Furrow-like invaginations of the yeast plasma membrane correspond to membrane compartment of Can1

The plasma membrane of living cells is considered to be laterally compartmented into domains of specialized composition and function. In the yeast Saccharomyces cerevisiae, two such non-overlapping membrane compartments (MCs) have been distinguished. The first, MCC, contains the arginine permease Can1 and the second, MCP, contains the proton ATPase Pma1 (Malinska et al., 2003). MCC consists of evenly distributed isolated patches. Based on confocal localizations of Can1-GFP and Pma1-GFP, the size of the MCC patches was estimated to be ∼300 nm. To date, more than 20 cortical proteins have been found to colocalize with the MCC, nine of which are integral to the plasma membrane (Grossmann et al., 2008; Maier et al., 2008; Deng et al., 2009). In addition to Can1, three other permeases with targeting pathways that are dependent on specific lipids, Fur4, Tat2 and heterologous HUP1 (Chlorella kessleri), were identified as MCC constituents (Malinska et al., 2004; Grossmann et al., 2006; Grossmann et al., 2007). The compartmentation of the plasma membrane into MCC and MCP is highly stable (Malinska et al., 2004), but the transporters dock within MCC patches in a reversible, membrane-potential-dependent manner. In addition to its specific protein composition, filipin-stained plasma membrane sterols were shown to accumulate in MCC (Grossmann et al., 2007).

The biological function of MCC was unclear until a delayed internalization of MCC-accumulated Can1 was reported recently, suggesting a protective function of MCC patches in protein turnover (Grossmann et al., 2008). Among soluble proteins sharing the MCC distribution, Pil1 and Lsp1, long-chain-base-responsive inhibitors of protein kinases Pkh1 and Pkh2 (Zhang et al., 2004), were postulated to form eisosomes, which are organelles with a proposed role in endocytosis (Walther et al., 2006). MCC markers show a homogenous distribution in the plasma membrane of pil1Δ cells, with rare enlarged spotted accumulations. Electron microscopy analysis of these cells suggested that these spots correspond to large aberrant plasma membrane infoldings (Walther et al., 2006).

In wild-type budding yeast, two specialized structures containing invaginated plasma membrane are described: one usually referred to as `finger-like' and the other as `furrow-like'. It is believed that the cortical patches of fibrous actin form around finger-like plasma membrane invaginations, 150- to 250-nm-deep tubular structures with a diameter of ∼50 nm (Mulholland et al., 1994) that concentrate in areas of polarized growth. The formation of a finger-like invagination in relation to the endocytic event has been explained in molecular detail by fluorescence (Kaksonen et al., 2005) and electron microscopy (Idrissi et al., 2008) studies. Actin patches and/or finger-like invaginations were also suggested to be places of cell wall synthesis (Kopecka and Gabriel, 1995).

By contrast, furrow-like invaginations are rather randomly distributed over the cell surface. On replicas of freeze-fractured cells they appear as straight linear plasma membrane depressions ∼300 nm long (Moor and Mühlethaler, 1963; Gross et al., 1978). Although various examples of similarly shaped plasma membrane structures in other organisms have been reported (see Discussion), the function of furrow-like invaginations remains unclear.

In this study, we present evidence that, in the plasma membrane of S. cerevisiae, furrow-like invaginations correspond to MCC patches. A genome-wide screen identified Nce102, a protein of unknown function integral to MCC, and Pil1, a primary component of eisosomes, as the main organizers of MCC composition and structure (Grossmann et al., 2008). Using various electron microscopy approaches, we show the absence of furrow-like invaginations in nce102Δ and pil1Δ deletion mutants. In addition, we report an altered morphology of MCC patches and the corresponding changes in furrow-like invaginations in two other deletion mutants, ypr050cΔ and mak3Δ. Finally, using immunolocalization of nanogold probes, Pil1 and an MCC marker Sur7 (Malinska et al., 2004), a protein of unknown function integral to the plasma membrane (Young et al., 2002), were shown to localize to the furrow-like plasma membrane invaginations. For the first time, we assign a lateral plasma membrane domain to a specific membrane structure.

In an attempt to describe the fine structure of the MCC patch, we performed an electron microscopy analysis of yeast mutants showing aberrant MCC distribution (Grossmann et al., 2008). First, we tested plasma membranes of pil1Δ and nce102Δ cells, exhibiting homogenously distributed MCC markers (Fig. 1D,G; compare with the wild type in Fig. 1A), for possible structural alterations.

Two different approaches were used for visualization of the plasma membrane. First, replicas of freeze-etched pil1Δ and nce102Δ mutants were prepared and compared with those of the wild type. In accordance with the reports published previously (e.g. Moor and Mühlethaler, 1963; Steere et al., 1980), two types of characteristic structures were recognized on the inner leaflet (P face) (Branton et al., 1975) of the plasma membrane in wild-type cells: (1) areas with a hexagonally ordered repetitive grain pattern, and (2) islets of smooth membrane with straight lines of furrow-like invaginations (Fig. 1B,C). The most striking difference observed in the plasma membrane of mutant cells was the absence (or far lower frequency) of furrow-like membrane invaginations (Fig. 1E,F,H,I). We found 2.5±0.2 invaginations per μm² on the surface of wild-type cells. This surface density was decreased to ∼20% (0.5±0.2 invaginations per μm²) in pil1Δ cells, and only rare invaginations were detected in the plasma membrane of nce102Δ cells.

Similarly, the furrow-like invaginations could not be detected on ultrathin resin sections of pil1Δ and nce102Δ cells. Curved membrane areas that corresponded to invaginations 200-300 nm long and ∼50 nm deep were regularly found in wild-type cells (Fig. 2). As reported previously, these structures were often accompanied by electron-lucent granules of glycogen (Coulary et al., 2001). By contrast, only a planar membrane bilayer without any invaginations was observed in both the mutants (data not shown). The occasional occurrence of large membrane-surrounded structures in pil1Δ cells (supplementary material Fig. S1) obviously corresponded to the large aberrant fluorescent patches observed by Walther and co-workers (Walther et al., 2006).

Next, we determined whether there was a direct relationship between the furrow-like invaginations and MCC patches in the plasma membrane of wild-type cells. Bearing in mind the resolution limit of fluorescence microscopy, we aimed to compare surface densities of these two structures. First, we counted MCC patches. Analysis of 3D stacks of confocal sections revealed a surface density of 1.2±0.2 patches per μm² in wild-type cells expressing the MCC marker Sur7-GFP.

Next, we performed a computer simulation of the furrow-like invagination pattern blurred by confocal imaging. We assumed invaginations of a uniform length of 250 nm to be marked by a specific fluorescent marker and spread randomly (Moreira et al., 2009) on a plane (Fig. 3A) with an overall density corresponding to the value measured on freeze-fractured cells (2.5 invaginations per μm², see above). The confocal imaging of this model set of microscopic objects was simulated by its convolution with the 2D point spread function (PSF) of the confocal microscope (Fig. 3B; see Materials and Methods for details). Individual foci were counted in the resulting `confocal image'. In ten images consisting of 1000 model invaginations each, we measured the density of 1.72±0.05 foci per μm², which was comparable with the measured surface density of MCC patches.

We searched for further morphological evidence that the furrow-like invaginations of the plasma membrane represent the ultrastructural correlates of MCC patches. As was apparent from thin sections of high-pressure-frozen cells, cisternae of cortical ER appeared to be next to, but not beneath these invaginations (Fig. 4A). To test the mutual position of cortical ER and MCC patterns, we coexpressed an MCC marker Sur7-mRFP with an ER reporter construct, consisting of GFP fused to a signal sequence for ER localization and the ER retrieval sequence His-Asp-Glu-Leu (ss-GFP-HDEL). A clear separation of both the compartments was detected (Fig. 4B-D). Either the MCC patches were localized in the ER-free zones of the plasma membrane, or a local minimum of ER-specific signal could be detected underneath, indicating that there are holes in the flat cisternae of cortical ER.

As a consequence of previous findings, we checked the plasma membrane ultrastructure in the yeast strain lacking YPR050C, which has been described as a “dubious ORF unlikely to encode a protein” (Fisk et al., 2006). During the evaluation of a visual screen focused on strains with alterations in MCC integrity (Grossmann et al., 2008), we found that ypr050cΔ cells show abnormally elongated MCC patches (Fig. 5A,B). Although the appearance of membrane invaginations remained unchanged on transversal ultrathin sections of these cells (data not shown), sections almost parallel to the cell surface (tangential sections), as well as the freeze-fracture cell surface replicas, revealed abnormally elongated furrow-like invaginations (Fig. 5C,D, respectively). Cells lacking MAK3 (YPR051W) showed the same phenotype (data not shown), most probably because of a significant overlap of MAK3 with YPR050C (Murthi and Hopper, 2005). We conclude that individual furrow-like invaginations correspond to the MCC patches.

To confirm the above conclusion, we decided to investigate the localization of protein components of MCC patches, as well as cytosolic eisosomes, at the ultrastructural level. Using several labeling protocols, we performed immunogold detection of the MCC and the eisosome markers Sur7 and Pil1.

When tracked by pre-embedding immunogold labeling, including the extractive methanol permeabilization step, Sur7-GFP was localized unequivocally to areas of curved plasma membrane, which corresponded to the furrow-like invaginations (Fig. 6). The labeling signal was specific, confined exclusively to the plasma membrane without a detectable background. Pil1-GFP showed a similar subcellular distribution. In addition, some protein localized to the cell interior, corresponding to a cytoplasmic pool of Pil1 (data not shown).

Post-embedding `on-section' labeling assuring better fine structure preservation revealed comparable results. In a first attempt using mild chemical fixation and LR White embedding (Mulholland et al., 1994), Sur7-GFP was localized to the plasma membrane invaginations. However, the fine structure of the plasma membrane was altered in these cells, as judged from the artificially tilted and profound invaginations (data not shown).

For unknown reasons, we did not obtain reproducible results using cryosectioning approach (Tokuyasu, 1980). The most reliable method for ultrastructural characterization of the protein composition of plasma membrane invaginations thus seems to be the approach combining high-pressure freezing (HPF), cryosubstitution (freeze substitution or FS) and low-temperature embedding. The use of modified substitution medium (see Materials and Methods for details) led to a poor membrane contrast in these samples. Nevertheless, using the HPF-FS approach, we were able to confirm the Sur7-GFP and Pil1-GFP localization observed on chemically fixed cells, but in the context of a well-preserved ultrastructure (Fig. 7A). In addition, a slight difference could be distinguished between localization of the two proteins; although Pil1 regularly localized along deeper parts of the furrow-like invagination, mainly at its negatively curved bottom (Fig. 7B-D), similar distribution of Sur7 was rare (Fig. 8A, top panel). Sur7 was rather detected at the most superficial parts of the invaginated plasma membrane, at the positively curved rim of this labeled structure (Fig. 8A, mid and bottom panels). Decoration of individual furrows could be followed on superficial, nearly tangential sections (Fig. 8B).

Phylogenetically, furrow-like plasma membrane invaginations represent a highly conserved structure. Grooves that are almost identical to those observed in the plasma membrane of S. cerevisiae have been found in several other yeast species, such as Candida albicans (Barug and de Groot, 1985) and Schizosaccharomyces pombe (Takeo, 1984), in other fungi, such as Lichinella stipatula (Büdel and Rhiel, 1987) and Spotothrix schenckii (Svoboda and Trujillo-Gonzalez, 1990), in bacteria, such as Micrococcus (Sleytr and Kocur, 1971; Sleytr et al., 1976) and in green algae, including several Chlamydomonas species (Clarke and Leeson, 1985). Similar structures were also observed in higher plants, such as Plumbago zeylanica (Southworth et al., 1997). We identify these plasma membrane invaginations with the patches of MCC observed in S. cerevisiae by fluorescence microscopy (Malinska et al., 2003). Several pieces of evidence are presented, which allow us to draw this conclusion.

First, furrow-like invaginations are absent in the mutants defective in MCC patch integrity. Drastically reduced numbers of invaginations were observed in the freeze-fractured plasma membrane of pil1Δ cells, which show only few aberrant accumulations of fluorescent MCC markers. In the membrane of nce102Δ cells with homogenous distribution of MCC transporters, but patchy appearance of Sur7 and Pil1 (Grossmann et al., 2008), flat, smooth, elongated lateral domains, but no invaginations were detected (Fig. 1H,I).

Next, the distribution and surface density of both structures were comparable. Similarly to the MCC patches (Moreira et al., 2009), furrow-like invaginations also seem to be randomly distributed in the membrane (Moor and Mühlethaller, 1963; Gross et al., 1978) and are most abundant in old mother cells (Takeo, 1984). In accordance with the MCC appearance on buds (Grossmann et al., 2008), few invaginations were found on dynamically developing membranes, such as young buds of S. cerevisiae and the central division ring in S. pombe (Takeo, 1984) or the germ tube of C. albicans (Miragall et al., 1986). MCC patches also do not colocalize with the cortical ER (Fig. 4). Direct counting revealed a ∼twofold higher density of invaginations compared with MCC patches. As a number of cortical proteins (Young et al., 2002; Malinska et al., 2004; Walther et al., 2006; Grossmann et al., 2008) and ergosterol (Grossmann et al., 2007) were colocalized with Sur7 in MCC patches by fluorescence microscopy, it is not likely that only some invaginations contain Sur7 and thus correspond to MCC patches. The Sur7-negative invaginations would then be structurally indistinguishable from MCC patches, but would have completely different protein and lipid composition. Instead, we show by computer simulation, that the observed discrepancy between the two densities can be entirely ascribed to the resolution limit of fluorescence microscopy: even if distributed in a plane perpendicular to the optical axis of the microscope, the neighboring invaginations were often fused in the simulated fluorescence microscopy image (Fig. 3). On a round cell surface, this effect is further increased owing to the poor axial resolution of the microscope.

Abnormally elongated invaginations were observed in protoplasts (Necas et al., 1969). Freeze-fracture data suggested that grooves 1-2 μm long or even longer arose by longitudinal fusion of short invaginations during the protoplast formation (Miragall et al., 1986). Accordingly, elongated, negatively stained MCC patches could be recognized in the plasma membrane pattern of fluorescing H⁺-ATPase Pma1-GFP on protoplasts prepared by zymolyase digestion (Malinska et al., 2004). We present here another example of morphologically altered plasma membrane: the membrane of intact ypr050cΔ cells with elongated MCC patches and/or invaginations (Fig. 5). As mentioned in the Results, the ypr050cΔ phenotype could in fact be caused by the absence of Mak3, the catalytic subunit of N-terminal acetyltransferase C (NatC) complex. NatC-mediated acetylation was shown to be required for the targeting of the small GTPase Arl3 to Golgi membranes (Behnia et al., 2004; Setty et al., 2004). None of the 12 soluble proteins known to colocalize with MCC (Grossmann et al., 2008) seems, however, to be itself a subject of NatC-mediated acetylation. Any role of the NatC complex in MCC patch fission would thus be an indirect one. For example, the impaired recruitment of vesicle-tethering factors to the Golgi in mak3Δ cells could lower the rate of vesicle trafficking. Elongated MCC patches in these cells would then appear as a consequence of decreased plasma membrane dynamics.

Finally, the identification of MCC patches with furrow-like plasma membrane invaginations was confirmed by the immunogold localizations of MCC- and eisosome-specific proteins, Sur7 and Pil1, respectively (Figs 6, 7, 8). In the well-preserved ultrastructure of high-pressure-frozen cells, a possible spatial separation of Sur7-containing MCC from Pil1-marked eisosomes was indicated. This observation could not be verified on double-labeled specimens because the signal intensity dropped substantially upon usage of antibodies conjugated with bulky gold particles.

Based on these results, we propose the following model of the formation of furrow-like invaginations: (1) Sur7-enriched, elongated, planar domains devoid of hexagonal arrays are assembled by Pil1 in the plasma membrane. Direct involvement of the Sur7-Pil1 interaction in this assembly is unlikely, however, because cortical patches of C. albicans Lsp1, a homologue of S. cerevisiae Pil1, were detected in a mutant strain of C. albicans lacking Sur7 (Alvarez et al., 2008), which is the only member of the Sur7 protein family in this yeast. The formation of Pil1 patches thus seems to be Sur7 independent. (2) Into the preformed patches, specific transporters are recruited by Nce102 (Grossmann et al., 2008). Specific lipids, such as ergosterol, are recruited with these transporters (Grossmann et al., 2007). (3) Although the participation of membrane proteins in the final step of the furrow formation cannot be excluded so far, this local sterol accumulation is already enough to promote the membrane deformation and could form the furrow-like invagination. It is worth mentioning that another lipid preferring curved membrane localization, phosphatidylethanolamine, is essential for targeting an original constituent of the MCC, Can1, to the plasma membrane (Opekarova et al., 2002).

In the past decade, the view of the plasma membrane as a highly dynamic, laterally compartmented cellular organelle, rather than a fluid, homogenous mosaic, has become generally accepted. In light of our findings, partitioning of the plasma membrane into lateral domains is now finally resolvable by direct observation of intact cells. In the freeze-fractured flat plasma membrane of the nce102Δ mutant, which is devoid of any invaginations, MCC patches appear to be smooth, elongated areas within an otherwise particle-rich surface. Without any obvious relationship to the cytoskeleton (Malinska et al., 2004), they represent autonomous membrane domains with a specific composition and function.

Yeast strains used in this study are listed in supplementary material Table S1. If not stated otherwise, cells were grown in a rich medium (YPD; 2% peptone, 1% yeast extract, 2% glucose) at 30°C on a shaker. Cells expressing CAN1-GFP and ss-GFP-HDEL were cultured in arginine- and leucine-free synthetic medium (0.67% Difco yeast nitrogen base without amino acids, 2% glucose and essential amino acids), respectively.

Living yeast cells (Figs 1 and 4: exponentially growing culture, OD₆₀₀ 0.5-1.0; Fig. 5: overnight culture) were washed briefly in 50 mM potassium phosphate buffer (pH 6.3; KPi), immobilized by a thin film of 1% agarose in KPi and observed at 30°C. Specimens were viewed using LSM510-META confocal microscope (Zeiss) with a ×100 PlanApochromat oil-immersion objective (NA=1.4). Fluorescence signals of GFP and mRFP (excitation 488 nm/Ar laser, and 543 nm/HeNe laser) were detected using band-pass 505-550 nm, and long-pass 580 nm emission filters, respectively.

For the computer simulations, the sets of model objects were generated in Matlab software (The MathWorks). Invaginations were simulated as randomly positioned, randomly oriented, 250-nm-long rods. The only distribution constraint applied was the zero overlap of neighboring objects; 1000 objects per image were generated. We used the confocal image of fluorescent latex bead (175 nm in diameter; Molecular Probes) as a 2D point spread function (PSF) of the microscope. All the microscope settings, the objective, pinhole, image sampling, and the excitation and emission wavelengths were set identically to those for the real images. Model objects were convolved with PSF using Matlab.

Exponentially growing cells (OD 0.5-1.0) were pre-incubated with 30 mM final concentrations EGTA and 10 μg/ml pepstatin A (Sigma) for 5 minutes on a shaker at room temperature. The cells were fixed in 3.7% formaldehyde in PEM buffer (0.1M PIPES, 5 mM EGTA, 5 mM MgCl₂; pH 6.9) for 45 minutes at room temperature and washed twice with PEMI (PEM + 10 μg/ml pepstatin A; all centrifugation steps were performed at 200 g). The cell walls were digested using 20 μg/ml zymolyase (zymolyase-20T, Seikagaku) with 40 μg/ml pepstatin A in KCP buffer (0.1 M K₂HPO₄ + 0.1M citric acid, pH 5.9) for 20-40 minutes at room temperature. The cells were washed twice in PEMI, permeabilized with methanol (30%, 50%, 70% and 90% in water, for 2 minutes each) and rehydrated through a methanol series again. After two washing steps in PEMI, the cells were incubated with primary antibody (in PEMI) for 1 hour at room temperature, washed three times in PEMI and incubated in the secondary antibody conjugated with ultra small gold for 1 hour. Post-fixation (4% glutaraldehyde in PEM, 15 minutes) was used to prevent the loss of signal during the subsequent intensification (silver enhancement kit, Aurion). The labeled cells were pelleted in 10% gelatine, fixed again in 4% glutaraldehyde (in water, 1 hour at 4°C), washed and cut into small pieces. The samples were dehydrated at room temperature through a graded ethanol series (30%, 50%, 70%, 90% and 96%), infiltrated subsequently in 1:1 (v:v) ethanol:epon/durcupan mixture (2 hours at room temperature), 2:1 mixture (overnight, 4°C), and the pure resin (1 hour, and then for 2 hours at room temperature). Infiltrated samples were placed in epon/durcupan-filled molds and polymerized at 60°C for 3 days.

Living yeast cells (overnight culture) were concentrated by suction filtration (McDonald and Müller-Reichert, 2002) onto a filter and this was then placed onto a YPD agar plate. The yeast paste was scraped from the filter, put on a flat specimen carrier (Leica, 1.2 mm cavity diameter) and quickly frozen in a Leica EM PACT high-pressure freezer (Studer et al., 2001).

Frozen samples in the carriers were transferred under liquid nitrogen to freeze substitution medium (different for each purpose – see below) in cryovials and placed in a Leica AFS machine. Cells were freeze substituted at –90°C for 3 days. Thereafter, the temperature was elevated to –50°C (5°C per hour) and samples were kept for about 12 hours at this temperature. After this period, the specimens were washed four times with fresh pre-cooled acetone at –50°C and then infiltrated with Lowicryl HM20, following one of the following protocols.

For the best structure preservation, the FS medium consisted of 3% glutaraldehyde (70% stock; Sigma), 0.1% uranyl acetate (20% methanolic stock; Polysciences) and 1.3% H₂O in acetone (glass distilled; Polysciences) (O'Toole et al., 2002). For immunolabeling detection, the FS medium was modified to contain 0.1% uranyl acetate and 1% H₂O in acetone. After the acetone wash at –50°C, the samples were infiltrated subsequently in 3:1, 1:1, 1:3 (v:v) acetone:HM20 mixtures for 2 hours each at –50°C; then incubated for 2 hours at –50°C in 100% HM20 and finally placed in fresh resin and polymerized with UV for 48 hours at –40°C and ∼3days at 20°C.

For immunogold labeling, sections on formvar-coated gilded copper grids were blocked in 5% bovine serum albumin (BSA) in phosphate-buffered saline (PBS) for 30 minutes, and incubated for 1 hour on droplets of primary antibody diluted in 1% BSA-PBS. Afterwards, the grids were washed on droplets of PBS for 15 minutes, and incubated with ultra small gold-conjugated secondary antibody for 2 hours. In controls, the primary antibody was omitted. After three washes in PBS, the grids were post-fixed in 8% glutaraldehyde for 15 minutes, washed on droplets of distilled water, silver-enhanced (Aurion), air-dried and contrasted.

Primary antibodies: rabbit anti-Sur7 (Pineda Antikörper-Service, Germany; diluted 1:1000 for pre-embedding detection) was designed to recognize C-terminal (cytoplasmic) epitope of the Sur7 molecule; rabbit anti-GFP (Fitzgerald; diluted 1:100 for pre-embedding detection and 1:50 for post-embedding detection) was used to localize C-terminally tagged Sur7-GFP and Pil1-GFP. In post-embedding procedures, this antibody was pretreated with 0.5 mg/ml purified yeast mannan for 30 minutes to suppress the non-specific cell wall binding (Rossanese et al., 1999). Secondary antibody used was ultra small goat anti-rabbit IgG-gold (Aurion; diluted 1:100).

Cells from overnight culture were harvested by centrifugation (1 minute at 1500 g) and washed in KPi buffer (pH 5.5). A 2 μl aliquot of the concentrated cell suspension was loaded onto a gold carrier and frozen rapidly in liquid nitrogen. The sample was cut with a cold knife (≤–185°C), etched for 4 minutes (–97°C; pressure ≤1.3×10^–5 Pa) in a CFE-50 freeze-etch unit (Cressington, Watford, UK), shadowed (1 nm Pt/C, 45°; 10 nm C, 90°), and cleaned in fresh 70% H₂SO₄ for 16 hours (Rachel et al., 2002).

Ultrathin sections (70 nm; 60 nm for serial sections) were cut with a Leica EM UC6 or Ultracut S ultramicrotome equipped with a diamond knife (35°; 45° for Epon; Diatome) and placed on formvar-coated grids (copper; gilded copper for immunolabeling). Sections were contrasted with a saturated aqueous solution of UA for 1 hour, washed, air-dried and examined in a FEI Morgagni 268(D) transmission electron microscope at 80 kV. Images were captured with Megaview II CCD camera. Some of the freeze-fracture micrographs (Fig. 1) were acquired on Philips CM 12 (FEI, The Netherlands) operated at 120 kV and equipped with a slow CCD camera (TVIPS, Tietz, Gauting/Munchen).