Research ArticleCell Biology

Structural basis for the interaction between the cell polarity proteins Par3 and Par6

See allHide authors and affiliations

Science Signaling  13 Feb 2018:
Vol. 11, Issue 517, eaam9899
DOI: 10.1126/scisignal.aam9899

Building a polarity scaffold

Cell types often exhibit asymmetric distributions of proteins, lipids, and organelles that are linked to specialized functions, and loss of polarization precedes pathological conditions such as cancer. The cell polarity proteins Par3 and Par6 and the kinase aPKC are critical for generating asymmetry in cells. Par3 and Par6 contain multiple PDZ domains, which bind to short motifs to enable protein-protein interactions. Using biochemical analyses, cultured cells, and fly embryos, Renschler et al. identified a PDZ domain–binding motif in Par6 that bound to two of the three PDZ domains in Par3. These results suggest that one Par3 protein could bind to two Par6 proteins, which could serve as the basis for polarity-generating protein scaffolds in cells.

Abstract

Polarity is a fundamental property of most cell types. The Par protein complex is a major driving force in generating asymmetrically localized protein networks and consists of atypical protein kinase C (aPKC), Par3, and Par6. Dysfunction of this complex causes developmental abnormalities and diseases such as cancer. We identified a PDZ domain–binding motif in Par6 that was essential for its interaction with Par3 in vitro and for Par3-mediated membrane localization of Par6 in cultured cells. In fly embryos, we observed that the PDZ domain–binding motif was functionally redundant with the PDZ domain in targeting Par6 to the cortex of epithelial cells. Our structural analyses by x-ray crystallography and NMR spectroscopy showed that both the PDZ1 and PDZ3 domains but not the PDZ2 domain in Par3 engaged in a canonical interaction with the PDZ domain–binding motif in Par6. Par3 thus has the potential to recruit two Par6 proteins simultaneously, which may facilitate the assembly of polarity protein networks through multivalent PDZ domain interactions.

INTRODUCTION

Most cell types in multicellular organisms require an unequal distribution of proteins, lipids, and mRNA for their function (1). This phenomenon is referred to as cell polarity. During development, the differential segregation of cell fate determinants (anterior-posterior polarity) forms the basis for asymmetric cell division and ultimately the generation of body axes and cell diversity. In differentiated cells, partitioning of the cell membrane into functionally discrete domains enables cells to execute their specialized functions within an organism. Epithelia are a classic example of polarized cells and display apicobasal polarity. Whereas the apical side faces the exterior or lumen, the basal domain is directed toward the interior extracellular matrix and interstitial space in tissues and organs. The lateral cell surface allows neighboring epithelial cells to adhere and to communicate through cell-cell junctions. However, epithelial cells can also lose their polarity and acquire a migratory, mesenchymal character. Although this process is vital for embryonic development and tissue repair, it is also a key event in cancer metastasis. Elucidating the basic mechanisms underlying cell polarity thus has broad implications for our understanding of developmental and oncogenic processes.

Although polarized cells vary substantially in their morphology and function, polarity depends in all cell types on the Par (partitioning-defective) protein complex that is conserved across the animal phyla (2). The Par complex constitutes the major molecular machinery for creating the mutually exclusive signaling networks that cover large areas of the cell cortex in polarized cells. In epithelia, the Par complex is a key polarity regulator at the apical surface and engages in intricate signaling networks with the Crumbs complex and cell adhesion proteins, such as E-cadherin and nectins [Shotgun (Shg) and Echinoid (Ed), respectively, in Drosophila], to form and maintain cell junctions (1, 3).

The Par complex comprises atypical protein kinase C (aPKC) and two scaffolding proteins, Par3 [Bazooka (Baz) in Drosophila] and Par6 (Fig. 1A). Although a single set of genes encodes the Par complex in invertebrates, this gene family has expanded to two genes encoding Par3 and Par3L, two genes encoding aPKC isoforms (PKCλ/ι and PKCζ), and three genes encoding Par6α, Par6β, and Par6γ in vertebrates (46). Par6 and aPKC heterodimerize with high affinity through their N-terminal PB1 (Phox and Bem1) domains (Fig. 1A) (7, 8). In addition, Par6 contains a Crib (Cdc42/Rac-interactive binding) motif that is N-terminal to a PDZ protein-protein interaction domain. Par3 contains an N-terminal PB1-like homo-oligomerization domain (9), three central PDZ domains, and a C-terminal region with an aPKC phosphorylation site (1013). Whether Par3 interacts directly with Par6 has remained controversial. The Par3 PDZ1 domain has been described to heterodimerize with the Par6 Crib-PDZ domain (Fig. 1A) (1416). However, these reports disagree on whether the Par6 Crib motif is essential for the Par3:Par6 interaction. Moreover, the in vivo relevance of this interaction has been debated (16), and last, the Par3:Par6 interaction has been reported to be indirect, requiring aPKC as a linker molecule (7, 10). Given these conflicting observations, we sought to investigate the mechanism of Par3:Par6 association on a structural and functional level.

Fig. 1 The Par6 C terminus contains a class II PBM that interacts with the Par3 PDZ1 domain.

(A) Reported interactions (arrows) within the Par (partitioning-defective) complex. The controversial PDZ:PDZ interaction is indicated with a dashed arrow. The Par6 PDZ-binding motif (PBM) identified in this study is boxed in red. aPKC, atypical protein kinase C; PB1, Phox and Bem1; NTD, N-terminal domain; KBM, kinase binding-motif; KD, kinase domain. (B) Amino acid sequences of the eight C-terminal residues of vertebrate and invertebrate Par6 and known cell adhesion interaction partners (2529, 6668). Conserved hydrophobic (φ) residues are in dark pink for the 0 and −2 positions and in light pink for the −3 position, whereas polar residues at the −1 position are in blue. Organism abbreviations are expanded in fig. S3. PBMs used in this study are underlined. (C) Overlay of a representative region of the 1H,15N–heteronuclear single-quantum coherence (HSQC) spectra of the dPar3 PDZ1 domain in the absence (black) or presence of increasing stoichiometric amounts of dPar6 PBM as indicated. Dashed lines indicate the directions of chemical shift changes. Chemical shift assignments are shown for the most affected peaks. Peaks broadened beyond detection upon ligand binding are underlined. Residue numbers correspond to the respective nuclear magnetic resonance (NMR) or x-ray construct (see table S3) for clarity. Data are shown for 4 of 10 titration points (see table S4). Each spectrum represents at least 16 accumulated experiments that were acquired with 1024 and 128 points in the 1H and 15N dimensions, respectively. dPar3, Drosophila Par3; dPar6, Drosophila Par6; ppm, parts per million. (D) Sequence conservation of the Par6 PBM shown as position weight matrix (bottom) (65) based on the sequences shown in fig. S3.

PDZ domain interactions are a recurring theme in polarity and cell-cell adhesion protein networks. PDZ domains fold into a six-stranded antiparallel β-sheet capped by two α-helices (17) and predominantly bind with low affinity [typically in the micromolar range (18, 19)] to short (4 to 15 amino acids), disordered sequences at the C termini of target proteins, so-called PDZ-binding motifs (PBMs). This interaction results in an augmentation of the PDZ β-sheet by an antiparallel β-strand formed by the PBM. The four C-terminal residues constitute the specificity-determining core PBM (20) that can be categorized into three classes depending on the amino acid at the −2 position relative to the C terminus: X-[S/T]-X-φ-COO (class I), X-φ-X-φ-COO (class II), and X-[D/E]-X-φ-COO (class III), with X being any amino acid and φ being a hydrophobic residue (21). In addition, PDZ domains can recognize internal peptide sequences (22) or homo- or heterodimerize in some cases (1416).

Here, we uncovered the molecular basis of the Par3:Par6 interaction. We identified a previously unrecognized, highly conserved PBM in Par6 that was important for Par6 localization in vivo and essential for Par3:Par6 interaction in vitro. We found that both the Par3 PDZ1 and PDZ3 domains, but not the PDZ2 domain, associated through canonical PDZ:PBM interactions with the Par6 PBM and showed that one Par3 protein could recruit two Par6 molecules simultaneously. In cells, we found that the Par6 PBM was critical for Par3-mediated membrane recruitment of Par6 in cultured Drosophila Schneider 2R (S2R) cells and played a role in Par6 localization to the cell cortex in fly embryonic epidermal cells. In summary, our results provide important structural and functional details on the role of Par3 and Par6 in organizing polarity complexes.

RESULTS

Par6 contains a PBM that associates with the PDZ1 domain in Par3

To elucidate the molecular details of the Par3:Par6 interaction, we performed nuclear magnetic resonance (NMR) binding studies. This method is a powerful tool to study biomolecular interactions (23, 24) because the measured chemical shifts of atomic nuclei (that is, the peak positions in NMR spectra) are highly sensitive to changes in the local chemical environment. In a standard experiment, an unlabeled binding partner is titrated against an isotope-labeled, NMR-visible protein, resulting in concentration-dependent changes of the chemical shifts for the residues contacting the ligand [chemical shift perturbation (CSP)] that can be used to obtain dissociation constants (Kd’s) (fig. S1, A to C). Together with chemical shift assignments, CSP studies enable the mapping of the ligand-binding surface on a protein structure with close to atomic resolution.

The Par3 PDZ1 heterodimerizes with the Par6 PDZ domain (1416). However, this interaction has been debated (7, 10). To investigate the mechanism of the Par3:Par6 interaction, we first recorded 1H,15N-correlation NMR spectra of the 15N-labeled PDZ1 domain of human Par3 (hPar3). Surprisingly, we observed that this PDZ domain was largely unfolded in isolation (fig. S2A). The domain architectures of the Par complex proteins are highly conserved in metazoans, suggesting that Par3 and Par6 interact in a similar manner (2). We therefore investigated next whether the Drosophila melanogaster or Caenorhabditis elegans Par3 PDZ1 domains interact with the respective Par6 PDZ or Crib-PDZ domains. Although both the Drosophila and C. elegans Par3 PDZ1 domains were folded in isolation, we observed only few changes in the NMR spectra of the 15N-labeled Par3 PDZ1 domains upon addition of the respective unlabeled Par6 PDZ (fig. S2, B and C) or upon addition of the unlabeled Par3 PDZ1 domains to the respective 15N-labeled Par6 Crib-PDZ domains (fig. S2, D and E). Similarly, neither the Drosophila Par3 (dPar3) PDZ2 nor the PDZ3 domain induced substantial CSPs in the 15N-labeled Drosophila Par6 (dPar6) Crib-PDZ domain in our NMR experiments (fig. S2, F and G). We thus conclude that the Par3 PDZ domains do not associate directly with the Par6 PDZ domain.

In searching for an alternative binding mechanism, we revisited the Par6 protein sequence and identified a previously unrecognized class II (φ-X-φ-COO) PBM at its C terminus (Fig. 1B). This motif is highly conserved in metazoans (Fig. 1B) except in nematodes (fig. S3), and we therefore focused our further efforts on the dPar3 and dPar6 proteins. To assess whether the identified Par6 PBM could interact with the Par3 PDZ domains, we performed NMR CSP experiments. Upon addition of an unlabeled peptide containing the eight C-terminal residues of dPar6, we observed large peak shifts (more than one peak width) and line broadening for numerous residues in the 15N-labeled dPar3 PDZ1 domain (Fig. 1C), effects that would be expected for two proteins that specifically interact with each other. Thus, the dPar3 PDZ1 domain can directly interact with the dPar6 PBM in vitro. In support of this notion, epithelial cell polarity also critically depends on interactions of the Par3 PDZ domains with cell adhesion proteins through PBMs that are similar to the Par6 PBM (Fig. 1, B and D) (2529).

The Par6 PBM is important for interaction with Par3 in vitro and in cultured cells

To investigate the importance of the Par6 PBM for binding to Par3, we performed in vitro glutathione S-transferase (GST) pulldown experiments using a recombinant GST-tagged dPar3 fragment containing all three PDZ domains and Sumo-tagged dPar6 variants (Fig. 2A). The Par3 PDZ1-3 domains efficiently pulled down wild-type Par6 (Fig. 2B, lane 10). By contrast, deletion of the PBM (ΔPBM; Fig. 2B, lane 12), the region C-terminal of the PDZ domain (PB1-CribPDZ; Fig. 2B, lane 14), or the region C-terminal of the Crib motif (PB1-Crib; Fig. 2B, lane 16) essentially abolished Par6 binding to the Par3 PDZ1-3 domains. GST alone did not pull down any of the Par6 constructs in control experiments (Fig. 2B, lanes 9, 11, 13, and 15). These pulldown experiments thus confirm our NMR experiments and show a direct interaction of the Par6 PBM with the Par3 PDZ domains in vitro.

Fig. 2 The Par6 PBM is essential for Par3 interaction and Par3-mediated cortical targeting.

(A) Schematic representation of the dPar3 and dPar6 constructs used for glutathione S-transferase (GST) pulldown experiments and Par6 recruitment assays. (B) GST pulldown experiments using GST or GST-tagged dPar3 PDZ1-3 module incubated with wild-type (WT) or truncated Sumo-tagged dPar6 as indicated. Input and associated Par6 were detected along with GST and GST-dPar3 PDZ1-3 by Coomassie staining. Par6 proteins in the pulldowns are highlighted with asterisks. The pulldown is representative of at least three independent experiments. MW, molecular weight; Crib, Cdc42/Rac-interactive binding motif. (C to J) Representative fluorescence images of Schneider 2R cells transiently transfected with GFP-Par6 variants in the absence (C to F) or presence of RFP-Par3 (G to J). Images are representative of three independent sets of cells. At least 30 cells per set were evaluated. Scale bars, 5 μm. GFP, green fluorescent protein; RFP, red fluorescent protein; DAPI, 4′,6-diamidino-2-phenylindole.

To explore whether the Par6 PBM is important for Par3 interaction in cells, we transiently transfected Drosophila S2R cells with wild-type green fluorescent protein (GFP)–tagged Par6 or deletion constructs (Fig. 2A) in the absence (Fig. 2, C to F) or presence of red fluorescent protein (RFP)–tagged Par3 (Fig. 2, G to J). All Par6 variants were cytosolic in the absence of Par3 (Fig. 2, C to F). However, in the presence of Par3, wild-type Par6 showed a strong colocalization with Par3 at the cell cortex (Fig. 2G). A Par6 mutant lacking the PDZ domain (ΔPDZ) was still recruited to the plasma membrane in the presence of Par3 (Fig. 2H) with only a small fraction remaining cytosolic. By contrast, Par3-mediated membrane targeting was reduced for the Par6 mutant that lacked the PBM (ΔPBM; Fig. 2I) or the mutant lacking both the PDZ domain and the PBM (ΔPDZΔPBM; Fig. 2J). To assess whether Par3-mediated membrane targeting of Par6 depended on endogenous aPKC, we performed Par6 recruitment assays in S2R cells in which aPKC was knocked down by RNA interference (RNAi; fig. S4A). We observed that wild-type Par6 was still efficiently recruited to the cell cortex in the presence of Par3 (fig. S4B). By contrast, membrane targeting was compromised for all Par6 mutants (fig. S4, C to E). aPKC knockdown thus only affects the localization of the Par6 ΔPDZ mutant (fig. S4D compared to Fig. 2H) but not that of the ΔPBM and ΔPDZΔPBM mutants. These findings suggest that the PDZ domain may have an additional function in Par6 localization when aPKC is present. In line with our NMR binding studies and in vitro assays, these cellular results show that the PBM of Par6 plays an important role in the interaction with Par3 and in Par3-mediated localization of Par6 to the cell cortex.

The PBM is functionally redundant with the PDZ domain in Par6 localization in vivo

To explore the in vivo relevance of our findings, we generated transgenic flies expressing GFP-tagged Par6 under a constitutive promoter (30) and analyzed the embryos at stage 11 (Fig. 3, A to D). As expected, we observed that wild-type GFP-Par6 accumulated predominately at the cell-cell contacts of epidermal cells and colocalized with Par3 and aPKC (Fig. 3, A and E). Whereas deletion of the Par6 PBM had no significant effect on Par6 localization (Fig. 3, B and E), deletion of the PDZ domain resulted in a mild but consistent mislocalization of Par6 (Fig. 3, C and E). By contrast, deletion of both the PDZ domain and the PBM caused a strong mislocalization of this Par6 mutant to the cytosol (Fig. 3, D and E). All GFP-Par6 variants were expressed at similar amounts in fly embryos (fig. S5A). Thus, both the PDZ domain and the PBM contribute to the correct localization of Par6 in vivo, with deletion of both domains resulting in Par6 mislocalization. However, this effect may reflect an indirect association between Par3 and Par6 through aPKC as suggested previously (7, 10) and/or Par6 recruitment by other epithelial cell polarity regulators such as Crumbs or Stardust that bind to the Par6 PDZ domain (22, 3135).

Fig. 3 Recruitment of Par6 to the cell cortex in fly epithelia depends on both the PBM and the PDZ domain.

(A to D) Localization of aPKC, Par3, and GFP-Par6 variants in the epithelia of stage 11 fly embryos. Scale bars, 5 μm. Images are representative of 20 embryos for each group. (E) Pearson’s correlation coefficients (PCCs) of Par6, aPKC, and Par3 fluorescence intensities were determined using the Costes’ approach (50). The average PCCs estimated for four embryos of each Par6 variant are plotted with their SDs. Statistical significance was determined by a two-tailed t test and is indicated by horizontal brackets. *P = 0.0112 (ΔPDZ compared to WT for Par6 colocalization with Par3), **P = 0.0079 (ΔPDZ compared to WT for Par6 colocalization with aPKC), ***P = 0.0015 (ΔPDZΔPBM compared to WT for Par6 colocalization with Par3), and ***P = 1.8382 × 10−4 (ΔPDZΔPBM compared to WT for Par6 colocalization with aPKC).

Of these other polarity regulators, Stardust is expressed at earlier stages and thus is already present at stage 6 in fly embryos, whereas Crumbs is only expressed at later stages (36, 37). We found that localization of the various Par6 constructs in stage 6-7 fly embryos (when gastrulation occurs) was similar to their localization in stage 11 embryos (fig. S5, B and E). Notably, the ΔPDZΔPBM mutant was mislocalized in both stage 6-7 and stage 11 embryos. These findings demonstrate that Par6 recruitment to the cell cortex in vivo critically depends on both the PDZ domain and the PBM and suggest that these domains may be functionally redundant for Par6 localization in vivo.

In par6-null embryos (par6Δ226), which normally die at late embryonic stages (38), we found that expression of wild-type GFP-Par6 or GFP-Par6 ΔPBM restored viability and aPKC and Par3 localization (fig. S6, A and B), whereas expression of GFP-Par6 ΔPDZ or GFP-Par6 ΔPDZΔPBM could not rescue the lethality of the par6-null allele. The Par6 mutant lacking the PBM had a mild localization defect but did not affect the localization of Par3 or aPKC in stage 11 embryos (fig. S6, B and C). This finding confirms that the PBM contributes to the correct localization of Par6 in vivo under physiological conditions and suggests that it is redundant with the PDZ domain in the establishment of epithelial polarity.

The dPar3 PDZ1 domain forms a canonical PDZ:PBM complex with Par6

To gain structural insight into the Par3:Par6 complex, we crystallized a dPar3:dPar6 fusion construct comprising the Par3 PDZ1 domain and the C-terminal PBM octapeptide of Par6 (VKDGVLHL; the core PBM is underlined) and solved the x-ray structure of the PDZ1:PBM complex (table S1). Both the Par3 PDZ1 domain and the Par6 core PBM were well defined, with the PDZ1 domain adopting the typical PDZ fold and the core PBM forming the canonical antiparallel β-strand with the PDZ1 β2-strand (Fig. 4A and fig. S7A). The C-terminal carboxylate group was involved in an extensive hydrogen bond network with the backbone of the PDZ1 β1-2 loop (Fig. 4A). The hydrophobic residues at the 0 and −2 positions of the PBM were deeply buried in a hydrophobic pocket formed by the carboxylate-binding loop (Leu19); the β2 (Leu21, Ala23, Leu24, Pro25), β3 (Leu33), and β6 (Leu84, Val86) strands; and the α2 helix (Val71, Leu75, Leu79) of the PDZ1 domain (Fig. 4A and fig. S7B). Overall, our structure was highly similar to other PDZ:PBM complexes such as the PDZ3 domain of INADL (inactivation-no-after-potential D like protein) in complex with a phage display-derived class II PBM peptide (Fig. 4B) (39) and Par3 PDZ3:PBM complexes (fig. S7, C and D) (40, 41). Outside the core PBM, we observed less well-defined electron density for the Par6 peptide and three residues of the Gly-Ser–linker (fig. S8, A and B), indicating an additional antiparallel β-strand formed by the last amino acids of the Gly-Ser–linker and the −7 and −6 PBM residues. However, the consistently high B factors (fig. S8, C and D, and table S1) suggest that the residues outside the core PBM are not stably structured.

Fig. 4 The dPar3 PDZ1 and PDZ3 domains interact with dPar6 through canonical PDZ:PBM interactions.

(A) Interaction network of the Par6 core PBM (dark pink) with the Par3 PDZ1 domain (gray). Residues involved in hydrogen bonds (dashed lines) and side-chain interactions are shown in stick representation with carbon, nitrogen, and oxygen atoms colored in gray (for PDZ1) or dark pink (for PBM), blue, and red, respectively. The structure statistics for the dPar3 PDZ1:dPar6 PBM complex are detailed in table S1. (B) The Par3 PDZ1:Par6 PBM complex is highly similar to the INADL PDZ3 in complex with a phage-derived class II PBM peptide [Protein Data Bank code: 4Q2N (39)] shown in green (backbone root mean squared displacement = 1.64 Å). (C) Overlay of a representative region of the 1H,15N-HSQC spectra of the Par3 PDZ3 Δβ2-3loop domain in the absence (black) or presence of increasing stoichiometric amounts of dPar6 PBM as indicated. Data are shown for 4 of 13 titration points (see table S4). Each spectrum represents at least 16 accumulated experiments that were acquired with 1024 and 128 points in the 1H and 15N dimensions, respectively. (D) Binding surface of the dPar6 PBM mapped onto a PDZ3 homology model and colored with a linear gradient from white [chemical shift perturbation (CSP) ≤ 0.05 ppm; cut-off] to red (CSP = 0.35 ppm). Unassigned residues are shown in dark gray, whereas residues broadened beyond detection in the PDZ3 are in dark red. The Par6 core PBM is shown in yellow and was modeled by superposition of the dPar3 PDZ1:PBM and PDZ3 structures.

To validate key interactions of the Par6 PBM with the Par3 PDZ1 domain in our crystal structure, we substituted each of the three C-terminal positions in the dPar6 peptide with Ala [Par6 L349A (L−2A), H350A (H−1A), and L351A (L0A)] and tested the Par3 binding capabilities of the PBM mutants in NMR experiments (fig. S8E). Consistent with our crystal structure, both the L−2A and the L0A mutations led to an almost complete loss of Par6 binding, and the H−1A mutation lessened CSPs in the Par3 PDZ1 domain as compared to the wild-type PBM (fig. S8E compared to Fig. 1C). In summary, our results demonstrate that Par6 associates with Par3 through a canonical PDZ:PBM interaction that crucially depends on the −2 and 0 positions of the Par6 PBM.

The Par3 PDZ1 and PDZ3 domains both recognize the Par6 PBM

We noticed that the residues in the PDZ1 domain contacting the 0 and −2 PBM positions are well conserved in all three Par3 PDZ domains (fig. S9). We therefore also tested the binding specificities of the Par3 PDZ2 and PDZ3 domains toward the Par6 PBM. To this end, we performed NMR binding studies with the individual dPar3 PDZ2 and PDZ3 domains and the dPar6 PBM peptide. In contrast to the PDZ1 domain, we observed virtually no CSPs for the Par3 PDZ2 domain upon addition of the Par6 PBM (fig. S10A). The Par3 PDZ2 domain thus has binding specificities that are distinct from the Par3 PDZ1 domain.

The PDZ3 domain contains a long, disordered insertion in the β2-3 loop that is unique to dPar3 (fig. S9). Because this extension severely compromised spectral quality, we used a PDZ3 domain with a truncated β2-3 loop (PDZ3 Δβ2-3loop) for our NMR studies (fig. S9). As for the PDZ1 domain, addition of the Par6 PBM induced numerous, large CSPs in the 15N-labeled PDZ3 Δβ2-3loop (Fig. 4C). This finding demonstrates that the truncated PDZ3 domain is functional and interacts readily with the dPar6 PBM. To map the observed CSPs onto the surface of the PDZ3 domain, we obtained chemical shift assignments for the PDZ3 domain and generated a homology model of the Par3 PDZ3 domain. In this model, the PDZ3 domain binds the Par6 PBM in a canonical manner mainly involving the carboxylate-binding loop, the β2 strand, and the α2 helix (Fig. 4D).

To obtain quantitative insights into the Par3 PDZ interactions with the Par6 PBM, we determined Kd’s by NMR line shape fitting for the chemical shift titrations of the dPar3 PDZ1 and PDZ3 domains with the Par6 PBMs. Because the PDZ2 domain did not exhibit CSPs upon addition of the Par6 PBM, we did not fit these data. We found that the Par6 PBM bound to the PDZ1 domain with a moderate affinity of 216 ± 4 μM and to the PDZ3 domain with a higher affinity of 54 ± 1 μM for the PDZ3 domain [figs. S11 (A to C) and S12 (A to C) and table S2]. Next, we quantitatively addressed the importance of the three C-terminal positions in the Par6 PBM for PDZ1 binding (figs. S8E and S13). We found that mutation of the C-terminal position (L0A) weakened the affinity of the PDZ1 domain for the Par6 PBM by ~18-fold (table S2 and fig. S13, A and B), mutation of the −1 position (H−1A) weakened the affinity by ~4-fold (table S2 and fig. S13, C and D), and mutation of the −2 position (L−2A) weakened the affinity by ~11-fold, respectively (table S2 and fig. S13, E and F). These results confirm that these mutations in the core PBM compromise binding to the Par3 PDZ1 domain and that these residues are thus important determinants for the interaction with the Par3 PDZ1 domain.

Overall, Kd’s of up to a few hundreds of micromolars have been observed for physiologically relevant PDZ interactions (18, 19). Together with our analyses of cultured S2R cells (Fig. 2 and fig. S4) and fly embryos (Fig. 3 and figs. S5 and S6), our NMR analyses show that the interaction of the Par6 PBM plays a role in Par3 PDZ binding in vitro and is physiologically relevant in vivo.

Par3 can interact with two Par6 proteins simultaneously in vitro

In polarized cells, Par complexes form micrometer-sized clusters that cover the apical plasma membrane, but how Par complexes assemble into these sizeable protein networks is largely unknown. A prerequisite for forming higher-order networks is that the proteins involved are multivalent and contain multiple independent binding sites and engage in a multitude of weak (micromolar affinity) interactions. To test whether the first and third PDZ domains of Par3 fulfilled this requirement, we recorded 1H,15N-correlation spectra of the 15N-labeled dPar3 PDZ1-3 module that contains all three PDZ domains (PDZ1-3 Δβ2-3loop) and examined its Par6 binding capability. The NMR spectra of the Par3 PDZ1-3 module and the individual PDZ domains superimposed well (Fig. 5, A to C), and resonance assignments of the individual domains could be transferred to the PDZ1-3 module. This finding demonstrates that the individual PDZ domains in Par3 are structurally largely independent. Addition of unlabeled Par6 PBM to the PDZ1-3 construct resulted in CSPs comparable to the isolated PDZ domains [Fig. 5 (D to F) compared to Figs. 1C and 4C and fig. S10A]. This finding shows that the Par3 PDZ domains act as functionally independent entities within the PDZ1-3 module and that one Par3 protein can simultaneously interact with two Par6 proteins through its PDZ1 and PDZ3 domains in vitro (Fig. 5G). Par3 thus has the potential to engage in weak, multivalent interactions with Par6 and may thereby promote the assembly of large-scale clusters of Par complexes at the cell membrane in vivo.

Fig. 5 Par3 can interact with two Par6 proteins simultaneously.

(A to C) Overlay of 1H,15N-TROSY spectra of the dPar3 PDZ1-3 Δβ2-3loop module with the isolated PDZ1 (A), PDZ2 (B), and PDZ3 Δβ2-3loop domains (C). Each spectrum represents at least 16 accumulated experiments that were acquired with 1024 and 128 points in the 1H and 15N dimensions, respectively. (D to F) Overlay of 1H,15N-TROSY spectra of the dPar3 PDZ1-3 Δβ2-3loop module in the absence (black) or presence of dPar6 C-terminal peptide as indicated. Peaks not corresponding to the PDZ1 (D), PDZ2 (E), or PDZ3 Δβ2-3loop (F) domain within the PDZ1-3 module are opaque to highlight the changes of the individual domains within the entire module. The indicated resonance assignments correspond to the respective PDZ domain as shown in the top panels and correspond to the respective NMR or x-ray construct (table S2). Data are shown for four of six titration points. Each spectrum represents at least 48 accumulated experiments that were acquired with 1024 and 128 points in the 1H and 15N dimensions, respectively. (G) Schematic representation of the Par3:Par6 interactions as mapped by NMR spectroscopy highlighting the 1:2 stoichiometry of the complex in vitro.

The Par6 PBM can compete with the PBM of E-cadherin for Par3 binding

Cell adhesion proteins, such as cadherins and nectins, also interact with Par3 and contain conserved class II PBMs that are similar to the Par6 PBM (Fig. 1B and fig. S7C). The VE (vascular endothelial)–cadherin PBM interacts with the PDZ3 domain of murine Par3 with a Kd of ~6 μM (41) that is ~36- and ~9-fold tighter than the affinities of the dPar3 PDZ1 and PDZ3 domains for the Par6 PBM (table S2). To address whether the Par6 PBM could compete with such ligands for Par3 binding, we performed NMR CSP studies with the 15N-labeled dPar3 PDZ domains and added increasing amounts of unlabeled Shg (Drosophila E-cadherin) PBM to the individual domains. Similar to the Par6 PBM, both the Par3 PDZ1 (Fig. 6A) and PDZ3 (Fig. 6B) domains show large CSPs for numerous peaks in the 1H,15N-correlation spectra upon addition of the Shg (Drosophila E-cadherin) PBM. By contrast, the PDZ2 domain displayed few changes (fig. S10B). Line shape–fitting analyses for the Shg PBM chemical shift titrations of all three dPar3 PDZ domains yielded Kd values of 128 ± 4 μM for the PDZ1 domain, 954 ± 45 μM for the PDZ2 domain, and 0.6 ± 0.1 μM for the PDZ3 domain [table S2 and figs. S11 (C and D), fig. S14 (A and B), fig. S12 (C and D)]. The dPar3 PDZ domains thus have greater affinities for the PBM of Shg than for the Par6 PBM (albeit only by a factor of ~2 for the PDZ1 domain).

Fig. 6 The Par6 PBM can compete with Shg for Par3 binding.

(A) Overlay of 1H,15N-HSQC spectra of the dPar3 PDZ1 domain in the absence and presence of the Shotgun (Shg) PBM peptide as indicated. Each spectrum represents at least 20 accumulated experiments. (B) Overlay of 1H,15N-HSQC spectra of the dPar3 PDZ3 Δβ2-3loop domain in the absence and presence of the Shg PBM peptide as indicated. Data are shown for four of eight titration points for (A) and (B) (see table S4). Each spectrum represents at least 20 accumulated experiments. (C) Overlay of 1H,15N-HSQC spectra of the Shg PBM fused to immunoglobulin-binding domain B1 of streptococcal protein G (GB1) in the absence (black) or presence of dPar3 PDZ1 domain (purple) and upon stepwise addition of the Par6 PBM (red and orange). Arrows indicate the successive reversal of the chemical shifts from the Shg:PDZ complex toward the unbound Shg PBM. Each spectrum represents at least 40 accumulated experiments. (D) Overlay of 1H,15N-HSQC spectra of the Shg PBM fused to GB1 in the absence (black) or presence of the PDZ3 Δβ2-3loop domain. Data are shown for four of six (PDZ1) or seven (PDZ3) titration points. Each spectrum represents at least 40 accumulated experiments. (A to D) All experiments were acquired with 1024 and 128 points in the 1H and 15N dimensions, respectively.

To evaluate whether the PBM of Par6 and Shg compete for binding to the Par3 PDZ1 or PDZ3 domains, we performed a set of NMR experiments on the 15N-labeled Shg PBM. First, we recorded 1H,15N-correlation spectra of the peptide in the absence or presence of unlabeled dPar3 PDZ1 or PDZ3 domain. In both cases, this resulted in PDZ:Shg complex formation as indicated by the observed CSPs (Fig. 6, C and D). Subsequently, we added Par6 PBM to the PDZ:Shg complex in a stepwise manner and found that Shg peptide was released from both PDZ domains as indicated by chemical shift changes that reverted toward the unbound Shg peptide (Fig. 6, C and D). This finding demonstrates that the Par6 PBM can compete with Shg for Par3 PDZ binding. The relatively high stoichiometric amounts of Par6 PBM required to outcompete the Shg peptide from the PDZ3 domain reflect the large difference in binding affinities (table S2). Ultimately, the question of direct competition between the Par6 PBM and other PBM-containing ligands for Par3 binding in vivo would require determining the specific subcellular concentrations of Par3, Par6, and other binding partners; the binding affinities within the fully assembled Par complex; and the exact chronological order of binding events in cells. These are certainly highly interesting, although challenging topics for future studies.

The PDZ:PBM interaction is conserved in human Par3 and Par6 proteins

To explore whether the PDZ:PBM interactions that mediate Par3:Par6 association in Drosophila are conserved in the human proteins, we cotransfected human embryonic kidney (HEK) 293T cells with Flag-tagged hPar3 and Myc-tagged wild-type or truncated hPar6α. Full-length Par6α readily coprecipitated with Par3, which was abrogated by deletion of the Par6α PBM (ΔPBM) or deletion of both the Par6α PDZ domain and the PBM (ΔPDZΔPBM; Fig. 7A). By contrast, Par3 coimmunoprecipitated comparable amounts of Par6α ΔPDZ and wild-type Par6α. Consistent with our in vitro and in vivo data for the Drosophila Par3:Par6 interaction, the results of our coimmunoprecipitation experiments demonstrate that the Par6 PBM is also important for the interaction of the human proteins.

Fig. 7 The PDZ:PBM interaction is conserved in the human Par3:Par6 complex.

(A) Schematic representation of the human Par3 (hPar3) and hPar6α constructs used for coimmunoprecipitation (IP) experiments (top). Left: Cell lysates of human embryonic kidney (HEK) 293T cells cotransfected with human Flag-tagged Par3 and Myc-tagged Par6α variants were subjected to Flag-antibody IP followed by immunoblotting (IB) with Flag and Myc antibodies. Equal amounts of protein were confirmed by IB as shown (input). Western blots are representative of at least three independent experiments. Right: Signal intensities of immunoprecipitated Par6 normalized to the total cell lysates (input) in arbitrary units (a.u.). Data are means ± SD of three independent experiments. Statistical significance was determined by a two-tailed t test and is indicated by vertical brackets. *P = 0.0124 (Par6 ΔPBM compared to WT), **P = 0.0055 (Par6 ΔPDZΔPBM compared to WT). (B) Overlay of 1H,15N-correlation spectra of the hPar3 PDZ1 domain in isolation and fused to the hPar6α PBM. Each spectrum represents at least 80 accumulated experiments. (C and D) Overlay of 1H,15N-correlation spectra of the hPar3 PDZ2 (C) and the PDZ3 (D) domains in the absence and presence of hPar6α PBM peptide as indicated. Data are shown for four of five titration points. Each spectrum represents at least 24 accumulated experiments. (B to D) All experiments were acquired with 1024 and 128 points in the 1H and 15N dimensions, respectively.

To examine whether the mode of interaction and the specificities of the individual PDZ domains toward the Par6 PBM are conserved among human and dPar3, we performed NMR binding studies. Because the hPar3 PDZ1 domain was unfolded in isolation (fig. S2A), we fused the eight C-terminal residues of hPar6α to the C terminus of the hPar3 PDZ1 domain separated by a 15–amino acid Gly-Ser–linker. The PDZ1:PBM fusion had a well-dispersed 1H,15N-correlation spectrum demonstrating that the Par6α PBM induced the folding of the PDZ1 domain and hence interacted with the Par3 PDZ1 domain (Fig. 7B). Addition of unlabeled Par6α PBM peptide to the 15N-labeled hPar3 PDZ2 domain resulted in a few, although partially substantial, CSPs (Fig. 7C and fig. S10C) indicating that the hPar3 PDZ2 domain binds to the Par6α PBM. By contrast, the PDZ3 domain displayed numerous, large CSPs in the presence of Par6α (Fig. 7D). Mapping of the observed CSPs on the structures of the hPar3 PDZ2 and PDZ3 domains showed that both domains could engage the Par6α PBM through canonical PDZ:PBM interactions (fig. S10, C and D). Together, these results demonstrate that the Par3 PDZ:Par6 PBM interactions and the functions of the Par3 PDZ domains are largely conserved in the human and Drosophila proteins.

DISCUSSION

Here, we have identified a previously unrecognized PBM in Par6 that mediates canonical PDZ:PBM interactions with the PDZ1 and PDZ3 domains of Par3. This interaction mode is conserved among the human and Drosophila Par3 and Par6 proteins. We demonstrate that the PBM, but not the PDZ domain, of Par6 is essential for interaction with Par3 in vitro. Yet, the PBM seems to be redundant with the PDZ domain in Par6 localization in fly epithelia because the individual deletions only mildly reduce cortical localization in vivo, and deletion of both the PBM and the PDZ domain is required for almost complete mislocalization of Par6.

Overall, we did not find evidence of heterodimerization of the human, Drosophila, or C. elegans Par3 PDZ domains with the respective Par6 PDZ domain by NMR (fig. S2), in GST pulldown (Fig. 2B) or coimmunoprecipitation experiments (Fig. 7A), or in recruitment assays in S2R cells (Fig. 2, D and H). Our results thus provide important insights into Par complex assembly and contrast with previous reports that suggest that Par3 and Par6 associate through PDZ:PDZ interaction (14, 16, 29). In support of our findings, Par3 associated in epithelial cells with various cell adhesion proteins having class II PBMs that are similar to the Par6 PBM (Fig. 1B), and we showed that the dPar3 PDZ1 and PDZ3 domains also interacted with the major cell adhesion protein E-cadherin (Shg; Fig. 6, A and B). These interactions are crucial for the establishment and maintenance of cell junctions and epithelial cell polarity (3, 2529). Notably, as in the case of Par6, these adhesion protein interactions are conserved among human and Drosophila Par3. In summary, this supports the notion that class II PBMs constitute a general binding motif for the Par3 PDZ1 and PDZ3 domains to recruit polarity and cell adhesion proteins for the establishment and maintenance of cell polarity. Nematodes may constitute an exception because not only do their Par6 but also their cadherin (Hammerhead-1; HMR-1) proteins lack a detectable, C-terminal PBM (fig. S3). Future studies will thus be required to elucidate the mechanism of Par3 PDZ interactions and their role in cell polarity in nematodes.

Tandem arrangement of PDZ domains allows scaffolding proteins to assemble different components of a signaling cascade through multivalent interactions (42) based on the independent folding and function of interaction domains embedded in proteins (43). We found that the isolated Par3 PDZ1 and PDZ3 domains could independently interact with the Par6 PBM [Figs. 5 (D and F) and 7 (B and D)] and thus may be redundant. These functional redundancies allow each Par3 molecule to simultaneously recruit two Par6 proteins in vitro (Fig. 5G). Together with Par6:aPKC heterodimerization and Par3 homo-oligomerization, this provides a basis for the role of Par3 in enforcing the spatial segregation and the assembly of large-scale, self-organizing Par complex networks at the cell cortex in vivo.

The mode of Par3:Par6 interaction has remained controversial, which reflects the challenges of delineating the specific functions of proteins and their individual domains and motifs in cell polarity. Functional coupling, redundant interactions, differences in organism strains, cell types, and protein constructs and finally the existence of paralogs hamper in vivo analyses of polarity proteins (10, 44). Moreover, the composition of polarity complexes is dynamic and may depend on the cell type or developmental context (45). Last, different populations of polarity proteins could coexist within a single cell (46). These issues are obstacles for functional analyses because phenotypes may be obscured in mutational studies. Detailed structural analyses are thus essential to unambiguously determine the molecular basis of polarity complex formation. Moreover, the requirement for a free C terminus for PBM function suggests that C-terminal tagging for fluorescence microscopy or immunoblotting likely abrogates not only Par3:Par6 association but possibly also other Par6 PBM interactions that have yet to be identified.

The presence of at least one PDZ domain or a PBM in almost all polarity and cell adhesion proteins emphasizes the importance of PDZ:PBM interactions in the assembly of cell polarity and adhesion signaling networks. Our work illustrates the difficulties in predicting PDZ specificities, because the marked differences in Par6 and Shg recognition of the Par3 PDZ2 [Figs. 5 (D to F) and 6 (A and B), fig. S10B, and table S2] were unexpected on the sequence level (fig. S9). A detailed characterization of the functional specificities and redundancies of PDZ domains, as presented here, contributes to deciphering their contribution to polarity protein localization and function. Ultimately, elucidating the binding specificities of individual domains will help to predict PDZ domain function and provide a better understanding of the interaction networks underlying the establishment, maintenance, and loss of cell polarity and hence critical developmental and carcinogenic processes.

MATERIALS AND METHODS

Reagents and constructs

Par3 and Par6 constructs were cloned from C. elegans cDNA, S2R cell cDNA, or a synthetic gene fragment (Life Technologies) containing the Par3 PDZ domains (D. melanogaster). pK-myc-Par3b (hPar3) and pK-myc-Par6C (hPar6α) vectors were purchased from Addgene (plasmid #19388 and #15474). For NMR studies, gene fragments amplified by polymerase chain reaction were cloned into the following vectors: pETZ2.1a (His6-Z domain-TEV) for C. elegans. Par3 and Par6 constructs; pET-M41 (His6-MBP-TEV) for dPar3 PDZ domains, dPar3 PDZ1:dPar6 PBM fusion, dPar6 PDZ, and Crib-PDZ constructs; pET-M30 (His6-GST-TEV) for the dPar3 PDZ1-3 Δβ2-3loop, hPar3 PDZ domains, and the hPar3:Par6α PBM fusion constructs; and pRTDuet-GB1 (His6-GB1-TEV) for all Par6 and Shg PBM peptides. For GST pulldown experiments, a gene fragment containing the three dPar3 PDZ domains was cloned into a pETM30-HA (His6-GST-TEV-HA) vector, whereas full-length and truncated dPar6 constructs were cloned into a pET-M11-Sumo (His6-Sumo-TEV) vector.

For S2R cell transfection, dPar3 and dPar6 were cloned into pENTR vectors. Site-directed mutagenesis was carried out with the Par6 pENTR vector as the DNA template using the following primers: Par6 ΔPDZ, 5′-GTGCCGGAAACGCATGGTGGAGGTGGAGGTCCGGCCAATCAGCGC-3′ and Par6 ΔPBM, 5′-ACGATAATGGCCAGCGATTAAATCGATGGAGTGCTGCATTTG-3′.

Par3 and wild-type and mutant Par6 variants were subsequently subcloned into pUGW and pURW vectors from the Drosophila Genomics Resource Center (30), respectively. For coimmunoprecipitation studies, the Myc-tag in the pK-myc-Par3b vector was replaced by a Flag-tag using QuikChange mutagenesis. All deletions in the hPar6α gene were generated by QuikChange mutagenesis using the pK-myc-Par6C vector as the DNA template. For a complete list of constructs, see table S3.

Expression and purification of recombinant proteins

Recombinant proteins were expressed in Escherichia coli BL21-CodonPlus (DE3)-RIL cells (Stratagene) and purified by Ni affinity and size-exclusion chromatography. To facilitate peptide production, we fused the Par6 and Shg PBMs at their N termini to the immunoglobulin-binding domain B1 of streptococcal protein G (GB1) domain followed by a TEV protease cleavage site. For NMR binding experiments, nonspecific binding of GB1 alone was tested. For NMR studies, unlabeled His6-GB1-Par6 and His6-GB1-Shg peptides or dPar3 PDZ domains were expressed in LB medium, and 15N- or 13C,15N-labeled His6-GB1-Shg, Par3 PDZ or Par6 Crib-PDZ, and PDZ domains were expressed in M9 minimal medium with 15NH4Cl or 15NH4Cl and 13C-glucose as sole sources of nitrogen and carbon. All NMR constructs were buffer-exchanged into NMR buffer [20 mM Na phosphate (pH 6.5 or 7.5), 150 mM NaCl, 1 mM dithiothreitol (DTT), and 0.03% NaN3] for triple-resonance and CSP experiments.

Proteins for pulldown experiments were expressed as His6-GST-HA (human influenza hemagglutinin)–tagged dPar3 and His6-Sumo–tagged dPar6 proteins in E. coli BL21-CodonPlus (DE3)-RIL cells (Stratagene) in LB medium and purified by Ni affinity chromatography. Subsequently, buffer was exchanged by dialysis to 50 mM Na phosphate (pH 7.5), 150 mM NaCl, 1 mM DTT, and 10% glycerol for Par6 proteins or to 20 mM Na phosphate (pH 6.5), 150 mM NaCl, 1 mM DTT, and 10% glycerol for Par3 proteins.

Pulldown assays

dPar3 proteins containing an N-terminal His6-GST-HA-tag were incubated with glutathione beads (Macherey-Nagel) at 1 μM concentration together with 65 μM dPar6 proteins that contained an N-terminal His6-Sumo-tag for 1 hour at 4°C in pulldown buffer [50 mM Na phosphate (pH 7.5), 150 mM NaCl, 10% glycerol, and 2 mM DTT]. The beads were washed four times with pulldown buffer and specifically bound proteins eluted with pulldown buffer supplemented with 25 mM reduced glutathione. Eluted proteins were precipitated with 10% (w/v) trichloroacetic acid (Sigma) on ice followed by centrifugation. Protein pellets were resuspended in SDS loading buffer, resolved by SDS–polyacrylamide gel electrophoresis (PAGE), and detected by Coomassie staining.

S2R cell culture

Immunostainings of transfected S2R cells were carried out as previously described (12). Briefly, S2R cells were transfected using FUGENE (Promega) with Ubi::GFP-Par6 variants alone or with Ubi::RFP-Baz. Cells were fixed with paraformaldehyde and imaged with a Zeiss LSM 710 Meta confocal microscope. RNAi-mediated knockdown of aPKC was performed as described (47). Briefly, a ~300–base pair double-stranded RNA (dsRNA) fragment was taken up by the cells and processed to several small interfering RNAs (siRNAs). One dsRNA thus results in several siRNAs and thereby reduces the probability of off-targets. The efficiency of dsRNA-mediated gene knockdown of aPKC was confirmed by a Western blot of S2R cell lysates using aPKC (1:1000; Santa Cruz Biotechnology, sc-216) and mouse actin (1:1000; Santa Cruz Biotechnology, sc-47778) antibodies. The following oligonucleotides were used to amplify a fragment of aPKC mRNA for in vitro transcription into dsRNA: aPKC-dsRNA-F, 5′-ACTTCGCGTTCTCCGC-3′ and aPKC-dsRNA-R, 5′-TTGCTAGCTGGGTAAAATATTTTGA-3′.

HEK293T cell culture and coimmunoprecipitation

HEK293T cells were transfected with N-terminally Flag-tagged hPar3 in a pKFlag and N-terminally Myc-tagged hPar6α in a pKMyc expression vector using Lipofectamine (Invitrogen); grown in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum (Invitrogen), 2 mM l-glutamine (Invitrogen), penicillin (Invitrogen), and streptomycin (Invitrogen); and harvested 3 days after transfection. Cells were lysed in NET buffer [50 mM tris-HCl (pH 8.0), 150 mM NaCl, 0.1% (w/v) Triton X-100, and 1 mM EDTA] supplemented with proteinase inhibitor cocktail (Roche). The supernatants were incubated with Flag M2 antibody (2 μg/mg of cell lysate) and subsequently with GammaBind Plus Sepharose (GE Healthcare) for 1 hour each at 4°C. The beads were washed three times with NET buffer and once with Triton-free NET buffer. Immunoprecipitated proteins were eluted with SDS loading buffer, separated by SDS-PAGE, and immunoblotted using mouse Flag M2 (1:1000; Sigma) as primary and horseradish peroxidase (HRP)–coupled mouse immunoglobulin G (1:10,000) as secondary antibody (Thermo Fisher Scientific) or mouse Myc-HRP (1:5000; Life Technologies) antibody. Western blots were imaged by chemiluminescence. For quantitative analysis, signal intensities were determined for three independent experiments with ImageJ (Fiji) (48), and the immunoprecipitated proteins normalized to the total lysate (input). Two-tailed t tests were performed on the amalgamated data using the ttest2 routine in MATLAB v2017b (MathWorks) to estimate P values.

Fly stocks and genetics

Fly stocks were cultured on standard cornmeal agar food and maintained at 25°C. Transgenic flies of Ubi::GFP-Par6 variants were established using the PhiC31 integrase system with attP40 in a wild-type background (49). Fly embryo lysates were applied to SDS-PAGE and immunoblotted using rabbit GFP (1:500; Santa Cruz Biotechnology, sc-8334) and mouse actin (1:1000; Santa Cruz Biotechnology, sc-47778) antibodies to ensure uniform expression of the Par6 mutants. Western blots were imaged by chemiluminescence. To evaluate the function of GFP-Par6 and GFP-Par6 ΔPDZ, ΔPBM, and ΔPDZΔPBM mutants in a Par6-null background, the transgenes were backcrossed into a homozygous par6Δ226 mutant background (38). However, only the wild-type and ΔPBM transgenes can rescue the embryonic lethality of par6Δ226 mutant flies and be kept as stable stocks.

Immunohistochemistry

Drosophila stage 6-7 or stage 11 embryos were fixed in 4% formaldehyde phosphate buffer (pH 7.4), as previously described (30). The primary antibodies used for indirect immunofluorescence were as follows: mouse aPKC (PKCζ; 1:500; Santa Cruz Biotechnology, sc-17781), rabbit GFP (1:500; Santa Cruz Biotechnology, sc-8334), and guinea pig Baz (1:500). Secondary antibodies conjugated with Alexa 488, Alexa 568, and Alexa 647 (Life Technologies) were used at 1:400. Images of the epidermis were taken from stage 11 embryos on a Zeiss LSM 710 Meta confocal microscope or at stage 6-7 on a Leica SP8 confocal microscope. For colocalization analysis, we analyzed the images of four embryos for each Par6 variant (Fig. 3E and fig. S6C). Pearson’s correlation coefficients and SD were estimated after automatic thresholding using the Costes’ approach as implemented in the JACoP v2.0 plugin of ImageJ (Fiji) (48, 50, 51). Two-tailed t tests were performed using the ttest2 routine in MATLAB v2017b (MathWorks) to estimate P values.

NMR spectroscopy

CSP studies were performed with 75- to 100-μM samples of 15N-labeled protein in NMR buffer {20 mM Na phosphate [pH 6.5 (individual PDZ domains or Shg PBM) or 7.5], 150 mM NaCl, 1 mM DTT, and 0.03% NaN3} by recording 1H,15N–heteronuclear single-quantum coherence (HSQC) experiments on a 600-MHz Bruker Avance III spectrometer at 20°C for the Drosophila and C. elegans Par3 and Par6 PDZ domains and at 30°C for the human proteins. For CSP studies with the dPar3 PDZ1-3 Δβ2-3loop module, we recorded 1H,15N–transverse relaxation optimized spectroscopy (TROSY) experiments at 600 MHz and 30°C in 20 mM Na phosphate (pH 7.5), 150 mM NaCl, and 1 mM DTT. Backbone resonance assignment for the dPar3 PDZ domains was performed at 20° and 25°C for the hPar3 PDZ2 and PDZ3 domains by recording three-dimensional (3D) HNCACB, HNCOCACB, CCONH, and HNH-NOESY spectra at 600 or 800 MHz. NMR data were processed using the nmrPipe/nmrDraw software suite (52) and analyzed using XEASY (53) and Sparky (54). Figures displaying NMR spectra were produced with nmrView (www.onemoonscientific.com). Average CSPs used for binding site mapping were calculated in parts per million as [(Δδ1H)2 + (0.25 × Δδ15N)2]0.5, where Δδ is the difference in chemical shift at a ninefold stoichiometric excess of Par6 PBM to the respective reference in the absence of ligand. For binding site mapping, a homology model of the dPar3 PDZ3 domain was generated using HHPred (55), and Protein Data Bank (PDB) entries 2KOM and 2KOH were used for binding site mapping for the hPar3 PDZ2 (56) and PDZ3 (41) domains, respectively. A similar structure exists for the PDZ2 domain of rat Par3 (57).

2D line shape–fitting analysis

1H,15N-CSP studies for the dPar3 PDZ domains were quantified using TITAN (58) according to instructions and online documentation (www.nmr-titan.com and https://bitbucket.org/cwaudby/titan/wiki/Home). Spectra were acquired with 1024 and 128 points in the 1H and 15N dimensions, respectively, and processed with exponential window functions with a line broadening of 4 and 8 Hz. Spectra were zero-filled to 4096 and 1024 points in the 1H and 15N dimensions, respectively. To obtain comparable results for different ligands, we used the same cross peaks for the analysis of each PDZ domain (table S4). Errors were estimated with bootstrapping statistics on 100 replica. Figures for line shape analyses were prepared with TITAN.

X-ray crystallography

The dPar3 PDZ1:dPar6 fusion construct was crystallized in 0.1 M bis-tris (pH 6.5) and 2 M (NH4)2SO4. Diffraction data were collected at 100 K using a wavelength of 1 Å and a PILATUS 6M-F detector at the beamline PXII of the Swiss Light Source (Paul Scherrer Institute). Data were processed using XDS (59), and molecular replacement was performed using Phaser (60). The structure was finalized by iterative manual modeling with Coot (61) and refinement with Phenix (62). Each asymmetric unit contains two protein chains with virtually identical conformations (backbone root mean squared displacement of 0.031 Å for the PDZ1 domain and the core PBM). All figures displaying protein structures were generated using PyMOL (www.pymol.org).

Sequence alignment

Multiple sequence alignments were performed with MUSCLE (63) and displayed with ClustalX (64). The position-weighted matrix for the Par6 PBM was created with WebLogo 3.4 (65) using all Par6 PBMs displayed in fig. S3. The overall height of the stack indicates the sequence conservation at that position, whereas the height of a symbol within the stack indicates the relative frequency of each amino acid at that position.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/11/517/eaam9899/DC1

Fig. S1. CSP experiments.

Fig. S2. The Par3 PDZ1 domain does not interact directly with the Par6 PDZ domain.

Fig. S3. The Par6 C terminus contains a conserved class II PDZ-binding motif.

Fig. S4. Par6 localization in S2R cells upon aPKC knockdown.

Fig. S5. Expression of Par6 variants in fly embryos and localization of Par6 variants in transgenic flies during gastrulation.

Fig. S6. Par6 localization in Par6-null fly embryos.

Fig. S7. Crystal structure of the Par3 PDZ1:Par6 PBM complex.

Fig. S8. Ligand electron density and importance of the 0 and −2 positions of the Par6 PBM for PDZ domain interaction.

Fig. S9. Structure-based sequence alignment of the Par3 PDZ domains.

Fig. S10. Interaction of the dPar6 and Shg PBMs with the dPar3 PDZ2 domain and Par6:Par3 interactions for the human proteins.

Fig. S11. Line shape–fitting analyses of the dPar3 PDZ1 interactions with the PBMs of dPar6 and Shg.

Fig. S12. Line shape–fitting analyses of the dPar3 PDZ3 Δβ2-3loop interactions with the PBMs of dPar6 and Shg.

Fig. S13. Line shape–fitting analyses of the dPar3 PDZ1 interactions with mutant dPar6 PBMs.

Fig. S14. Line shape–fitting analyses of the dPar3 PDZ2 interaction with the Shg PBM.

Table S1. Statistics of x-ray data collection and model refinement for the dPar3 PDZ1:dPar6 PBM complex.

Table S2. Kd’s for the dPar3 PDZ domains and their interactions with the PBMs of dPar6 and Shg.

Table S3. Constructs used in this study.

Table S4. Cross-peaks and number of NMR titration points used for line shape analysis of the individual dPar3 PDZ domains.

REFERENCES AND NOTES

Acknowledgments: We thank A. Wodarz, E. Knust, P. Müller, and R. Sprangers for helpful comments on the manuscript. We are grateful to the Bloomington Drosophila Stock Center for clones and A. Streit and D. Lazzaretti for supplying cDNA. A. Neu and A. M. Liaci are acknowledged for help with x-ray data collection and crystal structure determination and D. Lazzaretti, C. Igreja, T. Raisch, J. Overbeck, P. Wurm, and C. Stollmaier for technical advice. Funding: This work was supported by the Max Planck Society (MPS; to S.W.) and the German Research Foundation (Deutsche Forschungsgemeinschaft) (DFG-KR3901/1-2 and SFB699-A13 to M.P.K.). F.A.R. is a recipient of an International Max Planck Research School fellowship funded by the MPS. Author contributions: T.P. and S.W. conceived the project. T.P. contributed the Par3 constructs. P.L.S., F.A.R., S.R.B., and S.W. recorded and analyzed NMR data. S.R.B. collected and analyzed crystallographic data. F.A.R. conducted GST pulldowns. A.M. and C.H. performed coimmunoprecipitation experiments. M.C.S.-S. cloned, expressed, and purified protein constructs for NMR analysis. L.K. conducted S2R cell culture assays. M.P.K. generated and imaged the fly embryos. S.W. wrote the manuscript with the help of the co-authors. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Structure coordinates for the dPar3 PDZ1:Par6 PBM complex are deposited in the PDB as 5OAK. Backbone chemical shift assignments for the dPar3 PDZ1, PDZ2, and PDZ3 and the hPar3 PDZ2 and PDZ3 domains are deposited in the Biological Magnetic Resonance Bank database under accession codes 27197, 27203, 27198, 27204, and 27205, respectively.
View Abstract

Navigate This Article