During the course of breast cancer progression, normally dormant tumour‐promoting effects of transforming growth factor β (TGFβ), including migration, invasion, and metastasis are unmasked. In an effort to identify mechanisms that regulate the pro‐migratory TGFβ ‘switch’ in mammary epithelial cells in vitro, we found that TGFβ stimulates the phosphorylation of Smad1 and Smad5, which are typically associated with bone morphogenetic protein signalling. Mechanistically, this phosphorylation event requires the kinase activity and, unexpectedly, the L45 loop motif of the type I TGFβ receptor, ALK5, as evidenced by studies using short hairpin RNA‐resistant ALK5 mutants in ALK5‐depleted cells and in vitro kinase assays. Functionally, Smad1/5 co‐depletion studies demonstrate that this phosphorylation event is essential to the initiation and promotion of TGFβ‐stimulated migration. Moreover, this phosphorylation event is preferentially detected in permissive environments such as those created by tumorigenic cells or oncogene activation. Taken together, our data provide evidence that TGFβ‐stimulated Smad1/5 phosphorylation, which occurs through a non‐canonical mechanism that challenges the notion of selective Smad phosphorylation by ALK5, mediates the pro‐migratory TGFβ switch in mammary epithelial cells.
In various breast cancer model systems, the transforming growth factor β (TGFβ) signalling pathway has been shown to have disparate functions during tumorigenesis (Massague, 2008). In normal mammary epithelial cells, TGFβ signalling maintains homoeostasis and elicits tumour‐suppressive responses such as growth inhibition and apoptosis (Siegel and Massague, 2003). But during the course of breast cancer progression, TGFβ signalling loses its tumour‐suppressive function and acquires the ability to drive tumour‐promoting responses such as proliferation, survival, migration, invasion, and metastasis (Jakowlew, 2006). The underlying signalling mechanisms which account for the switch in TGFβ signalling responses are not well characterized and thus the subject of intense investigation.
In the canonical view of TGFβ signalling, TGFβ signals through a heteromeric receptor complex composed of the type II TGFβ receptor kinase, TβRII, and the type I TGFβ receptor kinase, ALK5 (Feng and Derynck, 2005). The presence of TGFβ ligand results in receptor complex activation, thereby allowing ALK5 to phosphorylate Smad2 and Smad3, the so‐called TGFβ‐Smads, at their C‐terminal SSXS motifs. The selectivity of Smad2/3 phosphorylation by ALK5 is ensured by the kinase domain‐embedded L45 loop, which is thought to serve as the docking site for Smad2/3 (Feng and Derynck, 1997; Chen et al, 1998; Itoh et al, 2003). Following the association of phosphorylated Smad2/3 with Smad4, the Smad complexes translocate to the nucleus, where they modulate transcription of TGFβ‐responsive genes to produce a cellular response.
Although this simplistic Smad2/3‐dependent pathway accounts for many effects of TGFβ signalling, it cannot adequately explain how TGFβ signalling generates such a diverse and opposing set of responses. It is now recognized that the proteomic composition of a cell influences the cellular response to TGFβ signalling (Massague and Chen, 2000). Thus, cellular context controls not only the output response of the Smad2/3‐dependent pathway but also the activation of Smad2/3‐independent pathways that can regulate additional TGFβ responses (Derynck and Zhang, 2003).
This concept of context‐dependent control is exemplified by studies on TGFβ‐stimulated migration in breast cancer model systems. For example, previous studies suggest that TGFβ‐stimulated migration can be Smad2/3 dependent or independent (Dumont et al, 2003; Tian et al, 2003, 2004). Moreover, the ability of TGFβ to stimulate migration can be unmasked by overexpression of oncogenes such as HER2 (Seton‐Rogers et al, 2004; Ueda et al, 2004). Although these studies establish that TGFβ stimulates migration in breast cancer cells, few studies have investigated how cellular contexts that permit TGFβ‐stimulated migration affect the ability of TGFβ to signal through Smad proteins.
The TGFβ‐related bone morphogenetic proteins (BMPs) transmit signals through Smad1, Smad5, and Smad8, the so‐called BMP‐Smads, by activating type I receptors containing L45 loops that specify Smad1/5/8 phosphorylation (Feng and Derynck, 2005). Interestingly, TGFβ has been shown to stimulate Smad1/5 phosphorylation and regulate migration in endothelial cells (Goumans et al, 2002). Contrasting the canonical view of TGFβ signalling, these effects have been attributed to lateral signalling through the Smad1/5/8‐phosphorylating type I receptor, ALK1, which is thought to form a mixed ALK1–ALK5 receptor complex in the presence of TGFβ (Goumans et al, 2003). Although these ALK1‐based studies suggest that BMP‐Smads have a functional role downstream of TGFβ receptor complex signalling, the direct involvement of Smad1/5/8 during TGFβ‐stimulated migration has not been demonstrated. Moreover, apart from these endothelial cell‐based studies, few studies have investigated the prevalence, regulation, and potential relevance of TGFβ‐stimulated Smad1/5/8 phosphorylation in other cell contexts.
Herein, we establish the mechanism and function of TGFβ‐stimulated Smad1/5 phosphorylation. We provide evidence of a previously unrecognized ALK5 L45 loop‐dependent mechanism of TGFβ‐stimulated Smad1/5 phosphorylation. This phosphorylation event is essential to the initiation and promotion of potent TGFβ‐stimulated migration, and is particularly relevant to breast cancer progression as it occurs in permissive cellular environments such as those created by tumorigenic cells and their associated oncogenes.
TGF stimulates migration and Smad1/5 phosphorylation in 4T1 cells
To explore signalling pathways that regulate TGFβ‐stimulated migration, we used the tumorigenic mouse mammary epithelial line, 4T1, which is known to migrate in response to TGFβ (McEarchern et al, 2001). Although treatment of 4T1 cells with TGFβ strongly promoted migration, treatment with the related ligands, activin A or BMP7, had little to no effect on migration (Figure 1A). This effect did not result from proliferative effects of TGFβ as TGFβ had no effect on 4T1 proliferation (data not shown). We verified that TGFβ stimulated Smad3 phosphorylation, as assessed by western blot analysis using a phospho‐specific antibody that detects C‐terminally phosphorylated Smad3 (425 aa), Smad1 (465 aa), Smad5 (465 aa), and Smad8 (430 aa) (Figure 1B). Unexpectedly, TGFβ treatment also resulted in the appearance of a robust band that was similar in size to the band corresponding to phosphorylated Smad1 and Smad5, but not phosphorylated Smad8 in lysates of BMP7‐treated cells (Figure 1B). By showing that co‐depletion of Smad1 and Smad5 prevented the appearance of this band, we confirmed that TGFβ stimulated Smad1/5 phosphorylation (refer to Figure 3C). In contrast, activin A treatment, which stimulates Smad2 phosphorylation, resulted in only a slight increase in Smad1/5 phosphorylation (Figure 1B), suggesting that potent Smad1/5 phosphorylation by non‐BMP ligands is specific to TGFβ.
We next investigated the kinetics of Smad2 and Smad1/5 phosphorylation in a time course analysis. Although TGFβ‐stimulated Smad2 phosphorylation or BMP‐stimulated Smad1/5 phosphorylation persisted for at least 195 min, TGFβ‐stimulated Smad1/5 phosphorylation disappeared within 195 min (Figure 1C). Although the overall duration of TGFβ‐stimulated Smad1/5 phosphorylation was relatively short, Smad1/5, like Smad2 and Smad3, was phosphorylated within 5 min of TGFβ treatment (Supplementary Figure 1).
ALK5 kinase activity is required for TGF‐stimulated migration and Smad1/5 phosphorylation in 4T1 cells
The TGFβ‐specific effects and rapid Smad1/5 phosphorylation described above suggested a role for the type I TGFβ receptor, ALK5, during TGFβ‐stimulated migration and Smad1/5 phosphorylation. To test the requirement of ALK5, we generated 4T1 cells that stably express short hairpin RNA (shRNA) targeting nonspecific control sequence or ALK5 using the pSUPER‐retro system. Expression of two different ALK5 shRNAs resulted in an ∼65% decrease in ALK5 mRNA (Supplementary Figure 2A), potent inhibition of Smad2/3 and Smad1/5 phosphorylation following TGFβ treatment (Figure 2A; Supplementary Figure 2B), and inhibition of TGFβ‐stimulated migration (Figure 2B; Supplementary Figure 2C). Confirming that these effects were specific to ALK5 knockdown, expression of shRNA‐resistant wild‐type ALK5 (ALK5 WT) in the ALK5‐1 shRNA line rescued TGFβ‐stimulated migration and phosphorylation of Smad3 and Smad1/5 (Figure 2A and B). In contrast, expression of shRNA‐resistant kinase‐dead ALK5 (ALK5 KR) failed to stimulate phosphorylation of Smad3 and Smad1/5 (Figure 2A) to levels equivalent to ALK5 WT. Notably, a small increase in TGFβ‐stimulated Smad3 and Smad1/5 phosphorylation was detected in the ALK5 KR‐expressing line when compared with the expression of MSCV empty vector control (Figure 2A). However, this slight increase, which is most likely attributable to recruitment, scaffolding, and activation of low‐level endogenous ALK5 by ectopic ALK5, clearly failed to rescue TGFβ‐stimulated migration (Figure 2B). Confirming the importance of ALK5 kinase function, a small‐molecule inhibitor of ALK5 kinase activity, SB431542, blocked TGFβ‐stimulated migration and phosphorylation of Smad2 and Smad1/5 in a dose‐dependent manner (Figure 2C and D). Taken together, these data indicate that ALK5 kinase activity is required for TGFβ‐stimulated migration and phosphorylation of Smad2/3 and Smad1/5.
Depletion of Smad1 and Smad5 inhibits TGF‐stimulated migration in 4T1 cells
As functional inhibition of ALK5 decreased Smad1/5 phosphorylation and migration, we reasoned that Smad1/5 phosphorylation could mediate TGFβ‐stimulated migration. Consistent with this idea, even though BMP4 treatment alone failed to stimulate migration in MDA‐MB‐231 mammary epithelial cells, co‐treatment of TGFβ with BMP4, which produced higher levels of Smad1/5 phosphorylation than TGFβ treatment alone (Supplementary Figure 3A), enhanced the ability of TGFβ to stimulate migration (Supplementary Figure 3B).
To directly evaluate whether Smad1 and Smad5 functionally contribute to TGFβ‐stimulated migration in 4T1 cells, we silenced the expression of Smad1 or Smad5 using the lentiviral pSicoR‐puro system. Despite efficient and specific suppression by two independent Smad1‐targeting shRNAs, 1A or 1B, and two independent Smad5‐targeting shRNAs, 5B or 5C (Figure 3A), depletion of either Smad1 or Smad5 alone failed to block TGFβ‐stimulated migration (Figure 3B).
As Smad1 and Smad5, which are 90% homologous at the amino‐acid level, are functionally redundant in some contexts (Arnold et al, 2006), we reasoned that functional compensation could be occurring. Consistent with this notion, we detected comparable levels of phosphorylated Smad1/5 in TGFβ‐treated control, Smad1, and Smad5 shRNA‐expressing cells (Figure 3A).
Accordingly, we examined the effect of Smad1/5 co‐depletion on TGFβ‐stimulated migration. 5B+1B shRNA expression resulted in complete and specific reduction of Smad1/5 expression and TGFβ‐stimulated Smad1/5 phosphorylation (Figure 3C). This reduction in Smad1/5 expression resulted in an ∼85% decrease in BMP4‐induced transcriptional activity from the Smad1/5/8‐responsive BRE‐luciferase reporter (Supplementary Figure 4), confirming potent functional inhibition of Smad1/5 activity. Importantly, 5B+1B shRNA expression resulted in an ∼65% decrease in TGFβ‐stimulated migration when compared with control shRNA expression (Figure 3D), suggesting a partial requirement for Smad1/5 activity during TGFβ‐stimulated migration. In 1A+5C or 1B+5C shRNA cells, we observed complete inhibition of Smad1 expression, but incomplete inhibition of Smad5 expression and Smad5 phosphorylation (Figure 3C). This partial Smad1/5 co‐depletion corresponded to a moderate decrease (∼25–40%) in TGFβ‐stimulated migration when compared with control cells (Figure 3D). Taken together, these data indicate that maximal TGFβ‐stimulated migration in 4T1 cells is dependent on the total level of Smad1/5 expression and phosphorylation.
As Smad2 and Smad3 may also contribute to TGFβ‐stimulated migration, we tested the requirement of Smad2 or Smad3 using the lentiviral FG12 and pLKO.1 shRNA systems. Smad2 or Smad3 depletion alone failed to block TGFβ‐stimulated migration (Supplementary Figure 5A and B). Likewise, efficient Smad2/3 co‐depletion, as assessed by western blot analysis and Smad2‐responsive ARE‐luciferase or Smad3‐responsive SBE‐luciferase activity (Figure 3E; Supplementary Figure 5C and D), did not inhibit TGFβ‐stimulated migration (Figure 3F), suggesting that TGFβ‐stimulated Smad1/5 phosphorylation and migration are independent of Smad2/3 activity. Consistent with this view, expression of 5B+1B shRNA did not significantly alter TGFβ‐stimulated translocation of Smad3 to the nucleus or TGFβ‐induced Smad3 transcriptional activity (Supplementary Figure 5E and F). Taken together, these data indicate that Smad1/5 phosphorylation mediates TGFβ‐stimulated migration in a Smad2/3‐independent manner.
Selective Smad1/5 phosphorylation by ALK5 partially restores TGF‐stimulated migration in ALK5 shRNA‐expressing 4T1 cells
To confirm the involvement of Smad1/5 phosphorylation during TGFβ‐stimulated migration, we tested whether selective Smad1/5 phosphorylation could rescue TGFβ‐stimulated migration in the ALK5 shRNA background. To do this, we expressed a previously characterized ‘swap’ mutant of ALK5 (ALK5 3/6), in which the L45 loop of ALK5 was replaced by the corresponding L45 loop of ALK3 or ALK6, thereby permitting Smad1/5, but not Smad2/3 phosphorylation by ALK5 (Figure 4A). Consistent with published ‘swap’ mutant data (Feng and Derynck, 1997; Chen et al, 1998), expression of ALK5 3/6, but not ALK5 WT permitted TGFβ‐induced BRE‐luciferase activity (Figure 4B), which correlated with the ability of ALK5 3/6 to prolong TGFβ‐stimulated Smad1/5 phosphorylation when compared with ALK5 WT (Supplementary Figure 6). Importantly, ALK5 3/6 expression preferentially increased TGFβ‐stimulated Smad1/5 phosphorylation when compared with ALK5 KR, and partially restored TGFβ‐stimulated migration when compared with ALK5 WT (Figure 4C and D). These data reinforce the notion that Smad1/5 phosphorylation promotes TGFβ‐stimulated migration in 4T1 cells. Moreover, when taken together with the partial inhibition of TGFβ‐stimulated migration in Smad1/5 co‐depleted cells, these data indicate that Smad1/5 phosphorylation cooperates with other ALK5 L45 loop‐derived signalling events to promote maximal TGFβ–ALK5‐stimulated migration.
The ALK5 L45 loop is required for TGF‐stimulated Smad1/5 phosphorylation and migration
We next investigated potential mechanisms of TGFβ‐stimulated Smad1/5 phosphorylation. On the basis of the requirement of ALK5 activity, we hypothesized that TGFβ‐stimulated Smad1/5 phosphorylation arises from indiscriminate ALK5 L45 loop activity. To investigate this possibility, we expressed in the ALK5 shRNA background two different shRNA‐resistant ALK5 mutants (ALK5 D266A and ALK5 3A). These previously characterized mutants possess alanine substitutions within the L45 loop, thereby rendering ALK5 defective in Smad2/3 recruitment and phosphorylation (Figure 4A; Itoh et al, 2003). As with the ALK5 KR mutant, expression of the D266A and 3A mutants did not restore TGFβ‐stimulated Smad3 phosphorylation to levels equivalent to the expression of ALK5 WT (Figure 4C).
Notably, expression of the D266A and 3A mutants of ALK5 failed to restore TGFβ‐stimulated Smad1/5 phosphorylation (Figure 4C). These mutants also failed to rescue TGFβ‐stimulated migration (Figure 4D), demonstrating that TGFβ‐stimulated Smad1/5 phosphorylation and migration require the ALK5 L45 loop. When considered alongside our Smad protein depletion data, these data suggest that ALK5 L45 loop‐dependent migration does not result from L45 loop‐mediated Smad3 activation; rather, it stems from L45 loop‐mediated Smad1/5 phosphorylation.
As ALK5 L45 loop‐mediated Smad1/5 phosphorylation contrasts the widely accepted notion of selective Smad2/3 phosphorylation mediated by the ALK5 L45 loop, we sought to extend this observation to additional cell systems. As shown in Supplementary Figure 7, TGFβ‐stimulated Smad1/5 phosphorylation was also dependent on the ALK5 L45 loop in Mv1Lu‐derived R1B mink lung epithelial cells and PY‐4‐1 endothelial cells.
The requirement for the ALK5 L45 loop during TGFβ‐stimulated Smad1/5 phosphorylation suggested that ALK5 could directly phosphorylate Smad1 and Smad5. To address this possibility, we performed in vitro kinase assays using TGFβ receptor complexes immunoprecipitated from 293T cells co‐expressing HA‐tagged TβRII and HA‐tagged ALK5. As shown in Figure 4E, GST–Smad3, GST–Smad1, and GST–Smad5 were C‐terminally phosphorylated by anti‐HA immunoprecipitates containing ALK5 WT, but not ALK5 KR, suggesting that ALK5 directly phosphorylates Smad3, Smad1, and Smad5 in a kinase activity‐dependent manner. Importantly, anti‐HA immunoprecipitates containing ALK5 D266A (Figure 4F) could not phosphorylate GST–Smad1, indicating that an intact wild‐type ALK5 L45 loop is required for direct phosphorylation of Smad1/5 by ALK5. That GST–Smad1 phosphorylation occurred even in the absence of TGFβ treatment most likely reflects the ability of 293T cells to support high levels of basal Smad1/5 phosphorylation, which is due in part to ALK5 kinase activity (Supplementary Figure 8A; Daly et al, 2008), and the ability of co‐expressed TβRII and ALK5 to signal in a ligand‐independent manner (Feng and Derynck, 1996).
TGF‐stimulated Smad1/5 phosphorylation and migration in 4T1 cells is independent of BMP signalling and lateral type I receptor signalling
Although our data demonstrate the direct involvement of ALK5 during TGFβ‐stimulated Smad1/5 phosphorylation, additional mechanisms may be involved. Therefore, we asked whether TGFβ‐stimulated Smad1/5 phosphorylation and migration are secondary to activation of an autocrine BMP signalling loop. To test this possibility, we treated 4T1 cells with TGFβ in the presence or absence of soluble chimaeric antibody antagonists that neutralize the function of BMPs (BMPRIa‐Fc) or TGFβ (TβRII‐Fc). Although BMPRIa‐Fc and TβRII‐Fc effectively blocked BMP‐stimulated Smad1/5 phosphorylation and TGFβ‐stimulated Smad3 phosphorylation, respectively (Figure 5A), TGFβ‐stimulated migration and phosphorylation of Smad3 and Smad1/5 were blocked by TβRII‐Fc, but not by BMPRIa‐Fc (Figure 5A and B). These data indicate that TGFβ‐stimulated Smad1/5 phosphorylation and migration are independent of BMP‐derived signals. Additionally, the translation inhibitor, cycloheximide, did not block TGFβ‐stimulated Smad1/5 phosphorylation (Supplementary Figure 9), suggesting that de novo synthesis of secreted Smad1/5/8‐activating ligands is not required for this effect.
We next asked whether lateral signalling through Smad1/5/8‐phosphorylating type I receptors was involved in TGFβ‐stimulated Smad1/5 phosphorylation and migration. ALK1, ALK2, and ALK3, but not ALK6 mRNA expression was detected in 4T1 cells (Supplementary Figure 10). However, unlike endothelial cells, which are known to express high levels of ALK1 mRNA, 4T1 cells expressed low levels of ALK1 mRNA (Figure 5C). Indeed, we determined that 4T1 cells express three orders of magnitude less ALK1 mRNA than PY‐4‐1 endothelial cells, but relatively similar levels of ALK5 mRNA (Figure 5C). As such low ALK1 levels were beyond the limit of precise quantitation by real‐time RT–PCR (data not shown), thereby precluding our ability to assess and confirm ALK1 suppression using shRNA, we determined that ALK1 was unlikely to have a significant function.
We next investigated the potential involvement of ALK2 or ALK3. First, we verified significant ALK2 mRNA suppression by shRNA (Supplementary Figure 11A). In ALK2‐4 shRNA cells, we observed a significant reduction of Smad1/5 phosphorylation in response to BMP7, a known ALK2 ligand, but not to TGFβ (Figure 5D), thereby confirming specific suppression of ALK2. However, ALK2‐4 shRNA did not block TGFβ‐stimulated migration (Figure 5E). In ALK3‐23 shRNA cells, we verified ALK3‐specific mRNA suppression (Supplementary Figure 11B), and observed a significant reduction of BRE‐luciferase activity in response to BMP4, a known ALK3 ligand (Figure 5F). However, a decrease in TGFβ‐stimulated Smad1/5 phosphorylation or migration was not detected (Figure 5G and H). Taken together, these data suggest that lateral signalling through ALK1, ALK2, or ALK3 is not required for TGFβ‐stimulated Smad1/5 phosphorylation and migration in 4T1 cells. Consistent with this notion, we could not detect ALK3 in anti‐HA immunoprecipitates used to phosphorylate GST–Smad1/5 in vitro (Supplementary Figure 8B).
Tumorigenic mammary epithelial cells provide a permissive context for TGF‐stimulated Smad1/5 phosphorylation
TGFβ treatment has been shown to activate overexpressed Smad1 in Hs578T mammary epithelial cells (Liu et al, 1998). However, the prevalence of TGFβ‐stimulated Smad1/5 phosphorylation in mammary epithelial cells, particularly in the endogenous context, is unknown. Thus, we tested whether this phosphorylation event occurs in a small panel of mammary epithelial cell lines, including hTERT‐immortalized DU9910 cells, spontaneously immortalized MCF10A cells, and tumorigenic MDA‐MB‐231, 67NR, and MIV cells. Interestingly, we detected TGFβ‐stimulated Smad1/5 phosphorylation in tumorigenic, but not immortalized cells (Figure 6), suggesting that potent TGFβ‐stimulated Smad1/5 phosphorylation is dependent on mammary epithelial cell context.
TGF‐stimulated Smad1/5 phosphorylation is essential for establishing sensitivity to TGF‐stimulated migration
On the basis of preferential TGFβ‐stimulated Smad1/5 phosphorylation in tumorigenic mammary epithelial cells, we reasoned that breast cancer cells could provide a permissive environment for TGFβ‐stimulated Smad1/5 phosphorylation through amplification and activation of breast cancer‐associated oncogenes such as HER2. To address this possibility, we evaluated TGFβ‐stimulated Smad1/5 phosphorylation in control MCF10A cells or MCF10AN cells, which stably express a chimaeric HA‐tagged HER2 receptor, the dimerization and activation of which can be induced by the synthetic molecule, AP1510 (Muthuswamy et al, 2001). In control MCF10A cells, we observed low levels of TGFβ‐stimulated Smad1/5 phosphorylation in the presence or absence of AP1510 (Figure 7A). Although we detected slightly elevated levels of TGFβ‐stimulated Smad1/5 phosphorylation in untreated MCF10AN cells, which is most likely due to leaky chimaeric HER2 dimerization as evidenced by low‐level Erk phosphorylation (Figure 7A), we noticed potent TGFβ‐stimulated Smad1/5 phosphorylation in AP1510‐treated MCF10AN cells (Figure 7A). We also observed HER2‐enabled TGFβ‐stimulated Smad1/5 phosphorylation in independently generated MCF10A cells that overexpress full‐length human HER2 (Figure 7B), suggesting that this effect did not result from artefactual chimaeric HER2 activity. Taken together, these data demonstrate that HER2 overexpression or dimerization facilitates potent TGFβ‐stimulated Smad1/5 phosphorylation.
Previous work has shown that HER2 overexpression or dimerization provides a permissive cellular context for TGFβ‐stimulated migration (Seton‐Rogers et al, 2004; Ueda et al, 2004). Confirming this result, treatment of control MCF10A cells with TGFβ in the presence or absence of AP1510 did not result in a significant change in migration, whereas treatment of MCF10AN cells with TGFβ in the presence, but not the absence of AP1510 led to a significant increase in migration (Figure 7C). Interestingly, the magnitude of TGFβ‐stimulated migration was closely correlated with the potency of TGFβ‐stimulated Smad1/5 phosphorylation (compare Figure 7A with C), suggesting that Smad1/5 phosphorylation mediates the ability of HER2 to unmask an otherwise latent pro‐migratory function of TGFβ.
To test this idea, we suppressed Smad1 and Smad5 expression in MCF10AN cells by transiently transfecting control or Smad1/5 duplex siRNA. To assess the requirement of Smad3 in this cellular context, we also transfected Smad3 siRNA. These siRNAs specifically and efficiently suppressed target Smad protein expression (Figure 7D). Confirming the functional effect of Smad3 siRNA and the known requirement for Smad3 in the TGFβ cytostatic response (Siegel and Massague, 2003), Smad3‐depleted cells were defective in TGFβ‐stimulated growth inhibition (Figure 7E). Conversely, Smad1/5 depletion had no effect on TGFβ‐stimulated growth inhibition (Figure 7E). We then tested the effect of Smad depletion on migration of AP1510‐treated MCF10AN cells. Although TGFβ treatment led to a significant increase in migration in control siRNA‐transfected cells, TGFβ treatment had little to no effect in Smad3 or Smad1/5 siRNA‐transfected cells (Figure 7F), indicating that Smad3 and Smad1/5 are required for TGFβ‐stimulated migration in MCF10AN cells. Taken together, these data demonstrate that HER2 establishes sensitivity to TGFβ‐stimulated migration by facilitating potent TGFβ‐stimulated Smad1/5 phosphorylation.
In this report, we provide evidence for the existence, regulation, and pro‐migratory function of TGFβ‐stimulated Smad1/5 phosphorylation in breast cancer model systems. We present two main conclusions. First, TGFβ‐stimulated Smad1/5 phosphorylation occurs through an ALK5 L45 loop‐dependent mechanism. Second, Smad1/5 phosphorylation is critical to the initiation and promotion of sensitivity to TGFβ‐stimulated migration in mammary epithelial cells.
TGF‐stimulated Smad1/5 phosphorylation occurs through an ALK5 L45 loop‐dependent mechanism
We and others have clearly demonstrated in different cell types that ALK5 is required for TGFβ‐stimulated Smad1/5 phosphorylation (Goumans et al, 2003; Daly et al, 2008). However, the precise involvement of ALK5 remains controversial. TGFβ has been proposed to signal laterally through a mixed ALK1/2/3–ALK5 complex (Goumans et al, 2003; Daly et al, 2008). However, this proposed mechanism is under increased scrutiny due to recent evidence indicating that direct formation of mixed ALK5 receptor complexes is unlikely due to differences in ALK receptor‐binding modes (Groppe et al, 2008). Indeed, our data indicate that lateral type I receptor signalling is unlikely to be involved during TGFβ‐stimulated Smad1/5 phosphorylation, as evidenced by low‐level expression of ALK1 in 4T1 and MCF10A‐derived cells (data not shown) and by the inability of ALK2 or ALK3 depletion to block TGFβ‐stimulated Smad1/5 phosphorylation. Notably, pretreatment of 4T1 cells with dorsomorphin, a small‐molecule inhibitor of BMP signalling that blocks Smad1/5/8 phosphorylation by ALK2/3/6 (Yu et al, 2008), partially blocked TGFβ‐stimulated Smad1/5 phosphorylation (Supplementary Figure 11C). Although this result appears to suggest a requirement for ALK2/3/6, our data from ALK5 3/6 mutant studies suggest that the effect of dorsomorphin may not be attributable to the inhibition of ALK2/3/6 kinase activity, but rather to the prevention of Smad1/5 recruitment to the L45 loop of ALK receptors (Supplementary Figure 11D).
We provide evidence of a direct role for ALK5 in TGFβ‐stimulated Smad1/5 phosphorylation. The ALK5 L45 loop is required for TGFβ‐stimulated Smad1/5 phosphorylation in mammary epithelial, lung epithelial, and endothelial cells, suggesting that ALK5 L45 loop dependence is a conserved feature of TGFβ‐stimulated Smad1/5 phosphorylation in other cell types. Consistent with this idea, a TGFβ receptor complex composed of wild‐type TβRII and ALK5 is sufficient for Smad1/5 phosphorylation in vitro. Moreover, an intact ALK5 L45 loop is required for Smad1 phosphorylation in vitro. Although our results depart from the long‐standing view of selective Smad2/3 phosphorylation mediated by the ALK5 L45 loop, it corroborates the central role of the L45 loop during Smad recruitment and phosphorylation. Therefore, we conclude that ALK5 has a more direct function than previously recognized during TGFβ‐stimulated Smad1/5 phosphorylation.
The factors that regulate ALK5 L45 loop‐mediated Smad1/5 phosphorylation are unknown, but Smad2 and Smad3 are candidates based on their known L45 loop‐dependent recruitment to ALK5. However, Smad3 depletion in MCF10AN cells or Smad2/3 co‐depletion in 4T1 cells did not affect TGFβ‐stimulated Smad1/5 phosphorylation. Moreover, TGFβ‐stimulated Smad1/5 phosphorylation is detected in MCF10A‐HER2 cells despite significantly downregulated levels of Smad3 expression relative to control MCF10A‐vector cells (see Figure 7B). Thus, it is unlikely that Smad2/3 regulates TGFβ‐stimulated Smad1/5 phosphorylation.
Although we detected TGFβ‐stimulated Smad1/5 phosphorylation, we did not observe TGFβ‐induced transcriptional activity from the Smad1/5/8‐responsive BRE‐luciferase reporter (Supplementary Figure 4). The relatively short duration of TGFβ‐stimulated Smad1/5 phosphorylation most likely accounts for the lack of BRE‐luciferase activity. Supporting this view, optimal activation of a Smad‐responsive reporter requires at least 3–4 h of continuous ALK5 activity or Smad phosphorylation (Inman et al, 2002). Additionally, the ALK5 3/6 ‘swap’ mutant, which prolongs TGFβ‐stimulated Smad1/5 phosphorylation when compared with ALK5 WT, enables TGFβ‐induced BRE‐luciferase activity. Importantly, the lack of TGFβ‐induced BRE‐luciferase activity suggests that Smad1/5‐dependent TGFβ‐stimulated migration does not require some classical Smad1/5‐dependent BMP target genes. Indeed, BMP4, but not TGFβ upregulates Id1 and Id2 mRNA expression in 4T1 cells (data not shown). We are currently conducting microarray analysis to identify potential Smad1/5‐dependent TGFβ target genes as part of a larger effort to explore mechanisms of Smad1/5‐mediated migration.
Smad1/5 phosphorylation is critical to the initiation and promotion of sensitivity to TGF‐stimulated migration in mammary epithelial cells
Previous work has shown that activation of oncogenes, such as Ras and HER2, can unmask the pro‐metastatic, pro‐invasive, and pro‐migratory functions of TGFβ signalling (Oft et al, 1998; Seton‐Rogers and Brugge, 2004), but few mechanisms regulating this TGFβ switch have been identified. One potential mechanism involves prolonged signalling through receptor‐tyrosine kinase effectors, such as Rac1 and ERK (Seton‐Rogers et al, 2004; Wang et al, 2006). Another mechanism involves the subversion of Smad3 tumour‐suppressive function to pro‐invasive function (Sekimoto et al, 2007). We demonstrate the importance of a third mechanism involving TGFβ‐stimulated Smad1/5 phosphorylation.
One advantage of the TGFβ‐stimulated Smad1/5 phosphorylation response for pre‐malignant mammary epithelial cells is that it helps initiate the pro‐migratory response to TGFβ. Indeed, in MCF10AN cells, the ability of HER2 to elevate TGFβ‐stimulated Smad1/5 phosphorylation to functionally relevant levels underlies the known ability of HER2 to unmask TGFβ‐stimulated migration. Notably, HER2 is but one factor that promotes TGFβ‐stimulated Smad1/5 phosphorylation as functional inhibition of HER2 in 4T1 cells using shRNA‐mediated mRNA depletion or small molecule ErbB kinase inhibitors fails to block TGFβ‐stimulated Smad1/5 phosphorylation (data not shown).
A second advantage of the TGFβ‐stimulated Smad1/5 phosphorylation response is that it provides a Smad2/3‐independent pathway, which is capable of stimulating migration or invasion (Supplementary Figure 12). This may be of particular relevance to breast cancer progression as at least a subset of breast cancer cells can acquire resistance to the tumour‐suppressive or pro‐migratory functions of Smad3 (Chen et al, 2001; Dumont et al, 2003; Gomis et al, 2006). This notion is supported by our observation of Smad3‐dependent TGFβ‐stimulated migration and growth inhibition in MCF10AN cells, which are thought to model early‐stage cancer cells (Muthuswamy et al, 2001), and Smad2/3‐independent TGFβ‐stimulated migration and resistance to the TGFβ cytostatic response in tumorigenic 4T1 cells.
As a possible response to increasing resistance to Smad3‐dependent functions, some cancer cells may promote sensitivity to TGFβ‐stimulated migration through non‐canonical pathways (Bakin et al, 2000, 2002; Dumont et al, 2003; Northey et al, 2008). Our data from 4T1 cells indicate that the ALK5 L45 loop is required for two such pathways. One is Smad1/5‐dependent; the other is Smad1/5‐independent. The basis for Smad1/5‐independent pathways is not known but could involve XIAP and Dab2, which associate with ALK5 in an L45 loop‐dependent manner (Itoh et al, 2003). Thus, in the tumorigenic 4T1 cell context, Smad1/5 phosphorylation cooperates with additional ALK5 L45 loop‐derived TGFβ signals to promote maximal sensitivity to TGFβ‐stimulated migration.
The detection of high levels of TGFβ‐stimulated Smad1/5 phosphorylation in tumorigenic, but not immortalized, mammary epithelial cells suggests that elevated levels of Smad1/5 phosphorylation are functionally relevant during breast cancer progression. Supporting this notion, a recent study has reported intense phospho‐Smad1/5 staining in primary and metastatic tumour tissues derived from breast cancer patients with bone metastasis, yet little to no phospho‐Smad1/5 staining in non‐cancerous mammary epithelial tissue from the same patients (Katsuno et al, 2008). Although such immunohistochemical analysis cannot resolve whether phospho‐Smad1/5 staining results from TGFβ, BMPs, or other ligands, that intense phospho‐Smad2 staining is detected in the same primary and metastatic tumour tissues, but not in the non‐cancerous mammary epithelial tissue, strongly suggests that Smad1/5 phosphorylation is due, at least in part, to active TGFβ signalling in malignant mammary epithelia. Thus, this study supports the possibility that Smad1/5‐dependent TGFβ‐stimulated migration and invasion occur in vivo during breast cancer progression.
In summary, we propose that in permissive environments created by cancer cells or oncogene activation, TGFβ‐stimulated Smad1/5 phosphorylation requires the ALK5 L45 loop, promotes maximal TGFβ‐stimulated migration, and mediates the pro‐migratory TGFβ switch. We are particularly interested in confirming the role of the TGFβ–ALK5–Smad1/5 axis in physiologically and pathologically relevant model systems in the light of our observation that the pro‐migratory and growth inhibitory functions of TGFβ are separable by Smad1/5 (Figure 7). This interesting observation raises the possibility that TGFβ–ALK5–Smad1/5 axis‐targeting inhibitors can selectively block the pro‐tumorigenic functions of TGFβ.
Materials and methods
Cell culture, antibodies, plasmids, and reagents
A detailed description of these materials can be found in Supplementary data.
Generation of shRNA and/or ALK5‐expressing stable cell lines
Detailed information regarding retroviral and lentiviral expression vectors or shRNA vectors and sequences are available in Supplementary data. Two rounds of 24‐h viral transductions were performed using 0.45 μm‐filtered, 8 μg/ml polybrene‐spiked supernatants acquired 24 and 48 h after transfection of 293T cells with shRNA and helper constructs using Fugene6 (Roche). After 24 h recovery in appropriate culture media, infected cells were selected by treatment with 2 μg/ml puromycin or by FACS for GFP+ populations. GFP sorting typically resulted in 90–95% GFP+ cell populations.
siRNA reagents and MCF10AN transfection
Duplex siRNA including two non‐silencing controls, NSC‐1 and NSC‐2, Smad1–7, Smad5–8, and Smad3 smartpool were purchased from Dharmacon, and diluted to 20 nM stocks. MCF10AN cells (3 × 105) were plated overnight in six wells in growth media, and starved in 1.5 ml assay media (−p/s) immediately before two rounds of 5‐h Lipofectamine 2000 (L2000)‐based transfections with 500 μl of an siRNA:L2000:optiMEM mix. The siRNA:L2000 ratios were as follows: control knockdown (3 μl NSC‐1:3 μl NSC‐2:6 μl L2000), Smad1/5 knockdown (3 μl Smad1–7:3 μl Smad1–8:6 μl L2000), and Smad3 knockdown (6 μl Smad3:6 μl L2000). Cells were replenished with growth media, and used in subsequent assays 24 h later. ∼95% transfection efficiency was achieved under these conditions.
Western blot analysis
Here, 2–3 × 105 cells were seeded in six‐well format, and treated with vehicle (4 mM HCl+1 μg/μl BSA in water), 100 pM TGFβ, 20 ng/ml activin A, 33 ng/ml BMP2 or BMP4, and 100 ng/ml BMP7 in the presence or absence of 200 ng/ml TβRII‐Fc or 1 μg/ml BMPRIa‐Fc. In MCF10A‐based studies, cells were starved for 24 h in assay media before treatment with vehicle or 200 pM TGFβ unless noted otherwise. Lysates were harvested in ULB pH 7.5 (50 mM Tris, 150 mM NaCl, 50 mM NaF, 0.5% NP‐40 supplemented with 1 mM PMSF, 1 mM sodium orthovanadate, 1 mM DTT, and protease cocktail inhibitor), subjected to SDS–PAGE and western blot using the appropriate primary and secondary antibodies. In some exposures, vertical hash marks were added to indicate position of cropped, irrelevant bands.
Transwell migration assay
4T1 cells were typsinized, resuspended in culture media, and loaded into the upper chamber at 2–4 × 104 cells per 24‐well transwell insert (8 μm pore size; Corning). Transwell inserts were incubated at 37°C for 20 h in lower chamber culture media containing vehicle or 100 pM TGFβ. For MCF10A‐based studies, cells were starved for 24 h in assay media before loading into transwells. Lower chamber was filled with assay media containing vehicle or 200 pM TGFβ with or without 500 nM AP1510. After removing non‐migrating cells using cotton swabs, inserts were washed in PBS, fixed in 4% PFA, and stained with 0.5% toluidine blue in 4% PFA. For each condition, quantitation of cell migration was determined by averaging the total cell count of five random fields at × 100 magnification or by densitometry from triplicate membranes. Error bars represent standard error of the mean.
In vitro kinase assay
Assays were performed as described earlier (Macias‐Silva et al, 1996). Briefly, TGFβ receptor complexes were immunoprecipitated from 293T cells transiently transfected with TβRII–HA and ALK5–HA WT, ALK5–HA KR, or ALK5–HA D266A using anti‐HA antibody and protein‐G sepharose beads, washed three times with lysis buffer, pH 7.5 (20 mM Tris, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X‐100 supplemented with 1 mM PMSF, 1 mM sodium orthovanadate, 2 mM sodium fluoride, 2 mM β‐glycerophosphate, and protease cocktail inhibitor), then washed two times with kinase buffer, pH 7.4 (5 mM Tris, 1 mM magnesium chloride, and 0.1 mM calcium chloride). Receptor complexes were then incubated for 30 min at room temperature in kinase buffer containing 50 μM ATP and bacterially expressed GST–Smad. Samples were mixed with 4 × loading buffer, boiled, and then subjected to SDS–PAGE and western blot analysis using the appropriate primary and secondary antibodies.
Total RNA (1 μg), extracted from cells using Trizol (Invitrogen) or RNeasy Plus (Qiagen) kit, was used to generate cDNA using the Superscript III cDNA synthesis kit (Invitrogen). Real‐time RT–PCR was performed using iQ SYBR Green Supermix (Bio‐Rad).
Growth inhibition assay
siRNA‐transfected MCF10AN cells were starved in assay media for 24 h. Then, 4 × 104 cells were plated into 12 wells in assay media containing 500 nM AP1510. After 24 h, cells were treated with vehicle or 200 pM TGFβ, incubated at 37°C for 24 h, and pulsed with 2 μCi tritiated 3H‐thymidine for the final 4 h. After washing cells with PBS and fixing with 10% TCA, incorporated 3H‐thymidine was extracted with 0.2 N NaOH, added to liquid scintillation cocktail, and quantitated in a scintillation counter. For each condition, thymidine incorporation was quantitated by averaging the counts per minute (c.p.m.) of triplicate wells. Normalized TGFβ‐stimulated growth inhibition was determined by first calculating the percentage difference in c.p.m. between TGFβ‐treated and vehicle‐treated cells, then normalizing percentage differences to that of control cells. Error bars represent standard error of the mean.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Supplementary Materials and Methods
We thank Lynn Martinek of the Duke Comprehensive Cancer Center Flow Cytometry Shared Resource for FACS assistance. This study was supported by NIH grant 1R01GM083000‐01. IML was supported by the DOD BCRP predoctoral award W81XWH‐05‐1‐0260 and NIH grant 5T32GM007105. SHS was supported by the NSF graduate research fellowship and NIH grant T32ES07031. KAK was supported by the HHMI summer undergraduate research fellowship.
- Copyright © 2008 European Molecular Biology Organization