Structural Organization of Human Full-Length PAR3 and the aPKC–PAR6 Complex

The tripartite partition defect (PAR) polarity complex, which includes the proteins PAR3, atypical protein kinase C (aPKC), and PAR6, is a major regulator of cellular polarity. It is highly conserved and expressed in various tissues. Its largest component, PAR3, controls protein–protein interactions of the PAR complex with a variety of interaction partners, and PAR3 self-association is critical for the formation of filament-like structures. However, little is known about the structure of the PAR complex. Here, we purified non-filamentous PAR3 and the aPKC–PAR6 complex and characterized them by single-particle electron microscopy (EM). We expressed and purified an oligomerization-deficient form of PAR3, PAR3V13D,D70K, and the active aPKC–PAR6 dimer. For PAR3, engineering at two positions is sufficient to form stable single particles with a maximum dimension of 20 nm. aPKC–PAR6 forms a complex with a maximum dimension of 13.5 nm that contains single copies of aPKC. Thus, the data present a basis for further high-resolution studies of PAR proteins and PAR complex formation.

Polarity is established and regulated by a set of evolutionarily highly conserved proteins including a set of PAR (for partitioning defective) proteins. PAR proteins have been originally identified as factors required for polarization in C. elegans zygotes [8]. The PAR protein family includes a polarity complex (PAR complex) comprising PAR3, atypical protein kinase C (aPKC), and PAR6 [9,10]. In vertebrates, the PAR complex has been studied extensively in epithelia, where the proteins localize in the apical compartment near tight junctions and have been demonstrated to be central for the establishment and maintenance of apical-basal polarity [11]. The PAR complex is also critical for neural development, where it has been suggested to be essential for the differentiation of neurites into dendrites and the axon [12][13][14]. Moreover, growing evidence suggests a link between deregulation of PAR3 and cancer development favoring cell proliferation, epithelial-mesenchymal transition (EMT) and metastatic spread in a number of tumor types [15].
Moreover, PAR3 exhibits a microtubule binding and bundling activity [22], and the PAR complex has been implicated in the regulation of the microtubule and actin cytoskeleton as a critical step in neuronal development [12,13,23]. The semi-CRIB (Cdc42-and Rac-interactive binding) domain of PAR6 can bind to the active Rho GTPase CDC42 [9], which activates aPKC to phosphorylate PAR3 and cause PAR3 dissociation from the PAR complex [24].
Major attention has been paid to the ability of the proteins to enrich in certain cellular compartments as a hallmark of polarization [13,14,[25][26][27]. Particularly, PAR3 self-association may represent the molecular basis for the enrichment of the PAR complex at target sites. The first approximately 83 amino acids of PAR3 form an N-terminal domain (NTD) that has been demonstrated to exhibit a critical role in self-association of PAR3 [25,28]. Crystallization and cryo-EM studies of the rat PAR3 NTD fragment have shown that the isolated NTD forms protein helices with a regular pitch through a number of residues including T4, V13, and D70 providing lateral interactions, as well as R9 providing longitudinal interactions [29]. Mutation studies of the isolated PAR3 NTD furthermore suggested that mutations of V13 and D70 prevent oligomerization of NTD fragments [29]. However, the structural organisation of PAR3 clusters in vivo remains unknown.
Although the ability of the PAR complex to function in diverse cellular contexts such as epithelia and neurons would clearly be explained by knowing its structure, still little is known about its architecture. Here, we aimed at characterizing non-polymerized PAR complexes, the heterodimeric aPKC-PAR6 complex, and PAR3 alone.

Identification of PAR3, aPKC, and PAR6 Isoforms Expressed in Human Neural Cells
The genes of PAR6, PAR3, and aPKC subtype iota were amplified from human neural cells [30], and their identity was confirmed by sequencing ( Fig. 1). The sequences of the encoded aPKC and PAR6 proteins are identical to the canonical human isoforms (PKCι, GenBank accession code: NP_002731.4; PAR6α; GenBank accession code: NP_001032358.1), respectively. PAR3 is a novel isoform as a result of minor changes in alternative splicing. In comparison to the human PAR3 isoform 1 (GenBank accession code: NP_062565.2), the following differences were observed: (i) the neural PAR3 isoform possessed four additional amino acids after residue 269 with D269 changed to E immediately N-terminal to the PDZ1 domain; (ii) three amino acids were omitted between positions 739-743 in the aPKC-binding domain; and (iii) 37 residues were omitted between positions 1024-1062 C-terminal to the aPKC-binding domain. The omission of these 37 residues is also seen in isoform 4 of PAR3 (NCBI accession no. NP_001171716). All these modifications in the PAR3 isoform can be explained

aPKC and PAR6 form a Defined Complex
Wild-type tagged aPKC and untagged PAR6 were coexpressed in their full-length forms using a baculovirus/ insect cell system. Upon immuno-purification of aPKC via its N-terminal 3 × FLAG tag ( Fig. 2A), a dimeric complex was obtained. We confirmed by using a kinase assay that the purified aPKC-PAR6 complex was functionally active in phosphorylation (Fig. 2B). By size exclusion chromatography (SEC), the aPKC-PAR6 complex (theoretical molecular weight of 106 kDa assuming a 1:1 complex) and co-eluting unbound aPKC (theoretical molecular weight of monomer, 68 kDa) showed peaks in fraction 24 and 26, respectively, at elution volumes somewhat higher than expected for near spherical proteins (Fig. 2C, D).

aPKC Does Not Exhibit Self-Interaction
To assess whether or not the main component of the heterodimeric aPKC-PAR6 complex, aPKC occurs in monomeric form, we co-expressed aPKC with two different tags by replacing the 3 × FLAG sequence in the aPKC construct with an HA tag sequence, followed by co-expression of the two distinguishable tagged forms of aPKC and purification via the 3 × FLAG-tag. We expressed HA-aPKC together with 3 × FLAG-aPKC in the absence of PAR6, and we also expressed HA-aPKC and 3 × FLAG-aPKC in the presence of PAR6 to test whether the presence of PAR6 had an influence on the stoichiometry of the protein complex ( Supplementary  Fig. S1A, B). We measured the recovery of HA-aPKC in FLAG pulldown assays upon 3 × FLAG-aPKC,HA-aPKC coexpression ( Fig. 2E) and 3 × FLAG-aPKC,HA-aPKC,PAR6 co-expression ( Fig. 2F) using anti HA western blotting and anti FLAG western blotting as control. Only a minor fraction of HA-aPKC was recovered irrespective of the absence or presence of PAR6: when normalized to the input (supernatant), the elution yielded 0.5% recovery of HA-aPKC and 0.14% of HA-aPKC-PAR6 (Fig. 2E, F; for SDS-PAGE, see Fig. S1A, B) consistent with no noteworthy self-interaction and predominantly monomeric aPKC alone and in complex with PAR6. Thus, the predominant form of the aPKC-PAR6 complex is an assembly with a single copy of aPKC.

aPKC-PAR6 Forms Moderately Elongated Particles
For all particles, we employed gradient ultracentrifugation as final purification step, which provides an optimum sample quality for EM [31]. When aPKC-PAR6 was run on a 5-20% glycerol gradient, the peak of the protein complex ( Fig. 2G and Supplementary Fig. S1C) occurred in fraction 12 -14 (out of 38 fractions), which corresponds to an apparent Svedberg value of about 4.5S. By EM, raw images showed a monodisperse particle population (Fig. 2H). The maximum dimension of the particles is approximately 13.5 nm, and class averages with about 24 images per class show well-discernible finestructural details indicating a well-defined structure (Fig. 2I). In particular, the particles reveal an asymmetrical, compact, moderately elongated structure.
We also combined the Stoke's radii and sedimentation data to a molecular weight estimate (MW) using Erickson's approximation MW = 4.205(S • R S ) [32] based on Siegel and Monty [33], where S is the sedimentation in Svedberg Units and R s is the radius in nm. For the aPKC-PAR6 complex, this estimation yields a predicted MW of max. ~ 100 kDa (Table 1), which is consistent with a monomeric stoichiometry of the largest protein in the complex, aPKC, in addition to PAR6.

PAR3 V13D,D70K Forms a Stable Elongated Particle
Initial expression tests of wild-type PAR3 alone indicated that wild-type PAR3 was not stable after elution (data not shown); thus, we investigated a mutant form of PAR3. To this end, we expressed a form of PAR3 mutated at two positions (V13D and D70K) that was reported earlier [29]. Expression tests with PAR3 V13D,D70K alone (i.e., not in complex with PAR6-aPKC) showed that PAR3 V13D,D70K was stable, and protein degradation could be minimized (Fig. 3A,B), which provided the ability to purify PAR3 V13D,D70K in amounts sufficient for EM analysis. In SEC, PAR3 V13D,D70K peaked in fraction 19-20 at an elution volume corresponding to a Stokes radius of ~ 6.1 nm, separate from void (Fig. 3C,D and Table 1). Another UV peak (fraction 16) visible nearby the void volume, however, contained smaller amounts of PAR3 as evinced by western blotting (Fig. 3D). PAR3 V13D,D70K was subsequently run on a 5-20% glycerol gradient (Fig. 3B), where it peaked around fraction 12-13 out of 38 fractions corresponding to the ~ 7S region (approximate MW 133-182 kDa, compare Table 1). These results are consistent with a monomeric protein given the theoretical molecular weight of 141 kDa. EM images of PAR3 V13D,D70K showed monodisperse, moderately elongated single particles (Fig. 3E), and the 2D class averages of PAR3 V13D,D70K confirmed a well-defined structure with compact shape and maximum dimensions of ~ 20 nm (Fig. 3F).

Discussion
Herein, we aimed at characterizing the non-polymerized building blocks as smallest units of the PAR complex that is formed by the PAR proteins PAR3, aPKC, and PAR6. To avoid polymerization of PAR3, we took advantage of two point mutations in the NTD of PAR3, V13D, and D70K. These mutations had been previously reported to prevent the PAR3 NTD from self-association [28], and were described to abolish interactions in the isolated NTD fragment of rat Par3 by preventing lateral packing into a helix [28,29].
The sedimentation and SEC data we present herein are in favor of single copies of the largest subunit, PAR3, and of aPKC and PAR6 in the aPKC-PAR6 complex. For the aPKC-PAR6 complex, the presence of a single aPKC protein within the aPKC-PAR6 complex has independently been validated by the double tagging assay. In the aPKC-PAR6 heterodimer, aPKC and PAR6 interact via the PB1 domain present in both proteins [17]. In the crystal structure, the PB1 domains of aPKC and PAR6 form an asymmetric heterodimer, with one copy of the aPKC and PAR6, each [17]. Together with our pulldown data, these data support a 1:1 stoichiometry of aPKC and PAR6 as smallest unit.
It has been shown that PAR3 binds essentially two copies of PAR6 via its PDZ1 and PDZ3 domains, albeit at different affinities [18]. Both dissociation constants were however reported in the micromolar range [18]. Whether or not the local enrichment of the PAR proteins at the plasma membrane is sufficient to facilitate recruitment of two PAR6 copies (or two aPKC-PAR6 heterodimers) to the same PAR3 proteins under in vivo conditions at the cell membrane, will thus need further investigation.
By EM, we did not observe formation of specific multimers of the PAR3 protein, indicating that higher order assemblies that may have formed despite the engineering were not sufficient for visualization by EM. Further research will be required to investigate how the basic building blocks characterized here enrich into higher order PAR complex assemblies inside the cell. Especially, how the rather large PAR3 protein assembles into higher order complexes and whether or not PAR3 will adopt a helical assembly in vivo remains to be investigated. Future high-resolution cryo-EM reconstructions of the PAR complex based on these data are required to address these questions. Likewise, how the occurrence of alternative splicing of PAR3 shown here contributes to tissue-specific variants of the PAR complex will require more investigation. Overall, our current studies provide projection structures of the PAR components PAR3 and aPKC-PAR6 as a step toward a detailed structural and functional understanding of these components in the establishment and maintenance of cellular polarity.

Amplification of mRNA from Human Neural Cells
Full-length human PAR6, PAR3, and aPKC subtype iota (PKCɩ) were amplified with appropriate primers (Supplementary Table S1) and cDNA synthesized using mRNA derived from human neural cells as described previously [30]. The Maxima H minus first strand cDNA synthesis kit (ThermoFisher Scientific, Waltham, MA, U.S.A.) was used for cDNA synthesis. The vector pUC57 (ThermoFisher Scientific) was used to insert the DNA fragments using suitable restriction enzymes.

Plasmid Construction
PAR3 was subcloned into the vector pGS-BacA-21122 [34], a derivative of pACEBac1, which introduced a 3 × FLAG to the N-terminus of the expressed protein. PAR3 was studied as wild-type protein and as an engineered PAR3 V13D,D70K with two amino acid changes, V13D and D70K, in the NTD. We introduced a mutation causing a kinase-dead mutant in mouse [35], PKCι K283R , into the human gene upon sequence alignment of the human PKCι (GenBank accession code: NM_002740.5) and mouse PKCι (GenBank accession code: BC021630.1) [35]. The QuickChange Lightning Site-Directed Mutagenesis kit (Agilent Technologies, Santa Clara, CA, U.S.A.) was used for site-directed mutagenesis. By Cre recombination, composite bacmids containing PAR3, PKCι, and PAR6 were created from acceptor and donor plasmids as described previously [36] for multi-protein expression in insect cells using the Multibac system [37,38].
To study the dimeric aPKC-PAR6 complex, the coding sequence of aPKC was ligated into the acceptor vector pGS-BacA-21122 [34], and PAR6 was ligated into the donor vector pIDC. Furthermore, the 3 × FLAG tag from the aPKC plasmid was replaced by an HA tag to investigate aPKC self-oligomerization. The plasmids coding for PKCι and PAR6 were combined by Cre-LoxP reactions as outlined [36]. The plasmid constructs created in this study are listed in Supplementary Table S2. The sequence of relevant plasmid elements was confirmed by Sanger sequencing (Eurofins, Ebersberg, Germany or Macrogen Europe, Amsterdam, The Netherlands).

Size Exclusion Chromatography
The purified proteins were subjected to SEC on a Superdex 200 column (GE Healthcare, Little Chalfont, U.K.) for the dimeric aPKC-PAR6 and PAR3 V13D,D70K in 20 mM HEPES (pH 7.6) supplemented with 400 mM NaCl. The Gel Filtration HMW Calibration Kit (GE Healthcare) was used for calibration of the SECs. For calibration of the elution volume as a function of Stoke's radius R s , R s values reported in [32] were used.

Gradient Ultracentrifugation
The purified proteins were run in a 5-20% glycerol gradient for 17 h at 4 °C (20 mM HEPES, pH 7.6; 400 mM NaCl) at 352,996 × g for the dimeric aPKC-PAR6 complex or 274,824 × g for PAR3 V13D,D70K . The gradients were fractionated into 38 fractions with 5 drops per fraction by fractionation from the bottom of the gradient using a P-1 peristaltic pump (GE Healthcare) as described previously [40]. For estimation of the apparent sedimentation coefficients (S), commercial standards were used (Sigma). As glycerol gradient peaks typically span over multiple fractions, an apparent S value range is given for all particles. Proteins were visualized by Coomassie staining and verified by western blotting.

EM Image Acquisition
A volume of glutaraldehyde (Sigma-Aldrich) corresponding to a final concentration of 0.075% was added to the protein samples followed by incubation overnight at 4 °C before grid preparation. Negative staining samples were prepared using the sandwich carbon method with home-made carbon film and uranyl formate or uranyl acetate (2%) [41]. The images were taken in a Tecnai T12 electron microscope (FEI, Eindhoven, The Netherlands) with a Multiscan 794 CCD camera (Gatan, Pleasanton, U.S.A.) operated at 120 kV at a nominal magnification of 52,000 × , which corresponded to an apparent magnification of 63,160x. The pixel size on the specimen level was 3.8 Å/pixel.

EM Image Processing
The particles on the images were selected manually. Determination of defocus and astigmatism of the EM images was done by fitting contrast transfer function (CTF) curves to the power spectra of the images [42]. The particle images were extracted, corrected for CTF-effects, and merged. The data set characteristics are summarized in Table S3. The classification and averaging of particles followed standard methods [43] and were performed in the statistical framework R [44] with 3-10 rounds of particle alignment followed by principal component analysis and unbiased classification using hierarchical ascendant and k-means classification (Table S3).

PKC Kinase Activity
The activity of purified dimeric aPKC-PAR6 complex was tested using the PKC Kinase Activity Assay Kit (Abcam, Cambridge, U.K.) with a protein concentration dilution series of 5 ng, 10 ng, and 20 ng of the purified protein complex following the manufacturer's instructions, and the standard error of the mean (SEM) was used for visualization (n = 3 replicates). As positive control, 36 ng of control protein (Abcam) was used. A multimode plate reader (EnSpire, PerkinElmer, Waltham, MA, U.S.A.) was used to measure the reaction signal.