Comparative skeletal anatomy of neonatal ursids and the extreme altriciality of the giant panda

Abstract

Mammalian neonates are born at a wide range of maturity levels. Altricial newborns are born with limited sensory agency and require extensive parental care. In contrast, precocial neonates are relatively mature physically and often capable of independent function shortly after birth. In extant mammals, placental newborns vary from altricial to precocial, while marsupials and monotremes are all extremely altricial at birth. Bears (family Ursidae) have one of the lowest neonatal−maternal mass ratios in placental mammals, and are thought to also have the most altricial placental newborns. In particular, giant pandas (Ailuropoda melanoleuca) are thought to be exceptionally altricial at birth, and possibly marsupial‐like. Here we used micro‐computer (micro‐computed) tomography scanning to visualize the skeletal anatomy of ursid neonates and compare their skeletal maturity with the neonates of other caniform outgroups. Specifically, we asked whether ursid neonates have exceptionally altricial skeletons at birth compared with other caniform neonates. We found that most bear neonates are similar to outgroup neonates in levels of skeletal ossification, with little variation in degree of ossification between ursine bears neonates (i.e. bears of the subfamily Ursinae). Perinatal giant pandas, however, have skeletal maturity levels most similar to a 42−45‐day‐old beagle fetus (~70% of total beagle gestation period). No bear exhibits the skeletal heterochronies seen in marsupial development. With regards to skeletal development, ursine bears are not exceptionally altricial relative to other caniform outgroups, but characterized largely by the drastic difference between newborn and adult body sizes. A review on the existing hypotheses for ursids’ unique reproductive strategy suggests that the extremely small neonatal−maternal mass ratio of ursids may be related to the recent evolution of large adult body size, while life history characteristics retained an ancestral condition. A relatively short post‐implantation gestation time may be the proximal mechanism behind the giant panda neonates’ small size relative to maternal size and altricial skeletal development at birth.

Newborn ursine bears and caniform outgroups have well‐ossified skulls with varying degrees of cranial suture development. Perinatal giant pandas, however, lag behind in cranial skeletal development and most closely resemble a fetal dog in skeletal maturity in the skull.

Introduction

Extant mammals exhibit a wide range of behavioral and morphological maturity at birth (Martin & MacLarnon, 1990; Scheiber et al., 2017). Living placental mammal newborns vary from altricial to precocial (Martin & MacLarnon, 1985; Derrickson, 1992). Altricial placentals (e.g. carnivorans) are born with closed eyes, limited body fur coverage and little ability for independent locomotion; most precocial placentals (e.g. perissodactyls) are born with open eyes and extensive body fur coverage, and can move independently shortly after birth (Eisenberg, 1981; Martin & MacLarnon, 1990). In contrast, living marsupials and monotremes both have short gestation lengths and produce extremely altricial neonates (Tyndale‐Biscoe & Renfree, 1987; Weisbecker, 2011; Smith, 2015). Altriciality at birth has been reconstructed as the primitive condition for crown Mammalia (Werneburg et al. 2016), and this is consistent with the fossil record on non‐mammalian cynodont perinates (Hoffman & Rowe, 2018) and genetic evidence on the early evolution of lactation (Brawand et al. 2008).

At a given adult body size, precocial placental mammal species have larger neonatal and litter weight, smaller litter size, and longer gestation and postnatal growth periods to reach adult body mass (Leitch et al. 1959; Martin & MacLarnon, 1985; Derrickson, 1992; Scheiber et al. 2017). In contrast, altriciality tends to be associated with small neonatal and litter weight, larger litter size and shorter gestation (Martin & MacLarnon, 1985, 1990; Derrickson, 1992). Regardless of neonatal maturity levels, neonatal weight correlates tightly with adult body size (Leitch et al. 1959; Gittleman, 1986), thus neonatal size by itself is not one of the necessary criteria to define maturity levels at birth.

Like other carnivorans, bears (Ursidae) give birth to altricial newborns (Ewer, 1973). Additionally, at parturition ursid neonates are exceptionally small compared with the mother (Fig. 1; Gittleman, 1986; Sibly & Brown, 2009). Despite the negative allometric relationship between placental neonate and maternal body mass (Leitch et al., 1959), neonatal−maternal mass ratio of ursids is among the lowest of all placental mammals, ranging from 1 : 470 (calculated from Jones et al., 2009) to 1 : 600 (Thompson, 1942; Leitch et al. 1959). In contrast, the average is 1 : 49 in Canidae, 1 : 75 in Musteloidea, 1 : 84 for fissiped carnivorans and 1 : 26 for all placentals (Jones et al. 2009). Focusing on the Holarctic bears (brown/grizzly bears, black bears and polar bears), previous authors have predominantly attributed the small size of cubs to energetic constraints of gestation during winter hibernation (Ewer, 1973; Ramsay & Dunbrack, 1986).

Figure 1.

Log‐transformed neonatal vs. adult weight in fissiped carnivorans. Blue line represents the generalized least‐squared linear regression line. The shaded region denotes 95% confidence interval. Individual data points from Jones et al. (2009).

Among other life history traits, all extant bears except the sun bear (Helarctos malayanus) exhibit seasonal reproduction patterns and delayed implantation (Spady et al. 2007; Frederick et al. 2012). Whereas the reported total gestation length of most bears is longer than that of most other large‐bodied terrestrial carnivorans (Tables 1 and S1), the post‐implantation gestation length of ursids is inferred to be much shorter (Tsubota et al. 1987). Compared with other mammals of similar size, bears also have relatively long lactation periods, small litter size and long intervals between pregnancies (Table 1; Ramsay & Stirling, 1988; Gershelis, 2004). The combination of short post‐implantation gestation, long lactation and highly altricial cubs led Ramsay & Stirling (1988) to suggest that bears’ reproductive strategy is more similar to that of marsupials rather than that of placentals.

Table 1.

Gestation lengths, litter sizes, neonatal weight, adult weight, presence/absence of delayed implantation and average reproductive intervals for selected taxa

Species	Total gestation length (days)	Litter size	Neonatal weight (g)	Neonatal crown−rump length (mm)	Adult weight (g)	Delayed implantation?	Reproductive interval (years) averaged over population
Ailuropoda melanoleuca	122–163a (117n)	1–2a	90–131a	180j	75 000–160 000a (118 000n)	Yesb	2.2i
Ursus arctos	192–206c (246n)	2–3c	500c	230–280c	58 000–389 000c (196 290n)	Yesc	2.4–5.2i
Ursus maritimus	195–265d (235n)	1.85–1.87d	600d	300k	150 000–800 000d (371 700n)	Yesd	2.1–3.6i
Melursus ursinus	120–280e (203n)	1–2e	2000e	199m	90 000–140 000e (100 000n)	Yese	2.5i
Ursus americanus	180–240i (225n)	1.5–3i	333–356m	199m	50 100–86 000i (110 500n)	Yesi	2.0–3.2i
Vulpes lagopus	52f (53n)	3–25f	60–90f	210l	3000–5000f (3580n)	Yesg	/
Ailurus fulgens	112–158h (132n)	1–4h	110–130h	280h	4900h (5170n)	Yesn	/

Data from: (a) Chorn & Hoffmann (1978); (b) Zhang et al. (2009); (c) Pasitschniak‐Arts (1993); (d) DeMaster & Stirling (1981); (e) Web sources (2019a): https://nationalzoo.si.edu/animals/sloth-bear; (f) Audet et al. (2002); (g) Ferguson et al. (1996); (h) Roberts & Gittleman (1984); (i) Gershelis (2004); (j) Web sources (2018): https://www.chinahighlights.com/giant-panda/baby-panda.htm; (k) Web sources (2019b): https://seaworld.org/animals/all-about/polar-bear/care-of-young/; (l) average CRL length of a red fox newborn from Larivière & Pasitschniak‐Arts (1996); (m) Oftedal et al. (1993); sloth bear CRL assumed to be the same as black bear. Values in brackets from data presented in (n) Heldstab et al. (2018).

Here we use micro‐computed tomography (μCT) to visualize the skeletal anatomy of several ursid neonates and compare their skeletal maturity levels with those of other caniform outgroup neonates. Although skeletal anatomy has not been explicitly used to assess the relative altriciality of mammalian neonates (Ferner et al. 2017), we suggest it is a useful measure of maturity as it is directly connected to the functional aspects of neonatal survival such as locomotion and suckling. μCT imaging offers a non‐invasive, high‐resolution approach to capture the anatomical details of neonate skeletons (Hautier et al. 2012; Roston et al. 2013; Newton et al. 2018). We focus on the skeletal system and specifically ask: (i) if the minute size of ursid newborns is also reflected in exceptional skeletal altriciality relative to the neonates of other therian mammals, in particular those of their caniform relatives; and (ii) if relative altriciality varies between ursid species. We test several alternative hypotheses: (i) bear neonates resemble outgroup neonates (e.g. canids) in degrees of skeletal ossification, and only differ by the small size of newborns relative to the mother; (ii) bear neonates are more altricial than outgroup neonates, but still conform to a placental pattern of ossification; or (iii) bears significantly differ from other placentals in neonatal skeletal maturity and even exhibit marsupial‐like conditions. Our results provide new and detailed data on the development of internal anatomy among ursids, and we will discuss our results in the larger context of ursid life history strategy.

Materials and methods

Taxonomic sampling

We surveyed 15 caniform neonates/late‐stage fetuses (nine ursid and six outgroup specimens; Fig. 2; Table S2). Ursid specimens include grizzly (brown) bears (Ursus arctos, n = 2), black bear (Ursus americanus, n = 1), sloth bears (Melursus ursinus, n = 2), polar bears (Ursus maritimus, n = 2) and giant pandas (Ailuropoda melanoleuca, n = 2). Outgroup specimens include red panda (Ailurus fulgens, Ailuridae; Flynn et al. 2005, n = 1), coati (Nasua narica, Procyonidae, n = 1), African wild dog (Lycaon pictus, Canidae, n = 1) and arctic fox (Vulpes lagopus, Canidae, n = 1). All specimens above were obtained from the Smithsonian National Museum of Natural History (USNM). Two domestic dogs (Canis familiaris, Canidae, n = 2) from the North Carolina State College of Veterinary Medicine (NCSCVM) were also examined. One dog was clearly fetal; the other was late fetal to neonatal (hereafter neonatal). The level of ossification and dental calcification of the fetus most closely resembled a fetal 50−55‐day‐old beagle, approximately 5–10 days before birth (Evans, 1993; Tables S3 and S4). Detailed description on the external morphology of each specimen is presented in Supporting Information. A time‐calibrated phylogenetic tree for the studied specimens was drawn from data obtained from timetree.org (Kumar et al. 2017) and Slater & Friscia (2019) with the ggtree function (Yu et al. 2017) in R (R Development Core Team, 2019).

Figure 2.

Time‐calibrated phylogeny of caniform taxa studied, modified from timetree.org (Kumar et al. 2017) and Slater & Friscia (2019). The yellow node represents the ancestor of arctoids (Ursidae + Pinnepedia + Musteloidea; Finarelli & Flynn, 2006). Scale bar at the bottom represents geologic age in millions of years (MA). All silhouettes from http://phylopic.org

The two giant panda specimens were originally collected from the Smithsonian National Zoo. They were the cubs of Ling‐ling and Hsing‐hsing, which were gifted to the USA by the Chinese government in 1972 (Lunde, personal communication). One of the panda specimens (USNM 546337), collected in 1983, is labeled as ‘stillborn’ in the museum database, but its external morphology closely resembles that of a full‐term newborn panda (Fig. S1). The other panda specimen (USNM 464985) does not have developmental stage specified in the museum database. However, it was collected in 1987, 3 days after a ‘neonate’ panda specimen in the database (USNM 464984; the condition of this specimen was too poor to study) and together they were most likely cubs born in the same litter. Both specimens were therefore obtained at or shortly after birth and, as their external morphology resembles neonatal pandas, we estimate the two specimens are suitable representatives of giant panda at parturition.

μCT and 3D segmentation

Each specimen was scanned with a Nikon XTH 225 ST scanner at the Shared Materials Instrumentation Facility at Duke University (Table S2). One of the polar bears (USNM 282320) required two scans to capture the full body, and the separate TIFF stacks were stitched together with Fiji (Schindelin et al. 2012). The TIFF stack produced from scanning was imported into 3D segmentation software Avizo (8.1.1, Visualization Sciences Group) for volume rendering and surface mesh reconstruction. The Magic Wand and Paint Brush functions in the segmentation editor were used to extract skeletal elements from surrounding soft tissues and create a smooth surface mesh of the full skeleton. Tiff stacks and surface meshes for all specimens in this study are available at MorphoSource (https://www.morphosource.org/index.php; individual media numbers in Table S2).

Comparative analysis of degrees of ossification

After surface mesh generation, we recorded the presence/absence of all major cranial and postcranial bones examined during the fetal development of a beagle (Evans, 1993; Table S3). We also scored the development of nine major cranial sutures (sagittal, coronal, squamosal, lamboidal, metopic, intermaxillary, median palatine, transverse palatine, mandibular symphysis and basilar sutures) with a numerical system: 0 = an open suture with no contact between two bones; 1 = a partially closed suture, where two bones are superficially in contact in some areas but not others; 2 = a closed suture with suture lines, where two bones are superficially in full contact. All scores were added to produce an aggregate value defined as the suture index (SI; Table S5). In the post‐cranium, we assessed the relative degrees of ossification of individual skeletal elements based on its overall completeness and the presence/absence of key osteological landmarks outlined in Tables S6−S13.

Results

Cranium

All ursine bear (i.e. excluding the giant panda) and outgroup neonates have well‐ossified skulls with short faces and domed neurocrania (Fig. 3). The skulls do not appear enlarged relative to the body length. All major cranial bones are present at the time of birth. No individual interparietal ossification centers were observed at the posterior of the skulls (Koyabu et al. 2012), thus it appears that the interparietals had joined with the supraoccipitals before birth in all taxa examined.

Figure 3.

Left lateral and ventral views of (a, b) polar bear neonate (Ursus maritimus, USNM 240854), (c, d) domestic dog neonate (Canis familiaris, NCSCVM 1), (e, f) giant panda neonate (Ailuropoda melanoleuca, USNM 464985) and (g, h) fetal domestic dog (C. familiaris, NCSCVM 2). as, alisphenoid; bo, basioccipital; eo, exoccipital; fr, frontal; hy, hyoid apparatus; ma, maxilla; md, mandible; na, nasal; ot, otic bones; os, orbitosphenoid; pa, parietal; pe, petrosal; pm, premaxilla; so, supraoccipital; sq, squamosal; ty, tympanic rings; zy, zygomatic.

All taxa have open mandibular symphyses and basilar sutures at the time of birth. However, there is otherwise a wide range among the ursine bears and outgroups in the degree of suture closure (Table S5). For example, the male grizzly bear neonate (USNM 259097) and the neonatal dog have all the other major cranial sutures closed (SI = 16), while many such as the male sloth bear, the black bear, one of the polar bears and the arctic fox, have only a few closed sutures (SI 2−6). The suture development in the newborn dog observed here is consistent with the study of cranial suture development in terrier and Pekingese neonates (Starck, 1962).

Although both giant panda specimens had the upper neurocrania and occipital bones removed during necropsy, the occipital bones attached to the neonate specimen indicate that they are ossified at parturition. The remainder of the skulls appear more slender and more elongated in lateral view compared with other bears. All major cranial bones have ossified except for any of the hyoid bones (Fig. 3e, f), which shows partial ossification in all other taxa, including the fetal dog (Fig. 3). All examined cranial sutures in both giant pandas are open, a condition only observed in the fetal dog. A 45‐day beagle also has all major cranial sutures open (Evans, 1993).

Axial skeleton

All taxa have ossified vertebrae, ribs and sternebrae. The numbers of vertebrae and sternebrae are variable among species. No clear developmental distinction was noted between different sections of the vertebral column, thus the level of ossification was uniform throughout the vertebral column. No fusion is present between neural arches or between neural arch and centra, and ribs are not attached to the vertebral column. All sternebrae are unfused as well. No sternebrae were observed in the wild dog specimen, as they were presumably removed during necropsy.

The degree of axial skeleton ossification is generally similar between ursine bears and other carnivoran outgroups. Individual vertebral segments such as neural arch and centrum and distal ends of ribs are well ossified (Figs S2 and S3). In the giant panda and fetal dog, on the other hand, the morphology of individual vertebral elements is still largely undifferentiated, and the distances between individual vertebral elements are greater compared with other carnivorans given the relatively small sizes of individual ossification centers. In the giant panda, the proximal and central sections of the ribs are ossified, but the distal ends of the ribs remain poorly ossified. A similar condition is seen in the fetal dog (Fig. S4).

Pectoral and pelvic girdles

All ursine bear and outgroup neonates except the wild dog have well‐ossified scapulae, with complete scapular borders, acromia and scapular spines (Fig. 4). The glenoid cavities are well ossified with flat articular surfaces. Except for the domestic dogs, all taxa lack clavicles. In the pelvic girdle, all ursine bears and outgroup neonates except the wild dog have well‐ossified ilia, pubes and ischia with complete margins. In Lycaon, the scapular and pubic margins are still incomplete. No ursine bear or outgroup specimen exhibits fusion between individual pelvic bones and, in all taxa, the acetabulae are unfused at the time of birth (Fig. 4).

Figure 4.

pectoral and pelvic girdles of (a, b) Ursus arctos, (c, d) neonatal Canis familiaris, (e, f) Ailuropoda melanoleuca and (g, h) fetal C. familiaris. ac, acromion; ca, calcaneus; fe, femur; fi, fibula; hu, humerus; il, ilium; is, ischium; of, olecranon fossa; op, olecranon process; pu, pubis; ra, radius; sb, scapular border; ss, scapular spine; ta, talus; tb, tibia; tt, tibial tuberosity; ul, ulna. Each scale bar denotes length of 1 cm.

The scapulae of the giant pandas are poorly ossified, with missing acromia, incomplete scapular spines and dorsal scapular borders. The cranial and caudal borders of the scapulae are also incomplete. The degrees of ossification in the pelvic girdle resemble those in the pectoral girdle. The iliac crests, pubic and ischial margins are still largely incomplete. Like all the other carnivorans, individual pelvic bones are unfused, with unfused acetabulae. Overall, the giant panda’s degree of ossification in the pectoral and pelvic girdles is most similar to that of the fetal dog, although the latter lacks an ossified pubis (Fig. 4).

Appendicular skeletons

All ursine bears and outgroup neonates have robust and well‐ossified forelimb and hindlimb skeletons with little variation in the amounts of ossification. All long bone shafts have well‐ossified proximal and distal ends, though we observed no epiphyseal ossification in any of the long bones (Fig. 4). In the forelimb, the humeri lack humeral heads and distal epicondyle ossification. The olecranon fossae are well defined. The ursine bear neonates have well‐defined lateral supracondylar ridges relative to the outgroup neonates. The ulnae have visible olecranon and coronoid processes. In the hindlimb, the lesser trochanters are barely visible on the femora, while femoral heads, femoral necks and greater trochanters have not yet ossified. The tibial tuberosities are prominent (Fig. 4).

Like other areas of the skeleton, the degree of appendicular ossification of the giant panda lags behind other carnivoran neonates and resembles the fetal dog. All long bones are present, but the proximal and distal ends of the shaft are poorly ossified, and we found no epiphyseal ossification (Fig. 4). In the forelimb, the humeri lack well‐defined olecranon fossae, and no radial or coronoid fossae are visible. The ulnae are missing the olecranon and coronoid processes. The hindlimbs also lack major osteological landmarks found in other carnivoran neonates (Fig. 4). In terms of the overall robustness and degree of ossification, we noted no major differences between the fore‐ and hindlimb skeletons in any taxon studied.

All of the neonates examined in this study have ossified metacarpals, metatarsals, digits and claws at parturition. We observed no carpal bone ossification in any of the taxa, and the extra ‘thumb’ (radial sesamoid; Endo et al. 1999) is not found in the giant panda manus. The degree of ossification in tarsal bones is variable between species. All ursine bears only have rudiments of calcanea ossified at the time of birth, whereas all tarsal bones are absent in the giant panda. The neonatal dog, wild dog, coati and the red panda neonates both have well‐ossified calcanea with rudiments of astragali started to form, whereas the arctic fox has only prominent calcanea but no sign of talus development. Even the fetal dog, with poorly ossified digit bones in general, has an ossification center of the calcanea already visible.

Dentition

All ursine bears and outgroups have most deciduous tooth crowns mineralized inside alveolar bones (Fig. 5), with three incisors and one canine per quadrant in both upper and lower jaws at birth. The number of premolars is variable between individuals and between species (Table S12). In contrast, both giant panda specimens have no sign of crown mineralization at parturition (Fig. 5). In ursine bears and outgroups, the lower first incisors have started to erupt above the alveolar bones (Fig. 5; Table S13), and all neonates except the neonatal dog have upper first incisors beginning to erupt as well. The progress of eruption in other teeth is more variable between individuals and species. Most ursine bears have the lower second incisors beginning to emerge, and both sloth bears have both upper and lower second incisors started to erupt. Among the outgroups, the neonatal dog has all lower incisors starting to erupt, and the arctic fox and coati have both upper and lower incisors in the process of eruption. The wild dog has the upper first and second incisors and all lower incisors starting to erupt. The red panda has upper I1 and I2, all lower incisors and upper and lower PM2 and PM3 started to erupt as well (Table S13).

Figure 5.

Primary dentition of (a) Ursus arctos, (b) neonatal Canis familiaris, (c) Ailuropoda melanoleuca and (d) fetal C. familiaris. incisors, : canines, : premolars. Each scale bar denotes length of 1 cm.

Discussion

All ursine bear neonates examined here closely resemble caniform outgroup neonates in degree of ossification at parturition, except for minor differences in clavicle and talus ossification. Moreover, we found no major interspecific variation in skeletal maturity within ursine bear neonates, although there are some differences in cranial suture development, tarsal ossification and deciduous dentition eruption progress among individuals and species. Cranial suture development can be highly variable among carnivoran species during postnatal development (Goswami et al. 2013). Similarly, timing of dental eruption has been shown to vary among ursine bears (Marks & Erickson, 1966; Kenny & Bickel, 2005; Fosse & Cregut‐Bonnoure, 2014) and possibly within carnivorans in general. Therefore, we conclude that ursine bears are no more altricial than their close carnivoran relatives with respect to both cranial and postcranial skeletal development despite the fact that they are born at an exceptionally small body size relative to the mother.

In contrary, giant pandas are born with a level of skeletal maturity more embryonic than that of other caniform neonates, including all other ursid neonates examined here. Specifically, the giant panda perinates differ from other caniform neonates in having all cranial sutures open, no hyoid ossification, no primary dentition calcification, poorly ossified axial skeleton and pectoral and pelvic girdles, no tarsal ossification and poorly ossified postcranial long bones. The lack of primary dental mineralization and hyoid ossification is most similar to a 42–45‐day fetal beagle, which represents about 70−75% of total beagle gestation (Tables S3 and S4; Evans, 1993). With respect to the skeletal system, giant pandas are truly altricial relative to other carnivorans at birth. Nonetheless, the giant panda ossification level still conforms closely to the trajectory of other carnivorans at birth, suggesting that skeletal development in giant pandas is similar to other placental mammals. We found no evidence of any sequence heterochronies that characterize marsupial development, such as delayed neurocranium development relative to the oral‐facial apparatus (Smith, 2006), accelerated development of anterior axial skeleton (Weisbecker et al. 2008), and postponed ossification of the hindlimb and posterior axial skeleton (Weisbecker et al. 2008).

One caveat to our conclusions is that we only examined the skeletal system, and soft tissue development may exhibit different patterns in ursids. For example, Starck (1956) reported that the brain of one neonatal polar bear was notably less developed (‘lissencephalic’) relative to canids and procyonids at birth. Specifically, he reported that the cerebrum is relatively small, and the cerebral surface is completely smooth and lacking any sulci or gyrus development in the newborn polar bear. In contrast, newborn canids have a relatively much larger cerebrum with more complex surface sulci development (Starck, 1956). Given the small sample size of Starck (1956), we cannot affirm if his observations on nervous system development represent only a pathological case, which may be unlikely, nor can we determine if his observations apply to other bears or reflect differences in adult morphology. Currently, we have no data on newborn ursids’ nervous system or other organ systems such as the lungs, which exhibit different trajectories in marsupial mammals relative to placentals (Smith & Keyte, 2018).

Skeletal development of Ursidae and the unusual story of bear life history

Among placental mammals, a negative allometric relationship exists between neonatal and maternal body mass (Leitch et al. 1959). While individual species do sometimes deviate from this relationship in both directions (Fig. 1), Ursidae is the only family that uniformly produces disproportionately small newborns relative to the mother. What factors contributed to this extreme neonatal−maternal size difference in Ursidae? Two alternative but not necessarily mutually exclusive explanations have been proposed. First, it may be that neonate body size evolved to be disproportionately smaller than the typical carnivoran scaling relationship. Alternatively, adult body size may have increased while the absolute body size of the neonate remained largely constant. Here we examine existing hypotheses and evaluate the likelihood of these scenarios in the evolutionary history of bears.

One of the most commonly cited hypotheses attributes the unusually small size of bear cubs to maternal energetic constraints during gestation through winter (Ewer, 1973; Ramsay & Dunbrack, 1986). This hypothesis argues that during gestation under hibernation, the mother is fasting and draws energy from her fat storage. The fetus, however, is unable to catabolize maternal free fatty acids and, instead, the mother is forced to use her own proteins for alternative fetal gluconeogenesis pathways. Because maternal proteins are consumed during hibernation, the mother is forced to shorten gestation and produce relatively immature and small young. Once the cub is born, the mother can continue nourishing the cub through prolonged lactation, which utilizes her fat reserves and food intake with no significant risk to maternal health (Ramsay & Dunbrack, 1986; Gershelis, 2004). It is argued that extant bears that no longer live in a seasonal environment, such as the giant panda, produce small neonates because of phylogenetic constraints (Ramsay & Dunbrack, 1986; Gershelis, 2004).

Our comparisons of holarctic and sloth bear neonates revealed no major variation in skeletal maturity or size that may reflect any hibernation‐imposed energy constraints on fetal development. The fact that giant panda perinates are the smallest and most altricial of all carnivorans studied here contradicts the above hypothesis, as adult pandas are active year‐round and do not fast during gestation (Gershelis, 2004; Nie et al. 2015b). However, at this stage we cannot affirm if other organ systems will reveal potentially different developmental patterns among the neonates of hibernating and non‐hibernating ursids.

Phylogenetic data also challenge the above hypothesis for small neonatal size (Ramsay & Dunbrack, 1986). If non‐hibernating bears like the panda produce relatively small cubs due to phylogenetic constraint (Ramsay & Dunbrack, 1986), the traits of hibernation and small cub size should have evolved together in the last common ancestor of Ursidae. The dawn bear Ursavus elmensis (Stirling & Derocher, 1990) from Early Miocene Europe (Kurtén, 1976) is the earliest known Ursid, and several features suggest that Ursavus may not have exhibited seasonal denning behaviors seen in the holarctic bears today. First, it most likely did not undergo extended hibernation as Ursavus lived in a moist, subtropical environment (Kurtén, 1976), and any seasonal resource fluctuation would be moderate. Rather, it probably engaged in year‐round foraging like many tropical or temperate bears today. Moreover, the small size of Ursavus (similar to a fox terrier; Stirling & Derocher, 1990) means it has a much higher metabolic demand per body weight than its bigger‐bodied descendants today. Fasting during gestation in such a small animal would possibly jeopardize the survival of both the mother and the fetus. Finally, it must be noted that other mammals that undergo extensive fasting during gestation (e.g. some large mysticete whales; Ramsay & Dunbrack, 1986) do not appear affected by similar energetic constraints and still give birth to relatively large and highly precocial young, also under remarkably short gestation lengths (Roston et al. 2013). Therefore, the hibernation‐imposed constraint hypothesis is unlikely to be a general explanation for the small size of neonates in bears.

Almost all bears today are seasonal breeders. Many aspects of ursid reproductive physiology correlate strongly with seasonal changes in climate and resource availability (Gershelis, 2004; Spady et al. 2007), and litter size, reproductive intervals and cub survival depend heavily on maternal nutrition intake (Gershelis, 2004). For instance, increasing body fat content correlates with earlier parturition date, longer lactation time and larger cub size at den emergence in grizzly bears (Robbins et al. 2012). Female grizzly bears with insufficient body fat reserve at the beginning of hibernation may forgo embryo implantation completely (López‐Alfaro et al. 2013). In polar and American black bears, female body fat content also strongly influences implantation likelihood and cub size at den emergence (Robbins et al. 2012). Thus, instead of any physiological constraints imposed by gestation during hibernation, resource fluctuation may place an indirect constraint on reproduction in bears. The combination of exceptionally large adult body size, a generally omnivorous diet and seasonal environments may potentially limit the reproductive energy budget for females and facilitate selection for small initial reproductive investment. Testing this specific hypothesis, however, will require more information on ecology and the reproductive biology of tropical bears.

Other ecological factors may contribute to the evolution of small size neonates. For example, Sibly & Brown (2009) identified juvenile predation risk as a strong indicator of placental reproductive output: species often give birth to small litters of highly precocial neonates (e.g. terrestrial perissodactyls and cetaceans) in an environment with high juvenile predation risk, whereas species that give birth in a protected environment tend to produce a relatively large litter of small young. Bears may afford to produce relatively small and altricial offspring as usually parturition takes place in dens, where both the mother and the cubs are sheltered from the outside environment (Gershelis, 2004; Sibly & Brown, 2009) and, aside from conspecifics, bears have few natural predators.

On the other hand, it is possible that during the course of bear evolution the adult body size has increased considerably whereas the absolute neonatal size has remained largely unchanged. From an ancestral body size of 1.2 kg (the last common ancestor of Arctoidea; Fig. 2), bears underwent a dramatic increase in body size over the last 20 million years (Finarelli & Flynn, 2006). Today, members of the Ursidae are on average heavier than all other carnivoran families (Gittleman, 1985), with a wide range of size distribution from 27 kg in female sun bears to over 800 kg in large male polar bears (Stirling & Derocher, 1990). Numerous benefits accrue with a large body size, including better thermoregulation abilities, larger home range coverage, and more fat reserve to survive food shortage (Clutton‐Brock & Harvey, 1983). These traits may be especially advantageous in the highly seasonal environment of Pliocene and Pleistocene North America and Eurasia, in which modern ursid ancestors lived (Spady et al. 2007).

We hypothesize that as natural selection was actively facilitating larger adult biomass, the neonatal body size of bears lagged behind and remained similar to the conditions in other carnivoran outgroups. Indeed, over evolutionary time, adult body size is usually one of the first and fastest evolving morphological traits over other life history parameters (Gittleman, 1994; Webster et al. 2004). Webster et al. (2004) suggested that carnivoran families that have undergone recent body size increase have smaller litter size and neonate mass for their current body size. This legacy effect of ancestral body size on current life history may be particularly pronounced in Carnivora, because as an order it has a higher average rate of body size evolution than the mammalian average (Venditti et al. 2011).

In the case of bears, the small size ancestral body size, seen for example in Ursavus, may exert a residual impact on the neonatal size of modern bears. The tendency to retain ancestral life history characters would thus explain neonates that are disproportionately smaller relative to the adult body size (Fig. 1). Although bears are not the only carnivoran family with recent size increase (Webster et al. 2004), nor have they the fastest rate of body size evolution among carnivorans (fig. S2 in Venditti et al. 2011), they have arguably the largest order of magnitude change in body size from the ancestral state among Carnivora. The unusually small size of bear cubs relative to the mother may be a result between natural selection’s strong emphasis on large adult body mass and the legacy effect of the small body size in bear ancestors. Interestingly, in the other obligate herbivorous member of Ursidae, the cave bear (Ursus spelaeus; Veitschegger, 2017), body size seems to be evolving at a faster rate than brain size, the latter of which also directly correlates with gestation time, neonatal mass and litter size (Finarelli, 2010).

To summarize, hypotheses that link small neonatal size in bears to gestation energetic constraints during hibernation are not supported by current data on ursid newborn skeletal maturity or ursid phylogenetic patterns. Instead, we believe that the small neonatal/maternal size ratio may largely relate to the relatively rapid and extreme increase in body size in bears, with the retention of primitive life history parameters. This scenario requires no special modification of bear fetal development and instead relies on the evolution of large adult size, an observation well documented by the fossil record (Kurtén, 1976), and a lag in changes in the scaling relationship between carnivoran life history parameters and body size (Webster et al. 2004).

Life history factors in skeletal development of the giant panda

The degree of ossification of the giant panda at or around parturition markedly deviates from the neonates of other ursids and carnivoran outgroups. The giant panda perinate’s skeletal maturity level is most similar to a 42−45‐day‐old beagle fetus, or 70–75% of total beagle gestation with parturition on Day 60 (Evans, 1993). The giant panda, therefore, is born with an embryonic state in the skeleton relative to other caniforms, including other ursids. Compared with other bears, a giant panda fetus has a much shorter post‐implantation gestation time. Implantation is difficult to pinpoint in animals with delayed implantation but, based on peak urinary progesterone levels, it was inferred to occur roughly 30 days before parturition in giant pandas (Zhang et al. 2009). In comparison, timing of implantation was estimated to be approximately 60 days before parturition in grizzly, polar and black bears (Tsubota et al. 1987, 1998; Derocher et al. 1992). No scaling relationship between post‐implantation gestation length and adult body size among ursids is known, although Zhang et al. (2009) reported a positive correlation between the length of post‐fetal detection period and neonatal weight in the giant panda. If inferences on implantation time in various ursids are correct, these data suggest that the proximal explanation of notably small and altricial giant panda neonate is an extremely truncated developmental period relative to other bears.

Despite the shorter post‐implantation gestation time and the very small and relatively altricial neonate, the giant panda does not appear to have accelerated the development of any particular skeletal element as seen in marsupials (Clark & Smith, 1993; Smith, 1997; Weisbecker et al. 2008). Rather, the perinatal giant panda’s level of skeletal maturity still falls on a carnivoran‐like skeletal development trajectory, and its developmental pattern appears to be typical of placental mammals. It is simply born at an earlier point on the developmental trajectory. The panda may indeed be one of the most altricial placental mammals, but confirmation of its unique fetal development characteristics will require complete longitudinal sampling of panda fetuses.

Besides the exceptionally low neonatal−maternal mass ratio even among bears, the giant panda also has a slower postnatal growth rate relative to adult body weight among carnivorans (Gittleman, 1994). Various physiological and/or ecological factors may have contributed to these unique life history characteristics of the panda. One such candidate is the basal metabolic rate (BMR). Compared with other mammals of similar body size, the giant panda has significantly lower daily energy expenditure (Nie et al. 2015a; although see Fei et al. 2016 for contradictory BMR results). However, among therian mammals, BMR does not correlate with life history unless body size is accounted for (Harvey et al. 1991), and an unusual neonatal−maternal mass ratio is not observed in other mammals with exceptionally low BMR (Nie et al. 2015a).

A low‐nutrition diet is another explanation frequently invoked for the unusual features of giant panda life history (Gittleman, 1994); 99% of a panda’s diet consists of bamboo (Nie et al. 2015b), and the adaptation of bamboo foraging extends back to late Pliocene (Jin et al. 2007). Bamboo is known to have low caloric and protein content but is high in lignin and cellulose (Wei et al. 2014; Nie et al. 2015b). It has also been reported that female wild giant panda’s diet is consistently low in Ca : P ratio during the year except for a period shortly before giving birth (Nie et al. 2015b). The lack of sufficient calcium in the maternal diet may have negative consequences on fetal development during gestation, particularly in utero skeletal growth. However, recent reports suggest that pandas exhibit seasonal foraging migration and tend to target protein‐rich bamboo components (Nie et al. 2015b), and their macronutrient ratio (i.e. the relative percentage of protein, carbohydrate and fat) more closely resembles other carnivorans rather than herbivores (Nie et al. 2019). This ability to selectively consume certain macronutrients may explain why the giant panda has multiple craniofacial and dental adaptations for bamboo foraging (Wei et al. 2014), while its digestive tract and microbiome do not appear specialized for a herbivorous diet (Xue et al. 2015, but see Wei et al. 2014). At this stage, it is premature to draw a link between diet and the unique degree of ossification observed in giant panda neonates here. Like metabolism, dietary specialization alone does not predict life history variation among placental mammals (Gittleman, 1994). Moreover, the neonates of the cave bear resemble other ursine bears in degrees of ossification (Ehrenberg, 1973), although the neonatal−maternal mass ratio in that species is unknown (Fuchs et al. 2015).

Conclusion

In summary, we found that most ursine bears resemble Caniformia outgroups in degree of ossification at birth and cannot be considered altricial relative to other caniforms. On the other hand, bears have some of the lowest neonatal−maternal mass ratios among placental mammals, even after the negatively allometric relationship between neonatal and adult body size was accounted for (Fig. 1). No major interspecific variation of degree of ossification was observed between ursine bears that undergo hibernation and those that remain active year‐round. The giant panda, in contrast, demonstrates a more altricial skeletal development at or around parturition compared with other carnivorans, and a newborn panda is more similar to a beagle at about 70% of development in its degree of ossification. Overall, all ursid neonates exhibit highly placental‐like skeletal developmental patterns, and we observed no heterochronic shifts in the skeletal system of bears that characterize marsupial development.

Existing hypotheses on the extreme size contrast between bear neonate and adult mostly focus on direct or indirect physiological constraints on fetal development imposed by environmental seasonality and gestation during hibernation (Ramsay & Dunbrack, 1986). However, we found no variation in ursid skeletal development that might correlate with any such constraints. Instead, we suggest that bear neonates may appear disproportionately small compared with the mother because the adult bear body size has considerably increased in the last 20 million years of bear evolution (Kurtén, 1976), and the current size of bear neonates may be correlated partially with the small ancestral size of Ursidae (Webster et al. 2004).

Current explanations for the unusual life history of the giant panda may be inadequate, as neither metabolism nor an herbivorous diet can sufficiently account for the panda neonate’s extremely small size and embryonic skeletal maturity. Given its close phylogenetic affinity with other ursine bears and adaptations to an herbivorous diet like the giant panda, the cave bear may be a promising candidate to further shed light upon the relationship between phylogeny, herbivory and reproductive physiology among Ursidae.

Competing interests

The authors declare no competing interests.

Author contributions

P.L. and K.K.S. conceived the research; P.L. performed μCT and collected data; and P.L. and K.K.S wrote the paper.

Supporting information

Table S1. Gestation length of some large‐bodied fissiped carnivorans.

Table S2. Sex, developmental state, crown‐rump length, weight and selected μCT scanning parameters of all specimens used in the present study.

Table S3. Reference ossification sequence of a beagle (Evans 1993).

Table S4. Reference dental development sequence of a beagle (from Evans, 1993).

Table S5. Cranial suture development of caniform taxa in the present study.

Table S6. Axial skeleton ossification of caniform taxa in the present study.

Table S7. Pectoral girdle ossification level of caniform taxa in the present study.

Table S8. Forelimb ossification level of caniform taxa in the present study.

Table S9. Pelvic girdle ossification level of caniform taxa in the present study.

Table S10. Hindlimb ossification of caniform taxa in the present study.

Table S11. Autopod ossification of caniform taxa in the present study.

Table S12. Dental development of studied taxa at parturition.

Table S13. Dental eruption level of studied taxa at parturition.

Fig. S1. External morphology of representative species in this study.

Fig. S2 Skeletal reconstruction of: (a) Grizzly bear (Ursus arctos, USNM 259097); (b) Grizzly bear (Ursus arctos, USNM 256787); (c) Polar bear (U. sp. × maritimus, USNM 282320); (d) Polar bear (U. maritimus, USNM 240854); (e) Sloth bear (Melursus ursinus, USNM 399497); (f) Sloth bear (M. ursinus, USNM 399496); (g) American black bear (U. americanus, USNM 217911). Each scale bar denotes the length of 1 cm.

Fig. S3. Skeletal reconstruction of: (a) neonatal domesticated dog (Canis familiaris, NCSCVM 1); (b) Coati (Nasua narica, USNM 255308); (c) red panda (Ailurus fulgens, USNM 541401); (d) African wild dog (Lycaon pictus, USNM 395688); (e) Arctic fox (Vulpes lagopus, USNM 14031). Each scale bar denotes the length of 1 cm.

Fig. S4. Skeletal reconstruction of: (a) giant panda (stillborn) (Ailuropoda melanoleuca, USNM 546337); b) giant panda (neonatal) (A. melanoleuca, USNM 464985); (c) fetal domesticated dog (Canis familiaris, NCSCVM 2). Each scale bar denotes the length of 1 cm.

Data availability statement

Raw Tiff stacks and surface meshes for all specimens in this study are available at MorphoSource (https://www.morphosource.org/index.php; individual media numbers in Table S2). All raw data in this study are available at Dryad (https://datadryad.org/stash/share/EhhdvI5faKu2fsn0ewtrS-5Ukec6dRGaK0GBb4lQlKU).