A requirement for membrane cholesterol in the β-arrestin- and clathrin-dependent endocytosis of LPA₁ lysophosphatidic acid receptors

Lysophosphatidic acid (LPA, 1-acyl-2-lyso-sn-glycero-3-phosphate) is an abundant serum mitogen that evokes growth-factor-like responses in many cell types through activation of G-protein-coupled receptors (GPCR) (Moolenaar, 1999). LPA signaling affects a variety of cellular functions including: growth stimulation (cell proliferation and cell survival) (Fang et al., 2000; Goetzl et al., 2000; van Corven et al., 1992), induction of cytoskeletal rearrangements via Rho GTPases (Ridley and Hall, 1992), stimulation of serum-responsive genes (Hill et al., 1995), neurite retraction (Jalink et al., 1994), promotion of tumor cell migration/invasion (Stam et al., 1998) and the secretion of peptide growth factors (Hu et al., 2001; Pustilnik et al., 1999; Schwartz et al., 2001). Most of the effects of LPA are mediated through the activation of three members of the endothelial differentiation gene superfamily of receptors: LPA₁, LPA₂ and LPA₃ (Hla et al., 2001; Moolenaar, 1999). Upon LPA binding, both LPA₁ and LPA₂ activate the G_i, G_q and G_12/13 families of heterotrimeric G proteins; LPA₃ only activates G_i and G_q (Ishii et al., 2000). In addition to these well-characterized GPCRs, LPA also stimulates the orphan receptor, GPR23/LPA₄ (Noguchi et al., 2003) and the non-GPCR target, peroxisome proliferator-activated receptor γ (McIntyre et al., 2003). Given the complexity of cellular responses to LPA signaling and the potential role of LPA receptor subtypes in various cancers (Mills and Moolenaar, 2003), it is important to understand the mechanisms that regulate the activity of individual LPA receptors.

Upon agonist stimulation, most GPCRs are rapidly internalized into cells through a variety of different endocytic pathways. This facilitates either receptor downregulation or receptor resensitization (Marchese et al., 2003). Agonist stimulation usually leads to the rapid phosphorylation of serine/threonine residues located within cytoplasmically exposed regions of GPCRs (Ferguson et al., 1995). This subsequently induces the binding of β-arrestin proteins, which results in signal attenuation and often targets the GPCR to clathrin-coated pits for endocytosis (Lefkowitz and Shenoy, 2005). Internalized GPCRs transit through the endosomal system and are either sorted to lysosomes for degradation or become dephosphorylated by membrane-associated phosphatases and are recycled back to the plasma membrane (Ferguson, 2001). In addition to clathrin-mediated endocytosis, many GPCRs utilize a variety of clathrin-independent internalization mechanisms including cholesterol-dependent pathways such as caveolae (Chini and Parenti, 2004). Also, β-arrestins are not universally required for GPCR endocytosis as shown for the thrombin receptor, PAR1, whose association with β-arrestins is required for signal attenuation but not for its endocytosis (Paing et al., 2002). Thus, the mechanisms that regulate both signal attenuation and receptor endocytosis can vary from one GPCR to another.

We have previously shown that LPA₁ is probably internalized by clathrin-dependent endocytosis as dominant-negative mutants of dynamin 2 (K44A) and Rab 5 (S34N), which regulate clathrin-dependent trafficking, strongly inhibited LPA₁ endocytosis (Murph et al., 2003). However, a recent study showed that LPA stimulation of the phosphoinositide 3-kinase (PI3-K)/Akt pathway was dependent upon membrane cholesterol (Peres et al., 2003) suggesting a positive role for cholesterol-rich plasma membrane microdomains in LPA signaling. As cholesterol-enriched microdomains, such as caveolae, can also mediate receptor endocytosis, it is not clear what the relationship is between LPA signaling from cholesterol-rich microdomains and the endocytosis of LPA receptors. To address this question and to gain a better understanding about the regulation of LPA receptors, we investigated the role of membrane cholesterol, β-arrestins and clathrin in the signaling and endocytosis of the ubiquitously-expressed LPA₁ receptor.

Lysophosphatidic acid (1-oleoyl-2-hydroxy-sn-glycero-3-phosphate; LPA) was purchased from Avanti Polar Lipids (Alabaster, AL). Isoproterenol and cytochalasin D was obtained from Sigma Chemical Co. (St Louis, MO) and carbachol from Fluka Chemika-Biochemika. FLAG-tagged LPA₁ receptors were detected with mouse anti-FLAG antibodies (Sigma, St Louis, MO); HA-tagged β₂AR and HA-tagged M₁ muscarinic acetylcholine receptor (mAChR) were detected with mouse anti-HA antibodies (Covance, Berkeley, CA). Alexa 488-labeled transferrin (Alexa 488-Tfn), Alexa 594- and Alexa 488-conjugated goat anti-mouse were purchased from Molecular Probes (Eugene, OR). Monoclonal antibodies to clathrin heavy chain and monoclonal anti-actin antibodies were purchased from BD Transduction labs (San Jose, CA) and Santa Cruz Biotechnology (Santa Cruz, CA), respectively. Mouse anti-AP2 antibodies were purchased from Affinity Bioreagents (Golden, CO). FITC-labeled anti-CD59 was obtained from Chemicon (Temecula, CA). Methyl-β-cyclodextrin and water-soluble cholesterol complexes were purchased from Sigma. myo-[³H]inositol was purchased from American Radiolabeled Chemicals (St Louis, MO).

HeLa cells stably expressing the LPA₁ receptor (termed LPA₁/HeLa cells), native HeLa cells, wild-type (WT) MEF and β-arrestin 1/2 KO MEF cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 100 IU/ml penicillin, 100 μg/ml streptomycin (Media Tech, Herndon, VA) and 1 mM sodium pyruvate (Biosource International, Camarillo, CA) at 37°C with 5% CO₂. Cells were grown on glass coverslips (for immunolocalization) and transfected in six-well dishes, or were grown in 24-well dishes (for myo-[³H]inositol labeling) using Lipofectin or Lipofectamine 2000 (Invitrogen, Carlsbad, CA) according to manufacturer's directions. Plasmids encoding HA-tagged β₂AR, β-arrestin 1-GFP, β-arrestin-2-GFP, HA-tagged M₁ mAchR were transiently transfected at 1.0 μg/well (in six-well plates) and have been previously described (Paing et al., 2002; Scott et al., 2002).

siRNA oligonucleotides to clathrin were purchased from Dharmacon (Lafayette, CO) and have been described previously (Motley et al., 2003). LPA₁/HeLa cells were transiently transfected with 300 pmol (10 cm dish) or 100 pmol (24-well plate) of siRNA using Lipofectamine 2000 (Invitrogen, Carlsbad, CA). The transfection medium was replaced with complete medium (without penicillin/streptomycin) 5 hours later and the cells were incubated for 16 hours. The cells were transfected a second time as above and the medium was then replaced with serum-free medium (SFM) and incubated for an additional 16 hours before experimentation.

Cells were treated as described in the figure legends, 24-48 hours after transfection. Cells were then fixed in 2% formaldehyde in phosphate-buffered saline (PBS) for 10 minutes and rinsed with 10% fetal bovine serum (FBS) containing 0.02% azide in PBS (PBS-serum). Fixed cells were incubated with primary antibodies diluted in PBS-serum containing 0.2% saponin for 45 minutes and then washed (three times, 5 minutes each) with PBS-serum. The cells were then incubated in fluorescently labeled secondary antibodies diluted in PBS-serum containing 0.2% saponin for 45 minutes, washed three times with PBS-serum, washed once with PBS and mounted on glass slides as previously described (Murph et al., 2003).

For Alexa 594-Tfn and FITC-labeled anti-CD59 internalization, LPA₁/HeLa cells were briefly rinsed three times with 0.5% bovine serum albumin (BSA) in SFM and incubated in the same medium for 30 minutes at 37°C. The cells were then incubated with Alexa 594-conjugated human transferrin (50 μg/ml) or FITC-conjugated anti-CD59 (1 μg/ml) for 30 minutes at 37°C in the presence or absence of 10 μM LPA. Antibodies bound to the cell surface were removed by rinsing the cells with 0.5% acetic acid, 0.5 M NaCl, pH 3.0 solution (for Alexa 488-Tfn) or 100 mM glycine, 20 mM magnesium acetate, 50 mM KCl, pH 2.2 (for FITC anti-CD59) (Naslavsky et al., 2004). Cells were rinsed in complete medium, fixed and processed for fluorescence microscopy. For assessing uptake in the presence of methyl-β-cyclodextrin (5 mM) or nystatin (50 μg/ml), cells were pre-treated with DMEM supplemented with 0.5% BSA, with or without drugs, for 60 minutes prior to antibody and/or LPA addition. All images were acquired using an Olympus BX40 epifluorescence microscope equipped with a 60× Planapo lens and photomicrographs were prepared using an Olympus MagnaFire SP digital camera (Olympus America, Melville, NY). Images were processed with Adobe Photoshop 6.0 software.

Stably-transfected LPA₁/HeLa cells or transiently-transfected HeLa cells expressing M₁ mAChRs were grown on glass coverslips and treated with MβCD and/or water-soluble cholesterol as described in the figure legends. The cells were then incubated with 50 μg/ml Alexa 594-Tfn for 30 minutes in the presence or absence of 10 μM LPA or 1 mM carbachol, respectively. The cells were rinsed with a mild acid wash as described above, fixed with 2% formaldehyde in PBS and processed for immunofluorescence localization of LPA₁ using M1 mouse anti-FLAG IgG or M₁ mAChR using mouse anti-HA IgG followed by Cy2 secondary antibodies. The extent of LPA₁ or M₁ colocalization with internalized Alexa 594-Tfn was determined by quantifying the extent of pixel colocalization of GPCR staining with Alexa 594-Tfn fluorescence using Metamorph Imaging software (Universal Imaging, West Chester, PA) as described (Murph et al., 2003; Volpicelli et al., 2001). The background was subtracted from unprocessed images and the percentage of GPCR pixels that overlapped with Alexa-Tfn pixels was measured. The data is presented as the mean±s.e.m. of measurements from 20 cells per sample from a representative experiment that was performed three independent times with similar results.

Following 72 hours of siRNA treatment, cells were solubilized by addition of lysis buffer (1% NP-40, 1% sodium deoxycholate, 0.1% SDS, 0.15 M NaCl, 0.01 M sodium phosphate pH 7.2, 2 mM EDTA, 50 mM NaF, 0.2 M sodium orthovanadate, 0.02% azide, 100 μg/ml leupeptin and 0.1 mM PMSF) and incubated on ice for 60 minutes. The samples (12 μg protein per lane) were then separated by 10% SDS-PAGE and transferred to nitrocellulose. Clathrin heavy chain was detected using mouse anti-clathrin antibodies and actin was detected using monoclonal anti-actin antibodies. The binding of primary antibodies was detected by using an enhanced chemifluorescence detection kit (Amersham Biosciences, Piscataway, NJ).

LPA₁/HeLa cells or mouse embryo fibroblasts derived from wild-type or β-arrestin 1/2 null mice were plated at a density of 4.0×10⁴ cells/well into 24-well plates and transfected with plasmids encoding wild-type LPA₁ or M1 mAChRs alone or in combination with plasmids encoding wild-type β-arrestin 2 using Lipofectamine 2000. Transient transfection of plasmids encoding M₁ mAChRs was performed by using Lipofectin reagent. At 24 hours post-transfection, cells were labeled overnight with myo-[³H]inositol in inositol- and serum-free medium, treated as described in the figure legends and then processed for analysis of phosphoinositide hydrolysis by anion exchange chromatography as described (Paing et al., 2002).

LPA₁/HeLa cells were plated onto glass coverslips in 35 mm dishes at a density of 0.2×10⁶ cells per plate. After allowing cells to attach for 24 hours, the medium was changed to serum-free medium and the cells were incubated overnight (∼16 hours). The following day, the cells were treated as described in the figure legends and subsequently incubated with ice-cold 1% Triton X-100 in PBS on ice for 3 minutes prior to fixation with ice-cold 2% formaldehyde in PBS. LPA₁ or was localized using indirect immunofluorescence microscopy. To monitor the fate of surface LPA₁, LPA₁/HeLa cells were incubated on ice with mouse anti-FLAG antibodies for 30 minutes after LPA treatment to label only surface LPA₁. These cells were then extracted with 1% Triton X-100, fixed and processed for immunofluorescence localization as described above. Relative receptor expression was quantified by measuring receptor pixel intensity using MetaMorph imaging software and was normalized to DNA content, labeled with Hoescht dye.

LPA₁/HeLa cells were plated in 24-well dishes (Falcon) at a density of 0.4×10⁵ cells per well and grown overnight. Cells were then transiently transfected with no siRNA or with 100 pmol/well of clathrin-specific siRNA (Motley et al., 2003) using Lipofectamine 2000 reagent according to the manufacturer's instructions (Invitrogen, Carlsbad, CA). After 24 hours, the cells were again transfected with 100 pmol/well of clathrin-specific siRNA or no siRNA. 24 hours later, the cells were incubated in the presence or absence of 10 μM LPA for 45 minutes and fixed in 2% formaldehyde in phosphate-buffered saline (PBS) for 10 minutes and rinsed with 10% fetal bovine serum (FBS), containing 0.02% azide, in PBS (PBS-serum). Fixed cells were incubated with mouse anti-M1 FLAG primary antibody diluted in PBS-serum (250 μl/well) for 1 hour and then washed (three times, 5 minutes each) with PBS-serum. The cells were then incubated with horseradish peroxidase-conjugated goat anti-mouse IgG secondary antibody (Pierce Biotechnology, Rockford, IL) diluted in PBS-serum (250 μl/well) for 1 hour, washed three times with PBS-serum and washed three times with PBS. The cells were then incubated for 1 hour at 37°C with ABTS (2,2′-Azinobis [3-ethylbenzothiazoline-6-sulfonic acid]-diammonium salt) (Pierce Biotechnology). A 200 μl aliquot was then removed from each well, transferred to a 96-well plate and the absorbance read at 405 nm (corrected for blank). Internalization is expressed as the percent difference in surface LPA₁ between unstimulated cells and agonist-stimulated cells. The data are the mean±s.e.m. of six replicates/siRNA sample combined from two independent experiments.

HeLa cells, stably expressing FLAG-tagged LPA₁, were seeded in six-well plates at a density of 0.5×10⁶ cells per well, allowed to attach overnight and then incubated with serum-free DMEM for 24 hours prior to treatment. The cells were treated for 60 minutes, as described in the figure legend, rinsed twice with ice-cold PBS (pH 7.4) and then solubilized in ice-cold PBS (pH 7.4) containing 1% Triton X-100 and protease inhibitor cocktail (2 mM AEBSF, 1 mM EDTA, 130 μM bestatin, 14 μM E-64, 1 μM leupeptin and 0.3 μM aprotinin). Total cellular cholesterol was quantified using an Amplex Red Cholesterol Assay Kit (Molecular Probes; Eugene, OR), as indicated by the manufacturer. Briefly, cholesterol esters in the cell extracts are hydrolyzed by cholesterol esterase into cholesterol, which is then oxidized by cholesterol oxidase to yield H₂O₂, which is detected using 10-acetyl-3,7-dihydroxyphenoxazine (Amplex Red). In the presence of horseradish peroxidase (HRP), Amplex Red reacts with H₂O₂ to produce fluorescent resorufin. Fluorescence was measured with a fluorescence microplate reader using excitation at 560 nm and fluorescence detection at 590 nm and total cholesterol was calculated from a standard curve using purified cholesterol. Cellular cholesterol was normalized to total protein concentration, which was quantified by BCA Protein Assay (Pierce Biotechnology).

The data is expressed as the mean±s.e.m. from the indicated number of independent experiments performed in triplicate. Differences were analyzed by two-factor ANOVA followed by a Tukey's statistical significance test.

Our previous work suggested that LPA₁ was internalized by clathrin-dependent endocytosis based on the inhibitory effects of mutant dynamin 2 K44A and Rab 5 S34N (Murph et al., 2003). However dynamin K44A can also inhibit endocytosis from cholesterol-rich caveolae (Damke et al., 1994; Henley et al., 1998; Murph et al., 2003). To directly test whether LPA₁ used a clathrin-dependent pathway, we determined the effects of reducing the cellular abundance of the clathrin heavy chain, using small interfering RNAs (siRNAs), on the endocytosis of LPA₁. Fig. 1A shows the distribution of FLAG-tagged LPA₁ in stably transfected HeLa cells (LPA₁/HeLa cells). In untreated cells, LPA₁ is localized predominantly at the plasma membrane and to a lesser extent, at the Golgi complex, which probably represents newly synthesized LPA₁ en route to the plasma membrane. Treatment with 10 μM LPA results in a redistribution of LPA₁ to numerous punctate structures, which we have previously shown to colocalize with transferrin-receptor-positive endosomes (Murph et al., 2003).

To investigate the role of clathrin in LPA₁ endocytosis, we adapted a double-transfection procedure that was previously described (Motley et al., 2003) to knock down clathrin heavy chain amounts in HeLa cells to near undetectable levels. Using this procedure, we observed a 73% reduction in the abundance of clathrin relative to mock-transfected siControl cells (Fig. 1B) in LPA₁/HeLa cells. Treatment of cells with clathrin siRNA did not alter the abundance of actin. Endocytosis of LPA₁ was quantified by using a whole-cell ELISA, which measures the agonist-induced loss of cell surface LPA₁ (Kim and Benovic, 2002; Paing et al., 2002). In siControl cells, 10 μM LPA induced LPA₁ internalization (∼30%) and this was strongly inhibited in siClathrin cells (∼2%) (Fig. 1C). In siClathrin cells, both the agonist-stimulated internalization of β₂-adrenergic receptors (β₂ARs) (data not shown), which are known to use clathrin-dependent mechanisms (von Zastrow and Kobilka, 1992) and the constitutive endocytosis of Alexa transferrin was strongly inhibited (Fig. 1D).

As a negative control, we examined the effects of clathrin knockdown on the endocytosis of anti-CD59 antibodies bound to endogenous CD59, which is internalized via cholesterol-rich, detergent-resistant membranes and then merges with a clathrin-independent trafficking pathway that is regulated by the Arf6 GTPase as shown by Naslavsky et al. (Naslavsky et al., 2004). LPA₁/HeLa cells were transfected either with or without clathrin siRNAs and then incubated with FITC-labeled mouse anti-CD59 antibodies along with Alexa 594-Tfn for 30 minutes. The cells were acid-stripped to remove surface-bound anti-CD59 antibodies and Alexa 594-Tfn. In cells transfected with clathrin siRNAs, FITC-labeled anti-CD59 antibodies localized to pleomorphic tubulovesicular structures (Fig. 1D, Anti-CD59) similar to those described (Naslavsky et al., 2004). As expected, these same siRNA-treated cells did not internalize Alexa 594-Tfn (Fig. 1D, Alexa 594-Tfn). Taken together, these results indicate that LPA₁ is internalized by clathrin-mediated endocytosis.

Clathrin-mediated endocytosis of many GPCRs is also dependent upon their association with the multi-functional β-arrestins (Ferguson et al., 1996; Marchese et al., 2003). β-arrestin binding is initiated through the agonist-induced phosphorylation of cytoplasmic serine/threonine residues in the GPCR by G protein receptor kinases (GRKs) such as GRK2 (Pitcher et al., 1998). β-arrestin binding promotes both receptor desensitization, by preventing receptor-G protein coupling and clathrin-dependent endocytosis of the receptor. To determine whether β-arrestins are required for LPA₁ endocytosis, we compared agonist-stimulated internalization of LPA₁ and β₂AR in mouse embryo fibroblasts (MEFs) derived from either wild-type or β-arrestin 1 and 2 null mice (Kohout et al., 2001) (Fig. 2). Wild-type MEFs were transiently transfected with plasmids encoding either LPA₁ or β₂ARs and then incubated in the presence or absence of agonist. In the absence of agonist treatment, both receptors were primarily localized to the plasma membrane in a diffuse pattern (Fig. 2A, untreated). Upon agonist treatment for 30 minutes, both LPA₁ and β₂ARs redistributed to small punctate endosomal structures dispersed throughout the cell. The labeling of these structures was not observed in non-permeabilized cells, thus indicating that they were internal endosomal structures (Fig. 1A, non-permeabilized). In contrast to wild-type MEFs, agonist treatment of β-arrestin 1/2 KO MEFs expressing either LPA₁ or β₂ARs did not lead to their endocytosis (Fig. 2B, +agonist). Expression of wild-type β-arrestin-2-GFP in the knockout cells restored agonist-induced endocytosis of both LPA₁ (Fig. 3A, 30 minutes) and β₂ARs (data not shown), thus indicating that β-arrestins were required for the endocytosis of LPA₁ as well as β₂ARs.

Previous studies have shown that agonist stimulation of different GPCRs leads to the translocation of cytosolic β-arrestin proteins to the plasma membrane (Barak et al., 1997). For some GPCRs such as β₂ARs, this association with β-arrestins is transient and is not observed following receptor endocytosis, whereas other GPCRs, such as angiotensin AT1a receptors and vasopressin receptors, maintain a stable association with β-arrestins even on endosomes after endocytosis (Oakley et al., 1999). To determine whether LPA₁ formed a transient or stable association with β-arrestins, we examined the distribution of LPA₁ and β-arrestin-2-GFP after 0, 2 and 30 minutes of LPA treatment (10 μM) (Fig. 3A). In untreated cells, LPA₁ localized to the plasma membrane and β-arrestin-2-GFP localized in a diffuse cytoplasmic pattern (Fig. 3A, Untreated). After 2 minutes of LPA treatment, LPA₁ localized to small punctate structures, which partially colocalized with β-arrestin-2-GFP (Fig. 3A, inset, arrows). However, many punctate structures contained LPA₁ but did not contain β-arrestin-2-GFP, particularly in the larger and more pleotropic structures. Following 30 minutes of LPA treatment, LPA₁ localized to heterogeneously sized endosomal structures, but β-arrestin-2-GFP returned to the diffuse cytoplasmic pattern observed in untreated cells (Fig. 3A, 30 minutes). This suggested that β-arrestins dissociate from LPA₁ receptors at or near the cell surface and do not form a stable association with β-arrestin proteins, as defined by Oakley et al. (Oakley et al., 1997). Taken together, these data indicate that β-arrestins are critical for the endocytosis of LPA₁ and that LPA₁ only transiently associates with β-arrestins at the cell surface.

As mentioned above, β-arrestin binding to activated GPCRs leads to signal attenuation (Lefkowitz and Whalen, 2004). We next investigated whether β-arrestins were important for the desensitization of LPA₁. We examined the ability of LPA₁ receptors, which activate G_i, G_q and G_12/13 signaling pathways (Fukushima et al., 1998), to promote phosphoinositide (PI) hydrolysis, via G_q stimulation of phospholipase C, in the wild type and β-arrestin 1/2 KO MEFs (Fig. 3B). LPA₁-transfected wild type and β-arrestin 1/2 KO MEFs were labeled with [³H]inositol and the accumulation of inositol phosphates was determined in untreated cells and cells treated with 10 μM LPA for 60 minutes at 37°C. LPA treatment increased the accumulation of [³H]inositol phosphates in wild-type MEFs by ∼2.5-fold. However, stimulation of LPA₁-transfected β-arrestin 1/2 KO MEFs led to a 4.3-fold increase in inositol phosphate accumulation, suggesting that β-arrestins are important for attenuation of LPA signaling. To further test this, we determined the effects of re-expression of wild-type β-arrestin 2 on inositol phosphate accumulation in LPA₁-transfected β-arrestin 1/2 KO MEFs (Fig. 3B). Co-transfection of wild-type β-arrestin 2 and LPA₁ in the β-arrestin 1/2 KO MEFs reduced the magnitude of LPA-induced inositol phosphate accumulation to 3.1-fold, which was similar to that observed in LPA₁-transfected WT MEFs (2.5-fold). Taken together, these observations strongly support the notion that β-arrestin association with LPA₁ receptors is important for signal attenuation and for clathrin-mediated receptor internalization.

Having established that the agonist-induced endocytosis of LPA₁ was mediated by a β-arrestin- and clathrin-dependent pathway, we next investigated the role of membrane cholesterol in LPA₁ signaling and trafficking. As mentioned, membrane cholesterol has been shown to be important for LPA stimulation of PI3-kinase/Akt signaling (Peres et al., 2003). To address this question, we first examined the effects of cholesterol extraction with methyl-β-cyclodextrin (MβCD) and the effects of cholesterol disruption with the cholesterol-binding drug, nystatin, on LPA stimulation of phosphoinositide hydrolysis, which is stimulated by Gq signaling. As a first step, we measured the effects of these cholesterol-perturbing drugs and the effects of water-soluble cholesterol:MβCD complexes on the cellular abundance of cholesterol in LPA₁/HeLa cells by using a quantitative cholesterol measurement assay (see Materials and Methods) (Table 1). Control LPA₁/HeLa cells contained 17.10±0.13 μg cholesterol/mg protein and treatment with 5 mM MβCD for 1 hour reduced cellular cholesterol by 62% to 6.52±0.05 μg cholesterol/mg protein. Addition of 10 mM cholesterol, as a water-soluble MβCD complex, for 1 hour after MβCD extraction, increased cellular cholesterol levels to approximately twice that observed in control cells (34.2±0.78 μg cholesterol/mg protein). In contrast, treatment of LPA₁/HeLa cells with 50 μg/ml nystatin for 1 hour slightly elevated the amount of cellular cholesterol (20.8±1.5 μg cholesterol/mg protein) and addition of water-soluble cholesterol to nystatin-treated cells increased cellular cholesterol amounts by approximately 2.5-fold relative to control LPA₁/HeLa cells (42.9±1.3 μg cholesterol/mg protein). This is consistent with the notion that nystatin merely binds sterols but does not extract them from cells.

We next examined the effects of these cholesterol-perturbing drugs on LPA₁ stimulation of phosphoinositide hydrolysis, which is promoted by Gq stimulation of phospholipase C. LPA stimulation of native HeLa cells (Fig. 4A, HeLa) resulted in a small 1.9-fold increase in accumulation of [³H]inositol phosphates, whereas stimulation of LPA₁/HeLa cells resulted in a large increase in PI hydrolysis (∼14-fold) (Fig. 4A, lane 1). Treatment of LPA₁/HeLa cells with 5 mM MβCD reduced LPA stimulated PI hydrolysis to 3.9-fold (72% inhibition) (Fig. 4A, lane 2). Treatment of LPA₁/HeLa cells with 50 μg/ml nystatin did not significantly affect agonist-dependent PI hydrolysis (Fig. 4A, lane 4). Addition of water-soluble cholesterol:MβCD complexes, which contained 10 mM cholesterol, to MβCD-treated or nystatin-treated cells greatly increased the extent of basal accumulation of labeled inositol phosphates (Fig. 4A, lanes 3 and 5). Addition of water-soluble cholesterol led to a greater increase in the basal level of inositol phosphate accumulation as compared to that observed in LPA-stimulated samples, which consequently decreased the fold induction of PI hydrolysis by LPA. This could be due to enhanced G_q signaling that is independent of LPA₁ receptors. These results suggested that the presence of membrane cholesterol was important for LPA₁ stimulation of PI hydrolysis.

As a control, we examined the effects of MβCD on the stimulation of PI hydrolysis by an unrelated G_q-coupled receptor, the M₁ muscarinic acetylcholine receptor (M₁ mAChR). For these experiments, HeLa cells were transiently transfected with plasmids encoding either wild-type LPA₁ or M₁ mAChRs. Whereas 5 mM MβCD inhibited LPA₁-mediated PI hydrolysis (Fig. 4B, compare lanes 3 and 4) as expected, it did not significantly reduce the extent of PI hydrolysis in response to agonist stimulation (1 mM carbachol) of M₁ mAChR-expressing cells (Fig. 4B, compare lanes 5 and 6). Immunofluorescence microscopy indicated that the transfection efficiencies of the LPA₁ and M₁ mAChR plasmids were comparable and were approximately 40% (data not shown). These results indicated that the reduction of LPA₁-mediated PI hydrolysis by MβCD was not due to inhibition of either Gα_q or phospholipase C activities, but rather was due to a specific inhibition of LPA₁ function. Taken together, these results suggest that the presence of plasma membrane cholesterol is critical for LPA₁-dependent signaling to phospholipase C (Fig. 4).

Next, we investigated whether membrane cholesterol was also important for LPA₁ endocytosis. We compared the effects of MβCD on the agonist-induced endocytosis of LPA₁ and M₁ mAChRs, which are also internalized by clathrin- and β-arrestin-dependent mechanisms (Vogler et al., 1999). Stably-transfected LPA₁/HeLa cells or transiently-transfected HeLa cells expressing M₁ mAChRs were pre-incubated in the presence or absence of 5 mM MβCD for 1 hour. These cells were then incubated with the respective agonists and Alexa 594-Tfn for 30 minutes. In untreated cells, LPA₁ and M₁ mAChRs localized to the plasma membrane, whereas Alexa 594-Tfn labeled pleomorphic endosomal structures (Fig. 5A, untreated). Agonist stimulation induced the endocytosis of both receptors into endosomal structures that colocalized with internalized Alexa 594-Tfn (Fig. 5A, +agonist). To quantify GPCR endocytosis, we measured the extent of GPCR (either LPA₁ or M₁) colocalization with the internalized Alexa 594-Tfn using Metamorph image analysis (Murph et al., 2003; Volpicelli et al., 2001). In control cells, treatment with 10 μM LPA increased LPA₁ and Alexa 594-Tfn colocalization by 3.3-fold relative to untreated cells (Fig. 5B, Control). Pre-incubation with 5 mM MβCD reduced this agonist-induced colocalization by more than 50% (Fig. 5B, compare black bars in control cells to cells treated with 5 mM MβCD). Addition of 10 mM water-soluble cholesterol restored LPA₁ and Alexa 594-Tfn colocalization to near control levels (Fig. 5B, MβCD + cholesterol). In contrast to LPA₁, agonist treatment (1 mM carbachol, 30 minutes) stimulated a twofold increase in M₁ mAChR colocalization with Alexa 594-Tfn, which was not inhibited by MβCD (Fig. 5C). These data suggested that plasma membrane cholesterol is important for both LPA₁ endocytosis and LPA₁ signaling.

Although the data above indicates that membrane cholesterol is essential for LPA₁ endocytosis, the data in Fig. 1, 2, 3 indicate that LPA₁ is internalized by clathrin- and β-arrestin-dependent mechanisms, which are distinct from cholesterol-dependent endocytic pathways (Nichols, 2003). To further investigate this apparent difference, we examined whether LPA₁ localized to detergent-resistant membrane domains, which are enriched in both cholesterol and glycosphingolipids (Razani et al., 2002). We examined the effects of LPA stimulation on the resistance of LPA₁ to extraction with Triton X-100. LPA₁/HeLa cells were incubated with 10 μM LPA for different times before extraction with ice-cold 1% Triton X-100 and indirect immunofluorescence (Fig. 6A). LPA₁ staining in unstimulated cells was greatly reduced following Triton X-100 extraction (Fig. 6A, 0 minute). In contrast, the cell-associated LPA₁ staining, which remained after detergent extraction, was increased with time of agonist stimulation (Fig. 6A). We quantified the detergent-resistant LPA₁ staining associated with the cells by measuring the pixel intensity of LPA₁-specific fluorescence using MetaMorph image analysis (see materials and methods) and normalizing this value to DNA content as assessed by Hoescht dye labeling (Fig. 6B). This analysis showed that detergent extraction of unstimulated cells reduced the level of cell-associated LPA₁ staining to 5% of that observed in untreated and non-extracted cells. Cell-associated LPA₁ staining progressively increased with time of LPA treatment such that after 30 minutes of LPA stimulation approximately 38% of LPA₁ immunoreactive staining remained after detergent extraction, relative to control cells (Fig. 6B, solid circles). Extraction of membrane cholesterol with 5 mM MβCD prior to LPA stimulation blocked the LPA-induced increase in detergent resistance of LPA₁ (Fig. 6B, open triangles). As detergent resistance of proteins can also be increased by their association with the actin cytoskeleton, we examined the effects of inhibiting actin polymerization with cytochalasin D (5 μM) on the detergent resistance of LPA₁ (Fig. 6B, open squares). We observed no noticeable difference between the detergent resistance of LPA₁ in cells pre-treated with cytochalasin D and untreated cells, after brief exposure to LPA (i.e. 0 to 8 minutes). A slight delay in the rate of increase of LPA₁ detergent resistance was observed between 10 and 20 minutes of LPA treatment in cells that were pre-treated with cytochalasin D, but the extent of detergent resistance was the same in both untreated and cytochalasin D-treated cells after 30 minutes of LPA stimulation. These data indicate that LPA treatment promotes the association of LPA₁ with detergent resistant membranes and that this process is inhibited by cholesterol extraction.

Given that LPA treatment for 30 minutes promotes the endocytosis of LPA₁ into transferrin receptor⁺ endosomes, it is likely that some of the detergent resistant LPA₁ receptors observed after longer LPA treatment reside in endosomes. Studies have shown that transferrin receptor⁺ endosomes are enriched in cholesterol (Hao et al., 2002). To investigate the effects of agonist stimulation on the detergent resistance of surface LPA₁ receptors, we labeled surface LPA₁ with mouse anti-FLAG antibodies on ice prior to detergent extraction (Fig. 6C,D). The LPA₁ expressed in LPA₁/HeLa cells contains an N-terminal FLAG epitope tag that is accessible to the extracellular medium. In the absence of detergent, mouse anti-FLAG antibodies labeled only the cell surface in control cells (Fig. 6C, Control). Triton X-100 extraction removed most of the surface-bound antibody (only 15% of control, un-extracted cells remained) (Fig. 6C,D, 0 minute). The detergent resistance of surface LPA₁ increased with time of agonist treatment up to 45% of control levels after 6 minutes and then declined after 15 minutes and 30 minutes of LPA treatment (Fig. 6C,D). These results suggest that the LPA₁ receptor associates with detergent-resistant membranes following agonist stimulation, both at the cell surface and following endocytosis in cholesterol-rich endosomes.

As β-arrestins are required for the clathrin-mediated endocytosis of LPA₁, we examined whether membrane cholesterol was important for the association of β-arrestins with LPA₁ or with β₂AR, as a control. Preliminary experiments showed that, in HeLa cells that were transiently transfected with plasmid encoding either LPA₁ or β₂ARs along with β-arrestin-2-GFP, both LPA₁ and β₂ARs transiently recruited cytosolic β-arrestin-2-GFP to punctate plasma membrane structures after 2 minutes of agonist stimulation (Fig. 8A,B, Control). Double-labeling experiments showed that β-arrestin-2-GFP extensively colocalized with the plasma membrane clathrin adaptor, AP2 (Fig. 7), suggesting that brief LPA stimulation led to the recruitment of β-arrestin-2-GFP to cell surface clathrin-coated pits. After 30 minutes of agonist stimulation, both LPA₁ and β₂ARs localized to endosomal structures, but β-arrestin 2 GFP returned to a cytosolic distribution (data not shown). This is consistent with published reports showing that β₂ARs transiently associate with β-arrestins (Oakley et al., 1999).

We next examined the effects of cholesterol extraction on the surface recruitment of β-arrestin-2-GFP by LPA₁ and β₂AR after 2 minutes of agonist stimulation (Fig. 8). β-arrestin-2-GFP localized in a diffuse cytoplasmic pattern in unstimulated cells and both LPA₁ and β₂AR were localized to the plasma membrane (data not shown). After 2 minutes of LPA treatment, β-arrestin-2-GFP colocalized with LPA₁ in punctate spots at the cell surface (Fig. 8A, Control, left panels), which also colocalized with AP2 (see Fig. 7). Similarly, after 2 minutes of isoproterenol treatment of β₂AR-expressing cells, β-arrestin-2-GFP localized to punctate spots at the cell surface (Fig. 8B, Control, right panels). Pre-incubation with 5 mM MβCD for 60 minutes completely inhibited the recruitment of β-arrestin-2-GFP to the cell surface by LPA₁ and β-arrestin-2-GFP remained in a diffuse cytosolic distribution (Fig. 8A, MβCD, see inset). Addition of 10 mM water-soluble cholesterol restored the ability of LPA₁ to recruit β-arrestin-2-GFP to the cell surface in MβCD-treated cells (Fig. 8A, MβCD-cholesterol). In contrast, incubation with MβCD did not inhibit β-arrestin-2-GFP recruitment to punctate surface spots by β₂ARs (Fig. 8B, MβCD, see inset). Addition of water-soluble cholesterol did not alter the surface recruitment of β-arrestin 2-GFP by β₂ARs (Fig. 8B, MβCD-cholesterol).

To quantify these effects, we determined the percentage of cells that showed β-arrestin recruitment to the plasma membrane after 2 minutes of agonist stimulation of either LPA₁ or β₂AR (Fig. 8C). In control cells and in cholesterol repleted cells (i.e. MβCD-cholesterol), β-arrestin-2-GFP was recruited to the cell surface in ∼80% of cells expressing either LPA₁ or β₂AR. Only 4% of LPA₁-expressing cells showed surface recruitment of β-arrestin-2-GFP in MβCD-treated cells. In contrast, approximately 50% of β₂AR-expressing cells exhibited surface recruitment of β-arrestin-2-GFP. Taken together, these results indicate that LPA₁ is much more dependent upon plasma membrane cholesterol for the recruitment of β-arrestin than β₂ARs. This also provides a possible link between membrane cholesterol and clathrin-dependent endocytosis of LPA₁ as β-arrestin is critical for endocytosis of LPA₁.

All members of the GPCR superfamily share the ability to rapidly respond to agonist stimulation and then to undergo desensitization (Lefkowitz and Shenoy, 2005). Many of these GPCRs are also rapidly internalized into cells via one of several distinct endocytic pathways. However, the mechanisms that regulate GPCR desensitization and determine the specific endocytic pathway used for internalization vary from receptor to receptor. In this study, we found that LPA₁ receptors are internalized by a clathrin- and β-arrestin-dependent pathway, but that they also require plasma membrane cholesterol for receptor signaling and for their subsequent clathrin-dependent endocytosis. Our results indicate that the key requirement of membrane cholesterol for LPA₁ endocytosis is for the association of LPA₁ with β-arrestin, which promotes both signal attenuation and clathrin-dependent endocytosis of the receptor.

Caveolae and other detergent-resistant membrane domains are cholesterol- and glycosphinoglipid-rich, are sites of active signal transduction and have been implicated in the activation of heterotrimeric G proteins, Ras signaling and eNOS signaling (Razani et al., 2002). Several lines of evidence suggest that LPA₁Rs associate with cholesterol-rich, detergent-resistant membranes and that this is important for LPA₁-dependent signaling. First, cholesterol extraction with MβCD strongly inhibited LPA₁ induction of phosphoinositide hydrolysis, via Gα_q-mediated stimulation of phospholipase C (Fig. 4). Gα_q has been shown to be enriched in caveolae (Huang et al., 1997; Oh and Schnitzer, 2001), which supports the notion that LPA₁ stimulates PI hydrolysis by associating with Gα_q in detergent-resistant membranes. Re-addition of cholesterol to MβCD-treated cells increased both the basal and LPA-stimulated levels of PI hydrolysis (Fig. 4A). The fact that MβCD extraction did not affect PI hydrolysis by the M₁ mAChR suggests that cholesterol depletion does not impair either Gα_q or phospholipase C activity per se, but that LPA₁ stimulation of PI hydrolysis is particularly sensitive to cholesterol depletion (Fig. 4C). Two possible explanations for the difference between LPA₁ and M₁ mAChRs are that either LPA₁ exclusively couples to Gα_q that is localized to detergent-resistant membrane domains or that membrane cholesterol is required for the physical association of LPA₁ with Gα_q. Second, we found that LPA stimulation enhanced the resistance of both surface and total LPA₁ to extraction with TX-100 detergent (Fig. 6). Resistance to detergent extraction is a common property of proteins that are associated with caveolae and other cholesterol-rich membrane regions (Razani et al., 2002). The detergent-resistance of surface LPA₁ increased during the first 6 minutes of LPA treatment and then declined. This is consistent with a transient association of LPA₁ with detergent-resistant microdomains prior to β-arrestin- and clathrin-dependent endocytosis. Disruption of the actin cytoskeleton with 5 μM cytochalasin D did not alter the agonist-induced detergent resistance of LPA₁ suggesting that the increased detergent resistance of LPA₁ was not due to its association with the actin cytoskeleton. However, cholesterol extraction completely prevented the agonist-induced detergent resistance of LPA₁, which is consistent with the notion that LPA₁ associates with cholesterol-rich membrane microdomains.

Interestingly, total detergent-resistant LPA₁ staining increased with time of LPA treatment even after longer periods of agonist stimulation (Fig. 6B, 30 minutes). We have previously shown that about 35-40% of surface LPA₁ receptors are internalized into transferrin receptor⁺ endosomes after 30 minutes of LPA treatment (Murph et al., 2003). We hypothesize that some of the detergent-resistant LPA₁ staining observed after longer LPA treatment resides in transferrin receptor⁺ endosomes, which are known to be enriched in cholesterol (Hao et al., 2002). Finally, a recent study showed that LPA stimulation of phosphoinositide 3-kinase and the downstream effector kinase, Akt, was inhibited by MβCD treatment in Vero cells (Peres et al., 2003). Collectively, these data suggest that LPA₁ association with cholesterol-rich plasma membrane regions is critical for LPA-induced signaling through Gα_q. Given that a pool of Gα_q is present in cholesterol-rich caveolae, we hypothesize that the localization of LPA₁ to detergent-resistant membranes is important for their association with the pool of Gα_q that is localized to these domains.

In support of a role for cholesterol in LPA₁ endocytosis, we found that cholesterol extraction inhibited LPA₁ association with β-arrestin and the subsequent clathrin-dependent endocytosis of the receptor. Many different GPCRs interact with β-arrestins, which is important for the proper regulation of receptor function (Lefkowitz and Whalen, 2004). Phosphorylation of specific serine/threonine residues in either the cytoplasmic tail or the third intracellular loop, by G protein receptor kinases, leads to the recruitment of β-arrestin proteins, which in turn block G protein/receptor coupling (desensitization) and also promote clathrin-dependent endocytosis (Ferguson et al., 1996; Ferguson et al., 1995). Our data show that wild-type LPA₁ receptors transiently recruit β-arrestin-2-GFP to discrete AP2⁺ structures at the cell surface, in an agonist-stimulated fashion (Fig. 3), but do not colocalize on endosomes with β-arrestin-2-GFP (Figs 3 and 8). β-arrestins promote clathrin-dependent endocytosis of GPCRs by localizing receptors to clathrin coated pits through an interaction of β-arrestins with both clathrin heavy chain and the μ₂ subunit of the AP-2 clathrin adaptor complex (Goodman et al., 1996; Laporte et al., 1999). Using MEFs derived from β-arrestin 1 and 2 double-knockout mice, we showed that both signal attenuation and endocytosis of LPA₁ is dependent upon β-arrestin (Figs 2 and 3). Using an RNA interference approach to reduce the cellular abundance of clathrin heavy chain, we showed that knockdown of clathrin inhibited the internalization of LPA₁, transferrin receptors, but not the internalization of the GPI-anchored protein, CD59, which localize to cholesterol-rich membrane regions (Fig. 1). Taken together, these data indicate that LPA₁ receptors are internalized by β-arrestin- and clathrin-dependent endocytosis.

The most significant finding of these studies was that cholesterol extraction inhibited β-arrestin recruitment to the plasma membrane by LPA₁ and the subsequent endocytosis of these receptors (Figs 5 and 8) and that re-addition of cholesterol to MβCD treated cells restored both of these functions. As β-arrestin binding to LPA₁ precedes receptor endocytosis, we hypothesize that cholesterol is required for the association of LPA₁ with β-arrestins and that it is the lack of β-arrestin binding that leads to the inhibition of LPA₁ endocytosis. This is a novel and previously unappreciated role for membrane cholesterol in the recruitment of β-arrestins. We speculate that other GPCRs that localize to caveolae may also associate with β-arrestin in a cholesterol-dependent manner. β₂ARs localize to caveolae in cardiomyocytes in the absence of agonist but move out of caveolae and into clathrin-coated pits upon ligand binding (Ostrom et al., 2001; Rybin et al., 2000). As β₂AR endocytosis requires β-arrestin binding, it is probable that β-arrestin also binds to these receptors in caveolae. Whether the association of β-arrestin with β₂ARs in cardiomyocytes is cholesterol-dependent remains to be determined.

In contrast to LPA₁, cholesterol extraction did not inhibit the endocytosis of M₁ mAChRs (Fig. 5), which also follow a β-arrestin- and clathrin-dependent pathway (Vogler et al., 1999) nor did cholesterol extraction inhibit the association of β₂ARs with β-arrestins. This suggests that the cholesterol dependence of β-arrestin recruitment is a unique property of LPA₁. Cholesterol may be important either for the direct recruitment and binding of β-arrestins to LPA₁Rs or for the recruitment of kinases such as GRKs that phosphorylate agonist-stimulated LPA₁. Indeed, GRK4 and GRK6 are palmitoylated, which is required for their membrane association (Premont et al., 1996; Stoffel et al., 1994), and palmitoylation has been shown to target many proteins to cholesterol-rich membranes including SNARES (Salaun et al., 2005), flotillins (Neumann-Giesen et al., 2004) and RGS16 (Hiol et al., 2003). Recent work has shown that both LPA stimulation and activation of protein kinase C with phorbol esters promotes LPA₁ phosphorylation (Avendano-Vazquez et al., 2005).

Are there physiological contexts where changes in cellular cholesterol modulate LPA signaling? One intriguing example may be prostate cancer cells, whose growth is potently stimulated by LPA. Cholesterol is elevated in these cells and contributes to their enhanced growth (Heemers et al., 2001; Mills and Moolenaar, 2003; Zhuang et al., 2002), perhaps by augmenting LPA signaling. Future studies should provide the answer to this and other questions about this novel process.

We are greatly indebted to Stefano Marullo for providing GFP-tagged β-arrestin plasmids and to Robert J. Lefkowitz for providing wild-type and arrestin 1/2 null mouse embryo fibroblasts. We thank Giang Nguyen and Launa Scaccia for help during the initial stages of this work, and Nael McCarty and Julie Donaldson for providing comments on the manuscript. This work was supported by National Institutes of Health grant HL 67134 (to H.R.).