Parasitic fish embryos do a “front-flip” on the yolk to resist expulsion from the host

Significance

Embryos are sometimes considered to be resistant to natural selection. In parasitic embryos, however, the embryo may be under strong selection to survive in the host. In bitterling fish, which parasitize freshwater mussels, a suite of evolutionary adaptations has evolved which allow bitterling embryos to remain safely inside the mussel. We find that these adaptations are coordinated by a “front-flip” of the embryo on the yolk sac. Our data suggest that this front-flip is a modification of gastrulation, and ensures that the embryo, which has the shape of an arrowhead or anchor, is optimally positioned to resist being ejected by the mussel. The bitterling embryo provides an example of how natural selection can modify early development.

Abstract

Embryonic development is often considered shielded from the effects of natural selection, being selected primarily for reliable development. However, embryos sometimes represent virulent parasites, triggering a coevolutionary “arms race” with their host. We have examined embryonic adaptations to a parasitic lifestyle in the bitterling fish. Bitterlings are brood parasites that lay their eggs in the gill chamber of host mussels. Bitterling eggs and embryos have adaptations to resist being flushed out by the mussel. These include a pair of projections from the yolk sac that act as an anchor. Furthermore, bitterling eggs all adopt a head-down position in the mussel gills which further increases their chances of survival. To examine these adaptations in detail, we have studied development in the rosy bitterling (Rhodeus ocellatus) using molecular markers, X-ray tomography, and time-lapse imaging. We describe a suite of developmental adaptations to brood parasitism in this species. We show that the mechanism underlying these adaptions is a modified pattern of blastokinesis—a process unique, among fish, to bitterlings. Tissue movements during blastokinesis cause the embryo to do an extraordinary “front-flip” on the yolk. We suggest that this movement determines the spatial orientation of the other developmental adaptations to parasitism, ensuring that they are optimally positioned to help resist the ejection of the embryo from the mussel. Our study supports the notion that natural selection can drive the evolution of a suite of adaptations, both embryonic and extra-embryonic, via modifications in early development.

Brood parasites pass their parental care to other species. This reproductive strategy has evolved repeatedly among birds such as cuckoos and cowbirds, social insects, and fishes and involves a range of behavioral and morphological adaptations for successfully deceiving and exploiting the host (1). Developmental adaptations to brood parasitism in birds include an increased movement by the embryos that strengthens their musculoskeletal system and thus helps the chicks hatch (2). Among teleosts, brood parasitic cuckoo catfish (Synodontis multipunctatus) develop strong teeth and jaws during late embryogenesis (3) which facilitate the post-hatching predation of host embryos in the pouch of their mouth-brooding hosts. However, the earlier stages of embryonic development in many brood parasites are similar to closely related non-parasitic species (3). This is consistent with the more general hypothesis in evolutionary developmental biology that earlier-to-middle stages of development are more resistant to natural selection than are later stages (reviewed in ref. 4).

Bitterlings (Acheilognathinae) are a monophyletic lineage of cyprinid fishes containing at least 70 species which are all brood parasites of freshwater mussels (Unionida) (5). Female bitterlings have long ovipositors that they use to deposit their eggs into the mussel gills through their exhalant siphon. Males then fertilize the eggs by releasing sperm near the inhalant siphon of the mussel. Bitterling embryos hatch in 1 to 2 d but reside within the mussel gills for several weeks, competing with the host mussel for oxygen and food, and eventually exiting from the mussel as free-living offspring (6). Mussels have evolved responses to mitigate the fitness costs that bitterling embryos impose; notably, they are able to flush bitterling embryos out of the gill chamber with a sudden, high-velocity pulse of water (7).

The rosy bitterling (Rhodeus ocellatus) belongs to a highly derived clade within bitterlings (5). It possesses at least two developmental traits that enable it to effectively parasitize mussels and resist expulsion (8, 9). First, the fertilized egg has a characteristic shape, resembling a light bulb with an enlarged, spherical vegetal hemisphere and a small, tubular animal hemisphere (8, 10, 11), in contrast to the smaller, more spherical eggs of other cyprinids (10). Second, the yolk sac develops a pair of yolk sac extensions (YSEs) immediately after hatching, eventually giving the yolk the shape of an anchor (8) (Fig. 1). The surface of the YSEs is covered by specialized unicellular tubercles of uncertain function (8).

Fig. 1.

Bitterling embryos become lodged in the gill chamber of the host mussel by means of their expanded yolk sac extensions. (A) Bitterling embryos (indicated by arrows) are aligned in a “head-down” position in the interlamellar space of the host mussel (Anodonta anatine). The interlamellar space are divided by interlamellar junctions (ilj) into water tubes (w). (B) With the growth of the wing-like yolk sac extensions, the circular area of embryo axial plane increases significantly. (C) The water tubes of the gravid mussel are dilated by the enlarged YSEs of the brood parasite bitterling embryos. Neighboring water tubes (w2 and w3) merge into one chamber to house the embryo. (D) The embryos are in a secure space, closely attached by the gill filaments, up to the yolk regression period when the yolk sac extensions begin to disappear.

In this study, we have investigated whether other aspects of bitterling development may have been modified in response to selection for successful brood parasitism. It is known (12) that post-hatching bitterling embryos are always oriented in the mussel gill water tube in a head-down position (Fig. 1). However, it is not known how this consistent orientation arises, and how it comes to be in the same orientation as the specializations of the yolk mass, and the site of chorion rupture during hatching. In order to identify a mechanism underlying these phenomena, we examined the process of blastokinesis as a possible candidate. Blastokinesis is a specialized feature of development in bitterlings (13). Blastokinesis has also evolved independently in insects [and possibly in non-insect hexapods (14)] where it has been suggested to function to stir up the yolk, thereby facilitating the supply of yolk nutrients to the embryo (13–16). In insects, it involves/entails a complex set of morphogenetic movements that i) invert the embryo and then ii) return it to its original conformation by means of a “back-flip.” Crucially, during blastokinesis, the positional relationship between the embryo and the yolk changes as tissues move during embryogenesis.

Here, we want to examine whether blastokinesis in bitterlings might be a major, unidentified early embryo adaptation to brood parasitism (early in the sense that it takes place during gastrulation). We have addressed this question in the rosy bitterling by means of time-lapse photography, histology, microCT, electron microscopy, and expression profiling of the developmental genes fgf8a (a fibroblast growth factor) and msx3 (a muscle segment homeobox gene). The fgf8a gene is essential for the dorsal–ventral patterning of mesoderm during gastrulation and neurulation (17) and msx3 is important for dorsal–ventral differentiation of the neural ectoderm (18). These genes are expressed continuously during primary and secondary gastrulation, allowing us to examine the morphogenetic movements of ectoderm and mesoderm during blastokinesis. To examine whether blastokinesis has other consequences relevant to brood parasitism, we tracked the expression of the epidermal marker krt8 (keratin 8) and the hatching gland marker ctslb (cathepsin Lb) in the bitterling and compared it with a non-parasitic cyprinid, the zebrafish. The krt8 gene is a reliable marker of keratinocyte differentiation due to its involvement in intermediate filament organization and keratinization (19, 20). The expression of ctslb gene in the hatching gland is related to the important role of cathepsins in the lysosomal protein degradation required for enzymatic hatching (21). We further analyzed the cellular composition of YSEs and the ultrastructure of their specialized tubercles, to investigate the association of the yolk tubercles with mussel gills tissue and their possible function in helping the embryo resist being flushed out by the host mussel.

Results

Bitterling Embryos Remain Lodged in the Gill Chamber of the Host Mussel by Means of their Expanded Yolk Sac Extensions.

In the host mussel, the main parasitic lodging site of the bitterling embryo is the interlamellar space (Fig. 1A). This space is divided into parallel water tubes by the interlamellar junction (Fig. 1A). The interlamellar space of a parasitized mussel is distinctly dilated by the YSEs of the bitterling embryos. With the growth of YSEs, the width of the embryo increases, and its axial plane expands 1.5 to 3 times (Fig. 1B). At the site occupied by the bitterling embryo, 2 to 3 water tubes are merged to one large chamber (Fig. 1C). The gill filaments are closely attached to the yolk sac of the bitterling embryo, leaving no space for embryo movement before the YSEs regress (Fig. 1D). Therefore, the bitterling post-hatching embryo is held in the gill water tube in this position for 3 to 4 wk (8) until the yolk mass has been consumed and the YSEs have disappeared.

Bitterling Blastokinesis Involves a “front-flip” of the Embryo on the Yolk.

Blastokinesis starts before yolk plug closure and continues until tail-bud formation. During bitterling blastokinesis, the embryo and extraembryonic tissue exhibited a coordinated migration in a consistent direction which we liken to the front flip of a diver. This movement started from yolk plug closure and continued until tail-bud formation, as observed through time-lapse recording (Movies S1 and S3–S5) and analyzed using the Particle Image Velocimetry (PIV) and trajectory tracking techniques (Fig. 2). During the front-flip, the embryo’s head migrates from the animal pole (12 o’clock position) to the vegetal pole (6 o’clock position); the tail relocates from 3 o’clock to 12 o’clock (Fig. 2 A and B). Overall, the head region migrates slightly further than the tail region (Fig. 2C). The head and tail region experience a burst of rapid movement with a speed exceeding 0.35 µm/s (Fig. 2D) when the yolk shape changes (Movie S1). During epiboly, the ventral margin of the blastoderm migrates a longer distance and at a faster speed compared to the dorsal margin (Fig. 2 D and E).

Fig. 2.

The front-flip of bitterling embryo and extraembryonic tissue during the blastokinesis period. (A) The PIV analysis of tissue morphogenetic movements. During epiboly, embryonic tissue migrates in three regional directions: dorsal axis extension, blastoderm epiboly, and convergence of the ventral blastoderm toward the dorsal midline. Afterward, embryo and extraembryonic tissue show a consistent direction of migration akin to a front-flip. The migration of extraembryonic tissue during the elongate period exhibiting bidirectional movement toward the animal pole and the vegetal pole. (B) The timeline, with the schematic of embryo movements. The animal pole (AP) is at 12 o’clock and vegetal pole (VP) at 6 o’clock. At the beginning, the dorsal is at 3 o’clock and ventral is at 9 o’clock, but their location changes with the front-flip of the embryo. The arrowhead indicates the location of head region on the clock panel. (C) The trajectory tracking of regional migration. Trajectories of the blastoderm ventral margin (blue) and dorsal margin (green) during epiboly, and trajectories of head region (red) and tail region (cyan) of embryo during the blastokinesis movement are labeled. (D) Velocity and (E) migration distance of these four regions calculated from trajectories.

Molecular markers allowed us to examine the front-flip that embryos perform on the yolk surface. At the onset of blastokinesis, the yolk plug is positioned dorsal to the vegetal pole. The gene fgf8a is expressed in marginal cells around the entire periphery of the blastoderm during epiboly (Fig. 3 A–C). At the mid-gastrula stage [stages according to Yi et al. (8)], a dorsoventral gradient in the expression of fgf8a appears in the periphery of the blastoderm (Fig. 3B). The strongest expression is in the embryonic shield (the teleost equivalent of Spemann’s organizer) in the dorsal midline. The dorsoventral expression gradient persists up to the closure of blastopore, when it surrounds the yolk plug at the 90% epiboly stage (Fig. 3C). Additionally, the expression domains of fgf8a in the hindbrain change from two transverse bands (Fig. 3B) to a fused region in the midline (Fig. 3C), reflecting a narrowing of the body attributable to the axial convergence (Fig. 2A) at the start of the front-flip.

Fig. 3.

Bitterling embryos “do a front-flip” on the yolk mass during blastokinesis. (A–F) The schematic of each stage presents the dynamic of the embryo front-flip, with the rostral direction indicated by arrows. (A) During the blastula period, the blastoderm (bd) forms a cap on the yolk mass (yk). (B) At the mid-gastrula stage, the dorsoventral axis of the embryo is indicated by the dorsoventral gradient in fgf8a expression and the dorsal midline is labeled by fgf8a shield (sh) expression; the head region is marked out by the presumptive hindbrain region (ph). (C) At the beginning of blastokinesis, the hindbrain (hb) is at the animal pole, while the yolk plug (yp) is located dorsal to the vegetal pole. (D) As blastokinesis advances, the embryo rolls forward on the vegetal pole. With the bending of rostro-caudal axis, the head region (midbrain, mb) is opposite to the tail (t) region. (E) With the formation of somites (s), the embryo elongates, the head region (forebrain, fb and hindbrain, hb) continues migrating towards the vegetal pole, and the tail (t) approaches the animal pole. (F) At the end of blastokinesis, the embryo head is at the vegetal pole, the tail is at the animal pole. (G) When the bitterling embryo rolls on the animal pole, the neural keel (nk) has formed rostrally, while the neural plate (np) more caudally is less converged. The lateral thickenings (lt) and median thickening (mt) of the neural plate are still visible adjacent to the notochord (no). (H) The conserved expression pattern of the marker genes fgf8a and msx3 in zebrafish embryos, notice the laterally expanded hindbrain expression domain of fgf8a in zebrafish at the 90% epiboly stage. (I) Schematic presents difference in morphogenesis between the bitterling and zebrafish during the blastokinesis period. The arrow indicates the translocation of head region on the clock panel.

As blastokinesis advances, the neurula stage embryo has migrated half-way up the yolk mass, being symmetrically doubled over either side of the animal pole (Fig. 3D). The expression of msx3 marks the boundary between the dorsal (neural) ectoderm and the ventral (non-neural) ectoderm (Fig. 3D). A narrow neural keel is indicated by msx3 expression (Fig. 3D). At more caudal axial levels, the neural primordium has not yet developed a neural keel (Fig. 3G).

When the head of the bitterling embryo approaches the vegetal pole and the tail approaches the animal pole (Fig. 3E), expression of fgf8a is found throughout newly formed somites. At the end of the blastokinesis (the somite-6 stage), the head of the embryo arrives to the vegetal pole, a movement driven by axial elongation of the body during somitogenesis. The expression of fgf8a in the segregated somites becomes confined to the anterolateral margin of the maturing somite (Fig. 3F).

Zebrafish development presents a similar pattern of epiboly, except that the zebrafish yolk plug is located at the vegetal pole (Fig. 3 H and I). Furthermore, the fgf8a hindbrain expression domain in zebrafish is broader than in bitterling at the same 90% epiboly stage (the hindbrain width/egg diameter ratio is 62.3% in the zebrafish and 10.2% in the bitterling; Fig. 3 C and H). In zebrafish, the head extends from the animal pole (12 o’clock position) to the vegetal pole in a similar manner, but the zebrafish headstalls at the 9 o’clock position (compared with the bitterling head that goes further and arrives at 6 o’clock; Fig. 3I). Moreover, the zebrafish tail stays at the vegetal pole and extends to the head, resulting in a C-shaped trunk profile at the 12-somite stage (Fig. 3H). By contrast, the bitterling tail migrates toward the animal pole and the trunk is straight at the 6-somite stage (Fig. 3 F and I).

Embryos of R. ocellatus Are Consistently Oriented in a Heads-down Position in the Gill Chamber.

Aldridge (12) has shown that the orientation of the embryo relative to the egg is not random, but is consistently oriented such that the embryos are positioned head-down in the gills. He stated, for the species Rhodeus amarus, that the tail develops toward the open base of the demibranch in over 99% of cases (12). We confirm this observation here for R. ocellatus: Of 255 embryos examined, all were in this heads-down orientation. This is unlikely to be an effect of gravity, because pre-hatching embryos on Petri dishes held in various positions (animal pole upwards, downward, or horizontal) did not change their heads-down position or the heads toward the animal pole position (13). A suite of early events and characteristics in bitterling development sharing this same orientation include the primary embryonic axis and extraembryonic features such as the position and orientation of the YSEs and the site of chorion rupture (hatching). Our hypothesis is that these characters are coordinated according to the orientation of the front-flip on the yolk mass which enables the embryo to combine a proper positioning of the animal pole at fertilization on the one hand, with the necessity of hatching toward the blind end of the water tube to safely anchor in the mussel gills, on the other hand.

The Site of Chorion Rupture Is Consistently at the Vegetal Pole.

Using time-lapse video recording of live embryos, we documented the hatching of embryos (Movie S2). The process is very rapid, taking only 1 to 2 min for the embryo to break free out of the chorion. The newly hatched embryo is motionless until 1 to 2 d after hatching when it responds to stimuli by twitching. We find that the embryo hatches from the vegetal pole of the chorion in all 255 individuals examined. A possible explanation for the rupture of the chorion is an increase in tensile forces exerted on it by the rapidly growing distal tip of the YE (Fig. 4A) as we shall now consider.

Fig. 4.

Bitterling embryos hatch mechanically without hatching enzymes. (A) The embryo hatches with the rostral yolk extension (black arrow) protruding from the vegetal side of the chorion (white arrow). (B) The expression of krt8 is particularly strong at the distal tip of yolk extension (YE) and the yolk sac extensions (YSEs) before and after hatching. Notably, at the Hatching stage (29.5 hpf), krt8 was highly expressed in the anterior tip of the YE (regions 1 and 3), while no expression or low expression levels were observed in the lateral yolk extension ridge (region 2). The cell density in region 1 was not lower than that in region 2, excluding the possibility that the intense staining was arose through differences in cell density. (C) The ubiquitous krt8 expression during zebrafish (Danio rerio) embryonic development. (D) The specific expression of ctslb marking out the hatching gland cells (hg) of the zebrafish embryo. There is no expression of ctslb in bitterling embryos during the hatching event (SI Appendix, Fig. S1).

The Bitterling Has a Mechanical Mode of Hatching in Place of a Hatching Gland.

Since the YE may play a key role in hatching, and therefore in the orientation of the embryo, we examined its cellular composition, as well as that of the yolk sac itself, and the yolk sac extensions. We examined the expression of the gene krt8, a marker of keratinocyte differentiation. We find that its expression is restricted to the distal tip of the YE and YSEs, both before and after hatching (Fig. 4B). By contrast, krt8 is expressed in the zebrafish embryo in the epidermis over the whole embryo up to 48 hpf (Fig. 4C). We suggest that the strong, localized expression of krt8 on the distal tips of the YE and YSEs is related to keratinocyte differentiation which contributes to mechanical strength needed to rupture the chorion. Additionally, Chang (11) proposed that “the egg membrane is ruptured solely by means of the growing force of the embryo itself,” indicating that as the embryo grows, its length gradually exceeds the limit that the chorion can withstand. Our observations support this notion by demonstrating that the growth of the rostral tip of YE contributes to the increase in embryo length (SI Appendix, Fig. S2). Whether it is the aggregation of keratinocytes in the rostral tip of the YE or the morphogenesis of the YE itself, we suggest that it is the internal pressure surpassing the tension of the chorion that likely determines the mechanical nature of hatching (mechanical in that it involves physical forces rather than enzymatic activity).

We hypothesize that hatching in the bitterling is a purely mechanical process mediated by the YE, and there is no involvement of a hatching-gland enzyme. We find no expression of the hatching gland marker gene ctslb in bitterling embryos during the hatching event (10- and 35-somite stages, SI Appendix, Fig. S1) compared to intense expression of ctslb in pre-hatching stages in the zebrafish (Fig. 4D). In agreement with our finding, Chang and Wu’s (13) analysis of the hatching process in the bitterling based on specific histological staining concluded that “there are no integumentary serous glands such as found in many fish embryos.” Although the ctslb is not expressed during the pre-hatching to post-hatching period in bitterling, we were able to amplify the ctslb gene from 14-d-old bitterling larvae, indicating its presence in the bitterling genome and its expression during late development. This provides further evidence against the existence of enzymatic hatching in bitterling embryos.

The Yolk Sac Extensions Can Serve as Adhesive Anchors that Prevent Ejection of the Embryo from the Mussel.

The shape of the yolk changes during embryonic development in many fish species. For example, a caudal protrusion of the yolk ball to form a cylindrical yolk extension (YE) is a shared trait of the order Cypriniforms (22) (carps and minnows). The rosy bitterling shows an additional, highly characteristic modification of the yolk: a pair of wing-like yolk sac extensions (YSEs) that protrude dorsally from the yolk ball (Fig. 5A). The YSEs develop after hatching.

Fig. 5.

Microstructure and histochemical profile of the YSEs. (A) The YSEs are covered by the enveloping layer (EVL) cells. The apical surface of EVL cell is furnished with “microridges” (mr). (B) The thickened YSEs are composed by internal yolk platelets, basal multiply-layered epidermal cells (bc), and remain covered by the apical peridermal cells (ac). The ciliated surface (cs) of mussel gill epithelium is closely attached by the unicellular tubercles of the apical cells. (C) The unicellular tubercle of the apical peridermal cells (ac) contains vesicles that stain intensely with toluidine blue and periodic acid Schiff reagent. The basal cells (bc) are separated from the yolk ball by the basement membrane (bm). (D) The TEM profile of the tubercle apical cells (ac) presents high-density vesicles (v) at the bulged zone, and rough endoplasmic reticulum (RER) with Golgi apparatus (Ga) at the basal zone.

The YSEs are composed (from deep to superficial) of yolk platelets and an epithelial covering composed of basal and apical cell layers (Fig. 5B). The basal layer, separated from the yolk cells by its basement membrane, is the embryonic epidermis. During development, it becomes three- or four-layered at the apex of the YSE and remains covered by the apical cells. The regional thickening of the epidermis is the cellular foundation for the formation of the wing-like projections (Fig. 5B).

The apical layer of YSEs is the periderm or the enveloping layer (EVL) in teleosts. The apical surface of EVL cells is furnished with microridges (Fig. 5A); these are epithelial projections formed by F-actin networks (23). The unicellular tubercles of YSE are specialized EVL. The apical surface of these tubercles protrudes outward from the center, with the microridges present at the rim. At the “bulged” zone, there are high-density vesicles that stain with toluidine blue and periodic acid-Schiff (Fig. 5C). This histochemical picture is consistent with mucus or other glycoproteins (Fig. 5C). The well-developed Golgi apparatus and rough endoplasmic reticulum in the basal zone support the idea that these cells are synthetically active (Fig. 5D).

We have observed damage to the mussel gills that have been parasitized by the rosy bitterling. Specifically, histological analysis demonstrates damage at the interface between the tubercles of the YSEs on the one hand, and the mussel gill epithelium on the other (Fig. 5B). Taken together, our findings suggest that the EVL covering the YSEs may secrete an adhesive substance, possibly to help the embryo resist ejection from the mussel gill chamber. This hypothesis is supported by the fact that the specializations of the EVL disappear at the beginning of the free-swimming stage (24).

Discussion

We have described a suite of embryonic adaptions to the parasitic lifestyle of the rosy bitterling. Since the time of Darwin, it has been a tenet of evolutionary developmental biology that at stages of embryonic development during early organogenesis are shielded from natural selection (4, 25, 26). However, co-evolutionary relationships such as host-parasite and predator-prey arms races can impose exceptionally strong selective pressures on both parties (27). We suggest that the co-evolutionary relationship between the parasitic bitterling and its host mussel involves strong positive selection on the bitterling embryo to evolve adaptations that help it resist being flushed out of its relatively secure shelter by the water current in the gill chamber. Consistent with this hypothesis, we describe and confirm that the development of the rosy bitterling includes several adaptations that may help it remain safely lodged in the gills.

These adaptations include the unique pattern of blastokinesis in the bitterling, which we have explored using molecular markers. Bitterling blastokinesis presents an interesting case of embryonic body inversion related to axial convergence. We have shown that in the rosy bitterling, yolk plug closure occurs on the dorsal side of the vegetal pole, and during epiboly, the ventral blastodermal lip migrates toward the vegetal pole faster than the dorsal lip, indicating convergence toward the dorsal midline. In neurulation, the narrow expression domain of msx3 in the neural keel and the caudal to rostral transition (between neural plate and neural keel) reflect ectodermal convergence. Moreover, gene expression patterns differ between the rosy bitterling and zebrafish, such as the fused hindbrain domain of fgf8a expression in the bitterling compared to lateral expansion in zebrafish at 90% epiboly. Blastokinesis in the bitterling is clearly different from what has been described as blastokinesis in hexapods. In contrast to the back-flip seen in insects (14) (e.g., the milkweed bug, Oncopeltus fasciatus), we describe blastokinesis in the bitterling as being like a front-flip on the yolk mass.

When bitterlings deposit their eggs in the mussel, the eggs are aligned in the female ovipositor as a single row, with the vegetal pole of each egg toward the opening. Thus, there is some imposition of a consistent orientation of the egg during oogenesis. A funnel-shaped micropyle at the animal pole connects the chorion with the egg surface (Fig. 6); it is the only passage for the sperm to fertilize the egg (28). As we have shown here, the embryo is initially oriented with its blastoderm at the animal pole, and then undergoes a front-flip on the yolk mass during blastokinesis, so that it ends up with its rostral end at the vegetal pole. We argue that the distal tip of the YE ruptures the chorion mechanically (that is, by physical force) at the vegetal pole; the head of the embryo then hatches out from the chorion and remains oriented toward the blind end of the mussel water tube. The end result of these events is that the post-hatching embryos are aligned in a row in the head-down position (Fig. 6).

Fig. 6.

The head-down position of bitterling embryo in the mussel water tube. The animal-vegetal polarity of eggs has been determined before oviposition. The bulb-shaped eggs are aligned in the mussel’s gills with the tubular animal pole toward the suprabranchial chamber. The spermatozoa aggregate around the funnel-shaped micropyle at the animal pole. The result of blastokinesis is the embryo hatched out from its chorion in a head-down position, with its head towards the blind end of the gill water tube.

This consistent polarity of bitterling embryos in the gills is likely to be adaptive in two respects. First, we suggest that this arrangement allows bitterling embryos to make the most effective use of the interlamellar space. Thus, if the orientation of embryos were random, some embryos would be stuck in a head-to-head arrangement for a prolonged period that might lead to competition for space and oxygen. Second, we suggest that the head down position may help the embryos resist being flushed out by a sudden burst of high-velocity water flow (6). If this is correct, it could be that the “arrowhead” or “anchor” shape of the embryo indeed acts physically like an anchor.

In the eulamellibranch gill, the water runs into the partitioned water tubes through ostia. Inside each water tube, water moves dorsally and empties into the suprabranchial chamber, then out to the environment from the excurrent siphon (29, 30). From our observation, when hatched embryos are capable of movement, touching them on the tail tip will introduce a fast and straight movement away from the stimulus. In summary, the head down position may allow embryos to i) resist ejection by being anchored to the gill filament via their YSEs (during the early developmental period) and ii) swim against the water current (when capable of movement).

To summarize, the bitterling embryo provides an example of how small modifications in yolk shape and gastrulation movements profoundly affect the development of the embryo and its extraembryonic adnexa. These subtle modifications are amplified to produce an adaptive variation in blastokinesis. This is consistent with the concept of the developmental penetrance of adaptations (31, 32) by which evolutionary adaptations appearing at later stages of development can result from the modification of early stages of development. Interestingly, these adaptations appear not during organogenesis, but earlier, during gastrulation and during the formation of extra-embryonic structures, both of which are known to be highly variable in the vertebrates (33).

Our understanding of brood parasitism comes primarily from studies on birds and social insects but is not restricted to those lineages (1). Relatively benign brood parasites can be accepted by the hosts and their relationship evolves toward tolerance (34). Highly virulent brood parasites are often involved in rapid coevolutionary arm races with their hosts through reciprocal resistance to counter-adaptations (35). Yet, brood parasites typically adapt through behavioral and morphological trickery in terms of egg mimicry, deceptive oviposition or enforcement, and offspring traits related to competition rather than any documented modification of early embryo development (1, 3, 34). While the bitterling is considered a relatively benign parasite (6), it has evolved major adaptations and its hosts have evolved strong counter-adaptations (36). Here, we show how selection for parasitism has penetrated to early developmental stages. This demonstrates that natural selection at a given stage may act on gene regulatory networks active at earlier stages (32). The action of those early development changes can be observed later in development (31) but, as we demonstrate here, can also play a critical immediate role in successful embryo development and survival.

Materials and Methods

R. ocellatus Embryos.

R. ocellatus embryos of synchronized developmental age were collected at the Institute of Vertebrate Biology in Brno, Czech Republic, following in vitro fertilization (IVF). Embryos of various developmental stages (SI Appendix, Table S1) were fixed in 4% paraformaldehyde, dehydrated step-wise in methanol, and stored in 100% methanol at −20 °C. The developmental staging of the embryos was based on Yi et al. (8) Two or three replicates were used for each stage.

Danio rerio Embryos.

Embryos of the zebrafish Danio rerio (AB/TL line) were collected in the fish facility of the Institute of Biology Leiden. The eggs were fertilized by 1:1 spawning (single crossing) at the beginning of the light period (14 h light, 10 h dark). The fertilized eggs were collected and incubated in egg water (containing 60 μg/mL “Instant Ocean” sea salts) at 28.5 °C. After collection, the embryos were immediately fixated in 4% paraformaldehyde. Fixed embryos were dehydrated step-wise in methanol and stored in 100% methanol at −20 °C. The developmental stages were determined according to Kimmel et al. (37). For each stage (SI Appendix, Table S1), ten replicates were used.

Primer Design.

Zebrafish (Danio rerio) is relatively closely phylogenetically related to the rosy bitterling (the same family, Cyprinidae). We BLAST searched the zebrafish genes of our interest (fgf8a, msx3, krt8, and ctslb) since the entire zebrafish genome is known and aligned it to other species in the Cyprinidae (e.g., Carassius auratus and Cyprinus carpio). We searched for conserved regions. After the gene of interest was found, forward and reverse primers of approximately 20 base pairs were designed in the conserved regions that would provide fragments of at least 600 bp with a preferred melting temperature between 55 °C and 60 °C and a favorable CG percentage (around 30 to 40%). These primers were ordered from Sigma Aldrich and used to amplify the target gene from cDNA from reverse transcribed form R. ocellatus total RNA (see SI Appendix, Table S2 for primer sequences and accession number of target genes). These PCR products were cloned in the TOPO-TA PCRII Vector (Invitrogen) and transformed into E. coli bacteria.

Probe Synthesis.

The probe templates were then obtained using PCR with M13-pUC primers and used for RNA polymerase suited to the sequence (T7/sp6, SI Appendix, Fig. S3). Afterward these fragments were run-through a 1% agarose gel to verify whether the fragment length was correct. The sequence of the template was determined by Sanger sequencing performed at Baseclear (Leiden) to verify that the probes contained the correct gene, and the product was compared using a BLAST with the family Cyprinidae. After this procedure, the probes were purified and made ready for use.

Whole-mount In Situ Hybridization.

Whole-mount in situ hybridization (WISH) was performed as described for Danio rerio (38), with a few specific modification for R. ocellatus. Specifically, the Proteinase K (10 µg/mL) digestion time for R. ocellatus is longer than for D. rerio. For R. ocellatus embryo at blastula and gastrula stages, the digestion time was 1 min; for early somitogenesis stages, 5 min; for late somitogenesis (14 to 22 somites) up to 24 hpf, 15 min; up to 48 hpf, 30 min; up to 72 hpf, 45 min; and from 120 hpf to later stages, 90 min. After the treatment, the embryos were photographed using a Nikon smz1500 stereo microscope and stored in 1% paraformaldehyde at 4 °C.

Micro-computed Tomography.

Micro-computed tomography analysis was carried out at Naturalis Biodiversity Center (Leiden, the Netherlands) using a Xradia 520 Versa 3-D X-ray microscope (Zeiss) based on described methods (8) and visualized with Avizo software (Avizo 9.5, Thermo Scientific™).

Histology.

Samples were embedded in Technovit 7100 for routine histology followed by hameatoxylin and eosin, periodic acid Schiff, methenamine silver, and oil red O staining. The section thickness was 0.5 to 0.7 µm. The samples were embedded in epon for toluidine blue staining and semi-thin sectioned with a glass knife. For scanning electron microscopy, we used critical-point drying and scanned the samples with a JEOL SEM 7600. Ultra-thin sections were prepared for transmission electron microscopy and examined in a JEOL 1400 transmission electron microscope.

Time-lapse Recording.

R. ocellatus embryos at the desired developmental stage were mounted in glass view chamber. Stereo microscope (Nikon SMZ1500) was used for recording and the focal level has been adjusted manually to keep the embryo in focus. Photographs were captured with a Nikon DS-Fi1-L2 camera. Time-lapse images were assembled into video recordings using Fiji (39). To quantitatively describe the movement of embryos during blastokinesis, we used the Manual Tracking plugin developed by F. Cordelières (http://rsbweb.nih.gov/ij/plugins/track/track.html) for Fiji. The ventral and dorsal margins of the blastoderm, and the head region and tail region of the embryo were labeled and their trajectories during the blastokinesis movement were tracked. The distance and velocity information of each trajectory were then imported into Excel for statistical analysis.

Velocimetric estimation of overall morphogenetic movements was performed by particle image velocimetry (PIV) analysis using the PIVlab software package (40). The size of the vectors is proportional to the velocity of tissue movements. The analysis spanned four periods: 1) Epiboly period until the yolk plug closure. 2) Front-flip period where the embryo’s head region moved from the animal pole (12 o’clock position) to the vegetal pole (9 o’clock position). 3) Elongate period with continued head migration towards the vegetal pole (7 o’clock position) and tail relocation from 3 o’clock to 12 o’clock. 4) Rostral protrude period concluding at the head’s arrival at the 6 o’clock position before hatching. The deformation modes of each period were described by computing the temporal mean value of the velocity field. The resulting plots were imported to Adobe Illustrator to label the collective direction of vectors. Frame-by-frame PIV analysis was assembled into video (Movies S3–S5).

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Bitterling embryo do a front-flip on the yolk. A lateral view of time-lapse dataset from 70% Epiboly stage to somite-6 stage in 10 hours. The temporal resolution between each frame is 300 s.

Movie S2.

Bitterling embryo hatching out from the vegetal pole. A lateral view of time-lapse dataset during the hatching period in 1 hour. The temporal resolution between each frame is 90 s.

Movie S3.

Frame by Frame PIV analysis of the tissue movement during epiboly, from 00:00 to 02:25 (hours: minutes)

Movie S4.

Frame by Frame PIV analysis of the tissue movement during the front-flip period, from 02:25 to 05:45 (hours: minutes)

Movie S5.

Frame by Frame PIV analysis of the tissue movement during the ‘elongate’ and the ‘rostral protrude’ period, from 05:45 to 09:50 (hours: minutes)

Data, Materials, and Software Availability

MicroCT Raw data have been deposited in Dryad (https://doi.org/10.5061/dryad.qv9s4mwk3) (41). All other data are included in the manuscript and/or supporting information.