able to confirm here as well (File S1).

Figure 2.

(A) Percentage of Nasonia vitripennis alleles along the distal part of the left arm of chromosome 5 in N. vitripennis × N. giraulti F2 hybrid males with N. giraulti cytoplasm ( = VG[G]). Black circles represent the proportion of alleles found in surviving adults and gray boxes represent the proportion of alleles expected to be found in individuals that die during development (i.e., the inverse proportion of the survivors). Triangle = proportion of N. vitripennis alleles in the “large larvae” class (see Fig. 1), diamond = proportion of N. vitripennis alleles in the “small larvae” class. The x-axis shows the distance of the represented SNP markers (in Mbp) from the left end of the chromosome. Bolded lines on x-axis designate positions of markers M1 and M3. (B) Distal part of left arm of chromosome 5. Vertical lines within the chromosome represent map positions 5.01–5.09 defined by Niehuis et al. (2010). Markers M1 and M3 (∼3 Mbp apart) as well as marker M2 are shown as solid lines extending above the chromosome. The two reciprocal recombinant types are represented above (M1g:M3v) and below (M1v:M3g) the chromosome. The survival rate of the specific type of recombinant individuals is given in enclosed boxes as observed percentage of total expected number of individuals of each recombinant type. Symbols in the chromosome represent candidate genes described here (star = Ndufa11) and from Gibson et al. (2010) (square, Atp5i; diamond = Ndufs8; and triangle, Atp5h).


We analyzed MTRD at the distal end of the left arm of chromosome 5 in the small and large class of VG[G] larvae at day 6 post-oviposition. In the class of large larvae, only 5.4% of the individuals carried the N. vitripennis allele at marker M1 (X2-test, P < 0.05; Fig. 2A), a percentage similar to that observed in adult F2 hybrid males. In contrast, in the class of small larvae, 80% of the individuals carried the N. vitripennis allele of this marker (X2-test; P < 0.05; Fig. 2A).


We found no significant difference between the observed and expected number of recombinant individuals (between markers M1 and M3) when the distal-most marker (M1) carried the N. giraulti allele (M1g:M3v; observed = 83, expected = 87.8; X = 0.26, P > 0.003; Fig. 2B). In contrast, we found significantly fewer recombinant individuals than expected when the distal-most marker carried the N. vitripennis allele (M1v:M3g; observed = 2, expected = 87.8; X = 83.8, P < 0.003; Fig. 2B). This pattern held statistically for 13 of the remaining 15 tested pairs of markers (File S3), with the N. giraulti-distal genotype observed at the expected frequency and the N. vitripennis-distal genotype found far less often than expected. We found more individuals of the N. giraulti-distal type than expected with a recombination between marker M1 (Fig. 2B) and a marker ∼1.81 Mbp (∼2.9 cM) closer to the centromere (observed = 45, expected = 27.2; X = 11.65, P < 0.003; File S3). The other marker pair not showing statistical significance was the most distal one of the chromosome, M1 and M2 (Fig. 2B). We found fewer individuals than expected of both recombinant types (M1g:M2v: 1 vs. 2; M1v:M2g: 0 vs. 1). The counts are too low to be statistically compared, but this pattern is exactly what would be expected if the causal gene were located between these two markers as opposed to being more distal than marker M1.


Given that the OXPHOS pathway is likely impaired in hybrids (Ellison et al. 2008), we assessed genes (OGS 1.2) on the left end of chromosome 5 to determine potential candidates for being causally involved in hybrid mortality. Although three nuclear-encoded OXPHOS genes were identified near this end of the chromosome in a previous attempt to identify candidate genes, these do not fall within the most heavily biased region of the chromosome (Gibson et al. 2010). However, we subsequently identified a fourth nuclear-encoded gene (Ndufa11) involved in the OXPHOS pathway and that is located in the region of chromosome 5 that is most strongly affected by MTRD (OGS ID: NV13533; Scaffold14: 121419–122670; File S4). This gene is located ∼120 kb from the left end of the assembled chromosome and is nested between markers M1 and M2 (Fig. 2B). The gene comprises three predicted exons and exhibits a total of 14 substitutions between N. vitripennis and N. giraulti; nine of which result in amino acid changes (Files S5 and S6). On the basis of these data, we estimated a dN / dS ratio of 0.73 between N. vitripennis and N. giraulti, which is considerably higher than that for any other nuclear-encoded OXPHOS gene analyzed to date (previous highest ratio was 0.45 for Atp5h; Gibson et al. 2010).


The evolution and coexistence of species in sympatry depends on the establishment of barriers to gene flow (Mayr 1942). Hybrid incompatibility represents one important postzygotic barrier to gene flow and is the strongest intrinsic barrier acting in hybrids between N. vitripennis and N. giraulti (Breeuwer and Werren 1995). The foremost goal in speciation genetics is to identify the genes and genic interactions that underlie barriers to gene flow, because it allows us to unravel patterns in the genetics of hybrid breakdown (e.g., to identify pathways that might be particularly prone to disruption in hybrids) and to investigate the evolutionary forces that have driven the divergence of genes, ultimately leading to incompatible interactions in hybrids. Taken together, these answers from diverse species groups will help to determine what, if any, commonalities there are in the speciation processes that have led to the diversity of life seen today. Here we have begun answering these questions in Nasonia F2 hybrid males.

To date, the incompatibilities of F2 hybrid males between N. vitripennis and N. giraulti have been inferred to mainly involve mortality during larval development. However, the physical characteristics and genetic makeup of hybrid larvae remained to be investigated. Our result from genotyping larvae confirms that “large” hybrid larvae almost always carry the N. giraulti allele in the candidate region on chromosome 5 and that these individuals tend to survive to adulthood, whereas the small larvae almost always carry the N. vitripennis allele in this region and tend to die before eclosion. These results imply that the incompatibility locus involved in the observed mortality is also very likely to be involved in the observed larval growth. It is not clear whether this locus causes the larvae to stop growing, which in turn prevents them from surviving due to other extrinsic environmental factors, or if it causes both retarded growth and mortality independently of one another. Koevoets et al. (2012) recently showed that F2 hybrid male larvae from crosses between N. vitripennis and N. longicornis (a species closely related to N. giraulti) suffer from an elevated mortality rate and that this mortality rate is further elevated when the hybrids experience temperature stress. This indicates that, at least in that particular cross, the genic incompatibilities can depend on environment. However, it is not known how environmental factors affect the mortality in hybrids of N. vitripennis and N. giraulti. Future studies should determine how the larval size and mortality are connected.

The pattern and frequency of specific recombinant VG[G] F2 hybrids demonstrates the natural rescue effect of a recombination event that replaces the incompatible N. vitripennis allele with the compatible N. giraulti allele on the distal end of the left arm of chromosome 5. This pattern across successively closer marker pairs indicates that the causal locus is located near the distal-most marker that we studied (i.e., M1; Fig. 2B). Interestingly, we observed fewer individuals than expected of both types of recombinants between the two most distal markers and we identified a nuclear-encoded OXPHOS gene, Ndufa11, very near the distal end of chromosome 5, about halfway between these final two markers (Fig. 2B). The pattern of recombination discussed earlier is fully consistent with the hypothesized involvement of this OXPHOS gene in the observed hybrid mortality.

An incompatibility in the OXPHOS pathway could produce a pattern of incompatibilities like that observed in these Nasonia hybrids. As mitochondrial biogenesis proceeds and energy demands increase during early ontogeny, more OXPHOS complexes will be produced using incompatible gene products, which in turn might amplify the effects of the genic incompatibilities as development proceeds, resulting in larval instead of embryonic mortality. These incompatibilities could also potentially explain the larval growth patterns observed because individuals with a reduced capacity to produce cellular energy would likely suffer from delayed development/growth.

Previous work has hypothesized that nuclear-encoded genes that interact with mitochondrial-encoded genes may be subjected to directional selection to compensate for deleterious mutations that occur in the mitochondrial genome (Blier et al. 2001; Rand et al. 2004). This is hypothesized to occur because of the limited ability of selection to act on the mitochondrial genome (due to its small effective population size and lack of recombination) in conjunction with the typically elevated substitution rate of the mitochondrial genome. Indeed, Oliveira et al. (2008) showed that the substitution rate of protein-coding genes of the mitochondrial genome in Nasonia is ∼35 times higher than that of nuclear-encoded genes. Although the dN / dS ratio of 0.73 for Ndufa11 is below the strict cutoff value to infer directional selection (i.e., dN / dS > 1), this elevated ratio could indicate relaxed selective constraint on this gene relative to other genes in the pathway. A signature of relaxed selective constraint acting on a gene is consistent with a hybrid incompatibility gene, because this type of gene would be expected to be more divergent than others in the pathway (thereby increasing chances for being incompatible with other genes in hybrids) while still being constrained in its function within the pathway (i.e., the gene is still essential to the pathway and therefore the genic incompatibilities may lead to defects in its functioning). On the basis of our results, Ndufa11 is a promising candidate for being involved in both the larval mortality and the reduced larval growth of VG[G] F2 hybrid males, although future studies will be required to determine the mechanistic connection between these two phenotypes and to ascertain the specific role of this gene in hybrid breakdown in Nasonia.


This material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No. 2011131209 awarded to BREP. ON acknowledges the Alexander von Humboldt Foundation for a Feodor Lynen postdoctoral research stipend. We thank C. Desjardin for valuable bioinformatic assistance in evaluating scaffolds of the Nasonia genome assembly. The authors have no conflict of interest to declare.