Incompatibility between mitochondrial and nuclear genomes during oogenesis results in ovarian failure and embryonic lethality

Mitochondria are essential organelles that produce ATP through oxidative phosphorylation and participate in a number of other cellular processes (Cloonan and Choi, 2013; Tait and Green, 2010; Vyas et al., 2016). Although mitochondria have their own genomes, the vast majority of mitochondrial proteins are encoded in the nucleus (∼1500) (Meisinger et al., 2008). Studies in a number of organisms have shown that incompatibility between mitochondrial and nuclear genomes can have deleterious effects, and can contribute to reproductive isolation between populations (Burton and Barreto, 2012; Gibson et al., 2013; Hoekstra et al., 2013; Lamelza and Ailion, 2017; Ma et al., 2016; Narbonne et al., 2012; Sloan et al., 2017; Spirek et al., 2014). However, in only a few cases have the specific genes responsible for mito-nuclear incompatibility been identified (Chou et al., 2010; Lee et al., 2008; Meiklejohn et al., 2013; Singh and Brown, 1991; Spirek et al., 2014). It is known that mitochondrial dysfunction can severely compromise female fertility, and that maternal inheritance of sub-functional mitochondria can reduce embryonic survival (Bentov et al., 2011; Demain et al., 2016; Ge et al., 2012; Tilly and Sinclair, 2013). An important unresolved issue is how incompatible interactions between specific alleles of mitochondrial and nuclear genes contribute to female reproductive failure.

We and others have previously described a specific mito-nuclear incompatibility between alleles from two closely related Drosophila species (Holmbeck et al., 2015; Meiklejohn et al., 2013; Montooth et al., 2010). A strain with the mitochondrial genome from D. simulans strain w⁵⁰¹ (simw⁵⁰¹) and with the nuclear genome from the D. melanogaster strain Oregon-R, hereafter denoted as (simw⁵⁰¹); OreR, had delayed development, disrupted larval metabolic rate, compromised locomotion, bristle defects and reduced fecundity (Hoekstra et al., 2013; Montooth et al., 2010; Meiklejohn et al., 2013). These phenotypes were most severe at higher temperatures, similar to other strains with compromised mitochondrial function, which is consistent with metabolism being a temperature-sensitive process (Clarke and Fraser, 2004; Ghosh et al., 2013; Hoekstra et al., 2013). In contrast, a strain with the same D. simulans mitochondria but with the D. melanogaster Austria nuclear genome, (simw⁵⁰¹); AutW132, was phenotypically normal during larval development. Also phenotypically normal were combinations of the Oregon-R mitochondria (ore) with either D. melanogaster Austria or Oregon-R nuclear genomes, hereafter (ore); AutW132 and (ore); OreR. The molecular basis for the mito-nuclear incompatibility in the (simw⁵⁰¹); OreR strain was shown to be an allelic interaction between the mt-tRNA^tyr encoded in the simw⁵⁰¹ mitochondria and the mt-tRNA^tyr synthetase (Aatm) encoded by the D. melanogaster OreR nuclear genome (Meiklejohn et al., 2013). The simw⁵⁰¹ mitochondrial mt-tRNA^tyr polymorphism changes a G:C to G:U in the stem of the tRNA anticodon arm, and the OreR nuclear polymorphism changes a highly conserved alanine to valine at position 275 next to the synthetase ATP binding pocket Aatm^A275V. Consistent with the predicted effects of this interaction on mitochondrial protein translation, this incompatibility decreases oxidative phosphorylation activity specifically for only those complexes that require mitochondrial-translated proteins (Meiklejohn et al., 2013).

The incompatible (simw⁵⁰¹); OreR strain serves as a model for deciphering how mitochondrial dysfunction contributes to human disease. Mutation of the human ortholog of Aatm, YARS2, as well as other mt-tRNA synthetases cause a spectrum of heritable diseases (Riley et al., 2010). A current challenge is to understand why mutations in different mt-tRNA synthetases result in different clinical presentations (Jiang et al., 2016; Konovalova and Tyynismaa, 2013). The results from the (simw⁵⁰¹); OreR strain are consistent with the idea that mito-nuclear incompatibility in specific individuals may contribute to the variability in clinical phenotypes. Moreover, the reduced fecundity of the (simw⁵⁰¹); OreR flies suggest the mito-nuclear incompatibility may impair gametogenesis and embryonic survival, but the cellular basis for this reproductive failure has not been investigated.

In this study, we investigate the impact of mito-nuclear incompatibility on Drosophila oogenesis and female fertility. Each Drosophila ovary is composed of ∼16-20 ovarioles, which contain an array of progressively more mature egg chambers (Bilder and Haigo, 2012; Lin and Spradling, 1993) (Fig. 1A,B). Egg chambers are composed of one oocyte and 15 germline sister nurse cells surrounded by an epithelial sheet of somatic follicle cells (Fig. 1B). These cells are descendants of germline stem cells (GSCs) and somatic follicle stem cells (FSCs) that reside in the germarium at the tip of the ovariole (Fig. 1B). Egg chambers are formed and bud off from the germarium as the transit-amplifying FSC daughter cells surround the germline cells. These egg chambers then migrate posteriorly down the ovariole as they mature through 14 morphologically defined stages (Fig. 1B). During early oogenesis, mitochondria greatly increase in number, with some transported into the oocyte, while others remain in nurse cells and are rapidly transferred into the oocyte during later oogenesis (Cox and Spradling, 2003; Hill et al., 2014). In fly strains that are heteroplasmic for different mtDNA haplotypes, the events of early oogenesis are associated with selection and inheritance of functional mitochondria (Ma et al., 2014). Similar mitochondrial proliferation, transport and selection also occur during mouse oogenesis, and defects in these processes can negatively impact oogenesis and embryonic survival in both fly and mouse (Cox and Spradling, 2003; Lei and Spradling, 2016; Mishra and Chan, 2014; Pepling, 2016; Van Blerkom, 2011).

Here, we find that mito-nuclear incompatibility during oogenesis has pleiotropic cell and developmental consequences that compromise egg production and embryonic survival. Overall, the results provide a cellular basis for how mito-nuclear incompatibility can reduce organismal fitness and potentially contribute to reproductive barriers. More broadly, our findings are relevant to understanding the impact of mito-nuclear incompatibility on human female infertility, inter-generational inheritance of metabolic defects, and mitochondrial replacement therapy.

Previous results suggested that the mito-nuclear incompatible (simw⁵⁰¹); OreR strain had reduced fecundity relative to the other mito-nuclear combinations, which was more severe at the non-permissive temperature of 28°C (Hoekstra et al., 2013; Meiklejohn et al., 2013). To specifically evaluate the contribution of female infertility to this reduced fecundity, and to eliminate the contribution of male sterility, we outcrossed (simw⁵⁰¹); OreR females to males from a yellow (y) white (w) lab strain at 25°C and 28°C. Fecundity was measured by counting the number of pupae in the next generation, and compared with the number of offspring from y w females and females from strains previously shown to have compatible mito-nuclear combinations – (simw⁵⁰¹); AutW132, (ore); AutW132 and (ore); OreR, all of which were outcrossed to y w males (Table 1). The number of offspring was not significantly different among the (simw⁵⁰¹); AutW132, (ore); AutW132 and y w females at either 25°C or 28°C (Table 1). Whereas the (ore); OreR females had significantly fewer offspring than y w females at 25°C and 28°C, they had significantly more offspring than (simw⁵⁰¹); OreR females at 28°C, which were completely infertile at this temperature (Table 1). These results indicate that the OreR nuclear genotype contributes to a reduced female fecundity, which is further enhanced by the simw⁵⁰¹ mitochondrial genotype and higher temperature.

We evaluated whether the low fecundity of the (simw⁵⁰¹); OreR and (ore); OreR females was because of a reduced egg production. We raised mito-nuclear females at 25°C or 28°C, crossed them to y w males and then measured egg lay rate (oviposition) on different days of adulthood (Fig. 1C). At 25°C, all four types of mito-nuclear females had a similar oviposition rate over 3-12 days of adulthood, suggesting this temperature is permissive for egg production (Fig. 1D). This result differs from those of Meikeljohn et al., who reported that (simw⁵⁰¹); OreR and (ore); OreR have lower oviposition rates relative to (simw⁵⁰¹); AutW132 and (ore); AutW132 at 25°C, perhaps because we measured oviposition rate over 1 h in the morning, whereas Meikeljohn et al. calculated oviposition over an entire day (Meiklejohn et al., 2013). In contrast, at 28°C the egg production from both (ore); OreR and (simw⁵⁰¹); OreR was very low, especially in older females, and was significantly different from both (simw⁵⁰¹); AutW132 and (ore); AutW132 (Fig. 1E). The low oviposition rate of both the (ore); OreR and (simw⁵⁰¹); OreR females relative to the AutW132 nuclear strains suggests that the OreR nuclear genotype contributes to a temperature-sensitive decline in egg production. Moreover, the significantly lower egg lay rate of (simw⁵⁰¹); OreR compared with (ore); OreR in younger females, and the lack of an effect of the simw⁵⁰¹ mitochondria in the AutW132 nuclear background, suggests that the mito-nuclear incompatibility of (simw⁵⁰¹); OreR further compromises egg production.

To define the problem with egg production in (ore); OreR and (simw⁵⁰¹); OreR, we first examined the gross morphology of the ovaries and ovarioles in the adult females on day three of adulthood (Fig. 1A,B). Ovariole development starts during the early pupal stage when somatic cells form a stem cell niche around germ cells, which are destined to become adult germline stem cells (GSCs) (Gancz et al., 2011). The adult ovaries of (ore); OreR and (simw⁵⁰¹); OreR females were smaller than those of (simw⁵⁰¹); AutW132 or (ore); AutW132 (Fig. S1A) and were composed of significantly fewer ovarioles, a phenotype that was exacerbated at 28°C and most severe in the (simw⁵⁰¹); OreR strain (Fig. S1B). These results suggest that (ore); OreR and (simw⁵⁰¹); OreR females have temperature-sensitive defects in gonadogenesis, resulting in a reduced adult ovariole number that contributes to the lower egg production rates.

We next examined stages of oogenesis in adult females by confocal microscopy of ovarioles labeled with the fluorescent DNA dye DAPI. When females were raised at 25°C, on day three of adulthood all four strains had normal germaria and distributions of egg chambers representing different stages of oogenesis (Fig. 2A-D, Fig. S2A). When raised at 28°C, the (simw⁵⁰¹); AutW132 and (ore); AutW132 females again had normal germaria and stages of oogenesis (Fig. 2E,F, Figs S2B and S3A-B″). In contrast, in the (ore); OreR and (simw⁵⁰¹); OreR females raised at 28°C, ∼70-80% of the ovarioles were missing egg chambers from early and mid-stages of oogenesis (Fig. 2G,H, Figs S2B and S3C-D″). It is known that in response to metabolic and other stresses a vitellogenic checkpoint results in reduced egg chamber production and autophagy of egg chambers during stages 7-9 (Pritchett et al., 2009). Some ovarioles in both (ore); OreR and (simw⁵⁰¹); OreR had degenerating egg chambers that were labeled with Lysotracker, indicating that they were undergoing autophagy (data not shown). Unlike the vitellogenic checkpoint, however, these degenerating chambers were also seen during earlier stages of oogenesis, including stage 1. In many ovarioles, stages 1-7 were completely absent and germaria were directly attached to stage 8 or later egg chambers, with no evidence of nascent stage 1 egg chambers (germarium region 3) (Fig. 2G,H, Fig. S3C-D″). This last phenotype suggests that in the (ore); OreR and (simw⁵⁰¹); OreR females after an initial period of normal oogenesis egg chamber production from the germarium ceases in many ovarioles. These failures in oogenesis are consistent with the observed temperature-sensitive decline in egg oviposition rate in the (ore); OreR and (simw⁵⁰¹); OreR females (Fig. 1E).

To further define the defect in egg chamber production, we examined the earliest stages of oogenesis in the germarium. To measure germline stem cell (GSC) number, we labeled ovaries with antibodies against the germline protein Vasa and the Hu li tai shao (Hts) protein, the Drosophila adducin ortholog that associates with a spherical ‘spectrosome’ in the cytoplasm of GSCs (Lin et al., 1994). In a normal ovariole, anti-Hts labels spectrosomes in two or more GSCs that reside in the stem cell niche at the anterior tip of the germarium (Losick et al., 2011). The GSC daughter cells, called cystocytes, divide four times with incomplete cytokinesis to form interconnected germline cysts. These cysts contain a branched cytoskeletal body known as the fusome, which also labels with anti-Hts (Lin et al., 1994). Most germaria from all four mito-nuclear strains raised at 25°C had the normal two or more GSCs (Fig. 2I). When raised at 28°C, most germaria from (simw⁵⁰¹); AutW132 and (ore); AutW132 females also had two or more GSCs (Fig. 2J, Fig. S3A-B″). In contrast, many germaria from the (simw⁵⁰¹); OreR and (ore); OreR at 28°C had only one or zero GSCs and a reduced number of cystocytes, a temperature-dependent reduction of early germline cells that was significantly more severe in the (simw⁵⁰¹); OreR strain (Fig. 2J-M, Fig. S3C-D″). Labeling of germaria with an antibody against the cleaved Caspase Dcp-1 indicated that GSCs and their daughter cystocytes were undergoing programmed cell death (PCD) at an elevated rate in the (simw⁵⁰¹); OreR and (ore); OreR females specifically at 28°C (Fig. 2N-P, Fig. S2C-D) (McCall and Peterson, 2004). The (simw⁵⁰¹); OreR females had the highest frequency of germline PCD, consistent with them having the lowest GSC number at 28°C (Fig. 2J, Figs S2D and S3D-D″). TUNEL labeling confirmed that GSCs and cystocytes were indeed undergoing PCD (data not shown). These results suggest that death of GSCs and cystocytes contributes to the temperature-dependent decline in egg chamber production in the (simw⁵⁰¹); OreR and (ore); OreR females. The much lower frequency of this phenotype in the (simw⁵⁰¹); AutW132 and (ore); AutW132 strains suggests that the OreR nuclear genotype contributes to early germline cell death, which is further exacerbated by the mito-nuclear incompatibility in the (simw⁵⁰¹); OreR strain.

Many of the egg chambers produced by (ore); OreR and (simw⁵⁰¹); OreR females had more than the normal 15 germline cells (Fig. S3E-G). Given that there were not more, and in fact fewer, cystocytes per cyst in the germarium, the >15 germline cells per egg chamber is not the result of extra cystocyte divisions. Instead, this phenotype is likely the result of a follicle cell epithelium encapsulating two germline cysts during the formation of a single stage 1 egg chamber. This phenotype is similar to other mutants that have a deficit of transit-amplifying follicle cells in the germarium, suggesting that (ore); OreR and (simw⁵⁰¹); OreR have problems with early follicle cell proliferation or survival, or both (Cicek et al., 2016; Forbes et al., 1996).

The data suggested that (simw⁵⁰¹); OreR and (ore); OreR females share similar temperature-sensitive early oogenesis phenotypes. Progeny counts had indicated, however, that whereas (ore); OreR females had reduced number of progeny at 28°C, (simw⁵⁰¹); OreR females were completely infertile (Table 1). Therefore, the similar ovary phenotypes in these two strains failed to account for the difference in their female fecundity. Moreover, while progressive GSC loss was somewhat more severe in the (simw⁵⁰¹); OreR strain, these females did produce some normal-looking egg chambers, yet no progeny survived to the pupal stage. We therefore examined whether a difference in the survival of offspring from (simw⁵⁰¹); OreR and (ore); OreR females accounts for their difference in fecundity.

As before, we raised females from the four mito-nuclear strains at either 25°C or 28°C (the G0 generation) and then crossed them to y w males. We collected embryos from these females during different days of adulthood at 25°C or 28°C, allowed these G1 embryos to develop at the same temperature, and counted the fraction that hatched into larvae. When G0 mothers were raised at 25°C, G1 embryos from all four mito-nuclear strains had similar hatch rates (Fig. 3A). When mothers were raised at 28°C, however, the G1 embryos from (simw⁵⁰¹); OreR mothers had hatch rates that were significantly lower than those in the other three strains (Fig. 3B). To assess whether this is a temperature-sensitive process in the mother or embryo, or both, we again raised females at 25°C or 28°C, but this time shifted their embryos to the reciprocal temperature after a 1 h egg lay. When mothers were raised at 25°C and embryogenesis was at 28°C, all four strains had similar high hatch rates (Fig. 3C). When mothers were raised at 28°C and embryogenesis was at 25°C, however, the G1 embryos from (simw⁵⁰¹); OreR mothers again had hatch rates that was significantly lower than the other three mito-nuclear strains (Fig. 3D). These results suggest that the hatch rate of embryos is dependent on the temperature of the (simw⁵⁰¹); OreR mothers.

Although the embryos from these crosses inherit the simw⁵⁰¹ mitochondria from their mother, they are heterozygous for Aatm^V/Aatm^A because their y w fathers are homozygous for the compatible Aatm^A allele (Table 2). The incompatible Aatm^V allele has previously been shown to be recessive to the compatible Aatm^A allele for larval development. Nevertheless, in our experiments inheritance of the Aatm^A allele from the y w father failed to rescue the maternal effect on embryonic hatch rate. We could not cross to (simw⁵⁰¹); OreR males because they had reduced fertility at 28°C. Therefore, to directly assess whether embryos with a (simw⁵⁰¹); OreR (Aatm^V/Aatm^V) genotype are temperature sensitive, we outcrossed (simw⁵⁰¹); OreR females to fertile (ore); OreR males, which have the Aatm^V allele (Fig. S4A). The hatch rate of homozygous Aatm^V/Aatm^V embryos was not affected by the developmental temperature, but was dependent on the developmental temperature of the (simw⁵⁰¹); OreR mother (Fig. S4B). These results suggest that the incompatible (simw⁵⁰¹); OreR females have no surviving progeny because of a strict maternal effect that is primarily dependent on the genotype and temperature of the mother.

We next addressed whether the incompatibility between the nuclear Aatm^V and mitochondrial tRNA^Tyr (G:U) alleles in (simw⁵⁰¹); OreR was the cause of the maternal effect. To do this, we conducted a series of rescue crosses using strains with different Aatm transgenes on the third chromosome that have either the incompatible OreR allele (Aatm^V), the compatible AutW132 allele (Aatm^A) or the OreR Aatm^V allele mutated to the compatible Aatm^A allele (Aatm^V275A) (Fig. S5) (Meiklejohn et al., 2013). These rescue strains were heterozygous on the second chromosome for a deletion of the endogenous Aatm gene over the CyO balancer, which has the compatible Aatm^A allele (Meiklejohn et al., 2013). This resulted in progeny females that had different doses of the Aatm^A, Aatm^V or Aatm^V275A alleles and either ore or simw⁵⁰¹ mitochondria. We then tested these females (the G0 mothers) for the maternal effect on embryonic hatch rate (Fig. S5). We could not obtain adult females that only had the Aatm^V allele with the simw⁵⁰¹ mitochondria when raised continuously at 28°C because of insufficient rescue of larval development. Therefore, for all the crosses, we shifted larvae from 25°C to 28°C at second instar to increase survival and obtain adult females. These G0 adult females were crossed to y w males, allowed to lay eggs for 1 h at 28°C, and their embryos shifted to 25°C to specifically assay the maternal effect. The results indicated that only one copy of the Aatm^A allele or Aatm^V275A allele was sufficient to rescue the temperature-sensitive maternal effect on embryonic hatch rate (Fig. 3E). Conversely, mothers that inherited only the Aatm^V allele with simw⁵⁰¹ mitochondria displayed a strong temperature-sensitive maternal effect on hatch rate (Fig. 3E). Thus, it appears that, similar to the previous results for larval development, the compatible Aatm^A is dominant to the incompatible Aatm^V for the maternal effect on embryogenesis. These rescue results suggest that it is the incompatibility between the nuclear Aatm^V and mitochondrial tRNA^Tyr (G:U) alleles that is responsible for the temperature-sensitive maternal effect on offspring survival.

To define the temperature-sensitive period for the (simw⁵⁰¹); OreR maternal effect, we performed reciprocal temperature-shift experiments. The G0 females from all four mito-nuclear strains began development at either 25°C or 28°C, but were then shifted to the reciprocal temperature at different times of their larval, pupal or adult life (Fig. 4A). The resulting adult G0 females were crossed to y w males, allowed to lay eggs on day five of adulthood, and the hatch rate of their G1 embryos measured. Embryonic hatch rate from (simw⁵⁰¹); OreR mothers was significantly lower than the other strains when these females were shifted from permissive (25°C) to restrictive (28°C) temperature before early pupal development (Fig. 4B). The reciprocal shift from restrictive (28°C) to permissive (25°C) resulted in lower embryonic hatch rates when the (simw⁵⁰¹); OreR mothers were shifted after larval development (Fig. 4C). Together, these reciprocal shift experiments suggest that the temperature-sensitive period of the (simw⁵⁰¹); OreR maternal effect begins during early pupal development, a period that coincides with ovariole morphogenesis and the onset of oogenesis (Gancz et al., 2011).

Curiously, shifting the temperature of (simw⁵⁰¹); OreR females during adulthood did not alter the maternal effect on hatch rate (Fig. 4B,C). However, this maternal effect was assayed only on day five of adulthood. Therefore, to further investigate the dynamics of the temperature-sensitive maternal effect, we performed reciprocal temperature-shift experiments with females on day one of adulthood, and then measured hatch rate of their embryos from day three to 12 of adulthood (Fig. 5A). As expected, the (simw⁵⁰¹); OreR females that were shifted from 25°C to 28°C initially laid embryos with a very high hatch rate (Fig. 5B). This hatch rate then gradually declined to significantly lower levels by day five of adulthood, and continued to decline to very low levels until day 12 (Fig. 5B). Conversely, embryos from mothers shifted from 28°C to 25°C initially had a very low hatch rate. Despite shifting these mothers to permissive temperature on day one of adulthood, this low hatch rate persisted until day eight, with hatch rates rising only by day nine (Fig. 5B).

The nine-day lag between the temperature shift and the rise in hatch rate is approximately equal to the time it takes in oogenesis for a GSC daughter cell in the germarium to develop into a mature stage 14 egg (Fig. 1B) (Lin and Spradling, 1993). This similarity in timing suggested that the temperature-sensitive period for the mito-nuclear incompatibility may coincide with events of early oogenesis in the germarium. Indeed, early oogenesis is a crucial time for mitochondrial dynamics when germline mitochondria undergo a period of rapid proliferation and increase greatly in number, with some mitochondria actively transported into the differentiating oocyte to form part of a distinct cluster called the Balbiani body (Cox and Spradling, 2003; Hill et al., 2014; Lei and Spradling, 2016; Pepling, 2016; Pepling et al., 2007). To address whether these processes were affected, we labeled mitochondria with antibodies against ATP5a, a subunit of the F1 ATP synthase complex, and examined them by confocal and super-resolution microscopy (Godbout et al., 1993). This analysis suggested that the (ore); OreR and (simw⁵⁰¹); OreR females at 28°C have normal mitochondrial number in the germarium (Fig. 6A,B). In many (simw⁵⁰¹); OreR germaria, however, mitochondria formed a more tubular network, although this was difficult to quantify (Fig. 6B). Mitochondrial labeling of (simw⁵⁰¹); OreR was similar to all the other mito-nuclear strains in both somatic and germline cells of later stage egg chambers (Fig. 6C-H). Thus, the combined data suggest that mito-nuclear incompatibility does not result in observable reductions in the number of mitochondria in developing egg chambers. It remains possible, however, that early oogenesis is a temperature-sensitive period for mitochondrial quality that later manifests as reduced embryonic hatch rate, consistent with previous evidence that the mito-nuclear incompatibility in (simw⁵⁰¹); OreR results in reduced mitochondrial oxidative phosphorylation capacity (Meiklejohn et al., 2013).

We next examined the eggs laid by (simw⁵⁰¹); OreR mothers to determine how maternal mito-nuclear incompatibility inhibits embryonic hatch rate. At 28°C, (simw⁵⁰¹); OreR mothers laid eggs that were smaller with shells of unusual morphology (Fig. 7A-C, Fig. S6A-C). The A-P axis length of these eggs was significantly shorter than those from other mito-nuclear mothers by ∼20%, without an increase in egg width, indicating that eggs laid by (simw⁵⁰¹); OreR have reduced volume (Fig. 7D,E). This reduction in egg length without a significant change in width is similar to other mutants that have incomplete ‘dumping’ of nurse cell cytoplasm into the oocyte during late oogenesis (Bilder and Haigo, 2012; Cooley et al., 1992). An examination of late stages of oogenesis indicated that the (simw⁵⁰¹); OreR females did indeed have a very high fraction of egg chambers with incomplete nurse cell dumping (Fig. 7F). In addition, many of the eggs laid by (simw⁵⁰¹); OreR mothers had soft, gelatinous eggshells, a structure that is synthesized by somatic follicle cells. These follicle cells developmentally amplify the copy number of eggshell protein (Chorion) genes late in oogenesis (Calvi, 2006; Calvi et al., 1998; Spradling and Mahowald, 1980). However, labeling of ovaries with the nucleotide analog EdU indicated that developmental amplification of chorion genes was not impaired, suggesting that problems with eggshell synthesis are downstream of developmental gene amplification (Fig. S6D-F) (Calvi and Lilly, 2004; Calvi et al., 1998). Many of these eggs had unusual branching morphology of their eggshell dorsal appendages, with some of them wrapping laterally around the egg. Labeling of egg chambers with antibodies against the dorsal determinant Gurken, however, did not provide evidence for a disruption of D-V patterning (Fig. S6G-I). Together, these phenotypes suggest that some of the (simw⁵⁰¹); OreR female infertility may be caused by defects in both germline nurse cells and somatic follicle cells.

To gain further insight into the (simw⁵⁰¹); OreR maternal effect, we next examined stages of embryogenesis. To focus specifically on the maternal effect, mothers were raised at 28°C, mated to y w males, and the embryos then allowed to develop at 25°C. Many of the eggs from (simw⁵⁰¹); OreR mothers appeared unfertilized. To quantify sperm entry into the oocyte, we labeled with antibodies against the paternally-supplied centrosomal protein Centrosomin (Cnn) (Eisman et al., 2015; Megraw et al., 1999). This analysis revealed that at 28°C ∼36% of eggs from (simw⁵⁰¹); OreR mothers were unfertilized, a fraction that was significantly higher than the 5% unfertilized fraction produced by control y w mothers (Fig. 7G). Although the fraction of unfertilized eggs from (ore); OreR mothers at 28°C was also elevated (∼17%), this number was not significantly different from y w (Fig. 7G). Sperm enter the egg through a hollow cone called the micropyle, a specialized structure of the vitelline membrane (Nonidez, 1920; Suzanne et al., 2001). For some eggs from (simw⁵⁰¹); OreR mothers, the micropyles were occluded by aberrant eggshell structures, at least partially explaining the fertilization defect (Fig. S6J-L).

Although 64% of embryos from (simw⁵⁰¹); OreR mothers were fertilized, hatch rate was close to zero. For y w embryos, the duration of embryogenesis was ∼22 h with hatch rates of over 90%, whereas almost all embryos from (simw⁵⁰¹); OreR mothers failed to hatch even after 36 h. While embryos arrested at distributed stages, ∼70% of fertilized embryos had defects in syncytial nuclear divisions during the first 2 h of embryogenesis (Fig. 7H,K). These embryos frequently had unevenly spaced nuclei and bridging/lagging mitotic chromosomes, indicative of syncytial nuclear positioning and division defects (Fig. 7K). Taken together, these data suggest that mito-nuclear incompatibility in the (simw⁵⁰¹); OreR mothers results in severe pleiotropic effects on egg morphology, fertilization and embryonic cell divisions.

In this study, we show that a specific mito-nuclear incompatibility results in a suite of cell and developmental phenotypes that contribute to complete temperature-sensitive infertility of (simw⁵⁰¹); OreR females. The data indicate that both (ore); OreR and (simw⁵⁰¹); OreR have germline cell death and a decline in egg production with female age, suggesting that the OreR nuclear genotype contributes to reduced female fecundity (Hoekstra et al., 2013; Meiklejohn et al., 2013; Montooth et al., 2010). These phenotypes were more severe for (simw⁵⁰¹); OreR than (ore); OreR, and were absent in (simw⁵⁰¹); AutW132, suggesting that mito-nuclear incompatibility enhances this ovarian failure. The allelic mito-nuclear incompatibility in (simw⁵⁰¹); OreR strain caused a unique temperature-sensitive maternal effect that resulted in complete infertility. Our results provide a cell and developmental framework for understanding how mito-nuclear incompatibility can contribute to reproductive barriers among divergent populations and cause human disease.

The (ore); OreR and (simw⁵⁰¹); OreR females both had a temperature-sensitive reduction in ovariole number and a progressive ovarian failure caused by egg chamber degeneration and programmed cell death of germline cells in the germarium, suggesting that the OreR nuclear genotype contributes to these phenotypes (Fig. 8). It has been previously reported that dividing germline cystocytes undergo a Caspase-dependent cell death in response to metabolic stresses (Drummond-Barbosa and Spradling, 2001; Ikeya et al., 2002; Morrison and Spradling, 2008). Activated Caspase labeling of GSCs was surprising, however, given that it has been reported that GSCs and other stem cells are resistant to apoptosis after irradiation (Xing et al., 2015). It is possible that mitochondrial stress is a more potent inducer of apoptosis in stem cells than the genotoxic stress caused by irradiation. Given that ovarioles form around presumptive GSCs during early pupal stages of development, the reduction in ovariole number suggests that GSCs may also be lost before adulthood (Belles and Piulachs, 2015; Gilboa and Lehmann, 2006; Song et al., 2002).

The similar phenotypes of (simw⁵⁰¹); OreR and (ore); OreR in the ovary contrast with their different phenotypes in larvae reported by earlier studies, which showed that (simw⁵⁰¹); OreR, but not (ore); OreR, has temperature-sensitive defects during larval development (Hoekstra et al., 2013; Meiklejohn et al., 2013). The main effect of the OreR nuclear genome to disrupt oogenesis, but not larval development, is consistent with the known tissue-specific characteristics of other metabolic dysfunctions in development and disease, and the sensitivity of oogenesis to metabolic stress, including that caused by mitochondrial dysfunction (Benkhalifa et al., 2014; Bentov et al., 2011; Boots et al., 2016; Grindler and Moley, 2013; McCall, 2004; Morrison and Spradling, 2008; Sieber et al., 2016). Although our rescue experiments showed that the allelic incompatibility between the Aatm^V allele and simw⁵⁰¹ mitochondria causes the maternal effect embryonic lethality, preliminary observations suggest that the ovarian failure phenotype is multigenic and not just determined by alleles of Aatm. Changing the OreR nuclear background through rescue crosses ameliorated the ovarian failure phenotype. Females that inherited the simw⁵⁰¹ mitochondria and only the incompatible Aatm^V allele still had an ovarian failure phenotype, but this phenotype was less severe than in the parental (simw⁵⁰¹); OreR (Aatm^V/V) females. The rescue-cross females with only the incompatible Aatm^V allele, but with ore mitochondria, did not have an ovarian failure phenotype, consistent with the interpretation that mito-nuclear incompatibility enhances ovarian failure. Moreover, whereas (simw⁵⁰¹); OreR and (ore); OreR had similar early oogenesis phenotypes, these phenotypes were more severe in (simw⁵⁰¹); OreR, and absent in (simw⁵⁰¹); AutW132, suggesting that they are enhanced by the mito-nuclear incompatibility. The degree to which alleles of Aatm and other genes contribute to the OreR nuclear main effect on ovarian failure is a complex genetic issue that will be addressed by future mapping studies.

Importantly, the (simw⁵⁰¹); OreR females had the unique temperature-sensitive phenotypes of abnormal egg morphology, fertilization failure and maternal-effect embryonic lethality (Fig. 8). The embryos from these females had aberrant nuclear cleavage division and severe chromosome segregation errors during the first 2 h of embryogenesis. This embryonic lethality was a strict maternal effect dependent on the temperature of the mother, which rescue experiments suggest is caused by incompatibility between the nuclear Aatm^V and mitochondrial tRNA^tyr G:U alleles. The temperature-shift experiments suggested that early oogenesis is a temperature-sensitive period, which is known to be a crucial time for mitochondrial proliferation, transport and selection, and alleles of mitochondrial proteins can even influence early oocyte specification (Cox and Spradling, 2003; Hill et al., 2014; Hurd et al., 2016; Ma et al., 2014; Teixeira et al., 2015). Our analysis, however, did not reveal defects in mitochondrial number, transport or oocyte specification. It is likely, therefore, that this mito-nuclear incompatibility impairs mitochondrial quality, consistent with previous evidence that (simw⁵⁰¹); OreR mitochondria have reduced oxidative phosphorylation complex activity and an aberrant morphology in muscle (Holmbeck et al., 2015; Meiklejohn et al., 2013). Thus, temperature-sensitive events during early oogenesis in (simw⁵⁰¹); OreR mothers may result in persistent sub-functional mitochondria that cannot support energy-demanding processes during later oogenesis and early embryogenesis (Sieber and Spradling, 2015; Sieber et al., 2016).

The aberrant eggshell synthesis suggests that some of the processes affected by this incompatibility are in somatic follicle cells. The fertilization failure of these eggs may also be the result of aberrant eggshell and vitelline membrane synthesis by follicle cells (Fig. 8). The incomplete transfer of nurse cell cytoplasm and short egg phenotype resembles that of other ‘dumpless’ mutants, and suggests that mito-nuclear incompatibility also impairs germline functions (Bilder and Haigo, 2012; Frydman and Spradling, 2001; Mahajan-Miklos and Cooley, 1994). Rapid transfer of nurse cell cytoplasm depends on actin polymerization and contraction, and is sensitive to ATP/ADP ratio, perhaps explaining why it is sensitive to mitochondrial dysfunction (Huelsmann et al., 2013). A recent report indicates that compromised function of the muscular ovariole sheath can also affect nurse cell dumping and results in short eggs with reduced ooplasm (Andersen and Horne-Badovinac, 2016). Thus, it is also possible that the reduced nurse cell dumping is a non-autonomous effect of an ovariole sheath myopathy.

The mito-nuclear incompatibility in the (simw⁵⁰¹); OreR mothers resulted in an inter-generational lethality of their embryos (Fig. 8). Many embryos had unevenly spaced nuclei and aberrant mitotic chromosome segregation during the first 2 h of embryogenesis. Nuclear spacing and migration is ATP dependent and is enacted by aster microtubules and actin cytoskeleton (Telley et al., 2012). Thus, this energy-demanding process may be sensitive to cellular energy deficits, similar to actin-based nurse cell dumping and muscle function. These aberrant nuclear division and spacing phenotypes may be the result of maternal inheritance of sub-functional mitochondria or insufficient maternally-supplied metabolites. Although we favor the interpretation that the ovarian and embryonic phenotypes are caused by mitochondrial failure in somatic and germline cells of the ovary, it is also possible that they are influenced by non-autonomous effects caused by perturbation of female physiology and neuro-endocrine axis.

Our data are relevant to how mito-nuclear incompatibility can contribute to reproductive barriers between species (Burton et al., 2006; Chang et al., 2015; Gibson et al., 2013; Ma et al., 2016; Meiklejohn et al., 2013; Narbonne et al., 2012; Paliwal et al., 2014; Spirek et al., 2014). Our results reveal the cellular and developmental bases for the severe effects of this genomic incompatibility on female fertility in a high-temperature environment – a higher-order gene×gene×environment (G×G×E) effect, and suggest that it has the potential to strongly contribute to reproductive barriers among different populations via pleiotropic effects on female fertility (Hoekstra et al., 2013). Although our data show in general how mito-nuclear incompatibilities can generate barriers to reproduction, it is clear that nuclear-nuclear incompatibilities are the major barrier to reproduction between D. simulans and D. melanogaster (Ferree and Barbash, 2009; Phadnis et al., 2015). A recent report indicated that incompatibility between nuclear and mitochondrial genomes can also compromise female fertility in mouse, although the allelic basis for this incompatibility is not known (Ma et al., 2016). Together with our findings, these data suggest that mito-nuclear interactions may contribute to female infertility in humans.

Mutations in the human ortholog of Aatm, YARS2, cause myopathy, anemia and optic neuropathy (Jiang et al., 2016; Jordanova et al., 2006; Riley et al., 2010), and it has recently been reported that incompatibility between alleles of YARS2 and a specific mitochondrial haplotype worsens Leber's hereditary optic neuropathy (Jiang et al., 2016; Konovalova and Tyynismaa, 2013). Mutation of the human mitochondrial histidine tRNA synthetase HARS2 is known to cause ovarian dysgenesis, but an effect of YARS2 on female fertility has not been reported. Our results suggest that polymorphisms in YARS2 may have undocumented effects on human female fertility that may depend on mitochondrial haplotype, and emphasize that mito- nuclear interactions should be considered when interpreting mechanisms of mitochondrial disease (Lu et al., 2015; Storkebaum et al., 2009). A careful consideration of possible mito-nuclear interactions is also important for informing best practices for mitochondrial replacement therapies that introduce nuclei from two parents into the oocyte cytoplasm of a donor female (Craven et al., 2010; Kang et al., 2016).

The Drosophila mito-nuclear strains and rescue strains used in this study were previously described and are available in Table S1. The control strain, y w^67c23, was obtained from the Bloomington Stock Center and sequenced to confirm that it is homozygous for the Aatm^A allele.

Female fecundity in Table 1 was assayed by crossing five virgin females to three y w males over 24 h at either 25°C or 28°C. After removing the adults, the vials were kept at the same temperature and the number of pupae were counted on day 7. Ten biological replicates were conducted for each genotype and temperature.

For oviposition and ovary analyses of Figs 1, 2 and Figs S1, S2, flies were allowed to lay eggs in vials for 6 h at 25°C. After the adults were removed, the vials were either kept at 25°C or shifted to 28°C. The offspring in these vials develop into the G0 mothers that were subsequently assayed. Adult G0 mothers were kept at either 25°C or 28°C, mated to y w males, conditioned on wet yeast, and their oviposition rate and ovaries analyzed beginning on day 3 of adulthood. The ovarian failure phenotypes in Figs 2, 6 and 7, Figs S1, S2, S3 and S6 were analyzed on day 3 of adulthood. For oviposition rate, 50 newly emerged females were mated to 25 y w or (ore); OreR males, conditioned for 2 days on grape plates with wet yeast. Oviposition rate was measured over 1 h in the morning and flies were transferred to fresh yeast daily.

The G0 mothers were crossed to y w males and the hatch rate of the G1 embryos was analyzed at 36 h after egg lay (Fig. 3). To assay maternal-embryonic temperature sensitivity, G0 mothers were allowed to lay eggs for one hour, and then those G1 offspring were kept at the same temperature or shifted to the reciprocal temperature.

In the reciprocal temperature-shift experiments in Fig. 4, the G0 females were allowed to develop at 25°C or 28°C until they reached the indicated developmental stages and were then shifted to the reciprocal temperature, followed by measurement of hatch rate of their embryos as described. For rescue crosses, (ore); OreR and (simw⁵⁰¹); OreR virgin females were crossed to the males from thee different rescue strains. The G0 females were raised at 25°C until second-instar larvae and were then shifted to 28°C for further development.

For Fig. 5, the G0 adult females were mated to y w males and the temperature was shifted on day one of adulthood, followed by measurement of G1 hatch rate beginning on day 3.

For embryo size measurements, embryos at 0∼2 h of development were dechorionated and measured along their A-P and D-V axis using Openlab imaging software (Fig. 7). Nurse cell dumping was quantified by estimating the A-P axis length ratio of nurse cell compartment to oocyte in stage 12-14 egg chambers.

Ovaries and embryos were fixed and labeled as previously described (Calvi and Lilly, 2004). The following antibodies and concentrations were used: mouse anti-Hts [1:20 dilution, Developmental Studies Hybridoma Bank (DHSB)]; rabbit anti-cleaved Dcp-1 (1:100 dilution, Cell Signaling Technology, 9578); mouse anti-ATP5a (1:100 dilution, Abcam, ab14748); rat anti-Vasa (1:100 dilution, DHSB); and mouse anti-Gurken (2 µg/ml, Hybridoma bank). EdU labeling of amplicon foci was carried out as previously described (Calvi and Lilly, 2004; Paranjape and Calvi, 2016).

Secondary antibodies were Alexa Fluor 488-conjugated goat anti-mouse IgG (4 μg/ml, Invitrogen, 169425); Alexa Fluor 488-conjugated goat anti-rabbit IgG (4 μg/ml, Invitrogen, 1705869); Alexa Fluor 568-conjugated goat anti-mouse IgG (4 μg/ml, Invitrogen, 1698376); Alexa Fluor 568- conjugated donkey anti-rabbit IgG (4 μg/ml, Life Technologies, 1668655); and FITC-conjugated goat anti-rat IgG (2 μg/ml, Life Technologies, A18866). DNA was labeled with 1 μg/ml DAPI (Invitrogen).

For embryo labeling, 0∼2 h embryos were collected on grape plates and rinsed with 0.01% Triton-X 100, dechorionated with 1:1 mixture of bleach and 0.2% NaCl/ 0.02% Triton-X 100, and then slow fixed using a modification of previous protocols (Sullivan et al., 2000). Briefly, the dechorionated embryos were brushed into a glass vial and then 1 ml heptane was added followed by 1 ml of 3.7% formaldehyde in PEM buffer. After shaking the vial for 15 s, the embryos were allowed to fix at room temperature for 15 min. The bottom layer was removed and 1 ml of methanol was added. After vortexing for 10 min, the embryos that did or did not sink were collected separately and hand devitellinized. We did this because alterations to vitelline membrane synthesis in the incompatible strain made even fertilized embryos partially resistant to methanol devitellinization, and to quantify the fertilization rate independent of devitellinization efficiency. All the embryos were rehydrated with PBTA solution and labeled with rabbit anti-Cnn antibody (1:100, a generous gift from T. Kaufman, Indiana University, Bloomington, USA). The unfertilized embryos were quantified based on lack of Cnn labeled centrosomes.

Ovaries and embryos were mounted in Vectashield and imaged with Leica-SP5 scanning confocal microscope. For live embryo imaging, embryos were rinsed with PBS and mounted on supported slides in PBS and micropyles were imaged by DIC with Leica DMRA2 fluorescence microscope. The germaria mitochondrial morphology images were captured on an OMX super resolution microscope (Applied Precision).

One-way ANOVA was used for comparison of GSC number or GSC death among all four strains on day three of adulthood. One-way ANOVA was also used to compare the four strains for developmental temperature-shift experiments in Fig. 4, and for rescue experiments in Fig. 3E. For Fig. 5, significance was assessed by two-way ANOVA of temperature and age. For all other analyses, two-way ANOVA of female age and genotype was used to compare the four strains. ANOVA was followed by Tukey's or Bonferroni posttests to evaluate additivity of variables (e.g. female age and genotype) and correct for false discovery rate (FDR), respectively. All statistical analysis was carried out using GraphPad Prism5 statistical packages.

We thank R. Eisman and T. Kaufman for advice and the Cnn antibody, and the Bloomington Drosophila Stock Center (BDSC) and Flybase for flies and crucial information. We also thank J. Powers of the IU Light Microscopy Imaging Center (LMIC) for help and advice, and J. Zhu, M Ailion, P Lamelza and members of the Calvi lab for comments on the manuscript.