GPI anchoring leads to sphingolipid‐dependent retention of endocytosed proteins in the recycling endosomal compartment

FR‐GPI and GPI‐anchored decay accelerating factor (DAF) are extensively co‐localized with TfR in the REC as previously observed by horseradish peroxidase (HRP)‐mediated quenching and immunoelectron microscopy (EM) methodologies, respectively (Mayor et al., 1998). Despite this extensive co‐localization, a significant fraction of the internalized GPI‐AP pool is consistently separate from TfR (Mayor et al., 1998). Accumulation of GPI‐APs in GEECs (S.Sabharanjak and S.Mayor, in preparation) provides an explanation for the lack of co‐localization of a fraction of internalized GPI‐APs with TfR‐containing endosomes (Mayor et al., 1998). This fraction of GPI‐APs co‐localizes with fluid‐phase markers but not with Tf or low‐density lipoproteins (LDLs) that are internalized into sorting endosomes (data not shown).

To identify the rate‐limiting step in the recycling of GPI‐APs to the cell surface, we measured the rates of exit of GPI‐APs from the GEECs and the REC (Figure 1). As described by Mayor et al. (1998), the rate of approach to steady state is a measure of the rate of exit from a kinetically defined compartment. Therefore, we measured the rate of approach to steady state of internalized FR‐ GPI ligand, N^α‐pteroyl‐N^ϵ‐(4′‐rhodamine‐thiocarbamoyl)‐l‐lysine (PLR), in the REC and the GEECs, identified as described in Materials and methods. The amount of FR‐GPI, as measured by the extent of PLR fluorescence in the peripheral GEECs (Figure 1A and B; arrows), remained unchanged between the 20 and 60 min time points, whereas there was a dramatic increase in PLR fluorescence in the REC between the 20 and 60 min time points (Figure 1A and B; arrowheads). Kinetic analyses of these data show that FR‐GPI reaches steady state in the GEECs much earlier than in the REC (Figure 1E). The kinetics of exit of FR‐GPI from the GEECs and the REC are well described as a first‐order process (Figure 1E, lines represent the theoretical fit; R ≥0.98), although we can not rule out undetected contributions from additional processes. The t_1/2 of exit from the GEECs is ∼13 min (k_e 0.05 min⁻¹) while that from the REC is ∼30 min (k_e 0.022 min⁻¹) (Figure 1F). Since the t_1/2 for recycling of FR‐GPI to the cell surface measured in the same experiment is also 30 min (k_e 0.023 min⁻¹; Figure 1F, see also Table II), these data show that the rate‐limiting step in the recycling of the internalized GPI‐APs is their exit from REC. The slow rate of recycling of GPI‐APs results in the accumulation of GPI‐APs in the REC for extended times, confirming that GPI‐APs are sorted/segregated from other recycling membrane components in the REC.

To examine whether this manifestation of endocytic sorting, namely retention in the REC, is mediated by the presence of the GPI anchor, we compared the extent of accumulation of FR‐GPI and two isoforms of FR, FR‐TM and FR‐TA, lacking the GPI anchor. FR‐TM has the folate‐binding portion fused to the transmembrane region of the Fc receptor and the cytoplasmic tail of the LDL receptor (Varma and Mayor, 1998), and FR‐TA bears the same transmembrane and cytoplasmic tail region as FR‐TM but has three critical tyrosines mutated to alanine, rendering it incapable of being recognized by a cytoplasmic sorting machinery (Matter et al., 1993). As described above and shown in Figure 2A, FR‐GPI continues to accumulate in TfR‐containing REC, while the distribution and amount of internalized transferrin (Tf) bound to TfR has already reached a steady state by 20 min (Mayor et al., 1993). On the other hand, FR‐TM and FR‐TA traffic in a morphologically and kinetically identical manner to TfR; there is no change in the amount and distribution of internalized PLR bound to FR‐TM (Figure 2B) and FR‐TA (data not shown) with respect to endocytosed Tf bound to its receptor. These data show that retention in the REC in preference to other recycling membrane proteins requires the presence of a GPI anchor.

To determine whether sphingolipids play a role in the endocytic retention of GPI‐APs, we used a well characterized sphingolipid synthesis inhibitor, FB₁, which inhibits sphingosine N‐acyltransferase (Figure 3; Wang et al., 1991; Mays et al., 1995; Stevens and Tang, 1997), to alter sphingolipid levels in cells. FB₁ treatment does not affect cell viability, and lipid analyses of FB₁‐treated cells show that there is a reduction in the amount of all major sphingolipid classes without affecting cholesterol levels relative to the untreated controls (Table I). Growing FRαTb‐1 cells in the presence of FB₁ resulted in the loss of endocytic retention of FR‐GPI (Figure 4A; compare filled and open circles); FR‐GPI was recycled to the cell surface at a rate similar to the rate of TfR recycling (t_1/2 ∼6–8 min), without significantly altering its rate of internalization (Table II). Addition of short‐chain ceramide, N‐hexanoyl‐d‐sphingosine (C₆‐ceramide), to FB₁‐treated cells in the continuous presence of FB₁ restored the rate of FR‐GPI recycling to control levels (Figure 4B; open squares), along with a restoration of sphingolipid levels (Table I).

FB₁ treatment leads to a build‐up of sphinganines with a concomitant decrease in ceramide levels (Wang et al., 1991; see also Figure 3). Either of these changes could affect signal transduction pathways in cells (Hannun, 1994) leading to an alteration in the kinetics of endocytic processes (Chen et al., 1995). Therefore, we used a different drug, d‐threo‐1‐phenyl‐2‐decanoylamino‐ 3‐morpholino‐1‐propanol (PDMP), which specifically blocks synthesis of a large class of sphingolipids, the glycosphingolipids (Inokuchi and Radin, 1987; Inokuchi et al., 1990), resulting in the accumulation of a different set of sphingolipid precursors from that found after FB₁ treatment (Felding‐Habermann et al., 1990). PDMP treatment also leads to acceleration in the rate of FR‐GPI recycling (Figure 4D and Table II). In another approach we tested whether the rate of FR‐GPI recycling was restored to control levels in FB₁‐treated cells supplemented with a sphingolipid, N‐palmitoyl‐dl‐erythro‐dihydrosphingosine (NPDS). The naturally occurring erythro isomers of dihydroceramides or their metabolites are incapable of mediating signaling activity in cells (Bielawska et al., 1993; Hannun, 1994; Chen et al., 1995). NPDS restores the rate of FR‐GPI recycling to that in control cells (Figure 4B; open triangles). These data suggest that sphingolipid signaling is unlikely to be involved in regulation of endocytic retention of FR‐GPI. However, since sphingolipid intermediates involved in signaling may remain uncharacterized, we can not rule out this possibility altogether.

To rule out further that the loss in endocytic retention of FR‐GPI is due to secondary effects of using sphingolipid‐depleting drugs, and that the restoration of retention is a result of uncharacterized signaling properties of exogenous sphingolipid analogs, we examined the recycling of FR‐GPI in a cell line, LY‐B, mutant for the serine palmitoyl transferase enzyme (Hanada et al., 1998; see Figure 3). LY‐B cells do not accumulate any known signaling‐competent sphingolipid precursors, as the block is in the first step of sphingolipid biosynthesis. These cells are unable to make the sphingoid base from endogenous sources, and therefore rely on serum‐derived lipids to supplement their sphingolipid pools (Hanada et al., 1998). Growing cells in sphingolipid‐deficient medium for 48 h leads to sphingolipid levels that are 85% lower than in wild‐type cells or cells supplied with exogenous sources of sphingolipid, without affecting cell viability or the content of other major phospholipids (Hanada et al., 1998) and cholesterol (Fukasawa et al., 2000). The rates of FR‐GPI recycling in sphingolipid‐replete and sphingolipid‐ deficient conditions were measured in LY‐B cell lines expressing FR‐GPI. In cells grown under sphingolipid‐deficient conditions, recycling of FR‐GPI occurred with a t_1/2 of 6–8 min, as compared with 27 min in cells grown under conditions that restore normal levels of the lipid (Figure 4C and D). In contrast, the change in the rate of FR‐GPI internalization in LY‐B cells was relatively insignificant upon sphingolipid depletion: cells depleted of sphingolipids internalized FR‐GPI ∼1.3‐fold more quickly than replete cells. Thus, the loss of endocytic retention of FR‐GPI is not due to secondary effects of the drug treatments, and its restoration by addition of exogenous lipids is unlikely to be due to activation of signaling pathways.

To test whether sphingolipid depletion affects recycling of other GPI‐APs, we determined the rate of DAF recycling in an FB₁‐treated DAF‐expressing CHO cell line, DAFTb‐1 (Mayor et al., 1998). The rate of DAF recycling was accelerated upon treatment with FB₁ without substantially affecting the rate of internalization of DAF (Figure 5; Table II), confirming that the effect of sphingolipid depletion on endocytic retention of FR‐GPI is not specific to the type of GPI‐AP molecule being studied.

To ensure that FB₁ treatment or addition of exogenous ceramides to FB₁‐treated cells did not result in any qualitative alteration in the endocytic pathway of the FR‐GPI, we analyzed the co‐localization of endocytosed FR‐GPI with internalized TfR. Figure 6 shows that endocytosed FR‐GPI and TfR were extensively co‐localized in the REC in untreated (A and B), FB₁‐treated (C and D) or C₆‐ceramide‐replenished cells (E and F). PDMP treatment or growth of the LY‐B cell line under conditions of sphingolipid depletion did not affect co‐localization of FR‐GPI and TfR in the REC either (data not shown). The majority of internalized FR‐GPI and TfR were co‐localized in the REC, showing that upon alteration of sphingolipid levels, exit from the REC remains the rate‐limiting step in recycling. These data together show that reduction in sphingolipid levels results in the loss of endocytic retention of GPI‐APs in the REC; replenishment of sphingolipids results in a restoration of endocytic retention.

To test whether the defect in recycling of GPI‐APs observed upon sphingolipid depletion is specific to GPI‐APs or is a result of a generalized acceleration in recycling rates of all proteins from the REC, we measured the recycling kinetics of wild‐type human TfR and a C‐terminally truncated TfR (δ3‐59TfR) that lacks the clathrin internalization motif. The rates of TfR and δ3‐59TfR recycling were not affected by treatment with FB₁ (Figure 7A). The rate of recycling of endogenous hamster TfR in LY‐B cells grown in sphingolipid‐deficient medium was the same as that in sphingolipid‐replete medium (Figure 7B); the rate constants measured are similar to those of the human TfR in FRαTb‐1 and δ3‐59TfR in δ3‐59TRVb cells. These data show that sphingolipid depletion does not lead to a generalized increase in the exit rate of proteins from the REC, since neither those that interact with clathrin coats (TfR) nor those that do not (δ3‐59TfR) have altered recycling kinetics. Furthermore, the extent and distribution of the TfR in REC were also unaltered (Figure 6). Thus, sphingolipid depletion specifically abolishes the endocytic retention of GPI‐APs in the REC, and does not have a non‐specific effect on the kinetics or morphology of the endocytic pathway of other proteins.

Fluorescent derivatives of folic acid, N^α‐pteroyl‐N^ϵ‐(4′‐fluorescein‐thiocarbamoyl)‐l‐lysine (PLF) and PLR were synthesized as described previously (McAlinden et al., 1991), except that PLR was synthesized using lissamine rhodamine sulfonyl chloride (Fluka Chemie AG, Buchs, Switzerland) instead of fluorescein isothiocynate. PLR binds saturably to FR isoforms and is resistant to release at pH 5.0, well below the acidic pH of endosomes. Nigericin and all fluorochromes for conjugation, except Cy3 and Cy5, (Amersham Pharmacia Biotech) were obtained from Molecular Probes (Eugene, OR). Tf was purified (Mayor et al., 1993) and all protein‐based reagents were labeled with fluorophores according to the manufacturer's instructions. LDL was prepared from fetal bovine serum and labeled with 1,1′‐dioctadecyl‐3,3,3′,3′)‐tetramethylindocarbocyanine perchlorate [’DiI', DiIC18(3)] as previously described (Mayor et al., 1993). All dichroic mirrors, excitation and emission filters were from Chroma Technology Corp. (Brattleboro, VT). All reagents were obtained from Sigma Chemical Co. (St Louis, MI) unless indicated otherwise. Fetal bovine serum (FBS) was obtained from Life Technologies (Rockville, MD). PDMP was obtained from Matreya, Inc. (Pleasant Gap, PA). PI‐PLC (phosphatidylinositol‐specific phospholipase C) was a gift from Martin Low (Columbia University, New York, NY) and Ram Vishwakarma (National Institute of Immunology, New Delhi). A mouse monoclonal antibody (mAb) against the human folate receptor, Mov19, was a gift from Centocor Corp. (Malvern, PA). A mouse mAb against human DAF (human anti‐CD55) was purchased from PharMingen (San Diego, CA). Anti‐human transferrin receptor antibody and Nutridoma‐SP were obtained from Boehringer Mannheim (Roche Diagnostics, Mumbai, India).

DAF‐, FR‐GPI‐, FR‐TM‐ and δ3‐59TfR‐expressing CHO cells (DAFTb‐1; FRαTb‐1, FR‐TM‐Tb‐1 and δ3‐59TRVb cells, respectively) were maintained as described previously (Mayor et al., 1998; Varma and Mayor, 1998). FR‐GPI‐expressing LY‐B lines were obtained by transfecting LY‐B cells (Hanada et al., 1998) with FR‐GPI cDNA in plasmid pMEP4 as previously described (Mayor and Maxfield, 1995). Transfected LY‐B lines were maintained in folate‐free HF‐12 supplemented with 5% FBS, 50 μg/ml hygromycin and 100 μg/ml penicillin‐ streptomycin. cDNA for FR‐TA was obtained and transfected into TRVb‐1 cells (FR‐TA‐Tb‐1 cells) exactly as described for FR‐TM (Varma and Mayor, 1998).

All microscopy was carried out on cells grown on poly(d‐lysine)‐coated 35 mm cover‐slip bottom dishes as previously described (Mayor et al., 1998). Fluorescent labels were made up to appropriate concentrations in labeling medium (folate‐free HF‐12, 0.3 mg/ml NaHCO₃, 5% dialyzed serum, 15 mM HEPES pH 7.2). Cells were washed in medium 1 (150 mM NaCl, 5 mM KCl, 1 mM CaCl₂, 1 mM MgCl₂, 20 mM HEPES pH 7.4). Fluorescent folic acid analogs were removed from their receptors by multiple washes with ascorbate buffer (140 mM sodium ascorbate, 65 mM ascorbic acid, 1 mM CaCl₂, 1 mM MgCl₂ pH 4.5).

Cells expressing FR isoforms were plated in medium containing 5% dialyzed FBS 3 days prior to labeling. Cell surface labeling was carried out by incubation of the cells with the indicated ligands for 1 h at 0°C in labeling medium. Excess label was removed by washing with medium 1 prior to microscopy. Total endosomal receptor labeling was carried out by incubation of cells with indicated ligands for 3 h at 37°C and then the surface receptors were stripped off their ligands by treatment with chilled ascorbate buffer (4 × 15 min at 0°C). This procedure removes >95% of the cell surface receptors without affecting the labeling of the endocytosed receptors. Imaging of cells was carried out in the presence of 10 μM nigericin in a high‐potassium buffer (120 mM KCl, 5 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 20 mM HEPES pH 7.4) to neutralize endosomal pH.

Co‐localization studies were carried out by labeling cells with 10 nM PLF and 10 μg/ml Alexa‐568‐Tf for 3 h at 37°C followed by removal of surface receptors by PI‐PLC treatment and imaging on a wide‐field system using a 60× objective. Appropriate dichroic and emission filters were used so that there was no detectable cross‐over of fluorescence emission for the two fluorophores when imaged singly.

Cells were depleted of sphingolipids by growing them in the presence of 40 μg/ml FB₁ for 64–68 h before labeling. Sphingolipid analogs were added by making 1:1 molar mixtures of lipid with defatted bovine serum albumin. C₆‐ceramide and NPDS were added at 1 and 5 μM, respectively, to the FB₁‐containing culture medium 12 h before imaging. PDMP treatment was done by growing the cells in 10 μM PDMP in the culture medium for 64–68 h. Sphingolipid depletion in LY‐B lines was carried out exactly as previously described (Hanada et al., 1998), except that 1% dialyzed Nutridoma was used in the depletion medium. Dialyzed Nutridoma has similar effects on cell growth as Nutridoma (Hanada et al., 1998).

Fluorescence imaging was carried out using a wide‐field imaging system [Nikon Eclipse TE300 or Diaphot TMD equipped with a digital CCD camera with a TEK‐512×512 D‐series back‐illuminated chip (Princeton Instruments, Princeton, NJ)] controlled using Metamorph software. For all experiments 20× (0.75 n.a.) or 60× (1.4 n.a., oil) objectives (Nikon, Japan) were used. Fluorescence images were processed using routines in MetaMorph software following previously described procedures (Mayor et al., 1993).

To measure the amount of endocytosed FR‐GPI, internalized PLF (or PLR)‐fluorescence from individual cells was normalized against surface‐bound Cy3‐Mov19 (or the second‐color folate). In the case of DAF, internalized Cy5‐conjugated anti‐DAF fluorescence was normalized against surface‐bound primary antibody detected via Alexa‐488‐conjugated goat‐anti‐mouse secondary antibody. In the case of TfR, internalized Cy5‐conjugated Tf fluorescence bound to TfR was normalized against the number of cells in a field or surface receptors detected via Alexa‐488‐ or fluorescein isothiocyanate (FITC)‐conjugated secondary antibody against primary antibody bound to surface TfR.

In each experiment, normalized fluorescence was calculated from 12–16 fields from duplicate dishes for each time point, and the weighted mean value and the uncertainty in the mean were determined on a per field basis considering each field (consisting of 5–30 cells) to be an independent event, as previously described (Mayor et al., 1993).

In general, the rates of recycling (or exit rates) and internalization were obtained by determining the rates of approach to steady state of labeling in the whole cell (or in the GEECs or REC), under conditions where the cell surface receptors were labeled to saturation at all times in the experiment. The recycling rate, k_e, was determined by fitting the data to a single exponent of the form f = a + b [1 − exp(−k_et)], using the method of least squares provided in SigmaPlot version 4 (SPSS Inc.). The rate of internalization of FR‐GPI or DAF was determined from the first‐order steady‐state kinetic equation exactly as described in Mayor et al. (1993).

The rate of recycling of FR‐GPI to the cell surface was measured as previously described (Mayor et al., 1998), using either of two assays with identical results. In both assays, FR‐GPI‐expressing cells were incubated with saturating amounts of the fluorescent ligand for the indicated times at 37°C until a steady state of labeling was reached. Subsequently, in one of the assays (where PLF was used as the fluorescent ligand), Cy3‐conjugated Mov19 was used to normalize the amount of surface receptor expression. In the other assay, surface‐bound fluorescent ligand (PLF or PLR) was removed with chilled ascorbate buffer and labeled with saturating amounts of the second‐color ligand on ice to normalize for surface receptor expression. The rate of DAF recycling was measured exactly as described (Mayor et al., 1998), except that complete Cy5‐conjugated anti‐DAF mouse IgG instead of Fab fragments was used to label the recycled DAF. Identical rates were obtained using intact antibody or Fab (data not shown).

The efflux kinetics of FR‐GPI from REC (or the GEECs) were determined using an assay similar to that used for measuring the rate of FR‐GPI recycling from the whole cell, except that REC (or GEEC)‐associated PLR fluorescence was quantified at each time point. For this purpose, the REC in each cell was identified by co‐labeling cells at each time point with a 20 min pulse of Cy5‐Tf (10 μg/ml) followed by a 10 min chase in a Cy5‐Tf‐free medium containing 20 μM deferroxamine. This protocol ensures that Tf label is present mainly in the REC (Presley et al., 1993). Surface‐bound PLR was removed using chilled ascorbate buffer. The cells were then mildly fixed on ice with 1% paraformaldehyde. The surface receptors were then bound to PLF and normalized for surface receptor expression. The cells were imaged for PLR, PLF and Cy5‐Tf using appropriate filters. The REC in the Cy5‐Tf image was manually marked using the ‘draw region’ routine available in Metamorph software, and the PLR fluorescence co‐localized with this structure was quantified. The rate of exit of FR‐GPI from GEECs was determined from the same experiment by subtracting the REC‐associated PLR fluorescence from the total internalized PLR fluorescence. In separate experiments we determined that the PLR fluorescence in peripheral endosomes (not associated with the REC) was co‐localized with the fluid‐phase marker, FITC‐conjugated dextran, but not with endocytosed Tf and LDL (data not shown).

The rate of recycling of TfR in FRαTb‐1, LY‐B and δ3‐59TRVb cells was measured as described previously (Mayor et al., 1993), except that in LY‐B lines expressing endogenous hamster TfR, 20 μg/ml of Tf was used to label the receptors.

Lipids were extracted and purified from FRαTb‐1 as described (Ariga et al., 1988) with a few modifications. Cells (∼2.5 × 10⁸) were sequentially extracted as described (Stevens and Tang, 1997). Neutral and acidic lipids were resolved on silica HP‐TLC plates that had been pre‐run in chloroform‐methanol‐water (60:35:8, by volume) and heated to 100°C for 30 min. After application of standards and samples, the lipids were resolved by running the HP‐TLC plates first in chloroform‐ methanol‐acetic acid‐formic acid‐water (35:15:6:2:1, by volume) to ∼12 cm from the origin, followed by hexane‐diethyl ether‐acetic acid (65:35:2, by volume) to the top of the plate. The lipids were identified and quantified as described (Brown and Rose, 1992). They were visualized by charring with cupric acetate and identified by co‐migration with standards, and quantified by densitometric scanning and comparison with standards (Ariga et al., 1988).