The most obvious pattern of these data is the rarity of gynodioecious animals. A maximum of nine animal species may be gynodioecious, and likely fewer, given that most of these species have been poorly studied, as noted above. There is good evidence (Dunn 1975b) that only one animal species, E. prolifera, has been truly documented to be gynodioecious. This is in stark contrast to the situation in plants wherein approximately 8% of species are gynodioecious (Richards 1997).
The paucity of gynodioecious animals has been noted previously, in passing (Ghiselin 1974; Jarne and Charlesworth 1996; Schärer 2009), but few explanations for why gynodioecy should be almost unheard of in animals have been offered. Schärer (2009) suggests that animals may be less prone to mitochondrial manipulation (i.e., “cytoplasmic male sterility”) than plants because of the small size of animal relative to plant mtDNA. In barnacles, Yusa et al. (2012) suggest “outcrossing and female-biased sex allocation in hermaphrodites probably prevent gynodioecy from evolving.” Given that hermaphroditic animals comprise approximately 65,000 species (∼5% of all animals) and are present in approximately 70% of animal phyla (Jarne and Auld 2006), it is amazing that only a single verified case of gynodioecy has been found. In-depth surveys should be undertaken within the primarily hermaphroditic animals groups (e.g., pulmonate snails, annelids, platyhelminthes, etc.) to seek out species with gynodioecy, and a theoretical framework outlining why the evolution of hermaphroditism from dioecy should not include a gynodioecious intermediate (as in plants; Charlesworth and Charlesworth 1978) should be explored (but see below for a possible explanation).
The second obvious pattern is that androdioecy is derived from both hermaphroditism and dioecy in approximately equal numbers, even although dioecy is much more common in animals. Interestingly, androdioecious species derived from hermaphroditic ancestors (e.g., barnacles) do not likely reflect genetic sex determination. For example in many of the barnacles, androdioecy has been derived from protrandrous, simultaneous hermaphroditism (see Table 1) by the reduction of growth during the initial male phase to produce small, “complemental” males, and in the one fish genus, Serranus, androdioecy was derived from simultaneous hermaphroditism in that the largest fish eliminate female investment and become male-only (Petersen and Fischer 1986).
TRANSITIONS FROM HERMAPHRODITISM
As noted above, predictions about the evolution of separate sexes from hermaphroditism stemming from the flowering plants suggest a strong bias for a gynodioecious intermediate rather than an androdioecious one (Charlesworth and Charlesworth 1978; Charlesworth 2006). This pattern was not observed among animals (Table 2A). Animals show only one transition from hermaphroditism to gynodioecy and 16 transitions from hermaphroditism to androdioecy (Table 2A), which is a significant deviation from an equal split between these two “intermediate” reproductive systems (Table 2B). However, Weeks et al. (2006a) point out that the evolution of androdioecy from hermaphroditism in animals is quite different from that proposed for flowering plants. The transition from hermaphroditism to androdioecy in flowering plants has been discussed in terms of the avoidance of inbreeding (Lloyd 1975; Charlesworth and Charlesworth 1978). However, in the hermaphroditic progenitors to these androdioecious animals, selfing is not likely a selective factor due to anatomical considerations that make self-fertilization extremely unlikely (Charnov 1982; Petersen and Fischer 1986; but see Carlon 1999 for a discussion of selfing in corals). Thus, in these species, the avoidance of inbreeding is not likely to explain the evolution of androdioecy.
Instead, it appears that these androdioecious species are expressing sexual differences based on differences in body size. In barnacles, small (i.e., “dwarf” or “complemental”) males are specifically serving a different purpose than the hermaphrodites, being a ready source of sperm to nearby and larger hermaphrodites (Yusa et al. 2012). There are some who argue that these smaller males are essentially following the same sequential reproductive maturity as the Lysmata shrimp noted above (i.e., maturing as males when small and then growing to simultaneous hermaphrodites; Callan 1941; Crisp 1983). If so, then this switch may indicate that individuals can gain higher relative fitness as males when they are small because sperm are cheaper to produce than eggs, but then would do better to switch to hermaphrodites when larger and can afford to produce eggs (Charnov 1982). Others argue that the smaller males are a distinct morph to the larger hermaphrodites (Gomez 1975). In either case, the expression of the sexes is based on different roles of the hermaphrodites and males that correspond to different sizes and population densities (Ghiselin 1969; Charnov 1982). Similarly, in the one fish genus, Serranus, individuals are hermaphroditic when small to intermediate in size but then switch to all-male allocation when they are at the largest sizes (Smith 1965; Hastings and Petersen 1986; Petersen and Fischer 1986). Additionally, in the one coral species (G. australensis), the pattern is reversed, with small corals being male and larger ones are hermaphrodite (Kojis and Quinn 1981). Again, these patterns indicate that differential selective pressures on differently sized individuals drive sexual expression (i.e., differential sex allocation strategies with size; see Ghiselin 1969; Charnov 1982) rather than any benefit being derived from the avoidance of inbreeding in the single-sexed (i.e., male) individuals. Therefore, in these animals the processes selecting androdioecy would be distinct from those argued by Charlesworth and others (Lloyd 1975; Charlesworth and Charlesworth 1978; Charlesworth 1984). However, these species should not be seen as evidence that the hypothesis proposed for the evolution of androdioecy from hermaphroditism in flowering plants is incorrect. Rather, this body of theory is specific to flowering plant species that regularly self-fertilize and not to species in which sexual expression is affected by body size (Ghiselin 1969; Charnov 1982).
TRANSITIONS FROM DIOECY
The evolution of hermaphroditism from separate sexes has not been widely debated, but in the few works that have considered this transition, no one has suggested that such a transition is a mere reversal of the predictions for a hermaphroditism to dioecy transition (i.e., that an intermediate stage is likely to evolve first and that this stage would be gynodioecy). Pannell (1997, 2002) has suggested that androdioecy might be a likely intermediate strategy when evolving hermaphroditism from dioecy in a structured metapopulation in which reproductive assurance is strongly advantageous (e.g., early colonizing species) but in which outcrossing is still advantageous when population size is adequately large to allow an appreciable probability of locating a suitable mate. Similarly, Wolf and Takebayashi (2004) predict androdioecy to be likely to evolve from dioecy when there is pollen limitation (similar to the notion of reproductive assurance noted above) and the hermaphrodites are significantly female biased (i.e., allocating most of their reproductive energy to egg/ovule production). Indeed, there were significantly more cases of the evolution of androdioecy than gynodioecy from dioecious progenitors (Table 2B), which generally fit the expectations of these two models, although we cannot say that these data directly support a “reproductive assurance” cause for these observed transitions.
Weeks et al. (2006a; 2009) make a more specific “constraint” argument for why androdioecy rather than gynodioecy should be more commonly derived from dioecy. If hermaphroditism is selectively favored in any dioecious species (e.g., for reproductive assurance in an early-colonizing species), Weeks et al. argue that the most likely hermaphrodite to evolve from a dioecious female progenitor with strong sexual dimorphism would be a female-biased hermaphrodite that allocates limited resources to male function and lacks the ability to outcross with other hermaphrodites. Their argument is that in dioecious species with many sex-specific phenotypes (and presumably a similarly high level of sex-limited genes), one might expect that a transition to effective expression of both sexes would be highly improbable. The argument relies on the assumption that the minimum amount of mutational “gains” needed to become a functional hermaphrodite would likely be in a female genetic background wherein a “gain” mutation would allow the production of viable sperm in the gonad that would place the sperm into contact with the eggs. In a sexually dimorphic species, the notion that such a hermaphroditic mutant could also simultaneously “gain” the myriad secondary sexual characteristics allowing the hermaphrodite to also fully function as a male would be highly improbable.
An example of the number of gains required to evolve a hermaphrodite from a female progenitor has been elucidated in the nematode Caenorhabditis (Ellis and Guo 2011). In Caenorhabditis, only two mutations appear to be needed to produce a viable hermaphrodite from a female. Reduced activity of the tra-2 gene produces “females” that produce a small amount of sperm when young but then switch to egg production when older (Baldi et al. 2009). These modified females had female phenotypes in every other respect but this small amount of sperm production. However, the sperm was not functional, and thus these mutants were termed “pseudohermaphrodites” (Baldi et al. 2009). A second mutation that affects expression of the swm-1 gene was necessary to “activate” the sperm and allow the mutant hermaphrodites to self-fertilize (Baldi et al. 2009; Ellis and Guo 2011). These authors make a convincing argument that the “sperm activation” mutation would be effectively neutral and could have drifted to such a level as to allow the second (tra-2) mutation to be favored in these nematodes in which reproductive assurance was advantageous for early colonization (Ellis and Guo 2011). These studies show that a female-biased hermaphrodite with limited male gamete production and an inability to outcross through male function can evolve in a straightforward, two-step pathway.
In the above example, a more complete hermaphrodite (i.e., a hermaphrodite that combined all aspects of both males and females into an outcrossing hermaphrodite that functioned equally well in both male and female function) would require many more mutations, and thus be much less likely to evolve. Indeed, no such “fully functional” hermaphrodite has evolved within the nematodes even although there have been several transitions to hermaphroditism from dioecy in this group (Kiontke et al. 2004; Mayer et al. 2007; Yusa et al. 2012), which underscores the difficulty of such a “complete” transition. Therefore, androdioecy should be more likely to evolve from sexually dimorphic, dioecious ancestors than gynodioecy because the number of gains needed to produce a functional hermaphrodite from a male genetic background would be much higher due to the mechanical needs to produce a more complex gamete (e.g., yolking, shell production, storage, etc.) as well as the behavioral requirements of successful offspring production (e.g., egg laying, maternal care, etc.).
For example in dioecious clam shrimp (the ancestral state at the family level; Weeks et al. 2009), males produce ameboid sperm that fertilize the females’ eggs externally in a “brood chamber” on the dorsal surface of the female (Weeks et al. 2004). A hermaphrodite developed from a female genetic background would need to produce sperm within the tubular gonad typifying clam shrimp (Scanabissi and Mondini 2000) to be capable of self-fertilization in the absence of males. On the other hand, a hermaphroditic clam shrimp developed from a male genetic background would need to gain the ability to produce, yolk, and shell the eggs, develop a brood chamber, gain the ability to store eggs in the brood chamber (i.e., by attaching them to extensions of the phyllopod appendages), and develop the digging behavior needed to bury the eggs in the pond bottom (Zucker et al. 2002). Unless all of these phenotypes are regulated by the same regulatory pathway, it is highly unlikely that all of these gains could occur simultaneously within an otherwise male genetic background. Thus one should expect the evolution of hermaphrodites that are essentially female with a small amount of sperm production, which is what is observed in these shrimp (Zucker et al. 1997; Weeks et al. 2005, 2006a, 2009).
Using this argument, the relative likelihood of transitioning to androdioecy versus gynodioecy from a dioecious progenitor is directly related to the level of sexual dimorphism in that progenitor: a high level of sexual dimorphism should bias transitions toward androdioecy whereas limited (or no) sexual dimorphism should produce no bias for one intermediate reproductive mode over the other. Transitions from monomorphic dioecious ancestors, then, should be equally likely in a female or male genetic background resulting in no transitional bias to either androdioecy or gynodioecy.
These constraint arguments were largely upheld in these data. Fifteen total animal genera transitioned from sexually dimorphic dioecious ancestors (Table 2A). In all 15 cases, this transition was from dioecy to androdioecy and none were to gynodioecy, which is highly significantly different from random expectation of an equal number of transitions (Table 2B). These data perfectly fit the constraint hypothesis outlined above. The transitions from sexually monomorphic dioecious ancestors were not as clear: only four genera transitioned from sexually monomorphic dioecious progenitors, and all four transitioned to gynodioecy (Table 2A) which is not significantly different from an equal transition probability, but is close to significant (Table 2B). The small number of gynodioecious transitions does not allow much power to make a conclusion, but we can say that gynodioecy is not constrained from evolving from dioecy among sexually monomorphic dioecious ancestors whereas it does appear to be so from sexually dimorphic dioecious ancestors.
Weeks et al. (2006a) also predicted that hermaphrodites derived from a sexually dimorphic ancestor should be “female biased” (i.e., show primarily female secondary sexual characteristics with only a small amount of sperm production). Indeed, in all the androdioecious species derived from dioecious ancestors in which relative allocation patterns between male and female gametes have been reported, the hermaphrodites closely resemble females with only minor amounts of reproductive effort devoted to male gamete production (Weeks et al. 2006a; Chasnov 2010). This observation was clearly stated by Maupas (1900, p. 133), when comparing hermaphrodites to dioecious females in nematodes: “We are thus entitled to state that parthenogenesis and hermaphroditism, when they developed in these nematodes, only exerted a modifying influence on the products of the genital apparatus. The rest of the organism remained absolutely invariable.” In nematodes, the hermaphrodites generate a small amount of sperm during the last larval stage (Wood 1988) that they store for later use before irreversibly switching to egg production in the adult stage (Ward and Carrel 1979; Hodgkin 1988; Kimble and Ward 1988). Similarly, androdioecious branchiopods produce a small amount of sperm in an otherwise female gonad (Zucker et al. 1997; Scanabissi and Mondini 2002; Weeks et al. 2005) and in most other respects resemble the females of closely related dioecious species (Weeks et al. 2008). The fish (Harrington 1963) and decapod shrimp (Bauer 2006) species also show female-biased allocation. The three hermaphroditic insects (Hughes-Schrader 1928) also show a female-biased allocation, but this unique breeding system, in which the male tissue is provided by the polyspermic fertilization of the egg (Royer 1975), may not fit within the constraint arguments of Weeks et al. (2006a). As suggested above, these overall patterns can be explained by assuming a constraint on the development of a hermaphrodite that can competently perform as a male whereas simultaneously being competent as a female when there are numerous traits required to be competent in both male and female roles.
Although the optimal scenario for reproductive assurance is likely the development of self-compatible hermaphrodites that are competent in outcrossing via both male and female function, it is likely that a second level of constraint will disallow the development of such a hermaphrodite when evolving from a sexually dimorphic, dioecious ancestor. Consider a hermaphrodite that evolved from a female, as outlined above. Such a hermaphrodite should bias allocation toward female gamete production, producing only the minimal amount of sperm required to “assure” fertilization when outcrossing opportunities are low. If outcrossing through male function requires the development of many secondary sexual phenotypes (e.g., an intromittent organ, male mating behaviors, clasping structures for pairing, etc.), it is unlikely that such a female-biased hermaphrodite can simultaneously develop all of these structures to be competent to outcross through male function. Thus, it is likely that the initial hermaphrodite would be minimally competent in the male role, as noted above. The problem that such a “partially male competent” hermaphrodite has is that the further development of a male role would require competing with fully competent, male-only individuals for outcrossings. In this scenario, a hermaphrodite that might develop one (or at most a few) secondary sexual traits would not be likely to compete effectively with fully functional males for outcrossing opportunities. Thus, one might expect that such piecemeal development of a suite of secondary sexual characters would not be advantageous enough to allow the spread of such a “partially male” hermaphrodite that is competing with males with the full suite of secondary sexual traits. Therefore, female-biased hermaphrodites may be constrained to either coexist with males (i.e., androdioecy) or completely replace both males and females to form fully selfing, hermaphroditic populations that cannot outcross. The former alternative should be in species which populate a range of habitats, some of which require reproductive assurance and some of which do not (Pannell 1997, 2002), whereas the latter alternative would be for species in which reproductive assurance is of the utmost importance, or in which outcrossing, per se, is no longer beneficial. Indeed, in the better studied animal androdioecious systems (i.e., nematodes, clam shrimp, tadpole shrimp, and Kryptoledias fish), populations have been documented to be either a mixture of males and self-compatible hermaphrodites that cannot outcross through male function or all-hermaphroditic self-fertilizing populations (Sassaman 1991; Turner et al. 1992a,b; Weibel et al. 1999; Zierold et al. 2007; Weeks et al. 2008; Chasnov 2010).
CONCLUSIONS AND FUTURE DIRECTIONS
The overall patterns are clear. Animals are estimated to have an order of magnitude fewer gynodioecious than androdioecious species likely due to a bias of transitions from dioecy and hermaphroditism to androdioecy. Animals evolving single sexes from hermaphroditism are not avoiding inbreeding; rather these single-sexed individuals have a different ecological role to play than their hermaphroditic counterparts. Animals developing hermaphroditism from single-sexed, sexually dimorphic progenitors appear to be constrained to develop a female-biased hermaphrodite that cannot outcross through male function. Animals developing hermaphroditism from single-sexed, sexually monomorphic progenitors do not experience this same constraint.
A possible explanation for the few observed androdioecious or gynodioecious intermediates deriving from the “primitive” animal phyla (i.e., Cnidaria or Porifera) may be that these sexually monomorphic species may easily be able to develop a fully functional hermaphrodite directly from a dioecious progenitor because such a hermaphrodite does not require many “gains” to be fully functional in both sex roles. Such a hermaphroditic mutant may then quickly sweep through the population/species eliminating both single-sexed competitors if a combined sexual type is advantageous. This could explain the fairly high levels of both dioecy and hermaphroditism in both groups without many intermediate types (Kerr et al. 2011), and also why determining which reproductive system is ancestral to either phylum has proven problematic (Eppley and Jesson 2008; Iyer and Roughgarden 2008; Kerr et al. 2011).
Further understanding of these reproductive transitions would greatly benefit by a series of studies. First, gynodioecy should be extensively examined within the Animalia. A start would be to assess the validity of the reports of the nine instances of apparent gynodioecy noted in Table 1. Several of the gynodioecious Cnidaria and Porifera noted in Table 1 have vertically transmitted, unicellular symbionts, and it would be interesting to assess if these symbionts may be able to manipulate sex expression in these gynodioecious animals in a similar fashion to that noted for mitochondria in gynodioecious plants (Schärer 2009). Additionally, gynodioecy should be sought in the primarily hermaphroditic animals (e.g., pulmonate gastropods, flatworms, gastrotrichs, etc.) because it is quite likely that gynodioecy has evolved at least once (if not repeatedly) among these diverse species.
Second, further phylogenetic assessments of the primitive animals (e.g., sponges and cnidarians) would greatly improve our understanding of reproductive transitions in both groups (e.g., Kerr et al. 2011). Such studies would then allow the mapping of breeding system onto a detailed phylogeny so that we could assess the breeding system transitions in these interesting groups. In particular, it would be useful to note how many transitions between hermaphroditism and dioecy have occurred in these groups and whether there is a bias for one transition over another. Of course one additional benefit to such studies would be the potential to resolve the ancestral reproductive state of the Animalia (Eppley and Jesson 2008; Iyer and Roughgarden 2008).
Third, this field needs to develop a solid theoretical foundation predicting the transitions between dioecy and hermaphroditism in animal taxa. At the moment, the most plausible selective agent driving the evolution of hermaphroditism from dioecy is reproductive assurance (Pannell 1997; Wolf and Takebayashi 2004). However, we have no competing hypotheses nor do we have a single study that reveals a benefit of reproductive assurance for hermaphroditic mutants deriving from single-sexed progenitors in animals.
Finally, to assess the constraint hypothesis, the genetic underpinning of male gamete production in androdioecious species derived from dioecious progenitors needs to be undertaken, similar to that conducted in nematodes (Baldi et al. 2009; Ellis and Guo 2011). This argument relies on the constraint of simultaneously “gaining” many primary and secondary sexual traits of the alternate sex to become a fully functional hermaphrodite (Weeks et al. 2006a; 2009). Thus, the female-biased androdioecious species (e.g., Triops, Eulimnadia, and Kryptolebias) should have relatively simple mutational changes to produce the sperm needed to be a functioning hermaphrodite, as noted in Caenorhabditis (Baldi et al. 2009). Assessments of the genetic changes allowing sperm production in species from these genera would allow a thorough examination of this constraint hypothesis.