The human membrane folate receptor (MRF) exists in three isoforms of which the GPI-anchored α- and β-isoforms are expressed in placental tissue 10. MFR-α is also expressed in the buccal carcinoma cell line, KB cells. Folate uptake is essential for the survival of all dividing cells. Folate is a co-factor for DNA replication enzymes as well as a substrate in thymidine synthesis. Developmental abnormalities are seen in experimental models when FR functions are obstructed with the use of antibodies or anti-sense RNA as well as in populations wherein genetic mutations in MFR-α are present 1516. Folate uptake mediated via GPI-anchored receptors enables cell survival in media containing low amounts of folate 17. Chimeric FRs bearing a transmembrane protein anchor are inefficient in folate uptake and are also not subjected to the same regulatory mechanisms as the GPI-anchored form, which indicates that the GPI-anchorage directs an optimal endocytic pathway for folate absorption to be used by FR 18.

GPI-anchored prion proteins are the causative agents of neurodegenerative spongiform encephalopathies wherein the cellular form of the prion protein, PrP^C, is converted to the scrapie isoforms, PrP^Sc, resulting in the deposition of amyloid plaques on neurons 192021. The conversion of PrP^C to PrP^Sc occurs presumably in endocytic compartments 2223. Consistent with this observation, lysosomotropic agents like chloroquine as well as cysteine protease inhibitors are able to supress the conversion of PrP^C to PrP^Sc 24. Similar to the FR, replacement of the GPI-anchor with transmembrane anchoring sequences results in reduced scrapie formation 25. Recent hypotheses also suggest a role for prion proteins as modulators of innate as well as acquired immunity by functioning as receptors for complexes of viral RNA, viral capsid proteins and uric acid 26.

Aerolysin produced by Aeromonas hydrophila has been shown to bind to Thy-1 2728, specifically to the GPI-anchor 13. Erythrocytes from PNH patients, a disease condition where the synthesis of GPI-anchors is impaired, are resistant to intoxication by aerolysin 29. Aerolysin is a pore-forming toxin with pore formation enhanced by the co-operative binding of about seven toxin molecules to cell membranes 30. Oligomerization of the toxin to the heptameric form is a greatly accelerated process when the toxin is bound to the cell surface as compared to the association of the toxin molecules in vitro 31. It is likely that the presence of GPI-APs in clusters at the plasma membrane is responsible for the binding and insertion of the toxin subunits in the plasma membrane 31. Gordon et al. have demonstrated a role for FR-GPI as a receptor for the Clostridium septicum alpha toxin 14. PIPLC sensitive binding of the Helicobacter pylori to putative GPI-anchored receptors has also been postulated 32.

A significant role for GPI-APs as virus receptors has emerged from several studies. FR has been identified as a receptor for filoviruses like the Ebola virus (EBV) and Marburg virus (MBV). Expression of recombinant FR in virus insensitive cell lines renders them susceptible to infection by EBV and MBV 12. Similarly, GPI-anchored DAF has been shown to act as a co-receptor for coxsackieviruses B3 and A21 3334. DAF has also been identified as the receptor for Enterovirus 70 (EV70) in leukocytes 35 and in HeLa cells 36 as well as echovirus 7 37. Stable transfection of NIH3T3 cells with DAF leads to binding and replication of EV70, suggesting a role for DAF in the endocytosis of EV70 36. Anti-DAF antibodies were effective in preventing EV70 binding to cells. In a similar manner, pretreatment of HeLa cells with PIPLC resulted in inhibition of echovirus 7 attachment to cells. Transfection of CHO cells with DAF lead to binding and infection by multiple echovirus serotypes demonstrating a role of DAF as a virus receptor 37.

GPI-APs are also thought to be co-receptors with CR3 in mediating non-opsonic phagocytosis of Mycobacterium kansasii. Antibodies against GPI-APs inhibit the non-opsonic phagocytosis of M. kansasii specifically. CR3 mediated phagocytosis of opsonized zymosan particles as well as opsonized M. kansasii is not inhibited under these conditions 38. GPI-anchored CD55 (DAF) and CD66e (Carcinoembryonic antigen; CEA) are recruited by diffusely adhering Escherichia coli (Afa/Dr DAEC) strains during infection of CaCo-2 cells. The native diffuse distribution of CD55 and CD66e on the apical membrane is altered to a clustered distribution surrounding the adherent bacteria. The adhesion of bacteria can be inhibited if the interactions with these GPI-APs are prevented by pre-treatment with anti-CD55 and anti-CD66e antibodies 39. The GPI-anchor and a short consensus region 3 (SCR3 domains) of DAF are critical for the internalization of E. coli expressing the Dr-adhesins 40. GPI-anchored CD66e (also referred to as carcinoembryonic antigen, CEA) also functions as a cell surface receptor for Neisseria gonorrhoeae expressing the Opa⁵² adhesin 41.

Signal transduction via GPI-APs is also thought to be synergistic with the activation of the T-cell receptor complex. (reviewed in 42). Recently, stimulation of T-cells and natural killer (NK) cells by IL-18 was shown to be mediated by GPI-anchored CD48 43. IL-18 receptor, a complex of IL-18Rα and IL-18Rβ, binds to the protein as well as glycan anchor of GPI-anchored CD48 and thereby results in tryosine kinase activation and eventually production of interferon-gamma (IFN-gamma). These effects are prevented when cells are pretreated with phospholipase C (PIPLC) 43.

Cellular prion proteins also participate in signaling events leading to the acquisition of the metastatic phenotype. In gastric cancer cell lines, PrP^c was found to transactivate the transcription factor MMP11 by signaling via the MEK/ERK pathway leading to enhanced metastasis in vivo 44.

In COS-1 cells, GPI-anchored Cripto has been shown to control the intra-endosomal sorting and trafficking of the TGF-b family member, Nodal 45. GPI-anchored Cripto facilitates localization of Nodal (endocytosed for 40 min) to the limiting membrane of GFP-Rab4 positive endosomes. Signaling via Nodal is also facilitated by the interaction of Cripto and Nodal mediated by the EGF domain of Nodal.

These studies demonstrate that GPI-APs can participate in signaling events at the plasma membrane as well as in endosomal membranes.

GPI-APs are thus involved in various physiological processes, with the endocytosis of GPI-APs being an important regulatory mechanism in several of these processes. Clearly, in the case of the FR as well as the prion protein, access to certain endocytic compartments is mediated by the GPI-anchored isoforms of these proteins. The mechanism of intracellular trafficking might be responsible for efficient folate-uptake. It is likely that pathogens are able to exploit the normal endocytic pathways of GPI-APs to access intracellular compartments that may be distinct from the degradative lysosomal compartments. Likewise, endocytic compartments accessed by the cellular prion protein may be the sites of formation of pathogenic prion particles. In this review, we have chosen the FR, prion proteins and the urokinase type plasminogen activator receptor (uPAR) to examine the various endocytic processes that serve to internalize GPI-APs.

GPI-anchored FR was shown to recycle in cells between acid resistant (intracellular) and acid-sensitive (extracellular) pools presumably via endosomal compartments 53. Replacement of the GPI-anchor of FR with a transmembrane one resulted in endocytosis via clathrin coated pits but the chimeric receptor was unable to deliver the endocytosed folate to the cytoplasm with the same efficiency as the GPI-anchored FR 18. Moreover, the cytoplasmic accumulation of folate via the transmembrane anchored was not inhibited by the confluence status of the cells. The transmembrane anchored FR was therefore not subjected to cellular regulatory processes in the same manner as the native GPI-anchored FR 18. Presumably, replacement of the GPI-anchor with a transmembrane one resulted in altered intracellular destinations of the FR where the efficiency of folate absorption was considerably reduced.

GPI-anchored FR was thought to be organized within caveolae in cholesterol-dependent clusters (detected using primary and secondary antibody staining techniques54) and a mechanism for folate-uptake based on this clustering was proposed (termed as "potocytosis"), whereby, transient closure and acidification of the caveolae would facilitate folate-uptake and release into the cytoplasm. This model was proposed as an explanation of the fact that FR recycles in MA104 cells 53, resulting in an intracellular and an extracellular pool. Yet these authors were not able to conclusively present an endosomal location for the intracellular pool of recycling receptors.

The potocytosis model of caveolar organization and internalization of GPI-APs were revised when it was found that cross-linking of GPI-APs lead to their selective localization into caveolae 50. Immunodetection procedures employing primary and secondary antibodies resulted in the formation of clusters of GPI-APs in caveolae. Interestingly, the cross linking of GPI-APs is not prevented by standard paraformaldehyde based fixation techniques. The native distribution of FR visualized by labeled primary antibodies was uniform at the plasma membrane without any quantitative enrichment in any class of cell surface invaginations 50.

Cross-linking induced caveolar sequestration of GPI-APs has been confirmed in several other studies 5556. However, the native organization of GPI-APs has been shown to be independent of caveolae 5057.

Studies on endocytosis of non-cross linked GPI-APs, using monovalent ligands have been possible with the FR 5859. Sabharanjak et al have demonstrated that FR bound by a fluorescent monovalent analogue of folic acid (PLF, 59) is internalized into a distinct set of early endosomes that may include fluid phase markers but are distinct from the endosomes derived from the clathrin-coated pits. GPI-anchored FR is endocytosed at very early time points (typically 2 minutes) into

G

e

e

c

Depletion of specific lipids like cholesterol resulted in a dramatic redirection of the FR into endosomes containing endocytosed TfR. Since the segregation of GPI-AP and transmembrane anchored proteins like TfR is evident at very early time points (2 minutes, early relative to the recycling periods of these receptors), it is clear that these sorting events are mediated at the plasma membrane. Interestingly, endocytosis via the Cdc42 regulated pathway was also sensitive to lipid perturbations, like cholesterol depletion, which altered the cell surface organization of GPI-APs in lateral inhomogeneities termed as "lipid rafts" 5960. These elegant experiments demonstrate that internalization of GPI-APs occurs via distinct cell surface invaginations apart from clathrin-coated pits and caveolae 59, and is dependent on lateral associations of GPI-APs with cholesterol and sphingolipids in the exoplasmic leaflet of the plasma membrane. In the end, the efficiency of folate absorption by the GPI-anchored FR is determined by these endocytic events.

Recently Chen et al 61 demonstrated that FR endosomes are delivered to endosome recycling compartment in KB and HeLa cells via microtubules. Using antibodies for dynein and kinesin I, these authors verified the role of dynein and kinesin I in traffic-mediating of FR endosomes towards the minus or plus end of microtubules, respectively. Metabolic depletion of endosomal cholesterol led to increased FR endosome motility in KB cells. This increase in FR endosome motility was proposed to be a consequence of altered associations with Rab GTPases. Cholesterol-depletion was found to lead to decreased association with Rab7 and dynein and increased association with Rab4 and kinesin (KIF3), which ultimately caused the enhanced endosome motility. These results support the earlier work by Chatterjee et al demonstrating how the depletion of cellular cholesterol and sphingolipids leads to increased recycling of GPI-APs from recycling endosomal compartments 62.

Ultimately, studies such as those detailed above help clarify the endocytosis events mediating folate-uptake and will prove valuable for the general comprehension of the efficiency of cellular folate absorption.

Another piece of evidence that demonstrates that caveolin is not involved in the mechanism for internalization of GP-APs with membrane inhomogeneites is derived from studying the cholera toxin B subunit (CT-B). CT-B, which binds to the ganglioside GM1 63, has been shown to be endocytosed via multiple endocytic pathways including caveolae (when stimulated with okadiac acid or lactosylceramide treatments) as well as a non-clathrin non-caveolae mediated endocytic pathway 64. These CT-B positive endosomes are formed even in cells derived from caveolin-1 null mutant mice and demonstrate tubulovesicular and ring shaped morphologies similar to those of GEECs 5964. The ultra-structural characterization of these early endosomes shows that these are tubular structures, 65 which are capable of including cell surface bound cholera toxin, fluid phase markers like HRP and GPI-APs. In fact, in cav-1 null mouse embryonic fibroblasts, a major fraction of internalized CT-B was found to be co-localized with GPI-APs at 2 minutes post-internalization and not with transferring, a marker for the endosomes derived from clathrin-coated pits. Taken together, these results indicate that inclusion of membrane inhomogeneities consisting of GPI-APs, cholesterol, sphingolipids and GM1 can be organized in the absence of caveolin and that these structures do not utilize caveolar invaginations for endocytosis. It must be noted, however, that CT-B shows limited specificity for GM1 66 and is capable of binding other glycolipids as well which is probably why the molecule can enter cells via multiple pathways. In order to use CT-B as a selective marker for GPI-AP associated membrane regions, a distinction between the binding sites on the plasma membrane for CT-B is essential.

Tubular endosomes have been shown to be regulated by another GTPase, ARF-6 6768. ARF-6 and Tac (IL-2 receptor alpha subunit) were shown to traffic from the plasma membrane to a perinuclear endocytic compartment which was distinct from endosomes containing clathrin-coated-pit derived cargo (TfR) 6768. These endosomes also display tubular morphologies and were shown to facilitate endocytosis and recycling of the major histocompatibility complex class I (MHC-I) protein 69. Naslavsky et al have proposed that the ARF-6 regulated tubular endosomes also included GPI-APs like CD59 in HeLa cells 70. Inclusion of GPI-APs and MHC-I into the clathrin-independent tubular endosomes was also shown to be dependent upon cholesterol. Sequestration of plasma membrane cholesterol by filipin resulted in reduced endocytosis of these proteins. It is unclear whether fluid phase cargo like dextrans, representing large scale endocytosis, which is a characteristic feature of GEECs, is also included in these tubular endosomes. However, a distinction has emerged between the ARF6-regulated tubular endosomes 6768 and GEECs from the work of Kalia et al. 71. These researchers have shown that GEECs capable of endocytosing fluid phase markers, as well as GPI-APs, are not regulated by ARF6 in CHO cells and thus represent a completely distinct pathway. Over-expressed wild type or dominant negative ARF6 was not co-localized with GEECs and did not affect the delivery of endocytosed GPI-APs to the recycling endosomes containing TfR 71. The acidic nature of the GEECs (pH 6.0, 71) is also a feature that sets them apart from other endosomes like sorting endosomes, and pH neutral caveosomes 72. Also, proteins internalized via the ARF6-regulated endocytic pathway and clathrin-mediated endocytosis do not enter into a common compartment in HeLa cells even at later time points (30 minutes) post-endocytosis 70. Internalization of CD59 is dependent upon Rab22a activity 70, whereas the endocytosis of GPI-APs in CHO cells is regulated by Cdc42 59. It is therefore possible that the ARF6 regulated pathway, although cholesterol-dependent, represents a distinct pathway from that described by Sabharanjak et al in CHO cells. Alternatively, existence of multiple endocytic pathways may be a cell-type specific feature, and these studies reflect the inherent differences between HeLa (epithelial) and CHO (fibroblasts) cells in maintaining membrane flux. It would be interesting to understand whether these differences are instrumental in regulating physiological responses in various cell types. Also, the cholesterol sequestration methods used in these studies are different. Using metabolic inhibitors of cholesterol synthesis like statins is expected to reduce the overall levels of cholesterol in cells but would presumably not perturb the inherent distribution patterns of cholesterol in various cell membranes as well as in lateral membrane inhomogeneities. Filipin, on the other hand, binds to cholesterol in the plasma membrane and it is yet unclear whether this process in itself disrupts the partitioning of plasma membrane cholesterol between raft-associated and free cholesterol.

Urokinase type plasminogen activator (uPA) is an essential component of the plasmin generation system. Plasmin is a protease that facilitates cell mobility by digesting the integrin family of anchoring and adhesion proteins (73, reviewed in 74). uPA activity is in turn regulated by its inactivators, plasminogen activator inhibitors-1 and -2 (PAI-1 and -2, also referred to as

ser

p

in

uPAR bound to uPA is known to interact with α₅β₁ integrin as well as EGFR and thereby facilitate signal transduction via phosphorylation of focal adhesion kinase (FAK) and EGFR (reviewed in 11 and 78). The downstream effectors of these signaling events then facilitate cell migration 78. Gangliosides seem to moderate these interactions in an interesting manner. Squamous carcinoma cells (SCC12), showed increased mobility in scratch as well as chemotaxis-induced migration assays post-stimulation with uPA, when these cells were depleted of gangliosides 79. Conversely, increased levels of gangliosides GT1b and GM3 reduced the mobility of cells when challenged with uPA. Significantly, increased levels of GT1b also blocked the phosphorylation of FAK, which is normally facilitated by the interaction of uPa:uPAR with α₅β₁ integrin. Likewise, increased levels of GM3 inhibited phosophorylation of EGFR when cells were stimulated with uPA 79. These results show that GT1b and GM3 are capable of disrupting the interactions of uPAR with α₅β₁and of the uPAR: α₅β₁complex with EGFR 79. It may be envisaged that the presence of uPAR in plasma membrane inhomogeneities can be influenced by the plasma membrane levels of gangliosides. uPAR present in cholesterol-sphingolipid organized rafts would presumably be incapable of interacting with α₅β₁ integrin and EGFR. Thereby the rafts would serve as a regulatory sequestering mechanism that governs uPA-mediated transmembrane signaling and chemotaxis events. Alternatively, the aggregation of uPAR into rafts may also cause the endocytosis of these receptors into GEECs, or similar endosomes in other cell systems, thereby preventing uPA-induced cell motility.

It is unclear whether uPAR can be endocytosed via GEECs (or similar CDC42-regulated endocytic pathways) when cellular levels of gangliosides, and possibly cholesterol, are elevated (see Figure 1). However the existence of such a mechanism is possible. The inclusion of GPI-APs into laterally-organized membrane rafts seems to facilitate not just endocytosis but modulation of cell surface events, like receptor engagement and signal transduction. Engagement of uPAR:uPA complexes with PAI-1 also results in disengagement of uPA:uPAR and vitronectin 77, leading to a loss of anchorage and increased mobility of cells. It is unclear whether these functions of uPAR are also modulated by cholesterol and sphingolipid levels in the plasma membrane but it would be interesting to examine these phenomena, especially in metastatic cells.

Given that cellular lipids can modulate the signaling interactions of uPA:uPAR complexes, it is likely that raft-facilitated endocytosis of uPAR may serve as spatial-temporal regulatory mechanism to control cellular signaling events.

Cell surface resident prion proteins were shown to traffic through endocytic intermediates and this step was shown to be necessary for conversion of PrP^C to PrP^Sc 22. In N2A neuroblastoma cells, transfected chicken PrP (chPrP) was shown to be endocytosed in endocytic structures that also showed uptake of a fluid phase marker 80. Clathrin-coated pits were shown to be instrumental in the endocytosis of chPrP, since the endocytosis of prion proteins was shown to be inhibited by treatment with hypertonic medium 81. This is in contradiction with the endocytic pathway demonstrated for the endocytosis of FR-GPI 59, where the fluid phase filled endosomes were shown to be distinct from those derived via clathrin-coated pits.

Prion protein endocytosis is also stimulated by binding of cupric (Cu²⁺) ions to the extracellular domain 82. Copper-binding stimulated the endocytosis as well as recycling of GPI-anchored prion proteins for a sustained period (60 min.), indicating that perhaps the prion proteins act as receptors for Cu²⁺ and facilitate its cellular uptake. Quite noticeably, mutant prion proteins incapable of binding copper or zinc ions are not endocytosed 83. Mutations in the copper binding octarepeats of the sequence PHGG(G/S)WGQ results in loss of copper-binding sites as well as reduction in copper-induced endocytic trafficking of prion proteins 83.

Furthermore, endocytosis of prion proteins was found to be dependent upon dynamin and to occur independently of co-expressed GFP-GPI 84. In co-transfected SN56 cells, dynamin K44A mutant expression reduced the endocytosis of PrP^c whereas GFP-GPI was endocytosed into recycling endosomes despite an over-expression of dominant-negative dynamin. Co-localization of PrP^c with GFP-Rab5-Q79L in early endosomes suggested that these proteins are internalized via the clathrin-coated pit-derived dynamin-dependent endocytic processes. GFP-GPI is seen to be endocytosed and trafficked to the recycling endosomes despite the expression of Rab5-Q79L. These results propose that additional factors may influence the endocytosis of PrP^C and the GPI-anchor itself may not be a sole deterministic factor.

The existence of a potential cell surface receptor, a 66kD protein, for PrP^c has been suggested from complementary hydropathy studies 85. Recently, the LDL-receptor-related protein (LRP-1) has been shown to act as a receptor for PrP^c 868788. These results are in concord with the fact that cellular or over-expressed prion proteins are found in transferrin and Rab5 positive early endosomes, explaining the differences in the endocytic trafficking of other GPI-APs like FR-GPI and GFP-GPI 5984. However, these results contradict earlier experiments with transmembrane anchored prion proteins 2589 wherein conversion to the infectious and neurodegerative scrapie isoform was prevented when the GPI-anchor was replaced with transmembrane anchoring sequences that directed the chimeric protein(s) to be internalized via clathrin-coated pits.

It is possible that the endocytic compartments involved in prion protein mediated copper uptake may be distinct from those that are involved in the conversion of cellular prion proteins to the scrapie isoform. Alternatively, the cell surface organization of prion proteins into cholesterol-sphingolipid domains (rafts) may influence the endocytic pathway and thereby the decision to be converted to the disease-inducing isoform(s). Since lateral interactions with LRP-1 would necessitate that prion proteins be organized in monomers (or cause a dissociation of raft included multimers), internalization into the clathrin-mediated endocytic pathway would possibly preclude lateral interactions between PrP^C and PrP^Sc. It is likely that GPI-anchored prion proteins organized into lipid raft-included multimers may get internalized into acidic compartments similar to FR-GPI. These endosomes being acidic (pH ~6.0) may be the sites for the conformational changes responsible for conversion of PrP^cinto PrP^Sc as can be seen from in vitro conformational change studies 9091. Since protein-protein interactions between the PrP^C and PrP^Sc isoforms can possibly be facilitated by multimeric organization into cholesterol-sphingolipid ordered rafts and inclusion into acidic endosomes, it is likely that GEEC-like endosomes may be the sites for this conversion. Inclusion into the clathrin-coated pit pathway stimulated by the binding of the metal cations Cu²⁺ and Zn²⁺ may actually serve to be a disease preventive mechanism by promoting the dissociation of multimeric prion proteins into monomers. These hypotheses need to be evaluated in vivo using biophysical techniques to identify raft-associated and monomeric species of prion proteins.

A comparison of the endocytic pathway followed by PrP^C and PrP^Sc by simultaneous visualization of both proteins would probably shed some light on the endocytic destinations of these proteins. Differential regulation of endocytic pathways in various cell types can also account for the observed differences in endocytic trafficking of prion proteins. It is unclear whether all endocytic pathways remain active in all cell types, especially in terminally differentiated cell types such as neurons.

Assuming an unbiased cell surface distribution of all membrane resident proteins, it is theoretically valid to assume that LRP and LRP-1 may be included in cholesterol-sphingolipid organized rafts for some time on the cell surface. However, during internalization of these proteins as well as uPAR and prion proteins bound to them respectively, the peptide signal in the cytoplasmic tail for clathrin recruitment would be a more dominant factor than the interactions with laterally-situated cholesterol and sphingolipids. Therefore it can be safely assumed that LRP and LRP-1 cannot be endocytosed via the GEECs. Moreover, conditions that inhibit clathrin-coated pit-mediated endocytosis inhibit the endocytosis of prions (and therefore LRP-1) but not of co-transfected GFP-GPI 84.

Conversely, GPI-APs which span only the extracellular leaflet of the plasma membrane are likely to leave the rafts and can possibly be endocytosed via clathrin-coated pits. However, cellular and physiological evidence goes against this assumption. Quantification of endocytosed FR-GPI and TfR fluorescence shows that a large fraction of FR-GPI is not co-localized with TfR at 2 minutes post-internalization 59. This ratio is significantly reduced (showing more co-localization of the two proteins) only when cellular cholesterol is depleted metabolically. Nonetheless, absorption of folate into the cytoplasm is significantly compromised in cholesterol-depleted cells, which indicates that the monomeric form or GPI-anchored folate receptor may not be the default state of these proteins in living cells 95.

Prion proteins directed towards clathrin coated pits are not capable of forming amyloid plaques whereas GPI-anchored prions do result in amyloid plaques in an endocytic compartment, sensitive to cholesterol depletion and chloroquine 25. Likewise, the folate transport function of transmembrane-anchored FR is not regulated in the same manner to FR-GPI 18.

Although the endocytosis of monomeric GPI-APs via clathrin-coated pits cannot be ruled out absolutely with the state-of-the-art evidence, the functional correlates argue against this mechanism as being the dominant one for endocytosis of "raft-included" GPI-APs. Since cholesterol-sphingolipid ordered rafts are dynamic structures, it has been extremely difficult to establish distinct populations of raft-organized multimers and monomers of GPI-APs in living cells, visible at the resolution limit of light microscopy. That said, the differences in the morphologies of the two populations of endosomes in living cells suggest that the membrane deformation required to form the tubular GEEC invaginations can possibly be induced by the coalescence of cholesterol-sphingolipid ordered rafts. Clathrin-coated pits, on the other hand, are formed from the selective interaction of clathrin and other associated proteins with membrane segments containing membrane-spanning proteins with clathrin-recruitment signal sequences. The inherent differences in the mechanisms of these endocytic processes sufficiently indicate exclusion of cholesterol-sphingolipid ordered rafts from clathrin-coated pits and of LRP-1 from GEECs. As further support of this, Nichols 96 has demonstrated the exclusion of lipid rafts from clathrin-coated pits. It has been found that the cholesterol-dependent association of glycosphingolipid (GM1) labeled with cholera toxin B subunit did not internalize within clathrin-coated pits.

Biochemical isolation of GEECs has not been achieved yet. Membrane fragment isolation techniques are always subject to contamination from other internal membrane systems, and therefore any claims about lipid composition of selective membrane fragments becomes a controversial exercise. It may be possible to isolate GEECs by inhibiting clathrin-mediated endocytosis and then pulsing the cells with a fluorescent, non-cross-linking GPI-AP ligand for a short time to achieve selective labeling of GEECs to help isolation. However, to the best of our knowledge, this has not been achieved as yet.