Fish Fingers: Digit Homologues in Sarcopterygian Fish Fins

JOURNAL OF EXPERIMENTAL ZOOLOGY (MOL DEV EVOL) 308B:757 768 (2007) Fish Fingers: Digit Homologues in Sarcopterygian Fish Fins ZERINA JOHANSON 1, JEAN JOSS 2, CATHERINE A. BOISVERT 3, ROLF ERICSSON 2, MARGARETA SUTIJA 4, AND PER E. AHLBERG 3 1 Department of Palaeontology, Natural History Museum, London, UK 2 School of Biological Sciences, Macquarie University, Sydney, New South Wales, Australia 3 Department of Physiology and Developmental Biology, Evolutionary Biology Centre, Uppsala University, Uppsala, Sweden 4 Hubrecht Laboratory, The Netherlands Institute of Developmental Biology, Utrecht, The Netherlands ABSTRACT A defining feature of tetrapod evolutionary origins is the transition from fish fins to tetrapod limbs. A major change during this transition is the appearance of the autopod (hands, feet), which comprises two distinct regions, the wrist/ankle and the digits. When the autopod first appeared in Late Devonian fossil tetrapods, it was incomplete: digits evolved before the full complement of wrist/ankle bones. Early tetrapod wrists/ankles, including those with a full complement of bones, also show a sharp pattern discontinuity between proximal elements and distal elements. This suggests the presence of a discontinuity in the proximal-distal sequence of development. Such a discontinuity occurs in living urodeles, where digits form before completion of the wrist/ankle, implying developmental independence of the digits from wrist/ankle elements. We have observed comparable independent development of pectoral fin radials in the lungfish Neoceratodus (Osteichthyes: Sarcopterygii), relative to homologues of the tetrapod limb and proximal wrist elements in the main fin axis. Moreover, in the Neoceratodus fin, expression of Hoxd13 closely matches late expression patterns observed in the tetrapod autopod. This evidence suggests that Neoceratodus fin radials and tetrapod digits may be patterned by shared mechanisms distinct from those patterning the proximal fin/limb elements, and in that sense are homologous. The presence of independently developing radials in the distal part of the pectoral (and pelvic) fin may be a general feature of the Sarcopterygii. J. Exp. Zool. (Mol. Dev. Evol.) 308B:757 768, 2007. r 2007 Wiley-Liss, Inc. How to cite this article: Johanson Z, Joss J, Boisvert CA, Ericsson R, Sutija M, Ahlberg PE. 2007. Fish fingers: digit homologues in Sarcopterygian fish fins. J. Exp. Zool. (Mol. Dev. Evol.) 308B:757 768. The tetrapod limb develops proximo-distally in three distinct regions: the humerus/femur (stylopod), the radius, ulna/tibia, fibula (zeugopod) and most distally, the hands and feet (autopod). The autopod can be further divided into the more proximal wrist and ankle elements (carpus/tarsus, referred to as the mesopodium) and the more distal digits (or acropodium) (Wagner and Chiu, 2001). The acropodium is generally considered to be restricted to the tetrapod radiation and absent from sarcopterygian fish paired fins, both fossil and living forms. The full mesopodial complement itself includes proximal, central, and distal elements. The most proximal mesopodial elements are the ulnare/ fibulare and intermedium, which can be homologized uncontroversially to elements in the sarcopterygian fish fin, and the tibiale/radiale. These are phylogenetically stable elements that Grant sponsor: Australian Research Council Discovery; Grant number: DP-663874; Grant sponsor: Swedish Research Council; Grant numbers: 2003-6992; and 2003-3308. Correspondence to: Zerina Johanson, Department of Palaeontology, Natural History Museum, Cromwell Road, London, UK SW7 5BD. E-mail: z.johanson@nhm.ac.uk Received 26 April 2007; Revised 24 July 2007; Accepted 6 August 2007 Published online 11 September 2007 in Wiley InterScience (www. interscience.wiley.com). DOI: 10.1002/jez.b.21197 r 2007 WILEY-LISS, INC.

758 Z. JOHANSON ET AL. undergo only one major evolutionary transformation, fusion into astragalus and calcaneum in the foot of amniotes. The most distal mesopodial elements, the distal tarsals/carpals, primitively align one-to-one with the digits. These are also relatively stable across phylogeny although their number is sometimes reduced. Between these stable proximal and distal regions of the mesopodium lies a less stable middle region comprising the so-called central bones, which are highly variable in number. Clear one-to-one alignment is lacking between the distal tarsals/ carpals and the centrals or more proximal bones (Fig. 1). Recent discoveries of Late Devonian stem-group tetrapods (Coates and Clack, 90; Coates, 96; Clack, 2005) have added greatly to our knowledge of the origin of limbs. These show not only that the first limbed vertebrates had supernumerary digits (seven or eight), but also that they lacked a full complement of wrist/ankle bones. The ankles are most informative, as these are well ossified and Fig. 1. Ankle evolution in early tetrapods. (A) Phylogeny of representative early tetrapod hind limbs showing stepwise elaboration of ankle. In the upper series of drawings each ankle is disassembled to show the line of discontinuity (dashed line) between the digits plus distal tarsals and the more proximal elements. (B) Diagrammatic representation of the hypothesis that elaboration of the mesopodium occurred through proliferation of elements from both sides toward the line of discontinuity. The dotted-line structure anterior to the first digit of Ichthyostega (shown only in the lower drawing) is a poorly ossified element, possibly, but not definitely, representing another digit. Acanthostega and Ichthyostega redrawn from Coates, 96; Proterogyrinus from Panchen and Smithson, 87, originally from Holmes, 84. All modified by addition of gray tone to centrals and distal tarsals. Not to scale. Identification of ankle bones in Acanthostega and Ichthyostega follows Coates, 96. Coates et al. (2002) claimed that distal tarsals are absent in these genera, but the elements formerly interpreted as distal tarsals in Acanthostega were not reidentified. We cautiously reject this new interpretation, because we feel that the shape and position of the disputed elements justifies their original identification as distal tarsals, particularly in Ichthyostega where the ankle is preserved in articulation. We agree, however, that the three most posterior digits in Acanthostega and the two most posterior in Ichthyostega lack distal tarsals. If the interpretation of Coates et al. (2002) is accepted, this modifies our evolutionary hypothesis in B by implying that the proliferation of mesopodial elements in early tetrapods occurred exclusively distal to the line of discontinuity. fi, fibula; fib, fibulare; in, intermedium; ti, tibia; tib, tibiale.

DIGIT HOMOLOGUES IN SARCOPTERYGIAN FISH FINS 759 do not present difficulties of distinguishing between merely unossified and genuinely absent elements (Fig. 1A). In both the very primitive Acanthostega and the somewhat more derived Ichthyostega, the two or three most posterior digits articulate directly with the fibulare with no intervening distal tarsals. Furthermore, Ichthyostega has only a single central bone, contrasting with the three or four centrals seen in basal crown-group tetrapods (Coates, 96). No ossified central is preserved in the ankle of Acanthostega, but the alignment of surrounding elements suggests that one or at most two centrals were present. In Tulerpeton, the most crownward of the Devonian stem-group tetrapods, the wrist contains few elements and is interpreted as lacking a complete series of distal carpals (the ulnare articulating directly with two or three digits), whereas the ankle approaches the fully elaborated condition of basal crown-group tetrapods (Lebedev and Coates, 95). Thus, when the tetrapod autopod first evolved, digits (the acropodium) appeared before the complete mesopodium (Wagner and Chiu, 2001; Coates, 2003; Clack, 2004). Furthermore, the later elaboration of the mesopodium, which involved an increase in the number of centrals and distal tarsals/carpals, occurred immediately on either side of the aforementioned line of pattern disjuncture between proximal and distal elements. This suggests that a developmental discontinuity existed between the zeugopod-plus-proximalmesopodium and the autopod, with the elaboration of the mesopodium occurring through distal addition of elements (centrals) to the zeugopod complex and proximal addition of elements (distal carpals/ tarsals) to the autopod. Such a discontinuity occurs during autopodial development of urodele amphibians, where the paddle-like manus/pes characteristic of amniotes does not occur (Burke and Alberch, 85; Fabrezi and Barg, 2001). Instead, digits develop as individual buds, before the condensation of the mesopodium. For example, in Desmognathus aenus, metacarpal II is the first element of the autopodium to condense and chondrify and is well separated from the only other limb condensations, the radius and ulna (Franssen et al., 2005). Accepting that a fully developed mesopodium is not essential for the addition of digits in the development of an autopodium allows us to reassess the origin of digits among fish representatives of the Sarcopterygii, the most closely related fish to the digited tetrapods (Acanthostega, Ichthyostega, Tulerpeton, and all more crownward taxa). These representatives include both fossil and living taxa (e.g., Ahlberg and Johanson, 98; Zhu and Yu, 2002), providing access to palaeontological, morphological, molecular, and ontogenetic information. Of interest here is the relationship in the fin between the proximal wrist/ankle homologues and the fin radials and any potential developmental discontinuity between these. Among living sarcopterygian fishes, information can be obtained by examining morphological and genetic patterns of pectoral fin development in the Australian lungfish, Neoceratodus forsteri, the only living sarcopterygian fish available as a realistically tractable laboratory animal (see Fig. 2 for its phylogenetic position). Comparisons of these patterns with those of tetrapods allow us to draw certain inferences about development and element homologies in those fossil sarcopterygians that are more closely related to tetrapods than lungfishes (i.e., fish members of the Tetrapodomorpha; Ahlberg and Johanson, 98; also Fig. 2). The results illuminate the transition from fins to limbs, and in particular the origin of digits. MATERIALS AND METHODS Developing lungfish were raised from eggs collected from lungfish spawning ponds at Macquarie University (protocols approved by the Macquarie University Animal Ethics Committee, approval 2003/001). Stages used in the study covered the early development of the pectoral fins, lungfish post-hatching stages 43 51. Some fish were fixed in 10% neutral buffered formalin, stained for cartilage and cleared according to methods described by Dinkergus and Uhler ( 77). These were photographed with an Olympic BH2 phase contrast (Olympus, Japan) (Fig. 3A E, I) or an Olympic BX50 microscope (Olympus, Japan) (Fig. 3J K). Whole mount in situ hybridization (ISH) was carried out on paraformaldehyde (4%) fixed specimens according to methods by Wilkinson ( 92), modified to increase penetration of the probe. Fins of Neoceratodus develop with a particularly impenetrable outer epithelial layer. To try to alleviate this, proteinase K treatment was substituted by dimethyl sulfoxide /methanol (1:1) for 8hr at 41C which allowed the penetration of a small-medium sized probe only. The DIG-labeled ISH probes (sense control and antisense) for Hoxd13 were constructed from a 179 bp fragment

760 Z. JOHANSON ET AL. Fig. 2. Hypothesis of homology for different sarcopterygian pectoral fins/forelimbs. Note the contrast between the highly conserved proximal region (dark blue, pink) and the variable distal region (yellow), and the prevalence of trichotomies and polychotomies in the latter. In Neoceratodus, the second axial element forms from the fusion of radius and ulna; note that the more distal pink elements in Neoceratodus form distal to the radius and are thus probably not equivalent to the intermedium of other sarcopterygians. In Acanthostega the ulnare is not ossified, but its hind-limb counterpart (fibulare) articulates directly with the posterior digits. Latimeria and Neoceratodus from Panchen and Smithson, 87, others redrawn from Shubin et al., 2006. Sauripterus and Tiktaalik slightly modified. Not to scale. amplified by reverse transcription polymerase chain reaction using gene-specific primers (fwd- TCCCGGGTTACATAG ACATG and rev- ATCCC CTGCATACGA TTTCC-outside but bordering the homeodomain). The sense control probe for Hoxd13 expression was completely negative for all stages of fin development examined. Following ISH, the pectoral fins were removed and mounted in glycerol for imaging with an Olympus microscope BX50 (Fig. 3F H). RESULTS The pectoral fin of Neoceratodus forsteri Ontogenetic development The pectoral fin of adult Neoceratodus is dominated by a medial axis comprising a series of cartilaginous elements, with preaxial and postaxial radials extending outward from this axis. During early development, axial homologues of

DIGIT HOMOLOGUES IN SARCOPTERYGIAN FISH FINS 761 the humerus, ulna, and one more distal element of the axis begin to form before a small condensation representing the radius appears (Fig. 3A, Ra). At a later stage, the radius fuses with the ulna (Ul). This developmental sequence implies that the radius and ulna do not form as a branching event from the humerus (Joss and Longhurst, 2001; Coates et al., 2002; Cohn et al., 2002). The third axial condensation can be homologized with the ulnare, but unlike Devonian tetrapodomorphs, in which this third condensation, like the second, is also paired, in N. forsteri it is a single element. In other words, there does not seem to be an equivalent to the tetrapodomorph intermedium (Joss and Longhurst, 2001; Johanson et al., 2002). Within the fin, each new axial element condenses from within a continuous field of cells, representing precartilaginous mesenchymal cells or presumptive chondroblasts (Fig. 3A, pcc), before extracellular matrix deposition. Joint formation separates cells within this field into discrete elements (Johanson et al., 2002; Fig. 3A, larger black arrow), with the joint developing to form a strip of lightly chondrified cells that will subsequently contribute to appositional growth of the axial elements. In older individuals, these joints will narrow as appositional growth slows and the joint becomes more functional, allowing movement between elements that are by now in close juxtaposition. The axis continues to develop by adding at least six new elements before the condensation of radials (Fig. 3J); the latter form first preaxially, one per axial segment starting with the ulnare (Fig. 3K) and continuing distally to include at least four preaxial radials before the condensation of radials postaxially (Joss and Longhurst, 2001; Fig. 3L). As noted, the axial elements of the pectoral fin develop within a field of precartilaginous mesenchymal cells that connects each of these elements to one another. A comparable field also connects additional elements added to the radius (Fig. 3B and C, pcc) and to each of the other developing radials (Fig. 3E, pcc). But, in marked contrast, the first, most proximal element of each preaxial and postaxial radial beyond the radius (Fig. 3B, pr.ra, pt.ra) does not form from within a prechondrogenic field of mesenchymal cells surrounding the axis, and lacks an initial cellular connection to the axis (Fig. 3B E; small arrowheads). This absence of cells between the axis and the radials indicates an absence of the initial joint formation within the prechondrogenic field responsible for separating new, distinct elements within the main axis (compare Fig. 3A, black arrow and Fig. 3B E). By comparison, the radials themselves develop subsequently by segmentation, and can be observed forming in the same manner as elements within the main fin axis, by initial joint formation (compare Fig. 3A and E, area marked by larger black arrow). Also, cell outlines (representing regions of joint formation) are clearly seen between additional elements within the preaxial and postaxial radials themselves, but not between the proximal part of the radial and the axis (compare Fig. 3C and D). It is not until later developmental stages that growth of the preaxial and postaxial radials proceeds proximally, toward the axis, where a functional joint surface subsequently develops ( j, Fig. 3I; Joss and Longhurst, 2001; Fig. 21.1e). It is important to note that the late development of this joint supports the interpretation that when the radials first develop, they do so separately with no initial connection to the fin axis (i.e., not as segmental events). This lack of cellular connection includes the third axial element, representing the homologue of the ulnare in the proximal part of the tetrapod wrist (Fig. 3A and B, Uln). We suggest that this pattern of development, where the fin radials develop independently from the Neoceratodus fin axis, including the ulnare, is directly comparable with the discontinuous or interrupted development between the proximal wrist and digits in early tetrapods and living urodeles. Moreover, the elements associated with the first preaxial radial, the radius (part of the zeugopod rather than the autopod) do not develop independently, indicating that radial development has changed between the more proximal part of the fin (zeugopod homologue) to the more distal (potential autopod homologues). This is supported by the expression pattern of Hoxd-13 in Neoceratodus pectoral fins, described below. Gene expression patterns Hoxd13 expression in tetrapod limb development shows two distinct phases: an early phase that is also present in zebrafish, and a second, late phase expanding across the distal region of the limb bud that was once thought to be unique to tetrapods, but has been described recently in the distal pectoral fins of the basal actinopterygian Polyodon (Sordino et al., 95; Deschamps, 2004; Tarachini and Duboule, 2006; Davis et al., 2007; Fig. 2). In Neoceratodus fin development, only a single late phase of Hoxd13 expression has so far been

762 Z. JOHANSON ET AL. detected. Failure to detect an early phase may reflect genuine absence or may result simply from the technical difficulties associated with probe penetration in this animal (see Materials and Methods). Whatever the case, the key question for present purposes is whether the observed late A B C D E Uln Ra Uln F I j j j Ul Uln G st.48 J st.49 K st.49 H st.50 L st.50 st.51

DIGIT HOMOLOGUES IN SARCOPTERYGIAN FISH FINS 763 expression phase can be homologized with the late expression phase in tetrapods and Polyodon. In Neoceratodus, expression begins just before the condensation of radials beyond the radius, and is primarily in the mid-to-distal anterior (preaxial) region of the fin (Fig. 3F); this occurs after the stylopod, zeugopod, and several further elements of the fin axis have developed. Expression next shifts to occur throughout the fin, strongest in the mid-to-distal regions but excluded from the epidermis and, most notably, from the condensing elements in the main axis of the fin (Fig. 3G). In older fins, the expression of Hoxd13 shifts again, to be strongest in the posterior (postaxial) region (Fig. 3H). This expression pattern prefigures the condensation of the radials. As the expression of Hoxd13 shifts from anterior to posterior (preaxial to postaxial), some preaxial (anterior) radials begin condensation in the region of the fin where Hoxd13 expression had first appeared (compare Fig. 3F with Fig. 3K). Later, soon after Hoxd13 expression has ceased, the first postaxial (posterior) radials begin to appear in the same domain or region of the fin as Hoxd13 had been last expressed (compare Fig. 3H with Fig. 3K, black arrow, Fig. 3L). Thus, fin radial development is linked closely to the Hoxd13 field, and to an expansion of this field, which is comparable with the development of digits in tetrapods (Dolle et al., 93; Tarachini and Duboule, 2006, late-phase expression, discussed in more detail below) but not comparable with the early phase of expression in tetrapods, Polyodon and zebrafish, where Hoxd13 is only expressed posteriorly (Sordino et al., 95; Deschamps, 2004; Davis et al., 2007). We conclude tentatively that the observed Hoxd13 expression phase in Neoceratodus is homologous with the late phase of tetrapods and Polyodon. In addition to the characteristics of the expression pattern itself, our conclusion is supported by the phylogenetic position of Neoceratodus within a bracket formed by Polyodon (a primitive actinopterygian; Davis et al., 2007) and tetrapods, both of which have a late expression phase. This implies on grounds of parsimony that a late expression phase should be expected in Neoceratodus. Fin structure in other Sarcopterygii A comparison of radial development in Neoceratodus with digit development in tetrapods, particularly that of urodeles, reveals similarities in sequence of tissue condensations which suggest that these structures may be homologous. Our initial work on gene expression patterns in the pectoral fin of Neoceratodus, focusing on Hoxd13, also shows that this gene is expressed late in ontogeny in the areas where the radials are about to condense, much in the same way as 5 0 Hox genes are expressed immediately before digit condensation in tetrapods. The recently documented presence of a comparable late 5 0 Hox expression phase in the basal actinopterygian Polyodon suggests that the presence of such a phase is primitive for the Osteichthyes as a whole. We thus have some evidence that a conserved molecular mechanism initiates radial condensation in Neoceratodus and digit condensation in tetrapods, lending further support to the hypothesis that radials and digits are homologous. However, to test this hypothesis we must also examine the range of paired fin structures among the Sarcopterygii, to determine whether the lungfish and tetrapod structures form part of a phylogenetic continuum (are homologous) or have separate origins (are convergently acquired). Lungfish seem to be the closest living relatives of tetrapods, with coelacanths as the sister group of the lungfish-tetrapod clade (Zhu et al., 2001; Brinkmann et al., 2004). Among the living lung- Fig. 3. Neoceratodus forsteri, pectoral fins, preaxial toward top of each image. (A E, I L) Wholemount fin cleared and stained with alcian blue. (F H) Expression patterns of Hoxd13. (A) Early stage fin, showing development of axial elements and the radius. Large black arrow indicates joint formation between two axial elements. (B) Later stage fin, showing preaxial and postaxial radials developing. (C) Closeup of B, including radials developing from radius (pcc and black arrows) as well as preaxial radials, comparing independent origin of proximal segment (no connecting precartilaginous mesenchymal cells between axial element and radial; black arrowheads) and subsequent development within each radial, where radial elements are connected by these precartilaginous cells (includes radials developing from radius). (D) Closeup of postaxial radials in (B), developing independently from the axial element, as indicated by absence of connecting precartilaginous mesenchymal cells (black arrowheads). (E) Closeup of postaxial radials, showing original precartilaginous mesenchymal cells (to the right of figure, originating independently from axial element). To the left of the figure, larger black arrow indicates joint formation and formation of new radial element more distally. Hu, humerus; j, joint; pcc, precartilaginous mesenchymal cells; pr.ra, preaxial radials; pt.ra, postaxial radials; Ra, radius; scc, scapulocoracoid; Ul, ulna; Uln, ulnare. Scale: B, I 5 250 mm; A, C E, G 5 100 mm; F H, J K 5 1.0 mm.

764 Z. JOHANSON ET AL. fishes, Neoceratodus is uniquely primitive in retaining preaxial and postaxial radials; this pinnate pattern of the fin endoskeleton was established already in Devonian stem-group lungfish (Ahlberg, 89; Ahlberg and Trewin, 95). However, a large number of fossil sarcopterygian fishes belong to the tetrapod stem group (Tetrapodomorpha; Ahlberg and Johanson, 98) and are thus related more closely to tetrapods than Neoceratodus (Fig. 2). Comparison of the paired fins in these different groups suggests that the primitive condition for Sarcopterygii is a short fin, probably with four axial elements (Fig. 2, coelacanth and osteolepiform). The radius and ulna primitively remain separate, and postaxial radials (if present) are restricted to the distal part of the fin (Ahlberg, 89). Thus, although Neoceratodus faithfully reflects the primitive lungfish condition, this is itself derived relative to the primitive sarcopterygian condition; the fully formed Neoceratodus pectoral fin is modified from the primitive sarcopterygian pattern in having an elongated axis, fused radius1ulna, and a long series of both preaxial and postaxial radials. Tetrapods, by contrast, retain a separate radius and ulna, and seem to have only three axial elements (humerus, ulna, ulnare). Fish members of the tetrapod stem group such as Eusthenopteron (Fig. 2) similarly have fins with an independent radius, a low number of axial elements, and postaxial radials (if present) restricted to the distal part of the fin. Of the two fishes that are phylogenetically closest to tetrapods, the elpistostegids Tiktaalik and Panderichthys, Tiktaalik (Fig. 2) has five axial elements of which all carry preaxial radials and three also carry postaxial or terminal radials (Shubin et al., 2006). Panderichthys has three axial elements and has been described as lacking postaxial and terminal radials (Vorobyeva, 2000), but work in progress on a computed tomography scan model suggests that terminal radials are in fact present (Boisvert, 2006). Because the paired fin skeleton of Neoceratodus differs in many ways from the ancestral condition for tetrapods, we cannot use simple positional homology criteria to directly equate its radials with tetrapod digits, despite their developmental similarities. However, the radials and digits may be expressions of a conserved patterning component that has been retained in both lineages even though some other aspects of fin architecture have changed. If so, they should form part of a phylogenetic continuum, that is the intervening branches on the tree (the different tetrapodomorph fish groups) should also show comparable structures. This is obviously difficult to determine, as the question comes down to whether particular radials in fossil fin skeletons had non-independent development with a cellular connection to the nearest axial element, like the radius of Neoceratodus, or independent development like the other preaxial and postaxial radials of Neoceratodus. Fortunately, the morphology of the fin skeletons supplies a tentative answer. There is a notable contrast between the radials in the proximal and distal parts of sarcopterygian fins (Fig. 2). In the proximal part each axial element carries a single preaxial radial, giving an axial series of (to use tetrapod forelimb terminology) humerus, ulna, and ulnare, with the radius and intermedium as preaxial radials. These elements can be identified without ambiguity in all tetrapodomorph fishes as well as in the coelacanth Latimeria (Millot and Anthony, 58), and with the exception of the intermedium also in Neoceratodus; there is no reason to doubt that they constitute a conserved region (Fig. 2; blue and pink). On the developmental evidence from tetrapods and Neoceratodus, we suggest that this region comprises axial elements plus non-independent radials (radius, intermedium) and would not be associated with late Hoxd13 expression. In the distal part of the fin, by contrast, the number of radials is extremely variable and trifurcations and polyfurcations are common (Fig. 2: yellow). Such patterns are more readily explained as the product of an array of independently developing radials that only establish contact with the axial elements late in ontogeny (compare with Fig. 3B E, I). We infer from this pattern that the radials in the most distal part of all known sarcopterygian paired fins probably originated as independent structures, and were associated with a Hoxd13 expression pattern comparable with Neoceratodus. This interpretation, apart from being consonant with the observed morphologies, provides the most parsimonious explanation for the developmental similarities between lungfish and tetrapods. A further potential test of the hypothesis is provided by the living coelacanth Latimeria, which has distal polytomous radial fans in its fins and should thus also show distal Hoxd13 expression just before, and associated with the development of the radials in this fan. As noted above, the presence of an expanded, late-phase Hoxd13 expression pattern in the distal regions of the Polyodon pectoral fin (Davis et al., 2007) and

DIGIT HOMOLOGUES IN SARCOPTERYGIAN FISH FINS 765 in the tetrapod autopod also suggests, phylogenetically, that a comparable expression pattern will be present in Latimeria, and indeed, all sarcopterygians. Even without access to embryos it may be possible to determine this indirectly by a detailed investigation of the regulatory regions of Hoxd13 and upstream genes in Latimeria. If our hypothesis is correct, the distal array of independent radials/digits forms a terminal fan in sarcopterygians with a short fin axis (coelacanths, rhizodonts, osteolepiforms, tetrapods) but separate preaxial and postaxial series in forms with a long fin axis (lungfishes, porolepiforms, and to a lesser extent the elpistostegid Tiktaalik). The short-axis condition is probably primitive (Ahlberg, 89). The longer axis of Tiktaalik appears to be autapomorphic for that genus, as no comparable axis extension is known in any other tetrapodomorph fish or tetrapod. The size of the radial array is highly variable: it is large and elaborate in rhizodonts (Davis et al., 2004) and in forms with a long fin axis, but difficult to identify in osteolepiforms such as Eusthenopteron and Gogonasus (Andrews and Westoll, 70; Long et al., 2006) where it may comprise only a few small terminal radials (Fig. 2). Our reason for inferring its presence in osteolepiforms is essentially phylogenetic; welldeveloped distal radial arrays are present in rhizodonts and elpistostegids, which bracket the osteolepiforms phylogenetically, and it seems probable that mechanisms of fin patterning would be conserved among such closely related taxa. DISCUSSION Tetrapod autopod development In most extant tetrapod groups (all amniotes and anuran amphibians), autopod development occurs as the limb bud extends distally and then flattens into a paddle shape (Shubin and Alberch, 86; Franssen et al., 2005). Within this paddle, cartilaginous elements form as a continuous structure, divided into individual elements by joint formation, with all carpal/tarsal elements forming before digits, as indicated by their relatively darker staining with alcian blue (e.g., for anurans, Fabrezi and Barg, 2001; for turtles, Burke and Alberch, 85 and references therein for other tetrapods). Carpal/tarsal elements and digits are connected via prechondrogenic cells, which are more lightly stained and more loosely connected than the condensing elements (Burke and Alberch, 85; Shubin, 91). Cell apoptosis occurs between the digits, resulting in their final separation (Tickle, 96). In the developing urodele limb, by comparison, the paddle-like structure does not form and instead, digits develop as individual buds (Franssen et al., 2005). Notably, in urodeles these digits develop before the condensation of the mesopodium (Hanken, 86; Shubin and Alberch, 86; Blanco and Alberch, 92; Hinchcliffe, 2002; Shubin, 2002; Franssen et al., 2005; Fröbisch et al., 2007). For example, in Desmognathus aenus, metacarpal II is the first element of the autopodium to condense and chondrify, whereas in Ambystoma mexicanum, metacarpal I and digit I are the first elements to condense (Franssen et al., 2005). In D. aenus, metacarpal II is well separated from the only other limb condensations, the radius and ulna, with no apparent connection via prechondrogenic cells (Franssen et al., 2005; Fig. 3C). In Ambystoma, Shubin and Alberch ( 86; p 341) were not able to say from where elements of their digital arch developed. They suggested that connections between this arch and more proximal elements were secondary, as the arch elements developed before the more proximal elements. A comparable secondary connection between Neoceratodus radials and the fin axis, to form joint surfaces, is described above; interestingly, Shubin and Alberch ( 86; p 355) also describe the postaxial radials of Neoceratodus as developing de novo. Given that amniotes are the sister group of urodeles1anurans (e.g., Trueb and Cloutier, 91; Ruta et al., 2003; Brinkmann et al., 2004), the straightforward phylogenetic interpretation of this pattern would be that the shared amnioteanuran condition, proximal-to-distal chondrification within a paddle, is primitive for tetrapods, with the urodele condition being derived (e.g., Wagner and Larsson, 2007). However, the developmental and molecular evidence from Neoceratodus described above, in conjunction with the fossil evidence, suggests otherwise. The earliest phase of tetrapod evolution shows a gradual elaboration of the mesopodium after the establishment of the digital arch (Coates, 2003). Furthermore, this elaboration involves the addition of new centrals and distal carpals/tarsals, elements that lie immediately on either side of the transverse discontinuity line (Fig. 1, dashed line) across which it is impossible to trace one-toone relationships between elements. This suggests that the elaboration may have comprised two mirror-image processes of terminal addition of

766 Z. JOHANSON ET AL. new elements, one at the distal end of the limb proper, producing centrals, and one at the proximal ends of the digits, producing distal carpals/ tarsals (Fig. 1B). The fossil evidence thus implies a degree of digit independence in early tetrapods. We argue that the early establishment of digits during urodele limb development is a retained primitive character that has been lost independently in anurans and amniotes. We note, however, that Cohn et al. (2002) questioned the occurrence of branching mechanisms in the more distal elements of the amniote limb skeleton. A piece of foil placed within a developing chick wing directly in front of the most recently condensing element failed to prevent more distal elements from forming, suggesting some degree of independence of the latter from the former. Independence of the digits may thus be a general feature, masked by the proximal-to-distal chondrification sequence within the amniote paddle. Further evidence of this independence may be seen in patterns of digit ossification in fossil and living tetrapods. As recently reviewed by Fröbisch et al. (2007), the distal parts of digits begin to ossify before more proximal elements, including the mesopodium, in a variety of amniotes including mammals and turtles. Although Fröbisch et al. (2007) were uncertain as to why this deviation from the normal proximal-distal sequence of ossification occurred, it is very similar to the disjunct development seen in urodeles and Neoceratodus. A combination of ossified digits with an unossified carpus (and sometimes tarsus) is also widespread among early tetrapods, occurring in temnospondyls such as Trimerorhachis (Case, 35) and Apateon (Fröbisch et al., 2007), as well as in nectrideans such as Ptyonius (Bossy, 76) and seymouriamorphs such as Discosauriscus (Spinar, 52). Both nectrideans and seymouriamorphs are probable stem amniotes, whereas Trimerorhachis is a stem amphibian (Ruta et al., 2003), suggesting that a propensity for this ossification pattern is primitive for crown-group tetrapods. Apateon, an early Permian branchiosaurid temnospondyl that probably belongs to the urodele stem group, and is represented by numerous larval and juvenile specimens, shows precocious ossification of the digits in what seems to be a fundamentally urodele-like pattern of development (Fröbisch et al., 2007). We suggest that the presence of early digit ossification in crown-group tetrapods represents a heterochronic shift of early digit formation/digit independence, from chondrogenesis in sarcopterygian fishes and urodeles to ossification. These observations imply that the autopod evolved before the origin of tetrapods, represented by the more distal region of the sarcopterygian fin, with the origin of digits simply representing a modest repatterning of this region rather than the origin of a new structure. Patterns of gene expression During tetrapod limb development, Hoxd genes are expressed during an early phase associated with early limb development (stylopod/zeugopod), and a late phase of expression, associated with the autopod. Tarchini and Duboule (2006) recently reviewed the underlying control mechanisms for these early and late phases of Hoxd expression. The early phase is controlled by an enhancer positioned telomeric (3 0 ) to the Hoxd cluster, and a regulatory mechanism located centromeric (5 0 )to the cluster. The late phase is also controlled by two enhancers, both centromeric (5 0 ) to the Hoxd cluster. In the early phase, the telomeric enhancer is responsible for the activation of Hoxd genes from 3 0 to 5 0, whereas the centromeric regulatory mechanism restricts the more 5 0 Hox genes posteriorly by inhibiting the expression of 5 0 Hoxd genes in the anterior part of the limb bud (Tarachini and Duboule, 2006). This posterior restriction is followed by Sonic Hedgehog (Shh) expression within the domain of these genes. Shh expression subsequently modulates, but is not required for, late-phase Hoxd expression, which is controlled by the centromeric enhancers GCR (global control region) and Prox. These enhancers preferentially activate the more 5 0 genes and result in an expansion of these across the limb bud (Deschamps, 2004; Zákány et al., 2004; Tarchini and Duboule, 2006; Gonzalez et al., 2007). Thus, gene expression is inverted in the late phase, in the 5 0 to 3 0 direction. The 5 0 centromeric enhancers associated with the late phase are unrelated to the 3 0 enhancer and 5 0 inhibitor associated with early phase Hoxd expression. Tarchini and Duboule (2006) have proposed that these two phases reflect the different phylogenetic histories of proximal vs. distal limb structures, suggesting that the late phase controlling autopod formation was acquired in the first tetrapods, but was absent from the fish fin (also Gonzalez et al., 2007). This conclusion is supported by the finding that in the zebrafish pectoral fin, where only the early phase of Hoxd expression occurs, the weakest and last gene to be expressed is the most 5 0 in the cluster, Hoxd13 (Sordino et al., 95). This posterior

DIGIT HOMOLOGUES IN SARCOPTERYGIAN FISH FINS 767 expression of Hoxd13 in the zebrafish fin is not followed by the late-phase expansion of Hoxd13 expression across the fin bud. However, the zebrafish is a phylogenetically derived actinopterygian (teleost). Recently published gene expression studies on developing pectoral fins of the more basal actinopterygian fish, Polyodon, found that the late-phase expression of the 5 0 Hoxd genes, which had previously typified only the development of the tetrapod autopod, also occurs in Polyodon fin development (Davis et al., 2007). Intriguingly, it appears that the 5 0 centromeric enhancers directing the late-phase expression pattern of the Hoxd13 gene were lost in teleosts rather than acquired as an evolutionary novelty in tetrapods. It therefore must have been primitively present in Sarcopterygii and the most parsimonious expectation is that we should find it to be present in Neoceratodus. We would argue that the expression pattern of Hoxd13 observed across the Neoceratodus fin and strongly correlated with radial formation is homologous with late-phase expression patterns in tetrapods and Polyodon, where in the latter it also correlates with the condensation of radials (Davis et al., 2007). Early phase expression is either absent (unlikely) or is so far undetected due to technical difficulties with the rather intractable fins of Neoceratodus. Although 5 0 Hoxd genes, other than Hoxd13, remain to be examined in Neoceratodus, it would seem most likely that they too might be expected to display the reversed collinearity found in tetrapods and now the basal actinopterygian, Polyodon. Thus the autopod appears to have had its precursors firmly planted within the paired fins of early osteichthyans. This was also alluded to by Gonzalez et al. (2007), who suggested that, of the two centromeric enhancers, the global control region is the more ancient, and may have been responsible for a broad distal expression pattern of more 5 0 Hoxd genes in early tetrapods and sarcopterygians. The second centromeric enhancer, Prox, may have evolved subsequently. A phylogentically earlier appearance of the tetrapod autopod has also been suggested by Wagner and Larsson (2007), based on a suggested homology between the fin metapterygial axis and digit IV of tetrapods. The metapterygium, they argue, extends into the autopodial field and subsequently becomes modified into a digit, followed by a reiteration of this digit to produce more anterior and posterior digits, and by the evolution of digit identity. We disagree with this specific interpretation of autopodial evolution, because it does not seem to fit well with the palaeontological or molecular data. However, Wagner and Larsson (2007) make the important point that the presence of the autopod does not need to be linked to an autopodial morphology, but only needs to be a developmentally distinct module relative to the more proximal parts of the fin. This interpretation is now reinforced by the important contribution from Davis et al. (2007) on Polyodon. We believe that we too have identified such an independent module in sarcopterygians, represented by the disjunct development of fin radials/digits and patterns of Hoxd13 expression in extant taxa, and by a lack of distal regularity or homogeneity in the fin skeletons of fossil lobe-fins. ACKNOWLEDGMENTS Z.J. and J.J. were supported by the Australian Research Council Discovery grant DP-663874. Z.J. and J.J. thank Debra Birch and Nicole Vella (Microscopy Unit, Macquarie University) for all of their assistance. C.B. gratefully acknowledges Uppsala University for financing her PhD position. P.E.A. and C.B. acknowledge the support of the Swedish Research Council. LITERATURE CITED Ahlberg PE. 1989. Paired fin skeletons and relationships of the fossil group Porolepiformes (Osteichthyes: Sarcopterygii). Zool J Linn Soc 96:119 166. Ahlberg PE, Johanson Z. 1998. Osteolepiforms and the ancestry of tetrapods. Nature 395:792 794. Ahlberg PE, Trewin NH. 1995. The postcranial skeleton of the Middle Devonian lungfish Dipterus valenciennesi. Trans R Soc Edinb 85:159 175. Andrews SM, Westoll TS. 1970. The postcranial skeleton of Eusthenopteron foordi Whiteaves. Trans R Soc Edin 68:207 329. Blanco MJ, Alberch P. 1992. Caenogenesis, developmental variability, and the evolution in the carpus and tarsus of the marbled newt Triturus marmotus. Evolution 46:677 687. Boisvert C. 2006. Appendicular transformations during the fish-tetrapod transition: insights from the transitional form Panderichthys. In: Yang Q, Wang Y, Weldon EA, editors. Ancient life and modern approaches. Beijing: University of Science and Technology of China Press. p 137 138. Bossy KA. 1976. Morphology, paleoecology, and evolutionary relationships of the Pennsylvanian urocordylid nectrideans (Subclass Lepospondyli, Class Amphibia). Ph.D. thesis, Yale University, New Haven, Connecticut. Brinkmann H, Venkatesh B, Brenner S, Meyer A. 2004. Towards a solution of the tetrapod origin question: an approach based on concatenated nuclear-encoded genes. Proc Natl Acad Sci USA 101:4900 4905. Burke AC, Alberch P. 1985. The development and homology of the chelonian carpus. J Morphol 186:119 131.

768 Z. JOHANSON ET AL. Case EC. 1935. Description of a collection of associated skeletons of Trimerorhachis. Univ Mich Contrib Mus Paleo 4:227 274. Clack JA. 2004. From fins to fingers. Nature 304:57 58. Clack JA. 2005. The emergence of early tetrapods. Palaeogeog Palaeoclim, Palaeontol 232:165 169. Coates MI. 1996. The Devonian tetrapod Acanthostega gunnari Jarvik: postcranial anatomy, basal tetrapod interrelationships and patterns of skeletal evolution. Trans R Soc Edinb: Earth Sci 87:363 421. Coates MI. 2003. The evolution of paired fins. Theory Biosci 122:266 287. Coates MI, Clack JA. 1990. Polydactyly in the earliest known tetrapod limbs. Nature 347:66 69. Coates MI, Jeffery JE, Ruta M. 2002. Fins to limbs: what the fossils say. Evol Dev 4:390 401. Cohn MJ, Lovejoy CO, Wolpert L, Coates MI. 2002. Branching, segmentation and the metapterygial axis: pattern versus process in the vertebrate limb. BioEssays 24:460 465. Davis MC, Shubin NH, Daeschler EB. 2004. A new specimen of Sauripterus taylori (Sarcopterygii, Osteichthyes) from the Fammenian Catskill Formation of North America. J Vert Paleont 24:26 40. Davis MC, Dahn RD, Shubin NH. 2007. An autopodial-like pattern of Hox expression in the fins of a basal actinopterygian fish. Nature 447:473 476. Deschamps J. 2004. Hox genes in the limb: a play in two acts. Nature 304:1610 1611. Dinkergus G, Uhler LD. 1977. Enzyme clearing of alcian blue stained whole small vertebrates for demonstration of cartilage. Stain Tech 52:229 232. Dolle P, Dierich A, LeMeur M, Schimmang T, Schuhbaur B, Chambon P, Duboule D. 1993. Disruption of the Hoxd13 gene induces localized heterochrony leading to mice with neotenic limbs. Cell 75:431 441. Fabrezi M, Barg M. 2001. Patterns of carpal development among anuran amphibians. J Morphol 249:210 220. Franssen RA, Marks S, Wake D, Shubin NH. 2005. Limb chondrogenesis of the seepage salamander, Desmognathus aeneus (Amphibia: Plethodontidae). J Morphol 265:87 101. Fröbisch NB, Carroll RC, Schoch RR. 2007. Limb ossification in the Paleozoic branchiosaurid Apateon (Temnospondyli) and the early evolution of preaxial dominance in tetrapod limb development. Evol Dev 9:69 75. Gonzalez F, Duboule D, Spitz F. 2007. Transgenic analysis of Hoxd regulation during digit development. Dev Biol 306:847 859. Hanken J. 1986. Developmental evidence for amphibian origins. Evol Bio 20:389 417. Hinchcliffe JR. 2002. Developmental basis of limb evolution. Int J Dev Biol 46:835 845. Holmes R. 1984. The Carboniferous amphibian Proterogyrinus scheeleri Romer and the early evolution of tetrapods. Philos Trans R Soc Lond B 306:431 527. Johanson Z, Joss JMP, Wood D. 2002. The scapulocoracoid of the Queensland lungfish Neoceratodus forsteri (Dipnoi: Sarcopterygii): morphology, development and evolutionary implications for bony fishes (Osteichthyes). Zoology 107:93 109. Joss JMP, Longhurst T. 2001. Lungfish paired fins. In: Ahlberg PE, editor. Major events in early vertebrate evolution. London: Taylor & Francis. p 370 376. Lebedev OA, Coates MI. 1995. The postcranial skeleton of the Devonian tetrapod Tulerpeton curtum Lebedev. Zool J Linn Soc 114:307 348. Long JA, Young GC, Holland, Senden TJ, Fitzgerald EMG. 2006. An exceptional Devonian fish from Australia sheds light on tetrapod origins. Nature 444:199 202. Millot J, Anthony J. 1958. Anatomie de Latimeria chalumnae I: Squelette et Muscles. Paris: CNRS. Panchen AL, Smithson TR. 1987. Character diagnosis, fossils, and the origin of tetrapods. Biol Rev 62:341 438. Ruta M, Coates MI, Quicke DLJ. 2003. Early tetrapod relationships revisted. Biol Rev 78:251 345. Shubin NH. 1991. The implications of the bauplan for development and evolution of the tetrapod limb. In: Hinchliffe JP, Hurle JM, Summerbell D, editors. Developmental patterning of the vertebrate limb. New York: Plenum Press. p 411 421. Shubin NH. 2002. Origin of evolutionary novelty: examples from limbs. J Morphol 252:15 28. Shubin NH, Alberch P. 1986. A morphogenetic approach to the origin and basic organization of the tetrapod limb. Evol Biol 20:319 387. Shubin NH, Daeschler EB, Jenkins FA. 2006. The pectoral fin of Tiktaalik roseae and the origin of the tetrapod limb. Nature 440:764 771. Sordino P, van der Hoeven F, Duboule D. 1995. Hox gene expression in teleost fins and the origin of digits. Nature 375:678 681. Spinar Z. 1952. Revision of some Moravian Discosauriscidae. Rozpravy Ustredního Ustavu Geologického 15: 1 159. Tarachini B, Duboule D. 2006. Control of Hoxd genes collinearity during early limb development. Dev Cell 10:93 103. Tickle C. 1996. Genetics and limb development. Dev Genet 19:1 8. Trueb L, Cloutier R. 1991. A phylogenetic investigation of the inter- and intrarelationships of the Lissamphibia (Amphibia: Temnospondyli). In: Schultze H-P, Trueb L, editors. Origins of major groups of tetrapods: controversies and consensus. Ithaca, New York: Cornell University Press. p 223 313. Vorobyeva EI. 2000. Morphology of the humerus in the rhipidistian Crossopterygii and the origin of tetrapods. Paleontol J 34:632 641. Wagner GP, Chiu C-H. 2001. The tetrapod limb: a hypothesis on its origin. J Exp Zool 291:226 240. Wagner GP, Larsson HCE. 2007. Fins and limbs in the study of evolutionary novelties. In: Hall BK, editor. Fins into limbs: evolution, development and transformation. Chicago: University of Chicago Press. p 49 64. Wilkinson DG. 1992. In situ hybridization- a practical approach. Cary, NC: Oxford University Press. Zákány J, Kmita M, Duboule D. 2004. A dual role for Hox genes in limb anterior-posterior asymmetry. Science 304: 1669 1672. Zhu M, Yu XB. 2002. A primitive fish close to the common ancestor of tetrapods and lungfish. Nature 418: 767 770. Zhu M, Yu X-B, Ahlberg PE. 2001. A primitive sarcopterygian fish with an eyestalk. Nature 410:81 84.