To investigate jaw evolution in beloniform fishes, we reconstructed the phylogeny of 54 species using fragments of two nuclear (RAG2 and Tmo-4C4) and two mitochondrial (cytochrome b and 16S rRNA) genes. Our total molecular evidence topology refutes the monophyly of needlefishes (Belonidae) and halfbeaks (Hemiramphidae), but supports the monophyly of flyingfishes (Exocoetidae) and sauries (Scomberesocidae). Flyingfishes are nested within halfbeaks, and sauries are nested within needlefishes. Optimization of jaw characters on the tree reveals a diverse array of evolutionary changes in ontogeny. During their development, needlefishes pass through a “halfbeak” stage that closely resembles the adult condition in the hemiramphid halfbeaks. The reconstruction of jaw transitions falsifies the hypothesis that halfbeaks are paedomorphic derivatives of needlefishes. Instead, halfbeaks make up a basal paraphyletic grade within beloniforms, and the needlefish jaw morphology is relatively derived. The parallel between needlefish ontogeny and beloniform phylogeny is discussed, and clades amenable to future morphological analysis are proposed.
Phylogeny is integral to understanding the evolution of ontogeny. Without a firm understanding of a group's evolutionary relationships, the polarities of ontogenetic transformations are irretrievable. In particular, determining the role of heterochrony (changes in developmental timing during evolution) depends on phylogenetic patterns. Gould (1977) emphasized this point, which was further developed by Alberch et al. (1979), and rephrased in a cladistic context by Fink (1982). Two broad heterochronic patterns have been identified, paedomorphosis and peramorphosis. In taxa that exhibit paedomorphosis, descendant adults exhibit morphological features of the juveniles of their putative ancestors. In peramorphosis, descendant taxa extend the ontogeny of ancestors, so that adult features of the ancestor appear in the juveniles of descendants. In both cases, alterations of developmental timing produce parallels between ontogeny and phylogeny. Numerous studies have implicated heterochrony in the evolution of morphology in fishes (e.g.,Bemis, 1984; Winterbottom, 1990; Boughton et al., 1991; Johnson and Brothers, 1993; Zelditch et al., 2000). Gould (1977) points out that the endeavour to assess the relative frequencies of peramorphosis versus paeodomorphosis may be misplaced in a field with nearly limitless empirical potential. However, the examination of specific cases may still yield valuable insight into the relationship between phylogeny, ontogeny and ecology.
The order Beloniformes, a group that currently includes the needlefishes (Belonidae), halfbeaks (Hemiramphidae), flyingfishes (Exocoetidae), sauries (Scomberesocidae) and ricefishes (Adrianichthyidae) (Rosen and Parenti, 1981; Collette et al., 1984) is a good model system for investigating the evolution of ontogeny. Beloniform species exhibit striking variation in jaw structure that varies both ontogenetically and phylogenetically, and appears related to feeding. The most spectacular ontogenetic changes occur in belonid needlefishes. Larval needlefishes have short jaws of equal length. However, in juveniles, the lower jaw elongates first, producing a morphology that is distinctly reminiscent of a related family, the halfbeaks (Hemiramphidae)—indeed, needlefishes in this juvenile “halfbeak” form have been mistakenly described as new species of hemiramphids (Collette and Parin, 1970). Later, the upper jaw elongates as well, giving rise to the nearly equal length jaws of most adult needlefishes. These transformations appear linked to ecological shifts: juvenile needlefishes in the “halfbeak” morphology feed primarily on plankton, while most adult needlefishes are piscivorous (Boughton et al., 1991).
The similarity in jaw morphology between juvenile needlefishes and the closely related Hemiramphidae has provoked alternative interpretations. Severtsov (1927; summarized in Gould, 1977) thought that the ontogeny of needlefishes paralleled the phylogeny of beloniforms. He hypothesized that short-jawed ancestral flyingfishes gave rise to descendant halfbeaks, which in turn gave rise to the more advanced needlefishes. Needlefish ontogeny would thus be an example of the phenomenon of recapitulation or peramorphosis. Nichols and Breder (1928), on the other hand, building on the work of Schlesinger (1909) and Regan (1911) proposed an alternative beloniform phylogeny. In their scheme, needlefishes are the most basal family, and gave rise to hemiramphids, which in turn gave rise to flyingfishes. They hypothesized that hemiramphids are “fixed larval” (or paedomorphic) needlefishes (Nichols and Breder, 1928, p. 435). de Beer (1940) reported this finding in his book “Embryos and Ancestors” which emphasized the importance of paedomorphosis as an evolutionary pattern.
Clearly, differentiating between the paedomorphosis and recapitulation scenarios for beloniforms is only possible with a robust phylogenetic hypothesis for the group. Lovejoy (2000) presented a phylogenetic analysis based on 2 mitochondrial genes, a nuclear gene, and morphology for representatives of all 5 beloniform families. The result falsified Nichols and Breder's (1928) paedomorphic hemiramphid origin hypothesis, which predicted a basal position for needlefishes. Instead, hemiramphids were found to represent a basal paraphyletic grade, with needlefishes and flyingfishes relatively derived. The phylogenetic position of needlefishes therefore matched the prediction of Severtsov's (1927) recapitulation hypothesis.
Here, we present an extension and test of Lovejoy's (2000) findings. We have expanded the matrix by adding 14 ingroup taxa to the previous 39, which significantly improves taxonomic coverage of hemiramphids and flyingfishes. We have also added a novel source of character data, by sequencing a 1 kilobase fragment of the nuclear gene, Recombination Activating Gene 2, or RAG2. Our analysis is based on the nuclear RAG2 and Tmo-4C4 genes, and the mitochondrial cytochrome-b and 16S rRNA genes. We use the resultant phylogenetic hypothesis to explore the evolution of jaw ontogeny.
Fishes were collected in the field by ourselves or colleagues. Gill tissue was either frozen immediately in liquid nitrogen or preserved in 95–100% ethanol or buffer of 20% DMSO, 0.25 M EDTA at pH 8, saturated with NaCl (Seutin et al., 1991). Tissue preserved in buffer and stored at room temperature reliably yields amplifiable DNA (after storage for up to four years). Voucher specimens were preserved in 10% buffered formalin, transferred to 70% ethanol or 50– 55% isopropanol and deposited in museum collections (catalogue numbers for voucher specimens are listed with sequences in GenBank).
Samples represent all beloniform families, and with the addition of new taxa, 30 of 39 beloniform genera are represented in the study: ten of ten needlefish genera, three of four saury genera, nine of thirteen halfbeak genera, seven of seven flyingfish genera, and one of four ricefish genera. Whenever possible, sequences were collected from two individuals of each species, providing a total of 104 terminal taxa for analysis, representing 54 different beloniform species.
Both mitochondrial and nuclear genes were used for analysis. However, rather than sequencing a single complete mitochondrial gene, smaller segments of two separate genes, cytochrome b (cyt b) and 16S rRNA (16S), were examined. This decision was based on the expectation that sampling a range of genes, with different rates and patterns of molecular evolution, would provide phylogenetic information that spanned a broader range of taxonomic divergence. The nuclear gene, Tmo-4C4 (Tmo) is an anonymous, putative protein-coding locus identified and used for phylogeny by Streelman and Karl (1997). It provided resolution of families and genera within labroids, and was thus expected to provide useful information for deeper parts of the beloniform tree. Recombination Activating Gene 2 (RAG2) is a nuclear gene that appears to evolve slightly faster than Tmo-4C4 (Lovejoy and Collette, 2001), and has proven useful for species-level phylogenetic analyses (Sullivan et al., 2000; Lovejoy and Collette, 2001).
For each sample, approximately 25 mg of tissue was rinsed briefly in water, then DNA purified using Qiagen's spin-column tissue kit. Cells were lysed at 55° in 20 μl of Proteinase K (20 mg/ml) for three to six hours. Lysate was bound to the spin column membrane, and washed twice by centrifugation. DNA was then eluted by centrifugation twice with 200 μl of low salt buffer.
In the case of cyt b, 16S, and Tmo, template for sequencing was initially amplified using published PCR primers. For RAG2, primers were developed by comparing available vertebrate sequences. New primers were then designed for sequencing and additional amplifications (for details and primer lists, see Lovejoy, 2000 and Lovejoy and Collette, 2001). Generally, DNA was amplified in 50 μl reactions containing 1 μl of DNA, 3 mM MgCl2, 20 mM Tris HCl pH 8.4, 50 mM KCl, 200 μM dNTPs, 0.4 μM of each primer, and one unit of Gibco Taq polymerase. PCR amplifications were performed using the following conditions: 30 second denaturation at 95° to start, followed by 35 cycles of denaturation at 95° for 30 seconds, annealing at 48°–55° for 60 seconds, and 72° extension for 90 seconds, followed by a final extension of 72° for 5 minutes. In cases where faint secondary bands were detected, template was run in 1% agarose gels, then cut out and cleaned using PCR purification spin columns (Qiagen).
PCR products were cleaned using PCR product pre-sequencing purification kits (Qiagen or Amersham Life Science) and then sequenced using a variety of methods. In some cases, we manually sequenced using the Thermo Sequenase radiolabeled terminator cycle sequencing kit (Amersham Life Science). Other samples were cycle sequenced and run on an ABI 377 automated sequencer according to manufacturer specifications (Applied Biosystems, Inc., Foster City, CA). Some samples were sequenced at the University of Calgary Core DNA and Protein Services using an ABI 377 automated sequencer.
The Tmo, cyt b, and RAG2 sequences were aligned unambiguously using Lasergene (DNASTAR, Madison, WI) or Sequencher software (Genecodes, Ann Arbor, MI). As in Streelman and Karl's (1997) study, an open reading frame for Tmo was determined that produced amino acid translations with no stop codons in any sequences. Tmo is therefore considered a protein-coding gene, with positions determined by the hypothesized open reading frame. The 16S sequences were imported in ClustalX (Thompson et al., 1997) and an initial multiple alignment was conducted using the parameters: gap cost = 10, gap length = 10, DNA transition weight = 1. This alignment was used to generate a neighbour-joining distance tree (using K2P distances) using PAUP* (version 4.0b10, Swofford, 2002). The distance tree was used as a “guide tree” for subsequent multiple alignments in which gap cost and gap length were varied (from 1 to 20). Results were compared to a hypothesized model for secondary structure proposed by Orti et al. (1996) for piranhas, and alignments that inserted gaps in stable regions (stems) were excluded from further consideration. Of the remaining alignments (considered more biologically reasonable), sites and regions where alignment was ambiguous were removed. This rather conservative procedure excluded approximately 100 positions, leaving only regions of 16S that were conserved over the full range of taxa. The sacrifice of potential characters for clear topographical identity of sites (as per Brower and Schawaroch, 1996) was considered acceptable, since preliminary analysis of the mitochondrial protein-coding genes indicated that many characters for resolving recent nodes were available, while conservative characters for deeper parts of the tree would be at a premium. Additional analyses, incorporating the deleted characters, do not alter the findings reported here. New sequences determined for this study have been deposited in Genbank; see Appendix 1 for list of accession numbers.
All 2,516 characters (965 bp RAG2, 497 bp Tmo, 636 bp cyt b, and 415 bp 16S) were combined in a single matrix and the heuristic search algorithm of PAUP* (100 replicates of random taxon additions, TBR branch swapping) was used to search for most parsimonious trees. All characters were uniformly weighted. Oryzias (the ricefish representative) was used as an outgroup to root all trees. Nuclear, and mitochondrial datasets were also analyzed separately (using the same algorithm and settings) to investigate the contribution provided by each to the total molecular evidence hypothesis. For RAG2, sequence from the outgroup Oryzias was unobtainable due to PCR difficulties. To test the rooting of the tree, we conducted a separate RAG2 analysis with additional outgroups from Genbank, including Danio (NM_131385), Takifugu (AF108420), Campylomormyrus (AF201622), Gnathonemus (AF201628), Chitala (AF201626), and Gymnarchus (AF201629). To test the root of the Tmo tree, we included Genbank sequences for labroids deposited by Streelman and Karl (1997). The two mitochondrial genes were analyzed together because it was assumed that the small size of each fragment would prevent effective phylogeny reconstruction. Decay indices for nodes were calculated using TreeRot (Sorenson, 1996), and bootstrap proportions were calculated using PAUP* (100 replicates with 50 random taxon additions per replicate).
The evolution of jaw characters was examined by optimizing juvenile and adult conditions on the total molecular evidence tree using MacClade (Maddison and Maddison, 1992). Jaw states were derived from the literature and from personal observations of specimens by the authors. We examined the effects of slight changes in tree topology by optimizing characters on alternative topologies.
Figure 1 shows the single most parsimonious tree derived from an unweighted parsimony analysis of the total molecular dataset. The tree is 5,670 steps long, with a consistency index (excluding uninformative characters) of 0.27, and a retention index of 0.58. In most respects, the tree is similar to the total evidence topology presented in Lovejoy (2000). Of the five currently recognized beloniform families, only two are monophyletic: the flyingfishes (Exocoetidae) and sauries (Scomberesocidae), whereas the needlefishes (Belonidae) and halfbeaks (Hemiramphidae) are paraphyletic. Only a single ricefish (Adrianichthyidae) was included as an outgroup, prohibiting tests of adrianichthyid monophyly.
The monophyletic sauries, which include the genera Scomberesox, Cololabis, and Ellasichthys, are deeply nested within needlefishes. The clade including sauries and the needlefish genera Belone and Petalichthys appears well-supported by decay indices and bootstrap scores.
The relationships of halfbeaks are more complex, with the included genera forming three clades. Zenarchopterus, Hemirhamphodon, Nomorhamphus, and Dermogenys, which all practice internal fertilization and are distributed in freshwater and estuaries of the Indo-West Pacific, comprise a monophyletic group that is the sister clade to needlefishes/sauries. These halfbeak genera have been recognized as a separate family, Zenarchopteridae, based on evidence from the pharyngeal jaw apparatus (Tibbetts, 1992) and sperm ultrastructure (Jamieson and Grier, 1993). Meisner (2001) provides further anatomical evidence for the monophyly of this clade, with the addition of Tondanichthys (not included here). Based on the additional support of molecular data, we hereafter use the name Zenarchopteridae for this monophyletic group of halfbeaks.
The marine halfbeak genera Hemiramphus, Oxyporhamphus, and Euleptorhamphus compose a clade (hereafter referred to as the Hemiramphus clade) that is the sister to flyingfishes. Finally, the marine halfbeak genera Hyporhamphus and Arrhamphus make up a small clade that is the sister to needlefishes/sauries and Zenarchopteridae. The position of the Hyporhamphus/ Arrhamphus clade is not well-supported; in trees that are two steps longer, the clade may be grouped with the flyingfishes and other marine halfbeaks. However, the position of the Zenarchopteridae is well supported, as is the position of the marine Hemiramphus clade as the sister group to flyingfishes.
Separate analyses of the nuclear genes are largely congruent with the total molecular evidence hypothesis. Two analyses were conducted with RAG2. In the first, only the taxa from the total molecular evidence dataset were included (Fig. 2A). In the second, additional outgroups were used to root the tree. In this analysis (Fig. 2B), the Zenarchopteridae are grouped with the needlefish/saury clade; the Hemiramphus clade is grouped with the flyingfishes; and Hyporhamphus is the sister to the Hemiramphus clade and flyingfishes.
The two analyses of Tmo were stopped before completion, owing to the large number of equally parsimonious trees (>60,000). This is most likely a result of the small size of the Tmo dataset relative to the number of OTUs analyzed (>100). Strict consensus trees of the results (Fig. 2C,D) showed congruence with RAG2 and with the total molecular evidence hypothesis. These trees were similar to those with smaller numbers of taxa that were presented in Lovejoy (2000).
Separate analysis of the mtDNA genes resulted in four equally parsimonious trees. The consensus (Fig. 2E) disagrees in many ways with the total molecular evidence and nuclear gene trees. Although some of the more recent relationships seen in the total molecular evidence tree are reiterated (such as the close relationship between the saury genera and Petalichthys/Belone), many of the deeper parts of the tree are fundamentally different. For example, in the mtDNA trees, the freshwater South American Potamorrhaphis/Belonion clade (not shown) is the basal beloniform lineage, while in the total molecular evidence and nuclear trees, it is deeply nested in the tree. Lovejoy (2000) suggested that these differences may be the result of the high levels of homoplasy in cyt b 3rd codon positions. To examine this possibility, cyt b 3rd codon positions were downweighted 1/5, and 1/10th relative to other changes in additional mtDNA analyses. These alternative weighting schemes resulted in topologies that were more similar to the total molecular evidence and nuclear trees (Fig. 2F).
The general pattern of relationships indicated by the total molecular evidence analysis places halfbeaks as a series of paraphyletic lineages originating near the base of the tree. Needlefishes and sauries are nested within these halfbeak clades, as are flyingfishes. The taxonomic distribution and optimization of jaw characters on this topology is shown in Figure 3. The basal condition within beloniforms to the exclusion of ricefishes is the presence of an extended lower jaw in juveniles and in adults. The extended upper jaw appears in the needlefish/saury clade, and in some taxa (Xenentodon cancila, and Tylosurus crocodilus) the extended upper jaw is also present in juveniles. In flyingfishes, and in some halfbeaks such as Arrhamphus (and other genera not included in this study, e.g.,Chriodorus and Melapedalion), the extended lower jaw is lost in adults. In most flyingfish genera, the extended lower jaw in juveniles is also lost.
The expansion of the Lovejoy (2000) matrix by 1 kilobase of nuclear gene sequence and 14 new taxa confirms the main conclusions of that study regarding beloniform relationships. The family Belonidae is not monophyletic without the inclusion of sauries, traditionally regarded as a distinct family (Scomberesocidae). Halfbeaks are also non-monophyletic; however, salvaging the family Hemiramphidae is more problematic. Halfbeaks are divided into three main clades: Zenarchopteridae (the Indo-West Pacific freshwater halfbeaks), the Hemiramphus clade (which includes Euleptorhamphus and Oxyporhamphus), and the Hyporhamphus/Arrhamphus clade. Perhaps the most surprising result is the sister group relationship between Zenarchopteridae and needlefishes/sauries. This relationship is supported by high decay indices and bootstrap proportions, and by its presence in separate analyses of the nuclear genes. Similarly, the sister group relationship between the Hemiramphus clade and flyingfishes is quite strong. The position of the Hyporhamphus clade is less clear. In the most parsimonious tree, it is placed as the sister group to the zenarchopterids and needlefishes/sauries; however, in trees two steps longer (and in separate analyses of the nuclear genes), it is grouped with the Hemiramphus clade and flyingfishes. Increased taxonomic sampling of halfbeaks may resolve this issue; Hyporhamphus, the most taxonomically diverse halfbeak genus, with two distinct subgenera, is represented by only two species in this study.
The hypothesized division of halfbeaks into three groups has implications for patterns of morphological evolution. The fusion of the third pair of upper pharyngeal bones has been considered a synapomorphy of Hemiramphidae (Collette et al., 1984). In light of the topology presented here, these bones either fuse independently in two or more halfbeak lineages, or become unfused in needlefishes and flyingfishes Similarly, homoplasy is required in the evolution of the fourth upper pharyngeal toothplates. The plates are absent in halfbeaks and flyingfishes (Rosen and Parenti, 1981; Collette et al., 1984), thus the present topology requires their independent loss in the three halfbeak/ flyingfish lineages, or their reappearance in the needlefish/saury lineage. Before more definitive hypotheses concerning the evolution of the characters can be made, a complete anatomical survey needs to be undertaken for beloniforms, and incorporated into the matrix.
The inclusion of additional halfbeak and flyingfish taxa provides some interesting results concerning the evolution of gliding. Parin (1961), Collette et al., (1984), Dasilao et al. (1997), and Dasilao and Sasaki (1998) have presented phylogenetic hypotheses for flyingfishes. The results presented here are largely in agreement with these previous studies. Fodiator is generally considered the least sophisticated glider, along with Parexocoetus, and Exocoetus. These three genera are “monoplane” gliders, having greatly expanded pectoral fins but not pelvics. Cypselurus, Prognichthys, and Hirundichthys, and Cheilopogon (the Cypselurinae) are “biplane” gliders, and have expanded pelvic as well as pectoral fins, which help to control gliding stability (Davenport, 1992). Morphological studies place the monoplane gliders at the base of the exocoetid tree, suggesting a progressive refinement of gliding ability. The molecular data support this idea. Dasilao et al. (1997) also showed that, based on morphology, the halfbeak Oxyporhamphus, should be considered the basal flyingfish taxon. The molecular data presented here strongly disagree with this hypothesis, placing Oxyporhamphus deeply within the Hemiramphus clade. Several of the characters used by Dasilao et al. (1997) are related to gliding (in particular, a strengthened caudal complex for take-offs), and might be convergent features of gliding behavior.
Evolution of jaw ontogeny
The optimization of jaw characters on the total molecular evidence beloniform tree clearly falsifies the scenario proposed by Nichols and Breder (1928) and supported by de Beer (1940). These authors hypothesized that needlefishes are the basal members of Beloniformes, and that halfbeaks represent a more derived paedomorphic lineage. Instead, our hypothesis suggests that halfbeaks are relatively basal members of Beloniformes, and the needlefish morphology is relatively derived. The presence of elongate lower jaws in both juveniles and adults is relatively plesiomorphic, whereas the evolution of an elongate upper jaw in adults is a relatively derived condition. Collette et al. (1984) also supported this scenario, based on a phylogeny in which each of the traditional beloniform families was monophyletic. In the topology presented here, the basal paraphyletic position of the halfbeaks provides additional confidence in the optimization of the elongate lower jaw in juvenile and adult as the relatively plesiomorphic states.
The increased number of taxa and the more detailed presentation of jaw condition (Fig. 2) compared to Lovejoy (2000), makes possible a more complete compilation of evolutionary changes in beloniform jaw ontogeny. Figure 4 is a schematic summary of these transitions. The story of beloniform jaws is clearly not limited to the simple transition between the “half-beaked” and “needle-jawed.” The hypothesized patterns of transformation are complex. The generalized halfbeak ontogeny leads independently to ontogenies in which adults lose the elongate lower jaw (e.g.,Oxyporhamphus, Arrhamphus). In most flyingfishes, the elongate lower jaw is absent in juveniles. The generalized needlefish ontogeny leads to ontogenies in which juveniles have elongate upper and lower jaws (e.g.,Tylosurus crocodilus), and others in which the elongate upper jaw is absent in adults (Belonion). In sauries the transitions are particularly striking. Cololabis and Elassichthys have short upper and lower jaws as adults and juveniles, while Scomberesox has elongate upper and lower jaws in adults (Collette et al., 1984). Thus, the placement and relationships of sauries requires the loss of elongate upper and lower jaws, followed by their reappearance in Scomberesox.
What drives the evolution of jaw ontogeny? Boughton et al. (1991) summarized evidence for the importance of jaw shape for the feeding ecology of needlefishes. Needlefishes in the “halfbeak” stage often consume prey such as shrimp, mysids, amphipods, and phytoplankton, but after developing the elongate upper jaw, shift to a fish diet. The forceps-like adult needlefish morphology is well-suited to piscivory. Taxa such as Tylosurus crocodilus and Xenentodon cancila, which have no “halfbeak” juvenile stage, feed almost exclusively on fishes (Breder, 1932; Foster, 1974). Thus, it seems likely that changes in jaw ontogeny may be at least partially related to diet and food availability. Our phylogenetic approach suggests an additional consideration. While most needlefishes are marine, there are a number of freshwater taxa, including several freshwater endemic genera: Xenentodon from Southeast Asia, and Pseudotylosurus, Potamorrhaphis, and Belonion from South America. Most of these taxa display notable changes in ontogenetic patterns; however, not all are in the same direction. Xenentodon, and Potamorrhaphis have lost the “halfbeak” stage as juveniles, while Belonion maintains the halfbeak stage as an adult. The concentration of evolutionary change in ontogeny along freshwater lineages may be related to high diversity of food types in tropical freshwater habitats. For example, in South America, Belonion feeds mainly on zooplankton, insect larvae, and bryozoans, Potamorrhaphis on terrestrial insects (particularly ants), and Pseudotylosurus on fishes (Goulding and Carvalho, 1984). Specialization on alternative food resources might drive evolutionary shifts in jaw ontogeny and morphology. Several species of Strongylura are also freshwater inhabitants (S. krefftii, S. hubbsi, and S. fluviatilis), but data for the juvenile condition are not available for these taxa. The relationship between ecology and jaw ontogeny will remain speculative until more detailed studies of diet, broader surveys of ontogeny, and functional investigations of morphology are completed.
What role has heterochrony played in the evolution of beloniform jaw ontogeny? Severtsov (1927; summarized in Gould, 1977) proposed that needlefish phylogeny recapitulated beloniform ontogeny. The phylogenetic position of needlefishes relative to halfbeaks fits his prediction. The juvenile jaw morphology of the relatively derived needlefish lineage is similar to the adult morphology of the more basal halfbeak lineages. Thus, the ontogenetic sequence in needlefish of short-jawed larva to “halfbeaked” juvenile to “needle-jaw” adult parallels the phylogenetic transition of short-jawed outgroup to hemiramphid halfbeak to belonid needlefish. However, Lovejoy (2000) pointed out that von Baer's (1853) critique of recapitulation may apply—the similarities between needlefish juveniles and halfbeak adults may simply be a byproduct of limited divergence of the halfbeak adult from the halfbeak juvenile. In other words: (1) needlefish juveniles look like halfbeak juveniles due to shared ancestry, (2) halfbeak jaw structure does not change dramatically during the transition from juvenile to adult, thus (3) needlefish juveniles look like halfbeak adults. This “von Baerian” explanation does not require heterochrony (Gould, 1977).
Unfortunately, the discrete character-based approach used here does not allow differentiation between “von Baerian” and true recapitulation caused by changes in developmental timing. One alternative would be to compare ontogenetic trajectories, as did Boughton et al. (1991), who optimized upper and lower jaw length trajectories on a needlefish phylogeny. However, Zelditch et al. (2000, p. 1363) pointed out that one-dimensional features such as jaw length “…necessarily both develop and evolve along the same dimension so they always suggest parallelism between ontogeny and phylogeny.” These authors advocate a morphometric shape-based approach, which permits a comparative assessment of the roles of heterochrony and heterotopy (evolutionary change in spatial patterning of development) (Wray and McClay, 1989; Zelditch et al., 2000).
Beloniform fishes offer rich possibilities for such investigations. Our phylogeny highlights some limitations and opportunities. The transition to “needle-jaw” adults (transition E in Figs. 3 and 4) is deep within beloniform phylogeny. Reconstruction of hypothetical ancestral ontogenetic trajectories at this node will require extrapolation from large numbers of needlefish and zenarchopterid halfbeak species, and would be relatively speculative. However, a number of ontogenetic transitions are shallow in the tree. In particular, Belonion has lost the elongate upper jaw in adults, matures at a very small size, and is hypothesized to be paedomorphic (Collette, 1966). Detailed developmental analysis of this taxon and its sister clade Potamorrhaphis could be extremely informative. Similarly, analyses of Tylosurus and Xenentodon will provide data on the elimination of the juvenile “halfbeak” stage. We view our beloniform phylogeny as a valuable roadmap for further forays into the evolution of ontogeny in this model group of fishes.
From the Symposium Patterns and Processes in the Evolution of Fishes presented at the Annual Meeting of the Society for Integrative and Comparative Biology, 4–8 January 2003, at Toronto, Canada.
We thank the collectors of beloniforms and field assistants who made this study possible, many of whom are listed in Lovejoy (2000). In addition, we gratefully acknowledge the assistance of P. Heemstra, J. Fu, A. Summers, M. Littman, M. Sabaj, J. Albert, L. Ballance, and W. Crampton. R. Pitman generously provided nearly all of the flyingfish samples. L. Parenti, G.D. Johnson, and an anonymous reviewer provided valuable comments on the manuscript. We thank F. Santini and G. Ybazeta for organizing an excellent symposium. Support from NSERC (National Science and Engineering Research Council) and CFI (Canada Foundation for Innovation) made the study possible.