Introduction

Members of the transforming growth factor β (TGFβ) family play important roles in many essential cellular processes, including proliferation, differentiation, migration and apoptosis 1, 2, 3. The most prominent members of the TGFβ family are TGFβs, bone morphogenetic proteins (BMPs), activins, inhibins and nodal 4, 5, 6. Signaling is propagated via cell surface serine/threonine kinase receptors, followed by an intracellular cascade of events involving foremost Smads and their interacting partners. However, TGFβ members are also capable of inducing several alternative pathways such as those involving the small GTPases Ras, Rho, Cdc42, the phosphatases PP2A and Shc, and the mitogen-activated protein kinase (MAPK) kinase kinase TAK1 (TGFβ-activated kinase 1) 7. When TGFβ ligands reach the membrane of target cells, they bind directly to TGFβ type II receptors (TβRII), which leads to the recruitment of TGFβ type I receptors (TβRI). TβRII then trans-phosphorylates TβRI, enabling the TβRI kinase domain to act on cytoplasmic proteins and thereby propel downstream signaling actions. Receptor regulated (R-) Smads (Smad 1, 2, 3, 5 and 8) are direct targets of the ligand-receptor complex and substrates of the TβRI or other type I receptor kinases. Phosphorylated R-Smads form trimeric complexes involving two R-Smads and one common (Co-) Smad, Smad4. Once formed in the cytoplasm, these trimeric Smad complexes relocate to the nucleus in order to regulate transcription from a wide variety of promoters 8.

An important aspect of cellular signaling is the on-turning/off-turning mechanisms that regulate the strength and duration of ongoing signals. This switch can be regulated in various steps and multiple mechanisms through, for example, molecular interactions, chemical modifications, conformational changes, stability alterations, confined localization and time-dependent availability of co-factors. In recent years, research has begun to unravel the mechanisms that tune the TGFβ signaling pathway. These include interactions and modifications that regulate the stability of TGFβ's signaling machinery, receptors and Smads, and are known to be of major importance for maintaining proper physiological TGFβ signaling 9, 10, 11, 12. The key factors for regulating signaling component stability and balancing the incoming TGFβ signal are the inhibitory (I-) Smads (Smad 6 and 7)5, 9. I-Smads are induced during TGFβ signaling and negatively regulate the pathway in several steps. Initial findings showed that I-Smads competitively inhibit R-Smad phosphorylation by the type I receptors 13, 14. Further studies emphasized the capacity of I-Smads to recruit E3 ubiquitin (Ub) ligases and their co-factors to receptor complexes and to the Smads 9, 11. This feature of the I-Smads has highlighted the importance of degradation and of the Ub-proteasomal machinery in the regulation of TGFβ signaling.

Ubiquitination is a three-step enzymatic reaction where the activities of E1, E2 and E3 enzymes coordinate, in an energy-dependent manner, to covalently attach Ub moieties to lysine residues on substrate proteins 15, 16. Most E3 ligases catalyze the formation of long branches of Ub chains linked via lysine 48 (K48) and the C-terminal glycine of Ub. Proteins modified by K48-linked Ub chains are recognized by the proteasomes for proteolysis. Other post-translational modifications can regulate ubiquitination and connect protein stability to essential cellular processes. However, ubiquitination does not only lead to protein degradation. Protein mono-ubiquitination, oligo-ubiquitination and poly-ubiquitination via lysines other than K48 play roles in diverse cellular processes such as DNA repair, transcription and vesicular trafficking 16, 17. In addition, post-translational modifications such as the conjugation of small ubiquitin-like modifier (SUMO) target proteins in a similar fashion as ubiquitination and affect diverse protein characteristics 16, 18. SUMO modification is even thought to stabilize proteins because it can compete with ubiquitination for the same lysine residue in a given target protein. This versatility in protein modification has made its mark in the TGFβ field, with alternative forms of ubiquitination and sumoylation offering new insights into the molecular physiology of TGFβ signaling 8, 9.

Perturbing TGFβ signaling via mutations or deletions of TβRs or Smads has been linked to several human diseases, including cancer, fibrosis, vascular disorders and autoimmune diseases 3, 19. Many of the mutations lead to unstable protein products that are rapidly degraded after ubiquitination and shift the equilibrium of the signaling chain. This demonstrates the importance of having a controlled equilibrium in which degradation and stability of components of the TGFβ/Smad signaling pathway are finely regulated.

Regulation of TGFβ receptor stability

An early step of TGFβ signaling is the ligand-induced formation of tetrameric receptor complexes, each comprising two TβRII units and two TβRI units. Activated receptor complexes are internalized via either clathrin-coated pits or clathrin-independent caveolin-1-positive lipid rafts 20, 21. These two routes represent two distinct outcomes (Figure 1). Clathrin-mediated endocytosis carries the receptors to early endosomes, promotes sustained signaling and allows re-cycling of receptors back to the cell surface. Internalization via caveolae, on the other hand, will send the receptors to the ubiquitination machinery en route to lysosomes. It has been postulated that between these two endocytic paths rests an equilibrium, which can be shifted towards either a signaling or a degradation dominating position (Figure 1). TGFβ stimulation is not thought to affect the receptor internalization pattern towards any of these two endocytic pathways, and specific factors that determine the choice by the receptors remain poorly understood.

Figure 1
figure 1

Regulation of TGFβ receptor trafficking by ubiquitination and sumoylation. Effects of ubiquitination or sumoylation on receptor (type I, RI and type II, RII) biosynthesis and trafficking to the plasma membrane remain unknown (left side, question mark). Activated TGFβ receptors internalize to early endosomes and signal positively, leading to nuclear Smad accumulation. Type I receptor sumoylation (S) via the E2 enzyme Ubc9 promotes Smad-dependent signaling. Activated receptors also internalize through lipid raft membrane domains; and the inhibitory Smad7 (I-Smad) acts on this pathway. The serine/threonine kinase SIK1, the adaptor STRAP, the chaperone Hsp90 and several E2 and E3 ubiquitin ligases control the overall process of receptor downregulation. The deubiquitinase Uch37 can, via Smad7, reverse the ubiquitination of the type I receptor and promote positive signaling.

Full size image

Lipid rafts are important membrane domains that define the negative feedback loop of TGFβ signaling. The rafts are upon signaling occupied by TGFβ-induced I-Smads and E3 Ub ligases, which results in specific receptor degradation 9, 11, 20, 21. BMP signaling induces Smad6 and the HECT (homologous to E6-AP carboxyl terminus)-domain E3 Ub ligase Smad ubiquitylation regulatory factor 1 (Smurf1), whereas TGFβ1 induces Smad7 and either Smurf1 or Smurf2 22, 23, 24. The induction of these proteins by TGFβ serves the purpose of sustaining and increasing their cellular levels so that their inhibitory action can become effective when signaling progresses beyond a threshold. Alternatively, I-Smad induction could serve the purpose of building up their levels after their degradation, so that the system is ready for a possible subsequent signaling start.

I-Smads and Smurfs interact inside the nucleus via the proline-tyrosine (PY) motif of I-Smads and the WW domains of Smurfs, but evade out to the cytoplasm as signaling begins 25, 26. The C2 domain of Smurf2 can bind near the catalytic cysteine in its own HECT domain, which disturbs formation of the Ub thioester and blocks ubiquitination activity 27. This self-inhibition is relieved when Smurf2 binds to Smad7. The C2 domain of Smurfs also tethers the I-Smad/Smurf complex to lipid rafts at cell membranes where binding and ubiquitination of receptors are executed 28. In this complex, the N-terminal part of Smad7 facilitates Smurf2 activity by helping recruit the E2 conjugating enzyme UbcH7 29. Recently, it was shown that the heat shock protein 90 (Hsp90) is a modifier of Smurf2-mediated receptor ubiquitination, as inhibiting Hsp90 activity led to an increased receptor ubiquitination 30. This effect is due to the ability of Hsp90 to bind and chaperone the receptors. Earlier work has also established that the adaptor serine-threonine kinase receptor-associated protein (STRAP) stabilizes Smad7 and recruits it to the activated type I and type II receptor complex 31. STRAP contains WD40 repeats and associates with both TGFβ receptors and Smad7, thus promoting the inhibitory action of this I-Smad. However, a role of STRAP in the process of TGFβ receptor ubiquitination via Smurf E3 ligases has not yet been demonstrated.

In addition to the Smurfs, two other C2-WW-HECT domain E3 Ub ligases can bind to Smad7 and regulate receptor degradation: WW domain-containing protein 1 (WWP1) and neural precursor cell expressed, developmentally down-regulated 4-2 (NEDD4-2) 32, 33. Both proteins are structurally similar to the Smurfs and also work similarly to promote receptor downregulation, which results in decreased phosphorylation of R-Smads and negative regulation of TGFβ-mediated transcription. Another HECT domain E3 ligase implicated in TGFβ receptor downregulation is atrophin 1-interacting protein 4 (AIP4/Itch), which binds to Smad7 and enhances the recruitment of Smad7 to TβRI. This mechanism does not require the catalytic, ubiquitination activity of AIP4/Itch 34. However, at least under conditions of overexpression in epithelial cells, AIP4/Itch leads to Smad7 protein ubiquitination and degradation. Thus, AIP4/Itch may coordinate multiple steps during early TGFβ receptor internalization and signaling as discussed again below in the Smad section.

The I-Smad/Smurf ubiquitination pathway can be regulated by additional factors. Salt-inducible kinase (SIK) was recently shown to participate in the negative feedback loop of TGFβ signaling that controls receptor degradation 35. The SIK serine/threonine kinase is induced by incoming TGFβ signaling and enhances, via its Ub-associated (UBA) domain, the degradation of TβRI in a K48-linked ubiquitination-dependent manner. This is achieved in cooperation with Smad7, suggesting that SIK may enhance the turnover of ubiquitinated TβRI/Smad7 complexes (Figure 1). The kinase activity of SIK was shown to be important for proper TβRI degradation, but the direct substrate of this kinase remains to be determined. Smad7 also binds to the deubiquitinating enzyme UCH37, which can reverse the ubiquitination of TβRI, leading to its stabilization and increased TGFβ signaling 36. This interaction provides a balancing step against the ubiquitinating E3 ligases that act on the receptor (Figure 1). Another factor that regulates the degradation of TGFβ signaling components is Dapper2 37, 38. Dapper2 was shown to promote lysosomal degradation of TβRI and nodal type I receptor in zebrafish, resulting in repressed nodal signaling and inhibition of mesoderm formation 38. Dapper2 was further shown to downregulate TβRI signaling in mammalian cells 37. Whether Dapper2 cooperates with the I-Smad/Smurf machinery remains to be explored.

TβRI can also be sumoylated by Ubc9 39. SUMO is a Ub-like molecule that is attached to the lysines of substrate proteins by specific SUMO-ligating enzymes. TβRI sumoylation enhances recruitment and phosphorylation of Smad3. It remains to be investigated whether ubiquitination and stability of TβRI are regulated by sumoylation. Identification of the specific E3 ligase that sumoylates TβRI might shed light on this issue.

In summary, it is clear that the TGFβ receptors are ubiquitinated and degraded through the action of several co-operating protein complexes containing E3 ligases as well as other important regulators of protein degradation (Figure 1). The I-Smads seem to work as master regulators in many of these complexes, orchestrating both ubiquitination and de-ubiquitination. It is also clear that degradation of the receptors serves as an important regulatory mechanism for TGFβ signaling both in cell culture studies and in in vivo models. However, whether the receptors are degraded via the lysosomes, proteasomes or via both of these pathways is still an open question. Experimental data support that inhibition of both lysosomal and proteasomal activities can have effects on TGFβ receptor degradation 24, 35. This would imply a possible joint venture from both organelles to achieve complete receptor degradation.

Regulation of R-Smad stability

Activated TβRI phosphorylates intracellular R-Smad proteins. The stability and amounts of R-Smads are regulated by the Ub-proteasome machinery both at steady state and throughout the TGFβ signaling cycle (Figure 2). Regulation of degradation is an important measure to assure proper signaling strength and outcome. Although Smurfs are mainly transcriptionally induced by TGFβ signaling, their small constant pool in cells may have some effect on steady-state Smad stability as Smurfs can bind non-activated Smads. Smurf1 binds to the PY motif of Smad1 and Smad5 and induces their ubiquitination 40. Smurf2 can ubiquitinate Smad1 and Smad2 at steady state 41, 42. Smurf2 also binds Smad3, but without affecting its ubiquitination or stability. Although Smad3 is not a Smurf target, it can be stabilized at steady state by a proteasome inhibitor 43. This indicates that Smad3 is being degraded via ubiquitination under non-signaling conditions, which may help ensure correct amounts of Smad3 at the start of signaling. An E3 Ub ligase that can regulate basal Smad3 levels is the U-box-containing carboxyl terminus of Hsc70-interacting protein (CHIP) 44. CHIP was reported to ubiquitinate Smad3 for degradation at steady state and thus controls cellular sensitivity to TGFβ. In addition, CHIP can also bind and ubiquitinate Smad1 45. The scaffolding protein axin and its associated kinase glycogen synthase kinase 3-β (GSK3β) were also shown to participate in Smad3 basal stability 46. Axin/GSK3β mediates the phosphorylation of Smad3 at Thr66, which targets Smad3 for Ub-mediated degradation, but the E3 ligase involved is not yet fully defined. Additional factors that are involved in R-Smad steady-state degradation are yet to be identified. Research in this direction may help explain cell-type specific TGFβ responses and could provide interesting links to TGFβ-associated diseases.

Figure 2
figure 2

Regulation of R- and Co-Smad stability. The latent TGFβ with inactive TGFβ receptors (type I, RI and type II, RII) as well as inactive R-Smad and Co-Smad proteins are shown on the left side. Mature TGFβ binds to the receptor complex and activates phosphorylation of R-Smads by the type I receptor. Phosphorylated R-Smads bind to Co-Smad, followed by their nuclear accumulation and recruitment to specific chromatin loci on gene promoters/enhancers, which leads to mRNA synthesis. Smad complexes bound to chromatin exist always in association with additional sequence-specific transcription factors (TF). An unknown mechanism (question mark) dissociates the chromatin-bound Smad protein complex and the Smads recycle to the cytoplasm after being de-phosphorylated by nuclear phosphatases (like PPM1A). Different E3 ubiquitin ligases are shown in groups based on their mode of action towards specific Smad substrates. The ubiquitination and degradation of R-Smads or Co-Smad in the absence of signaling are not clearly established (left side, question marks). However, kinases like GSK3β phosphorylate R-Smads in the linker or MH1 domain, leading to their ubiquitination at steady state. CHIP, Jab1 and SCF ligases target the Co-Smad, whereas Smurfs and the related WWP1, NEDD4-2 target the Co-Smad with the help of the adaptor protein Smad7 (I-Smad). The HECT domain E3 ligase AIP4/Itch ubiquitinates Smad2, which promotes its phosphorylation by the type I receptor (RI). In the nucleus, SUMO ligases (PIAS and Ubc9) sumoylate (S) Smad3 and Smad4, leading to specific gene target regulation. Nuclear Smad4 can also be mono-ubiquitinated (U) by a yet unknown E3 ligase (question mark). Arkadia and SCF/Roc1 target receptor-phosphorylated R-Smads for degradation, while nuclear kinases (GSK3β, ERK1/2 and CDKs) phosphorylate the linker of nuclear R-Smads, leading to the recruitment of HECT-domain E3 ligases (Smurfs, WWP1, NEDD4-2) and R-Smad degradation. Whether the linker phosphorylation and R-Smad ubiquitination follow after R-Smad C-terminal dephosphorylation remains unknown (question mark). Nuclear ectodermin poly-ubiquitinates and degrades Smad4, whereas mono-ubiquitinated Smad4 might also escape degradation and be exported to the cytoplasm. All inhibitory arrows represent poly-ubiquitination followed by proteasomal degradation.

Full size image

Upon ligand-induced signaling, the R-Smads become phosphorylated at their C-terminal SXS-motifs by the type I receptor. Phosphorylated R-Smads bind to the Co-Smad and accumulate into the nucleus, where they will bind to chromatin 8. The Smurf protein levels elevate inside the cell upon signaling. BMP-induced Smurf1 can, as in the steady-state condition, ubiquitinate and degrade Smad1 and Smad5 40. In the case of Smad1, it is known that after phosphorylation at the C-terminal motif by the BMP type I receptors, sequential phosphorylations follow in the linker region of Smad1 by MAPK and GSK3β 47, 48. These linker phosphorylations lead to poly-ubiquitination and degradation of activated Smad1 via Smurf1 (Figure 2).

Phosphorylated Smad2 can be degraded via the proteasomes, as treatment with proteasome inhibitors could maintain higher phospho-Smad2 levels and showed better effects than treatment with phosphatase inhibitors 49. Upon signaling, Smurf2 shows enhanced interaction with and ubiquitination of Smad2 41, 50. On the other hand, the HECT domain E3 ligase AIP4/Itch that facilitates TβRI-Smad7 recruitment to negatively regulate TGFβ signaling (see the discussion before, and Ref. 34) can also positively influence this pathway due to its ability to promote TβRI-Smad2 association, ubiquitination of Smad2 and receptor-mediated phosphorylation of Smad2 51. On the basis of this mechanism, AIP4/Itch ubiquitinates Smad2 in a manner that promotes its C-terminal phosphorylation by the type I receptor. This mechanism explains why cells from mice lacking AIP4/Itch exhibit deficient TGFβ/Smad signaling, and emphasizes the complexity and plurality of the functional roles of ubiquitination in TGFβ signaling. Obviously, the AIP4/Itch-mediated regulation of Smad2 versus Smad7 is physically coupled with the association of corresponding protein complexes with the TGFβ type I receptor, and the detailed post-translational modifications that lead to one or the other ubiquitination pathway remain to be clarified. Similar to AIP4/Itch, the RING domain E3 ligase Cbl-b seems to be also required for Smad2 ubiquitination and proper C-terminal phosphorylation by TβRI, which explains why T lymphocytes from Cbl-b knockout mice are resistant to growth suppression and differentiation induced by TGFβ 52. The possibility for a functional link between Cbl-like E3 ligases and AIP4/Itch is open and might shed new light on the mechanism of positive regulation of TGFβ/Smad signaling by ubiquitination.

Smad3 is not directly affected by the Smurfs even though it contains the PY motif and binds Smurfs. Instead, the SCF/Roc1 E3 Ub ligase complex controls degradation of activated phospho-Smad3 53. SCF/Roc1 is suggested to terminate Smad3 nuclear signaling by ubiquitinating Smad3 and favoring its export to the cytoplasm for proteasomal degradation. It has also been shown that the nuclear RING domain E3 ligase Arkadia can ubiquitinate and degrade phosphorylated Smad2 and Smad3 54. The Arkadia-dependent ubiquitination step occurs after Smad transcriptional activation, which couples the degradation to signaling termination. But Arkadia is also suggested to be crucial for the ability of R-Smads to activate promoter transcription, as loss of Arkadia in embryonic cells showed accumulated hypoactive phospho-Smad2/3, and Arkadia (−/−) chimeras showed perturbed formation of foregut and prechordal plate 54.

In addition to being ubiquitinated, Smad3 has also been shown to be sumoylated by the sumoylation-specific E3 ligase, protein inhibitor of activated STAT (PIASy) 55, 56, 57. However, this Ub-like modification was reported to repress Smad transcriptional activity by inhibiting DNA-binding and stimulating nuclear export of Smad3 rather than affecting its degradation. Whether there are links between ubiquitination and sumoylation of Smad3 remains unclear.

In summary, it is very intriguing how all the E3 ligases and their co-factors work and possibly co-operate to regulate R-Smad stability (Figure 2). Studies that could sort out exactly how, when and where the different ubiquitination pathways start and interact would be of great importance for understanding the mechanisms that finely tune TGFβ signaling.

Regulation of Co-Smad stability

Phosphorylated R-Smads bind to the Co-Smad, Smad4, in a trimeric complex containing two R-Smads and Smad4. These complexes are mainly nuclear and active on specific promoter sites, where they bind DNA and recruit co-activators or co-repressors 8. Stability and degradation of Smad4 are under the influence of many different E3 ligases (Figure 2). Jab1 can interact with Smad4 and induce its ubiquitination and proteasomal degradation 58. SCFbTrCP1 is another E3 Ub ligase that has been shown to bind and catalyze ubiquitination of Smad4 59. Smad4 can also be degraded via CHIP-mediated ubiquitination together with Smad1, resulting in suppressed BMP signaling 45. The RING-type E3 Ub ligase ectodermin/TIF1γ also targets Smad4 for poly-ubiquitination and degradation, thereby restricting TGFβ, BMP and activin signaling 60. Ectodermin/TIF1γ was shown to be crucial for the development of ectoderm in Xenopus embryos through counteracting and balancing the activin-induced mesoderm formation. Furthermore, Smad7 can act as a bridge to support ubiquitination and degradation of Smad4 by recruiting C2-WW-HECT domain E3 ligases, such as Smurf1, Smurf2, WWP1 and NEDD4-2 61. Smad4 mutations that lead to instability have been reported in many different human cancers (see the discussion later). The E3 ligase SCFSkp2 has been strongly implicated to mediate the degradation of these mutants 62.

In addition to being poly-ubiquitinated and degraded by proteasomes, Smad4 has also been shown to be mono-ubiquitinated and sumoylated 57, 63, 64, 65, 66. Mono-ubiquitination of Smad4 occurs on Lys-507 in the MH2 domain and leads to improved transcriptional TGFβ signaling 63, but also promotes export of Smad4 to the cytoplasm 67. In addition to the so-called “universally conserved Smad lysine”, which corresponds to Lys-507 in Smad4, Lys-519 is a unique lysine residue in Smad4 and is well-conserved across species as suggested by recent phylogenetic evidence 68. Whether Lys-519 can also act as an acceptor site for mono-ubiquitination is currently unknown but remains possible as mutation of Lys-507 indicated the presence of a weak secondary mono-ubiquitination site in the Smad4 MH2 domain 63. Both Smad4 mono-ubiquitination by as yet uncharacterized E3 ligases and Smad3 poly-ubiquitination by SCF/Roc1 can be stimulated by the nuclear co-activator and acetyl-transferase p300 53, 67. The prevalence of mono-ubiquitination of wild-type Smad4 under physiological conditions may suggest the existence of a deubiquitinase that could remove the poly-Ub chains and that ensures the stable attachment of single Ub moiety to this protein. It is also possible that a deubiquitinase might act to remove the mono-Ub modification from Smad4, thus restoring the basal state of this protein.

Sumoylation targets Lys-113 and Lys-159 in Smad4 and is catalyzed by the E2 enzyme Ubc9 and members of the E3 SUMO ligase family PIAS 64, 65, 66, 69, 70, 71. Sumoylation of Smad4 seems to play dual roles during TGFβ signaling. On one hand, sumoylation leads to stabilization and nuclear accumulation of Smad4, which enhances TGFβ signaling 64, 65, 66, 71. On the other hand, sumoylation of Smad4 leads to repressed transcription 66, 70, 72. The total effects of sumoylation on TGFβ signaling are therefore complex and involve steps at receptor, R-Smad and Co-Smad levels.

In summary, R-Smads and Co-Smad are ubiquitinated and regulated by a large number of E3 ligases (Figure 2). This could indicate that regulation of ubiquitination is a very precise action. The E3 Ub ligases may work either alone or in sequential steps to give rise to accurate amounts of cellular Smads and a balanced pace of degradation. The various unique E3 ligases may also indicate that there can be an even greater diversity of differential ubiquitination patterns on the Smads than we are aware of today. This still remains to be explored. It is also possible that several E3 ligases with redundant tasks/functions all work individually to secure ubiquitination and degradation of the Smads.

Regulation of stability of I-Smads and other signaling targets

The I-Smads, Smad6 and Smad7, participate in the negative feedback mechanism of all known signaling pathways of the TGFβ superfamily 9. Although the I-Smads are reported to act at multiple points of the pathway, that is, at the receptor level, R-Smad and Co-Smad levels and even at the nuclear Smad complex level, it is not yet clear whether all of their actions are mediated by their ability to act as adaptors that carry Ub ligases to their target proteins, or also via complementary mechanisms. The best-studied I-Smad in this respect is Smad7, which acts not only as a mediator of ubiquitination and degradation of both the TGFβ receptors and the Co-Smad (Figures 1 and 2) 23, 24, 61, but also as a carrier of phosphatases or a simple disruptor of protein interactions between R-Smads and Co-Smad or between R-Smads and type I receptors 9. Furthermore, Smad7 is well established as a target of both Ub ligases and acetyl-transferases like p300/CBP, but the impact of such regulation of Smad7 on the receptor/R-Smad/Co-Smad pathway remains poorly understood 23, 24, 73, 74.

Here we will only highlight a few specific regulatory scenarios that involve Smad7 (Figure 3) beyond the level of receptor ubiquitination, which was discussed before. Smurf1 and Smurf2 bind to the PY motif in the Smad7 linker region and ubiquitinate Smad7, and this action may be direct, as shown by in vitro ubiquitination experiments 23, 24, 73. Alternatively, the discovery that Arkadia plays a primary role in Smad7 poly-ubiquitination and proteasomal degradation 74, 75 is compatible with a model whereby Smurfs bound to Smad7 might ubiquitinate another protein that eventually activates Arkadia, which leads to the degradation of Smad7. Whether other Smad7-bound E3 ligases such as WWP1 and NEDD4-2 participate in a similar regulatory loop remains to be clarified 32. In addition, Smad7 is acetylated by p300/CBP at the same lysine residue that alternatively can be ubiquitinated, thus leading to stabilization of Smad7 73. Acetylated Smad7 can then be targeted by histone deacetylases (HDACs), which deacetylate Smad7, allowing ubiquitination and promoting its proteasomal degradation 76. It is important to clarify the actual location in the cell where these protein complexes form, and exactly where Smad7 is acetylated, de-acetylated and finally degraded, that is, during receptor trafficking in the cytoplasm, in the nucleus or in both places. The emerging evidence that nuclear Smad7 directly antagonizes R-Smad/Co-Smad binding to DNA and inhibits transcriptional signaling by the latter proteins 77, 78 suggests it is highly possible that nuclear Smurfs or Arkadia may target chromatin-bound Smad7, thus releasing negative regulation from the Smad complex and permitting productive transcription, as we discuss below for the Ski/SnoN co-repressors (Figure 3).

Figure 3
figure 3

Negative TGFβ regulators in the nucleus. Three different regulatory scenarios are shown, which may be either linked or operating independently from each other. On the left-hand side, Ski/SnoN/Co-Smad complexes on chromatin maintain a default repressed state of a target gene of TGFβ. The incoming receptor-phosphorylated R-Smad brings along specific E3 ubiquitin ligases (APC/Cdh1, Smurf2 or Arkadia), all leading to ubiquitination and degradation of Ski and SnoN. Then R-Smads replace Ski/SnoN in the protein complex and stimulate transcription (mRNA). In the middle, a positive Smad transcriptional complex can be the target of inhibition by nuclear Smad7 (I-Smad), which displaces R-Smads as it competes for DNA-binding to the same DNA element. The I-Smad-bound chromatin complex is now transcriptionally repressed. On the right-hand side, nuclear E3 ligases (Smurfs and Arkadia) may also lead to degradation of the chromatin-bound I-Smad (mechanism not yet demonstrated, possibly activated by the incoming phosphorylated R-Smad, question mark), thus favoring formation of the positive transcriptional complex of R-Smad and Co-Smad. In all these scenarios the Co-Smad is shown as being constantly bound to chromatin, which may not be true, as the Co-Smad is also a dynamic and shuttling protein. Smad complexes bound to chromatin exist always in association with additional sequence-specific transcription factors (TF). All inhibitory arrows represent poly-ubiquitination followed by proteasomal degradation.

Full size image

In contrast to the extensive studies on Smad7, little is known about the role of Smad6 ubiquitination in TGFβ or BMP signal transduction. However, it has been shown that Smad6 plays an adaptor role by bridging Smurf1 and the transcription factor Runx2, thus leading to its ubiquitination and final proteasomal degradation 79. This mechanism is important for the fine tuning of BMP-induced bone differentiation. Smad6 also interacts with HDACs, and the recruitment of HDACs seems to be important for Smad6-mediated repression of transcription 80. In addition, Smad6 is di-methylated by the nuclear protein N-methyltransferase (PRMT) 1 on Arg-74, a modification that is possibly linked to the transcriptional activity of Smad6 and its roles in mediating protein ubiquitination 81. However, the latter hypothesis about the function of Smad6 dimethylation remains to be tested.

Another important mechanism that regulates protein stability during TGFβ signaling is R-Smad-dependent ubiquitination and proteasomal degradation of various proteins. In other words, R-Smads, like I-Smads, act as adaptors that carry the Ub ligases, such as Smurfs or Arkadia, to the target protein. Whether ectodermin/TIF1γ can also reach alternative substrates via its interaction with Smad4 and phosphorylated Smad2 remains an interesting open question 82. The best understood targets of R-Smad-mediated ubiquitination are the nuclear co-repressors Ski and SnoN, which bind to chromatin via their interaction with Smad4 and phosphorylated R-Smads 50, 83, 84, 85, 86, 87, 88. The current prevailing model favors a chromatin-tethered Ski/SnoN/Smad4 complex that represses specific target genes of TGFβ signaling (Figure 3). After phosphorylation by receptors, R-Smads enter the nucleus; they associate with Ub ligases such as the anaphase promoting complex (APC, which is best known for regulation of cell cycle progression), Smurf2 or Arkadia, and the concerted action of these Ub ligases leads to ubiquitination and proteasomal degradation of Ski or SnoN 50, 83, 84, 86, 87, 88. This leaves the target chromatin locus free from the repressors and allows the incoming R-Smads to establish positive regulation of RNA polymerase II and mRNA synthesis. It is interesting that in this regulatory node, Smad2 binds activated Smurf2 while Smad3 binds to the regulatory subunit of APC, Cdh1. The reason for utilizing diverse Ub ligases to achieve the de-repression by Ski/SnoN is a very interesting issue and not yet fully understood. Once again, we propose that this suggests a step-wise mechanism, whereby each ubiquitination step gradually modulates the activity of Ski/SnoN and also of additional regulators in the cognate chromatin-bound complexes, thus providing time-dependent and possibly dose-dependent control of gene expression downstream of TGFβ. It is also interesting to speculate that the Smad7-dependent repression on chromatin and the Ski/SnoN-dependent repression model may be physically and functionally linked, as they may represent a “single” multi-protein complex whose components are sequentially modified by post-translational modifications, such as ubiquitination (Figure 3).

The action of R-Smads as adaptors that carry Ub ligases to target proteins is not unique to the regulation of the Ski/SnoN co-repressors. In fact, this has been a rather early observation in the field, as Smad3 mediates proteasomal degradation of important proteins such as the multiadaptor protein human enhancer of filamentation (HEF1), which is ubiquitinated by the APC/Cdh1 and AIP4/Itch ligases, and the structural protein ELF, which is ubiquitinated by the PRAJA Ub ligase 10, 89. It is likely that there are other target proteins that are regulated by Smads based on their function as carriers and activators of Ub ligases. Whether Smads mediate a similar mode of regulation based on sumoylation or other enzymatic processes is open for future investigations.

Relevance of stability regulation of TGFβ pathway components in diseases

The previous discussion has introduced the complexity of negative regulatory mechanisms that impinge on several layers of the TGFβ signaling pathway and provide means for tight and timely control of the flow of molecular information all the way from the plasma membrane to the nucleus. Deregulation of any of the above-mentioned proteins impacts on either normal embryonic development or, alternatively, on a pathogenic process of human disease. As already discussed above, the first demonstration of ubiquitination as a regulatory mechanism of TGFβ pathways came from studies of Smad4 mutants in human tumors 90, 91.

The best evidence that perturbations of the stability of proteins of the TGFβ pathways can be linked to human disease came from studies of tumors. We will therefore discuss mainly cancer here, but will also briefly outline examples from other diseases. TGFβ plays complex roles in cancer progression by acting as a tumor suppressor in normal epithelial tissue homeostasis and during early benign adenoma stages, while acting as a pro-tumorigenic factor that promotes immune evasion, local invasiveness and metastasis at the late stages of tumor progression 2, 3. The influence of negative regulators of this pathway on cancer progression is therefore dependent on tumor type and disease stage. The tumor-suppressive action of TGFβ includes its ability to cause cell cycle arrest and promote apoptosis of normal epithelial cells or benign adenomas 9, 12. In human cancers, the core components of the pathway (i.e. receptors or Smad proteins) can become inactivated via germ-line or somatic mutations. Misexpression or aberrant activity of regulators of pathway component stability can lead to the same functional outcome, namely pathway inactivation 3. Thus, abnormally high expression of a Ub ligase that negatively regulates Smad signaling would lead to TGFβ pathway inactivation in a certain tumor. This indeed has been documented for E3 ligases such as Smurf1, Smurf2, WWP1, NEDD4-2 and ectodermin/TIF1γ 12. More specifically, in pancreatic cancer, the genomic locus encompassing the Smurf1 gene is frequently amplified and Smurf1 has been proposed as the major gene in this locus that may act oncogenically 92. In esophageal squamous cell carcinomas Smurf2 is abnormally expressed, and the higher its level the worse the prognosis of these patients 93. The WWP1 gene is amplified to high copy numbers in prostate and breast cancers, leading to inactivation of TGFβ signaling and oncogenic transformation 94, 95. A similar scenario has been reported for NEDD4-2 in prostate and bladder cancers 96. Ectodermin/TIF-1γ, the Ub ligase that specifically targets Smad4, is misexpressed in colorectal and breast cancers. RNAi-mediated depletion of this protein from either colon or breast cancer cell lines that express endogenous levels of Smad4 leads to suppression of proliferation of these cells, suggesting restoration of TGFβ's tumor suppressor action 60. In contrast, depletion of ectodermin from tumor cell lines that do not express Smad4 has no impact on the growth of such cells, enforcing the notion that this specific regulator targets primarily Smad4 and not other components of the TGFβ pathway. Finally, in addition to these Ub ligases, adaptor proteins like STRAP, the stabilizer of Smad7-TGFβ receptor complexes, can also be misexpressed in human colon and lung cancers 97. The oncogenic role of STRAP though, in addition to acting as a TGFβ receptor inhibitor, can be due to its ability to interact with nuclear oncogenes like the Ewing sarcoma (EWS) oncoprotein 98.

Despite the fact that the above examples give the simple picture that misexpressed Ub ligases lead to oncogenic phenotypes by inactivating normal TGFβ pathways, the true situation is much more complex. Smurf1 and Smurf2 have been reported to play metastasis-promoting or tumor-suppressing roles in cancer cells. For example, Smurf1 acts as a Ub ligase for the small GTPase RhoA, which regulates actin dynamics at the leading edge of migratory cells, and this activity plays important roles in tumor cell invasiveness and metastatic potential 99, 100. On the other hand, Smurf2, which shows abnormally elevated levels in various tumor cells, arrests their proliferation due to induction of senescence, a mechanism that seems to be independent from the Ub ligase activity of Smurf2 101. Thus, distinct cellular substrates of the Smurfs, or specific regulation of their activity by as yet unknown regulatory proteins, may lead to dual roles in cancer progression that may depend either on the tissue type or on the stage of tumor progression.

This complexity of the roles of Ub ligases that target the TGFβ pathway in cancer and the distinct roles these proteins may play in normal versus tumor cells deserves further discussion. The best examples, as mentioned before, come from Smad2 and Smad4. The RING-domain ligase Arkadia has been proposed to degrade specifically phosphorylated Smad2 in the nucleus 54. However, the specific E3 ligases that mediate regulatory ubiquitination or degradative poly-ubiquitination of Smad2 and Smad4 in normal, non-cancer cells remain unknown 49, 63. There is more information regarding the poly-ubiquitination and proteasomal degradation of mutant versions of Smad2 and Smad4 that are abundant in various human tumors. Mutations in the N-terminal MH1 domain probably affect the critical folding of this domain, causing general instability of the protein and Ub-mediated proteasomal degradation 90, 91. Such mutations are R133C in Smad2 and L43S, G65V, R100T and P130S in Smad4. Among these mutant Smads, G65V and R100T Smad4 were shown to be targeted effectively by the SCF Ub ligase, which includes the F-box subunits Skp2 or βTrCP-1 for proteasomal degradation, whereas wild-type Smad4 was not targeted by this enzyme 59, 62. The ligases affecting other Smad4 or Smad2 mutants remain to be elucidated. Furthermore, mutations in the C-terminal MH2 domain of Smad2 and Smad4 can also lead to destabilization, poly-ubiquitination and degradation of these proteins in tumor cells 102, 103, 104. Such mutants include Q407R and L369R in Smad2 or a C-terminal deletion of 38 amino acid residues in Smad4. The mechanism of preferential degradation of these mutants has not yet been clarified, as many of them also interfere with the stability of their wild-type counterparts. It is possible that such mutant Smads activate specific Ub ligases that lead to both their own degradation and also the degradation of other proteins found in the common complexes.

Control of the TGFβ pathway by ubiquitination is a complex problem, as the Ub ligases that target this pathway can regulate both Smads and negative regulators of Smads. In other words, an E3 ligase that degrades a Smad tumor suppressor protein has the properties of an oncoprotein. However, if the same E3 ligase also degrades a Smad repressor, it could have tumor suppressor properties. Examples of such Smad repressors are Ski and SnoN, as described in the previous section. According to the current model, Ski and SnoN seem to play dual roles in cancer progression, despite the fact that previous knowledge favored an oncogenic role for these proteins 88, 105, 106, 107. The oncogenic potential of Ski and SnoN is known to be mediated by their ability to bind directly to nuclear Smad complexes and repress their transcriptional activities 85, 108, 109, 110, 111. Thus, when the synergistic activities of the APC, Smurf2 and Arkadia Ub ligases lead to poly-ubiquitination and degradation of Ski and SnoN, the tumor suppressor action of the TGFβ/Smad pathway is enhanced 50, 83, 84, 86, 87, 88. But Smurf2 and Arkadia can also target Smad2 and Smad4 for degradation, thus removing two central components of the tumor suppressor pathway of TGFβ 41, 42, 54, 61. How can we therefore build a comprehensive model for the role of these Ub enzymes in cancer progression? Possibly by studying the specific timing of pathway regulation by each one of these enzymes and their adaptor proteins like Smad7. Smad7, which regulates both nuclear Smad signaling 77, 78 and TGFβ receptor signaling in the cytoplasm 23, 24, 74, 75, is a key factor. Whether or not Smad7 exists in nuclear complexes together with Ski and SnoN remains currently unknown. It is also possible that each individual ubiquitination step is not directly associated with proteasomal degradation of the ubiquitinated protein. For example, Smad ubiquitination may regulate intermediate steps in the signaling pathway and thus be integrated with the action of other enzymatic regulators (e.g. nuclear acetyl-transferases, de-acetylases, phosphatases, kinases), ultimately leading to the transport of Smads to the centrosome, where their proteasomal degradation has been recently observed 47.

We will close this section by briefly highlighting evidence for the involvement of negative regulators of TGFβ signaling in fibrotic disease. Most of this evidence is based on the pivotal role the inhibitory Smad7 plays in various types of chronic inflammatory and fibrotic conditions such as inflammatory bowel disease, Crohn's syndrome, scleroderma, chronic respiratory inflammation and asthma 112. Scleroderma is a chronic fibrotic disease of the skin, and dermal fibroblasts from such patients exhibit elevated levels of the inhibitory Smad7 113. The normal circuitry of Smurf-dependent TGFβ receptor downregulation seems defective in such cells. The net result of this abnormal pathway is sustained and chronic autocrine activation of TGFβ signaling, which was previously known to contribute to the pathogenesis of scleroderma. Similar to fibrotic skin, inflammatory bowel disease is also characterized by abnormally high Smad7 levels, which perturbs TGFβ-mediated negative regulation of inflammation 114. The high Smad7 levels in pro-inflammatory immune cells from such patients are due to defective ubiquitination and degradation of Smad7 that correlates with its enhanced acetylation by p300. In a mouse model of kidney fibrosis the inverse scenario prevails, namely Smad7 levels are very low while Smurf1 and Smurf2 levels and ubiquitination activities are high, which possibly plays causative roles in the progression of renal fibrosis caused by ureteral obstruction 115. More recent evidence points to Arkadia as the primary mediator of Smad7 downregulation in the fibrotic kidney, a mechanism that also leads to the process of epithelial-mesenchymal transition of kidney epithelial cells to contractile myofibroblast-like cells 116, 117. In the same mouse model, elevated levels of Smurf2 directly led to SnoN loss, which promotes chronic responsiveness to TGFβ and stronger autocrine TGFβ signaling 118. In human fibrotic kidneys, high Smurf2 expression best correlates with very low Ski and SnoN expression 119. Similar to the kidney, progression of experimental liver fibrosis correlates well with increased levels of Smurf2 and decreased levels of SnoN 120. It is therefore evident that some of the central mechanisms by which Ub ligases regulate the action of the TGFβ pathway seem to play prominent roles in diverse fibrotic conditions.

Concluding remarks

In this article we focused on the regulation of protein stability of critical components of the TGFβ signaling pathways. We attempted to provide a comprehensive view of the current state of the art, while also pinpointing the many open caveats in our current understanding. This fact emphasizes the need for a more careful examination of the various mechanisms of receptor, Smad and co-repressor ubiquitination/sumoylation under physiological conditions. Explaining the timing of events and the specific contribution of each Ub ligase will be of major importance. New and more sensitive reagents are needed to achieve this goal. In addition, it is important to decipher carefully the subcellular location where the various post-translational modifications and the ultimate proteolysis take place. Equally important is to integrate the various activities of the Ub-based regulators to those of other enzymes such as the kinases, phosphatases, acetyl-transferases and de-acetylases, which modulate the functions of the same target proteins. This will finally lead to understanding the reasons why sometimes ubiquitination and sumoylation play positive and sometimes negative regulatory roles in TGFβ signaling. Although the action of certain negative regulators is already linked to human disease, we anticipate that future deeper knowledge might better explain which of the signaling proteins play tumor suppressor roles and which promote tumor progression. As all of these regulators carry enzymatic activities, a combinatorial use of drugs that target these enzymes may allow beneficial therapeutic regimes once the logic of action of these enzymes is finally understood.