A Wnt- and beta -catenin-dependent pathway for mammalian cardiac myogenesis.
ABSTRACT Acquisition of a cardiac fate by embryonic mesodermal cells is a fundamental step in heart formation. Heart development in frogs and avians requires positive signals from adjacent endoderm, including bone morphogenic proteins, and is antagonized by a second secreted signal, Wnt proteins, from neural tube. By contrast, mechanisms of mesodermal commitment to create heart muscle in mammals are largely unknown. In addition, Wnt-dependent signals can involve either a canonical beta-catenin pathway or other, alternative mediators. Here, we tested the involvement of Wnts and beta-catenin in mammalian cardiac myogenesis by using a pluripotent mouse cell line (P19CL6) that recapitulates early steps for cardiac specification. In this system, early and late cardiac genes are up-regulated by 1% DMSO, and spontaneous beating occurs. Notably, Wnt3A and Wnt8A were induced days before even the earliest cardiogenic transcription factors. DMSO induced biochemical mediators of Wnt signaling (decreased phosphorylation and increased levels of beta-catenin), which were suppressed by Frizzled-8Fc, a soluble Wnt antagonist. DMSO provoked T cell factor-dependent transcriptional activity; thus, induction of Wnt proteins by DMSO was functionally coupled. Frizzled-8Fc inhibited the induction of cardiogenic transcription factors, cardiogenic growth factors, and sarcomeric myosin heavy chains. Likewise, differentiation was blocked by constitutively active glycogen synthase kinase 3beta, an intracellular inhibitor of the Wntbeta-catenin pathway. Conversely, lithium chloride, which inhibits glycogen synthase kinase 3beta, and Wnt3A-conditioned medium up-regulated early cardiac markers and the proportion of differentiated cells. Thus, Wntbeta-catenin signaling is activated at the inception of mammalian cardiac myogenesis and is indispensable for cardiac differentiation, at least in this pluripotent model system.
- SourceAvailable from: sibcb.ac.cn[show abstract] [hide abstract]
ABSTRACT: To understand how the Wnt coreceptor LRP-5 is involved in transducing the canonical Wnt signals, we identified Axin as a protein that interacts with the intracellular domain of LRP-5. LRP-5, when expressed in fibroblast cells, showed no effect on the canonical Wnt signaling pathway by itself, but acted synergistically with Wnt. In contrast, LRP-5 mutants lacking the extracellular domain functioned as constitutively active forms that bind Axin and that induce LEF-1 activation by destabilizing Axin and stabilizing beta-catenin. Addition of Wnt caused the translocation of Axin to the membrane and enhanced the interaction between Axin and LRP-5. In addition, the LRP-5 sequences involved in interactions with Axin are required for LEF-1 activation. Thus, we conclude that the binding of Axin to LRP-5 is an important part of the Wnt signal transduction pathway.Molecular Cell 05/2001; 7(4):801-9. · 15.28 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: In the chick, heart mesoderm is induced by signals from the anterior endoderm. Although BMP-2 is expressed in the anterior endoderm, BMP activity is necessary but not sufficient for heart formation. Previous work from our lab has suggested that one or more additional factors from anterior endoderm are required. Crescent is a Frizzled-related protein that inhibits Wnt-8c and is expressed in anterior endoderm during gastrulation. At the same stages, expression of Wnt-3a and Wnt-8c is restricted to the primitive streak and posterior lateral plate, and is absent from the anterior region where crescent is expressed. Posterior lateral plate mesoderm normally forms blood, but coculture of this tissue with anterior endoderm or infection with RCAS-crescent induces formation of beating heart muscle and represses formation of blood. Dkk-1, a Wnt inhibitor of a different protein family, similarly induces heart-specific gene expression in posterior lateral plate mesoderm. Furthermore, we have found that ectopic Wnt signals can repress heart formation from anterior mesoderm in vitro and in vivo and that forced expression of either Wnt-3a or Wnt-8c can promote development of primitive erythrocytes from the precardiac region. We conclude that inhibition of Wnt signaling promotes heart formation in the anterior lateral mesoderm, whereas active Wnt signaling in the posterior lateral mesoderm promotes blood development. We propose a model in which two orthogonal gradients, one of Wnt activity along the anterior-posterior axis and the other of BMP signals along the dorsal-ventral axis, intersect in the heart-forming region to induce cardiogenesis in a region of high BMP and low Wnt activity.Genes & Development 03/2001; 15(3):316-27. · 12.44 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: Heart induction in Xenopus occurs in paired regions of the dorsoanterior mesoderm in response to signals from the Spemann organizer and underlying dorsoanterior endoderm. These tissues together are sufficient to induce heart formation in noncardiogenic ventral marginal zone mesoderm. Similarly, in avians the underlying definitive endoderm induces cardiogenesis in precardiac mesoderm. Heart-inducing factors in amphibians are not known, and although certain BMPs and FGFs can mimic aspects of cardiogenesis in avians, neither can induce the full range of activities elicited by the inducing tissues. Here we report that the Wnt antagonists Dkk-1 and Crescent can induce heart formation in explants of ventral marginal zone mesoderm. Other Wnt antagonists, including the frizzled domain-containing proteins Frzb and Szl, lacked this activity. Unlike Wnt antagonism, inhibition of BMP signaling did not promote cardiogenesis. Ectopic expression of GSK3beta, which inhibits beta-catenin-mediated Wnt signaling, also induced cardiogenesis in ventral mesoderm. Analysis of Wnt proteins expressed during gastrulation revealed that Wnt3A and Wnt8, but not Wnt5A or Wnt11, inhibited endogenous heart induction. These results indicate that diffusion of Dkk-1 and Crescent from the organizer initiate cardiogenesis in adjacent mesoderm by establishing a zone of low Wnt3A and Wnt8 activity.Genes & Development 03/2001; 15(3):304-15. · 12.44 Impact Factor
A Wnt- and ?-catenin-dependent pathway for
mammalian cardiac myogenesis
Teruya Nakamura*†, Motoaki Sano*†, Zhou Songyang‡, and Michael D. Schneider*†§¶?
*Center for Cardiovascular Development and Departments of‡Biochemistry and Molecular Biology,†Medicine,§Molecular and Cellular Biology, and
¶Molecular Physiology and Biophysics, Baylor College of Medicine, Houston, TX 77030
Edited by Eric N. Olson, University of Texas Southwestern Medical Center, Dallas, TX, and approved February 28, 2003 (received for review
September 17, 2002)
Acquisition of a cardiac fate by embryonic mesodermal cells is a
avians requires positive signals from adjacent endoderm, including
signal, Wnt proteins, from neural tube. By contrast, mechanisms of
mesodermal commitment to create heart muscle in mammals are
largely unknown. In addition, Wnt-dependent signals can involve
either a canonical ?-catenin pathway or other, alternative mediators.
Here, we tested the involvement of Wnts and ?-catenin in mamma-
lian cardiac myogenesis by using a pluripotent mouse cell line
system, early and late cardiac genes are up-regulated by 1% DMSO,
and spontaneous beating occurs. Notably, Wnt3A and Wnt8A were
induced days before even the earliest cardiogenic transcription fac-
tors. DMSO induced biochemical mediators of Wnt signaling (de-
creased phosphorylation and increased levels of ?-catenin), which
were suppressed by Frizzled-8?Fc, a soluble Wnt antagonist. DMSO
provoked T cell factor-dependent transcriptional activity; thus, induc-
inhibited the induction of cardiogenic transcription factors, cardio-
genic growth factors, and sarcomeric myosin heavy chains. Likewise,
differentiation was blocked by constitutively active glycogen syn-
thase kinase 3?, an intracellular inhibitor of the Wnt??-catenin path-
way. Conversely, lithium chloride, which inhibits glycogen synthase
kinase 3?, and Wnt3A-conditioned medium up-regulated early car-
diac markers and the proportion of differentiated cells. Thus, Wnt?
this pluripotent model system.
anterolateral regions of the embryo during late gastrulation (1, 2).
In this process, morphogenic movements and cardiac fate deter-
mination are believed to be regulated by multiple extracellular cues
factor-?, and fibroblast growth factors (FGFs) secreted from un-
derlying endodermal cells (3). This general model is based on a
as early heart formation is most easily observed in embryos that
develop outside the mother, becoming accessible to explant exper-
can differentiate into beating muscle cells in the absence of
endoderm (9, 10), implying that several discrete steps initiate heart
development: a specification step before gastrulation, which leads
to the appearance of myocardial precursor cells, and a subsequent
step during gastrulation, in which endoderm serves to enhance the
rate of myocyte differentiation and degree of heart tube morpho-
genesis. Therefore, it is important to elucidate when, where, and
how mesodermal cells are instructed to assume the cardiac fate for
understanding the entire body of mechanisms that operate later in
initial instructive events are largely unproven.
he earliest event in heart formation is commitment of meso-
dermal cells to a cardiogenic ‘‘fate’’ and their migration into
Wnt?Wg genes, related to wingless in Drosophila, encode a
number of secreted proteins that play critical roles in the develop-
ment of many organisms, especially in cell fate and patterning
(11–13). Notably, the prototype wingless itself collaborates with the
tube in flies (14–17). In the absence of Wnt proteins, cells under-
take active measures to maintain low levels of the Wnt signaling
protein, ?-catenin. Under baseline circumstances, ?-catenin is
phosphorylated at its N terminus by glycogen synthase kinase 3?
(GSK-3?), targeting ?-catenin for destruction by the ubiquitin–
proteosome pathway (18). Wnt binding to the Frizzled (Fz) family
of serpentine receptors activates an associated downstream com-
stabilizing ?-catenin. Accumulation of ?-catenin in the cytosol
results in its translocation to the nucleus (no specific control is
known at this step), its interaction with T cell factor (TCF)?
Wnt-responsive genes (19). Thus, beyond the role of membrane-
associated ?-catenin in adherens junctions, soluble ?-catenin has a
pivotal function in the canonical Wnt signal transduction pathway.
In addition, Wnt proteins also activate an alternative cascade,
involving protein kinase C and Jun N-terminal kinase (18, 20, 21).
Recent provocative studies using Xenopus and chick embryos
indicate that Wnt proteins are potent negative regulators of
heart muscle specification in those species (6–8, 22). Apart from
the a priori concern that such findings are not necessarily
predictive of mammalian biology, isoform and pathway differ-
ences both likely exist. For instance, in Xenopus, Wnt3A and -8
were inhibitors acting via GSK-3 (6), whereas Wnt11 was an
inducer of heart formation through the ?-catenin-independent
A genetic analysis of the Wnt family in mammals, even
from straightforward, given extraordinary diversity and overlap-
ping expression of the ligands and receptors. Besides Fz proteins,
lipoprotein-receptor-related proteins also couple Wnts to
?-catenin (24). In addition, no known promoter has sufficient
early, potent, and specific expression to undertake a conditional
deletion of Wnt receptors or conditional expression of Wnt
inhibitors in the presumptive heart-forming region. For these
reasons, multipotential cells that can be made to adopt the
as an interim step, for understanding the earliest mechanisms of
cardiac determination. By using one such system, mouse P19CL6
cells (25–31), we demonstrate an essential role for the Wnt??-
catenin pathway in mammalian cardiac myogenesis.
Materials and Methods
Materials and Reagents. P19CL6 mouse embryonic carcinoma cells
were kindly provided by I. Komuro (Chiba University, Chiba,
This paper was submitted directly (Track II) to the PNAS office.
Abbreviations: BMP, bone morphogenetic protein; GSK-3?, glycogen synthase kinase 3?;
MHC, myosin heavy chain; QRT-PCR, quantitative RT-PCR; TCF, T cell factor; CM, condi-
tioned media; CMV, cytomegalovirus; HA, hemagglutinin; LiCl, lithium chloride; FGF,
fibroblast growth factor.
?To whom correspondence should be addressed. E-mail: email@example.com.
May 13, 2003 ?
vol. 100 ?
no. 10 www.pnas.org?cgi?doi?10.1073?pnas.0935626100
Japan). Wnt3A, control conditioned media (CM), and the corre-
sponding expression plasmids were provided by S. Takada (Kyoto
by M. Parsons (University of Toronto, Toronto; ref. 33). TOP-
FLASH (a firefly luciferase reporter plasmid, driven by two sets of
three copies of the TCF binding site and herpes simplex virus
thymidine kinase minimal promoter) and FOPFLASH (identical
except for inactivating mutations of the TCF sites) were purchased
from Upstate Biotechnology (Lake Placid, NY). pRL-cytomega-
lovirus (CMV; a constitutive, CMV-driven control, encoding Re-
nilla luciferase) was from Promega, DMSO and lithium chloride
from R & D Systems.
Cell Culture, Differentiation, and Transfection. Cells were grown on
10-cm dishes in ?-MEM (Invitrogen) supplemented with 10% FBS
(HyClone), penicillin, and streptomycin. To induce differentiation,
cells were seeded at a 1:40 dilution with ?-MEM?10% FBS?1%
DMSO. For each experiment, cardiomyocyte differentiation was
apparent in the control cultures as spontaneous beating, starting at
days 9–10. To obtain stable transformants incorporating GSK-
3?A9-hemagglutinin (HA) vs. the vector control, cells were trans-
fected by using Lipofectamine 2000 (Invitrogen) and maintained in
medium containing 500 ?g?ml geniticin (Invitrogen). After 10–14
days, 60 colonies were picked and screened by RT-PCR. To detect
exogenous GSK-3? selectively, the 5? primer corresponded to the
N terminus of GSK-3? and the 3? primer to the Haemophilus
influenzae HA epitope tag.
of primers and probes corresponding to each mRNA are available
using the TaqMan One-Step RT-PCR Master Mix reagent (Ap-
a 7700 Sequence Detector System (Applied Biosystems). The copy
number for each transcript is expressed relative to that of glyceral-
Immunocytochemistry. Cells were seeded on glass cover slips and
paraformaldehyde and permeabilized with 0.2% Triton X-100 for
5 min. To detect phosphorylated ?-catenin, cells were incubated
overnight with rabbit Ab to phospho-?-catenin (Ser-33?Ser-37?
Thr-41; Cell Signaling Technology, Beverly, MA) in Tris-buffered
saline?3% BSA and then for 1 h with goat Ab to rabbit IgG
conjugated with Alexa Fluor 488 (Molecular Probes). Immuno-
staining for sarcomeric myosin heavy chains (MHCs) was per-
formed by using FITC-conjugated MF20 Ab (35). Nuclei were
counterstained with 4?,6-diamidino-2-phenylindole (DAPI).
Western Blot Analysis. Cells were seeded on six-well dishes (1.67 ?
ng?ml Fz?Fc chimeric protein. After 3 days, cells were harvested in
PBS at 4°C, centrifuged at 2,000 ? g for 5 min, and resuspended in
20 mM Tris?HCl (pH 7.5), 25 mM sodium fluoride, and 1 mM
EDTA, containing a protease inhibitor mixture (Roche Molecular
Biochemicals). Cells were incubated on ice for 20 min, followed by
30 strokes in a Dounce homogenizer, and centrifuged at 100,000 ?
g for 30 min. The supernatant was collected as the soluble,
cytoplasmic fraction and subjected to electrophoresis in 10%
SDS-polyacrylamide gels. Proteins were transferred to poly(vinyli-
dene difluoride) membranes, which were incubated sequentially in
Cruz Biotechnology) overnight at 4°C. Bound Ab was visualized by
using horseradish peroxidase (HRP)-conjugated goat Ab to mouse
IgG (Zymed) or donkey Ab to goat IgG (Santa Cruz Biotechnol-
ogy) and enhanced chemiluminescence reagents (Amersham Bio-
X-100?1% sodium deoxycholate?0.1% SDS) with 10 nM calyculin
A, 10 nM okadaic acid, and a protease inhibitor mixture. Western
blotting was done as described previously by using rabbit Ab to
phospho-?-catenin and HRP-conjugated goat Ab to rabbit IgG
(Santa Cruz Biotechnology).
Luciferase Assays. Cellsseededandculturedasdescribedpreviously
were transfected 1 day after plating by using Lipofectamine 2000 in
serum-free ?-MEM for 6 h. Transfections contained 0.5 ?g of
TOPFLASH or FOPFLASH plus 0.1 ?g of pRL-CMV as the
cotransfected control. Medium containing 10% FBS with or with-
out DMSO was changed 6 and 48 h after transfection. Cells were
lysed 4 days after DMSO treatment, and luciferase was assayed by
using the Dual-Luciferase system (Promega). Firefly luciferase
activity, indicating TCF-dependent transcription, was normalized
to the Renilla luciferase activity of each extract. TOPFLASH
activity induced by Wnt3A CM vs. control CM was measured after
dilution into 24-well dishes, and TOPFLASH, pRL-CMV, and
PGK-Wnt3A were cotransfected by calcium-phosphate precipita-
tion along with pcDNA3-GSK-3?A9-HA vs. the empty vector
changed to DMEM supplemented with 0.2% BSA. Forty-eight
hours after transfection, cells were lysed and luciferase activities
were assayed as described previously.
Statistical Analysis. Results, shown as the mean ? SE, were com-
pared by ANOVA followed by Scheffe ´’s test, with P ? 0.05
Wnt3A and -8 Are Early Responses in Differentiating P19CL6 Cells. We
first used quantitative RT-PCR (QRT-PCR) analyses to map the
temporal changes of cardiac-specific genes in differentiating
P19CL6 cells, in comparison to their potential regulators (Fig. 1A).
genes assayed. After 1% DMSO treatment, the cells expressed the
cardiac transcription factors Nkx2.5, GATA4, MEF2C, and Tbx5
as early stage markers within 3–8 days. Spontaneous beating
was visible at days 9–10, and continued for 4–8 days. Among these
markers, GATA4 and Nkx2.5 were expressed earliest (as early as
3 days after DMSO), and GATA4 protein was detected at 8 days
(Fig. 1B). ?-MHC was seen as early as day 8 and increased until day
with the presence of ?-MHC mRNA (Fig. 1A) and protein (Fig.
1B), MF20-positive differentiated myocytes were observed on day
12 (Fig. 1C).
Neither BMP2 nor BMP4 was expressed in the absence of
DMSO. Both were induced to low levels within 4–5 days, with
Studies have implicated endogenous BMPs as essential for cardio-
genesis in these cells (26, 27). By contrast to the delayed and
sustained expression of BMP2 and BMP4, Wnt3A and Wnt8A were
each induced as early as 2 days after DMSO treatment, with
-8A mRNA levels were quickly down-regulated on day 4, with low
or undetectable expression at all later times. Like Wnt3A and -8A,
thereafter (Fig. 1D). Wnt11, which is required for Xenopus cardio-
genesis (23), was not expressed (data not shown).
Nakamura et al.
May 13, 2003 ?
vol. 100 ?
no. 10 ?
The Wnt??-Catenin Pathway Is Activated at the Early Stage of Dif-
at the time of Wnt induction, we examined cytosolic ?-catenin, the
crux of the canonical Wnt signal transmission pathway (13, 37).
Cells treated with or without DMSO were harvested and lysed on
day 3, cytosolic protein was fractionated, and ?-catenin protein
levels were examined by Western blot (Fig. 2A). DMSO caused the
accumulation of soluble ?-catenin; accumulation of ?-catenin was
also observed in the total cell lysates (Fig. 2B). Simultaneous
treatment with 500 ng?ml Fz-8?Fc chimeric protein, an antagonist
for Wnt8A and potentially for other Wnts, decreased ?-catenin to
the basal level (Fig. 2 A and B), indicating that its accumulation is
regulated by an autocrine or paracrine circuit in this system,
involving endogenous Wnts. Conversely, as expected, phosphory-
lated ?-catenin (the form targeted for degradation) was decreased
tional activity was measured (Fig. 2D). DMSO treatment for 4 days
provoked a 9-fold increase in luciferase activity. No activation was
binding sites. Thus, Wnt induction by DMSO was functionally
coupled to activation of the ?-catenin pathway.
The Wnt??-Catenin-Signaling Pathway Was Required for Cardiac Dif-
in early cardiac myogenesis, we first monitored differentiation
induced by DMSO, with and without 200 ng?ml Fz-8?Fc. Treat-
(Bar ? 50 ?m.) (D) FGF8 expression by RT-PCR.
myogenesis. (A) Increased soluble ?-catenin. Cells were treated with (?) or
blot analysis. ?-Catenin was specifically decreased by Fz-8?Fc, an extracellular
antagonist of Wnt signaling. (B) Decreased phosphorylated ?-catenin and in-
creased total ?-catenin in whole-cell lysates. Cells were cultured as in A and
Western blotting was done by using Abs to phospho-?-catenin (Top) and total
?-catenin (Middle). (C) Decreased phosphorylated ?-catenin (green), shown by
immunostaining. (Left) Without DMSO. (Right) With DMSO. Nuclei were coun-
terstained with 4?,6-diamidino-2-phenylindole (blue). (Bar ? 50 ?m.) (D) TCF?
and transfected with TOPFLASH or FOPFLASH (inactive, mutant TCF sites) along
with pRL-CMV. Luciferase activity was determined after 4 days of treatment.
www.pnas.org?cgi?doi?10.1073?pnas.0935626100Nakamura et al.
ment with the soluble Wnt inhibitor prevented GATA4 and Tbx5
induction by DMSO, at least through day 6 (Fig. 3A). Likewise,
the Wnt pathway lies upstream to the induction of these three
cardiac differentiation factors (Fig. 3A). Similar results were ob-
tained by using Fz-4?Fc. Correspondingly, Fz-8?Fc decreased the
and suppression continued for at least 15 days (Fig. 3A).
signaling was responsible for Wnt-dependent cardiac myogenesis,
we used a constitutively active form of GSK-3? (pcDNA3-GSK-
(Fig. 3C), we confirmed that this vector inhibited Wnt3A-induced
transcription of the TCF reporter gene. We then obtained stable
vector, neor(Fig. 3D). GSK-3?A9 suppressed the induction of
cardiac transcription factors by DMSO through at least 15 days,
with identical results in independent lines (Fig. 3E and data not
shown). By contrast, cardiac differentiation was impaired in none
of the clonal isolates bearing the selectable marker neoralone,
assayed at equivalent passage number. Likewise, GSK-3? sup-
pressed the prevalence of sarcomeric MHC staining, whereas
(Bar ? 25 ?m.) (C) GSK3?A9 inhibits Wnt3A-induced TCF transcriptional activity. 293T cells were cotransfected with PGK-neo, PGK-Wnt3A, and pcDNA3-GSK-
transfected with the pcDNA3 expression vectors shown. Plasmid, 5 ng of pcDNA3-GSK-3?A9-HA as template. (E) GSK-3?A9 suppresses DMSO-induced cardiac
Nakamura et al.
May 13, 2003 ?
vol. 100 ?
no. 10 ?
findings with soluble Fz proteins and activated GSK-3? implicate
signaling by endogenous Wnts, via ?-catenin, as required for early
cardiac determination in this setting.
The Wnt??-Catenin-Signaling Pathway Enhances Cardiac Myogenesis.
We next explored the reciprocal possibility of promoting cardiac
differentiation in this system by supplying exogenous Wnt or
potentiating ?-catenin. We first confirmed that Wnt3A-
conditioned medium could activate TCF-dependent transcription
in P19CL6 cells (Fig. 4A Left). Although control CM had no effect
(from L cells stably transfected with just the neomycin-resistance
gene), Wnt3A-conditioned medium markedly enhanced the induc-
tive effect of DMSO on Nkx2.5, GATA4, MEF2C, and Tbx5 (Fig.
4A Right). ?-MHC, otherwise expressed no sooner than day 8, was
detected on day 6 in cultures receiving Wnt3A CM. Interestingly,
In accordance with the enhanced expression of cardiogenic factors,
day 12 by Wnt3A CM (Fig. 4B).
We next used LiCl, which binds and inhibits GSK-3? and, hence,
activates Wnt signaling selectively via the ?-catenin?TCF pathway
(13, 38). At 10 ?M, LiCl significantly increased the expression of
Nkx2.5, GATA4, Tbx5, BMP2, and BMP4 at day 5 and MF20-
positive cells at day 12 (Fig. 4C). LiCl itself had an inductive effect
in the absence of DMSO. However, cells treated with LiCl or
Wnt3A CM alone did not show spontaneous beating (data not
necessary for terminal differentiation, beyond just induction of the
factors shown here.
In summary, activation of the Wnt??-catenin-signaling cascade
was an early event in the cardiogenic differentiation of pluripo-
tent P19CL6 cells, as measured by Wnt3A and Wnt8A induction,
was analyzed by QRT-PCR (day 5). NaCl was added as the control.
The Wnt??-catenin-signaling pathway enhances cardiac myogenesis. (A) Wnt3A CM increases TCF-dependent transcription. P19CL6 cells cotransfected with
www.pnas.org?cgi?doi?10.1073?pnas.0935626100Nakamura et al.