Cellular requirements for PIN polar cargo clustering in Arabidopsis thaliana

Summary Cell and tissue polarization is fundamental for plant growth and morphogenesis. The polar, cellular localization of Arabidopsis PIN‐FORMED (PIN) proteins is crucial for their function in directional auxin transport. The clustering of PIN polar cargoes within the plasma membrane has been proposed to be important for the maintenance of their polar distribution. However, the more detailed features of PIN clusters and the cellular requirements of cargo clustering remain unclear. Here, we characterized PIN clusters in detail by means of multiple advanced microscopy and quantification methods, such as 3D quantitative imaging or freeze‐fracture replica labeling. The size and aggregation types of PIN clusters were determined by electron microscopy at the nanometer level at different polar domains and at different developmental stages, revealing a strong preference for clustering at the polar domains. Pharmacological and genetic studies revealed that PIN clusters depend on phosphoinositol pathways, cytoskeletal structures and specific cell‐wall components as well as connections between the cell wall and the plasma membrane. This study identifies the role of different cellular processes and structures in polar cargo clustering and provides initial mechanistic insight into the maintenance of polarity in plants and other systems.


Introduction
Auxin is a fundamental plant hormone that plays crucial roles in a plethora of developmental processes (Mockaitis & Estelle, 2008;Grones & Friml, 2015). The key mechanism in auxin action is its directional (polar) transport between cells for its differential distribution in plant tissues. This process underlies a plethora of developmental processes, such as embryonic axis establishment, root and shoot tropisma, meristem activities, and root and shoot organogenesis (Adamowski & Friml, 2015). Polarly localized members of the plant-specific family of PIN-FORMED (PIN) auxin transporters regulate both the rate and the directionality of this auxin transport that is essential to connect polarities at the individual cell level and the tissue and organ levels Sauer et al., 2006a;Glanc et al., 2018;Skokan et al., 2019;Mazur et al., 2020;Zhang et al., 2020). In animal, cell polarity is regulated through several conserved factors (Crumbs, Scribble and PAR) that are absent in known plant genomes (Geldner, 2009;Kania et al., 2014), and tight junctions separating the polar domains between neighboring epithelial cells (Nelson & Beitel, 2009). Instead of tight junctions, plants possess a cell wall, a crucial cellular component that provides structural integrity to plant tissues and controls cellular growth and architecture (Wolf et al., 2012a). Mechanical strains exerted on the cell wall are transmitted toward the plasma membrane (PM) through still undiscovered connections. Both PIN polarity and endocytic trafficking can be affected by mechanical or osmotic stresses (Nakayama et al., 2012;Zwiewka et al., 2015) and PIN internalization is also regulated by the cytoskeletonlinked Rho-like GTPase (Craddock & Yang, 2012).
How can polarity of the polar cargoes be maintained within the rather fluid PM in plant cells in the absence of diffusion barriers? An experimental and computational simulation revealed that within the PM polar domains, PIN proteins are recruited into nonmobile signal aggregates, called clusters, suggesting that PIN clusters may play a role in polar auxin transport. This phenomenon might be critical for polarity maintenance together with endosome-guided cargo recycling, super-polar targeting of PIN proteins to the center of polar domains, and PIN protein retrieval at the lateral cell side by clathrin-dependent endocytosis (Kleine-Vehn et al., 2008Glanc et al., 2019;Narasimhan et al., 2020). Clustering, which has been reported to regulate polarity in Schizosaccharomyces pombe (fission yeast), Saccharomyces cerevisiae (baker's yeast) and Caenorhabditis elegans cells (Dodgson et al., 2013), might be a universal way to regulate protein polarity, and its mechanism may well be conserved. However, the characteristics of PIN clusters and the cellular and genetic factors that regulate clustering of polar cargoes remain unclear.
Genetic and pharmacological interference with the deposition of cellulose, which is the major component of the cell wall, and mechanical disruption of the cell wall increased lateral diffusion and defects in the polar distribution of PIN proteins (Feraru et al., 2011). A connection between microtubule arrangements, the cell wall and PIN1 localization has also been suggested in the shoot apical meristem (Heisler et al., 2010). In addition, the cell wall has been shown to play a crucial role in immobilizing PM proteins and the cellulose deposition pattern in the cell wall to affect the trajectory and speed of PM protein diffusion (Martini ere et al., 2012). Although both cell wall and polar cargo clustering are important for lateral diffusion and polarity, their exact roles and mutual relationship remain unclear. Here, we dissected the cellular characteristics of PIN clusters by using different microscopy methods and identified the cellular factors involved in polar cargo clustering, such as lipid kinase pathways, the cytoskeleton, cell wall and cell wall-PM connection.

Confocal microscopy
Three days after germination, the Arabidopsis seedlings were mounted in liquid ½MS under a cover glass or under a small piece of growth medium agar in a chamber with a cover glass bottom. In most cases, the meristematic root zone was imaged, but other regions were imaged in some experiments as indicated . Confocal images were obtained with a Zeiss LSM 700 using a 9100/NA 1.46 oil objective lens. Green fluorescent protein (GFP) fluorescence was excited at 488 nm (laser power 10%), and emission was collected 493 nm. Data for Supporting Information Videos S1 and S2 were taken with a spinning disk microscope (Andor Spinning Disc System; Andor Technology, Belfast, UK) with an inverted observer (Zeiss) and a 9100/NA 1.4 oil objective. The reconstruction of confocal 3D images and videos were processed in IMARIS (v.7.7.4).

Vibratome root cutting
The 3-d-old seedlings were imbedded into 5% (w/v) low-melting analytical agarose (Promega). The agar with roots was cut into cubes and glued onto the blade of Vibratome VT 1200S (Leica Microsystems, Wetzlar, Germany). Sections of 200-400 µm thickness were cut and placed on slides mounted with water for immediate observation under the confocal microscope.

Postembedding immunogold transmission electron microscopy
Root tips of 3-d-old seedlings of wild type or chemically treated Arabidopsis seedlings were excised, immersed in 20% (w/v) BSA, and frozen immediately in a high-pressure freezer (EM PACT; Leica Microsystems). Freeze substitution was carried out in an EM AFS 2 (Leica Microsystems). Cells were freeze-substituted in dry acetone with 0.1% (v/v) glutaraldehyde for 4 d as follows: À90°C for 24 h, 2°C h -1 increase for 15 h, À60°C for 16 h, 2°C h -1 increase for 15 h, and À30°C for 8 h. At À30°C, the carriers were rinsed three times with acetone for 20 min each time. Samples were then slowly warmed to 4°C, stepwise infiltrated over 3 d at 4°C in hard-grade LR-white resin (London Resin) and embedded in capsules. Polymerization was done in an EM AFS (Leica Microsystems) with UV illumination over 6 d, starting at 0°C and ending at 37°C. Ultrathin sections of gold interference color were cut with an ultramicrotome (EM UC6; Leica Microsystems) and collected on Formvar-coated copper mesh grids. All immunolabeling steps were done in a humid chamber at room temperature. Grids were floated upside down

Research
New Phytologist on 25 µl aliquots of blocking solution (5% (w/v) BSA, 1% (w/v) fish skin gelatin (FSG) in PBS) for 20 min, followed by washing five times for 5 min each wash (1% (w/v) BSA in PBS). Grids were incubated in a 1 : 300 dilution (1% (w/v) BSA in PBS) of biotin-conjugated goat anti-GFP primary antibodies (Rockland 600-106-215) for 120 min, followed by washing five times for 5 min each wash (0.1% (w/v) BSA in PBS). The grids were then incubated with a 1 : 10 000 dilution of unconjugated rabbit antibiotin (Rockland 100-4198) bridging antibodies for 30 min, followed by washing five times for 5 min (0.1% (w/v) BSA in PBS). After a final incubation with protein A/10-nm gold (PAG 10nm ; Cell Biology, Utrecht University, the Netherlands), grids were sequentially washed twice for 5 min each time with 0.1% (w/v) BSA in PBS, PBS and double-distilled water. Control experiments consisted of treating sections with bridging antibodies and/or PAG 10nm alone. Sections were post-stained in an automatic EM AC20 contrasting system (Leica Microsystems) for 30 min in uranyl acetate at 20°C and for 7 min in the lead stain at 20°C. Grids were viewed with a transmission electron microscope (JEM1010; Jeol, Tokyo, Japan) operating at 80 kV with the Image Plate Technology from Ditabis (Pforzheim, Germany). For each sample, cluster numbers and distribution were calculated from the analysis of at least 19 images, at least five seedlings and 15 cells that had been analyzed.

SDS-digested freeze-fracture replica labeling
Sodium dodecyl sulfate-digested freeze-fracture replica labeling (SDS-FRL) was applied with some modifications from the method described for mammalian tissue samples M€ obius et al., 2010). Details of the modified method are provided in Methods S1.

Sampling and analysis of SDS-FRL data
Four to seven replicas were used for quantification of immunolabeling per area of interest that were apical (shoot apex-facing), lateral and basal (root apex-facing) domains of PIN2-Venus and PIP2a-GFP epidermal cells. Within these areas, profiles were selected at random and electron micrographs were taken at a magnification of 939 000 to 993 000. The magnification was verified by a calibration grid. Quantification was done either manually with the iTEM CE software (Olympus Soft Imaging Solutions, M€ unster, Germany) or semi-automatically using FIJI and MATLAB. The semi-automatic method was done as follows, in FIJI: black shadows were manually overdrawn with white, and then a threshold was applied manually to highlight gold particles only. The plugin 'Analyze Particles' was applied. The obtained list of coordinates was imported into MATLAB and particles were assigned to groups based on their maximal distance of 55 nm. Data were expressed as mean AE SD. To compare the density of immunoparticles in different domains and genetic lines, a Mann-Whitney U-test (P = 0.05) was applied. Statistical analysis was carried out in PRISM (GraphPad, La Jolla, CA, USA). At least 25 cells were analyzed.

Quantification of clusters from confocal microscopy
The clusters are distributed in cell membranes, generally perpendicular to the optical axis of the microscope. Hence, the quality of the images is limited by the axial (Z) resolution of the objective lens used. For confocal microscopy, a 9100/1.46 oil objective lens was used with a Zeiss LSM700 confocal microscope. The obtained pixel sizes were 0.089/0.089/0.313 µm (x/y/z). The axial asymmetry was ignored, assuming that elongations in Z are artificial. As the very small volumes made segmentation problematic, a blob detection algorithm was implemented to find the clusters. For computational convenience, the Maximum Intensity Projections (MIPs) of the axial image sections were used to identify the local maxima and to classify them according to their size. First, the brightest pixels (local maxima) were localized. Each of these local maxima is the absolute maximum of its local 3 9 3 group of pixels. In other words, the eight nearest neighbors around it receive dimmer intensities. The next step is to estimate the corresponding cluster size for each maximum. This is done by dilating the local domains around them successively, and testing if the pixels in these larger domains start to become brighter again. The considered domains have initially 3 9 3 pixels (the considered pixel and its nearest neighbors), then 5 9 5 pixels to incorporate the second neighbors, then 7 9 7 pixels, etc. The area across which no brighter pixels are found defines the estimated size of the clusters. For a radius of n pixels, a 2n + 1 9 2n + 1 pixel domain has to be considered. To analyze large sets of cell membranes, we developed a software program in MATLAB (MathWorks, Natick, MA, USA). A batch series involving the steps described above was executed on a series of regions of interest (ROIs) selected interactively by the user. These ROIs might have different sizes, but the surface concentrations are given. We expect to find some correlation of the surface cluster densities for cells of the same type, and some significant difference from one cell type to another.
At small sizes, the cluster detection was not reliable. A simple explanation might be that the numbers were not consistent because the observed structures were not the clusters of interest, the size of which we did not know in advance. At the large extreme, other structures were found instead, such as overlapped sections of the same membrane. Between these boundaries, clusters ranging from 0.44 µm (5 pixels) to 0.80 µm (9 pixels) at the confocal microscope were selected. We considered that any thresholding would bias the analyses, and for this reason, we excluded the use of the actual intensity values for blob detection. The way to do it was to compare intensity differences between neighbors, but not particular intensity values. In this way, the brightest spot of a 9 9 9-pixel domain will be classified as an 8pixel-diameter cluster, regardless of whether the whole group is brighter or dimmer. The only condition is that the intensities decrease consistently from the maximum away. This was very robust in general, but some false positives might arise if a bright pixel appears in a very dim area, such as the border of another cell in a dark area. For this reason, the projections were checked manually and some cases were discarded.

Lipid-protein blot overlay assay
The lipid-protein binding assay was performed as previously described (Tan et al., 2020b). In brief, the recombinant His-PIN2HL was expressed and purified from Escherichia coli. PIP strips (P-6001, Echelon Bioscience, Salt Lake City, UT, USA) membrane was blocked in the blocking buffer with 3% BSA in 19 TBST for 1 h. Purified His-PIN2HL (20 µl in 10 ml 19 TBST) was incubated with the membrane for 2 h. The membrane was washed three times for 5 min with 19 TBST, and then incubated with anti-His HRP-conjugated antibody (dilution 1 : 4000; Agrisera, V€ ann€ as, Sweden) for 2 h at room temperature. After washing three times for 5 min with 19 TBST, the bound protein was detected using SuperSignal western detection reagents (Thermo Fisher Scientific, Waltham, MA, USA) in an Amersham 600RGB molecular imaging system (GE Healthcare, Little Chalfont, UK).

Auxin transport assay
The auxin transport assay was performed in etiolated hypocotyls as previously described (Lewis & Muday, 2009;Tan et al., 2020a). First, 6-d-old etiolated seedlings were transferred to the MS medium plates. For 1-N-naphthylphthalamic acid (NPA) treatment, 6-d-old etiolated Col-0 seedlings were transferred to the MS medium plates supplemented with 5 µM NPA. MS medium/1.25% agar droplets with 500 µM 3 H-IAA were prepared and were placed on the upper part of decapitated hypocotyls of 6-d-old seedlings, one droplet per hypocotyl. Fifteen hypocotyls were regarded as one replicate, with three replicates per genotype. After incubation in the dark for 6 h, upperparts with 3 H-IAA droplet and roots of hypocotyls were cut off, and the rest of the hypocotyls were collected and frozen in liquid nitrogen, and then homogenized in the 1 ml scintillation solution. After incubation in the scintillation solution overnight, the samples were evaluated with a scintillation counter (Hidex 300XL). The sample with only 1 ml scintillation solution was also measured as a background control.

Clustering of polar cargos observed by confocal microscopy
In A. thaliana, the auxin transporters PIN1 and PIN2 are polarly localized proteins that form clusters at the PM (Kleine-Vehn et al., 2011). We examined PIN2-Venus clusters (Leitner et al., 2012) in three dimensions and compared them to the nonpolar membrane marker PIP2a-GFP (Plasma Membrane Intrinsic Protein 2a) (Cutler et al., 2000) as a control ( Fig. 1a,b,e). We selected the nonpolar PM aquaporin PIP2a based on its localization and functional differences from PINs: polarly localized PINs are located in different membrane microdomains from PIP2a and PIN directionally transports auxin, while PIP2a is internalized upon exposure to salt stress conditions (Chevalier & Chaumont, 2014). Video S1 shows the distribution of PIN2-GFP in a 3D rotation of a typical image stack. Live-cell imaging showed that the PIN2-GFP clustering appears to be stable over 13 min while the root tip grows through the field of view (Video S2). Due to the limited axial (Z) resolution of the objective lens in light microscopes, clusters appear longer in the z-direction. To show that this is a purely optical artefact and not caused by the shape of clusters, we visualized clusters of PIN2 and PIP2a in vibratome-cut cross-sections of the meristematic zone of the root tip (Fig. 1b). However, all quantifications of clusters were done in living samples and showed pronounced clusters in PIN2-Venus when compared to the nonpolar marker PIP2a-GFP or ROP6-GFP (Fig. 1e). To quantify the clusters from Zstack confocal images, we developed software in the MATLAB environment ( Fig. 1c,d). Using this software, we quantified cluster densities in different zones and at various developmental stages and found similar PIN2 cluster densities throughout the root cap, transition and meristem zones (Fig. S1). Clustering of the polar-localized PIN2 appeared to be specific to these New Phytologist (2021)

Clustering at different polar domains observed by postembedding immunogold electron microscopy
To analyze the distribution of proteins at a higher resolution, we examined PIP2a-GFP, PIN1-GFP and PIN2-Venus by postembedding immunogold electron microscopy (IEM) (Fig. 2a-c). With this procedure, the gold particle localization does not represent directly the epitope localization, because the technique is based on indirect antigen immunodetection. Hence, gold particles may be as far as 20 nm from its epitope, due to the size of the interspersed immunoglobulins and Fab 0 fragments. We defined clusters when the distance of more than two gold particles near their closest neighbor was <55 nm (Fig. S2a). This criterion is based on epitope-to-label distances, estimates from protein interactions and protein sizes themselves (Casuso et al., 2010). This quantification principle was applied to the subsequent IEM experiments. We could identify clear differences in cluster diameters and cluster density, namely both polar markers PIN1-GFP and PIN2-Venus had larger and denser clusters than the nonpolar marker PIP2a-GFP (Fig. 2d,e). The PIN family proteins are known to exhibit distinct membrane distributions in roots (Feraru & Friml, 2008;Luschnig & Vert, 2014). PIN1 and PIN2 represent basally and apically localized proteins, respectively. To dissect the density of polar cargoes in different polar domains at the subcellular level, we used IEM to quantify the particle numbers per cluster (Fig. 2f,h,j) as well as the overall number of immune particles (Fig. S2c-h). Clusters were assigned to apical or basal domains based on their positions relative to the cell wall (Fig. S2b). These parameters were compared at the apical (shoot apex-facing), basal (root apex-facing) and lateral surfaces in PIP2a-GFP (Fig. 2f), PIN1-GFP (Fig. 2h) and PIN2-Venus (Fig. 2j). In the nonpolar PIP2a-GFP control, the particles were almost evenly distributed along the membrane. In PIN1-GFP and PIN2-Venus, the particles and clusters were preferentially located at the basal and apical sides, respectively. At the lateral side, the PIN1 and PIN2 cluster densities decreased along a gradient from the basal and apical domains, respectively, as schematically depicted (Fig. 2g,i,k). In the case of PIN2-Venus, the number of particles per cluster also varied along this gradient (Fig. 2j). The results show that the distribution of the PIN clusters is most prominent at the polar domains and gradually decreases along the lateral side away from the polar domain.

Distribution of PIN proteins revealed by SDS-FRL
To obtain semiquantitative data on cluster density and arrangement within the whole PM sheet, we additionally applied SDS-FRL (Figs 2l,m,S3). This procedure facilitates the localization of integral membrane proteins beyond the limitations of thin-section electron microscopy and reveals views of large cell membrane surfaces. For technical reasons, root samples were fractured only in the longitudinal plane ( Fig. S3a) and, hence, apical and basal cell surfaces remained largely inaccessible for immunocytochemical investigations. Therefore, the density of immunoparticles (anti-GFP) in PIN2-Venus was analyzed within lateral domains of epidermal cells close to the apical surface (distance < 2 µm from the apical edge), in centrolateral regions, and close to the basal surface (distance < 2 µm from the basal edge) (Fig. 2n,o). On average, 21.1 AE 10.2 immunoparticles per µm 2 were detected in apical domains. The overall particle density was significantly reduced in central domains with 3.3 AE 0.8 particles per µm 2 and in domains close to the basal surface with 3.1 AE 1.0 particles per µm 2 (area analyzed each = 100 µm 2 ). In PIP2a, numbers were similar in all domains (apical 3.2 AE 1.8, central 2.5 AE 1.1, basal 2.7 AE 1.5) (Fig. 2n). Immunoparticles were nonhomogenously distributed throughout the epidermal PM, and tight clusters (particle doublets within 55 nm from each other; mean distance = 32.9 AE 10.6 nm) as well as large particle aggregations were often observed (Fig. 2m,o). Such clusters were absent in control experiments without primary antisera, indicating that the conjugated antibodies and gold particles themselves did not aggregate. The Venus epitope of the PIN2-Venus fusion protein only appeared at the protoplasmic side (P-face) of the membrane (Fig. S3c), which is characterized by a coarse appearance, and were absent in endodermis and stele cell types, where PIN2 is not expressed (Fig. S3d), demonstrating the immunolabeling specificity. The immunolabeling of nonpolar PIP2a appeared at the membrane P-face, verifying labeling specificity, and was almost evenly distributed within the different cell domains. In most cases (84.5 AE 10.2%) individual (scattered) particles were observed, while clusters (particle doublets within 55 nm from each other; mean particle distance = 31.5 AE 11.9 nm) were rarely observed (12.3 AE 9.2%). Only a few clusters (3.2 AE 3.7%) showed three to five particles (area analyzed each = 100 µm 2 ) (Fig. 2o). By contrast, PIN2 showed only 29.4 AE 10.6% of scattered, 15.9 AE 6.1% Quantification of the number of clusters with radii >55 nm in PIP2a-GFP (2.31 AE 1.7), PIN1-GFP (9.1 AE 1.2) and PIN2-Venus (6.5 AE 0.8). Ten representative images from each line were quantified. (f-k) Quantitative analysis of IEM particles in PIP2a-GFP, n = 18 (f, g), PIN1-GFP, n = 15 (h, i), and PIN2-Venus, n = 9 (j, k) at apical, basal and lateral polar domains. Average particle number per cluster (f, h, j). PIN1 and PIN2 both show significant differences in average particle number per cluster between apical and basal domains (h, j), and schematic cluster distribution (g, i, k) in single root cells are presented. (l) Specific PIP2a-GFP immunolabeling for epidermal cells at the plasma membrane leaflet facing the protoplasm (membrane P-face) in SDS-FRL. The PIP2a-GFP labeling in epidermal cells is similar in apical and in basal fields. Bars, 500 nm. (m) Characterization of PIN2-Venus distribution in SDS-FRL. Immunogold particles labeling GFP (10-nm gold). Many more particles can be detected in apical fields compared to basal fields. Particles are nonhomogeneously distributed in the plasma membrane, forming loose aggregations and tight clusters. For quantitative analysis, gold particles were marked in red. A 27.5-nm-radius circle around the particle center was drawn in blue (enlargement of the boxed area on the right). Clusters were defined as particles residing within overlapping blue circles (center-to-center distance ≤ 55 nm). Bar, 500 nm. (n) Quantification of immunogold particle densities in PIN2-Venus and PIP2a-GFP epidermal cells in SDS-FRL. In PIN2-Venus cells the number of particles per area was significantly higher in domains close to the apical surface (distance ≤ 2 µm from the apical edge) compared to central (2 µm around the midline) and basal domains (distance ≤ 2 µm from the basal edge) (apical: 21.1 AE 10.2, central: 3.3 AE 0.8, basal: 3.1 AE 1.0). In PIP2a-GFP cells numbers were similar in all domains (aplical 3.2 AE 1.8, central 2.5 AE 1.1, basal 2.7 AE 1.5). (o) Characterization of immunogold particle distribution in PIN2-Venus and PIP2a-GFP epidermal cells. Scattered, single particles; doublets, two particles with a center-to-center distance ≤55 nm; clusters, at least three particles with a nearest-neighbor distance ≤55 nm. PIN2: scattered: 29.4 AE 10.6, doublets: 15.9 AE 6.1, clustered: 54.7 AE 12.8; PIP2a: scattered: 84.5 AE 10.2, doublets: 12.3 AE 9.2, clustered: 3.2 AE 3.7. Values are means AE SD. Mann-Whitney U-test: *, P < 0.1; **, P < 0.01; ***, P < 0.001. (2021)

Research
New Phytologist of doublets and 54.7 AE 12.8 of clustered particles. Taken together, the SDS-FRL data were consistent with the confocal microscopy observations and post-embedding IEM quantification, further confirming the PIN protein clustering and PIN2 distribution gradient along the lateral sides.

PIN2 clusters are independent of several cell surface structures
The localization of PIN2 in distinct clusters at the PM suggests a possible association with the cell wall, the cytoskeleton or membrane structures. Plasmodesmata that function in cell-to-cell communication are among the structure types confined to discrete spots within the membrane (Brunkard et al., 2015). No PIN2 colocalization with plasmodesmata was found by PIN2 immunolocalization in the marker line 35S::P30-GFP (Kim et al., 2005) (Fig. S4a). Colocalization of PIN2 clusters could also not be detected by aniline blue staining (Stone et al., 1984) of callose, a plant polysaccharide deposited at plasmodesmata and with multiple functions in growth, development and stress response (De Storme & Geelen, 2014) (Fig. S4b). Moreover, Nile red staining was used to discover intracellular lipid droplets (Greenspan et al., 1985) to observe whether PIN clustering was associated with their lipid ester core rather than with the membrane phospholipid. Again, no PIN2 clusters colocalized with these droplets (Fig. S4c). There was no pronounced colocalization with other proteins known to be localized in membrane microdomains such as Remorin1.2 (Fig. S4d) using immunolocalization; however, it is unclear how well the immunolocalization protocol preserves the cluster localization of nonmembrane integral proteins. The absence of PIN colocalization with these features suggests that PIN clusters may represent a unique, so far unknown, structure distinct from many of the recognized cellwall and membrane structures.
To investigate whether the phospholipid composition of the membrane is involved in PIN2 clustering, we examined PIN2 clusters in lipid kinase mutants or after treatment with PIs and PtdOH biogenesis inhibitors. The phosphatidylinositol 4-phosphate (PtdIns4P) 5-kinases PIP5K1 and PIP5K2 have been reported to be redundantly required for the polar localization of PIN proteins. In the pip5k1 À/À pip5k2 À/À (pip5k1 pip5k2) double mutant, the PIN protein distribution is partially apolar (Ischebeck et al., 2013;Tejos et al., 2014). As PIP5K1 is located at the apical and basal membrane (Ischebeck et al., 2013;Tejos et al., 2014), we noted that YFP-PIP5K1 formed visible aggregates at the cell surface (Fig. 3b), suggesting a close link between the PIP5K function and polarity and clustering at the cell surface. To test whether the PIP5K function in polarity is related to clustering of PIN2, we examined PIN2 clusters in pip5k1 pip5k2 mutants (Fig. 3c). The number and density of the PIN2 clusters decreased significantly in pip5k1 pip5k2 ( Fig. 3d; Table S1), suggesting this phosphoinositide pathway contributes to PIN2-GFP clustering.

Research
New Phytologist crucial for clustering of polar cargoes; in particular, PtdIns(4) P, PtdIns(4,5)P 2 , the derivatives of PtdIns(4,5)P 2 as well as PIP5K, which is itself clustered at the PM, are required for maintaining PIN clusters.
Peripheral membrane proteins are recruited to the plasma membrane via lipid/protein or protein/protein interaction. To test whether PIN2 protein binds to the lipids directly, we performed the lipid-binding assay of His-tagged PIN2 hydrophilic loop (His-PIN2HL) using PIP strips. We found that His-PIN2HL binds to various phospholipids, including PtdIns(3)P, PtdIns(4)P, PtdIns(5)P, PtdIns(4,5)P 2 , PtdIns(3,5)P 2 and PtdOH (Fig. 3j). Together, these findings suggest that PIN2 protein might be recruited to some specialized microdomains by the direct interaction with phospholipids.

Cytoskeleton requirements for clustering
Proteins that associate with the plasma membrane can interact with and be immobilized by association with the cell wall, the cytoskeleton or both (McKenna et al., 2014). The microtubule cytoskeleton determines the pattern of cellulose deposition in the cell wall, which, in turn, has been shown to affect protein diffusion within the PM (Martini ere et al., 2012). Therefore, we examined the structural role of the cytoskeleton in PIN clustering. First, the cytoskeleton was disrupted by oryzalin treatment to depolymerize microtubules and the effect on PIN2-Venus was observed by IEM (Fig. 4a) and confocal microscopy (Fig. 4d). Quantitative analyses revealed that in both methods the cluster size and density of clusters decreased after disruption of the microtubule network (Fig. 4b,c,e). Moreover, we observed PIN2 clusters in several microtubule mutants. Katanin is a microtubule-severing protein with a role in microtubule organization and dynamics (Luptov ciak et al., 2017). SPIRAL2/ TORTIFOLIA 1 (SPR2/TOR1), a plant-specific microtubule-associated protein, regulates the orientation of cortical microtubules and the direction of organ growth as well as the severing activity of katanin (Buschmann et al., 2004;Wightman et al., 2013). TUBULIN ALPHA-4/TORTIFOLIA 2 (TUA4/TOR2) encodes the a-tubulin subunit that consists of microtubule polymers (Elliott & Shaw, 2018). The tortifolia1 (tor1-1, Buschmann et al., 2009), spiral2 (spr2-2, Shoji et al., 2004 and tortifolia2 (tor2-1, Buschmann et al., 2009) mutants display right-handed helical growth. The density of clusters was decreased and lateral diffusion of PIN2 was increased in the katanin mutant ktn1-5 (Lin et al., 2013), tor1-1, tor2-1 (Buschmann et al., 2009) and spr2-2 (Shoji et al., 2004) (Fig. 4f-l). In addition, it has previously been shown that microtubules can control PIN2 via the CLASP-SNX1 pathway, which controls PIN2 recycling (Ambrose et al., 2013). The study showed that elimination of microtubules with oryzalin could increase the amount of PIN2 in the lytic vacuoles and, hence, reduce the amount at the plasma membrane (Ambrose et al., 2013). This is consistent with our observation that there are fewer clusters after disrupting microtubules by oryzalin treatment. Besides changing the clustering, we also observed a stop in growth after the chemical treatments. To test whether a change in PIN clustering is not simply caused by a stop in growth, we treated roots with AlCl 3 , a known inhibitor of growth (Yang et al., 2014). We did not observe a change in clustering after 1, 2 or 3 h of AlCl 3 treatment (Fig. S5a). Disruption of the actin network with Latrunculin B treatment to inhibit actin polymerization did not reduce the clustering significantly (Fig. S5b,c). Thus, we conclude that microtubules are involved in the clustering of PIN2.

Cell-wall components required for clustering
The cell wall and its constituents, such as cellulose produced by the cellulose synthase catalytic subunit (CESA), are needed for PIN polarity maintenance (Feraru et al., 2011;Martini ere et al., 2012). To test whether cell-wall components might play a role in regulating PIN clusters, we utilized pharmacological treatments known to disrupt specific components of the cell wall, such as macerozyme (pectin depolymerase), cellulase (cellulolysis catalysis), isoxaben (synthesis inhibitor of cell-wall materials) and driselase (fungal enzyme mixture, including cellulase, pectinase, xylanase and mannanase). After application, PIN2-Venus clusters were decreased in IEM labeling (Fig. 5a). As a control, we performed the same chemical treatments using PIP2a-GFP-expressing plants to see whether the chemicals influence health of the cells or GFP-fluorescence and found a normal appearance of plasma membrane signal after all treatments (Fig. S6). Quantitative analysis indicated that disruption of cell-wall components led to significant decreases in both PIN cluster size and density (Fig. 5b,c). To confirm this observation, we used the SDS-FRL method to look at PIN2-Venus and wild type plants treated with EGCG, a specific inhibitor of a pectin methyl esterase (Wolf et al., 2012b). After 24 h of treatment with 50 µM EGCG, we measured a significantly lower particle density in the apical domain (EGCG: 6.0 AE 3.5, control: 14.6 AE 9.6 clusters µm À2 ) (Fig. 5d), as well as fewer clusters (EGCG: 0.20 AE 0.35 cluster µm À2 ) compared to the control (1.34 AE 1.20 cluster µm À2 ) (Fig. 5e). As suggested by the cellulase-induced decrease in the density of PIN2 clusters, we quantified PIN2 clusters and the PIN2 apical-to-lateral ratio in two cellulose mutants, cesa3 and cesa6 (Feraru et al., 2011) (Fig. 5f-h). We found decreased PIN2

Research
New Phytologist cluster density and coinciding PIN2 polarity defects in the cellulose mutants. Moreover, plasmolysis on PIN2-Venus with 0.5 M mannitol to disrupt connections between cell walls and PM followed by IEM also slightly decreased the PIN2 cluster sizes and densities (Fig. S7), probably due to the PIN protein redistribution and PIN2 cluster accumulation on hechtian strands, a stretched PM extending from the plasmolyzed protoplast to the cell wall (Fig. S7a). Together, we found that general cell-wall components, such as cellulose, pectin, as well as the connection between PM and cell wall, are important for PIN clustering.

The integrin-like protein NDR1 is involved in PIN2 clustering and polarity
The Arabidopsis NON-RACE-SPECIFIC DISEASE RESISTANCE1 (NDR1) is a PM-localized protein with conserved motifs, suggesting homology with mammalian integrins that are well-characterized proteins involved in adhesion and signaling (Knepper et al., 2011). The ndr1-1 mutant is defective in adhesion between the cell wall and PM (Knepper et al., 2011). To test whether this protein is involved in PIN2 clustering we performed a SDS-FRL experiment using an anti-PIN2 antibody to measure the cluster density in the ndr1-1 mutant compared to wild type. First, the overall particle density was similar to the wild type and showed the typical enrichment in the apical fields ( Fig. 6a,b). We counted the number of clusters in the apical field and found that there were significantly more clusters in the ndr1-1 mutant compared to the wild type (Fig. 6c). Furthermore, the ndr1-1 mutant appeared to have clusters composed of more particles compared with the wild type (Fig. 6d). In this experiment, we included two complementation lines, one line expressing NDR1 under its native promotor and an overexpression line using the 35S promotor. The native promotor line was able to rescue the phenotype only to a certain extent, while the overexpression line fully rescued the number of clusters as well as the size of clusters (Fig. 6c,d). The morphological phenotypes of the ndr1-1 mutant showed an increased frequency of abnormal roots (11% rootless) and cotyledon development (3% cotyledon defects) (Fig. 6e-g). These phenotypes were similar to those found in multiple pin mutants or mutants defective in PIN polarity , Friml et al., 2004. These results indicate that NDR1 and the connection between PM and the cell wall play important roles in PIN clustering and polarity maintenance.

Conclusions
Outside the plant kingdom, various mechanisms and processes have been shown to play a role in cell polarity maintenance and polar cargo distribution within PM polar domains, among which the tight junction-based diffusion barriers bordering the polar domains are prominent (Mellman & Nelson, 2008). Most plant cells have no similar diffusion barriers, but, unlike their animal counterparts, are surrounded by an extracellular matrixthe cell wall. Previously, we demonstrated that plant polar cargoes, exemplified by polarly localized PIN auxin transporters, are distributed at the polar domains where they are grouped into largely immobile clusters. Computational modeling suggests membrane superpolar exocytosis and lateral endocytosis may contribute to PIN polar localization and lateral diffusion . However, the cellular processes underlying the formation of PIN clusters as well as the role of these clusters in polarity maintenance are unknown. Here, we reveal multiple insights into PIN protein clusteringwe visualized it in different cells and at different polar domains, identified several necessary cellular structural features, and provided strong, correlative support for the importance of clusters in cell polarity.

Visualization and quantification of clusters in Arabidopsis
Real-time visualization of cluster formation and maintenance in plant cells is challenging due to the high Z-resolution required to monitor the apical and basal cell surfaces lying perpendicular to the focal plane. However, the quantitative analysis of clusters is necessary to obtain insights into when and how the clusters are formed and what their lifetime is. We report on the quantification of polar cargo clustering by confocal microscopy and highresolution electron microscopy methods, specifically post-embedding immunogold and SDS-FRL. The latter approach allows the immunolabeling of integral membrane proteins without diffusion restrictions and their quantification within large fields of the PM in a pseudo-3D manner. These approaches complemented each other to show the following: a clustering of the PIN1 and PIN2 polar cargoes, which contrasts with the more uniformly distributed nonpolar cargoes; a stronger bias towards clustering in the polar domains; and an inverse correlation between the clustering gradient and the distance from the polar domain.
The number of particles per cluster may affect the generation of polar distribution gradients. For instance, in fission yeast, the dual-specificity tyrosine phosphorylation-regulated kinase (DYRK)-type protein kinase pom1p associates with the PM at the cell tip and diffuses laterally on the membrane, while aggregating in clusters and dissociating from the membrane after cluster fragmentation (Saunders et al., 2012). We also identified a correlation between the number of PIN particles per cluster and

Research
New Phytologist the concentration of PIN along its polar gradient. It remains unclear whether this correlation implies functional roles for variations in PIN cluster size, such as altered lateral diffusion or stabilization within the PM.

PIs are important for PIN clustering
The PtdIns4P 5-kinases, PIP5K1 and PIP5K2, have been reported to be redundantly required for the polar localization of (a) (d)  (Ischebeck et al., 2013;Tejos et al., 2014) and the polar localization of PIP5K to possibly facilitate the localized synthesis of PtdIns(4,5)P 2 in both plants and animals. Notably, we observed that PIP5K1 forms visible aggregates at the cell surface similar to PIN clusters. Furthermore, PIN2 cluster density decreased significantly in the pip5k1 pip5k2 mutant, suggesting a close link between PIP5K lipid kinase function and polarity and polar cargo clustering. Additionally, pharmacological interference with PI-PLC function and with PI3K and PI4K signaling inhibited PIN2 clustering. Additional pathways known to regulate PIN polarity, such as the reversible phosphorylation of conserved PIN motifs (Huang et al., 2010;Zhang et al., 2010;Barbosa et al., 2014;Rakusov a et al., 2016), MAB4/ENP/NPY1-LIKE (Furutani et al., 2011), may similarly regulate PIN clustering, but this hypothesis remains to be tested. Overall, the results suggest that lipid kinase pathways are important for cluster

Research
New Phytologist maintenance; in particular, PtdIns(4)P, PtdIns(4,5)P 2 , the derivatives of PtdIns(4,5)P 2 as well as the corresponding enzyme PIP5K, which itself shows a clustered distribution, are required for both PIN clustering and polarity, further correlating these two processes.
Interestingly, recent studies have shown that PtdIns(4)P is required for the Remorin nanodomain organization (Gronnier et al., 2017;Legrand et al., 2019;Huang et al., 2019) and the association of D6PK to the PM depends on the phospholipid composition and PIP5K function (Barbosa et al., 2016), suggesting that PIs not only affect clustering of PINs but also other important PM-resident proteins.
In addition to PIs, the PM lipid composition such as phosphatidylserine (PS) ratios (Skotland & Sandvig, 2019), sphingolipid chain length (Markham et al., 2011), glycosyl inositol phosphorylceramide (GIPC) content (Gronnier et al., 2016) and sterol content (Tapken & Murphy, 2015) probably play an important role in nanodomain formation and funtion. Nonetheless, the exact involvement of different phospholipid membrane constituents and their possible clustered distribution within the PM are topics for further investigation.
Cell wall and connections to the PM are required for PIN clustering Colocalization studies and pharmacological and/or genetic interference with cellular structural components have suggested that PtdInsP-related constituents of the membranes, microtubule cytoskeleton and cell wall, as well as connections between cell wall and PM are required for PIN clustering. Plant developmental processes, such as embryo development, anisotropic growth, directional cell elongation as well as shoot and root growth are impaired in mutants deficient in lipid kinases, microtubules, cell wall or in ndr1-1 (Fig. 6e-g; Shoji et al., 2004;Buschmann et al., 2009;Persson et al., 2007;Feraru et al., 2011;Tejos et al., 2014;Luptov ciak et al., 2017;Hu et al., 2018). Coincidently, PIN2 clustering andpolar auxin transport (Fig. S8) are also altered in lipid kinase, microtubule and ndr1-1 mutants. Whereas the ndr1-1 mutant shows increased PIN2 clustering concomitantly with increased auxin transport (Figs 6c, S8), the pip5k1 pip5k2 and spr2-2 mutants show decreased PIN2 clustering but increased auxin transport (Figs 3d, 4h). Therefore, the casual connection between auxin transport and PIN2 clustering remains unclear and awaits further study. Structures surrounding the PM may act to restrict the diffusion of scaffold proteins and molecular fences, and perhaps also, through anchoring, promote their clustering within the PM (Mellman & Nelson, 2008). A variety of multicomponent linkages between the cell wall, PM and cytoskeleton are known in plants (Liu et al., 2015). Some fasciclin-like arabinogalactan (FLA) proteins are thought to function in nanodomain-cell-wall interactions (Tapken & Murphy, 2015). A membrane skeleton fence model was proposed to illustrate the regulation of PM organization (Kusumi & Sako, 1996). In the model, membrane skeleton/cytoskeleton structures control the mobility and assembly of membrane proteins in specialized domains. We speculate that this model might also describe the plant PM because several nanodomain proteins were reported to align with or bind to microtubules or actin in plants (Jarsch et al., 2014;Gui et al., 2015;Lv et al., 2017;Dan ek et al., 2020). In addition, cytoskeleton disruption leads to an increased diffusion rate or reduction of nanodomains (Lv et al., 2017;McKenna et al., 2019). Through genetic and pharmacological interference, we confirmed the involvement of microtubules in PIN clustering (Fig. 4). At the same time, increased lateral diffusion of PIN2 proteins in the microtubule mutants was also detected (Fig. 4l), suggesting that a membrane cytoskeleton fence confines the movement and assembly of PIN2 clusters. The cell wall confines the lateral diffusion of PM proteins (Martini ere et al., 2012). Cell-wall disruption by protoplasting increases the lateral diffusion of PM proteins, including PIN2 and PIN3 (Feraru et al., 2011;Martini ere et al., 2012;McKenna et al., 2019). Consequently, PIN2-GFP polarity measured by the apical to lateral signal ratio decreased in cellulose synthase mutants cesa3 and cesa6 (Fig. 5f,h).
NDR1 is an integrin-like protein, which is thought to mediate the PM-cell wall adhesions/connection (Knepper et al., 2011;Liu et al., 2015). NDR1 recognizes the extracellular matrix through its NGD motif, which is similar to the RGD motif (Knepper et al., 2011). Cell wall-PM adhesion is altered in the ndr1-1 mutant (Knepper et al., 2011). Several studies have shown that increasing the distance between the cell wall and PM by plasmolysis promoted the lateral diffusion of PM proteins, such as PIN2 (Feraru et al., 2011;Martini ere et al., 2012). Our data showed that PIN2 clustering increased in the ndr1-1 mutant (Fig. 6c,d). These findings demonstrated that the PM-cell wall adhesion/connection confines the mobility and assembly of PM proteins. In epithelial cells, RGD-containing ligands bound to the integrin receptor activate the RhoA-formin signaling pathway to generate PM nanodomains (Kalappurakkal et al., 2019). Thus, we speculate that there might exist a more complicated mechanism underlying PM-cell wall adhesion in the regulation of PIN clustering.
Certainly, the diversity of elements necessary for PIN clustering that has been identified here implies a complicated network of interactions to control this process (Fig. 7). However, interference with specific components of these complex structures may have resulted in their generalized impairment that might affect a specific linkage between the cell wall, PM and cytoskeleton.
Together, the use of diverse imaging methods combined with genetic and pharmacological approaches, interfering with diverse cellular processes, has provided insights into the distribution of polar cargoes (illustrated by PIN proteins) in plants.
Our results indicate that the clusters of polar cargos are unique structures presumably linked to overall polarity regulation. The strong dependence of clustering on the plant cell wall and its connection to the PM suggests that plants might have evolved mechanisms of polarity maintenance distinctive from those of metazoans.

Supporting Information
Additional Supporting Information may be found online in the Supporting Information section at the end of the article.        Methods S1 SDS-digested freeze-fracture replica labeling.
Video S2 PIN2-GFP root cap and epidermal cell clusters during root growth.
Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.