PTEN controls glandular morphogenesis through a juxtamembrane β-Arrestin1/ARHGAP21 scaffolding complex

PTEN controls three-dimensional (3D) glandular morphogenesis by coupling juxtamembrane signaling to mitotic spindle machinery. While molecular mechanisms remain unclear, PTEN interacts through its C2 membrane-binding domain with the scaffold protein β-Arrestin1. Because β-Arrestin1 binds and suppresses the Cdc42 GTPase-activating protein ARHGAP21, we hypothesize that PTEN controls Cdc42 -dependent morphogenic processes through a β-Arrestin1-ARHGAP21 complex. Here, we show that PTEN knockdown (KD) impairs β-Arrestin1 membrane localization, β-Arrestin1-ARHGAP21 interactions, Cdc42 activation, mitotic spindle orientation and 3D glandular morphogenesis. Effects of PTEN deficiency were phenocopied by β-Arrestin1 KD or inhibition of β-Arrestin1-ARHGAP21 interactions. Conversely, silencing of ARHGAP21 enhanced Cdc42 activation and rescued aberrant morphogenic processes of PTEN-deficient cultures. Expression of the PTEN C2 domain mimicked effects of full-length PTEN but a membrane-binding defective mutant of the C2 domain abrogated these properties. Our results show that PTEN controls multicellular assembly through a membrane-associated regulatory protein complex composed of β-Arrestin1, ARHGAP21 and Cdc42.


Introduction
PTEN (phosphatase and tensin homolog) is the second most commonly mutated tumor suppressor gene in human cancer (Cantley and Neel, 1999) and has a central role in multicellular morphogenesis (Martin-Belmonte et al., 2007;Jagan et al., 2013a;Deevi et al., 2016). While PTEN antagonizes the phosphoinositol 3-kinase (PI3K)/AKT pathway via its N-terminal phosphatase domain (Cantley and Neel, 1999), three-dimensional (3D) multicellular assembly was unaffected by forced variation of PI3K activity in colorectal organotypic model systems (Jagan et al., 2013a;Magudia et al., 2012). The PTEN domain structure includes an N-terminal phosphatase domain, a C2 domain, a C-terminal tail and a PDZ-binding domain. The C2 domain binds to membrane phospholipids by inserting a hydrophobic (CBR3) loop into the membrane bilayer and thereby provides a scaffold for juxtamembrane signaling (Lee et al., 1999). Furthermore, the PTEN C2 domain regulates polarized migration , multicellular morphology (Leslie et al., 2007;Jagan et al., 2013b) and has an important but poorly understood tumor suppressor function (Caserta et al., 2015).
Within complex systems, protein scaffolding enhances signaling efficiency by assembly of spatially distinct subcellular complexes for different cellular tasks (Weng et al., 1999;Pertz, 2010). The PTEN C2 domain binds the plasma membrane and interacts with the scaffold protein b-Arrestin1 (Lima-Fernandes et al., 2011) that in turn binds and suppresses ARHGAP21 (Anthony et al., 2011), a member of a highly conserved class of RhoGAPs (Bos et al., 2007;Anderson et al., 2008). ARH-GAP21 regulates the small GTPases, Cdc42  and RhoA (Anthony et al., 2011). These GTPases have overlapping, complementary functions required for mitotic spindle orientation and consequent control of the cell division axis, cytokinetic furrow positioning, daughter cell size and tissue morphogenesis (Morin and Bellaïche, 2011). Both Cdc42 and RhoA drive actin nucleation and cortical stiffening (Ma et al., 1998;Eisenmann et al., 2007) required for spindle orientation (Johnston et al., 2013). Furthermore, Cdc42 crosstalk with protein kinase c zeta [PRKCZ] (Noda et al., 2001;Durgan et al., 2011) localizes force generators within the cell cortex that act via astral microtubules to orientate the spindle (Hao et al., 2010). ARHGAP21 has high GAP activity for Cdc42  and its Pac-1 homologue regulates multicellular patterning in C. elegans by spatial regulation of Cdc42 (Anderson et al., 2008;Klompstra et al., 2015).
Here, we investigate PTEN spatiotemporal coordination of mammalian glandular morphogenesis through conserved juxtamembrane b-Arrestin1-ARHGAP21 interactions, using 3D colorectal cancer (CRC) model systems. To substantiate physiological relevance of these processes, we also investigate their role in morphogenesis of 3D multicellular organoids isolated from normal colon.
In particular, it collects signals from a cells' surroundings and changes where cells divide so new cells are produced in the right places. The control of cell division by PTEN is also thought to help limit the progression and spread of cancer.
PTEN can interact with another protein called b-Arrestin1, which behaves as a so-called scaffolding protein -in other words, one that helps groups of proteins to interact with each other. b-Arrestin1 has been found to control cell division via a series of other proteins, including ARHGAP21 and Cdc42. The relationship between PTEN and these other proteins in dividing cells is still not fully understood.
Javadi, Deevi et al. studied PTEN in human cells grown in the laboratory to show that a part of PTEN known as the C2 domain allows it to help organize cells by moving b-Arrestin1 to the outer edge of the cell -the cell membrane. This relocation allows b-Arrestin1 to interact with ARHGAP21 and Cdc42, and control cell division. Active Cdc42 changes the orientation of cell division, allowing cells to organize into single layers of regular cells and similar tightly controlled structures.
Further experiments revealed that these proteins are important to form tubes inside the glands of the gut. The C2 region of PTEN also helps to detect signals carried by fat molecules in the cell membrane, so these results provide a direct link between signaling and cell organization via PTEN. The work of Javadi, Deevi et al. provides new understanding of how PTEN links nutrient availability to cell organization during development and may also lead to new insights into the role of PTEN in limiting the growth of tumors. protein's densitometry value against its total lysate level, to investigate relative proportions of b-Arrestin1 and ARHGAP21 associated with membrane. We found greater b-Arrestin1 but lower ARH-GAP21 levels in Caco-2 than in ShPTEN membrane fractions ( Figure 1C,D). As b-Arrestins are known to localize to activated lysophosphatidic acid receptors [LPARs] (Urs et al., 2005;Li et al., 2009) that are expressed in Caco-2 and HCT116 cell membranes (Yun et al., 2005), we investigated effects of PTEN on lysophosphatidic acid (LPA)-induced membrane recruitment of b-Arrestin1. We found greater LPA-mediated membrane enrichment of b-Arrestin1 in Caco-2 and HCT116 cells than in PTEN-deficient ShPTEN or PTEN -/-HCT116 (PTEN -/-) subclones ( Figure 1E,F; Figure 1-figure supplements 3 and 4). We next used confocal microscopy to determine PTEN effects on b-Arrestin1 subcellular distribution in whole cells. We expressed the b-Arrestin1-mCherry fusion protein and mCherry only controls in PTEN-expressing and -deficient cells. We assessed colocalization with Alexa 488-labeled wheat germ agglutinin (WGA), a widely used fluorescent probe for cell and Golgi complex membranes (Crossman et al., 2015) by confocal microscopy. b-Arrestin1-mCherry was predominantly cytosolic in vehicle only (VO)-treated cells, in accord with cytosolic accumulation of unlabelled b-Arrestin1 in fractionation studies. On treatment with LPA, the b-Arrestin1-mCherry fusion protein colocalized with WGA at the plasma membrane in PTEN-expressing Caco-2 and HCT116 control cells ( Figure 1G (-)] treatment shown in (E) ;Caco-2 (+) vs (-)=1.74 ± 0.09; **p<0.01; ShPTEN (+) vs ShPTEN (-)=0.97 ± 0.08 vs 0.66 ± 0.07;*p<0.05 values expressed as fold change relative to Caco-2 (-) control. ShPTEN (-) vs Caco-2 (-)=0.66 ± 0.07;*p<0.05; ShPTEN (+) vs Caco-2 (+)=0.97 ± 0.08 vs 1.74 ± 0.09;**p<0.01. Green bars indicate ShPTEN cells. (G) Plasma membrane localization of the b-Arrestin1-mCherry fusion protein in Caco-2 control (top two panels) or ShPTEN (bottom two panels) after VO (-) or LPA (+) treatment. Red and green fluorescence emitted by m-Cherry and Alexa 488 labels correspond to b-Arrestin1 and WGA, respectively. Note WGA localization to plasma membranes and Golgi apparatus. Colocalization of b-Arrestin1 and WGA at the plasma membrane revealed by composite yellow signal in Merge, Z-stack images and by overlap of fluorescence intensity curves in line scans. Horizontal yellow bars indicate focal plane. All experiments in triplicate. Analyses by Student's paired t test or ANOVA with Tukey post hoc test. White scale bar 20 mm. Molecular weights indicated by arrows in blots. DOI: https://doi.org/10.7554/eLife.24578.003 The following source data and figure supplements are available for figure 1: Source data 1. Source data for Figure 1B. DOI: https://doi.org/10.7554/eLife.24578.010 Source data 2. Source data for Figure 1D. DOI: https://doi.org/10.7554/eLife.24578.011 Source data 3. Source data for Figure 1F.  While LPA had limited effects in ShPTEN cells that have residual low level PTEN ( Figure 1G), this treatment had no effects on b-Arrestin1-mCherry subcellular distribution in PTEN-null (PTEN -/-) cells ( Figure 1-figure supplement 5). mCherry only did not localize at the plasma membrane (data shown for control PTEN-expressing HCT116 and PTEN -/cells only; Figure 1-figure supplement 6). To exclude a non-specific effect of PTEN on ligand-mediated protein translocation to the cell membrane, we investigated 1,25(OH) 2 D 3 -mediated membrane localization of E-Cadherin (Pálmer et al., 2001) in Caco-2 and ShPTEN cells. We found that 1,25(OH) 2 D 3 treatment induced equivalent E-Cadherin translocation to the plasma membrane in PTEN-expressing and -deficient cells, compared to control VO treatment (data not shown). Collectively, these findings show that PTEN functions within a regulatory scaffolding network that couples b-Arrestin1 to ARHGAP21 at the plasma membrane.
PTEN promotes b-Arrestin1 and ARHGAP21 interactions through its C2 domain b-Arrestin1 has previously been shown to bind the PTEN C2 domain directly and modulate PTEN function (Lima-Fernandes et al., 2011). To investigate PTEN regulation of b-Arrestin1-ARHGAP21 interactions, we conducted co-immunoprecipitation (CoIP) studies and normalized b-Arrestin1-associated ARHGAP21 against total ARHGAP21 densitometry values in cell lysates. Here, we show greater b-Arrestin1-associated ARHGAP21 levels in PTEN-expressing Caco-2 or HCT116 cells versus ShPTEN or PTEN -/cells or IgG negative controls ( Figure 3A,B). To investigate involvement of PTEN catalytic and noncatalytic domains in these processes, we conducted transient expression studies of GFP-labeled full-length wild type (wt) PTEN or mutants ( Figure 3C), in PTEN-deficient cells. Mutants included full-length PTEN with a mutation at the CBR3 membrane-binding loop within the C2 domain (PTEN-MCBR3), full-length phosphatase-dead (PTEN C124S-based) constructs with mutations in key C-terminal phosphorylation sites, namely PTEN C124S-T383A (CS-T383A) and PTEN C124S-A4 (CS-A4 with S380A, T382A, T383A and S385A mutations combined). CS-T383A has been proposed to contain an unmasked C2 domain ) that effectively binds b-Arrestin1 while CS-A4 lacks b-Arrestin1 binding capacity (Lima-Fernandes et al., 2011). We also used the isolated PTEN C2 domain (C2) and a membrane-binding mutant of the C2 domain (C2-MCBR3). We found that expression of C2 enhanced b-Arrestin1-associated ARHGAP21 levels in CoIPs conducted in ShPTEN ( Figure 3D figure supplements 1 and 2). b-Arrestin1-associated ARHGAP21 levels were normalized against total ARHGAP21 expression in cell lysates. Collectively, these findings indicate that the membrane-binding function of PTEN is important for scaffolding ARHGAP21 and b-Arrestin1.
We then used an intramolecular bioluminescence resonance energy transfer (BRET)-based PTEN biosensor Misticone et al., 2016) to test if the full-length C124S C-terminal phosphorylation mutants ( Figure 3C) that have different b-Arrestin1 binding capacities (Lima-Fernandes et al., 2011), display different conformations. The biosensor contains PTEN sandwiched between the energy donor Renilla luciferase (Rluc) and the energy acceptor YFP. Changes in the BRET signal depend on the relative distance and orientation of the donor and acceptor proteins within the fusion and therefore provide readout for conformational change of PTEN in live cells (  orientation and inhibits single lumen formation. High-power (HP) spindle views (orange border) enlarge areas within white rectangles and show orientation angles (interrupted white arrows) of spindle planes (double-headed solid white arrows) toward gland centres (GCs). Normal spindle planes are orientated at approximately 90 0 angles relative to gland centres [GCs] (Jaffe et al., 2008). Summary SiRNA effects on spindle angles relative to GCs are shown in Figure  Source data 1. Source data for Figure 2B. DOI: https://doi.org/10.7554/eLife.24578.020 Source data 2. Source data for Figure 2D.         . Membrane fraction values were normalized against total expression of each protein in lysate. Furthermore, expression of C2 but not C2-MCBR3 also increased b-Arrestin1-associated ARHGAP21 levels in PTENdeficient cell membrane fractions. b-Arrestin1-associated ARHGAP21 levels were normalized against total ARHGAP21 in the membrane fraction ( Figure 4G-J). In PTEN -/cells, expression of C2 and full-length PTEN had greater effects on b-Arrestin1-associated ARHGAP21 levels than PTEN-MCBR3, C2-MCBR3 or control ( Figure 4I,J). PTEN-MCBR3 had small but significant effects on b-Arrestin1-associated ARHGAP21 levels in excess of control ( Figure 4I,J). Taken together, these data indicate that PTEN enhances b-Arrestin1 membrane recruitment and b-Arrestin-ARHGAP21 interactions, predominantly through its membrane-binding C2 domain.

PTEN controls mitotic spindle orientation and 3D morphogenesis by regulation of b-Arrestin1
To investigate PTEN coordination of morphogenic processes through its C2 domain, we conducted 3D organotypic culture studies. We found greater expression and membrane localization of b-Arrestin1 in 3D control PTEN-expressing Caco-2 cultures compared to ShPTEN cultures ( Figure 5A, PTEN controls 3D Caco-2 morphogenesis by noncatalytic coupling of b-Arrestin1, ARHGAP21 and Cdc42 Precise spatiotemporal coordination of Cdc42 activity is central to multicellular morphogenesis (Meitinger et al., 2013). To investigate PTEN non-catalytic regulation of Cdc42 via b-Arrestin1 and ARHGAP21, we conducted transfection and peptide inhibitor studies. Expression of the catalytically inactive PTEN CS-T383A construct that binds b-Arrestin1 but not the PTEN CS-A4 binding-defective mutant, enhanced Cdc42-GTP levels in PTEN -/cells ( Figure 6A,B). While Cdc42 can be inhibited by ARHGAP21 , competitive b-Arrestin1 binding to the GAP domain can release the active GTPase from ARHGAP21 inhibition (Anthony et al., 2011). To investigate the specific role of b-Arrestin1-ARHGAP21 interactions on Cdc42-dependent 3D morphogenesis, we used a cell-permeant 24-mer peptide analogue of the ARHGAP21 GAP domain that was designed to disrupt the b-Arrestin1-ARHGAP21 interaction ( Figure 6C) (Anthony et al., 2011). Here, we show that treatment with this b-Arrestin1-ARHGAP21 peptide binding inhibitor (pep24) but not a scrambled control peptide attenuated the association between b-Arrestin1 and ARHGAP21, resulting in lower levels of b-Arrestin1-associated ARHGAP21 ( Figure 6D       To avoid any compromise of experimental interpretation by intrinsic Caco-2 cancer cell mutations, we investigated our key observations from cell culture experiments in organoids formed from primary normal murine colon cells (Clevers, 2016). In this study, we cultured colonic crypt progenitor epithelium in Matrigel supplemented with growth factors as previously described (Sato et al., 2011). By these methods, we generated colorectal organoids with appropriate mitotic spindle orientation, apical membrane alignment, luminogenesis and epithelial organization, in 3D cultures ( Figure 7A-C). To investigate the role of b-Arrestin1-ARHGAP21 interactions on colorectal homeostasis, we assessed effects of pep24 vs control peptide treatment on 3D organoid morphogenesis. Here, we show that pep24 treatment perturbed mitotic spindle orientation, disrupted 3D glandular morphology and lumen formation in normal colorectal organoids ( Figure 7A-C). Conversely, control peptide treatment had no discernible effect on 3D glandular morphology ( Figure 7A-C). Collectively, these findings highlight a significant role for b-Arrestin1-ARHGAP21 interaction in multicellular morphogenesis of normal colorectal epithelium.
As well as protein abundance, stoichiometry and post-translational targeting machinery modulate the assembly of spatially restricted scaffolding complexes (Boisvert et al., 2012). LPAR is a lysophosphatidic acid (LPA) activated GPCR that couples heterotrimeric G proteins, to control membrane recruitment of b-Arrestin1 (Urs et al., 2005;Li et al., 2009), GTPase activity (Ueda et al., Source data 1. . Figure 6B -Cdc42-GTP after transfection -source data DOI: https://doi.org/10.7554/eLife.24578.076 Source data 2. Figure 6E -Peptide inhibitor treatment effects on Beta Arrestin1:ARHGAP21 binding in Caco-2 DOI: https://doi.org/10.7554/eLife.24578.077 Source data 3. Figure 6G Peptide inhibitor treatment effects on Cdc42-GTP in Caco-2 DOI: https://doi.org/10.7554/eLife.24578.078     (Nagasaki and Gundersen, 1996). To investigate these phenomena, we assessed spontaneous and LPA-mediated cell membrane localization of b-Arrestin1 and ARHGAP21 in subcellular fractions of PTEN-expressing and -deficient cells. We found proportionately greater differences of constitutive and LPA-mediated b-Arrestin1 membrane localization in PTEN-expressing Caco-2 vs PTEN-deficient ShPTEN cells, indicative of PTEN involvement in b-Arrestin1 membrane recruitment. Similarly, LPA promoted greater b-Arrestin1 membrane localization in PTEN-expressing HCT116 cells than in PTEN -/cells. In both cell types, we found reciprocal differences of ARHGAP21 membrane localization, consistent with b-Arrestin1-mediated suppression (Anthony et al., 2011). To investigate PTEN effects on b-Arrestin1 plasma membrane recruitment in whole cells, we used confocal microscopy to track the spatial distribution of transfected b-Arrestin1-mCherry against plasma membrane localization of Alexa 488-labeled WGA. Line scans of fluorescence intensities across image focal planes and high-resolution confocal z-stack reconstructions (Furia et al., 2014) revealed that PTEN enhanced LPA-mediated b-Arrestin1-mCherry colocalization with Alexa 488 -labelled WGA at the plasma membrane. LPA treatment had limited effects on b-Arrestin1-mCherry plasma membrane recruitment in ShPTEN cells that have residual low level PTEN but was ineffective in PTEN -/cells. We cannot attribute these findings to nonspecific fluorophore diffusion, since mCherry distribution was unaffected by PTEN status or LPA treatment. Furthermore, we can exclude PTEN nonspecific effects on cell membrane protein localization because 1,25 (OH) 2 D 3-induced plasma membrane recruitment of E-Cadherin (Pálmer et al., 2001) was equivalent in PTEN-expressing and -deficient cells. Collectively, these findings show that PTEN is an essential coregulator of plasma membrane recruitment of b-Arrestin1 and of b-Arrestin1:ARHGAP21 functional interactions.
b-Arrestin1 and ARHGAP21 coregulate GTPases (Anthony et al., 2011) that function as a signaling hub for diverse cytokinetic processes (Glover et al., 2008;Jaffe et al., 2008). We investigated PTEN regulation of GTPase activity through b-Arrestin1-ARHGAP21 scaffolding and conducted perturbation experiments in organotypic 3D culture models that are ideal for precise, image-based assays of multiscale epithelial homeostasis. We found that siRNA knockdown (KD) of b-Arrestin1 suppressed Cdc42 activation in PTEN-expressing cells and 3D cultures while siRNA ARHGAP21 KD had reciprocal effects by increasing Cdc42 activity in PTEN-deficient cultures. Furthermore, these perturbation experiments affected sequential layers of homeostatic controls. Suppression of b-Arrestin1 induced mitotic spindle misorientation, abnormal epithelial configuration, defective apical membrane positioning and formation of multiple lumens during assembly of PTEN-expressing 3D Caco-2 glandular structures. Conversely, ARHGAP21 KD increased Cdc42 activation, restored spindle orientation and rescued aberrant morphogenesis of isogenic PTEN-deficient 3D ShPTEN cultures. Because of the previously reported relationship between ARHGAP21 and RhoA (Lima-Fernandes et al., 2011), we assessed relationships between PTEN, b-Arrestin1, ARHGAP21 and RhoA. PTEN KD suppressed b-Arrestin1, enhanced ARHGAP21 and suppressed RhoA activation. Collectively, these findings indicate that PTEN orchestrates the cell division plane and apical membrane dynamics during multicellular morphogenesis by coordination of b-Arrestin1-ARHGAP21 functional interactions and GTPase activity.
C2 domains are phospholipid and protein binding modules involved in membrane recruitment and localization of signaling molecules (Corbalán-García and Gó mez-Fernández, 2010). Here, we show that PTEN C2 enhances b-Arrestin1 expression. While precise mechanisms remain unclear, PTEN C2 binds thioredoxin-1 (Meuillet et al., 2004) that regulates b-Arrestin1 in a context-dependent manner (Jia et al., 2014). In this study, PTEN C2 also promoted membrane enrichment of b-Arrestin1 in excess of total lysate concentrations and enhanced b-Arrestin1-ARHGAP21 interactions. Interestingly, wt PTEN promoted greater b-Arrestin1-associated ARHGAP21 expression in membrane fractions than the isolated C2 domain. Within its N-terminal domain, full-length PTEN contains a conserved polybasic phosphatidylinositol[ 4,5 ] biphosphate (PtdIns [ 4,5 P 2 ]) [PIP 2 ]-binding site that participates in membrane-targeting (Walker et al., 2004) and could complement PTEN C2 domainmediated interactions between scaffold complexes and membrane lipids. Similarly, in CoIP assays PTEN-MCBR3 promoted a small but significant increase of b-Arrestin1-associated ARHGAP21 expression in excess of C2-MCBR3. Taken together, our findings indicate that PTEN promotes b-Arrestin1-ARHGAP21 interactions predominantly through its C2 domain, although the PTEN N-terminal domain has a weak additional effect.
PTEN, b-Arrestins and GTPase-activating proteins modulate the activity of Rho GTPases CdcC42 and RhoA (Martin-Belmonte et al., 2007;Anthony et al., 2011;Bos et al., 2007;Anderson et al., 2008). Furthermore, the PTEN C2 domain has morphogenic properties (Leslie et al., 2007;Jagan et al., 2013b;Caserta et al., 2015). We investigated PTEN orchestration of multicellular gland assembly through its C2 domain, in organotypic culture studies. We found greater b-Arrestin1 and lower ARHGAP21 immunoreactivity in control Caco-2 vs ShPTEN 3D cultures in accord with our findings in cell monolayers. Transfection of ShPTEN cells with a C2 domain construct promoted b-Arrestin1 membrane localization, rescued mitotic spindle orientation, single central lumen formation and 3D multicellular morphogenesis. Conversely, expression of the membrane-binding mutant C2-MCBR3 domain did not rescue the morphology phenotype. ShPTEN cells stably express PTEN ShRNA that targets the phosphatase domain (Jagan et al., 2013a). To assess PTEN ShRNA specificity and test for potential off-target effects, we investigated effects of full-length ShRNA-resistant PTEN (ShR PTEN) on the integrated ShPTEN 3D morphology phenotype. Expression of ShR PTEN rescued 3D ShPTEN morphogenesis and thus confirmed shRNA functional specificity. Collectively, the above data show that PTEN has important noncatalytic morphogenic functions mediated through its C2 domain and b-Arrestin1 membrane targeting.
PTEN is frequently downregulated in human colorectal (Naguib et al., 2011) and other cancers, even in the absence of genetic loss or mutation (Salmena et al., 2008). Regulatory cues may have a central role in tumorigenesis (Song et al., 2012). Activated GPCRs modulate b-Arrestin1 conformation (Shukla et al., 2008), membrane recruitment (Urs et al., 2005;Li et al., 2009;Décaillot et al., 2011) and PTEN-b-Arrestin1 interactions (Lima-Fernandes et al., 2011). By these processes, GPCRs may influence b-Arrestin1-dependent GTPase activation, cytoskeletal dynamics and neoplastic multicellular patterning. Studies in cancer cell models provide useful mechanistic data (Hanahan and Weinberg, 2011) but intrinsic mutations may compromise physiological relevance. Studies of 3D multicellular organoids isolated from normal tissues have provided basic insights into normal tissue morphogenesis (Clevers, 2016). We have previously generated intestinal crypt organoids for study of multicellular assembly, patterning and lineage commitment (Campbell et al., 1993;Tait et al., 1994;Slorach et al., 1999). Suppression of b-Arrestin1-ARHGAP21 binding in organoid systems by pep24 treatment perturbed spindle orientation and apical membrane alignment to induce a multilumen phenotype, surrounded by disorganized epithelium. Collectively, these findings demonstrate the importance of b-Arrestin1-ARHGAP21 interactions in control of normal colorectal multicellular architecture.
Within the PTEN C2 domain, the CBR3 loop can localize cytoplasmic PTEN to early endosomes arranged along the microtubule cytoskeleton, by binding endosomal PIP 3 (Naguib et al., 2015). Restriction of PTEN to a punctate vesicular distribution along microtubules may enable dephosphorylation of PIP 3 signals generated by plasma membrane receptor tyrosine kinases and parcelled in endosomes (Naguib et al., 2015). However, it is difficult to envisage that wide endosomal distribution of scaffolding interactions along the microtubule cytoskeleton could regulate the compartmentalized focus of GTPase activity (Pertz, 2010) required for control of spindle dynamics and multicellular morphogenesis (Durgan et al., 2011). Hence, dephosphorylation of PIP 3 on endosomes and scaffolding of b-Arrestin1-ARHGAP21 may represent spatiotemporally distinct PTEN tumor suppressor functions.
This study shows that b-Arrestin1-ARHGAP21 interactions represent an essential component of the PTEN morphogenic pathway and sheds light on conserved developmental mechanisms. In C. elegans, PTEN/DAF18 conducts nutrient-sensing through its phosphatase domain (Ogg and Ruvkun, 1998) in a negative feedback loop with the insulin/IGF axis (Narbonne et al., 2015) and casein kinase II [CKII] (Liu Tj et al., 2001). CKII phosphorylates PTEN to induce the closed conformation (Rahdar et al., 2009;Torres and Pulido, 2001) that suppresses plasma membrane binding (Rahdar et al., 2009). We show that the PTEN membrane-binding C2 domain is essential for multicellular morphogenesis. Hence, our findings may provide a rationale for PTEN multifaceted control of embryonic development by nutrient-sensing (Ogg and Ruvkun, 1998) and regulation of morphogenic growth (Rouault et al., 1999) according to the available nutrient energy balance (Hietakangas and Cohen, 2009). Our study also has oncological relevance, since disruption of PTEN C2 domain-mediated b-Arrestin1-ARHGAP21 interactions drive evolution of morphology phenotypes in 3D cultures that are evocative of colorectal cancer (Deevi et al., 2016;Jaffe et al., 2008). Dissection of these phenomena may yield novel targets for therapy aimed at suppression of aggressive cancer morphology phenotypes that predict early metastasis.

Reagents and antibodies
All laboratory chemicals were purchased from Sigma-Aldrich, Dorset, England unless otherwise stated. RNAiMAX and X-tremeGENE transfection reagents were purchased from Thermofisher, Dublin, Ireland and Roche, Basel, Switzerland, respectively. Antibodies used in this study were anti-b-Actin ( (Anthony et al., 2011)] and scrambled control peptide were purchased from EZ Biolabs, Carmel, IN 46032 USA. Pep24 and control peptides were prepared in dimethyl sulfoxide (DMSO), according to manufacturer's instructions. For the pep24 experiments, cells were incubated in 2D or 3D cultures as outlined below for 48 hr, then treated with either 10 mM pep24 or 10 mM control peptide. Incubations were continued for 24 hr for assays of protein binding or Western blots in cell monolayers. In 3D morphogenesis assays, test and control peptides were added to the media in the above concentrations, changed at 48 hr intervals and effects on morphogenesis assessed at 4 days of culture.

Cell transfection
We carried out mammalian SiRNA and DNA transfections using RNAiMAX and X-tremeGENE transfection reagents respectively, according to manufacturer's protocols. Cells were plated at 2 Â 10 5 cells/35 mm dish for 24 hr, then transfected with 10 mM siRNA or 500 ng DNA/2 Â 10 5 cells for all respective siRNA oligonucleotides or DNA constructs. Cells were incubated with RNA/RNAiMAX or DNA/X-tremeGENE lipofectamine complexes for 48 hr, before lysis and probing. In 2 ShPTEN cells, the stably expressed PTEN ShRNA targets a 58 base pair region within the PTEN phosphatase coding region and C2 domain constructs are unaffected (Boehm et al., 2005). In membrane localization studies, PTEN-expressing and -deficient Caco-2 and HCT116 clones were transfected with b-Arrestin1-mCherry against mCherry only controls. In expression, co-immunoprecipitation and morphogenesis studies, ShPTEN cells were transiently transfected with empty vector (EV) only, the isolated PTEN C2 domain (C2) or a C2 domain construct mutated at the CBR3 membrane-binding loop [C2-MCBR3] (Lee et al., 1999) in pEGFP expression vectors.

Cell fractionation
We conducted these experiments using a subcellular fractionation kit (Thermo Fisher Scientific, Dublin, Ireland) according to manufacturer's protocol. Briefly, cells were trypsinized and lysed in cytoplasmic extraction buffer for 10 min at 4˚C, then centrifuged at 500 g for 5 min. The supernatant was collected as the cytoplasmic fraction while the pellet was resuspended in membrane extraction buffer, vortexed for 5 s and mixed gently for 10 min at 4˚C. The mix was centrifuged at 3000 g for 5 min and the supernatant was collected as the membrane fraction. In separate experiments, we conducted protein extraction, Western blotting and co-immunoprecipitation (Co-IP) assays in isolated cell membrane and cytosolic fractions. Equivalent amounts of membrane fraction and cytosol were loaded in immunoblots and Co-Ips.

Protein extraction and western blotting
As previously described (Jagan et al., 2013a;Jagan et al., 2013b), proteins were resolved using gel electrophoresis, followed by blotting onto nitrocellulose membranes. Membranes were probed using antibodies as indicated in the text. Experiments were repeated in triplicate.

Co-immunoprecipitation (Co-IP)
Cells were lysed on ice in buffer containing 100 mM Tris-HCl, pH 7.5, 1% Triton X-100, 5 mM EDTA, 5 mM EGTA, 50 mM NaCl, 5 mM NaF, 1 mM Na 3 VO 4 and protease inhibitor. Cell lysates were centrifuged (for 10 mins at 15,000 g) and protein concentrations were measured by the BCA method. 1000 mg of protein was precleared overnight with control IgG and 15 ml of Protein A/G Sepharose beads (Santa Cruz, Dallas, Texas, USA). The protein was then immunoprecipitated with the appropriate antibody-beads conjugate and incubated on a rotating wheel for 2 hr. The beads were collected by centrifugation and washed five times in wash buffer (50 mM Hepes, pH 7.4, 1% Triton X-100, 0.1%, SDS, 150 mM NaCl, 1 mM Na 3 VO 4 ). The beads were subsequently resuspended in 40 ml Laemmli sample buffer and processed for gel electrophoresis.

Proximity ligation assays
We assessed protein-protein proximities using the Duolink II red kit (Sigma-Aldrich, Dorset, UK) according to the manufacturer's instructions. Briefly, we transfected PTEN -/cells with GFP tagged-EV or -PTEN constructs and cultured the cells in Millipore eight well chambers. After 24 hr, we fixed the cells with 4% paraformaldehyde (PFA) at room temperature for 20 min. We then permeabilized the cells with 0.05% TritonX100 in PBS for 10 min. Cells were blocked with immunofluorescence (IF) buffer (Duolink, Sigma-Aldrich, Dorset), England for 2 hr according to manufacturer's instructions and incubated with primary antibody overnight at 4˚C. Cells were washed twice with buffer A, and incubated with PLA probe, ligase and polymerase according to the manufacturer's protocol. Finally, cells were washed with buffer B and slides were mounted with a cover slip using Duolink in situ mounting medium with DAPI.

BRET assays
BRET investigations were performed as described previously . Briefly, HEK cells were transfected with the indicated plasmids 24 hr after seeding. At 24 hr post transfection, cells were detached, resuspended in full media, and distributed into poly-l-orthinine coated white 96-well optiplates (Perkin Elmer). The following day, cells were washed with PBS and then overlayed with HBSS. Coelenterazine h was added to a final concentration of 5 mM and incubated for 3 min at 25˚C. BRET readings were collected using a Multimode Reader Mithras 2 LB 943 (Berthold Technologies). Substrate and light emissions were detected at 480 nm (Rluc) and 540 nm (YFP) for 1 s. The BRET signal was calculated by ratio of the light emitted by YFP and the light emitted by Rluc (YFP/Rluc). The ratio values were corrected by substracting background BRET signals detected when Rluc-PTEN was expressed. mBRET values were calculated by multiplying these ratios by 1000. DmBRET values are shown to demonstrate the shift in BRET signal compared to wt signal, which is set to zero, or between the two mutants (C124S-T383A and C124S-A4) that were tested.

Three-dimensional (3D) cultures and morphogenesis assays
Caco-2 and Caco-2 ShPTEN cells were cultured and embedded in Matrigel matrix (BD Biosciences, Oxford, UK), then imaged by confocal microscopy as we previously described (Jagan et al., 2013a;Jagan et al., 2013b). Briefly, 6 Â 10 4 trypsinized cells were mixed with Hepes buffer (20 mM, pH 7.4) and Matrigel (40%) in a final volume of 100 ml, placed in each well of eight-well multichambers (BD Falcon, Fisher Scientific, Dublin, Ireland), allowed to solidify for 30 min at 37˚C and subsequently overlayed with 400 ml of media/well. We imaged the 3D cultures at progressive stages of morphogenesis as previously described (Jagan et al., 2013a;Jagan et al., 2013b).

Colorectal organoid cultures
We used C57B/6 wild-type mice (1-6 weeks old) for experiments and conducted all animal procedures in accordance with local and national regulations. We isolated organoids as previously described (Tait et al., 1994;Slorach et al., 1999). Briefly, murine colons were opened longitudinally, cut into 0.5 cm fragments, washed 7-10 times in 1x HBSS (low calcium, low magnesium (Gibco-BRL), 2% D-glucose, 0.035% NaHCO 3 ) to remove all luminal contents. The fragments were then finely chopped with a scalpel and digested in HBSS solution containing collagenase and dispase I neutral proteases (Sigma-Aldrich, Dorset, UK) at 1 mg/ml for 20 min at room temperature on a shaking platform. Digestion was stopped by the addition of 30 ml DMEM/F12 culture medium (Life Technologies, Renfrew UK) supplemented with 5% FCS containing penicillin and streptomycin. Large fragments and muscle sheets were allowed to settle to the bottom of the flask. We removed the supernatant containing the organoids and centrifuged it for 3 min at 250 rpm, to pellet the organoids. We removed the supernatant and gently resuspended the organoid pellet in 20 ml of the DMEM/F12 solution. We repeated the centrifugation step 5-6 times until the pellet contained a homogeneously sized organoid preparation. Organoids thus prepared were resuspended in a 2x volume of Matrigel (growth factor reduced, phenol red free; BD Biosciences, Oxford UK) supplemented with 50 ng/ml murine EGF, murine Noggin 100 ng/ml (PeproTech, NJ, USA) and 1 mg/ml human R-Spondin, as indicated for organoid culture (Sato et al., 2011). Eight well multichambers were coated with a thin layer of undiluted Matrigel and allowed to solidify. Organoid preparations in Matrigel (100 ml suspension) were placed into each well, then overlaid with 250 mL/well culture medium (Dulbecco's modified Eagle medium/F12) supplemented with penicillin/streptomycin, 10 mmol/L HEPES, Glutamax supplements 1Â N2, 1 Â B27 [Invitrogen], 1 mmol/L N-acetylcysteine [Sigma]), 50 ng/ml murine EGF, Noggin 100 ng/ml and 1 mg/ml human R-Spondin (Sato et al., 2011). We cultured the organoids for 4 days with peptide treatments as defined.

Confocal immunofluorescence microscopy
Membrane and cytosolic localization of b-Arrestin1-mCherry or mCherry only were imaged against Alexa 488-labelled WGA, a widely used fluorescent marker that binds to cell membranes (Crossman et al., 2015) in Caco-2, ShPTEN, HCT116 and PTEN À/À cells, with or without LPA treatment. We used a Leica SP8 confocal microscope and Leica LAS-X software for line scanning of fluorescent images. Caco-2 ShPTEN (ShPTEN) glands and organoid cultures were incubated in 4% PFA for 20 min and processed for immunofluorescence as previously described (Jagan et al., 2013a;Deevi et al., 2016;Jagan et al., 2013b). Briefly, 3D cultures were fixed in PFA for 20 min at room temperature and permeabilized for 10 min in 0.5% Triton X-100 in PBS. The 3D cultures were rinsed with PBS/glycine buffer for 15 min to reduce autofluorescence and blocked by incubation in IF Buffer (PBS with 0.1% bovine serum albumin, 0.2% Triton X-100, 0.05% Tween-20)+10% goat serum, for 1-1.5 hr at room temperature. Primary antibodies were diluted in blocking buffer and incubated overnight at 4˚C. The 3D cultures were incubated with secondary antibodies and/or FITC-labeled phalloidin for 1 hr. DNA was stained using Vectashield mounting medium containing DAPI (Vector Scientific, Belfast, NI). Sequential scan images were taken the midsection of glands/organoids at room temperature using a Leica SP5 confocal microscope [RRID:SCR_012314] on a HCX PL APO lambda blue 63 Â 1.40 oil immersion objective at 1x or 2x zoom. Images were collected and scale bars added using LAS AF confocal software (Leica) [RRID:SCR_013673]. We assessed effects of transfection or treatment on signal intensity, spindle orientation, lumen formation and/or epithelial configuration in in 3D glands or organoids at 4 days of culture. Because imaging for apical protein kinase C zeta (PRKCZ) was unsuccessful in organoids, the apical domain was imaged using FITClabelled phalloidin as a marker of apical actin.

Assessment of mitotic spindle orientation
In cultured cells, centrosomes (Csms) were identified using anti-pericentrin (PCN) and microtubules by anti-a-Tubulin antibodies, respectively, and we identified chromosomal DNA by DAPI staining. We defined bipolar mitotic spindle architecture by convergence of microtubules towards each of 2 spindle poles, as we previously described (Deevi et al., 2016). Caco-2 and Caco-2 shPTEN glands were cultured in Matrigel for 4 days, fixed with 4% PFA and stained with anti a-Tubulin, PRKCZ and PCN primary antibodies. Gland midsections were imaged by confocal microscopy to identify cells containing well-formed mitotic spindles, during metaphase or anaphase. Lines connecting each spindle extremity were drawn using ImageJ and the line center was considered as the spindle midpoint. Angles between spindle planes and lines connecting spindle midpoints to gland centres were measured, as outlined previously (Deevi et al., 2016;Jaffe et al., 2008). We used a similar approach or imaging of organoid morphogenesis and identified apical domains and spindles using FITC-labeled phalloidin and anti-a-Tubulin antibodies, respectively.

Image processing and statistical analysis
Confocal microscopy images were processed using Leica Fw4000 Imaging software and cropped using Adobe Ilustrator [RRID:SCR_014198]. Confocal images were processed, merged and mean area quantified using LAS AF Leica imaging software, as previously described (Jagan et al., 2013a;Jagan et al., 2013b). We assessed lumen formation, spindle orientation and signal intensities using 50, 15 or 10 Â 3D Caco-2 or ShPTEN glandular cultures (glands) in triplicate for each experimental condition, respectively. We selected glands with mitotic figures for spindle orientation assays. Organoids were fewer in number and we assessed lumen formation and spindle orientation in 10 organoids per experimental condition, in triplicate. Multicellular structures with single central lumens were expressed as a percentage and spindle orientation angles relative to gland centres were calculated using ImageJ.

Data analysis
Descriptive statistics were expressed as the mean ± sem. Statistical analyses were by one or two-way ANOVA with the Tukey post hoc test or Student's paired t test using Graphpad Prism software (v5; Graphpad CA 92037 USA [RRID:SCR_002798]). Scatterplots and bar charts were used for display of quantitative numerical or categorical data.