Identification of a Novel TGFβ/PKA Signaling Transduceome in Mediating Control of Cell Survival and Metastasis in Colon Cancer

Background Understanding drivers for metastasis in human cancer is important for potential development of therapies to treat metastases. The role of loss of TGFβ tumor suppressor activities in the metastatic process is essentially unknown. Methodology/Principal Findings Utilizing in vitro and in vivo techniques, we have shown that loss of TGFβ tumor suppressor signaling is necessary to allow the last step of the metastatic process - colonization of the metastatic site. This work demonstrates for the first time that TGFβ receptor reconstitution leads to decreased metastatic colonization. Moreover, we have identified a novel TGFβ/PKA tumor suppressor pathway that acts directly on a known cell survival mechanism that responds to stress with the survivin/XIAP dependent inhibition of caspases that effect apoptosis. The linkage between the TGFβ/PKA transduceome signaling and control of metastasis through induction of cell death was shown by TGFβ receptor restoration with reactivation of the TGFβ/PKA pathway in receptor deficient metastatic colon cancer cells leading to control of aberrant cell survival. Conclusion/Significance This work impacts our understanding of the possible mechanisms that are critical to the growth and maintenance of metastases as well as understanding of a novel TGFβ function as a metastatic suppressor. These results raise the possibility that regeneration of attenuated TGFβ signaling would be an effective target in the treatment of metastasis. Our work indicates the clinical potential for developing anti-metastasis therapy based on inhibition of this very important aberrant cell survival mechanism by the multifaceted TGFβ/PKA transduceome induced pathway. Development of effective treatments for metastatic disease is a pressing need since metastases are the major cause of death in solid tumors.


Introduction
Colorectal cancer (CRC) is one of the most common malignancies with high incidence rates globally [1] and is the second highest cause of cancer related death among adults in the United States [2]. CRC can be cured by surgery and multimodal treatment in about half of the individuals with this disease (Stages I-III). However, metastasis to distant organs (Stage IV) is the most frequent cause of treatment failure [2]. Recent work has stressed on the importance of the development of inappropriate cell survival signaling for various steps in the metastatic process. Particularly noteworthy in the context of survival signaling in the metastatic process is the importance of aberrant cell survival to successful colonization at metastatic sites in distal organs [3]. Importantly, molecular mechanisms involved in the early stage of metastasis critical for diagnosis and therapy are not well understood [1]. Several key players including the Bcl-2, inhibitor-of-apoptosis (IAP) proteins XIAP and survivin, and the phosphoinositide 3-kinase (PI3K) -AKT/PKB which transmit anti-apoptotic signals in promoting cancer cell growth have been implicated in metastasis [4,5].
Tumor suppressor genes (TSG) contribute to the induction of apoptosis in response to stress. The failure to induce apoptosis in response to various types of cellular damage has been long recognized as contributing to oncogenesis. One example of TSG loss contributing to cancer formation and progression is TGFb signaling [2]. The TGFb signaling pathway has been contributing both negatively and positively in regulating growth inhibition, proliferation, replication, invasion, metastasis, apoptosis, immune surveillance and angiogenesis in a context dependent manner [6]. TGFb inhibitory/tumor suppressor responses are decreased with increasing progression and in late stage malignancies are often corrupted in a manner that supports invasion and metastasis [6]. While corruption of TGFb responses to support metastasis implies the presence of functional receptors in these cells, it is equally clear that there are substantial numbers of models ranging from transgenic mice to human cancer xenografts indicating that loss, or attenuation, of receptor expression in a wide variety of tumor types leads to increased malignancy [7,8,9,10]. These results suggest that some subgroups of cancers have pursued a pathway toward malignant progression involving the loss of TGFb receptor expression while others have, in a yet undetermined fashion, usurped TGFb signaling to drive malignant progression. There are several examples of TGFb receptor silencing in clinical samples indicating that TGFb receptor RI and/or RII (designated TGFbRI and TGFbRII respectively) downregulation are early events in oncogenesis and that loss of receptor expression by epigenetic silencing correlates to malignant progression in subgroups of several types of cancer [11,12].
We have shown that TGFb signaling in an early stage nonmetastatic colon carcinoma model leads to cell death in colon cancer cells in response to stress in association with inactivation of pAKT and inhibition of the expression of the IAP protein survivin [13]. Complex formation between survivin and another IAP protein XIAP in the cytoplasm in response to stress enables stabilization of XIAP and inhibition of its effector and executioner caspase targets to inhibit cell death [14]. Therefore, given the dominating role of PI3K/AKT signaling in cell survival mechanisms and the association of XIAP and survivin in cancer progression [5], we posited that loss of TGFb signaling may contribute to enhanced XIAP/survivin expression and consequently loss of TGFb signaling may be a key to an aberrant survival mechanism permitting metastatic growth at distal organ sites in contrast to the current view that TGFb TSG is primarily a gatekeeper to prevent oncogenesis.
TGFb signaling has been shown to activate cAMP-dependent Protein Kinase A (PKA) [15]. PKA plays a dominant role in the integration of multiple signal transduction networks [16] including the ability to disrupt the XIAP/survivin complex through phosphorylation of survivin on Ser 20 [17].
The molecular mechanisms involved in the TGFb mediated downregulation of IAP were investigated in order to determine the effects of TGFb receptor signaling on metastasis in a metastatic orthotopic model of colon carcinoma. We now report the identification of a novel TGFb/PKA/AKAP mediated transduceome that converges on XIAP function in controlling aberrant cell survival. Additionally, we have shown that TGFb receptor rescue in highly metastatic cells with epigenetic silencing of TGFb receptors leads to decreased metastases in a TGFb/PKA signaling pathway dependent mechanism in a highly metastatic orthotopic colon cancer models in vivo. Rescue of specific tumor suppressor aspects of the TGFb signaling pathway may provide therapeutic benefits without promoting the cell survival and metastatic effects of TGFb. If these TGFb tumor suppressor effects can be mimicked in late stage tumors, therapeutic value against metastatic CRC might ultimately be obtained.

TGFb activates PKA in colon cancer cells
A critical report from the Simeone laboratory noted that TGFb signaling activates PKA in Mv1Lu cells and mediates control of TGFb growth inhibitory responses [15]. TGFb activation of PKA was mediated by a Smad-dependent, cyclic AMP (cAMP)independent mechanism. This raised the possibility that the TGFb signaling mediated control of aberrant cell survival observed in the early stage TGFb-sensitive FET colon carcinoma model might involve the activation of PKA. We hypothesized that TGFb activates PKA in FET cells in a cAMP independent, Smad dependent mechanism. To this end, FET cells were treated with TGFb (5 ng/ml) for specified times, and a protein kinase assay was performed for measuring PKA activity ( Figure 1A). Following TGFb treatment, PKA activity increased by approximately 2-fold within 15 min and 4-fold at 1 h. It was observed that TGFb mediated PKA activation was completely abolished following pretreatment of cells with H89 (15 mM), a pharmacological PKA inhibitor [15]. TGFb mediated activation of PKA was also concentration dependent ( Figure 1B) with maximal activity in FET cells observed at 5 ng/ml TGFb. To confirm that activation of PKA was downstream of Smad activation by TGFb, we observed that pretreatment with H89 had no effect on phosphorylation of Smad2 by TGFb using immunoblot analysis ( Figure S1). We developed stable PKA catalytic a subunit shRNA knockdown in FET cells (designated FET PKACata KD) to validate the H89 response genetically ( Figure S2). TGFb treatment for the specified times was unable to activate PKA in the knockdown cells as opposed to robust activation in the parental FET cells ( Figure 1C). Activation of PKA by TGFb has been shown to be cAMP independent in Mv1Lu cells [15]. Classical PKA activation involves cAMP binding to the regulatory subunits of PKA to trigger dissociation of the catalytic subunits. An alternative mechanism of PKA activation involves the association of IkB with PKA catalytic subunits, thereby maintaining an inactive state of PKA. Upon IkB degradation, there is activation of PKA independent of cAMP activation [18]. We determined whether activation of PKA by TGFb is dependent upon cAMP activation in FET cells by treating cells with TGFb and measuring cAMP levels using a non-radioactive cAMP enzyme immunoassay ( Figure 1D). It was observed that TGFb was unable to increase cAMP production in contrast to Forskolin treatment which provided a significant increase in cAMP levels as expected. Next, we examined IkB protein following TGFb treatment for specified times to determine whether the activation of PKA by TGFb was due to IkB degradation ( Figure S3). The IkB levels remained unchanged following TGFb treatment. Therefore, PKA is activated by TGFb in a cAMP and IkB independent manner in FET cells in agreement with the report from Zhang et al (2004). To further confirm the functional significance of TGFb mediated PKA activation, transcription factor CREB was tested, which has been identified as a direct target of the cAMP-PKA signaling pathway [15]. We observed that TGFb was able to stimulate the phosphorylation of CREB ( Figure 1E) with no change in the total CREB levels (data not shown). Pretreatment of cells with H89 followed by TGFb exposure significantly reduced the phosphorylation of CREB. Zhang et al (2004) emphasized the role of activated Smad3 in TGFb mediated PKA activation. We developed stable shRNA Smad3 knockdown in FET cells (designated FET Smad3KD) ( Figure S4). TGFb treatment on these Smad3 knockdown cells was unable to activate PKA ( Figure 1F) indicating to the dependence of TGFb mediated PKA activation on Smad3 as reported earlier [15].
TGFb/PKA signaling mediated XIAP downregulation XIAP and survivin are well characterized IAP family members recently documented as metastatic genes [5]. Their concerted role in promoting cell survival and metastasis provides a strong rationale for targeting the expression and/or activity of these IAPs in advanced stages of cancer. An early event associated with PKA activation is phosphorylation of survivin on Ser 20 in the cytosol, but not in mitochondria [17]. This phosphorylation of survivin has been shown to disrupt the binding interface for XIAP, leading to XIAP degradation. Our data as well as earlier reports indicate that the PKA activation response is an early event following TGFb treatment and it attains maximal levels within 1 h of TGFb treatment. We hypothesized that TGFb causes a disengagement of aberrant cell survival by stimulating PKA mediated disruption of the XIAP/survivin complex. FET cells were treated with TGFb for the specified times followed by determination of XIAP and survivin (Figure 2A and S5). TGFb treatment downregulated both XIAP and survivin at the specified times. As survivin phosphorylation has already been shown to be linked to PKA activation, we focused our attention on the regulation of XIAP. Degradation of XIAP was blocked by H89 pretreatment prior to TGFb exposure for the specified times ( Figure 2B). To examine the specificity of the response, we compared the effect of TGFb on FET PKACata KD cells and parental FET cells ( Figure 2C). TGFb treatment had no effect on XIAP protein levels in FET PKACata KD cells indicating the dependence on the PKA catalytic subunit in TGFb mediated loss of XIAP protein. To confirm the specificity of TGFb/PKA pathway in this response, FET Smad3 KD cells were treated with TGFb for the specified times and XIAP protein expression was determined ( Figure 2D). Consistent with the Smad3 dependence of TGFb mediated PKA activation, stable shRNA knockdown of Smad3 abrogated XIAP downregulation. A proteasomal inhibitor (MG132) was used to determine whether loss of XIAP was dependent upon proteasomal degradation ( Figure 2E). It was observed that pretreatment of FET cells with MG132 (15 mM) for 1 h prior to TGFb treatment completely abolished the XIAP degradation indicating that XIAP is degraded within the proteasome. PKA is an important regulator of proteasomal activity and PKA has been shown to co-purify with the proteasome [19,20]. We hypothesized that TGFb/PKA mediated regulation of proteasomal activity may provide another layer of control of IAP expression (beyond the PKA induced phosphorylation of survivin on Ser 20 ). Earlier reports indicate that PKA phosphorylates the Rpt6 ATPase, which unfolds and transports substrates into the proteasome and this phosphorylation enhances the chymotrypsin proteasomal activities of the proteasome [21]. To this end, we determined whether TGFb mediated PKA activation is able to increase the chymotrypsin-like proteasomal activity using a proteasomal activity assay. TGFb treatment of FET cells selectively stimulated the chymotrypsin-like activity of the proteasome within 1 h ( Figure 2F). H89 pretreatment prior to TGFb exposure completely abrogated the basal level of the proteasomal chymotrypsin-like activity and completely abolished the stimulatory effects of TGFb on the proteasome.
Collectively, these data lead to the hypothesis that the TGFb/ PKA signaling regulates XIAP expression at multiple levels with activation of PKA leading to phosphorylation of key proteasomal components that promote proteasomal degradation of XIAP.

AKAP regulates TGFb/PKA mediated XIAP downregulation
There is an abundance of A-kinase anchoring proteins (AKAPs) which compartmentalize the PKA and other enzymes to the vicinity of specific cellular organelles for carrying out enzyme function. We hypothesized that AKAPs are a critical component in TGFb/PKA mediated XIAP downregulation. It is well documented that the subcellular localization of PKA regulatory subunits (PKARI and/or PKARII) is mediated through interactions with AKAPs [22]. AKAP inhibitor Ht31, a synthetic thyroid anchoring peptide has been shown to be a very potent competitive inhibitor of PKARII/AKAP interaction, thereby preventing PKA anchoring [23]. To ensure that the Ht31 was specifically targeting AKAP-PKA interactions, we performed PKA activity assays on FET cells to determine the extent of PKA activation following pretreatment with Ht31 (25 mM) prior to TGFb exposure ( Figure 3A). In agreement with earlier studies [15], Ht31 was able to block the TGFb mediated PKA activation. Since AKAP-PKA interaction appears to be a requirement for TGFb mediated PKA activation, we reasoned that these interactions might also be required for downstream signaling events leading to XIAP loss ( Figure 3B). Addition of Ht31inhibitor for 1 h prior to TGFb treatment for specified times completely abrogated the XIAP loss.
Since the AKAP family includes more than 50 members [24], the challenge was to identify an individual AKAP that was responding to the TGFb/PKA effects on XIAP. After initial screening for the presence of different AKAP family members in FET cells, we found AKAP149 protein expression in these colon cancer cells. AKAP149 (also termed D-AKAP1 and AKAP121 in mouse) has been identified as a membrane protein of the mitochondria [25]. It has been reported that mitochondrial AKAP121 is involved in targeting cAMP activated PKA to the outer mitochondrial membrane (OMM) and plays a role in mitochondrial biogenesis and survival [26]. We hypothesized based on the understanding that the mitochondrial XIAP and survivin move to the cytoplasm following a stress response that AKAP149 might be regulating TGFb/PKA mediated XIAP degradation. Expression of an AKAP149 small interfering RNA (siRNA) prevented the TGFb mediated PKA activation ( Figure 3C-D). Further, we observed that TGFb was unable to downregulate XIAP protein in AKAP149 siRNA knockdown cells compared to parental FET control cells ( Figure 3E). This pro- apoptotic function of AKAP149 is in contrast to the known function of mitochondrial AKAP121 (or its human homologue AKAP149) in promoting survival.
We next questioned whether the effect of AKAP149 was selective in this TGFb/PKA effect. While AKAP-PKA interaction is a global phenomenon, validation of AKAP149 as selectively regulating TGFb/PKA signaling would be a novel extension of our understanding of the TGFb/PKA mediated XIAP downregulation. AKAP220 has been shown to regulate protein phosphatase 1 (PP1) activity by coordinating the location of PKA and PP1 catalytic subunit [27]. We silenced the expression of AKAP220 in FET cells with AKAP220 siRNA. These transfected cells when treated with TGFb had no effect on either the induction of PKA activity or XIAP downregulation (data not shown). Taken together, AKAP scaffolding plays a pivotal role in the TGFb/ PKA pathway mediated XIAP downregulation with mitochondrial localized AKAP149 being necessary for the TGFb/PKA mediated response.

Involvement of PP2A in TGFb/PKA mediated XIAP downregulation
It has been observed that activation of PKA leads to the inactivation of AKT through dephosphorylation by protein phosphatase 2A (PP2A) [28]. This led to the hypothesis that TGFb mediated activation of PKA is responsible for the repression of AKT activation that we previously reported as a consequence of TGFb signaling [13]. Consequently, we determined PP2A activity in FET cells ( Figure 4A). TGFb treatment for specified times significantly increased the PP2A activity. PP2A selective inhibitor okadaic acid (OA, 50 nM) was able to completely abolish the PP2A activation when treated alone or pretreated prior to TGFb exposure. Next, we tested the role of TGFb in mediating AKT inactivation and XIAP loss ( Figure 4B). It was observed that TGFb mediated dephosphorylation of AKT is PP2A dependent as reflected by the ability of OA to block the TGFb mediated dephosphorylation of AKT. XIAP downregulation was also abolished by OA treatment. Although PP2A is a tumor suppressor involved in the control of several cell survival pathways [29], documentation of a role in subverting cell survival signaling within the TGFb pathway is a novel extension of PP2A function.
To further validate the role of PP2A in TGFb/PKA signaling, we generated stable shRNA PP2A catalytic subunit knockdown in FET cells designated FET PP2A Cat KD ( Figure 4C). TGFb treatment of FET PP2A Cat KD cells showed PKA activation indicating that PKA activation is upstream of PP2A activation ( Figure 4D). However, stable shRNA knockdown of the PP2A catalytic subunit completely abolished TGFb mediated XIAP inhibition and showed sustained phosphorylation of AKT as well ( Figure 4E). FET PKACata KD cells treated with TGFb also showed a sustained phosphorylation of AKT (data not shown).
While inhibition of AKT activation would have potential for many effects on cell survival, one effect that is of particular interest as related to TGFb/PKA function is in the context that it targets XIAP stability. A previous study reported that XIAP is phosphorylated by AKT on Ser 87 which leads to increased stabilization of XIAP and decreased apoptosis of ovarian cancer cells in response to cisplatin [30]. We reasoned that PP2A inhibition of AKT activity may represent an additional mechanism for the disruption of the stabilization of XIAP that is targeted by the TGFb/PKA pathway through enhancement of PP2A activity. To this end, FET cells were treated with TGFb or pretreated with different TGFb/PKA pathway inhibitors prior to TGFb exposure for specified times and immunoprecipitated for XIAP and immunoblotted for possible proteins bound to XIAP. We found an association of XIAP with pAKT ( Figure 4F). Following TGFb treatment, there was a significant dissociation of the XIAP-pAKT complex. However, pretreatment with H89 or Ht31 prior to TGFb exposure completely abrogated XIAP-pAKT dissociation of the two proteins indicating that TGFb/PKA signaling mediated inactivation of AKT destabilizes XIAP leading to its proteasomal degradation. Therefore, using inhibitor and stable knockdown studies, we underscore the novel finding that TGFb/PKA signaling mediates loss of XIAP protein by a PKA-PP2A dependent repression of AKT activation leading to XIAP loss.
TGFb receptor reconstitution leads to decreased metastatic colonization: Impact on TGFb/PKA signaling We have developed technology for orthotopic colonic implantation of human colon cancer lines in athymic mice that allows for reproducible quantitative analysis of metastasis to the liver and lungs [31,32,33]. The metastatic pattern displayed in this model system reflects the nature of metastatic spread in human patients. Highly metastatic GEO colon cancer cells were used to further understand the impact of TGFb/PKA signaling on metastasis. GEO cells have attenuated TGFbRI expression and thus attenuated TGFb tumor suppressor signaling as well [8]. We have utilized colonic orthotopic implantation of subcutaneously grown xenografts to characterize factors that influence the extent of metastasis to liver and lungs without affecting invasion at the primary tumor site [31,33]. GEO cells were found to be highly metastatic in this orthotopic model as reflected by metastatic colonization in 53% of the implanted animals ( Table 1). Rescue of receptor attenuation in GEO cells by transfection of a TGFbRI expression vector (designated GEORI) resulted in the reduction of metastatic incidence to about 20% of animals implanted without affecting invasion at the primary site as both GEO and GEORI transfected animals gave rise to an invasive primary tumor in 100% of implanted animals. GEO transplanted animals developed robust liver metastasis compared to GEORI ( Figure 5A). Hematoxylin and Eosin (H&E) staining showing liver metastasis in GEO cells is shown in Figure 5B. The characterization of GEO tumors showed increased cell survival signaling as reflected by lower TUNEL rates relative to GEORI tumors indicating a repression of metastatic colonization by TGFb tumor suppressor signaling is associated with repression of cell survival signaling in vivo ( Figure 5C-D). Further, no change in Ki67 IHC staining was observed between highly metastatic GEO and poorly metastatic GEORI primary tumors indicating no difference in proliferation rate in vivo between GEO and GEORI tumors (Fig 5E-F). The in vivo findings indicate that the restoration of TGFb receptor suppresses metastatic competence in GEO cells at the level of metastatic colonization as opposed to preventing invasion. A key to capitalizing on the repression of metastasis by TGFb signaling is the elucidation of the molecular mechanism by which metastasis is inhibited. Based on the decrease in metastatic incidence with TGFb receptor reconstitution leading to rescue of TGFb signaling, we hypothesized that TGFb/PKA mediated XIAP downregulation requires an intact TGFb signaling mechanism which is lost in the highly metastatic GEO cells. However, rescue of TGFb signaling by receptor reconstitution in GEORI cells leading to decreased metastatic incidence should increase TGFb/PKA signaling and its downstream effects on XIAP. GEO cells with attenuated TGFb signaling had no PKA activation with TGFb exposure ( Figure 5G). However, GEORI cells with functional TGFb signaling due to receptor reconstitution had a robust increase in PKA activation by TGFb which was about 4fold higher than the control. A second colon carcinoma cell line (CBS cells) with attenuated TGFbRII signaling [34] also showed reduced metastasis after rescue of receptor deficiency (data not shown). A similar response in PKA activity was observed in highly metastatic CBS cells compared to the poorly metastatic CBSRII cells after TGFb treatment in vitro ( Figure 5H).
Following this observation, we hypothesized that the highly metastatic GEO and CBS cells would be resistant to the TGFb/ PKA signaling mediated loss of XIAP protein. We reasoned that the inability of TGFb signaling to induce PKA activation in these cells due to receptor inactivation would also prevent its downstream effects on XIAP thereby making these cells more metastatic through increased pro-survival signaling. While GEO  and CBS both showed no response to TGFb treatment at specified times ( Figure 5I-J), GEORI and CBSRII cells showed XIAP downregulation for similar treatments ( Figure 5I-J). To validate the role of PKA activation in XIAP downregulation by TGFb, we used H89 pretreatment prior to TGFb exposure in GEORI and CBSRII cells ( Figure 5K). In line with our observations in FET cells, both GEORI and CBSRII cells pretreated with H89 showed abrogation of TGFb/PKA mediated XIAP downregulation. Thus, TGFb receptor restoration in deficient cells was able to reactivate the TGFb/PKA signaling pathway in poorly metastatic cells leading to XIAP loss. These cells mimicked the observation from FET cells, which has a functional TGFb/PKA signaling. Therefore, the in vitro results are consistent with the in vivo observation that TGFb receptor rescue reactivates the attenuated TGFb signaling in controlling cell survival and metastasis. Taken together, these findings along with clinical evidence of TGFb receptor silencing suggests that reconstitution of TGFb receptor expression could lead to inhibition of growth and/or induce apoptosis in highly progressed metastatic colon cancer cells by the TGFb/PKA signaling pathway.

TGFb/PKA signaling controls cell survival
Recently, we showed that TGFb signaling leads to cell death in FET cells in vitro in response to stress [13]. Based on this background and the current findings regarding AKAP-PKA transduceome signaling in XIAP downregulation, we hypothesized that the tumor suppressor effects of TGFb can be mediated through the TGFb/PKA signaling axis. Using FET cells as an established model, we tested the effects of TGFb treatment on cell death to further understand the role of the TGFb/PKA signaling pathway in abrogating cell survival. TGFb treatment for 48 h showed approximately 2-fold increase in cell death ( Figure 6A). To better understand the impact of PKA in the TGFb effect, cells were pretreated with low dose of H89 (1 mM) prior to TGFb treatment. Compared to control FET cells, inhibiting the PKA activation with H89 significantly decreased the cell death in these cells. To further confirm the role of PKA, FET PKACata KD cells were treated with TGFb for the specified times and cell death was assessed in comparison to FET cells ( Figure 6B). PKA catalytic subunit knockdown completely abrogated the cell death by TGFb treatment indicating that PKA activation is critical in TGFb mediated cell death.
Since AKAP149 was observed to be a critical player in the TGFb/PKA mediated XIAP loss, we reasoned that AKAP149 might also be critical for the TGFb/PKA mediated cell death. Silencing of AKAP149 expression using AKAP149 siRNA in FET cells followed by TGFb treatment to assess cell death was performed ( Figure 6B). AKAP149 knockdown completely abrogated cell death in these cells. The siRNA knockdown studies indicate that AKAP149 is selective for TGFb/PKA signaling in FET cells.
Since TGFb mediated PKA activation leads to activation of PP2A followed by AKT dephosphorylation and XIAP downregulation, we reasoned that inhibiting PP2A should also alter the cell death response to TGFb. For this reason, FET PP2A Cat KD cells were compared with parental FET cells for effects on cell death by TGFb signaling ( Figure 6B). Indeed, stable knockdown of PP2A catalytic subunit completely abrogated the PP2A mediated cell death.
GEO cells reconstituted with TGFbRI showed a decrease in metastatic capability. Consequently, we determined whether receptor reconstitution also causes an increase in functional TGFb tumor suppressor signaling leading to increased cell death in these cells. Comparison of GEO and GEORI cells with respect to cell death by TGFb treatment for 48 h was performed ( Figure 6C) and similarly, comparisons were performed with CBS and CBSRII cells as well ( Figure 6D). Death assays revealed a striking difference in their responsiveness to TGFb induced cell death. While the GEO and CBS cells were completely resistant to TGFb mediated cell death; the GEORI and CBSRII cells showed significant increases in cell death following TGFb treatment. We hypothesized that the increase in cell death due to receptor reconstitution was due to the restoration of TGFb/PKA signaling. To this end, GEORI and CBSRII cells were treated with TGFb or pretreated with H89 prior to TGFb exposure and cell death was assessed ( Figure 6E-F). In both GEORI and CBSRII cells, H89 was able to abrogate the TGFb response indicating the importance of TGFb/ PKA signaling in controlling cell survival in these cells.
We hypothesized that the TGFb/PKA mediated XIAP downregulation would require dissociation of the XIAP and survivin complex prior to XIAP degradation. We compared the endogenous levels of the XIAP/survivin complex in CBS and CBSRII cells and determined whether TGFb was able to dissociate the XIAP/survivin complex in a PKA dependent manner ( Figure 6G). CBS cells (TGFb signaling deficient) had higher levels of survivin bound to XIAP compared to CBSRII cells with restored TGFb signaling. Treatment of TGFb completely dissociated survivin from XIAP in CBSRII cells while survivin remained bound to XIAP in CBS cells. Blocking PKA activation by H89 significantly abrogated the TGFb mediated XIAP/survivin complex dissociation in accordance with the well documented role of PKA in dissociation of the XIAP/survivin complex [17]. Most XIAP/survivin complex studies have been performed using overexpression approaches as only a small fraction of the total cellular XIAP and survivin form a complex. We used the endogenous protein as opposed to the other approach since overexpression might lead to artifacts in the interaction. Therefore, TGFb was able to dissociate XIAP/ survivin complex through the TGFb/PKA signaling in poorly metastatic cells which have a restored TGFb signaling as opposed to their highly metastatic counterparts deficient in functional TGFb signaling.

TGFb signaling and aberrant cell survival
We have identified a novel mechanism of TGFb tumor suppressor signaling pathway that is capable of counteracting aberrant cell survival. This TGFb/PKA/AKAP149 dependent transduceome involves a TGFb initiated series of kinase and phosphatase events that sequentially converge on the inhibition of XIAP function in promoting cell survival and thereby permitting cell death in response to stress. The proposed mechanism involves the activation of PKA in a manner that is independent of cAMP, but AKAP and Smad3 dependent. PKA activation results in PP2A mediated dephosphorylation of pAKT followed by destabilization of XIAP resulting in its proteasomal degradation (Figure 7). The ability of XIAP to directly inhibit caspases in vivo [35] makes it a critical element in the control of apoptotic threshold in cancer cells. More recently, PKA was found to mediate compartmentalized regulation of survivin on Ser 20 selectively in the cytosol but not in mitochondria leading to the control of survival signals through degradation of the IAPs [17]. A critical aspect of this novel TGFb/PKA pathway is the implication that the TGFb mediated activation of an AKAP/PKA complex initiates a multifunctional cascade directed at disruption of XIAP mediated cell survival involving several different targets including PP2A and the proteasome along with cytosolic survivin.
The tetrameric PKA holoenzyme is a Ser/Thr kinase consisting of two catalytic (C) subunits bound to two (R) subunits (RIa/b and RIIa/b) that renders it inactive [16]. Classically, following elevation of cAMP, the C-subunits disengage from the R-subunits and actively phosphorylate proteins in the cellular vicinity [27]. The correct sub-cellular localization of the PKA holoenzyme within cellular compartments is a function of scaffolding and anchoring proteins AKAPs [27] which contribute to the spatiotemporal regulation of second messenger signaling events [24]. Similar to Zhang et al (2004), we found that AKAP-PKA interaction was required for PKA activation by TGFb. Earlier work has suggested that AKAPs are required for subcellular localization of PKA and interaction with Smads [15]. We identified a specific mitochondrial localized AKAP149, as having a critical role in TGFb mediated PKA activation. Knockdown of AKAP149 abrogated TGFb induced PKA activation and downstream signaling leading to XIAP loss. We demonstrated for the first time that AKAP149/PKA was directly involved in the TGFb's ability to induce cell death in colon cancer cells. Our study demonstrates that AKAP149 might have context dependent effects based on the mechanism of PKA activation. The cAMP mediated AKAP121/PKA signaling to mitochondria inhibits apoptosis [26]. However, we have demonstrated here that AKAP149 is also involved in a pro-apoptotic role during TGFb mediated PKA activation independent of cAMP as evidenced by AKAP149 siRNA studies.
An intriguing observation made in our studies which originally showed that TGFb signaling inhibits survivin expression was that TGFb signaling also mediated inactivation of AKT [13]. The activation of PKA has been shown to inactivate AKT through dephosphorylation by PP2A [28]. Consequently we tested the hypothesis that inactivation of AKT is dependent upon PP2A in our studies of TGFb/PKA mediated control of XIAP expression. PP2A is ubiquitously expressed as a trimeric complex composed by a C-catalytic subunit, an A-scaffolding subunit and a B-targeting subunit which is drawn from among several families of multiple isoforms [36] . PP2A has been observed to have a tumor suppressor function [29] playing a critical role in apoptotic cell death through differential interactions with Bcl2 and caspase3. TGFb has been shown to induce G1 arrest mediated through PP2A [37]. Studies with the PP2A selective inhibitor okadaic acid (OA) have shown that it acts as a tumor promoter in mouse skin carcinoma [38]. We demonstrated that TGFb mediated activation of PKA was upstream of PP2A activation by stable transfection of FET cells with PP2A catalytic shRNA or treatment with OA resulting in PP2A mediated XIAP degradation through inactivation of AKT. Although PP2A is a tumor suppressor involved in the control of several cell survival pathways, documentation of a role in subverting cell survival signaling within TGFb function is a novel extension of its function.
We have shown that XIAP is associated with pAKT and this association is disrupted following TGFb treatment ( Figure 4F). AKT along with some of its substrates constitute a major cell survival pathway [30]. Importantly, the anti-apoptotic functions of AKT have been proposed to be at the post-mitochondrial level. It has been shown to directly phosphorylate and inactivate caspase 9 and Bax [30]. XIAP inhibits death-signaling pathways also at the post-mitochondrial level [30] and has been documented as a physiological substrate of AKT. In HEK293 cells, AKT physically associates with XIAP and phosphorylates XIAP on Ser 87 , thereby stabilizing XIAP and preventing it from auto-ubiquitination [30].
We have now demonstrated that pAKT and XIAP are in a complex and TGFb is able to modulate the complex formation. Further, we observe that dissociation of complex formation by TGFb was inhibited by H89 and Ht31. This strongly supports the notion that TGFb exerts its effects on cell survival through the TGFb/PKA transduceome signaling.
In association with direct regulation of XIAP by TGFb/PKA signaling, we also demonstrated that chymotrypsin-like activity of the proteasome is induced by TGFb in a PKA dependent manner. XIAP can be auto-ubiquitinated and degraded when treated with DNA damaging agents like chemotherapeutic drugs [30]. The fact that XIAP has an E3-ligase which could act as a scaffold for EI or E2 ubiquitination enzymes is consistent with XIAP as a potential proteasomal target. PKA is an important kinase in proteasome phosphorylation and the regulation of metabolic function [20]. Collectively, these data lead to the indicate that the TGFb/PKA transduceome regulates XIAP and survivin expression at multiple levels with activation of PKA complex leading to phosphorylation of key proteasomal components such as Rpt6 [21] which might contribute to the ubiquitin-proteasome mediated degradation of XIAP as well as other related cell survival mediators.

TGFb receptor reconstitution suppresses metastasis
We have made the novel observation that reconstitution of TGFb receptor TGFbRI in highly metastatic GEO colon cancer cells rescues TGFb signaling and inhibited metastatic colonization from orthotopic xenografts. This observation raises the therapeutic possibility that TGFb can suppress metastatic competence after this advanced state of progression is attained in cells with attenuated type I and/or II TGFb receptor expression (designated TGFbRI or TGFbRII).
We have made the novel observation that rescue of attenuated receptor expression does inhibit metastatic capability in GEORI (reconstituted with RI) transplanted orthotopic models in vivo even though invasion at the primary tumor site was maintained (Table 1). This finding indicates that the effect of receptor reconstitution was on the ability to form progressively growing colonies at the distal site rather than preventing escape from the primary tumor. This result also suggested that the inhibition of distant organ colonization might be related to stress induced cell death associated with regenerated TGFb signaling in response to the foreign micro-environment for growth of the colon cancer cells in the liver and lungs. These results are significant because they raise the possibility that regeneration of attenuated TGFb signaling leading to activation of the TGFb/PKA transduceome would be an effective strategy to target the treatment of metastasis. As discussed in the introduction, epigenetic loss of TGFb receptors is a frequent occurrence in a broad array of different types of cancer. Thus, there is the potential for translating our novel observations into the clinic. To this end, we found that pancreatic, breast and colon cancer cell lines frequently exhibited loss of TGFbRII and/or RII due to transcriptional repression. We identified HDAC inhibition as mechanism for rescue of TGFb receptor expression in these histological types of cancer [39,40,41,42,43,44].
Studies on different cancer types provide extensive support for loss of TGFb receptor expression as an important contributor to tumor progression in subgroups of several types of cancer [11,12]. However, there is also evidence from additional clinical studies indicating that other subgroups exist within the same histopathological types of cancer that utilize corrupted TGFb signaling via TGFbRI and TGFbRII as a means of contributing to cancer progression [45]. Thus, various subgroups of breast and colon cancer have evolved different strategies for generating malignant progression. The existence of corruption of TGFb signaling as a mechanism for malignant progression does not preclude the clinical importance of loss of TGFb tumor suppressor signaling due to deficiency of receptor expression as a potential target for cancer therapy. In fact, the evidence for the importance of receptor loss in clinical cancer is, as indicated above, significantly more developed than the evidence for corrupted TGFb signaling which is largely restricted to in vitro investigations at this point.
In an effort to determine the relationship between the TGFb/ PKA transduceome and the capability of a tumor for metastatic spread, we investigated the molecular aspects of the TGFb/PKA transduceome signaling in the context of metastatic progression in highly metastatic GEO and CBS cells and their poorly metastatic counterpart GEORI and CBSRII where receptor reconstitution has led to a decrease in metastatic spread. Since metastatic deposits, like primary cancers, are faced with hypoxic stress and must face the additional challenge of growth in a foreign microenvironment, it seems likely that even established metastases must retain their aberrant survival characteristics.
Striking difference in TGFb/PKA signaling activity and downstream signaling leading to XIAP downregulation was observed when highly metastatic (GEO, CBS) and their poorly metastatic colon cancer counterparts (GEORI and CBSRII) were compared. Therefore, reconstitution of TGFb receptors in metastatic models with attenuated TGFb signaling leading to metastasis might also reconstitute TGFb/PKA signaling in established metastases which could affect their survival. These lines of evidence lead to the consideration that the TGFb/PKA transduceome is a critical component in steady-state suppression of abnormal survival signaling that prevents formation of metastases by disrupting the XIAP/survivin cell survival pathway.
In conclusion, understanding drivers for metastasis in human cancer is important for potential development of therapies to treat metastases (most prominent cause of death from solid cancers). The role of loss of tumor suppressor activities in the metastatic process is essentially unknown. Presently, loss of TGFb signaling is largely regarded as a driver for transition from a benign to a malignant state subsequently leading to metastatic spread by other mechanisms or even by usurping aspects of TGFb signaling for oncogenic mechanisms such as the acquisition of EMT. However, we have now shown that loss of TGFb tumor suppressor signaling is a driver for metastatic colonization of distant organs from the primary tumor. Importantly, loss of TGFb signaling does not affect invasion as reflected by retention of a 100% rate of invasion at the primary site of metastatic cells in which TGFb signaling has been rescued and metastatic potential inhibited. This indicates that while loss of TGFb signaling enables the transition from a benign to a malignant state its loss does not drive the earliest step of the metastatic process.
However, we have shown here that loss of TGFb signaling is necessary to allow the last step of the metastatic processcolonization of the metastatic site. Moreover, we have identified a novel TGFb tumor suppressor pathway that acts directly on a known cell survival mechanism that responds to stress with the survivin/XIAP dependent inhibition of caspases that effect apoptosis. This survival mechanism is apparently not necessary for malignant behavior at the primary site once the cancer cells have made the transition from the benign state since its inhibition by restoration of TGFb tumor suppressor signaling does not alter malignant behavior at the primary site.
Our work indicates the potential for developing anti-metastasis therapy based on inhibition of this very important aberrant cell survival mechanism by this novel, multifaceted TGFb induced pathway. Our previous body of work showing that HDACi's act at least in part through the rescue of TGFb receptor I and II expression and TGFb tumor suppressor activity in a wide variety of cancer cells indicates the clinical potential of this concept.
Since TGFb signaling responses can be supportive of events that support malignant progression such as EMT and suppression of TGFb signaling as a therapy for cancer has been supported by experimental evidence, therapeutic strategies for the inhibition of TGFb signaling in cancer therapy are currently being pursued. However, our work indicates that the relationship between TGFb signaling and tumor suppressor function may be governed by context even in the later stages of cancer. The concept that the disruption of the balance between TGFb tumor suppressor activity and the survivin/XIAP cell stress response enables metastatic colonization of distant organs raises the concern that therapies aimed at inhibition of TGFb signaling may be deleterious to at least a subset of cancer patients. For example, it has been shown in several types of cancer that progressive epigenetic silencing of TGFb receptor I and/or II correlates with malignant progression. Pharmacological inhibition of TGFb receptor signaling may promote metastases in these patients.

Materials and Methods
All experiments involving animals were approved by the University of Nebraska Medical Center Institutional Animal Care and Use Committee (IACUC #: 07-047-08-FC).

Cell Culture and Reagents
The FET [13] , GEO [8] and CBS [34] colon carcinoma cells are routinely maintained in a serum free (SF) medium containing insulin (20 ug/ml, Sigma) transferrin (4 ug/ml, Sigma) and EGF (5 ng/ml; R&D Systems) as previously described [46]. Cells were harvested after the addition of TGFb (5 ng/ml) at the specified times. The TGFbRII, AKAP149, AKAP220 and PKA and PP2A catalytic subunit antibodies were obtained from Santa Cruz Biotechnology. The XIAP antibody was obtained from Abcam and actin from Sigma. The survivin, TGFbRI, AKT, Smad3, pSmad3, CREB, and IkBa antibodies as well as the PKA inhibitor H89 were obtained from Cell Signaling Technology. The AKAP inhibitor Ht31 and the phosphatase inhibitor Okadaic acid were obtained from Calbiochem.

Western Blotting and Immunoprecipitation
Cells were lysed in a Tris-HCl based buffer containing 0.5% NP-40 and appropriate protease and phosphate inhibitors as described previously [13]. Protein concentration was determined by bicinhoninic acid assay (Pierce). Immunoprecipitation was performed with 500 ug protein aliquots using magnetic beads (Millipore) according to manufacturer's instructions.

RNA Interference Studies
Smad 3 (sc-38376-SH), PKA Catalytic a (sc-36240-SH) and PP2A Catalytic a (sc-43509-SH) shRNA was obtained from Santa Cruz. These vectors were co-transfected by electroporation as described previously [34] with a puromycin selection vector and FET clones were selected in SF medium containing 4 ug/ml puromycin and 0.2% FBS. AKAP149 (sc-40301) and AKAP220 (sc-105049) siRNA and Transfection Reagent was obtained from Santa Cruz and knockdown was performed according to the manufacturer's protocol.

PKA Activity Assay
PKA activity was measured using the PepTag non-radioactive protein kinase assay (Promega) using kemptide (LRRASLG) following the manufacturer's protocol. For quantitative determination of cellular cAMP, the non-radioactive Direct Cyclic AMP Enzyme Immunoassay kit (Assay Design) was used.

cAMP Assay
For quantitative determination of cAMP, non-radioactive Direct Cyclic AMP Enzyme Immunoassay kit (Assay Design) was utilized. Manufacturer's protocol was followed for the assay.

Proteasomal Activity Assay
The fluorometric Proteasome substrate III for chymotrypsin (SUC-LLVY-AMC) and Peptidyl-glutamyl peptide hydrolyzing (PGPH)-caspase-like activity proteasome substrate II (Z-LLG-AMC) and the corresponding inhibitors were obtained from Calbiochem. FET cells were plated at 800,000 cells/100 mm dish in SF media. On day 5, cells were treated with TGFb (5 ng/ml) with or without H89 (15 mM) pretreatment for 1 h and harvested in proteasomal assay buffer [50 mM HEPES (pH 7.5), 5 mM EDTA, 150 mM NaCl and 1% Triton X-100 containing 2 mM ATP]. After quantification of protein concentration, cell lysates were pretreated with 25 uM of the chymotrypsin-like activity inhibitor Benzyloxy carbonyl Leu Leu phenylalaninal Inhibitor (ZLLF-CHO) or the PGPH activity inhibitor Z-Gly-Pro-Phe-Leu-CH (Z-GPFL-CHO) followed by incubation with chymotrypsin or PGPH substrate for 1 h at 37uC as described by [47]. The chymotrypsin and PGPH activities were measured fluorometrically.

Orthotopic Implantation for In Vivo Metastasis
All experiments involving animals were approved by the University of Nebraska Medical Center Institutional Animal Care and Use Committee. The orthotopic implantation methodology has been described in details in the following publications from Brattain laboratory [31,32,33]. Briefly, the GEO and GEORI cells used in orthotopic assays were transfected with Green Fluorescence Protein (GFP). Exponentially growing GFP-labeled cells were inoculated subcutaneously onto the dorsal surfaces of separate BALB/c nude male mice. Once xenografts were established (,1 week), they were excised and minced into 1 mm 3 pieces. These pieces were then orthotopically implanted into other BALB/c nude mice. For operative procedures, animals were anesthetized with isofluorane inhalation. A 1 cm laparotomy was performed and both the cecum and ascending colon were exteriorized. Using 76 magnification and microsurgical techniques, the serosa was disrupted by scraping in two locations. The 1 mm 3 pieces of xenograft were sub-serosally implanted using an 8-0 nylon suture at the two points of serosal disruption. The bowel was then returned to the peritoneal cavity and the abdomen was closed with 5-0 vicryl suture. Subsequently, animals were anesthetized with a 1:1 mixture of ketamine (10 mg/ml) and xylazine (1 mg/ml) by intraperitoneal injection (0.01 ml/mg) and weekly GFP fluorescence imaging was performed for up to 5 weeksWeekly GFP fluorescence imaging was performed for up to 5 weeks. Excitation of GFP in the light box facilitated identification of primary and metastatic disease by direct near-real time visualization of fluorescence in live animals. Both GEO and GEORI cells gave rise to an invasive primary tumor in 100% of the animals implanted.

Hematoxylin and Eosin, TUNEL and Ki67 Staining
Approximately 50 d post-implantation, the animals were euthanized following proper IACUC protocol. The colon (with primary tumor), liver, lungs, and heart were harvested. Following the harvesting, organs were explanted, imaged, and immediately placed in 10% neutral buffered formalin fixative for 24 h. This was followed by tissues processing and embedding in paraffin. Slides were then cut for hematoxylin and eosin (H and E) and immunohistochemical characterizations. Serial sections were cut to complement the H and E sections and were stained with an IgG1 rabbit polyclonal antibody for Ki-67 (Dako North America, Inc., Carpinteria, CA). Slides from paraffin embedded tissue blocks were stained according to the Apotag (Oncor, Gaithersburg, MD) terminal nucleotidyl transferase-mediated nick end labeling (TUNEL) assay kit. Detailed methodology and quantitative analysis for these individual experiments has been described in details in the [32,33,49].