By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
C. Stournaras, Department of Biochemistry, School of Medicine, University of Crete, GR-71110 Heraklion, Greece Fax: +30 2810 394530 Tel: +30 2810 394563 E-mail: firstname.lastname@example.org
The integrated roles of small GTPases in executing the transforming growth factor β (TGFβ) signaling pathway have attracted increasing attention in recent years. In this review, we summarize recent findings on TGFβ signaling during receptor endocytosis, Smad trafficking and actin cytoskeleton remodeling, and emphasize the role of small GTPases in these processes. First, we give an overview of the different endocytic routes taken by TGFβ receptors, their impact on active TGFβ signaling versus degradation and their regulation by the small GTPases Rab, RalA/Ral-binding protein 1 and Rap2. Second, we focus on the mechanisms and regulation of Smad trafficking in the cytoplasm, through the nuclear pores and into the nucleus, and the contribution of Ran GTPase to these events. Third, we summarize the role of Rho small GTPases in early and late cytoskeleton remodeling in various cell models and diseases, and the positive and negative cross-talk between Rho GTPases and the TGFβ/Smad pathway. The biological significance of this exciting research field, the perspectives and critical open questions are discussed.
Transforming growth factor β (TGFβ) is the prototype member of a large, evolutionarily conserved, superfamily of pleiotropic cytokines that also includes activins, bone morphogenetic proteins (BMPs) and growth and differentiation factors, among others . TGFβ controls various physiological processes during embryogenesis and is an important homeostatic regulator in various cell types, for example, epithelial and endothelial cells in adult organisms [1–3]. TGFβ is a growth suppressor because of its cytostatic program . However, during the late stages of cancer and metastasis, TGFβ acts as a tumor promoter because of its ability to enhance processes such as epithelial to mesenchymal transition (EMT), cell motility and invasion, immunosuppresion, angiogenesis and extracellular matrix production [4–7].
All members of the TGFβ superfamily signal via a ‘canonical’ pathway that involves a heterotetrameric complex of two type I and two type II Ser/Thr kinase receptors on the plasma membrane and downstream cytoplasmic effector proteins termed Smads [8,9]. TGFβ promotes receptor oligomerization which leads to the phosphorylation of its type I receptor (TβRI) by the constitutively active type II receptor (TβRII). Activated TβRI (also called ALK5), phosphorylates Smad2 and Smad3 at their C-terminal SSXS motifs [8–10]. The R-Smads, in turn, oligomerize with the common partner Smad4 and rapidly translocate to the nucleus where they bind to the promoters of a large variety of target genes and regulate their expression in a positive or negative manner [8–10]. TGFβ target genes code for proteins involved in cell-cycle regulation, apoptotic regulation, extracellular matrix production, cytokine signaling, transcriptional regulation, differentiation control and autoinhibitory loops . The best understood example of a negative feedback loop involves Smad7, an inhibitory Smad, which blocks Smad phosphorylation by TβRI and directs receptor ubiquitination and degradation via the ubiquitin ligases Smurf1 and Smurf2, thus ensuring that the pathway is shut off [9,11].
Proteins have been identified which recruit Smads to the activated type I receptor for phosphorylation. Smad anchor for receptor activation (SARA) recruits Smad2 into the vicinity of the receptor. Phosphorylation of Smad2 increases its affinity for Smad4 and decreases its affinity for SARA, promoting the dissociation of Smad2 from SARA, unmasking a nuclear localization signal in Smad2 and allowing signaling to occur [12,13]. In the BMP pathway, endosome-associated FYVE-domain protein (Endofin) functions as a Smad anchor for receptor activation [14,15]. Interestingly, both SARA and Endofin are FYVE-domain-containing proteins  and localize predominantly to the early endocytic compartment [17–20], thereby underscoring the importance of this compartment in the signaling cascades of both pathways. Hrs, another FYVE domain protein, also localizes to the early endocytic compartment, binds to Smad2 via its C-terminal domain and cooperates with SARA to stimulate activin receptor-mediated signaling via the efficient recruitment of Smad2 to the receptor . It is therefore evident that receptor endocytosis is an important early step in TGFβ signal transduction.
To date, five emerging transport routes for proteins that become internalized have been identified: the clathrin-coated vesicle-mediated route (CCVMR), macropinocytosis/phagocytosis, the APPL route, the caveolar route and the nonclathrin and noncaveolar pathways . Therefore, it is clear that understanding the endocytic route followed by TGFβ receptor–ligand complexes will allow a systems-level molecular dissection of the signaling regulators of TGFβ.
Since the discovery and molecular cloning of Smad proteins, it has been known that Smads rapidly accumulate in the cell nucleus upon activation of the TGFβ receptors [23–26]. The original studies gave a static view of the pathway, whereby Smads were thought to reside firmly in the cytoplasm and translocate rapidly into the nucleus upon activation via receptor-mediated phosphorylation. Twelve years later, we appreciate that Smad proteins show a very dynamic behavior within the cell because they constantly shuttle in to and out of the nucleus .
Furthermore, both TGFβ receptor endocytosis and Smad trafficking seem to rely on interactions and cross-talk with the cytoskeleton, including microtubules and actin-based microfilaments . Such cross-talk facilitates the timely movement and accurate transport of signaling components to their various destinations. In addition, TGFβ signaling has a profound impact on the regulation of the actin cytoskeleton, which supports various physiological and developmental processes such as cell motility, differentiation changes and tissue organization . The regulatory enzymes of the Ras family, namely Rab, Ran and Rho GTPases are pivotal components in the regulation of TGFβ signaling during receptor endocytosis, Smad trafficking and cross-talk with the actin cytoskeleton, respectively . Here, we provide a detailed review of the specific and integrated roles of small GTPases in the control and execution of the TGFβ signaling pathway.
Interconnection between TGFβ signaling and receptor trafficking-regulation by small GTPases
Endocytosis has long been considered a way of terminating signaling processes via receptor degradation. This was challenged in recent years when activated epidermal growth factor receptors and their effectors were found in what was considered to be the endosomal compartment . It is now evident that endomembrane structures serve as signaling platforms , and there are signaling endosomes or hermesomes which may be specialized for this process . The endomembrane system is divided into functionally and compositionally specialized subdomains [32,33], which determine the strength and duration of signaling responses by controlling recruitment of the downstream effectors of signaling complexes and sorting events such as recycling and transport to the lysosomal compartment for degradation. The endocytic pathway itself is controlled by signaling, demonstrating the extent to which signaling and trafficking are interlinked [34,35]. Furthermore, transport from early to late endocytic compartments is controlled by the cargo, and activated receptors may alter the kinetics to modulate their signaling duration .
Is internalization required for TGFβ family signaling?
The presence of SARA, Hrs and Endofin in early endocytic compartments questions whether signaling can occur from the plasma membrane or whether internalization is required to bring activated receptors to the endosome which is enriched in SARA and Endofin. This issue remains controversial, reflecting differences in experimental approaches and their limitations. TβRII undergoes constitutive internalization in the absence of ligand via clathrin-coated pits. This process is dependent on a short sequence (I218-I219-L220) that conforms to the di-leucine family of internalization signals [37,38] and the direct binding of type I and II receptors to β2-adaptin . No di-leucine motifs have been found in type I receptors. Interestingly, the NANDOR box is well conserved throughout type I receptors  and appears to play a role in type I receptor endocytosis.
Indeed, TβRI (ALK5) is internalized rapidly via CCVMR [41,42]. Ligand stimulation has no effect on the initial internalization rate or receptor recycling . Using a range of techniques including potassium (K+) depletion, which inhibits clathrin-mediated endocytosis , and a dominant-negative form of the dynamin GTPase, K44A dynamin II, which inhibits both clathrin- and caveolar-mediated endocytosis , various groups have addressed the requirement for internalization in TGFβ signaling. Lu et al.  found no involvement, however, several other groups have demonstrated the need for internalization [45,46].
Further studies showed that TGFβ receptors localize to both raft and nonraft membrane domains and the internalization route dictates whether signaling or degradation will ensue [11,47]. Internalization of TGFβ receptors, via the CCVMR, into an EEA1- and SARA-positive endosome promoted signaling. However, internalization via the raft–caveolar pathway, where Smad7 and Smurf2 are localized, promoted ubiquitin-dependent receptor degradation and inhibition of this pathway led to receptor stabilization, suggesting that trafficking of receptors to the SARA-positive early endosome functions to sequester receptors from the rafts and caveolae, thereby stabilizing the receptors .
In support of the above model, hyaluronan, an extracellular matrix polysaccharide, attenuated TGFβ signaling by increasing the segregation of TGFβ receptors into a lipid raft–caveolar compartment , whereas ADAM12 (a disintegrin and metalloproteinase) facilitated signaling by inducing the accumulation of TβRII in early endosomal vesicles and counteracting the internalization of TβRII into a caveolin1-positive compartment . Likewise, interleukin-6 augmented TGFβ signaling by increasing partitioning of TGFβ receptors to the nonlipid raft fraction (early endosomal) . No significant caveolar internalization was observed in the study by Mitchell et al. , in which nystatin (used at lower, more specific doses) had no effect on receptor internalization and degradation. Moreover, TGFβ receptors did not exhibit considerable co-localization in compartments positive for caveolin-1 .
What about the role of endocytosis in the signaling of other members of the TGFβ receptor family? With regard to activin A signaling, an ALK4 mutant, Alk4W477A, that was unable to undergo activin-dependent internalization, retained the ability to signal, demonstrating that ALK4 can signal without receptor internalization . However, in another detailed study addressing the memory of Xenopus embryonic cells to activin A exposure, the critical step in determining the duration of activin A signaling was the time spent by the ligand/receptor complexes in the endo-lysosomal pathway. Activin A internalization was required for correct signaling, suggesting that the localization of ligand to the endosomes was also required for a signaling step upstream of Smad2 activation. Dynamin-dependent endocytosis was necessary to generate signaling complexes, whereas delayed targeting to the lysosome ensured the persistence of signaling by such internalized complexes .
In agreement with the results with endosomal signaling of activin A/receptor complexes in Xenopus, work in Drosophila has shown that mutations in spinster (spin) , hrs/vps27p  and vps25 , which impair endosome-to-lysosome trafficking, cause an increase in BMP signaling, accompanied in some cases by increased levels of Thick Veins (an ortholog of ALK3/6). By contrast, Spichthyin (Spict), the Drosophila ortholog of the SPG6 and ichthyin protein family, which causes segregation (without degradation) of Wit (an ortholog of BMPRII) in early endosomes (Rab5-positive compartment), inhibits BMP signaling . Further work in Drosophila revealed that Nervous Wreck interacts with Thick Veins and the endocytic machinery to attenuate BMP signaling. Because Nervous Wreck co-localizes with Rab11, the authors suggested that Nervous Wreck might regulate the rate at which vacant Thick Veins receptors are recycled back to the plasma membrane following activation and internalization . Indeed, as mentioned below, TGFβ receptors are recycled via a Rab11-dependent mechanism independent of ligand binding, possibly as a means of rapidly and dynamically regulating surface receptor number and thus sensitivity to TGFβ . Recent biochemical data has shed more light on the link between BMP signaling and endocytic trafficking. BMPRI and BMPRII appear to be continuously internalized via CCVMR endocytosis, and BMPRII is also endocytosed via a caveolae- and cholesterol-dependent route . Smad1/5 phosphorylation seems to occur at the plasma membrane; however, continuation of Smad1/5-dependent signaling requires internalization via the CCVMR. The BMP receptor population that resides in cholesterol-enriched, detergent-resistant membrane fractions is required for Smad-independent BMP signaling . However, downregulation of caveolin-1 via siRNA resulted in a loss of BMP-dependent Smad phosphorylation and gene regulation, and was not linked only to Smad-independent signaling .
Rab GTPases are master regulators of vesicular transport and are distributed in distinct intracellular compartments. Rab5 is a key regulator of endocytosis that, by interacting with multiple effectors , regulates organelle-tethering, fusion and motility. Rab7 localizes to the late endocytic compartment and controls the trafficking of late endosomes . Therefore, conversion of Rab5 to Rab7 controls the progression of cargo from the early to the late endocytic compartment, but the cargo itself can also modulate the kinetics of this transport step . Thus, inputs from the Rab5/7 machinery or cargo (activated growth-factor receptors) may modulate the extent of downstream signaling by altering early/late endosome transport kinetics, thereby allowing activated receptors to access and/or reside for longer in an environment that allows productive signaling, especially in the case of TGFβ/activin A pathways in which SARA is enriched in the early endosome.
Indeed, RIN1, a Rab5 guanine exchange factor (GEF), via the activation of Rab5, directs TβRs into an endocytic pathway that promotes TGFβ signaling through Smads  (Fig. 1A). Silencing of RIN1, in turn, reduces TβR signaling efficiency. A negative feedback loop exists, whereby TβR signaling induces SNAI1, which in turn represses RIN1 expression. Interestingly, RIN1 promotes clathrin-dependent endocytosis of RTKs, such as MET and epidermal growth factor receptor, through direct binding to activated receptors and the stimulation of Rab5 proteins . This serves principally to direct RTK receptors to degradation, thereby leading to reduced signaling [63–65]. The differential signaling outcome of RTK versus TβR signaling by RIN1, however, suggests that Rab5-mediated endocytosis is not inextricably linked to a particular signaling outcome. Multiple endocytic complexes, each containing RIN1 and Rab5, and also other distinct components, may help explain different signaling outcomes during and following receptor internalization.
An additional consideration is the length of time TGFβ family receptors reside in early endosomes  once trafficked there. This is important because the signaling outcome is proportional to the residence time in this compartment. Trafficking of TGFβ family receptors via early endosomes with extremely fast kinetics will most likely have a minimal enhancing effect on signaling compared with early endosomal trafficking that is accompanied by blocking of further trafficking. Indeed, several studies on activin A and BMPs have revealed that the enhancing effect on signaling of various proteins was dependent on how long the relevant receptors resided on early endosomes [52–54]. Whether conversion of Rab5 to Rab7 or other mechanisms are responsible remains open. Our previous results suggest that Rab5 cycling between the GTP and GDP forms may influence the length and intensity of TGFβ/activin signaling cascades by regulating TGFβ–activin type I/II receptor trafficking via the early endocytic compartment . Indeed, in endothelial cells, Rab5S34N, a Rab5 mutant locked in the GDP form, caused augmented Smad3-dependent transcription in the absence of ligand. Because RN-tre, a specific Rab5 GTPase-activating protein (GAP) that blocks plasma membrane endocytosis, did not influence Smad3-dependent transcription, we concluded that the effect of Rab5S34N should have been the consequence of decreased degradative or recycling trafficking, leading to an accumulation of constitutively formed TGFβ–activin type I/II receptor complexes on early endosomal membranes.
Certainly, the station after early endosomes in the trafficking route of TGFβ family receptors is critical. Recycling back to the plasma membrane will influence signaling differently compared with trafficking towards late endosomes/lysosomes. This issue has been investigated by overexpressing dominant-negative forms of Rab4 (Rab4S22N) and Rab11 (Rab11S25N) and assessing TGFβ receptor trafficking . Rab4 regulates recycling from sorting/early endosomes to the plasma membrane, whereas Rab11 controls recycling through the perinuclear recycling endosomes  and trans-Golgi network to plasma membrane transport . Only Rab11S25N caused significant intracellular retention of TGFβ receptors, in both the presence and absence of ligand. Because co-localization of TGFβ receptors with Rab11 has been reported , it seems that, after clathrin-dependent internalization, TGFβ receptors recycle (irrespective of their activation state) in a Rab4-independent and Rab11-dependent manner (Fig. 1A). To date, the effect of Rab4 and Rab11 mutants or siRNA silencing on TGFβ signaling has not been investigated. However, it is expected to influence TGFβ signaling, especially its developmental aspects.
RalA/ Ral-binding protein 1
RalA is a multifunctional GTPase that is activated by receptor-activated Ras via recruitment of Ral GEFs [68–70]. Activated Ral associates with the Ral effector Ral-binding protein 1 (RalBP1), a cytosolic protein that is recruited to membranes following Ral activation  and activates hydrolysis of GTP bound to Rac1 and Cdc42. RalA has been implicated in many intracellular trafficking events  from the regulation of the endocytosis of EGF and insulin receptors  to secretion . Indeed, RalA, via its effector protein RalBP1, interacts with the μ2 subunit of the AP-2 complex  as well as with REPS1  and POB1  which are EGF receptor substrates containing Eps15 homology (EH) domains. POB1 interacts directly with the EH-containing proteins epsin and eps15, which have been reported to be involved in the regulation of EGF and transferin receptor endocytosis [67,78,79]. Thus, activation of RalA by EGF and insulin suggests that RalA/RalBP1 and its interactions with the μ2 chain of AP-2, REPS1, POB1, epsin and eps15 act as a scaffold that conveys signals from receptors to the endocytic machinery, thereby regulating ligand-dependent receptor-mediated endocytosis. Moreover, REPS1 interacts with Rab11-FIP2  a Rab11 effector that may couple REPS1-containing vesicles originating from clathrin-coated vesicles (and the early endocytic compartment) to the recycling endosomes.
RalA and RalBP1 appear to be involved in activin A receptor trafficking and signaling (Fig. 1B). It has been shown that activin receptor interacting protein 2 (ARIP2) interacts with ActRIIs and regulates their endocytosis via a PDZ domain-mediated interaction, concentrating them in a perinuclear compartment. Thus, ARIP2 reduces the response to ligands by decreasing the levels of ActRII at the plasma membrane . ARIP2 triggers the endocytosis of ActRIIs via Ral/RalBP1. Indeed, ARIP2 associates with ActRIIA and RALBP1 via its PDZ domain and C-terminal region, respectively. Because ARIP2C, the C-terminal deletion mutant of ARIP2 that does not bind RalBP1, failed to induce ActRII endocytosis, it appears that endocytosis of ActRIIs by ARIP2 is RalA/RalBP1 dependent. Moreover, activin A activates GDP–GTP exchange in RalA . Activation of RalA/RalBP1 by activin A is calcium dependent, in contrast to activation by EGF and insulin, which occurs via a Ras-dependent cascade . Interestingly, because only ActRIIs among all the serine/threonine kinase receptors for BMP/TGFβ/activin have the PDZ-binding sequence (ESSL for ActRIIA and ESSI for ActRIIB) , PDZ protein-regulated endocytosis and sorting is expected to influence only ActRIIs. Because ActRIIs bind both activins and also nodal and BMP7, ARIP2 is likely to play a role in shaping the activin/nodal/BMP gradient by regulating the endocytosis of ActRIIs.
Rap2 is a member of the Ras family of small GTPases whose effector domain is almost identical to that of Ras, and can therefore bind most Ras effectors. Rap2 inhibits many Ras pathways including Ras-induced Raf activation at the plasma membrane . Rap2 also binds to the Ral GEFs, Ral GDS, RGL and RLF . These proteins are also Ras effectors and induce nucleotide exchange leading to the formation of active RalA. As discussed above, Ral has been implicated in activin A receptor trafficking and may be linked to the molecular mode of action of Rap2 in Xenopus, as explained below.
In a very elegant study in Xenopus embryos, Rap2 was shown to regulate activin/nodal signaling by modulating receptor trafficking  (Fig. 1C). In the absence of ligand, Rap2 directs activin/nodal receptors into a Rab11-dependent recycling compartment, thereby avoiding degradation and maintaining cell-surface levels of receptors. Upon ligand addition, Rap2 no longer directs the receptors for recycling, but rather competes with Smad7 and delays receptor degradation, thus enhancing signaling. Moreover, Rap2 is initially enriched in the dorsal region of the blastulae, then as gastrulation proceeds, it decreases dorsally and increases ventrally. However, Smad7 is expressed uniformly across the dorso–ventral axis in early gastrulation and as gastrulation proceeds, Smad7 is restricted to the ventral region. Thus, Smad7 and Rap2 levels appear to regulate Smad2 activation along the dorso–ventral axis of the developing embryo.
Growing evidence links the progression of TGFβ receptor signaling to key regulatory steps in endocytic trafficking. These steps involve the active regulation of GDP-to-GTP exchange by various small GTPases of the Rab/Ral and Rap families. These mechanisms ensure optimal signal transduction from active receptor complexes to activated Smads.
Intracellular Smad trafficking – the role of the Ran GTPase
Most current evidence on the mechanisms that govern the dynamic shuttling of Smad proteins in the cell is based on the behavior of engineered GFP–Smad2 and GFP–Smad4 fusion proteins which are stably expressed in human cells cultured in vitro. The evidence supports a model whereby Smads shuttle constantly, although each specific Smad seems to obey distinct kinetic properties during its movements . Mathematical modeling of Smad protein shuttling has recently suggested that the strength of Smad signaling depends directly upon the length of time a certain Smad molecule spends in the nucleus . Such kinetic analysis also emphasized that the nuclear export of Smads is highly regulated, whereas the nuclear import of Smads may act as a default pathway.
The evidence from the in vitro cell system is complemented by pioneering in vivo studies first developed in Xenopus embryos [88,89]. Continuous shuttling of Smad2 could be observed in developing Xenopus and zebrafish embryos . Furthermore, Smad2 and Smad4 proteins fused to fluorescent protein fragments fluoresce only when a Smad2–Smad2 homo-oligomer, Smad4–Smad4 homo-oligomer or Smad2–Smad4 hetero-oligomer forms inside the living cells of Xenopus embryos caused by trans-complementation of the fused fragments . Cells in the developing Xenopus embryo are responding to the TGFβ members nodal or activin and show accumulation of Smad4 homo-oligomers only in the cytoplasm, whereas Smad2 homo-oligomers and Smad2–Smad4 hetero-oligomers accumulate in the nucleus. These experiments demonstrated that Smad2–Smad4 oligomers can be observed in the nuclei of developing embryonic cells only when these cells reached the proper developmental stage. This observation suggested that factors independent of nodal/activin signaling regulate the ‘competence’ of the embryonic cell to accumulate nuclear Smad2–Smad4 oligomers. Smad trafficking may be classified according to the cellular compartment where this specific movement occurs. Thus, we can consider Smad trafficking in the cytoplasm, Smad trafficking through the nuclear pores and Smad trafficking inside the nucleus.
Smad trafficking in the cytoplasm
When Smad2 moves inside the cytoplasm it associates with the motor protein kinesin-1 and the integrity of the microtubular network is essential to support this type of motility . This new evidence is compatible with an older study that first identified an inherent ability of all Smad proteins to associate and localize on microtubules . Another motor-like protein that associates with Smad2 is the dynein light chain km23-1, which assists in the nuclear accumulation of Smad2, and also regulates trafficking of the TβRI . According to this new evidence, cytoplasmic Smads traffic towards the signaling receptors with the help of kinesin motors that slide on microtubules. The signaling receptors most likely reside on endosomes, as discussed above. However, cytoplasmic Smad trafficking towards the nucleus involves the dynein motor–microtubule machinery. Although it makes sense to consider microtubules as trafficking highways that facilitate the movement of Smad proteins, microtubules have also been shown to act as cytoplasmic traps for Smads . According to this model, connexin 43 is a regulatory protein that competes with Smads for binding to microtubules. However, the latter mechanism needs to be further clarified as it is important to understand which factor regulates the residence of Smads on microtubules versus their mobility along microtubules and towards neighboring cellular locations.
The association of Smads with microtubules provides additional insight into the functional regulation of these proteins. In dividing cells, such as those of the Xenopus embryo, Smads can associate with the spindle and decorate the metaphase chromosomes . This evidence is compatible with a role for microtubules in trapping Smads and protecting their integrity, thus delivering them safely to the daughter cells after mitosis. It remains unclear as to whether Smad signaling may also regulate mitosis or cytokinesis. However, in addition to protecting Smad integrity, microtubules may also guide a pool of Smads towards their ultimate turnover. The site of assembly of the microtubular network is known to be the centrosome, a subcellular structure in which Smads that are phosphorylated in their linker domain can also localize and undergo ubiquitin-dependent proteasomal degradation . It appears that Smads may slide along microtubules to reach the centrosomes and become degraded . Interestingly, when cells divide, the pool of linker-phosphorylated Smads that traffic towards the centrosome segregates together with other ubiquitinated proteins on the mitotic spindle towards only one of the two daughter cells . This mechanism ensures that proteins targeted for disposal go to only one of the two daughter cells, leaving the other relatively clear of such signaling byproducts. A deeper understanding of the role of microtubules in the regulation of Smad trafficking and signaling is clearly warranted.
Smad trafficking through nuclear pores
The entry of Smad proteins to the nucleus is regulated by specific interactions with transporters and nuceloporins. A lysine-rich nuclear localization signal (NLS) located in the N-terminal Mad homology 1 domain of all Smads binds to importin-β in the case of Smad3 and importin-α in the case of Smad4, while mutation of the NLS blocks the ability of these proteins to enter the nucleus [95–98]. Although the functional role of the Smad2 NLS has not yet been determined, the long Smad2 isoform that incorporates exon 3 fails to bind to importin-β, whereas the shorter Smad2 isoform that lacks exon 3 binds to importin-β similar to Smad3 . In addition, the importin moleskin mediates the nuclear entry of the Drosophila R-Smad Mad, and its human orthologues, importin-7 and importin-8, mediate the nuclear translocation of Smad1, Smad2, Smad3 and Smad4 in human cancer cells in response to BMP or TGFβ signaling . Future work may explain why Smads utilize multiple importins for their entry to the nucleus (Fig. 2).
Importins are known to move through the pore by consecutive contacts with the phenylalanine/glycine (F/G)-rich repeats of specific nucleoporins. Such step-wise translocation is energetically demanding and requires GTP expenditure. Similar to the role of Rab GTPases that control the trafficking of endocytic vesicles during TGFβ signaling in the cytoplasm (Fig. 1), the small GTPase Ran controls Smad3 trafficking via the nuclear pore (Fig. 2) . Ran is a small GTPase dedicated to the control of nucleocytoplasmic trafficking and chromosomal segregation during mitosis . A Ran activity gradient is established through the nuclear pore with high Ran–GDP concentrations in the cytoplasmic phase of the pore which gradually decrease along the pore . In the nuclear phase of the pore, the Ran-specific GEF RCC1 loads Ran with GTP, thus establishing a high Ran–GTP concentration in the nucleus. GDP-bound Ran drives the transport of Smad3 through the pore, whereas Ran–GTP induces the allosteric change needed to dissociate Smad3 from importin-β (Fig. 2) . Ran also mediates importin-β trafficking back into the cytoplasmic phase of the pore .
In addition to binding to importins, Smad2 can also bind directly to the F/G-rich repeats of nucleoporins Nup214 and Nup153 of the nuclear pore (Fig. 2) [103,104]. However, whether Smad3 and Smad4 bind to the nucleoporins directly or via the importins remains unclear [103,104]. In addition, it would be interesting to examine whether Smad2–nucleoporin interactions are regulated by the Ran GTPase gradient along the nuclear pore. Analysis of importin-7 and importin-8 as Smad carriers suggested that continuous Smad shuttling in the absence of ligand activation is independent of the action of transportins, and is presumably facilitated by direct contacts with nucleoporins . By contrast, when TGFβ receptor activation leads to R-Smad phosphorylation, nuclear import seems to depend on the activity of specific transportins. Thus, different mechanisms of nuclear import might operate at different stages of the TGFβ signaling pathway.
The cytoplasmic distribution of Smads in the resting cell seems to be regulated by the dominant role of Smad nuclear export [86,87]. Upon ligand-dependent signaling, nuclear Smad complexes prevail but eventually shuttle back to the cytoplasm, thus providing a way of dampening the strength of the signal or alternatively replenishing the cytoplasmic pool of Smads with molecules that are ready to become activated again, as long as the receptors remain active. The importance of nuclear export is underscored by the presence of nuclear export signals (NES) in all Smads examined to date. Smad4 carries a leucine-rich NES in its linker domain, which mediates export via exportin-1/chromosome region maintenance 1 (CRM1) (Fig. 2) [105,106]. Mutation of hydrophobic amino acids within the Smad4 NES or exposure of cells to the pharmacological inhibitor of CRM1 leptomycin-B, lead to an exclusive nuclear distribution of Smad4, independent of the presence or absence of ligand. Smad3 is exported from the nucleus in a CRM1-independent manner and an extended peptide surface of the MH2 domain has been identified as critical for this export by exportin-4 . In the case of Smad3, the role of Ran has been studied and it was clearly demonstrated that, similar to many other exported proteins, Ran supports the movement of Smad3 via the nuclear pore towards the cytoplasm (Fig. 2). The Smad3 NES has no obvious resemblance to a bipartite leucine-rich motif identified in the MH2 domain of Smad1, the R-Smad of the BMP pathways, which is thought to be recognized by CRM1 based on leptomycin-B inhibitor experiments . The role of Ran in mediating the export of proteins from the nucleus follows the inverse biochemical steps used for import of proteins to the nucleus [100,102]. Ran–GTP promotes the association of Smad3 with exportin-4 in the nuclear phase of the pore . Upon trafficking via the nuclear pore, Ran–GTP in complex with cargo is attacked by Ran GAP, which is associated on the cytoplasmic phase of the nuclear pore, and activates the GTPase activity of Ran so that GTP is hydrolyzed to GDP and orthophosphate (Fig. 2) [100,102]. This leads to conformational changes in Ran that facilitate disruption of the complex between exportin and its cargo, and the ultimate release of cargo to the cytoplasm.
Smad trafficking in the nucleus
Although a growing understanding of the mechanisms that guide bidirectional Smad trafficking in the cytoplasm and through the nuclear pores is now established, nothing is known about Smads trafficking within the nucleoplasm. Classically, the native, yet weak, ability of Smads to bind to DNA has suggested that upon entry to the nucleus, Smads might tether chromatin. However, the current dynamic shuttling model of Smads necessitates a more dynamic view of the nuclear residence of these proteins. The dynamic shuttling model disfavors long-lasting and very stable tethering mechanisms, however, it allows for the highly regulated formation of protein complexes between Smads and nuclear residents. In fact, nuclear Smads are known to bind to a high number of nuclear transcription factors and the role of such interactions in the timing and shuttling behavior of Smads remains unexplored . One nuclear factor that seems to fulfill the criteria for a tethering factor and which might coordinate the nuclear residence time of Smads and the process of transcription is the newly reported protein transcriptional coactivator with PDZ-binding motif (TAZ) . TAZ is a transcriptional regulator containing a WW domain and promotes the nuclear accumulation of Smads. Loss of TAZ perturbs the ability of Smads to accumulate in the nucleus. TAZ binds the transcriptionally active Smad complex and anchors it to ARC105, a central component of the transcriptional mediator complex. TAZ has a close homolog, the WW domain protein YAP, which might also be involved in a similar mechanism. Thus, we await significant developments in Smad nuclear trafficking that might provide a more comprehensive view of how the entry and exit of Smads from the nucleus coordinates with transcription. It will also be interesting to examine the role of additional nuclear small GTPases as regulators of nuclear Smad function, because this class of proteins offers a versatile regulatory system that empowers biological processes with the ability to switch on and off.
The role of small GTPases of the Rho subfamily in TGFβ-induced actin cytoskeleton remodeling
Actin cytoskeleton remodeling is one of the earliest cellular responses to extracellular stimuli [110–115]. Binding of ligands to the appropriate receptors triggers specific signaling cascades, which may generate rapid and long-term modifications of actin polymerization dynamics and microfilament organization [116–120]. Among the specific signaling effectors regulating actin architecture, the family of small Rho GTPases has a prominent role. Classically, plasma membrane receptors activate specific guanine-exchange factors often via phosphorylation, which leads to the subsequent activation of Rho GTPases . Rho GTPases have been implicated in many cellular processes, including actin and microtubule cytoskeleton organization, cell division, motility, cell adhesion, cell-cycle progression, vesicular trafficking, phagocytosis and transcriptional regulation [122,123]. Rho proteins cycle constantly between GTP-bound active forms and GDP-bound inactive forms, and this process is regulated by various factors including GEFs, guanine nucleotide dissociation inhibitors and GAPs . As well as contributing to physiological processes, Rho GTPases have been found to contribute to pathological processes including cancer cell migration, invasion, metastasis, inflammation and wound repair [122,123]. Although Rho proteins do not seem to be mutated in cancer cells, their expression is often elevated, indicating that Rho dysregulation promotes malignant phenotypes .
Rho proteins can be subdivided into three major groups: Rho (RhoA, RhoB, RhoC), Rac (Rac1, Rac2) and cdc42 proteins . Active Rho GTPases transmit signals via downstream effectors such as Rho coiled-coiled kinase 1 (ROCK1), p21-activated kinase 1 and neural Wiskott–Aldrich syndrome protein [126,127]. Activated p21-activated kinase 1 and ROCK1 phosphorylate and activate LIM-kinases 1 and 2, respectively [128–132]. Eventually, LIM-kinases 1 and 2 phosphorylate actin-depolymerizing proteins such as cofilin, destrin and actin-depolymerizing factor, which are inactivated and thus permit actin polymerization to occur [128–130,133,134].
Mechanisms of TGFβ-induced actin cytoskeleton remodeling – short- and long-term events
The ability of TGFβ to regulate actin cytoskeleton remodeling has been demonstrated in a variety of cell systems, and specific members of the Rho subfamily of small GTPases including RhoA, RhoB, Rac and cdc42 have been found to play essential roles (Fig. 3). The contribution of individual Rho GTPases and their downstream effectors in TGFβ-induced actin remodeling has been studied using a variety of experimental tools. These tools include constitutively active and dominant-negative mutants of Rho proteins or their target proteins, siRNA-mediated gene silencing or general inhibition of Rho function using molecules such as the C3 exoenzyme, which selectively ADP-ribosylates and inactivates low molecular mass G proteins of the Rho subfamily at an asparagine residue within the effector domain. Rho GTPase activation is generally measured by affinity precipitation using appropriate GST–fusion peptides that bind only to GTP-bound Rho proteins such as GTP–Rhotekin binding domain for RhoA and RhoB or GST–p21-activated kinase and GST–Wiskott–Aldrich syndrome protein for Rac1 and cdc42 . Changes in the actin cytoskeleton are monitored by immunofluorescence microscopy of rhodamin/phalloidin-labelled actin or by calculating the ratio of total versus polymerized actin by immunoblotting Triton-soluble (globular actin) and Triton-insoluble (filamentous actin) cell extracts .
TGFβ-induced cytoskeleton rearrangements involving Rho activation in EMT
The most extensively investigated TGFβ-induced cytoskeleton rearrangements are the differentiation of epithelial to mesenchymal cells, a process that is called epithelial to mesenchymal transition or transdifferenrtation (EMT) [137,138]. EMT is characterized by the dissolution of epithelial cell–cell junctions and reorganization of the actin cytoskeleton with the formation of focal adhesions and stress fibers, acquisition of a spindle-shaped morphology, delocalization of E-cadherin from cell junctions and elevated N-cadherin expression. This process is associated with embryonic tissue movements and also with cancer cell invasiveness and metastasis. The earliest event in TGFβ-induced EMT is the activation of RhoA which occurs within 5 min of TGFβ stimulation. This activation lasts for a short time (15 min to 3 h, depending on the cell type) and is followed by the activation of downstream target kinases such as ROCK. Rapid RhoA activation was found to operate in a variety of cell models of TGFβ-induced EMT under physiological or pathological conditions. (a) In the atrioventricular canal of the embryonic chicken heart, TGFβ was found to promote the conversion of endothelial cells to mesenchymal cells via a pathway that requires the activation of RhoA . (b) During tubulointerstitial fibrosis, TGFβ promotes the differentiation of tubular epithelium to mesenchymal cells via a biphasic activation of RhoA and its downstream target ROCK; a rapid and transient elevation of RhoA-GTP levels which was detectable as early as 1 min after TGFβ stimulation and lasted for 5 min, and a chronic elevation at 24 h of stimulation. Chronic activation was correlated with the upregulation of α-SMA gene expression via activating protein 1 (AP1) factors . (c) In proliferative vitroretinopathy, TGFβ leads to the transformation of retinal pigment epithelial cells to contractile fibroblasts via rapid activation of RhoA and Rac1 GTPases and their downstram effectors ROCK kinase, LIMK and cofilin, and the concomitant upregulation of α-SMA gene expression .
TGFβ-induced Rho GTPase activation and actin remodeling in various cell systems
In addition to their role in EMT, Rho GTPases and their downstream effectors are activated by TGFβ and contribute to cytoskeletal rearrengements in other cell systems. (a) TGFβ promotes the differentiation of neural crest cells into vascular smooth muscle cells via a rapid (5 min) activation of RhoA and ROCK1 . In this study, it was shown that inhibition of RhoA activity blocked Smad phosphorylation by TGFβ, suggesting that RhoA and Smads may cooperate in TGFβ signaling responses, a concept that is dicscussed thoroughly below. (b) TGFβ promotes the differentiation of rat pulmonary arterial smooth muscle cells via a rapid (2 min) activation of RhoA, which was followed by activation of its downstream kinases ROCK, PKN/PRK2 and p38 MAPK and the transcriptional upregulation of smooth muscle-specific genes such as α-SMA, SM-MHC and SM-22α via the cooperation of serum response factor, GATA and myosin enhancer factor 2 transcription factors  (Fig. 3). (c) In the human prostate carcinoma cell line PC-3U, TGFβ induces the rapid (5 min) formation of membrane ruffles via activation of RhoA and Cdc42 in a Smad-independent manner . In the same study, it was also shown that TGFβ induced long-term actin remodeling and stress fiber formation which required an active Smad pathway. Thus, this study revealed that the rapid and sustained changes in actin cytoskeleton reorganization that are observed in response to TGFβ are mechanistically distinct processes and could be mediated by separate nongenomic and transcriptional signaling pathways that are induced by the same stimulus. Short-term and sustained actin cytoskeleton remodeling has been investigated thoroughly in fibroblasts such as H-ras transformed NIH3T3 fibroblasts, mouse embryo fibroblasts and Swiss3T3 fibroblasts [145–148]. In fibroblasts transformed by inducible expression of the H-Ras oncogene, TGFβ induced the formation of new stress fibers from focal adhesions as early as 15 min post TGFβ addition and this reorganization was associated with an increase in the polymerization state of actin and in protein levels of RhoA and RhoB .
In Swiss3T3 fibroblasts, TGFβ induced rapid activation of both RhoA and RhoB small GTPases as early as 5 min post TGFβ1 addition which remained high for 3 h before decreasing . Activation of RhoA and RhoB was accompanied by phosphorylation of the downstream effectors LIMK2 and cofilin, whereas inhibition of ROCK1 completely blocked TGFβ1-induced LIMK2/cofilin phosphorylation and downstream stress fiber formation (Fig. 3). In these cells, TGFβ induced fibroblast to myofibroblast differentiation, which was evidenced by enhanced expression of α-SMA and the subsequent incorporation of α-SMA into microfilamentous structures . Fibroblast to myofibroblast conversion is a pathophysiological feature of various fibrotic diseases such as idiopathic pulmonary fibrosis, asthma and chronic obstructive pulmonary diseases [149–151]. Given that enhanced TGFβ concentrations have been detected in various fibrotic diseases, including idiopathic pulmonary fibrosis [152,153], sarcoidosis  and cystic fibrosis , understanding the mechanism that underlies this TGFβ-induced conversion may lead to the development of novel therapeutic approaches for these diseases.
Non-Smad and Smad pathways in Rho GTPase activation by TGFβ
Certain studies have examined the participation of MAP kinases or the phosphatidylinositol 3-kinase as downstream signaling effectors that cooperate with Rho GTPases or are activated by them to achieve TGFβ-induced cytoskeleton remodeling. Edlund et al.  showed that treatment of human prostate cancer cells with an inhibitor of the p38 MAP kinase (SB203580) at a concentration that was unable to block the activity of TβRI, as well as ectopic expression of a kinase inactive p38 mutant, abrogated the TGFβ-induced actin reorganization. The same group also showed that TGFβ-induced membrane ruffling and stress fiber formation in prostate cancer cells requires an active phosphatidylinositol 3-kinase pathway . Chen et al.  used neural crest stem cells to show that p38, p44/42 MAPK and phospatidylinositol-kinase inhibitors did not counterbalance the TGFβ induction of α-SMA expression in smooth muscle cell differentiation from stem cells. By contrast, in a different system of smooth muscle cell differentiation, Deaton et al.  showed that inhibition of p38 MAPK by SB203580 blocked the TGFβ1-mediated activation of α-SMA and other SMC marker genes (Fig. 3).
Although the extremely rapid activation of Rho GTPases in response to TGFβ stimulation implies the involvement of non-Smad pathways, in certain cases it was found that the Smad pathway may also play a role in the early activation of Rho proteins by TGFβ. By studying the signaling properties of a TβRI bearing a mutation in its L45 loop, which contains the Smad docking site , Vardouli et al.  demonstrated that interaction of TβRI with R-Smads is required for signaling towards Rho GTPases and the actin cytoskeleton. The role of Smads in TGFβ-induced actin remodeling is further supported by experiments in a cellular model lacking endogenous Smad3 expression (JEG3 choriocarcinoma cells) . In addition, TGFβ-induced Rho activation and cytoskeleton organization was abolished by overexpression of the inhibitory Smad7 protein which blocks the TGFβ/Smad signaling pathway [147,157]. This observation is in contrast to a study showing that Smad7 is required for TGFβ-induced activation of Cdc42 and the concomitant reorganization of the actin filament system, as discussed below.
The ability of TGFβ to affect both rapid and sustained actin cytoskeleton remodeling in various cell types [144,147,148] implies that genomic actions of TGFβ may be involved in long-term cytoskeleton reorganization (Fig. 3). In support of this, it was shown that treatment of Swiss3T3 fibroblasts with actinomycin D, a well-established inhibitor of active gene transcription, abolished TGFβ-induced actin reorganization in Swiss3T3 fibroblasts and HaCaT keratinocytes (E. Vasilaki, E. Papadimitriou, C. Stournaras and D. Kardassis, unpublished results). TGFβ also activates the expression of Rho GEF NET1 via the Smad pathway (Fig. 3) . In a recent study, it was shown that TGFβ induces the transcription of RhoB in mouse fibroblasts and human hepatocytes (Fig. 3) . In this study, Vardouli et al.  demonstrated that transcriptional upregulation by TGFβ was specific for RhoB, because the expression of RhoA was not affected, and that the TGFβ/Smad pathway activated the human RhoB but not the RhoA promoter. Expression of the endogenous RhoB gene in fibroblasts was also upregulated by overexpression of TGFβ-regulated Smads via adenovirus-mediated gene transfer . Similar observations have been made in HaCaT keratinocytes (Vasilaki et al., unpublished results). However, this genomic effect of TGFβ on RhoB gene expression is cell type-specific because it could not be observed in mink lung epithelial cells. In these cells, TGFβ promoted the accumulation of RhoB protein without a concomitant increase in RhoB mRNA levels .
Cross-talk between Rho GTPases and the TGFβ/Smad pathway
In addition to the established positive role of Rho GTPases in TGFβ-induced actin cytoskeleton remodeling, certain Rho proteins seem to play a negative role in TGFβ/Smad signaling when their expression is upregulated. In epithelial cells, RhoB overexpression antagonized TGFβ for the transcriptional activation of a Smad-responsive promoter, whereas dominant-negative RhoB mutant enhanced TGFβ signaling towards this promoter . In a different study and system, it was shown that ectopic expression of RhoB, but not RhoA, caused a decrease in the expression TβRII and in the activity of the TβRII promoter in HaCaT keratinocytes and pancreatic carcinoma cells, and antagonized the TGFβ-mediated anti-proliferative responses . Downregulation of the TβRII gene by RhoB was mediated by inhibition of AP1 transcription factors that bind to an AP1 site in the proximal TβRII promoter . Rho proteins might also play a positive regulatory role in TGFβ/Smad signaling, as demonstrated by Chen et al.  who showed that ectopic expression of a dominant-negative RhoA mutant in Monc-1 neural crest stem cells blocked the phosphorylation of Smad2 and Smad3 by TβRI, their translocation to the nucleus and the activation of a Smad-specific reporter gene. Chen et al.  also showed that general inhibition of Rho activity by C3 exotoxin attenuated Smad-mediated transactivation.
The positive role of R-Smads and the negative role of the inhibitory Smad7 in Rho GTPase activation by TGFβ is discussed above. A novel, positive role for Smad7 in actin remodeling was reported by Edlund et al. . This study showed that increased expression of Smad7 in prostate cancer cells was associated with increased mobilization of the actin filament system and activation of the Rho GTPase Cdc42 (Fig. 3). It also showed that the Smad7-induced rearrangement of actin cytoskeleton required the p38 MAPK pathway previously shown to act downstream of Cdc42 . Recently, a Par6–Smurf1–RhoA pathway was shown to operate in TGFβ-induced EMT. According to this model, Par6, a regulator of epithelial cell polarity and tight junction assembly, interacts with TGFβ receptors at sites of tight junctions and is phosphorylated by TβRI, an event that leads to its interaction with the E3 ubiquitin ligase Smurf1. By an unknown mechanism, this ubiquitin ligase recruits RhoA and ubiquitinates it, thus causing its proteosomal degradation and the dissolution of the tight junctions . This mechanism was later confirmed in TGFβ-induced atrioventricular cushion endocardial cell EMT .
Finally, recent screening of miRNA microarrays for miRNAs that are up- or downregulated by TGFβ in epithelial NMuMG cells, identified miR-155 as the most significantly activated miRNA. Knockdown of miR-155 suppressed TGFβ-induced EMT and tight junction dissolution, migration and invasion of these cells . Importantly, ectopic expression of miR-155 inhibited the synthesis of RhoA. Given that miR-155 levels are frequently elevated in invasive breast cancer, the new data indicated that miRNA-based strategies could be used for the treatment of breast cancer .
Conclusions and perspectives
The balance of evidence suggests that the endocytosis of TGFβ family receptors plays an enhancing role in TGFβ family signaling. However, the magnitude and duration of the effect on the signaling output depend on the cell type. Embryonic stem cells and differentiated cells of various types are not expected to conform to the same mechanisms. Indeed, it has been reported that there are fundamental differences in the endocytic sorting of TGFβ receptors between fibroblasts and epithelial cells . Moreover, in some cell types, endocytosis of TGFβ receptors might not be interconnected with signaling, as observed in some studies. However, many questions remain. Which endocytic routes are taken by TGFβ receptor complexes and what is the contribution of each pathway to the final signal? Of the five emerging transport routes, internalization of TGFβ receptors has been reported to occur via the CCVMR and caveolar routes. There are no studies regarding the contribution and significance of the other routes. What dictates which route the receptor will follow, and more interestingly which are the effectors/regulators with which TGFβ family receptors will interact along the various endocytic routes? These questions are more or less unanswered. It is anticipated that understanding the endocytic route followed by a receptor–ligand complex will allow for a more detailed dissection of the molecular mechanisms of TGFβ family signaling and the functional consequences thereof on cell responses.
As far as Smad trafficking is concerned, although the dynamic nature of such nucleocytoplasmic shuttling has been established, critical questions remain. These concentrate on the dynamics of the movement of single Smads versus oligomeric Smad complexes. The trafficking of pools of Smads that undergo specific post-translational modifications is a major area for future research. As explained above, nuclear trafficking of Smads is only now beginning to be elucidated because the dynamics of the nuclear architecture and of chromatin interactions are now amenable to precise experimental analysis. Finally, as the complexity and depth of understanding of TGFβ signaling increase, a most critical aspect of the whole signaling pathway remains the sequence of specific steps and the establishment of the complete time-lapse history of this cascade.
Small GTPases of the Rho/Rac/Cdc42 family control the early TGFβ signaling towards actin cytoskeleton reorganization via non-Smad pathways, whereas the late cytoskeletal events seem to be directed by specific cross-talk between Smad-mediated transcriptional events involving the upregulation of Rho proteins (RhoB), GEFs (NET1) or cytoskeletal proteins (i.e. α-SMA). This may be of extreme biological significance during the premalignant to malignant transition of cancer cells, which is characterized by Rho-mediated increases in cell motility and invasiveness, or during the pathogenesis of various fibrotic diseases. Cross-talk between TGFβ/Smad signaling and the nongenomic or the transcriptional regulation of Rho GTPases is beginning to be elucidated. However, the complexity of TGFβ signaling towards Rho-governed actin cytoskeleton reorganization and cellular responses leave several exciting open questions to be addressed.
DK and CS acknowledge funding by the Greek Secretariat for Research and Technology (PENED03EΔ688) and the Research Council of the Greek Ministry of Health (KESY03KA2396). TF and CM acknowledge funding from EndoTrack FP6 Integrated Project and PENED03EΔ688 and thank Savvas Christoforidis for comments on the manuscript. AM acknowledges funding by the Ludwig Institute for Cancer Research, the Atlantic Philanthropies/Ludwig Institute for Cancer Research Clinical Discovery Program, the Swedish Cancer Society, the Swedish Research Council and the Marie Curie Research Training Network (RTN) ‘EpiPlastCarcinoma’ under the European Union FP6 program.