Access to this full-text is provided by Springer Nature.
Content available from Nature Ecology & Evolution
This content is subject to copyright. Terms and conditions apply.
Nature Ecoogy & Evoution | Voume 7 | Juy 2023 | 1131–1140 1131
nature ecology & evolution
https://doi.org/10.1038/s41559-023-02074-0
Article
Extended embryo retention and viviparity in
the first amniotes
Baoyu Jiang 1 , Yiming He1, Armin Elsler 2, Shengyu Wang1,
Joseph N. Keating 1, Junyi Song1, Stuart L. Kearns2 & Michael J. Benton 2
The amniotic egg with its complex fetal membranes was a key innovation
in vertebrate evolution that enabled the great diversication of reptiles,
birds and mammals. It is debated whether these fetal membranes evolved
in eggs on land as an adaptation to the terrestrial environment or to control
antagonistic fetal–maternal interaction in association with extended
embryo retention (EER). Here we report an oviparous choristodere from
the Lower Cretaceous period of northeast China. The ossication sequence
of the embryo conrms that choristoderes are basal archosauromorphs.
The discovery of oviparity in this assumed viviparous extinct clade, together
with existing evidence, suggests that EER was the primitive reproductive
mode in basal archosauromorphs. Phylogenetic comparative analyses on
extant and extinct amniotes suggest that the rst amniote displayed EER
(including viviparity).
The amniotic egg is very different from the anamniotic egg of extant
amphibians, which lacks an eggshell and extraembryonic membranes.
The amniotic egg consists of a suite of fetal membranes, including the
amnion, chorion and allantois, as well as an external shell that can be
either strongly mineralized (as in rigid-shelled eggs) or weakly min-
eralized (as in parchment-shelled eggs). The extraembryonic mem-
branes enclose specific egg elements, regulate gas and fluid exchange
between the egg and the external environment, store nutrients and
collect waste1–3.
Where and how the fetal membranes of the amniotic egg evolved
has been debated, and two competing hypotheses have been proposed
(Fig. 1). The conventional, ‘terrestrial model’2 is that the precursor to
amniotes laid eggs on land, similar in many respects to the directly
developing eggs of a variety of extant amphibians, and the fetal mem-
branes were gradually acquired so that the egg could adapt to terres-
trial environments by retaining water inside and allowing oxygen and
carbon dioxide to pass through the eggshell. This widely accepted
model has been challenged by the ‘extended embryo retention model’
3–
7
, that the extraembryonic membranes appeared in the oviducts of
the amniotic ancestor as specializations to control fetal–maternal
interaction in association with extended embryo retention (EER). The
EER model could occur with the embryo either in a post-neurula stage
(oviparity)
3
or with live bearing (viviparity)
6,7
. Among extant amniotes,
turtles, crocodilians and birds generally lay eggs at an early develop-
mental stage (non-EER oviparity), whereas most squamates (lizards and
snakes) and mammals either display oviparity with EER or viviparity.
Evolutionary studies based on extant amniotes give equivocal results
about whether oviparity or viviparity arose first
8–13
. Circumstantial
evidence for the EER model is the near absence of fossils of amniotic
eggs before the Late Triassic period and the discovery of viviparity in
many extinct amniotes as old as the Early Permian period
14–17
. Support-
ers of the terrestrial egg model note that EER is absent in archelosaurs,
including chelonians, crocodiles and birds, as well as extinct dinosaurs,
pterosaurs and their ancestors14,18.
Whether the first amniote displayed EER or not is key to testing
between the two models. As EER occurs widely among extant lizards
and snakes (squamates) and mammals3, exploring the occurrence
of EER among oviparous primitive archosauromorphs is decisive to
determine the developmental stage of the first amniotic egg (Fig. 1).
In this study, we report an articulated embryo of the choristodere
Ikechosaurus sp. inside a parchment-shelled egg. The ossification
sequence of the embryo confirms that choristoderes are basal archo-
sauromorphs. The shell structure reveals that the aquatic choristodere
was oviparous and presumably came ashore to lay its eggs, like extant
Received: 4 November 2022
Accepted: 17 April 2023
Published online: 12 June 2023
Check for updates
1State Key Laboratory for Mineral Deposits Research, School of Earth Sciences and Engineering and Frontiers Science Center for Critical Earth
Material Cycling, Nanjing University, Nanjing, China. 2School of Earth Sciences, Life Sciences Building, Tyndall Avenue, University of Bristol, Bristol, UK.
e-mail: byjiang@nju.edu.cn
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution | Voume 7 | Juy 2023 | 1131–1140 1132
Article https://doi.org/10.1038/s41559-023-02074-0
orbits and gradually tapers to about the midpoint along the snout; the
interorbital bar is narrow; the jugal extends anteriorly to the midpoint
of the lacrimal; the postorbital region is flared; the temporal openings
lie largely above one another; and the parietal extends only about half
way along the posterior edge of the upper temporal opening
20
(see
the Supplementary Results for a detailed description of the skeleton).
The fact that Ikechosaurus is the only neochoristoderan in the region
supports this assignment19.
The phylogenetic position of Choristodera in Amniota remains
controversial, having been placed as a basal clade of archosauro-
morphs21, a sister group of archosauromorphs, or basal to archosau-
romorphs and lepidosauromorphs
22
. The ossification sequence of the
new embryo confirms that choristoderes are basal archosauromorphs
(see Supplementary Results for evidence that the choristodere embryo
is an archosauromorph). The high degree of ossification of the skeleton
indicates that the unhatched embryo was in a late developmental
stage. Ontogenetically, the animal was precocial or superprecocial23:
its well-developed skull with sharp teeth suggests that it was ready to
hunt and the relatively well-ossified pelvic girdle and hindlimbs that
it could run and swim soon after hatching.
We identified traces of a parchment-shelled egg around the tiny
skeleton, which is embedded in an incomplete, oval phosphate matrix,
demarcated by a 0.43–0.50-mm-wide halo. The halo contains mate-
rials from both the phosphate matrix and the enclosing mudstone
(Fig. 2d–g). The small size, embryonic pose and egg-shaped matrix
prove that this is an embryo inside an egg. The outer edge of the halo
is slightly meandering and locally folded. Loosely arranged, irregu-
larly shaped, flake-like structures (around 50-μm thick by estimate)
are locally present along the marginal area, surrounding pore-like
structures (Fig. 2b,c). These features indicate that the halo is preserved
eggshell, which is pliable and has a thin calcareous layer composed of
flake-like shell units and many pores24.
There are three types of amniotic eggs: membrane-shelled;
parchment-shelled; and rigid-shelled25,26. A mineral layer is not
sea turtles and crocodilians. The specimen, together with previous
evidence of viviparity in other taxa, demonstrates that an evolutionar-
ily labile reproductive strategy across oviparity to viviparity existed in
choristoderes, a basal clade of archosauromorphs, and potentially also
in other various aquatic vertebrates of the past, such as mesosaurs,
ichthyosaurs and sauropterygians. We run phylogenetic analyses on
extant and extinct amniotes to test whether EER and viviparity are the
ancestral conditions in Amniota.
Results
Viviparity and oviparity in choristoderes
Choristoderes are a clade of extinct diapsids that lived primarily in
Laurasia from the Middle Jurassic period to the Early Miocene epoch
(approximately 168–120 Myr). The gavial-like neochoristoderes were
top predators in freshwater bodies, competing with contemporane-
ous crocodiles
19
. The new specimen (MES-NJU 57004) was collected
from yellowish white, thinly laminated tuffaceous mudstone of the
Lower Cretaceous Jiufotang Formation ( Jehol Biota, approximately
125–120 Myr) in the Lamagou locality adjacent to Chaoyang City,
western Liaoning, northeast China. Many choristoderes have been
discovered in the Jiufotang Formation and the underlying Yixian Forma-
tion in western Liaoning, including the lizard-like Monjurosuchus and
Philydrosaurus, the long-necked Hyphalosaurus and the neochoris-
toderan Ikechosaurus. Some specimens of these choristoderes are
associated with eggs and embryos19.
The new specimen is a small skeleton (approximately 102.73 mm
long; Fig. 2a) that exhibits the typical pose of a vertebrate embryo:
curving and with the head contacting the tail
17
. The skeleton is dor-
soventrally flattened and exposed in ventral view covered by a thin
layer of ferric oxide. Computed tomography (CT) scans reveal that
the skeleton is nearly complete and all bony elements are articulated
except for the distal end of the tail, which was slightly displaced (Fig. 3
and Extended Data Figs. 1 and 2). The embryo shows many diagnostic
traits of Ikechosaurus: the snout is long, broad and flat in front of the
Water
Precursors to amniote
Anamniote
Amniote
EER eggs
Non-EER eggs
EER eggs
Viviparity
Viviparity
Viviparity
Non-EER eggs
Non-EER eggs
parchment-shelled
Non-EER eggs
rigid-shelled EER eggs
parchment-shelled
parchment-shelled parchment-shelled
Terrestrial model EER model
Land
Non-EER eggs
rigid-shelled
Most mammals
Many squamates
Many squamates
Most mammals
Birds
Most squamates
Monotremes
Crocodilians
A few chelonians
Most chelonians
Birds
Crocodilians
Most chelonians
Monotremes
A few chelonians
Most squamates
Extinct aquatic amniotes
Fig. 1 | The two competing theories for the evolution of the amniotic egg.
In the terrestrial model (left), non-EER oviparity (purple) was the primitive
condition; oviparity with EER and viviparity (blue) evolved multiple times in
amniotes. In the EER model (right), the evolutionarily labile reproductive mode
of EER across oviparity to viviparity (blue) was primitive, while non-EER oviparity
(purple) evolved multiple times in amniotes.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution | Voume 7 | Juy 2023 | 1131–1140 1133
Article https://doi.org/10.1038/s41559-023-02074-0
developed in membrane-shelled eggs (monotremes, a few squamates),
is thinner than an organic layer in parchment-shelled eggs (a few che-
lonians and most squamates) and is well developed and thicker than
the organic layer in rigid-shelled eggs (a few squamates, most chelo-
nians, crocodilians and birds)25. It could be argued that the described
choristodere specimen is an incomplete viviparous egg27,28. In extant
squamates, the egg of viviparous species lacks the calcified layer but is
enveloped with a very thin organic layer (commonly less than 10 μm),
whereas oviparous squamates lay eggs with a calcified outer layer
and a relatively thick organic layer (usually over 30 μm)29,30, as in this
specimen. A similar eggshell structure, with a thick organic layer (over
100 μm) and a very thin mineral layer (less than 10 μm), has also been
documented in isolated eggs of the basal choristoderan Hyphalosaurus
baitaigouensis31. These shell structures reveal that the aquatic choris-
toderes were oviparous and presumably came ashore to lay their eggs,
like extant sea turtles and crocodilians. Putative females of the neo-
choristoderan Champsosaurus possessed fused sacral centra and more
robust limb bones than males, perhaps adaptations for nesting on
land32. Other choristoderes were viviparous, such as one specimen of
H. baitaigouensis
28
and Monjurosuchus splendens
14
. The co-occurrence
of viviparity and oviparity in one nominal species, H. baitaigouensis,
indicates that it had a bimodal reproductive mode, that is, both vivi-
parity and oviparity occurred in a single species as in the basalmost
sauropsid Mesosaurus tenuidens and few extant squamates14,33.
Macroevolutionary study
The new specimen exemplifies the complexity of parity modes in some
early reptiles and provides information to constrain the evolution of
reproductive strategies in archosauromorphs, reptiles and amniotes
in general. We return to the question of whether EER is the ancestral
condition. To test this hypothesis, we collected data on reproductive
modes from representative taxa spanning the phylogenetic diversity of
extant amniotes, and from all extinct taxa where information is available
(Supplementary Table 1). The balance of data on fossil taxa inevitably
favours those that laid rigid-shelled eggs because those are preserved
more readily than eggs without mineralized shells. In particular, we
could find no examples of extinct synapsids with evidence of repro
-
ductive mode. However, we compensated for this by broadly sampling
extant taxa for which reproductive data are secure. Most oviparous
squamates lay membrane- and parchment-shelled eggs at the limb-bud
stage. In contrast, most amniotes that lay rigid-shelled eggs obligately
oviposit at an early developmental stage, for instance, at the blastula
stage in birds, the gastrula stage in chelonians and tuataras, and the
neurula stage in crocodilians
33–35
. In these forms, the thick calcite layer
delays the exchange rate of respiratory gases and may prevent the
development of the embryo before the eggs are laid
36
. Therefore, both
eggshell and developmental stage of the embryo at oviposition provide
information constraining the ancestral state of reproductive modes.
We coded each extant and extinct taxon for three characters:
(1) reproduction mode: viviparous, oviparous; (2) eggshell mineraliza-
tion: membrane-shelled, parchment, rigid; (3) EER: absent, present. EER
was defined as amniotes that lay eggs at the limb-bud stage or later33.
EER was identified in extinct taxa based on fossil adults that contain
embryos at the limb-bud stage or later or are associated with neonates,
which were commonly identified as evidence for viviparity previously
(Supplementary Table 1). We conservatively treated all fossils for which
we could not judge the stage of development of an egg at oviposition
as non-EER. Character 2 is dependent on character 1, so viviparous taxa
were coded as inapplicable (‘−’) for character 2. To resolve the problem
of character dependency, characters 1 and 2 were amalgamated into a
single structured Markov model (SMM) equipped with hidden states.
We also applied an analogous approach within a parsimony framework
using Sankoff (cost) matrices. We conducted an exhaustive multipli-
cative set of ancestral states analyses, accounting for: (1) different
Si K Ca K Mn K Fe K 0 666
0134
0 951
0 1,585
B
C
D–G
a b c
gd e f
Fig. 2 | Structure of the choristodere egg (MES-NJU 57004). a, An overview of
the embryo inside the egg. b,c, Photomicrographs show loosely arranged shell
units and pores (black arrows) preserved along the marginal zone of the egg.
d–g, Energy dispersive spectroscopy mapping of silica (d), calcium (e),
manganese (f) and iron (g) showing the inferred eggshell (white triangles), which
contains materials from both the phosphate matrix and the enclosing mudstone.
Scale bars, 1 cm (a), 100 μm (b,c), 1 mm (d–g).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution | Voume 7 | Juy 2023 | 1131–1140 1134
Article https://doi.org/10.1038/s41559-023-02074-0
phylogenetic time-scaling methods; (2) alternative tree topologies;
(3) exclusion of key fossils; (4) different ancestral state reconstruction
methods, evolutionary models and optimization criteria; (5) among lin-
eage rate heterogeneity; and (6) constraining extant node states based
on previous studies. Together, these totalled over 100,000 individual
analyses. Further details can be found in the Methods.
The ancestral state reconstruction results are unequivocal for
both the amalgamated character (reproduction mode + eggshell min-
eralization) and EER presence (Fig. 4). Viviparity with EER dominates
the deeper nodes, being the most likely condition for the roots of
Amniota (mean marginal maximum likelihood (ML)-based ancestral
state across 100 trees for the best-fitting model and other models
whose Akaike information criterion (AIC) difference was less than
2 compared to the best-fitting model for viviparity-ML: 99.5–100%;
EER-ML: 99.8–100%), Reptilia (viviparity-ML: 98.5–100%; EER-ML:
100%), Diapsida (sensu lato) (viviparity-ML: 100%; EER-ML: 100%),
Archelosauria (viviparity-ML: 98.5–100%; EER-ML: 99.6–100%) and
Archosauromorpha (viviparity-ML: 99.4–100%; EER-ML: 99.8–100%),
irrespective of which time-scaling approach or best-fitting model
(including models whose AIC differed from the best-fitting model by
less than 2) was considered (Supplementary Tables 3–6).
Using a different ancestral state reconstruction also does not
change the main conclusions: 10 of 18 maximum parsimony (MP)-
based ancestral state reconstructions recover viviparity at the ori-
gin of Amniota, with the remaining reconstructions mostly favour-
ing membrane-shelled eggs as the ancestral state of amniotes (all
MP-based reconstructions using accelerated transformation
(ACCTRAN) favour viviparity at the origin of Amniota; Supplementary
Tables 21–26). All MP-based ancestral state reconstructions favour
EER as the ancestral state of Amniota. Bayesian traits (BT) results are
consistent with ML-based ancestral state reconstructions (Supplemen-
tary Tables 28–31 and Extended Data Fig. 4). Fixing the nodes of both
Lepidosauria and Squamata to a non-viviparous state (Supplementary
Tables 28–31 and Extended Data Fig. 5) still results in viviparity being
the most likely condition for the origin of Amniota (mean BT-based
ancestral state for viviparity: 87.4–97.7%), only leading to increased
variability in the ancestral state estimates (Reptilia viviparity-BT:
79.8–97.2%; Diapsida (sensu lato) viviparity-BT: 72.3–99.7%) and
potentially suggesting that viviparity re-evolved in Archosauromor-
pha (Archelosauria viviparity-BT: 16.6–91.2%; Archosauromorpha
viviparity-BT: 48.9–96.1%).
Among archosaurs, the results reflect the possession of
rigid-shelled eggs by extant birds and crocodilians, which is con-
sistent with the proposal15 that the first dinosaurian eggs were
membrane-shelled, although the results are less clear. Most best-fitting
models favour rigid-shelled eggs at the root of Theropoda (rigid-ML:
81.6–98.5%; EER-ML: 0–1.5%); however, non-EER, membrane-shelled
eggs are most likely at the root of Saurischia (membrane-shelled-ML:
67.5–97.2%; EER-ML: 0.1–0.4%), Dinosauria (membrane-shelled-ML:
60.5–96.7%; EER-ML: 0.1–1.4%) and probably also Archosauria
(membrane-shelled-ML: 52–96.1%; EER-ML: 0.2–13.6%). The non-EER
results are consistent across all variant analyses but the reproduction
mode and eggshell mineralization results are somewhat equivocal
for later-diverging clades. The results from the best-fitting models
are compatible with parchment-shelled eggs as the ancestral state
for Saurischia (parchment-ML: 1–30.3%), Dinosauria (parchment-ML:
1.1–35.6%) and Archosauria (parchment-ML: 1.3–42.5%). If we also
consider simpler ML models with a worse fit (AIC difference greater
than 2), where character state transition rates are more con-
strained compared to the best-fitting models, the evidence for
parchment-shelled eggs for these clades increases. Note, however, that
parsimony-based analyses favour rigid-shelled eggs as the ancestral
state of the three clades. Bayesian analyses are generally consistent
with ML-based results but evidence for parchment-shelled eggs in
a b
sa
pt
q
qj
ot
pdt
cv
cvr
fi
fe
isc
isc
pub
pub
r
cdv
r
v
ce
so
sq
pa
po
ju
ept
prf
cv
pt
st
pof
po
fro
pa
cl
sc
hu
r
dv
cdv
e f
g h
na rccv na
na
fi
ti
isc
il
pub
fe
c d
cv
ot
ept
ju
sa
q
ot
pa
so
sq
qj
na
r
cdv
r
v
ce
so
mt
hdt cl
sc
mc
hu
ul/ra
il
ti
mt
sq
st
hdt
cl
sc
mc
hu
ul/ra
il
ti
sq
pa
po
ju
ept
fro
cv
na
mx
pmx
t
pal
lac
den
mx
prf
q
qj
ot
pdt
pmx
t
pal
lac
den
pof
an
Fig. 3 | CT scans of the choristodere skeleton (Ikechosaurus sp.) and its
reconstruction. a,b, Ventral (a) and dorsal (b) views of the reconstructed
skeleton. c–f, Coronal slices showing dorsal (c) and ventral (d) sections of the
posterior skull, proximal left forelimb and presacral axial bones (e), and pelvic
girdle, proximal axial bones and femur (f). g,h, Transverse slices show the
relationship of centra and neural arches in dorsal (g) and caudal (h) vertebrae.
cdv, centra of dorsal vertebrae; ce, cervical; cl, clavicle; cv, caudal vertebra; den,
dentary; ept, ectopterygoid; fe, femur; fi, fibula; fro, frontal; hdt, hand digit;
hu, humerus; il, ilium; isc, ischium; ju, jugal; lac, lacrimal; mc, metacarpal; mt,
metatarsal; mx, maxillary; na, nasal; ot, otic area; pa, parietal; pal, palatine; pdt,
pes digit; pmx, premaxillary; po, postorbital; pof, postfrontal; prf, prefrontal;
pt, pterygoid; pub, pubis; q, quadrate; qj, quadratojugal; r, rib; ra, radius;
sa, surangular; sc, scapula; so, supraoccipital; sq, squamosal; st, stapes; t,
tooth; ti, tibia; ul, ulna; v, vertebra. Scale bar, 10 mm.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution | Voume 7 | Juy 2023 | 1131–1140 1135
Article https://doi.org/10.1038/s41559-023-02074-0
these three clades is generally lower; fixing the node of Lepidosauria
and Squamata to a non-viviparous state leads to a higher variability in
the ancestral state estimates. These discrepancies emphasize that while
the ancestral state results are consistent for deep nodes, the results for
later-diverging clades should be treated with more caution. A similar
example is found in Lepidosauria for which parsimony and alternative
ML and BT evolutionary models do not recover viviparity as ancestral
(Supplementary Tables 3–26 and 28–45).
Kryptobaatar dashzevegi
Mammalia
Amniota
Reptilia
Diapsida (s.l.) Lepidosauria
Extinct marine
reptiles
Diapsida (s.s.)
Turtles
Archelosauria
Archosauromorpha
Archosauria
Dinosauria
Saurischia
Theropoda
Avialae
Macropus rufus
Elephas asiaticus
Oreodon culbertsoni
Maiacetus inuus Homo sapiens
Ornithorhynchus anatinus
Dolichorhynchops osborni
Polycotylus latipinnis
Mesosaurus tenuidens
Neusticosaurus peyeri
Lariosaurus sp.
Keichosaurus hui
Chaohusaurus geishanensis
Mixosaurus sp.
Besanosaurus leptorhynchus
Shonisaurus popularis
Qianichthyosaurus zhoui
Temnodontosaurus sp.
Leptonectes tenuirostris
Ichthyosaurus communis
Stenopterygius quadriscissus
Stenopterygius triscissus
Platypterygius australis
Platypterygius longmani
Yabeinosaurus tenuis
Sphenodon punctatus
Anepischetosia maccoyi
Podarcis siculus
Elaphe guttata
Sceloporus scalaris
Sceloporus clarkii
Furcifer lateralis
Diploglossus delasagra
Diplodactylus vittatus
Pelusios sinuatus
Acanthochelys radiolata
Pelodiscus sinesis
Adocus sp.
Dinocephalosaurus orientalis Desmatochelys padillai
Philydrosaurus proseilus
Ikechosaurus sp.
Monjurosuchus splendens
Pterodaustro guinazui
Hamipterus tianshanensis
Pterodactyloid
Ornithocheirid
Massospondylus carinatus
Absent
EER
Reproduction mode +
eggshell mineralization
Viviparity
Non-mineralized egg
Parchment-shelled egg
Rigid-shelled egg
Mississippian
320
300
280
260
240
220
200
180
160
140
120
100
80
60
40
20
0
Pennsyl-
vanian
Cisuralian
Carboni-
ferous Permian Triassic Jurassic Cretaceous Palaeogene Neo-
gene
Guadalupian
Lopingian
Lower
Middle
Upper
Lower
Middle
Upper
Lower
Upper
Palaeocene
Eocene
Oligocene
Miocene
Present
Saltasaurid
Titanosaurian
Oviraptor philoceratops
Heyuannia huangi
Deinonychus antirrhopus
Enantiornithine
Gobipipus reshetovi
Sinosauropteryx
Mussaurus patagonicus
Protoceratops andrewsi
Maiasaura peeblesorum
Telmatosaurus transsylvanicus
Chelonoidis carbonaria
Kinosternon baurii
Caretta caretta
Alligator mississippiensis
Crocodylus niloticus
Struthio camelus
Casuarius casuarius
Gallus gallus
Anas platyrhynchos
Phoenicopterus ruber
Columba palumbis
Strix nebulosa
Passer domesticus
Microtroodontid
Troodontid
Carsosaurus marchesetti
Plioplatecarpus primaevus
Phu phok embryo
Maiaspondylus lindoei
Fig. 4 | Phylogeny of amniotes, showing known reproduction mode and
eggshell mineralization, and the EER of 80 extant and extinct species (tips,
to right), and the inferred mean ancestral states for all branching points
(larger pie charts at the nodes). The dominant inferred state towards the root
(left) is viviparity with EER. This is a consensus tree based on a sample of 100 trees
time-scaled using the fossilized birth-death (FBD) tip-dating method (with node
age constraints for major clades) and component ARD model with a switch-on
dependency (CARD_sw) for eggshell mineralization and reproduction mode
(best-fitting ML model based on the AIC score; Supplementary Table 6). Note that
alternative evolutionary models do not recover viviparity as the ancestral state
in Lepidosauria (Supplementary Tables 3–6). Ikechosaurus sp. is shown in red
and indicated by a red arrow. Elephant and bird icons reproduced from PhyloPic
under a Creative Commons license CC BY 3.0 (elephant, T. Michael Keesey; bird,
Chloé Schmidt).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution | Voume 7 | Juy 2023 | 1131–1140 1136
Article https://doi.org/10.1038/s41559-023-02074-0
While the best EER model is consistently recovered as all-rates-
different (ARD), the best model to explain character state evolution
for reproduction mode and eggshell mineralization is more enigmatic
and changes depending on the distribution of branch lengths, the
time-scaling methods used and topology (Supplementary Tables 27
and 28–45). However, the component ARD model with or without a
switch-on dependency is always among the best-fitting ML-based
models irrespective of which time-scaling approach or topology is con-
sidered (Supplementary Tables 3–20). The component ARD model is
generally also found among the best-fitting BT-based models; however,
the evidence for alternative model hypotheses is generally lower for
analyses with fossilized Lepidosauria and Squamata node states due
to smaller-log Bayesian factors (BFs) (Supplementary Tables 28–45).
Discussion
It is widely accepted that viviparity evolved from oviparity through
EER in squamates and mammals
33–35
, and this may also have been the
case in the various aquatic vertebrates of the past, such as choristo-
deres, as well as mesosaurs, ichthyosaurs and some sauropterygians.
This implies that viviparity might have derived from membrane- or
parchment-shelled eggs with EER in these clades. Yet, our results
strongly suggest viviparity as the ancestral condition of amniotes and
the marine reptile clades.
This result is subject to a number of biases: first, the phylogenetic
placement of extinct marine reptiles within the Amniota is contested
37
.
Our results, however, are qualitatively unchanged if we revise the phy-
logeny to place extinct marine reptiles within the Lepidosauromorpha
or in other basal diapsid locations (Supplementary Tables 15–20 and
40–45). Completely removing the extinct marine reptiles from the
analyses has little impact on the results, with only models with a worse
fit and some of the models for which the states of both Lepidosauria and
Squamata were fixed a priori showing more uncertainty in the ances-
tral state reconstructions (Supplementary Table 11–14 and 36–39).
Further, the placement of Mesosaurus, coded ambiguously as either
viviparous or oviparous with non-mineralized eggs
14,17
near the root of
the Amniota does not affect these results; if it is removed, the results
remain qualitatively the same (Supplementary Tables 7–10 and 32–35).
Second, our dataset may be taphonomically biased. It could be
argued that well-skeletonized embryos in adult oviducts are more
likely to be preserved than membrane- or parchment-shelled eggs,
and that isolated eggs are difficult to assign to species. Consequently,
viviparity might be overrepresented by extinct viviparous clades in
our analysis. Given that Ikechosaurus is a derived choristodere
22
, the
discovery of oviparity in this presumably viviparous clade suggests
that either oviparity evolved from viviparity or viviparity evolved
from extinct oviparous ancestors with EER in choristoderes (the
latter hypothesis, however, is not backed by the fossil record). This
might also be true for extinct viviparous clades, such as mesosaurs,
ichthyosaurs and sauropterygians that belong to basal parareptiles
or amniotes in general and diapsids or lepidomorphs, respectively.
As fully aquatic amniotes could not come onto land to lay their eggs
and embryos in eggs could not survive in water, viviparity in these
extinct amniotes must have evolved from either semi-aquatic or
terrestrial viviparous ancestors (as implicated by our analyses), or
oviparous ancestors that displayed EER but whose eggs have not been
preserved in the fossil record27. In either case, these early amniotes
reflect EER as the primitive reproductive mode of amniotes (Fig. 1,
model on the right).
The absence of rigid-shelled eggs through the Carboniferous
period, Permian period and most of the Triassic period
2,14,15
has long
been noted: if such eggs truly existed widely during this time interval,
it could change our results entirely, pointing to rigid-shelled eggs as
the ancestral amniote condition. However, as rigid-shelled eggs are
more likely to be preserved than membrane- or parchment-shelled
eggs and perhaps the partially ossified and tiny bones of retained
embryos within the mother, this might not be a preservation bias but
a real absence
15
. Conversely, given the scarcity of fossil evidence on the
reproductive mode of Palaeozoic amniotes and the low fossilization
potential of membrane- or parchment-shelled eggs, the possibility
remains that the earliest amniote eggs were, indeed, membrane- or
parchment-shelled. The addition of more extinct taxa will provide an
important test of these results14,15.
The occurrence of parchment-shelled eggs associated with EER
in choristoderes extends the wide occurrence of this phenomenon
among amniotes and may be the primitive reproductive mode that
occurred before archosaurs (crocodilians, dinosaurs, birds) and
chelonians acquired non-EER oviparity and rigid-shelled eggs
1
. As
membrane- and parchment-shelled eggs associated with EER are also
common in extant mammals and squamates, this suggests that this
condition was common among early terrestrial amniotes, for example,
in the first 100 Myr of their evolution from the Carboniferous to the
Triassic, favouring a model that amniotic fetal membranes evolved
in association with EER. Similarly, a phylogenetic study of extant
tetrapods38 inferred that many of the structures that characterize
the Amniota (delayed deposition of eggs, large yolk mass, cellular
yolk sac and amnion) may have evolved in an aquatic environment in
association with delayed egg laying.
Methods
CT scanning and reconstruction
The specimen was scanned at the University of Bristol, UK, on the
Nikon X‐Tek H 225 ST X‐ray scanner at 225 kV and 188 μA (42.3 W) from
a rotating tungsten target, with 2-s exposure, 1× binning, 24-dB gain
and a 3-mm copper filter, slice thickness = 48.45 μm and total num
-
ber of slices = 1,142. Each scan captured 3,141 projections, with four
frames averaged per projection. The reconstructed scan data were
subsequently combined in VGStudio v.3 (https://www.volumegraphics.
com). A three‐dimensional (3D) model was created from the CT data
using the segmentation tools in Avizo v.9.1.1 Lite (Visualization Science
Group; https://www.fei.com/software/amira-avizo/). All scan data and
3D models are available in the supplementary data.
Scanning electron microscopy
Samples were examined using a JEOL 8530F Hyperprobe at the School
of Earth Sciences, University of Bristol, and a LEO 1530VP scanning
electron microscope at the Technical Services Centre, Nanjing Insti-
tute of Geology and Palaeontology, Chinese Academy of Sciences.
Both instruments were equipped with a secondary electron detector,
a back-scattered electron detector and an energy dispersive X-ray
spectrometer.
Phylogenetic macroevolutionary analysis
Data collation. Data were compiled on key reproductive parameters
for as many extinct amniotes as possible, distinguishing three char-
acters, each with two or three states: reproduction mode: (1) vivipar-
ity and (2) oviparity; eggshell mineralization: (1) non-mineralized,
(2) weakly mineralized, (3) rigid; and EER: (1) absent and (2) present.
These eggshell and parity characteristics have been documented
widely in the literature, and we indicate exact data sources in our
data compilation (Supplementary Table 1). We identified 59 extinct
taxa for which eggs or viviparity had been identified (6 mammals;
1 mesosaur; 6 turtles; 13 ichthyosaurs; 5 sauropterygians; 4 squamates;
3 choristoderes; 1 protorosaur; 2 crocodilians; 4 pterosaurs; 13 dino-
saurs; 1 bird). We added a further 21 extant taxa, making a total of 80.
As the basis for an initial analysis of ancestral states for Amniota and
subclades, we compiled a supertree for the 80 taxa, using a standard
genomic tree
39
as scaffold (Supplementary Table 2), supplemented by
recent cladistic analyses of extinct groups40. Seymouria baylorensis
and Diadectes sideropelicus were added as outgroups to a polytomy
including the Amniota.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution | Voume 7 | Juy 2023 | 1131–1140 1137
Article https://doi.org/10.1038/s41559-023-02074-0
Phylogenetic time-scaling. We obtained first and last appearance
dates (FAD and LAD, respectively) for each taxon in our analysis using
the Paleobiology Database
41
. We then time-scaled the supertree using
four different approaches: (1) the minimum branch length (mbl)
method42; (2) the equal branch length (equal) method43; (3) the FBD
tip-dating method
44–46
with only the root; and (4) with both the root
and the node ages of major extant clades constrained. The mbl and
equal methods are a posteriori time-scaling methods47 that avoid
zero-length branches by imposing a minimum branch length of 1 Myr
(mbl) or by taking an equal share from preceding non-zero-length
branches (equal). We generated 100 time-scaled trees with the time-
PaleoPhy function of the paleotree package48 with tip dates sampled
from a uniform distribution (dateTreatment = minMax) bounded by
FADs and LADs and the vartime argument set to 1 Myr. The FBD method
jointly considers speciation, extinction and fossil preservation rates
to estimate divergence times in a Bayesian framework
44–46
. We applied
a clockless tip-dating approach using MrBayes (v.3.2.7a)
49,50
where the
empty morphological matrix was generated using the createMrBayes-
TipDatingNexus function of the paleotree package
48
. Topology was
constrained to the input supertree; uniform priors bounded by the
respective FADs and LADs were used to calibrate the tip ages. To place
the time-scaled trees in absolute time, M. tenuidens (Early Permian,
Kungurian, 278.4 Myr) was used as the anchor taxon. For all node cali-
brations, we used offset gamma distributions with a shape parameter
of 3. Based on Benton et al.
51
for the root of the Amniota, we used an
offset gamma distribution with a mean age of 325.1 Myr, a minimum
age of 318 Myr and an s.d. of 4.099175 Myr. Additional major crown
clade calibrations
50
for the second FBD analysis were parametrized
as follows: Mammalia (minimum age = 164.9, mean age = 182.3402,
s.d. = 10.06911); Diapsida (255.9, 274.9603, 11.0045); Lepidosauria (238,
245.0047, 4.044152); Squamata (168.9, 188.2463, 11.16956); Archosau-
ria (247.1, 253.3423, 3.603972); and Aves (66, 75.91138, 5.722338). The
default FBD and clock priors52 provided by createMrBayesTipDat-
ingNexus were kept. We disallowed sampled ancestors (prset sam-
plestrat = fossiltip;). we ran the two FBD analyses four times, using
four chains per run, for 1,000,000,000 generations, sampling every
100,000th generation. We checked convergence using Tracer v.1.7.1
(ref. 53), ascertaining an effective sample size (ESS) of all parameters
exceeding 200 for combined traces. We used the obtainDatedPoste-
riorTreesMrB function of the paleotree package to obtain a sample of
100 time-scaled trees from the posterior, employing a burn-in of 50%.
Before the ancestral state reconstruction, we removed the outgroup
taxa S. baylorensis and D. sideropelicus.
Ancestral state estimation. Hierarchical character dependencies
have long presented a challenge in ancestral state estimation54–56.
In our analysis, character 1 and 2 are hierarchically related; the state
of character 2 (eggshell mineralization) is dependent on character
1 (reproduction mode) being in a specific state (state = oviparous).
This is equivalent to the classic tail colour problem of Maddison
54
.
Recently, Tarasov55 demonstrated that SMMs equipped with hidden
states solve the problem of modelling character complexes with hier-
archical dependencies. In doing so, he also demonstrated the invari-
ant nature of characters and states. These concepts are equivalent;
characters can be transformed into states and vice versa. We followed
the approach of Tarasov55 and amalgamated characters 1 and 2 into a
single SMM with 6 states (Extended Data Fig. 3). Ancestral states were
estimated under two variations of this model. SMM_ind assumes that
reproduction mode and eggshell mineralization evolve independently.
This is analogous to modelling the characters separately using two
independent models, except for the fact that under the SMM approach,
simultaneous changes to character 1 and character 2 are prohibited.
Alternatively, SMM_sw assumes ‘switch-on’ dependency. That is, char-
acter 2 (eggshell mineralization) can only change state if character 1
(reproduction mode) is in a specific state (state = oviparous).
The SMM_ind and SMM_sw approaches consider different models
on rate transition between the character states, resulting in eight evo-
lutionary models: a component equal-rate (ER) model (CER_ind and
CER_sw: transitions between states among component characters share
a single rate parameter); a component symmetrical model (CSYM_ind
and CSYM_sw: transitions between states among component char-
acters are symmetrical); a component ARD model (CARD_ind and
CARD_sw: transitions between states among component characters are
all different); and an equal rates model (ER_ind and ER_sw: transitions
between aggregated rates share a single rate parameter). The SMMs
used in this study are summarized in Extended Data Fig. 3.
We used an ML approach to estimate the ancestral states for the
SMMs, applying the asr_mk_model function (with the optimization
algorithm set to ‘optim’) of the castor R package57. To avoid optimiza-
tion problems, the input trees were scaled to a tree height of 1 before
the ML analyses. The transition matrix was fitted ten times and the
maximum allowed number of iterations per fitting trial was set to 500.
We used the ‘tip.priors’ argument to assign probabilities to amalga-
mated states. M. tenuidens is a special case. Current fossil evidence
does not enable us to determine confidently whether M. tenuidens was
viviparous or oviparous with membrane-shelled eggs14,17. Again, we
used the tip.priors argument to specify this uncertainty. For the CARD
models, which are not time-reversible, marginal ancestral likelihoods
were computed without rerooting the input tree. We used the AIC to
select the best-fitting model. We then calculated the mean marginal
ancestral states of the best-fitting model for each set of 100 input trees,
which were plotted on a consensus tree generated using the consensus.
edges function of the R package phytools58. Plots were generated using
the R package strap59.
The presence and absence of EER was also modelled using asr_mk_
model with default parameter settings, the optimization algorithm
set to optim, using the same input trees, and providing the two-state
character for EER via the ‘tip_states’ argument. We used an ER model
(EER ER) and an ARD model (EER ARD). Model selection was again car-
ried out using the AIC. The calculation and plotting of mean marginal
ancestral states followed the same practice as for the SMMs.
In addition to ML, we also ran an MP-based ancestral state recon-
struction for the same characters using the ancestral.pars function of
the phangorn package60,61. As the parsimony approach does not esti-
mate transition rates, the evolutionary models used in the ML approach
cannot fully be translated into a parsimony setting. We generated
three parsimony models for the amalgamated character (reproduc-
tive mode and mineralization): one ACCTRAN approach that allows
all character transitions and two most-parsimonious reconstructions
(MPR) approaches that attempt to model either independence of
reproduction mode and eggshell mineralization (similar to SMM_ind)
or a switch-on dependency (similar to SMM_sw) using Sankoff (cost)
matrices54,56. The number of required steps for transitions that were for-
bidden in the model was set to a value (step cost = 100) that would make
it practically impossible for the transition to occur. The ACCTRAN
62
approach as implemented by ancestral.pars does not allow for cost
matrices, thus requiring the MPR
63,64
approach. Contrast matrices were
used to assign probabilities to amalgamated states and account for
uncertainty in tip states. The MP-based ancestral state reconstruction
was repeated for the EER character using ACCTRAN.
Furthermore, we also ran a BT ancestral state reconstruction using
the package BayesTraits
65,66
for each set of 100 time-scaled input trees.
As with the ML and MP approaches, the input data was formatted to
account for uncertainty in (amalgamated) tip states. To avoid optimiza-
tion problems, input time-scaled trees were rescaled to have a mean
branch length of 0.01. We used the same SMM and EER used in the ML
approach. A reverse-jump, continuous time Markov67,68 Chain Monte
Carlo algorithm (rjMCMC) was applied to both homogeneous and
variable rate models, the latter allowing for shifts in the rate of evolu-
tion
σ2
v
on individual branches69,70. For the models with multiple
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution | Voume 7 | Juy 2023 | 1131–1140 1138
Article https://doi.org/10.1038/s41559-023-02074-0
transition rate parameters, we ran the multistate approach using the
rjMCMC method with an exponential (0, 10) hyperprior. For the
single-rate models (ER_ind, ER_sw, EER ER), a uniform (0, 10) distribu-
tion was set as the transition rate prior. Three independent MCMC
chains per model were run for 11,000,000 iterations and parameters
were sampled every 10,000 iterations; 1,000,000 iterations were
discarded as burn-in. We calculated the ML of the models using the
stepping stone sampler
71
implemented in BayesTraits. We sampled
1,000 stones and used 100,000 iterations per stone. Convergence was
assessed using the R package CODA72, ensuring that the smallest ESS
always exceeded 200 for the combined chains. Models were compared
using a log BF test
73
applied to the mean log MLs from the combined
three MCMC chains. To calculate the log BF, the homogeneous rate
ER_sw and homogeneous rate EER ER models served as the simple
comparison models for the SMMs and the EER models, respectively.
Mean ancestral states were calculated across the combined three
MCMC chains for each model and plotting followed the same practice
used for the ML approach. Given the uncertainty in the ancestral state
reconstructions for Lepidosauria and Squamata8–10,74–77, we reran our
Bayesian SMM analyses, this time fixing the nodes of the two clades to
a non-viviparous state.
Robustness tests. To test the robustness of our results, we reran all our
analyses dropping M. tenuidens and extinct marine reptiles, respec-
tively, from our input phylogeny. We also reran the mbl and equal
time-scaled trees with a modified phylogenetic position of the extinct
marine reptiles, adding them either as a sister taxon to Archelosauria,
Archosauromorpha or Lepidosauria21,78,79.
Reporting summary
Further information on research design is available in the Nature Port-
folio Reporting Summary linked to this article.
Data availability
We provide all data as supplementary data. The phylogeny used in
this study is shown in Fig. 4. The specimen studied (MES-NJU 57004)
is hosted at the School of Earth Sciences and Engineering, Nanjing
University. Correspondence and requests for materials should be
addressed to B.J. or M.J.B.
Code availability
All analyses in this study were conducted using readily available, pub-
lished programs that are cited in the text. The versions of the programs
are as follows: R v.4.1.0; ape v.5.5; castor v.1.6.7; paleotree v.3.3.25;
phangorn v.2.7.0; phytools v.0.7-70; strap v.1.4; and BayesTraits v.4.0.0.
References
1. Stewart, J. in Amniote Origins (eds Sumida, S. & Martin, K. L. M.)
291–326 (Academic Press, 1997).
2. Romer, A. S. Origin of the amniote egg. Sci. Monthly 85,
57–63 (1957).
3. Laurin, M. Embryo retention, character optimization, and the
origin of the extra‐embryonic membranes of the amniotic egg.
J. Nat. Hist. 39, 3151–3161 (2005).
4. Lombardi, J. Embryo retention and the origin of the amniote
condition. J. Morphol. 220, 368 (1994).
5. Laurin, M. & Reisz, R. R. in Amniote Origins (eds Sumida, S. &
Martin, K. L. M.) 9–59 (Academic Press, 1997).
6. Hubrecht, A. A. W. Memoirs: the fœtal membranes of the
vertebrates. J. Cell Sci. 2, 177–188 (1910).
7. Mossman, H. W. Vertebrate Fetal Membranes (Rutgers Univ.
Press, 1987).
8. Pyron, R. A. & Burbrink, F. T. Early origin of viviparity and multiple
reversions to oviparity in squamate reptiles. Ecol. Lett. 17,
13–21 (2014).
9. King, B. & Lee, M. S. Y. Ancestral state reconstruction, rate
heterogeneity, and the evolution of reptile viviparity. Syst. Biol.
64, 532–544 (2015).
10. Wright, A. M., Lyons, K. M., Brandley, M. C. & Hillis, D. M. Which
came irst: the lizard or the egg? Robustness in phylogenetic
reconstruction of ancestral states. J. Exp. Zool. B Mol. Dev. Evol.
324, 504–516 (2015).
11. Lee, M. S. Y. & Shine, R. Reptilian viviparity and Dollo’s law.
Evolution 52, 1441–1450 (1998).
12. Harrington, S. & Reeder, T. W. Rate heterogeneity across
Squamata, misleading ancestral state reconstruction and the
importance of proper null model speciication. J. Evol. Biol. 30,
313–325 (2017).
13. Holland, B. R., Ketelaar-Jones, S., O’Mara, A. R., Woodhams, M. D.
& Jordan, G. J. Accuracy of ancestral-state reconstruction for
non-neutral traits. Sci. Rep. 10, 7644 (2020).
14. Blackburn, D. G. & Sidor, C. A. Evolution of viviparous
reproduction in Paleozoic and Mesozoic reptiles. Int. J. Dev. Biol.
58, 935–948 (2014).
15. Norell, M. A. et al. The irst dinosaur egg was soft. Nature 583,
406–410 (2020).
16. Liu, J., Organ, C. L., Benton, M. J., Brandley, M. C. & Aitchison, J. C.
Live birth in an archosauromorph reptile. Nat. Commun. 8,
14445 (2017).
17. Piñeiro, G., Ferigolo, J., Meneghel, M. & Laurin, M. The oldest
known amniotic embryos suggest viviparity in mesosaurs. Hist.
Biol. 24, 620–630 (2012).
18. Blackburn, D. G. Evolution of viviparity in squamate reptiles:
reversibility reconsidered. J. Exp. Zool. B Mol. Dev. Evol. 324,
473–486 (2015).
19. Zhang, W. & Gao, K.-Q. Early Cretaceous evolution of
choristoderes in western Liaoning based on geographic and
stratigraphic evidence. J. Palaeogeogr. 16, 205–216 (2014).
20. Brinkman, D. B. & Dong, Z.-M. New material of Ikechosaurus
sunailinae (Reptilia: Choristodira) from the Early Cretaceous
Laohongdong Formation, Ordos Basin, Inner Mongolia,
and the interrelationships of the genus. Can. J. Earth Sci. 30,
2153–2162 (1993).
21. Scheyer, T. M. et al. A new, exceptionally preserved juvenile
specimen of Eusaurosphargis dalsassoi (Diapsida) and
implications for Mesozoic marine diapsid phylogeny. Sci. Rep. 7,
4406 (2017).
22. Matsumoto, R., Dong, L., Wang, Y. & Evans, S. E. The irst record
of a nearly complete choristodere (Reptilia: Diapsida) from the
Upper Jurassic of Hebei Province, People’s Republic of China.
J. Syst. Palaeontol. 17, 1031–1048 (2019).
23. Dial, T. R. & Carrier, D. R. Precocial hindlimbs and altricial
forelimbs: partitioning ontogenetic strategies in mallards
(Anas platyrhynchos). J. Exp. Biol. 215, 3703–3710 (2012).
24. Packard, M. J., Packard, G. C. & Boardman, T. J. Structure of
eggshells and water relations of reptilian eggs. Herpetologica 38,
136–155 (1982).
25. Pike, D. A., Andrews, R. M. & Du, W.-G. Eggshell morphology and
gekkotan life-history evolution. Evol. Ecol. 26, 847–861 (2012).
26. Hallmann, K. & Griebeler, E. M. Eggshell types and their
evolutionary correlation with life-history strategies in squamates.
PLoS ONE 10, e0138785 (2015).
27. Sander, P. M. Reproduction in early amniotes. Science 337,
806–808 (2012).
28. Ji, Q., Wu, X.-C. & Cheng, Y.-N. Cretaceous choristoderan
reptiles gave birth to live young. Naturwissenschaften 97,
423–428 (2010).
29. Stewart, J. R. et al. Uterine and eggshell structure and
histochemistry in a lizard with prolonged uterine egg retention
(Lacertilia, Scincidae, Saiphos). J. Morphol. 271, 1342–1351 (2010).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution | Voume 7 | Juy 2023 | 1131–1140 1139
Article https://doi.org/10.1038/s41559-023-02074-0
30. Heulin, B., Ghielmi, S., Vogrin, N., Surget-Groba, Y. & Guillaume,
C. P. Variation in eggshell characteristics and in intrauterine egg
retention between two oviparous clades of the lizard Lacerta
vivipara: insight into the oviparity-viviparity continuum in
squamates. J. Morphol. 252, 255–262 (2002).
31. Hou, L.-H., Li, P.-P., Ksepka, D. T., Gao, K.-Q. & Norell, M. A.
Implications of lexible-shelled eggs in a Cretaceous
choristoderan reptile. Proc. Biol. Sci. 277, 1235–1239 (2010).
32. Katsura, Y. Fusion of sacrals and anatomy in Champsosaurus
(Diapsida, Choristodera). Hist. Biol. 19, 263–271 (2007).
33. Whittington, C. M. et al. Understanding the evolution of
viviparity using intraspeciic variation in reproductive mode and
transitional forms of pregnancy. Biol. Rev. Camb. Philos. Soc. 97,
1179–1192 (2022).
34. Andrews, R. M. & Mathies, T. Natural history of reptilian
development: constraints on the evolution of viviparity.
BioScience 50, 227–238 (2000).
35. Andrews, R. M. in Reptilian Incubation: Environment, Evolution
and Behaviour (ed. Deeming, D. C.) 75–102 (Nottingham Univ.
Press, 2004).
36. Raerty, A. R., Evans, R. G., Scheelings, T. F. & Reina, R. D. Limited
oxygen availability in utero may constrain the evolution of live
birth in reptiles. Am. Nat. 181, 245–253 (2013).
37. Martínez, R. N., Simões, T. R., Sobral, G. & Apesteguia, S. A Triassic
stem lepidosaur illuminates the origin of lizard-like reptiles.
Nature 597, 235–238 (2021).
38. Starck, J. M., Stewart, J. R. & Blackburn, D. G. Phylogeny and
evolutionary history of the amniote egg. J. Morphol. 282,
1080–1122 (2021).
39. Simões, T. R. et al. The origin of squamates revealed by
a Middle Triassic lizard from the Italian Alps. Nature 557,
706–709 (2018).
40. Ford, D. P. & Benson, R. B. J. The phylogeny of early amniotes and
the ainities of Parareptiles and Varanopidae. Nat. Ecol. Evol. 4,
57–65 (2020).
41. Alroy, J. Paleobiology Database (National Science Foundation);
https://paleobiodb.org/#/
42. Laurin, M. The evolution of body size, Cope’s rule and the origin of
amniotes. Syst. Biol. 53, 594–622 (2004).
43. Brusatte, S. L., Benton, M. J., Ruta, M. & Lloyd, G. T. Superiority,
competition, and opportunism in the evolutionary radiation of
dinosaurs. Science 321, 1485–1488 (2008).
44. Stadler, T. Sampling-through-time in birth–death trees.
J. Theor. Biol. 267, 396–404 (2010).
45. Heath, T. A., Huelsenbeck, J. P. & Stadler, T. The fossilized birth–
death process for coherent calibration of divergence-time
estimates. Proc. Natl Acad. Sci. USA 111, E2957–E2966 (2014).
46. Zhang, C., Stadler, T., Klopfstein, S., Heath, T. A. & Ronquist, F.
Total-evidence dating under the fossilized birth–death process.
Syst. Biol. 65, 228–249 (2016).
47. Bapst, D. W., Wright, A. M., Matzke, N. J. & Lloyd, G. T. Topology,
divergence dates, and macroevolutionary inferences vary
between dierent tip-dating approaches applied to fossil
theropods (Dinosauria). Biol. Lett. 12, 20160237 (2016).
48. Bapst, D. W. paleotree: an R package for paleontological and
phylogenetic analyses of evolution. Methods Ecol. Evol. 3,
803–807 (2012).
49. Ronquist, F. & Huelsenbeck, J. P. MrBayes 3: Bayesian
phylogenetic inference under mixed models. Bioinformatics 19,
1572–1574 (2003).
50. Ronquist, F. et al. MrBayes 3.2: eicient Bayesian phylogenetic
inference and model choice across a large model space. Syst.
Biol. 61, 539–542 (2012).
51. Benton, M. J. et al. Constraints on the timescale of animal
evolutionary history. Palaeontol. Electron. 18.1.1FC, 1–106 (2015).
52. Matzke, N. J. & Wright, A. Inferring node dates from tip dates
in fossil Canidae: the importance of tree priors. Biol. Lett. 12,
20160328 (2016).
53. Rambaut, A., Drummond, A. J., Xie, D., Baele, G. & Suchard, M. A.
Posterior summarisation in Bayesian phylogenetics using Tracer
1.7. Syst. Biol. 67, 901–904 (2018).
54. Maddison, W. P. Missing data versus missing characters in
phylogenetic analysis. Syst. Biol. 42, 576–581 (1993).
55. Tarasov, S. Integration of anatomy ontologies and evo-devo
using structured Markov models suggests a new framework for
modeling discrete phenotypic traits. Syst. Biol. 68, 698–716 (2019).
56. Sanko, D. & Rousseau, P. Locating the vertices of a Steiner tree in
an arbitrary metric space. Math. Program. 9, 240–246 (1975).
57. Louca, S. & Doebeli, M. Eicient comparative phylogenetics on
large trees. Bioinformatics 34, 1053–1055 (2018).
58. Revell, L. J. phytools: an R package for phylogenetic
comparative biology (and other things). Methods Ecol. Evol. 3,
217–223 (2012).
59. Bell, M. A. & Lloyd, G. T. strap: an R package for plotting
phylogenies against stratigraphy and assessing their stratigraphic
congruence. Palaeontology 58, 379–389 (2015).
60. Schliep, K. P. phangorn: phylogenetic analysis in R. Bioinformatics
27, 592–593 (2011).
61. Paradis, E. Analysis of Phylogenetics and Evolution with R
(Springer-Verlag, 2011).
62. Swoord, D. L. & Maddison, W. P. Reconstructing ancestral
character states under Wagner parsimony. Math. Biosci. 87,
199–229 (1987).
63. Hanazawa, M., Narushima, H. & Minaka, N. Generating most
parsimonious reconstructions on a tree: a generalization of the
Farris–Swoord–Maddison method. Discrete Appl. Math. 56,
245–265 (1995).
64. Narushima, H. & Hanazawa, M. A more eicient algorithm for MPR
problems in phylogeny. Discrete Appl. Math. 80, 231–238 (1997).
65. Pagel, M., Meade, A. & Barker, D. Bayesian estimation of ancestral
character states on phylogenies. Syst. Biol. 53, 673–684 (2004).
66. Pagel, M., O’Donovan, C. & Meade, A. General statistical model
shows that macroevolutionary patterns and processes are
consistent with Darwinian gradualism. Nat. Commun. 13, 1113 (2022).
67. Pagel, M. Detecting correlated evolution on phylogenies:
a general method for the comparative analysis of discrete
characters. Proc. R. Soc. Lond. B Biol. Sci. 255, 37–45 (1994).
68. Pagel, M. & Meade, A. Bayesian analysis of correlated evolution of
discrete characters by reversible-jump Markov chain Monte Carlo.
Am. Nat. 167, 808–825 (2006).
69. Venditti, C., Meade, A. & Pagel, M. Multiple routes to mammalian
diversity. Nature 479, 393–396 (2011).
70. Baker, J., Meade, A., Pagel, M. & Venditti, C. Positive phenotypic
selection inferred from phylogenies. Biol. J. Linn. Soc. Lond. 118,
95–115 (2016).
71. Xie, W., Lewis, P. O., Fan, Y., Kuo, L. & Chen, M.-H. Improving
marginal likelihood estimation for Bayesian phylogenetic model
selection. Syst. Biol. 60, 150–160 (2011).
72. Plummer, M., Best, N., Cowles, K. & Vines, K. CODA: convergence
diagnosis and output analysis for MCMC. R News 6, 7–11 (2006).
73. Raftery, A. E., Gilks, W. R., Richardson, S. & Spiegelhalter, D. J.
in Markov Chain Monte Carlo in Practice (eds Gilks, W. R. et al.)
163–187 (Springer, 1996).
74. Griith, O. W. et al. Ancestral state reconstructions require
biological evidence to test evolutionary hypotheses: a case study
examining the evolution of reproductive mode in squamate
reptiles. J. Exp. Zool. B Mol. Dev. Evol. 324, 493–503 (2015).
75. Pyron, R. A. & Burbrink, F. T. Contrasting models of parity-mode
evolution in squamate reptiles. J. Exp. Zool. B Mol. Dev. Evol. 324,
467–472 (2015).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution | Voume 7 | Juy 2023 | 1131–1140 1140
Article https://doi.org/10.1038/s41559-023-02074-0
76. Recknagel, H., Kamenos, N. A. & Elmer, K. R. Common lizards
break Dollo’s law of irreversibility: genome-wide phylogenomics
support a single origin of viviparity and re-evolution of oviparity.
Mol. Phylogenet. Evol. 127, 579–588 (2018).
77. Gao, W. et al. Genomic and transcriptomic investigations of the
evolutionary transition from oviparity to viviparity. Proc. Natl Acad.
Sci. USA 116, 3646–3655 (2019).
78. Motani, R. et al. A basal ichthyosauriform with a short snout from
the Lower Triassic of China. Nature 517, 485–488 (2015).
79. Schoch, R. R. & Sues, H.-D. Osteology of the Middle Triassic
stem-turtle Pappochelys rosinae and the early evolution of the
turtle skeleton. J. Syst. Palaeontol. 16, 927–965 (2018).
Acknowledgements
We thank C. O’Donovan, M. Pagel, G. Ruxton and M. Sakamoto
for discussions during the course of this study, and M. Laurin and
G. Wagner for comments. B.J. was supported by the National
Science Foundation of China (award no. 42288201), Strategic
Priority Research Program (B) of the Chinese Academy of Sciences
(award no. XDB26000000) and Fundamental Research Funds for the
Central Universities (award no. 0206-14380137). M.J.B. was funded
by a Natural Environment Research Council (NERC) UK (grant no.
NE/P013724/1) and European Research Council Advanced Grant
(no. 788203). A.E. was funded by NERC UK grant nos. NE/L002434/1
and NE/P013724/1. This work was carried out using the computational
facilities of the Advanced Computing Research Centre, University of
Bristol (http://www.bris.ac.uk/acrc/).
Author contributions
B.J. and M.J.B. conceived the study. Y.H. and B.J. made the 3D CT
reconstruction. S.W. and S.L.K. carried out the scanning electron
microscopy and geochemical analyses of the specimen. A.E., J.N.K.,
J.S. and M.J.B. carried out the phylogenetic comparative analyses.
All authors contributed to data collection, interpreted the results and
wrote the manuscript.
Competing interests
The authors declare no competing interests.
Additional information
Extended data is available for this paper at
https://doi.org/10.1038/s41559-023-02074-0.
Supplementary information The online version
contains supplementary material available at
https://doi.org/10.1038/s41559-023-02074-0.
Correspondence and requests for materials should be addressed to
Baoyu Jiang.
Peer review information Nature Ecology & Evolution thanks
Susan Evans, Michel Laurin, Gunter Wagner and the other, anonymous,
reviewer(s) for their contribution to the peer review of this work.
Peer reviewer reports are available.
Reprints and permissions information is available at
www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with
regard to jurisdictional claims in published maps and
institutional ailiations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format,
as long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license, and indicate
if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons license and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this license, visit http://creativecommons.
org/licenses/by/4.0/.
© The Author(s) 2023
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution
Article https://doi.org/10.1038/s41559-023-02074-0
Extended Data Fig. 1 | Coronal CT slices of the embryo skull. a. Approximate central view. b. Approximate middle views. Abbreviations see Fig. 1.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution
Article https://doi.org/10.1038/s41559-023-02074-0
Extended Data Fig. 2 | A coronal CT slice shows axial and appendicular skeletons of the embryo. Abbreviations see Fig. 1.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution
Article https://doi.org/10.1038/s41559-023-02074-0
Extended Data Fig. 3 | See next page for caption.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution
Article https://doi.org/10.1038/s41559-023-02074-0
Extended Data Fig. 3 | Visualisation of the Structured Markov Models (SMMs)
used in this study. a. Diagrammatic representation of the SMMs. SMM-ind
= independent model; SMM-sw = switch-on dependency model. Numbers
represent amalgamated character states. Left digit represents viviparity = 1,
oviparity = 2. Right digit represents unmineralized egg = 1, parchment egg = 2,
rigid egg = 3. b. Index matrices for the eight SMM’s used in the ancestral state
analyses. Different greyscale values represent different rate categories. Equal
rates (ER) models: transitions between aggregated rates share a single rate
parameter; Component equal-rates (CER) models: transitions between states
among component characters share a single rate parameter; Component
symmetrical (CSYM) models: transitions between states among component
characters are symmetrical; Component all-rates-different (CARD) models:
transitions between states among component characters are all different.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution
Article https://doi.org/10.1038/s41559-023-02074-0
Extended Data Fig. 4 | Phylogeny of amniotes, showing known reproduction
mode + eggshell mineralisation, and EER of 80 extant and extinct species
(tips, to right), and inferred mean ancestral states for all branching points
(larger pie charts at nodes). The dominant inferred state towards the root (left)
is viviparity with EER. This is a consensus tree based on a sample of 100 trees
time-scaled using the FBD method (with node age constraints for major clades)
and component all rates different model without a switch-on dependency for
eggshell mineralisation and reproduction mode (CARD_ind.het) (best-fitting BT
model based on log BF score; see Supplementary Table 31a–d), which allows for
variable evolutionary rates on individual branches. Ikechosaurus sp. is indicated
in red and with a red arrow. Silhouettes as in Fig. 4.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Nature Ecoogy & Evoution
Article https://doi.org/10.1038/s41559-023-02074-0
Extended Data Fig. 5 | Same as Supplementary Fig. 4, but the nodes of
Lepidosauria and Squamata have been constrained to a non-viviparous
state. The best-fitting BT model for eggshell mineralisation and reproduction
mode presented here is a component equal-rates model without a switch-on
dependency for eggshell mineralisation and reproduction mode (CER_ind.het)
(see Supplementary Table 31d–g), which allows for variable evolutionary rates on
individual branches.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Content courtesy of Springer Nature, terms of use apply. Rights reserved
1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com
