Fetal adaptations for viviparity in amphibians

Authors

  • Marvalee H. Wake

    Corresponding author
    1. Department of Integrative Biology and Museum of Vertebrate Zoology, University of California, Berkeley, California
    • Correspondence to: Marvalee H. Wake, Department of Integrative Biology, 3060 VLSB, University of California, Berkeley, CA 94720-3140. E-mail: mhwake@berkeley.edu

    Search for more papers by this author

ABSTRACT

Live-bearing has evolved in all three orders of amphibians—frogs, salamanders, and caecilians. Developing young may be either yolk dependent, or maternal nutrients may be supplied after yolk is resorbed, depending on the species. Among frogs, embryos in two distantly related lineages develop in the skin of the maternal parents' backs; they are born either as advanced larvae or fully metamorphosed froglets, depending on the species. In other frogs, and in salamanders and caecilians, viviparity is intraoviductal; one lineage of salamanders includes species that are yolk dependent and born either as larvae or metamorphs, or that practice cannibalism and are born as metamorphs. Live-bearing caecilians all, so far as is known, exhaust yolk before hatching and mothers provide nutrients during the rest of the relatively long gestation period. The developing young that have maternal nutrition have a number of heterochronic changes, such as precocious development of the feeding apparatus and the gut. Furthermore, several of the fetal adaptations, such as a specialized dentition and a prolonged metamorphosis, are homoplasious and present in members of two or all three of the amphibian orders. At the same time, we know little about the developmental and functional bases for fetal adaptations, and less about the factors that drive their evolution and facilitate their maintenance. J. Morphol. 276:941–960, 2015. © 2014 Wiley Periodicals, Inc.

INTRODUCTION

Amphibians have performed many “natural experiments” in live-bearing and nutrition. Viviparity, with and without fetal nutrition supplied after yolk is exhausted, has evolved independently several times in each of the three orders of living amphibians. Intraoviductal live-bearing with maternal nutrition via cellular proliferations of the oviductal mucosa characterizes some of the viviparous species of salamanders, perhaps all of the species of live-bearing caecilians, and perhaps two viviparous frogs. However, unlike the situation in some lineages of sharks and rays, teleost fishes, and of course amniotes, no species of amphibians, so far as is known, has developed placentae or functional pseudoplacentae (with the interesting exception of the possible use of the dilated gills and a skin patch termed an “ectotrophoblast” in the caecilian Typhlonectes compressicauda; see discussion later, and summaries by Exbrayat, 2006a–d). Anamniotes lack the extra-embryonic membranes used for placentation by amniotes (except for the yolk sac), although the gills and skins of embryos and fetuses likely function in gaseous exchange (and perhaps nutrient uptake). Both parents and developing embryos and young have evolved a great diversity of modifications of structures and functions to facilitate live-bearing (e.g., summarized by Blackburn, 1992; Wake, 1992, 1993, 2003; Buckley, 2012). However, much of the literature deals with maternal mechanisms for viviparity (e.g., Greven, 2011; Wake, in press), and far less with those of the developing embryos and fetuses. Forty years ago, I wrote a paper on “fetal maintenance and its evolutionary significance” in the caecilian amphibians (Wake, 1977a); at that time, there was also active research on live-bearing in one frog species and two or three salamanders. While the research of that period still informs current work, new information has emerged as more species have been studied and a greater range of tools used. We still, though, have great gaps in our understanding of the evolution of the mechanisms of live-bearing, especially those of developing embryos and fetuses.

Amphibians have evolved a great diversity of derived reproductive modes, including several that involve live-bearing. The site of maintenance of the embryos varies, especially among frogs. Nearly all live-bearers are yolk dependent and give birth to either late larvae or metamorphosed froglets; very few species provide maternal (or paternal-see below) nutrients after the yolk is resorbed. Salamanders and caecilians that are live-bearers bear the developing young intraoviductally. Some salamander species and subspecies variously provide no maternal nutrition beyond that of the yolk and give birth either to late larvae or metamorphs, while others practice intraoviductal cannibalism, and one provides a nutrient material after yolk is exhausted. Caecilians also carry their developing young in their oviducts, but all of the species that do so provide nutrients after the yolk is resorbed, this has evolved multiple times within the order. Similarities and differences in morphology, development, endocrinology, physiology, and pattern of evolution among live-bearing amphibians are my focal interest.

Given this diversity of modes of live-bearing in the amphibian clade, and the number of times it has evolved independently, I examine the evolution of live-bearing in amphibians to understand the origin, development, and maintenance of similarities, that is, homoplasy, the situation in which similar features have evolved independently in lineages not associated with derivation from a common ancestral state, as well as differences in live-bearing modes.

The purpose of this review is the examination of fetal adaptations to viviparity, especially modes of fetal nutrition, but among amphibians (and many other taxa), the adaptations of the fetuses are largely inextricable from those of the parent. I have recently reviewed research on amphibian viviparity largely from the perspective of maternal biology (Wake, in press). I will only briefly summarize the relevant parental information. I concentrate now on the nature of embryonic and fetal modifications for viviparity, and the correlations/potential causations associated with the conditions in the parents. The literature cited is not exhaustive but intended to be illustrative.

Amphibians have devised a diversity of nonplacental ways of ingesting nutrients and accomplishing gaseous exchange. This fact leads to another key point: the definition of “viviparity,” or live-bearing, includes a number of different conditions. In general, these conditions fall into one of two large sets: one characterized by the developing embryos and larvae being maintained by a parent (usually female, but occasionally male) in or on its body, and being completely dependent for nutrition on the yolk provided in the egg before it is fertilized. They can be “born” at different states of late development including completion of metamorphosis, but often are not yet fully metamorphosed and must spend a period in water to complete metamorphosis. See Wake (1977a), Wake and Dickie (1998), Blackburn (1994), and Greven (2003b) for summaries. The second category is that in which the embryos and fetuses are maintained in or on the body of a parent (again, usually female, but male in a few species) with provision of nutrients following the resorption of the yolk, and the young are born fully metamorphosed and do not require an aquatic larval period. Consequently, in the literature “viviparity” is often used to include all modes of live-bearing, whereas the terms “ovoviviparity” and “larviparity” refer to conditions in which developmental stage at birth is plastic, often as larvae, and postyolk nutrition is not provided, and “euviviparity” and “pueriparity” to the condition of maternal nutrition and birth as metamorphs.

Such definitions beg the questions “what is a fetus?” and “what is fetal nutrition?” I infer that a fetus is the posthatching, postembryonic, prebirth, often metamorphic, maternally or paternally retained developing young. There are several components of fetal nutrition—the composition of the nutrient material (such as yolk, other parental secretions, and sibs), the way the fetus obtains the material, the way the fetus obtains O2, and so forth, for the energy to metabolize the nutrients, and the manner in which metabolic byproducts and waste are eliminated—that are spatially, temporally, morphologically, and physiologically separated in amphibians in ways that generally do not occur in placental vertebrates.

AMPHIBIAN MODES OF LIVE-BEARING

The great diversity of natural experiments in reproductive modes of frogs include ways of live-bearing. Some extreme but well-documented examples include males of Rhinoderma darwini that swallow the clutch of eggs they have just fertilized and maintain the developing embryos through metamorphosis in their vocal sacs; there is indirect evidence of potential paternal nutrition via secretion from the epithelium of the sacs until the young are born as metamorphs (Goicoechea et al., 1986). Females of Rheobatrachus silus ingested their just fertilized eggs and brooded them in their stomachs until they were “born” (= regurgitated) as juvenile frogs (Tyler and Carter, 1981). Tyler et al. (1982) found that the developing embryos secrete a particular prostaglandin that inhibits gastric secretion in the mothers' stomachs so that the young are not digested. In both of these cases, the genus includes the species with the behavior mentioned, and a single sister taxon that has basically the same behavior, but in which the young are born as advanced larvae that require a period in water to complete metamorphosis. Regrettably, the two stomach-breeding species are now considered extinct, and Rhinoderma is highly threatened, so the research on the biology of the systems likely will not be completed. Another example of frog live-bearing is that of Eleutherodactylus jasperi, a species in the highly speciose family Eleutherodactylidae, which, with its sister families, is characterized by having direct development [eggs fertilized usually externally, the clutch deposited on land (sensu lato), and development through metamorphosis such that hatchlings are terrestrial juveniles, rather than aquatic larvae]. E. jasperi was the lone exception in this large clade of direct-developers; females retained fertilized (internal fertilization is presumed) ova in their oviducts, and gave birth to 1–5 fully metamorphosed juveniles (Wake, 1978a; see figures), so it was ovoviviparous. E. jasperi also is presumed extinct.

Two other major modes of live-bearing reproduction in anurans include 1) back-brooding with fertilized eggs deposited either in a pouch formed of the skin of the mother's dorsum or embedded directly in the dorsal skin, with the young born either as late larvae or as metamorphs, in the terrestrial Gastrotheca and its relatives (Hemiphractidae) and in the aquatic frog Pipa (Pipidae) and 2) intraoviductal development, either without postyolk maternal nutrition but birth as fully metamorphosed froglets (e.g., Nectophrynoides) or with provision of nutrients after yolk resorption and birth as fully metamorphosed juveniles (e.g., Nimbaphrynoides). I compare and contrast similarities and differences at several levels of the hierarchy of biological organization in a search for generalizations about the origin and maintenance of the live-bearing homoplasious condition. I include new data for the two lineages of back-brooders, and also for the intraoviductal bearers; the latter include species in only a few genera of bufonids, the Eleutherodactylus aforementioned, and a dicroglossid (not yet described). I compare and contrast the biology of the frogs (viviparous and ovoviviparous) with that of live-bearing salamanders and caecilians.

Salamanders have evolved live-bearing only in a few species in two genera, Salamandra and Mertensiella, of the family Salamandridae, and perhaps in a plethodontid that occurs in Italy, in all cases, the live-bearing is intraoviductal. Salamandra presents diverse patterns of evolution of viviparity, with one species, S. atra, being obligately euviviparous and having a flexible gestation period that may be as long as 5 years (Browning, 1973; Joly et al., 1994), and several subspecies of the species Salamandra salamandra being ovoviviparous, one subset of which practice intraoviductal cannibalism as the postyolk mode of “maternal” nutrition and birth of metamorphs (see Bucklely et al., 2007), and the rest being yolk dependent and born at various stages of late larval development through metamorphosis.

In contrast, viviparity has evolved in caecilians at least four times independently, (Gower et al., 2008; Wake et al., 2011) including all dermophiids (Dermophis, Gymnopis, Schistometopum, Geotrypetes; taxonomy of Wilkinson et al., 2011), presumably all typhlonectids (Typhlonectes, Chthonerpeton, Potamotyphlus, Atretochoana (not yet observed in the latter), one species of Gegeneophis (of the seven genera in the Indotyphlidae), and all of the species of Scolecomorphus (Scolecomorphidae; reproductive mode not known for the other genus in the family, Crotaphatrema). So far as is known, caecilian viviparity always involves provision of maternal nutrition via the cells of the oviductal mucosa after yolk is fully resorbed and the embryo hatches, and birth after a gestation period of several months (e.g., Wake, 1993; Wake and Dickie, 1998). Caecilians have also experimented with posthatching and/or birth in a unique manner, that of hatchlings and perhaps newborns feeding on the proliferated lipid-rich skin of the mother (dermatophagy: e.g., Kupfer et al., 2006; Wilkinson et al., 2008), which is a mode of parental care that may be influenced by reproductive control mechanisms (see later and Gomes et al., 2013).

BACK-BROODING IN FROGS

The biology of the two distantly related back-breeding lineages, Gastrotheca and Pipa has a number of similarities associated with reproductive mode but also includes major differences at both upward and downward levels of the hierarchy of organization. The ecology of the animals could hardly be more different; Gastrotheca are terrestrial frogs; courtship and mating are not dependent on nearby water. Courtship ends with amplexus with the female in a head-down position as the male clasping her back fertilizes her eggs as they emerge singly from her cloaca. The male guides the eggs into the entrance of a permanent dorsal skin pouch on the back of the female (see Elinson et al., 1990 for summary). So far as is known, of the 62 species of Gastrotheca currently recognized, the majority give birth to fully metamorphosed froglets, the rest to advanced tadpoles (Weins et al., 2007). Pipa are aquatic northern South American frogs; far less is known about their reproductive biology than that of G. riobambae. Their courtship is complex, with turns in the water and amplexus with the female eventually assuming a head-down position and the male clasping her back inguinally, facilitating the male's fertilizing the eggs as they emerge from the female's cloaca and placing them onto the back of the female (Rabb and Rabb, 1960; Rabb and Snedigar, 1960; Weygoldt, 1976). The assumption of the female's head-down position and the male's guiding eggs on to her back have evolved independently as end points of courtship in both genera, apparently to facilitate the dorsal brooding habit—a homoplasious behavior. [See Wake (in press) for details.]

Gastrotheca

The reproductive biology and development of Gastrotheca riobambae, which gives birth from the pouch to advanced tadpoles approximately 100 days after the insertion of the fertilized ova, are well described, mostly by Eugenia del Pino and coworkers. The basic endocrinology and morphology of maternal skin modification for back-brooding are well established; I summarize this information. Jones et al. (1973) reported that estrogen administered at 12 weeks postfertilization causes formation of a pouch of dorsal skin in females; the skin on both sides of the dorsal midline enlarges, folds, and upper and lower layers fuse to form the pouch, with the reduction or loss of epidermal keratin, dermal poison glands, and dermal chromatophore units. The skin of the pouch regresses during the nonbreeding season. del Pino examined the endocrine cycle of the frogs, and found that during the follicular phase of the endocrine cycle approximately 100 ova mature (3 mm diameter; del Pino and Sánchez, 1977; de Albuja et al., 1983), and the skin of the pouch proliferates and vascularizes (del Pino et al., 1975; del Pino, 1980, 1983), clearly estrogen effects. She determined that progesterone conditions the pouch and maintains hyperemia and vascularization and initiates pouch closure well before ovum maturation and oviposition (del Pino, 1983; de Albuja et al., 1983). del Pino did not identify the evacuated egg follicles as corpora lutea (del Pino and Sánchez, 1977), but described their incomplete invasion by cells and their regression 30–50 days later, and she measured progesterone levels and their effects on the pouch, oogenesis, and development. Eggs and/or embryos are necessary for pouch maintenance (a fetal adaptation), as del Pino (1983) demonstrated by removing eggs from the pouch and substituting inert beads the size of the eggs. She gave the females either progesterone or human chorionic gonadotropin to delineate effects on behavior and on maintenance of the beads. Frogs with beads but not hormone did not exhibit the normal postmating behavior; frogs that received only progesterone did, so del Pino concluded that the behavior is hormonally mediated. The females incubated the progesterone-enhanced beads for about one week, then released them, covered with pouch epithelium. del Pino (1983) concluded that the physical presence of embryos or beads, as well as progesterone-like hormone, is necessary for formation of the pouch chambers for embryos, and continuous presence of the hormone is necessary for maintenance and incubation. At the same time, it is clear that growing embryos interact with the mother such that physical and perhaps physiological cues are involved in maintaining the pregnancy, a fetal adaptation.

The pouch is a permanent structure, but it regresses after birth, then recrudesces with the next seasonal/hormonal cycle. The length and placement of the pouch opening vary among species (see also Duellman and Maness, 1980). Data gathered in my lab on embryos and pregnant females of several species of Gastrotheca largely corroborate del Pino's observations of G. riobambae.

Early development of embryos of G. riobambae has been extensively studied [see del Pino (2000) and Elinson (1987) for reviews]. Cell division takes place atop the yolk such that the developing blastula is a flat layer of cells, unlike the total cleavage that occurs in most frogs. This pattern is presumed to characterize all members of the lineage. It is not known whether or how it correlates with back-brooding per se. Furthermore, Gastrotheca tadpole morphology is characterized by a fetal adaptation equally unique to the clade, the early development of a single pair of external “bell gills” that grow to envelop the body of the developing embryos/tadpoles. del Pino and Escobar (1981) found that in some species, the bell gills are derived from the first branchial arch, and cover less than 50% of the tadpole's body, but in others, the gills are derived from the first and second branchial arches and coverage ranges from less than 50% to complete 100% coverage of the embryo/tadpole body. The most complex and complete bell gills result from fusion of the two arches. In G. riobambae, enlargement and fusion of branchial arches I and II begins very early at Stage 14, when the heart is formed but not yet beating. As fusion becomes more complete and the aortic arches receive circulation, afferent and efferent vessels extend from each aortic arch and are incorporated into a stalk, so two stalks extend into the developing bell gill. Extensive branching of the vessels occurs such that the bell gills become well vascularized. In all the species of Gastrotheca they studied (their Table 2), the bell gills cover 100% of the tadpole's body, and all have gills formed from both arches; the five of their 27 species that have only one stalk include four vessels in the stalk, rather than two, indicating that the stalk is fused. By Stage 25, the gill stalks are mostly inside the opercular chamber, with the large bell gills protruding from the spiracle. The internal gills and the lungs are well developed but not completely. The gills contact the outer vitelline envelope and the thin jelly capsule of the egg/embryo as they grow to surround the tadpole, and are separated only by the envelope and the jelly from the maternal pouch epithelium. At the same time, the pouch skin becomes highly vascular, proliferates, and forms partitions between developing young. In G. riobambae, hatching and birth of tadpoles occurs at Stage 25, and the bell gill circulation ceases as the tadpole touches water, but the gills continue to protrude for about 24 h before retracting into the opecular chamber, where they reduce in size until they are fully resorbed. del Pino postulated that an “osmotic shock” of the fresh water induces the circulatory changes and cause elimination of the bell gills as respiratory organs, because if tadpoles are kept in Ringer's solution after birth, the gills and their circulation are maintained (del Pino et al., 1975). Our histological observations on seven species, four that del Pino did not study, largely corroborated their work, but we also note that the space between the bell gills and the body encloses fluid; we also noted that the tail is highly vascular in all the species with advanced tadpoles that we examined. The bell gills in particular have been suggested to be involved in gaseous exchange while appressed to the pouch epithelium. It has not been previously stated, but the tadpole tails of Gastrotheca are expanded and highly vascular in several species, at least, so they too might serve as respiratory structures.

Another component of fetal adaptation in G. riobambae, and presumably all members of the genus, is that embryos in the pouch excrete urea, as do free-living tadpoles following birth (Alcocer et al., 1992). Typically, embryos and aquatic tadpoles excrete mostly ammonia, but acquire full ureotely at metamorphosis (Balinsky, 1970). In G. riobambae, fluid surrounding embryos in the pouch had very high concentrations of urea, and that in the blood of pregnant and nonpregnant females and males lower concentrations, but present in all samples. Further, activity of arginase, a key enzyme in the ureotelic cycle, increased as embryos advanced in development (see Alcocer et al., 1992). Ureotely is maintained during both the 3-month pouch period, the 2–12 months tadpole phase in water before metamorphosis, and in adulthood. Alcocer et al. suspect that maintenance of urotely is an adaption to pouch life, and to development in limited amounts of water and crowded conditions. del Pino et al. (1994) found that urea is required for culture of embryos outside of the pouch; in fact, the concentration must be doubled postneurula for organogenesis. Although ureotely is generally considered to be an adaptation for terrestriality, various amphibians deal with it differently. The well-studied direct-developing Eleutherodactylus coqui is apparently ammonotelic and switches to ureotely toward the end of its prehatching terrestrial clutch life, as measured by inception of arginase activity, such that it hatches as a ureotelic metamorph (Callery and Elinson, 1996). However, the intrauterine larvae of S. salamandra excrete urea, but switch to ammonotely as free-living larvae, then back to ureotely as adults (Schindelmeiser and Greven, 1981). A few frogs (e.g., a foam nester, etc.) are ureotelic as tadpoles, and some even secrete uric acid as adults (Shoemaker and McClanahan, 1973). Obviously, too little is known of patterns of nitrogen excretion for most amphibians with derived modes of reproduction; the data at hand suggest there may be a diversity of fetal adaptations in this physiological mechanism.

A common component of derived modes of reproduction in amphibians is enlarged egg sizes and reduced egg numbers, especially in direct-developers but also in yolk-dependent live-bearers. We found that egg and clutch size vary considerably across the species of Gastrotheca that we examined. del Pino and Escobar (1983) considered G. riobambae eggs large at 3 mm diameter; we found that in a G. longipes (which gives birth to fully metamorphosed froglets) the diameter of back-brooded eggs that were at very early development is 11.5 mm. Also, del Pino and Escobar considered the clutch of approximately 100 eggs “small”; the clutch of the G. longipes we examined is 17. Consequently, I suggest that life history characters such as egg and clutch size and length of retention continue to evolve in various directions in the lineage, despite the plesiomorphic pouch deposition potentially constraining the evolution of those features in terms of size and physiological accommodation of the embryos, again indicating an arena of fetal adaptation that requires study.

Full ontogenetic sequences of development are not yet available for the species we examined, so our observations are based on samples of early, mid-, and late development in various species. We examine late stages to determine a mechanistic basis for birth as tadpoles or as metamorphosed froglets. Accumulating evidence suggests that corpus luteum maintenance and activity mediates the stage at which birth takes place, but more research is necessary. If corpora lutea regress at mid- to late tadpole development, tadpoles will be born shortly afterward and if the corpora lutea maintain function for a longer time, metamorphosis ensues, and froglets are born (see Wake, in press).

Pipa

Knowledge about back-brooding in Pipa is rather limited. Relatively little work has been done on the morphology involved, and virtually none on the endocrinology. Observations include those of Trueb and Massemin (2000) that some eggs do not embed adequately after deposition by the male and they fall off, but most develop in the skin of the female's back, which overgrows the eggs, leaving a small aperture in nearly all species. The skin thickens and becomes hyperemic and vascularized, forming a cup that holds the young. Pipa arribali is apparently the only species in which the skin completely covers the embryos, thought by Buchacher (1993) to protect the embryos as the pregnant females migrate on land among ponds. Fully metamorphosed froglets are born to members of the {P. aspera + [P. arrabali + (P. snethlageae + P. pipa)]} clade; (P. parva + P. myersi) + (P. carvalhoi), the more basal taxa, have free-living tadpoles (Trueb and Massemin, 2000). P. carvalhoi tadpoles resorb their yolk supply 2/3 through gestation, and are born as advanced tadpoles. Eggs are larger and fewer in Pipa species that give birth to metamorphosed froglets (Trueb and Massemin, 2000). Fernandes et al. (2011) suggested that there is nutrient transport from the vascularized maternal epithelium to that of the tadpole, but did not characterize the purported nutrient material.

Greven and Richter (2009) studied the morphology of the skin of the dorsum of Pipa carvalhoi in females with embryos and young at different stages of development. They carefully documented histological modifications during pregnancy. The skin incompletely overgrows each egg, so that the embryos lie in epidermal cups. The epidermis progressively becomes bilayered, thin, and lacking a stratum corneum; the dermis is loose and highly vascularized. Tadpoles lie head-down before birth, their tails waving through the open mouths of the skin cups (Greven and Richter, 2009; see their figures). They are born as well-developed, prelimb extension, tadpoles. Greven and Richter examined the morphology of the ovaries, but only in terms of oogenesis; they also discussed similarities of P. carvalhoi skin with that of G. riobambae as reported in the literature.

Our preliminary data for five species (P. pipa, P. parva, P. snethlageae, P. arrabali, P. carvalhoi) suggest that the Pipa ovarian cycle is similar to that of Gastrotheca. We have identified corpora lutea in P. parva and P. carvalhoi, and corpora atretica in P. snethlageae), but we do not have ontogenetic series, so I cannot say how long corpora lutea are maintained. I predict that the same cycle of follicular estrogen stimulation of skin cup formation followed by luteal secretion of progesterone to maintain the pregnancy and inhibit ovum development occurs. P. arrabali has the most extreme skin response of the species we have examined to date. It has extensive proliferation, vascularization, and so forth, and the complete skin covering over the developing embryo/tadpole/froglet, likely correlated with its tendency to be semiterrestrial. The tadpoles of Pipa have dilated, well-vascularized tail fins, which has been presumed to be a fetal adaptation for gaseous exchange during back-brooding. This is another aspect of maintenance that should be investigated in both lineages of back-brooding frogs. Much more extensive investigation of the ontogenies of the embryos and tadpoles in relation to the back-brooding phenomenon and interaction with the mother's biology awaits.

I predict, based on my preliminary data for both lineages of back-brooding frogs, that species vary considerably in the timing of corpus luteum regression, such that “young” may be born at a range of different stages and sizes of development from midlarvae to fully metamorphosed froglets in live-bearing frogs, whether borne in the dorsal skin or in the oviducts, a significant fetal adaptation. I suspect that this might also vary within species, depending on the specificity of regulation interaction with seasonal and/or other ecological effects. Elinson et al. (1990) reported that developmental rates and successes vary in G. riobambae when they are raised in labs in cities at different altitudes! (Compare with discussions of Nimbaphrynoides, S. atra, and some caecilians below.) Understanding the evolution of the phenomenon of back-brooding in frogs, especially the homoplasy represented by the two lineages, requires a hierarchical and integrative approach that considers molecular, cellular, developmental, physiological and endocrinological, behavioral, and ecological parameters, and it has only begun.

INTRAOVIDUCTAL RETENTION OF INTERNALLY FERTILIZED OVA

Oviductal viviparity in amphibians is relatively rare, especially in frogs and salamanders (Wake, 1989). Internal fertilization is a precondition for oviductal retention of fertilized eggs and is accomplished by distinctly different means in each of the three orders. The very few frogs that have internal fertilization use cloacal apposition or insertion of the vestigial tail (Ascaphus: see Sever et al., 2001); however, 90% of salamanders fertilize the ovulated eggs as a spermatophore is deposited on the substrate by the male during courtship and grasped in the female's cloaca, often for some time in the spermatheca, from which the sperm migrate to the ova. All caecilians apparently have internal fertilization via the male inserting the extruded posterior part of the cloaca (the intromittent organ) into the vent of the female during courtship, a mechanism of direct sperm transfer.

Some frogs that have internal fertilization lay their eggs and guard them as they develop, such as Eleutherodactylus coqui (Townsend et al., 1981). A few anuran species (mostly bufonids, and the extinct Eleutherodactylus) are ovoviviparous, retaining yolky fertilized eggs in the oviducts and giving birth to fully metamorphosed froglets [e.g., Nectophrynoides (see Wake, 1980a for summary) and E. jasperi (Wake, 1978a)].

Prolonged oviductal retention of fertilized eggs occurs in a few members of the Family Salamandridae among salamanders, and it is reported in a plethodontid (see later). Subspecies of S. salamandra retain internally fertilized ova; the embryos depend on their yolk for nutrition during development and are born as larvae, requiring an aquatic period during which further growth, feeding, and metamorphosis take place. Two other subspecies of S. salamandra are viviparous via intraoviductal siblicide, with birth following metamorphosis (summarized by Buckley et al., 2007). S. atra, which occurs (or did in the recent past) broadly throughout the Alps, is obligately viviparous, providing nutrients to its two embryos after their yolk is resorbed (summarized by Guex and Greven, 1994 and Joly et al., 1994).

Less information is available for caecilians, but even oviparous taxa appear to retain the fertilized ova for a time. For example, Ichthyophis glutinosus lays its eggs at neurulation, and the female guards them until the larvae hatch and wriggle to streams to continue developing through metamorphosis (e.g., Breckenridge et al., 1987; Sarasin and Sarasin, 1887−1890; Brauer, 1897, 1899, and see discussions by Wilkinson and Nussbaum, 1998; Laurin et al., 2000a,b; Wilkinson et al., 2002; Laurin, 2005). Also, a number of caecilian species in four different lineages retain the developing young through metamorphosis in their oviducts [see Gower et al., 2008; Wake (in press) for summaries]. Yolk dependence without maternal nutrition after yolk resorption (ovoviviparity) is not yet known in caecilians. All of the species that retain their young beyond neurula appear to have embryos that hatch in the oviducts when their yolk is exhausted. The oviductal mucosa of the maternal females proliferates a nutrient material that is ingested by the fetuses. A feature of obligate viviparity in all three orders of amphibians is that egg size (yolk provisioning) is reduced, concomitant with the longer gestation periods and extensive maternal nutrient secretion following yolk resorbtion (summarized in Wake, 1977a,b, 1980b, 1993; Wake and Dickie, 1998).

Fetal Adaptations for Oviductal Viviparity in Frogs

Just as internal fertilization (necessary but not sufficient for oviductal maintenance) has evolved in only a few species of frogs, intraoviductal development has evolved in even fewer lineages and species, and most are yolk dependent, such that the maternal female does not provide additional nutrients, but gives birth to metamorphosed froglets that often have a small amount of unresorbed yolk still in the gut, depending on the species. (None to date are known to give birth to tadpoles that require a free-living, aquatic period to complete metamorphosis.) This pattern characterizes the ovoviviparity of several species of the African genus Nectophrynoides (Bufonidae) and the (now extinct) Puerto Rican Eleutherodactylus jasperi (Eleutherodactylidae). In addition, there are reports of species [another Eleutherodactylus (Lynn, 1940), a craugastorid (McCranie et al., 2013), a dicroglossid (McGuire, personal communication), etc.] found with intraoviductal tadpoles or laid froglets, some still in their egg membranes.

Research on the reproductive biology of E. jasperi, now extinct and the sole viviparous species among that speciose genus, was summarized by Wake (1978a). Gravid females were collected between April and August, each with three to six large, mature eggs, 3.3–5 mm in diameter (Drewry and Jones, 1976; Wake, 1978a). The fertilized eggs are retained within a chamber formed of fused portions of the oviducts (Wake, 1978a). Approximately 33 days after amplexus, three to five tiny, fully metamorphosed froglets are born (Drewry and Jones, 1976). Like other species with direct development, Eleutherodactylus jasperi embryos have poorly developed mouthparts, lacking denticles, and adhesive organs (Wake, 1978a). Among “fetal adaptations,” the larval spiracle and external gills are present but transitorily, and rates of development are modified compared to those of direct-developing species (Wake, 1978a). The froglets retain a small egg tooth, characteristic of the lineage. They have a large, thin, highly vascularized fan-like tail throughout much of tadpole/froglet development, but resorbed by birth. The tail may function in intraoviductal gaseous exchange. The source of nutrition for developing embryos appears to be solely egg yolk; unresorbed yolk was still present in froglets after birth (Wake, 1978a). Thus E. jasperi is ovoviviparous.

Some aspects of the reproductive biology of several of the 13 species of Nectophrynoides have been explored, especially with the discovery of several new species that have very restricted ranges in Africa (e.g., N. asperginis) and are considered endangered; there is significant interest in captive breeding of these animals. All of the species are ovoviviparous; fertilization is internal and larvae are retained within the female, with juvenile toadlets being born. N. asperginis eggs are 2.4 mm in diameter (Poynton et al., 1998). Lee et al. (2006) reported that the clutch size is 24–28 eggs; Channing et al. (2006) state that clutch size varies from 5–13 offspring (causing me to wonder whether the former number is of mature ovarian ova, and the latter metamorphosed froglets). In captivity, births occurred between mid-December and the end of April; calling and amplexus occurred among frogs less than a year old (Lee et al., 2006). Thibaudeau and Altig (1999) report that the embryos develop two pairs of external gills, whereas gills are absent in other species in the genus. Wake (1980a) summarized features of the development of N. tornieri and N. vivipara from the literature and her own observations: the egg diameter in both species is 3.0–4.0 mm diameter; clutch size (numbers of oviductal tadpoles) in N. tornieri is 11–31 and 114–135 in N. vivipara (a significant difference). Both lack mouthparts (as does N. asperginis), although a “ridge” develops in N. tornieri where the labrum would be (Orton, 1949). The gut is short in each (correlated with a yolk resorption, nonfeeding biology); the tail in both species is reduced, shorter than snout-vent length and with low fins. Thibaudeau and Altig (1999) report that limb buds emerge simultaneously in N. tornieri. We have a comparative study of development under way in my lab; we are developing “normal tables” that reflect ontogenetic sequence changes, such as increased cephalization in Nectophrynoides relative to direct-developers (perhaps concomitant to their short developmental period), and those of mouth organization, the gut, and the skin. We see no increase in vascularization of the tail. We have scanning electron micrographs as well as histological preparations of the oviductal larvae, but not yet complete ontogenetic sequences.

Oviductal retention with internal nutrition provided after yolk is resorbed is documented only for Nimbaphrynoides occidentalis (formerly included in the genus Nectophrynoides). Francoise Xavier, in a series of excellent articles, described the ecology, reproductive cycle, development of the young, the state of the ovaries and oviducts, and the basic endocrinology of N. occidentalis. [See Xavier (1977, especially Fig. 1, 1986) for summaries]. As summarized by Wake (in press), Xavier found that N. occidentalis has internal fertilization by cloacal apposition; the gestation period is 9 months, specifically correlated with the cold and dry and warm and rainy seasons; the maternal female provides nutrient material to the maturing “tadpoles” after yolk is fully resorbed, and it is secreted by a region of the anterior end of the oviduct. Ovulation and fertilization take place in late October-early November, following an intergravidic period during which the ovarian follicles mature and secrete estrogens. After ovulation, corpora lutea form and secrete progesterone. The hormone inhibits new oocyte maturation and slows embryonic development, but maintains some secretory activity of the oviductal mucosa. After mating, the female goes underground during the dry season (approximately mid-October to mid-April) and emerges in April with the inception of the rainy season, dwells actively above ground for about 3 months. When she emerges, the corpora lutea begin to degenerate, progesterone secretion decreases, her uterine mucosa becomes more secretory and hyperemic, and the embryos undergo rapid growth through metamorphosis (Xavier, 1977, summarized in Fig. 1). The yolk of the small ova is presumably fully resorbed during the slow-growth embryonic period; subsequently, the developing tadpoles hatch and are able to feed on nutrient material secreted by the uterine mucosa (Vilter and Lugand, 1959). Birth occurs in June, in midrainy season, so that prey is abundant and approximately 7.5 mm total length (TL), 30–60 mg froglets able to feed immediately. Xavier demonstrated that estrogen conditions the pregnancy and progesterone maintains it, with the diminution of progesterone secretion facilitating the rapid growth-through-metamorphosis phase of tadpole development. The ovarian and uterine cycles are highly correlated with habitat conditions and seasonality and concomitantly, embryonic-tadpole-froglet development is mediated, involving a number of fetal adaptations.

The eggs, embryos, and froglets have a number of modifications concomitant with their intraovoductal period. The eggs are small (∼0.6 mm diameter), with relatively little yolk; clutch size is 4–35. The tight correspondence/control with the maternal female's endocrinology and morphology and her environment induces a lengthy period of early development (the first 5 months, during the dry season), then a rapid developmental phase through metamorphosis as the female emerges, feeds, and progesterone secretion diminishes. Also, tadpole morphology is unlike that of the majority of frog species in several regards: mouthparts are absent, and a set of epidermal papillae surround the mouth; Lamotte and Xavier (1972) suggested that they act as a “mop” for the attraction of the nutrient secretion. The spiracle is closed; the gut is reduced, short, and curved, and the tail is longer than the head and body, with low fins. See Xavier (1977) and Wake (1980a) for summaries. Such fetal adaptations that involve major changes in the developmental trajectory and nutritional provision have evolved in the species.

Fetal Adaptations for Viviparity in Salamanders

Oviductal live-bearing without and with maternal nutrition

Obligate viviparity is well documented for only one species of salamanders, S. atra. It has a 2–5 year gestation period and provision of maternal nutrition by ingestion of modified oviductal epithelial mucosal cells in an anterior region of the oviduct, the zona trophica (Browning, 1973; reviewed by Guex and Greven, 1994). As Wake (in press) summarized, viviparity also has been reported for the closely related Lyciasalamandra luschani, which gives birth to two fully metamorphosed young, one from each oviduct, but which lacks a zona trophica and, therefore, presumably does not supply maternal nutrition (Ozeti, 1979; Polymeni and Greven, 1992; Olgun et al., 2001); S. algira of Morocco is also reported to have viviparous populations (Donaire-Barroso et al., 2001). It is not clear whether either of these species is entirely yolk dependent or not; the authors defined “viviparity” solely in terms of giving birth to fully metamorphosed young, rather than the two characters (provision of maternal nutrition; birth of metamorphosed young) that I prefer see expressed. In addition, females of the distantly related European plethodontid Speleomantes sarrabusensis are reported to give birth to metamorphosed young (Lanza and Leo, 2001), but details of the mechanism by which this occurs are not available. I expect that most, if not all, of these taxa likely will prove to be ovoviviparous, yolk dependent, when their reproductive biology, including maternal and fetal modifications for live-bearing, is better known. Proteus anguinus (Family Proteidae) had been considered viviparous (e.g., Nusbaum, 1907), but recent observations in captivity of courtship, egg laying, and hatching of larvae counter that idea (Juberthie et al., 1996).

The ovoviviparity of several subspecies of S. salamandra has been relatively well studied. The developing embryos are typically retained in the oviducts until they reach mid- to advanced larval stages, when they are “born” in ponds, streams, ditches, water troughs, and so forth, feeding in the water where they complete metamorphosis. They are nutritionally dependent on their yolk while in the oviducts; it typically is resorbed by the time of birth. The young that hatch in the oviduct and are subsequently retained for some time before birth as larvae or near-metamorphs are thought to ingest lysing eggs and possibly even less well-developed siblings (Greven and Guex, 1994). The maternal oviductal mucosa becomes thick and hyperemic before ovulation, thought to condition the “environment” for retention of the embryos, a response to estrogen. Greven (1980, 2002) and Greven and Rüterbories (1984) extensively studied the oviduct before and during pregnancy, and Greven compared that of the ovoviviparous and viviparous species of Salamandra (1977, 1998, 2003a,b); he and Guex (1994) summarized data for S. salamandra. They found that the oviduct is regionalized in both species, and the lower part is modified as a chamber for development of the young (the “uterus”). However, they found that the “uterus” of the ovoviviparous forms did not exhibit the cyclical morphological changes that occur in the euviviparous S. atra, concomitant with the latter's production of nutrient material. Joly and Picheral (1972) found no effect on the course of gestation from removal of the ovaries in S. salamandra, and suggested that while the hypervascularization of the “uterus” is a response to estrogen, prostaglandins, and/or local stimuli may also be involved. Joly (1986) and Joly et al. (1994) also provided summaries of information about reproductive strategies in S. salamandra.

There appear to be relatively few modifications of the pattern of development of the embryos through to the larval stage in most of the subspecies (although few have been rigorously assessed) that would be specifically associated with live-bearing (but see later for the exception). For example, S. salamandra have relatively large clutches; at ovulation, all of the eggs receive a full set of jelly coats as is typical for most amphibians. Greven and coworkers (summarized by Greven and Guex, 1994) have carefully examined the egg membranes, the skin of the developing young, and of the oviductal mucosa to assess osmotic balance and ion transfer; there appear to be physiological differences, but not significant morphological modifications.

The obligate viviparity of S. atra, on the other hand, is well studied, but mostly dealing with the morphology and physiology of the maternal parent. To briefly summarize those aspects: Vilter and Vilter (1960, 1964) discovered that corpora lutea are present throughout pregnancy, but that their numbers, sizes, and activity decrease late in gestation (probably regulating growth and initiating oogenesis). Greven and coworkers (summarized by Greven and Rüterbories, 1984; Guex and Greven, 1994; Greven and Guex, 1994; Greven, 2002, 2011) and Joly (Lostanlen et al., 1976; summarized in Joly et al., 1994) thoroughly examined oviduct morphology and function in pregnant and nonpregnant females of the viviparous S. atra, especially the oviductal epithelium and its secretions, including the egg jelly and the nutrient material secreted by the mother (Greven, 2003a) and the specific region, the zona trophica, located in the anterior part of the duct, which produces the nutritive cells/material (Greven, 1984).

In contrast to the situation for S. salamandra, there are several modifications that can be construed as fetal adaptations for viviparity in S. atra. Greven and coworkers examined the development of the embryos and the way they obtain the nutrient material secreted by the zona trophica (Greven, 1984; Guex and Chen, 1986; Greven, 2003b). They note that it has long been known that only two embryos fully develop, one maintained in each of the oviducts. An important facilitator of the one-embryo-per-oviduct developmental mode, and potentially maternal nutrition, is that only one ovum, probably the first (or second) one receives a full set of jelly coats (summarized in Guex and Greven, 1994). Because jelly coats are necessary for effective fertilization, and then for prevention of polyspermy, only the single-embryo-per duct receives all of the maternal nutrients. Guex and Greven (1994) report that when the embryos first hatch, they practice oophagy, eating disintegrated, perhaps unfertilized ova (and possibly less advanced embryos). More advanced embryos develop in a posterior, aglandular, part of the oviduct but move about in the duct. Teeth develop during the “larval” phase of development in Salamandra (Clemen, 1978a,b; Clemen and Greven, 1994; see Fig. 5), but in addition a “fetal” dentition develops later on the medial components of the jaws of S.atra (Greven, 1984; Guex and Chen, 1986; Greven, 2003b). Its tooth series are aligned at a distinct angle in S. atra and are thought to be the armature used in scraping the epithelium and ingesting the epithelial nutrient material elaborated by the zona trophica, the mucosal cells detaching from the connective tissue below as a consequence of partial necrosis (Greven, 1984; Guex and Chen, 1986; Guex and Greven, 1994; Greven, 2003b but see Greven's (1998) comments cited later). Guex and Chen (1986) developed a model (see their Fig. 10) for tooth-jaw action for the presumed fetal feeding mode. Necrosis and regeneration of the epithelium continue to cycle in different parts of the zona trophica throughout the rest of the gestation period so that the young have abundant nutrition via epitheliophagy (Guex and Chen, 1986; Guex and Greven, 1994). However, Greven (1998) stated that the factors that induce and maintain the zona trophica are largely unknown, and he insisted that development of the zone “could not be induced by estrogen and/or progesterone,” apparently because of the decline in number and function of corpora lutea during the period of active feeding. He also stated (contra Guex and Chen, 1986) that direct stimulation of the zone by the larval dentition is improbable, and that secretion of stimulatory substances by the fetuses, “perhaps prostaglandins, may trigger the induction of the trophic epithelium.”

Finally, among potential fetal adaptations in Salamandra, it has been conjectured whether or not the embryonic/larval gills might be involved in nutrient uptake, at least that of glycoproteins, but the idea was dismissed by Greven (1998). The gills in S. salamandra are triramous and fimbriated, typical of those of many salamanders and are not large, suggesting that they may not be very active in uptake. Curiously, the large, highly fimbriated, well-vascularized gills of S. atra (see figures in Wunderer, 1910; Haefeli, 1971) have not been implicated in nutrient uptake, assumed to be only for gaseous exchange (but not tested). It is apparent that patterns of development of the embryos/young in both species should be carefully examined, following on the reports of the 1960s and 1970s, as initiated and summarized by Greven and Guex (1994), Guex and Greven (1994), Joly et al. (1994), and Greven (1998, 2002). Indeed, much research remains to be done on all aspects of the maintenance of pregnancy in Salamandra; I expect that, as suggested by some authors, hormonal mediation will prove to be important to a number of aspects of pregnancy and the development of fetal adaptations.

Live-bearing and intraoviductal cannibalism

A major variation on the theme of intraoviductal maintenance of embryos and fetuses is the intraoviductal cannibalism that has evolved, clearly recently, in two subspecies of S. salamandra in northern Spain (Dopazo and Alberch, 1994; Alcobendas et al., 1996; Dopazo and Korenblum, 2000; García París et al., 2003; Buckley et al., 2007). The subspecies are not each other's sister taxa (Buckley et al., 2007), so the patterns of development, including the cannibalism, constitute a homoplasious situation. Cannibalism is itself a major fetal adaptation for live birth; the modification of the ontogenetic trajectory of the cannibalistic morph in comparison to that of noncannibalistic, ovoviviparous conspecifics includes a number of modifications presumed to be adaptive. The evolution of the system and the ontogenies of the two morphs are summarized by Buckley et al. (2007). The two viviparous, cannibalistic subspecies, bernardezi and fastuosa, are striped, and give birth to fully metamorphosed young; their respective nearest ovoviviparous taxa, S. s. gallaica and S. s. terrestris, are spotted, and give birth to larvae that live, feed, and metamorphose in water. The latter subspecies, like most of the subspecies of S. salamandra, are ovoviviparous, completely dependent on yolk as the nutrient that supports their development. The cannibalistic subspecies are also yolk dependent, but as they resorb their own yolk, they begin to ingest sibling eggs, and even developing embryos that are less advanced than they (see figures in Buckley et al., 2007). This mode constitutes a preprovisioning, nonsecretory form of nutrient supply. The pattern of development in the cannibalistic subspecies differs from that of the ovoviviparous subspecies in a number of features, as noted above, time to birth is shorter, and they are born fully metamorphosed; furthermore, their ontogenetic trajectories are modified, presumably as adaptations for the cannibalistic mode. The heterochronic events include: jaws, jaw musculature, and teeth develop precociously; the gape is larger; the tail fins are lower (concomitant to the absence of an aquatic larval phase); and development through metamorphosis is completed much faster (90 days postfertilization), and before birth, in contrast to the 120 days of intraoviductal development and birth of larvae in their sister subspecies (Buckley et al., 2007).

The pattern of evolution of the subspecies, and, therefore, presumably of the evolution of viviparity, has been well studied using molecular markers (Alcobendas et al., 1996; García París et al., 2003). The viviparous populations are not geographically isolated, but occur within the continuous range of distribution of the species on the Iberian Peninsula and are surrounded by ovoviviparous populations. García París et al. (2003) developed a paleogeographic scenario and population model to explain the evolutionary history of the system and concluded that viviparity (the cannibalistic mode) arose only once in S. salamandra. However, given our new data (Buckley et al., unpublished) showing that the cannibalistic subspecies are not sister taxa, more work remains to be done. The reproductive strategies and life history traits appear to be correlated with modifications in the ontogenetic trajectories in the populations, and new adaptive pathways. Investigation of the ontogenies of the viviparous and ovoviviparous populations, pregnancy maintenance, and interactions at hybrid zones continue.

Fetal Adaptations for Viviparity in Caecilians

In contrast to the situation in frogs and salamanders, viviparity has evolved independently at least four times in caecilians; it characterizes almost one-third of the species for which we know, or can infer, reproductive modes (see, e.g., Wake, 1977a,b, 1989; Exbrayat, 2000, 2006c; Gower et al., 2008). All caecilians, so far as is known, have internal fertilization via the male's insertion of his everted posterior cloaca into the vent of the female and transferring sperm directly, so retention of fertilized ova is facilitated. The gestation period of live-bearers, known for only a few species, ranges from 6 to 11 months, depending on the species. Ova are reduced in size, approximately 2 mm diameter or less; clutch number also appears to be reduced relative to species with free-living larvae (Wake, 1977a,b, 1980b; Exbrayat and Hraoui-Bloquet, 2006). Ovoviviparity is not (yet) known for caecilians; all viviparous species apparently provide maternal nutrition after embryos resorb their yolk early in the gestation period (Wake, 1985, 1993; Wake and Dickie, 1998; Exbrayat, 2000, 2006a–d), and all give birth to apparently fully metamorphosed young. In Dermophis mexicanus, the gestation period is 11 months; in the population investigated, females are synchronous in having ova fertilized in June–July, an 11-month gestation period, and birth in May–June at the inception of the rains and increased prey abundance (Wake, 1980b). Exbrayat and coworkers (summarized by Exbrayat, 2006a–d) observed a shorter gestation period (7–9 months) in Typhlonectes compressicaudus, with events synchronized but to different months; nevertheless, they concluded that ecological parameters are correlates of gestation and timing of birth.

There are numerous fetal adaptations for viviparity in caecilians. Most, if not all of them, appear correlated with and likely precipitated by endocrine changes in the maternal female. They include a fetal dentition and epitheliophagy, heterochronic changes in the embryonic growth trajectory such that the feeding apparatus and gut develop earlier than in nonviviparous taxa, probable gill use for gaseous and perhaps nutrient exchange, a disjunct metamorphosis, and others.

Hatching from the egg membrane in D. mexicanus occurs concomitantly with yolk depletion and the development of the jaws and their musculature and a species-specific “fetal” dentition at approximately 3 months into the 11-month gestation period (Wake and Hanken, 1982). Furthermore, the oviductal mucosa just at the time of hatching proliferates and becomes hyperemic, highly vascularized, and secretory (Wake, 1993). The dentition [the shape of which changes dramatically over the gestation period (Wake, 1976, 1980c)] is apparently used to scrape the oviductal mucosa of its cells and their contents as the developing fetuses move about the lengths of the oviducts. The fetal teeth are shed very shortly after birth, with the adult dentition (characterized by tooth crown shape and position) erupting within approximately 48 h after birth (Wake, 1980c). Exbrayat (2000, 2006c,d) and Exbrayat and Hraoui-Bloquet (2006) found a similar phenomenon in Typhlonectes compressicauda. Wake (1977b, 1978b) and Hraoui-Blouquet and Exbrayat (1996) have described the development of the fetal teeth of Typhlonectes compressicauda; the shapes of their tooth crowns change only slightly during fetal ontogeny. It must be noted that the earlier descriptions of caecilian “fetal” teeth by Parker (1956) and Parker and Dunn (1964) included the assertion that the fetal dentition in caecilians was retained from ancestors and functionless. Current evidence indicates independent evolution of such a dentition in viviparous lineages and significant function in intraoviductal feeding. Furthermore, it is now apparent that many direct-developers have a “fetal” dentition that differs markedly from that of the adults and is shed at hatching (e.g., Perez et al., 2009), and skin feeders also have a modified juvenile dentition (Kupfer et al., 2006; Wilkinson et al., 2008). The diversity of preadult dentitions that may have developed independently in several lineages of caecilians is striking; they have substantial differences in tooth crown morphologies and distributions on the jaws, many/most apparently species-specific, as well as somewhat different functions. Even more noteworthy is the convergence of the development of “fetal” dentitions in the live-bearing caecilians and Salamandra, apparently involved in epitheliophagy in S. atra (Greven, 1984; Guex and Chen, 1986; Guex and Greven, 1994), and perhaps in cannibalism in subspecies of S. salamandra.

Unlike the situation in S. atra, there apparently is not a specialized region for nutrient proliferation in live-bearing caecilians; virtually the entire mucosa is involved (Wake 1970, 1972; Exbrayat, 1988; Hraoui-Bloquet et al., 1994; Exbrayat and Hraoui-Bloquet, 2006; Exbrayat, 2006a,c). The ovaries of pregnant females are characterized by the presence of several large corpora lutea, as well as maturing ova (Wake, 1968). The corpora lutea appear to be present during the entire gestation period, at least in D. mexicanus, and fully metamorphosed young are born. Exbrayat (2006a; Exbrayat and Estabel, 2006) found similar structure in T. compressicauda. Wake (1993) measured circulating progesterone during late pregnancy, and found it to be five times that present in nonpregnant females. Because of the activity of the fetuses, the proliferation of secretory mucosa becomes confined to crypts in the epithelium lining the oviducts as the luminal epithelium is ingested. The endocrinology of several aspects is not known: a) the inception of the fetal dentition and the fact that in D. mexicanus the tooth crown shapes vary with fetal “age” (Wake, 1980c), b) the mediation of the changes in composition of the nutrient material elaborated by the maternal oviductal mucosa (initially free amino acids and simple carbohydrates, complex amino acids and carbohydrates in midpregnancy, and proteins, carbohydrates, and lipids during the last months of gestation: Welsch et al., 1974; Wake, 1993; Wake and Dickie, 1998), and c) the regulation of the disjunct metamorphosis (Wake, 1986; Exbrayat, 2006a,b). Exbrayat and coworkers made very similar observations of pregnancy, ovarian and oviductal morphology, the fetal dentition, the biennial female cycle, and metamorphosis, in their study species, the aquatic typhlonectid Typhlonectes compressicaudus (see, in particular, Exbrayat and Collenot, 1983; Exbrayat, 1988; Hraoui-Bloquet et al., 1994). They also studied development (e.g., Exbrayat and Delsol, 1988; Sammouri et al., 1990) and have presented useful summaries (Exbrayat, 2000, 2006a–d; Exbrayat and Estabel, 2006).

In all caecilians except typhlonectids, so far as is known, the gills are external and triramous (diramous has been reported for Gegeneophis carnosus: Ramaswami, 1954), fimbriated structures. In viviparous species, such as Gymnopis multiplicata and Dermophis mexicanus, gills are well developed at hatching and grow for some 2–3 months after hatching (Wake, 1967, 1969). They are extended during active movement in the oviducts. They begin to resorb (or are constricted and drop off about midway through gestation, but nubs of the gills and an open spiracle can remain for some months (see Wake, 1967, 1969, 1982). The gills likely function in gaseous exchange, at least for the period of development during which they are large and fimbriated, but there is limited physiological evidence for that capacity.

In typhlonectids, all of which are presumed to be viviparous, the gills are large, sac-like, paired structures formed from the fusion of the three rami of the gills. The three aortic arches involved are homologous to those of other caecilians, of course, and have basically the same extensions into the body of the sac-like gill (Wake, personal observation; Hraoui-Bloquet and Exbrayat, 1994), but different external proximal associations (Wilkinson and Nussbaum, 1997). Typhlonectids retain their gills until, and sometimes shortly after birth; in the oviducts, they are often flattened against the oviduct mucosal wall and the body of the fetus, hypothetically an excellent position for gaseous exchange (Wake, personal observation). Some data on maternal-fetal gas exchange via the gills are available for Typhlonectes compressicauda. Toews and MacIntyre (1977) found that advanced fetuses have a much higher O2 saturation than maternal females, and suggest that the high hemoglogin saturation in fetuses would, therefore, be achieved even though maternal saturation is moderate, as a feature of maternal-fetal gas exchange. Garlick et al. (1979) reported similarly that Typhlonectes compressicauda has a shift in equilibrium and kinetics of oxygen binding by blood and hemoglobin, with the oxygen affinity of fetal blood higher than that of adult blood. Fetal and adult hemoglobins appear to be biochemically identical. Garlick et al. state that the fetal to maternal shift in O2 equilibrium is mediated solely by the different in ATP content of the blood cells, the maternal erythrocytes having three times as much ATP as fetal. It is not clear what impact, if any, on fetal survival this phenomenon might have, save for the likely increase in efficiency of O2 uptake by fetuses.

The unique modifications of the gills in typhlonectids are intriguing in terms of fetal adaptation and potential function. In fact, Exbrayat and Hraoui-Blaquet (1991, 1992a,b, 2006) concluded that the gills of Typhlonectes compressicauda are modified for uptake of nutrient material secreted by the oviductal mucosa, such that they are “pseudoplacental” and absorb nutrients in addition to those received through oral ingestion of oviduct mucosal cells and secretions. They followed and added to Delsol et al. (1986) and Exbrayat's earlier work (Exbrayat et al., 1981-1983; Exbrayat and Delsol, 1988). The ultrastructural morphology they describe, with club shaped and ciliated cells, the vascularization of the sac-like gills, their disposition appressed to the oviductal mucosa (which lacks an epithelium late in gestation owing to feeding (Exbrayat, 2006 c,d) and to the body of the fetus, all suggest at least gaseous exchange, but uptake of the large complex nutrient components characterized by Welsch et al. (1974) for the closely related typhlonectid Chthonerpeton and by Wake (1993) is more difficult to accomplish with such a system unless it is highly selective and can segregate different constituents. This is not to say that such uptake is impossible, but it requires demonstration. Obviously, gill use during gestation requires further investigation to test mechanisms of exchange to understand fetal maintenance during development.

It is also noteworthy that Delsol and coworkers (Delsol et al., 1981; Exbrayat et al., 1981-1983; Exbrayat and Delsol, 1985) proposed that a patch of ventral skin in embryos of T. compressicauda serves as an “ectotrophoblast” facilitating nutrient uptake (reported by Sammouri et al., 1990; Exbrayat, 2006d). It is present during presumed oral ingestion, and is short lived. Furthermore, Exbrayat and Hraoui-Bloquet (2006) report that intraoviductal fetuses ingest ova and even other fetuses, such that cannibalism may occur. [I have observed dead, desiccated fetuses in oviducts along with living ones in D. mexicanus, but have no evidence of fetuses ingesting eggs (none are present in the duct with the fetuses after hatching) or other fetuses.] It is indeed intriguing that the typhlonectids with their many adaptations for an aquatic or semiaquatic lifestyle and for viviparity might possess such a host of unique and highly derived structures for nutrient uptake. Further investigation will be worthwhile to increase our understanding of caecilian evolution.

Metamorphosis

A major element of development in live-bearing amphibians (as well as direct-developers) that may be a significant fetal adaptation is the modification of metamorphosis. Amphibian metamorphosis is usually thought, based on the common frog model, to be an abrupt set of morphological and physiological changes, often correlated with change from aquatic (larval) existence to terrestrial (adult) habitats. This generalization, however, begs the question of metamorphosis in permanently aquatic taxa, and that of the relatively prolonged and disjunct metamorphosis in many salamanders and probably most caecilians. Furthermore, a number of “larval” features may never develop in some direct-developing frogs, and perhaps other less well-studied direct-developing amphibians; loss and/or reduction of “larval” features has not received specific attention in studies of live-bearing amphibians, but there are hints (see later). The evolution of metamorphosis in amphibians has been considered in a number of studies, largely focused on specific issues or elements of morphology. These include, among many, Szarski's (1957) pioneering work on the origin of the larva and amphibian metamorphosis, Fritzsch's (1990) work on the evolution of metamorphosis of the rhombencephalon, the lateral line, and the inner ear (of frogs) as reflective of the process by which aquatic and terrestrial components of the life cycle are united, Reiss's (1996) examination of palatal metamorphosis in rhinatrematid caecilians as evidence for the monophyly of the Lissamphibia and his 2002 work on the phylogeny of amphibian metamorphosis, Alberch's (1989) and Wake's (1986, 1994) comments, all interesting perspectives on the nature of metamorphosis and its evolution. However, despite this work, there remain rather few modern reviews of metamorphosis that rigorously compare diverse lineages in all three orders of amphibians, and the “frog model” that emphasizes metamorphosis as an abrupt set of correlated events persists. In fact, some frogs (especially direct-developers), a number of salamanders, and probably most caecilians metamorphose over time, rather than abruptly, and the events involved may be disjunct and the period of metamorphosis prolonged. Szarski (1957) stated that most amphibians have two phases of metamorphosis, one at about hatching, the other the transformation of larvae to adults. He stated that (in caecilians) “The second metamorphosis is a very long process,” and that “The metamorphosis of Urodela is similarly prolonged. It is difficult to fix the exact limits of duration as there is a great individual variability in every species.” He contrasted these taxa with frogs, commenting that “Metamorphic processes progress at a much greater rate in Anura; they are more synchronized and more dependent on thyroid hormone.” His ideas remain prescient, even though we now have substantially more information about metamorphosis and its endocrinology. Wake (1966) commented on “differential metamorphosis” in salamanders, including plethodontids, in his discussion of paedomorphosis. He stated “Differential metamorphosis is a term that describes the paedomorphic pattern in which metamorphic processes are extended over a considerable period of time, with some elements completing metamorphosis early, others very late, and some not at all.” He implies significant disjunction of events, and considers the condition to be more characteristic of terrestrial plethodontids and “of great importance in providing an escape from specialization and access to new adaptive zones.” Paedomorphosis occurs only in salamanders among adult amphibians, so I will not consider it further, nor will I deal with the extensive literature on direct development (except to say that “loss of the larval stage” is a misnomer, the free-living larval phase is obviated, but the events of transformation from late embryo/larva to adult take place before hatching as a fully metamorphosed juvenile, another example of prolonged metamorphosis in frogs, salamanders, and caecilians). For caecilians, it has long been known that metamorphosis is prolonged and disjunct, as Sarasin and Sarasin (1890) and Breckenridge et al. (1987: p. 437: “Metamorphosis itself is gradual, spread over an appreciably long period, and is not as dramatic as in anurans.”), among others, have described for Ichthyophis glutinosus, a basal taxon that is oviparous with a lengthy larval period. Wake (1977a,b; Wake and Hanken, 1982) and Exbrayat and Hraoui-Bloquet (1994, 1995) briefly discuss the phenomenon, particularly as “complicated” by viviparity.

And therein lies my present concern: with respect to live-bearing frogs, salamanders, and caecilians, again, little attention has been paid to embryonic/fetal development, but the few data and discussions available in the literature, particularly regarding metamorphosis, are tantalizing. For frogs, Lamotte and Xavier's (1986) statements are direct; they make the point that even though growth and metamorphosis accelerate after the emergence of the female with the first rains and concomitant with her endocrinological and morphological changes, “Par rapport au developpement normal des tetards d'Anoures, l'evolution intrauterine des embryons de Nectophrynoides occidentalis presente en effet des particularities remarquables.” Metamorphosis is not the abrupt set of events as in most frogs, but has a particular sequence with “early” development of the hind limbs and slow loss of the tail, occurs over four stages of development (that Lamotte and Xavier characterize), and takes approximately 60 days, the last 2 months of the 9-month gestation period. They also indicate that the maternal nutrient secretion is not unlimited, given that members of small clutches are much larger at birth than are those of large clutches [members of clutches of fewer than five are 7.3–10.5 mm body length (and those of clutches of 2 are ∼9.5 mm); those of clutches of 15 are 6.3–7.5 mm (from Lamotte and Xavier's Fig. 23)]. Furthermore, apparently there are no traces of larval mouthparts (e.g., cartilages) during development (as occurs in some direct-developers); Lamotte and Xavier comment on and illustrate only the papillae that surround the mouth, and that they speculate may be involved in ingesting the maternal nutrient secretion. The spiracle also does not develop.

Similarly, Nectophrynoides species, ovoviviparous but with young born fully metamorphosed, lack adhesive organs, labial teeth, and horny beaks, which Orton (1949) equates with the similar losses in direct-developers and “other modifications of the life history.” She reports that N. tornieri has a trace of a larval lower lip, the jaw cartilages and musculature are “essentially larval in pattern,” and that general proportions of the head and body are “greatly modified,” tiny gill arches, gill slits, and external gills develop, as well as other such features. Orton (1949) speculated that changes in the development of the mouth might be correlated with some feeding capacity, but I suggest that they are simply the kinds of modifications/losses that occur in most direct-developing frogs that have been studied. The tails of Nectophrynoides spp. are long and slender (personal observation). The length of gestation for N. tornieri is 60–90 days (Orton, 1949); Lee et al. (2006) report that the gestation period for N. asperginis in captivity is 30–60 days. Eleutherodactylus jasperi, the New World ovoviparous frog that is born fully metamorphosed, has a different strategy from that of Nectophrynoides in several ways. The comparison is summarized by Wake (1978a), and I add more recent information about N. asperginis (Channing et al., 2006; Lee et al., 2006). E. jasperi has a reduced clutch (1–6 vs. 5–13 in N. asperginis, 9–35 in N. tornieri, and 114–135 in N. vivipara; eggs are larger (3–5 mm diameter vs. 0.9 mm in N. vivipara and 2.0 in N. tornieri and N. asperginis), and the tail is large, fan like, and well vascularized, in contrast to the slender tail of Nectophrynoides. Larval mouthparts are lost, but the egg tooth, typical of all Eleutherodactylus (and Craugastor), is retained. Gestation is very short in E. jasperi, approximately 33 days. Drewry and Jones (1976) observed “large embryos” in a female on April 25, 1974, and birth of froglets on May 14, so it is possible that metamorphosis is not particularly abrupt, given the short duration of the gestation period, but an ontogenetic sequence is necessary to provide data, this is not likely to occur, because the species is likely extinct. In the absence of any published information such as normal tables for any Nectophrynoides, I cannot comment on the nature of their metamorphosis.

A prolonged or disjunct metamorphosis occurs in the live-bearing Salamandra, in both S. salamandra and S. atra, but with different attributes. Buckley et al. (2007) described the heterochronic changes in the developmental trajectories of the derived, cannibalistic S. salamandra subspecies; the fact that the cannibalistic morphs are born fully metamorphosed, and in a shorter period of time but with a different course than the subspecies born as larvae, illustrates a way that metamorphosis may be altered by several (as yet unidentified) factors. Different modifications of development and metamorphosis occur in S. atra. The work by Wunderer (1910) and especially that of Haefeli (1971) makes the point that the length of the gestation period is correlated with altitude, and presumably environmental conditions. Haefeli's graphs of growth show increased growth during summer months, and slow growth (or none) during the winter, as well as a longer gestation at higher altitudes. His photographs of specimens show that metamorphosis itself (based on features of the gills, the skin, the tail, etc.) is slower at higher altitudes but also that the events of metamorphosis (gill development and loss, tail shape change) take place over many months, fewer at lower levels, but still over a substantive amount of time.

Caecilians have long been reported to have relatively lengthy metamorphoses, and this appears to characterize all species (but see later regarding reports for Typhlonectes). Basal taxa with hatchlings having a free-living larval period typically resorb gills early in the larval period, with small stubs and a spiracle that remains open for some time, and the tail fin takes weeks to fully resorb (e.g., Breckenridge et al., 1987; Sarasin and Sarasin, 1890 for Ichthyophis glutinosus); in addition, in Epicrionops, such osteological features as the squamosal and the maxilla develop late in the metamorphic period (personal observation), and the elements of the palate also reach their adult configuration late in metamorphosis (Reiss, 1996), and well after the inception of the gill and tail fin changes (personal observation). Relatively prolonged metamorphic trajectories appear to characterize direct-developers as well, so far as is known from the very few for which there are ontogenetic data [e.g., Hypogeophis rostratus (Brauer, 1897, 1899; Müller, 2005) and Caecilia orientalis (Pérez et al., 2009)]. Tentacle development, appearance, and protrusion occur “at metamorphosis,” but the timing of these events apparently takes days to weeks in the species examined (see Billo and Wake, 1987). Furthermore, skin modifications, gill loss, and other such features also take place over considerable developmental time [few studies present “stages” as correlated with real-time development but illustrations of the embryos-to-“hatchlings” indicates this (e.g., Fox, 1985, 1986; Sammouri et al., 1990)].

Two viviparous species have been reasonably well studied: Dermophis mexicanus (in the terrestrial Dermophiidae) and Typhlonectes compressicauda (a member of the aquatic Typhlonectidae). The former has a number of developmental features shared with terrestrial direct-developers (e.g. triramous gills); the latter several correlated with an aquatic mode of life; the two species also share a number of features that appear to characterize viviparous taxa. Exbrayat and Hraoui-Bloquet (1994, 1995) briefly examined what was then known about metamorphosis in caecilians, with particular reference to that of viviparous taxa. They make the point that the development (and retention, of some elements) of a number of features of viviparous embryos/fetuses has a different trajectory than that of oviparous taxa either with a free-living larval period or direct development. They briefly consider the gills, skin, ossification, the digestive tube, sensory organs, and so forth, and consider the differences among taxa to elements of heterochrony. Furthermore, they considered metamorphosis to be short in Gymnopis multiplicata with a lengthy postmetamorphic intraoviductal period before birth (not accurate; see Wake, 1967; Billo and Wake, 1987); furthermore, my unpublished data on G. multiplicata development and metamorphosis show it to be very similar to that of D. mexicanus (see later). Strikingly, they considered metamorphosis in T. compressicauda to be short, followed by a long (nearly half of the 6-month gestation period) “placental structure” intraoviductal period before birth. Given that a) the gills form the “placenta” and are retained until birth or shortly after, b) the fetal teeth are retained similarly, and likely are involved in feeding on the maternal nutrient secretions, c) the skin retains its fetal condition until birth, and d) several other such fetal/“larval” features are retained for varying lengths of the posthatching period (see Sammouri et al., 1990), I must conclude that metamorphosis in T. compresssicauda, and presumably all typhlonectids, is disjunct in terms of some events and quite prolonged [4 of the 6 months' gestation period, using Exbrayat's and Hraoui-Bloquet's (1995) data and definition (e.g., metamorphosis characterized by gill loss)].

Metamorphosis in Dermophis mexicanus includes a number of disjunct events, and occurs over the last 4 or 5 months of the gestation period. Gills are triramous, long, and filamentous during at least the first half (or more) of the gestation period; they are then resorbed, but the spiracle remains open until approximately the 8th month of gestation. The fetal teeth first develop at hatching, about 2.5–3 months into the gestation period, and increase in number of rows and of pedicels and crowns through the rest of the gestation period. They are lost at or shortly after birth, perhaps the signal of the end of metamorphosis. Tentacle begins development at 14–15 mm TL and many of the components are present in 58 mm TL fetuses, at which point the tentacle is perforate and open to the exterior. Additions and growth continue until birth; no new structures are added subsequently, although movement and growth continue for much of the first year postbirth [see Billo and Wake (1987) for data]. The trajectory in the closely related, also viviparous Gymnopis multiplicata is similar (Billo and Wake, 1987). Little is known about tentacle development in other species; however, the Sarasin's reports (1890) and Dünker et al. (2000) suggest that in at least two species of the oviparous Ichthyophis, tentacle development may occur later in the larval period, and that perforation does not occur until metamorphosis. More work must be done to determine whether these trajectories are mediated by developmental/reproductive modes.

Unquestionably, the data for Nimbaphrynoides, the ovoviviparous frogs, S. salamandra and S. atra, and the viviparous caecilians for which some ontogenetic information is available [e.g., Dermophis mexicanus (Wake and Hanken, 1982; Billo and Wake, 1987; Wake, 1994) and Typhlonectes compressicauda (see earlier)] suggest that metamorphosis is most prolonged and disjunct in the live-bearing taxa in all three orders of amphibians. Especially in the viviparous taxa in which the hatchlings/fetuses ingest secreted maternal nutrient material, the ontogenetic trajectories are modified such that jaws and their musculature and the fetal dentition, as well as the digestive tract, develop early, so that the fetus can feed during a lengthy span of the gestation period (some months in Nimbaphrynoides, S. atra, and the relevant caecilians). The same ontogenetic modification occurs in the subspecies of S. salamandra that are cannibalistic on sibling eggs and younger embryos. The mediation of metamorphosis of fetuses of live-bearers should be a rich avenue of study.

CONCLUSIONS

Based on this delineation of fetal adaptations for viviparity, it seems obvious that extensive homoplasy exists in the evolution of live-bearing. The acquisition and use of a “fetal” dentition in Salamandra and many caecilians (also associated with direct development and with skin feeding in the latter), the prolonged metamorphosis, and a number of other features in live-bearers in two or often all three of the orders of amphibians attests to this. Consequently one must assess the homoplasy of such systems that have similarities and differences at multiple levels of examination by adopting a hierarchical examination of phenomena and the mechanisms by which they arise and are maintained (Wake, 1991; Wake et al., 2011). At the same time, this review points out that many of the data for live-bearing members of the three orders are noncomparable across clades, because we have information for some components in some species, but not for others, both within and across clades. It also identifies a number of areas in which new research will provide information crucial to understanding the evolution and maintenance of viviparity in diverse lineages.

Some useful research areas regarding fetal adaptations include those already identified by Wake (in press) in terms of maternal adaptations for viviparity. It is essential, however, that research also include an integrative assessment of maternal and fetal interactions and their evolution. Such aspects include:

  1. The influences of maternal hormones, especially estrogen, progesterone, prolactin, and thyroxin, and how they interact with embryonic-fetal hormones have received limited study in viviparous taxa. Relatively little is known about hormones and their effects in gestation, development, metamorphosis, and birth; several hormones and their interactions have been virtually ignored to date. Study at several levels of the hierarchy of biology should be undertaken, and direct comparisons made of members of different clades.
  2. Embryos/fetuses develop clade-specific means of facilitating gaseous exchange [across modified gills, expanded tails, and body surfaces and the mucosa of the skin or oviduct, given their increased vascularity, and the nature of any nutrient uptake by gills and/or ectotrophoblast (see summaries by Exbrayat, 2000, 2006b–d)]. Similarly, the “new” structures for obtaining nutrients (e.g., fetal dentitions, tufts of skin around the mouth) that have evolved in intraoviductal embryos and fetuses, as well as the absence of certain “larval” features, deserve study. Regulation of the prolonged metamorphosis and the loss of fetal modifications during the gestation period and near birth, presumably influenced by maternal hormones, remain to be examined.
  3. Young maintained intraoviductally are born either as advanced larvae or metamorphs, as are those of back-brooders; species with obligate oviductal maternal nutrition typically give birth to fully metamorphosed young. The cellular, molecular, endocrinological, and environmental interactions that mediate the timing of birth require study.
  4. Ontogenetic studies for live-bearing taxa are very few; our assumptions about the evolution and maintenance of viviparity are based on limited data for only a few species. Normal tables that allow comparisons within clades to assess the evident heterochrony and homoplasy involved in the evolution of viviparity would be useful.

Maintenance of development in or on the body of a parent is a complex interaction of morphology, development, endocrinology, ecology, and historical contingency that is underappreciated, understudied and deserving of extensive attention. It is especially important that research be stimulated because the existence of many species with derived modes of reproduction is threatened. It is likely that the mechanistic bases of the evolution of live-bearing are consequences of different deployment of common substrates (e.g., morphologies, hormones), as well the development of unique attributes that achieve a common end point, depending on the levels of the hierarchy of organization of the phenomenon under study (Wake, in press). Integrative studies will advance our understanding of pattern, process, divergence, and rampant homoplasy in the evolution and maintenance of live-bearing.

ACKNOWLEDGMENTS

I thank the organizers of the symposium on Fetal Adaptations for Viviparity/Nutrition held at the 10th International Congress of Morphology for inviting me to participate and to produce a written synthesis of my work on amphibian adaptations for viviparity. I appreciate the work of many undergraduates who prepared material. I thank David Buckley for many discussions of the evolution of viviparity and David Wake for ongoing discussions about evolution.

Ancillary