BP, Evolutionary Studies Institute (formerly, Bernard Price Institute) of the University of the Witwatersrand (Johannesburg, South Africa); BP MN, Bernard Price Museum Number; NMQR, National Museum (Bloemfontein, South Africa).

Cluver [54], Grine et al. [42], Botha et al. [48], and Botha [46] have already provided comprehensive studies on the skeleton and ontogeny of Lystrosaurus. This contribution thus focuses on the new anatomical data obtained from comparing BP/1/9332, BP/1/4011, and NMQR 3636.

Based on their basal skull length (length from the basioccipital condyle to the anterior edge of the premaxilla), these specimens are the three smallest Lystrosaurus ever found [42]. Specimen NMQR 3636 is the smallest, with a basal skull length of 34.5 mm. The skull of BP/1/4011 is 43.0 mm long and that of BP/1/9332 is 44.0 mm long. None of the specimens preserves an egg tooth (caruncle). An egg tooth is normally present in neonate monotremes [4,55], thus it is possible that, here, it was lost, not preserved, or damaged during preparation. Both BP/1/4011 and BP/1/9332 possess small unerupted tusk buds visible in their maxillary alveolae, whereas NMQR 3636 has empty alveolae (Fig 2b, c). The mesethmoid bone, which supports the olfactory bulbs in life, is completely ossified in both BP/1/4011 and BP/1/9332. It is displaced in the former, whereas it is preserved in situ in the latter, suggesting that the median septal cartilage that supported it was better developed in BP/1/9332. Specimen NMQR 3636 preserves no mesethmoid. It may have been lost or completely cartilaginous and therefore not preserved. The occipital and basicranial bones, including the supraoccipital, tabulars, prootics, opisthotics, exoccipitals, and basioccipital, are loose and displaced in both NMQR 3636 and BP/1/4011. As a result, the occipital surface of the skull of these two specimens has been crushed, which gives it a posteriorly tapering profile in lateral view (Fig 2a). In contrast, the occipital bones are preserved in situ and in articulation in BP/1/9332, and its skull is undeformed as a result (Fig 2a). In Lystrosaurus, the co-ossification of cranial bones begins anteriorly and proceeds posteriorly, so that the occipital bones are the last to co-ossify [54]. This suggests that the development of BP/1/9332 was more advanced than that of the other two. Sclerotic ring plates are preserved in all three specimens, demonstrating that these bones formed very early in ontogeny.

The lower jaw in NMQR 3636 displays an unfinished intermandibular symphysis. Ventrally, on the mandible, a deep notch excavates the splenials at the midline, showing that the symphyseal suture between the two paired bones is incompletely co-ossified (Fig 2d). The edges of this notch are smooth on synchrotron images, which demonstrates that this feature is not due to post-mortem damage or erosion (Fig 2e). It likely accommodated Meckel’s cartilage [56]. Dorsally, the inter-dentary suture is still clearly open in NMQR 3636 (Fig 2e). In contrast, BP/1/4011 and BP/1/9332 reveal splenials that are completely co-ossified and the inter-dentary suture is mostly closed, although a midline suture is still visible at mid-height on the symphysis (Fig 2d, e).

Two of the three specimens preserve postcrania, i.e., BP/1/9332 and NMQR 3636. Since neither specimen has duplicate elements, each likely represents a single individual. Specimen NMQR 3636 is preserved in a curled-up position in a 73.0 mm long and 5.5 mm wide nodule. It preserves most of the vertebral column up to the last caudal vertebra, the left humerus, most of the right forelimb, including right humerus, radius, ulna and undetermined and disarticulated elements of the right manus, as well as most of the pelvis and the left femur (Fig 3a, b, d). The carpals are spherical due to the incomplete ossification of the hand. Both the humerus and femur have unossified epiphyses (Fig 3a-c). On the vertebrae, the centra and neural arch are separated on all vertebrae, and many vertebral elements are loose and displaced. In the pelvic region, the sacral vertebrae and connecting ribs are loose and disarticulated, resulting in the absence of a co-ossified sacrum. A few of the left and right dorsal ribs are preserved and are tightly packed around the skeleton.

Specimen BP/1/9332 is preserved in a splayed-out position (Fig 3e), typical of most Lystrosaurus articulated skeletons found in the Karoo Basin [49]. All bones of the skeleton are almost perfectly articulated, as synchrotron images show that no loose elements are preserved in the surrounding matrix. The most distal elements were lost to erosion. The forelimbs preserve both humeri and the proximal half of the left radius and ulna. Both hind limbs are missing except for a small fragment of the right femur (Fig 3e). The epiphyses of the humeri demonstrate a more advanced degree of ossification than in NMQR 3636 (Fig 3c). As in NMQR 3636, the centra and neural arches of the vertebrae are unfused. The pelvis is preserved in full anatomical articulation, which suggests a more advanced stage of development of the cartilages in this specimen compared to NMQR 3636 (Fig 3e). The sacrum includes six sacral vertebrae with ribs connecting them to the pelvic bones.

Osteohistological differences are minor between all three individuals (S6 Data). The bone microstructure is mostly spongy. No hatchling line is visible in the long bones of NMQR 3636 and BP/1/9332, consistent with previous observations made in juvenile Lystrosaurus specimens [46,47,57].

Working on the hypothesis that NMQR 3636 is the skeleton of a Lystrosaurus preserved in its egg (see discussion below), its estimated volume would be 115 cm³ and its mass would be 115 g (assuming a density equal to that of water). Given the range of body masses considered here for Triassic Lystrosaurus (see material and methods), its estimated egg sizes are largely consistent with other egg-laying amniotes (Fig 4). They fall in the upper range of those of reptiles for the largest body mass considered, while they fall slightly above them for smaller body masses. Monotremes have comparatively smaller eggs than other amniotes because of the limited amount of yolk they contain, which is compensated for by post-hatching milk feeding [4,32]. The values for Kayentatherium [21] make it fall far outside other amniotes on the scatter plot, well below the values for reptiles and monotremes (Fig 4).

All three specimens have similar skull lengths (approximately 4 cm) and were thus almost the same age at the time of death, likely perinates given their very small size and low level of skeletal ossification. They can be organised in a growth sequence, beginning with NMQR 3636, which is likely at the youngest developmental stage, owing to its incomplete mandibular symphysis, absence of tusk buds, possible absence of ossified mesethmoid, and lesser degree of occipital bone co-ossification. Specimen NMQR 3636 also has the shortest skull. Specimen BP/1/4011 is the second least mature of the three specimens because its splenials are co-ossified at the mandibular symphysis (unlike in NMQR 3636), but its occipital and basicranial bones remain loose (unlike BP/1/9332). Its skull length is intermediate between the other two specimens. Skeleton BP/1/9332 is the most developed of the three, as it has the longest skull and its occipital bones are sutured together. Its humeral epiphyses are also in a more advanced state of ossification than those of NMQR 3636. Its splayed-out position shows it is not preserved in an egg and had moved around for some distance before dying (Fig 3e).

Specimen NMQR 3636 was therefore likely younger at the time of death than the other two specimens, making it the least developed of all known Lystrosaurus skeletons. The evidence that supports NMQR 3636 was in ovo when it died include the dimensions of the nodule in which it is preserved, which are consistent with that of a non-avian amniote egg (Fig 4), and the curled-up position of the skeleton that outlines an ovoid shape consistent with that of an egg (Fig 1). Elsewhere, finding a perinate individual in this position would be sufficient to reasonably hypothesise in ovo preservation in other fossil amniotes, even in the absence of a preserved egg shell (e.g., [7,28]); however, in the Permo-Triassic Main Karoo Basin, many early juvenile, adult, and subadult tetrapods have been found fossilised in this position (e.g., [15,20,58,59]). Thus, the curled-up position alone is not enough to conclude that NMQR 3636 was in ovo.

The disarticulation of the pelvis, sacral vertebrae, and ribs suggests that the pelvic bones and cartilage were too weak to support its weight, unlike those of the splayed-out and more mature BP/1/9332. This is consistent with an in ovo occurrence. Furthermore, in NMQR 3636, there is a large gap on the ventral side of the lower jaw where the two splenial bones are not fully co-ossified medially. The edges of this gap are smooth and undamaged, indicating that it is a genuine anatomical feature (Fig 2c, d), that likely accommodated Meckel’s cartilage in life. In other tetrapods with a beak, i.e., chelonians and birds, the symphysis of the lower jaw always closes in ovo, in the last third of embryogenesis [60–65]. Synchrotron images of in ovo embryos of the turtle Centrochelys sulcata show that the mandibular symphysis becomes fully co-ossified between 45 and 55 days of embryogenesis [66–68]. This supports that NMQR 3636 was not yet close to the hatching stage. Egg-laying mammals have unfused mandibles when they are born, but they are highly altricial because they are fed milk before they complete their development [4], whereas Lystrosaurus most likely did not produce milk [32,69]. It is unlikely that a Lystrosaurus hatchling with an unfused, partly cartilaginous intermandibular symphysis would have been able to feed itself and survive, as its jaws would have been too weak to process hard food items. Altogether, all the evidence supports the conclusion that NMQR 3636 is preserved in a curled-up, in ovo position.

Comparing egg sizes amongst amniotes reveals that Lystrosaurus laid relatively large eggs for a non-avian amniote, regardless of the body size considered (Fig 4). This is remarkable as the egg size reconstructed and compared here is a minimum value based on an incompletely developed and partly damaged embryo. Thus, the following conclusions are based on a conservative estimate of Triassic Lystrosaurus egg size.

In extant tetrapods, larger eggs generally correlate with precociality and higher levels of parental investment in raising the hatchling [70,71]. Accordingly, the relatively large size of the egg in Triassic Lystrosaurus compared to other egg-laying, terrestrial amniotes (Fig 4) aligns with the rich fossil record of adult-juvenile aggregations of late Palaeozoic and early Mesozoic synapsids that has been traditionally interpreted as illustrating the early evolution of parental care in this lineage [19–21,52,72–74]. Large egg size also correlates with low predation and low competition in extant species [71,75], which is consistent with the depleted terrestrial biodiversity after the EPME, in which early Triassic Lystrosaurus were the only medium-sized herbivorous species, and predators large enough to attack them (e.g., Moschorhinus, Proterosuchus) were scarce [37]. Larger eggs are also less prone to desiccation, particularly soft-shelled ones, because their surface area to volume ratio is low [76,77]. This would have been advantageous in the arid post-extinction environment [49,78].

Lystrosaurus early juveniles or neonates are usually discovered as single specimens, but rare groups of two or three individuals arguably belonging to the same age class have also been found [14,16,79,80]. The only exception is an aggregation of nine more mature juvenile Lystrosaurus specimens, which are disarticulated and thus likely gathered post-mortem [49,79]. This would imply a rather small to medium-sized clutch, which, combined with the relatively large size of the eggs, would be consistent with precociality [11], as is indicated by the early mobility of the neonate individual BP/1/9332 hypothesised above. In addition, this specimen and BP/1/4011 both have closed mandibular symphyses with only an incipient suture visible (Fig 2d) indicating that they would have been able to process hard food items. Larger eggs and precociality would have been advantageous in the post-EPME environment as they would have enhanced the exploitation of the scarce food resources soon after hatching [71]. Precociality would also be consistent with Triassic Lystrosaurus reaching sexual maturity early, a trait favoured by natural selection as juvenile mortality was very high [46–48].

Neonates of the genus Diictodon, a dicynodont taxon from the Permian that survived the end-Capitanian extinction, are also found as single individuals or in pairs [19], whereas known aggregations of juveniles of other dicynodont species are significantly larger [74,81,82]. This may be because small clutch size could have been a recurring survival strategy during biological crises in this clade. Alternatively, large aggregations of dicynodonts (including Lystrosaurus) often involve more mature juveniles and/or subadults [74,79,82], thus it is possible that social dicynodonts were gathering in increasingly larger groups as they became older. More data is needed to address these possibilities.

In extant egg-laying mammals— monotremes—the limited yolk supply is compensated for by lactation. As a result, their eggs are comparatively small (Fig 4), and the hatchlings are highly altricial [4,30,33]. Consequently, the reconstructed range of egg size presented here rules out the possibility that Lystrosaurus produced milk for feeding (Fig 4). In contrast, the egg size to body mass ratio of the non-mammalian tritylodontid cynodont Kayentatherium is very low (Fig 4). Given its large clutch size (38 individuals based on a specimen from the Early Jurassic), it has been proposed that Kayentatherium had an essentially reptilian-like reproductive biology [21]. This, coupled with evidence of dental wear in neonates, indicated that hatchlings were partially capable of processing hard food items and that adult female Kayentatherium likely did not lactate [24]. The current results nuance this interpretation, as the very small size of the neonates in the aggregation suggests that Kayentatherium had very small eggs compared to Lystrosaurus and other amniotes (Fig 4), more consistent with the size of the eggs in monotremes [4]. Dental wear at birth is found in some extant rodents, which shows that this trait is not incompatible with lactation [83]. Moreover, as proposed by Hopson [29], it is probable that primitive milk was provided alongside other food items to sustain the young in the first evolutionary steps towards mammalian lactation. Under these circumstances, dental wear in neonate tritylodontid cynodonts is not unexpected. The evolution of primitive milk production at the evolutionary root of the Mammaliamorpha, including tritylodontids, would be consistent with the timing of some mutations involving genes coding for caseins and vitellogenin proteins (for the production of milk and reduction of egg yolk, respectively), and MSX2, a gene involved in the formation of mammary glands [34,69,84–86]. Tritylodontids also likely had hair [87–89], which further supports that they had mammary glands, given the well-established genetic, phylogenetic, morpho-anatomical, and ontogenetic connections between the two traits [32,34,69]. Thus, the previous observations made on Kayentatherium are, here, reconciled with a possibly more mammal-like reproductive biology than previously envisioned, including lactation. They, in fact, support that Kayentatherium was indeed more mammal-like in this respect than the more basal Lystrosaurus.