Structure-function analysis of human stomatin: A mutation study

Stomatin is an ancient, widely expressed, oligomeric, monotopic membrane protein that is associated with cholesterol-rich membranes/lipid rafts. It is part of the SPFH superfamily including stomatin-like proteins, prohibitins, flotillin/reggie proteins, bacterial HflK/C proteins and erlins. Biochemical features such as palmitoylation, oligomerization, and hydrophobic “hairpin” structure show similarity to caveolins and other integral scaffolding proteins. Recent structure analyses of the conserved PHB/SPFH domain revealed amino acid residues and subdomains that appear essential for the structure and function of stomatin. To test the significance of these residues and domains, we exchanged or deleted them, expressed respective GFP-tagged mutants, and studied their subcellular localization, molecular dynamics and biochemical properties. We show that stomatin is a cholesterol binding protein and that at least two domains are important for the association with cholesterol-rich membranes. The conserved, prominent coiled-coil domain is necessary for oligomerization, while association with cholesterol-rich membranes is also involved in oligomer formation. FRAP analyses indicate that the C-terminus is the dominant entity for lateral mobility and binding site for the cortical actin cytoskeleton.


Introduction
Stomatin is a 31 kDa monotopic integral membrane protein that is palmitoylated, forms homo-oligomers, and associates with cholesterol-rich membrane domains, also known as lipid rafts [1]. It was first identified in the "band 7" region of human erythrocyte membrane proteins [2][3][4][5]. Due to the lack of this protein in red cells of Overhydrated Hereditary Stomatocytosis (OHSt) patients, it was termed "stomatin" [4]. However, the stomatin knockout mouse was viable and did not show stomatocytosis [6]. The lack of this protein in OHSt erythrocytes appears to be due to mistrafficking during terminal erythropoiesis [7]. Human stomatin is ubiquitously expressed in all tissues; highly in hematopoietic cells, relatively low in brain [8,9]. It is associated with the plasma membrane and cytoplasmic vesicles of fibroblasts, epithelial Comparing the crystal structures of the PHB/SPFH core domains from Pyrococcus horikoshii and mouse stomatin [47,48], some differences are showing that are based on varying interacting domains. While the archaeal PHB/SPFH domains form trimers with head-to-tail orientation and interact via coiled-coils [47], the mouse domains interact with their C-termini to form banana-shaped dimers similar to BAR domains and the N-termini of adjacent dimers are bound to each other by hydrophobic interactions yielding linear oligomers [48]. Whichever structure may turn out to be physiologically relevant, both structures have revealed crucial amino acid residues, which imply functional roles.
In this study, we have generated point mutants by exchanging some of these crucial residues in human stomatin and created deletion mutants. We have expressed these in the human cell line A431, which is low in endogenous stomatin, and analyzed the subcellular localization, stability, dynamics, and biochemical properties such as oligomerization, cholesterol binding, and association with lipid raft-like detergent-resistant membranes (DRMs). We identified stomatin as a cholesterol binding protein and noticed that both the CRAC/CARC (residues 55-68) and ORA/CARC (residues 263-273) domains are equally important for the association with cholesterol-rich domains. This association with cholesterol-rich membranes also appears to be a condition for oligomer formation, while the coiled-coil domain is clearly essential for oligomerization. At the plasma membrane, lateral mobility measurements revealed the interaction of the C-terminus (residues 264-288) with the actin cytoskeleton. Our data are in accordance with the previously proposed function of stomatin as an integral scaffolding protein.

Materials
Research chemicals were obtained from Sigma. Cell lines A431 and COS-7 were purchased from ATCC. [ 3 H]photocholesterol was synthesized as described by Thiele et al. [54]. Monoclonal mouse antibody against c-myc was obtained from the Developmental Studies Hybridoma Bank (DSHB) and rabbit anti-GFP was from Abcam. The mouse monoclonal antibody against human stomatin (GARP-50) was described [2].

Cell culture
Human squamous epithelial carcinoma cells (A431) and the derived stably transfected cell lines were maintained in Dulbecco's modified Eagle's medium (DMEM) with high glucose, supplemented with 10% fetal bovine serum (Sigma) and antibiotics under standard conditions.

Confocal laser scanning microscopy (CLSM)
Confocal fluorescence microscopy of paraformaldehyde-fixed specimens was essentially performed by standard methods as described previously [10,11]. Pictures were acquired using the Zeiss LSM 510 META CLSM and the corresponding software.
Fluorescence Recovery After Photobleaching (FRAP) analysis A431 cell lines stably transfected with GFP-tagged WT or mutated stomatin were chosen that showed plasma membrane expression of the construct. The cells were seeded on 35 mm glassbottom dishes (MatTek Corp., Ashland, MA) and cultivated for 24 h. For cholesterol depletion or loading, cells were incubated with methyl-β-cyclodextrin (MβCD) or the cholesterol-MβCD complex in Hank's balanced salt solution (HBSS). For disruption of actin microfilaments, cells were treated with 10 μM cytochalasin D (cytoD) in medium for 30 min. FRAP analysis was performed on the Zeiss LSM 510 META confocal microscope equipped with a stage heater at 37˚, as described previously [51]. Analyses were repeated at least 20-times for each cell line. Fluorescence intensity data of whole cell ROI, bleached ROI, and background ROI were quantified using the Zeiss software package. Recovery curves were bleach-and background-corrected and the mobile fractions and recovery halftimes were calculated as described [55]. Recovery data, derived from !20 experiments, were fit to the equation y = y o + a(1-e -kt ) with GraphPad Prism scientific graphing software to calculate the rate constant k. The recovery halftime was calculated as t 1/2 = ln2/k.

Density gradient centrifugation
Cell membranes were prepared, solubilized, and subjected to density gradient ultracentrifugation as described [50]. Briefly, the membranes of confluent A431 clones (one 15-cm dish each) stably expressing GFP-tagged WT or mutant stomatin were solubilized in 200μl of 1% Triton X-100 in TNE buffer (20 mM Tris-Cl, pH 8.0, 130 mM NaCl, 5 mM EDTA) containing protease inhibitors (10 μg/ml aprotinin and leupeptin, 1 μg/ml pepstatin A, and 1 mM phenylmethylsulfonyl fluoride) at room temperature for 10 min. The solubilized proteins were applied on top of a linear density gradient (15-50% sucrose) and centrifuged in an SW40 rotor (Beckman) at 40,000 rpm and 4˚C for 19 h. Nineteen fractions were collected from the top and subjected to SDS-PAGE and Western blotting. The sucrose gradient was regularly checked by refractometry. Marker proteins were run in parallel.
Flotation analysis was carried out as described [10]. Briefly, cell membranes of confluent A431 clones (one 15-cm dish each) were prepared by homogenization and centrifugation. These were solubilized in 280 μl of ice-cold TNE buffer containing 1% Triton X-100 for 10 min. The lysate was adjusted to 50% sucrose by the addition of 490 μl of 80% sucrose in TNE and overlaid with 3 ml of 35% sucrose followed by 1.2 ml of 5% sucrose in the same buffer. After ultracentrifugation at 48,000 rpm and 4˚C in an SW 55 rotor (Beckman) for 18 h, 9 fractions were collected from the top and analyzed by Western blotting. Detergent-resistant membrane (DRM) components were identified in the low-density fractions 2 and 3, whereas Triton X-100-soluble proteins were found in the dense fractions 5 to 9.
Binding of photocholesterol COS-7 cells were transiently transfected with WT and mutated stomatin constructs, respectively, with or without myc-tags, and subjected to photocholesterol cross-linking according to the method of Thiele et al. [54,56]. Briefly, the cells were incubated in 10 cm dishes with 10 μCi [ 3 H]photocholesterol in DMEM at 37˚C for 15 min and irradiated for 1 min with a 500 W beam of filtered UV-light (λ > 310 nm) of a high-pressure mercury lamp at 4˚C to crosslink photocholesterol to cholesterol-binding proteins. The cells were lysed, stomatin was immunoprecipitated with anti-stomatin antibody GARP-50 and analyzed by SDS-PAGE and autoradiography. Protein expression levels were estimated by immunoblotting with GARP-50 and densitometry of stomatin bands.

Western blot analysis
Fractions collected from the sucrose gradients were analyzed by 10% SDS-PAGE and blotted onto nitrocellulose by standard methods as described previously [50]. HighPrecisionProtein™ Standards (Bio-Rad) were used as molecular weight markers. After blocking, the blots were incubated with anti-stomatin antibody GARP-50, anti-GFP, or anti-myc for 1 h. After washing, the blots were incubated with anti-mouse or anti-rabbit horseradish peroxidase conjugate and the SuperSignal™ chemiluminescent substrate (Pierce).

Statistical analysis
Densitometric values of Western blot or autoradiographic bands were obtained by using Ima-geJ software and analyzed statistically with the IBM SPSS program. Mean values, standard deviation, and standard error were calculated and p-values were determined by two-tailed ttest.

Mutation and expression of GFP-tagged stomatin
To study the relevance of some prominent structural features of stomatin, we deleted parts of the molecule and exchanged single amino acids. The recent elucidation of the 3-dimensional structure of the highly conserved PHB/SPFH core of archaeal and mouse stomatin [47,48] highlighted several amino acid residues that implicated a structural and functional role. Moreover, the finding that the conserved, long α-helix adjacent to the PHB/SPFH domain was involved in coiled-coil interactions [47] gave a hint as to the formation of stomatin oligomers. Therefore, we exchanged the residues in question and deleted the long α-helix to study their effects on the structure and function of stomatin. An overview of the deletions and amino acid exchanges that we have introduced into GFP-tagged stomatin is given in Fig 1. Generation and stable transfection of GFP-tagged stomatin mutants. All of the constructs shown in Fig 1 were stably transfected in the human carcinoma cell line A431 that is known for low expression of endogenous stomatin [51]. Single cell-derived clones were isolated, expanded and analyzed for stable expression of the GFP fusion proteins. To get a variety of expression levels, we pooled various isolated clones to populations. Subcellular localization of expressed GFP-tagged stomatin mutants. In a variety of cell lines, stomatin is localized to the plasma membrane (PM) and the late endosomal/lysosomal compartment [10,11,19,57]. Therefore, we started out by comparing our stable mutant cell lines with respect to these compartments.
The localization in A431 cells of WT stomatin, Stom(1-288)GFP, and the deletion mutants Stom(22-288)GFP (ΔN), Stom(1-263)GFP (ΔC), and Stom(d204-241)GFP (ΔCC), depicted in S1 Fig, appeared as previously shown for endogenous or myc-tagged stomatin in various cell lines [10,57], namely at the PM and cytoplasmic vesicles. However, several mutants were unable to target PM, namely Pro47Ser, Cys87Ser, Asp89Ala, Arg97Ala, Lys198Ala, and Pro270Ala. Other mutants, such as Ile57Ala, Tyr60Ala, Pro200Ala, and Pro245Ala showed reduced PM staining (S1 Fig). The complete localization data of the stomatin mutants are summarized in S2 Table. Cholesterol binding of wildtype and mutant stomatin Wildtype and mutant stomatin constructs with or without myc-tag were transiently expressed in COS-7 cells, incubated with [ 3 H]-labeled photocholesterol and irradiated with UV light for crosslinking. The expressed proteins were immunoprecipitated with anti-stomatin antibody and the covalently bound [ 3 H]photocholesterol was identified by autoradiography of SDS-PAGE gels. To estimate the expression levels of WT and mutant stomatin, immunoblotting with anti-stomatin antibody was performed. Radioactive bands were observed for WT stomatin and essentially all mutants (Fig 2).
Quantitative data were assessed by densitometry of autoradiography and Western blot bands (Table 1) and by calculating the ratios of these data. The expression levels of mutants varied largely, thereby changing the ratio of bound cholesterol to each stomatin mutant. The C-terminal deletion mutant Stom(1-263) bound only one third of WT-bound cholesterol (Table 1) suggesting the loss of a binding site at the C-terminal end (residues 264-288). Interestingly, the values of Stom(F269A) and Stom(P270) were quite different; while cholesterolbinding to F269A was slightly reduced, binding to P270A was enhanced almost twofold, when compared to WT. The myc-tagged constructs showed higher cholesterol-binding, particularly the Cys-mutants; however, the results of mutants Stom(C53S)myc and Stom(C87S)myc were not significant due to exceedingly varying ratios.

Oligomerization and DRM-association of GFP-tagged wildtype and mutant stomatin
Hallmarks of stomatin structure and function are the oligomeric nature [50,51] and association with lipid raft-like detergent-resistant membranes (DRMs) [10,51]. Therefore, we analyzed the GFP-tagged stomatin mutants to evaluate their ability to form oligomers and associate with DRMs.
Oligomerization of GFP-tagged wildtype and mutant stomatin. We analyzed oligomers by linear density gradient ultracentrifugation ( Table 2). Due to the large size of the oligomers, we used higher than usual concentrations, i.e. 15-50% sucrose (B) Hypothetical model of a monomeric wildtype stomatin in association with a biological membrane. Sidedness is marked by "in" (cytoplasmic) and "out" (extracellular or luminal). The color code denotes the domains as illustrated in (A). The green ball at the N-terminal region symbolizes the phosphorylation site at Ser-10; the "P" at the kink within the hydrophobic IM domain marks residue Pro-47, which is responsible for the monotopic membrane protein structure. The model is roughly drawn according to known and estimated sizes; the N-terminal region is αhelical (E. Umlauf, unpublished results), the PHB/SPFH core domain is 5 nm in length and 2 nm in height, while the coiled-coil domain is 6 nm long [47]. CARC denotes a reversed CRAC motif; there are three CARC motifs, two overlapping with the CRAC-like (CL) and one overlapping with the ORA motif. Schematic models of the most remarkable mutants are shown in S4 Fig. gradient. WT stomatin yielded oligomers in the range of 500-700 kDa but also a varying amount of mono-or dimers (70-150 kDa) estimated at 10-50% of total stomatin (n = 5). It has to be taken into account that the large GFP-tag may at least partially affect stomatin oligomerization or oligomer stability in solution; however, a few percent of monomer are also found in endogenous stomatin and untagged oligomeric mutants [50,51]. The N-and C-terminal deletion mutants, ΔN and ΔC, showed similar distributions as the previously described myc-tagged mutants [50], that is oligomerization like WT and inability to oligomerize, respectively (Fig 3, left panel, S2 Fig). The coiled-coil deletion mutant, ΔCC, yielded only mono-/ dimers. Cys30Ser showed roughly equal parts being mono-/dimeric and oligomeric, while the other Cys mutant, Cys87Ser, appeared similar but was largely degraded, as noted before [39]. Pro47Ser was mono-/dimeric ( DRM-association of GFP-tagged wildtype and mutant stomatin. We compared the stomatin mutants with respect to DRM-association using sucrose density step gradient ultracentrifugation (Fig 3, right panel). WT stomatin showed a distribution similar to endogenous stomatin [10,51], partly floating to the top but partly also being dissolved in the detergent Triton X-100 (Fig 3, right panel, and S3 Fig; Table 3). While the N-terminal truncation mutant (ΔN) behaved like WT, the C-terminal truncation (ΔC) led to absence from the floating DRMfraction, as previously shown for the untagged or myc-tagged proteins [10,51]. Deletion of the coiled-coil domain (ΔCC) did not interfere with DRM-association. Apparently, there is no need for stomatin to oligomerize in order to associate with cholesterol-rich membranes. Palmitoylation is thought to be necessary for DRM-association; however, both Cys30Ser and Cys87Ser partially floated to the top (  Table 2). The mono-/ dimeric ORA/CARC mutant Phe269Ala partially floated with DRMs in contrast to the untagged mutant reported previously [51]. The DRM-association data of all mutants are summarized in Table 3. Densitometry of Western blot bands obtained from gradient fractions showed the distribution of the DRM-associated WT or mutant stomatin (fractions 1-3), versus Triton X-100-soluble, non-DRM stomatin mutants (fractions 4-9). Relative amounts (% of total) are listed and visualized (

Lateral mobility of GFP-tagged wildtype and mutant stomatin
To interrelate the biochemical data with the situation in living cells, we analyzed the lateral mobility by Fluorescence Recovery After Photobleaching (FRAP) measurement of those which was estimated at 80 kDa, as predicted due to glycosylation [40]. https://doi.org/10.1371/journal.pone.0178646.g003 Structure-function analysis of stomatin  Structure-function analysis of stomatin mutants that were stably expressed at the PM (S1 Fig). Compared to WT, the C-terminal truncation mutant (ΔC) showed the largest mobile fraction (Fig 4, Table 4), in accordance with previous data [51]. ΔC neither formed oligomers nor associated with DRMs (Fig 3) and thus this lack of interaction may be the cause for the high lateral mobility in the membrane. Similarly, the second largest mobile fraction (MF) was observed for the CRAC-like mutant Arg62Ala that likewise did not oligomerize nor bind to DRMs. Comparing the data of all analyzed mutants, we found a reverse correlation between the size of MF and the ability to form oligomers or associate with DRMs ( Table 4). The N-terminal deletion mutant (ΔN) and WT stomatin are both able to oligomerize and bind to DRMs [51] and they presented the lowest MFs, whereas the highest MFs were observed with mutants that lacked both oligomerization and DRM-association, i.e. Arg62Ala and ΔC (Fig 4, Table 4). It is easily conceivable that oligomerization and DRM-association constitute two kinds of membrane interactions that restrict lateral mobility.

Lateral mobility of GFP-tagged wildtype and mutant stomatin as a function of membrane cholesterol and actin cytoskeleton
Stomatin binds cholesterol (Fig 2), generates oligomers, associates with DRMs (Fig 3), and interacts with cortical actin microfilaments [57]. While the sites of oligomerization and/or DRM-association have been localized to the ORA region [51], coiled-coil domain, and CRAC/ CARC motifs (Fig 3), the site of actin-binding is still unknown; however, a PDZ-protein binding motif near the C-terminal end [22] may be involved. To analyze the effects of the actin cytoskeleton and membrane cholesterol-levels on stomatin lateral mobility, we treated WT and mutant stomatin expressing cells with cytochalasin D (cytoD) and/or methyl-β-cyclodextrin (MβCD), with or without cholesterol. In general, treatment of WT and mutants with cytoD showed increased MF and recovery ( Fig 5, Table 5) suggesting that these proteins were dissociated from the cortical actin cytoskeleton or released from an actin-dependent PM compartment. According to the FRAP data, about 20% of WT stomatin were released from the immobile fraction by the treatment with cytoD and most likely were associated with the actin cytoskeleton. Similar values were found for ΔN, ΔCC, and Cys30Ser, while smaller differences were seen with Arg62Ala and Pro200Ala. There was no effect of cytoD on ΔC and Pro245Ala MFs, suggesting that these mutants were not associated with the actin cytoskeleton. Recovery halftimes showed a remarkable increase in the lateral diffusion of WT, ΔN, Pro200Ala, and Pro245Ala, while the other mutants presented marginal increases.
Cholesterol depletion of WT and mutants showed little changes in MF (Fig 5, Table 5). Only Pro245Ala presented a marked reduction in MF. In contrast, cholesterol depletion had some effect on the speed of recovery, particularly slowing down Cys30Ser and Pro200Ala. This effect is known and attributed either to the concomitant depletion of cholesterol and phosphatidylinositol(4,5)bisphosphate resulting in reorganization of the actin cytoskeleton [58][59][60] or formation of a solid-like phospholipid diffusion barrier [61]. The slower recovery after cholesterol depletion was partially reversed by concomitant cytoD treatment, indicating the involvement of cytoskeleton. For Cys30Ser, cholesterol depletion slowed down recovery (from 29.3 s to 49.0 s) but concomitant treatment with cytoD alleviated slowing to 37.3 s. Similarly, for Pro200Ala, the slowing of recovery due to cholesterol depletion (from 32.3 s to 50.6 s) was moderated (40.7 s) when cells were concomitantly treated with cytoD. However, rapid recovery was seen after cytoD treatment alone (19.3 s).
Cholesterol loading showed some reduction in MF, particularly of ΔC (-26%) and Arg62Ala (-17%). ΔC and Arg62Ala are mutants that do not associate with DRMs (Fig 3 and S3 Fig,  Table 3), probably due to the loss of cholesterol binding sites, and neither oligomerize. Concomitant cytoD treatment of the mutants partially reversed the low MFs, suggesting that the actin cytoskeleton was involved, except for ΔC, which showed unchanged MF (Fig 5, Table 5), thus confirming that ΔC is not interacting with the actin cytoskeleton. Cholesterol loading had some impact on recovery halftimes, increasing lateral mobility, particularly of Arg62Ala and Pro245Ala. This effect of cholesterol loading had been attributed to endocytosis [60]; possibly, the increased speed reflects the cholesterol flow in the PM caused by increased endocytosis. Concomitant treatment with cytoD resulted in extremely rapid recovery of ΔN (17.7 s).

Discussion
In this study, we used targeted mutation to evaluate the structural and functional significance of prominent domains and amino acid residues of the monotopic membrane protein stomatin. We selected these domains and residues on the basis of recent NMR and crystal structures of the PHB/SPFH domain [47][48][49] and our biochemical data [5,10,11,39,50,51]. Schematic models of WT and mutant stomatin showing the relevant domains and residues are depicted in   Table 4.
https://doi.org/10.1371/journal.pone.0178646.g004 Table 4. Lateral mobility of plasma membrane-bound GFP-tagged wildtype and mutant stomatin. Correlation of mobile fractions and recovery halftimes with the ability to oligomerize and/or associate with DRMs.

Structure-function analysis of stomatin
When we expressed the GFP-tagged mutants, we surveyed the subcellular localization and found that several mutants were unable to target the PM (S1 Fig, S2 Table). Other mutants, particularly Cys87Ser, were difficult to express at all. This may be due to gross structural changes and degradation or inefficient transport via the Golgi apparatus. Palmitoylation of Cys-87 has been shown to be essential for the attachment of the PHB/SPFH domain to the PM [16].  Table 5. https://doi.org/10.1371/journal.pone.0178646.g005 Structure-function analysis of stomatin Stomatin-like proteins are known to bind cholesterol, which modifies interactions with ion channels thereby regulating channel activities [33,62]. To study whether stomatin itself binds cholesterol, we have used the photochemical crosslinking approach [54] comparing WT stomatin with various mutants (Fig 2, Table 1). We clearly showed covalent binding of [ 3 H] photocholesterol to WT stomatin and all mutants (Fig 2). Comparing the relative amounts of bound cholesterol (Table 1), large differences were seen between different mutants. This may be explained by different affinities for cholesterol but also by large variations of mutant expression levels. The cholesterol-labeling data of all these mutants indicate that there is more than one cholesterol-binding site on stomatin. As pointed out, stomatin contains several domains or motifs that may function as cholesterol-binding sites [46], particularly in the CRAC and CARC regions. It is currently not clear, why myc-tagged stomatin shows higher cholesterolbinding than untagged stomatin (Fig 2, Table 1). Possibly, the modified C-terminus may cause a structural change that raises the affinity of cholesterol binding sites.
Apart from binding cholesterol, there are two other hallmarks of stomatin and similar monotopic proteins, such as flotillins and caveolins, namely the generation of oligomers and the association with lipid raft-like DRMs [10,50]. The basic mechanisms of the respective protein-protein and protein-lipid interactions are still unknown and probably need advanced methods for elucidation. Here, we used classical density gradient ultracentrifugation to estimate and compare the impacts of various mutations on oligomerization and DRM association (Fig 3, Tables 2 and 3, S2 and S3 Figs). Due to the relatively large GFP-tag, the changes seen were less pronounced than with untagged or myc-tagged proteins [51]; however, in principle, the data are in line with each other. One question arises, why in ΔC the lack of the ORA/ CARC domain (residues 263-273) is causing the absence of this mutant from DRMs, when other putative cholesterol-binding sites such as the IM domain and adjacent CRAC/CARClike region (residues 55-68) are still available. In this respect, is the C-terminal ORA/CARC domain more important for DRM-association than the upstream CRAC/CARC-like domain or is it necessary that these cholesterol-binding domains cooperate? Similarly, the exchange of the CRAC-like consensus residues Tyr-60 and Arg-62 led to absence of these mutants from DRMs despite the intact C-terminal ORA/CARC domain. This shows that both the upstream CRAC/CARC and downstream ORA/CARC domains are equally important and suggests that they do cooperate in cholesterol-rich membranes. Another question is, why should the lack of association with DRMs, as in the CRAC/CARC and ΔC mutants, lead to absence of oligomers? There may be a structural connection between the CRAC/CARC and ORA/CARC domains but it is currently unknown. In this context, several membrane proteins have been described to contain adjacent CRAC and CARC motifs, possibly for cholesterol-transfer [46]. Related to the above question is, how such a small domain as ORA/CARC (residues 263-273) may be causing two apparently unrelated effects, oligomerization and DRM association. Possibly, the  Structure-function analysis of stomatin major effect of the CRAC/CARC and ORA/CARC (ΔC) mutations is the disruption of stomatin binding to cholesterol-rich membranes/lipid rafts and this loss may in consequence prevent oligomerization due to unfavorable spatial alignment of monomers. Eventually, whole molecule structure analysis, e.g. cryoelectron microscopy, will give the answer to these questions. Deletion of the coiled-coil domain (ΔCC) resulted in the absence of oligomers thereby confirming the interacting function of this domain [47], whereas DRM association was unaffected (Fig 3, Tables 2 and 3, S2 and S3 Figs), probably because the CRAC/CARC domains, particularly the possibly affected downstream ORA/CARC domain, remained intact (S4 Fig). Exchange of Pro-47 led to lack of oligomers and absence of DRM association as predicted for a single-pass transmembrane glycoprotein [40] (S4 Fig). Exchange of Phe-91, which may be involved in linear oligomerization of the PHB/SPFH domain, similar to mouse residue Leu-91 [48], led to an unstable molecule, with the tendency to aggregate; it did not associate with Comparing the three phenotypes of stomatin mutants, expression at the PM, oligomerization, and DRM association, showed that there is some intricate correlation between them ( Table 6). Several mutants were unable to target the PM; however, those that were expressed at the PM neither required oligomerization nor association with DRMs. Some mutants, which did not oligomerize, were still able to associate with DRMs. However, all mutants that were oligomeric, also associated with DRMs. This suggests that DRM association precedes oligomerization or, in other words, DRM association is a necessary but not sufficient condition for stomatin oligomerization. It is conceivable that small, mono-or dimeric stomatin-cholesterol complexes may oligomerize to form large, stomatin-cholesterol/lipid aggregates. This is in line with the idea that small lipid rafts may coalesce to form large cholesterol-rich signaling platforms [63,64].
The in vitro biochemical characterization of oligomerization and DRM association does not directly reflect the situation of molecules in the living cell. Therefore, we complemented our biochemical data with Fluorescence Recovery After Photobleaching (FRAP) measurements to relate the lateral mobility of stomatin mutants to the phenotypes of varying oligomerization and/or DRM-association (Fig 4, Table 4). Of all analyzed mutants, the C-terminal deletion (ΔC) was outstanding with its high mobile fraction (MF) and rapid recovery. Because this mutant does not oligomerize nor associate with DRMs (Fig 3, Tables 2 and 3) [51], this lack of interaction with other monomers and/or cholesterol-rich membrane domains may cause this high lateral mobility. Comparing all FRAP data with the mutants' biochemical characteristics, we saw that the highest MFs were measured for mutants that neither oligomerize nor associate with DRMs, while the lowest MFs were found for mutants that generate oligomers and bind to DRMs (Table 4). It is quite evident that such molecular interactions are reducing the lateral mobility; however, a major player in restricting lateral mobility is the actin cytoskeleton [65,66] and stomatin is known to interact with cortical actin microfilaments [57].
To evaluate the effects of the membrane actin meshwork on lateral mobility of stomatin mutants, we treated the cells with cytoD to disrupt the microfilaments. Moreover, to estimate the influence of cholesterol-rich lipid domains, we manipulated membrane cholesterol levels ( Fig 5, Table 5). In general, treatment of cells with cytoD had the highest enhancing effect on MF of all mutants that had an intact C-terminus (Fig 5, Table 5), while it had no effect on mutants with lacking or impaired C-terminus (ΔC, Pro245Ala). This indicates that the C-terminal end (residues 264-288) contains an actin cytoskeleton binding site, as hypothesized [1].
Without an intact cortical cytoskeleton or a respective binding site, stomatin would be free to diffuse, as shown by the high MF and rapid recovery.
Depletion of cholesterol had little effect on MFs of mutants, except for the marked MF reduction of Pro245Ala and MF increase of ΔCC (Table 5); however, it resulted in slower recovery of WT and several mutants. This effect of cholesterol depletion appears to be a general phenomenon of all membrane proteins [60] and may be due to the reorganization of the actin cytoskeleton, because MβCD-treatment does not only sequester cholesterol but also phosphatidylinositol (4,5)bisphosphate [58,59], which is essential for actin filament organization. Alternatively, or additionally, a diffusion barrier of solid-like phospholipid domains may form in the absence of cholesterol [61] and thus restrict lateral mobility. Loading of cholesterol resulted in a marked increase in the immobile fraction, particularly of ΔC and Arg62Ala, two mutants that do not show oligomerization nor association with DRMs. In general, cholesterol loading shifts the subcellular distribution of membrane proteins and causes endocytosis [60]; possibly, the loss-of-function mutants ΔC and Arg62Ala are insensitive to membrane cholesterol flow and remain immobile, while mutants that bind cholesterol may follow the dynamic cholesterol gradient in the PM. In the case of Arg62Ala, the original MF was restored by cytoD, indicating that cytoskeleton was involved; however, the MF of ΔC was not restored by cytoD, according to the lack of actin-binding. Cholesterol loading may also lead to molecular crowding in the PM, increasing viscosity, and possibly immobilization of the overcrowded phase.
To sum up our data, we give here an overview of the induced structural changes and the resulting phenotypes (Table 7). We put the focus on six areas/domains of stomatin that appear functionally relevant, i.e. the N-terminus, IM-domain, CRAC/CARC motif, PHB/SPFHdomain, coiled-coil domain, and ORA/CARC domain. To illustrate the structural changes in stomatin, we show schematic models of the most remarkable mutant proteins (S4 Fig). Structure-function analysis of stomatin Oligomerization and DRM-association as described [51], see structural model (S4 Fig). Interaction with actin cytoskeleton, as WT.

IM
Cys30Ser Loss of palmitoylation Affected oligomerization; Reduced lateral speed after cholesterol-depletion.
Suggests a structural change impairing oligomerization. Cholesterol-depletion creates a barrier to Cys30Ala diffusion.
As to be expected from a normal transmembrane protein [40] (S4 Fig).

Loss of C-terminal domain
No oligomers; no DRM-association; largest MF, unchanged by cytoD; fastest diffusion; low cholesterol-binding (Table 1).
FRAP data indicate that the C-terminus is essential for actin-cytoskeleton-binding. Lack of DRMassociation is in line with low cholesterol-binding.
ORA characteristics as described [51], except for partial DRM-association that suggests a tag-effect. Reduced DRM-association is in line with reduced cholesterol-binding.
ORA characteristics as described [51]. Enhanced DRM-association is in line with stronger cholesterol-binding. Structure-function analysis of stomatin In conclusion, we have shown that stomatin is a cholesterol binding protein and that one binding site is located within the C-terminal region (residues 264-288). Two regions of stomatin, the putative cholesterol binding sites, CRAC/CARC (residues 55-68) and ORA/CARC (residues 263-273), are equally responsible for the association with cholesterol-rich membranes. We conclude that these two regions are either structurally connected or cooperating. Moreover, this association with cholesterol-rich membranes appears to be a condition for oligomer formation possibly by creating a membrane microenvironment that induces oligomerization. For oligomerization, the coiled-coil domain is clearly essential, while the CRAC/ CARC-residues Tyr-60, Arg-62, and the ORA/CARC domain, as well as Phe-91, may play an indirect role in oligomerization by their primary association with cholesterol-rich domains. The interaction of stomatin with the cortical actin cytoskeleton has been linked to the C-terminal region (residues 264-288).

S1 Fig. Subcellular localization of GFP-tagged wildtype and mutant stomatin.
Confocal laser scanning microscopy (CLSM) of A431 cells stably expressing GFP-tagged wildtype (WT) stomatin shows the normal, dual localization to the plasma membrane (PM) and late endosomal/lysosomal compartment. Deletion of the N-terminus, C-terminus, or coiled-coil domain did not alter this localization and distribution. The point mutants Cys30Ser, Arg62Ala, Phe91Ala, and Phe269Ala, also showed the normal localization and distribution. Lack of PM staining or largely reduced staining was observed for Pro47Ser, Cys87Ser, Asp89Ala, Arg97Ala, Lys198Ala, and Pro270Ala, while weak staining of PM and preferential staining of cytoplasmic vesicles was visible in cells expressing Ile57Ala, Tyr60Ala, Pro200Ala, and Pro245Ala.