untitled


How the Turtle Forms its Shell: A Paracrine
Hypothesis of Carapace Formation

JUDITH CEBRA-THOMAS1, FRASER TAN1, SEETA SISTLA1, EILEEN ESTES1,
GUNES BENDER1, CHRISTINE KIM2, PAUL RICCIO1, AND
SCOTT F. GILBERT1�
1Department of Biology, Swarthmore College, 500 College Avenue, Swarthmore,
Pennsylvania 19081
2Department of Biology, Bryn Mawr College, North Merion Avenue, Bryn Mawr,
Pennsylvania 19010

ABSTRACT We propose a two-step model for the evolutionary origin of the turtle shell. We
show here that the carapacial ridge (CR) is critical for the entry of the ribs into the dorsal dermis.
Moreover, we demonstrate that the maintenance of the CR and its ability to attract the migrating rib
precursor cells depend upon fibroblast growth factor (FGF) signaling. Inhibitors of FGF allow the CR
to degenerate, with the consequent migration of ribs along the ventral body wall. Beads containing
FGF10 can rearrange rib migration in the chick, suggesting that the CR FGF10 plays an important
role in attracting the rib rudiments. The co-ordinated growth of the carapacial plate and the ribs
may be a positive feedback loop (similar to that of the limbs) caused by the induction of Fgf8 in the
distal tips of the ribs by the FGF10-secreting mesenchyme of the CR. Once in the dermis, the ribs
undergo endochrondral ossification. We provide evidence that the ribs act as signaling centers
for the dermal ossification and that this ossification is due to bone morphogenetic proteins secreted
by the rib. Thus, once the ribs are within the dermis, the ossification of the dermis is not difficult to
achieve. This relatively rapid means of carapace formation would allow for the appearance of turtles
in the fossil record without obvious intermediates. J. Exp. Zool. (Mol. Dev. Evol.) 304B:558– 569,
2005. r 2005 Wiley-Liss, Inc.

The turtle shell is an evolutionarily novel
adaptation that functions in numerous ways. In
different species, the shell provides physical
protection, a shelter for aestivation or hibernation,
fat storage, calcium storage, and ionic buffering.
However, although the turtle shell is an apomor-
phy (new structure) that distinguishes the Chelo-
nian clade from all others, the elements
constructing the shell are not new. Unlike the
origins of chordates (wherein the notochord cells
and podocytes evolved) or the origin of vertebrates
(wherein the neural crest cell and osteocyte types
were first seen), the origin of the turtle shell is an
example of heterotopy (sensu Hall, ’99), wherein
pre-existing tissue types develop in new places.
The innovation of the turtle shell does not involve
the evolution of new tissue types, but how the
developmental instructions for making certain
tissues become used in new places.

The shell is composed of two main parts: the
dorsal carapace and the ventral plastron. Between
them, on the lateral sides, is a bridge. The shell is

not merely a box of dermal bone covering a pre-
existing reptilian body plan. Rather, the turtle
body has become extensively modified. Turtles
lack true lumbar vertebrae, and their thoracic
vertebrae (together with the sacrum and first
caudal vertebra) become part of the carapace. The
dorsal portion of the vertebrae fuses with the
midline of the shell, while the costal processes that
give rise to the ribs grow dorsolaterally into the
dermis, instead of moving ventrolaterally. Thus in
turtles, the dermis expands under the influence of
the ribs to form the rudiment of the carapace, and
the ribs later become part of the shell. Unlike any

Published online 20 June 2005 in Wiley InterScience (www.
interscience.wiley.com). DOI: 10.1002/jez.b.21059.

Received 10 December 2004; Accepted 2 May 2005

Grant sponsor: National Science Foundation; Grant number: IBN-
0079341 and IBN-0316025; Grant sponsor: Faculty Research Grants
from Swarthmore College; Grant sponsor: HHMI Grant to Swarth-
more College for Undergraduate Student Research.
�Correspondence to: Scott F. Gilbert, Department of Biology,

Swarthmore College, 500 College Avenue, Swarthmore, PA 19081.
E-mail: sgilber1@swarthmore.edu

r 2005 WILEY-LISS, INC.

JOURNAL OF EXPERIMENTAL ZOOLOGY (MOL DEV EVOL) 304B:558–569 (2005)



other vertebrate, the limb girdles of the turtles
develop within the ribs rather than outside
them (Rieppel, 2001; Ruckes, ’29; Zangerl, ’69;
Yntema, ’70).

Zangerl (’69) denoted three modes of carapace
development. The ‘‘thecal shell’’ mode, used by
hard-shelled turtle species, is characterized by
costal bones that develop in the dermis (immedi-
ately dorsal to the ribs) and extend to the marginal
bones at the lateral edges of the carapace. In
contrast, the Trionychidae lack marginal bones,
and their costal bones are much abbreviated. This
allows the upper dermal region to form the
characteristic ‘‘soft shell’’ morphology to give the
shell greater flexibility. Later in life, these turtles
often acquire bony ossicles in the dorsal regions of
the dermis, more distant from the ribs. Finally,
members of the marine genus Dermochelys do not
form any costal bones, generating instead thou-
sands of punctate ossicles in its dermis. This
allows for its more streamlined ‘‘leatherback’’
carapace. This report will focus on the costal bones
that form the major part of the turtle’s dorsal
shield in the typical hard-shelled turtles.

The carapacial ridge

The turtle egg is laid at the mid-gastrula stage.
While these early stages of turtle embryology
have not been extensively studied, turtle gastrula-
tion and somite formation are similar to those of
the chick (Pasteels, ’37, ’57; Ewert, ’85; Burke,
2004). The first sign that the organism is to
become a turtle rather than some other tetrapod
occurs at Yntema stage 14 (stages for Chelydra
Yntema, ’68; stage 14 is approximately equivalent
to Hamburger–Hamilton chick stage 24). At this
stage, there are the first signs of ridges on the
lateral surfaces of the embryo, dorsal to the limb
buds (Ruckes, ’29). At first, these ridges are seen
between the two limb buds, and only later do the
ridges extend anteriorly and posteriorly. This
structure has been named the carapacial ridge
(CR) (Burke, ’89a, b, ’91) and will eventually form
the outer edge of the carapace. The CR is formed
by a thickening of the ectoderm and is underlain
by a condensed somite-derived mesenchyme
(Nagashima et al., 2005). This is a typical config-
uration for an epithelial–mesenchymal interaction,
and the distributions of fibronectin and N-CAM in
the CR are similar to their locations in other
inductive sites such as the early limb bud or
feather primordia (Burke, ’89a, ’91).

Rucke’s (’29) observations of turtle embryos
described two important factors in the develop-
ment of the shell. First, there is an accelerated
lateral growth of the dorsal dermis of the trunk
compared to growth in the dorso-ventral plane.
Second, there is an apparent ‘‘ensnarement’’ of
the growing ribs by the CR. The involvement of
the ribs with the carapacial dermis results in their
growth in a predominantly lateral direction. The
limb girdles develop in typical tetrapod fashion,
but because of the growth trajectory of the ribs,
the pectoral girdle becomes ventral to and in-
cluded within the axial elements.

Yntema (’70) performed a series of somite
extirpation experiments on snapping turtles, con-
firming a somitic origin for the ribs and dermis of
the carapace. Post-otic somite pairs 12 through 21
are involved in forming the carapace in Chelydra.
A causal role for the CR in rib placement was
studied experimentally by Burke (’91). Surgical
methods were used to (a) remove the CR or (b)
prevent CR formation. In the first set of experi-
ments, in those cases where the CR did not
regenerate, the rib at the level of the surgery did
not grow in its normal trajectory but migrated
instead toward a neighboring region that did have
a CR. In the second set of experiments, tantalum
foil barriers between the somite and the lateral
plate mesoderm prevented CR formation. This
procedure had extreme results and surviving
embryos showed disruption of the body wall as
entire regions of the dermal carapace failed to
form. The ribs associated with these missing
regions interdigitated with the bones of the
plastron. Thus, the lateral development of the
turtle ribs appears to be directed by the CR, and in
the absence of the CR, these ribs grow ventrally
and enter the lateral plate, like the ribs of non-
chelonian vertebrates.

Loredo et al. (2001) were the first to analyze the
CR with molecular probes and found Trachemys
fgf10 expression in the mesenchyme condensed
beneath the CR. While they were able to detect
fgf8 expression in the apical ectodermal ridge of
the limb bud, they were unable to see fgf8
expression in the vicinity of the CR. Vincent
et al. (2003) found the turtle homologue of msx1
being expressed in the mesenchyme of the Emys
CR. This furthered the notion that the CR was
made through mesenchymal/epithelial interac-
tions similar to those that generate the limb bud.
By using RT-PCR, Kuraku et al. (2005) found
turtle orthologs of Sp5 and Wnt targets APDCC-1
and LEF-1 in the CR mesenchyme and ectoderm

TURTLE CARAPACE DEVELOPMENT 559



of the Chinese softshell turtle Pelodiscus. They
also found CRABP-1 expressed in the CR ecto-
derm; but they could not detect the expression of
either of the previously reported genes, msx-1 or
fgf10, in the CR mesenchyme of this species.
Species differences might be important in these
patterns because the costal bones of Pelodiscus
might form in different ways (Zangerl, ’69) and
the pattern of fgf10 distribution in the limbs of
Pelodiscus differed from the expression pattern
seen in the limbs of Trachemys.

Costal bones

The character and homology of the bony
elements of the turtle shell have a long history of
controversy enjoined by some of the great 19th-
century morphologists. Cuvier (1800) and Geof-
froy Saint-Hilaire (1818) agreed that the carapace
merely represented the expansions of the ribs and
vertebral spines. Carus (1834) was perhaps the
first to suggest that the carapace contained both
the endo- and the exoskeletal (dermal) tissue. He
proposed that the endoskeletal vertebrae, ribs,
and sternum were overlain by dermal ossifica-
tions. Rathke (1848), in an extensive monograph
on turtle development, confirmed the dual nature
of the carapace and Owen (1849) also correctly
recognized the presence of both dermal and
endochondral bone in the carapace. Gegenbaur
(1859) considered the ribs to be greatly expanded
transverse processes of the vertebrae, overlain
with dermal ossifications.

The controversy concerning the origin of the
carapacial bones continued to the end of the 20th
century. Goette (1899) made more detailed histo-
logical studies of the carapace and proposed that
the dermal carapace bones formed as outgrowths
of the periosteal collar around the ribs and
vertebrae. The studies of Kälin (’45) and Vallèn
(’42) confirmed Goette and showed the cartilagi-
nous ribs lying between two layers of the stratum
compactum in the thick dermis of the carapace.
These layers unite between the ribs, and costal
ossification is initiated within the double layer of
stratum compactum. Kälin (’45) thought that
discrete dermal ossifications later associated with
the ribs and spinous processes. This view that the
costal bones were derived from osteoderms (cuta-
neous bones) that secondarily fused with the ribs
and vertebrae was the predominant view among
paleontologists (Romer, ’56; Sukhanov, ’64; Car-
roll, ’88; Laurin and Reisz, ’95), and it gave rise to
the model of evolution wherein the turtle carapace

arose by the fusion of osteoderms from ancestral
Pareiassaurs (Lee, ’96, ’97).

Suzuki (’63) noted that in Trachemys, the
cartilaginous matrix of the ribs degenerates within
the periosteum and the ribs appear to induce the
ossification of the surrounding dermal cells.
Cherepanov (’97) similarly argued that osteo-
derms did not form in the carapace and proposed
that the costal bones originated as expansions of
the ribs and neural region of the vertebrae. The
observations of Gilbert et al. (2001) supported
Suzuki’s (’63) description of ossification of the
costal bones. Ossification of these elements was
seen to be initiated as spicules of bone extend-
ing from the intramembranous, periosteal collar
and surrounding the rib cartilages. The spicules
develop into the typical reticulated pattern of
trabeculae, and the ribs themselves grow by
apical apposition, invested in a dense ‘‘periosteo-
chondrogenetic’’ membrane. These observations
indicate that the ribs act as initiation centers
for the dermal ossification of costal bones. The
ossifying regions of the dermis extend towards
one another to eventually fuse. The data reported
in the present report confirm and extend
these observations and permit us to frame a
hypothesis to explain the rapid origin of the
turtle carapace.

We have been looking at how the carapace of the
turtle forms in the hard-shelled red-eared slider,
Trachemys scripta. While the data presented here
have enabled us to propose a model for carapace
formation, the research is still in a relatively early
stage. We envision two major stages in the
formation of the costal bones that form the bony
plates of the carapace. In the first stage, the rib
precursor cells enter the dermis rather than
migrating ventrally to form a rib cage (Fig. 1A
and B). Our data indicate that fibroblast growth
factor (FGF) signaling in the dorsolateral dermis
maintains the CR on each side of the turtle
embryo and that FGF signaling from the CR
alters rib precursor migration such that the ribs
grow dorsolaterally into the dermis (rather than
forming a rib cage). In the second process, the
developing ribs secrete bone morphogenetic pro-
teins (BMPs) that induce the costal bones that
form the plate of the carapace. Our studies suggest
that BMPs produced during the normal endochon-
dral ossification of the rib (Fig. 1C) induce
intramembranous bone formation in the dermis
surrounding it. The BMP signal is then propa-
gated by the developing bones. This wave of
ossification proceeds from each rib until the

J. CEBRA-THOMAS ET AL.560



ossified dermal regions meet and form the sutures
apposing the costal bones of the carapace.

MATERIALS AND METHODS

Whole-mount in situ hybridization

T. scripta eggs were purchased from the Kliebert
Turtle and Alligator Farm (Hammond, LA).
Embryos were dissected free of extraembryonic
membranes, fixed overnight in cold 4% parafor-
maldehyde in phosphate-buffered saline (PBS),
washed, dehydrated with methanol, and subjected
to whole-mount in situ hybridization as described
(Riddle et al., ’93) with a digoxygenin-labeled RNA
probe for turtle Fgf8 (Loredo et al., 2001).
Polyvinyl alcohol was added to the detection
solution to enhance the color reaction (Barth and
Ivarie, ’94).

Turtle explant cultures

Two-week T. scripta eggs (Greenbaum stage 15)
were disinfected by soaking them in 10% bleach,
followed by sterile dH2O and finally 70% ethanol.
This stage was chosen because the somatic
derivatives have become specified but the ribs
have not migrated out to any appreciable degree.
Fgf signaling is important for the specification of

the sclerotome (Huang et al., 2003), and we
wanted to be beyond that stage. Embryos were
dissected into sterile PBS and the extraembryonic
membranes were removed. Embryos were trans-
ferred to Hanks balanced salt solution (HBSS,
Sigma) with 50 mg/ml gentamycin (Gibco) and
2.5 mg/ml fungizone amphotericin B (Gibco). The
heads were removed and embryos were opened
along the ventral midline. The heart, digestive
tube, and mesonephros were removed.

The eviscerated trunk explants were cultured
ventral-side down on Transwell-Clear filters
(Costar) on top of Dulbecco’s modified Eagles
medium (DMEM, Sigma) supplemented with 2%
fetal calf serum (FCS, Mediatech), 50 mg/ml
gentamycin, 2.5 mg/ml fungazone, and 67 U/ml
nystatin (Sigma). For inhibition of FGF signaling,
explants were treated with 10 mM SU5402 (SU-
GEN Corp.) in culture media with 0.3% dimethyl
sulfoxide, while control explants were treated with
culture media containing 0.3% dimethyl sulfoxide
alone. Explants were cultured at 301C with 5%
CO2 for 3 days, and photographed using an
Olympus DP-12 digital camera and SZX12 stereo-
microscope. Explants were fixed in 4% parafor-
maldehyde and embedded in paraffin. Ten-
micrometer sections were rehydrated and stained
with 2 mg/ml Hoechst 33258 (Sigma).

Fig. 1. Rib involvement in carapace formation. (A) Stage 16/17 embryo of red-eared slider turtle, Trachemys scripta. The
drawing was rendered from life by Auguste Sonrel for Louis Agassiz’ 1857 volume Contributions to the Natural History of the
United States. III. The Embryology of the Turtle. The turtle itself may have been one of those collected by Agassiz’ friend Henry
David Thoreau, from Walden Pond. (B) Cross-section of a T. scripta embryo at approximately the same stage, showing the rib
(arrow) moving through the myotome to the CR (arrowhead). (C) Hall stain of a 90-day T. scripta hatchling showing regions of
endochondral bone formation and the initial induction of costal bones. The cartilage is stained blue, the bone is stained red, and
the costal bones can be seen forming around the proximal portion of the ribs. (B) and (C) are taken from Gilbert et al. (2001).

TURTLE CARAPACE DEVELOPMENT 561



Treatment of chick trunk explants
with FGF10 beads

Day 5 chick embryos (stage 26, Hamburger and
Hamilton, ’51) were dissected into HBSS with
50 mg/ml gentamycin. The head and ventral
organs were removed, and the dorsal trunk
explants were cultured on Transwell-clear nucleo-
pore membranes DMEM supplemented with 2%
FCS and 50 mg/ml gentamycin as described above.
Heparin-acrylic beads (Sigma) were manually
fractionated, rinsed 3 times with PBS, and soaked
with 0.25 mg recombinant human FGF10 (Re-
search Diagnostics, Inc.) as described by Weaver
et al. (2000). FGF10-coated beads were washed
with PBS and transferred to a small slit in one
flank adjacent to the somites. Control embryos
were exposed to washed heparin-acrylic beads.
Explants were cultured at 371C with 5% CO2 for
6 days, photographed as above, fixed, stained with
0.1% Alcian Green 2GX (Sigma) in acidic alcohol,
dehydrated, cleared in methyl salicylate, and
rephotographed.

Analysis of bone formation and BMP2
expression in the carapace

Immunohistochemistry with antibodies to PS1
(a gift of Dr. Carl-Henrik Heldin, Ludwig Institute
for Cancer Research, Sweden) and Hall’s stain for
cartilage and bone were performed on paraffin
sections of hatchling turtle carapace as previously
described (Clark et al., 2001). For RNA isolation,
the carapace of a 14-day old T. scripta hatchling
was dissected away from internal tissues and
incubated in 2% (w/v) trypsin (Type II, Sigma) in
PBS at room temperature for 50 min (modified
from Jiang et al., ’99). The epidermal scutes were
removed manually and the ribs were dissected
away from the surrounding dermal tissue. The
ribs and dermis were homogenized in TRIZOL
(Sigma) and RNA was isolated according to the
manufacturer’s recommendations. cDNA was pre-
pared with Superscript III M-MLV reverse tran-
scriptase using an oligo(dT) primer (Invitrogen).
Polymerase chain reaction was performed using
primers derived from the turtle bmp2 sequence
published by E. LeClaire (accession AY327846)
with an annealing temperature of 601C. The
sequences were gagctcccagaagcaagtgg (tbmp2F)
and ggcaccatatcctggtggg (tbmp2R). Primers for
the housekeeping gene GAPDH (George-
Weinstein et al., ’96) were used to control for
cDNA quality.

RESULTS AND DISCUSSION

Evidence that Fgf signals maintain the
carapacial ridge and direct the migration

of the rib precursor cells

The first sign that an organism is to become
a turtle rather than some other reptile is the
appearance of the CR at Yntema stage 14/
Greenbaum stage 15 (equivalent to Hamburger–
Hamilton chick stage 24) (Yntema, ’68; Green-
baum, 2002). Given the similarity of structure of
the CR to the limb bud (Burke, ’89b), Loredo et al.
(2001) searched for fgf8 and fgf10 expression in
the CR. These are the two paracrine factors
thought to be responsible for the formation and
expansion of the limb bud (see Ohuchi et al., ’99;
Lewandowski et al., 2000). While in situ hybridi-
zation with cloned turtle probes showed that fgf10
was expressed in the condensed mesenchyme
directly beneath the CR ectoderm (Fig. 2B), no
fgf8 expression in the CR was observed. We can
now report that turtle fgf8 expression is seen in
the immediate vicinity of the CR. This fgf gene is
not expressed in the CR itself; rather, turtle fgf8 is
expressed in the distal tip of the rib as it enters
the CR (Fig. 2A). If the positive feedback between
the rib and the CR is similar to that between the
mesoderm and apical ectodermal ridge of the
amniote limb, then the mutual induction between

Fig. 2. Expression of fg8 and fgf10 as the ribs enter the
CR. (A) Whole-mount in situ hybridization showing fgf8
expression in the distal tip of the ribs (arrow) as they enter the
CR. (B) Section in situ hybridization of fgf10 expression in
the carapacial ridge mesoderm. The rib can be seen entering
the CR. (B) is taken from Loredo et al. (2001).

J. CEBRA-THOMAS ET AL.562



these two Fgfs could serve as the mechanism by
which the growth of the rib and the carapace is co-
ordinated. Such a positive feedback loop is
mediated through the translocation of b-catenin
into the nucleus of the apical ridge ectoderm
(Kawasaki et al., 2001), and such a nuclear
translocation of b-catenin has recently been shown
in the Pelodiscus CR ectoderm (Kuraku et al.,
2005). We are also investigating the possibility
that other FGFs (such as those found in the apical
ectodermal ridge) may also be important in CR
initiation or outgrowth.

FGF signaling is critical for CR
maintenance and rib ensnarement

To determine if Fgf signaling is required for
the formation of the CR and the ensnarement of
the ribs into the dermis, we cultured embryonic
turtle dorsi in an inhibitor of Fgf signaling. Turtle
embryos staged at Yntema 14/Greenburg 15 were
isolated after repeated washing of the eggs in
ethanol and bleach. The embryos had developed
mature somites and early CR (Fig. 3A and B). The
head and viscera were removed and the embryo
explants were cultured on nucleopore membranes
for 3 days at 301C. When cultured in control
medium, the CR was maintained, the ribs entered
the CR, and the CR became elevated above the rest
of the dorsum in all cases (21/21, Fig. 3C). When
such embryonic explants were grown in medium
that also included 10 mM SU5402, an inhibitor of
Fgf signaling (Mohammadi et al., ’97; Mandler and
Neubuser, 2001), the CR degenerated, and the ribs
proceeded to grow out towards the flanks, as they
would in most amniotes (in 20/20 cases, Fig. 3D).
Transverse sections through control explants
show a well-developed CR above a region of
densely packed mesenchyme (Figs. 3E & G). In
contrast, sections through SU5402-treated ex-
plants show that the CR region has largely
reverted to simple epithelium and is underlain
with sparse mesenchyme (Figs. 3F and H).

The question then became whether the effect
of Fgf was two-fold (i.e., involved in both the

formation/maintenance of the CR and in the
migration of the rib precursors) or indirect (i.e.,
involved in forming the CR, while the CR uses
some other mechanism to ensnare the ribs). While
it is obvious that the Fgf inhibitor caused the CR
to degenerate (Fig. 3D), the direct effect of Fgf
signaling on the rib ensnarement had to be tested
some other way. To this end, we cultured 5-day
chick dorsi (HH 26) in the same manner as our
control turtle explants. We implanted FGF10-
coated beads or control (saline) beads into the
dermis between the somite and the lateral plate
and cultured for the explants for 6 days. In all
cases (11/11), the ribs on the side of the FGF10-
coated beads had migrated to the beads and had
stopped there (Fig. 4B and C). The FGF10-coated
beads appeared to have ensnared the ribs. In the
embryos implanted with the control beads, each
set of ribs ignored the beads (6/6), proceeding
outward towards the flanks (Fig. 4A). It therefore
appears that FGFs are involved in both maintain-
ing the CR and altering the migration of the rib
precursor cells.

Formation of the carapacial costal bones:
evidence for BMP signaling

The rib precursor cells enter the dermis of the
shell a short distance from their origin in the
vertebrae, and they grow laterally within the
carapacial dermis (Ruckes, ’29; Burke, ’89a, b;
Gilbert et al., 2001). Initially the ribs are cartila-
ginous, but they undergo normal endrochondral
ossification to become bone. As endochondral
ossification ensues, the ribs appear to become the
organizing centers for the costal bones that make
the plate of the carapace (Figs. 1C, 5A). There is a
1:1 correspondence between the ribs and the
costal bones of the carapace (Zangerl, ’69; Gilbert
et al., 2001). The costal bones form around the
ribs by intramembranous ossification (Kälin, ’45;
Burke, ’91; Gilbert et al., 2001). Thus, the
carapace is a composite of endochondral axial
skeleton (from the ribs) plus intramembranous
dermal bone. The costal bones begin to form as the

Fig. 3. Fgf signaling is critical for rib entry into the CR. (A) Trachemys scripta embryo at stage 15. (B) Dorsal explant of
stage 15 T. scripta embryo on a nucleopore membrane in preparation for culture. The CR is elevated and can be seen running
along both flanks (arrow). (C) T. scripta dorsum taken at stage 15 and cultured for 3 days. The CRs (arrow) have been
maintained and are elevated above the dorsal surface. The rib primordia have entered into them. The apical ectodermal ridge
can also be seen on the limb bud (arrowhead). (D) T. scripta embryo from (B) cultured for 3 days in medium containing 10 mM
SU5402. The CRs have degenerated, and the ribs have not entered the dermis, but have extended along the flank as they do in
other vertebrates. (E) Section through control explant stained with the nuclear dye Hoechst 33258 showing a fully developed CR
(arrow). (F) Section through an SU5402-treated explant showing that the CR region has reverted to simple epithelium. (G, H)
Higher-magnification view of the CR region from (E) and (F).

TURTLE CARAPACE DEVELOPMENT 563



J. CEBRA-THOMAS ET AL.564



ribs become encased in a thin tube of bone, and
trabeculae extend from this bony casing (Fig. 5B).
Later, spicules form between the rib and the
epidermis, forming a pattern reminiscent of the
formation of the mandible around Meckel’s carti-
lage (Suzuki, ’63; see Tyler and Hall, ’76;
Takahashi et al., ’91; Fig. 5A). The most intense
area of costal bone formation is initially located at
the proximal region of the ribs, where the ribs had
first entered the dermis.

One clue as to how the ribs might induce the
intramembranous ossification of dermis into bone
comes from the secretion of paracrine factors
during the endochondral ossification of the ribs
(Vortkamp et al., ’96). Indian hedgehog (Ihh)
secreted by the ribs’ prehypertrophic cartilage
induces BMPs in the perichondrium. Pathi et al.
(’99) demonstrated that in chick limbs, perichon-
drial BMP 2, 4, 5, and 7 are induced by
endogenous and ectopic Ihh, and Wu et al. (2001)
demonstrated the induction of BMP2/4 by Ihh in
chick jaw tissue. Both Ihh and BMPs are known to
induce bone formation in surrounding competent
cells (Barlow and Francis-West, ’97; Ekanayake
and Hall, ’97; Nakamura et al., ’99; Pathi et al.,
’99; Wang et al., ’99).

The competence of dermal cells to respond to
BMPs by producing intramembranous bone has
been demonstrated in adult dermal and periosteal
tissues. Indeed, the aberrant expression of BMP4
in dermal tissues is thought to be the cause of the
fibrodysplasia ossificans progressiva, a disease
wherein connective tissue is converted into bone
by BMPs ectopically secreted by lymphoid cells
(Shafritz et al., ’96; Gannon et al., ’97; Lanchoney
et al., ’98). The conversion of periosteal cells into
bone by purified BMP has also been used in tissue
engineering (Franceschi et al., 2000; Rutherford
et al., 2002). Conversely, ectopic bone formation
can also occur in the human dermis as a result
of deficiencies in the BMP antagonist, Noggin
(Lucotte et al., ’99; Marcelino et al., 2001; Brown
et al., 2002). This ability to induce bone in adult
tissues is an important consideration, since the
turtle forms most of its costal bones after it has
hatched.

Using Alcian blue and Alizarin red (Hall Stain),
we have confirmed Suzuki’s PAS-Schiff data and
have shown the formation of the costal bones
around the hypertrophic cartilage of the ribs
(Figs. 5A and B). Furthermore, we have coupled
this with immunohistological staining for the
presence of the phosphorylated (activated) form
of Smad1, a transcription factor activated by

Fig. 4. FGF10 can redirect rib outgrowth in chicken
embryo. Dorsal explants of 5 d chick embryos (HH26) were
placed in organ culture for 6 days. Either FGF10-soaked or
control beads (�) were placed dorsally on one side of the
embryos, and compared with the unoperated side. (A) In those
chick explants with the saline-coated beads, the nascent ribs
(arrows) developed normally. (B) In those explants with the
FGF10-coated beads, the ribs developed normally on the
unoperated side. On the operated side, the somitic rib-forming
cells grew towards the beads and did not progress past them.
(C) Alcian green staining shows the fusion of the condensing
chondrocytes near the FGF10 beads and the extension of the
ribs on the opposite side.

TURTLE CARAPACE DEVELOPMENT 565



Fig. 5. Formation of the costal bones. (A) Cross section of the carapace of a 106-day T. scripta turtle showing a rib near its
site of entry in the dermis. The bone originally forms on the epidermal side, reminiscent of mandibular bone. Hall stain makes
the cartilage blue and the bone red. (B) Section through a more posterior region of the carapace stained with Hall stain. The rib
cartilage (blue) is surrounded with bone (red) extending out as trabeculae. (C) Serial section showing the same rib stained with
PS1 antibody to localize regions of BMP signaling. Nuclear expression of phosphorylated Smad1 (brown) is seen in the
periosteum of the bone and in the immediately adjacent dermal cells. (D) Higher magnification of region boxed in (C). (E)
Section through a third, more posterior rib that is just initiating bone formation showing the extent of staining with the PS1
antibody. (F) Negative control antibody staining to a serial section of the same rib. (G) cDNA prepared from 14-day T. scripta
hatchling dermis (lanes 2 and 5) and ribs (lanes 3 and 6) was subjected to RT-PCR analysis using primers for BMP2 (lanes 1–3)
and GAPDH (lanes 4–6). Lanes 1 and 4 contain no added cDNA.

J. CEBRA-THOMAS ET AL.566



BMPs (Kurata et al., 2001; Faure et al., 2002). The
dermal cells around the rib show a high degree of
staining, as evidenced by their black nuclei and
brown cytoplasm (Figs. 5C–E). While the rib and
its perichondrium remain unstained, there is
intense staining in the periosteum and in the cells
adjacent to it. Moreover, a high level of staining is
observed in the cells that are in the area destined
to become bone. Such staining for phosphorylated
Smad1 was not seen in controls where the primary
antibody was absent (Fig. 5F).

Thus, it appears that BMP signaling from the
rib during endochondral ossification is able to
induce intramembranous ossification in the der-
mal cells surrounding them. Moreover, as the cells
ossify, they appear to transmit the BMP signal to
the cells surrounding them, thereby continuing a
cascade through which BMP would be produced by
the dermal cells as they ossify. To confirm this, we
separated ribs and dermis from 14-day T. scripta
hatchlings, prepared cDNA, and carried out RT-
PCR with primers for Trachemys bmp2 (Fig. 5G).
Bmp2 transcripts were represented in the cDNA
from both the dermis (lane 2) and ribs (lane 3).

CONCLUSION

Our present study suggests a two-step mechan-
ism for the production of the costal bones that
make the plate of the turtle carapace. First, FGF
signaling is responsible for the maintenance, and
possibly the initiation, of the CR and for the
ensnarement of the ribs into the dermis. We have
shown fgf10 expression in the CR mesenchyme
and fgf8 expression in the distal tip of the ribs
entering the mesenchyme. Moreover, by inhibiting
FGF signaling, we observed the degeneration of
the CR and the migration of the turtle ribs. This
supports the conclusions of Burke (’91) that the
CR might be providing chemotactic factors needed
for the continued lateral growth of the ribs. The
second step is a BMP-dependent stage wherein the
dermal cells respond to BMPs and perhaps other
bone-forming proteins elaborated by the ribs as
these ribs undergo endochondral ossification. We
had previously shown (Gilbert et al., 2001) that
the rib is the organizing center for each costal
bone. Our present study shows that the dermal
cells surrounding the rib are responding to BMPs
and that the location of BMP reception presages
bone formation around the ribs. We have also
shown that both the ribs and the surrounding
dermis express bmp2 transcripts, as expected if a
signaling cascade was initiated by the ribs.

Such a mechanism could explain the rapid rise of
turtles in the fossil record. The order Chelonia
emerges abruptly in the late Triassic with the
fossil species Proganochelys (Gaffney, ’90; Rieppel
and Reisz, ’99). This reptile had the characteristic
derived trunk morphology now associated with
turtles. Thus, the distinctive morphology of the
turtle appears to have arisen suddenly. We can
propose a hypothesis that may explain at least part
of how this might happen. The key innovation is to
getting the ribs into the dermis. Once there,
variation in the population might enable some
individuals to use this heterotopic placement of
ribs to form a shell. If they could form a positive
feedback loop between the rib and the CR (e.g.,
through Fgf10 and Fgf8), they could co-ordinate
rib and carapace growth. When the ribs undergo
normal endochodral ossification, the BMPs would
induce the costal bones that form the plate of the
carapace. (This may involve overpowering natural
inhibitors of BMPs that are secreted by the
dermis.) This mechanism, wherein the displace-
ment of a tissue allows it to induce structures at
new locations, has been proposed by Brylski and
Hall (’88) to account for the rapid emergence of
the fur-lined cheek pouches of pocket gophers. The
compatibility of our findings with those of the
turtle fossil record has been noted by paleontolo-
gists (Rieppel, ’01).

ACKNOWLEDGMENTS

This work was supported by the National
Science Foundation (Grants IBN-0079341 and
IBN-0316025 to S.F.G.) and by faculty research
grants from Swarthmore College. F.T., S.S., P.R.,
and G.B. were supported by an HHMI grant to
Swarthmore College for undergraduate student
research. We thank C.-H. Heldin for the PS1
antibodies, SUGEN Corporation for SU5402, and
Nicole Adelman and Julia Landry for help with
preparing the hatchling turtle rib and dermis
cDNA.

LITERATURE CITED

Barlow AJ, Francis-West PH. 1997. Ectopic application of
recombinant BMP-2 and BMP-4 can change patterning
of developing chick facial primordia. Development 124:
391–398.

Barth J, Ivarie R. 1994. Polyvinyl alcohol enhances detection
of low abundance transcripts in early stage quail embryos in
a nonradioactive whole mount in situ hybridization techni-
que. BioTechniques 17:324–327.

Brown DJ, Kim TB, Petty EM, Downs CA, Martin DM,
Strouse PJ, Moroi SE, Milunsky JM, Lesperance MM. 2002.

TURTLE CARAPACE DEVELOPMENT 567



Autosomal dominant stapes ankylosis with broad
thumbs and toes, hyperopia, and skeletal anomalies is
caused by heterozygous nonsense and frameshift mutations
in NOG, the gene encoding noggin. Am J Hum Genet 71:
618–624.

Brylski P, Hall BK. 1988. Ontogeny of a macroevolutionary
phenotype: the external cheek pouches of geomyoid rodents.
Evolution 42:391–395.

Burke AC. 1989a. Development of the turtle carapace: Impli-
cations for the evolution of a novel Bauplan. J. Morphol.
199:363–378.

Burke AC. 1989b. Epithelial–mesenchymal interactions in the
development of the Chelonian Bauplan. In: Splechtana H,
Hilgers H, editors. Trends in vertebrate morphology (Progr
Zool 35). Stuttgart: Fischer Verlag. p 206–209.

Burke AC. 1991. The development and evolution of the turtle
body plan: inferring intrinsic aspects of the evolutionary
process from experimental embryology. Am Zool 31:616–627.

Burke AC. 2004. Gastrulation in reptiles. In: Stern C, editor.
Gastrulation. Cold Spring Harbor, NY: Cold Spring Harbor
Press. p 205–218.

Carroll RL. 1988. Vertebrate paleontology and evolution. New
York: W.H. Freeman and Company.

Carus KG. 1834. Lehrbuch der vergeichenden Zootomie,
2nd edition. Leipzig: Ernst Fleischer.

Cherepanov GO. 1997. The origin of the bony shell of turtles
as a unique evolutionary model in reptiles. Russ J Herpet
4:155–162.

Clark K, Bender G, Murray BP, Panfilis K, Cook S, Davis R,
Murnen K, Tuan RS, Gilbert SF. 2001. Evidence for the
neural crest origin of turtle plastrum bones. Genesis
31:111–117.

Cuvier, G. 1800. Lec-ons d’Anatomie Comparée. Paris:
Boudouin.

Ekanayake S, Hall BK. 1997. The in vivo and in vitro effects
of bone morphogenetic protein-2 on the development of
the chick mandible. Int J Dev Biol 41:67–81.

Ewert MA. 1985. Embryology of Turtles. In: Gans C, Billett F,
editors. Biology of the Reptilia. New York: John Wiley.
p 74–255.

Faure S, de Santa Barbara P, Roberts DJ, Whitman M. 2002.
Endogenous patterns of BMP signaling during early chick
development. Dev Biol 244:44–65.

Franceschi RT, Wang D, Krebsbach PH, Rutherford RB. 2000.
Gene therapy for bone formation: in vitro and in vivo
osteogenic activity of an adenovirus expressing BMP7. J Cell
Biochem 78:476–486.

Gaffney ES. 1990. The comparative osteology of the
Triassic turtle Proganochelys. Bull Am Mus Nat Hist
194:1–263.

Gannon FH, Kaplan FS, Olmsted E, Finkel GC, Zasloff MA,
Shore E. 1997. Bone morphogenetic protein 2/4 in early
fibromatous lesions of fibrodysplasia ossificans progressiva.
Hum Pathol 28:339–343.

Gegenbaur C. 1859. Grundgüge der vergleichenden Anatomie.
Leipzig.

Geoffroy Sainte-Hilaire E. 1818. Philosophie Anatomique; Des
Organes Respiratoires sous le Rapport de la Détermination
et de l’Identité de leurs Pièces Osseuses. Paris: J.B.
Baillière.

George-Weinstein M, Gerhart J, Reed R, Flynn J, Callihan B,
Mattiacci M, Miehle C, Foti G, Lash JW, Weintraub H. 1996.
Skeletal myogenesis: the preferred pathway of chick embryo
epiblast cells in vitro. Dev Biol 173:279–291.

Gilbert SF, Loredo GA, Brukman A, Burke AC. 2001.
Morphogenesis of the turtle shell:the development of a
novel structure in tetrapod evolution. Evol Dev 3:47–58.

Goette, A. 1899. Über die Entwicklung des knöckeren
Rückenschides (Carapax) der Schildkröten. Z Wiss Zool
66:407–434.

Greenbaum, E. 2002. A standardized series of embryonic
stages for the emyd turtle Trachemys scripta. Can J Zool
80:1350–1370.

Hall BK. 1999. Evolutionary developmental biology, 2nd
edition. Boston: Kluwer Academic Publishers.

Hamburger V, Hamilton H. 1951. A series of normal stages
in the development of the chick embryo. J Morphol 88:
49–92.

Huang R, Stolte D, Kurz H, Ehehalt F, Cann GM, Stockdale
FE, Patel K, Christ B. 2003. Ventral axial organs regulate
expression of myotomal Fgf-8 that influences rib develop-
ment. Dev Biol 255:30–47.

Jiang T, Jung H, Widelitz, RB, Chong C. 1999. Self-organiza-
tion of periodic patterns by dissociated feather mesenchymal
cells and the regulation of size, number and spacing of
primordia. Development 126:4997–5009.

Kälin J. 1945. Zur Morphogenese des panzers bei den
Schildkröten. Acta Anat 1:144–176.

Kawasaki Y, Capdevila J, Buscher D, Itoh T, Rodriguez-
Esteban C, Izpisua Belmonte JC. 2001. WNT signals control
FGF-dependent limb initiation and AER induction in the
chick embryo. Cell 104:891–900.

Kurata T, Nakabayashi J, Yamamoto TS, Mochii M, Ueno N.
2001. Visualization of endogenous BMP signaling during
Xenopus development. Differentiation 67:33–40.

Kuraku S, Usuda R, Kuratani S. 2005. Comprehensive survey
of carapacial ridge-specific genes in turtle implies co-option
of some regulatory genes in carapace evolution. Evol Dev
7:3–17.

Lanchoney TF, Olmsted EA, Shore EM, Gannon FA, Rosen V,
Zasloff MA, Kaplan FS. 1998. Characterization of bone
morphogenetic protein 4 receptor in fibrodysplasia ossifi-
cans progressiva. Clin Orthop 346:38–45.

Laurin M, Reisz RR. 1995. A reevaluation of early amniote
phylogeny. Zool J Linn Soc 113:165–223.

Lee MSY. 1996. Correlated progression and the origin of
turtles. Nature 379:812–815.

Lee MSY. 1997. Pareiasaur phylogeny and the origin of
turtles. Zool J Linnean Soc 120:197–280.

Lewandoski M, Sun X, Martin GR. 2000. Fgf8 signalling from
the AER is essential for normal limb development. Nat
Genet 26:460–463.

Loredo GA, Brukman A, Harris MP, Kagle D, Leclair EE,
Gutman R, Denney E, Henkelman E, Murray BP, Fallon JF,
Tuan RS, Gilbert SF. 2001. Development of an evolutiona-
rily novel structure: fibroblast growth factor expression in
the carapacial ridge of turtle embryos. J Exp Zool 291:
274–281.

Lucotte G, Semonin O, Lutz P. 1999. A de novo heterozygous
deletion of 42 base-pairs in the noggin gene of a fibrodys-
plasia ossificans progressiva patient. Clin Genet 56:469–470.

Mandler M, Neubuser A. 2001. FGF signaling is necessary for
the specification of the odontogenic mesenchyme. Dev Biol
240:548–559.

Marcelino J, Sciortino CM, Romero MF, Ulatowski LM,
Ballock RT, Economides AN, Eimon PM, Harland RM,
Warman ML. 2001. Human disease-causing NOG missense
mutations: effects on noggin secretion, dimer formation, and

J. CEBRA-THOMAS ET AL.568



bone morphogenetic protein binding. Proc Natl Acad Sci
USA 98:11353–11358.

Mohammadi M, McMahon G, Sun L, Tang C, Hirth P, Yeh BK,
Hubbard SR, Schlessinger J. 1997. Structures of the
tyrosine kinase domain of fibroblast growth factor receptor
in complex with inhibitors. Science 276:955–960.

Nagashima H, Uchida K, Yamamoto K, Kuraku S, Usuda R,
Kuratani S. 2005. Turtle-chicken chimera: An experimental
approach to understanding evolutionary innovation in the
turtle. Dev Dyn 232:149–161.

Ohuchi H, Nakagawa T, Itoh N, Noji S. 1999. FGF10 can
induce Fgf8 expression concomitantly with En1 and R-fng
expression in chick limb ectoderm, independent of its
dorsoventral specification. Dev Growth Differ 41:665–673.

Owen R. 1849. On the development and homologies of the
carapace and plastron of the chelonian reptiles. Phil Trans
R Soc Lond (Biol) 1849:151–171.

Pasteels JJ. 1937. Etudes sur la gastrulation des vertébrés
méroblastiques. 11 Reptiles. Arch Biol Paris 48:105–148.

Pasteels JJ. 1957. Une table analytique du développement des
Reptiles. 1 Stades de gastrulation chez les Cheloniens et les
Lacertiliens. Ann Soc R Zool Belg 87:217–241.

Pathi S, Rutenberg JB, Johnson RL, Vortkamp A. 1999.
Interaction of Ihh and BMP/Noggin signaling during
cartilage differentiation. Dev Biol 209:239–253.

Rathke H. 1848. Über die Entwicklung der Schildkröten.
Braunschweig: Friederich Vieweg und Sohn.

Riddle RD, Johnson RL, Laufer E, Tabin C. 1993. Sonic
hedgehog mediates the polarizing activity of the ZPA. Cell
75:1401–1416.

Rieppel O, Reisz RR. 1999. The origin and early evolution
of turtles. Ann Rev Ecol Syst 30:1–22.

Rieppel O. 2001. Turtles as hopeful monsters. Bioessays
23:987–991.

Romer AS. 1956. Osteology of the reptiles. Chicago: University
of Chicago Press.

Ruckes H. 1929. The morphological relationships between the
girdles, ribs, and carapace. Ann NY Acad Sci 13:81–120.

Rutherford RB, Moalli M, Franceschi RT, Wang D, Gu K,
Krebsbach PH. 2002. Bone morphogenetic protein-trans-
duced human fibroblasts convert to osteoblasts and form
bone in vivo. Tissue Eng 8:441–452.

Shafritz AB, Shore EM, Gannon FH, Zasloff MA, Taub R,
Muenke M, Kaplan FS. 1996. Overexpression of an

osteogenic morphogen in fibrodysplasia ossificans progressi-
va. N Engl J Med 335:555–561.

Sukhanov VB. 1964. Subclass Testudinata. In: Rozhdest-
vensky AK, Tatarinov LP, editors. Nedra: Principles of
paleontology: amphibians, reptiles and birds. Nedra: Mos-
cow. p 354–438 (in Russian).

Suzuki HK. 1963. Studies on the osseous system of the slider
turtle. Ann NY Acad Sci 109:351–410.

Takahashi Y, Bontoux M, Le Douarin NM. 1991. Epithelio-
mesenchymal interactions are critical for Quox 7 expres-
sion and membrane bone differentiation in the neural
crest derived mandibular mesenchyme. EMBO J 10:
2387–2393.

Tyler MS, Hall BK. 1976. Epithelial influences on skeletogen-
esis in the mandible of the embryonic chick. Anat Rec
188:229–240.

Vallén E. 1942. Beiträge zur Kenntnis der Ontogenie und der
vergleichenden Anatomie des Schildkrötenpanzers. Acta
Zool Stockholm 23:1–127.

Vincent C, Bontoux M, LeDouarin NM, Pieau C, Monsoro-
Burq AH. 2003. Msx genes are expressed in the carapacial
ridge of turtle shell: a study of the European pond turtle
Emys orbicularis. Dev Genes Evol 213:464–469.

Vortkamp A, Lee K, Lanske B, Segre GV, Kronenberg HM,
Tabin CJ. 1996. Regulation of rate of cartilage differentia-
tion by Indian hedgehog and PTH-related protein. Science
273:613–622.

Wang YH, Rutherford B, Upholt WB, Mina M. 1999. Effects of
BMP-7 on mouse tooth mesenchyme and chick mandibular
mesenchyme. Dev Dyn 216:320–335.

Weaver M, Dunn NR, Hogan BLM. 2000. Bmp4 and Fgf10 play
opposing roles during lung bud morphogenisis. Develop-
ment 127:2695–2704.

Wu Q, Zhang Y, Chen Q. 2001. Indian hedgehog is an essential
component of mechanotransduction complex to stimulate
chondrocyte proliferation. J Biol Chem 276:35290–35296.

Yntema CL. 1968. A series of stages in the embryonic deve-
lopment of Chelydra serpentina. J Morphol 125:219–252.

Yntema CL. 1970. Extirpation experiments on the embryonic
rudiments of the carapace of Chelydra serpentina. J Morphol
132:235–244.

Zangerl R. 1969. The turtle shell. In: Gans C, Bellairs A d’A,
editors. The biology of the Reptilia, Vol. 1. New York:
Academic Press. p 311–319.

TURTLE CARAPACE DEVELOPMENT 569