Premaxillary crest variation within the Wukongopteridae ( Reptilia , Pterosauria ) and comments on cranial structures in pterosaurs

Cranial crests show considerable variation within the Pterosauria, a group of flying reptiles that developed powered flight. This includes the Wukongopteridae, a clade of non-pterodactyloids, where the presence or absence of such head structures, allied with variation in the pelvic canal, have been regarded as evidence for sexual dimorphism. Here we discuss the cranial crest variation within wukongopterids and briefly report on a new specimen (IVPP V 17957). We also show that there is no significant variation in the anatomy of the pelvis of crested and crestless specimens. We further revisit the discussion regarding the function of cranial structures in pterosaurs and argue that they cannot be dismissed a priori as a valuable tool for species recognition.

Most recently several specimens of the nonpterodactyloid clade Wukongopteridae were discovered in Middle to Late Jurassic Tiaojishan Formation in China (Zhou andWang 2010, Sullivan et al. 2014).These specimens have called attention due to the particular array of primitive and derived features that placed them as closely related to the derived Pterodactyloidea, although their actual phylogenetical position is still disputed (Wang et XIN CHENG, SHUNXING JIANG, XIAOLIN WANG and ALEXANDER W.A. KELLNER al. 2009, 2010, Lü et al. 2010, Andres et al. 2014).There is also some discussion about their diversity, with some authors considering them monospecific (Lü et al. 2010(Lü et al. , 2011a) ) while others regard them as more diverse (Wang et al. 2010, Lü et al. 2011b).
Among the anatomical features used to establish the wukongopterid diversity is the premaxillary crest.In some specimens, the crest is absent (Wang et al. 2010, Lü et al. 2011a) while in others it varies in shape and extension (Wang et al. 2010, Lü et al. 2011b, Cheng et al. 2016).More recently, a specimen with an associated egg, that lacked a crest, was reported (Lü et al. 2011a), raising the possibility that the presence and absence of such structure might indicate sexual dimorphism in pterosaurs.
The main purpose of this paper is to review and discuss the differences of the cranial crests found in the Wukongopteridae and revisit the debate concerning these head structures.We also briefly report on a new wukongopterid specimen (IVPP V 17957) that shows a premaxillary crest different from all others within this group.

DESCRIPTION AND COMPARISONS
Up to now, eight specimens referable to the Wukongopteridae have been published, all coming from the Middle to Upper Jurassic Tiaojishan Formation (northeast China).They have been considered to represent three genera, Wukongopterus, Darwinopterus, and Kunpengopterus (Wang et al. 2009, 2010, 2015, Lü et al. 2010, 2011a, b), or remained indeterminate (Cheng et al. 2016).Two additional individuals were also referred to this clade.One is the holotype of Changchengopterus pani, that lacks a skull (Lü 2009) and has been considered a putative wukongopterid (Wang et al. 2009).However, this very small and apparently ontogenetic young animal has comparatively shorter cervical vertebrae compared to other wukongopterids and therefore its exact phylogenetical position remains to be established.The second specimen was described by Zhou and Schoch (2011) and consists of a fairly complete skeleton lacking the skull.Based on the elongation of the cervical vertebrae, along with a long tail and developed fifth toe, it can confidently be assigned to the Wukongopteridae.
Recently, a new specimen (IVPP V 17957) was discovered from the same region and horizon as the holotypes of Wukongopterus, Kunpengopterus and Darwinopterus (Wang et al. 2009, 2010, Lü et al. 2010, 2011b).IVPP V 17957 consists of a partial skeleton with incomplete skull that can be regarded as a wukongopterid pterosaur based on the following characters: confluent nasoantorbital fenestra, quadrate inclined backwards, elongated cervical vertebrae, reduced cervical ribs, length of the wing metacarpal about half the length of the first wing phalange, elongated tail enclosed by rod-like bony extensions made by the zygapophyses.This new specimen shows a premaxillary crest that differs from previous wukongopterids and therefore is included in the present paper.A full description of this material will be done elsewhere.
Regarding the crested wukongopterids, there is some variation.Cheng et al. (2016) described a specimen (IVPP V 17959) that presents a crest starting anterior to the anterior margin of the nasoantorbital fenestra and that clearly does not reach the posterior region of the nasoantorbital fenestra what is confirmed by the lack of impression in the sediment matrix of this very compacted material (Figs. 1g,1h,2e).
The holotype of Darwinopterus modularis (ZMNH M8782) preserves an incomplete premaxillary crest that lacks the portion above the anterior margin of the nasoantorbital fenestra (Lü et al. 2010, Fig. 2g).According to the illustration provided in the original description (Lü et al. 2010), the crest becomes taller at the anterior part of the preserved portion and extends posteriorly above the orbit.The crest presents sub-vertical fiber-like structures and bears a serrated dorsal margin, that could indicate a soft tissue extension.
Compared to other wukongopterid skulls, Darwinopterus linglongtaensis (IVPP V 16049) preserves a rather large premaxillary crest (Figs. 1c,1d,2b), that is smaller in respect of the one of Darwinopterus modularis (ZMNH M8782) and the one of Darwinopterus robustodens (HGM 41HIII-0309A).It also shows a serrated dorsal margin, with the posterior portion partly preserved as an impression.The crest starts posterior to the anterior margin of the nasoantorbital fenestra and reaches the middle region of the orbit.Most of the dorsal margin of this structure is sub-parallel to the skull roof, but towards the anterior end it becomes taller forming a triangular structure in lateral view.The crest also shows fiber-like striations on the lateral surface that follow a sub-vertical pattern, becoming strongly curved anteriorly (Figs. 1d,2b).
The holotype of Darwinopterus robustodens (HGM 41HIII-0309A) preserves a nearly complete premaxillary crest with a serrated dorsal margin (Lü et al. 2011b, Figs. 1i, 1j, 2f).The crest starts anterior to the nasoantorbital fenestra and is about one third longer than the nasoantorbital fenestra.Based on the impression in the matrix, the crest extends posteriorly above the orbit, similar to the condition present in Darwinopterus linglongtaensis (Figs. 1c,2b).As the former, also this specimen shows fiber-like structures on the lateral surface of the crest that get curved and more marked towards the anterior end (Figs. 1j,2f).
The new specimen (IVPP V 17957) shows a crest that is quite different from the previous ones.It starts posterior to the anterior margin of the nasoantorbital fenestra, albeit more anterior than Darwinopterus linglongtaensis and extends posteriorly.It is much lower than in other crested wukongopterids, except for IVPP V 17959 and based on the preserved portion, reached the orbits (Figs. 1e,1f,2c).The lateral surface of the crest is smooth, differing from all other wukongopterid crests, showing a straight dorsal margin.The anterior end of the crest lacks an extensive dorsal projection as the one found in Darwinopterus linglongtaensis and Darwinopterus robustodens (Figs. 1d,1f,1g,2b,2c,2f).
In a very influential paper, Bennett (1992) argued that the large collection of Pteranodon specimens, a Late Cretaceous pterodactyloid from North America, could be divided into size classes suggesting sexually dimorphism.According to him, small individuals with small cranial crests show a pelvis with a large pelvic canal while the reverse is true for larger individuals (i.e., large cranial crest -small pelvic canal).However, there is no direct association of the skulls and pelves that could back this hypothesis (e.g., Kellner and Tomida 2000).Furthermore, it has been suggested that, at least theoretically, the differences found by Bennett (1992) could be accounted for ontogeny and taxonomy (Tomkins et al. 2010).In fact, a re-evaluation of several specimens attributed to Pteranodon has shown that in some cases there are sufficient morphological differences other than the shape and size of the cranial crest, supporting a larger taxonomic diversity within what can be called the Pteranodon-complex (Kellner 2010).
In any case, it should be noted that since both morphs of Pteranodon, which supposedly represent males and females, bear crests (Bennett 1992, Hone et al. 2012), it is not the presence but the expression in terms of sizes and shapes of these structures that might be regarded as sexually dimorphic.This last notion was recently supported by the exceptional discovery of the tapejarin tapejarid Caiuajara dobruskii in the southern part of Brazil, represented by dozens of individuals of different ontogenetic stages and potentially different sexes.It could be determined that right from a very early ontogenetic stage, this pterosaur bore a cranial crest (Manzig et al. 2014).
In order to explain the cranial crests or other over developed structures as sexually selected traits demands considerable sexual dimorphism (Padian and Horner 2011), but this has never been adequately established in pterosaurs.Despite the heated debate on this topic (e.g., Hone et al. 2012, Knell et al. 2013a, b, Padian and Horner 2013) the fact is that sexual dimorphism (and several other biological questions) are difficult to be tested without a large number of individuals that belong to the same or closely related populations (Kellner et al. 2013), what has only rarely been the case for pterosaurs (Grellet-Tinner et al. 2014, Manzig et al. 2014, Wang et al. 2014a).
Regarding these flying reptiles, an interesting discovery shed some light on the question of the cranial crests being sexually dimorphic.In a Cretaceous deposit from northwestern China, dozens of individuals of the pterodactyloid pterosaur Hamipterus tianshanensis of different ontogenetic stages were found (albeit not covering the same range as in the Caiuajara material).Since this Chinese material is associated with eggs, it is expected that males and females are present in this collection.Hamipterus tianshanensis bears a XIN CHENG, SHUNXING JIANG, XIAOLIN WANG and ALEXANDER W.A. KELLNER premaxillary crest that, in similar sized individuals, showed consistently two distinct morphotypes: one with larger and more robust crests, and the second with smaller and more delicate crests.These morphotypes were tentatively regarded as males and females, respectively (Wang et al. 2014a).Although the Chinese specimens collected so far are disarticulated and pelvic elements that might allow a more detailed assessment of crest-independent sexual dimorphism characters (Padian and Horner 2011) are rare, this occurrence constitute, to our knowledge, the best argument favoring sexual dimorphism expressed by cranial crests.Granted, it should be noted that in this taxon, the sizes of these structures are not particularly large as compared to other pterosaurs such as Thalassodromeus, Pteranodon, and Geosternbergia (Bennett 1992, Kellner and Campos 2002a, Kellner 2010).Once again, this is another example where the expression of the crest and not its presence shows potential for sexual dimorphism.
The hypothesis put forward by Lü et al. (2011a) and that was subsequently used as example for sexual selection in the fossil record (Knell et al. 2013a) defends that the wukongopterid Darwinopterus is sexually dimorphic, with males having a premaxillary crest that is absent in females.Their argument resides in two main points: the direct association of a specimen attributed (even if not explicitly) to the species Darwinopterus modularis, and the comparisons of the pelves of four individuals, two of which were regarded as males and two as females (Lü et al. 2011a

suppl. material).
There is no question that the main specimen (ZMNH M8802/IVPP V 18043) is crestless (Fig. 2d) and was associated with not only one, but two eggs, one still inside the body of the animal (Wang et al. 2015) and therefore can be confidently regarded as a female.However, there are other aspects of the study of Lü et al. (2011a) which are more problematic.The skull of the second supposed female (YH-2000) is badly preserved (see Lü et al. 2010: fig. 2f), lacking most of the dorsal portion, impeding the determination if a crest is present or not.In their original description, Lü et al. (2011a) implied that the wukongopterid specimen ZMNH M8802/IVPP V 18043 represented the species Darwinopterus modularis, failing to acknowledge the existence of other members of this clade (Wang et al. 2009(Wang et al. , 2010)).Oddly, in the same year, Lü et al. (2011b) reinterpreted one of the original males (HGM 41HIII-0309A; Lü et al. 2011a) as a new species, Darwinopterus robustodens (still not acknowledging previous studies on the Wukongopteridae).
We have examined the question about the sizes of the pelvis in more debt.Although Lü et al. (2011a) advocated the presence of two males and females, actually regarding the dimensions of the pelvis, there are only two specimens where this portion of the skeleton can be examined in more detail, one of each gender: ZMNH M8782 the holotype of Darwinopterus modularis, regarded as a male, and ZMNH M8802/IVPP V 18043, the indisputable female.According to Lü et al. (2011a), both pelves show quite similar dimensions, but the crested specimen (ZMNH M8782) is about 13% larger (based on the length of the humeri), a difference which, according to our estimates, might have been even greater (see Table I).Therefore, according to these authors, the pelvis of ZMNH M8782 is proportionally smaller than the one of the female (ZMNH M8802/IVPP V 18043).
We have measured the pelvis of the specimen IVPP V 16049 (holotype of Darwinopterus linglongtaensis, Wang et al. 2010; Table II), that bears a developed premaxillary crest and has reached about 68% the size of the holotype of Darwinopterus modularis (ZMNH M8782; Table I).If Lü et al. (2011a) are correct, IVPP V 16049 should represent a male individual.We have followed the procedures outlined by them regarding the measurements of the pelvis (see Lü et al. 2011a: suppl. material), which differ from the   procedures adopted by Bennett (1992).Comparing the size and proportions of the pelvic elements of these wukongopterid specimens, however, we could establish that the pelvis of IVPP V 16049 is proportionally larger and deeper than the one of the holotype of Darwinopterus modularis (ZMNH M8782; Table II).Moreover, when we compare the measurements of the indisputable female (ZMNH M8802/IVPP V 18043), the proportions of that pelvis is not much different from the holotype of Darwinopterus modularis (ZMNH M8782; Table II), meaning that it could not have had a wider pelvic canal (contra Lü et al. 2011a).Actually, if all three specimens represent the same species, IVPP V 16049, which bears a crest, has the deepest pelvis, implying in a proportionally larger pelvic canal (Table II), and according to previous studies on this subject (Bennett 1992, Lü et al. 2011a), would have to be interpreted as a female.And if so, just for the sake of argumentation, it would be unusual to consider these two specimens (ZMNH M8802/ IVPP V 18043 and IVPP V 16049) as representing the same species.
Based on the available wukongopterid skulls (including the new specimen -IVPP V 17957), there are at least three conditions regarding premaxillary crests: absent (Figs. 1a,1b,2a,2d), present and limited posteriorly to the anterior end of the nasoantorbital fenestra 2b,2c), and extended anterior to the nasoantorbital fenestra (Figs. 1i,1j,2f,2g).A fourth condition can also be advocated based on the specimen IVPP V 17959, where the cranial crest is very reduced (Figs. 1g,1h,2e).One could try to argue that, if not taxonomic, this variation could reflect different ontogenetic stages, with young individuals (Fig. 2a) being crestless, that would at a certain ontogenetic stage develop a crest above the nasoantorbital fenestra (Fig. 2b) and expand this crest anteriorly in older (and larger) individuals (Figs.2f, 2g).However, the crestless and smallest individual (based on the length of humerus, IVPP V 16047, Fig. 2a, Table I) has a larger (more than 3.6%) maximum wingspan than the second smallest individual (based on the length of humerus, IVPP V 16049, Fig. 2b, Table I).Even though considering the wingspan as individual differences, it still seems difficult that the appearance of the crest would be a sudden event in such close sized animals.In the best example of an ontogenetic series within pterosaurs that is represented by Caiuajara (admittedly very distantly related to the Wukongopteridae), there seems a continuum in the appearance and development of the cranial crest, present in this taxon at a very young ontogenetic stage (Manzig et al. 2014).Furthermore, there are other important morphological differences that separate Kunpengopterus sinensis (IVPP V 16047) and Darwinopterus linglongtaensis (IVPP V 16049) (see Wang et al. 2010).
Still pursuing the potential ontogenetic argument, the shapes and the sizes of the crests also do not seem to follow an ontogenetic pattern (compare Figs. 2b with 2c and 2f).Secondly, the shapes and the external surfaces of the crests (i.e., smooth or fibrous) also vary to a great degree independent of the sizes of the individuals (compare Figs. 1d with 1f and 1g).

CONCLUSION
The function of cranial crests in pterosaurs will be a matter of contempt until more material becomes available.Although the comparisons of the pelvic elements of wukongopterids is interesting, one cannot overemphasize the reduced number of specimens on which this and other studies are based.Variations introduced by taphonomy, allied with morphological differences as a result of ontogeny and other factors cannot be adequately assessed with three or so individuals.Notwithstanding these shortcomings, the explanation of the evolutionary function of cranial crests as sexual dimorphism (including mutual sexual selection, Hone et al. 2012) should not be used as the default option.As has been pointed out before (e.g., Kellner andXIN CHENG, SHUNXING JIANG, XIAOLIN WANG andALEXANDER W.A. KELLNER Campos 2002a, 2002b), it seems quite unlikely that once a pterosaur developed a cranial crest (as perhaps the so called "bizarre structures" in other reptiles), this structure could not have performed a variety of functions.Despite the limitation of the available data, it seems difficult not to acknowledge that these quite distinct anatomical features have played a major role in species recognition, as has been put forward for other reptiles (e.g., Padian and Horner 2011).This has been shown to be true at least for some material of the Pteranodon-complex and seems also to have been the case for the Wukongopteridae.Perhaps more detailed studies of deposits with large amount of specimens that might have been part of the same or closely related populations like the occurrences of Caiuajara dobruskii and Hamipterus tianshanensis (and others that might come to light) have the potential to provide a step further into the discussion of several paleobiological questions concerning flying reptiles, including sexual dimorphism and ontogeny (Kellner 2015).Meanwhile the variation in shapes and sizes of cranial crests that are found in pterosaurs, associated with other morphological features, should not be understated as being a powerful tool for understanding their diversity.

Figure 1 -
Figure 1 -The variation of cranial crest within the Wukongopteridae.Kunpengopterus sinensis (IVPP V 16047), (a) complete skull and (b) close-up of the area in the white frame of a, showing the flat crestless skull roof; holotype of Darwinopterus linglongtaensis (IVPP V 16049), (c) complete skull and (d) close-up of the area in the white frame of c; new specimen (IVPP V 17957), (e) complete skull (inverted), and (f) close-up of the area in the white frame of e; IVPP V 17959 (g) complete skull (inverted), and (h) close-up of the area in the white frame of g, with arrow 1 pointing at the low premaxillary crest and arrow 2 pointing at the crestless middle portion of the premaxilla; Darwinopterus robustodens (holotype, HGM 41HIII-0309A) (i) complete skull based on Lü et al. 2011b), and (j) close-up of the area in the white frame of i. Scale bar: 50mm in a, c, e, g, i; 10mm in b, d, f, h, j (see the colors in the online version).

TABLE I Measurements of the length of skull, humerus and ulna of the Wukongopteridae, from the smallest to the largest individual based on wingspan (in mm).
Kellner et al. 2013st, estimated value.Abbreviations: hu, humerus; maxws, maximized wingspan; sq-pm, squamosal to the tip of premaxilla; ul, ulna.The maximized wingspan (maxws=cor+hu+ul/ra+car+mcIV+ph1d4+ph2d4+ph3d4+ph4d4,Kellner et al. 2013; abbreviations and measurements see TableIII) of Kunpengopterus sinensis, Darwinopterus linglongtaensis, Wukongopterus lii, YH-2000, and Darwinopterus robustodens is based on complete forelimb elements plus the coracoid, while that of IVPP V 17959, ZMNH M8802/IVPP V 18043, and the holotype of Darwinopterus modularis is estimated by the ratio between the preserved elements and maximized wingspan compared to complete specimens.

TABLE II Measurements (in mm) and ratio of the pelvis and humerus of Darwinopterus modularis, Darwinopterus linglongtaensis, and one female wukongopterid.
~, estimate based on the third sacral vertebra; est, estimated value based on pictures.Abbreviations: hu, humerus; pd, pelvis depth from dorsal margin of the ilium to the ventral margin of the puboischiadic plate at level of the acetabulum; pl, pelvis length from anterior tip of preacetabular process of ilium to posterior tip of postacetabular process of ilium; pw, pelvis width between the acetabula; sl, length of sacral vertebrae.126XINCHENG, SHUNXING JIANG, XIAOLIN WANG and ALEXANDER W.A. KELLNER