The Ancestry of Modern Amphibia: A Review of the Evidence.

by Theodore H. Eaton.

INTRODUCTION

In trying to determine the ancestral relationships of modern orders of Amphibia it is not possible to select satisfactory structural ancestors among a wealth of fossils, since very few of the known fossils could even be considered possible, and scarcely any are satisfactory, for such a selection. The nearest approach thus far to a solution of the problem in this manner has been made with reference to the Anura. Watson's paper (1940), with certain modifications made necessary by Gregory (1950), provides the paleontological evidence so far available on the origin of frogs. It shows that several features of the skeleton of frogs, such as the enlargement of the interpterygoid s.p.a.ces and orbits, reduction of the more posterior dermal bones of the skull, and downward spread of the neural arches lateral to the notochord, were already apparent in the Pennsylvanian _Amphibamus_ (Fig. 1), with which Gregory synonymized _Miobatrachus_ and _Mazonerpeton_. But between the Pennsylvanian and the Tria.s.sic (the age of the earliest known frog, _Protobatrachus_) there was a great lapse of time, and that which pa.s.sed between any conceivable Paleozoic ancestor of Urodela and the earliest satisfactory representative of this order (in the Cretaceous) was much longer still.

The Apoda, so far as known, have no fossil record.

Nevertheless it should be possible, first, to survey those characters of modern Amphibia that might afford some comparison with the early fossils, and second, to discover among the known Paleozoic kinds those which are sufficiently unspecialized to permit derivation of the modern patterns. Further circ.u.mstantial evidence may be obtained by examining some features of Recent Amphibia which could not readily be compared with anything in the fossils; such are the embryonic development of the soft structures, including cartilaginous stages of the skeleton, the development and various specializations of the ear mechanism, adaptive characters a.s.sociated with aquatic and terrestrial life, and so on.

COMPARISON OF MODERN ORDERS WITH THE LABYRINTHODONTS AND LEPOSPONDYLS

[Ill.u.s.tration: Fig. 1. _Saurerpeton_ ( 1/2, after Romer, 1930, fig. 6); _Amphibamus_, the palatal view 2-1/4, from Watson, 1940, fig. 4 (as _Miobatrachus_), the dorsal view 2-1/2, from Gregory's revised figure of _Amphibamus_ (1950, Fig. 1); _Protobatrachus_, 1, from Watson, 1940, fig. 18, 19.]

In both Anura and Urodela the skull is short, broad, relatively flat, with reduced pterygoids that diverge laterally from the parasphenoids leaving large interpterygoid vacuities, and with large orbits. (These statements do not apply to certain larval or perennibranchiate forms.) The skull in both orders has lost a number of primitive dermal bones in the posterior part; these are: basioccipital, supraoccipital, postparietal, intertemporal, supratemporal, and tabular. The exoccipitals form the two condyles but there are no foramina for the 11th and 12th nerves, since these are not separate in modern Amphibia.

The opisthotic is missing in all except Proteidae (but see discussion of the ear). Although the skull is normally autostylic, a movable basipterygoid articulation is present among Hyn.o.biid salamanders and in at least the metamorphic stages of primitive frogs, and therefore should be expected in their ancestors. The vertebrae are, of course, complete; see discussion in later section. The quadratojugal, lost in salamanders, is retained in frogs, and conversely the lacrimal, absent in frogs, occurs in a few primitive salamanders. The situation in Apoda is different, but postfrontal and jugal should be noted as bones retained in this order while lost in the others.

Thus, in spite of minor differences, the above list shows that there are numerous and detailed similarities between Anura and Urodela with respect to the features in which they differ from the Paleozoic orders.

Pusey (1943) listed 26 characters which _Ascaphus_ shares with salamanders but not with more advanced frogs; a few of these might be coincidental, but most of them are of some complexity and must be taken to indicate relationship. The main adaptive specializations of Anura, however, including loss of the adult tail, extreme reduction in number of vertebrae, formation of urostyle, elongation of the ilium and lengthening of the hind legs, must have appeared at a later time than the separation of that order from any possible common stem with Urodela, although they are only partially developed in the Tria.s.sic _Protobatrachus_.

Turning to the Paleozoic Amphibia, there are two groups in which some likelihood of a relationship with modern order exists. In the Pennsylvanian Trimerorhachoidea (Labyrinthodontia, order Temnospondyli) some members, such as _Eugyrinus_, _Saurerpeton_, and notably _Amphibamus_ (Fig. 1) had short, broad heads, an expansion of palatal and orbital openings, posterior widening of the parasphenoid a.s.sociated with divergence of the pterygoids, a movable basipterygoid articulation, and reduction in size (but not loss) of the more posterior dermal bones of the skull. In recognition of Watson's (1940) evidence that these animals make quite suitable structural ancestors of frogs, Romer (1945) placed _Amphibamus_ in an order, Eoanura, but Gregory (1950) indicated that it might better be left with the temnospondyls. a.s.sociation of the urodele stem with this group does not seem to have been proposed hitherto.

The other group of Paleozoic Amphibia that has been considered probably ancestral to any modern type is the subcla.s.s Lepospondyli, containing three orders, Aistopoda, Nectridia and Microsauria. In these the vertebrae are complete (holospondylous), the centra presumably formed by cylindrical ossification around the notochord, and there is no evidence as to the contributions from embryonic cartilage units. It is important to note at this point that precisely the same statement can be made regarding the vertebrae of _adults_ of all three Recent orders, yet for all of them, as shown in a later section, we have ample evidence of the part played by cartilage elements in vertebral development. Therefore (a) we cannot say that there were no such elements in embryonic stages of lepospondyls, and (b) it would take more than the evidence from adult vertebrae to relate a particular modern order (for example, Urodela) to the Lepospondyli. Vague similarities to Urodela have been noted by many authors in the Nectridia, Aistopoda and Microsauria, but these are not detailed and refer mainly to the vertebrae. The skulls do not show, either dorsally or in the palate, any striking resemblance to those of generalized salamanders, and certainly most known lepospondyls are too specialized to serve as the source of Urodela. It is true that the elongate bodies, small limbs, and apparent aquatic habitus of some lepospondyls accord well with our usual picture of a salamander, but such a form and way of life have appeared in many early Amphibia, including the labyrinthodonts. The family Lysorophidae (Fig. 2), usually placed among microsaurs, is sufficiently close in skull structure to the Apoda to be a possible ancestor of these, but it probably has nothing to do with Urodela, by reason of the numerous morphological specializations that were a.s.sociated with its snakelike habitus.

[Ill.u.s.tration: Fig. 2. _Lysorophus tricarinatus_, lateral and posterior views 2-1/2, modified after Sollas, 1920, Figs. 8 and 12, respectively; palatal view after Broom, 1918, 1-1/2. For explanation of abbreviations see Fig. 3.]

McDowell's (1958) suggestion that it would be profitable to look among the Seymouriamorpha for the ancestors of frogs seems to be based upon a few details of apparent resemblance rather than a comprehensive view of the major characters of the animals. In most points which he mentions (limb girdles, form of ear, pterygoid articulation) the present writer does not see a closer similarity of frogs to Seymouriamorpha than to Temnospondyli.

Still other opinions have been expressed. Herre (1935), for instance, concludes "on anatomical, biological and paleontological grounds" that the orders of Urodela, Anura, Apoda and Stegocephali were all independently evolved from fish, but beyond citing the opinions of a number of other authors he does not present tangible evidence for this extreme polyphyletic interpretation.

More notable are the views of several Scandinavian workers (Save-Soderbergh, 1934; Jarvik, 1942; Holmgren, 1933, 1939, 1949a, b), of whom Jarvik, in a thorough a.n.a.lysis of the ethmoid region, would derive the Urodela from Porolepid Crossopterygii, and all other tetrapods from the Rhipidistia; Save-Soderbergh and Holmgren, the latter using the structure of carpus and tarsus, see a relationship of Urodela to Dipnoi, but accept the derivation of labyrinthodonts and other tetrapods from Rhipidistia. All of this work is most detailed and laborious, and has produced a great quant.i.ty of data useful to morphologists, but the diphyletic theory is not widely adopted; the evidence adduced for it seems to consist largely of minutiae which, taken by themselves, are inconclusive, or lend themselves to other interpretation. For instance Holmgren's numerous figures of embryonic limbs of salamanders show patterns of cartilage elements that he would trace to the Dipnoan type of fin, yet it is difficult to see that the weight of evidence requires this, when the pattern does not differ in any fundamental manner from those seen in other embryonic tetrapods, and the differences that do appear may well be taken to have ontogenetic rather than phylogenetic meaning. Further, the Dipnoan specialization of dental plates and autostylic jaw suspension, already accomplished early in the Devonian, would seem to exclude Dipnoi from possible ancestry of the Urodela, an order unknown prior to the Mesozoic, in which the teeth are essentially similar to those of late Paleozoic Amphibia, and the jaw suspension is not yet in all members autostylic.

THE EAR

[Ill.u.s.tration: Fig. 3. Occipital region of skulls of _Megalocephalus brevicornis_ ( 3/10, after Watson, 1926, as _Orthosaurus_), _Dvinosaurus_ ( 1/4, modified after Bystrow, 1938; the lower figure after Sushkin, 1936), and _Necturus maculosus_ ( 3, original, from K.

U., No. 3471).

Abbreviations Used in Figures

b'd.c.--basidorsal cartilage (neural arch) b'oc.--basioccipital ce._{1-4}--centrale_{1-4} ch.--ceratohyal clav.--clavicle clei.--cleithrum cor.--coracoid d.c._{1-4}--distal carpal_{1-4} diap.--diapophysis exoc.--exoccipital ep.--episternum hyost.--hyostapes i.--intermedium Mk.--Meckel's cartilage n.--notochord om.--omosternum op.--operculum opis.--opisthotic par.--parietal par. proc.--paroccipital process peri. cent.--perichordal centrum p'p.--postparietal prep.--prepollex pro.--prootic p'sp.--parasphenoid pt.--pterygoid p.t.f.--post-temporal fossa postzyg.--postzygapophysis qj.--quadratojugal qu.--quadrate ra.--radiale r.hy.--hyomandibular ramus of VII rib-b.--rib-bearer r.md.--mandibular ramus of VII sc.--scapula sc'cor.--scapulocoracoid s'd.--supradorsal cartilage s'd.(postzyg.)--supradorsal (postzygapophysis) soc.--supraoccipital sp.c.--spinal cord sq.--squamosal s'sc.--suprascapula s't.--supratemporal sta.--stapes ster.--sternum tab.--tabular uln.--ulnare v.a.--vertebral artery xiph.--xiphisternum I,IV--digits I and IV V, VII, X, XII--foramina for cranial nerves of these numbers (in Fig. 4, VII is the facial nerve) ]

In temnospondylous Amphibia the tympanum generally occupied an otic notch, at a high level on the skull, bordered dorsomedially by the tabular and ventrolaterally by the squamosal. In this position the tympanum could receive airborne sounds whether the animal were entirely on land or lying nearly submerged with only the upper part of its head exposed. Among those Anura in which the ear is not reduced the same is true, except that the tabular is lost. In Temnospondyli (Fig. 3) the posterior wall of the otic capsule was usually formed by the opisthotic, which extended up and outward as a b.u.t.tress from the exoccipital to the tabular, and sometimes showed a paroccipital process for the insertion, presumably, of a slip or tendon of the anterior axial musculature. The stapes, in addition to its foot in the fenestra ovalis and its tympanic or extrastapedial process to the tympanum, bore a dorsal process (or ligament) to the tabular, an "internal" process (or ligament) to the quadrate or an adjacent part of the squamosal, and a ligament to the ceratohyal. Some of these attachments might be reduced or absent in special cases, but they seem to have been the ones originally present both phylogenetically and embryonically in Amphibia.

Among typical frogs (Fig. 4) the base, or otostapes, is present and bony, the extrastapedial process (extracolumella, or hyostapes) is usually cartilaginous, the dorsal process (processus paroticus) is of cartilage or ligament, but the other two attachments are absent in the adult. The exoccipital extends laterally, occupying the posterior face of the otic capsule. Between it and the otostapes is a small disc, usually ossified, the operculum, which normally fits loosely in a portion of the fenestral membrane, and is developed from the otic capsule. The opercularis muscle extends from this disc to the suprascapula, in many but by no means all families of Anura.

[Ill.u.s.tration: Fig. 4. Diagram of middle ear structures in _Rana_ (upper figure, after Stadtmuller, 1936, and lower left after DeBeer, 1937), and _Ambystoma_ (lower right, after DeBeer, 1937); all 4. For explanation of abbreviations see Fig. 3.]

Among Urodela (Fig. 4) the middle ear cavity and tympanum are lacking, and the stapes (columella) consists of no more than its footplate and the stylus, which is attached to the border of the squamosal, thus corresponding to the "internal" process. In families in which individuals metamorphose and become terrestrial (Hyn.o.biidae, Ambystomidae, Salamandridae, Plethodontidae), an operculum and opercularis muscle appear in the adult, just as in frogs, except that in Plethodontidae, the most progressive family, the operculum fuses with the footplate of the stapes. Among neotenous or perennibranchiate urodeles there is no separate operculum or opercularis. The evidence given by Reed (1915) for fusion of the operculum with the columella in _Necturus_ appears inconclusive, in spite of the great care with which his observations were made. On the other hand, _Necturus_ and _Proteus_ alone among living salamanders have a distinct opisthotic on the posterior wall of the otic capsule (Fig. 3), as do the Cretaceous _Hylaeobatrachus_ and the Eocene _Palaeoproteus_. Probably these Proteidae should be regarded as primitive in this respect, although many other features may be attributed to neoteny.

There is a contrast between Anura and most Urodela in the relative positions of the stapes and facial nerve, as shown in DeBeer's (1937) diagrams. In the latter (_Ambystoma_) the nerve is beneath, and in the former (_Rana_) above, the stapes. Judging by figures of _Neoceratodus_, _Hypogeophis_, and several types of reptiles and mammals, the Urodela are exceptional. _Necturus_, however, has the nerve pa.s.sing above its stapes, and this may be primitive in the same sense as the persistent opisthotic. There can be, of course, no question of the nerve having worked its way through or over the obstructing stapes in order to come below it in salamanders; rather, the peripheral growth of neuron fibers in the embryo must simply pursue a slightly different course among the partially differentiated mesenchyme in the two contrasting patterns.

Although DeBeer (1937) shows in his figure of _Hypogeophis_ (one of the Apoda) an operculum, this is apparently a mistake. The stapes has a large footplate, and its stylus articulates with the quadrate, but no true operculum or opercularis has been described in the Apoda. The facial nerve pa.s.ses above the stapes. It does not seem necessary to regard the conditions in this order as related directly to those of either salamanders or frogs, but a reduction of the stapes comparable to that in salamanders has occurred.

The presence in both frogs and terrestrial salamanders of a special mechanism involving the opercularis muscle and an operculum cut out in identical fashion from the wall of the otic capsule behind the stapes seems to require some other explanation than that of a chance convergence or parallelism. Although the stapes and otic region are readily visible in a number of labyrinthodonts and lepospondyls, no indication of an operculum seems to be reported among them. But in the Tria.s.sic _Protobatrachus_ (Fig. 1), which is unmistakably a frog in its skull, pelvis and some other features, Piveteau (1937) has shown, immediately behind the foot of the stapes, a small bony tubercle, which he and Watson (1940) designated opisthotic. Very clearly it served for insertion of a muscle, and it is equally clear that the bone is a reduced opisthotic, carrying the paroccipital process already mentioned as characteristic of it in some temnospondyls. Since the remainder of the posterior wall of the otic capsule consists of cartilage, meeting the exoccipital, it may be that the opisthotic becomes the operculum in frogs. _Protobatrachus_ was too far specialized in the Anuran direction, although it still had a tail, and the forelegs and hind legs were nearly the same size, to be considered a possible ancestor of the Urodeles. But at one stage in the general reduction of the skull in the ancestry of both groups, a condition similar to that in _Protobatrachus_ may have characterized the otic region, long before the Tria.s.sic.

In the argument thus far we have considered terrestrial, adult amphibians, since it is only in these that either the normal middle ear and tympanum, or the opercular apparatus, is present. But among the urodeles several neotenic types occur (this term applies also to the perennibranchs). For most of these there is nothing about the otic region that would be inconsistent with derivation, by neoteny, from known families in which adults are terrestrial; for example, _Cryptobranchus_ could have had a Hyn.o.biid-like ancestor. But this, as mentioned above, does not hold for the Proteidae, which possess an opisthotic of relatively large size, distinctly separate from the exoccipital and prootic. Either this bone is a neomorph, which seems improbable, or there has not been in the ancestry of this particular family an episode of reduction comparable to that seen in the terrestrial families, where there is an operculum instead of a normal opisthotic. Therefore the Proteidae probably are not derived from the general stem of other salamanders, but diverged sufficiently long ago that the bones of the otic region were reduced on a different pattern.

They need not be removed from the order, but, in this respect, recognized as more primitive than any other existing Urodela or Anura. A recent paper by Hecht (1957) discusses many features of _Necturus_ and _Proteus_, and shows that they are remote from each other; his evidence does not seem to prove, however, that they were of independent origin or that they need be placed in separate families.

VERTEBRAE AND RIBS

Development of the vertebrae and ribs of Recent Amphibia has been studied by Gamble (1922), Naef (1929), Mookerjee (1930 a, b), Gray (1930) and Emelianov (1936), among others. MacBride (1932) and Remane (1938) provide good summaries. In this section reference will be made to the embryonic vertebral cartilages by the names used for them in these studies, although the concept of "arcualia" is currently considered of little value in comparative anatomy.

[Ill.u.s.tration: Fig. 5. Development of Anuran vertebrae. Upper left, late tadpole of _Xenopus laevis_; lower left, same just after metamorphosis; upper right, diagram of general components of primitive Anuran vertebra.

(After MacBride, 1932, Figs. 35, 38, 47D, respectively.) Lower right, section through anterior portion of urostyle, immediately posterior to sacral vertebra, in transforming _Ascaphus truei_ (original, from specimen collected on Olympic Peninsula, Washington). All 20 approx.

For explanation of abbreviations see Fig. 3.]

The centrum in Anura (Fig. 5) is formed in the perichordal sheath (_Rana_, _Bufo_) or only in the dorsal portion thereof (_Bombinator_, _Xenopus_). The neural arch develops from the basidorsal cartilages that rest upon, and at first are entirely distinct from, the perichordal sheath. Ribs, present as separate cartilages a.s.sociated with the 2nd, 3rd and 4th vertebrae in the larvae of _Xenopus_ and _Bombinator_, fuse with lateral processes (diapophyses) of the neural arches at metamorphosis, but in _Leiopelma_ and _Ascaphus_ the ribs remain freely articulated in the adult. Basiventral arcualia have been supposed to be represented by the hypochord, a median rod of cartilage beneath the shrinking notochord in the postsacral region, which at metamorphosis ossifies to produce the bulk of the urostyle. Fig. 5, lower right, a transverse section taken immediately posterior to the sacral ribs in a transforming specimen of _Ascaphus_, shows that the "hypochord" is a ma.s.s of cartilage formed in the perichordal sheath itself, and very obviously is derived from the ventral part of postsacral perichordal centra; there are, then, no basiventral arcualia, and the discrete hypochord shown in MacBride's diagram (Fig. 5, upper right) of a frog vertebra does not actually occur below the centrum, but only below the notochord in the postsacral region.

[Ill.u.s.tration: Fig. 6. Development of Urodele vertebrae. Upper figures, _Triton_: at left, larva at 20 mm., at right, diagram of components of vertebra (from MacBride, 1932, figs. 17, 47C). Middle figures, _Molge vulgaris_ larva: left, at 18 mm.; middle, at 20-22 mm.; right, at 25 mm.

(from Emelianov, 1936, figs. 33, 36, 38 respectively). Lower figures, _Necturus maculosus_ larva: left, at 21 mm.; right, at 20 mm. (from MacBride, 1932, figs. 41.5, 41.3 respectively, after Gamble, 1922). All 20 approx. For explanation of abbreviations see Fig. 3.]

In Urodela (Fig. 6) the pattern of vertebral and rib development is more complex, and there has been much controversy over its interpretation.

Neural arches and perichordal centra form in the same manner as in frogs, but with the addition in certain cases (_Triton_) of a median supradorsal cartilage, which gives rise to the zygapophyses of each neural arch. Difficulty comes, however, in understanding the relationship of the ribs to the vertebrae. Each rib, usually two-headed, articulates with a "transverse process" that in its early development seems to be separate from both the vertebra and the rib, and is therefore known, noncommittally, as "rib-bearer." This lies laterally from the centrum, neural arch, and vertebral artery; upon fusing with the vertebra it therefore encloses the artery in a foramen separate from the one between the capitulum and tuberculum of the rib (the usual location of the vertebral artery). At least four different interpretations of these structures have been suggested:

(1) Naef (1929) considered the rib-bearer a derivative of the basiventral, which, by spreading laterally and dorsally to meet the neural arch, enclosed the vertebral artery. He then supposed that by reduction of the rib-bearer in other tetrapods (frogs and amniotes) the vertebrarterial foramen and costal foramen were brought together in a single foramen transversarium. The implication is that the Urodele condition is primitive, but it cannot now be supposed that Urodela are ancestral to any other group, and the rib-bearer is most probably a specialization limited to salamanders. This does not, of course, invalidate the first part of his interpretation.

(2) Remane (1938), noting that rib insertions of early Amphibia are essentially as in Amniota, argued that the rib-bearer is not from the basiventral but is a neomorph which originates directly from the neural arch and grows ventrally. This he inferred mainly from Gamble's (1922) observation on _Necturus_, but his a.s.sumption that _Necturus_ is more primitive than other salamanders (such as the Salamandridae), where the pattern differs from this, is not necessarily correct. Rather, the perennibranchs are distinguished mainly by their neotenous features, and their development is likely to show simplifications which are not necessarily primitive. The suggestion of a "neomorph" ought not to be made except as a last resort, for it is simply an acknowledgment that the author does not recognize h.o.m.ology with any structure already known; sometimes further information will make such recognition possible.

(3) Gray (1930), using _Molge taeniatus_, concluded that the normal capitulum of the rib was lost, but that the tuberculum bifurcated to make the two heads seen in Urodela, thus accounting for the failure of the costal foramen to coincide with that of the vertebral artery. This answer, too, seems to entail an unprovable a.s.sumption which should not be made without explicit evidence.

(4) Finally, Emelianov (1936) regarded the rib-bearer as a rudimentary _ventral_ rib, on account of its relationship to the vertebral artery, and considered the actual rib to be a neomorph in the _dorsal_ position characteristic of tetrapod ribs in general. This argument would fit the ontogenetic picture satisfactorily, provided that (_a_) there were some evidence of ventral, rather than dorsal, ribs in early Amphibia, and (_b_) we accept the invention of another neomorph in modern Amphibia as an unavoidable necessity. Emelianov's conclusion (p. 258) should be quoted here (translation): "The ribs of Urodela are shown to be upper ribs, yet we find besides these in Urodela rudimentary lower ribs fused with the vertebral column. The ribs of Apoda are lower ribs. In Anura ribs fail to develop fully, but as rare exceptions rudiments of upper ribs appear."

Of these various interpretations, that of Naef seems to involve the minimum of novelty, namely, that the rib-bearer is the basiventral, expanded and external to the vertebral artery. It is not necessary to take this modification as the ancestral condition in tetrapods, of course. The basiventral (=intercentrum) would merely have expanded sufficiently to provide a diapophysis for the tuberculum as well as the (primitive) facet for the capitulum. No neomorph appears under this hypothesis, which has the distinct advantage of simplicity.

Figures of early stages in vertebral development by the authors mentioned show that the basidorsals chondrify first, as neural arches, while a separate ma.s.s of mesenchyme lies externally and ventrally from these. This mesenchyme may chondrify either in one piece (on each side) or in two; in _Molge_ the part adjacent to the centrum is ossified in the 20-mm. larva, and subsequently unites with the more dorsal and lateral cartilaginous part, while the rib, appearing farther out, grows inward to meet this composite "rib-bearer." In _Necturus_ the mesenchyme below the neural arch differentiates into a cartilage below the vertebral artery (position proper to a basiventral), a bridge between this and the neural arch, and a rib, the latter two chondrifying later than the "basiventral" proper. In the "axolotl" (presumably _Ambystoma tigrinum_) the rib-bearer grows downward from its first center of chondrification at the side of the neural arch (Emelianov, 1936).