Wnt signaling pathways have fundamental roles in animal development and tumor progression. Here, employing Xenopus embryos and mammalian cell lines, we report that the degradation machinery of the canonical Wnt pathway modulates p120-catenin protein stability through mechanisms shared with those regulating β-catenin. For example, in common with β-catenin, exogenous expression of destruction complex components, such as GSK3β and axin, promotes degradation of p120-catenin. Again in parallel with β-catenin, reduction of canonical Wnt signals upon depletion of LRP5 and LRP6 results in p120-catenin degradation. At the primary sequence level, we resolved conserved GSK3β phosphorylation sites in the amino-terminal region of p120-catenin present exclusively in isoform-1. Point-mutagenesis of these residues inhibited the association of destruction complex components, such as those involved in ubiquitylation, resulting in stabilization of p120-catenin. Functionally, in line with predictions, p120 stabilization increased its signaling activity in the context of the p120–Kaiso pathway. Importantly, we found that two additional p120-catenin family members, ARVCF-catenin and δ-catenin, associate with axin and are degraded in its presence. Thus, as supported using gain- and loss-of-function approaches in embryo and cell line systems, canonical Wnt signals appear poised to have an impact upon a breadth of catenin biology in vertebrate development and, possibly, human cancers.
Wnt signaling pathways engage in multiple aspects of development and tumor progression (Grigoryan et al., 2008; Moon et al., 2004; Nusse, 2005; Reya and Clevers, 2005). In what is defined as canonical Wnt signaling, β-catenin (also known as catenin beta-1) is generally considered the key element transmitting Wnt signals into the nucleus (McCrea et al., 1993; Moon et al., 2004; Sokol, 1999). However, recent vertebrate work from our group and others has indicated that a structurally related protein, p120-catenin (also known as catenin delta-1), also responds to Wnt signals, and unexpectedly, that a number of established β-catenin gene targets are co-regulated by p120-catenin in conjunction with the transcriptional repressor Kaiso (Park et al., 2005; Spring et al., 2005). Despite these findings, with some recent debate centered on kaiso (Iioka et al., 2009; Ruzov et al., 2009a; Ruzov et al., 2009b), an important unanswered question concerns the nature of the precise mechanism by which p120-catenin is regulated in response to the presence or absence of Wnt signals, what similarities such mechanisms might share with those known to regulate β-catenin and whether additional p120-subfamily members might also be subject to Wnt regulation.
The metabolic stability of β-catenin is regulated by destruction complex components including GSK3β (glycogen synthase kinase-3β), CK1α (casein kinase I isoform alpha), axin and APC (adenomatous polyposis coli) (Amit et al., 2002; Behrens et al., 1998; Hedgepeth et al., 1999; Ikeda et al., 1998; Lee et al., 2003). In the absence of Wnt ligand(s), β-catenin is phosphorylated by CK1α and GSK3β in association with a larger scaffolding complex including axin and APC. Phosphorylated β-catenin is then recognized and ubiquitylated by β-TrCP (β-transducin repeat-containing protein), a substrate-recognition E3 subunit of ubiquitin ligase, resulting in complex degradation through the ubiquitin-mediated proteasome pathway (Aberle et al., 1997; Liu et al., 1999; Winston et al., 1999). By contrast, when extracellular Wnt ligands bind to the transmembrane and context-dependent receptor–co-receptor proteins, Fz (Frizzled) and LRP (low-density lipoprotein receptor-related protein), intracellular Dsh (Dishevelled) becomes activated by unknown mechanisms (Zeng et al., 2008). Among other possibilities, activated Dsh together with LRP is thought to sequester axin to the inner plasma membrane and inhibit the ability of GSK3β to phosphorylate β-catenin directly, thereby promoting β-catenin stabilization (Bilic et al., 2007; Cselenyi et al., 2008; Schweizer and Varmus, 2003; Tamai et al., 2004; Wehrli et al., 2000; Zeng et al., 2008; Zeng et al., 2005). This pool of β-catenin responds to additional signals before accessing the nucleoplasmic space, where β-catenin relieves repression otherwise conferred by the HMG (high-mobility group) proteins LEF (leukocyte enhancing factor) or TCF (T-cell factor) (Esufali and Bapat, 2004; Gottardi and Gumbiner, 2004; Mosimann et al., 2009; Wu et al., 2008). Genes activated by the β-catenin–LEF (TCF) complex (canonical-Wnt target genes) are numerous, with well-known examples including the gene encoding the homeobox protein siamois (Xenopus), as well as MYC and CCND1 (encoding cyclin-D1) (Brannon et al., 1997; He et al., 1998; Nusse, 1999; Tetsu and McCormick, 1999). In addition to the extensive involvement of the Wnt pathway in embryogenesis and adult-tissue homeostasis, pathological pathway activation is linked to multiple human diseases, such as those characterized by bone abnormalities and cancer (Chen and Alman, 2009; Oving and Clevers, 2002).
Members of the p120-catenin subfamily include p120-catenin, ARVCF-catenin, δ-catenin and p0071 (Hatzfeld, 2005; McCrea and Park, 2007). p120 subfamily members bear limited structural resemblance to the β-catenin subfamily members β-catenin and plakoglobin (γ-catenin). The most obvious similarity is that each contains a central Arm (Armadillo) domain consisting of either nine (p120 subfamily members) (Choi and Weis, 2005) or 12 (β-catenin and plakoglobin) Arm repeats. Through such Arm domains, members of each catenin subfamily were first found to bind to the cytoplasmic tails of cadherin cell–cell adhesion proteins. However, while p120 subfamily members competitively associate with cytoplasmic cadherin membrane-proximal regions, where they contribute to cadherin stabilization (Reynolds and Carnahan, 2004; Xiao et al., 2007), β-catenin or plakoglobin bind more distally and confer other attributes to the complex, such as indirect cytoskeletal association (Abe and Takeichi, 2008; Drees et al., 2005; Zhurinsky et al., 2000). In addition to binding and modulating cadherins and engaging in nuclear activities, p120 subfamily members have now been well recognized to modulate small GTPases such as RhoA and Rac (Anastasiadis, 2007; Casagolda et al., 2010; Elia et al., 2006; Fang et al., 2004; McCrea and Park, 2007; Perez-Moreno et al., 2008; Wildenberg et al., 2006; Wolf et al., 2006). p120 subfamily members can further be distinguished from β-catenin or plakoglobin in that each transcript bears multiple potential translational start sites and arises from differential splicing events. These characteristics have added layers of regulatory complexity to studies of p120 subfamily proteins (Mo and Reynolds, 1996; Yanagisawa et al., 2008).
Aside from the central Arm domains, which themselves share only modest homology, catenin amino- or carboxy-terminal regions bear yet less resemblance across distinct members, even within a subfamily. These end domains engage in inter-protein associations, while in an intra-molecular manner, might modulate Arm-domain interactions (Castano et al., 2002; Choi et al., 2006; Gottardi and Gumbiner, 2004; Mo et al., 2009; Solanas et al., 2004). In the context of the canonical Wnt pathway, the amino-terminal domain of β-catenin is well known to harbor conserved GSK3β and CK1α sites that, when phosphorylated, allow for β-TrCP recognition (Aberle et al., 1997; Liu et al., 1999; Winston et al., 1999). Thus, this domain participates in determining the extent and activity of the cytoplasmic:nuclear signaling pool of β-catenin. Additionally, the β-catenin subfamily member plakoglobin contains this destruction box, and its protein stability is likewise modulated by canonical Wnt signaling (Karnovsky and Klymkowsky, 1995; Kodama et al., 1999).
Previous work arising from our and independent groups has outlined a new role for p120-catenin in nuclear signaling. One context examined has been the response of p120 to canonical Wnt signals, where, together with β-catenin, p120 modulates expression of select Wnt gene targets such as Siamois (Xenopus), CCND1 (encoding cyclin D1) and MMP7 (encoding matrilysin). At such promoters, p120 recognizes and associates with kaiso, a transcriptional repressor of the BTB/POZ zinc-finger family that recognizes consensus sequences (KCSs) in DNA (Park et al., 2005; Spring et al., 2005). Once formed, the p120–Kaiso complex is thought to dissociate from the gene promoter. Kaiso-directed gene repression is thus relieved, and increased transcriptional activity ensues. In addition to Kaiso–KCS interactions with DNA, for which conflicting reports have arisen in certain gene contexts (Iioka et al., 2009; Ruzov et al., 2009a; Ruzov et al., 2009b), Kaiso further recognizes methyl-CpG islands present in gene control regions that are associated with repressive states (Daniel, 2007; Daniel et al., 2002; Prokhortchouk et al., 2001; Yoon et al., 2003). However, at these sites in particular, no indications have yet arisen that p120-catenin acts to relieve Kaiso-mediated repression, and further uncertainty concerns the relationship between the roles of Kaiso at CpG-island versus sequence-specific DNA-binding sites.
Our most recent work has focused upon upstream elements of the p120–Kaiso pathway. A central question that we address in this report is the specific means by which p120-catenin signaling activity is controlled and the extent to which the mechanisms involved are shared with β-catenin. Using both vertebrate embryo and cell line systems, we find that regulatory overlaps do indeed exist, as examined in-depth for p120-catenin, that are further likely to extend to additional p120 subfamily members, such as ARVCF-catenin and δ-catenin. Thus, we propose that Wnt pathway signals impinge through shared mechanisms on the stability of distinct p120 subfamily members, producing a networked catenin response relevant to development and, potentially, disease.
GSK3β modulates p120-catenin stability
Based upon our prior work that suggested similarities in p120-catenin and β-catenin regulation (Park et al., 2006), we directly examined the involvement of established components of the destruction machinery for β-catenin in p120 regulation. Studies were initially performed to identify the impact of expressing exogenous GSK3β in early Xenopus embryos. As our prior findings showed that inhibiting GSK3 through LiCl increased endogenous p120-catenin levels, it was expected that ectopic expression of GSK3β should decrease p120-catenin levels (Park et al., 2006). Indeed, native GSK3β and CK1α reproducibly reduced the level of epitope-tagged Xenopus p120 isoform-1 (longest isoform; translation beginning at the most upstream ATG site of p120), whereas CK1ε or a kinase-dead form of GSK3β (KD) (negative-control), did not have statistically significant effects (Fig. 1A, and data not shown). Unexpectedly, lower as opposed to intermediate or sometimes even high GSK3β expression (5 pg versus 10 pg, or in some cases 100 pg mRNA) produced modestly greater reductions in ectopic or endogenous p120-catenin isoform-1 levels (Fig. 1A and supplementary material Fig. S1A,B), as discussed later in relation to other findings (Fig. 5 and Discussion). Exogenous epitope-tagged p120-catenin was used in these initial studies given that a large proportion of endogenous p120 is bound to cadherin, which, as established for β-catenin, would be expected to exhibit less sensitivity to Wnt-pathway regulation. We next conducted in vitro kinase assays with p120, CK1 and GSK3β. As occurs for β-catenin (Liu et al., 1999), phosphorylation upon p120 was faint but reproducible, suggesting that it is a direct GSK3β and/or CK1α target (supplementary material Fig. S2 and data not shown).
We next tested for a functional relationship between p120 and GSK3β using a phenotypic assay in vivo. p120-catenin was overexpressed, and, as anticipated, gastrulation failures arose in a significant fraction of embryos (Fang et al., 2004). However, when a carefully titrated dose of GSK3β was co-injected with p120, we observed significant rescue of the gastrulation phenotypes (Fig. 1B; compare the first two bars). Kinase-dead GSK3β did not produce such rescue effects (GSK3β KD, negative control), whereas a titrated dose of axin displayed modest rescuing activity. Consistent with their respective rescuing effects, GSK3β or axin acted to reduce exogenous p120 levels when they were coexpressed with p120 in vivo (Fig. 7A; compare lanes 1–3).
To complement over-expression assays, we next employed loss-of-function tests. As morpholinos to deplete GSK3 have not been characterized for work in Xenopus, we instead employed a proven short-hairpin shRNA to block GSK3 function in HeLa cells (Kim et al., 2006). As predicted, p120-catenin protein levels were elevated upon the depletion of endogenous GSK3, as assessed using immunofluorescence analysis as well as immunoblotting (Fig. 1D). Noteworthy in our immunofluorescent images is the increased presence of p120 in both the cytoplasmic and nuclear compartments, relative to negative-control shRNA-transfected cells. To better resolve the signaling pool of p120-catenin, we employed cell fractionation of the cytoplasmic and nuclear populations. In common with β-catenin, cytoplasmic p120-catenin was stabilized when cells were incubated in the presence of the GSK3 inhibitor LiCl, whereas a lesser effect was seen for p120 in the nuclear fraction (Fig. 1C). Additionally, the sensitivity of p120 to GSK3 was supported in pulse-chase data, where LiCl or MG132 treatment prolonged the half-life of endogenous p120-catenin (Fig. 1E). Taken together, these data (Fig. 1) suggest that p120 is subject to modulation by GSK3β.
p120-catenin is regulated by the ubiquitin-proteasome pathway
Given that β-catenin is phosphorylated by CK1α and GSK3β and degraded through the proteasomal pathway, we tested whether p120 is similarly regulated. To assist in following the signaling pool of p120, we employed MDA-MB-231 and MDA-MB-435 cells, which are largely E-cadherin deficient, in addition to 293T cells, which express E-cadherin. MDA-MB-231, MDA-MB-435 and 293T cells were incubated with the proteasome inhibitor MG132, and endogenous p120 was detected by immunoblotting. Consistent with the pulse–chase data from MG132-treated HeLa cells (Fig. 1E), p120-catenin was stabilized by MG132 in a dose-dependent manner in all cell lines tested (Fig. 2A). To detect endogenous p120-catenin, we employed two previously characterized monoclonal antibodies against p120, 6H11 and pp120 (Ireton et al., 2002). 6H11 recognizes an amino-terminal epitope, selectively resolving the isoform generated from the most-upstream translational start site, known as isoform-1. pp120 resolves additional isoforms, as it is directed against a carboxyl-terminal epitope present across most p120 translation products. In the cell lines used here, pp120 detects two products thought to be isoform-1 and -3, based upon their relative SDS-PAGE migration (Mariner et al., 2004; Paredes et al., 2007) and reactivity to 6H11 (isoform-1) versus pp120 (isoform-1 and -3, etc.). Inhibition of the proteasome pathway through dose-dependent titration of a chemical inhibitor (MG132) raised p120 isoform-1 levels in all cell lines tested, with an observable yet lesser impact upon isoform-3, as was evident in MDA-435 cells (Fig. 2A). This response will be addressed later in more depth (Fig. 7). Likewise, as CK1α is a priming kinase for GSK3β in the destruction process for β-catenin, we assayed the effect of incubating Xenopus embryos in the presence of the CK1 inhibitor D4476. Consistently, D4476 exposure increased exogenous p120 levels, as was expected and observed for β-catenin (positive control) (Fig. 2D). In keeping with our amphibian embryo and exogenous expression results, endogenous p120 isoform-1A in mammalian cells was stabilized when MDA-435 or 293T cells were treated with D4476 (Fig. 2C). Again analogous to β-catenin (Aberle et al., 1997), p120 ubiquitylation was clearly evident upon its coexpression with HA–ubiquitin in HeLa cells (Fig. 2B). Likewise, as predicted if dependent upon GSK3β, such p120 ubiquitylation dropped when cells were incubated in the presence of LiCl (Fig. 2B). As expression of exogenous p120 recruits the transcription repressor Kaiso from the nucleus to the cytoplasm (Kelly et al., 2004; Spring et al., 2005), we next asked whether endogenous p120 stabilization correlates with a similar outcome. As expected, in the coordinate presence of MG132 and LiCl, Kaiso relocalization (Fig. 2E) was coincident with known effects upon p120 stabilization (Fig. 1C and Fig. 2A), although it is important to note that these inhibitors produce many effects aside from p120 stabilization. LiCl or MG132 alone produced a reproducible but considerably weaker Kaiso relocalization, consistent with their lesser protection of p120 relative to the combined treatment (data not shown). Together, data from Figs 1 and 2 suggest that endogenous p120-catenin is subject to some of the same regulatory processes established for β-catenin in the context of canonical Wnt signaling.
Association of destruction complex components with p120-catenin
Several in vivo and in vitro reports show that β-catenin associates with axin, β-TrCP, GSK3β and CK1α (Ikeda et al., 1998; Liu et al., 1999; Liu et al., 2002). Based upon the implied functional relationships we observed for p120-catenin (Figs 1 and 2), we tested whether p120 might likewise engage in such associations. As a considerable fraction of endogenous p120 is complexed with cadherins as opposed to being within an accessible signaling pool, co-immunoprecipitations were performed from Xenopus embryo extracts using ectopically expressed proteins. We first tested for the association of p120 with GSK3β, and, although weakly apparent in immunoblots, the p120–GSK3β complex was reproducibly resolved relative to controls and GSK3β KD (Fig. 3A, see also Fig. 4B). We next assessed axin, which interacts with multiple components of the destruction complex for β-catenin. In precipitations conducted in either direction (reverse precipitation not shown), axin associated with p120 (Fig. 3B; Kaiso negative control). p120 was also found in complex with CK1α, the priming kinase that acts upon β-catenin immediately before GSK3β (Fig. 3C; kaiso and C-cadherin serving as negative controls). Additionally, we immunoprecipitated the E3 ubiquitin ligase β-TrCP, resolving its association with p120 relative to negative controls (Fig. 3D). To test endogenous p120 binding to GSK3β, we immunoprecipitated endogenous GSK3β and resolved the p120–GSK3β complex in 293T and MDA-435 cells (Fig. 3E). Thus, in an in vivo context, p120-catenin directly or indirectly interacts with some of the key proteins known to regulate β-catenin stability.
The amino-terminal domain of p120 associates with GSKβ and CK1α
The amino-terminal domain of β-catenin provides the platform for its successive association with CK1α and GSK3β and harbors the serine and threonine sites that become phosphorylated (Liu et al., 1999; Orford et al., 1997). As a first step in assessing p120-catenin structure–function relationships, we searched for potential GSK3β phosphorylation sites. Three evolutionally conserved GSK3β phosphorylation regions were resolved, with one being amino-terminal and two residing in the Armadillo repeat domain. Interestingly, the amino-terminal region of p120 contained a 3DSX3SX2SX3S15 motif comparable to the GSK3β-sensitive site present in β-catenin (Fig. 5A) (Winston et al., 1999). To map the interaction of p120 with GSK3β and CK1α, we generated a series of p120 deletion constructs (a–f) (Fig. 4A) and performed co-immunoprecipitations from Xenopus embryo extracts. Constructs containing the amino-terminal region of p120 associated with GSK3β (fl, b, c and e), whereas p120 constructs lacking this region (a, d and f) did not (Fig. 4B). We then examined ubiquitylation of the p120 deletion mutants following expression in HeLa cells. Correspondingly, only those lacking the amino-terminal region of p120 were negative for ubiquitylation (Fig. 4C). Based on such findings, we next tested whether the CK1α priming kinase of β-catenin (primes for GSK3β) interacts with the analogous region of p120-catenin. Employing a similar co-immunoprecipitation strategy, we observed that CK1α associates with the amino-terminal domain of p120 (c), but not with a construct (d) lacking this region (Fig. 4D). Thus, in keeping with our earlier findings, these results suggest that p120 and β-catenin engage in shared protein interactions reflecting their similar biochemical and likely functional responses to Wnt signals.
p120 point mutant differs in destruction-complex responsiveness
To test whether the potential GSK3β or CK1α sites identified in the amino-domain of p120 are in fact relevant to its sensitivity to destruction complex components including GSK3β, CK1α and β-TrCP, we generated a compound point mutant (p1204SA: Ser→Ala: Ser6, Ser8, Ser11, Ser15) (Fig. 5A). Full-length p120-catenin (fl: isoform-1) versus p1204SA mutant were expressed in Xenopus embryos, and their protein stability examined in response to coexpressed GSK3β. Whereas p120-fl levels were consistently reduced upon GSK3β coexpression (Fig. 5B, compare lanes 1 and 2; see also Fig. 1A and Fig. 7A), the levels of the mutant were, by contrast, not lowered (Fig. 5B, compare lanes 3 and 4).
In fact, surprisingly, the levels of p120 were reproducibly increased upon p1204SA coexpression with GSK3β. Apparently, in producing p1204SA, a direct or indirect protective effect of GSK3β was unmasked. This possibility might also have been reflected in the unexpected lesser impact upon p120 levels and degradation of intermediate (10 pg or in some cases higher) GSK3β levels relative to lower (5 pg) GSK3β levels (Fig. 1A and supplementary material Fig. S1A,B). We hypothesize that, at intermediate or higher GSK3β levels, wild-type p120 might become susceptible to direct or indirect GSK3β effects that are protective in nature (phosphorylation or other events), whereas lower GSK3β activity largely acts at the negative-regulatory sites that we resolved. Indeed, distinct from a further predicted conserved site (Ser199) just upstream of the Arm domain of p120, other potential but more distal GSK sites have been reported in p120 (Xia et al., 2003). Evidence supporting the modification by GSK3 of such additional predicted sites was not obvious, however, in our in vitro kinase assays (supplementary material Fig. S2, and data not shown). Alternative possibilities include GSK3β phosphorylation of LRP, leading to inhibition of the destruction complex (either axin or GSK3β) and β-catenin stabilization (Cselenyi et al., 2008; Tamai et al., 2004; Zeng et al., 2008; Zeng et al., 2005). Whatever the underlying mechanistic explanation, mutation of the conserved four amino-terminal serine residues of p120 (p1204SA: Ser→Ala: Ser6, Ser8, Ser11, Ser15) protects p120 from GSK3β-mediated degradation, consistent with their relevance in determining p120 levels and activities.
In addition to candidate kinase sites, study of the primary sequence of p120 pointed to a conserved potential β-TrCP recognition motif, DSEXXS (Orford et al., 1997; Winston et al., 1999). In common with β-catenin, this motif resides adjacent to the putative amino-terminal CK1α–GSK3β phosphorylation site of p120 (Fig. 5A). Indeed, using a co-immunoprecipitation and immunoblotting approach, we consistently resolved the interaction of p120 with β-TrCP, whereas our phosphorylation mutant p1204SA exhibited considerably reduced β-TrCP co-association (Fig. 5C, compare lane 2 with lanes 4 and 6). These observations are consistent with the known biology of β-catenin, wherein phosphorylation is required for β-TrCP recruitment (Liu et al., 1999). Given this outcome, we suspected that p1204SA would prove less susceptible to ubiquitylation relative to wild type and tested this possibility in HeLa cells through coexpression with HA–ubiquitin. Indeed, while p120 displayed a strong ubiquitylation signal, poly-ubiquitylation upon p1204SA was greatly reduced. Interestingly, mono- or di-ubiquitylated forms showed lesser differences when comparing p120 versus p1204SA (Fig. 5D). Consistent with reduced ubiquitylation, pulse–chase data indicated that p1204SA has an extended half-life relative to native p120 (approximately 3 hours versus 1.5 hours) (Fig. 5E). At the functional level, p120-catenin has a number of in vivo activities, including the capacity to relieve Kaiso-mediated repression of target genes containing sequence-specific binding sites (KCS). Thus, we next examined whether the transcriptional activity of p1204SA is higher than that of wild-type p120, using Wnt-11 as a direct endogenous gene readout (Kim et al., 2004). Employing semi-quantitative RT-PCR (Fig. 5F), and real-time PCR (Fig. 5G), p1204SA promoted Wnt-11 expression (relief of kaiso-mediated repression) to a greater extent than wild-type p120-catenin, even though p1204SA was reproducibly expressed at lower levels (Fig. 5B, compare lanes 1 and 3). Furthermore, the p1204SA mutant proved more effective in promoting the expression of xSiamois (supplementary material Fig. S3), another direct p120–Kaiso (as well as β-catenin–TCF–LEF) gene target (Park et al., 2005). These data indicate that stabilized p120 (p1204SA) exhibits more potent gene-regulatory effects than wild-type p120-catenin, analogous to the β-catenin context (Lyons et al., 2004; Yost et al., 1996).
Shared mechanisms modulate the metabolic stability of multiple p120-catenin subfamily members
p120-catenin is the prototypical member of the p120-catenin subfamily that comprises p120-catenin itself, ARVCF-catenin, δ-catenin and p0071-catenin (McCrea and Park, 2007). These p120-catenin family members have certain shared features, such as a central Armadillo domain (nine repeats as opposed to the 12 in β-catenin or plakoglobin), their association with cadherin membrane-proximal regions (β-catenin and plakoglobin instead bind membrane-distally) and their regulation of small GTPases (β-catenin and plakoglobin apparently lack this functionality). Given our p120-catenin findings, we wished to know whether other p120 subfamily members might interact and respond to components of the destruction complex. Probing Xenopus embryo extracts, we first tested whether one of the main constituents of the destruction complex, axin, had the capacity to diminish ARVCF and δ-catenin levels. Serving as positive controls, β-catenin and p120 levels were reduced in the presence of axin (Fig. 6A and Fig. 1B). Likewise, ARVCF and δ-catenin were significantly decreased upon axin coexpression, whereas negative controls (xKazrin and EWS) showed no response to axin (Fig. 6A). We further asked whether ARVCF-catenin and δ-catenin associate with axin, as we demonstrated above for p120, and as known to occur for β-catenin. Indeed, in common with p120-catenin and β-catenin (positive controls), both ARVCF and δ-catenin associated with axin, relative to negative controls such as xDyrk and xKaiso (Fig. 6B and 6C). Although full-length δ-catenin exhibited less association with axin than did β-catenin, shorter forms of δ-catenin, which likely arose from endogenous proteolytic processing or incomplete translation, showed strong association with axin (Fig. 6C).
Our previous work showed that p120-catenin binds and is stabilized by the intracellular protein Frodo, allowing the p120-catenin–kaiso pathway to modulate certain Wnt target genes (Park et al., 2006). We thus tested whether depletion of Frodo might also have an impact upon β-catenin but found that, although p120 levels became lowered as expected, β-catenin levels were maintained (Fig. 6D). This was consistent with our subsequent finding that Frodo does not appear to bind to β-catenin, in contrast to its known interaction with p120 (Fig. 6E). In summary, our results indicate that components of the destruction complex known to act upon β-catenin are further involved in regulating the levels of p120 subfamily members and that Frodo has a more obvious impact upon the p120-catenin than the β-catenin trajectory of the Wnt pathway. These data leave open the possible existence of additional unknown modulators of p120 subfamily members in response to Wnt signals.
Canonical Wnt signals modulate p120-catenin isoform-1 levels
Generally, destruction of β-catenin is blocked upon activation of the canonical Wnt pathway. In addition to more recently reported mechanisms (Wu et al., 2008), canonical Wnt signals activate intracellular Dishevelled and LRP5 or LRP6, which inhibit the destruction complex by membrane recruitment of the core component axin, followed by LRP phosphorylation and further axin recruitment (Cselenyi et al., 2008; Zeng et al., 2008). Membrane recruitment of axin, together with associated GSK3β, permits enhancement of the cytoplasmic signaling pool of β-catenin by a mechanism that remains somewhat unclear, facilitating its nuclear entry and activation of Wnt target genes such as xSiamois, MYC and CCND1. Extending this mechanism conceptually to p120-catenin, we wished to confirm that Wnt signals increase p120 levels through inhibition of the destruction complex. We co-injected p120-catenin together with GSK3β in the presence versus absence of the Wnt8 ligand, which exhibits canonical and, in some contexts, non-canonical activity, and also assessed the impact of coexpressing Frodo with GSK3β. Although the role of Frodo in Wnt signaling remains unclear, it associates with Dsh (Gloy et al., 2002), and we reported earlier its positive modulation of p120-catenin levels (Park et al., 2006). As expected, Wnt8 as well as Frodo reproducibly protected p120-catenin from the negative effects of GSK3β (Fig. 7A, compare lanes 5 and 7 with lane 2). Furthermore, consistent with earlier results showing the partial capacity of axin to rescue developmental phenotypes arising from overexpression of p120 (Fig. 1B), we found that axin promoted p120-catenin degradation (Fig. 7A, compare lane 3 with 1). The ability of Wnt8 to counter axin-mediated destruction of p120 was shown to be dose dependent (Fig. 7C). Frodo, by contrast, protected p120 from the effects of axin (Fig. 7A, compare lane 8 with 3), as Frodo also protected p120 from the negative impact of coexpressed GSK3β (Fig. 7A, compare lane 5 with 2). At the low levels of GSK3β and axin that are expressed exogenously, effects upon endogenous β-catenin were not observed (whereas an effect of higher GSK3β expression upon endogenous β-catenin is shown in Fig. 7B).
To evaluate the protective effects of differing Wnt ligands, we expressed Wnt8, Wnt 11 or Wnt5a together with GSK3β and p120-catenin. Wnt8, as anticipated, reduced GSK3β-facilitated destruction of p120 (Fig. 7B). Wnt8 is generally thought to activate the canonical Wnt/β-catenin pathway, whereas Wnt11 and Wnt5a are portrayed as non-canonical in most contexts. However, Wnt11 and Wnt5a have also been shown to activate the canonical pathway in axis development of Xenopus embryos (Kofron et al., 2007; Tao et al., 2005). Indeed, in our hands, even endogenous β-catenin experienced a modest protective benefit when Wnt11 or Wnt5a was coexpressed with GSK3β (Fig. 7B). As anticipated, relative to Wnt8 and Wnt11, the non-canonical Wnt5a had a lesser capacity to protect HA–p120 from the effect of GSK3β (Fig. 7B).
Owing to the complexity of reported Wnt ligand and Frizzled receptor effects in Xenopus embryos during axis specification (among others), we employed a knockdown approach in conjunction with a cell line system to investigate the response of p120-catenin to other Wnt signaling components such as axin and LRP5 and LRP6. MDA-231, HeLa and 293T cells were used to test the effect of knocking-down axin-1 and axin-2 using proven siRNAs (Pan et al., 2008). Complementing our overexpression experiments conducted in Xenopus embryos, depletion of axin-1 and axin-2 stabilized endogenous p120-catenin (Fig. 7D). We further examined the effect of knocking down LRP5 and LRP6. Being positive modulators of canonical Wnt signaling, we would expect knockdown of LRP5 and LRP6 to produce an impact opposite to that of depletion of axin. Indeed, using a previously characterized siRNA directed against LRP5 and LRP6 (Pan et al., 2008), p120 levels were reduced in MDA435 and HeLa cells (Fig. 7F,G). Interestingly, the pp120 antibody, which detects most p120-catenin isoforms, revealed that the longer isoform of p120-catenin, presumably p120-catenin isoform-1, was more responsive to depletion of LRP5 and LRRP6. As isoform-2 and other higher isoforms initiate translation at a primary sequence position beyond the destruction motif of p120, these endogenous data appear to support the view that the most amino-terminal region of p120, present in isoform-1, favors its regulation by conical Wnt signals. This view was additionally supported through use of proteasome and CK1 (D4476) inhibitors, which resulted in increased levels of p120 isoform-1 (Fig. 2). Consistently, lesser effects on isoform-3 were seen following proteasome (MG132) or CK1 inhibition (Fig. 1C, Fig. 2A,C).
To further assay possible differences between the p120 isoforms, we expressed full-length xp120 (similar to human isoform-1) or an N-terminal deletion mutant (ΔNp120; similar to human isoform-4) in 293T cells. While a prior report pointed to the occurrence of two Xenopus p120 isoforms (iso1 similar to hp120 isoform-1; and iso2 similar to hp120 isoform-3), the Xenopus iso2 (similar to hp120 isoform-3) is not yet functionally characterized, whereas Xenopus isoforms similar to hp120 isoform-2 or isoform-4 have not yet been reported. As noted earlier, ΔNp120 lacks the potential GSK phosphorylation site S199, and it does not associate with either GSK or CK1 (Fig. 4B,D), and nor is there evidence for ubiquitylation (Fig. 4C). Thus, while human isoform-3 and -4 are functionally distinct (Roczniak-Ferguson and Reynolds, 2003; Yanagisawa et al., 2008), we largely used ΔNp120 (Xenopus surrogate isoform-4) in comparing Wnt pathway responses relative to those of isoform-1. When comparing Xenopus isoform-1 (full-length) and ΔNp120, only isoform-1 responded to CK1 or proteasome inhibition (Fig. 7H). Finally, we evaluated mouse isoform-1 versus isoform-3 upon GSK3 inhibition using another agent, BIO, in 293T cells. Isoform-1 became modestly but reproducibly stabilized, whereas isoform-3 appeared unresponsive under the same conditions (Fig. 7I). Inhibition of CK1similarly produced a reproducible, if modest, response in isoform-1 (Fig. 7I). Collectively, our data indicate that p120 and β-catenin, and probably additional p120 subfamily members, share regulatory mechanisms responsive to Wnt pathway activity. Furthermore, with respect to p120, it is isoform-1 that is most clearly modulated, in keeping with its capacity to associate with components of the destruction complex and to be phosphorylated by CK1 or GSK3.
The canonical Wnt pathway has been viewed as having β-catenin as the primary signal transduction mediator in response to Wnt pathway activation. While plakoglobin (γ-catenin), a member of the β-catenin subfamily, has been implicated in context-dependent Wnt gene regulation (Karnovsky and Klymkowsky, 1995; Sadot et al., 2000), little emphasis has been directed toward the possible roles of p120-catenin subfamily members.
p120 isoform-1 protein levels are subject to regulation by the destruction complex
Our recent work has indicated that the levels of p120 protein respond to canonical Wnt signals through Frodo, resulting in the modulation of certain Wnt/β-catenin target genes harboring both Kaiso- and TCF-binding sites (Park et al., 2006; Park et al., 2005). We reasoned that additional pathway components might act upon p120-catenin, prompting us to address potential mechanisms. In this report, we indicate that protein components involved with the physiologic destruction of β-catenin also promote p120-catenin degradation – with the prime examples being axin and GSK3β. In mapping studies, the amino-terminal region of p120-catenin was found to be required for conferring sensitivity to, and for interaction with, destruction complex proteins (CK1α and GSK3β). Importantly, we observed that the degradation mechanism applies principally to p120 isoform-1, which harbors amino-terminal GSK3 and CK1 sites that are not present in isoforms arising from translational initiation events at later primary-sequence positions. We mutated four potential phospho-residues predicted to comprise a CK1α and/or GSK3β consensus region. In contrast to the wild type, the levels of these constructs when coexpressed with GSK3β were no longer reduced. Consistent with the existence of an adjoining putative ubiquitylation site dependent upon prior CK1α or GSK3β phosphorylation, the p120 point-mutant lost most of its capacity to bind to β-TrCP and was largely free of ubiquitylation. Furthermore, we observed that components of the canonical Wnt pathway such as certain Wnt ligands and LRP5 and LRP6, in addition to the little-understood component Frodo, confer protective effects upon p120, whereas a core component of the destruction complex, axin, negatively modulates p120 in both Xenopus and cell line systems. Our results point to the view that the levels of p120 protein are regulated through mechanisms analogous to those operating on β-catenin in vivo. Intriguingly, other members of the p120-catenin subfamily, namely δ-catenin and ARVCF-catenin, were also responsive (apparent substrates) of the destruction complex. While we did not examine these latter catenins in depth, we note that, in common with β-catenin and p120, each contains a number of conserved potential GSK3β sites. Indeed, independent support for the responsiveness of δ-catenin to GSK3β and ubiquitylation has arisen recently from the laboratory of K. Kim (Oh et al., 2007). In that work, δ-catenin responded positively to GSK knockdown. Likewise, in keeping with our findings centered upon p120, δ-catenin became ubiquitylated in the presence of MG132, with ubiquitylation being reduced upon GSK inhibition. Mutation of a potential GSK3 phosphorylation site (Thr1078 residue) likewise resulted in less ubiquitylation, although interestingly, the site was mapped to the carboxyl-, as opposed to the amino-terminal, region of δ-catenin. Together with our data presented here and that published previously (Oh et al., 2009), we envisage that the degradation machinery of the canonical Wnt pathway modulates multiple members of the catenin family, with multiple catenins forming a Wnt-responsive network extending considerably beyond β-catenin alone.
Interestingly, an independent group recently reported the involvement of p120-catenin in canonical Wnt signaling (Casagolda et al., 2010). In response to Wnt-ligand, p120 was found to promote Dishevelled phosphorylation through the association of p120 with CK1ε and E-cadherin (cadherin-1). This was shown to result in increased stability of β-catenin. This study complements our own as cadherin-dissociated p120 (perhaps more than one isoform) represents a distinct signaling-pool origin versus that which we propose arises from Wnt-pathway inhibition of the destruction complex, which exhibits selectivity toward isoform-1. In all cases, our collective work supports the concept that p120-catenin participates in canonical Wnt signaling.
Although our results demonstrate responsiveness of p120 (and probably also ARVCF- and δ-catenin) to canonical Wnt pathway components and signals, central questions remain to be addressed. First is the issue of why, in addition to β-catenin, p120 and other subfamily members would be modulated coordinately. In the case of p120-catenin, we among other groups have gathered evidence that p120 acts in combination with β-catenin at shared and developmentally significant gene promoters, such as Xenopus Siamois and to a lesser extent Wnt11. In the case of Siamois, for example, gene activation occurs to an additively larger extent when de-repression by β-catenin of TCF or LEF occurs in combination with p120-mediated de-repression of the Kaiso transcriptional repressor (Kim et al., 2004; Park et al., 2005). Thus, the Wnt/p120-catenin and Wnt/β-catenin pathways could be required for coordinate regulation of Wnt signaling in context dependent manners. Requiring further study to resolve, independent work questions whether Siamois and Wnt11 are gene targets of Kaiso (Ruzov et al., 2009a; Ruzov et al., 2009b), whereas another independent study has supported the responsiveness of Siamois to kaiso depletion in the presence of weak Wnt pathway stimulation (Iioka et al., 2009). Whatever the outcome of this discussion, our results here, while directed toward upstream p120 regulation, continue to be consistent with Siamois and Wnt11 gene responsiveness to p120 and Kaiso (Fig. 5F and 5G, and supplementary material Fig. S3). We hypothesize that the Wnt/p120-catenin pathway, and probably the Wnt/ARVCF- and Wnt/δ-catenin pathways that require further elucidation, act in parallel with the Wnt–β-catenin signaling trajectory to modulate certain canonical Wnt gene targets (or cytoplasmic downstream effectors) in a context-dependent manner. Interestingly, δ-catenin has also been reported to modulate Kaiso function at gene promoters (Bienz, 1998; Rodova et al., 2004), and ARVCF appears present in some cell and tissue nuclei (Mariner et al., 2000). While there is no known action of ARVCF in the nuclear compartment, we have recently resolved its association with Kazrin, a structurally novel and incompletely characterized cortical or junctional protein, that we and others find additionally shuttles into the nucleus (Cho et al., 2010) (Borrmann et al., 2006; Groot et al., 2004; Sevilla et al., 2008). Together, a coordinate response of catenins to the destruction machinery of the canonical Wnt pathway could be imagined to result in a networked gene-regulatory outcome downstream of Wnt signals. With respect to potential upstream modulators of catenins, Frodo and the closely related Dapper functionally and physically interact with Dishevelled, and each appears to modulate Wnt signals positively or negatively in context-dependent manners (Cheyette et al., 2002; Gloy et al., 2002; Hikasa and Sokol, 2004; Teran et al., 2009). Furthermore, Frodo acts in downstream Wnt-signaling capacities (Hikasa and Sokol, 2004). In this study, we did not resolve a β-catenin–Frodo complex, although we confirmed the expected interaction of p120-catenin with Frodo (Park et al., 2006). As Frodo binds to Dishevelled, we conjecture that Frodo (and Dapper) exhibits some selectivity for the p120 signaling trajectory in upstream Wnt contexts and that analogous, but presently undefined, proteins might work together with Dishevelled to modulate the trajectory of other catenins.
As noted earlier, the initial p120-catenin transcript (pre-RNA) is subject to inter- and intra-exonic splicing events that allow for the generation of distinct polypeptides, as does the presence and use of four alternative translation start sites in humans (van Hengel and van Roy, 2007). Recent evidence indicates that differing isoforms confer distinct, or even opposing, functions with respect to cell adhesion and invasion (Liu et al., 2009; Miao et al., 2009; Yanagisawa et al., 2008). The p120 isoform-1, initiated from the first translational start site, encompasses the destruction motif subject to Wnt-pathway regulation and, correspondingly, is more responsive to depletion of LRP5 and LRP6, or the inhibition of proteasomes, GSK3 or CK1. In the context of ectopically expressed p120 isoforms, isoform-1 was considerably more sensitive than p120 isoform-3 or an N-terminal deletion mutant ΔNp120 (the latter similar to human isoform-4). Endogenous Xenopus isoforms equivalent to human isoform-2 and -4 have not yet been reported or characterized. However, in certain cell lines, we observed an additional response of isoform-3, which might conceivably arise from its association with isoform-1, indirect effects upon the cadherin–catenin complex or the existence of additional destruction motifs within p120. Although exogenous isoform-3 and ΔNp120 (similar to human isoform-4) displayed similarly unapparent or mild responses to inhibition of GSK, CK1 or the proteasome, these isoforms have distinctive features, including their relative distribution in differing intracellular localizations (Roczniak-Ferguson and Reynolds, 2003). Relative to other p120 isoforms, expression of isoform-1 is associated with cells having mesenchymal characteristics and also associated with epithelial cells that have undergone progression toward invasive and transformed cell states (Yanagisawa et al., 2008). We imagine that regulatory events determining the choice of translational initiation sites would provide another layer of control for the responsiveness of p120 to canonical Wnt signals. At present, little is known concerning how the particular translational initiation sites in the p120 transcript are selected or, similarly, what governs the differential splicing decisions of the p120 transcript, which together produce multiple p120 isoforms.
Apart from nuclear functions, p120 subfamily members play additional key roles at cell–cell junctions and other locations, regulating cadherin stability and trafficking (Davis et al., 2003; Ireton et al., 2002; Xiao et al., 2003), as well as modulating small GTPases (e.g. Rac1 and RhoA; see Introduction). Thus, in considering the potential consequences of Wnt signaling for members of the p120 subfamily, we must include possible effects that are not transcriptional in nature but, instead, represent more immediate effects upon cell adhesion, motility or cytoskeletal activity.
Materials and Methods
Standard recombinant DNA techniques were used to construct pCS2-based plasmids harboring Xenopus laevis: Myc–p120-catenin, HA–p120-catenin (Fang et al., 2004), Myc–β-catenin, HA–β-catenin, Myc–ARVCF, HA–ARVCF (Fang et al., 2004), Myc–δ-catenin and HA–δ-catenin (Gu et al., 2009). Myc–GSK3β-pXT7 and catalytically inactive mutant (K85R) (S. Y. Sokol, Mount Sinai School of Medicine) (Dominguez et al., 1995), HA–CK1α-pCS2 (Xi He, Harvard Medical School) (Liu et al., 2002), Myc–axin-pCS2 (P. S. Klein, University of Pennsylvania) (Hedgepeth et al., 1999), xWnt8 (S. Y. Sokol, Mount Sinai Scholl of Medicine) (Sokol et al., 1991), xWnt11 (R. Keller, University of Virginia) (Du et al., 1995; Ku and Melton, 1993) and xWnt5a (R. Harland, University of California, Berkley) (Wallingford et al., 2001) were kindly provided as indicated. GSK3β was subcloned into the pCS2-HA vector for coimmunoprecipitation. Generated previously were pCS2 plasmids harboring Myc–p120 deletion constructs, Myc–Frodo and HA–Frodo (Park et al., 2006). The quadruple p120-catenin point mutant (S6, S8, S11 and S15→4SA) was generated by PCR amplification of HA–p120-catenin (in vector pCS2) as a template with 5′ mutated primer as follows: p1204SA-F, 5′-GGAATTCATGGATGAGCCAGAGGCTGAAGCTCCGGCCGCTATATTGGCCGCAGTGAGAGCT-3′; and p1204SA-R, 5′-AAGGCCTGACACGCTGATCTTCAGCATCACCAAGATTCAGTGATCCTCCAGCACTTACGGA-3′.
Embryo culture, microinjections and in vitro transcription
Fertilization, embryo culture and microinjections were performed in accordance with standard methods (Fang et al., 2004). Embryos were microinjected with capped mRNA synthesized in vitro (mMessage mMachine, Ambion). All pCS2-based constructs were linearized by using NotI before in vitro transcription. Gastrulation phenotypes, principally failure of blastopore closure, were visually scored at embryonic stages 11–12.
Immunoprecipitation and immunoblot analysis
Immunoprecipitation and immunoblotting employed monoclonal antibodies directed against Myc (9E10), HA (12CA5), p120-catenin (mouse or human pp120, BD Transduction; mouse or human 6H11, Santa Cruz) and GSK3β (mouse, BD Transduction). Polyclonal antibodies included β-catenin and p120-catenin raised in our laboratory (Park et al., 2006) and Frodo provided by the laboratory of S. Y. Sokol. For immunoprecipitations, procedures were largely performed as described previously (Fang et al., 2004). Whole-embryo lysates were prepared using 0.5% Triton X-100 buffer [0.5% Triton X-100, 10 mM HEPES (pH 7.4), 150 mM NaCl, 2 mM EDTA, 2 mM EGTA], inclusive of a proteinase inhibitor cocktail (Sigma). Interference from the IgG heavy chain in immunoblot analyses was reduced by employing TrueBlot™ anti-mouse IgG IP Beads and Mouse TrueBlot ULTRA:horseradish peroxidase (HRP) anti-mouse IgG (eBioscience). Embryos were lysed in 0.5% Triton X-100 buffer (20 μl per embryo), in the presence of a proteinase inhibitor cocktail (Sigma). After centrifugation (5488 g, 20 minutes), the supernatant fraction was denatured in 5× SDS sample buffer (200 mM Tris-Cl [pH 6.8], 40% glycerol, 8% SDS, 0.08% Bromophenol Blue), followed by incubation at 95°C for 5 min. Half-embryo equivalents were resolved by SDS-PAGE and transferred onto nitrocellulose membranes. Immunoblotting and antibody incubations took place in 2% bovine serum albumin–TBST [25 mM Tris-HCl (pH 7.8), 125 mM NaCl, 0.5% Tween20]. SuperSignal WestPico (Pierce Biotechnology) reagents were utilized to detect HRP-conjugated secondary antibodies.
Mammalian cell culture and immunofluorescence staining
HeLa, 293T and MDA-MB-435 cells were purchased from ATCC and maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. MDA-MB-231 cells were kindly provided by D. Yu (UT MD Anderson Cancer Center). Using Fugene6, HeLa cells were transiently co-transfected with Myc–p120-catenin, along with Venus–sh-random or Venus–sh-GSK3 (Kim et al., 2006). Twenty four hours after transfection, cells were fixed with 4% PFA and immunostained with antibody against Myc (9E10). Fluorescence images (Nikon ECLIPSE E800 microscope) were recorded using SPOT advanced software. For immunofluorescence staining of kaiso (Fig. 2E), HeLa and MDA-231 cells were grown on glass coverslips. Cells were transiently transfected with constructs such as Myc–kaiso and treated with MG132 and/or LiCl. Cells were fixed with 4% PFA for 10 minutes, blocked with 5% goat serum in PBS and immunostained with antibody against Myc.
Transfection and RNA interference
siRNA oligonucleotide sequences directed against the transcripts encoding axin1 and 2 or LRP5 and 6 were gathered from published reports (Pan et al., 2008) and synthesized by Applied Biosystems. The proprietary duplex negative control siRNA was purchased from Ambion. Duplex oligonucleotides were directed against the target sequences: axin1, 5′-GGCGAGAGCCATCTACCGAAA-3′; axin2, 5′-GCAGACGATACTGGACGATCA-3′; LRP5, 5′-CCAACGACCTCACCATTGTCT-3′; and LRP6, 5′-AGACATTGTTCTGCAGTTAGA-3′. For transfection of siRNA alone into six-well plates, Lipofectamine RNAiMax (Invitrogen) was employed (50 pmol per well). When DNA plasmids were transfected along with siRNA, Lipofectamine 2000 (Invitrogen) was used. After transfection and incubation for 48–72 hours, cell lysates were collected using M-PER mammalian protein extraction reagent (Thermo Fisher Scientific) and proteins detected by immunoblotting.
HeLa-S3 cells were transiently transfected with (HA epitope tagged) pCS2-HA–p120 or pCS2-HA–p1204SA. After 24 hours, cells were washed and preincubated for 1 hour with Met- and Cys-free DMEM made to 10% in FBS. The medium was then replaced with fresh Met- and Cys-free medium containing 1.48×106 Bq of [35S]-Met or -Cys and incubated for 1hour. The cells were washed and incubated in complete medium, and 2 mg of cell lysates were immunoprecipitated with anti-HA-7 agarose (Sigma). The samples were subjected to 8% SDS-PAGE followed by autoradiography.
We thank Sergei Y. Sokol (Mount Sinai School of Medicine), Xi He (Harvard Medical School), Dihua Yu (UT MD Anderson Cancer Center) and Peter S. Klein (University of Pennsylvania) for providing helpful constructs or cell lines, and Mireia Dunach and Antonio Garicia de Herreros for sharing work before publication. This work was supported (P.D.M.) by the NIH (RO1GM52112) and a Texas Advanced Research Project Grant (003657-0008-2006). DNA sequencing and other core facilities were supported by a University of Texas M.D. Anderson Cancer Center National Cancer Institute Core Grant (CA-16672). J.Y.H. was supported from a William Randolf Hearst Foundation Student Research Award and a Developmental Award from the UT SPORE in Lung Cancer (P50 CA070907). Deposited in PMC for release after 12 months.
Supplementary material available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.067199/-/DC1
- Accepted September 6, 2010.
- © 2010.