Prenatal development in pterosaurs and its implications for their postnatal locomotory ability

(a). Egg shape and size

Changes in egg shape and size were investigated by plotting predicted egg mass (EM) against egg elongation ratio (EER), which is the length of egg along the major axis/length of egg along the minor axis, for 37 eggs of Hamipterus tianshanensis [6,8], including three with embryos, and two eggs each containing a complete, near-term embryo of an ornithocheirid from the Lower Cretaceous Yixian Formation of China [1,2] (electronic supplementary material, table S2). We included data for a relatively small egg (IVPP V18938) also from the Hami locality [6], that probably represents a species of pterosaur distinct from that of H. tianshanensis (figure 1; electronic supplementary material, table S2). See the electronic supplementary material for further discussion of the identity of this fossil.

(b). Predicting egg mass

EM was predicted using the same dataset (electronic supplementary material, table S2) as that used to calculate the EER. EM was predicted from egg length (L) and egg breadth (B) using the methodology of Hoyt [18], where EM = 0.56 L × B². The constant K (0.56) is derived from symmetrical crocodilian eggs [19] which provide a better model for pterosaur eggs than asymmetric bird eggs [4,5,14]. Wang et al. [7] applied a correction factor to their estimate of EM based on experimentation with an egg of the extant ratsnake Elaphe taeniura. Such approaches have merit, but in this case, required further development to avoid a number of potential pitfalls. For example, it is not clear whether the eggshell of E. taeniura has comparable biomechanical properties (elasticity, compliance) compared to the eggshell of pterosaurs. Moreover, experimental conditions for the egg of E. taeniura, including residence time in an aquatic regime, the chemical composition of the fluid in which the egg was immersed and key aspects of compressional regimes including speed of and geometry of compression and the compliance of surrounding sediment (or absence of it), are likely to have differed dramatically from those experienced by pterosaur eggs. In the absence of more comprehensive studies, it is simpler to apply a standardized method of predicting EM to all pterosaur eggs [5,15] and avoid complications stemming from the introduction of arbitrary correction factors.

(c). Morphometric analyses

Data were compiled from the literature (see sources listed in the electronic supplementary material, tables S1, S3 and S4) and personal examination (by D.M. Unwin 2015, 2017) of fossil material. The entry for the composite of H. tianshanensis was based on an aggregation of data for embryos 11, 12 and 13 of this pterosaur. These are closely comparable in size and thought to represent the same developmental stage [8]. Missing values were estimated by multiplying the length of the relevant element of IVPP V13758, a similarly sized embryo of a closely related taxon, by 1.05. This value was generated by calculating the mean for seven cases where the length of the element in IVPP V13758 could be directly compared to that for the composite: (∑(element length IVPP V13758/element length composite))/7 = 1.05. Similarly, forelimb lengths for IVPP V18943, embryo 7 and IVPP V18942 hatchling were estimated on the basis of the humerus/forelimb index for IVPP V13758. Note that this relationship shows near isometry in ornithocheirids (figure 2; electronic supplementary material, table S3).

(d). Determining prenatal patterns of ossification

(a). Egg size and morphology

More than 300 eggs have been reported for H. tianshanensis, an ornithocheirid pterosaur from the Lower Cretaceous of China ([6,8]; electronic supplementary material, table S1). The eggs exhibit a relatively large size range, the largest reaching 151% the length and four times the predicted mass of the smallest (electronic supplementary material, table S2). The most elongated eggs are those with the smallest predicted mass and as size increases the eggs become less elongated, with the heaviest examples showing the lowest EERs. This change in shape is primarily achieved by a disproportionate increase in width (168%) compared to length (151%). The eggs containing the three smallest embryos (IVPP V 18941–3 embryos 5, 11, 12) and thought to represent the same developmental stage [8] are closely comparable in size to each other and have EERs (1.91–2.21) that fall within the middle of the range (1.70–2.63). Eggs of a closely related ornithocheirid from the Yixian Formation of China [1,2] exhibit low EERs (1.29–1.75), comparable to, or lower than that for, Hamipterus eggs (figure 1).

The parchment-like shell structure of the eggs of extant lizards and snakes allows them to absorb water during incubation [29]. In lizards, this results in an increase in mass, over incubation, of 150–200%, accommodated by an increase in egg length and breadth of 23–60% depending on species [30–32], with breadth tending to increase more rapidly leading to a decrease in the EER. The presence of the same pattern in Hamipterus, in which the largest eggs show the lowest EERs, suggests that the range of egg sizes and egg elongation exhibited by this pterosaur [6,8] reflects a sample from differing stages of incubation: the smallest eggs with highest EERs having died at an early stage of incubation, while the largest eggs with the lowest EERs had been incubated for longer and were probably near term when they died.

(b). Size distribution and morphometrics

Reported lengths for the humerus of embryos and a hatchling of H. tianshanensis exhibit a relatively large size range: 13.3 mm in the three smallest embryos (numbers 11–13), 20% longer in embryo no. 7 and 40% longer in a hatchling ([8]; electronic supplementary material, table S3). This corresponds to estimated forelimb lengths of between 124 and 175 mm and wingspans of 0.26–0.36 m (figure 2). Based on regression analyses of recent mass estimates for pterosaurs [33,34], the mass of the Hamipterus hatchling is likely to have been at least double that of the smallest embryos (numbers 11–13). The relatively large size range subtended by the embryos and hatchling of Hamipterus is further emphasized by comparison with data for prenatal and early postnatal growth stages in other pterosaurs (figure 2; electronic supplementary material, tables S1 and S4). In species such as Aurorazhdarcho micronyx, immature individuals of which span the exact same size range as individuals of Hamipterus [21], there is a significant change in skeletal proportions. The humerus, for example, exhibits negative allometry with regard to forelimb length, a relationship that is common within pterosaurs, although ornithocheirids are exceptional in that the humerus–forelimb index shows a near isometric relationship over a large range of sizes (electronic supplementary material, table S3). The key point here is that the smallest individual of Au. micronyx, which corresponds in size to embryos 11–13 of Hamipterus, represents a distinctly different, younger growth stage from that of the largest and presumably oldest individual, which corresponds in size to the Hamipterus hatchling (figure 2).

(c). Ossification sequences in prenatal pterosaurs

The single known embryo of P. guinazui (figure 3a; electronic supplementary material, table S1) [3,22], generally thought to be near term, or full term [3,4,15,22], is closely comparable in terms of degree of ossification to hatchlings of the same species (MIC V241, MMP 1168; [28]) and immature individuals of Pterodactylus kochi [21] and Rhamphorhynchus muensteri [35], although the phalanges of manus digits I–III are seemingly unossified in the embryo (figure 3b). The same elements, and the metacarpals, tarsals and pedal phalanges, are unossified in embryos of an ornithocheirid pterosaur from the Yixian Formation ([1,2]; electronic supplementary material, table S1), suggesting a slightly earlier developmental stage. The skeletons of Hamipterus embryos 11–13 are markedly less well ossified (figure 3a), with no evidence for a series of elements, including the rostrum, teeth, sacral and caudal vertebrae, ribs, gastralia, carpals, manus digits I–III, the pelvis, tarsals or pes digits I–IV. Those elements that are present (cervical and possibly dorsal vertebrae, coracoid, metacarpals I–III, distal wing-phalanges, metatarsals) appear more poorly ossified compared to the same structures in IVPP V13758 or MIC V246. This is also true for some limb elements such as the humerus which, in embryo 13, seems less clearly defined than that of the Hamipterus hatchling or other embryos (figure 2).

(d). Comparison with embryological development in extant archosaurs

Comparison with the well-established prenatal developmental sequence for Al. mississippiensis [23,24] shows that the Yixian Formation ornithocheirid embryos correspond most closely to developmental stages 23/24 of Al. mississippiensis, while the Pterodaustro embryo corresponds to stages 24/25 (figure 3c). Hamipterus embryos 11–13 show closest congruence with developmental stages 21/22 of Al. mississippiensis: the teeth have not yet erupted and elements which ossify at later stages of development in Al. mississippiensis (sacral and caudal vertebrae, ribs, gastralia, pelvis, carpals, manus digits I–III, tarsals and pedal phalanges) are seemingly unossified in these embryos. Notably, the tibia/fibula, which ossifies at stage 18 in Al. mississippiensis also appears to be absent. By contrast, metacarpal IV and proximal phalanges of manus digit IV (i.e. the wing-finger) are ossified in the Hamipterus embryos. This might suggest a slightly later stage of development as, in Al. mississippiensis, the homologous elements ossify in stages 22 and 23, respectively. However, an equivalency to Al. mississippiensis stages 22/23 for the Hamipterus embryos is less congruent than for stages 21/22 and probably reflects a pattern of ossification that is heterochronic relative to that of Al. mississippiensis and interpreted here (see below) as flight related. Irrespective of whether embryos 11–13 correlate to stages 21, 22 or 23, it is clear that within a developmental framework based on Al. mississippiensis, these embryos represent an earlier developmental stage than the Pterodaustro or ornithocheirid embryos.

Comparison with the ossification sequence for the quail, C. coturnix (electronic supplementary material, table S5) [25–27] yielded the same pattern as those for Al. mississippiensis with regard to the differing stages of development reached by the pterosaur embryos included in this analysis. Hamipterus embryos 11–13 conform most closely to day 8/9 of embryonic development in C. coturnix, the Yixian ornithocheirid embryos to day 12, or possibly later, and the Pterodaustro embryo to day 16 (term). Sequence incongruencies reflect the relatively late stage of mineralization of the vertebral column in Coturnix [26], and a shift to an earlier stage of ossification of manus digit IV in pterosaurs.

(a). Extending the fossil record of prenatal development in pterosaurs

Current interpretations of the fossil record of early development in pterosaurs, heavily reliant upon qualitative assertions [17], identify almost all finds as perinatal—either near-term embryos and eggs, or hatchlings [1–4,6–8,14,15,22]. Multiple lines of evidence used in this study provide a much firmer basis for, and greater precision in, the assignment of records to particular growth stages and show that some eggs and embryos should be reassigned to earlier stages. This extends our knowledge of prenatal development in pterosaurs (figure 4) and has important consequences for understanding the conditions under which eggs were incubated and how they developed, prenatal ossification sequences in pterosaurs and how they compare to those of other tetrapods, and the implications of those sequences for the locomotory ability of pterosaur hatchlings.

(b). A new map of pre- and perinatal development in pterosaurs

Two eggs preserved in association with a specimen of Darwinopterus modularis (ZMNH M8802) provide direct evidence for egg size and morphology at oviposition [5,7]. The EER for these eggs (1.45–1.56) is relatively low, but may reflect a species-specific morphology or taphonomic compression of the fossil. The most elongate, relatively low predicted mass of eggs of Hamipterus (figures 1 and 4) probably represent the earliest stages of incubation prior to any significant skeletogenesis.

Embryos 11–13 of Hamipterus appear to represent a ‘mid-term’ stage of development in which skeletogenesis had begun, but was relatively incomplete (figure 4). This is consistent with the mid-range degree of elongation and masses of eggs containing these embryos (figure 1), the relatively small size of the embryos (figure 2) and the lower degree of ossification, compared to late term embryos (figure 3b). This roughly corresponds to developmental stages 21/22 in Al. mississippiensis (figure 3c) and day 8/9 in Coturnix (electronic supplementary material, table S5) which is approximately 50% of the incubation period for these taxa [24,26,36]. Hamipterus embryo 5, enclosed in an egg with a relatively high EER (2.1) and with a seemingly very poorly ossified skeleton [8], may represent a slightly earlier stage of development compared to embryos 11–13.

Compared to Hamipterus embryo 13, the scapula of embryo 4 and the humerus of embryo 7 are approximately 20% longer ([8]; electronic supplementary material, table S2). The latter two specimens probably represent late term embryos that were accommodated in a larger, less elongate egg, comparable to those of the Yixian ornithocheirid (IVPP V13753; JZMP 03-03-2) that exhibit a relatively low EER (figure 1). Hamipterus eggs with comparable EERs exhibit cracking and crazing of the outer surface of the eggshell (electronic supplementary material, figure S1; [6]) that probably reflects the accommodation of volumetric changes following water uptake.

Comparison with hatchling and immature pterosaurs and prenatal developmental sequences for Al. mississippiensis and C. coturnix confirm the conclusions of earlier studies that the Yixian ornithocheirid embryos were at a late developmental stage when they died [1,2,4,14,15] but, seemingly, not quite as advanced as the Pterodaustro embryo. The latter has been recognized as well developed [3,4,14,15,22,37,38] and is more specifically identified here as very near term or at the point of hatching when it died.

Many of the Hamipterus eggs described by Wang et al. [6,8] lack evidence of embryonic skeletal remains and seem to be collapsed. Dimpling of the shell may reflect dehydration postmortem [6], but might also represent empty egg shells after hatching, particularly for shells that are crumpled [8], or seem to have been slit open [8], as observed in hatched eggshells of extant lizards [39]. One example (figure 4j [8]) exhibits an EER (1.83) typical of late term eggs.

Two individuals of P. guinazui (MIC V241, MMP 1168) closely comparable in size and skeletal development to MIC V246, and interpreted as ‘at a very early stage of postnatal development’ [28], are confirmed here as highly immature and probably hatchlings. A small humerus preserved on block IVPP V 18942 has been identified as belonging to a hatchling of Hamipterus [8]. This is consistent with its size, degree of ossification and relative proportions (the length of the deltopectoral crest compared to the length of the humerus is only slightly greater than in the Hamipterus embryos [8]), although the possibility that it represents an isolated element from a pre-hatching individual cannot be entirely excluded.

(c). Implications of prenatal development for postnatal flight ability in pterosaurs

Precise identification of the developmental stage of embryos is critical for inferring the locomotory ability of hatchlings [17,40]. It has been argued, primarily on the basis of embryos 11–13, that the more advanced state of ossification of hind limb compared to forelimb elements and the incomplete ossification of key flight muscle attachment sites, such as the processus scapularis of the scapula and deltopectoral crest of the humerus, means that hatchlings of Hamipterus would not have been able to fly [8,13]. However, as shown here, embryos 11–13 probably represent a mid-term stage of development; consequently, the relatively poor ossification of forelimb bones and muscle attachment sites almost certainly reflects initial stages in the mineralization of the skeleton, and not the skeletal morphology of near-term embryos, as, for example, Pterodaustro (MIC V246). Indeed, patterns of ossification evident in the prenatal pterosaur fossil record described here (figure 4) provide new support for the idea of super-precocial flight ability in pterosaurs.

Hamipterus embryos 11–13 show two key features. First, a relatively greater degree of ossification of the principal fore and hind limb long bones, which form the main wing spars, compared to the rest of the skeleton (figure 3b). With the exception of the tibia–fibula, all the principal wing-spar elements are at least partially ossified including the metacarpus and proximal phalanges of manus digit IV which, in Al. mississippiensis and C. coturnix ossify at later stages of development (figure 3; electronic supplementary material, table S5), as is typical for amniotes [41]. This pattern seems to have been underpinned by a developmental heterochrony unique to pterosaurs that involved a shift in the timing of ossification of metacarpal and digit IV to a much earlier stage of development (figure 3b). Second, embryos 11–13 exhibit relatively elongate fore limbs, the estimated length of which is more than nine times the length of the humerus. Principally, this was achieved through hyper-elongation of metacarpal IV and manus digit IV which reach 114% and 165% of humerus length, respectively (electronic supplementary material, table S3). The early ossification, and elongation, of flight-related skeletal elements, as demonstrated by the mid-term growth stages of Hamipterus, shows that unique components of the pterosaur bauplan were present at the earliest stages of skeletal formation and acted as anatomical precursors to the final stages of prenatal development.

Terminal stages of embryonic development, represented by MIC V246 [3,22,37,38], IVPP V 13758 [2,4,15], JZMP 03–03-2 [1,4,15], and the humeri of a near-term embryo (no. 7) [8] and a hatchling of Hamipterus (figure 2), have multiple features that point towards flight ability in hatchlings. First, extensive ossification of all elongate structures contributing to the flight apparatus that are likely to have experienced significant loads in bending during flight. These include dorsal and sacral vertebrae, the limb girdles and diaphyses of long bones that form the wing spars. This stiffening of the skeletal components of the flight module is analogous to ossification sequences in Al. mississippiensis [24], the hatchlings of which are also highly precocial locomotors, but is in sharp contrast to most extant birds where, prior to hatching, only the central region of the diaphysis of long bones is ossified [42].

Second, inferences regarding the implied lack of development of key flight muscles, based on the absence or poor development of osteological features, are insecure for two reasons: (i) muscle attachment sites do not need to be ossified in order to function effectively [43]. In tension, cartilage can accommodate loads comparable to those for bone [44]; consequently, it cannot be assumed, a priori, that an incomplete deltopectoral crest directly implies a relatively small m. pectoralis, the principal wing depressor; and (ii) the relative size and shape of the deltopectoral crest of embryos 7, 11–13 and the hatchling (figure 2) is smaller than that of adult Hamipterus, but it is directly comparable in terms of shape and relative size to the deltopectoral crest of other pterosaurs including individuals of Anurognathus and Aurorazhdarcho that are widely considered to have been flight capable [9,14,20,21,33,34].

Third, the relative elongation of long bones contributing to the wing spars, their relative proportions to each other and the relative elongation of the fore limb of mid and late term embryos compare closely to the same indices for mature, flight capable individuals of ornithocheirids (electronic supplementary material, table S2). This is in sharp contrast to most birds and all bats where fore limb proportions comparable to those of adults, and flight ability, are only achieved at a relatively late stage of postnatal development [45,46].

(d). Parental care in pterosaurs

Parental care in pterosaurs has been proposed by several authors [3,8–13,47,48], although the particular modes of care (e.g. defending nest, protecting young) are rarely specified. The requirement for postnatal care has often been directly related to the idea that hatchling pterosaurs were not flight capable, as suggested for Hamipterus [8]. However, as shown here, available evidence favours the idea that the hatchlings of Hamipterus and other pterosaurs were flight capable at a very early stage in their postnatal development. It may be that early postnatal growth stages received parental care, in the form of protection from predators, as, for example, in crocodilians [49], but such a behaviour is difficult to demonstrate and, for the present, there is no direct evidence to suggest that pterosaur hatchlings required parental care.

The fossil accumulation at Hami, comprising eggs, embryos, hatchlings, juveniles and adults, has been interpreted as evidence for gregariousness in pterosaurs, a component of which might have involved care of young [8]. Alternatively, however, the nesting site of Hami may have attracted Hamipterus and seemingly another pterosaur, because it provided a suitable substrate within which to deposit clutches and the persistence of this sedimentary environment through time encouraged multiple nesting events. Gregariousness (in the sense of individuals interacting with each other) is not a required outcome of such locations and a better analogy might be turtle nesting sites [50].