Cytoskeletal Components Define Protein Location to Membrane Microdomains*

The plasma membrane is an important compartment that undergoes dynamic changes in composition upon external or internal stimuli. The dynamic subcompartmentation of proteins in ordered low-density (DRM) and disordered high-density (DSM) membrane phases is hypothesized to require interactions with cytoskeletal components. Here, we systematically analyzed the effects of actin or tubulin disruption on the distribution of proteins between membrane density phases. We used a proteomic screen to identify candidate proteins with altered submembrane location, followed by biochemical or cell biological characterization in Arabidopsis thaliana. We found that several proteins, such as plasma membrane ATPases, receptor kinases, or remorins resulted in a differential distribution between membrane density phases upon cytoskeletal disruption. Moreover, in most cases, contrasting effects were observed: Disruption of actin filaments largely led to a redistribution of proteins from DRM to DSM membrane fractions while disruption of tubulins resulted in general depletion of proteins from the membranes. We conclude that actin filaments are necessary for dynamic movement of proteins between different membrane phases and that microtubules are not necessarily important for formation of microdomains as such, but rather they may control the protein amount present in the membrane phases.

Living cells need borders and molecular compartments for biochemical reactions and storage of metabolites. The plasma membrane therefore is a prerequisite for the evolution of different life forms. It consists of a phospholipid bilayer into which proteins and special lipid species such as sterols, sphingolipids, and glycolipids are inserted. The first complex model of plasma membrane was proposed in 1972 by Jonathan Singer and Garth Nicolson (1), replacing the concept of the plasma membrane as a strict protein-lipid-protein sandwich that was generally accepted until then. In Singer and Nicolson's model, the cell membrane is a two-dimensionally oriented viscous solution in which the membrane constituents are orientated in the most thermodynamically favorable manner, hiding hydrophobic hydrocarbon chains inside the lipid bilayer and exposing polar and ionic groups to the aqueous phase. This fluid mosaic model also implied that membrane proteins as well as lipid components are distributed in a homogeneous lipid bilayer at long range, but they can form specific aggregates and phases at short range, which were also termed "lipid rafts" or membrane microdomains.
Over the past 30 years, it has become evident that the plasma membrane is not such a homogeneous structure as it was initially proposed. We now know that the lipid bilayer is asymmetric (2) and that the free diffusion of membrane proteins is restricted by their interactions with intracellular and extracellular components (3). More recently, Simons and Ikonen suggested that large ordered phases, enriched with cholesterol and sphingolipids, emerge within the plasma membrane and that they function as platforms for enrichment of certain proteins while excluding others (4). This current membrane model suggests that the mixture of sterols and polar lipids within the plasma membrane can appear in two distinct phases: liquid disordered (L d ) and liquid ordered (L o ) phase (5). In this view, the so-called membrane microdomains are considered to be part of the L o phase. Based on work on model membranes, it is suggested that lateral segregation of components into L d and L o phases occurs spontaneously (6) with the self-associating properties between sterols and highly saturated hydrocarbon chains of phopsho-and sphingolipids as the main driving force (7). Additionally, it is suggested that also specific lipid-protein and protein-protein interactions are essential for the formations of membrane domains as well as for stabilization of smaller nanodomains which subsequently may cause formation of larger platforms. In contrast to the animal cells, in plants these membrane microdomains seem to be rather immobile (8), possibly due to their attachment to the outer cell wall. More recently, it became obvious that membrane microdomains within a single cell are highly diverse and of different compositions (9). Generally, in the plant model, organisms' plasma membrane microdomains turned out to be important in plant defense (10,11), cell polarity (12,13), and general signaling properties of the plasma membrane (14,15).
The cytoskeleton was identified as an essential cellular component with important roles in membrane topography, bordering, trafficking, and organelle movement (16). Single particle tracking in mammalian cells revealed that the transferrin receptor and macroglobulin receptor demonstrate normal Brownian diffusion but only within a specific membrane compartment (17). Two hypothetical models were proposed in order to explain this phenomenon (supplemental Fig. 1). Direct interactions between transmembrane proteins and cytoskeleton are suggested to creates a barrier, called "fence," where cytosolic parts of transmembrane proteins collides with cytoskeletal components, limiting their diffusion to certain areas. These molecules can jump over the "fence" to a neighboring compartment, possibly due to the dynamic nature of the interaction of membrane proteins and cytoskeleton, where they are again temporally trapped (17). This phenomenon was recently described also in A. thaliana where the interplay between membrane microdomains and microtubules plays a role in secondary cell wall formation (reviewed in (18)). The second model assumes, additionally, that particular transmembrane proteins are anchored to and lined up along cytoskeleton and act as "pickets" to arrest free diffusion of other membrane components, including nontransmembrane proteins, within the enclosed compartment (19).
For plants, the composition of these sterol-rich membranes phases was analyzed in several biochemical studies (14, 20 -22). Thereby, low-density preparations of plasma membrane fractions after treatment with nonionic detergents (DRM 1 fractions) were considered as a biochemical representation enriched in cellular membrane ordered phases or microdomains. Proteomic studies in mammalian cells consistently reported that the DRM fraction is highly enriched with several cytoskeletal proteins such as actin, tubulin, myosin, dynamin, actinin, and supervillin (23)(24)(25). Additionally, the level of phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2), a lipid connecting the plasma membrane to actin filaments, was also significantly elevated in DRM preparations (26). Treatment with microtubule and actin depolymerizing agent results in drastic loss of many signaling proteins from these DRM fractions prepared from adult rat cardiac myocytes (27) or human embryonic retinal cells (28).
Based on this knowledge, we propose two hypothetical models for the relationship between cytoskeleton and membrane microdomains for plant cells: (i) Actin filaments and microtubules could be important in the membrane phase separation or formation of the membrane microdomains themselves. In this case, disruption of the cytoskeleton would cause a lack of phase segregation in the plasma membrane. (ii) The cytoskeleton is only important for the incorporation of specific protein into the sterol-enriched regions but not for the general formation of these phase separations. This view implies that phase separations or membrane microdomains would still be present after cytoskeleton disruption but their protein composition can be different. Another possible scenario is (iii) that cytoskeletal elements serve as anchors for membrane microdomains at particular position in the plasma membrane, so the absence of these anchors would cause the increased mobility of microdomains (supplemental Fig. 1).
The primary aim of this study was to characterize the interplay between cytoskeletal components and different membrane phases (microdomains) in A. thaliana suspension cell cultures. To reach this goal, biochemical and proteomic approaches were combined with confocal microscopy and activity assays measuring the influence of actin or tubulin disruption on the composition, localization, and biochemical properties of the sterol-enriched membrane microdomains. Thereby, for biochemical analyses, low-density detergentresistant membrane fractions are analyzed as containing cellular sterol-rich membrane compartments. Seeds of an A. thaliana dual-labeled line for actin and tubulin were obtained as a courtesy of AG Persson from the Max Planck Institute of Molecular Plant Physiology in Golm, Germany. This line allowed observation of the structure of microtubules and actin filaments in the same plant at the same time. The dual-labeled line was created by transformation of Col-0 with pART27 vector containing GFP-FABD2 fusion construct under 35 S promoter together with a gene for kanamycin resistance (29). This line was subsequently crossed with another line that was transformed with a Gateway vector containing TUA5 fused to mCherry-RFP under 35 S promoter together with genes for kanamycin resistance (30).

Plant Lines Expressing Proteins Fused to GFP/YFP/RFP-Seeds
Seedlings-Sterilized seeds of YFP-tagged remorin lines were grown on 1/2 Murashige-Skoog medium containing 1.5% agarose, 1% sucrose, and kanamycin. Seedlings transferred on plates were left for 5 days in a light-dark cycle with a 16-h photoperiod at 23°C. Prior to the performed experiments, seedlings were carefully moved into the cavity of a six-well plate filled with 1/2 liquid MS medium with 1% sucrose and placed on a rotary shaker (50 rpm).
Cell Suspension Cultures-Cell suspension cultures were established from A. thaliana Col-0 leaf callus as described (31). They were grown in flasks in full JPL medium under constant light conditions at 80 -100 mol m Ϫ2 s Ϫ1 at 23°C with rotary shaking at 120 rpm. Cells were subcultured every 7 days. In order to perform experiments, cell cultures were pooled from and independent flask and subsequently divided into desired numbers of replicates for treatment. In general, treatment experiments were carried out in at least three biological replicates, and results are presented as an average of these.
Treatment with Cytoskeleton Depolymerizing Agents-Experiments were performed with cytochalasin D, which is a potent inhibitor of actin polymerization. Cytochalasin D binds to G-actin and prevents polymerization of actin monomers (32), as well as to the barbed end of actin filaments, which inhibits the association of hew monomers at that end (33). The second drug used in the experiments was oryzalin (3,5-dinitro-N4, N4-dipropylsulfanilimide), which is a selective preemergence, surface-applied dinitroaniline herbicide with rather low toxicity to animals and fungi. This drug acts as a microtubule depolymerizing agent, and it perturbs all microtubule-related processes in plant cells (34,35). Both inhibitory effects were shown to be reversible (36). Stock solutions of cytochalasin D and oryzalin were dissolved in DMSO. The following experimental conditions were applied: Cytochalasin D treatment: Cells treated with cytochalasin D may react to the loss of the links between the plasma membrane and actin filaments by creating more filaments that might compensate for the initial depolymerizing effect of cytochalasin D (37). In order to evaluate this effect, time points instead of different concentrations were used. Cell cultures were treated with cytochalasin D in concentration of 500 nM for different periods: 30 min, 60 min, 90 min, and 120 min. The concentration used was far lower than the minimal concentration known to disrupt cytoplasmic streaming (38). Control samples were treated with respective volumes of DMSO solution. The concentration of DMSO in final solutions was not higher than 0.1%. After treatment, cells were filtered and immediately frozen in liquid nitrogen.
Oryzalin treatment: According to the literature, oryzalin has an impact on the normal structure of the microtubules already at nanomolar concentrations but often requiring long exposure (34,35). In order to perform experiments at similar time frames as the cytochalasin D treatment, we selected increasing oryzalin concentrations applied for the duration of 1 h. Cell cultures were treated with different concentrations of oryzalin (1 M, 2 M, 5 M, and 10 M) for 1 h. Instead of a disrupting agent, respective amount of DMSO was added to control samples. The concentration of DMSO in final solutions was not higher than 0.1%. After treatment, cells were filtered and immediately frozen in liquid nitrogen.
Enzyme Activity Measurements-Activity of the plasma membrane H ϩ -ATPase in DRM and DSM fractions was measured by the malachite green colorimetric assay (39). The standardization of different samples was based on total protein content, which was evaluated by the NanoOrange Protein Quantitation Kit (Invitrogen, Germany) according to the instructions. Specific inhibitors of P-type and V-type ATPases were used to distinguish the specific activity of selected enzyme from the background activity of the sample: Na 3 VO 4 as plasma membrane ATPase inhibitor, EDTA as inhibition of Ca 2ϩ -ATPases, NaN 3 as inhibition of mitochondrial ATPases, and bafilomycin A1 as inhibitor of V-ATPases (40).
In order to perform enzymatic activity measurement, 1 g of extracted protein was used per reaction mixture. The plate was placed on a shaker at 37°C. Reaction was stopped after 1 or 2 h by adding MG-AM-TRIT [three volumes of MG solution (2 mM malachite green in Milipore water) and one volume of AM solution (0.034 M ammonium molybdate in 4 N HCl, 0.1% Triton X-100) solution to each well. Addition of sodium citrate to a final concentration of 3% inhibited the color development. The change in phosphate concentration was measured by absorption using the wavelength of 630 nm.
Preparation of Membrane Fractions for Proteomic Analysis-In general, microsomal fractions, plasma membrane, and DRM/DSF fractions were prepared as described in detail (41). Frozen plant material was powderized, mixed with two volumes of extraction buffer (100 mM Hepes-KOH (pH 7.5); 250 mM sucrose; 10% (w/v) glycerol; 5 mM EDTA; 5 mM ascorbic acid; 0.6% (w/v) PVP K-25; freshly added: 5 mM DTT; 1 mM PMSF; protease inhibitor mixture (SIGMA, Germany)), and stirred gently on ice until complete homogenization was achieved. The homogenate was then filtered through one layer of Miracloth (Merck, Germany), and cell debris was spun down at 10,000 g for 8 min. The supernatant was taken and centrifuged at 100,000 g for 30 min. After this step, microsomes were acquired in the pellet and soluble proteins (SOL fraction) in supernatant (Fig. 1). The microsomal fraction was solubilized in resuspension buffer (5 mM potassium phosphate; 0.33 M sucrose; 3 mM KCl; 0.1 mM EDTA; freshly added: 1 mM DTT, 1 mM PMSF) and then deposited onto a two-phase system of polyethyleneglycol and dextran (41). After gentle mixing and centrifugation at 1,000 g for 10 min, the lower green phase contained intracellular membranes (IM fraction) while the upper phase contained cell plasma membrane (PM fraction). The latter was collected, diluted with five volumes of resuspension buffer and centrifuged again at 200,000 g for 1 h. The obtained pellet was then resuspended in TNE buffer (25 mM Tris-HCl pH 7.5; 150 mM NaCl; 5 mM EDTA; 1 mM DTT), and protein concentration was measured using Bradford assay (42). At this step, an aliquot of the plasma membrane protein was taken for analysis ( Fig. 1).
Solubilized in TNE buffer, the plasma membrane was subsequently treated with Triton X-100 in a concentration between 0.5 and 1% and in a ratio of proteins to Triton of 1:15. The detergent treatment was applied for 30 min on ice with shaking. Detergent-resistant and detergent-soluble membranes created after Triton X-100 were separated by centrifugation for 18 h through a sucrose step gradient (from the bottom: 1.8 M, 1.6 M, 1.4 M, 0.15 M). After centrifugation, a low-density band was observed as a white ring at the interface of the 0.15 M and 1.4 M sucrose phases (41). A fraction of 1 ml containing the ring was collected as DRM fraction, and 1 ml from the bottom of the gradient was collected as DSM fraction.
Recovery of Proteins from the Sucrose Gradient-For mass spectrometry: Proteins from collected fractions were purified by methanolchloroform method. Initially, four volumes of methanol, one volume of chloroform, and three volumes of water were added to the sample, and after each addition, the sample was vortexed. Next, the sample was centrifuged for 15 min at room temperature. The upper phase was removed, and the lower phase was diluted with three volumes of FIG. 1. Protein fractionation and analysis scheme. During isolation of DRM other protein fractions were collected, including soluble proteins, internal membranes, and DSM, which were all subsequently analyzed by a combination of different biochemical and mass spectrometry methods.
methanol. Protein was then pelleted by centrifugation for 15 min at maximum speed (20,800 g).
For enzymatic measurements: The recovery of proteins from collected sucrose gradient fractions was based on dissolving the dense sucrose with large volumes of buffer TNE and subsequently pelleting membranes by ultracentrifugation at 200,000 g for 1 h.
In Solution Protein Digestion and Desalting-Purified proteins from preceding steps were dissolved in 6 M urea/2 M thiourea, pH 8, which causes denaturation of proteins. Subsequently, disulfide bonds were released by treatment with DTT, and spontaneous oxidation was prevented by carbamidomethylation with iodoacetamide. Lys-C and trypsin were applied for the digestion of proteins and creation of low-molecular-weight tryptic peptides as described (41). After trypsin digestion, samples were desalted on C18 Stop And Go Extraction tips (C18 Empore Disks, 3 M) (43), dried in a vacuum concentrator, and resuspended in 1% trifluoroacetic acid, 5% acetonitrile just before loading onto the liquid chromatography column.
LC-MS/MS Analysis of Peptides-Digested and desalted peptides were analyzed by LC/MS/MS using nanoflow HPLC (Easy nLC, Thermo Scientific) and an Orbitrap hybrid mass spectrometer (LTQ-Orbitrap XL, Thermo Scientific) as mass analyzer. Peptides were eluted from a 15 cm long, 75 m wide analytical column filled with C18 beads (Reprosil C18, Dr. Maisch GmbH, Germany) on a linear gradient running from 5% to 80% acetonitrile in 90 min at a flow rate of 250 nl/min and sprayed directly into the mass spectrometer. Proteins were identified by tandem mass spectrometry (MS/MS) by information-dependent acquisition of fragmentation spectra of multiplecharged peptides. Up to five data-dependent MS/MS spectra were acquired in the linear ion trap for each FTMS full scan spectrum acquired at 60,000 full-width half-maximum resolution settings with an overall cycle time of 1 s.
For raw file peak extraction and protein identification, we used MaxQuant software (44) (version 1.4.2.1) with A. thaliana TAIR10 protein sequences (35,386 entries) used by the search engine Andromeda (45). The following additional parameters were applied: trypsin as cleaving enzyme; 10 ppm peptide mass tolerance; 0.8 Da MS/MS tolerance; maximal two missed cleavages allowed, threshold for validation of peptides set to 0.01 using a decoy database, acetylation at N-terminal of proteins, and oxidation of methionine set as variable modification, carbamidomethylation of cysteine set as a fixed modification, "label-free quantification" marked, "match between runs" marked, minimum peptide length of six amino acids, and retention time correlation enabled (time window of 2 min). Spectra of identified peptides are available in supplemental Fig. 2. Known contaminants, such as keratins, were excluded from further analysis.
Label-Free Protein Intensities Calculation-Peptide lists derived from MaxQuant (evidence.txt) were processed by an R-based script (cRacker (46)) allowing for automated and standardized processing. Peptides with the assigned MaxQuant score below 30 and peptides that were quantified in less than 90% of all fraction-specific samples were filtered out. The principal steps of the normalization and quantitation were: (i) All peptide intensities within each sample were normalized to fraction of total ion-intensity sums. (ii) To compensate effects of normalization due to the number of detected peptides per sample, intensity was proportionally corrected to the reciprocal number of identified peptides in each sample using the integrated normalization option in cRacker. (iii) Peptide intensities with missing values in more than the permitted numbers of samples within replicates of each fraction were filtered out. (iv) Outliers exclusion was used to lower the influence of extreme peptide intensity values on the average protein intensity. (v) Remaining normalized peptide intensities were z-score scaled and mean averaged.
Protein Gel Blots-Protein concentration in different fractions after chloroform/methanol extraction was evaluated by NanoOrange (Invit-rogen) according to the instructions and equal amounts of proteins (10 g) per sample were loaded onto 10% polyacrylamide gels. SDS-polyacrylamide gel electrophoresis and protein gel blots were performed as described (47). The membrane was blocked in blocking solution overnight at 4°C and was subsequently washed three times with PBS-T (0.2% Tween-20, 10 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , and 2 mM KH 2 PO 4 , pH 7.4) and later incubated with the primary antibody ␣GFP at the concentration of 1:10,000 in PBS-T for 1 h at room temperature. After additional washing steps, the secondary antibody alkaline phosphatase coupled anti-mouse IgG was used at the concentration of 1:10,000 in PBS-T for 1 h at room temperature. For detection of the signal, blots were incubated in alkaline phosphatase-detection buffer (pH 9.5) with addition of 5-bromo-4-chloro-3Јindolyphosphate p-toluidine salt and nitro-blue tetrazolium chloride.
Extraction of Lipids-After density gradient centrifugation for the separation of DRM and DSM, a volume of 1 ml for the low-density band (DRM) as well as 1 ml from the bottom of the gradient (DSM) was transferred to the clean 15 ml falcon tube. For each 1 ml of sample, 5 l of 5 mM 3-hydroxy-5-cholestane (cholestanol) was added, which reflected 10 g of this compound and served as an internal standard. Lipids were extracted by the chloroform-methanol method (48). To each sample 3.75 ml of 1:2 (v/v) CHCl 3 :CH 3 OH were added, then 1.25 ml of CHCl 3, and finally 1.25 ml of water. Tubes were vortexed thoroughly after each solvent addition. The final mixture was centrifuged at 200 g for 5 min at room temperature, which resulted in a separation into upper aqueous and bottom organic phases. The aqueous solution was removed using a Pasteur pipette, and the organic phase was washed with "authentic upper phase" (prepared by performing blank chloroform-methanol extraction on water). The cleaned organic phase was subsequently transferred in batches of 500 l to a smaller tube and dried in a vacuum concentrator. By repeatedly drying down small volumes, the spread of extracted lipids around the whole inner surface of the tube was minimized.
Metabolite Profiling-The profiling of soluble metabolites was performed as described previously (49) by gas chromatography coupled to electron impact ionization/time-of-flight mass spectrometry (GC-EI/TOF-MS). Dried lipid species were methoxyaminated and trimethylsilylated manually prior to GC-EI/TOF-MS analysis. Analysis was performed using an Agilent 6890N24 gas chromatograph (Agilent Technologies, Boeblingen, Germany) with splitless injection onto a FactorFour VF-5ms capillary column, 30 m length, 0.25 mm inner diameter, 0.25 m film thickness (Varian-Agilent Technologies) connected to a Pegasus III time-of-flight mass spectrometer (LECO Instrumente GmbH, Moenchengladbach, Germany). Retention indices were calibrated by addition of a C 10 , C 12 , C 15 , C 18 , C 19 , C 22 , C 28 , C 32 , and C 36 n-alkane mixture to each sample (50).
GC-EI/TOF-MS chromatograms were acquired, visually controlled, baseline corrected, and exported in NetCDF file format using Chro-maTOF software (Version 4.22; LECO, St. Joseph, MI). GC-MS data processing into a standardized numerical data matrix and compound identification were performed using the TagFinder software (51,52). Compounds were identified by mass spectral and retention time index matching to the reference collection of the Golm metabolome database (http://gmd.mpimpgolm.mpg.de/) (53)(54)(55). Guidelines for manually supervised metabolite identification were the presence of at least three specific mass fragments per compound and a retention index deviation within 1.0%. All mass features of an experiment were normalized by sample fresh weight, internal standard (C 22 ), and were then maximum scaled.
Confocal Microscopy-Fluorescence imaging shown in the study was performed as described below. The fluorescent signal of GFP, YFP, and mCherry was recorded using Leica TCS-SP2 system coupled with DM6000 inverted light microscope and HCX PL APO 63.0 ϫ 1.20 W CORR objective. The excitation was achieved using 488 nm argon laser for GFP, 514 nm argon laser for YFP, and 561 nm diode laser for mCherry. Resulting emission was separated using triple dichroic TD 488/543/633 for GFP, dichroic RSP 500 for YFP, and double dichroic DD488/543 for mCherry. Separated light was collected and amplified using PMT detectors between 502-550 nm for GFP, 505-550 for YFP, and 577-650 nm for mCherry. Images were acquired with resolution of 1,024 ϫ 1,024 pixels. Pinhole was set between 0.295 and 1 Airy unit.
Statistical Analyses and Data Visualization-All statistical tests on proteomic data, such as hierarchical clustering and k-means clustering, were automated and performed by the R-based script cRacker (46). Information on subcellular location was taken from SUBA3 (56), and functional annotation was done according to MAPMAN (57). ANOVA analysis was used to test for significant differences between quantitative values on the protein level in the various samples and treatments. Differences with the p value Ͻ .05 in comparison to the control samples were taken as significant. Statistical assessments on sterol and metabolite data, such as hierarchical cluster analysis, heat map, and other data visualizations, were performed with the multiexperiment viewer software MeV (Version 4.6.2) (58, 59).

RESULTS
The aim of this study was to characterize effects of cytoskeletal disruption on protein distribution between low-density sterol-rich microdomains (as biochemically extracted in the form of DRM) and general other higher-density membrane phases (DSM).
Cytoskeletal Disruption Does Not Affect Sterol Content of the Membrane-Two major phytosterols, sitosterol and campesterol, were identified and quantified in DRM and DSM fractions, with generally higher values in DRM fractions. Generally, these sterols contribute to about 20% of the plasma membrane lipids. No other typical membrane sterols were identified but rather organic acids, sugars, alkanes, and many unidentified compounds containing an alkyl group were characterized in the GC-TOF analysis. In both fractions, a variety of soluble compounds like ethyl-hexanoic acid, hydroxy-pyridine, octadecane, diethylene glycol, boric acid, or phosphoric acid were detected ( Fig. 2A). These other compounds may have masked the detection of other lower abundant lipids. Most of these substances, except for ethanolamine, were considered as copurifying contaminants since they were not known to be membrane components.
The cytoskeletal disruption did not result in statistically significant relative and absolute changes in any of the identified substances. Both sterols grouped in the same hierarchical cluster of compounds. The level of sterols in DRM as well as in DSM fractions remained unaltered in response to cytochalasin D and oryzalin treatment, and as expected, sterol levels were higher on average in DRM fractions (Fig. 2B). Thus, we conclude that disruption of actin filaments or disruption of microtubules does not cause a drastic change in membrane sterol composition.
General Proteomic Comparison of Collected Fractions-Detergent-resistant membrane preparations are prone to copurifying nonsterol-dependent proteins and may be considered an enrichment but not a purification of sterol-rich membrane microdomains (60). In order to robustly establish a filter for identification of true effects of cytoskeletal disruption on changes in protein distribution between cellular membrane phases, reference protein lists of know sterol-dependent proteins, as well as from intracellular membrane proteins and soluble proteins, were established (60). In the cytochalasin D experimental series, 557 proteins were identified in DRM fractions (five replicates), a total of 918 proteins were identified in DSM fractions (four replicates), and 588 and 1,558 proteins were identified from intracellular membranes (IM, two replicates) and as soluble proteins (SOL, four replicates). For oryzalin treatment experiments, 449 proteins were identified in detergent-resistant membranes (five replicates), 996 proteins in detergent-soluble membranes (five replicates), 1,313 proteins in intracellular membranes (four replicates), and 776 proteins as soluble proteins from two biological replicates (supplemental Fig. 2A). Separation quality of cellular fractions was confirmed by the fact that in almost all fractions collected, a large number of proteins was identified in a single fraction only (unique proteins). The largest overlap between collected fractions was represented by the proteins from neighboring compartments, for example, 154 or 204 proteins overlapping between DRM and DSM fractions while only 37 or one protein was in common for detergent-resistant membranes and soluble fraction (supplemental Fig. 3A).
The identified proteins were then assigned to expected cellular location based on SUBA3 (56). As expected, in both drug treatment series, the majority of proteins identified in the DRM fraction had a plasma membrane location, while proteins identified in the IM fraction largely were of mitochondrial, plastidial, and endoplasmic reticulum location. The SOL fraction contained the highest amount of soluble proteins (supplemental Fig. 3B). Furthermore, filtering out all proteins that were found in more than one fraction enriched the fractionspecific protein groups at an even higher degree (unique proteins in supplemental Fig. 3B). In further analyses, proteins were excluded if their abundance was highest in IM or SOL fractions, even when detected in DRM or DSF fractions. This filtering procedure was validated in detail in a previous study (60).

Effects of Cytoskeletal Disruption on Membrane Protein
Distributions-To confirm the specificity of cytochalasin D and oryzalin treatment on the disruption of actin filaments and tubulins, respectively, the disappearance of actin or tubulin protein in the DRM fraction was monitored (Fig. 3A). Cytochalasin D treatment depleted the sample from ACT3 within 30 min of treatment, but a less strong effect was observed for ACT2. Oryzalin treatment efficiently depleted different tubulin isoforms from the DRM fraction in a concentration-dependent manner.
Secondly, specificity of the drug treatment was monitored using the dual-labeled line for actin and tubulin (Fig. 3B). Cytochalasin D treatment of 1 h was already enough to significantly disrupt the majority of actin filaments, but some actin strands still remained visible. Microtubule structures were not affected by the cytochalasin D treatment. Oryzalin treatment at a concentration of 5 M for 1 h completely disintegrated typical microtubule structures, while actin filaments remained intact. Thus, the drug treatments applied in our experiments specifically affected only one component of the cytoskeleton.
Cluster analysis of the abundance profiles of DRM proteins in response to cytochalasin D time course treatment revealed five major response groups (supplemental Fig. 4A). The largest number of proteins (199 proteins) was found in cluster 3, with a tendency for an increase of abundance in response to the treatment. However, these were mainly cytosolic proteins (66%) and very few plasma membrane proteins (6%). The cluster with the highest fraction of plasma membrane proteins (46%) was cluster 1 (144 proteins), with a tendency of decreased abundance in the DRM fraction. Clusters with a high content of cytosolic proteins (clusters 2 and 3) particularly contained proteins involved in protein synthesis (ribosomal proteins). Such proteins are known copurifying proteins in DRM preparations (60). In contrast, clusters 1, 4, and 5, with high fraction of membrane proteins particularly, were rich in proteins with transport and signaling functions (these functions made up 18 and 15% of all proteins in cluster 1).
Similarly, oryzalin treatment over the range of concentrations resulted in differential responses of proteins, which could be separated into five response clusters (supplemental Fig. 4B) containing 51 (cluster 4) to 113 proteins (cluster 2). Proteins in cluster 1 showed an increased abundance with increased concentration of oryzalin, but these were particularly proteins with cytosolic location such as ribosomal proteins. Proteins in two clusters with a high fraction of membrane proteins (62% in cluster 2 and 67% in cluster 4) displayed a decreased abundance profile in DRM with increasing oryzalin concentration but at different slope. These clusters were particularly enriched in proteins with signaling function (3% in cluster 2) or transporters (40% in cluster 4). However, cytochalasin D or oryzalin treatment of cell cultures did not affect global protein abundance levels and distribution of proteins identified in the IM and SOL fractions (supplemental Fig. 5). These findings suggest that the observed drug-induced protein abundance changes for proteins identified in DRM and DSM fractions were mainly due to local redistributions within the membrane and not due to global redistributions between compartments based on disturbed membrane trafficking.
In order to identify candidate DRM proteins significantly affected by the cytoskeletal disruptions, three filtering criteria were applied to the above data set: (i) proteins were required to be identified in DRM fractions in at least one replicate of this study, (ii) proteins were required to show significant response to drug treatment (ANOVA p value cutoff .05 after multiple testing (61)), and (iii) have evidence for sterol-dependent membrane location from previous work (22,60,62). Based on these criteria, for cytochalasin D treatment, from the 557 proteins identified in DRM fractions, only 50 proteins had previously been confirmed also as sterol-dependent proteins (Fig. 4). For oryzalin treatment, out of 449 proteins identified in DRM fractions, 66 proteins were also characterized as steroldependent. From this overlap, 22 proteins were found to significantly display an altered abundance distribution between DRM and DSF upon cytochalasin D treatment, and 13 proteins were found to significantly change their distribution profile in response to oryzalin. In most cases, cytoskeleton disruption, particularly treatment with cytochalasin D, induced a reduced abundance of proteins in sterol-enriched DRM fractions with an accompanied increased abundance of that protein in the sterol-depleted DSM fraction (DRM/DSM ratios in Table I). This may indicate that these proteins within the membrane moved from sterol-enriched phases into steroldepleted membrane phases, particularly after disruption of actin filaments.
Out of the 22 proteins with significantly altered distribution between DRM and DSM upon cytochalasin D treatment (Table I; supplemental Fig. 6A), there were eight proteins involved in membrane transport, four signaling proteins, two oxidases, two proteins with functions in stress responses, a remorin protein, and few others with unassigned functions. Among the 13 proteins with significant response to oryzalin, we found seven transporters, three signaling proteins, a remorin, one protein involved in abiotic stress responses, and one protein with unspecified role in plant development (Table I; supplemental Fig. 6B). Among all these proteins, nine proteins showed altered distribution between DRM and DSM upon both treatments, namely disruption of actin filaments as well as upon disruption of microtubules (supplemental Fig. 7). Thereby, six proteins (AHA1, AHA2, ERD4, RLK902, AT3G61260, AT3G02880) were depleted from DRM in both drugs, one protein (PATL1) increased in DRM fractions upon both drugs, and two proteins (AVP1 and VHA-A3) showed opposite behavior upon actin or microtubule disruption.  (Table I), we identified members of the H ϩ -ATPase family (AHA1, AHA2, AHA11). To study the biochemical effect of altered membrane phase recruitment, enzymatic activity assays were carried out in control and drug-treated cells. Thereby, the activity measurements of the sterol-rich DRM fractions were compared with activity measurements of total plasma membrane to relate differences in activity to differences in submembrane location (Fig. 5). Cytochalasin D treatment lead to a depletion of all three isoforms of the H ϩ -ATPase from the sterol-rich DRM preparations (Fig. 5A), and this was accompanied with an increase in the measured ATPase activity in that fraction, especially in cytochalasin D treatments of 1 h (Fig. 5B). Activity measured in total plasma membrane remained unaltered upon the disruption of actin filaments and was generally lower than in sterol-rich DRM fractions. These findings suggest that actin filament disruption leads to recruitment of H ϩ -ATPase molecules from more rigid DRM fractions to DSM (supplemental Fig. 6A, Table I), and this relocation is associated with an increased activity of the protein remaining in DRM. In contrast, oryzalin treatment also removed ATPase isoforms from the DRM fraction (Fig. 5C), but no accompanied increase in DSM was observed (supplemental Fig. 6B, Table I). The disruption of microtubules with oryzalin lead to a decreased H ϩ -ATPase activity in the DRM fraction and increased activity levels in the whole plasma membrane (Fig. 5D). It seems that microtubule disruption removed the plasma membrane ATPases from DRM fractions without associated relocation to DSM. In general, based on the enzyme activity measurements in combination with the resulting abundance distribution between DRM and DSM (Table I), we conclude that alterations in the location of proteins within membrane phases may signif-  icantly affect their activity either through changes in abundance or by creating activating/inactivating interaction environment.
Cytoskeletal Disruption Affects Location of a Receptor Kinase Within the Plasma Membrane-The receptor kinase AT3G02880 was found to be depleted from DRM upon treatment with cytochalasin D as well as upon treatment with oryzalin (supplemental Fig. 8A, Table I). The effects of drug treatment on location of AT3G02880 were analyzed using Arabidopsis seedlings expression a GFP-fusion under the CaMV35S promoter (supplemental Fig. 8B). In untreated seedlings, AT3G02880 showed more or less uniform distribution across the plasma membrane. Interestingly, the uniform distribution of fluorescent signal was exempt in line-shaped regions, resembling the cellular pattern of microtubule strands. We suggest that these exempt darker areas follow the distribution of microtubules below the membrane bilayer. This statement is supported by the finding that we did not record such distribution in oryzalin-treated seedlings, whereas it was still visible after treatment with cytochalasin D. Occasionally, locations in punctate patterns (arrows in supplemental Fig. 8C) were visible.
Treatment of cells with cytochalasin D did not lead to large overall changes in location pattern of AT3G02880 but enhanced the appearance of the punctate location. Treatment with oryzalin resulted in a more uniform pattern, losing the cytoskeleton negative print in the location pattern and resulting in smaller punctate structures.
Cytoskeletal Disruption Affects Location of Remorins-Remorins are plasma membrane proteins that were shown to

Summary of proteins with increased or decreases abundance in DRM fractions upon cytoskeletal disruption with cytochalasin D or oryzalin
locate to sterol-rich membrane microdomains using cell biological and biochemical approaches (9,63), and they are highly enriched in DRM preparation (22,60,63). In our experiments, cytoskeletal disruptions led to altered distributions of remorin AtREM1.2 between DRM and DSM (Fig. 6A). These findings from the large-scale proteomic approach were confirmed by protein gel blot analysis of four fractions across the sucrose gradient separating sterol-rich DRM from sterol-depleted DSM (Fig. 6B). Particularly, when actin filaments were disintegrated by cytochalasin D treatment, remorin abundance clearly decreased in DRM fractions and increased in DSM. Disruption of microtubules by oryzalin also led to a depletion of remorins AtREM1.2 and AtREM1.3 from DRM, but the accompanied increase in DSM was less prominent compared with the observations in cytochalasin D treatment (Fig. 6C). Location of AtREM1.2 in untreated Arabidopsis seedlings was found in punctate patterns, as observed previously (9,63). Treatment with cytochalasin D completely disrupted this punctate location pattern to a more uniform distribution with smaller punctate patterns of lower intensity. In contrast, treatment with oryzalin enhanced punctate location of AtREM1.2 to bigger dots of higher intensity.

DISCUSSION
This work aimed at a systematic analysis of the role of cytoskeletal components for sterol-rich membrane phase composition based on treatments with actin-or tubulin-disrupting agents. We used a proteomic screening approach to identify cytoskeleton-dependent proteins among established microdomain proteins. Enzyme assays and microscopy were then used to relate the biochemical findings to protein activities and location.
As an important prerequisite for the experimental design and the conclusions drawn, using the actin and microtubule dual-labeled line, we confirmed that the conditions used for the treatments affect only the respective cytoskeletal component and not the other. Thus, we may be able to differentiate between specific role of actin filaments or microtubules for membrane phase composition in plants.
Cytoskeleton and Protein Trafficking-All the plasma membrane constituents undergo a mechanism called secretion by which proteins, lipids, and other molecules are trafficked and delivered to their final destinations. Usual migration of proteins starts in the endoplasmic reticulum (ER) where they are synthesized. Then they are moved into the Golgi-apparatus where they obtain modifications (reviewed in (64)  proteins occurs at the trans-Golgi network from where the carriers are being transported to the targeted compartment.
It is likely that because of the close proximity, the ER and Golgi organelles do not require cytoskeletal elements to exchange carrier vesicles with each other. This was clearly proven by fluorescence recovery in the Golgi after photobleaching of fluorescent protein fusions of Golgi-localized proteins in the presence of the trafficking inhibitor brefeldin A (65,66), in the presence of microtubule-disrupting agents (65,66) and in cells treated with actin-disrupting agents (67). Thus, short distance membrane flow in the early secretory pathway is not impaired in the absence of functional cytoskeletal components. The secretory route from ER and trans-Golgi network to the destination membranes has not been revealed in detail. It is also unknown if trafficking to the plasma membrane is a direct process or if it includes some intermediate compartments endosomal compartments (68,69). Although there are indications that the trans-Golgi network to plasma membrane trafficking involves exocyst complex (70), exosomes, and probably multivesicular bodies (71), it still needs to be revealed if the cytoskeleton is necessary for this process.
In contrast, participation of microtubules and actin filaments in endocytosis from the plasma membrane is quite well documented (72). Actin may create aided invaginations by specific protein-protein interactions (73)(74)(75). Importance of actin as part of the endocytotic machinery was also evaluated with the use of chemical inhibitors of actin polymerization (76). Also, microtubules were proven to play important role in both clathrin-and nonclathrin-mediated endocytosis (77).
A question that arises while analyzing the data set obtained here is whether the observed effects on the composition of detergent-resistant membrane fractions could be a result of a disturbed protein and membrane trafficking from/to the ER and Golgi apparatus. If the common trafficking pathways to the plasma membrane were disturbed, we would expect to observe some large global changes in the composition of different intracellular fractions (IM, SOL; see Fig. 1). For example, if vesicles carrying cargo destined to the plasma membrane or extracellular matrix are trapped within the cytoplasm, endoplasmic reticulum, Golgi apparatus or vacuole, a drop in protein abundance should be observed in target compartments, and an increased abundance is expected either in intracellular membrane fraction (IM) or in fraction of soluble proteins (SOL). However, the cytoskeletal disruption by cytochalasin D did not induce major global changes in protein abundances of proteins identified in the IM and SOL fractions (supplemental Fig. 5). This may indicate that observed changes in the abundance of specific proteins in the detergent-resistant membrane fraction was likely not a result of generally disturbed membrane trafficking or cell metabolism but due to rearrangements within the plasma membrane phases. Oryzalin treatment induced a large scatter of abundances of proteins identified in IM fractions, indicating a pos-sible contribution of tubulins in protein trafficking from/to DRM fractions.
Proteins Significantly Affected by Cytoskeletal Disruption-By application of the filtering criteria, 22 proteins previously defined as sterol dependent showed a significant relocation in response to cytochalasin D and 13 proteins rearranged in response to oryzalin and nine proteins showed a significant response to both actin and microtubule disruption. These nine proteins were also previously identified as proteins with stimulus-induced abundance changes in DRM fractions either in response to flagellin (10) or during cold acclimation (78).
We observed that in response to cytochalasin D (actin filament disruption), 86% of the proteins were reduced in the DRM fraction, often displaying an associate increased abundance in the DSM fraction (Table I). In case of oryzalin treatment (microtubule disruption), a reduced abundance in DRM was observed only in 54% of the cases, and often there was no associated increase in protein abundance in DSM fractions (Table I). Thus, disruption of actin and tubulin resulted in differential distribution between the DRM and DSM fractions, although in both cases a depletion of DRM fractions was observed. AVP1 (AT1G15690) is a type I H ϩ -pyrophosphatase and observed with depletion from DRM and associated enrichment in DSM in response to cytochalasin D treatment and enrichment in DRM fractions in response to oryzalin treatment. Activity of AVP1 depends on cytosolic level of potassium and is moderately sensitive to inhibition by calcium (79). Overexpression of AVP1 also enhances nutrient uptake by affecting the abundance and activity of the plasma membrane H ϩ -ATPase (80). Initially, AVP1 was considered to be a bona fide vacuolar marker (79,81). However, recent studies with the use of double epifluorescence and immunogold labeling experiments showed the colocalization of AVP1 and plasma membrane located aquaporin PIP1 in plasma membrane from Arabidopsis (82). In the protein-protein interaction network of the proteins with altered DRM abundance upon cytoskeletal disruption, AVP1 seems to take a hub position connecting cytoskeletal elements with other ATPases such as AHA1, AHA2, AHA11, or VHA-A3 (supplemental Fig. 7). AHA1 (AT2G18960) and AHA2 (AT4G30190) are the two most dominant plasma membrane proton pumps that generate proton gradient across the membrane in an ATP-dependent manner. Their activity results in creation of proton motive force that is used by other membrane components as a source of energy for their functions. Phosphorylation was found to be crucial mechanism for regulation of both AHA1 and AHA2 (40,(83)(84)(85). AHA1 and AHA2 were both identified as DRM constituents in several proteomic analyses (62,78) and were additionally found as significantly enriched in DRM fractions after flg22 treatment (10). Cytoskeletal disruption lead to a depletion of AHA1 and AHA2 from DRM fractions but enzymatic assays revealed opposite effects of relocation within the membrane on H ϩ -ATPase activity.
Remorins-The remorin AtREM1.2 (AT3G61260) was identified as significantly responsive to actin as well as microtubule disruption. In both cases, a decreased abundance in DRM fraction was observed, but only when actin filaments were disrupted, REM1.2 showed an accompanied increased abundance in DSM. Remorins are considered as membrane microdomain markers, and they were repeatedly identified in DRM fractions by proteomic analyses (10,22,60,62,78). Immunolocalization, as well as GFP-tagging, revealed that remorins from group 1b (86) to which AtREM1.2 belongs, accumulate in distinct patches within plasma membrane. Also, in Medicago truncatula, MtSYMREM1 localizes in immobile domains in the plasma membrane of root hair cells (87). Here, we generally confirmed the patchy membrane location of AtREM1.2. Recent evidence, however, indicates that remorin location is quite complex and can exhibit isoform-specific patterns restricted to organs or developmental stages (9). The patchy location of AtREM1.2 in hypocotyls was abolished upon actin filament disruption, leading to uniform fluorescence with exception of the regions where microtubule strands run below the membrane bilayer.
We conclude that intact actin filaments are necessary for AtREM1.2 location to the patchy structures of membrane microdomains. Actin-dependent patchy membrane location has been described in mammalian cells also for acetylcholine receptor (in the membrane of Chinese hamster ovary that was lost when intracellular components including cytoskeleton were removed) (88).
Enhancement of patchy AtREM1.2 patterns was observed upon disruption of microtubules (Fig. 6), but the distance between the observed patches became larger. It therefore seems that disturbance of microtubules resulted in relocalization of remorin containing membrane phases or microdomains within the plasma membrane.
Contrasting Effects of Actin and Tubulin Disruption-Disruption of actin filaments or disruption of microtubules structures in many cases resulted in differential responses of the respective proteins. One general finding was that, although both drug treatments resulted in depletion of candidate proteins from DRM fractions, only for cytochalasin D treatment a respective increase in abundance of these proteins in the DSM was observed (supplemental Table 1). These observations were made initially by proteomic analysis of cell suspension cultures and then confirmed using microscopy on young seedlings. We consider the results from both systems as similar and comparable since general membrane and cytoskeleton interaction effects were studied and no specific responses in context of differentiation and cell contact. However, the cytoskeleton may be particularly important in inducing polar protein location in differentiated cells.
Treatment of the cells and seedlings with oryzalin also resulted in decreased abundance of proteins in DRM frac-tions, but the expected increased abundance in DSM was not measured (supplemental Table 2). These opposite effects of actin and microtubule disruption were confirmed by activity measurements for H ϩ -ATPase isoforms (Fig. 5), for location of receptor kinase AT3G02880 (supplemental Fig. 8), and for remorin AtREM1.2 (Fig. 6). The effects of actin disruption on remorin protein membrane location were reproduced using latrunculin B as another actin-disrupting agent (supplemental Fig. 9). Thereby, disruption of actin as well as the disappearance of the spotty localization pattern for remorin were similar as under cytochalasin D treatment.
In general, it needs to be kept in mind that the plasma membrane is a very dynamic structure. There is evidence that in plants multiple types of membrane domains with distinct composition coexist at the same time (9), and some of these could be formed in a stimulus-dependent manner or when the specific changes in lipid or protein composition occurred (87). Thus, biochemical approaches as carried out here by using changes in composition of a single biochemical fraction (here DRM) and the distribution of proteins between fractions (DRM and DSM) as readout for dynamic relocation of proteins within the membrane will not be able to distinguish these different specific microdomains observed in vivo. However, the approach validated identification of candidate proteins that undergo dynamic submembrane locations. CONCLUSIONS Various proteins found with altered distributions between DRM and DSM upon cytoskeletal disruption could be confirmed through resulting differences in activity or location patterns. Based on our results, we suggest that actin filaments are necessary for dynamic movement of proteins between different membrane phases. The effects of actin were confirmed by similar location patterns using two different actin-disrupting agents (latrunculin B and cytochalasin D). We further suggest that microtubules are not necessarily important for formation of microdomains as such, but they control the density and size of these structures (supplemental Fig.  10). Thus, if density and size of membrane microdomains changed within a cell (as, for example, observed for remorin), this could also result in an overall lower abundance of these microdomain proteins in membrane preparations and in DRM fractions. Our suggestion that microtubules may be responsible for control of size and density of membrane microdomains is further supported by the fact that more proteins with increased DRM/DSM ratio were found upon microtubule disruption treatment compared with treatment with actin-disrupting agents.