Geographic patterns of postzygotic isolation between two closely related widespread dung fly species (Sepsis cynipsea and Sepsis neocynipsea; Diptera: Sepsidae)

2.1 Study organism

Sepsis cynipsea is the most abundant sepsid in north-central Europe, where it occurs in sympatry with the rare S. neocynipsea in some mountainous regions such as the Swiss Alps. In North America, where S. cynipsea does not occur, S. neocynipsea occupies a similar warm-adapted temperature niche as S. cynipsea in Europe (Pont & Meier, 2002). The mating system of S. cynipsea has been described in detail (Blanckenhorn, Morf, Mühlhäuser, & Reusch, 1999; Blanckenhorn et al., 2002; Parker, 1972a,b; Puniamoorthy et al., 2009; Ward, 1983; Ward, Hemi, & Rösli, 1992), whereas relatively little is known about S. neocynipsea (Eberhard, 1999; Puniamoorthy et al., 2009; Rohner et al., 2016). More detailed descriptions of the maintenance of the flies in our laboratory and their behaviour can be found in Giesen et al. (2017).

2.2 Crossing scheme

We caught gravid females from six sites (i.e., populations) to establish 5–15 iso-female lines (i.e., offspring of a single field-caught female) per population in the laboratory. Sympatric populations were collected in Switzerland (Zurich, CH1: 47°24′0.60″N, 8°34′23.97″E, sampled in June 2013 (2013.6); and Sörenberg, CH2: 46°49′23.72″N, 8°1′54.59″E, in 2013.7), where both species co-occur. We further obtained S. cynipsea from two parapatric European areas (Ludwigshafen, Germany, EU1: 49°28′41.25″N, 8°22′21.65″E, in 2014.7; Stirling, Scotland, EU2: 56°6′59.47″N, -3°56′12.83″W, in 2013.5), where S. neocynipsea only has been observed in adjacent regions (Ozerov, 2005; Pont & Meier, 2002). The other S. neocynipsea were collected from two allopatric North American populations (Fort Hall, Idaho, NA1: 43°1′59.69″N, −112°26′17.91″W; Lamar Valley, Wyoming, NA2: 44°52′6.67″N, -110°10′28.72″W, both in 2013.6), where S. cynipsea does not exist.

Flies from iso-female lines, which at the time had been kept for at least half a year (ca. eight generations) in the laboratory, were used in conspecific parental crosses within each of the four populations per species as the baseline. They were further used in two replicate reciprocal population crosses of European and North American S. neocynipsea (conspecific allopatric cross-continental crosses). Moreover, reciprocal heterospecific parental crosses of three biogeographic types were set up with two population replicates each: European sympatry, European parapatry and cross-continental allopatry (Table 1). One population replicate consisted of 15–20 replicate crossings of iso-female lines, which were conducted over extended periods of time (weeks to months).

Hybrid F1 and F2 flies for our assessments were generated by randomly combining up to 20 flies of each sex from various iso-female lines of a given population and species to be paired with so treated flies from the other species and sex, always reciprocally. Matings in these settings were thus necessarily heterospecific (no choice). Potentially lower replicate numbers of F₁ and F₂ hybrid offspring were expected due to natural difficulties in obtaining hybrids. For backcrosses, we targeted a sample size of six replicates per pairing, as we set up two reciprocal cross-types (female hybrid with male parental – F₁xP, and female parental with male hybrid – PxF₁) to detect possible sex-specific effects.

We obtained fecundity and egg-to-adult viability measurements for (a) con- and heterospecific parental crossings (P; N ≈ 18 per crossing replicate), (b) F₁ hybrid crossings using the offspring resulting from heterospecific crossings (N ≈ 10 per crossing replicate), (c) F₂ hybrid crossings (N ≈ 6 per crossing replicate) and (d) backcrosses of F₁ hybrid offspring with the parental species (BC; N ≈ 6, per crossing replicate). All crosses were done reciprocally, randomized over extended periods of time.

2.3 Fecundity and egg-to-adult viability assessment

One cross consisted of one female and two males (either con- or heterospecific). This grouping should simulate natural conditions, increasing hybridization probability according to Puniamoorthy (2014) and limiting the probability of total failure due to male sterility. Virgin flies were assigned randomly after adult eclosion to a round 50-ml glass vial containing fresh cow dung as oviposition substrate in a plastic rectangular dish (42 (L) × 21 (W) × 16 (H) mm²) plus some grains of sugar. We scored (a) female adult age when she first laid eggs into the dung [in days], (b) her first clutch size, (c) the number of emerged offspring, plus (d) female and (e) male mating partner head widths as measures of their body size (Blanckenhorn, Reusch, & Mühlhäuser, 1998; Rohner et al., 2016). From these observations, we further derived (f) the probability that females of a given cross-produced eggs (p(eggs)) and (g) offspring (p(offspring)) [both binary: Y/N]. Lastly, we calculated (h) egg-to-adult viability as the proportion of offspring emerging from a clutch. Note that given our set-up with lots of heterospecific crosses, we expected low overall numbers of fertile eggs and offspring numbers.

2.4 Statistical analyses

All traits were analysed separately, with female head width as covariate because body size typically affects fecundity measures (Honek, 1993), in SPSS Statistics version 23 using univariate generalized linear models, with binomial errors when outcome variables were binary or proportional (only significant covariates are reported in the Results). A given trait was analysed as a function of species (pure C: S. cynipsea, pure N: S. neocynipsea; hybrid CN or NC; or, analogously, N_EU, N_USA, hybrids N_EUN_USA and N_USAN_EU – females always named first), biogeographic type nested within species (sympatric versus parapatric versus allopatric), all as fixed factors, and population nested within biogeographic type and species as random effect. Additionally planned baseline comparisons were performed to compare the two parental species (C versus N), the continental S. neocynipsea populations (N_EU versus N_USA), and the direction of heterospecific mating (CN versus NC; N_EUN_USA versus N_USAN_EU). F₁, F₂ hybrids and backcrosses were analysed analogously. Lastly, we compared the (F₁ or F₂) hybrid offspring with their parental conspecific crosses.

3.1 Baseline comparison between S. cynipsea and S. neocynipsea

Most fecundity traits did not differ between the two species (Figure 1; Supporting Information Table S1). Mean egg-to-adult offspring viability was ~30% (χ²(1, 6) = 1.41, p = 0.235, Figure 1e). Females laid their first eggs after ca. 6 days (χ²(1, 6) = 2.04, p = 0.154, Figure 2a), with a mean clutch size of ~45 eggs (χ²(1, 6) = 1.34, p = 0.247, Figure 1b). In total, ca. 68% of all females laid eggs (p(eggs): χ²(1, 6) = 0.04, p = 0.846, Figure 1a). Offspring number was around ~22 individuals (χ²(1, 6) = 0.37, p = 0.542, Figure 1d). The only difference between the species was evident in the probability of producing offspring, which was higher for S. cynipsea (p(offspring) = 0.70 ± 0.13) than for S. neocynipsea (0.54 ± 0.15, χ²(1, 6) = 3.87, p = 0.049, Figure 1c), despite the latter species being larger (neocynipsea females: 0.79 ± 0.01 mm, males: 0.76 ± 0.01 mm; cynipsea females: 0.74 ± 0.01, males: 0.71 ± 0.01; χ²(1, 6) > 7.77, p < 0.001, Figure 2b,c).

3.2 Baseline comparison between continental S. neocynipsea

Differences between European and North American S. neocynipsea were only found in the female's age of first reproduction (χ²(1, 2) = 26.51, p < 0.001), with North American S. neocynipsea females (6.55 d ± 0.69) being older when first laying a clutch than those from Europe (4.91 d ± 0.18), even though females were of the same size (0.79 ± 0.01, χ²(1, 2) = 0.00, p = 0.999). All other traits were similar: both laid around ~45 eggs per clutch (χ²(1, 2) = 0.82, p = 0.365, Figure 3b), with p(eggs) around ~68% (χ²(1, 2) = 0.21, p = 0.646, Figure 3a). From these clutches ~23 offspring emerged (χ²(1, 2) = 0.15, p = 0.702, Figure 3d), p(offspring) being ~55% (χ²(1, 2) = 0.60, p = 0.440, Figure 3c). The egg-to-adult viability found on both continents was ~26% (χ²(1, 2) = 1.10, p = 0.295, Figure 3e).

3.3 Intercontinental S. neocynipsea crosses over three generations

Although no differences in crossing direction across continents were detectable for S. neocynipsea, p(eggs) was higher when females from North America and males from Europe were crossed (~59%) relative to the reciprocal cross (~35%, χ²(1, 2) = 3.13, p = 0.007, Figure 3a). In the intercrossed F₁ generation, only female age at first reproduction differed significantly in crossing direction, showing a difference of 1 day (χ²(1, 2) = 8.33, p = 0.004). Intercontinental F₂ hybrids showed an asymmetry in hybridization direction with crosses of European females and North American males resulting in no eggs and no offspring whatsoever, although the overall sample size was very low (N = 3, Figure 3e). Backcrosses showed no significant differences across all measured traits (Supporting Information Table S1; Figure 3).

3.4 Fitness of heterospecific crosses and their hybrid offspring

In parental heterospecific crosses, the probability of producing eggs was affected by hybridisation direction (CN versus NC), with CN crosses (0.41 ± 0.09) having lower p(eggs) than NC crosses (0.64 ± 0.09, χ²(1, 6) = 8.05, p = 0.005; Figure 1a). Only this trait was also affected by biogeographic type (χ²(4, 6) = 32.96, p < 0.001), with sympatric crosses producing fewest eggs (p(eggs) ~27%), while parapatric (~65%) and allopatric (~70%) crosses produced many more. At the same time, p(eggs) marginally differed between con- (~68%) and heterospecific (~53%) parental crosses (χ²(1, 19) = 2.93, p = 0.087; Figure 1a). The same pairwise comparison revealed expectedly higher egg-to-adult viability in conspecific (~30%) than heterospecific (~4%) crosses (χ²(1, 14) = 50.70, p < 0.001, Figure 1e), as well as corresponding differences in the probability of producing offspring (conspecific ~63%; heterospecific ~8%; χ²(1, 15) = 10.15, p < 0.001; Figure 1c; Supporting Information Table S1).

Crossing of F₁ hybrid offspring produced F₂ hybrid offspring only in the NC direction, with CN crosses generating no offspring in most crosses (χ²(1, 3) = 5.10, p < 0.001, Figure 1d), and with corresponding differences in the probability of producing offspring (Figure 1c) and egg-to-adult viability (Figure 1e). Consequently, egg-to-adult viability from F₁ hybrid crosses was strongly reduced relative to the conspecific parental crosses (χ²(1, 10) = 30.64, p < 0.001, Figure 1e). F₁ hybrid females and males resulting from heterospecific CN and NC crosses differed in size, yielding a significant effect of crossing direction (females: χ²(1, 4) = 4.91, p = 0.027, Figure 2b; males: χ²(1, 4) = 6.56, p = 0.010, Figure 2c). Interestingly, F₂ hybrids of both sexes were significantly smaller than the parental females (females: χ²(1, 6) = 60.11, p < 0.001, Figure 2b; males: χ²(1, 6) = 33.87, p < 0.001, Figure 2c).

We found the same hybridisation asymmetry in the F₂ hybrid generation. Adults emerged only from one of 10 clutches in the CN direction, while p(offspring) ~38% resulted from NC crossings (no statistics calculable due to too many missing values; Figure 1c). This similarly affected offspring number (χ²(1, 2) = 7.58, p = 0.006, Figure 1d) and egg-to-adult viability (no statistics calculable; Figure 1e). Compared to the conspecific crosses (~30%), F₂ hybrid viability was significantly decreased to ~10% (χ²(1, 6) = 4.99, p = 0.025, Figure 1e).

3.5 Haldane's rule in backcrosses of F₁ hybrids with the parental species

To test for sex-specific deleterious hybridization effects, we reciprocally crossed F₁ hybrids with the parental species. Egg-to-adult viability was reduced when hybrid males mated with parental females, but not when hybrid females were mated with parental males (χ²(1, 12) = 67.37, p < 0.001, Figure 1e); egg (χ²(1, 12) = 16.57, p < 0.001, Figure 1b), and offspring numbers including zeroes (χ²(1, 12) = 54.30, p < 0.001, Figure 1d) were similarly decreased for the former cross.

4.1 Baseline comparison of the two species

Our study revealed minor differences in fecundity and fertility traits between S. cynipsea and S. neocynipsea, as could be expected due to their ecological similarities and their relatively low heterospecific genetic differentiation (Pont & Meier, 2002; Su et al., 2008). However, the species exhibit significant differences in body size, S. neocynipsea being larger than S. cynipsea, in agreement with a previous study (Rohner et al., 2016). All other fecundity and fertility traits showed few differences, especially when controlling for the body size differences, which contrasts with the more pronounced but nevertheless merely quantitative (rather than qualitative) differences in mating behaviour between the two species (Giesen et al., 2017). Roughly two-thirds of all females of both species laid a first clutch containing ~45 eggs after ~6 days. Thus, a considerable proportion of females was not able to produce any eggs and therefore might be sterile, again in accordance with previous studies of S. cynipsea (Blanckenhorn et al., 2002; Teuschl & Blanckenhorn, 2007). This might well be a consequence of our protracted laboratory breeding, as freshly caught field females are typically more fertile (~80%–90%; P.T. Rohner, personal communication). Egg-to-adult survival was also low (~30%), again possibly explained by our samples stemming from long-term laboratory iso-female lines. Despite some likely effects of inbreeding across all populations held in the laboratory for multiple generations, this low baseline fertility of our iso-female lines ultimately did not impede our ability to detect differences between con- and heterospecific crosses in the fecundity traits investigated, our main goal here, as these were substantial.

4.2 Heterospecific crosses resulting in F₁ hybrid offspring

As expected, fertility of heterospecific crosses was strongly reduced to roughly 3.5% in both hybridization directions (CN versus NC), documenting clear mating and fertilization incompatibilities between the two species (Turelli, Barton, & Coyne, 2001). This reduction in F₁ viability and fecundity is a typical hybridization pattern evident in several taxa (see Dowling & Secor, 1997; Fitzpatrick, Fordyce, & Gavrilets, 2009) and confirms our expectation of intrinsic postcopulatory incompatibilities between the species effected by processes such as disrupted sperm transfer (Arthur & Dyer, 2015), cryptic female discrimination of heterospecific sperm (Eberhard, 1991) or high mortality of hybrid offspring following zygote formation (Dowling & Secor, 1997; Mayr, 1942). Despite low fertility, egg numbers were similar in heterospecific and conspecific crosses, although the likelihood of producing eggs was significantly lowered for the CN (41%) relative to the NC hybridization direction (64%) and the conspecific crosses (68%). This suggests disruption of egg production by heterospecific males due to postmating, prezygotic barriers, for instance via incompatibilities between sperm and egg surface proteins preventing fertilization (Palumbi, 1999), toxicity of S. neocynipsea sperm for S. cynipsea eggs (Rice, 1996) or male failure to induce proper female ovulation (Sirot, Wong, Chapman, & Wolfner, 2015), all potentially reflecting sexual conflict. Moreover, egg production in the CN direction approached zero in sympatric relative to para- and allopatric crosses. This again supports our conjecture that heterospecific sperm possibly disrupts fertilization, which in sympatry might be reinforced by natural selection due to the continuous contact of the two species (Coyne & Orr, 1989, 2004; Turelli et al., 2001; Yukilevich, 2012), which we also found for some behavioural traits (Giesen et al., 2017).

4.3 Reduced hybrid fitness across generations with F₂ hybrid breakdown

Similar to the results of the heterospecific crosses, intercrossed F₁ and F₂ hybrids had very low egg-to-adult viability. However, at least egg numbers of F₁ hybrid intercrosses were not reduced but in fact somewhat higher than those of the con- and heterospecific crosses (Figures 1b and 3b), suggesting hybrid vigour (Todesco et al., 2016; Wolf, Takebayasi, & Rieseberg, 2001). Hybrid vigour admittedly can be expected when comparing conspecific crosses based on potentially inbred iso-female line populations with the resulting heterospecific crosses. Nevertheless, egg and offspring production was significantly reduced in intercrosses of F₂ hybrids, demonstrating strong hybrid breakdown. Furthermore, emergence of F₂ and F₃ hybrid offspring approached zero, re-emphasizing fertilization difficulties similar to the heterospecific crosses (Dowling & Secor, 1997; Palumbi, 1999; Rice, 1996; Turelli et al., 2001). This again appears to document a reinforcement pattern, as sympatric crosses produced no F3 offspring at all, while some offspring resulted in parapatric crosses. Yet, no F3 offspring were produced in allopatric crosses either, and the overall sample sizes were low at this stage due to intrinsic incompatibilities between the species.

More interestingly, our data demonstrate an asymmetry in hybridization direction (CN versus NC) affecting offspring number and hence fertility of the F₁ generation. If the mating partners stemmed from crosses of S. cynipsea females with S. neocynipsea males, they were not able to produce any adult F₂ offspring, similar to the asymmetry in egg production of the parental generation. As the less abundant species is more likely to mate heterospecifically (Matute, 2014), S. neocynipsea females have a better chance to mate and produce fertile F₁ hybrid offspring with S. cynipsea males than vice versa (Matsuda, Ng, Doi, Kopp, & Tobari, 2009; Sawamura, Sato, Lee, Kamimura, & Matsuda, 2016). This asymmetry was still evident in intercrosses of the F₂ hybrids and explains our low sample sizes for the later hybrid generations, as it was very hard to obtain F₂ and F₃ hybrid offspring.

4.4 Haldane's rule—male sterility in F₁ hybrid offspring

Our study highlights a breakdown of hybrid offspring production between S. cynipsea and S. neocynipsea in terms of decreased egg production and egg-to-adult viability (Figure 1), signifying strong but not absolute suppression of hybridization. In addition, hybrids performed differently depending on sex as predicted by Haldane's rule, according to which the heterogametic sex, here the male, is mostly sterile (Haldane, 1922). This was already visible in the first hybrid generation, but also in the backcrosses of F₁ hybrid offspring with the parental species. Egg-to-adult viability, as well as egg and offspring numbers, revealed a certain degree of sterility when hybrid males were forced to mate with S. cynipsea or S. neocynipsea females, whereas female hybrids showed little to no such effects in the converse situation, clearly demonstrating hybrid female fertility (Figure 1b). Our results thus agree well with Haldane's rule and Darwin's corollary thereof as already documented in several other taxa (Dowling & Secor, 1997; Turelli & Moyle, 2007). However, while fitness of the heterogametic sex is almost entirely suppressed via genetic incompatibilities between hybrids and parental species, the homogametic sex sometimes also shows lowered fitness (e.g., in flycatchers: Alatalo, Eriksson, Gustafsson, & Lundberg, 1990). In this bird example, natural selection acted mainly via sterility of the heterogametic sex (in this case females), while reduced fitness of the homogametic sex (the males) was driven by sexual selection (Qvarnström, Alund, McFarlane, & Sirkiä, 2016). We can conclude that natural selection, either by directly acting on compatibility mechanisms or indirectly on other traits, is the main force driving selection on hybrids via reduced survival, with sexual selection as an additional precopulatory force as demonstrated by Giesen et al. (2017). A next step will be determining the extent of hybridization and its underlying evolutionary forces in nature using genomic tools.

4.5 Comparing the biogeographic cross-types

Strongest reproductive isolation is typically expected in sympatric species pairs following reinforcement and/or some sort of purging of genetic incompatibilities due to natural selection (Coyne & Orr, 2004; Turelli & Moyle, 2007; Via, 2001, 2009). Contrary to a number of precopulatory behavioural traits (Giesen et al., 2017), we obtained only weak evidence for such reinforcement in sympatric areas (the Swiss Alps) for the (postcopulatory) fecundity traits assessed here (egg production only). It thus appears that precopulatory behavioural traits, which are inherently plastic and can potentially evolve fast, are of overriding importance in preventing hybridisation in nature in this and other systems (Eberhard, 1996; Puniamoorthy, 2014; Puniamoorthy et al., 2009; Beysard, Krebs-Wheaton, & Heckel, 2015; Schmidt & Pfennig, 2016).

4.6 Cross-continental S. neocynipsea hybridization

We here also explored differences between continental S. neocynipsea populations featuring some morphological differences (Rohner et al., 2016). Although European and North American populations are not in contact in nature due to their allopatric distribution, they are recognized as the same species based on similar morphology and behaviour (Giesen et al., 2017; Pont & Meier, 2002). However, population genetic differentiation between the continents is substantial (op. cit.), indicating that allopatric speciation is ongoing, mediated by genetic drift and/or adaptation to new habitats, as we found substantial evidence for hybridization barriers (Figure 3) similar to those between the two species (Figure 1). North American S. neocynipsea required more time than European flies to lay their first clutch. This might merely reflect ecological or geographic differences, as body size was controlled for. Egg production and egg-to-adult viability were lower in crosses of North American females with European males (Figure 3). Therefore, sperm from cross-continental males also appears to disturb fertilization in some way (Arthur & Dyer, 2015), and/or incompatibilities are evolving, at least in one direction, perhaps as by-products of other ecological adaptations (Turelli & Moyle, 2007). Intercrossing of F₂ offspring showed signs of hybrid breakdown too, but this was based on only very few replicates.

However, in contrast to the heterospecific situation in Figure 1, backcrosses between hybrid offspring and parental continental S. neocynipsea populations showed few fitness reductions in terms of egg-to-adult viability, such that both hybrid sexes appear to be equally reproductively fertile; genetic incompatibilities according to Haldane's rule were not detected at this level either (Figure 3). Overall, our cross-continental S. neocynipsea crosses indicate some degree of genetic differentiation similar to that between the two species (Figure 1), with associated incompatibilities and fitness decrements following hybridization. These incompatibilities are presumably due to genetic drift alone but potentially may also involve direct or indirect natural or sexual selection (Rohner et al., 2016), as flies from both continents still recognize each other as the same species behaviourally (Giesen et al., 2017).