Activation of Lyn Tyrosine Kinase through Decreased Membrane Cholesterol Levels during a Change in Its Membrane Distribution upon Cell Detachment*

Background: Lyn tyrosine kinase is anchored to the plasma membrane. Results: Cell detachment changes Lyn distribution from the low density to the high density membrane fractions and activates Lyn through decreased membrane cholesterol levels. Conclusion: Cell detachment-induced Lyn activation plays a role in survival of suspension cells. Significance: Our results provide novel insight into signal transduction upon cell detachment. Cellular membranes, which can serve as scaffolds for signal transduction, dynamically change their characteristics upon cell detachment. Src family kinases undergo post-translational lipid modification and are involved in a wide range of signaling events at the plasma membrane, such as cell proliferation, cell adhesion, and survival. Previously, we showed the differential membrane distributions among the members of Src family kinases by sucrose density gradient fractionation. However, little is known about the regulation of the membrane distribution of Src family kinases upon cell detachment. Here, we show that cell detachment shifts the main peak of the membrane distribution of Lyn, a member of Src family kinase, from the low density to the high density membrane fractions and enhances the kinase activity of Lyn. The change in Lyn distribution upon cell detachment involves both dynamin activity and a decrease in membrane cholesterol. Cell detachment activates Lyn through decreased membrane cholesterol levels during a change in its membrane distribution. Furthermore, cholesterol incorporation decreases Lyn activity and reduces the viability of suspension cells. These results suggest that cell detachment-induced Lyn activation through the change in the membrane distribution of Lyn plays an important role in survival of suspension cells.

Cellular membranes are assemblies of lipids and proteins and serve as signaling platforms. They consist of various membrane domains, which are specialized by their properties and linked to their specific functions. Lipid rafts, receptor foci, and organelles are separated by chemical, physical, or microscopic characteristics (1)(2)(3)(4)(5).
Src family kinases, a family of nonreceptor tyrosine kinases, include at least eight highly homologous proteins, c-Src, Lyn, Fyn, c-Yes, c-Fgr, Hck, Lck, and Blk, and are involved in cell survival, transformation, and metastasis (6 -8). Src family kinases have an N-terminal Src homology (SH) 3 4 domain that contains lipid modification sites, an SH3 domain and an SH2 domain, which mediate protein-protein interactions, and a tyrosine kinase catalytic domain at the C-terminal region (9). Src family kinases are located at the cytoplasmic side of the plasma membrane via post-translational lipid modification. We have shown that the state of lipid modification at their SH4 domains makes a difference in trafficking among Src family kinases (10 -13).
Microscopic and biochemical analyses showed that Src family kinases are present in both lipid raft and non-raft membranes (14,15). There is increasing evidence of the relationship between the membrane distribution and the activity of Src family kinases. A previous study showed that Lyn, a member of Src family kinases, is activated in lipid raft membranes (16). The distribution of Src family kinases to non-raft membranes is associated with cell transformation (7). Furthermore, some types of membrane domains in the plasma membrane, such as lipid rafts and caveolae, are internalized following cell detachment (17,18). Cell detachment elevates production of cAMP, resulting possibly from lipid raft internalization (19). Cell detachment also induces activation of the Src family kinases c-Src and c-Yes (20). Although cell detachment plays a role in cell membrane functions, little is known about the relationship between the kinase activity and the membrane distribution of Src family kinases upon cell detachment.
In this study, we separated membranes of HeLa S3 cells by sucrose density gradient fractionation without detergent (10), which can split membrane segments into 10 fractions by their density, probably due to the proportions of lipids to proteins (1, as described (23). To establish cells stably expressing shRNA against luciferase, Lyn, Fyn, or Lyn plus Fyn, HeLa S3 cells were co-transfected with the shRNA expression vector and a plasmid containing the hygromycin-resistant gene and selected in 250 g/ml hygromycin. HeLa S3/c-Src-HA cells were generated for tetracycline-inducible c-Src-HA expression (28). Transient transfection was performed using linear polyethyleneimine (25 kDa; Polysciences) (29).
Adherent and Suspension Cultures-For adherent culture, cells were seeded on tissue culture dishes and cultured in Iscove's modified Dulbecco's medium containing 5% bovine serum (BS). For suspension culture, adherent cells were detached by treatment with 0.25% trypsin for 2 min at 37°C and then cultured on poly(2-hydroxyethyl methacrylate) (poly-HEMA)-coated dishes in RPMI 1640 medium containing 5% BS. Poly-HEMA-coated dishes were prepared as described previously (30,31). In brief, 3% (w/v) poly-HEMA (Sigma) was dissolved in 95% ethanol at 37°C. Culture dishes were filled with poly-HEMA solution, and then ethanol was evaporated under air blowing for 1 h. To support cell attachment at low concentrations of serum, culture dishes were coated with fibronectin. In brief, dishes were incubated with 50 g/ml fibronectin (BD Biosciences) in phosphate-buffered saline (PBS) at room temperature for 1 h and then washed gently with water. For suspension culture of HCT116 cells, cells were trypsinized and cultured in a spinner flask with RPMI 1640 medium containing 5% BS. THP-1 cells were grown in suspension in culture dishes with Iscove's modified Dulbecco's medium containing 5% fetal bovine serum (FBS).
Preparation of M␤CD-Cholesterol-Cholesterol-loaded methyl-␤-cyclodextrin (M␤CD-cholesterol) was prepared as described previously (32). In brief, 35 mg of cholesterol (Sigma) was solubilized in 150 l of isopropyl alcohol/chloroform (2:1 v/v) and then 7.63 ml of 100 mM M␤CD was added at 80°C. After solubilization of cholesterol, the solution was filtered through a 0.2-m pore size membrane. M␤CD-cholesterol was diluted in serum-free medium (1:10 v/v) before use.
Microscopy-Immunofluorescence staining was performed as described previously (24,28,31). In brief, cells were fixed in PBS containing 2% paraformaldehyde for 20 min at 37°C, permeabilized with PBS containing 0.1% Triton X-100 for 3 min at room temperature, blocked in PBS containing 0.1% saponin and 3% bovine serum albumin, and then sequentially incubated with a primary and a secondary antibody for 1 h each. Confocal and Nomarski differential-interference-contrast images were obtained using a Fluoview FV500 laser scanning microscope (Olympus). Filipin staining of cell surface cholesterol was performed as described (17). In brief, cells were fixed with 4% paraformaldehyde for 10 min at room temperature, quenched with 50 mM NH 4 Cl for 10 min, incubated with 50 g/ml filipin III (Cayman Chemical Co.) for 15 min, and washed with PBS. Filipin staining of whole cells was performed according to the protocol of the manufacturer. In brief, cells were fixed with 4% paraformaldehyde for 10 min at room temperature, washed with PBS containing 0.1% Triton X-100 three times, and then stained with filipin III. Filipin fluorescence and phase contrast were obtained using an inverted Zeiss Axiovert S100 micro-scope, and the images were analyzed using ImageJ software (National Institutes of Health).
Western Blot Analysis-Whole cell lysates prepared in SDSsample buffer containing the tyrosine phosphatase inhibitor Na 3 VO 4 (10 mM) were separated by SDS-PAGE and electrotransferred onto polyvinylidene difluoride membranes (PVDF, Millipore). Immunodetection was performed as reported previously (10,26). Results were analyzed using an image analyzer ChemiDoc XRSplus (Bio-Rad). Intensity of chemiluminescence was measured using Quantity One software (Bio-Rad).
Sucrose Density Gradient Fractionation-Sucrose density gradient fractionation was performed as described previously (10). In brief, cells were swollen in hypotonic buffer (40 mM Hepes, pH 7.4, 1 mM MgSO 4 , and 10 mM Na 3 VO 4 ) containing protease inhibitors, followed by homogenization with 20 strokes of a tight-fitting 1-ml Dounce homogenizer. After adjusting the sucrose concentration to 250 mM, postnuclear supernatants were recovered by centrifugation at 1,000 ϫ g for 5 min. The resultant postnuclear supernatants (900 l) were loaded at the top of a discontinuous sucrose gradient, composed of successive layers of 800 l of hypotonic buffer containing 1.5, 1.2, 1.0, 0.8, and 0.5 M sucrose. After centrifugation at 100,000 ϫ g for 85 min in a P55ST2 rotor (Hitachi, Tokyo, Japan), 10 fractions of 490 l were collected from the top of the tube and analyzed by Western blotting. All steps were carried out at 4°C.
Cell Viability Assay-8 ϫ 10 4 cells (per assay) were cultured on poly-HEMA-coated dishes with RPMI 1640 medium containing 0.05 or 5% BS or fibronectin-coated dishes with Iscove's modified Dulbecco's medium containing 0.05% BS. Cells were cultured for 3 days with or without M␤CD-cholesterol treatment, and the number of viable and dead cells was counted using trypan blue.
Dynasore Treatment-Cells cultured in adhesion for 2 days were treated with 100 M dynasore (Abcam) or DMSO (solvent control) for 2 h and then cultured in suspension for further 2 h with 100 M dynasore or DMSO.
Tumor Xenograft Experiment-6-Week-old female athymic BALB/c nude mice were obtained from Japan SLC, Inc. (Shizuoka, Japan). HeLa S3/Lyn-FH cells cultured in suspension for 2 days were injected subcutaneously into both flanks or intraperitoneally into the abdominal cavity. Seven days later, the mice were sacrificed to obtain the established tumors. The tumor cells were subjected to sucrose density gradient fractionation. All experiments and procedures were approved by the Chiba University Institutional Animal Care and Use Committee.

Activation of Lyn Tyrosine Kinase upon Cell Detachment-
To examine the effect of cell detachment on the activity of Src family kinases, we used HeLa S3 cells, a variant of HeLa cells that can proliferate in both adherent and suspension cultures. Autophosphorylation levels of endogenous Src family kinases were increased in suspension, although several autophosphorylation bands of the Src family members were overlapped in both culture conditions (Fig. 1A). To distinguish each member of the Src family kinases, we stably overexpressed epitopetagged c-Src, Lyn, and Fyn in HeLa S3 cells. Consistent with the results that cell detachment increases the kinase activity of endogenous c-Src (20,33), the autophosphorylation level of overexpressed c-Src (c-Src-HA) was increased in suspension culture (Fig. 1B). Intriguingly, we found that stably overexpressed Fyn (Fyn-FH) and Lyn (Lyn-FH) also increased their autophosphorylation levels in suspension culture, compared with those in adherent culture ( Fig. 1, C and D). Immunoprecipitation of endogenous p53/56Lyn substantiated that the autophosphorylation levels of endogenous Lyn (p53 and p56) are increased in suspension culture (Fig. 1E). In addition, an increase in the autophosphorylation level of Lyn was observed upon cell detachment by trypsinization or metal chelation (Fig.  1F). A time course analysis showed that the autophosphorylation level of Lyn was greatly increased within 10 min after cell detachment (Fig. 1G). In vitro kinase assays, using acid-denatured enolase as an exogenous substrate, verified that the kinase activity of Lyn is increased in suspension culture (Fig. 1H). These results indicate that the kinase activity of Lyn is increased upon cell detachment.
Shift of the Main Peak of Lyn Distribution to the High Density Fractions upon Cell Detachment-To examine whether cell detachment changed the characteristics of the membrane segments that harbor Lyn, we compared Lyn distributions in sucrose density gradient fractionation between adherent and suspension cultures. The membrane distribution of endogenous Lyn in adherent cells showed the bimodal peaks as follows: the main peak is located in Fr. 4 and the other is in Fr. 7 and 8 ( Fig. 2A). Intriguingly, the main peak of Lyn distribution was shifted from Fr. 4 to Fr. 7 and 8 in suspension culture, and it returned to Fr. 4 when the suspension cells were recultured in adhesion ( Fig. 2A). Like endogenous Lyn, the main peak of the distribution of stably overexpressed Lyn was shifted from Fr. 4 to Fr. 7 and 8 after cell detachment (Fig. 2C). As with an increase in Lyn activity (Fig. 1F), a change in Lyn distribution occurred within 10 min after cell detachment (Fig. 2C). Meanwhile, the distributions of the plasma membrane protein desmoglein, the Golgi protein ␤-1,4-galactosyltransferase, the endosomal protein transferrin receptor, and the endoplasmic reticulum protein calnexin (CNX) were largely unchanged between adherent and suspension cultures (Fig. 2B). These results suggest that the density of the membrane segments that harbor Lyn drastically changes in HeLa S3 cells upon cell detachment.
Furthermore, in HCT116 human colorectal cancer cells, the main peak of Lyn distribution was shifted from Fr. 4 to Fr. 7 and 8 upon cell detachment (Fig. 2D). In THP-1 human monocytic cells, the main peak of Lyn distribution was located in Fr. 4 during suspension culture in the presence of 5% FBS and also shifted to Fr. 8 when the cells were cultured in serum-free medium (Fig. 2E). These results suggest that the change in Lyn distribution between Fr. 4 and Fr. 7 and 8 is commonly seen among different cell types.
To examine Lyn distribution in tumor cells in vivo, HeLa S3/Lyn-FH cells cultured in suspension were implanted subcutaneously into a nude mouse (Fig. 2F). After xenograft tumors of HeLa S3/Lyn-FH cells were formed in vivo, we collected them for analyzing Lyn distribution in sucrose density gradient fractionation. We found that the main peak of Lyn distribution in xenograft tumors was shifted from Fr. 7 and 8 to Fr. 4 ( Fig.  2G), which was similar to the result of re-adhesion in vitro ( Fig.  2A). This result suggests that the formation of solid tumors in vivo is equivalent to the in vitro re-adhesion of suspension cells to culture dishes in lowering the density of the membrane segments that harbor Lyn. In addition, we also injected HeLa S3/Lyn-FH cells cultured in suspension into the abdominal cavity in order to attempt to incubate suspension cells in vivo, but we could not obtain them from the abdominal cavity.
Lack of Association between the Peak Shift in Lyn Distribution and the Trafficking of Lyn-Newly synthesized Lyn is accumulated in the Golgi region and then trafficked to the plasma membrane (24). To examine the effect of Lyn trafficking on the membrane distribution of Lyn, we transiently transfected HeLa S3 cells with Lyn-WT and then compared the Lyn distribution between 12 and 18 h post-transfection. Despite the trafficking of Lyn from the Golgi region to the plasma membrane ( Fig. 3A), the main peak of the membrane distribution of Lyn was invariably located in Fr. 4 in adherent culture (Fig. 3D). Upon cell detachment, the main peak of Lyn distribution was in part shifted from Fr. 4 to the high density fractions (Fig. 3E). To further substantiate that Lyn trafficking was unrelated to the change in Lyn distribution upon cell detachment, we used the protein synthesis inhibitor cycloheximide (CHX), which reduces the amount of newly synthesized Lyn in the Golgi region (Fig. 3, B and C) (24). Reduction of Golgi-localized Lyn by CHX did not change the major peak of Lyn distribution in both adherent and suspension cultures (Fig. 3, D and E). These results suggest that Lyn trafficking is not associated with the change of Lyn distribution upon cell detachment.

Role of Palmitoylation in the Membrane Distribution of Src
Family Kinases-Our previous studies showed that the membrane distribution of Lyn differs from that of c-Src (10,11). We then examined whether cell detachment affected the distribution of individual Src family kinase members in sucrose density gradient fractionation. Like Lyn, the main peaks of the distributions of endogenous c-Yes and stably overexpressed Fyn (Fyn-FH) were located in Fr. 4 in adherent culture and shifted to Fr. 7 and 8 in suspension culture (Fig. 4A). In contrast, the main peak of the membrane distribution of overexpressed c-Src (c-Src-HA) was located in Fr. 1 and 2 in both adherent and suspension cultures (Fig. 4A). Because Lyn, c-Yes, and Fyn, except for c-Src (a nonpalmitoylated kinase), are palmitoylated at their SH4 domains, we hypothesized that palmitoylation is critical for the distribution change of Lyn upon cell detachment. To test this hypothesis, we used Lyn(C3S), a Lyn mutant that is not palmitoylated (Fig. 4B) (11). Similar to the result of c-Src (Fig. 4A), the main peak of the distribution of Lyn(C3S) was located in Fr. 1 and 2 in both adherent and suspension cultures (Fig. 4C). To further substantiate the importance of palmitoylation in Lyn distribution, we transfected HeLa S3 cells with Lyn(SH4)-GFP, a GFP chimeric mutant that is fused with the Lyn SH4 domain at the N terminus (Fig. 4B). A peak of the membrane distribution of Lyn(SH4)-GFP was located in Fr. 4 and 5 in adherent culture and shifted to Fr. 7-9 in suspension culture (Fig. 4D). These results were similar to those of endogenous Lyn and overexpressed Lyn (Fig. 2, A and C), although approximately one-third of Lyn(SH4)-GFP somehow resided in Fr. 1 and 2 in both adherent and suspension cultures. Thus, these results suggest that palmitoylation is important for the change in the membrane distribution of Src family kinases upon cell detachment.
Effect of Cholesterol Depletion on Lyn Distribution-Cell detachment is known to decrease the amount of cholesterol in the plasma membrane ( Fig. 5A) (17). To examine whether cholesterol depletion from the plasma membrane affected Lyn distribution upon cell detachment, we treated adherent cells with the cholesterol-depleting reagent methyl-␤-cyclodextrin (M␤CD) in serum-free medium (Fig. 5, B-D). Intriguingly, M␤CD treatment increased the amount of Lyn in Fr. 5 and 6 and decreased that in Fr. 3 and 4 (Fig. 5B), although it did not seem to change the distributions of the membrane markers and the microscopic localization of Lyn (Fig. 5 SEPTEMBER 19, 2014 • VOLUME 289 • NUMBER 38 JOURNAL OF BIOLOGICAL CHEMISTRY 26331 trols). Next, we treated suspension cells with M␤CD-cholesterol inclusion complexes for cholesterol incorporation into cell membranes (Fig. 5, E-G). Incorporation of cholesterol transferred some of the amounts of Lyn from Fr. 5 and 6 to Fr. 2 and 3, although it did not change the microscopic localization of Lyn (Fig. 5, E and G). Incorporation of cholesterol also shifted the membrane distribution of calnexin to the lower density fractions (Fr. 1-4) (Fig. 5E), suggesting that cholesterol incorporation may reduce the density of overall membranes. These results suggest that cholesterol depletion from the plasma membrane may contribute to a change in Lyn distribution to some extent in suspension cells.

Membrane Density and Lyn Activation upon Cell Detachment
Effect of Dynamin Inhibition on Lyn Distribution-Because cell detachment is also known to change the localization of caveo-   SEPTEMBER 19, 2014 • VOLUME 289 • NUMBER 38

Membrane Density and Lyn Activation upon Cell Detachment
lin (18), we compared the membrane distributions between caveolin and Lyn. Cell detachment shifted the main peak of caveolin distribution from Fr. 8 to Fr. 4 and 5 (Fig. 6A). In Fr. 4 and 5, the peaks of the membrane distributions of the recycling endosomal proteins transferrin receptor and Rab11 were located in both adherent and suspension cultures (Figs. 2B and 6B). Following cell detachment, caveolin was transported from the plasma membrane to Rab11-positive endosomes (Fig. 6C), in agreement with previous studies (18). These results suggest that cell detachment induces the translocation of caveolin into the membrane segments that constitute recycling endosomes. In contrast to the caveolin distribution, the main peak of Lyn distribution was shifted from Fr. 4 to Fr. 7 and 8 upon cell detachment (Fig. 2, A and C). We treated suspension cells with CHX to accumulate Lyn in the plasma membrane (Fig. 3, B and C) and examined whether Lyn was internalized upon cell detachment. Lyn present in the plasma membrane did not appear to be internalized or co-localized with caveolin upon cell detachment (Fig. 6D), suggesting that upon cell detachment the process of the change in Lyn distribution may be different from that in caveolin distribution.
The internalization of caveolin by cell detachment requires dynamin-mediated membrane fission (18). To examine whether Lyn distribution was affected by dynamin-mediated internalization, dynamin activity was inhibited using dynasore during cell detachment. Dynasore treatment increased the amount of caveolin in Fr. 7 and 8 (Fig. 6E) and decreased the accumulation of caveolin in the perinuclear region (Fig. 6F), confirming that dynamin activity was required for the caveolin internalization from the plasma membrane to recycling endosomes. It should be underscored that dynasore treatment increased the amount of Lyn in Fr. 4 and 5 in suspension cells   (Fig. 6, G and H), although the microscopic localization of Lyn and the membrane marker distributions were largely unchanged (Fig. 6, F and G; see also Fig. 2B as controls). These results suggest that the change in Lyn distribution involves dynamin activity upon cell detachment. bimodal peak locations of Lyn in Fr. 4 and Fr. 7 (Fig. 7, A and C), similar to those in Lyn distribution in adherent culture (Fig.  7C). Upon treatment of suspension cells with dynasore and cholesterol, the distributions of the membrane markers were similar to those in cholesterol-treated suspension cells ( Fig. 7A; see Fig. 5E) and the microscopic localization of Lyn was not much different from that in control cells (Fig. 7B). These results suggest that the changes in Lyn distribution may involve both dynamin activity and a decrease in membrane cholesterol upon cell detachment.

Membrane Density and Lyn Activation upon Cell Detachment
Like Lyn, the main peak of Fyn distribution was also detected in Fr. 4 in dynasore-and cholesterol-treated suspension cells ( Fig. 7D; see Fig. 4A as control), suggesting that the characteristics of the membrane segments containing Lyn are similar to those containing Fyn upon cell detachment or that Lyn and Fyn are located in the same membrane segments.
Relationship between Lyn Distribution and Lyn Activation-To examine whether the activity of Src family kinases was influenced by their membrane distributions, we compared the distributions of the autophosphorylation signal of endogenous Src family kinases in adherent cultures. Although the amounts of Lyn, c-Yes, and Fyn in Fr. 4 were lower than those in Fr. 7 and 8 in adherent culture (see Figs. 2 and 4), the autophosphorylation signal of endogenous Src family kinases in Fr. 7 and 8 was stronger than that in Fr. 4 (Fig. 8A). Furthermore, the level of the autophosphorylation of overexpressed Lyn present in Fr. 7 was higher than that in Fr. 4 in adherent culture (Fig. 8B). These results suggest that Src family kinases present in Fr. 7 and 8 have higher activity than that in Fr. 4.
Next, we transfected cells with a constitutively active form of Lyn (Lyn⌬C-HA), which lacks the negative regulatory site of Lyn. Despite the high activity of Lyn⌬C-HA (Fig. 8C), the main peak of Lyn⌬C-HA distribution was located in Fr. 4 and 5 in adherent culture (Fig. 8D). Like Lyn-WT, a fraction of Lyn⌬C-HA was shifted to Fr. 7 and 8 upon cell detachment ( Fig.  8D; see Fig. 3, D and E). These results suggest that the change in Lyn distribution may be independent of Lyn activity.
The membrane distributions of the regulators of Src family kinases, EGFR, Csk, and SHP-2, were shown in Fig. 8E. Whereas the main peaks of Csk and SHP-2 were located in Fr. 1 and 2 in both adherent and suspension cultures, the main peak of EGFR was shifted from Fr. 4 to 7 and 8 upon cell detachment as with the change in Lyn distribution. Although Lyn, EGFR, and CNX (see Fig. 2B) were detected in Fr. 7 and 8 in suspension cells, Lyn and EGFR, but not CNX, were co-localized mainly on the plasma membrane (Fig. 8F). These results suggest that EGFR is included in the membrane segments that harbor Lyn upon cell detachment. Thus, we hypothesize that activation of Lyn upon cell detachment takes place via the change in the characteristics of the membrane segments.
Inhibitory Role of Cholesterol in Lyn Activity upon Cell Detachment-The change in Lyn distribution upon cell detachment is suggested to involve cholesterol depletion and dynamin activity (Figs. 5-7). To examine how Lyn is activated upon cell detachment, adherent cells were incubated in serum-free medium containing M␤CD. We revealed that cholesterol depletion drastically enhanced the autophosphorylation level of Lyn after serum stimulation (Fig. 9A). In sharp contrast, cholesterol incorporation decreased the autophosphorylation level of Lyn in suspension cells, and the decreased level was not recovered until at least 23 h after serum stimulation (Fig. 9B). The increase in the autophosphorylation level of Lyn upon cell detachment was not affected by inhibition of dynamin activity (Fig. 9C). These results suggest that, upon cell detachment, cholesterol depletion from the plasma membrane plays a critical role in the change in Lyn distribution and the increase in Lyn activity.
Effect of Cholesterol Incorporation on the Viability of Suspension Cells-HeLa S3 cells can proliferate in suspension culture (31,34), although suspension culture is known to induce apoptosis in epithelial cells (35). Because the activity of Src family kinases is involved in survival of suspension cells (8,20,33,36), we established a HeLa S3 cell line stably expressing shRNA against Lyn (Fig. 10A). The autophosphorylation level of endogenous Src family kinases in Lyn knockdown cells was decreased in adherent culture but was not in suspension culture (Fig. 10B). Because Fyn was also activated upon cell detachment (Fig. 1, C  and D), Lyn knockdown might be compensated by activated Fyn in suspension culture. We thus established Lyn/Fyn double-knockdown cells, in which the autophosphorylation level of endogenous Src family kinases was obviously decreased in both adherent and suspension cultures (Fig. 10, A and C). In suspension culture, control cells and Lyn/Fyn double-knockdown cells were fully viable in medium containing 5% serum but underwent cell death in serum-free medium (Fig. 10D). However, in medium containing 0.05% serum, the viability of Lyn/Fyn double-knockdown cells was lower than that of control cells in suspension culture (Fig. 10D).
Next, we examined whether inactivation of Lyn by cholesterol incorporation affected cell viability in suspension culture. Incorporation of cholesterol into HeLa S3 cells in suspension culture drastically decreased the viability of cells cultured in medium containing 0.05% serum (Fig. 10E). However, the decrease in viability of suspension cells was attenuated by Lyn expression in medium containing 0.05% serum or by culture in medium containing 5% serum (Fig. 10E). In addition, the viabil- ity of adherent cells, which were seeded on fibronectin-coated dishes, in medium containing 0.05% serum was not affected by cholesterol incorporation (Fig. 10F). These results suggest that cholesterol incorporation causes decreased viability of suspension cells through inhibition of Src family kinases at low concentrations of serum.

DISCUSSION
In this study, we show that cell detachment changes the membrane distribution of palmitoylated Src family kinases from the low density to the high density membrane fractions and activates their kinase activity. The change in Lyn distribution upon cell detachment involves both dynamin activity and a decrease in membrane cholesterol levels. Cell detachment activates palmitoylated Src family kinases, such as Lyn and Fyn, through decreased membrane cholesterol levels during the change in their membrane distribution.
Cellular membranes consist not of a monotonous structure but a mass of segments that are composed of various proportions of lipids and proteins (3,5). Density gradient fractionation is a method to separate small membrane particles prepared from cellular membranes into different fractions based on their densities (1). Because membrane particles are prepared without detergent, the distribution of a membrane-anchored protein of interest can be determined by the densities, or the constitutions, of the membrane segments that harbor the protein of interest. The peak position of the membrane distribution of the protein of interest would reflect the expected value of the density of the membrane segments. Membrane protein markers for the plasma membrane, the Golgi, the endoplasmic reticulum, and endosomes show their specific distributions ranging between the low density and the high density fractions (Fig. 2B). Because density gradient fractionation was capable of detecting the cell detachment-induced change in the membrane distribution of Lyn, which our microscopic analyses were unable to recognize (Figs. 2-7), density gradient fractionation is an appropriate method to scrutinize the characteristics of proteinanchored membrane segments.
The membrane distribution of Lyn shows the bimodal peaks located in the low density fraction (Fr. 4) and the high density fractions (Fr. 7 and 8) in our density gradient fractionation assays. We initially hypothesized that the bimodal peaks are attributed to the microscopic localizations of Lyn to the Golgi and the plasma membrane, because newly synthesized Lyn is first accumulated to the cytoplasmic side of the Golgi membranes and then trafficked to the plasma membrane (24). However, we found that the observations of the microscopic localizations of Lyn are not linked to the main peak of Lyn distribution in Fr. 4 in adherent culture and the peak shift to Fr. 7 and 8 in suspension culture (Fig. 3). Given that Lyn, a palmitoylated Src family kinase, and c-Src, a nonpalmitoylated Src family kinase, are distributed differently in density gradient fractionation (10,11), we then hypothesized that the bimodal peaks are explained by palmitoylation of Lyn. To verify the hypothesis, we compared the membrane distributions of Lyn(C3S) and Lyn(SH4)-GFP with that of Lyn-WT. Our results suggest that palmitoylation is required for Lyn distribution between the low density (Fr. 4) and the high density fractions (Fr. 8) but is not attributed to the bimodal peaks (Fig. 4). These results implicate that the bimodal peaks of Lyn may be due to the difference in constitutions of the membrane segments.
Cell detachment is known to induce the translocation of proteins, such as Bax and YAP (37,38), and to internalize cholesterol (17) and caveolar vesicles (18) from the plasma membrane. Cholesterol accumulation in the plasma membrane changes the shape of erythrocytes from spherical to flat (39), and cholesterol depletion changes the shape of cells on fibronectin from flat to spherical (40). Considering that cell detachment induces cell rounding, probably due to the surface tension (41), we speculate that the spherical change in detached cells would induce the dissociation of cholesterol from the plasma membrane. In fact, cholesterol depletion by M␤CD, which would form high density membranes, increases the amount of Lyn in Fr. 5 and 6 by decreasing the amount of Lyn in Fr. 2-4 (Fig. 5). However, treatment with dynasore during cell detachment shows that dynamin activity is partly involved in the shift of the main peak of Lyn distribution from Fr. 4 and 5 to Fr. 7 and 8 (Fig. 6). Because cell detachment increases the amount of Lyn in Fr. 6 -8 by decreasing the amount of Lyn in Fr. 2-4, we consider that cholesterol depletion and dynamin activity are both involved in the change in Lyn distribution upon cell detachment. Actually, the membrane distribution of Lyn in suspension cells treated with cholesterol and dynasore was similar to that in adherent cells (Fig. 7). Given that caveolae accumulate some membrane molecules in the plasma membrane before the inward budding of caveolar vesicles (42), we assume that the FIGURE 8. Relationship between Lyn distribution and Lyn activation. A and E, HeLa S3 cells cultured in adhesion or suspension for 2 days were subjected to sucrose density gradient fractionation, and each fraction was analyzed by Western blotting (WB) with the indicated antibodies. B, HeLa S3/Lyn-FH cells cultured in adhesion were separated by sucrose density gradient fractionation. Fr. 4 and 7 of the fractions were subjected to Western blot analysis with the indicated antibodies. Autophosphorylation levels of Lyn-FH were quantitated as described in Fig. 1D. Results were expressed relative to the autophosphorylation level of Lyn-FH in Fr. 4. The data represent the mean Ϯ S.D. from three independent experiments. C, whole cell lysates of HeLa S3/Lyn-FH cells and HeLa S3 cells transfected with Lyn-WT or Lyn⌬C-HA were subjected to Western blot analysis with the indicated antibodies. D, HeLa S3 cells transfected with Lyn⌬C-HA were cultured for 18 h. Cells were cultured in adhesion or in suspension for the last 2 h of culture. Cells were subjected to sucrose density gradient fractionation, and each fraction was analyzed by Western blotting with anti-Lyn antibody. Molecular size markers are shown in kDa. F, HeLa S3/Lyn-FH cells cultured in suspension for 2 days were doubly stained with anti-Lyn (green) and anti-EGF receptor or anti-CNX (red) antibodies. Scale bars, 10 m. change in Lyn distribution upon cell detachment could be attributed to elimination of some components of caveolar vesicles in conjunction with cholesterol depletion from the membrane segments that harbor Lyn.
Cell detachment changes the main peak of Lyn distribution from Fr. 4 to 7 and 8 through cholesterol depletion and dynamin activity (Figs. 5-7). We show that Lyn is activated not by dynamin-mediated vesicle budding but by cholesterol depletion upon serum stimulation (Fig. 9). As with the change in Lyn distribution, the main peak of EGFR distribution changes from Fr. 4 to 7 and 8 upon cell detachment (Fig. 8). EGFR is stimulated by serum containing growth factors, thereby activating Src family kinases (43). EGFR is reported to be activated by cholesterol depletion but be inactivated by cholesterol incorporation (44). Taken together, these findings raise the intriguing possibility that cell detachment decreases the cholesterol content in the membrane segments that harbor Lyn and EGFR to promote the activation of Lyn upon serum stimulation.
In adherent cells, the main peak of Lyn distribution was present in Fr. 4 rather than Fr. 7 and 8 ( Fig. 2A). Importantly, the level of the activity of Src family kinases, including Lyn, present in Fr. 7 was higher than that in Fr. 4 even in adherent cells (compare Fig. 8A with 2A and Fig. 8B), which suggests that decreased membrane cholesterol plays a significant role in Lyn FIGURE 10. Effect of cholesterol incorporation on the viability of suspension cells. A-C, HeLa S3 cells stably expressing shRNAs against luciferase (Luci), Lyn, Fyn, or Lyn plus Fyn were cultured in adhesion or suspension for 2 days. Whole cell lysates were subjected to Western blot analysis (WB) with the indicated antibodies. Molecular size markers are shown in kDa. D-F, cell viability was estimated by trypan blue staining. The data represent the means Ϯ S.D. from three independent experiments. Asterisks indicate significant differences (***, p Ͻ 0.001; **, p Ͻ 0.01; NS, not significant) calculated by Student's t test. D, HeLa S3 cells stably expressing shRNAs against luciferase or Lyn plus Fyn were cultured in suspension for 3 days with serum-free medium or medium containing 0.05 or 5% serum. E, parental HeLa S3 or HeLa S3/Lyn-FH cells were cultured in suspension with medium containing 0.05 or 5% serum for 2 days. The cells were treated with M␤CD-cholesterol for 1 h and cultured in suspension with medium containing 0.05 or 5% serum for 23 h. F, HeLa S3 cells cultured on fibronectin-coated dishes for 2 days with medium containing 0.05% serum were treated with M␤CD-cholesterol for 1 h and cultured in adhesion with medium containing 0.05% for 23 h.
activation. Considering that a decrease in membrane cholesterol shifted Lyn distribution to higher density fractions, we hypothesize that an increase of membrane fluidity by decreased membrane cholesterol may augment the friction between individual Lyn molecules in higher density membrane fractions, thereby increasing the level of Lyn autophosphorylation for its activation.
Unlike Lyn, the main peak of c-Src distribution is found in Fr. 1 and 2 in both adherent and suspension cultures (Fig. 4), and two regulatory molecules of Src family kinases, SHP-2 and Csk (45,46), are also found in Fr. 1 and 2 (Fig. 8). Given that SHP-2 is reported to regulate c-Src activation upon cell detachment (20), we hypothesize that activation of Lyn and c-Src is regulated differently upon cell detachment.
In the absence of cell-matrix interactions, epithelial cells undergo apoptosis, referred to as anoikis (35). To survive in the circulation, metastatic cells are considered to acquire anoikis resistance through the epithelial-mesenchymal transition (47,48). Src family kinases, especially Lyn (49,50), are capable of promoting several aspects of tumor progression and metastasis (6). Following cell detachment, cholesterol depletion from the plasma membrane changes the characteristics of Lyn-anchored membrane segments to enhance the kinase activity of Lyn upon serum stimulation (Fig. 9), and incorporation of cholesterol into cells in suspension decreases both Lyn activity and cell viability (Fig. 10). Although the relationship between the level of serum cholesterol and cancer mortality is controversial (51,52), our results implicate that inhibition of palmitoylated Src family kinases through cholesterol incorporation into Lyn-anchored membrane segments might be important for a novel cancer therapy.