Reproductive isolation caused by azoospermia in sterile male hybrids of Drosophila

Abstract Recently diverged populations in the early stages of speciation offer an opportunity to understand mechanisms of isolation and their relative contributions. Drosophila willistoni is a tropical species with broad distribution from Argentina to the southern United States, including the Caribbean islands. A postzygotic barrier between northern populations (North America, Central America, and the northern Caribbean islands) and southern populations (South American and the southern Caribbean islands) has been recently documented and used to propose the existence of two different subspecies. Here, we identify premating isolation between populations regardless of their subspecies status. We find no evidence of postmating prezygotic isolation and proceeded to characterize hybrid male sterility between the subspecies. Sterile male hybrids transfer an ejaculate that is devoid of sperm but causes elongation and expansion of the female uterus. In sterile male hybrids, bulging of the seminal vesicle appears to impede the movement of the sperm toward the sperm pump, where sperm normally mixes with accessory gland products. Our results highlight a unique form of hybrid male sterility in Drosophila that is driven by a mechanical impediment to transfer sperm rather than by an abnormality of the sperm itself. Interestingly, this form of sterility is reminiscent of a form of infertility (azoospermia) that is caused by lack of sperm in the semen due to blockages that impede the sperm from reaching the ejaculate.


Different types of barriers can be critical to speciation. In
Drosophila, studies on the rate at which different barriers evolve have shown that, on average, prezygotic isolation evolves faster than postzygotic isolation (Coyne & Orr, 1989, 1997, with premating barriers evolving faster than postmating prezygotic and postzygotic isolation being even slower (Turissini, McGirr, Patel, David, & Matute, 2018). However, the average rate of evolution of such barriers among species is not necessarily indicative that premating mechanisms are always more relevant in establishing isolation. For example, among Hawaiian species of Drosophila, the strength of premating versus postmating barriers can be dependent on sympatry versus allopatry status of the species (Carson, Kaneshiro, & Val, 1989;Kaneshiro, 1976;Kang, Garner, Price, & Michalak, 2017).
Among populations of Drosophila montana, there is evidence that premating mechanisms contribute to isolation, but premating isolation increases with distance between populations, while postmating isolation is independent of distance, suggesting its important role in the early stages of speciation (Garlovski & Snook, 2018). While mechanisms of isolation have been studied extensively, they have been most commonly studied using species in which isolation is already fully established, thus making it difficult to differentiate between barriers that might evolve postspeciation from those that might have contributed to reduce gene flow in early stages of speciation. The identification of isolating barriers among diverging populations or partially isolated subspecies that have not yet reached a full-species status can help address questions on the role of different isolating mechanisms in speciation. Moreover, it has become increasingly evident that proper identification of the speciation phenotype aids in understanding not only the speciation process but also its genetic basis (Mullen & Shaw, 2014). In turn, fine phenotypic characterization is crucial to functionally annotate genes.
Drosophila willistoni is a nonhuman commensal that uses flowers and fruits as substrates (Markow & O'Grady, 2008). The species was once believed to continuously spread from the southern United States to South America (D. w. willistoni), with a different subspecies (D. w. quechua) restricted to the west of the Andes in a narrow geographical area near Lima, Peru. It has been recently found that D. w. willistoni is subdivided into two partially isolated populations (subspecies) that are reproductively isolated from each other: D. w. willistoni in North America, Central America, and northern Caribbean islands, and D.
w. winge in South America and southern Caribbean islands (Mardiros et al., 2016). When a female of D. w. willistoni mates with a male of D.
w. winge, the resulting males are sterile, but the females are fertile. In the reciprocal cross, all offspring are fertile. It has also been previously determined that copulation duration is similar for sterile hybrid males and parental species and that the external male genitalia show no differences between the subspecies. Further, examination of the internal genitalia found no evidence of major atrophy in the hybrids relative to parental species, and the sterile males produced motile sperm but failed to place sperm within the female reproductive storage organs after mating (Civetta & Gaudreau, 2015). Whether hybrid male sterility due to failure to transfer sperm is unique to the two populations previously assayed (Civetta & Gaudreau, 2015) remains unclear.
Moreover, we lack clear phenotypic characterization of what causes sterile male hybrids' failure to transfer sperm and whether any form of assortative mating, or postmating prezygotic incompatibility, reduces gene flow between these two different subspecies.
Here, we use strains derived from different populations of the two subspecies (i.e., D. w. willistoni: Guadeloupe, Puerto Rico, and D. w. winge: Uruguay and Saint Vincent). We found assortative mating among individuals of the same populations and no evidence of noncompetitive postmating prezygotic isolation. Using a series of interrupted mating assays to track the fate of sperm and ejaculate of sterile male hybrids, we find that the sterile males manage to transfer an ejaculate that triggers the expected responses of elongation and expansion of the female uterus. However, the ejaculate is devoid of sperm. We identify a large mass forming a bulge at the basal end of the testes (i.e., the seminal vesicle) in sterile males that appears to impede the movement of the sperm toward the sperm pump, where sperm normally mixes with secretions produced by the accessory glands to produce the ejaculate. This mechanical impediment to transfer sperm represents a novel form of hybrid male sterility in Drosophila.  (Mardiros et al., 2016).

| Drosophila stocks and maintenance
Throughout the experiments, flies were kept in either 8 oz.
Emerging adult flies used in mating experiments, in interrupted mating assays or to produce hybrids, were collected under light CO 2 anesthesia as newly emerged every 4 hr to ensure virginity.
Virgin females and naïve males were separated, maintained at a density of 20 to 30 flies per vial, and aged for 5 to 6 days posteclosion before being used.

| Premating isolation and fecundity
We measure premating isolation among strains of the same subspecies (i.e., wil(G) × wil(P); win(S) × win(U)) as well as different subspecies (i.e., wil(G) × win(U); wil(P) × win(S)) using multiple-choice mating experiments (Jennings et al., 2011). Virgin females and naïve males from different strains were transferred without anesthesia from collection vials into vials with either red-or blue-colored food and allowed to feed overnight. The dyes were alternated to account for possible dye effects. Flies were then placed together in bottles containing CYMA food supplemented with yeast in groups of 30 males and 30 females per strain for a total of 120 flies. Mating pairs were observed in the morning until half of all possible matings had occurred, but for no longer than an hour, mating pairs were removed and identified based on the color of their abdomen (Casares et al., 1998;Gilbert & Starmer, 1985). The experiments were run over 3-4 replicates (different days) of each cross. We an- We tested fecundity of crosses between individuals of the same population as well as between individuals of different subspecies.
We followed a protocol described in Gomes and Civetta (2014).
Briefly, five 5-to 6-day-old naïve males and virgin females were placed together for 48 hr in a vial containing CYMA food. Males were removed after 48 hr and females transferred to a fresh vial five days after the initial setup. Females were discarded after 5 days and progeny counted from both vials 23 days after the initial setup. Each cross was replicated at least 5 times.

| Interrupted matings
Virgin male and female pairs of the same strain/population were aspirated into vials without anesthetization. Similarly, sterile hybrid males were paired with females of each population used to generate them.
Vials containing a single male and female pair were observed continuously for a period of 3-4 hr. Mating pairs were stopped by freezing at either 2 or 6 min into copulation and stored frozen for later dissection. The frozen couples were retrieved from the freezer and allowed to briefly thaw over the course of a few minutes. All dissections were done in a drop of 1 × phosphate-buffered saline (PBS). The frozen copulating pair was gently separated using forceps and the male was checked for an ejaculate mass on the tip of its aedeagus, in case the separation had pulled it from the female, before being discarded. The removal of the female reproductive tract was done following a protocol described by Adams and Wolfner (2007). Briefly, forceps and pins were used to separate cuticular tissue and open the abdomen. Once the uterus was visible, we gently removed it from the abdomen without disturbing the ejaculate when present. Pins were used to clean excess tissue; no coverslips were placed over samples, and images were captured using an inverted Olympus CKX41 microscope. The samples were checked for morphological shifts in the female's uterus and the presence of a darker mass that denoted an ejaculate. When present, ejaculates were removed from the female's uterus using dissecting pins and placed on a drop of NucBlue Fixed Cell DAPI stain (Thermo Fisher Scientific). A coverslip was placed over the drop containing the ejaculate and incubated in the dark for 30 min. The samples were observed for the presence of sperm under both phase contrast and UV light using a Zeiss AX10 microscope.

| Testes' measures of sterile and fertile males
Males from each strain as well as sterile hybrid males were collected and kept in vials containing CYAM medium with no more than 20 males per vial to avoid crowding. The males were aged to either 5 or 10 days old before being frozen to preserve them. Frozen males were allowed to briefly thaw over the course of a few minutes and dissected in a drop of 1 × PBS. Forceps were used to grip the thorax, while the other pinched the male's abdomen just above the aedeagus. Then, by simply pulling on the gripped abdomen the entire male's reproductive tract was pulled out into a drop of 1 × PBS. The testes were isolated using dissecting pins and moved onto a fresh drop of 1 × PBS. Testes' images were captured using an inverted Olympus CKX41 microscope, and the image processing and analysis software Image J (https://imagej.nih.gov/ij/) was used to measure an area toward the basal end of the testes where mature sperm is found and which we refer to as the seminal vesicle. The area can be approximately identified by an apical pinch followed by a basal pinch or torsion. One testis from each male was measured.

| Statistical analysis
Data from the interrupted matings were analyzed using Fisher's exact test on nominal variables (i.e., presence vs. absence). One-way analysis of variance (ANOVA) was used to compare differences in fecundity and on area measures of testes. If significant differences were detected by ANOVA, Scheffe's post hoc test was used to determine whether the variance was statistically significant between specific samples (e.g., Drosophila strains). All analysis was done using SPSS software.

| Positive assortative mating among populations and no evidence of noncompetitive postmating prezygotic isolation
We found significant deviation from random mating when we fit a GLMM with replicas as a random variable and analyzed the data by grouping assays using strains of different subspecies (heterotypic: wil(P) × win(S) and wil(G) × win(U); p < .001) or by grouping assays that used strains of the same subspecies (homotypic: win(S) × win(U) and wil(G) × wil(P); p < .001). Similarly, significant results were found when the analysis was conducted independently for each pair of strains, except for the homotypic cross wil(G) × wil(P) ( Table 1) Table 1). Altogether, premating isolation does not appear to be driven by the subspecies status of the strains assayed but rather by its population origin.
We compared overall fecundity of females crossed to males of the same population to estimates from crosses between individuals of different subspecies. This is a noncompetitive setting, as females were not offered an opportunity to doubly mate with both males of the same population and of a different population or subspecies. We found significant differences among crosses in fecundity (F 7,37 = 9.8; p < .001). Crosses among individuals of the same populations were nonsignificantly different from crosses between individuals of different subspecies (Figure 1).

| Sterile hybrids and fertile males transfer seminal products that trigger morphological changes in the uterus
Interrupted copulations showed significant differences between males from parental population and sterile male hybrids in proportion of seminal fluid masses present at two minutes interruptions (parentals = 87.5%; sterile males = 33.3%; Fisher's exact test p = 1.3 × 10 −8 ), but by six minutes, the differences were nonsignificant (parentals = 97.7%; sterile males = 86.5%; Fisher's exact test p = .068) ( Table 2). This result shows that sterile males take longer, but ef-

| Failure to transfer sperm by sterile hybrids is caused by a testis's blockage at its basal end
Visual inspection of the testes identified a region toward its basal end that looked enlarged and more oval in sterile hybrid males

| D ISCUSS I ON
We found evidence of deviation from random mating for both homotypic and heterotypic crosses and premating isolation among populations of D. willistoni that is not determined by the subspecies status. Given the costly consequences of producing sterile male hybrids in crosses among populations of different subspecies, we expected premating barriers might exist due to selection against maladaptive hybridization in heterotypic crosses (Coyne & Orr, 2004). However, the occurrence of premating isolation in crosses involving populations of the same subspecies is rather unexpected. While it is possible that premating reproductive barriers to gene flow are important among populations of D. willistoni regardless of their subspecies status, a caveat is our small sample size and the use of laboratory strains that do not allow us to determine to what extent the partial but significant levels of isolation have been a consequence of laboratory conditions. While we consider it unlikely, if the levels of premating isolation we detected among populations arose or became stronger in laboratory stocks then clearly this form of isolation is not an important contributor to reproductive isolation between subspecies.
If positive assortative mating among males and females of the same populations truly reflects a condition found in natural populations, then premating isolation is clearly important, but it is not a fixed condition. Contrary to premating isolation, postmating postzygotic isolation (i.e., unidirectional male hybrid sterility) is a fixed condition between subspecies (i.e., D. w. willinstoni and D. w. winge) that is unlikely to have been created in laboratory settings for two reasons: (a) The isolation mechanism is fixed in a pattern that is geographi-  (Cordeiro & Winge, 1995;Dobzhansky, 1975).
An interesting observation regarding the levels of premating isolation detected between populations is the significant isolation between geographically distant populations (i.e., Saint Vincent and Uruguay) but not between geographically closer populations of the same subspecies (i.e., Guadeloupe and Puerto Rico). Isolation by distance rather than between geographically closer populations suggests that allopatry might have facilitated the evolution of positive assortative mating among populations, rather than premating barriers being reinforced upon possible secondary contacts among more geographically closer populations. This pattern of increased isolation by distance is preliminary given the small sample sizes but reminiscent of observations made in other populations of Drosophila (Garlovski & Snook, 2018;Jennings et al., 2014) and, in our case, suggests that postmating isolation might be particularly important as a barrier during early divergence of these two subspecies of D. willistoni.
We have shown that sterile males manage to trigger changes in the morphology of the female's uterus that is not different from the changes induced by fertile males. While the transfer of the seminal fluids seems to be slower in sterile males, their ability to cause the tially diverged between these subspecies. Therefore, we cannot rule out the possibility that competitive postmating prezygotic isolation (e.g., conspecific sperm precedence) might exist between these two subspecies.
Problems in sperm transfer have been reported before, for example, Drosophila simulans females mate for a shorter period of time with Drosophila sechellia males than with conspecific males and very few sperm are transferred (Price, Kim, Gronlund, & Coyne, 2001 Thus, hybrid male sterility at early stages of speciation caused by blockage originating during tissue development rather than divergence of the germ line is unexpected. The type of failure to transfer sperm due to a blockage (azoospermia) we report for flies is reminiscent of cases of sterility in humans (Jarvi et al., 2010)  It is feasible that the origin of the enlargement in sterile hybrids be a consequence of subtle abnormalities at or around the time when the seminal vesicles are formed. Given that smooth musculature grows over the testes, it is possible that the musculature layer of the vas deferens grows, but the interior lumen is hindered by a defect in the seminal vesicle.
Here, we have shown that noncompetitive postmating prezygotic isolation is not a barrier to hybridization between D. w. willistoni and D. w. winge, but incomplete premating isolation is detectable among populations regardless of subspecies status. We have characterized a unique form of hybrid male sterility that involves an impediment of the male's ability to transfer sperm. Detail characterization of the "speciation phenotype" is crucial in guiding future attempts to understand its genetic basis.

ACK N OWLED G M ENTS
We would like to thank Dr. Bahar Patlar for her assistance with statistical data analysis. HD and NS were partially funded by NSERC USRA scholarships. This work was supported by an NSERC Discovery Grant to AC.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.