Comparative pelvic development of the axolotl (Ambystoma mexicanum) and the Australian lungfish (Neoceratodus forsteri): conservation and innovation across the fish-tetrapod transition

Background The fish-tetrapod transition was one of the major events in vertebrate evolution and was enabled by many morphological changes. Although the transformation of paired fish fins into tetrapod limbs has been a major topic of study in recent years, both from paleontological and comparative developmental perspectives, the interest has focused almost exclusively on the distal part of the appendage and in particular the origin of digits. Relatively little attention has been paid to the transformation of the pelvic girdle from a small unipartite structure to a large tripartite weight-bearing structure, allowing tetrapods to rely mostly on their hindlimbs for locomotion. In order to understand how the ischium and the ilium evolved and how the acetabulum was reoriented during this transition, growth series of the Australian lungfish Neoceratodus forsteri and the Mexican axolotl Ambystoma mexicanum were cleared and stained for cartilage and bone and immunostained for skeletal muscles. In order to understand the myological developmental data, hypotheses about the homologies of pelvic muscles in adults of Latimeria, Neoceratodus and Necturus were formulated based on descriptions from the literature of the coelacanth (Latimeria), the Australian Lungfish (Neoceratodus) and a salamander (Necturus). Results In the axolotl and the lungfish, the chondrification of the pelvic girdle starts at the acetabula and progresses anteriorly in the lungfish and anteriorly and posteriorly in the salamander. The ilium develops by extending dorsally to meet and connect to the sacral rib in the axolotl. Homologous muscles develop in the same order with the hypaxial musculature developing first, followed by the deep, then the superficial pelvic musculature. Conclusions Development of the pelvic endoskeleton and musculature is very similar in Neoceratodus and Ambystoma. If the acetabulum is seen as being a fixed landmark, the evolution of the ischium only required pubic pre-chondrogenic cells to migrate posteriorly. It is hypothesized that the iliac process or ridge present in most tetrapodomorph fish is the precursor to the tetrapod ilium and that its evolution mimicked its development in modern salamanders.


Background
Around 395 million years ago, the first tetrapods (fourlegged vertebrates) appeared, having evolved from lobefinned fish [1,2]. This fish-tetrapod transition was marked by many morphological transformations and ecological adaptations ranging from the evolution of fingers and toes [3][4][5] to new modes of respiration, hearing [6,7] and locomotion [8,9]. One of the major changes in locomotory habit is that of a shift from fish principally using their pectoral fins and lateral undulation to swim to tetrapods relying much more heavily on their hindlegs to swim and walk [10]. This shift from 'front-wheel drive' to 'back wheel drive' locomotion was enabled by the evolution of a large, weight-bearing pelvic girdle in tetrapods. In lobefinned fishes, the pelvic girdle is composed of a crescentric pubis often connected through cartilage at the midline but lacking an ilium to connect it and, consequently, the whole fin, to the vertebral column [11]. In tetrapods, not only is an ilium present and fused to the vertebral column through a sacral rib, but an ischium is also present posterior to the pubis. The pubis and ischium from both halves of the girdle are fused along their midlines and, hence, the girdle is weight-bearing [2]. Tetrapod legs are also oriented laterally compared to the posterior orientation of fish pelvic fins. This reflects the orientation of the acetabulum, which is located on the lateral face of the pelvis in tetrapods but at the posterior end of the pelvis in fish. The morphology of Paleozoic lobe-finned fish pelves is known from descriptions of Eusthenopteron [12] and other fish members of the tetrapod stem group [13,14], as well as from stem dipnoans [15] and a porolepiform [11], and that of the earliest tetrapods by descriptions of Acanthostega [16] and Ichthyostega [8,17]. This provides a good picture of the general morphology on either side of the transition ( Figure 1) but little information as to how the transformation occurred (Figure 1). In order to elucidate this, the pelvic girdle of the transitional fish Panderichthys was studied [18] but it is very fish-like and unfortunately does little to answer the following questions: Is the iliac process present in Eusthenopteron a precursor of the ilium of tetrapods? How did the ilium become connected to the sacral rib? How did the ischium evolve? And how did the acetabulum move during the transition?
In the absence of more informative fossils, we can use the Extant Phylogenetic Bracketing method [19] as a basis for framing hypotheses about how the transformation occurred. By comparing the development of the pelvic girdle in modern representatives of groups on either side of the Figure 1 Phylogeny spanning the fish-tetrapod transition, showing stepwise transformation of the pelvic morphology. In lungfishes (Dipnoi) there is no iliac ramus, but a low ridge that can be homologized with the iliac process of other tetrapodomorph fishes. In the extant Neoceratodus the homologues of tetrapod iliac musculature attach to a low ridge anterodorsal to the acetabulum. In fish members of the tetrapod stem group (exemplified here by the rhizodont Gooloogongia and the osteolepiform Eusthenopteron), overall pelvic morphology is similar to that of lungfishes but an iliac process is present. In the stem tetrapods Acanthostega and Ichthyostega an ischium is present and overall pelvic morphology approaches that of extant salamanders such as Ambystoma. The boundary between ishium and pubis is approximate in stem tetrapods and axolotl as the elements are not separated by sutures. Anterior to the left. Phylogeny based on [4], Griphognathus and Neoceratodus redrawn from [15], Gooloogongia from [46], Eusthenopteron from [12], Acanthostega from [16], Ichthyostega from [17], Ambystoma from Figure 2, Mus musculus from [47]. All reconstructions by CAB. transition, we can determine which aspects are similar and, thus, likely to have been conserved from fish to tetrapod, and which appear to be tetrapod innovations. This approach not only illuminates the evolution of pelvic development, but provides clues for the interpretation of the adult morphologies of transitional fossils. Heterochronies in developmental sequences have proven useful in understanding evolutionary change and are, thus, of particular interest in this context [20,21]. The Australian lungfish (Neoceratodus forsteri) is a morphologically conservative member of the Dipnoi and is the only lobe-finned fish available for developmental studies. Fortunately, its pelvic girdle is very similar to that of Paleozoic dipnoans [15] and a good representation of the general lobe-finned fish pelvic shape (Figure 1). The Mexican axolotl (Ambystoma mexicanum) is a commonly used laboratory animal, which, like other salamanders, has a pelvic morphology very similar to that of early tetrapods. In both species, a cartilaginous anterior process (pubic process in Neoceratodus and ypsiloid cartilage in Ambystoma) is present. This process is absent, or arguably unpreserved, in both Devonian lungfish and in Devonian tetrapods, and so cannot be assumed to be homologous in the two groups. However, the pubis and acetabulum are uncontroversially homologous [22] and will be used here as landmarks for the developmental comparison.
The skeletal components of the pelvic girdle do not exist in isolation, but are intimately linked both developmentally and functionally to the muscles that attach there. In order to understand the transformation of the pelvis at the fish-tetrapod transition, it is thus important to consider not just the skeleton but also the musculature. Accordingly, we present here a comparison of muscular development in Neoceratodus and Ambystoma, as a complement to the skeletal study. Establishing a robust comparative framework for the muscular data is, however, a more complicated matter than for the skeletal data because the musculature of the pelvis and hind limb is far more complex than the skeleton and the mapping of homologies between taxa correspondingly more difficult. Establishing muscle homologies has been an important focus of comparative anatomy in the 19th and early 20th century [23][24][25][26] and its importance is now being put back into focus as a way of understanding and explaining evolutionary change [27,28]. Despite the adult pelvic musculature having been described for two of the living sarcopterygians, the coelacanth Latimeria chalumnae [29] and the Australian lungfish Neoceratodus forsteri [15], as well as for several salamanders [23,24,30] and several studies having been published on hindlimb muscle homologies [25,28], no attempt has been made to establish detailed muscular homologies across the fish-tetrapod transition. We based our mapping of muscle homologies on close representatives of the taxa used in the developmental study. As a representative for the axoltol, we used the mudpuppy Necturus maculosus [31] because of its close phylogenetic position and similarities to Ambystoma mexicanum [32,33]. We present a detailed comparison of the three musculatures, developed from published descriptions, with a set of proposed homologies (Tables 1, 2 , 3, 4, 5). The proposed homologies are based on the points of origin and insertion of the muscles as well as their function, following established principles for such comparisons [25,27,28].

Salamander and lungfish larvae
Albino larvae of the Mexican axolotl (A. mexicanum) were purchased from the Ambystoma genetic stock center at the University of Kentucky, USA. They were fixed in paraformaldehyde overnight and stored in 100% methanol. The youngest larvae of the series used in this article were staged using the extended table of development developed by Nye et al. [34]. Given the absence of an adequate development table, older larvae were staged according to total length (in cm). Larvae of the Australian lungfish (N. forsteri) were raised in captivity from eggs collected in the lungfish spawning ponds at Macquarie University (protocols approved by the Macquarie University Animal Ethics Committee, approval # 2003/001). The embryos were left to hatch and develop for approximately five months, being fed on brine shrimp and bloodworms (for older larvae). The youngest larvae used in this study were staged using the developmental table developed by Kemp [35]. Fish older than stage 55 (latest stage of the table) were staged according to pelvic fin length and were given a stage number corresponding to the same developmental progress as between stages 54 and 55. The larvae were euthanized with Tricaine (MS-222) and fixed overnight in 4% paraformaldehyde with a pH of 7.4.

Alcian blue and Alizarin red staining No acid, Kimmel protocol
Lungfish were cleared and stained using a protocol without acid developed by the Kimmel Laboratory (University of Oregon, USA) for zebrafish and modified by Catherine Anne Boisvert. The larvae were eviscerated and then washed in Tris/MgCl 2 before being transferred to Alcian stain solution (0.02% alcian in 71% EtOH and 25 mM MgCl 2 in Tris pH 7.5 aqueous solution) for a period ranging from three to seven days. The specimens were then rehydrated through a series of ethanol in 100 mM Tris pH 7.5 and 25 mM MgCl 2 for 30 minutes each, after which they were bleached in 3% H 2 O 2 and 0.5% KOH for 20 to 22 hours with a change of solution.       Muscles were digested away at room temperature in a solution of 1% pancreatin in 35% saturated sodium borate for 17 to 51 hours. The specimens were washed for one hour in 25% glycerol and 0.1% KOH and then stained for bone in a solution of 0.02% Alizarin stain in 10% glycerol and 0.5% KOH for two to four days. Excess stain was removed by placing them in a solution of 50% glycerol and 0.5% KOH for at least a day after which they were stored in 100% glycerol with a few crystals of thymol to avoid fungal growth. All steps were carried out on a gyrating platform at a low setting.

Zebrafish protocol for optical tomography
Axolotls of stages 54, 55 and 1.5 cm were cleared and stained following a protocol developed by Silke Berger from the Currie Laboratory (Australian Regenerative Medicine Institute, Monash University) for Optical Tomography on zebrafish and modified by Catherine Anne Boisvert. The larvae were eviscerated and dehydrated before being bleached in a mixture of formamide, SSC (sodium chloride citrate) and H 2 O 2 under a light source for 20 to 25 minutes. They were then washed in PBS, dehydrated to 75% ethanol and stained in Alcian stain solution (EtOH, glacial acetic acid and 0.01 mg/ml Alcian blue) for 22 hours. The specimens were washed in 80% EtOH/Tris/MgCl 2 and rehydrated and washed in dH 2 O. They were then stained for 11 to 24 hours in Alizarin stain (0.1 mg/ml in 0.5% KOH aqueous solution) and washed in PBS. They were then transferred to an increasing series of glycerol in PBS and stored in 100% glycerol with a few crystals of thymol. All steps were carried out on a gyrating platform at a low setting.

Taylor and VanDyke protocol
Ambystoma mexicanum specimens of sizes 2.0 cm, 2.5 cm, 3.0 cm, 3.5 cm and 4.0 cm were cleared and stained according to a protocol developed by Taylor and VanDyke [36] and modified by Catherine Anne Boisvert. The sa-lamanders were eviscerated and washed in 80% EtOH/ Tris/MgCl 2 before being stained in Alcian blue stain (0.3 mg/ml Alcian stain in 80% EtOH and glacial acetic acid) for three days. They were neutralized in a saturated solution of sodium borate and bleached for one hour and forty minutes in a solution of 0.5% KOH and H 2 O 2 . Muscles were removed in a solution of 2.25 mg/ml trypsinin saturated sodium borate. They were then stained in Alizarin red solution (0.1 mg/ml in 0.5% KOH aqueous solution) for two to three days, rinsed in dH 2 O and transferred to an increasing series of glycerol in water. They were stored in 100% glycerol with a few crystals of thymol.
All steps were carried out on a gyrating platform at a low setting.

Klymkowsky and Hanken protocol
Neoceratodus larvae from stages 50 and 51 were stained as whole-mounts according to a protocol modified from Klymkowsky and Hanken [37]. The larvae were refixed overnight in Dent's fixative and bleached for 29 hours in Dent's bleach. The specimens were then rehydrated and washed in 'saline cocktail' (PBS, 0.4% Triton X-100) before being blocked in 'serum cocktail' (PBS, 0.4%Triton X-100, 2% bovine serum albumin (BSA), 5% dimethylsulfoxide (DMSO)) for one hour. The specimens were then incubated with the primary antibody against skeletal muscle (Hybridoma gene bank 12/101, 3.7 mg/ml IgG 1 ) diluted 1:50 in 'serum cocktail' for five days at room temperature. They were then washed and reblocked in 'serum cocktail' overnight. The larvae were then incubated in the secondary antibody (488 goat antimouse Alexa antibody by Molecular Probes/Invitrogen; 2 mg/ml) diluted 1:150 in 'serum cocktail' for two days at room temperature in the dark. They were then washed in 'serum cocktail' and 'saline cocktail' before being dehydrated to 100% methanol and transferred to an increasing series of BABB (benzyl alcohol/benzyl

Development of the pelvic girdle in the Australian lungfish
Kemp's staging table for Neoceratodus [35] stops at stage 55 and no staging table currently exists for older stages. Older larvae have been given a stage starting at number 56 according to general growth. These stage numbers are indicated between single quotation marks to differentiate them from the published staging table [35]. At stage 49, the pelvic girdle is already present as two narrow bands of cartilage curving mesially but not meeting at the midline (Figure 2A). In the following stages, the pelvic cartilage gradually thickens ( Figure 2B). During stages 52 and 53, the pelvis gradually elongates anteriorly and the two halves start fusing at the midline ( Figure 2C). During stages 54 and 55, the halves of the pelvic girdle progressively fuse anteriorly at the midline but not between the acetabula. A gap is visible between the acetabula until it closes at stage '56' and thickens at stage '59' (Figure 2D). At stage '60' , the pubic process appears as a small triangular projection much thinner than the rest of the pelvis. At stage '61' , the pubic process is much longer and forms the extremity of a large triangle formed by the pelvis itself ( Figure 2E and F). At this stage, cartilage anterior to the acetabula is thick and a crescentric arch of strongly chondrified tissue is visible, bridging the left and right sides of the pelvic girdle. This morphology is very similar to that of a larger juvenile as described by Young et al. [15].

Development of the pelvic girdle in the axolotl
Stage 54 [34] marks the beginning of pelvic development in the axolotl with the appearance of the pubis as a condensation anterior to the future acetabulum (indicated as the area between the pubic condensation and the femur) ( Figure 2G). An anlage of the femur is also present at this stage posterior to the pubis condensation. (Figure 2G lateral ). In the least developed 1.5 cm larva, the ischium has developed posterior to the pubic cartilage ( Figure 2H ventral ). The pubis is still only a small condensation of cells at the acetabulum but the ilium is now present (Figure 2H lateral ). In another 1.5 cm larva, the pubis is larger, having extended anteriorly from the acetabulum and towards the midline as a thick condensation. The ischium is also longer, reaching farther posteriorly (Figure 2I ventral ). The ilium is longer, projecting postero-dorsally towards the vertebral column ( Figure 2I lateral ) but not reaching it. In 2.0 cm larvae, the pubis is much larger and almost complete anteriorly ( Figure 2J ventral ). It is pierced by the obturator foramen on either side but the halves do not meet in the middle. The ischium is also larger but not complete; each half of the pelvic girdle is still triangular in shape, the anterior extremity being widest. At this stage, the ilium is much longer but far from reaching the vertebral column ( Figure 2J lateral ). Cartilage of the pubis, ischium and ilium continue to thicken and grow in 3.0 cm larvae but there is little change in overall shape. Both sides of the pelvic girdle are still unfused and roughly triangular in shape. The major difference is the development of the sacral rib and the elongation of the ilium towards it ( Figure 2K). In the most developed larvae of the series (a 3.5 cm larva), the ypsiloid cartilage is complete and both sides of the pubis and ischium are almost fused ( Figure 1L ventral ). The ilium almost connects with the sacral rib and has started ossifying (white zone close to its base) but is not calcified yet, preventing the Alizarin red from binding ( Figure 1L lateral ). A schematic comparative representation of pelvic development in Neoceratodus and Ambystoma, aligned using the acetabulum as a fixed landmark, is shown in Figure 3. Table 1 displays equivalences of pelvic muscles in Latimeria, Neoceratodus and Necturus. Comparison of the musculature of Neoceratodus and Necturus is relatively straightforward and most muscles present in Neoceratodus can be associated with one or several present in Necturus. This is not the case for Latimeria, where text in italics indicates an incomplete equivalence. Often, the point of insertion is much more distal than in Neoceratodus or Necturus and it is often on lepidotrichia rather than endoskeletal fin elements. In some instances this may be the result of an incomplete dissection and it is possible that some of the muscles written in italics actually are equivalent to those of Neoceratodus and Necturus noted in plain text on the same row. However, these identifications should be regarded as tentative until a re-examination of Latimeria is made. For all tables, the names of the muscles were taken from their original description and have been directly translated from French for Latimeria.

Adductor muscles
All adductor muscles in Latimeria insert very distally on the fin, making each one an incomplete equivalent to those of Neoceratodus and Necturus. However, equivalences between Neoceratodus and Necturus are easily established. Only two muscles do not have equivalents. The mesial adductor of Neoceratodus was described by Young et al. [15] as linking muscles of both fins from either side. This is unique for Neoceratodus. In Necturus, the puboischiofemoralis externus, a superficial adductor originating from the pubis and ischium and inserting at the base of the femur, cannot be equated with any muscle present in Neoceratodus.

Abductor muscles
Abductor muscles equivalences are displayed in Table 2. Again, all muscles described in Latimeria insert very distally on the fin, either on lepidotrichia or on muscles inserting onto lepidotrichia. An incomplete equivalence can be established between the pelvic abductor of Latimeria and the iliofibularis of Necturus with no equivalent in Neoceratodus. The iliofemoralis of Necturus originating at the base of the ilium and inserting on the caudal edge of the femur does not have an equivalent in Latimeria or Neoceratodus. Table 3 presents muscles that either have an adductor or abductor function. The radial flexors of Neoceratodus can be equated to the shank flexors and extensors of Necturus.

Supinators and pronators
Many supinators and pronators have been described for Latimeria and most of them can be equated to lepidotrichial flexors and radial-axials present in Neoceratodus. Lepidotrichial flexors do not have equivalents in Necturus since lepidotrichia have been lost during the finlimb transition and are, therefore, absent in all tetrapods. The radial-axials of Neoceratodus and their equivalents in Latimeria cannot be directly equated to muscles in Necturus, lacking pre-and post-axial radials in a 'fish configuration' , but given that fish distal radials are precursors to digits [3,38], it is possible that those muscles were the precursors of tetrapod digit musculature. The only supinator muscle in Latimeria that has no equivalent in Neoceratodus is the supinator of the second layer, second group which originates from the pelvic girdle and inserts on preaxial radials and preaxial lepidotrichia. The popliteus of Necturus, wrapping around the knee, does not have direct equivalents in Latimeria or Neoceratodus since this muscle is specific to the tetrapod configuration of a knee joint.
Hypaxial musculature Table 5 displays hypaxial muscles attaching to the pelvic girdle. No complete equivalences can be established despite the fact that Latimeria, Neoceratodus and Necturus all have a specific pelvic component of the hypaxial musculature. In Latimeria, the hypaxial muscle inserts onto the tip of the pubic ramus but neither the hypaxial musculature in Neoceratodus nor the rectus abdominis insert on the pelvic girdle. Two hypaxial muscles inserting on the ischium are present in Necturus but not in Latimeria and Neoceratodus. These are involved in tail movement and are specific to tetrapods in more or less elaborated ways [39].

Development of the pelvic musculature in the Australian lungfish
At stage 50, there is no trace of pelvic musculature. Only the hypaxial musculature is visible in lateral and ventral views ( Figure 4A). At stage 51, the deep ventral abductor depressor (D. V. abd. depr.), the deep ventral adductor depressor (D. V. add. depr.) and the dorsomesial adductor levator (Dm add. lev.) appear ( Figure 4B) as short and thin muscles. These muscles are longer and thicker ( Figure 4C) at stage 52 and are accompanied by radial flexors (R.f.) over five or six axial elements and radial-axials (R.-A.) on the pre-axial side of the second element. Radial flexors proximal on the fin are more developed than the ones more distal and these muscles seem to develop in a proximo-distal direction following the development of axial elements. The deep ventral abductor depressor continues to develop anteriorly at stage 54 and the superficial ventrolateral adductor (S. vl. add.) is now present between the dorsal side of the pelvis and the postaxial edge of the fin ( Figure 4D). No new muscles appear at stage 56. However, the dorsomesial adductor levator is more developed and its fibers extend more distally. The radial flexors are also longer, and the radial-axials extend on all six axial elements pre-axially and appear for the first time on post-axial radials ( Figure 4E). At stage 61, proximal muscles (D. V. abd. depr. and Dm. add. lev.) are fully developed and begin being covered by the superficial ventromesial abductor (S. vm. abd.), which originates at a very anterior position on the pelvic girdle ( Figure 4F). Radial-axials on the post-axial side of the fin are now present through its entire length. Lepidotrichial flexors (L.f.) are present through the entire pre-axial side of the fin but are only

Development of the pelvic musculature in the axolotl
The very first stage of this series is at the pelvic bud stage 55 [34] when only the rectus abdominis (R. a.) and the caudofemoralis (Cfe) are present, both of which are extensions of the hypaxial musculature ( Figure 5A). Those muscles flank the emerging limb bud and at the following stage (1.5 cm), the ischiofemoralis (Isfe), the puboischiofemoralis internus (Pisfe int) and the iliofemoralis are all present ( Figure 5B l ) (all muscles inserting into the limb). At 2.0 cm, the caudofemoralis muscle now extends all the way to the base of the femur and is a lot thicker. The puboischiotibialis (Pist) is now covering the ischiofemoralis and the caudopuboischiotibialis (Cpist) is now visible, linking the puboischiotibialis to bands of muscles extending towards the caudal vertebrae ( Figure 5C l ). Fibers of the iliofemoralis now extend caudally, suggesting that its point of origin on the ilium extends dorsally in connection with the dorsal extension of the iliac cartilage ( Figure 2). Shank flexors (Figure 5C l ) and extensors ( Figure 5C v ) are now present, as well as the pubotibialis (pt), visible in the middle of the thigh in the ventral view. The 2.5 cm larva is very similar to the previous stage except for the distal leg muscles, which are missing along with the rest of the leg in this individual, having fallen victim to cannibalism (a common behavior in axolotl). The puboischiotibialis is thicker and extends farther anteriorly, the iliofemoralis extends a little more dorsally and the caudopuboischiotibialis can be clearly seen connecting the puboischiotibialis to muscle fibers extending to the caudal vertebrae ( Figure 5D l ). In the ventral view, the puboischiofemoralis externus (Pisfe ext) is now present, its origin overlapping with that of the puboischiofemoralis internus ( Figure 5D v ). Due to species variation in the pelvic shape, fibers of the puboischiofemoralis externus are more parallel to the hypaxial musculature than those of Necturus but can be distinguished from it by its posterior point of origin. The ischioflexorius (Isfl) is probably present at this stage as a very faint band of muscle originating from the ischium and extending distally on the post-axial side of the thigh (Figure 5D vd ). The last stage of this series is very similar to 2.5 cm. At 3.0 cm, the ischioflexorius is thicker and more visible and shank extensors and flexors as well as the popliteus are present (Figure 5E d ). All other muscles seem to have reached their adult conformation ( Figure 5E) but the ischiocaudalis, the iliotibialis, ilioextensorius and iliofibularis cannot be seen in any view. They are all small, deep muscles, which may be covered by more superficial muscles, or they might be absent altogether in Ambystoma.

Discussion
Comparative pelvic development and hypotheses of pelvic evolution One of the major questions about the evolution of the tetrapod pelvic girdle during the fish-tetrapod transition has been the apparent shift in position of the acetabulum. In sarcopterygian fish, the fin articulation is located posteriorly on the pelvic girdle [11,12,15] whereas in early tetrapods, it is lateral [16,17] (Figure 1). This seems to imply that the acetabulum has moved across the lateral face of the pelvis during the transition. However, the fact that the acetabular region is the first part of the pelvis to develop in both Neoceratodus and Ambystoma suggests that the acetabulum should instead be regarded as a fixed landmark (grey line, Figure 3). Such a change of perspective makes the evolutionary transformation of the pelvis much easier to understand. As shown in the descriptions above, the principal difference in early pelvic development between salamander and lungfish is that in the salamander, the pubis grows anteriorly and the ischium posteriorly from the acetabular region, whereas in lungfish the pubis grows anteriorly and the ischium is absent; in other words, chondrogenic cells proliferate both anteriorly and posteriorly in the salamander, whereas in the lungfish they only proliferate anteriorly. The presence of an ischium and the lateral position (and orientation) of the acetabulum are, therefore, developmentally coupled morphological states, presumably resulting from a change in molecular signalling in the immediate ancestors to tetrapods: there is no need to postulate a 'migration' of the acetabulum. Figure 3 schematizes pelvic development in lungfish and axolotl with this hypothesis in mind. The grey line represents the position of the acetabula for both species.
Another major question about the evolution of the pelvic girdle in tetrapods concerns the origin of the ilium and its relationship, if any, to the presence of an iliac process in Eusthenopteron, Goologongia and other fish members of the tetrapod stem group [11,12,14,15]. The fact that the ilium of salamanders slowly extends dorsally during development, only contacting the sacral rib at a late stage when the pelvis is more or less fully formed, means that the ilium passes through a protracted developmental stage when it closely resembles the iliac process of these fishes ( Figure 1). Together with the wide phylogenetic distribution of the iliac process, which suggests that it is a general character for the 'fish' part of the tetrapod stem group [11,12,14,15], this provides strong circumstantial evidence for the homology of the two structures.
Muscle homologies in Coelacanth, Lungfish and a salamander, and the evolution of pelvic musculature Here, the muscles of Necturus are described as a model for basal salamanders. There are almost twice as many muscles originating from the pelvic girdle in salamanders as in lungfish (13 versus 7) but six of these muscles are either originating or inserting on the ischium, which is a purported novelty in tetrapods. Very few pelvic muscles present in Necturus could not be associated with a muscle present in Neoceratodus. Among the exceptions is the puboischiofemoralis externus, originating mostly on the anterior process of the pubis, which is probably not homologous to the pubic process of Neoceratodus. However, the puboischiofemoralis is the only muscle with an ischiatic origin that could not be compared to a muscle in Neoceratodus. In all other cases, muscles originating from the mesial surface of the pelvic girdle in Neoceratodus had the same function and were very similar in terms of insertion point to muscles originating from the ischium in Necturus. This suggests that the ischium originated as an inflation of the posteromesial face of the pubis, created by increased proliferation of chondrogenic cells posterior to the acetabulum, an interpretation that also fits well with the data from skeletal development (Figures 2, 3). If the ischium is, in fact, a posterior projection of the pubis, the muscle groups paired up in Tables 1, 2, 3, 4 and 5 are likely to be truly homologous and the muscles themselves would not have changed substantially during the fish-tetrapod transition. This is consonant with the results from the comparative study of muscle development (see below). As for muscles originating from the ilium in Necturus, almost all seem to correspond to muscles originating from the posterodorsal surface of the pelvic girdle in Neoceratodus. Most known pelvic girdles of fossil tetrapodomorph fish have a lateral ridge or process on their posterodorsal surface [11,12,14,15]. Most of these ridges are very slight but given the fact that muscles originating on the ilium in Necturus originate at the very base of it, a homologous muscle could in principle originate from a small process situated at the same position on the pelvic girdle of a fish. This gives further support to the hypothesis that the iliac process of Eusthenopteron and other fish members of the tetrapod stem group is the precursor to the ilium of tetrapods.

Comparative muscular development
In both Ambystoma and Neoceratodus, the hypaxial musculature develops first, followed by deep musculature originating from the pelvic girdle and inserting proximally onto the fin/limb. All muscles that have been equated in the tables above develop in the same order except for the caudofemoralis, which is the first pelvic muscle to develop in axolotls. Its homologue, the superficial ventrolateral abductor, develops last. This suggests a large heterochronic shift in the appearance of this muscle. The other, less dramatic, exception is the ischioflexorius, which develops slightly earlier in sequence than its counterpart, the superficial ventromesial adductor. The deep ventral abductor depressor, equivalent to the puboischiofemoralis internus, and the deep ventral adductor depressor, equivalent to the ischiofemoralis is the first homologous pair to appear. The dorsomesial adductor levator then appears in Neoceratodus and the iliofemoralis, in Ambystoma. These muscles have not been homologized on the basis of origin and insertion points but their developmental sequence might indicate some homology. The dorsomesial adductor levator of lungfish develops in the same sequence as the iliofemoralis and the radial-axial develop simultaneously to the caudopuboischiotibialis. Homologous pairs of slightly more superficial muscles appear next: the radial flexors/ shank flexors and the superficial ventrolateral adductor/ puboischiotibialis. The first of the superficial muscles to form in Ambystoma is the puboischiofemoralis externus, which was not homologized to any muscle in lungfish and is followed by the ischioflexorius. Its homologue (superficial ventromesial adductor) develops slightly later in lungfish, after the appearance of the superficial ventromesial abductor (synchronous with the development of the pubotibialis). This heterochronic shift might be functional in nature or an artefact of the whole mount staining method where discerning incompletely differentiated muscles can be difficult. Lepidotrichal flexors and the popliteus develop last. This might be due to the superficial nature of those muscles. Despite large differences in pelvic morphology, development of both the cartilage and muscles of these species is thus very similar. One important difference is that while the order of appearance is the same, there is more lag between consecutive muscle appearances in Neoceratodus. While several muscles appear at once in Ambystoma, new muscles appear one by one in Neoceratodus.
Further insight can be gained from the direction in which muscles develop. For example, the iliofemoralis is one of the first pelvic muscles to appear in 1.5 cm long Ambystoma larvae, the stage at which the ilium starts to extend dorsally (Figure 2I lateral ). This muscle continues to extend dorsally, presumably following the ilium. The same thing is true for muscles attaching to the pubis: in both Neoceratodus and Ambystoma, muscles originating from the pubis extend anteriorly through development, presumably following the anterior extension of the pubic cartilage. Similarly, radial-axial and lepidotrichial flexors of the fins of Neoceratodus start to develop pre-axially before developing post-axially, in a manner resembling that of the appearance of radials on the fin [38]. In summary, the sequence and mode of development of the pelvic musculature appears to be substantially conserved between Neoceratodus and Ambystoma, suggesting that the morphological transformation from pelvic fin to tetrapod hind limb was accomplished without major heterochronic reorganization of muscle development.

Conclusions
Despite large differences in pelvic morphology, the development of the pelvic girdle in Neoceratodus and Ambystoma is strikingly similar. Most pelvic muscles can be homologized between the two species and homologous muscles develop in the same order. Deep muscles develop first, followed by some muscles unique to either lungfish or salamanders and superficial muscles develop last. The only exception to this gradual development from deep to superficial is the appearance of the caudofemoralis of the axolotl at the very beginning of development whereas the superficial ventrolateral abductor of the lungfish develops last, along with the rest of the superficial musculature. The caudofemoralis originates from caudal vertebrae and is involved in tail motion. Its early appearance might be due to the very large reliance on undulatory motion of larval axolotls. Its early development might also be a consequence of the derived mode of development recently described for tetrapod pelvic appendages [40]. In lungfish, pelvic fin musculature develops from both a myotomal extension from the body wall and the migration of somite derived mesenchymal cells. However, in tetrapods, including Ambystoma, pelvic limb muscle formation is solely dependent on somite derived mesenchymal cells; it is argued that this mode of development would allow for earlier initiation and easier heterochronic reorganization of muscle development.
With regards to the skeleton, if we consider the acetabulum to be a fixed landmark, the evolution of the ischium simply required pre-chondrogenic cells of the pubis to migrate posteriorly as well as anteriorly in order to form it. This would explain why very similar muscles originating from the inside of the pubis in the lungfish originate from the ischium in the salamander. The ischium does not ossify in neotenic forms, such as the axolotl. However, in several salamander taxa, the ischium ossifies whilst the pubis remains unossified (Personal observation, CAB). This is reminiscent of the condition in early tetrapods [2] and might have initially been caused by the greater reliance on muscles originating from the ischium such as the puboischiotibialis. Ossification would have occurred in the ischium as a result of muscular tension which is known to promote ossification [41,42]. It would be interesting to combine the results of this study and recent biomechanical studies [8,43] to test those hypotheses. As for the origin of the ilium, some very similar muscles originating from the posterodorsal surface of the pelvis in Neoceratodus originate from the base of the ilium in Ambystoma, suggesting that the lateral ridge or process present in most tetrapodomorph fish is the precursor of the tetrapod ilium ( Figure 1). This precursor would then have extended dorsally to eventually reach and articulate with the sacral rib, as observed by the development of both iliac cartilage and muscles. In order to test those hypotheses, the origin and insertion points of the muscles of Latimeria, Neoceratodus and Ambystoma should be verified using three-dimensional models generated from undissected computed tomography (CT)-scanned specimens. This will provide a more robust comparative morphological framework against which to analyze triple-stained (double immunostained for muscles and nerves plus alcian stained for cartilage) growth series of Neoceratodus and Ambystoma by optical tomography (OPT), which allows for three-dimensional three-color visualization. As a complement to these techniques for investigatng extant morphologies, the application of propagation phase contrast synchrotron microtomography (PPC-SRμCT) to the study of fossil bone histology is opening up new possibilities for detecting muscle attachments on the pelvic bones of fossil fishes and early tetrapods [44]. We expect that data extracted by these novel methods will cast a great deal of light on the transformation of the pelvic appendage at the fish-tetrapod transition and perhaps help resolve the question of the origin of tetrapod locomotion raised in several recent publications [8,43,45].