Breeding Young as a Survival Strategy during Earth’s Greatest Mass Extinction

Studies of the effects of mass extinctions on ancient ecosystems have focused on changes in taxic diversity, morphological disparity, abundance, behaviour and resource availability as key determinants of group survival. Crucially, the contribution of life history traits to survival during terrestrial mass extinctions has not been investigated, despite the critical role of such traits for population viability. We use bone microstructure and body size data to investigate the palaeoecological implications of changes in life history strategies in the therapsid forerunners of mammals before and after the Permo-Triassic Mass Extinction (PTME), the most catastrophic crisis in Phanerozoic history. Our results are consistent with truncated development, shortened life expectancies, elevated mortality rates and higher extinction risks amongst post-extinction species. Various simulations of ecological dynamics indicate that an earlier onset of reproduction leading to shortened generation times could explain the persistence of therapsids in the unpredictable, resource-limited Early Triassic environments, and help explain observed body size distributions of some disaster taxa (e.g., Lystrosaurus). Our study accounts for differential survival in mammal ancestors after the PTME and provides a methodological framework for quantifying survival strategies in other vertebrates during major biotic crises.

nor the dicynodont skull they documented in the Katberg Formation, to be sufficient evidence to change the placement of the currently recognized PTB in the Karoo Basin. The radiometric date supports a latest Permian age (mid-Changxingian) 60 metres below the traditional PTB, which is consistent with the traditional placement of the boundary. As Gastaldo and Neveling 20 note, estimates of sedimentation rates are imprecise due to various assumptions adopted when using modern day sedimentation rates as analogues, as well as factors such as hiatuses and compaction, and so should be treated with caution. A radiometric date closer to the currently positioned PTB would be more appropriate when attempting to correlate the non-marine PTB with the marine record.
In addition to their radiometric date, Gastaldo et al. 18 recovered a partial dicynodont skull from the base of the Katberg Formation that they identified as a Permian "dicynodontoid," and they used this identification to argue against the synchroneity of the extinction fauna given its apparently young stratigraphic position. This argument is invalid on phylogenetic grounds. The dicynodont clade Dicynodontoidea includes lystrosaurid and kannemeyeriiform dicynodonts best known from the Triassic, as well as a paraphyletic assemblage of basal taxa from the Late Permian (i.e., the species previously assigned to the wastebasket taxon Dicynodon 21 ). Thus, the stratigraphic range of Dicynodontoidea extends from the Late Permian to near the Triassic-Jurassic boundary, when kannemeyeriiform dicynodonts are generally accepted to have become extinct, and a specimen identified only as a dicynodontoid cannot offer any biostratigraphic resolution below this roughly 60 million year time interval. We may assume that Gastaldo et al. 18 are using "dicynodontoid" as a shorthand for "basal dicynodontoid" (i.e., a Dicynodon-grade dicynodontoid, not a lystrosaurid or kannemeyeriiform), since this would imply a Late Permian age. If this is the case, their identification is questionable because their published photographs of the specimen do not present any characters that can be used to definitively identify it as a Dicynodon-grade dicynodontoid to the exclusion of Lystrosauridae or Kannemeyeriiformes. The specimen could just as easily represent Lystrosaurus, which would be completely consistent with an Early Triassic age and the traditional placement of the PTB. Finally, Gastaldo et al.'s 18 treatment of the specimen ignores the possibility of a previously unrecognised survival of a Permian dicynodontoid into the Early Triassic. The Permian Lystrosaurus species L. maccaigi is known only from the Permian in South Africa 22 , but it appears to have survived into the Early Triassic in Antarctica. Therefore, the survival of a Permian dicynodontoid into the Early Triassic would not be completely unprecedented 23,24 .
Based on these lines of evidence, we retain the traditional placement of the non-marine PTB in the Karoo Basin. This placement is supported by biostratigraphic research based on hundreds of well-provenanced specimens 1,11 , as well as work in other basins (e.g. Europe 25 , China 14,15 , Russia 26 , Australia 27 and Antarctica 28 ), and we consider it to represent the terrestrial faunal turnover associated with the PTME. Regardless of the final consensus, the results in our study will remain unaffected by the position of the PTB in the terrestrial realm. The purpose of this study was to examine life history changes during a massive vertebrate faunal turnover, which is represented by the extinction and origin of vertebrates associated with the currently defined PTB in the Karoo Basin.

Growth Marks
Detailed descriptions of the bone histology of the Permo-Triassic therapsids used in this study have been published elsewhere 29-32 and will not be repeated here. However, the occurrence of growth marks in these taxa requires a more detailed explanation. Growth marks include annuli and Lines of Arrested Growth (LAG). Annuli represent a temporary decrease in growth and contain slower forming bone tissues such as lamellar or parallel-fibred bone. They are either avascular or poorly vascularised and have few, flattened osteocyte lacunae. LAGs indicate a temporary, but complete cessation in growth and are represented by a cement line. Experiments on living taxa have shown that growth marks are deposited annually 33,34 . In Permian therapsids ( Supplementary Figs 1-3), cortical bone is punctuated by numerous growth marks (usually three or more) that interrupt the bone tissues prior to the onset of slower-forming bone tissues and is interpreted here as evidence of multi-year growth to somatic and reproductive maturity. Exceptions to this pattern can be found in Triassic therapsids  where two, but usually one or no growth marks are found before a decrease in growth rate is observed.
The Early Triassic dicynodonts Myosaurus gracilis, Lystrosaurus murrayi, and L. declivis deserve special mention. Myosaurus is a tiny emydopoid (BSLmax= 46.22 mm) from the Lystrosaurus Assemblage Zone (LAZ). A humerus was sectioned from the second largest known Myosaurus specimen (BP/1/4269) from South Africa. A larger specimen is known from Antarctica, but we do not consider it here due to possible taxonomic differences with the South African species. As this is the first time that the bone histology of Myosaurus has been examined, we describe it in some detail here ( Supplementary Fig. 4a). This taxon exhibits fibrolamellar bone with numerous, haphazardly arranged, globular osteocyte lacunae in a woven-fibred bone matrix. The bone tissue is relatively well-vascularised (6.2%) and the vascular canals are arranged in a reticular network. These features indicate high bone deposition rates. The presence of a slight change in bone tissue to slower forming peripheral parallel-fibred bone indicates that this individual was likely a subadult, possibly a late subadult. The lack of secondary remodelling allows an examination of the entire life history of the animal and reveals the absence of growth marks from the cortex, indicating rapid growth to somatic maturity within as little as one year. This is particularly noteworthy as its closest Permian counterparts include Cistecephalus, Dicynodontoides and Diictodon, all of which are relatively small dicynodonts (although larger than Myosaurus) and contain numerous annuli or LAGs throughout ontogeny.
The Permian Lystrosaurus species reveal fairly extensive remodelling in the inner and mid-cortex, particularly in larger specimens. However, multi-year growth ( Fig. 1a and Supplementary Fig. 2b) to maturity at large body size can still be inferred by the number and spacing of growth marks. The Early Triassic Lystrosaurus murrayi and L. declivis (Supplementary Figs 4b-h) are the most abundant vertebrates known from the LAZ by an order of magnitude. In contrast to Permian Lystrosaurus species, they exhibit few growth marks. A single annulus only appears in some individuals from 59% and 63% BSLmax in L. declivis and L. murrayi, respectively. However, the presence of such annuli is not consistent, being present in only one element of an individual, or completely absent, even in some of the larger specimens. One annulus was observed in the largest known specimen of L. murrayi and a L. declivis specimen that is 82% BSLmax. As presence of these annuli are highly variable and generally absent, it was not possible to estimate growth curves for these species. Instead, we superimposed the cross-section from a younger individual containing a growth mark, onto that of the largest for each species. We estimate that only one growth mark is missing from the largest Lystrosaurus samples, suggesting that the largest individuals were two years old at the time of death. Although presumably ontogenetically older than the smaller individuals (given their BSL of 185 mm and 213 mm respectively), transitions to slower forming bone tissues and an outer circumferential layer ('OCL') are absent in these largest individuals, indicating that somatic maturity had not been reached. However, resorption cavities extending into the midcortex, although less extensive compared to the Permian species, are observed in larger individuals and small scattered secondary osteons are observed throughout the cortex, indicating that they had deviated from the juvenile stage (as secondary remodelling generally increases with age). In order to maintain high abundances in the unstable Early Triassic environment, we propose that Lystrosaurus compensated for high mortality rates by reproducing at younger ontogenetic stages compared to their Permian relatives.

BSL Distributions of Lystrosaurus and Other Permo-Triassic Dicynodonts
BSL distributions of the two Permian Lystrosaurus species were approximately normal. This differed markedly from the two Triassic members of the genus, which had a far lower representation of individuals >70% of the BSL ranges of their respective species (21 and 27% versus 2 and 8% of individuals, respectively; Supplementary Fig. 6). Whereas the Permian species, L. curvatus and L. maccaigi had similar BSL distributions (X 2 =0.295, df=1, p=0.587), as did the Triassic L. declivis and L. murrayi (X 2 =0.427, df=1, p=0.513), both Permian taxa differed from both Triassic taxa (X 2 =4.023 to 10.879, df=1, p=0.001 to 0.045). For comparison, we evaluated BSL distributions in another 14 Permian dicynodont taxa for which we had BSL data for a minimum of 13 specimens each ( Supplementary Fig. 7). In all cases, except for Pelanomodon moschops (n=12) and Dicynodon lacerticeps (n=30), BSL distributions for these taxa were similar to those of the Permian Lystrosaurus species, and differed from the Triassic Lystrosaurus species in having a higher representation of larger-sized individuals (>70% of the BSL range; Supplementary Table 1). Two taxa (Oudenodon bainii and Rhachiocephalus magnus) had higher representation of large-bodied individuals compared with other Permian taxa. These results are consistent with results from the bone histology, which indicate lower survival rates in Early Triassic species. The results of Pelanomodon moschops and Dicynodon lacerticeps is interesting and requires further investigation.
One dicynodont from the Middle Triassic, Kannemeyeria simocephalus, showed a BSL distribution similar to that of the Permian, but not Triassic, dicynodonts, hinting at a return to pre-extinction life history strategies. Interestingly, multiple growth marks prior to the shift to slower growth rates are observed in numerous Middle Triassic therapsids (Fig. 2, Supplementary Fig. 4i, 5e, g and Supplementary Appendix 1), supporting the hypothesis that these taxa had returned to a Permian-like life history. However, since the BSL distribution is based on only one taxon (Kannemeyeria), with only a few individuals sampled (n=12), this hypothesis requires further investigation and a greater sample of Middle Triassic specimens.

Simulated Population Structures, Growth Rates and Extinction Rates
Simulated demographic differences across size classes for two types of species (Type 1 and Type 3 survivorships) are shown in Supplementary Fig. 8. These plots compare predicted size distributions of six potential life history patterns (long-and short-lived, onset of breeding at smaller or larger size classes, and producing higher quantities of offspring, respectively) in a static environment, i.e. assuming zero variability in K or in vital rates through time. As expected, survival rates were typically higher for longer-lived than shorter-lived species. However, especially amongst Type 3 populations, a strategy of breeding from a smaller size class (younger age) resulted in lower survival rates -most notably amongst larger size classes. This result is consistent with the trend seen in Triassic Lystrosaurus, which had fewer individuals >70% BSLmax. Results of matrix model projections for these populations strongly support this interpretation: at stable size distributions, the proportion of individuals in larger size classes was substantially lower in younger-breeding populations than others (Supplementary Fig. 8e and f). These differences are interpreted to be statistically significant based on non-overlapping 95% confidence limits between younger-versus older-breeding populations. The patterns were similar for models in which environmental variability was moderate (±10% of K) and high (±50% of K).
The outcome of these differences in demographics was that population growth rates were higher amongst simulated populations in which the onset of reproduction was at smaller size classes (in most models 30% versus 50% of the size range; Supplementary Tables 2 and 3).
Although this difference was seldom significant, extinction vortices revealed that younger breeding clearly has an advantage for animals living in variable environments and having curtailed life expectancies. In almost all cases, especially for hypothetical populations with reduced survivorships and life expectancies, extinction rates were lower than in populations which reproduced only later in life, including those with a higher absolute reproductive output (Supplementary Fig. 9; Supplementary Table 2). These scenarios mimic the shorter lifespans and lower relative abundances of larger-sized individuals of Lystrosaurus species from the Early Triassic, as inferred from histological and BSL data presented here. Thus, results of these models indicate that onset of reproduction at earlier ages is a life history strategy that was likely favoured in terrestrial vertebrate ecosystems shortly after the extinction event. Faster growth, younger breeding ages, and resultant shortened generation times are all factors that would have played an important role in the survival of Lystrosaurus through the end-Permian mass extinction and its dominance in Early Triassic vertebrate communities.        Kannemeyeria simocephalus (red).The majority differ from the Lystrosaurus pattern, in which individuals from this period were more evenly distributed amongst size classes than in the Early Triassic. An exception to this pattern is D. lacerticeps. K. simocephalus emerged only later, and BSL distributions for this taxon suggest an evolutionary shift back to pre-extinction life history strategies in more stable Middle Triassic environments. Dotted lines: above 70% BSLmax.  ; (b, d, f), Type 3 survivorships. Symbols are means and error bars depict 95% confidence intervals estimated over 1,000 simulations. For these plots, environmental variability and potential effects on demographic parameters is assumed to be zero. Blue: long life expectancy; Red: short life expectancy. Circles: late breeding, low fecundity; Triangles: early breeding, low fecundity; Squares: late breeding, high fecundity. Fx; fertility of size class x; gx; age-specific survivorship; log Wx; proportion of individuals in size class x at stable size distributions. % of BSLmax, % of maximum known basal skull length.