Perceived dissimilarities.

Apparent dissimilarities cited in the literature include (i) the shape and orientation of the postglenoid process of the squamosal, described as twisted and vertically oriented in Herpetocetus, but transversely oriented and posteriorly reclined in Caperea [12]; (ii) the size and shape of the pterygoid hamulus, thought to be broadly triangular in Herpetocetus but small and almost indistinct in Caperea and balaenids [12]; (iii) the size of the pterygoid exposure on the ventral side of the skull, which is relatively small in cetotheriids, but large in Caperea [12]; (iv) the attachment of the anterior process of the periotic to the body of the periotic, which is strong in cetotheriids but tenuous in Caperea [12]; (v) the shape and location of the lateral tuberosity of the periotic, thought to be small, shelf-like and variably positioned in Herpetocetus, but massive and projecting far anteriorly in Caperea [12]; (vi) the shape of the anterior process of the periotic in medial view, described as polymorphic in Herpetocetus, but L-shaped in Caperea [12]; and (vii) the shape of the ascending process of the maxilla, described as short and parallel-sided in Caperea, Balaena and Balaenella, but not cetotheriids [11]. Below, each of these points is discussed in turn.

1. (i). The postglenoid process of Caperea is more transversely oriented in ventral view than that of Herpetocetus, but it is not perpendicular to the sagittal plane. Instead, adult individuals of Caperea consistently show a slight twisting of the postglenoid process in ventral view: clockwise on the left, anticlockwise on the right [4]. This twisting also occurs in somewhat more pronounced form in Miocaperea, where it is clearly evident despite substantial erosion of both postglenoid processes (Fig 9). Though still not as marked as in Herpetocetus, the orientation of the postglenoid process in Miocaperea is therefore consistent with a twisted ancestral postglenoid morphology. We agree that the articular surface in Caperea is more inclined than in Herpetocetus in lateral view, but note that the degree of inclination is relatively slight and exaggerated by the natural anterior slant of the Caperea skull (as judged from the orientation of the orbit) when resting on a horizontal surface. This is in stark contrast to balaenids, in which the articular surface is markedly more horizontal in lateral view. Erosion of the articular surface and uncertainty about the in vivo orientation of the skull in Miocaperea (owing to the apparent posterior orientation of the orbits in lateral view) currently hinder a confident assessment of the slope of the postglenoid process in this species.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Orientation of the postglenoid process in neobalaenines.

(a) Caperea marginata, NMNZ MM002235; (b) Miocaperea pulchra, SMNS 46978, both in posteroventral view.

https://doi.org/10.1371/journal.pone.0164059.g009

1. (ii). Caperea and balaenids strikingly differ from most other mysticetes in the shape of the pterygoid hamulus. Thus, instead of being finger-like, the hamulus of Caperea is developed as a small, triangular, somewhat hook-shaped projection (Fig 10); by contrast, that of balaenids is expanded laterally into a broad, robust horizontal blade [24]. Given these differences, we disagree that the condition of the hamulus in Caperea and balaenids represents a shared state, but note that Miocaperea does not preserve the hamuli, thus leaving open the question of character homology. It is interesting to note that Herpetocetus also shows a relatively unusual morphology of the hamulus, with the latter being triangular and somewhat more confluent with the remainder of the pterygoid than in other mysticetes [12]. This loss of “distinctiveness” could plausibly be interpreted as a first step towards a state similar to Caperea.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 10. Position and shape of the pterygoid hamulus of Caperea marginata.

(a) Basicranium of NMNZ MM002235 in posteroventral view; (b) pterygoid hamulus of the same specimen, in ventral view.

https://doi.org/10.1371/journal.pone.0164059.g010

1. (iii). Caperea is unusual in having an extremely large ventral exposure of the pterygoid, with the latter–uniquely among mysticetes–entirely surrounding the foramen pseudovale [19]. In this arrangement, Caperea differs from all other described mysticetes (including Miocaperea), in which the foramen pseudovale instead appears to be at least partially surrounded by the squamosal [19]. The extremely enlarged exposure of the pterygoid in Caperea therefore represents a phylogenetically uninformative autapomorphy.
2. (iv). Like the ventral exposure of the pterygoid, the narrow connection between the anterior process and the body of the periotic in Caperea [12] is autapomorphic, and thus phylogenetically uninformative. Miocaperea instead shows the widespread plesiomorphic condition of a solidly attached anterior process (Fig 1).
3. (v). The position of the lateral tuberosity in cetotheriids is variable, both inter- and intraspecifically, which has led to the phylogenetic value of this feature being questioned [12]. This is particularly true for Herpetocetus, in which the lateral tuberosity can occur posterolateral, lateral or anterolateral to the anterior pedicle of the tympanic bulla, depending on the specimen and species. Nevertheless, it appears that one of the extremes of this continuum–the plesiomorphic position of the tuberosity posterolateral to the anterior pedicle–only occurs in extremely juvenile individuals (e.g. SDNHM 38689), suggesting that change in position may be ontogenetic [12]. Most Herpetocetus specimens instead have a tuberosity that is located lateral (e.g. H. morrowi: SDNHM 34155; H. transatlanticus: USNM 182962) or anterolateral (e.g. H. bramblei: UCMP 82465; H. morrowi: SDNHM 63690; Herpetocetus sp.: NMNS PV19540) to the anterior pedicle (Fig 11). In Caperea, the lateral tuberosity is developed as a broad, relatively massive shelf extending all along the anterior process, well past the level of the anterior pedicle [4]. By contrast, Miocaperea preserves a more Herpetocetus-like state, in which the lateral tuberosity is located further posteriorly and approximately lateral to the anterior pedicle (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 11. Periotic of Herpetocetus spp. in medial view.

(a) Herpetocetus morrowi, SDNHM 63690; (b) H. bramblei, UCMP 82465; (c) H. transatlanticus, USNM 182962.

https://doi.org/10.1371/journal.pone.0164059.g011

The question of whether the shape of the lateral tuberosity is comparable between neobalaenines and Herpetocetus is more problematic. The lateral tuberosity of Herpetocetus is relatively variable in both size and shape, ranging from small and triangular (e.g. H. transatlanticus: USNM 182962) to proportionally large and rounded (e.g. H. bramblei: UCMP 82465; Herpetocetus sp.: NMNS PV19540). In all cases, however, the tuberosity is bent laterally, articulates with the adjacent squamosal and, in general, is oriented anterolaterally in ventral view, relative to the long axis of the anterior process (Fig 11). The lateral tuberosity of Caperea provides the closest match for that of Herpetocetus in also being oriented anterolaterally and in articulating with the squamosal [4], yet at the same time clearly differs in being considerably more massive. Miocaperea currently offers little to clarify this situation–partly, because the anterior face of the lateral tuberosity remains covered in matrix, and partly because the rim of the squamosal surrounding the anterior process is crushed.

1. (vi). The shape of the anterior process is demonstrably variable among cetotheriids: two-bladed, or L-shaped, in Kurdalagonus mchedlidzei, Herpetocetus transatlanticus, H. bramblei and, possibly, Brandtocetus chongulek (Fig 11); and triangular or squared in Piscobalaena nana, H. morrowi and Metopocetus durinasus. In Caperea, the anterior process is also L-shaped [4]. In Miocaperea, the outline of the anterior process is partially obscured by the lateral lamina of the pterygoid; it appears, however, that the anterior border of the process, if not L-shaped, is at least concave (Fig 1). Such polymorphism can be a challenge to phylogenetics, but does not in itself invalidate a particular character. In this case, the outline of the anterior process is admittedly not clear-cut, but we note that an irregularly-shaped anterior process is relatively rare among mysticetes, and currently confined to cetotheriids, neobalaenines and eomysticetids [6, 30].

One previous study questioned the homology of the L-shaped anterior process in Caperea and cetotheriids, pointing out that the irregular anterior border in Caperea is formed entirely within the process, whereas in H. transatlanticus it arises from an interaction of the anterior process and the pars cochlearis [12]. This distinction is, however, somewhat arbitrary, and in itself variable. Thus, the L-shape is clearly formed entirely within the anterior process in K. mchedlidzei (NMRA 10476/1), B. chongulek (TNU skull 2), H. bramblei (UCMP 82465; Fig 11) and a periotic of Herpetocetus sp. from the Lee Creek Mine, North Carolina, USA (USNM 360765). The condition in the holotypes of H. transatlanticus (USNM 182962) and Miocaperea is less clear, but largely depends on where one draws the line between the pars cochlearis and the anterior process. In our view, it is entirely reasonable to argue that the L-shape, or concavity, in both is also entirely formed by the anterior process.

1. (vii). The shape of the ascending process of the maxilla in adult Caperea and balaenids is difficult to determine because its size is restricted by the anterior telescoping of the supraoccipital (see below). Nevertheless, a clearly parallel-sided ascending process is evident in several neonate (NMNZ MM002262, MM002898) and juvenile Caperea (NMNZ MM002254; Fig 12), as well as Miocaperea, which–irrespective of the length of the process–therefore seem to share this condition with most cetotheriids and balaenopterids [29]. A parallel-sided ascending process of the maxilla also occurs in some individuals of Balaena mysticetus [11] (e.g. LACM 54464), but in other specimens, including neonates (e.g. LACM 54485), it is much more triangular (Fig 12). Other balaenids, including Balaenula astensis (MSNTUP I12555), neonate specimens of Eubalaena spp. (e.g. CNPMAMM 746 and LACM 54763), Morenocetus parvus [29] and, contrary to previous claims [11], Balaenella brachyrhynus (NMB 42001), also have a triangular or rounded ascending process with posteriorly convergent medial and lateral borders, and thus clearly differ from balaenopterids, most cetotheriids and neobalaenines in this regard (Fig 12).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 12. Ascending process of the maxilla in balaenids and Caperea.

(a) Balaenella brachyrhynus, NMB 42001, in anterolateral view; (b) neonate of Balaena mysticetus, LACM 54485, in anterolateral view; (c) Balaenula astensis, MSNTUP I12555, in lateral view; (d) juvenile of Caperea marginata, NMNZ MM002254, in anterolateral view. Note the missing nasals of B. brachyrhynus, as well as the relatively obtuse posterior angle defining the ascending process of balaenids.

https://doi.org/10.1371/journal.pone.0164059.g012

Proposed balaenoid synapomorphies.

In terms of proposed similarities between Caperea and balaenids, one recent study listed 15 morphological features which it argued were shared by both, and concluded that the weight of the available evidence therefore spoke against a relationship of Caperea with cetotheriids [11]. Specifically, these features included:

“(1) massive elongation of supraoccipital; (2) supraoccipital covering the parietal and excluding the parietal to be exposed in dorsal view; (3) anterior end of supraoccipital covering the posterior portion of the interorbital region of the frontal; (4) parietal not extending anteriorly to the posteromedial elements of the rostrum; (5) short ascending process of the maxilla that may be squared in some individuals; (6) premaxilla evident laterally to the nasal; (7) lack of parietal exposure at cranial vertex; (8) development of a concave posterior wall of the temporal fossa; (9) zygomatic process of the squamosal strongly shortened; (10) low tympanic cavity; (11) low conical process of the tympanic bulla; (12) dorsoventrally oriented mandibular condyle; (13) presence of a depression or a groove for mylohyoidal muscle on the medial side of the dentary; (14) fusion of cervical vertebrae; and (15) long baleen.” [11: 15]

Of these 15 purported balaenoid synapomorphies, seven directly or indirectly describe the same feature, namely, the anterior extension of the supraoccipital shield. Thus, as the supraoccipital extends anteriorly (1), it excludes the parietal both from the vertex (7) and from dorsal view (2), and covers the interorbital region of the frontal (3). As it extends forward, the supraoccipital leaves no room for the rostral bones (maxilla, premaxilla and nasal) to project posteriorly on to the vertex, thus precluding overlap of the rostral bones with the parietal (4) and resulting in a short ascending process of the maxilla (5), with the latter–like the nasal–also being unable to extend posterior to the premaxilla (6). All seven of these characters (in particular chars 1–5 and 7) are therefore linked by a simple, reciprocal geometrical relationship: the more the supraoccipital extends anteriorly, the less room there is for the rostral bones to telescope backwards, and vice versa (Fig 13). Functionally, this relationship is presumably constrained by the relatively short intertemporal region of crown mysticetes, which reduces the space available for telescoping; the need to maintain enough of an attachment surface for the semispinalis capitis muscle to support the skull; and the need to position the external nares as far posteriorly and/or dorsally as possible to facilitate breathing.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 13. Reciprocal relationship between the ascending process of the maxilla and the supraoccipital shield.

(a) Eubalaena; (b) Piscobalaena; (c) Balaenoptera, all in dorsal view. All skulls are scaled so that the anteroposterior distance between the antorbital notch and the posterior border of the exoccipital is the same. Arrows indicate the dominant mode of telescoping. The anterior telescoping of the supraoccipital (dark grey) is accompanied by a shortening of the ascending processes of the maxillae (light grey), and vice versa.

https://doi.org/10.1371/journal.pone.0164059.g013

Balaenids, neobalaenines and cetotheriids provide perfect case studies: in the former two, the supraoccipital occupies much of the intertemporal portion of the cranium at the expense of the rostral bones; by contrast, exactly the opposite is true in cetotheriids, where the pronounced posterior elongation of the rostral bones confines the supraoccipital shield to the posteriormost portion of the cranium (Fig 13). Note, however, that in balaenids the ascending process of the maxilla may be genuinely (i.e. ancestrally) short, as judged from the often relatively large exposure of the frontal on the vertex (e.g. in Balaenella brachyrhynus and neonate Eubalaena glacialis, LACM 54763). This condition is especially obvious in the oldest described balaenid, Morenocetus parvus [31], thus making it the only right whale in which the ascending process of the maxilla can be coded without its relative position being obviously compromised by the supraoccipital. In extant balaenopterids, the anterior telescoping of the supraoccipital and the concurrent posterior shift of the rostral bones are more balanced, resulting in an anteriorly truncated supraoccipital shield that meets the equally truncated, squared ascending processes of the maxillae roughly halfway along the vertex.

Overall, the above examples demonstrate that the first seven features cited in support of grouping Caperea with balaenids are interdependent, and thus–for taxa showing pronounced telescoping–should be coded only once to avoid incidental character weighting. In the present analysis, all of these features are therefore subsumed in char. 90, “Anteriormost point of supraoccipital in dorsal view”, which we accept as a potential balaenoid synapomorphy. Consider, however, that neobalaenines differ from balaenids in the detailed arrangement of their skull vertex: whereas the frontal has virtually disappeared from view behind the nasals in both Caperea and Miocaperea (Fig 14), it has escaped obliteration by the supraoccipital in balaenids by insertion between the posterior portions of the rostral bones (Fig 12) (pl. 42 in [32]). This difference in vertex architecture suggests that the pronounced telescoping of the supraoccipital and attendant changes in balaenids and neobalaenines are a result of evolutionary convergence. Miocaperea further differs in having posteriorly convergent nasals, accompanied by equally convergent ascending processes of the maxilla and premaxilla (Fig 14). By contrast, the lateral margins of the nasals in balaenids, and indeed Caperea, are nearly parallel (unclear in the holotype of Balaenella brachyrhynus, in which the nasals are missing, contra [33]; Fig 12). This condition is consistent with the common ancestor of Caperea and Miocaperea having posteriorly convergent maxillae (as seen in other cetotheriids), which were later shortened and made less convergent by the anterior telescoping of the supraoccipital.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 14. Detailed view of the vertex showing the dorsal outline of the nasals.

(a) Caperea marginata, NMNZ MM002235; (b) Miocaperea pulchra, SMNS 46978. Note the subparallel lateral and medial margins of the nasal in C. marginata, compared to posteriorly convergent margins in M. pulchra.

https://doi.org/10.1371/journal.pone.0164059.g014

Out of the remaining characters, we concur that a short zygomatic process of the squamosal (9) and long baleen (15) could be seen to unite Caperea and balaenids–assuming that the presence of long baleen is coded in lieu of the presence of an arched rostrum. Nevertheless, the case for a homologous reduction in the size of the zygomatic process is somewhat speculative, given the rather disparate morphologies of neobalaenines and balaenids in this regard (Fig S4 in [4]). In Caperea and, as far as can be told, Miocaperea, the zygomatic process is reduced to the point of obliteration, but seems to be oriented anteriorly judging from the position of the (usually juxtaposed) postorbital process of the frontal. By contrast, the zygomatic process of balaenids is typically much better developed, twisted in anterior view, and oriented anterolaterally. Further evidence is needed to demonstrate that these conditions are plausibly homologous.

It is unclear how to interpret the statements that Caperea and balaenids share (8) a “concave posterior wall of the temporal fossa” (p. 15 in [11]), as the direction of the concavity was not specified. Confusingly, the corresponding character (number 76, “Strong concavity in temporal fossa posterior to emergence of supraorbital process of frontal”; p. 10 of the suppl. material of [11]) appears to be coded as “present” (state 1) for Caperea, Zygorhiza kochii, Aetiocetus weltoni and balaenopterids (including grey whales), but not a single balaenid, thus invalidating this character as a potential balaenoid synapomorphy. Likewise, we are unsure about the correct interpretation of the presence of a “low tympanic cavity” (10). We agree that the tympanic bulla of Caperea is dorsoventrally flattened in medial view, but the same cannot necessarily be said for balaenids, in which this compression appears to be limited to the main ridge (Fig 15). Peripolocetus vexillifer is an exception, but even in this case the dorsoventral compression of the bulla appears less marked than in Caperea. Additional data on the bulla morphology of Miocaperea and archaic balaenids, such as Morenocetus and Peripolocetus, are necessary to test whether a flattened bulla might indeed represent a shared feature.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 15. Comparison of the tympanic bulla of Caperea, balaenids and Herpetocetus in medial view.

(a) Caperea marginata, OM VT227; (b) Eubalaena australis, NMNZ MM000226; (c) Balaenula astensis, MSNTUP I12555; (d) Balaena mysticetus, USNM 15695; (e) Balaenella brachyrhynus, NMB 42001; (f) Herpetocetus transatlanticus, USNM 182962. E. australis and B. astensis have been mirrored to facilitate comparisons.

https://doi.org/10.1371/journal.pone.0164059.g015

We disagree with the claim that the presence of a low conical process of the tympanic bulla is demonstrably homologous in balaenids and Caperea, but not Herpetocetus (11) [11]. All three of these taxa appear to show a comparable degree of reduction and dorsal flattening of the conical process, and thus do not allow an a priori distinction into different character states. A more stringent test of homology might be whether the conical process of fossil neobalaenines is taller than in Caperea and, if so, whether its morphology more closely resembles that of cetotheriids or balaenids. At present, there is no material that could provide such insights. We also disagree that Caperea and balaenids share a dorsally oriented articular condyle of the mandible (12); rather, the condyle in Caperea appears to point posterodorsally (Fig 16), but we note that its mandible is difficult to orient (owing to the pronounced curvature of the mandibular body), and that the in situ position of the condyle needs to be ascertained via dissection. Finally, we agree that the presence of a mylohyoid groove or depression (13) and fusion of the cervical vertebrae (14) characterise both Caperea and balaenids. We also note, however, that the attachment of the mylohyoid is less clearly developed in Caperea, and that an apparent mylohyoid groove also occurs in at least some cetotheriids (e.g. Herpetocetus morrowi [12], and likely also Piscobalaena nana). Likewise, incipient fusion of the cervical vertebrae is present in some specimens of Herpetocetus, with C2 and C3 being incipiently fused in the type specimen of H. morrowi (UCMP 124950) [12], and C2–4 being partially fused in Herpetocetus sp. from Japan (NMNS PV-19540). Both characters could thus potentially also unite Caperea with (certain) cetotheriids.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 16. Mandible of Caperea marginata in medial view.

(a) NMNZ MM002235; (b) detailed view of the posterior (ramus) portion of the same specimen.

https://doi.org/10.1371/journal.pone.0164059.g016

In summary, only three of the 15 characters cited as balaenoid synapomorphies, namely, numbers 1, 9 and 15, unequivocally support a neobalaenine-balaenid clade. The remainder either code for the same feature (2–7), are equivocal or do not apply (8, 10 and 12), or are shared by neobalaenines, balaenids and at least some cetotheriids (11, 13 and 14). This relatively weak morphological support for Balaenoidea is outweighed by the morphological evidence supporting a neobalaenine-cetotheriid clade, as well as the fact that molecular analyses consistently group Caperea with extant balaenopteroids [8, 9]. As a result, we therefore here strongly reaffirm our previous referral of neobalaenines to Cetotheriidae [4], and our identification of Caperea as the last survivor of this once diverse family.