MINIREVIEW
Bone morphogenetic protein signaling in stem cells ) one
signal, many consequences
Toni U. Wagner
Physiological Chemistry I, University of Wuerzburg, Germany
BMP signals in stem cells
Bone morphogenetic protein (BMP) signals have tre-
mendous effects on all kinds of cells. Most striking
and defining, however, are the reactions that stem and
progenitor cells show upon exposure to BMP ligands.
Various stem cell types utilize BMP signals in a multi-
tude of ways in order to define their fates. The integra-
tion of this pathway with a variety of other signals is
still poorly understood, but recent discoveries strongly
suggest that stem cell niches, areas with a certain sig-
nal–molecule cocktail, are responsible for the final out-
come of BMP signaling, be it by modulating the ligand
itself, or the cascade transducing the signal within
the cell. Regulation of BMP signaling is seen at all
molecular levels: ligands, receptors, transducers, tran-
scription complex composition and chromatin state.
The present review focuses on data gathered on the
role of BMP signaling in selected stem cell systems.
Due to space limitations, numerous stem cell niches
described to be influenced by BMP are not reviewed.
We try to focus on publications that represent the wide
variety of effects induced by BMP signals.
The BMP signaling cascades
The basic BMP signaling process is started by homo-
or heterodimeric BMP ligands. Upon binding to type I
receptors, formation of a heteromeric complex with

neural stem ⁄ progenitor cell; PGC, primordial germ cell; STAT3, signal transducer and activator of transcription-3; TGF-b, transforming growth
factor b.
2968 FEBS Journal 274 (2007) 2968–2976 ª 2007 The Author Journal compilation ª 2007 FEBS
version, the type II receptors then phosphorylate the
type I receptors, which subsequently activate R-Smads
(Smads 1, 5 and 8 for BMP ligands) by serine-threonine
phosphorylation [1,2]. R-Smads are transcription fac-
tors, which need activation (usually by phosphoryla-
tion) and subsequent multimerization in order to
become active and accumulate in the nucleus. Once
activated, the R-Smad is able to bind to the Co-Smad
(Smad4) and translocate to, or accumulate in the nuc-
leus [3]. There, together with a wide variety of cofac-
tors, target gene transcription is usually activated when
R-Smads are in play.
Of course, this very linear pathway is far from real-
ity. BMP receptors have been shown to convey signals
not only by Smad phosphorylation, but also through
p38 activation [3]. Furthermore, transduction can as
well occur through so-called repressor-Smads and
Smads specific for transforming growth factor b
(TGF-b) signals such as Nodal (Table 1) [2].
BMP signals are strongly influenced by many addi-
tional parameters, such as the mode of oligomerization
of receptors even prior to ligand-binding and resulting
differences in downstream targets have also been clar-
ified in much more detail [1]. Regulation is further
fine-tuned by BMP receptor regulation through degra-
dation and dephosphorylation [4], different modes of
endocytosis [5], interaction with other pathways and

be noted that recent studies in nonhuman primate
ES-cell cultures [19] as well as in human ES-cell culture
systems [20,21], have demonstrated complete independ-
ence of LIF and STAT3.
Back in the mouse system, feeder cells and serum
can be omitted if BMP2 ⁄ 4 and LIF are present in the
medium [22], resulting in a very defined two-factor sys-
tem to study pluripotency. The same work demonstra-
ted that the downstream target genes primarily
responsible for the pluripotency maintenance effect of
BMPs under these conditions are the inhibitors of
DNA-binding (ID) genes. ID gene transcription was
previously shown to be enhanced by a Smad1–Smad4
complex directly binding GC-rich elements in combina-
tion with Smad-binding elements (SBE, sequence:
GTCT) present in the ID1 promoter region [23]. ID
gene expression is further enhanced by another well
known pluripotency associated factor called Nanog
[24] in a not yet understood way.
Adding to the picture are data obtained by micro-
array-based analysis [25] of murine stem cells, in which
Table 1. BMP and TGF-b signal transducer molecules of the Smad family and their respective function [1,2].
Name Type Ligands Receptors (type I) Function
Smad1 R-Smad AMH, BMP2 ⁄ 4 ⁄ 7 ALK1 ⁄ 2 ⁄ 3 ⁄ 6 Transcriptional regulation
Smad2 R-Smad Activin, Nodal, TGF-b ALK4 ⁄ 5 ⁄ 7 Transcriptional regulation
Smad3 R-Smad Activin, Nodal, TGF-b ALK4 ⁄ 5 ⁄ 7 Transcriptional regulation
Smad4 Co-Smad all – Co-Smad needed for All R-Smads
Smad5 R-Smad AMH, BMP2 ⁄ 4 ⁄ 7 ALK1 ⁄ 2 ⁄ 3 ⁄ 6 Transcriptional regulation
Smad6 I-Smad – All Decoy Smad, inhibition of Smad interactions
Smad7 I-Smad – All Decoy Smad, inhibition of Smad interactions

port of pluripotency by BMP signals is not only highly
dose-dependent, but also needs to be counter-regulated
(e.g. by STAT3 and Nanog).
Recapitulating, the BMP-signaling pathway promotes
pluripotency only indirectly by driving expression of ID
genes in a Smad1-dependent manner. ID-proteins block
neural differentiation of ES-cells by sequestering tran-
scription factors needed to initiate commitment to this
lineage. Concurrently, the differentiation induction
effects of BMP are counteracted by STAT3 and Nanog,
which are able to suppress activation of Smad1-target
genes necessary for differentiation into mesodermal and
endodermal cell fates.
In other words, the essence of defined medium
murine ES-cell culture appears to be the simultaneous
action of STAT3 and Smad in a certain ratio. Down-
stream, negative regulation of differentiation pro-
grammes for mesodermal and endodermal fates
(mediated by STAT3) as well as neuro-ectodermal line-
ages (controlled by IDs) is initiated, resulting in the
blockage of any kind of differentiation (Fig. 1).
Taking these data from the culture system, it is inter-
esting to look at studies on BMP signaling proteins
in early embryonic development. Although Smad4
– ⁄ –
mouse embryos do not successfully undergo gastrula-
tion and die before embryonic day 7.5, it was possible
to derive ES-cell lines from the inner cell mass of these
mutants [27]. As the only Co-Smad, Smad4 is abso-
lutely necessary for any Smad-linked BMP signal con-

free mouse embryonic stem cells. Parallel activity of STAT3 and
Smad leads to inhibition of differentiation programs induced by the
other pathway, thereby upholding the pluripotent state of the cells.
The balance is easily broken upon signal increase in any direction.
BMP signaling in stem cells T. U. Wagner
2970 FEBS Journal 274 (2007) 2968–2976 ª 2007 The Author Journal compilation ª 2007 FEBS
tolerate lack of ALK3 in absence of the inhibitor.
Although functional proof is missing, the authors
found up-regulation of ALK1 and ALK2 in these cells
and suggest these receptors to be able to compensate
ALK3 loss. Bringing the different aspects together,
alternative BMP signaling pathways all seem to be able
to support pluripotency, but a complete loss of BMP
signal transduction is not compatible with stemness. To
truly clarify this situation, additional studies using con-
ditional depletion of all combinations of the suggested
transduction ways are needed.
Yet another BMP signal influencing factor associ-
ated with pluripotency has been identified, namely
growth and differentiation factor-3 (GDF3) [30].
GDF3 is exclusively expressed in the undifferentiated
state in both mouse and human ES-cell culture. GDF3
is a secreted factor, which is able to bind to and
thereby inactivate BMP4. Reduction of GDF3 expres-
sion in murine ES-cells lead to increased independence
of LIF but, at the same time, to a lack of mesodermal
and endodermal differentiation-ability in vitro. In vivo,
GDF3 is expressed during early embryogenesis in mice,
notably in the inner cell mass. Protein localization
shows extracellular distribution throughout the blasto-

bHLH transcription factors such as Neurogenin1 and
Mash1, both responsible for neurogenesis [32]. Addi-
tionally, BMP exposure results in down-regulation of
Olig2 expression [33]. Olig2 in turn inhibits formation
of the GFAP superactivator complex STAT3–p300–
Smad1 [34], thus clearing the path for neuronal differ-
entiation for cells exposed to low amounts of BMPs.
The p300–Smad1 complex is target for yet another reg-
ulatory input. Neurogenin has been shown to compete
with STAT3 for its recruitment. Although the GFAP
promoter is hyper activated when bound by Smad1–
p300–STAT3, the neuroD promoter is strongly driven
by binding of Smad1–p300–Neurogenin [35], giving
BMP signals a role in neurogenesis as well. Other stud-
ies [36] have shown that LIF or BMP4 alone are also
able to drive GFAP expression in neurosphere cul-
tures. Phenotypically, the resulting GFAP
+
cells gener-
ated by either LIF or BMP4 differ strongly: whereas
LIF induces GFAP expressing NSCs to become elon-
gated and stay proliferative, BMP4 application results
in cell-cycle exit and a star-like cell-morphology. Fur-
thermore, LIF treatment leads to an upkeep of progen-
itor features, such as prolonged culture ability and the
potential to undergo neural differentiation, whereas
BMP4 decreases both. These results stress that BMP
signaling is indeed an antiproliferative and differenti-
ation inductive signal for neural stem cells, again (as
shown and discussed for murine ES-cells) modulated

explanted chicken otic vesicles leads to fewer hair
cells [38]. Inhibition of BMP signals by Noggin
results in increasing numbers of hair-cells. Cell-cycle
and apoptosis analyses of these experiments reveal
that BMP4 not only suppresses expression of prosen-
sory markers (itself including), but also drives the
proliferative sensory precursor cells into apoptosis. In
the same context, Noggin application is able to
expand the sensory patches without increasing pro-
genitor proliferation. This suggests that BMP4 has a
double function in restriction of hair-cell number:
block of differentiation and stop of progenitor prolif-
eration, with both effects leading to apoptosis. The
molecular mechanism at work in these progenitor
cells has not yet been addressed, but might include
ID protein mediated block of bHLH factors such as
NeuroD, which are responsible for correct progres-
sion of final differentiation steps.
BMP signals lead to proliferation, apoptosis
and cell-cycle arrest within the eye
Studies in the chick embryo have revealed a role for
BMP4 in eye development. Implantation experiments
of beads soaked in BMP4 have shown that BMP is
able to induce programmed cell death (apoptosis).
Blocking the BMP pathway by using Noggin-leaded
beads does not lead to over proliferation but, on the
contrary, restricts growth and, when applied for longer
periods, will result in reduced size of the optic cup
[39]. Surprisingly, apoptosis is inhibited at the same
time. In line with this, BMP4, even though responsible

of mesodermal progenitor cells (MPCs) is set at gastru-
lation. A BMP signal gradient originating from the
posterior end of the embryo establishes a boundary
between trunk and tail domains. Determination of the
two trunk domains is probably due to higher levels of
Nodal signaling and weakening of BMP signaling by
antagonists such as Chordin or Follistatin, which are
most likely genetically downstream of Nodal. The
MPCs are then able to integrate and interpret the
signal strengths of Nodal and BMP by entering
the somite regions at different somites. Strong Nodal
drives them in early, at somite 1, whereas strong BMP
delays their entry until somite 16. MPCs receiving both
weak Nodal and weak BMP input enter in-between, at
somite 9.
So far, it has not been possible to truly visualize gra-
ded signal activities in living embryos. Furthermore,
the cellular machinery for signal integration also
remains elusive. Whether this process is strictly nuc-
lear, transcriptional control of sets of target genes or
happens in the cytoplasm where translocation and acti-
vation of transducer molecules is modulated in is also
unclear. There is a significant gap between the avail-
able in vivo knowledge coming from analyses of pheno-
types as a result of BMP signal strength and in vitro
knowledge about intracellular processes downstream of
signal triggering.
There is another example of stemness decisions by
Nodal versus BMP signaling emerging from human
embryonic stem cell research. Testing human ES-cells

conditions as described before.
BMP as an initiator of the germ line
Cells of the germ line are unique in many aspects.
Their DNA and differentiation state have to be con-
trolled over generations. They are extremely mobile
and on their long way from being defined to arriving
in the gonad they are exposed to, and ignore,
practically all extracellular cues used for building
and patterning the embryo. Generally, with the
appearance of primordial germ cells (PGCs), the
germ line is usually the earliest cell-lineage that
is determined in the embryo. In many lower ani-
mals, such as flies and fish, they are defined by
maternally deposited factors. Strikingly, they have
even been shown to be pluripotent after in vitro
expansion [43].
In the mouse embryo, PGCs are formed during
embryonic day 6 at the posterior proximal epiblast
through a location dependent mechanism. Using gene
knockouts, the molecules responsible were identified as
BMPs. The primary induction of PGCs is driven by
BMP4 [44], whereas the number of PGCs is guided by
BMP2, BMP4 and BMP8b in synergistic action [45].
As demonstrated by in vitro culture assays [30], inde-
pendent BMP signals originating from the visceral
endoderm and the extra embryonic ectoderm are
necessary for proper PGC induction. The receiving
receptors were found by knockout experiments, where
no PGCs were present in ALK2
– ⁄ –

T. U. Wagner BMP signaling in stem cells
FEBS Journal 274 (2007) 2968–2976 ª 2007 The Author Journal compilation ª 2007 FEBS 2973
outcomes. This has been seen when Smad1 and STAT3
interact to guide astrocyte fates. For embryonic
stem cell cultures, it is more likely an effect of negative
regulation of target genes: here, STAT3 and its targets
seem to block transcription of BMP signal targets and
vice versa. Certainly, the signal integration process
has to be extended beyond the transducer molecules.
One such level of signal modulation occurs outside
of the signal receiving cell, by secretion of blocking
ligands such as Noggin, Chordin and Follistatin or
GDF3. Another way to modulate ligand ⁄ receptor inter-
action is by receptor localization within the membrane
[46]. Additionally, the makeup of BMP-receptor com-
plexes is crucial for the choice of the transduction route
and resulting cellular response. BMP2, for example,
can bind a type I receptor (e.g. BRIa) with high affinity
[47] and induce subsequent BRII recruitment to the
complex. This mode of complex formation leads to
signal conveyance via the p38 MAPK pathway. Alter-
natively, a preformed receptor complex including
type I and type II receptors is able to bind BMP2, in
which case transduction will occur through Smad
activation [48].
There is further evidence for higher-order cross-talk
between BMP signaling proteins and those of other
pathways close to the membrane (e.g. with PI3-Kin-
ase [49] or Dullard [4]), in cytoplasmic complexes
(e.g. Smad1 and Enofin [50]) and nuclear complexes

Dimeric secreted proteins such as noggin,
chordin, follistatin and GDF3 bind and therby
inactivate BMP dimers [16]
Cell membrane Predimerization of type I receptors,
type I–type II receptors
Receptor heterooligomers induced by ligand
binding signal via Smads, preformed complexes
signal via p38 [31]
Receptor complex inhibiton BAMBI pseudoreceptors block receptor activation
[1]
Receptor–adapter junction R-Smads are kept from interacting
with receptor complexes
Smad7 binds to type I receptors and thus blocks
R-Smads from being activated by the receptors
[1]
Adapters inhibit receptor activation FKBP12 inhibits type I receptor phosphorylation
[1]
Cytoplasm Smad expression and covalent modification Sumoylation by PIAS, ubiquitylation by Smurf1 ⁄ 2
[1]
Competition for Smad4 binding Smad6 sequesters Smad4, thereby blocks Smad1
binding and nuclear accumulation [1]
Nucleus Co-factor-availability, corepressors and coactivators
can form a complex
Ngn competes with Stat3 for Smad1–p300
complexes in glial differentiation [21]
Nuclear import, export and retention MAPK phosphorylation of Smads leads to export
from the nucleus [2]
Dephosphorylation of Smads to end a signal Several phosphatases targeting the SXS
C-terminal regions of Smad1 ⁄ 2 ⁄ 3 [12–15]
BMP signaling in stem cells T. U. Wagner

Moustakas A (2003) Nuclear factor YY1 inhibits trans-
forming growth factor beta- and bone morphogenetic
protein-induced cell differentiation. Mol Cell Biol 23,
4494–4510.
9 Hu MC & Rosenblum ND (2005) Smad1, beta-catenin
and Tcf4 associate in a molecular complex with the
Myc promoter in dysplastic renal tissue and cooperate
to control myc transcription. Development 132, 215–225.
10 Dahlqvist C, Blokzijl A, Chapman G, Falk A, Dannaeus
K, Ibaˆ n
˜
ez CF & Lendahl U (2003) Functional notch
signaling is required for BMP4-induced inhibition of
myogenic differentiation. Development 130, 6089–6099.
11 Blokzijl A, Dahlqvist C, Reissmann E, Falk A, Moliner
A, Lendahl U & Iba
´
n
˜
ez CF (2003) Cross-talk between
the notch and TGF-beta signaling pathways mediated
by interaction of the notch intracellular domain with
Smad3. J Cell Biol 163, 723–728.
12 Sapkota G, Knockaert M, Alarco
´
n C, Montalvo E,
Brivanlou AH & Massague
´
J (2006) Dephosphorylation
of the linker regions of Smad1 and Smad2 ⁄ 3 by small

tional activation in embryonic stem cells. J Cell Biol
138, 1207–1217.
19 Sumi T, Fujimoto Y, Nakatsuji N & Suemori H (2004)
STAT3 is dispensable for maintenance of self-renewal in
nonhuman primate embryonic stem cells. Stem Cells 22,
861–872.
20 Humphrey RK, Beattie GM, Lopez AD, Bucay N, King
CC, Firpo MT, Rose-John S & Hayek A (2004) Mainte-
nance of pluripotency in human embryonic stem cells is
STAT3 independent. Stem Cells 22, 522–530.
21 Daheron L, Opitz SL, Zaehres H, Lensch WM,
Andrews PW, Itskovitz-Eldor J & Daley GQ (2004)
LIF ⁄ STAT3 signaling fails to maintain self-renewal of
human embryonic stem cells. Stem Cells 22, 770–778.
22 Ying QL, Nichols J, Chambers I & Smith A (2003)
BMP induction of Id proteins suppresses differentiation
and sustains embryonic stem cell self-renewal in colla-
boration with STAT3. Cell 115, 281–292.
23 Lopez-Rovira T, Chalaux E, Massague J, Rosa JL &
Ventura F (2002) Direct binding of Smad1 and Smad4
to two distinct motifs mediates bone morphogenetic
protein-specific transcriptional activation of Id1 gene.
J Biol Chem 277, 3176–3185.
24 Chambers I, Colby D, Robertson M, Nichols J, Lee S,
Tweedie S & Smith A (2003) Functional expression
cloning of Nanog, a pluripotency sustaining factor in
embryonic stem cells. Cell 113, 643–655.
25 Sekkai D, Gruel G, Herry M, Moucadel V, Constantin-
escu SN, Albagli O, Tronik-Le Roux D, Vainchenker
W & Bennaceur-Griscelli A (2005) Microarray analysis

31 Nakashima K, Yanagisawa M, Arakawa H, Kimura N,
Hisatsune T, Kawabata M, Miyazono K & Taga T
(1999) Synergistic signaling in fetal brain by STAT3-
smad1 complex bridged by p300. Science 284, 479–482.
32 Nakashima K, Takizawa T, Ochiai W, Yanagisawa M,
Hisatsune T, Nakafuku M, Miyazono K, Kishimoto T,
Kageyama R & Taga T (2001) BMP2-mediated altera-
tion in the developmental pathway of fetal mouse brain
cells from neurogenesis to astrocytogenesis. Proc Natl
Acad Sci USA 98, 5868–5873.
33 Mekki-Dauriac S, Agius E, Kan P & Cochard P (2002)
Bone morphogenetic proteins negatively control oligo-
dendrocyte precursor specification in the chick spinal
cord. Development 129, 5117–5130.
34 Fukuda S, Kondo T, Takebayashi H & Taga T (2004)
Negative regulatory effect of an oligodendrocytic bHLH
factor OLIG2 on the astrocytic differentiation pathway.
Cell Death Differ 11, 196–202.
35 Sun Y, Nadal-Vicens M, Misono S, Lin MZ, Zubiaga
A, Hua X, Fan G & Greenberg ME (2001) Neurogenin
promotes neurogenesis and inhibits glial differentiation
by independent mechanisms. Cell 104, 365–376.
36 Bonaguidi MA, McGuire T, Hu M, Kan L, Samanta J
& Kessler JA (2005) Lif and bmp signaling generate
separate and discrete types of gfap-expressing cells.
Development 132, 5503–5514.
37 Hua H, Zhang Y, Dabernat S, Kritzik M, Dietz D,
Sterling L & Sarvetnick N (2006) BMP4 regulates
pancreatic progenitor cell expansion through Id2. J Biol
Chem 281, 13574–13580.

action of BMP4 and BMP8B signaling pathways. Proc
Natl Acad Sci USA 98, 7858–7862.
46 Nohe A, Keating E, Underhill TM, Knaus P & Petersen
NO (2003) Effect of the distribution and clustering of
the type I A BMP receptor (ALK3) with the type II
BMP receptor on the activation of signalling pathways.
J Cell Sci 116, 3277–3284.
47 Gilboa L, Nohe A, Geissendo
¨
rfer T, Sebald W, Henis
YI & Knaus P (2000) Bone morphogenetic protein
receptor complexes on the surface of live cells: a new
oligomerization mode for serine ⁄ threonine kinase recep-
tors. Mol Biol Cell 11, 1023–1035.
48 Nohe A, Hassel S, Ehrlich M, Neubauer F, Sebald W,
Henis YI & Knaus P (2002) The mode of bone morpho-
genetic protein (BMP) receptor oligomerization deter-
mines different BMP-2 signaling pathways. J Biol Chem
277, 5330–5338.
49 Ghosh-Choudhury N, Abboud SL, Mahimainathan L,
Chandrasekar B & Choudhury GG (2003) Phosphati-
dylinositol 3-kinase regulates bone morphogenetic
protein-2 (bmp-2)-induced myocyte enhancer factor
2a-dependent transcription of bmp-2 gene in
cardiomyocyte precursor cells. J Biol Chem 278,
21998–22005.
50 Shi W, Chang C, Nie S, Xie S, Wan M & Cao X (2007)
Endofin acts as a Smad anchor for receptor activation
in bmp signaling. J Cell Sci 120, 1216–1224.
BMP signaling in stem cells T. U. Wagner