| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ludwig Institute for Cancer Research, Box 595, S-75124 Uppsala, Sweden
1Correspondence and current address: Division of Cellular Biochemistry, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands. E-mail: ptdijke{at}nki.nl
 |
ABSTRACT |
|---|
 TOP ABSTRACT THE TGF-{beta} SUPERFAMILY AND...SIGNALING VIA TGF-{beta}...INTRACELLULAR SIGNALING BY TGF...CONCLUSIONS AND PERSPECTIVESREFERENCES |
|---|
Key Words: activin • bone morphogenetic protein • signal transduction • Smad • transforming growth factor-ß
|
THE TGF-ß SUPERFAMILY AND ITS BINDING PROTEINS |
|---|
|
TOPABSTRACTTHE TGF-{beta} SUPERFAMILY AND...SIGNALING VIA TGF-{beta}...INTRACELLULAR SIGNALING BY TGF...CONCLUSIONS AND PERSPECTIVESREFERENCES |
|---|
, 2
). BMPs are potent
inducers of bone and cartilage formation and play important
developmental roles in the induction of ventral mesoderm,
differentiation of neural tissue, and organogenesis (reviewed in refs 3
, 4
). Activins, named after their initial identification as activators
of follicle-stimulating hormone (FSH) secretion from pituitary glands,
are also known to promote erythropoiesis, mediate dorsal mesoderm
induction, and contribute to survival of nerve cells (reviewed in ref 5
). Several growth factors belonging to the TGF-ß superfamily play
important roles in embryonic patterning and tissue homeostasis. Their
inappropriate functioning has been implicated in several pathological
situations like fibrosis, rheumatoid arthritis, and carcinogenesis.
Distinct in vivo expression patterns of TGF-ß
isoforms
TGF-ß1, TGF-ß2, and TGF-ß3 are highly similar in their
biological activities in vitro. However, they differ in
their in vivo expression patterns (reviewed in ref 6
), which
largely explains the unique isoform-specific phenotypes displayed by
the TGF-ß knockout mice. TGF-ß1-deficient mice that are born alive
undergo early postnatal death due to excessive infiltration of
inflammatory lymphocytes and macrophages into several organs
(7
8
9)
. Half of the mice lacking TGF-ß1 die in
utero due to defects in vasculogenesis and hematopoiesis
(10)
. It is thought that maternally supplied TGF-ß1, in
combination with redundant expression of TGF-ß isoforms, contributes
to normal development of the embryos. TGF-ß1 null mice born from
TGF-ß1-deficient mothers display abnormal cardiac development
(11)
. TGF-ß2 knockout mice show multiple developmental
malformations of tissues and organs, leading to perinatal death
(12)
. TGF-ß3 null mice die shortly after birth,
displaying delayed pulmonary development and defective palate
development (13
, 14)
.
Analysis of the genes encoding the three TGF-ßs revealed that each
isoform is controlled by differentially regulated promoters (reviewed
in ref 15
). The existence of different genes encoding functionally
similar proteins, yet controlled by differentially regulated promoters,
might provide an important mechanism to ensure tissue-specific and
spatio-temporal expression patterns of the different TGF-ß isoforms,
thereby resulting in proper embryonic development and tissue
homeostasis.
In addition, posttranscriptional control mechanisms contribute to
regulation of the production of TGF-ß. TGF-ß1, TGF-ß2, and
TGF-ß3 that have long, GC-rich 5'-untranslated regions and
intramolecular duplex loops located in close proximity to the
transcriptional start site negatively modulate TGF-ß expression,
possibly by binding cell type-specific cytoplasmic molecules
(16)
.
Control of TGF-ß bioactivity
Members of the TGF-ß superfamily are synthesized as large
precursor molecules that are proteolytically processed in the Golgi
apparatus by the convertase family of endoproteases, of which furin is
a member. Furin cleaves the precursor into a mature TGF-ß and
amino-terminal precursor remnant, also termed latency-associated
protein, or LAP (reviewed in ref 17
and references therein; 18). LAP
remains noncovalently linked to TGF-ß and prevents binding of mature
TGF-ß to its receptors. These so-called small latent TGF-ß
complexes are significantly more stable than bioactive TGF-ß. Within
the Golgi, LAP covalently interacts with latent TGF-ß binding
proteins (LTBPs) to form large, latent complexes (reviewed in ref 19
and references therein). Four different LTBP homologs, LTBPs 1–4, have
been characterized, including several alternative splice variants
(reviewed in refs 19
, 20
). LTBPs function to enhance secretion and
stability of the TGF-ß-LAP complex, ensure correct folding of
TGF-ß, and target the latent TGF-ß complex to the extracellular
matrix of certain cells and tissues for storage or to the cell surface
where activation takes place (reviewed in refs 17
, 19
and references
therein). LTBP-1, -2, and -4 contain RGD sequences that are integrin
binding sites (20
, 21)
. It has been shown that large
latent TGF-ß complexes directly interact with integrin
v/ß1 at the cell
surface (21)
, which may enable TGF-ß to activate
integrin signaling. Activation of latent TGF-ß into biologically
active mature TGF-ß is controlled by proteases like plasmin or
cathepsin, which cleave LAP (Fig. 1
), and by binding of LAP to mannose-6-phosphate receptors (17, 19 and
references therein). Lipoprotein Lp(a) inhibits activation of
plasminogen to plasmin and thereby negatively affects activation of
TGF-ß (reviewed in ref 22
and references therein). Other mechanisms
that have been implicated in activation of latent TGF-ß are
deglycosylation of LAP (23)
, exposure to reactive oxygen
species (24)
, or acidic cellular microenvironments
(25)
. In addition, an important activator of TGF-ß
in vivo appears to be thrombospondin-1, which induces a
conformational change of LAP and thereby activates TGF-ß
(26)
.
|
Signaling diversity by generation of heterodimeric ligands
After proteolytic cleavage of the mature carboxyl-terminal parts,
biologically active TGF-ß1, -ß2, or -ß3 homodimers are generated.
Homodimers are most abundant, but TGF-ß1.2 (27
, 28)
and
TGF-ß2.3 (28)
heterodimers have been identified in
vivo. For activin, four different ß chains have been
identified—ßA, ßB, ßC, and ßE (reviewed in ref 29
and
references therein)—of which ßA and ßB are known to form homo- as
well as heterodimers. Whether heterodimers can also be formed with the
ßC and ßE chains and whether these different isoforms exert
different biological activities is not clear. Inhibins, which are
heterodimers of inhibin chains and activin ß chains, are
functional antagonists of activin signaling (reviewed in ref 5
). In the
case of BMPs, BMP2/7 and BMP4/7 heterodimers are much more potent in
the induction of ventral mesoderm (30)
, as well as in bone
induction (31)
than their respective homodimers. The
coexpression of individual BMPs in several tissues suggests that
heterodimer formation might occur in vivo, and heterodimers
of BMP2 and BMP-7 have been isolated from bovine bone
(32)
. The mechanism underlying potentiated signaling by
heterodimers compared to homodimers has not been investigated, but
might be due to formation of receptor complexes consisting of, for
example, two different type II and two different type I receptors,
which may activate downstream signaling in a more potent manner.
TGF-ß superfamily binding proteins
After production of bioactive TGF-ß superfamily members, several
extracellular proteins can bind and modify their activity (Table 1
). The extracellular matrix proteoglycans decorin and biglycan have been
described to inhibit TGF-ß activity (33)
. Transgenic
mice lacking biglycan expression show dramatically decreased bone mass,
suggesting an important role for biglycan in the constitution of bone,
possibly by its function as a source for storage of TGF-ß superfamily
members (34)
.
|
The serum protein 2-macroglobulin associates with circulating
TGF-ß superfamily members; via its interaction with the
2-macroglobulin receptor, it takes care of clearance of the growth
factors from serum (35
, 36)
. Fibronectin can bind TGF-ß
and promote its bioactivity, possibly by changing it from a latent to a
bioactive conformation (37)
, and collagen IV may store
bioactive TGF-ß in the extracellular matrix (38)
.
Factors known to play important roles during embryonic development are
follistatin, noggin, chordin, and Short gastrulation (Sog), the
Drosophila homologue of chordin (39)
. These
proteins are secreted by the Spemann organizer and compete with BMPs to
exert opposing effects on developmental fates of mesoderm and ectoderm
(reviewed in ref 40
and references therein). Mice deficient in noggin
expression display excessive cartilage formation and failure of joint
specification during skeletogenesis (41)
. The TGF-ß
superfamily member Xnr3 has been reported to interfere with BMP
signaling in the Xenopus Spemann organizer via an
intracellular antagonistic mechanism that prevents BMP synthesis
(reviewed in ref 40
).
Cerberus, DAN, and gremlin have recently been identified as antagonists
of BMP signaling during early embryogenesis (42)
. Cerberus
can also counteract signaling by activin and nodal. These proteins,
like noggin and chordin, have been shown to interact directly with the
growth factors, thereby preventing interaction of the ligands with
their signaling receptors (reviewed in refs 40
, 42
). Follistatin
counteracts activin signaling by a similar mechanism (reviewed in ref 36
), whereas the mechanism by which follistatin interferes with BMP
signaling is unclear; BMP–follistatin complexes were found to interact
normally with BMP receptors (43)
.
Interaction of chordin or Sog with BMPs is controlled by
metalloproteases of the astacin family, to which tolloid, the
Drosophila homologue of BMP1, Xenopus xolloid,
and zebrafish Ztld, belongs, and they are able to proteolytically
cleave chordin and Sog, thereby liberating bioactive BMPs (44; reviewed
in ref 45
).
An extensive number of membrane-bound or transmembrane TGF-ß
binding proteins are known to interfere with TGF-ß action. The type I
and type II serine/threonine kinase receptors, which will be discussed
in detail below, are directly involved in TGF-ß signal transduction.
TGF-ß receptor III (TßR-III; also known as betaglycan) and endoglin
are transmembrane proteins that share high sequence homology in their
short cytoplasmic tails rich in serine residues (46
, 47)
.
Betaglycan has several high-affinity binding sites for TGF-ß1,
TGF-ß2, and TGF-ß3 (48)
and facilitates binding of
TGF-ß to their type II receptors, a property that is especially
important for TGF-ß2, which has low intrinsic affinity for TßR-II
(Fig. 2
) (49
, 50)
. In contrast, soluble betaglycan acts as an
antagonist of TGF-ß bioactivity by preventing the interaction of
TGF-ß with the signaling receptors (46
, 51)
.
|
The role of endoglin in modulation of TGF-ß signaling appears to be
different from betaglycan. In contrast to betaglycan, endoglin can
interact with TGF-ß1, TGF-ß3, activin A, BMP-2, and BMP-7 and
requires appropriate type I or type II receptors for efficient ligand
binding (52)
. Ectopic expression of endoglin in monocytes
results in a selective abrogation of TGF-ß1-induced growth inhibition
and fibronectin synthesis (53)
. Furthermore, mutations in
the genes encoding endoglin or ALK-1, a putative TGF-ß type I
receptor in endothelial cells, form the basis for the vascular disorder
hereditary hemorrhagic telangiectasia type 1 and type 2, respectively
(54
, 55)
, suggesting that endoglin and ALK-1 act in a
common signal transduction pathway.
In many vascular endothelial cells and hematopoietic progenitor
cells, TGF-ß1 and TGF-ß3 are equipotent in mediating signaling, but
TGF-ß2 has weaker biological effects (47
, 56)
. This
difference in potency between different TGF-ß isoforms can be
accounted for by the lack of betaglycan in these cells. Differential
activities of TGF-ß isoforms in other cell types have been observed,
but their modes of action have not been correlated to receptor binding
properties (57
, 58)
. Several cell membrane-associated
TGF-ß binding proteins, including certain
glycosyl-phosphatidylinositol (GPI) -anchored proteins (59
, 60)
, have been identified that are endowed with specific
affinity for certain TGF-ß isoforms, but their role in TGF-ß
signaling is not clear. Furthermore, a 60 kDa TGF-ß binding protein
has been characterized (61
, 62)
that exhibits TGF-ß1
binding specificity and interferes with binding of TGF-ß to the
signaling receptors (62)
.
|
SIGNALING VIA TGF-ß SUPERFAMILY RECEPTORS |
|---|
|
TOPABSTRACTTHE TGF-{beta} SUPERFAMILY AND...SIGNALING VIA TGF-{beta}...INTRACELLULAR SIGNALING BY TGF...CONCLUSIONS AND PERSPECTIVESREFERENCES |
|---|
.
TGF-ß1, TGF-ß3, and activins initially bind to their corresponding
type II receptors, after which the type I receptors are recruited into
the signaling complex. Type I and type II receptors have an intrinsic
affinity for each other, which contributes to stability of the
heteromeric complex (63)
. TGF-ß2, which does not bind to
TßR-II or TßR-I alone, can interact with and stabilize transiently
formed complexes of TßR-I and TßR-II (64)
. BMPs have
low affinity for type II or type I receptors individually, and
high-affinity binding requires formation of a heteromeric type I/type
II receptor complex (65
, 66)
. At present, only a few type
I and type II receptors for the TGF-ß superfamily are known. However,
depending on the ligand, multiple heterotetrameric complexes consisting
of two type I and two type II receptor homodimers are possible
(67)
. In view of the existence of heterodimeric ligands,
it is conceivable that two different type I and two different type II
receptors constitute the receptor complex, thereby creating
combinatorial signaling.
TGF-ßs form heteromeric complexes between TßR-II and TßR-I in
most cell types (68
, 69)
. In endothelial cells TGF-ß can
bind and signal through activin receptor-like kinase 1 (ALK-1); ALK-1
has therefore been implicated as an endothelial-specific TGF-ß type I
receptor (T. Imamura, P. ten Dijke, and K. Miyazono, personal
communication). Activins signal via combinations of ActR-II or ActR-IIB
and ActR-IB (68
, 70)
. Activins have also been shown to
interact with ActR-I (68
, 71)
, although this receptor
appears to play a minor role in activin signaling. BMPs interact with
ActR-II, ActR-IIB, or BMPR-II, in combination with ActR-I, BMPR-IA, or
BMPR-IB (65
, 66
, 72)
. For AMH, only a type II receptor has
been identified (73)
. A potential AMH type I receptor is
ActR-I, which is coexpressed with AMHR-II in the Müllerian duct
and can mediate AMH-specific repression of FSH-induced aromatase
activity (A. Themmen, personal communication). ALK-7 is an orphan
receptor (74)
. A novel type I receptor, designated zALK-8,
was recently cloned and shown to be widely expressed throughout early
zebrafish development (75)
. Mammalian homologues of zALK-8
as well as of zebrafish TARAM-A, which is a type I receptor involved in
induction of anterior dorsal mesoderm (76)
, have not yet
been identified. The ligands for these receptors have not been
elucidated.
Activation and regulation of serine/threonine kinase type I and
type II receptors
Type I receptor kinases become activated upon phosphorylation by
the constitutively active type II receptor kinase (Fig. 2)
. Thus, type
I receptors act downstream of type II receptors (77)
.
Consistent with this finding, type I receptors were found to determine
the specificity of the intracellular signals induced by different
TGF-ß superfamily members, whereas the L45 region in type I receptors
was found to be important in determining the specificity of type I
receptors (78
79
80)
.
The phosphorylation status of TßR-II has been shown to influence
TGF-ß signaling. TßR-II is a constitutively active kinase that
requires phosphorylation of Ser213 (located outside the kinase domain)
to mediate its autocatalytic effect. Ligand-dependent
autophosphorylation of TßR-II on either Ser409 or Ser416
differentially contributes to regulation of its activity, leading to
stimulation or inhibition of TGF-ß signaling, respectively
(81)
.
Differential kinetics in biosynthesis, ligand-induced internalization,
and down-regulation of type I vs. type II receptors have been
attributed to differential modulation of TGF-ß signaling
(82
83
84
85)
. The half-life and processing of TßR-II in the
endoplasmic reticulum are considerably shorter than that of TßR-I
(82
83
84
85)
. Furthermore, homo- and heteromeric receptor
complexes have distinct endocytotic fates (84)
. Rapid
modulation of TßR-II expression levels as well as ligand-bound
TßR-I/TßR-II complexes may be important for fine-tuning of
signaling by TGF-ß.
Diversity in receptor availability might be provided by the
existence of alternative receptor transcripts, as identified for
ActR-IIA (86)
, ActR-IB (87)
, ActR-IIB
(70)
, AMHR-II (73)
, and TßR-II
(88)
. In the case of ActR-IIB, differential affinity of
activin for the alternatively spliced gene products has been observed.
In addition, differences in the cytoplasmic domains offer possibilities
for differential modes of signaling (70)
.
Several growth factors are known to bind to the same type II
receptor—for example, activins and BMPs, which share ActR-II and
ActR-IIB (72
, 89)
. This enables signaling by several
growth factors via a limited scala of available receptor types, whereas
it can lead to competition of different ligands for the same type II
receptors, thereby fine-tuning signaling via type I receptors. This is
the case for activin and inhibin, since inhibin is known to counteract
many of activin functions by competing for binding to the activin type
II receptor (90)
. A similar mechanism has been suggested
to underlie the inhibitory action of zebrafish antivin, most closely
related to mouse lefty, upon activin signaling (91)
, and
appears to contribute to the functional antagonism observed between
activin and OP-1 in Tera-2 embryonal carcinoma cells (89)
.
Whether different ligands might also compete for binding to common type
I receptors is not clear at present.
Cytoplasmic TGF-ß receptor interacting proteins distinct from
Smads
A number of cytoplasmic proteins are now known to interact
with the kinase domain of type I and type II receptors, thereby
manipulating or mediating their signaling capacities (Table 2
). TRIP-1, a protein that contains five WD domains, is known to bind and
become phosphorylated by TßR-II. The interaction requires a
functional TßR-II kinase (92)
. Overexpression of TRIP-1
attenuates TGF-ß-induced PAI-1 transcriptional response, but not
TGF-ß-induced cyclin A response (93)
. The WD
domain-containing protein STRAP can associate with TßR-I and
TßR-II. STRAP and the inhibitory Smad7 synergistically block
TGF-ß-mediated transcriptional activation (94)
. Another
WD-40 repeat protein B, a subunit of protein phosphatase 2A, has
been reported to interact with and become phosphorylated by TßR-I.
B potentiates the antiproliferative effect mediated by TGF-ß
(95)
. TßR-I-associated protein-1 (TRAP-1) interacts
specifically with the activated TßR-I. However, its functional role
in TGF-ß signaling is unclear (96)
.
|
The immunophilin FKBP12 interacts with type I serine/threonine kinase
receptors (97)
through an FK506-like Leu-Pro motif
preceding the kinase domain (98)
. FKBP-12 inhibits
phosphorylation of the type I receptors by the type II receptors,
possibly by sterical hindrance, and may function to prevent leaky
signaling in the basal cell state (99)
. FKBP12 dissociates
from TßR-I after ligand-induced phosphorylation of TßR-I by
TßR-II (97
, 98)
.
Farnesyltransferase (FT) has also been shown to interact with
and become phosphorylated by activin and TGF-ß type I receptors
(99)
. However, farnesyl transferase activity was shown to
be dispensable for TGF-ß signaling (100)
.
Type I receptors as determinants of downstream signal propagation
Activation of type I receptors by type II receptors occurs by
phosphorylation of serine and threonine residues within a juxtamembrane
region preceding the kinase domain of type I receptors that is
characterized by repeats of glycine and serine residues (also termed
the GS domain) (77)
. Specificity of signal propagation to
Smad molecules, the main downstream components in serine/threonine
kinase receptor signaling, is determined by type I receptors (reviewed
in refs 29
, 101
). More specifically, a region flanking ß strands 4
and 5 in the kinase domain, designated L45 loop, confers a high degree
of specificity in Smad interaction with the type I receptors (78
, 80)
. The L45 loop is highly conserved among type I receptors
with comparable signaling specificity, but differs significantly
between BMP and TGF-ß or activin type I receptors (79
, 80)
. Swapping the L45 loop between TßR-I and BMPR-IB, followed
by ligand-stimulation, was shown to result in exchange of Smad1 vs.
Smad2 recognition specificity, accompanied by a switch in
transcriptional responses (79
, 80)
. However, differential
effects between the two BMP type I receptors have been described
(102
103
104)
. CDMP-1 and CDMP-2, which both interact with
BMPR-IB and BMPR-II, differentially regulate osteogenesis (105
, 106)
. Thus, in addition to a role for the L45 loop, other
mechanisms must contribute to specify downstream signaling after
activation of type I receptors.
It has been shown that the cytoplasmic juxtamembrane region of TßR-I
is involved in specifying signal transduction, since mutations of
certain serine or threonine residues within this region selectively
impair TGF-ß-mediated growth inhibition, but do not affect
TGF-ß-induced PAI-1 or fibronectin synthesis (107)
.
However, in vivo phosphorylation of these sites by TßR-II
has not yet been reported. Mutation of Ser165, located amino-terminal
of the GS domain, results in potentiation of TGF-ß-mediated growth
inhibition and extracellular matrix induction but diminishes
TGF-ß-induced apoptosis (108)
. Thus, additional residues
in TßR-I appear to modulate distinct pathways triggered by TGF-ß.
In certain cell systems it has been observed that the expression level
of type I receptors correlates with extracellular matrix production and
gene transcription, whereas the expression level of type II receptors
correlates to the growth inhibitory response, suggesting participation
of type I and type II receptors in different signaling pathways.
However, the observations made are all compatible with an alternative
interpretation, which is that all signals are initiated by activation
of the type I receptor and that different activation thresholds of the
signaling pathways leading to growth inhibition vs. gene transcription
underlie the apparent differential modes of signaling by type I and
type II receptors (109)
.
|
INTRACELLULAR SIGNALING BY TGF-ß SUPERFAMILY MEMBERS |
|---|
|
TOPABSTRACTTHE TGF-{beta} SUPERFAMILY AND...SIGNALING VIA TGF-{beta}...INTRACELLULAR SIGNALING BY TGF...CONCLUSIONS AND PERSPECTIVESREFERENCES |
|---|
, 111)
, was isolated from a genetic screen for
enhancers of a weak Dpp maternal phenotype. Subsequently, three
Mad homologues, called Sma-2, Sma-3,
and Sma-4, were picked up in C. elegans by
searching for genes whose mutations caused the same small body-size
phenotype as observed for mutant Daf4, a type II
serine/threonine kinase receptor (112)
. At present, eight
different Sma- and Mad-related proteins have been identified in
mammals, and are termed Smads (Fig. 3A
). They can be subdivided into three distinct subclasses:
receptor-activated Smads (R-Smads), common-partner Smads (Co-Smads),
and inhibitory Smads (anti-Smads). R-Smads and Co-Smads are homologous
in their amino- and carboxyl-terminals, called the MH1 and MH2 domains,
respectively. These domains are connected by a proline-rich linker
region. R-Smads contain SSXS phosphorylation motifs in their
own carboxy termini.
|
Smad1, Smad5, and Smad8 are involved in BMP signaling and become
phosphorylated by ActR-I, BMPR-IA, or BMPR-IB (113
114
115
116
117)
.
Smad2 and Smad3 are mediators of TGF-ß and activin signaling and
interact with and become phosphorylated by TßR-I or ActR-IB
(118
119
120
121)
. Consistent with these findings, injection of
BMP-Smads into Xenopus animal caps results in
ventralization of mesoderm (115
, 116
, 122
, 123)
, and
ectopic expression of Smad2 in Xenopus embryos induces
dorsal mesoderm and secondary axis formation analogous to
activin/Vg-1-like responses (122
, 123)
. Smad2 null mice
lack anterior-posterior specification and fail to develop mesoderm
(124
, 125)
. The phenotype of Smad2 knockouts is strikingly
similar to Nodal knockouts and suggests that they cooperate in
regulation of gastrulation (124)
. Mutant mice lacking
Smad3 expression have been reported to be viable and fertile; however,
Smad3 -/- animals were found to exhibit limb malformations
(126)
, have a defect in immune function
(127)
, and develop colon carcinomas 4–6 months after
birth (128)
. Smad3 null cell lines showed strongly
impaired TGF-ß responsiveness and indicated an important role for
Smad3 in TGF-ß-mediated inhibition of cellular proliferation
(126)
. The Smad3 knockout phenotype observed by Zhu and
co-workers (128)
classifies Smad3 as a tumor suppressor
gene, although Smad3 mutations have not yet been detected in
tumorigenesis (129)
.
TGF-ß-mediated phosphorylation of Smad1 has been observed in breast
tumor cells, but the receptor by which this phosphorylation is mediated
has not been determined (130)
. Smad5 phosphorylation in
response to TGF-ß was detected in CD34+ cells (131)
.
ALK-1 has been shown to bind TGF-ß and to phosphorylate and activate
BMP-receptor-regulated Smads (113, 114; T. Imamura, P. ten Dijke, and
K. Miyazono, personal communication), and thus could possibly mediate
TGF-ß-induced Smad phosphorylation in these cells.
Three Co-Smads have been characterized: mammalian Smad4/DPC4 (deleted
in colon carcinoma; ref 132
), C. elegans Sma-4
(112)
, and Drosophila Medea
(133)
. Co-Smads play a critical role in signaling by
TGF-ß superfamily members, illustrated by the absence of TGF-ß
signaling and R-Smad functioning in Smad4-deficient tumor cell lines,
as well as by the synergistic action of Smad4 with R-Smads in
transcriptional activation of target genes and induction of mesoderm
(121
, 134
135
136
137)
. The importance of Smad4 functioning is
also shown by its tumor suppressor activity; mutation or deletion of
Smad4 is often associated with tumorigenesis (132
, 138
139
140
141)
. Cell lines expressing Smad4-mutants found in human
cancers failed to transcriptionally activate transfected luciferase
reporters containing Smad binding elements (140)
. Mice
lacking expression of Smad4 show a BMP knockout phenotype,
characterized by growth retardation and gastrulation defects
(142)
. However, the abnormalities in mesoderm development
of Smad4 null mice could be overcome by the presence of wild-type
extraembryonic tissue, indicating that Smad4 is not critical for early
gastrulation (142)
. Although several explanations are
possible, one reasonable argument is that a yet unidentified Co-Smad is
involved in mediating early embryonic differentiation induced by
members of the TGF-ß superfamily. At later stages of development,
Smad4 is important for anterior specifications (142)
.
Additional indications for the possible existence of novel Co-Smads
comes from studies in Drosophila where Medea, in
contrast to Mad, is dispensable for oogenesis
(143)
. An explanation for the fact that Mad knockouts have
a more severe phenotype than Medea knockouts might be that Mad is able
to signal certain responses independent of Medea. Furthermore,
mutations in Medea efficiently block signaling by Saxophone (Sax) but
fail to fully suppress signaling via Thick Veins (Tkv), both of which
are Dpp type I receptors (133). Although Medea might be abundantly
present to allow Tkv signaling even when hypomorphic alleles are
expressed, the inability to fully abrogate Tkv signaling might also
reside in the fact that another, yet unidentified, Co-Smad is involved
in Tkv signaling or that Tkv can signal independent of Medea. Besides
the XSmad4 ortholog, another Smad4/Medea-like protein was recently
identified in Xenopus (144, 145; C. Hill, personal
communication)
Mechanisms controlling Smad activation
R-Smads occur as monomers in the cytoplasm (146)
and,
after ligand-induced type I receptor activation, specifically interact
with the L45 loop in the kinase domain of type I receptors (79
, 80)
. Recruitment of R-Smads to the TGF-ß receptors is promoted
by the action of SARA, a FYVE zinc finger domain-containing protein
that interacts with both TGF-ß receptors and Smad2 or Smad3 (Fig. 2)
(147
, 148)
. The highly conserved L3 loop of R-Smads, which
protrudes from the MH2 core domain as observed in the Smad4-MH2 crystal
structure (139)
, is involved in specifying type I
interactions with TßR-I and BMPR-I (150)
. This loop
varies between Smad1 and Smad2 in only two amino acids, and exchanging
these residues between Smad1 and Smad2 enables their binding to TßR-I
and BMPR-I, respectively (79
, 149)
. The -helix 1 of
Smad1, located near the L3 loop in the 3 dimensional structure, has
been shown to be important for the interaction of Smad1 with ALK-1
(114)
.
After type I receptor interaction, R-Smads become activated by
phosphorylation of their carboxyl-terminal SS(M/V)S motifs (Fig. 3B
) (119
, 150
151
152)
, which triggers their
homodimerization and heteromerization with Co-Smads (134
, 153)
or other R-Smads (121
, 146)
. This leads to the
speculation that homomeric R-Smads, heteromeric R-Smads, and
R-Smad/Co-Smad complexes might control different biological responses.
In contrast to R-Smads, Co-Smads do not contain the carboxyl-terminal
phosphorylation motif SSXS and are not phosphorylated by the activated
type I receptor (119
, 120)
. Although the L3 loop of
Co-Smads shares high amino acid similarity with the L3 loop of R-Smads,
Co-Smads do not interact with type I receptors (120)
;
instead, their L3 loops are important for heteromerization with
phosphorylated R-Smads (139)
. Resolution of the crystal
structure of the Smad4-MH2 domain and analyses of Smad-deletion mutants
(139)
have indicated that the MH2 domains of R-Smads and
Co-Smads are involved in formation of homo-and heteromeric trimers.
After association of R-Smads with Co-Smads, the heteromeric complexes
are translocated into the nucleus (see Fig. 5
).
|
Domains controlling nuclear translocation
Whereas ligand-induced nuclear translocation of R-Smads is not
dependent on expression of Co-Smads (136)
, Co-Smads
require association with activated R-Smads in order to enter the
nucleus (136
, 151)
. A carboxyl-terminal truncation of
Smad4, corresponding to a mutation of Smad4 found in pancreatic tumors
(132)
and identified in an inactive form of Mad in
Drosophila (111)
, dominant-negatively
interferes with nuclear transport of all R- and Co-Smads
(137)
. In contrast, a comparable mutation in R-Smads does
not affect their nuclear localization (137)
. This
indicates that the carboxyl-terminal part of Co-Smads is required for
nuclear trafficking (137)
, whereas phosphorylation of
R-Smads, though required for their functionality, is not obligatory for
nuclear transport. How these mutant Co-Smads interfere with the nuclear
transport function of R-Smads needs to be further studied. Removal of
the NH2 terminus and linker region of R-Smads
results in strong, ligand-independent nuclear localization (122
, 137)
, whereas an additional carboxyl-terminal truncation of the
remaining MH2 domain results in cytoplasmic accumulation at the nuclear
membrane (137)
. Presumably, ligand-induced
heteromerization results in a conformational change of the Smad
molecules, thereby exposing nuclear translocation signals in R- and
Co-Smads. Whether the cytoplasmic residence of R-Smads and Co-Smads in
the absence of ligand is due merely to autoinhibition of the MH2
effector domain by association with the MH1 domain (154)
,
or whether a cytoplasmic tether exists that anchors R- and Co-Smads in
the cytoplasm in the absence of ligand, is not clear
(137)
.
R-Smads and Co-Smads as sequence-specific transcriptional
activators
Once in the nucleus, R-Smads and Co-Smads are involved in
transcriptional regulation of target genes (Fig. 3C
)
(reviewed in ref 155
), and the MH1 and MH2 domains
differentially contribute to this property. Direct binding of Smads to
DNA was first shown by the interaction of the Drosophila
Mad-MH1 domain with specific sequences in the promoter of the Dpp
target gene vestigial (156)
. A prerequisite
for the interaction of Mad, Smad3, but not Smad4, with DNA is the
dissociation of the MH2 domain from the MH1 domain, which is achieved
by ligand-induced phosphorylation of the SSXS motif in R-Smads
(154
, 156)
. Thus, the MH2 domain can exert an
autoinhibitory action on the MH1 DNA binding domain.
In several promoters of genes induced by TGF-ß signaling like
type VII Collagen, JunB, and
PAI-1 (157
, 158
, 160)
, unique elements have
been identified with which Smad3 and Smad4 physically interact. The
sequences of the Smad binding elements (SBEs) in these promoters are
highly related and indicate an AGAC-containing motif as binding site
for Smad3 and Smad4. This sequence was also picked up in a random
screening for Smad interacting DNA sequences (161)
, and
mutation of these SBEs in the PAI-1 promoter impairs TGF-ß
responsiveness (160)
.
Crystalization of the Smad3 MH1 domain bound to an optimal Smad binding
sequence revealed that the so-called ß-hairpin loop, protruding from
the MH1 core domain, interacts with DNA (162)
. The
ß-hairpin loop is highly conserved among R-Smads and Co-Smads.
However, Smad2 contains an extra exon preceding this region, encoding
30 amino acids, that might interfere with the correct conformation of
the ß-hairpin loop; in contrast to Smad3 and Smad4, Smad2 has not yet
been shown to interact with DNA. Removal of exon 3 in Smad2 enables its
efficient binding to DNA (163
, 164)
.
In several TGF-ß target genes, multiple copies of SBEs can be
identified, often located in close proximity to sites for other
transcription factors (158
, 160
, 165
166
167)
. Smads have
been shown to mediate transcriptional activation by interaction with
other regulatory promoter units, like the TPA-responsive elements
(TREs)/AP-1 sites in the collagenase promoter, which actually overlap
in sequence with SBEs (158
, 166)
. Smads cooperate with the
transcription factor TFE3 in the regulation of the PAI-1
gene (168)
. The p21/WAF1/Cip1 and p15/INK4B gene promoters
lack SBEs, but TGF-ß transcriptionally regulates these genes via the
Sp1 site (169
170
171)
. It has been suggested that the
interaction of Smads with additional transcription factors such as AP-1
(Fig. 4
; 166
, 172
, 173
), forkhead activin signal transducers
(FAST; 165, 167, 174), or Sp1 (169
, 171)
confers
additional DNA binding specificity to the Smad-containing
transcriptional complex (see below).
|
R-Smads and Co-Smads as transcriptional (co-)activators
Studies with Xenopus explants, in which injection of
the Smad2-MH2 domain fully induced dorsalization of mesoderm and
secondary axis formation analogous to activin signaling
(122)
, suggested an effector function for MH2 domains.
Furthermore, transcriptional activation properties of R- and Co-Smads
were shown by fusion constructs of Smad1-MH2 or Smad4-MH2 domains with
Gal4-DNA binding domains (175)
. Heteromerization of
Co-Smads with R-Smads is obligatory for their transcriptional
activation properties (136
, 153
, 176)
.
Several domains in Smad4 have now been identified to contribute to its
important role in signaling by members of the TGF-ß superfamily. As
mentioned above, the MH1 domain enables the interaction of
R-Smad/Co-Smad complexes with DNA, which is of particular importance
for Smad2-Smad4 heteromeric complexes (136)
since Smad2 is
not endowed with DNA binding capacities. The MH1 domain autosuppresses
the MH2 effector domain (154)
, and heteromerization of
Co-Smads with phosphorylated R-Smads has been suggested to release the
inhibitory association (177)
. Studies in which chimeras
between Smad1 and Smad4 were tested revealed that the proline-rich
linker region of Smad4 contains a domain that is important for its
transcriptional activity mediated by the MH2 domain (178)
.
This Smad activation domain (SAD) cooperates functionally with the
transcription coactivator MSG1 (178
, 179)
. MSG1 strongly
enhances Smad4 transcriptional activity, provided that Smad4 is present
in a heteromeric complex with R-Smads (179)
. Smad
heteromerization possibly introduces a configuration that allows
trans-activators to associate and endow Smad4 with
transcriptional potencies (179)
.
Smads interact with transcription factors
The first evidence that Smads interact and cooperate with
transcription factors comes from the interaction observed between
Smad2, Smad4, and Xenopus FAST-1 (XFAST-1), making a complex
that interacts with the activin responsive element (ARE) in the
Xenopus Mix2 promoter (see Figs. 4
, 5
) (136
, 180
, 181)
. Human FAST-1 and mouse FAST-2 (mFAST-2) have been
identified as well, but share little sequence homology with XFAST-1
(165
, 167
, 174)
. FAST-1 efficiently interacts with DNA,
and Smad2 and Smad4 contribute in providing additional DNA interactions
as well as essential transcriptional activation properties (136
, 167
, 174
, 180
, 181)
. Phosphorylation of the Smad2 SSMS motif is
required for interaction of -helix 2, exposed from the MH2-trimer
structure, with FAST-1 (136
, 167
, 180
, 181)
.
Whereas Smad2/Smad4/mFAST-2 complexes drive transcriptional activation
of the goosecoid reporter, Smad3/Smad4/mFAST-2 complexes were found to
suppress promoter activation, presumably because Smad3 and Smad4 share
the same DNA binding site (165)
. This suggests that the
relative expression levels of Smad2 and Smad3 in a cell might determine
the outcome of certain biological responses induced by TGF-ß.
Additional evidence for differential roles of Smad2 and Smad3 comes
from observations in human keratinocytes, where TGF-ß stimulates
Smad2 and Smad3 to comparable extents while activin predominantly
activates Smad3 (182)
. In human lung epithelial cells,
Smad3 more potently induces apoptosis compared to Smad2
(183)
.
Cooperation of Smads with the AP-1 complex
Characterization of Smad binding sites in a collagenase
promoter revealed that Smads can mediate transcription via
12-O-tetradecanoyl-13-acetate (TPA) -responsive elements (TREs; 120,
121, 158), to which the transcription factor AP-1, a dimer of c-Jun and
c-Fos, binds. These AP-1 binding sites overlap in sequence with
consensus Smad binding sites (158)
. Smad3 directly
interacts with TRE, and Smad3 and Smad4 activate the TGF-ß-inducible
transcription of TRE-Luc reporter in the absence of c-Jun and c-Fos
(166)
. However, TGF-ß-induced reporter activity was
further enhanced in the presence of c-Jun-c-Fos heterodimers
(166)
. After TGF-ß-induced Smad activation, Smad3 and
Smad4 can interact with JunB, c-Jun, and JunD as well as with c-Fos
(166
, 172)
. By footprint analysis and electromobility
shift assays, it was shown that c-Jun and Smad3 can bind simultaneously
to overlapping sequences in the TRE (166)
.
Crystallographic analysis of the Smad3 MH1 domain bound to DNA
indicated that a conformational change of the NH2
terminus of c-Fos would allow the multimeric Smad-AP-1 complex to
interact with a single AP-1 site (162)
. This cross talk
might be an important level of coordination between Smads and MAPK/JNK
signaling.
Interaction with the coactivators p300/CBP
Insight into how Smads mediate their transcriptional activating
function was recently unraveled by the interaction of R-Smads with the
coactivators p300 and CREB binding protein (Fig. 4)
(CBP; 176,
184–187). These coactivators contribute to transcriptional activation
by loosening the chromatin structure or by increasing the affinity of
certain transcription factors for DNA through their intrinsic (or
associated) acetyltransferase activity. In addition, they act as a
bridging factor between transcription factors and the basal
transcription machinery (184 and references therein). In view of the
importance of Smad4 for transcriptional activation of the Smad-p300/CBP
complex, Smad4 might function as a coactivator as well
(176)
. The interaction between the MH2 domain of R-Smads
and CBP/p300 is triggered by ligand-induced phosphorylation of the
carboxyl-terminal SSXS motif (176
, 184
, 186)
.
Phosphorylation most likely induces a conformational change in the
R-Smad molecule, which then results in exposure of the CBP/p300 binding
site; deletion of the NH2 terminus and part of
the carboxy terminus, leaving the CBP binding site intact, enhances the
affinity of Smad3 for CBP/p300 (184)
. The
carboxyl-terminal region of CBP is required for Smad interaction
(176
, 184
, 186)
and can associate with other transcription
factors like the adenoviral oncoprotein E1A. E1A is known to interfere
with TGF-ß-mediated signaling (188
, 189)
and is shown to
abrogate Smad-mediated transcriptional activation (176
, 184
, 186)
, most likely by direct competition with Smads for binding
to CBP/p300 (188
, 189)
.
Negative regulation of Smad function
Activin and BMPs have been shown to compete which each other in
Xenopus mesoderm development due to limiting amounts of
Smad4 (190). In view of the limited amounts of CBP/p300 in the cell
(184)
and the observed competition between different
signal transduction pathways at the level of coactivator availability
(191), competition between members of the TGF-ß superfamily as
observed in embryogenesis can also be expected to occur at the level of
CBP/p300.
Interaction of the first zinc finger domain of the nuclear protein
Evi-1 with the MH2 domain of Smad3 interferes with binding of Smad3 to
DNA (Fig. 4)
, thereby repressing transcriptional activity of Smad3 as
well as growth inhibition (192). Evi-1 shows a spatial and temporal
expression pattern during mouse embryogenesis, suggesting it could play
a regulatory role during development (193).
Inhibitory Smads
Mammalian Smad6, Smad7, and Drosophila Dad have been
characterized as inhibitors of TGF-ß signal transduction (Fig. 5
) (194
195
196
197
198)
. Whereas their amino-terminal domains are
highly diverse and share only weak similarity to other Smads, they are
homologous to the R- and Co-Smads in their MH2 domains. However,
inhibitory Smads lack the carboxyl-terminal SSXS phosphorylation motif,
which may enable them to stably associate with type I receptors and to
interfere with receptor binding and activation of R-Smads.
Phosphorylation of the type I receptors by the type II receptors is
essential for inhibitory Smad association (194
, 196
, 199)
and the MH2 domain of anti-Smads suffices in exhibiting the inhibitory
effect (194
, 196
, 199
, 200)
, as manifested by inhibition
of R-Smad phosphorylation and abolition of their heteromerization with
Co-Smad and nuclear translocation, as well as abrogation of growth
factor-induced transcriptional responses and growth inhibition
(194
195
196
, 198
199
200
201
202)
. It has been reported that Smad6 can
exert its inhibitory effect via an alternative mode that involves
BMP-induced association of Smad6 with phosphorylated Smad1, thereby
competing with activated Smad1 for heteromerization with Smad4
(200)
(Fig. 5)
.
The expression of the inhibitory Smads is quickly induced upon
stimulation by TGF-ß superfamily members and thereby provides an
autoinhibitory mechanism in TGF-ß signaling (196
, 197
, 201
, 203)
. The human Smad7 promoter contains an SBE consensus site
that, when fused to a luciferase gene, induces luciferase activity
after TGF-ß stimulation (G. Brodin and R. Heuchel, personal
communication). Thus, after receptor activation, R- and Co-Smads might
be directly involved in transcriptional regulation of the Smad7 gene.
The MH1 domain of inhibitory Smads might play a role in specification
of their inhibitory actions: Xenopus and mouse Smad7, which
share 96% identity in their carboxyl-terminal domain but only 51% in
the amino-terminal region, exert differential effects in inhibition of
TGF-ß and activin signaling (199). Mouse Smad7 inhibits
TßR-I-induced phosphorylation of Smad2 and Smad3 (194
, 196
, 199)
, BMPR-IA and BMPR-IB-mediated phosphorylation of Smad1 and
Smad5 (199), and ActR-I mediated phosphorylation of Smad1 (199); it
therefore appears to be a common inhibitor of R-Smad activation.
Whereas Smad6 has been implicated in inhibition of TGF-ß signaling
(194
, 204)
, it appears to play a more pronounced role in
inhibition of BMP signaling (Fig. 5)
(195
, 200
, 201
, 203
, 205
, 206)
. The mechanism of action for inhibitory Smads in TGF-ß
superfamily signaling is evolutionarily well conserved; it has recently
been shown that Dad, which is a Dpp-inducible inhibitory Smad in
Drosophila (197), interacts with Tkv and thereby prevents
binding and phosphorylation of Mad by Tkv (198).
The in vivo relevance of Smad6 and Smad7 is shown in several
systems. Smad6 and Smad7 were initially identified as genes induced by
shear stress in vascular endothelial cells (204)
. Smad7
inhibits activin-induced dorsal mesoderm formation and interferes with
inhibition of activin-mediated growth arrest and apoptosis in B-cells
(206)
. Smad6 and Smad7 antagonize BMP-induced mesoderm
development, have neural-inducing potencies in Xenopus, and
inhibit BMP-induced growth arrest and apoptosis in B cells (200
, 205
206
207)
.
Alternative functions for Smad6 and Smad7 may exist as well. In
Xenopus, XSmad6 was found to be partially or completely
restricted to the nuclei in most cells (207)
. In mammalian
cells, Smad7 has been located in the nucleus in the absence of ligand,
but rapidly accumulated in the cytoplasm after TGF-ß stimulation
(202)
. In accordance with its importance for inhibition of
R-Smad signaling, the inhibitory Smad carboxyl-terminal tail also
harbors domains required for transport across the nuclear membrane
(202)
. A nuclear localization of inhibitory Smads might be
required to allow phosphorylation of R-Smads after receptor activation;
the mechanism whereby Smad7 is translocated to the plasma membrane
after TGF-ß receptor activation is unknown. An interesting
possibility, which remains to be elucidated, is that inhibitory Smads
in addition function in transcriptional regulation. Differential
compartmentalization of the inhibitory Smads, in combination with
induction of their expression after signaling by TGF-ß superfamily
members, and the fact that inhibitory Smads can selectively eliminate
particular R-Smad pathways provide a tightly controlled cell-autonomous
regulatory mechanism.
Modulation of Smad signaling by MAP kinase pathways
The mitogen-activated protein kinase kinase kinase TAK1 (TGF-ß
activated kinase 1) has been shown to be phosphorylated and activated
upon TGF-ß or BMP-4 stimulation (208)
, and can mediate
transcriptional activation of a luciferase reporter driven by a
TGF-ß-inducible element of the PAI-1 promoter containing three AP-1
sites. This effect is strongly enhanced by the TAK1-activator TAB1
(209)
. The in vivo relevance of the cooperative
functioning of TAK1 and TAB1 was demonstrated in Xenopus
mesoderm development in which TAK1 and TAB1 promoted ventral mesoderm
induction and perturbed neural differentiation, thereby substituting
BMP signaling (210)
. The human X-chromosome-linked
inhibitor of apoptosis protein (XIAP) can interact with both BMP
receptors as well as with TAB1 and enhances ventralization of
Xenopus embryos in a TAB1-TAK1-dependent manner
(211)
.
Using the transmembrane and cytoplasmic domain of BMPR-IA as bait in
yeast two-hybrid screens, BRAM1 (BMP receptor-associated molecule) was
identified. The carboxyl-terminal region of BRAM1 is responsible for
specific interaction with BMPR-IA and does not associate with the
kinase domain of TßR-I (212)
. Furthermore, this region
in BRAM1 is also responsible for interaction with TAB1
(212)
. Expression of GST-coupled BMPR-IA, HA-tagged BRAM1,
and myc-tagged TAB1 in COS7 cells showed that the three proteins form a
ternary complex. BRAM1 might function as an adaptor protein for
positioning TAB1 in close proximity of BMPR-IA kinase domain
(212)
. The role of BRAM1 in activation of the TAB1/TAK1
signaling cascade, as well as its possible effect on Smad-receptor
interaction and Smad-activation is not clear at present.
A downstream phosphorylation target in the MAPK cascade triggered by
TAK1 is mitogen-activated protein kinase kinase 4
(MKK4)/stress-activated protein kinase/extracellular signal-regulated
kinase SEK1 (208
, 213)
, which is involved in the
stress-activated protein kinase (SAPK)/c-Jun
NH2-terminal kinase (JNK) pathway, finally
leading to activation of c-Jun. In that JNK activity induced by
hematopoietic progenitor kinase-1 (HPK-1) can be perturbed by
expression of a kinase inactive form of TAK1, there are indications
that TGF-ß might exert JNK activation via the HPK1-TAK1-SAPK/JNK
route (214)
. The small Rho-like GTPases Rho, Rac, and
cdc42, which signal via the SAPK/JNK pathway, have also been shown to
be important for certain aspects of TGF-ß signaling (215
, 216)
.
The SAPK/JNK pathway has been shown to affect TGF-ß signaling via the
Smad-pathway, since overexpression of dominant negative members of the
SAPK/JNK pathway prevent both TGF-ß- and Smad4-mediated
transcriptional response (217)
. The interface of SAPK/JNK-
and Smad-signaling might be at the level of interaction between c-Jun
and Smad3, since these transcription factors have been shown to
associate after TGF-ß stimulation and synergize in activating a
transcriptional reporter containing AP-1 binding sites (163
, 166)
.
Other levels of interaction between MAPK pathways and Smad signaling
have been elucidated. In the linker regions of R-Smads that connect the
MH1 DNA binding domain with the MH2 transcriptional domain of
receptor-activated Smads, several ERK recognition sites (PXSP, or PXTP
motifs) (Fig. 3B
) and JNK motifs (X
