The Crumbs complex is an important determinant of epithelial apical-basal polarity that functions in regulation of tight junctions, resistance to epithelial-to-mesenchymal transitions and as a tumour suppressor. Although the functional role of the Crumbs complex is being elucidated, its regulation is poorly understood. Here, we show that suppression of RNF146, an E3 ubiquitin ligase that recognizes ADP-ribosylated substrates, and tankyrase, a poly(ADP-ribose) polymerase, disrupts the junctional Crumbs complex and disturbs the function of tight junctions. We show that RNF146 binds a number of polarity-associated proteins, in particular members of the angiomotin (AMOT) family. Accordingly, AMOT proteins are ADP-ribosylated by TNKS2, which drives ubiquitylation by RNF146 and subsequent degradation. Ablation of RNF146 or tankyrase, as well as overexpression of AMOT, led to the relocation of PALS1 (a Crumbs complex component) from the apical membrane to internal puncta, a phenotype that is rescued by AMOTL2 knockdown. We thus reveal a new function of RNF146 and tankyrase in stabilizing the Crumbs complex through downregulation of AMOT proteins at the apical membrane.
Coordination of apical-basal polarity in epithelial cells requires the convergence and interplay of different protein complexes that form a diverse, intricate and tightly regulated network with links to a variety of cell processes that include protein trafficking, cell signaling, gene expression and asymmetric cell division (Bose and Wrana, 2006; Giepmans and van Ijzendoorn, 2009; Neumuller and Knoblich, 2009; Varelas et al., 2010). Basolateral adherens junctions play a key role in establishing cell–cell contacts between epithelial cells that are crucial for maintaining the integrity of the epithelium, whereas the tight junction, comprising ZO-1 (also known as TJP1), claudins and occludin, forms a key barrier with adjacent cells that prevents the free diffusion of solutes across the epithelium. The Crumbs complex [comprising CRB3, PALS1 (also known as Mpp5), PATJ (also known as INADL) and MUPP1 (also known as MPDZ)], as well as the Par complex (PAR6, PAR3 and aPKC; note PAR proteins are also known as PARD proteins in mammals, where there is more than one isoform) play important roles in establishing and maintaining tight junctions and are linked to the cytoskeleton to control cell shape, whereas the Scribble complex defines the basolateral domain (Assemat et al., 2008). Through mutual exclusion, these complexes are precisely positioned in epithelial layers to ensure proper size and orientation of the apical versus basolateral domains (Shin et al., 2006).
Apical-basal polarity is not only essential for the proper function of normal epithelia, but is also linked to tumour suppressive capabilities. Whereas loss or mislocalization of tight and adherens junction components are hallmarks of invasive and metastatic epithelial cancers (Martin, 2014; Vasioukhin, 2012), the role of the Par and Crumbs complexes is less straightforward. The Crumbs complex suppresses tumorigenesis and metastasis in mouse models (Karp et al., 2008), whereas PAR6 and PKCζ are often overexpressed and sometimes phosphorylated in human cancers (Huang and Muthuswamy, 2010). Indeed both signalling through Par6 and loss of the Crumbs complex participate in epithelial-to-mesenchymal transition (EMT), which is important for tumour cell invasive activity (Ozdamar et al., 2005; Varelas et al., 2010; Viloria-Petit et al., 2009). These observations further emphasize the importance of tight regulation of polarity determinants in normal epithelial cells and during tumour progression.
Balanced spatial-temporal regulation of polarity complexes is crucial for establishing normal epithelial function. Thus, although the Crumbs complex is important for tight junction assembly, overexpression of CRB3 (in association with PALS1 binding) delays tight junction formation and leads to a disorganized apical domain (Fogg et al., 2005; Lemmers et al., 2004; Roh et al., 2003; Wodarz et al., 1995). In addition, PAR6 can have either a context-dependent stimulatory effect on tight junction assembly or an inhibitory effect on tight junction maintenance. Thus, whereas a clear picture is emerging describing the events that must occur en route to establishing normal apical-basal polarity in epithelial cells, comparatively less is known about how protein turnover or stability, and trafficking contribute to maintenance of this regulatory network. Identifying factors that contribute to these processes might thus reveal pathways that are deregulated in transitions to a disease state, such as cancer.
To identify such regulatory factors, we performed a small interfering RNA (siRNA) screen to identify genes involved in tight junction maintenance (see Results for details). One hit from this screen was the ubiquitin ligase RNF146, which has a RING finger domain important for ubiquitylation, and a WWE domain that recognizes poly-ADP-ribosylated (parsylated) substrates (DaRosa et al., 2015; Zhang et al., 2011). In addition, tankyrase (TNKS, which has two isoforms TNKS1 and TNKS2), a known interaction partner of RNF146, binds and parsylates AXIN leading to ubiquitylation by RNF146 and subsequent degradation; this in turn activates Wnt signalling (Callow et al., 2011; Zhang et al., 2011). Here, we demonstrate how RNF146 and its known partner TNKS2 function to regulate maintenance of the Crumbs complex in epithelial cells. We show that interference with either RNF146 or TNKS2 disrupts cell polarity by triggering a relocation of the Crumbs complex component PALS1, from the cell junctions to internal puncta. Furthermore, we demonstrate that AMOTL2, which regulates apical assembly of the Crumbs complex, is a key target of RNF146–TNKS that is parsylated by TNKS2 and subject to ubiquitin-dependent degradation mediated by RNF146. Consequently, continual turnover of AMOTL2 through the RNF146–TNKS pathway is required to facilitate the correct localization of the Crumbs complex. We thus uncover a new mechanism of apical-basal polarity regulation involving parsylation-dependent control of protein stability and trafficking.
Knockdown of RNF146 results in mislocalization of PALS1 and ZO-1
We developed a high-throughput immunofluorescence-based siRNA screen to identify regulators of tight junction maintenance using Eph4 cells, which are a non-transformed mouse mammary epithelial cell line. Cells were simultaneously transfected and plated at a relatively high density (approaching 100% confluence). In this condition, junctions form shortly after plating, and thus subsequent siRNA-mediated protein ablation assesses target gene function during junctional maintenance. We marked tight junctions using ZO-1 and then assessed junction maintenance upon systematically knocking down ∼4000 targets that correspond to genes with domains of known function. The details of the screen design will be described elsewhere, but one target identified as a putative regulator of tight junction maintenance was RNF146, a known E3 ubiquitin ligase (Fig. 1A). To validate these findings we used a deconvolved siRNA set (four individual siGENOME siRNAs) specifically targeting Rnf146 (Fig. 1B,C), as well as one ON-Target siRNA (Fig. 1D,E: siRNF146-2). This revealed that all five siRNAs suppressed Rnf146 to varying degrees and led to corresponding disruption of junctional ZO-1 staining. Of these, one siRNA from the deconvolved set (siRNF146-1) and the ON-Target siRNA (siRNF146-2) were particularly effective, displaying >75% knockdown of Rnf146 with concomitant junctional disorganization or loss of ZO-1 (Fig. 1E). We therefore used these two siRNAs for subsequent studies, initially focusing on the localization of other polarity complexes. We observed that upon Rnf146 knockdown, PALS1, a Crumbs complex component, was highly disorganized and no longer localized to the apical domain of the cell periphery (Fig. 1E). Of note, the disturbance in PALS1 was more dramatic than observed for ZO-1. To further confirm the phenotype caused by Rnf146 knockdown, we also examined human bronchial epithelial (HBE) cells. Similar to Eph4 cells, loss of RNF146 led to disorganized PALS1 and ZO-1 localization (Fig. S1A). Of note, knockdown of Rnf146 had marginal effects on E-cadherin, PARD6B or SCRIB localization in Eph4 cells (Fig. S1B), suggesting the role of RNF146 in cell polarity is restricted to apical complexes. We next used these two RNF146-targeting siRNAs to assess the functional role of RNF146 in apical-basal polarity, investigating transepithelial resistance (TER; Fig. 1F). This revealed that knockdown of the known regulators of apical-basal cell polarity, ZO-1, a key component of the tight junction, or CRB3, an important constituent of the Crumbs complex in epithelial cells, decreased TER by 59% and 82%, respectively. Similarly, knockdown of Rnf146 also decreased TER by 46%, demonstrating that epithelial integrity was compromised.
RING and WWE mutants of RNF146 fail to rescue siRNF146-mediated junctional defects
RNF146 is a RING-finger-domain-containing E3 ubiquitin ligase that also incorporates a WWE domain that recognizes substrates that have been poly-ADP-ribosylated (parsylated) by tankyrase (Callow et al., 2011; Zhang et al., 2011). Therefore, to explore structure–function relationships in RNF146, we stably expressed siRNA-resistant wild-type (WT) or mutant variants of Rnf146 in Eph4 cells (Fig. 2A) and examined polarity markers upon knockdown of endogenous Rnf146 expression. Stable expression of WT Rnf146 rescued junctional stability and apical Crumbs localization, as marked by ZO-1, as well as PALS1 localization, thus further validating that the Rnf146 knockdown phenotype is not due to an off-target effect (Fig. 2B). Next, we examined whether Rnf146 lacking either the RING-finger domain or the WWE domain rescued the knockdown phenotype. As shown in Fig. 2B, expression of either of these deletion mutants failed to rescue the Rnf146 knockdown phenotype. This suggests that both ubiquitylation and binding to parsylated substrate(s) are important for the function of RNF146 in cell polarity. As ectopic expression appeared to be cytoplasmic (Fig. 2B), we next identified the localization of endogenous RNF146. Antibody specificity was confirmed through immunofluorescence and western blotting. As expected RNF146 levels were elevated through inhibition of TNKS (both TNKS1 and TNKS2, using XAV939) (Fig. 2C). RNF146 was mainly cytoplasmic in Eph4 cells, and localized to apical and basolateral compartments, as assessed through co-staining with PALS1, ZO-1 and SCRIB (Fig. 2D). Interestingly, RNF146 displayed a diffuse localization and colocalization with any of these junctional markers was not enriched. This is consistent with the multifunctional role of RNF146 in targeting a broad range of cellular substrates (Kang et al., 2011; Levaot et al., 2011; Li et al., 2015; Zhang et al., 2011).
Tankyrase inhibition relocalizes PALS1 from the junction to internal puncta
To determine whether tankyrase proteins are also important in tight junction and Crumbs complex assembly, we used the TNKS inhibitor XAV939 (Huang et al., 2009) in EpH4 and HBE cells. This revealed that inhibition of TNKS alone led to loss of junctional PALS1 and relocation to internal puncta with or without Rnf146 knockdown. In addition, Rnf146 ablation resulted in a similar relocation of PALS1 even without TNKS inhibition (Fig. 3A, Eph4, high magnification; Fig. S1C, Eph4, low magnification; Fig. 3B, HBE, high magnification). In striking contrast, TNKS inhibition only marginally affected ZO-1 localization. Taken together, these results indicate that RNF146 and TNKS function together to promote maintenance of the Crumbs complex.
AMOT proteins interact with RNF146 and TNKS
To determine how RNF146 and TNKS proteins function to regulate Crumbs complex assembly, we used LUMIER (Barrios-Rodiles et al., 2005) to systematically map interactions between known components of the polarity complex and either RNF146 or TNKS2. For this, we built a Flag-tagged prey collection comprising 74 cDNAs encoding proteins known to regulate or associate with epithelial junction proteins. Renilla-luciferase-tagged RNF146, mutant variants or TNKS2 were then used as baits for immunoprecipitation followed by luciferase assays as described previously (Barrios-Rodiles et al., 2005). Because RNF146 is an ubiquitin ligase, substrates of this protein are likely rapidly ubiquitylated and subsequently degraded upon interaction with RNF146. RNF146 is also known to be autoubiquitylated after association with a substrate. We therefore used an RNF146 point mutant in the RING finger domain, H53A (denoted HA) (Fig. 2A), which was previously shown to be important for substrate ubiquitylation (Callow et al., 2011). Furthermore, given that TNKS inhibition also disrupted the Crumbs complex, we employed an RNF146 mutant within the WWE domain, R153A (denoted RA), which disrupts binding to parsylated proteins (Zhang et al., 2011). The RING and WWE domain deletion mutants were also included to complement these point mutants. This strategy thus stabilizes the interaction of RNF146 with potential substrates and identifies parsylation-dependent binding partners.
To test our strategy, we analyzed AXIN2, which is a substrate of RNF146 that binds in a parsylation-dependent manner. This showed interactions with TNKS2, as well as RNF146, with the latter stabilized by RING finger mutants, but lost upon mutation of the WWE domain (Fig. S2A). We therefore focused on RNF146 and TNKS2 partners that displayed similar binding patterns to AXIN2. This revealed that the angiomotin (AMOT) isoforms AMOT130, AMOTL1 and AMOTL2 all bound strongly to both RNF146(HA) and TNKS2, whereas interactions with the WWE domain mutants [RNF146(RA) and ΔWWE] were perturbed (Fig. S2A). To complement the LUMIER analysis, we also employed BioID to identify differential associations with WT RNF146 and the RING and WWE point mutants. BioID employs bait tagged with BirA encoding a promiscuous biotin ligase, which biotinylates neighbouring proteins. Biotinylated prey can then be isolated through streptavidin purification and identified through mass spectrometry (Roux et al., 2012). This revealed similar patterns of associations between RNF146 and a number of putative parsylation-dependent RNF146 targets. Most notably, RNF146 associated with TNKS1 and TNKS2, as well as the AMOT proteins, all of which were strongly stabilized by the RING finger mutant, RNF146(HA). We also detected interactions with PATJ and MUPP1, which are additional components of the Crumbs complex (Fig. S2C,D). The combined results of the LUMIER and BioID screens are summarized in a network graph, which shows interactions within the LUMIER polarity-associated collection, as well as new interactions identified through both LUMIER and BioID (Fig. 4A). This reveals that in addition to the AMOTs, RNF146 makes extensive connections with polarity complex components, many of which we confirmed by co-immunoprecipitation and immunoblotting (Fig. S2B). Thus, in addition to Crumbs complex components, RNF146 also interacted with multiple components of the Par complex (PARD6A, PARD6G and PRKCZ) and the Scribble complex (SCRIB, LLGL1 and LLGL2), as well as prominent interactions with JUB (also known as AJUBA), MLLT4 and PXN (Fig. S2B). Of these, interactions between RNF146 and AMOTs, which were also recently reported (Wang et al., 2015), were intriguing given the key role AMOT proteins play in regulating apical polarity complex assembly and localization (Wells et al., 2006).
AMOT proteins have been proposed to have an important role in endothelial tight junction stabilization (Bratt et al., 2005; Zheng et al., 2009). In addition, AMOT proteins have been shown to associate with tight junction and Crumbs complex proteins in epithelial cells, and although a role for AMOT proteins was proposed for tight junction stability, overexpression of AMOT proteins has been shown to redistribute Crumbs complex components from junctions to internal puncta (Wells et al., 2006). AMOT proteins are expressed from three highly conserved genes (Amot, AmotL1 and AmotL2), which all can undergo alternative splicing (Bratt et al., 2002; Ernkvist et al., 2006; Moreau et al., 2005). Although all Amot genes are known to produce different transcripts, Amot p80 and Amot p130 are the only two splice variants that have been extensively studied. Therefore, to assess binding of AMOT isoforms to RNF146, TNKS1 and TNKS2, we performed manual LUMIER assays, in which luciferase-tagged baits were co-transfected with Flag-tagged AMOT proteins. The luciferase activity in the immunoprecipitates was quantified and the NLIR value, a function of binding as previously described (Braun et al., 2009), was determined. AMOT130, AMOTL1 and AMOTL2 were all shown to bind RNF146(HA), TNKS1 and TNKS2, whereas the binding of the 80 kDa isoform of AMOT (AMOT80) was negligible (Fig. 4B). Co-expression of HA-tagged RNF146(HA) with Flag-tagged AMOTs followed by immunoprecipitation similarly showed that RNF146(HA) bound AMOT130, AMOTL1 and AMOTL2, but not AMOT80 (Fig. 4C). Therefore, RNF146 and TNKS interact with AMOT proteins, with especially strong interactions with AMOTL2.
RNF146 and TNKS2 regulate AMOTL2 stability through parsylation and ubiquitylation
As AMOT proteins were identified as putative substrates of both TNKS and RNF146, we hypothesized that TNKS2, which bound AMOTs, could parsylate AMOT proteins, thus leading to ubiquitylation by RNF146 and degradation. To explore this, we first assessed the binding of AMOTs to RNF146 harbouring either the deletion of the WWE domain or the mutation of a key residue in the WWE domain, R163A (RA). We then tested binding in the presence of chemical inhibition of TNKS proteins (XAV939). All three strategies dramatically decreased binding of AMOT proteins to RNF146, as shown by LUMIER (Fig. 5A). In addition, immunoprecipitation was used to validate the role of parsylation in facilitating RNF146 binding to AMOTs (Fig. 5B,C). Because AMOTL2 was the strongest binding partner to RNF146, TNKS1 and TNKS2 when compared to other AMOT proteins, we chose this protein for subsequent biochemical analysis. To determine whether AMOTL2 could be ubiquitylated by RNF146, sequential Flag immunoprecipitation in the presence of HA-tagged ubiquitin was employed. Ubiquitylation of AMOTL2 was increased when both exogenous RNF146 and TNKS2 were present; this effect was abrogated by a deletion of the RING finger of RNF146 (Fig. 5D). Total levels of exogenous Flag–AMOTL2 also declined with either RNF146 or TNKS2 expression, whereas co-expression of both RNF146 and TNKS2 led to almost undetectable levels of AMOTL2. This effect was also eliminated by the deletion of the RING finger of RNF146 (Fig. 5D). Similar results were obtained upon analysis of AMOT130 and AMOTL1 (data not shown).
We next assessed AMOTL2 parsylation by TNKS2. For this, an in vitro parsylation assay was performed, which revealed that AMOTL2 was parsylated in the presence of recombinant TNKS2 (Fig. 5E), which was blocked by a concentration of XAV939 as low as 0.1 μM (Fig. S3A). Parsylation of AMOTL2 was confirmed by exogenous expression of Flag-tagged AMOTL2 in the presence of exogenous TNKS2 followed by immunoprecipitation with antibodies to either poly-ADP-ribose (Fig. 5F) or Flag (Fig. 5G). Again, XAV939 potently inhibited parsylation of AMOTL2 when co-expressed with exogenous TNKS2 (Fig. S3B). Steady state levels of AMOTL2 were then assessed after knockdown of RNF146 with or without treatment with XAV939 in Eph4 cells. AMOTL2 levels were increased by XAV939 treatment (Fig. 5H) in a dose-dependent manner (Fig. 5E) and further increased upon concomitant Rnf146 knockdown (Fig. 5H). Antibody specificity was also confirmed through knockdown of AMOTL2 (Fig. 5H). AMOT130 and AMOTL1 were also increased upon XAV939 treatment in a dose-dependent manner; however, a further increase was not observed upon Rnf146 knockdown (Fig. S3D,E), with antibody specificity confirmed through knockdown of Amot or AmotL1. Similar results were obtained when HBE cells were subjected to XAV939 treatment (Fig. S3C,E). Collectively, these results demonstrate that, as in a recent report (Wang et al., 2015), AMOT proteins, and in particular AMOTL2, serve as substrates for TNKS and are subject to RNF146-dependent ubiquitylation and degradation.
Ectopic AMOT expression induces PALS1 redistribution from the junctions to internal puncta
Based on our finding that inhibition of RNF146 or TNKS causes PALS1 redistribution from the cell periphery to internal puncta (Fig. 3C; Fig. S1), and that RNF146 and TNKS2 regulate AMOT stability, we next examined whether overexpression of AMOT proteins could cause a similar disruption of PALS1 localization in Eph4 cells. This revealed that Clover-tagged AMOTL1 and AMOTL2, and to a lesser extent AMOT130, were localized to prominent internal puncta (Fig. 6A) and all caused dissolution of peripheral PALS1 staining, with redistribution to internal puncta (Fig. 6A), whereas very little effect was observed on ZO-1 localization (Fig. 6A).
Similar alterations occurred in HBE cells (Fig. 6B). Interestingly, in HBE cells, overexpressed Clover–AMOT130 was typically observed as a filamentous pattern, with PALS1 sometimes colocalized to these structures (Fig. 6B, lower panel, arrow 1). In contrast, overexpressed Clover–AMOTL1 and Clover–AMOTL2 were typically localized to internal puncta. Furthermore, although PALS1 sometimes colocalized with AMOTL1 in small internal puncta (Fig. 6B, lower panel, arrows 2), colocalization with AMOTL2 was more prominent (Fig. 6B, lower panel, arrows 3). These results were verified using Flag-tagged AMOTs (data not shown). Thus, overexpression of AMOT proteins is sufficient to interfere with the proper apical localization of PALS1.
AMOTL2 knockdown rescues junctional defects induced by RNF146 or TNKS ablation
Based on the above results, we hypothesized that knockdown of AMOT expression might rescue the defect in PALS1 localization that is caused by inhibition of RNF146 or TNKS. As shown above, XAV939 treatment induced relocatization of PALS1 from the periphery to internal puncta, which was enhanced by simultaneous Rnf146 knockdown (Fig. 7). Furthermore, although knockdown of Amot or AmotL1 had no effect on PALS1 relocalization induced by either RNF146 or TNKS inhibition (Fig. S4A), knockdown of AmotL2 (35%; Fig. S4B) strongly rescued the phenotype. Thus, simultaneous knockdown of Rnf146 and AmotL2 restored prominent junctional PALS1 staining, as did AmotL2 knockdown in cells treated with 1 μM XAV939 (Fig. 7). Given that the alteration in PALS1 localization induced by TNKS inhibition by XAV939 is dose dependent, we also tested higher doses of XAV939. This revealed that AmotL2, but not Amot or AmotL1, knockdown (56%, 77% and 82%, respectively; Fig. 8B), rescued junctional PALS1 localization even at 10 μM XAV939 (Fig. 8A). Taken together, these results suggest that interference with the TNKS–RNF146 pathway leads to excess AMOTL2 in epithelial cells that in turn disrupts the Crumbs complex and leads to redistribution of PALS1 to internal puncta. Thus, in polarized epithelia, RNF146 and TNKS2 play a key role in regulating steady-state levels of AMOTL2 to maintain the integrity of the epithelium.
Regulation of Crumbs complex assembly and trafficking remain relatively poorly understood. However, coordinated spatial expression of components of the Crumbs complex mediates proper tight junction formation and epithelial polarization (Campbell et al., 2009; Lemmers et al., 2004; Roh et al., 2003; Straight et al., 2004), resistance to induction of EMT (Varelas et al., 2010) and suppression of tumour metastasis (Karp et al., 2008). This is also crucial in development, where CRB3 is required for airway cell differentiation (Szymaniak et al., 2015). Thus, in discovering new regulators of Crumbs complex maintenance, the potential also exists to uncover new tumour suppressor networks and cell fate determinants. Here we identify a new regulatory mechanism governing junctional localization of the Crumbs complex that involves ADP-ribosylation-triggered degradation of AMOTL2 through the TNKS–RNF146 pathway.
Our initial identification of RNF146 as a regulator of the Crumbs complex arose from a high-throughput screen that revealed loss of RNF146 disturbed tight junction organization, as marked by ZO-1. This was accompanied by disturbances in transepithelial resistance (Fig. 1F), and analysis of the Crumbs complex, as revealed by PALS1 staining, revealed that RNF146 loss profoundly perturbed apical localization with PALS1 displaying prominent localization to internal puncta. Given the predominance of this phenotype we propose that the main function of the RNF146–TNKS pathway in cell polarity is through regulation of the Crumbs complex, with tight junction maintenance occurring as a secondary consequence. Furthermore, through systematic proteomics strategies, we showed that the TNKS–RNF146 pathway regulates AMOT protein stability that in turn mediates localization of the Crumbs complex. Our findings are thus consistent with previous studies showing that AMOTs colocalize with PALS1 (and also PAR3) in endosomes in MDCK cells that induces a secondary effect on tight junction dynamics (Wells et al., 2006). A secondary effect on tight junction maintenance is consistent with the importance of the Crumbs complex in tight junction assembly (Campbell et al., 2009; Lemmers et al., 2004; Roh et al., 2003; Straight et al., 2004). Furthermore, the adherens junctions and the scribble complex, as marked by E-cadherin and SCRIB, respectively, were relatively unaffected by RNF146 knockdown (Fig. S1B), supporting a more apical role for RNF146–TNKS in junctional maintenance.
Little is known about trafficking of Crumbs complex components. In Drosophila, Crb has been shown to be a cargo of the retromer complex, which is involved in recycling of endosomal membrane proteins, and loss of retromer results in a dramatic decrease in Crb expression (Pocha et al., 2011). Consistent with this, in MDCK cells, AMOT80 colocalizes with PAR3 and PALS1, and also with Rab11-positive recycling endosomes in cells that do not show a high degree of polarization or that are migratory, but not in highly polarized cells (Heller et al., 2010). Therefore, trafficking of Crumbs is likely a key mechanism to control its function in dynamic epithelia. Here, we showed that excess AMOTL2, caused by either disruption of TNKS–RNF146-mediated degradation, or by overexpression, leads to loss of junctional PALS1 and its accumulation in internal puncta. This indicates that appropriate AMOT levels are required to maintain the dynamic trafficking of the Crumbs complex and it will be of interest to determine whether AMOT-dependent trafficking follows similar itineraries. We attempted to elucidate the identity of these PALS1-positive puncta through PALS1 co-staining with markers such as EEA1 (early endosome), RAB7 (late endosome), RAB11 (recycling), LAMP1 (lysosomal), LC3B (autophagosome) and CAV1 (caveolae) after XAV939 treatment. However, colocalization was not observed for any of the tested markers (data not shown). Studies are ongoing to identify these structures.
AMOT proteins were originally discovered to be involved in endothelial cell migration, junction assembly and tube formation (Troyanovsky et al., 2001; Wang et al., 2011; Zheng et al., 2009). AMOT proteins, but not AMOT80, were also found to be associated with the actin cytoskeleton (Ernkvist et al., 2008; Gagne et al., 2009; Hultin et al., 2014). Furthermore, these individual AMOT isoforms have distinct cellular functions. AMOT80 in particular stably localizes at cell junctions (Wells et al., 2006), and in endothelial cells is associated with a migratory phenotype (Ernkvist et al., 2008), whereas AMOT130, through its N-terminal domain that is lacking in AMOT80, has a more rapid junctional localization and is associated with endothelial stabilization (Ernkvist et al., 2008). AMOTL1 and AMOTL2 have also been shown to localize to tight junctions (Patrie, 2005). Here, we showed that RNF146 interacts with AMOT130, AMOTL1 and AMOTL2, but not AMOT80, and demonstrated that TNKS–RNF146-dependent targeting of AMOTL2 is crucial in polarized Eph4 and HBE cells for the correct junctional localization of the Crumbs complex. It is therefore tempting to speculate that AMOT80, by avoiding TNKS–RNF146 targeting, provides an effective pathway to inhibit junctional Crumbs localization in dynamic epithelia.
TNKS1 has also been localized to the lateral membrane and tight junctions during junction assembly in epithelial cells, with parsylation leading to cytosolic translocation, and subsequent ubiquitylation and degradation (Yeh et al., 2006). In agreement, a belt of ADP-ribosylation that colocalizes with the tight-junction-associated protein vinculin, has been discovered in epithelial cells and can be ablated through TNKS inhibition (Lafon-Hughes et al., 2014). Consistent with ADP-ribosylation more broadly regulating junctional components beyond AMOT, we observed that RNF146 interacts with a number of proteins associated with apical-basal polarity, including additional Crumbs complex components such as PATJ. Interestingly, RNF146 was localized to the cytoplasm, in all apical-basal compartments (Fig. 2C,D). Taken together, these results suggest that, in polarized epithelial cells, TNKS interacts with polarity complex proteins to mediate their ADP-ribosylation and, in the case of AMOT130, AMOTL1 and AMOTL2, degradation through RNF146-mediated ubiquitylation. Based on its cytoplasmic localization, these events are likely to occur during trafficking before membrane localization, and could prevent AMOT proteins from reaching the apical junction. In Eph4 and HBE cells, we show that this in turn is crucial for establishing apical–basal polarity and functional tight junctions by restricting AMOT-dependent trafficking of the Crumbs complex to internal puncta.
Interestingly, RNF146 and TNKS have previously been shown to have pro-tumorigenic and cell survival effects and TNKS inhibitors suppress colorectal adenoma growth in mice (Waaler et al., 2012). The pro-oncogenic effects of TNKS are proposed to involve targeting of the WNT pathway inhibitor AXIN, as well as the tumour suppressor PTEN and, as has been shown more recently, the transforming ability of YAP (Callow et al., 2011; Li et al., 2015; Wang et al., 2015; Zhang et al., 2011). RNF146 has also been shown to mediate cell survival and DNA repair after DNA damage through its association with PARP-1 (Andrabi et al., 2011; Kang et al., 2011). Our results suggest that, by targeting AMOT proteins, the TNKS–RNF146 pathway might also possess key tumour suppressor activity that acts through stabilization of the Crumbs complex and tight junctions in polarized epithelial cells. Indeed other studies have identified AMOT proteins as potentially oncogenic, as knockdown of AMOT impedes the invasion, proliferation and tumour forming ability of breast cancer cells and NF2−/− Schwann cells (Lv et al., 2015; Ranahan et al., 2011; Yi et al., 2011). In addition, AMOTL2 knockdown decreases tumour growth and invasion in colorectal cancer cells (Mojallal et al., 2014) and AMOT potentiates RAS–MAPK signaling through sequestration of RICH1, a negative regulator of RAC1 and RAS–MAPK (Yi et al., 2011). As AMOT proteins also mediate endothelial cell migration during neovascularization (Aase et al., 2007; Ernkvist et al., 2009), it is not surprising that, in a mouse model of mammary tumour growth, AMOT-targeting antibodies, as well as DNA vaccination against Amot, have both been shown to decrease tumour vascularization (Holmgren et al., 2006; Levchenko et al., 2008). Although it is tempting to speculate on the potential efficacy of TNKS inhibitors based on their function in downregulating Wnt-, YAP- and AKT-mediated transcriptional programmes, it is impossible to ignore potential undesirable consequences from the upregulation of AMOT proteins. Further studies will be required to delineate the contexts in which TNKS inhibitors could function in restricting cancer growth versus potentially promoting cancer cell invasion.
In summary, this study has uncovered a new role for the RNF146–TNKS pathway in regulating apical-basal polarity by stabilizing the junctional Crumbs complex through ADP-ribosylation-dependent targeting of AMOT proteins.
MATERIALS AND METHODS
Constructs and siRNAs
siRNAs used in this study were obtained from Dharmacon (GE Healthcare) and are listed in Table S1. Rnf146 was subcloned into the gateway entry vector pDONR223 using primers listed in Table S3. For the generation of point mutations, site-directed mutagenesis was performed using a Quikchange II kit (Agilent) using primers listed in Table S2. For domain deletions, PCR was performed using primers listed in Table S2. For LUMIER and follow-up analysis, bait and prey were obtained as follows. Some Flag-tagged constructs were previously created and obtained directly as outlined in Table S5. For other cDNA, vectors were obtained and used as a template to clone genes into the gateway entry vector pDONR223 for subsequent epitope tagging; cDNA templates or entry clones were obtained from OpenFreezer (Olhovsky et al., 2011) or Addgene and primers used for this cloning are listed in Table S3. Llgl2 template cDNA (pEGFP2-LGL2) was obtained from William Nelson (Stanford, CA). BP gateway cloning was performed in accordance with the manufacturer's instructions (ThermoFisher). Epitope tagging was performed using LR gateway cloning (ThermoFisher) in accordance with the manufacturer's directions. Destination vectors were pCMV5-based and contained the following tags: N-terminal Renilla luciferase, N- and C-terminal 3XFlag (3F), and N-terminal 3XHA (3HA). For stable overexpression of RNF146, 3F-RNF146 variants were cloned into the pCAGIP vector using the primers listed in Table S3. For the creation of Clover-tagged expression vectors, Amot130, AmotL1 and AmotL2 cDNA was cloned into pClover-C1 using the primers listed in Table S3. All mutations, gateway vectors, pCAGIP vectors, Clover-tagged expression vectors and vectors received for direct use in LUMIER (Table S5) were verified by sequencing.
Cell culture, transfections, TER assays and qPCR
Eph4 mouse mammary epithelial cells were cultured in DMEM (Hyclone/GE Healthcare) with 10% fetal bovine serum (FBS) as described previously (Varelas et al., 2010), and human bronchial epithelial (HBE) cells were cultured in MEM with EBSS (Hyclone/GE Healthcare) with 10% FBS, subculturing at ∼90% confluency. HBE cells were obtained from Dieter Gruenert (University of California at San Francisco, CA). Cells were routinely checked for mycoplasma contamination. Transient transfections were performed in either eight-well chamber slides (BD Biosciences) (for immunofluorescence) or in 24-well plates (for RNA collection or on coverslips for immunofluorescence). Reverse transfections were performed, using Dharmafect 1 (Thermo Scientific) for siRNA or using Lipofectamine 3000 (Invitrogen) for plasmids. Transfections were performed at high density using 200,000 EpH4 cells or 240,000 HBE cells for 24-well plates or 100,000 Eph4 or 120,000 HBE cells for eight-well chamber slides. Medium was changed after 16 h. XAV939 (Selleckchem, cat. no. 51180) treatment was performed at 16 and 42 h post transfection. For stable transfections, 62,500 cells were transfected with 1 μg of DNA and 8 μl of lipofectamine 3000 (Invitrogen) in six-well plates. Plasmids included pCAGIP as a vector control, and pCAGIP-containing cDNA encoding variants of Rnf146. On day 1 after transfection, the medium was changed. Selection with 5 μg ml−1 puromycin (EMD) was initiated on day 4 and proceeded indefinitely every 2–3 days for 1 month before experiments were performed on pools of stable transfectants. For proteasome inhibition, cells were treated with 10 μM each of MG132 Alln. For TER assays, 140,000 cells were reverse transfected as above, and plated on 3.0 μm transwell inserts and cultured in 12-well plates containing normal medium. Medium was changed in both the upper and lower compartment of the chamber after 16 h. Resistance across the membrane was measured using an EvoMX Epithelial Voltohmmeter combined with an Endohm-12 chamber (Precision Instruments). For real-time PCR (qPCR), RNA was collected with RNeasy kits (Qiagen) and reverse transcribed using Superscript III (Thermo Fisher), with both random primers and OligoDT as primers. qPCR was performed by assessing comparative CT; ΔCT values were obtained through amplification of reverse transcribed cDNA with SYBR Green (Roche). Relative concentration is expressed relative to that of housekeeper genes HPRT and GAPDH or HPRT alone. Primer sequences are listed in Table S4.
Cells transfected with siRNA or treated with XAV939 as above were fixed with 4% paraformaldehyde and permeabilized with 0.2% Triton X-100 in PBS. Antigen demasking was performed with 1% SDS in PBS for 10 min. Samples were blocked with 2% BSA in PBS plus 0.1% Tween20, and primary antibody was incubated in this blocking buffer overnight at 4°C; after washing, samples were incubated with fluorescent secondary antibody and DAPI for 1 h at room temperature, washed and mounted with DABCO and mowiol. Antibodies used are listed below. Images were captured using a DMIRE2, Leica microscope with a spinning disk confocal scanner (CSU10, Yokogawa) with an EMCCD digital camera (Hamamatsu, C9100-13). Image acquisition and analysis was performed with Volocity software (Perkin Elmer).
LUMIER was performed in accordance with Barrios-Rodiles et al. (2005) in 96-well poly-D-lysine-coated plates (Corning) using HEK293T cells and polyfect (Qiagen) as the transfection reagent. Renilla luciferase-tagged RNF146 (WT, H53A, R163A and H53A/R163) as well as TNKS2 were used as baits. Flag-tagged prey included a novel collection of cDNA encoding known proteins that regulate or interact with apical-basal polarity complexes (Table S5). Results represent the average of two to four replicates. Larger scale LUMIER assays were performed in six-well dishes with luciferase-tagged bait (RNF146 H53A, H53A/ΔWWE, H53A/R163A, TNKS1, TNKS2) and Flag-tagged prey (AMOT80, AMOT130, AMOTL1, AMOTL2) as above except that luminescence readings were performed in duplicate. For all LUMIER assays, the normalized LUMIER interaction intensity ratio (NLIR) values are expressed as: the luminescence for immunoprecipitation of bait+prey or bait alone divided by the luminescence of the total lysate for bait+prey or bait alone (Braun et al., 2009).
Immunoprecipitation and immunoblotting
For immunoprecipitation, HEK293T cells were transfected with cDNA using polyfect (Qiagen), and protein was collected using TNTE (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA) lysis buffer with 0.5% Triton X-100 and protease inhibitors. Immunoprecipitation was performed with antibodies listed below for 1 h at 4°C followed by incubation with Sepharose (GE Healthcare, cat. no. 17-0618-02) for 1 h at 4°C. Samples were washed five times with TNTE buffer containing 0.1% Triton X-100. For double immunoprecipitation experiments, immunoprecipitates were collected as above, and precipitated Sepharose complexes were boiled in 1% SDS for 20 min. Samples were washed with TNTE lysis buffer and supernatants were subjected to a second immunoprecipitation step as above. Immunoprecipitates or total lysates collected above were subjected to immunoblotting. In addition, protein samples were collected from Eph4 or HBE cells using RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1%NP-40, 1 mM EGTA, 0.1% SDS, 0.5% sodium deoxycholate plus protease inhibitors) and used for immunoblotting analysis. Lysates in Laemmli buffer were subjected to protein electrophoresis using 7.5–10% SDS-polyacrylamide gels and transferred to 0.45 μm nitrocellulose membrane (Bio-Rad). Membranes were blocked with 5% milk or 4% BSA in Tris-buffered saline with 0.1% Tween (TBST) and incubated in blocking buffer overnight with primary antibodies (listed below). Membranes were washed and incubated with secondary horseradish peroxidase (HRP)-conjugated antibodies (GE Healthcare), washed again and developed with Dura or Femto chemiluminescent reagent (Thermo Scientific). Images were collected with a Chemidoc MP imaging system with Image Lab software (Bio-Rad).
The following antibodies were used for immunoblotting, immunofluorescence and immunoprecipitation. Anti-Flag (M2, monoclonal, Sigma-Aldrich, cat. no. F3165, concentration=3.5-3.7 μg μl−1) was used at a dilution of 1:10,000 for immunoblotting, 1:500 for immunofluorescence and 1:2000 for immunoprecipitation. Anti-Flag–HRP conjugate (Sigma-Aldrich, cat. no. A8592) was used at a dilution of 1:10,000 for immunoblotting. Anti-HA–HRP conjugate (Roche, cat. no. 12013819001) was used at a dilution of 1:10,000 for immunoblotting. Anti-poly-ADP-ribose (Cedarlane, cat. no. 4336-APC-050) was used at a dilution of 1:1000 for both immunoblotting and immunoprecipitation. Anti-luciferase (EMD, cat. no. MAB4400) was used at a dilution of 1:1000 and 1:10,000 for immunoblotting, and for probing immunoprecipitates and total lysates, respectively. Anti-AMOTL2 (FroggaBio, cat. no. AP8860C) and anti-AmotL1 (Abcam, cat. no. ab84049) were used at a dilution of 1:1000, and anti-Amot (Santa Cruz Biotechnology, cat. no. sc-82491) was used at a dilution of 1:500 for immunoblotting. Anti-αTubulin (Sigma-Aldrich, cat. no. T6199) was used at a dilution of 1:2000 for immunoblotting. Anti-PALS1 (Sigma-Aldrich, cat. no. HPA000993-10) was used at a dilution of 1:500 for immunofluorescence. Anti-ZO-1 (Santa Cruz Biotechnology, cat. no. sc-33725), anti-human ZO-1 (BD Biosciences, cat. no. 610966), anti-E-Cadherin (BD Biosciences, cat. no. BD610182), anti-Par6B (Santa Cruz Biotechnology, cat. no. sc-67393), and anti-SCRIB (Santa Cruz Biotechnology, cat. no. sc-28737) were used at a dilution of 1:1000 for immunofluorescence. Anti-RNF146 (Abnova, cat. no. H00081847-B01P) was used at a dilution of 1:1000 for immunoblotting and 1:100 for immunofluorescence.
Ubiquitylation and parsylation assays
To assess ubiquitylation, immunoprecipitation and immunoblotting was performed as outlined above after co-transfection of HEK293T cells with cDNA expression vectors and HA–ubiquitin. For parsylation assays, immunoprecipitation was performed as above. Immunoprecipitates were washed with reaction buffer (50 mM Tris-HCl pH 8.0, 4 mM MgCl2 and 0.2 mM DTT) and incubated with 25 μM Biotin-NAD (Cedarlane, cat. no. 4670-500-01) and 0.5 μg recombinant TNKS2 (rTNKS2) (Sigma-Aldrich, cat. no. SRP-0423) for 45 min at room temperature. When XAV939 was added to the reaction, Biotin-NAD, rTNKS2 and XAV939 were pre-incubated together for 45 min at room temperature before addition to immunoprecipitates. Samples were boiled in SDS and a second immunoprecipitation step was performed (as described above) for all parsylation assays.
Proximity biotinylation coupled with mass-spectrometry – BioID
Gateway cloning (ThermoFisher) was used to clone cDNA encoding RNF146 WT, H53A, R163A and H53A/R163A from entry vectors described above into the pDEST 5′ BirA*-FLAG-pcDNA5-FRT-TO destination vector. These vectors were used for stable expression of BirA-tagged RNF146 variants in Flp-In-T-REx HEK293 cells (Invitrogen), by transfection as described above with subsequent selection using hygromycin. Parental Flp-In T-REx HEK293 cells (n=2), and stable cells expressing BirA*–FLAG fused to a green fluorescent protein (GFP; n=2) were used as negative controls for the BioID experiments and processed in parallel to the cells expressing RNF146 bait. Cells were grown in 15-cm dishes and treated with 1 μg ml−1 of doxycycline and 50 μM biotin for 24 h. Subsequently, cells were washed and harvested in ice-cold PBS and frozen at −80°C until purification. Cell pellets were lysed in RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP-40, 1 mM EGTA, 1 mM MgCl2, 0.1% SDS and 0.5% sodium deoxycholate plus protease inhibitors). The lysates were sonicated at 4°C using three 5-s bursts at 35% amplitude with 3 s pauses. Samples were treated with 250 U TurboNuclease (BioVision) for 30 min followed by centrifugation at 20,000 g. Lysates were then incubated with streptavidin–Sepharose (GE Healthcare, cat. no. 17-5113-01) for 3 h at 4°C and subsequently washed with 2% SDS (in 25 mM Tris-HCl pH 7.4), followed by one wash with RIPA, one wash with TNTE plus 0.1% Tween-20 (see above), and three washes with 50 mM ammonium bicarbonate (ABC). Following the final wash, 1 μg of sequencing-grade trypsin (Promega) was added and digestion was allowed to proceed overnight at 37°C. An additional 500 ng of trypsin was added and the samples were further incubated at 37°C for 3 h. Samples were centrifuged and the supernatant was collected. Formic acid was added to a final concentration of 5%. Samples were dried with a centrifugal evaporator, reconstituted with water and dried again. Samples were reconstituted in 5% formic acid for processing. Mass spectrometry and data analysis was carried out as described previously (Lambert et al., 2014). Briefly, using an Eksigent Autosampler, 5 μl of the tryptic peptides were loaded at 400 nl min−1 on to a 75 μm×12 cm fused silica capillary tubing packed with 3 μm C18 (ReproSil-PurC18-AQ). Peptides were subjected to nano liquid chromatography electrospray ionization tandem mass spectrometry (nano-LC-ESI-MS/MS), using a 90-min reversed phase (5-35% acetonitrile, 0.1% formic acid) buffer gradient, delivered at 200 nl min−1 and analysed on a Orbitral-Elite (Thermo). The instrument performed a 250 ms MS1 survey scan from 350–1500 Da followed by ten 200-ms MS2 candidate ion scans from 100–2000 Da with a 10-s dynamic exclusion.
Mass spectrometry data analysis
Raw mass spectrometry files were stored, searched and analysed using the ProHits laboratory information management system (LIMS) (Liu et al., 2010). The RAW data files were converted into MGF format and subsequently converted into an mzML format using ProteoWizard (3.0.4468). The mzML files were searched using Mascot (v2.3.02) and Comet (2014.02 rev.2), as described previously (Lambert et al., 2014). The spectra were searched against a total of 72,230 proteins consisting of the NCBI human and adenovirus complements of the RefSeq database (v57, forward and reverse sequences), supplemented with ‘common contaminants’ from the Max Planck Institute (http://220.127.116.11:8080/share.cgi?ssid=0f2gfuB) and the Global Proteome Machine (GPM; http://www.thegpm.org/crap/index.html). The database parameters were set to search for tryptic cleavages, allowing up to two missed cleavage sites per peptide, MS1 mass tolerance of 15 ppm with charges of 2+ to 4+ and an MS2 mass tolerance of ±0.15 amu. Deamidated asparagine or glutamine and oxidized methionine were selected as variable modifications. The results from each search engine were analyzed through TPP (the Trans-Proteomic Pipeline, v4.7) via the iProphet pipeline. SAINTexpress version 3.3 was used with default parameters to calculate statistical significance of each potential protein–protein interaction relative to control samples. Only proteins identified with minimally two unique peptides ions and a minimum iProphet probability of 0.95 were considered. To increase the stringency in the identification of true positives, the four controls were compressed to two prior to running SAINTexpress by selecting the two highest spectral counts for each prey protein for modeling. All RAW mass spectrometry data and downloadable identification are deposited in the MassIVE repository housed at the Center for Computational Mass Spectrometry at UCSD (http://proteomics.ucsd.edu/ProteoSAFe/datasets.jsp). The dataset has been assigned the MassIVE ID MSV000079493 and is available for FTP download at: ftp://MSV000079493@massive.ucsd.edu. The dataset was assigned the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) identifier PXD003581.
We would like to thank Frederick Vizeacoumar and the robotics facility for help with the LUMIER screen and Bob Varelas for constructs used as prey in the LUMIER screen.
C.I.C. performed or helped perform all experiments, contributed to experimental design and writing of the manuscript. P.S.-T. helped generate samples for BioID, performed the mass spectrometry, and analyzed the results and contributed to writing the manuscript. M.B.-R. helped with the LUMIER assay, provided technical assistance throughout the study and critically read the manuscript. A.D. helped with the LUMIER screen. A.-C.G. contributed to the BioID experiments and critically read the manuscript. J.L.W. designed and supervised the study and helped write the manuscript.
This work was supported by the Canadian Institutes of Health Research Foundation Scheme [grant numbers FDN-143301 to A.C.G., FDN-143252 to J.L.W.], and a Canadian Breast Cancer Foundation (Ontario region) fellowship (to C.I.C.). J.L.W. is a CIBC Chair in Breast Cancer Research.
RAW mass spectrometry data was deposited in the MassIVE repository housed at the Center for Computational Mass Spectrometry at UCSD (http://proteomics.ucsd.edu/ProteoSAFe/datasets.jsp), with the MassIVE ID MSV000079493. It is available for FTP download at: ftp://MSV000079493@massive.ucsd.edu. The dataset was assigned the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) identifier PXD003581.
The authors declare no competing or financial interests.