P120-catenin is the prototypic member of a subfamily of Armadillo repeat domain (Arm domain) proteins involved in cell-cell adhesion. Interestingly, all members of the p120 subfamily have also been observed in the nucleus, suggesting that they have additional roles that have yet to be determined. Here, we have developed a novel model system for studying the nucleocytoplasmic shuttling capabilities of p120. We show that simultaneous deletion of both of the conventional nuclear localization sequences (NLSs) in p120 had little effect on its nuclear localization. Instead, the Armadillo repeat domain was essential, and deletion of Arm repeat 3 or Arm repeat 5 eliminated nuclear entry despite the presence of both NLSs. In addition, deletion of Arm repeat 8 resulted in constitutive nuclear localization of p120-3A in both E-cadherin-positive and -negative cell lines. Thus, the core shuttling functions are dependent on the Arm domain. We have also identified two regions within the N-terminus of p120 that modulate nuclear shuttling dynamics of p120. In cadherin-deficient cells, normal epithelial morphology could be restored by both WT-E-cadherin and p120 uncoupled E-cadherin mutants, but only WT-E-cadherin strongly reduced nuclear localization of p120. Moreover, structural changes in p120 that reduced its affinity for E-cadherin increased p120 nuclear localization. Thus, reduced shuttling in the presence of E-cadherin is principally due to sequestration, a condition that is probably dynamic under normal circumstances but completely lost in metastatic cells that have downregulated E-cadherin. Notably, Arm repeats 3 and 5 are necessary for both E-cadherin binding and nuclear translocation, indicating that these repeats have dual roles. Surprisingly, in the absence of E-cadherin there was significant colocalization of cytoplasmic p120 with elements of the tubulin cytoskeleton, particularly in perinuclear locations. Depolymerizing microtubules with nocodazole increased nuclear p120, whereas stabilizing tubulin with taxol reduced nuclear p120 and strongly increased p120 association with microtubules. Thus, p120 has intrinsic nucleocytoplasmic shuttling activity that is modulated, in part, by extrinsic factors such as cadherin binding and interactions with the microtubule network.
p120-catenin is the prototypic member of a subfamily of Armadillo (Arm) repeat domain proteins involved in cell-cell adhesion. It was originally identified as a Src substrate (Reynolds et al., 1989) and was subsequently shown to modulate cell-cell adhesion through interaction with the cytoplasmic domain of cadherins (Daniel and Reynolds, 1995; Ireton et al., 2002; Reynolds et al., 1994; Shibamoto et al., 1995; Staddon et al., 1995; Thoreson et al., 2000). The characteristic number and spacing of the Arm repeats define p120 subfamily members and suggest that they evolved from a common ancestor. The seven members of this subfamily fall into two functionally distinct groups. Members of the the first (p120, ARVCF, delta-catenin and p0071) interact via their Arm domains with classical cadherins. The second consists of the plakophillins (Plakophillins I, II and III), which interact via their amino-terminal head domains with desmosomal cadherins (Anastasiadis and Reynolds, 2000). The Arm domain consists of a series of linked ∼42 amino acid repeats folded in a helical conformation. One side of the helix forms a charged groove responsible for the vast majority of protein-protein interactions (Graham et al., 2000; Huber et al., 1997; Huber and Weis, 2001; Kobe, 1999). For example, the Arm repeats of β-catenin interact in different cellular locations with cadherins, the tumor suppressor adenomatous polyposis coli (APC) and lymphoid enhancer factor 1 (LEF-1)/T-cell factor (TCF) transcription factors (Shapiro, 2001). Although each of these interactions is mediated by different sets of Arm repeats, the sets overlap such that interactions with the various binding partners are mutually exclusive. Thus, by virtue of their unique interaction capabilities, Arm domains confer unusual functional flexibility and are found in a variety of proteins with complex and diverse functions.
A common trait among Arm domain proteins is their ability to translocate to the nucleus. Examples include importin-α (Gorlich et al., 1994; Weis et al., 1995), APC (Neufeld and White, 1997; Neufeld et al., 2000; Rosin-Arbesfeld et al., 2000), β-catenin (Behrens et al., 1996; Korinek et al., 1997; Molenaar et al., 1996; Morin et al., 1997) and small G protein GDP dissociation stimulator (SmgGDS) (Lanning et al., 2003). Indeed, most - if not all - Arm domain proteins shuttle between the cytoplasm and the nucleus, suggesting that Arm domains may have initially evolved as nucleocytoplasmic shuttling proteins and later evolved additional functions such as the adhesive roles exhibited by p120 and β-catenin. Thus, in addition to their roles in cell-cell adhesion, p120 and its relatives are likely to have role(s) in the nucleus.
The nucleocytoplasmic shuttling of proteins requires active nuclear import and export pathways best described for proteins containing classical nuclear localization (NLS) and exclusion (NES) sequences. Although some Arm proteins contain conventional NLSs, it appears that in many cases they can mediate their own import via mechanisms involving their Arm domains. For example, β-catenin lacks conventional NLSs but may bind to nucleoporins directly through its Arm repeats and appears to compete with the nucleoporin binding sites used by importin-β (Fagotto et al., 1998; Yokoya et al., 1999). APC also enters the nucleus by a mechanism that is dependant on its Arm domain (Galea et al., 2001; Rosin-Arbesfeld et al., 2000). Thus, it may be that Arm domains in general confer nucleocytoplasmic shuttling activity.
The role(s) of p120 subfamily members in the nucleus are not yet known. Certain plakophillins appear to be exclusively nuclear in cells lacking desmosomes, suggesting roles that are independent from adhesion (Mertens et al., 1996). Plakophilin II interacts and colocalizes with components of the largest subunit of RNA polymerase III (e.g. RPC155) and with the transcription factor TFIIIB, suggesting key nuclear functions (Mertens et al., 2001). Although p120 is usually found at cell-cell junctions, nuclear p120 has been observed in neural crest cells deficient in N-cadherin or connexin 43 (Xu et al., 2001), in E-cadherin-deficient carcinoma cells and in fibroblasts (van Hengel et al., 1999). The increased nuclear presence in E-cadherin-deficient cells suggests a role for nuclear p120 in enhancing the metastastic phenotype associated with E-cadherin downregulation. A potential functional clue is that p120 binds directly to Kaiso (Daniel and Reynolds, 1999), a BTB/POZ family zinc finger protein that represses transcription and binds preferentially to methylated DNA (Daniel et al., 2002; Prokhortchouk et al., 2001). The functional consequence of p120 interaction with Kaiso is still unknown but the interaction of Kaiso with its DNA target sequences appears to be inhibited by p120 (Daniel et al., 2002).
p120 is expressed in cells as multiple isoforms derived by alternative splicing (see Fig. 2A) (Keirsebilck et al., 1998). Isoforms 1-4 use different ATG start sites that produce proteins that differ by the length of the amino terminal domain. In addition, other splicing events give rise to exons A, B and C. The various isoforms are differentially expressed in various cell types, but exact functional consequences are largely unknown. Several domains have been identified in p120, including the coiled-coil domain found only in isoform 1, and the phosphorylation domain found in all isoforms except type 4. The p120 isoforms 1-3 contain two putative NLS sequences, whose roles are not yet clear (Aho et al., 2002; Mariner et al., 2000). The first is a bipartite NLS embedded in the N-terminal phosphorylation domain and is thus not present in isoform 4 (Aho et al., 2002). The second is localized within a loop structure in Arm repeat 6 and is conserved among all p120 isoforms and family members (Anastasiadis and Reynolds, 2000). In addition, exon B contains an NES that can mediate efficient nuclear exclusion of p120 isoforms (van Hengel et al., 1999), but many cells express p120 isoforms that do not contain exon B.
In carcinoma cells that have downregulated E-cadherin, there is a significant elevation of p120 levels in the cytoplasm and nucleus, a phenomenon that may play a role in the metastatic phenotype. Here, we have identified specific intrinsic and extrinsic factors that regulate p120 nucleocytoplasmic shuttling and clarify the mechanisms underlying this activity.
Materials and Methods
Cell culture and cell treatments
A431, HT 29, A431D, SKBr3, MiaPaca II, L-cell, Hela and MDA435 cells were grown in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% of fetal bovine serum (FBS), and 100 U/ml penicillin and 100 mg/ml streptomycin. Phoenix 293 cells were grown in DMEM supplemented with 10% heat-inactivated FBS. To block nuclear export, cells were incubated for 6 hours in the presence of 10 ng/ml leptomycin B (LMB) (Sigma). The microtubule dynamics were disrupted by incubating cells in the presence of nocodazole (5 μg/ml, Calbiochem) or taxol (10 μM, Sigma) for 8 hours.
All deletion mutants were generated by ExSite directed mutagenesis using Pfu Turbo (Stratagene) and subcloned into a shuttle vector (pMS) from which all constructs were moved into a retroviral expression vector, pLZRS-MS-Neo. The sites of NLS mutations were as follows (based on p120-1A): NLS 1 (amino acids 306-321), NLS 2 (amino acids 615-629) and NLS 1&2 (amino acids 306-321 and 615-629). The exact sites of Arm deletions were previously reported (Ireton et al., 2002). E-cadherin and p120-uncoupled E-cadherin 764AAA mutant (Thoreson et al., 2000) were expressed from pLZRSMS-Neo as previously described (Ireton et al., 2002). All deletion mutants were sequenced to confirm accuracy.
Viral production and retroviral transduction
For viral production Phoenix 293 packaging cell lines were transfected by calcium-phosphate precipitation and selected with 5 μg/ml puromycin as previously described (Ireton et al., 2002). To harvest virus, Phoenix cells at 75% confluence were incubated overnight with fresh media. The following day media were collected and filtered through a 45 μm syringe filter. For retroviral tranduction, sparsely plated cells (40% confluency) were incubated overnight with freshly harvested virus containing 4 μg/ml of polybrene. To generate stable cell lines, cells transduced with the LZRS-MS-Neo virus were selected with Geneticin (G418) and used as a polyclonal population.
Cells were plated at 4×104 cells per well in six-well dishes containing glass coverslips and grown for 24 to 48 hours. The cells were fixed in freshly made 4% paraformaldehyde followed by two 5 minute washes in PBS/glycine. The cells were then permeabilized with 0.2% Triton-X100 in PBS for 5 minutes and blocked with 3% milk before incubation with primary antibodies to p120 (8D11, 2 μg/ml; pp120, 0.5 μg/ml, Transduction Laboratories), E-cadherin (0.5 μg/ml, Tranduction Laboratories) or α-tubulin (1:1000, Sigma). The signal was detected with isotype specific Alexa 488- and Alexa 594-conjugated secondary antibodies (1:600, Molecular Probes). Slides were mounted using the ProLong antifade (Molecular Probes) and examined with a Zeiss Axioplan 2 microscope using the 63× oil immersion objective. Digital images were captured with an ORCAER camera and selected images were processed by deconvolution microscopy using the OpenLab software (Improvision).
Scoring of nuclear staining intensity
A scoring method was established to facilitate quantitative comparisons of various p120 variants with respect to nuclear localization. Random fields were selected and the cells assigned to one of three groups on the basis of their staining intensity in the cytoplasm and nucleus. The first group consisted of cells in which staining was brighter in the cytoplasm than in the nucleus (predominantly cytoplasmic, PC), the second group consisted of cells that stained equally in the nucleus and cytoplasm (NC), and the third group consisted of cells that stained brighter in the nucleus than the cytoplasm (predominantly nuclear, PN). The cell counts were expressed as a percentage of total cells counted.
A novel model system for studying nuclear localization of p120
We and others have observed significantly elevated levels of p120 in the nucleus of E-cadherin-deficient cells, many of which are highly metastatic (Mariner et al., 2000; van Hengel et al., 1999). To generate model systems for studying p120 nuclear localization and its consequences, we initially examined (by immunofluorescence microscopy) endogenous p120 localization in several well established cell lines. In a wide range of cadherin negative cell lines, such as L-cells (fibroblast), SkBr3 (breast carcinoma) and MiaPacaII (pancreatic carcinoma), p120 was predominantly localized in the cytoplasm and at lower levels in the nucleus (Fig. 1A,B,D). Interestingly, in some cell lines that were E-cadherin-negative but which expressed other types of cadherins such as N-cadherin (Hela, Fig. 1C,G), a substantial amount of nuclear p120 was nonetheless present, suggesting that such cadherins may not be as effective as E-cadherin in sequestering p120 at the membrane. In addition, the intensity of nuclear p120 staining varied from cell to cell and was generally more robust in sparsely plated cells. Treatment with Leptomycin B (LMB), a fungal drug that inhibits the chromosome maintenance region 1 (CRM1)-mediated nuclear export, resulted in marked retention of p120 in the nucleus (Fig. 1E,F,G,H). These data show that the cytoplasmic and nuclear populations of p120 in a wide variety of cell lines are dynamic and subject to nuclear shuttling activity.
The same results were observed using several p120 monoclonal antibodies and a variety of different fixative conditions. Elevated cytoplasmic and nuclear p120 were also observed in other E-cadherin-deficient cell lines (e.g. SW480, MDA231, PancI and A431D) (not shown).
On the basis of these observations, we have devised a novel model system for structure-function analysis of p120 nuclear localization. E-cadherin-deficient cells were used because of the enhanced levels of p120 in the nucleus. To allow the exclusive tracking of exogenously introduced p120, we chose human cell lines as recipients for murine p120 cDNA constructs. Human and murine p120 are 96% identical at the amino acid level and are functionally indistinguishable. We previously generated and described a p120 monoclonal antibody (mAb 8D11) that binds murine p120 with high affinity but fails completely to recognize human p120 (Wu et al., 1998). Thus, we can selectively identify and localize the products of exogenous murine p120 constructs without detection of the endogenous human p120. An important feature of the system is the relatively low levels of the exogenously expressed p120. The LZRS-based retroviral vector over-expresses murine p120 by only ∼3-5-fold over endogenous levels. Thus, p120 mutations that alter the shuttling process are likely to be physiologically relevant because p120 is not artificially forced into the nucleus by overwhelming the shuttling systems.
Nuclear localization of p120 is independent of conventional nuclear localization sequences
The structure of p120 is indicated in Fig. 2A. Different p120 isoforms are generated by the use of alternative translational start sites and of alternatively spliced exons A, B and C. Several functional domains have been identified in p120, including the coiled-coil domain (CC), the phosphorylation domain (PD) and the armadillo domain. p120 has two putative classical NLS sequences, one embedded in the PD and the other in Arm repeat 6. Their role(s) are unclear because each has been individually deleted with little or no effect on p120 nuclear localization (Aho et al., 2002; Anastasiadis et al., 2000). It is possible, however, that the presence of a single NLS is sufficient to promote nuclear entry. To definitively address this issue, we simultaneously deleted both NLSs in p120-3A, an isoform that normally enters the nucleus efficiently. Fig. 2B shows the location of p120-3A (i) and p120-3A mutants after deletion of the N-terminal NLS (ii), the Arm 6 NLS (iii) or both NLSs simultaneously (iv). None of these mutations eliminated p120 nuclear translocation. We conclude that p120 nuclear localization is independent of conventional NLSs and is instead regulated by alternative mechanisms.
The N-terminus modulates the subcellular distribution of p120
To identify sequences that control nuclear shuttling, detailed p120 structure-function analyses were performed. The amino-terminal end was examined first and a scoring system (described in Materials and Methods) was developed to quantify patterns of nuclear and cytoplasmic p120 staining in cells (Fig. 3). Naturally occurring murine p120 isoforms were introduced into A431 cells (parental cell line containing E-cadherin) or A431D cells (Lewis et al., 1997), (a variant A431 cell line lacking cadherins). p120 isoforms 1A, 3A and 4A differ only in their N-terminus (Fig. 3A). When introduced into parental A431 cells, all isoforms bound E-cadherin and localized to junctions (Fig. 3B i, ii and iii). In the E-cadherin-deficient A431D cells, however, the isoforms behaved differently from one another. Representative examples are shown (Fig. 3B, iv, v and vi).
The overall distribution of cells with cytoplasmic and/or nuclear p120 localization was quantified (Fig. 3C) so as to accurately represent localization of p120 in cell populations. p120-1A was predominantly cytoplasmic in 100% of infected cells (as illustrated in Fig. 3Biv), whereas p120-3A was predominantly nuclear in 41% of cells, and evenly nuclear and cytoplasmic in 59% (as in Fig. 3Bv). p120-4A had a random distribution over the cytoplasm and nucleus (Fig. 3Bvi), with 82% of cells showing equal nuclear and cytoplasmic staining and 18% of cells showing predominant cytoplasmic staining.
The ability of the different isoforms to shuttle between the nucleus and cytoplasm was assessed by blocking nuclear export with LMB (Fig. 3Bvii-ix and Fig. 3C). After treatment with LMB p120-1A localization shifted from predominantly cytoplasmic to predominantly nuclear in 34% of cells and evenly nuclear and cytoplasmic in 66% of cells. Likewise, p120-3A cells scored in the predominantly nuclear p120 category increased from 41% to 92%. Interestingly, p120-4A was relatively insensitive to LMB (Fig. 3B, compare vi and ix, and Fig. 3C).
Because p120-4A lacks the amino-terminal regulatory domain (containing most p120 phosphorylation sites), we revisited these experiments using mutants of p120-3A that retain the amino-terminal regulatory domain but contain phenylalanine (F), or alanine (A) substitutions at previously identified Y or S/T phosphorylation sites, respectively. p120-3A constructs lacking the seven known tyrosine phosphorylation sites in this region (p120-3A-7F) or the five serine/threonine phosphorylation sites mapping to this region (p120-3A5A) still localized in the nucleus (data not shown). In addition, larger amino acid deletions spanning the phosphorylation domain (the following amino acids were deleted, based on p120-1A nomenclature: 234-258, 259-280, 281-305 and 306-321) were generated and also found to localize efficiently to the nucleus (data not shown). Thus, it appears that the major phosphorylation sites within the regulatory region, and indeed extensive segments within the phosphorylation domain, are not major factors in determining p120 nuclear localization.
To exclude the possibility of cell-type-dependent effects, these experiments were repeated in SW480, MDA 435 and SKBr3 cells, where similar observations were made (not shown). The LMB-induced effects were striking in SW480 and A431D cells, and modest in SKBr3 and MDA435 cells. These results show that the N-terminus regulates the subcellular distribution and nuclear shuttling of the p120 isoforms in the E-cadherin-negative cell lines. However, absence of the N-terminus (i.e. p120-4A) does not prevent p120 from entering the nucleus, suggesting important role(s) for the Arm domain.
Specific Armadillo repeats are necessary for p120 nuclear localization
To determine the role of the Arm domain, each of the ten Arm repeats was individually deleted in p120-3A and the mutants localized in A431D cells in the absence and presence of LMB. Representative staining is shown for each mutant in Fig. 4A and quantification in Fig. 4B. Deletion of Arm repeats 3 and 5 completely abolished nuclear localization of p120-3A in all cells (Fig. 4A, iii and v), and blockade of export by incubation with LMB did not rescue the defect (Fig. 4A, viii and x). Thus, repeats 3 and 5 appear to be necessary for nuclear import, despite the presence of both putative conventional NLSs.
Deletion of arm repeat 8 significantly increased levels of p120 in the nucleus (Fig. 4A, xiii) and there was little further effect following LMB treatment (Fig. 4A, xviii). Sequence analysis revealed a putative consensus NES that might explain this effect. Alternatively, deletion of Arm 8 could uncouple p120 from a cytoplasmic anchor and result in higher steady state nuclear p120 levels.
Deletion of repeats 1, 2, 4, 7, 9 and 10 significantly decreased nuclear levels of p120 but this effect could at least partially be rescued by blocking nuclear export with LMB, suggesting that although these repeats contribute to efficient nuclear import of p120, they are not essential. Deletion of arm repeat 6 was the only Arm repeat mutation that had no apparent effect on nuclear levels of p120-3A (Fig. 4A, xi and xvi), and this mutant behaved similarly to WT-p120-3A (compare with Fig. 3B, v and viii, and Fig. 3C).
Binding to E-cadherin inhibits nucleocytoplasmic shutting of p120
To examine the role of E-cadherin in the regulation of nucleocytoplasmic shuttling of p120 we treated several E-cadherin-expressing cell lines with LMB. As shown in Fig. 5A, p120 is localized primarily at the membrane in A431 (Fig. 5Ai) and HT 29 cells (Fig. 5Aiii), and treatment with LMB did not significantly enhance nuclear staining of p120 (Fig. 5A, ii and iv), showing that E-cadherin strongly suppresses p120 translocation to the nucleus. However, it is unclear whether this is because of physical sequestration, or other events associated with the ability of E-cadherin to organize epithelial morphology.
To distinguish between physical sequestration and the tissue-organizing capabilities of E-cadherin as regulators of p120 nuclear localization, we first compared p120 localization in SKBr3 cells transduced with either WT-E-cadherin or E-cadherin (764AAA). Both constructs can rescue cell-cell adhesion and morphology (Fig. 5Ciii and Fig. 5Diii), but E-cadherin (764AAA) contains a triple alanine mutation in the p120 binding domain and is therefore physically uncoupled from p120. SKBr3 cells were used in these experiments because nucleocytoplasmic shuttling of p120 is very robust in this cell line. In the absence of E-cadherin (e.g. parental SKBr3 cells), p120 localized to both the nucleus and cytoplasm, and LMB increased nuclear p120 levels substantially (Fig. 5B, compare i and ii). SKBr3 cells transduced with E-cadherin formed cell-cell contacts characteristic of epithelial cells (Fig. 5Ciii). p120 was strongly recruited to junctions and largely excluded from the nucleus (Fig. 5Ci). As expected, treatment with LMB did not significantly alter p120 distribution (Fig. 5C, compare panels i and ii), indicating that the nucleocytoplasmic shuttling of p120 is strongly reduced by re-expression of E-cadherin. Expression of the p120-uncoupled E-cadherin (764AAA) also rescued epithelial morphology, but p120 remained stranded in the cytoplasm (Fig. 5D, i and iii). Despite the rescue of epithelial junctions and morphology, p120 nuclear localization and its nucleocytoplasmic shuttling activity was essentially the same as that of the E-cadherin-negative parental SKBr3 cells (Fig. 5D, i and ii). Similar results were obtained in A431D cells and L-cells. These data indicate that sequestration, rather than morphologic cues, is the main mechanism limiting p120 nucleocytoplasmic shuttling.
E-cadherin binding and nuclear import required overlapping sets of p120 Arm repeats
We showed previously that p120 Arm repeats 1-5 and 7 are necessary to bind E-cadherin (Ireton et al., 2002). Interestingly, comparison of Arm repeats necessary for E-cadherin binding and those necessary for nuclear import revealed a considerable overlap (Fig. 6A). Because binding to E-cadherin significantly limits p120 nucleocytoplasmic shuttling, we asked whether p120 mutants that fail to bind E-cadherin could shuttle efficiently in cells where E-cadherin is expressed normally. As might be expected, when E-cadherin-uncoupled p120 mutants were expressed in E-cadherin-positive A431 cells, they exhibited nucleocytoplasmic shuttling characteristics that were similar to those observed in E-cadherin-negative cells. Most of this data is not shown because the observations are almost identical to those in Fig. 4, which shows the localization of these mutants in E-cadherin-negative cells. The results for Arm 3 (Fig. 6B, i, ii, v and vi) and Arm 5 (Fig. 6B, iii, iv, vii and viii) deletion mutants, however, illustrate the point that repeats 3 and 5 are necessary for both E-cadherin binding and nuclear translocation. The cells shown are tightly adherent and contain high levels of functional E-cadherin. Nonetheless, p120 mutants containing deletions in Arm repeats 3 or 5 are both stranded in the cytoplasm and unable to translocate to the nucleus. Thus, the principle mechanism by which E-cadherin limits WT-p120 nuclear translocation is by sequestration, but E-cadherin binding is also likely to block other unknown interactions mediated by Arm repeats 3 and 5, which are clearly essential for nuclear translocation, whether cadherins are present or not.
E-cadherin binding is not sufficient to prevent nucleocytoplasmic shuttling
Interestingly, in the E-cadherin-positive A431 cells, the ability to bind E-cadherin was not by itself sufficient to block nucleocytoplasmic shuttling for several of the p120 Arm repeat deletion mutants. Fig. 7 contrasts the fate of WT-p120-3A with the Arm 8 and Arm 10 deletion mutants when expressed in A431 cells with and without LMB. WT p120-3A was sequestered at junctions and was not significantly affected by nuclear export blockade, confirming that E-cadherin can efficiently inhibit nuclear shuttling of p120 (Fig. 7A,B). By contrast, the Arm 8 mutant localized at the membrane, the cytoplasm and prominently to the nucleus (Fig. 7C), despite the fact that it can bind E-cadherin, and the nuclear localization was not affected by LMB treatment (Fig. 7D), further implicating the loss of a nuclear export function for this mutant. The Arm 10 mutant was particularly striking because it was localized primarily at junctions and at low levels in the cytoplasm in the absence of LMB (Fig. 7E), but localized in the nucleus when LMB was added (Fig. 7F). Previous work showed that the affinity of the Arm 10 deletion mutant for E-cadherin is slightly reduced relative to that of wild-type p120 (Ireton et al., 2002). The increased nuclear presence of this mutant after LMB treatment suggests that even slightly reduced affinity for E-cadherin can substantially increase the pool of p120 available for nucleocytoplasmic shuttling.
Microtubule status modulates nuclear shuttling of p120
Despite the strong influence of E-cadherin, there were other factors that appeared to influence the localization in the nucleus. In several E-cadherin-negative cell lines, p120-1A and p120-3A localized at the nuclear periphery and formed bright rings around the nucleus. To identify structures associated with p120 at this site, we selected markers for microtubules, intermediate filaments, the Golgi apparatus and the nuclear envelope, all of which show various degrees of perinuclear staining. Surprisingly, a significant portion of p120-1A (not shown) and 3A (Fig. 8A, i and ii) associated with the perinuclear staining colocalized with α-tubulin. p120-1A and 3A also partially colocalized with the nuclear envelope (mAb 414, not shown) but were not significantly colocalized with endoplasmic reticulum (GRP/Bip 78) or Golgi (GM13) markers. In A431D cells treated with LMB, p120-1A relocalized from the cytoplasm to the nucleus and the perinuclear p120 staining was strongly reduced (Fig. 8B, i-v). These data suggest that the reduced perinuclear staining is at the expense of increased nuclear p120 resulting from the LMB-induced export blockade. Thus, the perinuclear and nuclear pools probably exchange, and a pool of p120 may dock on these perinuclear structures under certain conditions.
To clarify the relationship between p120 and microtubule staining, we colocalized p120 and microtubules in the presence and absence of the microtubule-disrupting drug nocodazole (Fig. 8C). Before microtubule disruption, there was extensive p120 and tubulin colocalization (Fig. 8C, i and ii). Nocodazole treatment reduced α-tubulin staining (iv) and induced significant translocation of p120-1A to the nucleus (Fig. 8Ciii). We then repeated these experiments in the absence and presence of taxol, a microtubule-stabilizing drug (Fig. 8D). Panels i and ii show significant colocalization of p120 and α-tubulin in A431D cells overexpressing p120-3A, as well as prominent p120 staining in the nucleus. After stabilization with taxol, the microtubule immunoreactivity increased substantially and the microtubules formed prominent basket-like structures on either side of the nucleus (Fig. 8Diii). Interestingly, p120-3A staining in the nucleus was strongly reduced, and was replaced by significant colocalization with the stabilized microtubule structures (Fig. 8D, iii and iv).
These experiments indicate a close relationship between cytoplasmic p120 and the microtubule system, and suggest that the microtubule system may modulate p120 nucleocytoplasmic shuttling. During the course of these experiments, it became clear that p120 and microtubules interact significantly in many cells and under many circumstances. Collaborative studies have now been conducted to address issues such as whether p120 binds microtubules directly or indirectly, and the potential roles for these interactions. These issues are beyond the scope of this manuscript and are being submitted elsewhere (P. Z. Anastasisadis et. al., unpublished).
p120 actively shuttles between the cytoplasm and nucleus in E-cadherin-deficient cells. Here, we show that p120 nuclear import is not dependent on its conventional NLSs, but is instead mediated by the Arm domain and by sequences in the amino-terminus that include but do not require the amino-terminal NLS. The irrelevance of the conventional NLSs suggests that p120 nucleocytoplasmic shuttling is independent of the classical NLS shuttling mechanisms mediated by α- and β-importins. The extensive panel of p120 mutants used in this study provides a unique insight into the roles of individual repeats in the context of the overall shuttling function of the Arm domain. Different repeats mediate the various functions necessary for coordinated nuclear import and export. Repeats 3 and 5 were necessary for nuclear import, whereas Arm 8 deletion significantly increased p120 levels in the nucleus. Overall, our data point to an intrinsic ability of p120 to shuttle between the nucleus and cytoplasm, and suggest several mechanisms that further regulate this role.
Nuclear import and export of proteins is a highly regulated process that is best understood for members of the importin (also known as karyopherin) family of proteins. Importin-α and -β are the prototypical import receptors that are essential and sufficient for the import of proteins bearing classical consensus NLSs (Gorlich et al., 1995; Moroianu et al., 1995a; Radu et al., 1995). Interestingly, importin-α itself contains an Arm domain, which mediates the binding to NLSs on cargoes (Adam and Gerace, 1991; Weis et al., 1995). Importin-β contains so-called HEAT repeats, which target the importin- α/β complex and its NLS-containing cargo to the nuclear pore (Gorlich et al., 1995; Gorlich et al., 1996; Moroianu et al., 1995a; Moroianu et al., 1995b). p120 has two sequences that conform to consensus NLSs, but mutation of these residues separately and together had no effect on the nuclear levels of p120. Instead, deletions of Arm repeats 3 and 5 eliminated p120 nuclear localization, even though these constructs leave intact both of the conventional NLSs. Thus, nuclear import of p120 requires an intact Arm domain and is probably independent of the classical mechanism associated with binding to importin-α. A role for importins cannot be completely ruled out given that many isoforms of importin-α have been identified, and in some cases the substrate specificities are different from conventional NLSs (McBride et al., 2002; Miyamoto et al., 1997; Nadler et al., 1997; Sekimoto et al., 1997). In addition, nuclear import can be directly mediated by various importin-β isoforms (Nagoshi et al., 1999; Nagoshi and Yoneda, 2001; Takizawa et al., 1999; Xiao et al., 2000). Nonetheless, the dependency of p120 nuclear import on its Arm domain is consistent with recent data on other Arm domain proteins (e.g. β-catenin and APC) (Fagotto et al., 1998; Galea et al., 2001; Rosin-Arbesfeld et al., 2000; Yokoya et al., 1999) and adds to a growing consensus that Arm repeat domains have intrinsic shuttling activity. Owing to structural similarity between the Arm and HEAT repeats, it has been proposed that β-catenin mediates its own nuclear import by directly interacting with the nucleoporins (Fagotto et al., 1998; Yokoya et al., 1999). Taken together, our data strongly suggest that p120 nuclear import is also independent of classical import mechanisms and mediated principally by its Arm domain.
Protein export out of the nucleus is mediated by members of the importin-β family, of which CRM1 is the best described. CRM1 recognizes leucine-rich NES's and its activity is selectively inhibited by LMB (Fornerod et al., 1997; Fukuda et al., 1997; Ossareh-Nazari et al., 1997; Stade et al., 1997). An LMB-sensitive classical NES has previously been characterized in the alternatively spliced exon B and shown to reduce nuclear levels of the p120-3AB isoforms (van Hengel et al., 1999). However, we find that p120 isoforms 1A and 3A, both of which lack exon B, also accumulate in the nucleus when cells are treated with LMB. This effect suggests the presence of a novel NES sequence or an indirect effect on p120 mediated through a binding partner that is sensitive to CRM1-mediated export from the nucleus. Consensus NES's resembling the one identified in exon B are present in p120 Arm repeats 2, 7, 8 and 10, but deletion of repeats 2, 7 and 10 decreased nuclear levels of p120. Deletion of Arm repeat 8, however, strongly increased nuclear levels of p120 in both E-cadherin-negative and -positive cells, suggesting that the consensus NES in repeat 8 is crucial.
Our data suggests at least two regulatory regions in the N-terminus of p120 that affect the nuclear steady-state levels of p120 in different ways. p120-1A, which contains both the coiled-coil and the phosphorylation domain, localizes primarily to the cytoplasm (in cadherin-deficient cells) and is observed in the nucleus only after nuclear export blockade with LMB. By contrast, p120-3A, an isoform that lacks the coiled-coil domain, is both cytoplasmic and nuclear. The different nuclear shuttling dynamics of isoforms 1A and 3A suggests that the coiled-coil domain either impedes nuclear import of p120-1A or enhances its nuclear export. As coiled-coil domains frequently mediate protein-protein interactions, this domain may function by anchoring p120 to unknown cytoplasmic binding partners, or by interacting with proteins that promote export. Alternatively, it may promote conformational changes that reduce interactions with import machinery.
The second regulatory region lies within the N-terminus of p120-3A that contains the phosphorylation domain, but lacks the coiled-coil domain. Our data shows that although the N-terminus of p120-3A is not essential for nuclear localization, it strongly enhances the process. An obvious possibility is that phosphorylation may regulate p120 interaction with proteins that mediate nuclear import. However, mutation of tyrosine and serine/threonine phosphorylation sites within the phosphorylation domain had no effect on the steady-state levels of p120 in the nucleus, suggesting that the N-terminus may regulate p120 nuclear levels independently of its phosphorylation state. A more detailed structure-function analysis of the N-terminus will be required to determine the minimal region necessary for modulating nuclear steady-state levels of p120-3A. A previous study showed that deletion of the phosphorylation domain decreased nuclear levels of p120, whereas deletion of the NLS embedded in the phosphorylation domain had no effect (Aho et al., 2002). These results are mostly consistent with ours, except for an observation that p120-1A localizes to the nucleus in the absence of LMB. This latter discrepancy probably reflects the use in their study of very high level p120 expression by transient transfection from CMV-based promoters, although cell-type-specific differences might also contribute. Although exact roles of the various p120 isoforms are unknown, it is interesting to note that none of the splicing events affect binding to cadherins, but most of them affect nuclear translocation. For example, exon B contains an active NES (van Hengel et al., 1999), and isoforms 1, 3 and 4 have significantly different nuclear shuttling abilities.
Our data clearly indicate that E-cadherin negatively regulates nuclear levels of p120. Re-expression of E-cadherin in cadherin-negative cell lines inhibits p120 nucleocytoplasmic shuttling and sequesters p120 at the membrane. By contrast, re-expression of an E-cadherin mutant unable to bind p120 (i.e. E-cad-764AAA) has no effect on the nucleocytoplasmic shuttling of p120. Although E-cad-764AAA adhesiveness is not as strong as WT-E-cadherin (Thoreson et al., 2000), it is nonetheless quite effective at restoring epithelial-like cell-cell junctions and morphology. Thus, the inhibition of p120 nuclear translocation by E-cadherin is primarily due to sequestration rather than its ability to organize epithelial morphology. Interestingly, our structure-function analysis of the p120 armadillo domain revealed a significant overlap between the regions of p120 necessary for nuclear import and those necessary for binding to E-cadherin. In particular, repeats 3 and 5 are necessary for both cadherin binding and nuclear import. Thus, E-cadherin binding may also mask the binding of cytoplasmic proteins that mediate p120 translocation to the nucleus. Similar mutually exclusive interactions are common in Arm domain proteins and have been extensively characterized for β-catenin (see Introduction).
Our data suggest that regulation of the affinity of p120 for E-cadherin could be an important mechanism for modulating nuclear import. p120 ΔArm10 localized exclusively with E-cadherin in the absence of LMB but showed significant nuclear localization after the addition of LMB. This behavior differed from that of wild-type p120, which binds E-cadherin with higher affinity (Ireton et al., 2002) and was not significantly redistributed to the nucleus after LMB addition. Thus, small differences in the affinity of the p120-E-cadherin interaction may significantly increase the pool of p120 available for nuclear translocation. The concept that repeats 3 and 5 are necessary for both cadherin binding and nuclear translocation is consistent with this notion, as conditions that promote dissociation of p120 from E-cadherin would be expected to have the dual consequence of releasing p120 and of exposing its import-associated repeats. These binding interactions may work in a mutually exclusive fashion, such that release from the complex results in rapid commitment to alternative binding partners involved in nuclear shuttling. Thus, although we could not find a role for p120 phosphorylation, the contribution of the regulatory domain to nuclear shuttling may reflect, in part, modifications that control p120 affinity for cadherins.
In E-cadherin-negative cell lines, p120-1A and -3A concentrated around the nuclear periphery and were also found in a bright ring resembling the nuclear envelope. Surprisingly, a significant portion of this staining colocalized precisely with the microtubule network. LMB treatment reduced the perinuclear staining significantly at the expense of increased nuclear staining. This effect was specific for p120 because the tubulin staining was not affected by LMB. Thus, the perinuclear and nuclear pools of p120 appear to be interchangeable. In addition, destabilizing the microtubule network with nocodazole nearly eliminated the perinuclear staining and resulted in increased p120 staining in the nucleus. By contrast, stabilizing microtubules with taxol had the opposite effect. Concomitant with a striking accumulation of microtubules in basket-like structures on either side of the nucleus, nuclear p120 staining decreased and was replaced by co-staining with the taxol stabilized microtubules. Thus, it is clear that the microtubule-like staining represents direct or indirect binding of p120 to microtubules. Together, these data suggest that the microtubule system is closely linked to mechanism(s) controlling p120 nuclear shuttling.
The involvement of the microtubules in the regulation of nuclear trafficking has previously been shown for other nuclear proteins including p53 (Giannakakou et al., 2002; Giannakakou et al., 2000), Smad (Dong et al., 2000) and MIZ (Ziegelbauer et al., 2001). p120 might dock on microtubles and then move into the nucleus in response to signaling events -a mechanism reminiscent of Hedgehog (HH) nuclear signaling in which a complex of proteins including a transcription factor Cubitus interruptus (CI) and a kinesin-related protein Costal2 associate with the microtubules until the HH signal disrupts the complex, allowing CI to enter the nucleus (Robbins et al., 1997; Sisson et al., 1997). Interestingly, yeast two-hybrid analyses have turned up kinesins (J. Daniel and A. B. Reynolds, unpublished) that might mediate the indirect interaction of p120 with microtubules. Kinesins are typically motor proteins involved in, among other things, the transport of proteins and organelles on microtubules (Schliwa and Woehlke, 2003). Minimally, our data reveal a close relationship between p120 and the microtubule system, which appears to modulate p120 nuclear shuttling. A more detailed account of the potential functional consequences of the p120-kinesintubulin interactions has been submitted separately (P. Z. Anastasiadis et. al., unpublished).
In summary, we have shown that the nucleocytoplasmic shuttling of p120 is regulated by intrinsic and extrinsic mechanisms that together determine the extent of its nuclear localization. Our data show clearly the relationship between E-cadherin loss and increased nuclear p120, and suggest further a role for microtubule interactions. An intriguing hypothesis is that dysregulation of p120 nuclear trafficking contributes to the metastatic phenotype associated with E-cadherin-deficient carcinomas. To clarify this hypothesis, it will be crucial to determine the exact roles of p120 in the nucleus and in nucleocytoplasmic shuttling.
This work was supported by the NIH grants CA55724 and CA83068. Agnes Roczniak-Ferguson is a recipient of a postdoctoral fellowship from the Canadian Institutes of Health Research.